Description
In marketing, a product is anything that can be offered to a market that might satisfy a want or need. In retailing, products are called merchandise. In manufacturing, products are bought as raw materials and sold as finished goods. Commodities are usually raw materials such as metals and agricultural products, but a commodity can also be anything widely available in the open market.
ABSTRACT
Title of Dissertation:
Marketing of Digital Products Nevena Taneva Koukova, Doctor of Philosophy, 2005
Dissertation directed by:
Professor Brian T. Ratchford and Professor P. K. Kannan Department of Marketing
My dissertation comprises of three essays that theoretically and empirically investigate the marketing of digital products, which are information products such as newspapers and books sold in both physical and electronic form. In the first essay, we study product form bundling, defined as marketing two or more forms of the same product as a package. We show experimentally that, regarding information products, the usage situations communicated to consumers moderate the effect of the availability of bundle discount on the purchase likelihood for the product form bundle. We also compare the effect of different pricing strategies for information products. When no bundle discount is offered, the likelihood of buying both forms of an information product, holding the sum of their prices constant, can be increased by pricing the electronic form lower than the print form rather then pricing both at the same level. In the second essay we compare two product strategies that can be used in marketing digital products. Under standard mixed bundling companies offer full content in print and electronic form and the bundle of the two, while under content unbundled
mixed bundling companies offer full content in print form, unbundled content in electronic form, and the bundle of the two. Which strategy is more attractive for a company to pursue? We model the profits under these two strategies and outline conditions in which one or the other leads to higher profit. We apply our analytical framework to data from a field experiment implemented on the website of a book publisher. The third essay investigates the attractiveness of complete product lines of items such as books and newspapers. We employ a choice experiment in which a sample of consumers is presented with hypothetical product scenarios asked to make a choice. The data is used to develop a profit-maximizing configuration of products and prices. Similar approaches to the product line pricing problem have been employed for conventional products, but not when bundling of different forms of a product is an option, and not when the different products may be complements rather than substitutes.
MARKETING OF DIGITAL PRODUCTS
by Nevena Taneva Koukova
Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park in partial fulfillment Of the requirements for the degree of Doctor of Philosophy 2005
Advisory Committee: Professor Brian T. Ratchford, Chair Professor P. K. Kannan, Chair Professor Roger R. Betancourt Professor Rebecca W. Hamilton Professor Joydeep Srivastava
©Copyright by Nevena Taneva Koukova 2005
ACKNOWLEDGEMENT I would like to thank my advisors, Dr. Brian Ratchford and Dr. P. Kannan, for their excellent guidance and help on my dissertation and my other research projects. I would also like to thank the other dissertation committee members, Dr. Rebecca Hamilton, Dr. Joydeep Srivastava and Dr. Roger Betancourt, for their helpful comments and suggestions at various stages of my dissertation.
ii
TABLE OF CONTENTS List of Tables ......................................................................................................................v List of Figures ...................................................................................................................vii Chapter I: Dissertation Overview ....................................................................................... 1 Chapter II: Essay One. Product Form Bundling - Implications for Marketing Digital Products............................................................................................................................... 4 Abstract ................................................................................................................... 4 Introduction............................................................................................................. 5 Research Background ............................................................................................. 9 Study Objectives ................................................................................................... 13 Theoretical Framework and Hypotheses .............................................................. 14 Study 1 .................................................................................................................. 25 Study 2 .................................................................................................................. 29 Discussion ............................................................................................................. 38 Managerial Implications ....................................................................................... 40 Conclusion ............................................................................................................ 42 Chapter III: Essay Two. Bundling and Pricing Strategies for Digital Products ............... 44 Abstract ................................................................................................................. 44 Introduction........................................................................................................... 45 Research Background ........................................................................................... 47 Bundling Model .................................................................................................... 49 Data and Estimation Procedure............................................................................. 59 Results................................................................................................................... 64
iii
Discussion ............................................................................................................. 69 Managerial Implications ....................................................................................... 72 Conclusion ............................................................................................................ 73 Chapter IV: Essay Three. Bundling and Unbundling of Electronic Content.................... 75 Abstract ................................................................................................................. 75 Introduction........................................................................................................... 76 Research Background ........................................................................................... 78 Theoretical Background........................................................................................ 82 Methodology ......................................................................................................... 86 Study Design and Data Collection........................................................................ 87 Results................................................................................................................... 90 Discussion ........................................................................................................... 102 Managerial Implications ..................................................................................... 106 Conclusion .......................................................................................................... 107 References........................................................................................................................131
iv
LIST OF TABLES Table 1: Purchase Probabilities for Bundle: Product, Discount and Usage Conditions . 109 Table 2: Perceived Appropriateness of Generated Usage Situations.............................. 110 Table 3: Study 2 - Mean Percentage of Points Allocated to Bundle .............................. 111 Table 4: Study 2 ANOVA Results.................................................................................. 112 Table 5: Profit Comparison under CUMB and TMB Strategies..................................... 113 Table 6: Actual Purchase Rates under the Two Mixed Bundling Strategies.................. 114 Table 7: Estimation Results - Content Unbundled Mixed Bundling .............................. 115 Table 8: Estimation Results - Traditional Mixed Bundling............................................ 116 Table 9: Simulation Results – Profit under the Two Mixed Bundling Strategies. ......... 117 Table 10: Conjoint Design and Price Levels .................................................................. 118 Table 11: Random Parameters Logit Model Results (conjoint choice experiment)....... 119 Table 12: Book Category – Market Simulation Results ................................................. 120 Table 13: Newspaper Category – Market Simulation Results........................................ 121 Table 14: Book Category – Market Simulation Results (Incomplete Product Line)...... 122 Table 15: Newspaper Category – Market Simulation Results (Incomplete Product Line) ......................................................................................................................................... 123 Table 16: Book Category – Market Simulation Results (Incomplete Product Line, No Bundle Discount) ............................................................................................................ 124
v
Table 17: Newspaper Category – Market Simulation Results (Incomplete Product Line, No Bundle Discount) ...................................................................................................... 125 Table 18: Book Category - Forms’ Attribute Perceptions. ............................................. 126 Table 19: Newspaper Category - Forms’ Attribute Perceptions..................................... 127
vi
LIST OF FIGURES Figure 1: Study 2 Results................................................................................................ 128 Figure 2: Optimal Price and Total and Marginal Revenue from Electronic Chapters.... 129 Figure 3: Optimal Price and Total and Marginal Revenue from Full Electronic Book.. 130
vii
CHAPTER I: DISSERTATION OVERVIEW
Marketing digital products is appealing because of several reasons: 1) the marginal cost of reproducing and distributing them is almost zero (Bakos and Brynjolfsson, 1999), 2) they can be easily organized, searched and stored, and 3) the whole buying experience trough Internet can be more interactive and tailored to the specific needs and preferences of the customer (Peterson et al., 1997). Although many publishers make available content online, others (e.g. marketers of full length books, reference materials, music and video) are still cautious in distributing content in digital form over the Internet (Kannan and Jain, 2003). There are good reasons for this cautious approach, such as piracy of content and bandwidth constraints. However, less valid reasons also prevent managers from offering digital content, such as uncertainty about what product and pricing strategies to utilize. How should companies market digital products? What product, pricing and communication strategies should they employ? My dissertation attempts to tackle these interesting and practically relevant questions. It comprises of three essays organized as follows. In Essay 1 we focus on consumer reactions to bundling of information products, and contrast these with consumer reactions to bundling of conventional products. In particular, we concentrate on communication and pricing strategies that might enhance the viability of selling the information goods as a bundle. We show experimentally that communicating different possible usages for the product forms may increase the likelihood of buying the items as a bundle. We also investigate consumer reactions to
1
alternative pricing plans for product forms. Our respondents readily accept bundle discounts on conventional items that can be inventoried, but are generally unresponsive to bundle discounts on different forms of information goods that have the same content. However, emphasizing different uses for the forms does increase their responsiveness to bundle discounts. We conclude that communicating the appropriateness of different forms for different usage situations can be a key to a successful bundling strategy for information goods; further, it is more beneficial for companies to price the electronic form lower than the print form instead of offering bundle discounts. In Essay 2 we model the profits under two strategies a content provider may employ: traditional mixed bundling, in which the product line consists of print book, PDF book and the bundle of the two, and content unbundled mixed bundling, which is selling print book, PDF chapters and the combination of the two. While appealing because of low marginal costs and the likely catering to emerging consumer needs, offering electronic products may lead to additional revenue generation but can also result in product cannibalization. Our results suggest that offering unbundled PDF chapters is more profitable than offering only the full PDF books under certain conditions. We empirically support our predictions with actual data gathered in an online experiment executed on the website of a publisher of academic books. We present interesting insights on how customers from different subject domains react to electronic products and how they are likely to behave when different product and pricing schemes are employed. In Essay 3 we further extend the previous essays and investigate the attractiveness of complete product lines of information good. For example, the publisher of Wall Street Journal can offer the following product line of subscription options: print WSJ, WSJ
2
online, separate sections of WSJ online (e.g. Money & Investment, Technology), print WSJ and online WSJ bundle, and print WSJ and online section bundle. We employ a choice experiment in which a sample of consumers is presented with hypothetical product offerings at various prices and asked to make a choice. The data is used to develop a profit-maximizing configuration of products and prices. Similar approaches to the product line pricing problem have been employed for conventional products, but not when bundling of different forms of a product is an option, and not when the different products may be complements rather than substitutes.
3
CHAPTER II: ESSAY ONE. PRODUCT FORM BUNDLING - IMPLICATIONS FOR MARKETING DIGITAL PRODUCTS Abstract In this paper we study product form bundling, defined as marketing two or more forms of the same product as a package. Although we are primarily interested in information products (e.g. Wall Street Journal: print and online forms), we contrast consumer reactions to bundling these products with reactions to bundling conventional products (e.g. margarine: stick and tub forms). We show experimentally that the usage situations communicated to the consumers moderate the effect of the availability of bundle discount on the purchase likelihood for the product form bundle. Our subjects readily accept bundle discounts on conventional items that can be inventoried, but are generally unresponsive to bundle discounts on bundles of information products that have the same content. However, emphasizing different uses for the information product forms does increase the responsiveness to bundle discounts. Also, we examine the impact of different relative prices for the forms within the information bundle on the bundle purchase likelihood, and compare this to offering discounts. Our study provides useful insights into the conditions under which the marketer is better off pricing the forms differentially than pricing them equally and offering bundle discounts.
4
Introduction It is a common practice in consumer packaged goods categories to market products in different forms, for example, margarine (stick and tub forms) and soap (bar and liquid forms). In marketing different forms of the same product the focus has been on offering forms appropriate for various usage situations. For example, margarine is sold in both stick and tub form and customers can choose the form that better suits their needs and usage situations. Recently, marketers of information products have been applying similar strategies as well by making the information products available in electronic form in addition to the print/traditional form. Specifically, newspapers (e.g., Wall Street Journal) market print and electronic forms of their content, publishers (e.g., US Government Printing Office) offer print books and reports as well as the corresponding electronic versions, and recording companies (e.g., EMI Recorded Music) have started offering online music in addition to music in traditional forms (CDs, tapes, etc.) all catering to the diverse needs of customers. While bundling different forms of the same product is occasionally seen in the consumer packaged goods market (e.g. bundling Eucerin body cream and lotion; Dial liquid soap and Dial soap bars), one would generally not expect similar bundling strategies in the information goods market as the content being sold in the different forms is generally the same. However, newspapers such the Wall Street Journal not only offer the print and electronic forms separately as single products, but also as a bundle of different forms. Likewise, publishers of scholarly content such as National Academies Press have started to offer bundles of print and PDF forms in addition to offering them separately. Many scholarly journals also market subscriptions of both print and electronic
5
forms. New formats introduced by dualdisc.com provide the capability to sell music content in CD format as well as DVD-Audio or surround sound as a bundle on the same disk (Business Week, 2004). While different forms of products such as margarine and soap may have roughly the same production costs, information products are unique in that the marginal cost of reproducing and distributing the electronic form is often much lower than the one of the print form. Consequently, it is not surprising that marketers are interested in selling bundles of electronic and print forms, as it implies significant additional revenue. As broadband service penetration increases, similar opportunities in selling bundles of streaming videos and DVDs, and music bundles (mp3 form, CD, mobile tunes) become possible. In order to realize this additional revenue potential, however, companies need a clear understanding of how to market, communicate and price product form bundles. It is in this context that we present our research. In this paper, we first focus on consumers’ reactions to alternative forms of information products, and examine conditions under which they are more likely to purchase the product form bundle. In particular, we concentrate on communication and pricing strategies that might enhance the viability of selling the information goods as a product form bundle. We replicate our study in the consumer packaged goods category to establish a comparative baseline. It is well known that there are different types of utility a good may provide – elementary (or basic), place, form, time, and possession (Macklin 1922). In our study we distinguish between content utility (similar to elementary or basic) – utility provided by the product itself – and form utility – utility provided by the specific form in terms of convenience, ease of use and the like – of the product forms that
6
play a critical role in how the bundle is viewed. Content utility tends to be duplicated across different forms of an information product, making it less attractive to buy them as a bundle. From this viewpoint, the content utility of either of the information forms or the bundle is the same in magnitude. However, in the conventional packaged goods categories, the content utility of the bundle is generally more than that of either of the individual forms. While content utility remains the same in the case of an information product bundle making it less attractive to purchase, extant research would suggest that suitability of different forms of an information product for use in different situations will decrease the perceived similarity between them (Ratneshwar and Shocker 1991) and that bundles composed of complements have a higher purchase intent versus bundles of similar products (Estelami 1999), thus making the information bundle appealing. However, we find experimentally that communicating different possible usage situations for the forms, in and of itself, is not sufficient to increase the probability of selling the bundle of information product forms. Yet, for conventional packaged goods categories, such communication strategies are effective in increasing the bundle purchase probabilities. In investigating consumer reactions to alternative pricing plans for different forms of information products and different forms of conventional products, we find that our experimental subjects readily accept bundle discounts on conventional items that can be inventoried, but are generally unresponsive to bundle discounts on different forms of information products that have the same content. It is only when different uses for the forms are emphasized along with bundle discount that respondents’ bundle purchase probabilities of information product forms increase significantly. This suggests that
7
emphasizing the appropriateness of the forms for different usage situations is as important as bundle discounts in successfully marketing bundles of information products. Finally, we examine the impact of different relative prices for the forms within the information bundle on the purchase of information bundle, and compare this to offering discounts. Our study provides useful insights into the conditions under which the marketer is better off pricing the different forms differentially than pricing them the same and offering bundle discounts. The contribution of the paper is three-fold. First, our study offers a useful approach in terms of examining product forms and the attractiveness of information product bundles under bundle discounts, differential pricing of the forms, and usage situation communication from the perspective of content and form utilities. We are not focusing on the effect of the above factors per se but on their interaction effect in influencing consumer choice. Second, our study complements the findings in extant literature on bundling by providing insights into conditions under which those findings are directly applicable to product forms and instances where those results may not hold. Specifically, our study extends the findings to situations where items of bundles (forms) can be perceived as substitutes as well as complements across different consumers. Third, our paper provides substantive insights in an area of emerging importance – bundling of information products – by providing guidelines for communication and pricing strategies for increasing the attractiveness of information bundles and expanding revenue opportunities for marketers. This is managerially very relevant given that the consumer spending for online content in the US grew to $853 million in 2004, an increase of 14 percent over the same period last year (Online Publishers Association Report, 2004).
8
The paper is structured as follows. First, we define and discuss product form bundling in the context of extant literature and position our study. Second, we present a theoretical framework and derive hypotheses. Third, we discuss two experiments that were designed to empirically test our hypotheses. Finally, we highlight the implications of our study and suggest areas for future research.
Research Background Product Form Bundling Extant literature defines bundling as selling goods in packages (Adams and Yellen 1976), marketing products as a package at a special price (Guiltinan 1987), and selling products at a single price (Yadav and Monroe 1993). Distinction is also made between price bundling, defined as the sale of two or more separate products in a package at a discount without any integration of the products, and product bundling, which is the integration and sale of two or more separate products or services at any price (Stremersch and Tellis 2002). We define product form bundling as marketing two or more forms of the same product as a package. In the context of product categories, we make a distinction between information product form bundles in which the same information is presented in different forms and the forms are bundled together (e.g. electronic book and print book), and conventional product form bundles consisting of different physical forms of the same product (e.g. stick of margarine and margarine tub). Information bundles have some unique characteristics: the marginal cost of producing the electronic form after producing the print is negligible, consumers have reasonable knowledge of this cost structure, and, in
9
some cases, the individual products are perishable (e.g. after reading the news online consumers do not benefit from the print version of the newspaper to read the same news). With respect to conventional bundles, the cost structure is different - the marginal costs associated with the forms are positive - and the products usually are not immediately perishable and can be stockpiled. From a conceptual point of view, product form bundling seems similar to other types of bundling discussed in extant literature. On one hand, it could be viewed as a special case of product bundling – that is, the integration and sale of two or more separate products, the separate products being different forms of the same product; on the other hand, it is a type of price bundling - the sale of two or more separate products in a package at a discount. However, there are two important distinctions in the case of product form bundling. First, the product forms can range from being perceived as substitutes to being perceived as complements to each other and degree of heterogeneity across consumers on this dimension is much higher than what is encountered in other bundling situations. As an illustration, consider the bundling of print and electronic forms of newspaper subscriptions. The Wall Street Journal has been recently promoting the two forms of the newspaper as a bundle. The ads suggest that people might use the two forms for different purposes and/or situations (e.g. keeping up with the daily news versus doing research). If customers perceive WSJ Online as a substitute to WSJ, they will subscribe to one of them; if they view the two form of the journal as complementing each other, they may be willing to subscribe to both. Second, in the case of information product bundles, the marginal cost of producing one form after producing the other form is close to zero, which is quite different from the scenarios encountered in the extant bundling literature.
10
Offering different forms of a product is also similar to having more than one package of the same item available. Packaging of items explicitly sold at retail into a small and a large package, for example, can be viewed as commodity bundling and thus as a mechanism for price discrimination, or it can be viewed as a mechanism for providing different levels of distribution services through the shifting of distribution costs across market boundaries (Betancourt 2004). The small package provides a high level of the distribution service assurance of product delivery in the desired form bundled with the items explicitly sold that are contained in the small package (Betancourt 2004). Offering a print and an electronic form of a product can be considered as two different levels of the distribution service assurance of product delivery in the desired form, and each level is bundled with the content of the product. The difference between offering different product forms and offering different product packages is in the cost structure associated with the two offers. While the marginal costs of producing the different packages are positive and similar to each other, this is not necessarily true with respect to the different product forms - the marginal cost of the electronic form, for example, is close to zero. Consequently, the implications of offering different product forms can be different from the implications of offering different types of packaging both from buyers’ and sellers’ point of view. Because of the cost structure the buyers are likely to expect a relatively lower electronic form price as compared to the print form price, and it may be profitable for the sellers to charge relatively lower price for the electronic form. In sum, offering different forms of a conventional product is the same as offering different packaging types, while offering different forms of an information product is not necessarily the same as offering different packaging types because of the associated cost
11
structure. Thus, we use the term product forms as it is more general. Also, our focus is on bundling product forms (or types of packaging in the case of conventional products), and not on bundling an item with the specific service level it provides. While there is extensive literature focusing on conditions favorable for bundling – negative correlations in consumer reservation prices (Stigler 1963; Schmalensee 1984), complementarity in consumption (Telser 1979), uncertainty in the valuations of the quality of the goods (Kenney and Klein 1983) – contingent valuation of interrelated products in the bundle is very relevant for product form bundling. Contingent valuations have been examined at both conceptual (e.g. Guiltinan 1987) and analytical (e.g. Venkatesh and Kamakura 2003 (VK) levels. Under assumptions of homogeneous degree of contingency or complementarity, VK find that moderate to strong substitutes should be offered as separate products; the same is applicable for complements if the marginal costs are moderate relative to the market’s maximum willing to pay. They also find mixed bundling to be optimal for weak substitutes/complements if the marginal costs are not too high (Venkatesh and Kamakura 2003). It is not clear to what extent these results will hold given the heterogeneity in customer perceptions of the product forms as being substitutes or complements (contingency) or under the case of very low marginal costs. Thus, our study on bundling product forms might complement the above findings. In addition, the heterogeneity in consumer contingency/ complementarity perceptions may also suggest that consumers may be amenable to suggestions from marketers in manipulating such perceptions. Extant research based on mental accounting and related framing effects have provided interesting insights into consumer reaction to bundling (Thaler 1985; Johnson
12
1999). The findings relevant for relative pricing of the product forms and its impact on bundle choice suggest that (1) in terms of evaluation process, there is evidence that buyers use anchoring and adjustment (Yadav 1994), (2) the evaluation of a bundle depends on the price leader as well (Yadav 1995), and (3) savings offered on the bundle have a greater relative impact than savings offered on the individual items (Yadav and Monroe 1993). In the context of information products, the price of the form that consumers are more familiar with – print – could form an anchor against which other forms are evaluated. Also, given the marginal cost of one form (say, an e-book) is close to zero, the relative pricing of the forms may have to account for consumers’ expectations regarding the relative prices of product forms.
Study Objectives Our study objectives are as follows. First, we examine the concept of product form bundling from the viewpoint of content utility and form utility with the objective of answering the questions faced by a manager marketing information product form bundles – how can he/she increase consumers’ choice of bundles through appropriate communication to impact consumers’ perception of complementarity? And how should he/she price the individual product forms and the bundle? We contrast information product form bundles with conventional product form bundles to gain better insights into consumers’ reactions to bundling. Our second objective is to explain our findings in the light of extant literature in bundling and highlight the similarities and variances that help us better understand product form bundling. Given that most of the behavioral studies so far focus on bundles composed of complements (e.g. computer and printer), on investigating how consumers evaluate such bundles, and on the optimal strategies for
13
price and discount information presentation, the fact that we study bundle items for which marketers can influence the degree of contingency/complementarity should provide results that complement extant findings. Thus, our overall focus is not only on understanding how the bundling of interrelated products influences consumers’ purchase decisions but also on how these decisions can be affected through marketing communication and differential pricing.
Theoretical Framework and Hypotheses Perceived Complementarity From the firm’s point of view, complementarity is traditionally defined by referring to the sign of the cross-product elasticity of demand - if the sign is positive, products are substitutes; in the opposite case they are complements (Sarvary and Parker 1997). If different forms are substitutes, the purchase of one item lowers the value of alternative forms (contingent valuation), and therefore makes bundling the items less attractive. If different forms are complements, bundling becomes more attractive because the consumer is in effect purchasing a system of products that enable and enhance the functionality of each other (Estelami 1999). Consumers may perceive a high degree of complementarity based on economies of time and effort in purchasing the bundled items together (search economies), improved satisfaction because of the bundle, and/or improved total image of the bundle (Guiltinan 1987; Simonin and Ruth 1995). With regard to the attractiveness of bundling to firms, complementarity between products can cause bundling to be profitable (Telser 1979) and bundles composed of complements have higher purchase intent versus bundles of similar or unrelated products (Harlam et al. 1995). There is evidence that complementarity positively affects bundle reservation
14
prices as well (Gaeth et. al 1990). Applied to product form bundling, we would expect a positive relationship between the degree of perceived complementarity between the bundle components and the purchase likelihood for the bundle. On one hand, if consumers believe that the individual product forms can be used interchangeably, they would buy only one of the forms and not the product form bundle. On the other hand, if consumers think that there is additional utility in having both product forms versus just one of them, they would choose the product form bundle, ceteris paribus. Hypothesis 1: There is a positive relationship between the degree of perceived complementarity between the product forms in a bundle and the purchase likelihood for the product form bundle. We use the above hypothesis as a baseline and build on it to explain how exactly communicating different usage situations influences consumers’ complementarity perceptions and thus consumer choice. We elaborate more on this in the discussion section of Study 1.
Usage Situation Usage situations correspond to activities and conditions for which products are created and marketed (Fennell 1978). In order to analyze the effect of usage situations on the acceptance of product form bundles, we draw upon extant literature (Macklin 1922) to make a distinction between VC, the utility of the content (material in a book, the margarine or soap itself), and VF, the utility that the specific form adds to the content
15
(print or electronic form of a book, stick or tub form of margarine)1. A consumer’s utility of a product form of an information product is not independent of whether the consumer also buys another form. This is because the consumer obtains the same content when buying the two product forms and she does not get extra utility from having the same content twice. For example, if a consumer buys a reference book bundle, she gets the same content twice, but receives content utility only once. However, the consumer can read the print edition while traveling by metro/train, and search the electronic edition if looking for some specific information. Consequently, she can derive two separate form utilities from the bundle – that is, the incremental utility from acquiring the second form together with the first is positive. How great this incremental utility is depends on whether the alternative forms are viewed as complements. In our framework, the consumer’s purchase decision for an information product that comes in two formats can be represented as follows: choose the maximum of Value of Product Form i: Value of Product Form j: Value of Bundle: VC + VFi - Pi VC + VFj - Pj VC + VFi + VFj|Fi – (1 – d)*(Pi + Pj),
where d is the bundle discount, and VFj|Fi is incremental form utility of Form j given that Form i is purchased.2 Note that VC is counted only once in valuing the bundle because the same content is shared by both forms. Let Form i be the form that provides the highest value to the consumer (e.g., VFi - Pi > VFj - Pj). Then the bundle will be chosen if VC + VFi + VFj|Fi – (1 – d)*(Pi + Pj) > VC + VFi - Pi, or if VFj|Fi + d(Pi + Pj) > Pj. If there are two
We assume that VC and VF are independent and additive for expositional purposes, without any loss in generality. 2 Specifically VFj|Fi = VFi + Fj - VFi, where VFi + Fj is form utility from buying both items.
1
16
items in the bundle, and at least one of them has a positive valuation, we can state a general condition under which a consumer would buy both items if there were no discount: (1a) ProbINFO = Prob (VFj|Fi > Pj),
where j is the item that provides less value, i is the item that provides more value, and Prob refers to probability that a consumer will buy both items. The incremental form utility of the second item must exceed its price if it is to be purchased along with the first. If there is a bundle that is offered at a discount, the condition for buying the bundle becomes: (2a) ProbINFO = Prob (VFj|Fi + d (Pi + Pj) > Pj) = Prob (d > (Pj - VFj|Fi)/(Pi + Pj)).
The consumer will buy the bundle only if the discount is large enough to overcome the fact that she only receives form utility for the less preferred item. If VFj|Fi > Pj she would buy this item without a discount, as shown in Equation 1a. While information bundles suffer a redundancy of content, this is ordinarily not the case for conventional bundles. In the case of conventional bundles, consumers normally derive content utility from both product forms because they can first use Form i and then Form j.3 For example, if a consumer buys a bundle consisting of a pack of four sticks of margarine and a margarine tub, she can first use the sticks, store the tub, and then use the tub. Consequently, she derives two content values. In terms of form value, the consumer can use the margarine stick for baking purposes, while the tub for spreading margarine on bread or bakery, thus deriving form utility from both product forms
3
We assume here that conventional products do not deteriorate in quality as they are stocked. For example, rapidly perishable items can be an exception to this. As one example, large containers of soft drinks can be thought of as a bundle (they are a multiple of smaller sizes). A consumer might not value these highly because most of the contents may become flat before the consumer has a chance to consume a high proportion of the contents of the container.
17
(positive marginal utility from acquiring the second product form together with the first). For conventional bundles, the value of the bundle becomes 2VC + VFi + VFj|Fi – (1 – d)*(Pi + Pj), while the values of individual items remain as above4. Using the notation defined above, the probability of buying both forms if there were no discount becomes: (1b) ProbCONV = Prob (VC + VFj|Fi > Pj).
By comparing to 1b to 1a, the probability of buying the conventional product form bundle even when there is no discount will tend to be higher than for the information product form bundle. Similarly, as shown in Equation 2b below, the discount needed to induce consumers of the conventional products to buy a bundle will tend to be smaller because of the opportunity to obtain more content. (2b) ProbCONV = Prob (VC + VFj|Fi + d(Pj + Pi) > Pj) = Prob (d > (Pj - VC - VFj|Fi)/(Pi + Pj)) Hypothesis 1a: Because both forms of a conventional product form bundle provide content utility, consumers are more likely to select a conventional bundle than an information bundle at a given discount level. By directly comparing consumers’ reactions to the two types of product form bundles at specific discount levels we intend to show that the conventional and information bundles are valued differently by consumers because of the different content utility they provide (single or double).
Usage Situation and Bundle Discount When consumers are suggested usage situations in which the two product forms are differentially appropriate (one form is more appropriate for some situations, and the
For sake of exposition, we assume that the different forms of the conventional product have equivalent content and each provide the same content value Vc. Also, we assume that the content utility of the bundle is the sum of the content utilities of the individual forms, with marginal utility remaining the same whatever be the stage of consumption. These simplifying assumptions do not affect our general results.
4
18
other form is more appropriate for other situations), consumers are more likely to view them as complements. Consequently, by manipulating different possible usages for the two forms in the bundle, the form utility of one or both product forms can be changed. Specifically, if communications succeed in convincing consumers that different forms are appropriate for different uses of the product, and therefore more complementary, the value of VFj|Fi can be increased to, say, V*Fj|Fi. As can be seen from equations 1 and 2 this increases the value of the bundle in all cases. Numerous brand management teams apply usage-oriented advertising campaigns to expand the use of their products, for example “Eat Campbell’s Soup with formal family dinners or for breakfast” (Wansink and Ray, 1996). In our study we go beyond this conceptually and consider not the number (we keep it constant) but the nature of the usage situations presented to the consumers. We elaborate on this in the stimuli development section. << Insert Table 1 about here >> The probabilities of buying the information and conventional product form bundles subject to usage situations and level of discount are presented in Table 1. With respect to information bundles, we expect the usage situations considered when making a purchase decision to moderate the effect of bundle discount on purchase likelihood. Specifically, when consumers regard the two product forms as equally appropriate for the same usage situations, the forms will be substitutes and VFj|Fi ? 0. The bundle will be relatively unattractive and there will be no significant difference in their purchase
19
likelihood for the bundle whether a bundle discount is offered or not.5 The discount itself does not motivate consumers to buy the bundle because they are paying for the same content twice and the marginal utility they receive from the second form together with the first is very low or even zero. On the other hand, when consumers are considering different possible usages for the product forms, the likelihood of buying the bundle should be significantly higher in the discount as compared to the no discount case. Specifically, when consumers regard the two product forms as appropriate for different usage situations, the forms will be perceived as more complementary than in the previous case and VFj|Fi ? VFj. The bundle discount can motivate consumers to buy the bundle in this case – not only the marginal utility they get from the second form together with the first is positive but also the money outlay is lower. In the absence of a discount consumers are likely to find the bundle unattractive because it does not increase content utility (see Equation 1a). The bundle discount can compensate for this (see Equation 2a). Hypothesis 2 (information product form bundles): a) When consumers are suggested different usage situations for the product forms, the purchase likelihood for the bundle when bundle discount is offered is significantly higher than when bundle discount is not offered.6 b) There is no effect of availability of bundle discount when consumers are suggested same usage situations for the two product forms. With respect to conventional product form bundles the predictions on the effect of
In our experiments we have chosen to test a 25% level of bundle discount because this level is commonly used in practice (e.g., Buy a pair of shoes and get the second pair at 50% off). As shown in Equations 2a and 2b, the choice of a bundle is sensitive to the size of the discount, and this conclusion may not hold for very deep discounts. 6 As can be seen from Equation 2a, the appropriateness of this hypothesis is conditional on the discount being large enough to compensate for the bundle not offering an increase in content utility.
5
20
usage situation and bundle discount on purchase likelihood for the bundle are different – we expect a main effect of both usage situation and bundle discount but no interaction. Because consumers receive more content utility when acquiring the bundle, their willingness to pay for the bundle is higher than that for each individual item. If a bundle discount were offered, consumers would perceive the bundling of product forms as a quantity discount, and thus be more likely to buy the product form bundle. If they are suggested to use the two forms in different usage situations (versus same usage situations) VFj|Fi may still increase because the items could be viewed as more complementary. Thus, consumers may perceive positive marginal utility from owning Form j together with Form i, and consequently, be more likely to buy the product form bundle. Hypothesis 3 (conventional product form bundles): a) Consumers are more likely to buy the product form bundle when they are suggested different usage situations for the two product forms as compared to when they are suggested the same usage situations for the two product forms. b) Consumers are more likely to buy the product form bundle when they are offered bundle discount as compared to when they are not offered bundle discount.
Relative Price of Print vs. Electronic Forms Another factor that might influence the purchase likelihood for the product form bundle is the relative price of the two product forms. The more interesting case is the one of the information product form bundle. In this case, because consumers receive just one content utility from the two different forms, the marginal utility having Form j together
21
with Form i is relatively low and sometimes close to zero. Second, electronic information products are unique in that the marginal cost of reproducing and distributing them is often much lower than the cost of the print form, even close to zero (e.g. Bakos and Brynjolfsson 2000). In our empirical work we examine the purchase likelihoods for the bundle when the information product forms are priced differentially and compare the following cases: Case 1 - price of the two forms is equal, and Case 2 - price of the print form is higher than price of the electronic form. Note that the bundle prices are kept constant. To establish a benchmark, assume that the prices of the two forms are equal: Pi = Pj = PA, so the price of the bundle is (1-d)*(2 PA), where d is the discount. Then the bundle will be chosen if VC + VFi+Fj – (1 – d)*(2 PA) > Max (VC + VFi - PA, VC + VFj PA), where VFi + Fj = VFi + VFj|Fi is form utility from buying both items (see footnote 1). If VFi > VFj it follows that the bundle is chosen if VFj|Fi > (1-2d)PA, or if the incremental value of j exceeds its incremental cost. A similar condition can be worked out for the case in which VFj > VFi. Adding the conditional probabilities of these two cases gives the probabilities of bundle choice expressed in conditions 5 and 6 in Table 1, panel B (condition 5 expresses the special case of d = 0). Let the price of item i increase by kPA, and the price of item j decrease by an equal amount, so that the sum of the prices of the two items is still 2PA. Now, because of the shift in relative price, i is more attractive only if VC + VFi – (1 + k) PA > VC + VFj – (1 - k) PA, or if VFi – 2kPA > VFj. If this condition holds, the bundle will be chosen if VC + VFi+Fj– (1 – d)*(2PA) > VC + VFi – (1 + k) PA, which reduces to: VFj|Fi > (1 - k - 2d)PA. A similar expression can be worked out for the case in which j is more attractive (VFi –
22
2kPA < VFj). Adding conditional probabilities of these two cases gives the probabilities of bundle choice expressed in conditions 7 and 8 in Table 1, panel B (condition 7 expresses the special case of d = 0). Comparison of the conditions in panel B of Table 1 leads to predictions about the effects of discounts and relative prices on the acceptance of the bundle. First, because it lowers the incremental cost of the less preferred item, the discount makes condition 6 > 5 and 8 > 7, and there should be a main effect of discount.7 Whether a change in relative prices will lead to increased purchases of the bundle depends on how the relative valuation of the two items varies across consumers. As shown in conditions 7 and 8, when VFi – 2kPA > VFj the probability of buying both items increases because the less valued item becomes cheaper, and it can also be shown that the probability of buying the bundle will increase as long as VFi – kPA > VFj.8 If this condition does not hold, the bundle becomes less attractive when there are different relative prices, and the net effect of changes in relative price on bundle choice depends on the distribution of consumer valuations.9 Given that most consumers employ the print versions of books and newspapers, it seems reasonable to assume as a maintained hypothesis that most consumers place a relatively high valuation on the print version.
To compare conditions 6 and 5, let Fj|i(V’) represent the density of consumer valuations VFj|Fi between 0 and V’, and Fi|j(V’) be a similar density for VFi|Fj. Then the share of consumers who would buy the bundle with a discount, but not otherwise, is (Fj|i(PA)- Fj|i((1-2d)PA))(S) + (Fi|j(PA)- Fi|j((1-2d)PA))(1-S), where S is the share of consumers for whom VFi>VFj. A similar condition can be worked out for comparison of 8 & 7. 8 Let 0 < VFi – VFj < 2kPA, in which case the bundle is chosen if VFi|Fj = VFi+Fj - VFj > (1+k-2d)PA. Let VFi VFj =?, so VFi+Fj - VFi +? > (1+k-2d)PA, or VFj|Fi= (1+k-2d)PA- ?. If ? > kPA, the incremental value of i exceeds its cost, and the bundle becomes more attractive relative to the case of equal prices. If ? < kPA, the incremental value of i is not sufficient to cover its cost. 9 To compare conditions 7 and 5, use the same notation for the density of consumer valuations as in footnote 5, let S’ be the share of consumers with valuations VFi – VFj > kPA, S* be the share of consumers with valuations 0 < VFi – VFj < kPA, and 1- S’ – S* be the share with valuations VFj > VFi. Then the share of consumers who would buy the bundle compared to the case of equal relative prices is (Fj|i(PA)- Fj|i((1k)PA))(S’) - (Fj|i((1+k)PA)- Fj|i(PA))S*- (Fi|j((1+k)PA)- Fi|j(PA))(1-S’- S*). A similar comparison between 8 and 6 could be obtained from subtracting 2d PA from the price terms. Since 8 vs. 6 involves different parts of the price distribution compared to 7 vs. 5, the two comparisons need not give the same results.
7
23
Hypothesis 4 (information product form bundles): a.) Across both discount conditions, if the preference for the traditional (print) version (i) over the electronic version (j) is sufficiently large (VFi - VFj > kPA), consumers are more likely to buy the product form bundle when the traditional print product is priced at a premium and the electronic product is priced at a discount as compared to when the products are priced at the same level. b.) Consumers are more likely to buy the bundle when a discount is offered. The intuition behind Hypothesis 4a is straightforward. The more item prices match consumer valuations, the easier it is to induce consumers to buy both items. If VFi is very high relative to VFj, setting equal prices will make j unattractive, but leave the consumer a surplus on i. It becomes difficult to induce the consumer to buy both items. Conversely if VFi is only slightly higher than VFj, setting a high price on i will make it unattractive, but will leave the consumer with a surplus on j. If VFi is very high relative to VFj, offering different relative prices can have the same effect as offering a bundle discount (compare conditions 6 and 7 in Table 1). However, as they extract more surplus, different relative prices will bring more revenue. We can also present an alternative intuition for Hypotheses 4 from the perspective of consumers’ evaluation process. Recall that under the no discount case, the price of the bundle is the same across the same price (case 1) and differential price (case 2) conditions. Given that consumers have higher preference for the print version, it becomes the anchor for evaluating the bundle (see Yadav 1993). Under the same price condition, consumers have a “gain” on the anchor form (print) in the bundle, while they have a “loss” on the less preferred form (electronic), as the individual prices of the different forms could act as reference points in evaluating the bundle. Since perceived loss has a
24
greater impact than the perceived gain (Kahneman and Tversky 1979), the bundle is still evaluated negatively overall. In the differential price case, the prices are more in line with consumers’ expectations (driven by their valuation) and there is no significant gain/loss on the anchor form and the other form. Thus, the bundle is evaluated more favorably. We argue that the price level of the less preferred form relative to consumer valuation in case 1 (same prices) heightens the poor evaluation of the form and the bundle, while the lower price in case 2 does not have such an effect on the bundle evaluation. Giving a price discount for the bundle can help overcome the negative evaluations and improve the evaluation and purchase of the bundle. In the following sections we present the results of two studies designed to test the hypotheses outlined above.
Study 1 The main objective of this study is to test Hypothesis 1. Additionally, we introduce a bundle discount variable and investigate whether the level of bundle discount affects the purchase likelihood for the product form bundle. Specifically, we are interested in whether the bundle discount per se motivates consumers to buy the bundle. Method Subjects. Subjects were 80 undergraduate students enrolled in an introductory marketing course. They were awarded extra course credit for their participation. Product categories and materials. We used two information product categories and two conventional product categories, specifically books (print and electronic) and journal subscriptions (print and electronic), and margarine (stick and tub) and soap (bar
25
and liquid). Advertisements for fictitious brands were developed to describe these products. Design and procedure. We employed a 2 (level of discount: 0%, 25%) by 2 (category type: information, conventional) between subjects design with 2 product replicates. The order of presentation of the product categories was counterbalanced within each category type. The prices were presented in absolute values (price of form 1, price of form 2, bundle price). Every subject evaluated two product advertisements for either book and newspaper, or margarine and soap. Subjects reviewed each ad and stated their likelihood of purchasing the three options (the bundle and the two individual product forms). Purchase likelihood was measured in two different ways: by allocating 100 points between the three options10, and by indicating how likely they would be to buy the three options (1 = “very unlikely,” 9 = “very likely”). Next the subjects were asked to rate the perceived complementarity of the two product forms using a 3-item Likert scale11 (1 = “strongly disagree,” 7 = “strongly agree”). Finally, several control measures were taken including attitude towards the ad (bad/good, unpleasant/pleasant, worthless/valuable, and unfavorable/ favorable; MacKenzie and Lutz 1989; Mick 1992) and demographics (e.g. gender and computer usage).
Results
10
We do not include the “buy nothing” option in the 100-point allocation among the different product forms because we are interested in which form(s) our subjects are going to choose after they have decided to buy the product. Thus we try to control for other factors that may influence the product purchase decision and are not of specific interest in this study. 11 “It would be more useful to have both the paperback book and the electronic book than just the paperback book”, ”There is additional value in having both the paperback book and the electronic book as compared to having only one of them” and “It would be more useful to have both the paperback book and the electronic book than just the electronic book.”
26
The reliabilities of the perceived complementarity scales are between 0.81 and 0.92 in the four categories (N = 40 per category). The mean levels of perceived complementarity (after averaging the three items of the scale) between the product forms are as follows: XBOOK = 3.21 (St. Dev. =1.67), XNEWS = 3.17 (1.62), XMARG = 3.56 (1.81) and XSOAP = 4.19 (2.10). To test Hypothesis 1 we use linear regression in which the independent variable is the perceived complementarity between the product forms in the bundle, and the dependent variables are the points allocated to the bundle and the likelihood of buying the bundle (separate model for each product category and each dependent variable). Our regression results indicate that in the book, margarine and soap categories there is a positive relationship between the degree of perceived complementarity and the purchase likelihood for the bundle (Likelihood of buying the bundle: bBOOK = 0.535, bMARG = 0.746, bSOAP = 0.731, all p’s<0.01; Points allocated the bundle: bBOOK = 0.466, bMARG = 0.675, bSOAP = 0.626, all p’s<0.01), thus providing support for H1. In the newspaper category the relationship between perceived complementarity and the likelihood of buying the bundle is also positive (bNEWS=0.297, p=0.063), and the relationship between perceived complementarity and points allocated to the bundle is positive but not significant (bNEWS=0.178, p>0.1). In sum, our results generally provide support for Hypothesis 1. We use one-way ANOVA to investigate the effect of bundle discount on the purchase likelihood for the product form bundle. In the book and newspaper categories, there is no significant difference between the points allocated to the bundle and the likelihood of buying the bundle in the discount and no discount conditions (all p’s>0.44).
27
A discount of 25% does not make consumers more likely to buy information bundles. The perceived incremental value of the second form must therefore be less than 50% of the value of the first form. This is to be expected since both forms have the same content with its utility derived in one form or the other, but not in both. In the margarine and soap categories the situation is different – in both categories respondents are more likely to buy the bundle and allocate more points to the bundle in the discount versus the no discount conditions (Likelihood of buying the bundle: XMARG/HIGH = 4.90 versus XMARG/LOW = 2.74, F1,38 =6.20, p<0.05; XSOAP/HIGH = 6.62 versus XSOAP/LOW = 3.42, F1,38 =13.39, p<0.01; Points allocated the bundle: XMARG/HIGH = 32.33 versus XMARG/LOW = 14.47, F1,38 =4.14, p<0.05; XSOAP/HIGH = 43.75 versus XSOAP/LOW = 23.42, F1,38 =4.58, p<0.05). This indicates that, because consuming one form does not reduce the value of the other, consumers perceive the bundling of conventional product forms as a quantity discount and are more willing to buy the bundle if a discount is offered. In general, consumers were significantly more wiling to buy both forms of the conventional products than the information products, supporting Hypothesis 1a.
Discussion On the basis of these results, it appears that the purchase behavior of the consumers is related to the level of perceived complementarity between the different forms in the bundle, and can be positively affected by offering a nominal discount of 25% in the conventional but not in the information product cases. In Study 1 we do not communicate any usage situations to the consumers. We only present them with a short product description and with the item and bundle prices. We use these results as a
28
baseline and design a second study that investigates what happens when we communicate various usage situations to the consumers. Our goal is to show that although the bundle discount itself does not motivate consumers to buy information bundles, it can be effective when different usage situations for the forms are presented.
Study 2 Study 2 tests hypotheses 1a, 2, 3 and 4. In the usage situation manipulation we present consumers with possible usages for the two product forms in the following way: in the same usage situation condition we advertise situations for which the two forms are equally appropriate, while in the different usage situation condition we suggest usage situations for which the two forms are differentially appropriate. Below we describe how the stimuli were developed. Stimuli Development Product categories. As in Study 1, we use two information and two conventional product categories - book and newspaper, and margarine and coffee, respectively. Usage situations. We employed extant substitution-in-use (SIU) methods to generate usage situations appropriate for each product form (see Srivastava et al. 1984). The SIU approach (Stefflre 1971) is an iterative procedure for constructing product specific usage-situational taxonomies. Using the SIU approach, the perceived similarity between the different alternatives for the usage situations of interest and the resulting product-market structures can be investigated (Ratneshwar and Shocker 1991; Srivastava, Leone and Shocker 1981). In our study, first, a sample of consumers (N = 15) generated a set of usage
29
situations for the two forms in each product category. Then, a second sample (N = 10) evaluated the appropriateness of each product to each usage situation on a yes/no/don’t know scale. Finally, a structured questionnaire with product forms and usages was developed and administered (N = 67). The subjects judged the appropriateness of each product form for each usage situation on a 5-point Likert scale (not appropriate to very appropriate) including a “don’t know” option. << Insert Table 2 about here >> The perceived appropriateness of the generated usage situations for each product form is presented in Table 2. For book, newspaper, and coffee product categories, one of the two forms is perceived as more appropriate for some usage situations and the other product form is perceived as more appropriate for other usage situations. For example, in the book product category, a print book is perceived as more appropriate than an electronic book for giving as a present (XPRINT – XELECT = 2.24, p<0.05), reading for pleasure (MD = 1.76, p<0.05), reading while traveling (MD = 1.74, p<0.05) and reading it to other people (MD = 1.45, p<0.05). On the other hand, an electronic book is perceived as more appropriate than a print book for e-mailing pages/chapters (MD = 2.43, p<0.05), searching (MD = -0.95, p<0.05) and copying citations/paragraphs (MD = 0.60, p<0.05). Both product forms of the above three categories are perceived as equally appropriate for several usage situations. For example, in the book product category, the perceived appropriateness of getting them on a short notice and using them for archival purposes are not significantly different across product forms. Materials. Advertisements for fictitious brands were used to describe these products and manipulate the usage situations. We selected the usage situations in which
30
one product form is perceived as significantly more appropriate than the other for designing the different usage situation manipulation, and the situations in which the product forms are equally appropriate for designing the same usage situation manipulation. The number of usage situations is balanced across conditions and everything else in the advertisements is kept constant. Thus, we presented four usage situations in total in each advertisement: in the same usage situation condition we listed four situations in which the two forms are equally appropriate, while in the different usage situation condition Form 1 is more appropriate for two of the usage situations and Form 2 is more appropriate for the other two usage situations. With respect to the margarine product category, there was no significant difference in how the two product forms were perceived for most of the situations. Therefore, we used the same four usage situations to manipulate the same/different conditions, just changing the wording – stating that form 1 is more appropriate for two of the usage situations and form 2 is more appropriate for the other two usage situations in the different usage situation condition, while stating that the forms are appropriate for all four situations in the same usage situation condition. We consider this a more conservative manipulation and elaborate on this in the results section.
Pretest 1 We pretested the two advertisements for each of the four categories using a between-subjects design in which each subject saw only one advertisement (N=160). Based on MacKenzie and Lutz (1989) and Mick (1992), we use eight semantic differential scales to measure attitude towards the ad (bad/good, unpleasant/pleasant,
31
worthless/valuable, and unfavorable/favorable) and ad credibility (unconvincing/convincing, biased/unbiased, unbelievable/believable and noninformative/informative). Across the categories, the alphas varied between 0.85 and 0.93 for attitude towards the ad, and between 0.74 and 0.85 for ad credibility. The results showed that the attitudes towards the ads and the ad credibility were not significantly different across the manipulated same/different usage situation conditions. Therefore, differences in responses to the advertisements across conditions cannot be attributed to differences in attitudes towards the ad/ad credibility.
Pretest 2 We performed a second pretest of our stimuli (N=75) to assess the extent to which the product forms were perceived as complements using the same complementarity scale as in Study 1. The results indicated that our subjects viewed the product forms as more complementary in the different usage situation condition (vs. the same usage situation condition) in the book, newspaper, coffee and margarine product categories (XBOOK/DIFF = 4.32 versus XBOOK/SAME = 3.00, F1,35 =5.21, p<0.05; XNEWS/DIFF = 4.25 versus XNEWS/SAME = 2.96, F1,35 =5.33, p<0.05; XMARG/DIFF = 4.77 versus XMARG/SAME = 3.48, F1,36 =4.33, p<0.05; XCOFF/DIFF = 5.46 versus XCOFF/SAME = 4.25, F1,36 =5.31, p<0.05). Consequently, we decided to proceed with the study.
Method Subjects. 406 undergraduate students enrolled in several marketing courses participated in the experiment (N = 240 in the information bundle categories, and N=166
32
in the conventional categories)12. They were awarded extra course credit for their participation. Design and procedure. We employed a 2 (usage situations: same, different) by 2 (level of discount: 0%, 25%) by 2 (category type: information, conventional) between subjects design with 2 product replicates within each category type (book and newspaper, and margarine and coffee). Additionally, for the information category type, we introduce another between subjects factor, relative price, which was varied so that the print form price was either equal to the electronic form price or higher than the electronic form price. The prices were presented in dollar amounts. We used the same experimental procedure and measures as in Study 1.
Summary of Study 2 Results As in study 1, respondents allocated 100 points between the individual items and the bundle to indicate their likelihood of purchasing each alternative. Table 3 presents the average share of the points that were allocated to the bundle in each of the experimental conditions; these shares can be interpreted as average probabilities of choosing the bundle. In accord with Hypothesis 1a, the table shows that the average probability of choosing conventional bundles is considerably higher than the average probability of choosing information bundles in all conditions. Consistent with Hypothesis 2, the discount and different usage conditions only have a major effect on the probability of choosing information product bundles when they are combined. On the other hand, consistent with Hypothesis 3, the discount and different usage conditions are both associated with higher choice probabilities for conventional bundles across all conditions.
Our use of the additional relative price manipulation necessitated a larger sample for the information products.
12
33
Finally, the high price for print and low price for electronic items is associated with a higher incidence of preference for the bundle, as predicted by Hypothesis 4, only in the no discount case. We present a more complete analysis of these results, including formal hypothesis tests, in the following section. << Insert Table 3 about here >>
Detailed Results We first discuss the manipulation check, and then present results of hypothesis tests for conventional products, followed by information products. Manipulation check. The reliabilities of the perceived complementarity scales were between 0.82 and 0.92. We ran one-way ANOVAs for each product category with USAGE as the between-subjects factor and perceived complementarity as the dependent variable. In all four categories the perceived complementarity in the different usage situations condition was significantly higher than the perceived complementarity in the same usage situations condition (USAGE: XBOOK/DIFF = 3.84 versus XBOOK/SAME = 3.32, F1,239 =4.33, p<0.05; XNEWS/DIFF = 3.99 versus XNEWS/SAME = 3.46, F1,238 =4.46, p<0.05; XMARG/DIFF = 4.65 versus XMARG/SAME = 3.86, F1,78 =4.27, p<0.05; XCOFF/DIFF = 4.56 versus XCOFF/SAME = 3.81, F1,84 =5.16, p<0.05). Thus, we successfully manipulated the same/different usage situations variable and could proceed with the analyses. The results of the analyses employed in testing Study 2 hypotheses are presented in Table 4 (GLM procedure results) and Figure 1 (the interactions in a graphical form). Since it is bounded at zero and 100, the measure of points allocated to the bundle employed as our dependent variable may not be normally distributed. Accordingly we
34
applied a logit transform in which the dependent measure is defined as ln ((points + .5)/(100 – points + .5)) in conducting hypothesis tests.13 << Insert Table 4 and Figure 1 about here >> Conventional products. According to Hypothesis 3 we expect a main effect of both usage situation and bundle discount on likelihood of buying the bundle. As anticipated, two significant main effects were found in the analysis of the margarine and coffee data - USAGE (margarine: F1,76 =6.01, p<0.05; coffee: F1,82 =4.09, p<0.05) and DISCOUNT (margarine: F1,76 =9.24, p<0.01; coffee: F1,82 =23.65, p<0.05). The two-way interaction between USAGE and DISCOUNT is not significant in both product categories (p>0.1). Consequently, our results in the margarine and coffee categories provide support for Hypotheses 3a: consumers are more likely to buy the product form bundle when they are suggested different usage situations for the two product forms as compared to when they are suggested the same usage situations for the two product forms, and for Hypothesis 3b: consumers are more likely to buy the product form bundle when they are offered a bundle discount as compared to when they are not offered a discount. In the margarine case just stating that one form was more appropriate than the other for a given situation was sufficient to produce the desired manipulation. In the sense that it is not reinforced by a general perception that the different forms are more appropriate for different usage situations, this manipulation can be regarded as conservative. Evidently it is possible to create a perception of appropriateness for a given usage situation by simply presenting information that this is the case.
13
The addition of .5 makes it permissible to take logs in cases where points = 0 or 100. While we employed the logit transform, models using the raw points variables gave very similar results.
35
While not shown in Table 4, we also tested the hypothesis of equal choice probabilities for electronic and information products. As one might expect from the large differences in average choice probabilities between conventional and information products in Table 3, this hypothesis was rejected (F1,669 = 76.69, p<0.001). Information products. According to Hypothesis 2 we expect the usage situations considered when making a purchase decision to moderate the effect of bundle discount on purchase likelihood. To test this hypothesis, we use the GLM procedure in SAS in which the dependent measure is the points allocated to the bundle, and USAGE, DISCOUNT and RELATIVE PRICE are the between subjects factors. As anticipated, a significant USAGE by DISCOUNT interaction was found in both categories (F1,232 =6.03, p<0.05 for book, and F1,232 =3.91, p<0.05 for newspaper). The first planned contrast revealed that when the consumers are presented with different usage situations, they are significantly more likely to buy the bundle in the discount versus no discount condition (Fcontrast1,232 =13.28, p<0.01 for book, and Fcontrast1,232 =13.64, p<0.01 for newspaper), and the second planned contrast revealed that when consumers are suggested the same usage situations there is no significant difference in their purchase likelihood in the discount as compared to the no discount condition (p’s>0.37). Consequently, our results provide support for Hypotheses 2a and 2b. Finally, the last planned contrast (testing Hypothesis 4) showed that there is no significant difference between the likelihood of buying the bundle when the traditional print form is priced at a premium and the electronic form is priced at a discount, and the likelihood of buying the bundle when the traditional and the electronic forms are priced at the same level (p’s>0.42). This is consistent with the data in Table 3, which show similar
36
choice probabilities for both conditions. A post hoc contrast comparison revealed that our prediction is valid in the case of no bundle discount in the book category (Fcontrast1,232 =3.78, p=0.05): as shown in Table 3, the average probability of buying both print and electronic books is .1195 when the electronic form has a lower price, but only .0610 when prices are equal. However a similar comparison was insignificant in the other cases. The results for Hypothesis 4 can be explained as follows. Recall that our test of Hypothesis 4 was based on the assumption that the print form is valued significantly more than the electronic version. Specifically we assumed VFi - VFj> kPA, where k is increase in price of the print version (decrease on price of the electronic) as a proportion of price. In our study k = .25, so the print version must be worth at least 25 percent more than the electronic form as a proportion of the average price to make the bundle more attractive. However, a substantial proportion of our respondents would likely choose the electronic version at equal prices for print and electronic: at equal prices the average number of points allocated to the electronic book was 37 of 100; the average for the electronic newspaper was 40 of 100 (these averages did not differ much between discount conditions). In the no discount condition, the corresponding averages for print were 56 and 53 respectively.14 These results suggest that VFi and VFj (the utilities derived from the form per se) are not very different for many consumers. If so, since content utility can be obtained from any of the alternate forms, these consumers are likely to be better off buying just the electronic version when it has a low price rather than buying the bundle, which is what happened in our study. At a low price for the electronic version, the proportion buying only electronic form increased by about 10 percent for books, and 13 percent for newspaper. These consumers likely did not choose the bundle because the
14
In the discount condition, about ten percent more respondents chose the bundle rather than print only.
37
print version became too expensive relative to its incremental value. Thus our test of Hypothesis 4 appears to fail because the electronic form is valued more than we anticipated. Additionally, an interaction between the content and form utility may explain the results associated with Hypothesis 4.
Discussion The results of our study provide useful insights into how consumers react to product form bundles. In the context of information products, since consumers obtain content utility only once in buying the bundle (content utility is fixed going from either of the form to the bundle), the product forms tend to be considered overall more as substitutes. Even with differing degrees of perceived complementarity/contingency across our subjects, the purchase probabilities of the bundle do not increase with just price discounts on the bundle. The bundle purchase probabilities increase only when price discounts are accompanied by communication about different usage situations to impact the consumers’ perceived complementarities between the two forms in the bundle. These results may be more specific to the valuations of our student respondents, who we observed not to value form utilities very much in comparison to content utility, but they also point to the fact that different usages/usage situations of the forms (and form utilities) have to be emphasized clearly in order to highlight the value of the information product bundle, no matter which consumer group we market the bundle to. The important take-away is that such communication strategies help significantly in the case of information product bundles and, thus, marketers have a viable strategy to harvest the potential extra revenues in selling bundles.
38
In the case of conventional product form bundles on the other hand, the absolute probabilities of purchasing bundles are much higher as compared to information product bundles, even though the forms are perceived as strong substitutes. This is not surprising as consumers obtain content utilities from both forms in the bundle (content utility increases in going from either of the form to the bundle), and hence, in some sense, it is not a proper comparison. Price discounts for conventional product bundles increase the purchase probabilities of the bundles as they work as quantity discounts, and communication about different usages also seem have to a positive impact. An interesting observation from our study is that, although the product forms seem to be moderate/strong substitutes, bundle purchase probabilities are significantly high enough to make mixed bundling a viable and profitable strategy in both categories, albeit for different reasons. While in the case of conventional product form bundles, mixed strategies could be arguably optimal in the presence of quantity discounts, in the case of information product form bundles, mixed strategies can be very profitable given the very low marginal cost of the additional form. In both categories we observe that the communication strategies play an important role in supporting mixed bundling strategies. In this regard, our results are somewhat different from the traditional bundling results, which suggest a pure component strategy in the presence of strong substitutes and low marginal costs (Venkatesh and Kamakura 2003), and provide additional insights for the case of product form bundles. Our results also show some support for differentially pricing the information product forms. While we show that this is better than pricing the forms equally if consumers have a relatively high preference for one item as compared to the other, in our
39
experiment we could not enforce this condition. Nevertheless, the hypothesis was supported in the case of books under the no discount case. An interesting interpretation of the result is that the prices of the individual product forms could play an important role in bundle evaluation. If the price of the less preferred form is more than the consumers’ valuation for the form, there could be a magnified negative effect on the bundle evaluation. This implies that price information can have similar effects as non-price information studied in extant literature (e.g., Yadav and Monroe 1993). While this notion needs to be tested more formally, it does highlight the importance of pricing the individual product forms appropriately in a mixed bundling strategy.
Managerial Implications How should companies market information bundles? In practice, online publishers use various product and pricing strategies. For example, Wall Street Journal offers its online edition at 40% of the price of the print edition, and the discount for choosing the bundle is 17%. Business Week and Fortune offer free online content for their print subscribers only, and Newsweek and Washington Post offer their online editions for free to any interested readers. Finally, National Academies Press, a publisher selling books online, offers electronic copies for most of its books at about 75% of the print book price, and an average bundle discount of 40%. In general, the prevailing marketing practice in the case of information products has been to price differentially the various forms and to offer bundle discounts. With the exception of WSJ, none of these organizations stresses different usage situations for the print and electronic forms. However, our results indicate that information providers should try to
40
communicate different usages for their product forms to make them to be perceived as more complementary forms rather than just substitutes. We provide evidence that discounts commonly used in practice work only when different usage situations for the product forms are communicated. The message that we would like to convey is that stressing different usages for the product forms may be needed to motivate consumers to buy the bundle even when a discount is present. We find that communication of different usage situations, or other attempts to enhance the value of alternative forms of information products can increase the salience of the form utilities and may be needed because of the inherent duplication of content that is provided by the different forms of information products. If consumers do not perceive the form utilities to be very significant, then the content utilities can render the forms to be very strong substitutes. Something has to be done to induce buyers to purchase a second form in the face of this perception of substitutes due to content. One strategy is to make the functionalities of the forms very different and educate the consumers of the relative usage situations of the different forms. This will make the form utilities much more significant. The other option is to change the content of the two forms – for example, online newspapers are updated often during the day and thus the content is more dynamic as compared to the static content of the print form. Similarly, electronic versions of books can be updated and distributed to consumers for free. Electronic versions can also provide links to other relevant content that print versions cannot provide. However, this will increase the cost of the electronic form, and the cost side should be taken into account before considering such a strategy. Our results also suggest that if one form is significantly preferred much more than
41
the other, charging a high price for the more preferred form, and a low price for the less preferred form may be an effective strategy. This might well be as effective as giving discounts while providing higher revenues to the marketer. These results have particular relevance in the bundling of online content such streaming video and DVDs, online music, CDs, and DVD-Audio, where it is not clear how consumers’ valuation for the different forms are distributed. They also suggest as the online channel matures and quality of digital content improves, differential pricing strategies and discounting strategies should change significantly.
Conclusion The study contributes to the marketing literature in several ways. To our knowledge, this is the first attempt to study the effects of product form bundling on consumer preferences and consumption behavior. We provide a useful approach using the perspective of content utility and form utility, and we model the effect of both price and non-price information in bundle evaluation process and draw conclusions about their impact. We also introduce a usage situational perspective in studying bundling issues. The SIU approach has been proven successful in investigating the influence of usage situational variables on consumers’ behavior and is helpful in the present setting as well. Our study complements the findings in extant literature on bundling by providing insights into conditions under which those findings are directly applicable to product forms and instances where those results may not hold. We show that the perceived complementarity among the bundle components is extremely important and suggest how product form bundles might be marketed. Finally, we provide evidence that the conventional and
42
information product form bundles are different from a conceptual point of view and discuss a mechanism that can explain this difference. Several possible ways for extending the present study are worth mentioning. First, explaining the distinction between conventional and information product form bundles is worth future effort. Consumers may see bundling of conventional product forms as a quantity discount, while bundling of information product forms as bundling of complementary items. Second, it will be interesting to see how exactly the consumers evaluate product form bundles. Some authors suggest an anchoring and adjustment model for bundle evaluation (i.e. Yadav 1994), while others argue for an averaging model whereby component ratings are balanced into an overall evaluation (i.e. Gaeth et al. 1990). This issue is especially important in designing and pricing of electronic products and bundling them with traditional products. Different pricing and price presentation strategies should be employed in the above cases of bundle evaluation. A third interesting research issue comes from the fact that the digital form offers the potential of an augmented version of the traditional form. Consequently, companies may be able to increase the level of complementarity of the forms in information bundles by somehow changing or improving the electronic form, and thus positively influence bundle sales. Finally, it is also important to support/verify our findings using empirical data from online content providers. Future research should address and clarify such strategies.
43
CHAPTER III: ESSAY TWO. BUNDLING AND PRICING STRATEGIES FOR DIGITAL PRODUCTS Abstract Content providers such as publishers of books, newspapers and magazines have started to offer products in electronic form, and even individual sections, in addition to their print products. For example, some book publishers offer print and PDF books, while others offer print books and individual PDF chapters. Which strategy is more attractive for a company to pursue? We model the profit under these two strategies and outline conditions in which one or the other leads to higher profit. We apply our analytical framework to data from a field experiment implemented on the website of a book publisher.
44
Introduction
“U.S. consumer spending for online content grew 14% in 2004 reaching an all-time annual high of $1.8 billion…While there is no doubt that the market remains strong, with only 11.6% of the Internet population purchasing content online in Q4 2004, there is still significant room for growth .”
(Michael Zimbalist, President, Online Publishers Association15)
The emerging technological solutions and the Internet channel unequivocally change the marketing of information. In the music category, for example, tapes and CDs give way to music downloads and online subscription services, and new formats as those introduced by dualdisc.com allow content providers to sell music in CD format plus DVD-Audio or surround sound as a bundle on the same disk. A similar trend is observed in the publishing industry where print products such as newspapers and magazines are offered in electronic form too, and consumers are even encouraged to buy both forms (e.g., Wall Street Journal). The electronic products can be unbundled at no extra cost into separate sections to be sold individually or re-bundled with the traditional print products. Although not yet common, publishers do offer full and unbundled electronic content in addition to print content, allowing customers to buy, for example, individual PDF chapters or entire PDF books and reports separately or together with the print editions (e.g., US Government Printing Office). While appealing because of low marginal costs and the likely catering to emerging consumer needs, selling electronic products may lead to additional revenue generation but can also result in product cannibalization. The National Academies Press (NAP), for instance, publishes over 200 books a year on a wide range of topics in science, engineering, education and health, including full length books, reports and reference materials. In 1996 NAP started posting most of
15
Online Paid Content U.S. Market Spending Report 2004, Online Publishers Association, March 2005, www.online-publishers.org
45
its titles online as low (fax) quality content free for anyone to browse, search and sample. Although majority of the consumers continued buying print books, many website visitors began utilizing the online capability to download chapters, some even demanding higher quality electronic content at a price. Thus, NAP decided to consider two alternative product line strategies – on one hand, introducing books in PDF form in addition to the current print books; on the other hand, marketing individual PDF chapters with the print books. It is in this context that we execute our study. In this paper, we attempt to model the profits under two feasible strategies a content provider may employ: traditional mixed bundling, in which the product line consists of print book, PDF book and the bundle of the two, and content unbundled mixed bundling, which is selling print book, PDF chapters and the combination of the two. Although traditional mixed bundling is a special case of content unbundled mixed bundling in which consumers are constrained to buy all the chapters (instead of one or more chapters), we model them separately in order to provide better insights on what determines the optimal prices and profits. Our goal is to provide insights on which strategy is more beneficial for a company to pursue and how the electronic content should be priced to extract optimal profit. Further, we empirically test our analytical predictions with data from an online experiment executed on the website of NAP. Our study contributes to the marketing theory and practice in the following manner. While there is an extensive body of research suggesting that mixed bundling is the optimal strategy when there is asymmetry in the reservation prices for the bundle components (e.g. Adams and Yallen 1976; Schmalensee 1984; McAfee, McMillan and Whinston 1989), few studies focus on cases in which different mixed bundling strategies
46
are feasible and evaluate their attractiveness. Also, we look at form bundling and content unbundling simultaneously, which makes our approach unique and adds to the extant literature. Finally, while previous studies often treat bundle components as either complements or substitutes or unrelated products, we allow heterogeneous contingent valuations of one component given the other in the population. From a substantive point of view, we show empirically that the content unbundled mixed bundling strategy is more profitable than the traditional mixed bundling strategy under certain conditions. Our findings have significant practical implications, recommending how to profitably design and price electronic content. The paper is structured as follows. First, we review related research and position our study. Second, we model the profits under traditional and content unbundled mixed bundling strategies, and derive which one is better under various conditions. Third, we report the empirical results and conclude with the implications of the study to the extant literature and practice.
Research Background Bundling is marketing two or more products and/or services as a package at a special price (Guiltinan 1987). Demand side incentives favoring bundling include negative correlations in reservation prices (Stigler 1963; Schmalensee 1984), complementarity in consumption (Telser 1979), and uncertainty in the valuations of the quality of the goods (Kenney and Klein 1983). With respect to supply side incentives, large scale bundling of information goods, for example, can be profitable because it creates economies of aggregation when their marginal costs are low (Bakos and
47
Brynjolfsson 1999; 2000). Companies can employ a pure bundling strategy whereby only the bundle is offered, or a mixed bundling strategy, in which the bundled items are also sold separately. Extant research has looked at product and pricing strategies for bundles, including optimal bundle price and composition (Hanson and Martin 1990; Venkatesh and Mahajan 1993), how the degree of heterogeneity in the reservation prices affects optimal bundle pricing (Jedidi et al. 2003) and conditions favoring bundling/unbundling of industrial systems (Wilson, Weiss and John 1990). In terms of contingent valuations in bundling decisions, studies have compared interrelated with independently valued products in a bundle. For example, Venkatesh and Kamakura (2003) model the optimal bundling strategies for interrelated products under monopoly and suggest that moderate or strong substitutes should be offered separately; the same is applicable for complements if the marginal costs are moderate relative to the market’s maximum willingness to pay. A seller would gain by mixed bundling for weak substitutes/complements if the variable costs are not too high. Regarding information goods (e.g., music, weather forecasts, websites), large scale bundling is approximately optimal if consumers’ values of subsequent goods do not decrease too quickly; otherwise pure bundling is optimal even when there are strong negative or positive correlations of values across goods (Geng et al. 2005). To summarize, the literature on bundling identifies the conditions for profitable bundling and specifies optimal bundling strategies in various situations. What is missing in this stream of research is how to handle situations in which different mixed bundling strategies are feasible. With respect to information products, companies can implement
48
both traditional and content unbundled mixed bundling strategies, and which one is more profitable does not have a straight-forward answer. Different consumers may have different contingent valuations for the form bundles, and different contingent valuations for the unbundled electronic units (e.g., chapters within a book). Consequently, the existing premises on optimal bundling strategy and pricing can not be applied directly. Our predictions and empirical findings are relevant and insightful from both theoretical and managerial point of view.
Bundling Model In our model the seller is a monopolist who offers a wide selection of book titles as hardcover or paperback books. Subject areas are very diverse, ranging from agriculture and education to medicine and engineering. The product and pricing strategy the publisher is employing at the moment is segment-based and reflects the needs, buying power and quality expectations of the consumers in the different subject domains. The publisher is currently choosing between two possible product line strategies: 1) Content Unbundled Mixed Bundling - selling PDF book chapters individually and together with the print books, and 2) Traditional Mixed Bundling - selling full PDF books individually and together with the print books. Next, we express optimal pricing of electronic content and profit under each strategy, and then compare the two profits.
Content Unbundled Mixed Bundling Model basics
49
We assume that a book has J chapters and PC is the price per chapter
(P1 = K = PJ
= PC ) . V ji is the valuation of chapter j
( j = 1K J ) by consumer i
(i = 1K I ) . The chapters are rank-ordered in terms of their valuation starting from the
highest, e.g. V1i ? V2i ? ... ? V Ji . The number of chapters a consumer buys depends on the valuation of the chapters, namely consumer i buys chapter j if V ji ? PC and V j +1,i < PC . Thus, consumer i will buy exactly QC rank-ordered chapters if VQi ? PC and VQ +1,i < PC .
Revenue from selling electronic chapters In this section we first express the revenue from selling electronic chapters at individual level and then the aggregate the electronic chapters revenues across consumers. At this point we only consider offering electronic chapters to consumers; the print book sales and the associated revenue and costs are introduced in the next section. The optimal price per chapter and the total and marginal revenues from chapters at individual level are graphically displayed in Figure 2. << Insert Figure 2 about here >> We assume that the demand for chapters is linear. The valuation for the chapters, VC , is presented on the y-axis, and the cumulative number of chapters at every valuation point, QC , is presented on the x-axis. N C is the number of chapters with positive valuation (e.g., the consumer will buy N C chapters if the price is zero). The optimal price per chapter PC is equal to the valuation of the marginal chapter by the consumer. Let V Hi be the valuation of the chapter with the highest valuation for a consumer. The quantity of chapters bought and the marginal revenue from chapters can be expressed as follows:
50
(1)
QC = N C ? PC
NC VHi
(2)
MRC =
? (QC PC ) = ? PC
? ?? ? N C ? PC
??
??
NC VHi
? PC
? ? ? ? PC ? N ? ? = N C ? 2 PC C = 0 VHi
1 Solving equation 2 for PC , the optimal price per chapter is PC* = V Hi and the optimal 2
1 * revenue is RC = V Hi N C . Thus, the optimal price depends on the valuation of the highly 4
valued chapter, while the optimal revenue depends on both the valuation of the highly valued chapter and the number of chapters with positive valuation for the consumer. Aggregated across consumers, the number of chapters bought and the marginal revenue from chapters are expressed below: (3)
I I I ? N Ci N Ci ? ? ? = ? = ? N P Q N P ? ? ? Ci C? Ci C ? Ci ? VHi ? i =1 i =1 VHi i =1 ? i =1 I
(4)
? i =1 MRCi = ?
I
? ?? ? QCi ? PC ?
?
??
I
?
? ?=
I
? ?? ? ? N Ci ? PC ?
?? i =1
i =1
??
I
I
N Ci VHi
? PC
? PC
? ? ? ? PC ? I I N ? ? = ? N Ci ? 2 PC ? Ci = 0 i =1 i =1 V Hi
Let N = ? N Ci k F and S = ?
i =1
i =1
N Ci k F , where k is the number of potential VHi
consumers on the market and F is the fraction of consumers for which VHi ? PC . Then
N is the average demand for chapters at zero price for the customers who buy at least one
chapter, and S is the average slope of the demand for the customers who buy at least one chapter. Thus, the optimal price per chapter is PC* =
1 N S and the optimal revenue 2
51
* is RC =
1 (N )2 S . Thus, the higher the average demand for chapters at zero price for the 4
customers who buy at least one chapter, and the lower the average slope of the demand for the customers who buy at least one chapter, the higher the optimal price and revenue. This means that companies can charge a relatively higher price per chapter when consumers value positively relatively more chapters, and when the decrease in valuation from the most preferred to the least preferred chapter is relatively slow.
Revenue from selling electronic chapters and the print book Here we add the revenue from selling the print book and the associated costs. The price of the print book is assumed exogenous to the model as it was set when the book was initially introduced to the market. At present the company is deciding on whether to offer the electronic chapters in addition to the print book and at what price. Consumers do not receive an extra discount for buying the print book plus electronic chapters bundle and the bundle price is simply the sum of the two individual prices, PP and QC PC respectively. Consumers can buy the print book, one or more electronic chapters, or the print book plus one or more chapters. The profit function under content unbundled mixed bundling can be expressed as follows: (5)
I ?? I N Ci CUMB N P ? CUMB = PP Q P + ?? ? C? ? ? Ci i =1 V Hi ?? i =1
? ? CUMB ? ? FC , ? PC ? ? cQ P ? ?
CUMB where QP is the number of print books sold under CUMB, c is the print book variable
cost and FC is the fixed cost. We assume that the variable cost of electronic chapters is equal to zero. Differentiating the profit with respect to the chapter price, the optimal chapter price and profit are as follows: 52
(6)
CUMB I I ? ? CUMB N ? QP = ( PP ? c) + ? N Ci ? 2 PC ? Ci = 0 ? PC ? PC i =1 i =1 V Hi
(7)
* = PC
1 2 kFS
CUMB ? ? ? QP + ? ( ) k F N P c ? ? P ? PC ? ?
(8)
?
* CUMB
CUMB ? ? QP 1 ?? I ? ? = ??? N Ci ? ? ?( PP ? c) ? 4 ?? i =1 ? PC ? ? ? ? 2
2
? ? ? ?
?V
i =1
I
N Ci
Hi
CUMB + ( PP ? c) QP ? FC
CUMB 2 ? ? ? ? QP 1 ? 2 CUMB ?(k F N ) ? ?( PP ? c) = ? FC ? ? + ( PP ? c) QP 4k F S ? ? PC ? ? ? ? ?
The optimal price per chapter and profit under content unbundled mixed bundling is affected by several parameters. First, the higher (lower) the average demand for chapters at zero price for the consumers who buy and the smaller (bigger) the average slope of demand for the consumers who buy, the higher (lower) the optimal price and profit. From the consumers who enter the market for chapters, if more consumers have positive valuation for the chapters in a book and if the decrease in valuation from the most preferred to the less preferred chapter is relatively small across these consumers, publishers can charge higher price per chapter and achieve higher profit level. Second, the optimal price per chapter and the optimal profit also depend on how much the demand for print books is affected by a change in the price of the electronic chapters. If an increase in the chapter price leads to an increase (decrease) in the print book sales, the optimal chapter price is higher (lower) as compared to the case when a change in the chapter price does not affect print book sales. In terms of optimal profit, it is the magnitude of the rate of change of the print book demand function given a change in the chapter price that is important, not the direction – the more a change in the chapter price
53
affects the print book sales, the lower the profit as compared to the case when a change in the chapter price does not affect print book sales.
Traditional Mixed Bundling Strategy Model basics If the electronic chapters in a book are bundled and only the full electronic book is offered, consumer i will buy the full electronic book only if the valuation of the electronic book Vi exceeds its price PEB ( Vi = ? V ji ? PEB ). The aggregate electronic
j =1 J
book revenue is equal to PEB k [1 ? F ( PEB )] , where F ( PEB ) is the fraction of consumers for whom the valuation for the electronic book is smaller than the price of the electronic book ( Vi < PEB ) and k is the number of potential customers on the market.
Revenue from selling the full electronic book In this section we express the revenue from selling the full electronic book. Selling the print book is not yet considered in the model. The optimal price for the electronic book and the total and marginal revenues are graphically presented in Figure 3. << Insert Figure 3 about here >> The valuation of the electronic book by the different potential consumers, VEB , is presented on the y-axis, and the number of potential consumers, k , is presented on the xaxis. The optimal price, PEB , is equal to the valuation of the electronic book by the marginal consumer. The quantity of electronic books bought and the marginal revenue from electronic books can be expressed as follows:
54
(9)
QEB = k F (VEB ? PEB )
(10) MREB =
? [k F (VEB ? PEB ) PEB ] ? [1 ? F (VEB < PEB )] = k [1 ? F (VEB ? PEB )]+ k PEB =0 ? PEB ? PEB
Without loss of generality, we assume that the valuations of the consumers for the full electronic book are uniformly distributed between VL and VH, and that VL= 0. Consequently, 1 ? F (VEB ? PEB ) = 1 ? F ( PEB ) = 1 ?
PEB VH ? PEB = and VH VH
? [1 ? F ( PEB )] 1 =? . Solving equation 10, the optimal price of the full electronic book ? PEB VH
k VH 1 * * is PEB = . Thus, the higher the upper bound = V H and the optimal revenue is REB 4 2 of the electronic book valuation in the population, the higher the optimal price and revenue. The revenue also depends on the number of potential consumers.
Revenue from selling the full electronic book and the print book Again the price of the print book is assumed exogenous to the model as it was set when the book was initially introduced to the market. At present the company is considering whether to offer the electronic book in addition to the print book and at what price. Consumers do not receive extra discount for buying the print book plus electronic book bundle and the bundle price is just the sum of the two individual prices, PP and PEB respectively. Consumers can buy the print book, the electronic book or the bundle of the two books. The profit function under traditional mixed bundling can be expressed as follows: (11)
TMB TMB ? TMB = PP QP + k [1 ? F ( PEB )] PEB ? c QP ? FC ,
55
TMB is the number of print books sold under TMB, c is the print book variable where QP
cost and FC is the fixed cost. We assume that the variable cost of the electronic book is equal to zero. Differentiating the profit with respect to the electronic book price, the optimal electronic book price and profit are as follows: (12)
TMB ? ? TMB ? QP ? [1 ? F ( PEB )] = ( PP ? c) + k [1 ? F ( PEB )]+ k PEB =0 ? PEB ? PEB ? PEB
(13)
* = PEB
VH 2k
TMB TMB ? ? VH VH ? QP ? QP + ? ( ) = + ( ? ) k P c P c ? ? P P ? PEB ? 2 2 k ? PEB ?
(14)
?
* TMB
TMB 2 ? ? VH ? 2 ? ? QP TMB ?k ? ?( PP ? c) = ? ? + ( PP ? c) QP ? FC = 4k ? ? PEB ? ? ? ? ?
2
TMB ? ? QP k VH VH ? TMB = ? ( ? ) P c ? P ? + ( PP ? c) Q P ? FC 4 4k ? ? PEB ?
Similar to the content unbundled mixed bundling case, the optimal electronic book price and profit under traditional mixed bundling is affected by several parameters. First, the higher (lower) the upper bound of the electronic book valuations distribution, the higher (lower) the optimal price and profit. If consumers have relatively high valuation for the electronic book, publishers may charge higher electronic book price, resulting in higher profit. Second, the optimal electronic book price and the optimal profit also depend on how much the demand for print books is affected by a change in the electronic book price. If an increase in the electronic book price leads to an increase (decrease) in the print book sales, the optimal electronic book price is higher (lower) as compared to the case when a change in the electronic book price does not affect print book sales. In terms of optimal profit, similar to the optimal profit under CUMB, it is the
56
magnitude of the rate of change of the print book demand function given a change in the electronic book price that is important, not the direction – the more a change in the electronic book price affects the print book sales, the lower the profit as compared to the case when a change in the electronic book price does not affect print book sales.
Comparing TMB with CUMB
In this section we compare the optimal profit under the two bundling strategies. The fixed costs and the variable costs per print book are assumed to be equal across the two conditions. After subtracting (12) from (8) and doing some regrouping we get: (15)
?
* CUMB
??
* TMB
CUMB 2 ? ? ? ? QP 1 ? 2 CUMB ?(k F N ) ? ?( PP ? c) = ? FC ? ? + ( PP ? c) QP 4k F S ? ? PC ? ? ? ? ?
TMB 2 ? ? VH ? 2 ? ? QP TMB ?k ? ?( PP ? c) ? ? ? ? ( PP ? c) QP + FC = 4k ? ? PEB ? ? ? ? ? CUMB 2 ? ? ? ? QP 1 ? 2 ?(k F N ) ? ?( PP ? c) )+ ? ? 4k F S ? ? PC ? ? ? ? ?
= ( PP ? c) ( Q
CUMB P
?Q
TMB P
TMB 2 ? ? VH ? 2 ? ? QP ?k ? ?( PP ? c) ? ? ? 4k ? ? PEB ? ? ? ? ?
* ?* CUMB > ? TMB when:
(16)
CUMB TMB ? QP ? QP ? ? PP ? c ?
? 1 ? ? + 4k F S ?
CUMB 2 ? ? ? ? ? QP VH 2 ?(k F N ) ? ?( PP ? c) ? ?? ? PC ? ? 4 k ? ? ? ?
?? k ? 2 ? ? Q TMB P ?? ? ?? ? P ?c? ? ?P ? EB ? ? ?? P
? ? ? ?
2
? ?>0 ? ?
57
When CUMB is more profitable than TMB? This section outlines the conditions in which Content Unbundled Mixed Bundling is more profitable than the Tradition Mixed Bundling strategy. We look at three expressions in order to predict whether CUMB is more profitable than TMB (17a – 17c): (17a)
CUMB TMB CUMB TMB QP ? QP > 0 which is true when QP > QP
CUMB TMB QP and QP are the print sales under CUMB and TMB, respectively. These are
purchases by customers who have a strong preference for print; that is, customers who receive a higher value from the print form as compared to the value of the electronic form (accounting for their prices).
CUMB CUMB ??kFN kFN ? ? ? k F N ? ? ? QP ? QP (17b) ? ?. ? P ?c ; P ?c? ? P ?c? ? ?? ? ?P ? ? > 0 which is true when ? P ? ? C P ? ? P C ? P ? ? ?
2 2
CUMB ? QP is the rate of change of the print book demand given a change in the electronic ? PC
chapter price. We consider whether the absolute value of the rate of change is bigger or smaller than the average demand for chapters at zero price divided by the contribution margin of the print book. (17c)
TMB ? k ? ? ? QP ? ? ? ? ?P ?c? ? ?P EB ? P ? ? 2
TMB ? ?k ? QP k ? ? ? ? ? ; ? which is true when > 0 ? ? ? ? P P ? c P ? c EB P P ? ? ?
2
TMB ? QP is the rate of change of the print book demand given a change in the electronic ? PEB
book price. We look at whether the magnitude of the rate of change is bigger or smaller than the number of potential consumers on the market divided by the contribution margin of the print book. << Insert Table 5 about here >> 58
In Table 5 we present a comparison of the profits under content unbundled and traditional mixed bundling under all possible combinations of conditions 17a, 17b and 17c. For example, if print book sales under CUMB (vs. TMB) are higher, the rate of change of the print books sales given a change in the electronic chapter price is relatively small, and the rate of change of the print books sales given a change in the electronic book price is relatively high (line b, Table 5), CUMB is always more profitable. We expect the two rates of change of print book sales given changes in the PDF chapters price and electronic book price to be relatively small and equations 17b and 17c to be positive, and consequently - to fall within profit comparisons d or h (lines d and h, Table 5). Thus, CUMB will be more profitable than TMB if profit comparison equation d or profit comparison equation h in Table 5 is true. In sum, we predict that CUMB will be more profitable than TMB when equation 16 has a positive sign. The conditions in which equation 16 has a positive sign are outlined in Table 5. Overall, which strategy is more profitable depends on the extent to which it differentially affects the print book sales, combined with the extent to which it can extract higher profit given the optimal electronic product price that minimizes print sales cannibalization to electronic sales. If the publisher markets more than one book, or more than one item in general, the individual item profit functions can be aggregated across items and then compared on an aggregate level. This will not change the nature and the predictions of the model.
Data and Estimation Procedure Data
59
To empirically test our predictions, we use data provided by NAP. These are actual purchases in an online experiment involving intercepting customers on the publisher’s website. Although NAP had been offering free browsing of its books for years, customers could buy only print (paperback or hardcover) books before the experiment. During the experiment, customers who already had a print copy of a title (for which a companion PDF version was available) in their shopping carts, were intercepted as they clicked on the check-out button and assigned to one of two conditions. About 500 titles were used in both conditions. Experimental Condition 1: Content Unbundled Mixed Bundling Strategy Details of the PDF book including its price per chapter (in dollars) were presented to the consumers, and they had the option to check out a sample PDF before making a decision. In this condition respondents could buy one or more PDF chapters, the print book, or a combination of the two forms. Prices of PDF chapters for the participating titles were set at different levels relative to the pro-rated price of the print books (at 50%, 75%, 100% and 125%). For example, a print book containing 10 chapters priced at $50 would have a pro-rated price per chapter equal to $5. Then, the PDF chapter price could be set at 50%, 75%, 100% or 125% of the pro-rated price per chapter. The books were randomly assigned to the electronic price conditions; for example, all print books priced at $50 were randomly assigned to the four electronic price conditions. The reason for assigning books and not customers to the electronic chapters price conditions was to treat all the participants equally and avoid situations in which the same book would be offered as PDF chapters at different price to different consumers. We have data on 3256 customers who participated in the experiment.
60
Experimental Condition 2: Traditional Mixed Bundling Strategy Details of the PDF book including its price (in dollars) were presented to the consumers, and they had the option to check out a sample PDF as well. Participants could buy the PDF book, the print book, or the bundle of the two. The prices of the PDF books were set at different levels relative to the price of the print books (at 25%, 50%, 75%, 100%, and 110%). As in condition 1, books and not consumers were randomly assigned to the electronic book price levels. The dataset contains 950 customers.
Estimation Procedure
In this section we discuss the estimation procedure we employ to estimate the parameters of interest. We use the estimated coefficients to perform a simulation and investigate what will the probability of buying each alternative be when the PDF price as percentage of print varies. The probabilities are the predicted market shares of the various alternatives under different PDF pricing scenarios, which allows us to compute revenues and profits. We use a discrete choice framework to empirically investigate consumer choice. Specifically, we fit Random Parameters Logit Model (RPLM) to the experimental data (separately for the two conditions) and estimate the parameters of interest. This model is appealing for several reasons. First, the logit models in general are conceptually attractive as they are grounded in economic theory and have excellent empirical performance (Guadagni and Little 1983). Second, the RPLM accounts for heterogeneity across consumers in both brand preferences and responses to marketing variables (Jain, Vilcassim and Chintagunta 1994). Finally, it is empirically tractable and can be computed
61
using readily available computer software (e.g. NLOGIT 3.0 version of LIMDEP). The utility function of consumer i for alternative j in the choice set is equal to: (18)
U ji = ?1 ji CAT1 + ... + ? 5 ji CAT5 + ? 6 ji PRC PRINT + ? 7 ji PCTPDF + ? ji ,
where CAT1 to CAT5 are dummy variables for the five subject categories containing more than 5% of the observations in the sample, PRC PRINT is the price of the print book in dollars and PCTPDF is the price of the PDF item expressed as a percentage of the price of the print book (25% to 125%). The probability of consumer i choosing alternative j is equal to:
(19)
P ( j | vi ) =
exp(? j f ji + ? ji x ji )
? j =1 exp(? j f ji + ? ji x ji )
J
,
where j = 1...J alternatives in the choice set of consumer i (i = 1...I ) ,
? j is a vector of nonrandom (fixed) coefficients, ? ji is a coefficient vector that is randomly distributed across individuals, vi enters ? ji ,
f ji is a vector of choice varying attributes of choices, multiplied by ? j , x ji is a vector of choice varying attributes of choices, multiplied by ? ji ,
? ji is assumed to be distributed iid extreme values, and
vi is a random term with mean vector zero and covariance matrix I. We use the RPL procedure in NLOGIT 3.0 (LIMDEP) to estimate the models. The procedure works as follows: first, the mean vector and the covariance matrix of the coefficient vector ? ji are estimated on a sample of consumers drawn from the population using maximum simulated likelihood estimation; then the likelihood function of a
62
consumer’s sequence of choices is approximated by simulation given the mean vector and the covariance matrix of the coefficient vector. Consequently, the expected parameters for each individual customer are computed conditional on this customer’s observed choices. We account for heterogeneity in two ways. First, under both CUMB and TMB strategies, we estimate separate alternative specific constants for print, PDF and bundle for the consumers in the five book categories containing more than 5% of the observations in the sample – agriculture, behavioral science, education, general interest and medicine. Second, under CUMB, we use random parameters for the effect of
PCTPDF (PDF price as percentage of print price) on print and bundle sales, assuming that
they are normally distributed in the population. This allows us to get insights into the category-specific effects and control for them when estimating the price effects. We use fixed coefficients for the effect of print price on choice, and for the effect of PCTPDF on PDF sales16. Under TMB, we use fixed parameters for all print price and PCTPDF effects on choice17. Our smaller sample size is the likely reason for the non-significant standard deviation of the parameters distributions.
Simulation
Next, we use the RPL model parameters to perform a simulation in which we vary the price of the PDF books and chapters and compare the CUMB and TMB profits under various pricing scenarios. Separately for each strategy and pricing scenario, we compute
16
We estimate another model allowing these coefficients to vary across consumers as well. The model in which they are fixed fits the data better. 17 We estimate another model allowing these coefficients to vary across consumers. The model in which they are fixed fits the data better.
63
the utility and the predicted probability of the alternatives in the choice set for every consumer in our sample using equations 18 and 19. Then, using variable print book cost information provided by the publisher, the profit is computed as follows: (20) ? = (PPRINT + PBUNDLE )(PRC PRINT ? VC PRINT ) + (PPDF + PBUNDLE ) PRC PDF
Finally, we average the profits across the sample. We then compare the profits under CUMB and TMB at various price levels, and suggest optimal product line and pricing.
Results
Descriptive Statistics The actual purchase rates under the two mixed bundling strategies are presented in Table 6. Overall, under content unbundling mixed bundling 59.57% of the customers held on to their print book purchases, 10.08% switched to PDF, 4.13% bought both forms, while the rest 26.22% abandoned their carts. Under traditional mixed bundling 47.37% of the customers held on to their print book purchases, 12.32% switched to PDF, 5.89% bought both forms, while the rest 34.42% abandoned their carts. The exact breakdown under the specific price conditions is presented in Table 6 as well. Note that in the second part of the table for each condition the bundle percentages are added to the print and PDF percentages in order to see the total print and PDF sales for each condition. The number of observations for each pricing level is displayed as well. << Insert Table 6 about here >>
Estimation Results – Content Unbundled Mixed Bundling The results under Content Unbundled Mixed Bundling are presented in Table 7.
64
Consumers buying books in the behavioral science and education categories have a significantly higher preference for the print form and are more likely to buy it as compared to consumers buying books in the other categories ( ? = 0.39 and ? = 0.52 respectively, p’s < .05). They are also less likely to buy PDF chapters, although the coefficients are only marginally significant ( ? = -0.34, p = .09, and ? = -0.31, p = .18). The other investigated category specific effects are not significantly different. The print price is significantly affecting consumers’ probability of buying PDF chapters ( ? = -0.02, p < .05), suggesting that consumers are less likely to buy PDF chapters from the more expensive books (books with relatively higher print price). Finally, PCTPDF is affecting the choice probabilities in the following manner: when the PDF price increases, consumers are significantly more likely to buy print books ( ? = 0.90, p < .05), significantly less likely to buy PDF chapters ( ? = -0.55, p < .05), and marginally less likely to buy the bundle ( ? = -5.80, p = .09). The derived standard deviations of the PDF percentage parameters for the print and bundle choices are significant as well (1.75 and 3.89 respectively, p < .05), suggesting that the effects vary across consumers. The RPL model allows us to estimate individual specific PCTPDF parameters and account for heterogeneity when making predictions. << Insert Table 7 about here >>
Estimation Results – Traditional Mixed Bundling The estimation results under Traditional Mixed Bundling are presented in Table 8. Similar to the results under CUMB, consumers buying books in the behavioral science and education categories are more likely to choose the print book ( ? = 0.50, p = .08, and
65
? = 0.97, p < .05) as compared to the customers buying books in the other categories.
Consumers buying books in the behavioral science category are less likely to buy the PDF book ( ? = -2.34, p < .05) as compared to the customers buying books in the other categories. The other investigated category specific effects are not significantly different. Additionally, the print price is significantly affecting the probability of buying print ( ? = 0.01, p = .07) and the bundle ( ? = -0.04, p < .05), suggesting that consumers are more likely to buy the print book and less likely to buy the bundle regarding the relatively more expensive print books. Finally, as the PDF book price increases, the probability of buying the PDF book decreases ( ? = -1.25, p < .05). << Insert Table 8 about here >> In sum, the CUMB and TMB estimation results seem intuitive. The models that we use fit the data well and allow us to estimate parameters that can be further used to perform simulations and make predictions.
Simulation Results Using the estimated coefficients from the CUMB and TMB models, we perform a simulation by choosing different PDF price levels (PDF price as percentage of print price) and computing the individual utilities, purchase probabilities and profits using equations 18, 19 and 20, and average the profit across individuals. The simulation results for the average profit per consumer under various price levels are presented in Table 9. The simulation results suggest that in the case of Traditional Mixed Bundling the PDF book should be priced the same as the print book in order to get the highest profit ($12.74), while in the case of Content Unbundled Mixed Bundling the PDF chapters
66
should be priced at 125% of the price of the print book ($15.21). Content Unbundled Mixed Bundling seems to generate a higher profit as compared to Traditional Mixed Bundling when priced at levels above 75% of the price of the print book. Under Content Unbundled Mixed Bundling consumers are more likely to buy the print book and less likely to choose the “no choice” option, which helps generate higher profit. Although more consumers buy the bundle under TMB, this extra revenue from the current print customers do not compensate for the customers who decide not to make a purchase. We explore this issue more in discussion section. << Insert Table 9 about here >>
CUMB/TMB vs. Print Books Only Here we use as a baseline the case under traditional mixed bundling where the price of the print book is the same as the price of the PDF book. In this case consumers willing to buy the book will buy it in the form they prefer as the prices are the same. As the baseline strategy is in fact the optimal strategy under TMB, this implies that introducing PDF books with the same price as the print books will not change the overall profit for the company. Introducing PDF chapters priced at 125% of the price of the print books will result in higher profit as compared to offering only books in print form (19% increase). Consequently, under the conditions considered in our empirical investigation, offering individual chapters is more beneficial for the company then either offering print books only or print and full electronic books.
67
Relating the Empirical Results to the Predictions of the Bundling Model Here we compare the optimal conditions under CUMB and TMB with the predictions of the model. First, which strategy is more profitable depends on which one leads to higher print sales (equation 17a). The predicted print book market share under CUMB when the PDF chapters are priced at 125% of the print book price is 71%, while the predicted print book market share under TMB when the PDF book is priced at 100% of the print book price is 48.1%. Consequently, equation 17a holds. Second, which strategy is more profitable depends on whether the absolute value of the rate of change of the print book sales with respect to changes in the PDF chapter price is smaller than the average demand for chapters at zero price divided by the contribution margin of the print book (equation 17b). The rate of change is 0.90, the average contribution margin is about $18, and the average demand for chapters at zero
CUMB price is 3.591. Consequently, ? Q P ? PC ? [? k F N (PP ? c ); k F N (PP ? c )] and
equation 17b holds. Finally, which strategy is more profitable depends on whether the absolute value of the rate of change of the print book sales with respect to changes in the PDF book price is smaller than the number of potential consumers on the market divided by the contribution margin of the print book. The rate of change is equal to -0.245, and the average contribution margin is $18. As there are 1,000 potential customers on the market according to our assumption (which is an underestimation relative to the number of respondents in our experiment), equation 17c holds and
TMB ? QP ? PEB ? [? k (PP ? c ); k (PP ? c )].
Consequently, our results suggest that we have to expect the profit under CUMB
68
to be higher than the profit under TMB if
CUMB TMB ? QP ? QP ? ? PP ? c ?
? 1 ? ? + 4 kFS ?
?? k F N ? 2 ? ? Q CUMB P ?? ? P ?c? ? ?? ? ? ? ? ? PC ?? P
? ? ? ?
2
? V ?? k ? 2 ? ? Q TMB P ? ? > H ?? ?? ? ? ? 4 k ?? PP ? c ? ? ? PEB ? ? ?
? ? ? ?
2
? ? (line ? ?
d in Table 1). If there are 1,000 potential customers on the market, the first term of the above equation is equal to 12.7218. The second term of the equation is equal to 62.0919 (average print book price = $31.33). Finally, the third term is equal to 71.6020. Thus, the sum of the first two terms (74.81) is bigger than the third one and the predicted profit under CUMB is bigger (vs. TMB). This is in line with our empirical results.
Discussion
Our study highlights several important issues. First, there are significant subject category effects suggesting that consumers buying books in education and behavioral sciences have a stronger preference for print as compared to consumers buying in the other subject domains, and are less likely to either switch from print to PDF or buy the bundle. On the one hand, these consumers may not have extensive experience with electronic products, or may not use computers as often as other consumers. On the other hand, the print form may be more appropriate for their specific usage situations. Thus, publishers interested in selling electronic products to this customer segment may study why these consumers are more likely to stick with the print form and less likely to buy the PDF form.
18 19
Term 1 = [(0.71-0.481)*1,000]/18=12.72 Term 2 = [1/639.89]*[(3,591/18)*(3,591/18)-(0.9*0.9)]=62.09; S=(1/(2*31.33*1.25/14)*(3,591+18*0.9)=639.89 20 Term 3 = [(62.93/4000)*[(1000/18)*(1000/18)-(-0.234)(-0.234)]=71.60; VH=(2*31.33*1000)/[1000+18*(-.234)]=62.93
69
Second, our study reveals that for relatively more expensive books consumers are more likely to stick with the print form and less likely to switch to the PDF form or buy the bundle of the two forms. Although quite unintuitive at first, there may be good reasons for such a result. On the one hand, consumers may have relatively higher valuation for the more expensive print books because they are likely to be hardcover books with pictures and graphics, and the print form may be perceived as a better choice with regard to these attributes than the PDF form; customers may even be likely to display such books in their offices or at home (e.g., coffee table books) or have other reasons to prefer the print form. On the other hand, consumers may be less willing to pay a high price for PDF books as they know that their marginal costs are negligible. For example, keeping everything else constant, if a print book costs $25.00 and the PDF book price is set at $18.75 (75% of the price of the print book), consumers may be more likely to buy the PDF book vs. the print book as compared to the case where the print book costs $75 and the PDF book price is set at $56.25 (75% of the price of the print book). Although the price of the electronic book is the same percentage of the price of the print book in both cases, consumers are likely to be aware that the marginal costs of the electronic books are negligible and they would not expect to be charged a big dollar amount even for expensive print books. Thus, owning electronic products may be unappealing to consumers in certain subject domains, and there may be an upper bound on how much consumers are willing to pay for electronic content in terms of a dollar amount. It may be unreasonable to determine the electronic form price as a percentage of the print form price, and other pricing mechanisms may be more appropriate in some cases.
70
Third, our simulation predicts that, under traditional mixed bundling, as the price of the PDF book approaches and exceeds the price of the print book, the share of the nochoice option increases and the share of the print option does not change. This is not the case under content unbundled mixed bundling where as the chapter price increases, the share of the no-choice option decreases while the share of the print book option increases. Recent studies in the marketing domain may provide insight into this finding. Dhar (1997) argues that consumers tend to focus more on the comparative characteristics among the alternatives provided than on their own utilities. On the one hand, expanding the choice set by adding an attractive alternative increases the preference for the nochoice option when respondents can choose only one alternative (Dhar 1997). On the other hand, respondents are less likely to defer choice when both attractive alternatives could be selected (Dhar 1997). If they have to make a choice between print and full PDF books, consumers may focus on the comparative characteristics of the two forms and evaluate which form is better on various attributes such as layout, browsing, image quality, etc. Because both forms have their advantages and disadvantages, and because price is often perceived as a signal of product quality, consumers may be more likely to find the two alternatives equally attractive and defer choice when the prices are similar as compared to when the PDF price is significantly lower than the print price. When the full PDF book price is significantly lower than the print price consumers may be more likely to buy the PDF book as it becomes relatively more attractive because of its low price, or buy the cheaper bundle and benefit from the advantages of both forms. It seems that under content unbundled mixed bundling it is easier for consumers to make a choice as they are less likely to select the no-choice option. Thus, content unbundled mixed
71
bundling may be more profitable than traditional mixed bundling also because it is easier for consumers to make a choice and more consumers enter the market. Finally, there is a self-selection bias in our sample because consumers are intercepted in the experiment after they have already decided to buy the print book. Consequently, we are offering a conservative test with respect to the quantity of electronic chapters being sold at the optimal price level under CUMB, and CUMB may be even much more profitable than TMB. The reason is that consumers for whom the reservation price for the print book is lower than the price of the print book and will not buy the print book, may be attracted to buy one or several chapters. We are not able to capture these purchases in our data because we intercept consumers after they have decided to buy the print book. Our third essay accounts for the revenue from such consumers.
Managerial Implications
Out study provides valuable insights to content providers on how to successfully design and price information products in addition to their print products, assuming that the providers have reasons not to or do not want to change the price of the print products. Content unbundled mixed bundling is a viable strategy to pursue in certain market conditions as it leads to higher profit. The unbundled electronic units should be priced at a premium compared to the pro-rated price of the print book in order to achieve the optimal level of profit. The demand for electronic versions of products such as books is sizeable and it justifies the extension of the existing product lines. Although our study does not look at this issue, introducing electronic products may attract new consumers
72
and expand the market as well. It may be more beneficial for the publishers to consider different pricing schemes for the different book categories. Our results highlight that consumers buying books in certain subject domains are less likely to buy electronic products, and more likely to stick with the traditional print form. Thus, in order to penetrate these segments, content providers may focus on educating the consumers on the benefits the electronic form provides, and employ strategies that encourage first trial. Publishers may find it also interesting that the more expensive print books are less likely to be substituted with electronic books. As these are likely to be high margin books that bring substantial income, introducing electronic version of such books will not necessarily hurt this revenue stream.
Conclusion
In this paper, we model the profits under two strategies a content provider may employ: traditional mixed bundling, in which the product line consists of print book, PDF book and the bundle of the two, and content unbundled mixed bundling, which is selling print book, PDF chapters and the combination of the two. Our study suggests that offering unbundled PDF chapters is more profitable than offering only the full PDF books when the print books prices do not change when PDF content is introduced. We empirically support our predictions with actual data gathered in an online experiment executed on the website of a publisher of electronic books. We offer important insights on how customers from different subject domains react to electronic products and how they are likely to behave when different product and pricing schemes are employed.
73
Our study contributes to the literature on bundling as it offers guidelines on the optimal strategy when different mixed bundling strategies are feasible. Further, it looks at both form bundling and content unbundling, which is different than the literature so far and provides additional insights. Last but not least, our data includes purchases of books in a wide range of subject domains, thus allowing us to generalize our findings and recommendations (while controlling for some category specific effects). Our findings highlight important questions and unresolved issues that future research may address. It will be interesting to discover why consumers from different subject domains have distinctive preferences and inclination with regard to electronic products. Further, future studies may clarify why consumers are less likely to make a choice when they are offered print and electronic products with similar prices. Providing further insights may help companies successfully design and implement penetration strategies regarding electronic content, resulting in higher customer satisfaction and repeated purchases. Third, it will be interesting to investigate whether offering individual book chapters in print form is a viable strategy for a publisher to pursue. Although such a product strategy may be too expensive because of the costs associated with unbundling print content, it may be appealing under certain conditions.
74
CHAPTER IV: ESSAY THREE. BUNDLING AND UNBUNDLING OF ELECTRONIC CONTENT
Abstract
This paper investigates the attractiveness of product lines of items such as books and newspapers available in conventional and electronic media. For example, the publisher of Wall Street Journal can offer the following product line of subscription options: print WSJ, WSJ online, separate sections of WSJ online (e.g. Money & Investment, Technology), print WSJ and online WSJ bundle, and print WSJ and online section bundle. We employ a choice experiment in which a sample of consumers is presented with hypothetical product offerings at various prices and asked to make a choice. The data is used to develop a profit-maximizing configuration of products and prices. Similar approaches to the product line pricing problem have been employed for conventional products, but not when bundling of different forms of a product is an option, and not when the different products may be complements rather than substitutes.
75
Introduction
The Internet as an information channel has facilitated the development and dissemination of electronic forms of traditional print products such as books and newspapers. Electronic information products are unique in that the marginal cost of reproducing and distributing them is often much lower than the cost of the print products. Hence, content providers may be interested in selling the electronic form by itself or together with the print form to get higher revenue. Further, the electronic products can be unbundled at no extra cost, and the unbundled components can be re-bundled with the traditional print products. For instance, Wall Street Journal is currently offered as print WSJ, WSJ online, and the bundle of the two subscription options. Theoretically, the publisher can also sell individual sections of WSJ in electronic form, and even re-bundle the electronic units with the print WSJ (e.g., offer a subscription to the print WSJ and the financial section in electronic form as a package). Thus, if consumers switch from the traditional print products to the cheaper electronic products or units, the publisher may encounter product cannibalization and lost revenues. On the contrary, if consumers subscribe to the print products and electronic products/units bundles, the publisher may extract extra revenue. The focus of the present study is on product lines of items available in conventional and electronic media. Specifically, we discuss a method for determining optimal product line design and pricing for an item available in print and electronic form. In addition, we examine the effect of both price and non-price attributes on consumers’ preferences and choice decisions. In our study we use choice experiment and choicebased conjoint analysis to obtain estimates of customer valuations of the price and non-
76
price attributes of interest. We present respondents with hypothetical choice situations in which the attributes are varied to determine how choice fluctuates with changes in the levels of attributes. Then, accounting for response and preference heterogeneity, we estimate random parameters logit models and use the results to derive optimal product mix and pricing strategy. The output of the study is a general methodology for setting product lines and prices for items available in conventional and electronic media. The contribution of the study is as follows. It extends the research on bundling by examining bundling of interrelated products in the context of product forms, which is a new and managerially relevant application. There are three unique characteristic of the above problem that makes it different from what have been studied in the literature so far. First, the product forms can range from being perceived as substitutes to being perceived as complements to each other, and these contingent valuations may vary across the consumers. Second, the forms are not only interrelated but are likely to contain the same information; thus, by choosing the bundle, consumers acquire two items with the same content. Third, there are various bundles that can be offered – print newspaper and online newspaper bundle, print newspaper and online newspaper section(s) bundle, and online sections bundle. Consequently, the predictions of the literature in terms of optimal bundle composition and pricing may not be applied directly. We study how consumers perceive electronic and conventional print products, and provide insight on how demand fluctuates with changes in price and other attributes. Similar approaches to the product line pricing problem have been traditionally employed for conventional products, but not when bundling of different forms of a product is an option, and not when the different products may be complements rather than substitutes.
77
The third essay extent the second essay in several ways. First, we consider all feasible print and electronic options simultaneously, thus being able to investigate the market expansion when full electronic products and electronic units are offered. Second, we incorporate in our design different price levels for the print products and different levels of bundle discount in addition to the electronic products price levels, which results in more realistic and generalizable conclusions. Finally, the choice-based conjoint experiment we execute allows us to better control the data collection process and generates panel data with many observations per consumer to better model the consumer choice process. The paper is structured as follows. First, we review the relevant marketing and economics literature and position the study. Second, we present the theoretical foundations and the methodology of our study. Third, we discuss the experimental design and data collection procedure. Fourth, we report our empirical findings and their practical implications.
Research Background
Bundling is the strategy of marketing two or more products and/or services as a package at a special price (Guiltinan 1987). Two common approaches are pure bundling whereby only the product bundle is offered, and mixed bundling, in which the bundled items are also sold separately. Hanson and Martin (1990) provide an overview on why the sellers are motivated to bundle, discussing both demand and supply side incentives. The demand side incentives that make bundling profitable include negative correlations in reservation
78
prices (Stigler 1963; Schmalensee 1984), complementarity in consumption (Telser 1979), and uncertainty in the valuations of the quality of the goods (Kenney and Klein 1983). Mixed bundling is more beneficial than pure components or pure bundling, and it is the optimum bundling strategy when there is asymmetry in the reservation prices for the bundle components (Adams and Yallen 1976; Schmalensee 1984; McAfee, McMillan and Whinston 1989). In terms of supply side incentives for bundling, Bakos and Brynjolfsson (1999, 2000) demonstrate that large scale bundling of information goods can be very profitable because it creates economies of aggregation when their marginal costs are very low. In the marketing domain, empirical studies focus on issues such as bundling and unbundling of industrial systems (Wilson, Weiss and John 1990), determining bundle prices and composition (Hanson and Martin 1990), and optimal bundle pricing (Venkatesh and Mahajan 1993). Additionally, several analytical articles discuss contingent valuations in bundling decisions, contrasting interrelated from independently valued products in a bundle (e.g. Venkatesh and Kamakura 2003). In sum, the above-presented research identifies the conditions for profitable bundling and specifies the optimal bundling strategies in various situations. Sellers are motivated to bundle because, on the demand side, it can help them extract consumer surplus, and, on the supply side, it can increase producer surplus or lead to cost savings.
Positioning Our Study
Most of the research so far suggests that mixed bundling is the optimal solution for producers of various goods (Adams and Yallen 1976, Schmalensee 1984, McAfee,
79
McMillan and Whinston 1989, etc.). This policy enables the seller to reduce effective heterogeneity among those buyers with high reservation prices for both goods, while still selling at a high markup to those buyers willing to pay a high price for only one of the goods (Schmalensee 1984). With regard to bundling of information goods such as magazines and journals, recent articles argue that pure bundling is an optimal strategy for a monopolist marketing a large number of information goods given negligible marginal costs and bulk sales (Bakos and Brynjolfsson 1999, 2000). Geng et al. (2005) further refine their results and suggest that mixed bundling is approximately optimal if consumers’ values to subsequent goods do not decrease too quickly, otherwise pure bundling is optimal even when there are strong negative or positive correlations of values across goods. Our study contributes to this stream of research in the following manner. First, we investigate the attractiveness of various bundled and unbundled forms of print and electronic content, which is an important area with significant managerial implications. Second, we address the problem both from a seller’s and a buyer’s point of view. We not only study how consumers are likely to behave when offered bundled/unbundled content in different forms, but also use a methodology that companies may apply in designing their product lines and pricing them optimally after accounting for consumer preferences. Third, the application in the book and newspaper categories that we discuss can be easily extended to other categories such as magazines and music. Finally, our results have interesting implications in the area of personalization of product offering and customization in the digital goods domain. There are three recent papers that focus on issues similar to the ones in our study.
80
Next we briefly outline these studies in order to provide comparisons with ours. First, Venkatesh and Kamakura (2003) model the optimal bundling and pricing strategies for interrelated products under monopoly. Our study is different from theirs because we investigate optimal bundling strategies in the case of different forms of the same product and various combinations of these forms, while they consider bundling of distinctive products that are interrelated (e.g. computer and printer). Moreover, we use choice-based conjoint analysis with stated choices to estimate the parameters of interest and further include them in our optimization problem while Venkatesh and Kamakura (2003) rely on simulation-based analytical solution. In the second study, in a context similar to ours, Venkatesh and Chatterjee (2003) focus on bundling of print and electronic magazines, including separate sections of these magazines. Our study complements their analytical findings by including offering more options to the consumers. While they consider print magazine, online magazine and separate modules of the online magazine and all possible combinations of these in the product line but not bundling them (for example, they do not include a bundle of print magazine with an online magazine at a price lower than the sum of the component prices), we allow bundling of the print item with the online item, and bundling of the print item with the online modules at a price lower than the sum of the prices of the bundle components. The third paper by Jedidi et al. (2003) suggests a model for capturing continuous heterogeneity in the joint distribution of reservation prices for products and bundles. Their results suggest that the product line pricing policy depends on the degree of heterogeneity in the reservation prices. Our study further extends theirs by considering non-price attributes in addition to product and bundle price. Additionally, Jedidi et al (2003) study bundles of conventional products such as the Times Magazine
81
and investigate the optimal price for the magazines and the bundle of the two, while we study the same information content under different product forms and include all feasible bundled, unbundled and re-bundled options in the choice set. In sum, our paper extends the bundling research by investigating bundling of different product forms containing the same information. Additionally, we provide optimal product line design and pricing strategies while incorporating consumer preference for bundled and individual print and electronic forms of a product, and after adding different combinations of re-bundled electronic units with print products to the choice set.
Theoretical Background
Next we discuss our expectations about the attractiveness of bundling of electronic content with traditional print content, and of unbundling of electronic content and/or re-bundling it with traditional print content.
Product Form Complementarity in Choice Complementarity has been traditionally characterized by referring to the sign of the cross-price elasticity of demand - if the sign is positive (negative), products are substitutes (complements). In marketing, complementarity is often defined relative to product-specific utilities and the corresponding consumer needs (Chernev 2005). Products are substitutes if both can satisfy the same need to the consumer, they are complements if they are consumed jointly (but not necessarily simultaneously) in order to satisfy a need (Lattin and McAlister 1985; Henderson and Quandt 1958). Previous
82
research suggests that complementarity between products can cause bundling to be profitable (Telser 1979). Bundles composed of complements have higher purchase intent versus bundles of similar or unrelated products (Harlam et al. 1995), and complementarity positively affects bundle reservation prices (Gaeth et. al 1990). From consumers’ point of view, if the product forms are substitutes, the purchase of one form lowers the value of the alternative form, and therefore makes bundling the items less attractive; if the forms are complements, bundling becomes more attractive. Consumers may perceive a high degree of complementarity based on search economies, improved satisfaction because of the bundle, and/or improved total image of the bundle (Guiltinan 1987; Simonin and Ruth 1995). By and large, we expect a positive relationship between the degree of perceived complementarity between the bundle components and the purchase likelihood for the bundle. On one hand, if consumers believe that the individual product forms can be used interchangeably, they will buy only one of the forms and not the product form bundle. On the other hand, if consumers perceive additional utility in having both product forms versus just one of them, they will choose the product form bundle.
Attractiveness of Bundling of Electronic Content with Traditional Print Content Analytical results suggest that in general moderate or strong substitutes should be offered separately; the same is applicable for complements if the marginal costs are moderate relative to the market’s maximum willing to pay (Venkatesh and Kamakura 2003). Also, a seller gains by mixed bundling for weak substitutes/complements if the variable costs are not too high (Venkatesh and Kamakura 2003). Finally, although the
83
product line pricing policy depends on the degree of heterogeneity in the reservation prices for the items, firms would benefit from using a mixed bundling strategy (Jedidi et al. 2003). Building upon these findings, we would expect mixed bundling strategy be the optimal solution. This statement is supported by both analytical solutions and related empirical results. However, there are three unique features that characterize our problem: consumers are likely to be heterogeneous in the way they see the different product forms on the substitute-complement continuum, the marginal costs of the electronic products are almost negligible, and the market’s willingness to pay for some electronic products may be significantly lower than the one for conventional print products. Consequently, our results may suggest a different optimal solution. In terms of optimal prices, Jedidi et al. (2003) advocate that a uniformly high price strategy for all products and bundles is optimal when the heterogeneity in the reservation prices is high; otherwise, a hybrid strategy is optimal. Thus, it is an empirical question on how to price the same print and electronic content as individual products and as a bundle. In sum, the findings of the extant literature do not offer guidelines on to how to design and optimally price product lines consisting of different forms of the same product. Our study aims at providing insights into this question with important implications for practitioners.
Attractiveness of Unbundling of Electronic Content and Re-Bundling with Print Content In a different context (industrial systems), Wilson, Weiss and John (1990) identify the following cases favoring unbundling: larger unit margins from unbundling, market growth from unbundling, new market segments from unbundling, and inferior but cheaper systems from unbundling. With respect to unbundling of electronic content (e.g.
84
offering individual sections of WSJ online), it would be an appealing strategy if the companies could attract new customers who are more price-sensitive and otherwise would not buy the full product, thus leading to market growth. If unbundling causes cannibalization of print sales to unbundled content, the unbundling strategy could result in lower profits. Consequently, attracting new market segments and market growth would favor unbundling of electronic content. In terms of pricing of unbundled units, contrary to conventional wisdom on mixed bundling, Venkatesh and Chatterjee (2003) claim that low-priced components (e.g. individual sections) should be targeted at consumers who have low value for the information content. Thus, it might be optimal to pursue such a pricing strategy: on one hand, trying to attract new customers by offering unbundled content at a low price, and on the other hand, extracting extra profits by offering full print content plus individual sections in electronic form to the consumers who see the different forms as complementing each other.
Product Form Attractiveness in Choice Dhar (1997) argues that consumers tend to focus more on the comparative characteristics among the alternatives provided than on their own utilities. When the choice task is to choose one alternative from a choice set, consumers are more likely to defer choice when the alternatives are equally attractive. When the task allows the choice of more than one alternative from a choice set of two attractive alternatives, the preference for a no-choice option decreases. Thus, if the consumers perceive one product form better then the other on most of the attributes (e.g. form 1 is superior to form 2), we would expect them to be more likely to choose the superior form. On the other hand, if the consumers perceive form 1 better then form 2 on some of the attributes, while form 2 85
better than form 1 on the rest of the attributes (form 1 and form 2 are equally attractive), we would expect them to be more likely to defer choice when they have to select just one alternative. Finally, if they are given the option to choose more than one form (e.g. buy a bundle), consumers may decide to buy both form if they are equally attractive. Thus, we would expect the relative attractiveness of the product forms to influence choice in the following manner: on one hand, the more differentially attractive the two forms are, the higher the probability to defer choice; on the other hand, the higher the difference between the attractiveness of the superior product form and the bundle, the more likely consumers will be to select the bundle.
Methodology
We use a choice-based conjoint framework to study our questions of interest. Since the early 1970s, conjoint analysis has received considerable academic and industry attention (Green and Srinivasan, 1990). Conjoint analysis and the related technique of experimental choice analysis represent the most widely applied methodologies for measuring and analyzing consumer preferences (Carroll and Green 1995). The choicebased conjoint is a relatively new type of conjoint analysis and is considered to better approximate actual decision processes as compared to the traditional ratings or rankingsbased conjoint analyses21. Consequently, more realistic aggregate level estimates are expected. In the next sections we describe in detail our experimental design, the data collection procedure and our modeling approach.
For an excellent review and empirical comparison between ratings-based and choice-based conjoint models see Elrod, Louviere and Davey (1992).
21
86
Study Design and Data Collection
We use two product categories in our study – a book (“Vault Guide – insider information on industries, careers, and employers”) and a newspaper (Wall Street Journal). We have selected these two products because they are very popular among our sample, MBA students, and our respondents are well-experienced with them. Additionally, current MBA students are an important target audience for both publications and this increases the external validity and applicability of our results. In the conjoint experiment the respondents evaluated various scenarios for each product and made a choice. We vary the following factors in the scenarios: print (absent, present), full electronic (absent, present), electronic unit/section (absent, present), bundle (absent, present) and price level (low, medium, high). The conjoint factors and the price levels are presented in Table 10, panels A and B. We took the actual print and electronic/PDF form prices of the Vault Guide and WSJ for the medium price level; the electronic unit price was computed by dividing the electronic/PDF price by three22 and then adding a 15% premium. The price of the print product and electronic product bundle is the sum of the two individual product prices discounted 15%, while the price of the print product and electronic section bundle is the sum of the two individual product prices discounted 7%23. Finally, for all product offers, the low and the high price levels represent a 15% discount/premium compared to the medium price level. We chose a 15% gap between the different pricing levels to assure that the respondents are able to register such price differences while keeping the face validity of the prices of the various items.
22 23
There are three sections in the product. We decreased the bundle price by only 7% in this case to preserve the face validity of our prices and avoid situations in which the print product costs much more than the print product/electronic section bundle (but still employ three different price levels).
87
<< Insert Table 10 about here >> For each of the two product categories the participants were presented with a choice situation (e.g. that they are considering subscribing to WSJ) and asked to consider sixteen independent choice scenarios with three options in each (e.g. Option 1: Subscribe to the Online WSJ for $39.50; Option 2: Subscribe to the Print WSJ for $114.50; Option 3: Will not subscribe to either of the two). The options in each scenario were randomly selected from a full factorial design in which product offer and price were varied. Participants were asked to select the option that they were most likely to pursue in such a situation. After evaluating all the scenarios for the first category, the same was repeated in the second product category. At the end the participants answered question related to the relative attractiveness of the product forms on five attributes (image quality, browsing, layout, convenience of use and archival quality) and demographics. Eighty seven full and part time MBA students (60% part time students; 37% female) enrolled in graduate level marketing courses participated in our study. Each participant evaluated sixteen choice scenarios in the book category and sixteen choice scenarios in the newspaper category (the order was counterbalanced).
Modeling Consumer Preferences We use a discrete choice framework to model consumer choice. Specifically, we fit Random Parameters Logit Model (RPLM) to the conjoint data and to estimate the parameters. This model is appealing for several reasons. First, the logit models in general are conceptually attractive as they are grounded in economic theory and have excellent empirical performance (Guadagni and Little 1983). Second, when desired, the RPLM can
88
account for heterogeneity across consumers in both brand preferences and responses to marketing variables (Jain, Vilcassim and Chintagunta 1994). Finally, it is empirically tractable and can be computed using readily available computer software (e.g. NLOGIT 3.0 version of LIMDEP). Accounting for heterogeneity across consumers is an important consideration when analyzing consumer purchase behavior from panel data - consumers with the same demographic and socioeconomic characteristics (e.g. income, family size) when confronted with a given set of covariates (e.g. price, feature advertisements) may exhibit different choice behavior due to differences in overall brand preferences (intercept/preference heterogeneity) and/or variations in their responses to these covariates (slope/response heterogeneity; Jain, Vilcassim and Chintagunta 1994). The above argument applies to our data as well. For example, we may observe different choice behavior by consumers with the same age, education and income level because of differences in overall preference for print vs. electronic content, and because of variations in their responses to marketing mix variables such as price. Thus, the RPLM allows us to control for the unobserved heterogeneity among individual consumers when estimating various parameters and using them to make predictions. The RPLM is a one level multinomial logit for individuals i = 1, …, N. P ( j | vi ) = exp(? j f ji + ? ji x ji ) ,
?m=1 exp(?m f mi + ? mi xmi )
J
where U ( j, i) = ? j f ji + ? ji x ji + ? ji , j = 1, …, J alternatives in the i’s choice set,
? j is a vector of nonrandom (fixed) coefficients, ? ji is a coefficient vector that is randomly distributed across individuals, vi enters ? ji ,
89
f ji is a vector of choice varying attributes of choices, multiplied by ? j , x ji is a vector of choice varying attributes of choices, multiplied by ? ji ,
? ji is assumed to be distributed iid extreme values,
vi is a random term with mean vector zero and covariance matrix I. We use the RPL procedure in NLOGIT in LIMDEP to estimate the models. The procedure works as follows: first, the mean vector and the covariance matrix of the coefficient vector ? ji are estimated on a sample of consumers drawn from the population using maximum simulated likelihood estimation; then the likelihood function of a consumer’s sequence of choices is approximated by simulation given the mean vector and the covariance matrix of the coefficient vector. Consequently, the expected parameters for each individual customer are computed conditional on this customer’s observed choices.
Results
Conjoint Experiment << Insert Table 11 about here >> The estimation results are presented in Table 11. We include as attributes dummy variables for the following conjoint factors (1-present, 0-absent): print, electronic full, electronic unit and bundle; the price variable contains the actual prices the respondents evaluated with each product offer. We estimate a Random Parameters Logit model for each product category in which the preference parameters are assumed to be normally distributed across the sample. The results are consistent across the two categories: consumers value very positively the print attribute (the part worth is 8.485 in the book
90
category and 11.333 in the newspaper category) and the full electronic attribute (7.306 and 6.793 respectively), while the part worth of the electronic unit attribute is relatively smaller (3.063 and 1.452 respectively). The bundle attribute part worth is negative and relatively high in both categories (-6.647 and -3.384 respectively), and the price parameter is negative (-0.261 and -0.133)24. Interestingly, the derived standard deviations of the means of all five parameters are significant in both categories (varying from 0.107 and 0.025 for the means of the price parameters to 5.249 and 6.222 for the means of the print parameters). This suggests that that there is significant heterogeneity in how much people value the different product forms and their combinations, and how much price changes affect their choices. In terms of the price response heterogeneity, at least 95% of the consumers have negative responses to a price increase in the book category (their price parameters fall within two standard deviations from the parameter means and are below zero) and at least 99% of the consumers have negative responses to a price increase in the newspaper category (their price parameters fall within three standard deviations from the parameter means and are below zero). The part worth of the print attribute is positive for 68% to 95% of the consumers (their print parameters fall within one to two standard deviations from the parameter mean), and the part worth of the full electronic attribute is positive for 95% to 99% of the consumers (their full electronic parameters fall within two to three standard deviations from the mean). About 68% to 95% of the consumers have positive electronic unit part worth in the book category while the same is true for less than 68% of the consumers in the newspaper category. Finally, with respect to how much consumers value the bundle factor, the situation is very
We also estimate another model in each category in which the price coefficients are not constrained to be the same across the alternatives. The model we report fits the data better.
24
91
different across the two categories: the bundle part worth is negative for more than 99% of the respondents in the book category, while the same is true for less than 68% of the respondents in the newspaper category. Note that we can not directly compare the preference parameters across the two product categories because of the scaling of the parameters in the estimated models. We can compute what price makes consumers indifferent between buying one product offer or the other. For example, in the book category, the average consumer will be indifferent between buying the print book and the full electronic book when the print book price is $29.00 and the electronic book price is $24.36 [exp(8.485–0.261*29.00) = exp(7.3060.261*24.36)]. Also, he/she will be indifferent between buying the print book and the print book plus electronic book bundle when the print book price is $29.00 and the bundle price is $31.49 [exp(8.485–0.261*29.00) = exp(8.485+7.306-6.6470.261*31.49)]. In the newspaper category, the average consumer will be indifferent between subscribing to the print WSJ and the online WSJ when the print WSJ price is $99.50 and the online WSJ price is $65.67 [exp(11.333–0.133*99.50) = exp(6.7930.133*65.67)]. Also, he/she will be indifferent between subscribing to the print WSJ and the print WSJ plus online WSJ bundle when the print WSJ price is $99.50 and the bundle price is $125.14 [exp(11.333–0.133*99.50) = exp(11.333+6.793-3.384-0.133*125.14)]. In sum, the conjoint experiment results are very consistent across the two investigated product categories. Because there is significant attribute preference and price response heterogeneity in our sample, the random parameters logit model allows us to estimate the part worth of the conjoint factors and their distributions. Next we use the above results to evaluate all possible market scenarios (combinations of different product
92
offers at different price levels), compute the market share of the various product offers and compare overall profits.
Market Shares and Profits In this section we present the market shares and profit under various market scenarios. We first look at offering a complete product line (print product, electronic product, electronic unit, print product and electronic product bundle, print product and electronic unit bundle), and then compare the results with the case where incomplete product line is offered (print product, electronic product, print product and electronic product bundle) with or without bundle discount. The goal is to empirically compare the various approaches a company may pursue and discover the most profitable product line configuration and pricing strategy.
Complete product line If the content provider decides to go with all available product offers, we need to compare 243 (five product offers with three price levels each) market scenarios. We use the estimated individual level part worths for the five conjoint factors to compute the utility of the available profiles for every consumer under each market scenario. Then we compute the individual probability of choosing each product profile and average the results across consumers for every market scenario. Finally we calculate the total revenue, costs and profit for each of the 243 market scenarios. We assume that the variable costs for the electronic products are equal to zero, and we consider different levels of print variable costs (from print variable cost equal to 40% of the medium price
93
level of the print products to print variable cost equal to 90% of the medium price level of the print products). We use the medium price level as a base as this is the actual market price the publishers of WSJ and Vault Reports are charging. We do not consider fixed costs in our analysis as they just shift the profit function and do not change its slope. The most profitable market scenarios in case of complete product line under the different variable cost levels are presented in Tables 12 and 13. << Insert Table 12 about here >> The book category market simulation results are displayed in Table 12. The top three market scenarios with respect to overall profit are listed within each considered cost level (print variable cost is 40%, 50%, 60%, 70%, 80% and 90% of print price). Together with the three most profitable scenarios, we also present the top two market scenarios in which the print book price is at its medium price level (which is the current price the publisher of Vault Report is charging for the book). For example, when the variable cost of the print book is assumed to be 40% of the price of the print book (first five lines in Table 4), our results show that the most profitable market scenarios is #181 (market shares is brackets): print book at high price (31.1%), electronic book at low price (23.4%), electronic unit at high price (14.1%), print book and electronic book bundle at low price (9.4%), print book and electronic unit bundle at low price (3.9%), and no choice option (18.1%); the total revenue is equal to $20.94 (calculated if there is only one consumer on the market), cost is $5.15, and the profit is $15.79. From the scenarios in which the print book is at medium price, the market scenario that brings the highest profit is #100 (it is ranked number 12 within all market scenarios): print book at medium price (42.0%), electronic book at low price (20.1%), electronic unit at high price (12.5%), print
94
book and electronic book bundle at low price (6.5%), print book and electronic unit bundle at low price (2.5%), and no choice option (16.3%); the total revenue is equal to $20.70 (assuming that the total market consists of one consumer), cost is $5.93, and the profit is $14.77. When comparing the top three most profitable market scenarios and the top two most profitable market scenarios in which the print book is at medium price across the considered print variable cost levels, we notice that the results are quite consistent – market scenarios #181, #182 and #172 are always the top three overall, and market scenarios #100 and #91 are always the top two in which print is at medium price. Thus, in the book category, the publisher generates the highest profit when the print book and the electronic unit are offered at high price, and the electronic book and the two bundles are offered at low price. The reason is that, although the publisher is forgoing print book revenue when the print book price is high, the company is compensating this loss with the increased revenues from the electronic book and the two bundle options which are priced at a low level and attract more buyers. As the electronic products have zero variable costs, this scenario results in lower overall revenue but higher profit as compared to the scenario when the print book price is at medium. The higher electronic unit price allows the company to price discriminate against consumers who value only a portion of the print book, and get the maximum revenue from this segment. << Insert Table 13 about here >> The newspaper category market simulation results are displayed in Table 13. As before, we list the top three market scenarios with respect to overall profit within each considered cost level, and the top two market scenarios in which the print newspaper
95
price is at its medium price level (which is the current price the publisher of Wall Street Journal is charging). For example, when the variable cost of the print newspaper is assumed to be 40% of the price of the print newspaper (first five lines of Table 13), we can see that the most profitable market scenarios is again #181 (market shares is brackets): print newspaper at high price (2.5%), electronic newspaper at low price (51.7%), electronic unit at high price (10.3%), print newspaper and electronic newspaper bundle at low price (25.9%), print newspaper and electronic unit bundle at low price (2.1%), and no choice option (7.5%); the total revenue is equal to $51.26 (assuming that the total market consists of one consumer), cost is $12.14, and the profit is $39.12. Regarding the scenarios in which the print newspaper is at medium price, the market scenario that brings the highest profit is also #100 (it is ranked number 4 within all market scenarios): print newspaper at medium price (5.5%), electronic newspaper at low price (51.3%), electronic unit at high price (10.1%), print newspaper and electronic newspaper bundle at low price (24.2%), print newspaper and electronic unit bundle at low price (1.7%), and no choice option (7.1%); the total revenue is equal to $51.21 (assuming that the total market consists of one consumer), the cost is $12.53, and the profit is $38.68. Comparing the most profitable market scenarios overall and most profitable market scenarios in which the print newspaper is at medium price across the considered print variable cost levels, the results are generally consistent – market scenarios #181, #184, #187, #172, #178 and #208 are among the top overall, and market scenarios #100 and #91 are the top two in which print is at medium price. Thus, in the newspaper category, the publisher generates the highest profit when the print newspaper and the
96
electronic unit are offered at high price, the electronic newspaper is offered at low price, and the two bundles are offered at low, medium or high price (depending on the cost structure). As in the book category, although the publisher is forgoing print newspaper revenue because of its high price, the company is compensating this with the increased revenues from the two bundle options which are priced low and attract more buyers. Because the electronic products have zero variable cost, this scenario results in lower overall revenue but higher profit as compared to the scenario when the print newspaper price is medium. The higher electronic unit price enables the company to price discriminate against consumers who value only a portion of the print newspaper.
Incomplete product line (traditional mixed bundling with bundle discount) In this case the publisher offers the print product, the electronic product and the bundle of the two. As each of the three product offers has three price levels, we compare 27 possible market scenarios. The results are presented in Table 14 (book category) and Table 15 (newspaper category). Again the results are consistent across the two product categories – the publisher generates the highest profit when the print product is offered at high price, the electronic product at low price, and the bundle is at low price. In most of the cases one of the top three profiles in terms of overall profit is the one in which the print product is at medium price, and the electronic product and the bundle are at low price, which is very similar to what the publisher of WSJ is pursuing currently (print at medium price, online at low price and bundle at medium price, as in market scenario #85). The publisher of Vault Reports is offering the print and electronic books at medium price each, and no discount on the bundle is given.
97
<< Insert Table 14 and Table 15 about here >>
Incomplete product line (traditional mixed bundling with no bundle discount) This case is similar to the previous one with the only difference being that consumers do not get additional discount for choosing the bundle and the price they are charged for the bundle is the sum of the two individual prices. We consider nine possible market situations in total. The results are presented in tables 16 and 17 (book and newspaper categories respectively). In the book category, the market scenario in which the print book is at high price and the electronic book is at low price is consistently the most profitable scenario, followed by the scenario in which the print price is at medium level and the electronic price is at low level. With regard to the newspaper category, the electronic product should be offered at low price, while the price of the print product depends on the variable print costs as percentage of the print price – when the costs are relatively low, the print product should be offered at low price, but when the costs are relatively high, the print product should be offered at high price. << Insert Table 16 and Table 17 about here >>
Comparing the complete and incomplete product line profits Looking at the overall profits of the most appealing market scenarios in the book category listed in Tables 12, 14 and 16, we can conclude that the publisher generates the highest profit when the incomplete product line with bundle discount is chosen (Table 14); this holds regardless of the print variable costs. For example, at the 70% cost level, the overall profits for market scenario #181 are as follows: $12.12 (incomplete product
98
line with bundle discount, Table 14), $11.93 (complete product line, Table 12) and $10.82 (incomplete product line with no bundle discount, Table 16). The results are consistent for all most profitable market scenarios and cost levels. With regard to the newspaper category, the results are very similar with one exception – the complete product line is the most profitable strategy when the print variable costs are relatively high (e.g. 80%-90% of the medium print price). For example, at the 80% cost level, the overall profits for market scenario #184 are as follows: $27.36 (complete product line, Table 13), $27.04 (incomplete product line with bundle discount, Table 15) and $26.75 (incomplete product line with no bundle discount, Table 17). When the variable print costs are relatively low or medium level (e.g. 40%-50%-60%-70%), the results are as in the book category: incomplete product line with bundle discount is the best option, followed by the complete product line, and then by the incomplete product line with no bundle discount. Thus, the incomplete product line with discount leads to the highest profit in both product categories. We further elaborate on this result in the discussion section.
Relative Form Attractiveness and Choice Recall that the respondents evaluated the relative attractiveness of print book as compared to electronic (PDF) book and print newspaper as compared to electronic (online) newspaper on five attributes – image quality, browsing, layout, archival quality and convenience of use (7-point scale, 1-electronic better than print, 7-print better than electronic). They also provided information on the importance of each of these attributes when making a choice within the specific product category (7-point scale, 1-not at all
99
important, 7-very important). We mean-centered the five scales and computed four index variables for each of the two categories: print form index, electronic form index, difference index (bundle vs. superior form) and difference index (print form vs. electronic form). The example that follows is an illustration how the index variables were computed. BOOK Image Quality (IQ) Browsing (B) Layout (L) Archival Quality (AQ) Convenience of Use (CU) 7 3 2
Attribute 1 2 4 4 Perception Mean Centered (a) -3 -2 0 0 Importance (b) 2 5 5 2 Print Index (PI) = IQ(a*b) + B(a*b) + L(a*b) + AQ(a*b) + CU(a*b) Print Index (PI) = (-3)*2 + (-2)*5 + 0*5 + 0*2 + 3*2 = - 10 Electronic Index (EI) = - Print Index Electronic Index (EI) = 3*2 + 2*5 + 0*5 + 0*2 + (-3)*2 = 10 Bundle Index (BI)= Sum of Max (Print, Electronic) for IQ, B, L, AQ, CU Bundle Index (BI)= 3*2 + 2*5 + 0*5 + 0*2 + 3*2 = 22 Difference Index 1 (DI1) = Bundle Index – Max (Print Index, Electronic Index) Difference Index 1 (DI1) = 22 – 10 = 12 Difference Index 2 (DI2) = |Print Index – Electronic Index| = |10 – (-10)| = 20
We estimate a random parameters logit model for each product category using the following utility functions: U of Print Product (PP) = bPP + bPRC *Price + bPI*PI U of Electronic Product (EP) = bEP + bPRC * Price + bEI_1*EI U of Electronic Unit (EU) = bEU + bPRC * Price + bEI_2 * EI U of Print Product & Electronic Product (PPEP) = bPPEP + bPRC * Price + bDI1_1*DI1 U of Print Product & Electronic Unit (PPEU) = bPPEU + bPRC * Price + b DI1_2* DI1 U of No-Choice = bPRC* Price + bDI2* DI2 << Insert Table 18 and Table 19 about here >>
100
The results are displayed in Table 18 (book category) and Table 19 (newspaper category)25. Our goal is to discover whether the relative product form attractiveness influences choice over and above the preference for the different product offers and their price. In the book category the probability of choosing the print product is significantly positively related to the Print Index (bPI = 0.029, p < .05), for both categories the probability of choosing the electronic product is significantly positively related to the Electronic Index (bEI_1 = 0.029 for the book category and bEI_1 = 0.011 for the newspaper category, p’s < .05), and the probability of choosing the electronic unit is significantly positively related to the Electronic Index (bEI_2 = 0.013 for the book category and bEI_1 = 0.007 for the newspaper category, p’s < .05). With respect to the two difference indices, the results reveal that the difference in attractiveness between the bundle and the more attractive product form significantly affect the probability of buying the print product/electronic product bundle in the book category only (bDI1_1 = 0.107, , p < .05); the difference in attractiveness of the two forms significantly affects the probability to defer choice also in the book category only (bDI2 = 0.023, p < .05). For the book category we estimate individual level parameters for the intrinsic preference for the print form, the PDF form and the bundle. We then regress the differences in the constants for pairs of alternatives for each individual for differences in the indexes. Specifically, we compute the difference between the print and the PDF book alternative specific constants, and then regress the difference on the Difference Index 2; also, we compute the difference between the alternative specific constants of the bundle
25
In the book category we estimate a model in which the means of the price parameter and the alternative specific constants for print book, PDF book and print book/PDF book bundle are allowed to vary and are assumed to be normally distributed, while in the newspaper category only the price parameter is allowed to vary. These are the models that provide the best fit to the data.
101
and the individual form, and then regress this difference on the Difference Index 1. In both cases the regression coefficients are negative but marginal or non-significant (0.007, p>.10, and -0.022, p = 0.099). Nevertheless, these results provide weak evidence that the indexes are related to the brand specific constants: the higher the value of the indexes, the lower the difference between the brand specific constants. This suggests that the relative attractiveness of the product forms explains at least a portion of the difference in the intrinsic preference for these product forms. In sum, controlling for the preference for the various product offers and their prices, the more consumers find the specific form attractive and superior to the other form, the more likely they are to choose it. Furthermore, the probability of buying the bundle in the book category is positively associated with the difference in the attractiveness between the bundle and the more attractive product form.
Discussion
Several issues are important from both theoretical and managerial point of view. Why is offering the incomplete product line with discount more profitable than offering the complete product line? Our book category simulation results suggest that when the electronic units are introduced, customers who would otherwise buy the print book or the full electronic book switch to the electronic units option. Further, the no choice option market share decreases only slightly as the unbundled electronic units do not attract a significant number of new customers. Consequently, there is cannibalization of print and full electronic books sales to unbundled electronic units sales, and this loss in revenue is not compensated by the revenue from the new customers who enter the market when the
102
electronic units are offered. In the newspaper category the results follow a similar pattern. When the electronic units are introduced, consumers who would otherwise buy the full electronic product choose a single unit instead. Again, the no-choice option share decreases only slightly. This results in loss of revenue because of the cannibalization of full electronic product sales to electronic units sales, and because of the limited market expansion from selling electronic units. Why is the incomplete product line with no discount worse than the other two strategies? In the book category most of the consumers who buy the bundle in the incomplete with discount case buy print only when there is no bundle discount. Because of this, the company is better off charging a discounted bundle price as it brings extra revenue without incurring extra cost. Regarding the newspaper purchases, most of the consumers who buy the bundle in the incomplete product line with discount now switch to print newspaper or electronic newspaper only. As offering the electronic newspaper as a part of the bundle does not involve extra costs, it is more profitable for the company to sell the bundle. Thus, our simulation results show that when the unbundling of electronic content does not lead to attracting a sizeable amount of new customers or getting extra revenue from the current customers, it is more profitable for the company not to pursue such an approach. The optimal solution is to offer the print products at a high price to extract more revenue from the customers who have a high valuation for this form, and to offer the electronic products and the bundle of the two forms at low price. This way the company can still persuade the more price sensitive consumers to enter the market and buy the cheaper electronic products, and extract additional revenue from those consumers
103
who value the bundle and buy both forms when the bundle is offered at a discount. The optimal pricing strategy is as follows: book category – print book at $33.40, electronic book at $25.20, and bundle at $43.80 (33% effective bundle discount); newspaper category – print newspaper at $114.50, electronic newspaper at $34.50, and bundle at $105.20 (42% effective bundle discount). The conclusions of Essay 2 and Essay 3 are complementing each other. Recall that in Essay 2 consumers do not receive discount for buying the bundle. Therefore, it is more profitable for the company to offer unbundled electronic units (content unbundled mixed bundling) as compared to full electronic books (traditional mixed bundling). In Essay 3 we also compare these two strategies, although they are not exactly the same as in Essay 2: offering the complete product line (which is similar to CUMB but not the same, see explanation bellow) vs. offering the incomplete product line without bundle discount (which is TMB in Essay 2). The results in Essay 3 are similar to those in Essay 2 – the profit is higher when unbundled electronic units are offered (but lower than TMB with discount), while TMB with no discount is the least profitable. How do our findings relate to the extant literature on bundling? Previous studies suggest that mixed bundling is more beneficial than pure bundling and pure components is the optimal bundling strategy when there is asymmetry in the reservation prices for the bundle components (Adams and Yallen 1976; Schmalensee 1984; McAfee, McMillan and Whinston 1989). Further, with respect to interrelated products and when the marginal costs are not too high, Venkatesh and Kamakura (2003) argue that while pure components strategy is optimal for moderate to strong substitutes and complements, mixed bundling is optimal for weak substitutes/complements. We extend the above
104
findings by further clarifying which bundling strategy is more appealing when different mixed bundling strategies are feasible. We show that, applied to product forms, mixed bundling is the optimal strategy when full print and electronic products are offered; in the case of separate electronic units companies are better off pursuing pure bundling (which is offering the full electronic product). Also, similar to Wilson et al. (1990) we show that when there is no market growth from unbundling and new market segments are not attracted by the unbundled components, product bundles should not be unbundled. In terms of optimal pricing, Jedidi et al. (2003) argue that when the reservation prices of the bundle components are highly heterogeneous, a uniform high price for all products and bundles is optimal; otherwise a hybrid strategy may work better. In addition, Venkatesh and Chatterjee (2003) claim that low-priced components should be targeted at consumers who have low value for the information content. Our findings suggest that, in the case of mixed bundling of different forms of information products, it is preferable to charge high price for the print product but low price for the full electronic product and the bundle. As the costs associated with the electronic products are negligible, it is beneficial for the company to sell them at a relatively low price to consumers who have a low value for the information content and attract them to make a purchase. Thus, we show that it is not always optimal to charge a high price for the individual products under mixed bundling, and the cost structure may play a role as well. In sum, we show that with respect to product forms, the mixed bundling strategy of offering the individual product forms and the bundle is the optimal product strategy for a content provider to pursue. Adding to the existing literature on bundling of interrelated products, we show that mixed bundling is optimal when the valuations of the bundle
105
components vary significantly across consumers, as well as the contingent valuation for the bundle. We also show that the cost structure may play an important role in bundling decisions, especially when determining the optimal pricing strategy – it may allow the seller to generate higher profit in the case of mixed bundling versus pure components (incomplete product line with discount versus incomplete product line with no discount) even when the bundle discount is sizeable. Specifically, companies are better off selling the bundle at a discounted price as compared to selling only the individual forms – from a seller’s point of view the bundle does not involve extra cost over and above the cost of the print form but can generate extra revenue over and above the print form revenue.
Managerial Implications
Our study has important managerial implications in the areas of product line design and pricing. We offer guidelines to publishers and other content providers on how to profitably market digital content together with traditional print content. Companies should implement mixed bundling regarding lines of information products, charging a high price for the print products and offering the electronic products and the bundle at a low price. Additionally, we outline a choice-based conjoint methodology that can be applied to a wide range of products including magazines, journals, music, etc. Second, our study has implications in the area of market segmentation and targeting. Based on the estimates that can be obtained about the valuations of the different product forms, companies can segment their customers and target their product offering and pricing accordingly. For example, consumers who have low valuation for the information content may be encouraged to try and buy lower priced electronic products,
106
while customers with high valuation for the information content may be encouraged to buy the bundle. Further, targeting can be applied at individual level as well resulting in personalization of product offering and customization of digital goods.
Conclusion
In this essay we use choice experiments and choice-based conjoint analysis to obtain estimates of customer valuations of the price and non-price attributes of interest. We present respondents with hypothetical choice situations in which the attributes are varied to determine how choice fluctuates with attribute changes. Then, controlling for consumer heterogeneity, we estimate random effects logit models and use the results of the conjoint analysis to derive the optimal product mix and pricing. Consequently, we are able to come up with recommendations for optimal product line design and pricing strategy. Our empirical investigation and conclusions are conditional on the content provider being a monopolist. It will be interesting to investigate what the optimal product line and pricing strategies will be when competition exists on the marketplace. In some conditions it may be optimal to offer unbundled electronic units, and complete product line may be the market equilibrium. The rational behind such an equilibrium solution may be that offering more product options results in higher customer satisfaction and likelihood of repeat purchase. Another interesting question is how the market norms affect the expectations of the consumers on how the print content, the electronic content and the bundle are priced. Although in our study these norms are taken into account when modeling the valuations of the consumers for the different product offers and
107
consequently in the market simulations, it will be beneficial for the marketing theory and practice to further investigate the long term effects of offering low priced electronic products and form bundles on the valuation of information products, as well as the unbundling of electronic content.
108
TABLES Table 1: Purchase Probabilities for Bundle: Product, Discount and Usage Conditions
Panel A. Product, Discount and Usage Conditions Discount No Discount d% Discount Predictions Bundle INFO CONV INFO CONV Same Usage (1a) Pr (VFj|Fi > Pj) (1b) Pr (VC+VFj|Fi> Pj) (2a) Pr (VFj|Fi + d(Pi + Pj) > Pj) (2b) Pr (VC+VFj|Fi+d(Pi + Pj) > Pj) INFO: 1a=3a, 1a=2a, 2a<4a, 3a<4a CONV > INFO, all conditions Panel B. Discount and Relative Price Conditions – Information Products Only Price of each item same (PA) No Discount d % Discount Predictions (5) Pr (VFj|Fi > PA | VFi >VFj) + Pr (VFi|Fj > PA | VFj > VFi) (6) Pr (VFj|Fi > (1 - 2d)PA | VFi > VFj) + Pr (VFi|Fj > (1 - 2d)PA | VFj > VFi) 7>5, 8 > 6, 6>5, 8>7 Item i higher priced: Pi = (1+k)PA, Pj = (1-k)PA (7) Pr (VFj|Fi > (1-k) PA | VFi-2k PA >VFj) + Pr (VFi|Fj > (1+k)PA | VFi-2k PA < VFj) (8) Pr (VFj|Fi > (1 – k - 2d)PA | VFi - 2kPA >VFj) + Pr (VFi|Fj > (1 + k - 2d)PA | VFi - 2kPA < VFj) Different Usage (3a) Pr (V*Fj|Fi > Pj) (3b) Pr (VC+V*Fj|Fi> Pj) (4a) Pr (V*Fj|Fi + d(Pi + Pj) > Pj) (4b) Pr (VC+V*Fj|Fi+d(Pi + Pj) > Pj) CONV: 1b<3b, 2b<4b, 1b<2b, 3b<4b
109
Table 2: Perceived Appropriateness of Generated Usage Situations
Usage Situation MD** Rank e-mail pages/chapters -2.425 1 give as a present 2.238 2 Form 1: read for pleasure 1.762 3 PAPER BOOK read while traveling 1.738 4 Form 2: read to others 1.452 5 ELECTRONIC search -0.952 6 BOOK copy citations/paragraphs -0.600 7 need on a short notice 0.167 8 archive -0.024 9 read the news -2.548 1 Form 1: read old articles -1.902 2 PAPER read for pleasure -1.690 3 NEWSPAPER read during lunchtime/breaks 1.476 4 Form 2: archive -1.171 5 ONLINE read at home 0.805 6 NEWSPAPER search 0.756 7 read to others 0.463 8 follow the stock market -0.375 9 e-mail articles 0.310 10 for cooking 1.071 1 Form 1: for baking 0.976 2 STICK of shortening in cakes 0.921 3 MARGARINE take on a picnic -0.350 4 Form 2: eat with pancakes 0.262 5 TUB of spread on a tray -0.093 6 MARGARINE spread on bread for breakfast -0.048 7 for frying fish/meat -0.025 8 wash dishes 3.000 1 Form 1: wash your hands in public places 2.372 2 LIQUID SOAP wash delicate laundry 1.930 3 Form 2: take on a picnic 0.767 4 BAR of SOAP take a shower at home -0.628 5 take a shower while traveling -0.349 6 wash your hands at home 0.302 7 take a shower in the gym 0.140 8 when looking for a good cup of coffee -2.770 1 offer to your guests at home -2.580 2 Form 1: while traveling 1.130 3 INSTANT COFFEE when in a hurry 1.040 4 Form 2: on a picnic -0.480 5 GROUND COFFEE during a meeting -0.420 6 while in class -0.190 7 in the middle of the day -0.170 8 *p<0.05 **The mean difference calculated as mean appropriateness of Form 1 minus mean appropriateness of Form 2. Product t stat -12.825 10.408 7.638 6.581 6.203 -4.421 -2.399 0.510 -0.100 -11.790 -7.963 -7.275 6.173 -4.566 3.778 4.036 2.649 -1.684 1.915 4.079 3.992 3.441 -1.236 1.097 -0.371 -0.184 -0.097 17.037 11.715 8.908 2.760 -2.071 -1.044 2.383 0.374 -9.222 -9.259 3.191 2.982 -1.297 -1.305 -0.548 -0.848 Rank 1 2 3 4 5 6 7 8 9 1 2 3 4 5 7 6 8 10 9 1 2 3 4 5 6 7 8 1 2 3 4 6 7 5 8 1 2 3 4 6 5 8 7 * * * * * * *
* * * * * * * *
* * *
* * * * * * * * * *
110
Table 3: Study 2 - Mean Percentage of Points Allocated to Bundle Panel A: Bundle Discount and Usage Conditions Discount Bundle No Discount INFO CONV 25 % Discount INFO CONV Predictions Product Book Newspaper Margarine Coffee Book Newspaper Margarine Coffee Same Usage (1a) 9.77% (1a) 9.17% (1b) 15.10% (1b) 12.30% (2a) 9.18% (2a) 11.36% (2b) 26.85% (2b) 31.00% INFO: 1a=3a, 1a=2a, 2a<4a, 3a<4a CONV > INFO, all conditions Different Usage (3a) 8.28% (3a) 6.35% (3b) 23.75% (3b) 17.61% (4a) 20.08% (4a) 18.90% (4b) 49.50% (4b) 48.19% CONV: 1b<3b, 2b<4b, 1b<2b, 3b<4b
Panel B: Bundle Discount and Relative Price Conditions Discount No Discount INFO 25 % Discount Predictions
a
INFO
Booka Newspaperb Bookc Newspaperd
Equal Prices (5) 6.10% (5) 6.88% (6) 15.68% (6) 16.23% 7>5, 8 > 6, 6>5, 8>7
Print High. Electronic Low (7) 11.95% (7) 8.63% (8) 13.58% (8) 14.07%
Prices for book in equal condition: print = electronic= $29.99, bundle = $59.98; in highlow condition: print = $37.49, electronic = $22.49, bundle = $59.98. b Prices for newspaper in equal condition: print = electronic = $38.99, bundle = $77.97; in high-low condition: print = $48.79, electronic = $29.19, bundle = $77.97. c Prices are same as a except bundle = $44.99. d Prices are same as b except bundle = $58.49.
111
Table 4: Study 2 ANOVA Results
Source USAGE (USE) DISCOUNT (DISC) REL. PRICE (RPRC) USE * DISC USE * RPRC DISC * RPRC USE * DISC * RPRC Error Total
SS 26.67 40.99 0.72
df
MARGARINE MS F 1 26.67 6.01 1 40.99 9.24 1 0.72 0.16
Sig. SS 0.02 22.53 0.00 130.31 0.69 9.76
df
COFFEE MS F 1 22.53 4.09 1 130.31 23.65 1 9.76 1.77
Sig. 0.05 0.00 0.19
337.24 616.25
76 80
4.44
451.87 928.23
82 86
5.51
Source SS USAGE (USE) 16.92 DISCOUNT (DISC) 32.68 REL. PRICE (RPRC) 2.97 USE * DISC 27.11 USE * RPRC 5.24 DISC * RPRC 16.84 USE * DISC * RPRC 0.91 Error 1042.35 Total 1145.02
DV: Ln of points allocated to the bundle
df 1 1 1 1 1 1 1 232 240
BOOK MS 16.92 32.68 2.97 27.11 5.24 16.84 0.91 4.49
F 3.77 7.27 0.66 6.03 1.17 3.75 0.20
Sig. 0.05 0.01 0.42 0.01 0.28 0.05 0.65
NEWSPAPER SS df MS F 1.49 1 1.49 0.33 48.46 1 48.46 10.54 1.02 1 1.02 0.22 17.96 1 17.96 3.91 2.17 1 2.17 0.47 9.89 1 9.89 2.15 0.94 1 0.94 0.20 1066.50 232 4.60 1148.43 240
Sig. 0.57 0.00 0.64 0.05 0.49 0.14 0.65
112
Table 5: Profit Comparison under CUMB and TMB Strategies.
Print Sales a.
CUMB TMB ? QP > QP ? ? ?;
CUMB ? QP ? ? PC
TMB ? QP ? ? PEB
* ?* CUMB > ? TMB
IF
? ? ? Q CUMB P ?? ? ?P ? C ?? ? ? kFN ? ? ? ?? ? ? ( P ? c) ? ? ? P ?
2 2
? ?
? ? kFN ? ? kFN ;+?? ?U? ? PP ? c ? ? PP ? c ?
? ? ?k ? ? k ? ? ? ? ?; P ? c ? U ? P ? c ;+?? P ? ? ? ? P
CUMB TMB ? QP ? QP ? ? PP ? c ?
? VH ?+ ? 4k ?
?? ? Q TMB P ?? ? ?P ? EB ??
? ? k ? ? ?? ? ? ? P ?c? ? ? ? P
2
2
? 1 ?> 4k F S ? ?
? ? ? ?
b.
CUMB TMB ? QP > QP ?
? ? kFN kFN ? ? ; ? ? PP ? c PP ? c ?
? ? ?k ? ? k ? * ? Always ? * ? ? ?; P ? c ? U ? P ? c ;+?? CUMB > ? TMB P ? ? ? ? P
c.
Q
CUMB P
>Q
TMB P
? ? ? kFN ? ? kFN ? ? ? ? ?; P ? c ? U ? P ? c ;+?? P ? ? ? P ?
? ?k k ? ? ; ? ?P ?c P ?c? P ? ? P
? ?k k ? ? ? ? P ?c;P ?c? P ? ? P
CUMB TMB ? QP ? QP ? ? PP ? c ?
? 1 ?> ? 4k F S ?
? 1 ?+ ? 4k F S ?
? ? ? Q CUMB P ?? ? ?P ? C ? ?
? ? kFN ? ? ? ?? ? ? ( P ? c) ? ? ? P ?
? ? ? ?
2
2
? V ?+ H 4k ? ?
?? k ? 2 ? ? Q TMB P ? ?? ? P ?c? ? ?? ? P ? P EB ? ? ? ?
?? k ? 2 ? ? Q TMB P ? ?? ? P ?c? ? ?? ? P ? EB ? ? ?? P
2
? ? ? ?
? ? ? ?
2
? ? ? ?
? ? ? ?
? ? ? ?
? ? ? ?
d.
Q
CUMB P
>Q
TMB P
? ? kFN kFN ? ? ? ? P ?c ;P ?c? P ? ? P
? ? ? kFN ? ? kFN ? ? ? ? ?; P ? c ? U ? P ? c ;+?? P ? ? ? P ?
CUMB TMB ? QP ? QP ? ? PP ? c ?
?? k F N ? 2 ? ? Q CUMB P ? ? ?? ? ( P ? c) ? ? ? ? P ? C ? ? ?? P
2
2
? V ?> H 4k ? ?
2
e.
Q
CUMB P
?Q
TMB P
TMB 2 ? V H ?? ? Q P ? ?k ? ? k ? ? ? U +? ? ? ; ; ? ? ? ? ? ? ? ? ? P ? c P c ? PEB ? P ? 4k ? ? ? ? P ? ? ?
? k ? ?? ? P ?c? ? ? P ?
? ? Q TMB ? Q CUMB P ?>? P ? PP ? c ? ? ?
? ? ? ?
2
? 1 ?+ ? 4k F S ?
2
? ? ? Q CUMB P ?? ? ?P ? C ? ?
2
? ? kFN ? ? ? ?? ? ? ( P ? c) ? ? ? P ?
2
f.
Q
CUMB P
?Q
TMB P
? ? kFN kFN ? ? ? ? P ?c ;P ?c? P ? ? P
? ? ? kFN ? ? kFN ? ? ? ? ?; P ? c ? U ? P ? c ;+?? P ? ? ? P ?
? ? ?k ? ? k 1 ? ? ? ?; P ? c ? U ? P ? c ;+?? ? P ? ? ? P ? 4k F S
? ? k F N ? 2 ? ? Q CUMB P ? ? ?? ? ( P ? c) ? ? ? ? P ? C ? ? ?? P
? V ?+ H 4k ? ?
?? ? Q TMB P ?? ? ?P ? EB ??
? ? k ? ? ?? ? ? ? P ?c? ? ? ? P
? ? Q TMB ? Q CUMB P ?>? P ? P c ? ? P ? ?
g.
CUMB TMB QP ? QP
? ?k k ? ? ? ? P ?c;P ?c? P ? ? P
? ?k k ? ? ; ? ?P ?c P ?c? P ? ? P
* Always ? * CUMB < ? TMB
h.
Q
CUMB P
?Q
TMB P
? ? kFN kFN ? ? ? ? P ?c ;P ?c? P ? ? P
1 4kF S
?? k F N ? 2 ? ? Q CUMB P ? ? ?? ? ( P ? c) ? ? ? ? P ? C P ? ? ? ?
? ? ? ?
2
? ? Q TMB ? Q CUMB P ?>? P ? PP ? c ? ? ?
? VH ?+ ? 4k ?
?? k ? 2 ? ? Q TMB P ? ?? ? P ?c? ? ?? ? P ? P EB ? ? ? ?
? ? ? ?
2
? ? ? ?
113
Table 6: Actual Purchase Rates under the Two Mixed Bundling Strategies
Content Unbundled Mixed Bundling PDF Price 0.50 0.75 1.00 1.25 Total Print 57.21% 54.11% 63.55% 62.88% 59.57% PDF 11.21% 13.59% 9.47% 6.67% 10.08% Bundle 4.95% 3.93% 4.87% 3.13% 4.13% Nothing 26.64% 28.37% 22.11% 27.32% 26.22% Total 21.17% 25.86% 23.42% 29.55% 100.00%
Traditional Mixed Bundling PDF Price 0.25 0.50 0.75 1.00 1.10 Total Print 43.57% 45.20% 51.60% 48.57% 40.00% 47.37% PDF 16.43% 17.65% 8.33% 5.00% 11.43% 12.32% Bundle 7.14% 4.95% 6.73% 5.00% 5.71% 5.89% Nothing 32.86% 32.20% 33.33% 41.43% 42.86% 34.42% Total 14.74% 34.00% 32.84% 14.74% 3.68% 100.00%
114
Table 7: Estimation Results - Content Unbundled Mixed Bundling
Variable Agriculture Choice Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Coefficient 0.486 0.232 -0.860 0.391 -0.341 -0.873 0.515 -0.311 -0.672 0.262 -0.209 0.739 0.129 0.057 0.235 0.000 -0.017 -0.006 0.902 -0.553 -5.802 SE 0.350 0.412 1.146 0.173 0.202 0.590 0.192 0.229 0.569 0.306 0.358 0.907 0.142 0.163 0.384 0.006 0.006 0.021 0.285 0.179 3.364 b/St.Er. 1.389 0.565 -0.750 2.262 -1.691 -1.479 2.684 -1.357 -1.181 0.855 -0.584 0.814 0.907 0.353 0.612 0.038 -2.672 -0.279 3.169 -3.092 -1.725 2.142 1.985 3245 43 -3290.938 -4498.525 2415.175 23 0.000 P[|Z|>z] 0.165 0.572 0.453 0.024 0.091 0.139 0.007 0.175 0.238 0.393 0.559 0.416 0.364 0.724 0.541 0.970 0.008 0.780 0.002 0.002 0.085 0.032 0.047
Behavioral Science
Education
General Interest
Medicine
Print Price
PDF Price
PDF Price
Derived SD of parameter distributions Print 0.819 1.753 Bundle 1.958 3.887 Number of observations Iterations completed Log likelihood function Restricted log likelihood Chi squared Degrees of freedom Prob[ChiSqd > value] =
115
Table 8: Estimation Results - Traditional Mixed Bundling
Variable Agriculture Choice Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Coefficient -0.317 -0.732 -0.487 0.504 -2.339 -0.235 0.974 -0.003 0.321 0.252 -0.266 0.085 -0.099 -0.285 -0.661 0.011 -0.001 -0.035 -0.245 -1.252 -0.815 SE 0.337 0.551 0.812 0.283 1.036 0.542 0.267 0.424 0.481 0.290 0.454 0.480 0.187 0.258 0.384 0.006 0.008 0.012 0.262 0.397 0.508 b/St.Er. -0.940 -1.329 -0.599 1.780 -2.257 -0.434 3.649 -0.007 0.668 0.869 -0.587 0.177 -0.531 -1.102 -1.723 1.802 -0.151 -2.992 -0.934 -3.156 -1.605 950 7 -1056.12 -1316.98 -1088.57 P[|Z|>z] 0.347 0.184 0.549 0.075 0.024 0.665 0.000 0.995 0.504 0.385 0.557 0.860 0.596 0.270 0.085 0.072 0.880 0.003 0.351 0.002 0.109
Behavioral Science
Education
General Interest
Medicine
Print Price
PDF Price
Number of observations Iterations completed Log likelihood function No coefficients Constants only
116
Table 9: Simulation Results – Profit under the Two Mixed Bundling Strategies. A. Content Unbundled Mixed Bundling
PDF Price as % Print Price 0% 25% 50% 75% 100% 110% 125% Actual Market Share PDF Bundle No Choice 0.166 0.214 0.273 0.161 0.061 0.301 0.133 0.021 0.284 0.105 0.014 0.257 0.082 0.014 0.231 0.075 0.014 0.222 0.065 0.015 0.210 0.094 0.017 0.243 Profit PDF 0.000 0.833 1.376 1.631 1.708 1.709 1.690 1.609
Print 0.347 0.477 0.561 0.624 0.672 0.689 0.710 0.646
Print 6.295 8.655 10.188 11.310 12.185 12.477 12.862 11.653
Bundle 3.910 1.435 0.607 0.471 0.536 0.581 0.655 0.629
Total 10.205 10.923 12.170 13.412 14.429 14.767 15.207 13.890
B. Traditional Mixed Bundling
PDF Price as % Print Price 0% 25% 50% 75% 100% 110% 125% Actual Market Share PDF Bundle No Choice 0.203 0.079 0.277 0.165 0.072 0.306 0.132 0.064 0.335 0.105 0.057 0.362 0.082 0.050 0.387 0.075 0.047 0.397 0.064 0.043 0.412 0.120 0.059 0.349 Profit PDF 0.000 1.302 2.084 2.475 2.588 2.577 2.517 2.141
Print 0.441 0.457 0.469 0.477 0.481 0.481 0.481 0.472
Print 7.921 8.219 8.431 8.562 8.618 8.623 8.611 8.455
Bundle 1.231 1.390 1.490 1.536 1.537 1.527 1.502 1.492
Total 9.152 10.911 12.005 12.572 12.743 12.727 12.630 12.088
117
Table 10: Conjoint Design and Price Levels
A. Conjoint Factors Product offer
Print 1 0 0 1 1
Print Product (PP) Electronic Product (EP) Electronic Unit/Section (EU) Print Product & Electronic Product Bundle (PPEP) Print Product & Electronic Unit Bundle (PPEU) B. Price Levels WSJ Subscription (26 weeks) Print WSJ Electronic WSJ Electronic WSJ Section Print WSJ & Electronic WSJ Print WSJ & Electronic WSJ Section Vault Guide Print Book Electronic Book Electronic Section Print Book & Electronic Book Print Book & Electronic Section
Full Electronic 0 1 0 1 0
Electronic Bundle Price Unit Level 0 0 L, M, H 0 0 L, M, H 1 0 L, M, H 0 1 L, M, H 1 1 L, M, H
Low 86.50 34.50 13.20 105.20 94.90 Low 25.20 25.20 9.70 43.80 33.20
Medium 99.50 39.50 15.20 120.90 109.20 Medium 29.00 29.00 11.10 50.40 38.20
High 114.50 45.50 17.40 139.10 126.60 High 33.40 33.40 12.80 58.00 44.00
118
Table 11: Random Parameters Logit Model Results (conjoint choice experiment)
Book Category Coefficient SE b/SE Random parametes in utility functions Bundle -6.647 0.534 -12.457 Print 8.485 0.878 9.666 Price -0.261 0.028 -9.378 Full Electronic 7.306 0.793 9.210 Electronic Unit 3.063 0.419 7.314 Bundle Print Price Full Electronic Electronic Unit Variable P[|Z|>z] 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Derived standard deviations of prameter distributions 2.236 0.365 6.127 5.249 0.515 10.193 0.107 0.013 8.494 1.621 0.296 5.469 2.316 0.230 10.068
Number of observations Log likelihood function Restricted log likelihood Chi squared (df=10) R-sq Adjusted R-sq Newspaper Category
4176 (87 respondents x 16 choices x 3 options) -904.67 -2486.96 3164.58 (p=0.000) 0.636 0.635
Coefficient SE b/SE Random parametes in utility functions Bundle -3.384 0.572 -5.921 Print 11.333 1.376 8.235 Price -0.133 0.014 -9.575 Full Electronic 6.793 0.619 10.968 Electronic Unit 1.452 0.380 3.819 Bundle Print Price Full Electronic Electronic Unit
Variable
P[|Z|>z] 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Derived standard deviations of prameter distributions 4.005 0.565 7.083 6.222 0.720 8.646 0.025 0.004 6.157 3.253 0.502 6.475 3.515 0.357 9.852
Number of observations Log likelihood function Restricted log likelihood Chi squared (df=10) R-sq Adjusted R-sq
4176 (87 respondents x 16 choices x 3 options) -783.75 -2490.55 3413.59 (p=0.000) 0.685 0.684
119
Table 12: Book Category – Market Simulation Results
Market Share PDF Print Book Print Book No Revenue Unit and and Choice PDF Book PDF Unit 0.141 0.144 0.164 0.125 0.146 0.141 0.144 0.164 0.125 0.146 0.141 0.144 0.164 0.125 0.146 0.141 0.144 0.164 0.125 0.146 0.141 0.144 0.164 0.125 0.146 0.141 0.144 0.164 0.146 0.125 0.094 0.099 0.092 0.065 0.064 0.094 0.099 0.092 0.065 0.064 0.094 0.099 0.092 0.065 0.064 0.094 0.099 0.092 0.065 0.064 0.094 0.099 0.092 0.065 0.064 0.094 0.099 0.092 0.064 0.065 0.039 0.016 0.037 0.025 0.024 0.039 0.016 0.037 0.025 0.024 0.039 0.016 0.037 0.025 0.024 0.039 0.016 0.037 0.025 0.024 0.039 0.016 0.037 0.025 0.024 0.039 0.016 0.037 0.024 0.025 0.181 0.182 0.175 0.163 0.159 0.181 0.182 0.175 0.163 0.159 0.181 0.182 0.175 0.163 0.159 0.181 0.182 0.175 0.163 0.159 0.181 0.182 0.175 0.163 0.159 0.181 0.182 0.175 0.159 0.163 20.939 20.502 20.442 20.699 20.446 20.939 20.502 20.442 20.699 20.446 20.939 20.502 20.442 20.699 20.446
Rank
Market Scenario # 181 182 172 100 91 181 182 172 100 91 181 182 172 100 91 181 182 172 100 91 181 182 172 100 91 181 182 172 91 100
Print Book
PDF Book
Cost
Cost % Profit of Print Price
1 2 3 12 15 1 2 3 15 18 1 2 3 22 23 1 2 3 30 33 1 2 3 37 38 1 2 3 37 38
0.311 0.322 0.306 0.420 0.413 0.311 0.322 0.306 0.420 0.413 0.311 0.322 0.306 0.420 0.413 0.311 0.322 0.306 0.420 0.413 0.311 0.322 0.306 0.420 0.413 0.311 0.322 0.306 0.413 0.420
0.234 0.236 0.225 0.201 0.194 0.234 0.236 0.225 0.201 0.194 0.234 0.236 0.225 0.201 0.194 0.234 0.236 0.225 0.201 0.194 0.234 0.236 0.225 0.201 0.194 0.234 0.236 0.225 0.194 0.201
5.149 15.790 5.069 15.433 5.051 40.00% 15.390 5.926 14.773 5.813 14.634 6.436 14.503 6.337 14.165 6.314 50.00% 14.127 7.408 13.292 7.266 13.181 7.723 13.216 7.604 12.898 7.577 60.00% 12.865 8.889 11.810 8.719 11.727
20.939 9.011 11.929 20.502 8.872 11.630 20.442 8.840 70.00% 11.602 20.699 10.371 10.328 20.446 10.172 10.274 20.939 20.502 20.442 20.699 20.446 20.939 20.502 20.442 20.446 20.699 10.298 10.641 10.139 10.363 10.103 80.00% 10.339 11.852 8.847 11.625 8.821 11.585 9.354 11.406 9.096 11.366 90.00% 9.076 13.078 7.368 13.334 7.365
Market Scenario Print Book PDF Book PDF Unit Print Book Print Book # PDF Book PDF Unit 91 Medium 100 Medium 172 High 181 High 182 High Low Low Low Low Low Medium High Medium High High Low Low Low Low Low Low Low Low Low Medium
120
Table 13: Newspaper Category – Market Simulation Results
Rank Market Scenario # 181 208 19 100 91 181 100 208 172 91 181 172 182 100 91 181 184 182 100 91 187 178 184 100 101 187 178 184 106 97 Print WSJ Low Medium Medium Medium Medium Medium High Print WSJ Online WSJ Market Share Online Print WSJ Print WSJ No Revenue Unit and and Choice Online WSJ Online Unit 0.103 0.123 0.096 0.101 0.115 0.103 0.101 0.123 0.116 0.115 0.103 0.116 0.105 0.101 0.115 0.103 0.108 0.105 0.101 0.115 0.109 0.124 0.108 0.101 0.103 0.109 0.124 0.108 0.107 0.122 0.259 0.284 0.206 0.242 0.241 0.259 0.242 0.284 0.258 0.241 0.259 0.258 0.269 0.242 0.241 0.259 0.142 0.269 0.242 0.241 0.067 0.067 0.142 0.242 0.249 0.067 0.067 0.142 0.060 0.060 0.021 0.021 0.011 0.017 0.017 0.021 0.017 0.021 0.020 0.017 0.021 0.020 0.003 0.017 0.017 0.021 0.044 0.003 0.017 0.017 0.062 0.061 0.044 0.017 0.003 0.062 0.061 0.044 0.042 0.041 Market Scenario 178 181 182 184 187 208 0.075 0.095 0.056 0.071 0.069 0.075 0.071 0.095 0.073 0.069 0.075 0.073 0.078 0.071 0.069 0.075 0.092 0.078 0.071 0.069 0.098 0.096 0.092 0.071 0.073 0.098 0.096 0.092 0.086 0.084 Print WSJ High High High High High High 51.256 51.985 52.621 51.213 50.834 51.256 51.213 51.985 50.539 50.834 51.256 50.539 50.228 51.213 50.834 51.256 44.963 50.228 51.213 50.834 41.559 41.200 44.963 51.213 50.571 41.559 41.200 44.963 41.794 41.435 Online WSJ Cost Cost % of Print Price Profit
1 2 3 4 11 1 2 3 4 7 1 2 3 4 8 1 2 3 8 13 1 2 3 16 23 1 2 3 21 24 Market Scenario 19 91 97 100 101 106 172
0.025 0.026 0.133 0.055 0.055 0.025 0.055 0.026 0.025 0.055 0.025 0.025 0.026 0.055 0.055 0.025 0.035 0.026 0.055 0.055 0.044 0.044 0.035 0.055 0.058 0.044 0.044 0.035 0.100 0.099 Online WSJ Low Low Low Low Low Low Low
0.517 0.451 0.498 0.513 0.504 0.517 0.513 0.451 0.508 0.504 0.517 0.508 0.519 0.513 0.504 0.517 0.579 0.519 0.513 0.504 0.619 0.607 0.579 0.513 0.515 0.619 0.607 0.579 0.605 0.594
12.137 13.174 13.913 12.534 12.442 15.171 15.667 16.468 15.071 15.553 18.205 18.085 17.796 18.801 18.663 21.239 15.401 20.762 21.934 21.774 13.821 13.712 17.601 25.068 24.609 15.548 15.426 19.801 18.044 17.890
40.00%
39.120 38.811 38.708 38.679 38.392 36.086 35.546 35.518 35.468 35.281 33.051 32.454 32.432 32.412 32.171 30.017 29.561 29.466 29.279 29.060 27.738 27.487 27.361 26.145 25.961 26.010 25.773 25.161 23.750 23.544
50.00%
60.00%
70.00%
80.00%
90.00%
Online Print WSJ Print WSJ Unit Online WSJ Online Unit High Medium Medium High High High Medium Low Low High Low Low High Low Low Low Low Low Medium Low Low
Online Print WSJ Print WSJ Unit Online WSJ Online Unit Low Low Medium Low Low Low
Low Medium High Low High Low Low High Low Low High Medium Low High High Medium High Low
121
Table 14: Book Category – Market Simulation Results (Incomplete Product Line)
Market Share PDF Print Book Book and PDF Book 0.308 0.259 0.328 0.228 0.269 0.308 0.328 0.259 0.228 0.269 0.308 0.328 0.259 0.337 0.269 0.308 0.328 0.337 0.259 0.269 0.308 0.328 0.337 0.259 0.269 0.308 0.328 0.337 0.259 0.269 0.109 0.072 0.041 0.122 0.025 0.109 0.041 0.072 0.122 0.025 0.109 0.041 0.072 0.011 0.025 0.109 0.041 0.011 0.072 0.025 0.109 0.041 0.011 0.072 0.025 0.109 0.041 0.011 0.072 0.025
Rank
Market Scenario # 172/181/182 91/100/101 184 208 85 172/181/182 184 91/100/101 208 85 172/181/182 184 91/100/101 169/187 85 172/181/182 184 169/187 91/100/101 85 172/181/182 185 169/187 91/100/101 85 172/181/182 184 169/187 91/100/101 85
Print Book
No Choice
Revenue
Cost
Cost % of Print Price
Profit
1 2 3 4 7 1 2 3 4 6 1 2 3 4 6 1 2 3 4 6 1 2 3 5 6 1 2 3 5 6
0.365 0.479 0.407 0.400 0.514 0.365 0.407 0.479 0.400 0.514 0.365 0.407 0.479 0.426 0.514 0.365 0.407 0.426 0.479 0.514 0.365 0.407 0.426 0.479 0.514 0.365 0.407 0.426 0.479 0.514
0.217 0.190 0.224 0.250 0.192 0.217 0.224 0.190 0.250 0.192 0.217 0.224 0.190 0.226 0.192 0.217 0.224 0.226 0.190 0.192 0.217 0.224 0.226 0.190 0.192 0.217 0.224 0.226 0.190 0.192
21.748 21.764 20.311 21.153 20.818 21.748 20.311 21.764 21.153 20.818 21.748 20.311 21.764 19.707 20.818 21.748 20.311 19.707 21.764 20.818 21.748 20.311 19.707 21.764 20.818 21.748 20.311 19.707 21.764 20.818
5.503 6.395 5.197 6.055 6.252 6.879 6.497 7.994 7.568 7.815 8.255 7.796 9.593 7.606 9.378 9.631 9.095 8.873 11.191 10.941 11.007 10.394 10.141 12.790 12.504 12.38 11.69 11.41 14.389 14.07
40.00%
16.244 15.369 15.113 15.099 14.566 14.868 13.814 13.770 13.585 13.003 13.493 12.515 12.172 12.101 11.440 12.117 11.215 10.833 10.573 9.877 10.741 9.916 9.566 8.974 8.314 9.365 8.617 8.298 7.375 6.751
50.00%
60.00%
70.00%
80.00%
90.00%
Market Scenario #
Print Book PDF Book Print Book PDF Book Low Low Low Low Low Medium Medium Low High Low Medium Low
85 Medium 91/100/101 Medium 169/187 High 172/181/182 High 184 High 208 High
122
Table 15: Newspaper Category – Market Simulation Results (Incomplete Product Line)
Market Share Online Print WSJ No WSJ and Choice Online WSJ 0.557 0.498 0.576 0.423 0.515 0.581 0.576 0.557 0.521 0.515 0.581 0.576 0.557 0.521 0.515 0.581 0.576 0.521 0.557 0.515 0.581 0.655 0.576 0.704 0.643 0.704 0.655 0.581 0.685 0.643 0.214 0.237 0.256 0.263 0.283 0.277 0.256 0.214 0.306 0.283 0.277 0.256 0.214 0.306 0.283 0.277 0.256 0.306 0.214 0.283 0.277 0.162 0.256 0.078 0.134 0.078 0.162 0.277 0.067 0.134 0.085 0.111 0.104 0.147 0.134 0.114 0.104 0.085 0.145 0.134 0.114 0.104 0.085 0.145 0.134 0.114 0.104 0.145 0.085 0.134 0.114 0.138 0.104 0.150 0.120 0.150 0.138 0.114 0.125 0.120
Rank
Market Scenario # 19 28 91/100/101 55 127 172/181/182 91/100/101 19 208 127 172/181/182 91/100/101 19 208 127 172/181/182 91/100/101 208 19 127 172/181/182 184 91/100/101 178 85 178 184 172/181/182 97/106 85
Print WSJ
Revenue
Cost
Cost % of Print Price
Profit
1 2 3 4 6 1 2 3 4 5 1 2 3 4 5 1 2 3 4 6 1 2 3 4 5 1 2 3 4 5
0.143 0.154 0.064 0.167 0.068 0.028 0.064 0.143 0.028 0.068 0.028 0.064 0.143 0.028 0.068 0.028 0.064 0.028 0.143 0.068 0.028 0.046 0.064 0.068 0.102 0.068 0.046 0.028 0.123 0.102 Print WSJ
54.129 55.407 52.320 56.666 53.445 51.609 52.320 54.129 52.633 53.445 51.609 52.320 54.129 52.633 53.445 51.609 52.320 52.633 54.129 53.445 51.609 43.523 52.320 38.369 45.145 38.369 43.523 51.609 41.307 45.145
14.216 15.560 12.737 17.099 13.975 15.170 15.921 17.771 16.646 17.469 18.204 19.105 21.325 19.976 20.963 21.238 22.290 23.305 24.879 24.456 24.272 16.484 25.474 11.615 18.805 13.07 18.54 27.31 17.01 21.16 Print WSJ
40.00%
39.912 39.847 39.583 39.567 39.469 36.439 36.399 36.358 35.987 35.976 33.405 33.215 32.804 32.657 32.482 30.371 30.030 29.328 29.250 28.988 27.337 27.039 26.846 26.754 26.340 25.30 24.98 24.30 24.29 23.99 Print WSJ Online WSJ Low Low High Medium Low
50.00%
60.00%
70.00%
80.00%
90.00%
Market Scenario #
Online Print WSJ WSJ Online WSJ
Market Scenario #
Online WSJ
19 Low Low Low 28 Low Medium Low 55 Low High Low 85 Medium Low Medium 91/100/101 Medium Low Low 97/106 Medium Low High
127 Medium Medium 172/181/182 High Low 178 High Low 184 High Low 208 High Medium
123
Table 16: Book Category – Market Simulation Results (Incomplete Product Line, No Bundle Discount)
Rank Market Scenario # 163-189 82-108 1-27 28-54 109-135 163-189 82-108 1-27 190-216 109-135 163-189 82-108 1-27 190-216 109-135 163-189 82-108 190-216 1-27 109-135 163-189 82-108 190-216 1-27 109-135 163-189 82-108 190-216 1-27 109-135 Print Book Market Share PDF Print Book No Book and Choice PDF Book 0.337 0.272 0.218 0.163 0.200 0.337 0.272 0.218 0.252 0.200 0.337 0.272 0.218 0.252 0.200 0.337 0.272 0.252 0.218 0.200 0.337 0.272 0.252 0.218 0.200 0.337 0.272 0.252 0.218 0.200 0.010 0.013 0.015 0.010 0.008 0.010 0.013 0.015 0.007 0.008 0.010 0.013 0.015 0.007 0.008 0.010 0.013 0.007 0.015 0.008 0.010 0.013 0.007 0.015 0.008 0.010 0.013 0.007 0.015 0.008 0.226 0.192 0.167 0.183 0.216 0.226 0.192 0.167 0.263 0.216 0.226 0.192 0.167 0.263 0.216 0.226 0.192 0.263 0.167 0.216 0.226 0.192 0.263 0.167 0.216 0.226 0.192 0.263 0.167 0.216 Revenue Cost Cost % of Print Price Profit
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
0.427 0.523 0.600 0.645 0.575 0.427 0.523 0.600 0.478 0.575 0.427 0.523 0.600 0.478 0.575 0.427 0.523 0.478 0.600 0.575 0.427 0.523 0.478 0.600 0.575 0.427 0.523 0.478 0.600 0.575
19.686 20.593 21.281 20.776 19.913 19.686 20.593 21.281 18.708 19.913 19.686 20.593 21.281 18.708 19.913 19.686 20.593 18.708 21.281 19.913 19.686 20.593 18.708 21.281 19.913 19.686 20.593 18.708 21.281 19.913
5.066 6.215 7.138 7.591 6.770 6.332 7.769 8.923 7.036 8.463 7.599 9.323 10.707 8.443 10.156 8.865 10.876 9.850 12.492 11.848 10.132 12.430 11.257 14.276 13.541 11.40 13.98 12.66 16.061 15.23
40.00%
14.620 14.378 14.143 13.185 13.142 13.354 12.824 12.359 11.672 11.450 12.087 11.270 10.574 10.265 9.757 10.821 9.716 8.858 8.790 8.064 9.554 8.162 7.451 7.005 6.372 8.288 6.609 6.044 5.221 4.679
50.00%
60.00%
70.00%
80.00%
90.00%
Market Scenario #
Print Book PDF Book Print Book PDF Book Low Medium Low Medium Low Medium Sum Sum Sum Sum Sum Sum
1-27 Low 28-54 Low 82-108 Medium 109-135 Medium 163-189 High 190-216 High
124
Table 17: Newspaper Category – Market Simulation Results (Incomplete Product Line, No Bundle Discount)
Rank Market Scenario # 1-27 28-54 55-81 82-108 109-135 1-27 28-54 82-108 55-81 109-135 1-27 82-108 28-54 163-189 109-135 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 82-108 1-27 163-189 109-135 28-54 163-189 82-108 1-27 190-216 109-135 163-189 82-108 190-216 109-135 1-27 Print WSJ Market Share Online Print WSJ No WSJ and Choice Online WSJ 0.611 0.562 0.494 0.676 0.627 0.611 0.562 0.676 0.494 0.627 0.611 0.676 0.562 0.719 0.627 0.676 0.611 0.719 0.627 0.562 0.719 0.676 0.611 0.671 0.627 0.719 0.676 0.671 0.627 0.611 0.112 0.102 0.090 0.081 0.076 0.112 0.102 0.081 0.090 0.076 0.112 0.081 0.102 0.056 0.076 0.081 0.112 0.056 0.076 0.102 0.056 0.081 0.112 0.053 0.076 0.056 0.081 0.053 0.076 0.112 0.093 0.125 0.169 0.125 0.164 0.093 0.125 0.125 0.169 0.164 0.093 0.125 0.125 0.151 0.164 0.125 0.093 0.151 0.164 0.125 0.151 0.125 0.093 0.196 0.164 0.151 0.125 0.196 0.164 0.093 Revenue Cost Cost % of Print Price Profit
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
0.184 0.212 0.247 0.119 0.134 0.184 0.212 0.119 0.247 0.134 0.184 0.119 0.212 0.074 0.134 0.119 0.184 0.074 0.134 0.212 0.074 0.119 0.184 0.081 0.134 0.074 0.119 0.081 0.134 0.184 Print WSJ
48.721 48.400 47.887 42.070 41.150 48.721 48.400 42.070 47.887 41.150 48.721 42.070 48.400 37.078 41.150 42.070 48.721 37.078 41.150 48.400 37.078 42.070 48.721 35.696 41.150 37.078 42.070 35.696 41.150 48.721
11.754 12.478 13.413 7.934 8.330 14.693 15.598 9.918 16.766 10.413 17.631 11.902 18.717 7.750 12.496 13.885 20.570 9.042 14.578 21.837 10.334 15.869 23.508 10.644 16.661 11.63 17.85 11.97 18.74 26.45 Print WSJ
40.00%
36.967 35.922 34.475 34.136 32.820 34.029 32.802 32.152 31.121 30.737 31.090 30.169 29.683 29.328 28.655 28.185 28.152 28.037 26.572 26.563 26.745 26.201 25.213 25.052 24.490 25.45 24.22 23.72 22.41 22.27 Print WSJ Online WSJ Sum Sum Sum
50.00%
60.00%
70.00%
80.00%
90.00%
Market Scenario # 1-27 28-54 55-81 82-108
Online Print WSJ WSJ Online WSJ Sum Sum Sum Sum
Market Scenario #
Online WSJ
Low Low Low Medium Low High Medium Low
109-135 Medium Medium 163-189 High Low 190-216 High Medium
125
Table 18: Book Category - Forms’ Attribute Perceptions.
Variable Price Print Book (PP) Electronic Book (EP) Print Book & Electronic Book Bundle (PPEP) Electronic Book Unit (EU) Print Book & Electronic Book Unit Bundle (PPEU) Print Form Index (PI) PDF Form Index (EI_1) PDF Form Index (EI_2) Difference Index 1 (DI1_1): Bundle vs. Superior Difference Index 1 (DI1_2): Bundle vs. Superior Difference Index 2 (DI2): Print vs. Electronic Price Print Book (PP) Electronic Book (EP) Print Book & Electronic Book Bundle (PPEP) Number of observations Iterations completed Log likelihood function Restricted log likelihood Chi squared Degrees of freedom Prob[ChiSqd > value] Coefficient SE b/St.Er. P[|Z|>z] Random parameters in utility functions -0.058 0.017 -3.392 0.001 4.339 0.615 7.055 0.000 3.743 0.579 6.464 0.000 -5.861 1.775 -3.302 0.001 Nonrandom parameters in utility functions 1.861 0.291 6.392 0.000 1.768 0.671 2.635 0.008 0.029 0.010 2.840 0.005 0.029 0.009 3.357 0.001 0.013 0.003 4.036 0.000 0.107 0.033 3.232 0.001 0.018 0.013 1.319 0.187 0.023 0.006 3.612 0.000 Derived st. dev. of parameter distributions 0.091 0.010 9.107 0.000 2.289 0.254 8.998 0.000 2.291 0.253 9.068 0.000 3.461 0.852 4.063 0.000 4176 (87 respondents x 16 choices) 68 -917.41 -2429.63 3024.44 16 0.000
126
Table 19: Newspaper Category - Forms’ Attribute Perceptions.
Variable Price Print Newspaper (PP) Electronic Newspaper (EP) Print & Electronic Newspaper Bundle (PPEP) Electronic Newspaper Unit (EU) Print Newspaper & Electronic N. Unit Bundle (PPEU) Print Form Index (PI) PDF Form Index (EI_1) PDF Form Index (EI_2) Difference Index 1 (DI1_1): Bundle vs. Superior Difference Index 1 (DI1_2): Bundle vs. Superior Difference Index 2 (DI2): Print vs. Electronic Price Number of observations Iterations completed Log likelihood function Restricted log likelihood Chi squared Degrees of freedom Prob[ChiSqd > value]
Coefficient
SE
b/St.Er.
P[|Z|>z] 0.000 0.000 0.000 0.000 0.819 0.007 0.166 0.000 0.005 0.402 0.855 0.783
Random parameters in utility functions -0.042 0.007 -5.868 Nonrandom parameters in utility functions 2.747 0.661 4.159 2.586 0.292 8.867 3.148 0.847 3.718 0.042 0.186 0.229 1.994 0.744 2.680 0.008 0.006 1.386 0.011 0.003 3.645 0.007 0.003 2.822 -0.011 0.013 -0.837 0.003 0.014 0.183 -0.001 0.004 -0.276
Derived st. dev. of parameter distributions 0.029 0.003 8.666 0.000 4176 (87 respondents x 16 choices) 18 -1068.52 -2433.21 2729.37 13 0.000
127
FIGURES Figure 1: Study 2 Results (significant effects only) A. USAGE by DISCOUNT
Mean LN of Points Allocated to the Bundle
Mean LN of Points Allocated to the Bundle
Category: MARGARINE/COFFEE
0.0
Category: BOOK/NEWSPAPER
-2.4 -2.6 -2.8 -3.0 -3.2 -3.4 -3.6 -3.8 -4.0 0 1
-.5
DIFFERENT USAGES
-1.0
DIFFERENT USAGES
-1.5
-2.0
-2.5
SAME USAGES
SAME USAGES
-3.0 0 1
NO DISCOUNT
DISCOUNT
NO DISCOUNT
DISCOUNT
B. DISCOUNT by RELATIVE PRICE
Mean LN of Points Allocated to the Bundle
Category: BOOK
-2.6 -2.8 -3.0 -3.2 -3.4 -3.6 -3.8
DISCOUNT
NO DISCOUNT
-4.0 -4.2 0 1
PRICE PRINT = PRICE ELECT
PRICE PRINT > PRICE ELECT
128
Figure 2: Optimal Price and Total and Marginal Revenue from Electronic Book Chapters.
VC VH PC MR VL ½ NC NC QC TR
129
Figure 3: Optimal Price and Total and Marginal Revenue from Full Electronic Book.
VEB VH PB MR VL VPB k TR
130
REFERENCES March 2005 Online Paid Content U.S. Market Spending Report (2005), http://www.online-publishers.org/pdf/paid_content_report_030905.pdf. Aaker, David and Kevin Keller (1990), “Consumer Evaluations of Brand Extensions,” Journal of Marketing, 54 (January): 27-41. Adams, William J. and Janet L. Yellen (1976), “Commodity Bundling and the Burden of Monopoly,” Quarterly Journal of Economics, 90 (August): 475-98. Bakos, Yannis and Erik Brynjolfsson (2000), “Bundling and Competition on the Internet: Aggregation Strategies for Information Goods,” Marketing Science, 19 (1): 63-82. Bakos, Yannis and Erik Brynjolfsson (1999), “Bundling Information Goods: Pricing, Profits and Efficiency”, Management Science, 45 (12), 1613-1630. Betancourt, Roger R. (2005), The Economics of Retailing and Distribution, Edward Elgar Publishing, Chapter 6. Carroll, J. Douglas and Paul Green (1995), “Psychometric Methods in Marketing Research: Part I, Conjoint Analysis”, Journal of Marketing Research, 32 (November), 385-391. Chakravarti, Dipankar, Rajan Krish, Paul Pallab and Joydeep Srivastava (2002), “Partitioned Presentation of Multi-Component Bundle Prices: Evaluation, Choice and Underlying Processing Effects,” Journal of Consumer Psychology, 12 (3): 215 - 29. Chernev, Alexander (2005), “Feature Complementarity and Assortment in Choice”, Working paper, Kellogg School of Management, Northwestern University, Evanston, IL. Dhar, Ravi (1997), “Consumer Preference for a No-Choice Option,” Journal of Consumer Research, 24 (September), 215-231. Estelami, Hooman (1990), “Consumer Savings in Complementary Product Bundles,” Journal of Marketing Theory and Practice, Summer: 107-14. Fennell, Geraldine G. (1978), “Consumers’ Perceptions of the Product-Use Situations,” Journal of Marketing, 42 (April): 39-47. Gaeth, Gary, Irwin Levin, Goutam Chakraborty and Aron M. Levin (1990), “Consumer Evaluation of Multi-Product Bundles: An Information Integration Analysis,” Marketing Letters, 2: 47-57. Geng, Xianjun, Maxwell B. Stinchcombe and Andrew B. Whinston (2005), “Bundling Information Goods of Decreasing Value,” ? forthcoming in Management Science. Green, Paul and V. Srinivasan (1990), “Conjoint Analysis in Marketing: New 131
Developments with Implications for Research and Practice”, Journal of Marketing, 54 (4), 3-19. Guadagni, Peter M. and John D.C. Little (1983), “A Logit Model of Brand Choice Calibrated on Scanner Panel Data, Marketing Science, 2 (3), 203-239. Guiltinan, Joseph P. (1987), “The Price Bundling of Services: A Normative Framework,” Journal of Marketing, 51 (April): 74-85. Hanson, Ward and R. Kipp Martin (1990), “Optimal Price Bundling”, Management Science, 36 (2), 155-174. Harlam, Bari A., Aradna Krishna, Donald R. Lehmann and Carl Mela (1995), “Impact of Bundle Type, Price Framing and Familiarity on Purchase Intention for the Bundle”, Journal of Business Research, 33, 57-66. Henderson, James M. and Richard E. Quandt (1958), Microeconomic Theory: A Mathematical Approach, New York: McGraw-Hill Book Company, p. 29. Jain, Dipak C, Vilcassim, Naufel J, and Pradeep K. Chintagunta (1994), “A RandomCoefficients Logit Brand-Choice Model Applied to Panel Data,” Journal of Business & Economic Statistics, 12 (July), 317-328. Jedidi, Kamel, Sharan Jagpal and Puneet Manchanda (2003), “Measuring Heterogeneous Reservation Prices for Product Bundles”, Marketing Science, 22 (1), 107-130. Johnson, Michael D., Andreas Herrmann and Hans H. Bauer (1999), “The Effects of Price Bundling on Consumer Evaluations of Product Offerings,” International Journal of Research in Marketing, 16: 129-42. Kahneman, Daniel and Amos Tversky (1979), “Prospect Theory: An Analysis of Decision Under Risk,” Econometrica, 47: 163-91. Kenney, Roy W. and Benjamin Klein (1983), “The Economics of Block Booking,” Journal of Law and Economics, 26: 497-540. Lattin, James M. and Leigh McAlister (1985), “Using a Variety-Seeking Model to Identify Substitute and Complementary Relationships Among Competing Products,” Journal of Marketing Research, 22 (August), 330-339. MacKenzie, Scott B. and Richard J. Lutz (1989) “An Empirical Examination of the Structural Antecedents of Attitude Toward the Ad in an Advertising Pretesting Context,” Journal of Marketing, 53 (April): 48-65. Macklin, Theodore (1922), Efficient Marketing for Agriculture, The Mackmillian Co. New York: 25.
132
McAfee, R. Preston, John McMillan and Michael Whinston (1989), “Multiproduct Monopoly, Commodity Bundling, and Correlations of Values,” Quarterly Journal of Economics, 104 (2): 371-83. Mick, David Glen (1992), “Levels of Subjective Comprehension in Advertising Processing and their Relations to Ad Perceptions, Attitudes and Memory,” Journal of Consumer Research, 18 (March): 411-24. Ratneshwar, S. and Allan D. Shocker (1991), “Substitution in Use and the Role of Usage Context in Product Category Structures,” Journal of Marketing Research, 28 (August): 281–95. Sarvary, Miklos and Philip M. Parker (1997), “Marketing Information: A Competitive Analysis,” Marketing Science, 16 (1): 24-38. Schmalensee, Richard (1984), “Gaussian Demand and Commodity Bundling,” Journal of Business, 57 (1): S211-30. Simonin, Bernard L. and Julie A. Ruth (1995), “Bundling as a Strategy for New Product Introduction: Effects on Consumers’ Reservation Prices for the Bundle, the New Product, and Its Tie-in,” Journal of Business Research, 33: 219-230. Srivastava, Rajendra K., Mark I. Alpert and Allan D. Shocker (1984), “A CustomerOriented Approach for Determining Market Structures,” Journal of Marketing, 48 (Spring): 32-45. Srivastava, Rajendra K., Robert P. Leone and Allan D. Shocker (1981), “Market Structure Analysis: Hierarchical Clustering of Products Based on Substitution-in-Use,” Journal of Marketing, 45 (Summer): 38-48. Stefflre, Volney R. (1971), New Products and New Enterprises: A Report of an Experiment in Applied Social Science, Irvine, CA: University of California. Stigler, George J. (1963), “United States v. Loew’s, Inc: A Note on Block Booking,” Supreme Court Review, 153 (1963): 152-57. Stremersch, Stefan and Gerard J. Tellis (2002), “Strategic Bundling of Products and Prices: A New Synthesis for Marketing,” Journal of Marketing, 66 (January): 55-72. Telser, L. G. (1979), “A Theory of Monopoly of Complementary Goods,” Journal of Business, 52 (2): 211-30. Thaler, Richard (1985), “Mental Accounting and Consumer Choice,” Marketing Science, 4 (3): 199-214. Venkatesh, R. and Rabikar Chatterjee (2005), “Bundling, Unbundling and Pricing of Hybrid Products: The Case of Magazine Content,” Working Paper, University of 133
Pittsburg, PA. Venkatesh, R. and Vijay Mahajan (1993), “A Probabilistic Approach to Pricing a Bundle of Products or Services”, Journal of Marketing Research, 30 (November), 494-508. Venkatesh, R. and Wagner Kamakura (2003), “Optimal Bundling and Pricing under a Monopoly: Contrasting Complements and Substitutes from Independently Valued Products”, Journal of Business, 76 (2), 211-231. Wilson, Lynn O., Allen M. Weiss and George John (1990), “Unbundling of Industrial Systems”, Journal of Marketing Research, 27 (May), 123-138. Yadav, Manjit S. (1995), “Bundle Evaluation in Different Marketing Segments: The Effects of Discount Framing and Buyers’ Preference Heterogeneity,” Journal of the Academy of Marketing Science, 23 (3): 206-15. Yadav, Manjit S. (1994), “How Buyers Evaluate Product Bundles: A Model of Anchoring and Adjustment,” Journal of Consumer Research, 21 (September): 342-53. Yadav, Manjit S. and Kent B. Monroe (1993), “How Buyers Perceive Savings in a Bundle Price: An Examination of a Bundle’s Transaction Value,” Journal of Marketing Research, 30 (August): 350-58. Wansink, Brian and Michael L. Ray (1996), “Advertising Strategies to Increase Usage Frequency,” Journal of Marketing, 60 (January): 31-46. www.businesswire.com (2004), “U.S. Consumer Spending for Online Content Totals $853 Million in the First Half of 2004, According to Online Publishers Association Report”, Nov. 15.
134
doc_327658533.pdf
In marketing, a product is anything that can be offered to a market that might satisfy a want or need. In retailing, products are called merchandise. In manufacturing, products are bought as raw materials and sold as finished goods. Commodities are usually raw materials such as metals and agricultural products, but a commodity can also be anything widely available in the open market.
ABSTRACT
Title of Dissertation:
Marketing of Digital Products Nevena Taneva Koukova, Doctor of Philosophy, 2005
Dissertation directed by:
Professor Brian T. Ratchford and Professor P. K. Kannan Department of Marketing
My dissertation comprises of three essays that theoretically and empirically investigate the marketing of digital products, which are information products such as newspapers and books sold in both physical and electronic form. In the first essay, we study product form bundling, defined as marketing two or more forms of the same product as a package. We show experimentally that, regarding information products, the usage situations communicated to consumers moderate the effect of the availability of bundle discount on the purchase likelihood for the product form bundle. We also compare the effect of different pricing strategies for information products. When no bundle discount is offered, the likelihood of buying both forms of an information product, holding the sum of their prices constant, can be increased by pricing the electronic form lower than the print form rather then pricing both at the same level. In the second essay we compare two product strategies that can be used in marketing digital products. Under standard mixed bundling companies offer full content in print and electronic form and the bundle of the two, while under content unbundled
mixed bundling companies offer full content in print form, unbundled content in electronic form, and the bundle of the two. Which strategy is more attractive for a company to pursue? We model the profits under these two strategies and outline conditions in which one or the other leads to higher profit. We apply our analytical framework to data from a field experiment implemented on the website of a book publisher. The third essay investigates the attractiveness of complete product lines of items such as books and newspapers. We employ a choice experiment in which a sample of consumers is presented with hypothetical product scenarios asked to make a choice. The data is used to develop a profit-maximizing configuration of products and prices. Similar approaches to the product line pricing problem have been employed for conventional products, but not when bundling of different forms of a product is an option, and not when the different products may be complements rather than substitutes.
MARKETING OF DIGITAL PRODUCTS
by Nevena Taneva Koukova
Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park in partial fulfillment Of the requirements for the degree of Doctor of Philosophy 2005
Advisory Committee: Professor Brian T. Ratchford, Chair Professor P. K. Kannan, Chair Professor Roger R. Betancourt Professor Rebecca W. Hamilton Professor Joydeep Srivastava
©Copyright by Nevena Taneva Koukova 2005
ACKNOWLEDGEMENT I would like to thank my advisors, Dr. Brian Ratchford and Dr. P. Kannan, for their excellent guidance and help on my dissertation and my other research projects. I would also like to thank the other dissertation committee members, Dr. Rebecca Hamilton, Dr. Joydeep Srivastava and Dr. Roger Betancourt, for their helpful comments and suggestions at various stages of my dissertation.
ii
TABLE OF CONTENTS List of Tables ......................................................................................................................v List of Figures ...................................................................................................................vii Chapter I: Dissertation Overview ....................................................................................... 1 Chapter II: Essay One. Product Form Bundling - Implications for Marketing Digital Products............................................................................................................................... 4 Abstract ................................................................................................................... 4 Introduction............................................................................................................. 5 Research Background ............................................................................................. 9 Study Objectives ................................................................................................... 13 Theoretical Framework and Hypotheses .............................................................. 14 Study 1 .................................................................................................................. 25 Study 2 .................................................................................................................. 29 Discussion ............................................................................................................. 38 Managerial Implications ....................................................................................... 40 Conclusion ............................................................................................................ 42 Chapter III: Essay Two. Bundling and Pricing Strategies for Digital Products ............... 44 Abstract ................................................................................................................. 44 Introduction........................................................................................................... 45 Research Background ........................................................................................... 47 Bundling Model .................................................................................................... 49 Data and Estimation Procedure............................................................................. 59 Results................................................................................................................... 64
iii
Discussion ............................................................................................................. 69 Managerial Implications ....................................................................................... 72 Conclusion ............................................................................................................ 73 Chapter IV: Essay Three. Bundling and Unbundling of Electronic Content.................... 75 Abstract ................................................................................................................. 75 Introduction........................................................................................................... 76 Research Background ........................................................................................... 78 Theoretical Background........................................................................................ 82 Methodology ......................................................................................................... 86 Study Design and Data Collection........................................................................ 87 Results................................................................................................................... 90 Discussion ........................................................................................................... 102 Managerial Implications ..................................................................................... 106 Conclusion .......................................................................................................... 107 References........................................................................................................................131
iv
LIST OF TABLES Table 1: Purchase Probabilities for Bundle: Product, Discount and Usage Conditions . 109 Table 2: Perceived Appropriateness of Generated Usage Situations.............................. 110 Table 3: Study 2 - Mean Percentage of Points Allocated to Bundle .............................. 111 Table 4: Study 2 ANOVA Results.................................................................................. 112 Table 5: Profit Comparison under CUMB and TMB Strategies..................................... 113 Table 6: Actual Purchase Rates under the Two Mixed Bundling Strategies.................. 114 Table 7: Estimation Results - Content Unbundled Mixed Bundling .............................. 115 Table 8: Estimation Results - Traditional Mixed Bundling............................................ 116 Table 9: Simulation Results – Profit under the Two Mixed Bundling Strategies. ......... 117 Table 10: Conjoint Design and Price Levels .................................................................. 118 Table 11: Random Parameters Logit Model Results (conjoint choice experiment)....... 119 Table 12: Book Category – Market Simulation Results ................................................. 120 Table 13: Newspaper Category – Market Simulation Results........................................ 121 Table 14: Book Category – Market Simulation Results (Incomplete Product Line)...... 122 Table 15: Newspaper Category – Market Simulation Results (Incomplete Product Line) ......................................................................................................................................... 123 Table 16: Book Category – Market Simulation Results (Incomplete Product Line, No Bundle Discount) ............................................................................................................ 124
v
Table 17: Newspaper Category – Market Simulation Results (Incomplete Product Line, No Bundle Discount) ...................................................................................................... 125 Table 18: Book Category - Forms’ Attribute Perceptions. ............................................. 126 Table 19: Newspaper Category - Forms’ Attribute Perceptions..................................... 127
vi
LIST OF FIGURES Figure 1: Study 2 Results................................................................................................ 128 Figure 2: Optimal Price and Total and Marginal Revenue from Electronic Chapters.... 129 Figure 3: Optimal Price and Total and Marginal Revenue from Full Electronic Book.. 130
vii
CHAPTER I: DISSERTATION OVERVIEW
Marketing digital products is appealing because of several reasons: 1) the marginal cost of reproducing and distributing them is almost zero (Bakos and Brynjolfsson, 1999), 2) they can be easily organized, searched and stored, and 3) the whole buying experience trough Internet can be more interactive and tailored to the specific needs and preferences of the customer (Peterson et al., 1997). Although many publishers make available content online, others (e.g. marketers of full length books, reference materials, music and video) are still cautious in distributing content in digital form over the Internet (Kannan and Jain, 2003). There are good reasons for this cautious approach, such as piracy of content and bandwidth constraints. However, less valid reasons also prevent managers from offering digital content, such as uncertainty about what product and pricing strategies to utilize. How should companies market digital products? What product, pricing and communication strategies should they employ? My dissertation attempts to tackle these interesting and practically relevant questions. It comprises of three essays organized as follows. In Essay 1 we focus on consumer reactions to bundling of information products, and contrast these with consumer reactions to bundling of conventional products. In particular, we concentrate on communication and pricing strategies that might enhance the viability of selling the information goods as a bundle. We show experimentally that communicating different possible usages for the product forms may increase the likelihood of buying the items as a bundle. We also investigate consumer reactions to
1
alternative pricing plans for product forms. Our respondents readily accept bundle discounts on conventional items that can be inventoried, but are generally unresponsive to bundle discounts on different forms of information goods that have the same content. However, emphasizing different uses for the forms does increase their responsiveness to bundle discounts. We conclude that communicating the appropriateness of different forms for different usage situations can be a key to a successful bundling strategy for information goods; further, it is more beneficial for companies to price the electronic form lower than the print form instead of offering bundle discounts. In Essay 2 we model the profits under two strategies a content provider may employ: traditional mixed bundling, in which the product line consists of print book, PDF book and the bundle of the two, and content unbundled mixed bundling, which is selling print book, PDF chapters and the combination of the two. While appealing because of low marginal costs and the likely catering to emerging consumer needs, offering electronic products may lead to additional revenue generation but can also result in product cannibalization. Our results suggest that offering unbundled PDF chapters is more profitable than offering only the full PDF books under certain conditions. We empirically support our predictions with actual data gathered in an online experiment executed on the website of a publisher of academic books. We present interesting insights on how customers from different subject domains react to electronic products and how they are likely to behave when different product and pricing schemes are employed. In Essay 3 we further extend the previous essays and investigate the attractiveness of complete product lines of information good. For example, the publisher of Wall Street Journal can offer the following product line of subscription options: print WSJ, WSJ
2
online, separate sections of WSJ online (e.g. Money & Investment, Technology), print WSJ and online WSJ bundle, and print WSJ and online section bundle. We employ a choice experiment in which a sample of consumers is presented with hypothetical product offerings at various prices and asked to make a choice. The data is used to develop a profit-maximizing configuration of products and prices. Similar approaches to the product line pricing problem have been employed for conventional products, but not when bundling of different forms of a product is an option, and not when the different products may be complements rather than substitutes.
3
CHAPTER II: ESSAY ONE. PRODUCT FORM BUNDLING - IMPLICATIONS FOR MARKETING DIGITAL PRODUCTS Abstract In this paper we study product form bundling, defined as marketing two or more forms of the same product as a package. Although we are primarily interested in information products (e.g. Wall Street Journal: print and online forms), we contrast consumer reactions to bundling these products with reactions to bundling conventional products (e.g. margarine: stick and tub forms). We show experimentally that the usage situations communicated to the consumers moderate the effect of the availability of bundle discount on the purchase likelihood for the product form bundle. Our subjects readily accept bundle discounts on conventional items that can be inventoried, but are generally unresponsive to bundle discounts on bundles of information products that have the same content. However, emphasizing different uses for the information product forms does increase the responsiveness to bundle discounts. Also, we examine the impact of different relative prices for the forms within the information bundle on the bundle purchase likelihood, and compare this to offering discounts. Our study provides useful insights into the conditions under which the marketer is better off pricing the forms differentially than pricing them equally and offering bundle discounts.
4
Introduction It is a common practice in consumer packaged goods categories to market products in different forms, for example, margarine (stick and tub forms) and soap (bar and liquid forms). In marketing different forms of the same product the focus has been on offering forms appropriate for various usage situations. For example, margarine is sold in both stick and tub form and customers can choose the form that better suits their needs and usage situations. Recently, marketers of information products have been applying similar strategies as well by making the information products available in electronic form in addition to the print/traditional form. Specifically, newspapers (e.g., Wall Street Journal) market print and electronic forms of their content, publishers (e.g., US Government Printing Office) offer print books and reports as well as the corresponding electronic versions, and recording companies (e.g., EMI Recorded Music) have started offering online music in addition to music in traditional forms (CDs, tapes, etc.) all catering to the diverse needs of customers. While bundling different forms of the same product is occasionally seen in the consumer packaged goods market (e.g. bundling Eucerin body cream and lotion; Dial liquid soap and Dial soap bars), one would generally not expect similar bundling strategies in the information goods market as the content being sold in the different forms is generally the same. However, newspapers such the Wall Street Journal not only offer the print and electronic forms separately as single products, but also as a bundle of different forms. Likewise, publishers of scholarly content such as National Academies Press have started to offer bundles of print and PDF forms in addition to offering them separately. Many scholarly journals also market subscriptions of both print and electronic
5
forms. New formats introduced by dualdisc.com provide the capability to sell music content in CD format as well as DVD-Audio or surround sound as a bundle on the same disk (Business Week, 2004). While different forms of products such as margarine and soap may have roughly the same production costs, information products are unique in that the marginal cost of reproducing and distributing the electronic form is often much lower than the one of the print form. Consequently, it is not surprising that marketers are interested in selling bundles of electronic and print forms, as it implies significant additional revenue. As broadband service penetration increases, similar opportunities in selling bundles of streaming videos and DVDs, and music bundles (mp3 form, CD, mobile tunes) become possible. In order to realize this additional revenue potential, however, companies need a clear understanding of how to market, communicate and price product form bundles. It is in this context that we present our research. In this paper, we first focus on consumers’ reactions to alternative forms of information products, and examine conditions under which they are more likely to purchase the product form bundle. In particular, we concentrate on communication and pricing strategies that might enhance the viability of selling the information goods as a product form bundle. We replicate our study in the consumer packaged goods category to establish a comparative baseline. It is well known that there are different types of utility a good may provide – elementary (or basic), place, form, time, and possession (Macklin 1922). In our study we distinguish between content utility (similar to elementary or basic) – utility provided by the product itself – and form utility – utility provided by the specific form in terms of convenience, ease of use and the like – of the product forms that
6
play a critical role in how the bundle is viewed. Content utility tends to be duplicated across different forms of an information product, making it less attractive to buy them as a bundle. From this viewpoint, the content utility of either of the information forms or the bundle is the same in magnitude. However, in the conventional packaged goods categories, the content utility of the bundle is generally more than that of either of the individual forms. While content utility remains the same in the case of an information product bundle making it less attractive to purchase, extant research would suggest that suitability of different forms of an information product for use in different situations will decrease the perceived similarity between them (Ratneshwar and Shocker 1991) and that bundles composed of complements have a higher purchase intent versus bundles of similar products (Estelami 1999), thus making the information bundle appealing. However, we find experimentally that communicating different possible usage situations for the forms, in and of itself, is not sufficient to increase the probability of selling the bundle of information product forms. Yet, for conventional packaged goods categories, such communication strategies are effective in increasing the bundle purchase probabilities. In investigating consumer reactions to alternative pricing plans for different forms of information products and different forms of conventional products, we find that our experimental subjects readily accept bundle discounts on conventional items that can be inventoried, but are generally unresponsive to bundle discounts on different forms of information products that have the same content. It is only when different uses for the forms are emphasized along with bundle discount that respondents’ bundle purchase probabilities of information product forms increase significantly. This suggests that
7
emphasizing the appropriateness of the forms for different usage situations is as important as bundle discounts in successfully marketing bundles of information products. Finally, we examine the impact of different relative prices for the forms within the information bundle on the purchase of information bundle, and compare this to offering discounts. Our study provides useful insights into the conditions under which the marketer is better off pricing the different forms differentially than pricing them the same and offering bundle discounts. The contribution of the paper is three-fold. First, our study offers a useful approach in terms of examining product forms and the attractiveness of information product bundles under bundle discounts, differential pricing of the forms, and usage situation communication from the perspective of content and form utilities. We are not focusing on the effect of the above factors per se but on their interaction effect in influencing consumer choice. Second, our study complements the findings in extant literature on bundling by providing insights into conditions under which those findings are directly applicable to product forms and instances where those results may not hold. Specifically, our study extends the findings to situations where items of bundles (forms) can be perceived as substitutes as well as complements across different consumers. Third, our paper provides substantive insights in an area of emerging importance – bundling of information products – by providing guidelines for communication and pricing strategies for increasing the attractiveness of information bundles and expanding revenue opportunities for marketers. This is managerially very relevant given that the consumer spending for online content in the US grew to $853 million in 2004, an increase of 14 percent over the same period last year (Online Publishers Association Report, 2004).
8
The paper is structured as follows. First, we define and discuss product form bundling in the context of extant literature and position our study. Second, we present a theoretical framework and derive hypotheses. Third, we discuss two experiments that were designed to empirically test our hypotheses. Finally, we highlight the implications of our study and suggest areas for future research.
Research Background Product Form Bundling Extant literature defines bundling as selling goods in packages (Adams and Yellen 1976), marketing products as a package at a special price (Guiltinan 1987), and selling products at a single price (Yadav and Monroe 1993). Distinction is also made between price bundling, defined as the sale of two or more separate products in a package at a discount without any integration of the products, and product bundling, which is the integration and sale of two or more separate products or services at any price (Stremersch and Tellis 2002). We define product form bundling as marketing two or more forms of the same product as a package. In the context of product categories, we make a distinction between information product form bundles in which the same information is presented in different forms and the forms are bundled together (e.g. electronic book and print book), and conventional product form bundles consisting of different physical forms of the same product (e.g. stick of margarine and margarine tub). Information bundles have some unique characteristics: the marginal cost of producing the electronic form after producing the print is negligible, consumers have reasonable knowledge of this cost structure, and, in
9
some cases, the individual products are perishable (e.g. after reading the news online consumers do not benefit from the print version of the newspaper to read the same news). With respect to conventional bundles, the cost structure is different - the marginal costs associated with the forms are positive - and the products usually are not immediately perishable and can be stockpiled. From a conceptual point of view, product form bundling seems similar to other types of bundling discussed in extant literature. On one hand, it could be viewed as a special case of product bundling – that is, the integration and sale of two or more separate products, the separate products being different forms of the same product; on the other hand, it is a type of price bundling - the sale of two or more separate products in a package at a discount. However, there are two important distinctions in the case of product form bundling. First, the product forms can range from being perceived as substitutes to being perceived as complements to each other and degree of heterogeneity across consumers on this dimension is much higher than what is encountered in other bundling situations. As an illustration, consider the bundling of print and electronic forms of newspaper subscriptions. The Wall Street Journal has been recently promoting the two forms of the newspaper as a bundle. The ads suggest that people might use the two forms for different purposes and/or situations (e.g. keeping up with the daily news versus doing research). If customers perceive WSJ Online as a substitute to WSJ, they will subscribe to one of them; if they view the two form of the journal as complementing each other, they may be willing to subscribe to both. Second, in the case of information product bundles, the marginal cost of producing one form after producing the other form is close to zero, which is quite different from the scenarios encountered in the extant bundling literature.
10
Offering different forms of a product is also similar to having more than one package of the same item available. Packaging of items explicitly sold at retail into a small and a large package, for example, can be viewed as commodity bundling and thus as a mechanism for price discrimination, or it can be viewed as a mechanism for providing different levels of distribution services through the shifting of distribution costs across market boundaries (Betancourt 2004). The small package provides a high level of the distribution service assurance of product delivery in the desired form bundled with the items explicitly sold that are contained in the small package (Betancourt 2004). Offering a print and an electronic form of a product can be considered as two different levels of the distribution service assurance of product delivery in the desired form, and each level is bundled with the content of the product. The difference between offering different product forms and offering different product packages is in the cost structure associated with the two offers. While the marginal costs of producing the different packages are positive and similar to each other, this is not necessarily true with respect to the different product forms - the marginal cost of the electronic form, for example, is close to zero. Consequently, the implications of offering different product forms can be different from the implications of offering different types of packaging both from buyers’ and sellers’ point of view. Because of the cost structure the buyers are likely to expect a relatively lower electronic form price as compared to the print form price, and it may be profitable for the sellers to charge relatively lower price for the electronic form. In sum, offering different forms of a conventional product is the same as offering different packaging types, while offering different forms of an information product is not necessarily the same as offering different packaging types because of the associated cost
11
structure. Thus, we use the term product forms as it is more general. Also, our focus is on bundling product forms (or types of packaging in the case of conventional products), and not on bundling an item with the specific service level it provides. While there is extensive literature focusing on conditions favorable for bundling – negative correlations in consumer reservation prices (Stigler 1963; Schmalensee 1984), complementarity in consumption (Telser 1979), uncertainty in the valuations of the quality of the goods (Kenney and Klein 1983) – contingent valuation of interrelated products in the bundle is very relevant for product form bundling. Contingent valuations have been examined at both conceptual (e.g. Guiltinan 1987) and analytical (e.g. Venkatesh and Kamakura 2003 (VK) levels. Under assumptions of homogeneous degree of contingency or complementarity, VK find that moderate to strong substitutes should be offered as separate products; the same is applicable for complements if the marginal costs are moderate relative to the market’s maximum willing to pay. They also find mixed bundling to be optimal for weak substitutes/complements if the marginal costs are not too high (Venkatesh and Kamakura 2003). It is not clear to what extent these results will hold given the heterogeneity in customer perceptions of the product forms as being substitutes or complements (contingency) or under the case of very low marginal costs. Thus, our study on bundling product forms might complement the above findings. In addition, the heterogeneity in consumer contingency/ complementarity perceptions may also suggest that consumers may be amenable to suggestions from marketers in manipulating such perceptions. Extant research based on mental accounting and related framing effects have provided interesting insights into consumer reaction to bundling (Thaler 1985; Johnson
12
1999). The findings relevant for relative pricing of the product forms and its impact on bundle choice suggest that (1) in terms of evaluation process, there is evidence that buyers use anchoring and adjustment (Yadav 1994), (2) the evaluation of a bundle depends on the price leader as well (Yadav 1995), and (3) savings offered on the bundle have a greater relative impact than savings offered on the individual items (Yadav and Monroe 1993). In the context of information products, the price of the form that consumers are more familiar with – print – could form an anchor against which other forms are evaluated. Also, given the marginal cost of one form (say, an e-book) is close to zero, the relative pricing of the forms may have to account for consumers’ expectations regarding the relative prices of product forms.
Study Objectives Our study objectives are as follows. First, we examine the concept of product form bundling from the viewpoint of content utility and form utility with the objective of answering the questions faced by a manager marketing information product form bundles – how can he/she increase consumers’ choice of bundles through appropriate communication to impact consumers’ perception of complementarity? And how should he/she price the individual product forms and the bundle? We contrast information product form bundles with conventional product form bundles to gain better insights into consumers’ reactions to bundling. Our second objective is to explain our findings in the light of extant literature in bundling and highlight the similarities and variances that help us better understand product form bundling. Given that most of the behavioral studies so far focus on bundles composed of complements (e.g. computer and printer), on investigating how consumers evaluate such bundles, and on the optimal strategies for
13
price and discount information presentation, the fact that we study bundle items for which marketers can influence the degree of contingency/complementarity should provide results that complement extant findings. Thus, our overall focus is not only on understanding how the bundling of interrelated products influences consumers’ purchase decisions but also on how these decisions can be affected through marketing communication and differential pricing.
Theoretical Framework and Hypotheses Perceived Complementarity From the firm’s point of view, complementarity is traditionally defined by referring to the sign of the cross-product elasticity of demand - if the sign is positive, products are substitutes; in the opposite case they are complements (Sarvary and Parker 1997). If different forms are substitutes, the purchase of one item lowers the value of alternative forms (contingent valuation), and therefore makes bundling the items less attractive. If different forms are complements, bundling becomes more attractive because the consumer is in effect purchasing a system of products that enable and enhance the functionality of each other (Estelami 1999). Consumers may perceive a high degree of complementarity based on economies of time and effort in purchasing the bundled items together (search economies), improved satisfaction because of the bundle, and/or improved total image of the bundle (Guiltinan 1987; Simonin and Ruth 1995). With regard to the attractiveness of bundling to firms, complementarity between products can cause bundling to be profitable (Telser 1979) and bundles composed of complements have higher purchase intent versus bundles of similar or unrelated products (Harlam et al. 1995). There is evidence that complementarity positively affects bundle reservation
14
prices as well (Gaeth et. al 1990). Applied to product form bundling, we would expect a positive relationship between the degree of perceived complementarity between the bundle components and the purchase likelihood for the bundle. On one hand, if consumers believe that the individual product forms can be used interchangeably, they would buy only one of the forms and not the product form bundle. On the other hand, if consumers think that there is additional utility in having both product forms versus just one of them, they would choose the product form bundle, ceteris paribus. Hypothesis 1: There is a positive relationship between the degree of perceived complementarity between the product forms in a bundle and the purchase likelihood for the product form bundle. We use the above hypothesis as a baseline and build on it to explain how exactly communicating different usage situations influences consumers’ complementarity perceptions and thus consumer choice. We elaborate more on this in the discussion section of Study 1.
Usage Situation Usage situations correspond to activities and conditions for which products are created and marketed (Fennell 1978). In order to analyze the effect of usage situations on the acceptance of product form bundles, we draw upon extant literature (Macklin 1922) to make a distinction between VC, the utility of the content (material in a book, the margarine or soap itself), and VF, the utility that the specific form adds to the content
15
(print or electronic form of a book, stick or tub form of margarine)1. A consumer’s utility of a product form of an information product is not independent of whether the consumer also buys another form. This is because the consumer obtains the same content when buying the two product forms and she does not get extra utility from having the same content twice. For example, if a consumer buys a reference book bundle, she gets the same content twice, but receives content utility only once. However, the consumer can read the print edition while traveling by metro/train, and search the electronic edition if looking for some specific information. Consequently, she can derive two separate form utilities from the bundle – that is, the incremental utility from acquiring the second form together with the first is positive. How great this incremental utility is depends on whether the alternative forms are viewed as complements. In our framework, the consumer’s purchase decision for an information product that comes in two formats can be represented as follows: choose the maximum of Value of Product Form i: Value of Product Form j: Value of Bundle: VC + VFi - Pi VC + VFj - Pj VC + VFi + VFj|Fi – (1 – d)*(Pi + Pj),
where d is the bundle discount, and VFj|Fi is incremental form utility of Form j given that Form i is purchased.2 Note that VC is counted only once in valuing the bundle because the same content is shared by both forms. Let Form i be the form that provides the highest value to the consumer (e.g., VFi - Pi > VFj - Pj). Then the bundle will be chosen if VC + VFi + VFj|Fi – (1 – d)*(Pi + Pj) > VC + VFi - Pi, or if VFj|Fi + d(Pi + Pj) > Pj. If there are two
We assume that VC and VF are independent and additive for expositional purposes, without any loss in generality. 2 Specifically VFj|Fi = VFi + Fj - VFi, where VFi + Fj is form utility from buying both items.
1
16
items in the bundle, and at least one of them has a positive valuation, we can state a general condition under which a consumer would buy both items if there were no discount: (1a) ProbINFO = Prob (VFj|Fi > Pj),
where j is the item that provides less value, i is the item that provides more value, and Prob refers to probability that a consumer will buy both items. The incremental form utility of the second item must exceed its price if it is to be purchased along with the first. If there is a bundle that is offered at a discount, the condition for buying the bundle becomes: (2a) ProbINFO = Prob (VFj|Fi + d (Pi + Pj) > Pj) = Prob (d > (Pj - VFj|Fi)/(Pi + Pj)).
The consumer will buy the bundle only if the discount is large enough to overcome the fact that she only receives form utility for the less preferred item. If VFj|Fi > Pj she would buy this item without a discount, as shown in Equation 1a. While information bundles suffer a redundancy of content, this is ordinarily not the case for conventional bundles. In the case of conventional bundles, consumers normally derive content utility from both product forms because they can first use Form i and then Form j.3 For example, if a consumer buys a bundle consisting of a pack of four sticks of margarine and a margarine tub, she can first use the sticks, store the tub, and then use the tub. Consequently, she derives two content values. In terms of form value, the consumer can use the margarine stick for baking purposes, while the tub for spreading margarine on bread or bakery, thus deriving form utility from both product forms
3
We assume here that conventional products do not deteriorate in quality as they are stocked. For example, rapidly perishable items can be an exception to this. As one example, large containers of soft drinks can be thought of as a bundle (they are a multiple of smaller sizes). A consumer might not value these highly because most of the contents may become flat before the consumer has a chance to consume a high proportion of the contents of the container.
17
(positive marginal utility from acquiring the second product form together with the first). For conventional bundles, the value of the bundle becomes 2VC + VFi + VFj|Fi – (1 – d)*(Pi + Pj), while the values of individual items remain as above4. Using the notation defined above, the probability of buying both forms if there were no discount becomes: (1b) ProbCONV = Prob (VC + VFj|Fi > Pj).
By comparing to 1b to 1a, the probability of buying the conventional product form bundle even when there is no discount will tend to be higher than for the information product form bundle. Similarly, as shown in Equation 2b below, the discount needed to induce consumers of the conventional products to buy a bundle will tend to be smaller because of the opportunity to obtain more content. (2b) ProbCONV = Prob (VC + VFj|Fi + d(Pj + Pi) > Pj) = Prob (d > (Pj - VC - VFj|Fi)/(Pi + Pj)) Hypothesis 1a: Because both forms of a conventional product form bundle provide content utility, consumers are more likely to select a conventional bundle than an information bundle at a given discount level. By directly comparing consumers’ reactions to the two types of product form bundles at specific discount levels we intend to show that the conventional and information bundles are valued differently by consumers because of the different content utility they provide (single or double).
Usage Situation and Bundle Discount When consumers are suggested usage situations in which the two product forms are differentially appropriate (one form is more appropriate for some situations, and the
For sake of exposition, we assume that the different forms of the conventional product have equivalent content and each provide the same content value Vc. Also, we assume that the content utility of the bundle is the sum of the content utilities of the individual forms, with marginal utility remaining the same whatever be the stage of consumption. These simplifying assumptions do not affect our general results.
4
18
other form is more appropriate for other situations), consumers are more likely to view them as complements. Consequently, by manipulating different possible usages for the two forms in the bundle, the form utility of one or both product forms can be changed. Specifically, if communications succeed in convincing consumers that different forms are appropriate for different uses of the product, and therefore more complementary, the value of VFj|Fi can be increased to, say, V*Fj|Fi. As can be seen from equations 1 and 2 this increases the value of the bundle in all cases. Numerous brand management teams apply usage-oriented advertising campaigns to expand the use of their products, for example “Eat Campbell’s Soup with formal family dinners or for breakfast” (Wansink and Ray, 1996). In our study we go beyond this conceptually and consider not the number (we keep it constant) but the nature of the usage situations presented to the consumers. We elaborate on this in the stimuli development section. << Insert Table 1 about here >> The probabilities of buying the information and conventional product form bundles subject to usage situations and level of discount are presented in Table 1. With respect to information bundles, we expect the usage situations considered when making a purchase decision to moderate the effect of bundle discount on purchase likelihood. Specifically, when consumers regard the two product forms as equally appropriate for the same usage situations, the forms will be substitutes and VFj|Fi ? 0. The bundle will be relatively unattractive and there will be no significant difference in their purchase
19
likelihood for the bundle whether a bundle discount is offered or not.5 The discount itself does not motivate consumers to buy the bundle because they are paying for the same content twice and the marginal utility they receive from the second form together with the first is very low or even zero. On the other hand, when consumers are considering different possible usages for the product forms, the likelihood of buying the bundle should be significantly higher in the discount as compared to the no discount case. Specifically, when consumers regard the two product forms as appropriate for different usage situations, the forms will be perceived as more complementary than in the previous case and VFj|Fi ? VFj. The bundle discount can motivate consumers to buy the bundle in this case – not only the marginal utility they get from the second form together with the first is positive but also the money outlay is lower. In the absence of a discount consumers are likely to find the bundle unattractive because it does not increase content utility (see Equation 1a). The bundle discount can compensate for this (see Equation 2a). Hypothesis 2 (information product form bundles): a) When consumers are suggested different usage situations for the product forms, the purchase likelihood for the bundle when bundle discount is offered is significantly higher than when bundle discount is not offered.6 b) There is no effect of availability of bundle discount when consumers are suggested same usage situations for the two product forms. With respect to conventional product form bundles the predictions on the effect of
In our experiments we have chosen to test a 25% level of bundle discount because this level is commonly used in practice (e.g., Buy a pair of shoes and get the second pair at 50% off). As shown in Equations 2a and 2b, the choice of a bundle is sensitive to the size of the discount, and this conclusion may not hold for very deep discounts. 6 As can be seen from Equation 2a, the appropriateness of this hypothesis is conditional on the discount being large enough to compensate for the bundle not offering an increase in content utility.
5
20
usage situation and bundle discount on purchase likelihood for the bundle are different – we expect a main effect of both usage situation and bundle discount but no interaction. Because consumers receive more content utility when acquiring the bundle, their willingness to pay for the bundle is higher than that for each individual item. If a bundle discount were offered, consumers would perceive the bundling of product forms as a quantity discount, and thus be more likely to buy the product form bundle. If they are suggested to use the two forms in different usage situations (versus same usage situations) VFj|Fi may still increase because the items could be viewed as more complementary. Thus, consumers may perceive positive marginal utility from owning Form j together with Form i, and consequently, be more likely to buy the product form bundle. Hypothesis 3 (conventional product form bundles): a) Consumers are more likely to buy the product form bundle when they are suggested different usage situations for the two product forms as compared to when they are suggested the same usage situations for the two product forms. b) Consumers are more likely to buy the product form bundle when they are offered bundle discount as compared to when they are not offered bundle discount.
Relative Price of Print vs. Electronic Forms Another factor that might influence the purchase likelihood for the product form bundle is the relative price of the two product forms. The more interesting case is the one of the information product form bundle. In this case, because consumers receive just one content utility from the two different forms, the marginal utility having Form j together
21
with Form i is relatively low and sometimes close to zero. Second, electronic information products are unique in that the marginal cost of reproducing and distributing them is often much lower than the cost of the print form, even close to zero (e.g. Bakos and Brynjolfsson 2000). In our empirical work we examine the purchase likelihoods for the bundle when the information product forms are priced differentially and compare the following cases: Case 1 - price of the two forms is equal, and Case 2 - price of the print form is higher than price of the electronic form. Note that the bundle prices are kept constant. To establish a benchmark, assume that the prices of the two forms are equal: Pi = Pj = PA, so the price of the bundle is (1-d)*(2 PA), where d is the discount. Then the bundle will be chosen if VC + VFi+Fj – (1 – d)*(2 PA) > Max (VC + VFi - PA, VC + VFj PA), where VFi + Fj = VFi + VFj|Fi is form utility from buying both items (see footnote 1). If VFi > VFj it follows that the bundle is chosen if VFj|Fi > (1-2d)PA, or if the incremental value of j exceeds its incremental cost. A similar condition can be worked out for the case in which VFj > VFi. Adding the conditional probabilities of these two cases gives the probabilities of bundle choice expressed in conditions 5 and 6 in Table 1, panel B (condition 5 expresses the special case of d = 0). Let the price of item i increase by kPA, and the price of item j decrease by an equal amount, so that the sum of the prices of the two items is still 2PA. Now, because of the shift in relative price, i is more attractive only if VC + VFi – (1 + k) PA > VC + VFj – (1 - k) PA, or if VFi – 2kPA > VFj. If this condition holds, the bundle will be chosen if VC + VFi+Fj– (1 – d)*(2PA) > VC + VFi – (1 + k) PA, which reduces to: VFj|Fi > (1 - k - 2d)PA. A similar expression can be worked out for the case in which j is more attractive (VFi –
22
2kPA < VFj). Adding conditional probabilities of these two cases gives the probabilities of bundle choice expressed in conditions 7 and 8 in Table 1, panel B (condition 7 expresses the special case of d = 0). Comparison of the conditions in panel B of Table 1 leads to predictions about the effects of discounts and relative prices on the acceptance of the bundle. First, because it lowers the incremental cost of the less preferred item, the discount makes condition 6 > 5 and 8 > 7, and there should be a main effect of discount.7 Whether a change in relative prices will lead to increased purchases of the bundle depends on how the relative valuation of the two items varies across consumers. As shown in conditions 7 and 8, when VFi – 2kPA > VFj the probability of buying both items increases because the less valued item becomes cheaper, and it can also be shown that the probability of buying the bundle will increase as long as VFi – kPA > VFj.8 If this condition does not hold, the bundle becomes less attractive when there are different relative prices, and the net effect of changes in relative price on bundle choice depends on the distribution of consumer valuations.9 Given that most consumers employ the print versions of books and newspapers, it seems reasonable to assume as a maintained hypothesis that most consumers place a relatively high valuation on the print version.
To compare conditions 6 and 5, let Fj|i(V’) represent the density of consumer valuations VFj|Fi between 0 and V’, and Fi|j(V’) be a similar density for VFi|Fj. Then the share of consumers who would buy the bundle with a discount, but not otherwise, is (Fj|i(PA)- Fj|i((1-2d)PA))(S) + (Fi|j(PA)- Fi|j((1-2d)PA))(1-S), where S is the share of consumers for whom VFi>VFj. A similar condition can be worked out for comparison of 8 & 7. 8 Let 0 < VFi – VFj < 2kPA, in which case the bundle is chosen if VFi|Fj = VFi+Fj - VFj > (1+k-2d)PA. Let VFi VFj =?, so VFi+Fj - VFi +? > (1+k-2d)PA, or VFj|Fi= (1+k-2d)PA- ?. If ? > kPA, the incremental value of i exceeds its cost, and the bundle becomes more attractive relative to the case of equal prices. If ? < kPA, the incremental value of i is not sufficient to cover its cost. 9 To compare conditions 7 and 5, use the same notation for the density of consumer valuations as in footnote 5, let S’ be the share of consumers with valuations VFi – VFj > kPA, S* be the share of consumers with valuations 0 < VFi – VFj < kPA, and 1- S’ – S* be the share with valuations VFj > VFi. Then the share of consumers who would buy the bundle compared to the case of equal relative prices is (Fj|i(PA)- Fj|i((1k)PA))(S’) - (Fj|i((1+k)PA)- Fj|i(PA))S*- (Fi|j((1+k)PA)- Fi|j(PA))(1-S’- S*). A similar comparison between 8 and 6 could be obtained from subtracting 2d PA from the price terms. Since 8 vs. 6 involves different parts of the price distribution compared to 7 vs. 5, the two comparisons need not give the same results.
7
23
Hypothesis 4 (information product form bundles): a.) Across both discount conditions, if the preference for the traditional (print) version (i) over the electronic version (j) is sufficiently large (VFi - VFj > kPA), consumers are more likely to buy the product form bundle when the traditional print product is priced at a premium and the electronic product is priced at a discount as compared to when the products are priced at the same level. b.) Consumers are more likely to buy the bundle when a discount is offered. The intuition behind Hypothesis 4a is straightforward. The more item prices match consumer valuations, the easier it is to induce consumers to buy both items. If VFi is very high relative to VFj, setting equal prices will make j unattractive, but leave the consumer a surplus on i. It becomes difficult to induce the consumer to buy both items. Conversely if VFi is only slightly higher than VFj, setting a high price on i will make it unattractive, but will leave the consumer with a surplus on j. If VFi is very high relative to VFj, offering different relative prices can have the same effect as offering a bundle discount (compare conditions 6 and 7 in Table 1). However, as they extract more surplus, different relative prices will bring more revenue. We can also present an alternative intuition for Hypotheses 4 from the perspective of consumers’ evaluation process. Recall that under the no discount case, the price of the bundle is the same across the same price (case 1) and differential price (case 2) conditions. Given that consumers have higher preference for the print version, it becomes the anchor for evaluating the bundle (see Yadav 1993). Under the same price condition, consumers have a “gain” on the anchor form (print) in the bundle, while they have a “loss” on the less preferred form (electronic), as the individual prices of the different forms could act as reference points in evaluating the bundle. Since perceived loss has a
24
greater impact than the perceived gain (Kahneman and Tversky 1979), the bundle is still evaluated negatively overall. In the differential price case, the prices are more in line with consumers’ expectations (driven by their valuation) and there is no significant gain/loss on the anchor form and the other form. Thus, the bundle is evaluated more favorably. We argue that the price level of the less preferred form relative to consumer valuation in case 1 (same prices) heightens the poor evaluation of the form and the bundle, while the lower price in case 2 does not have such an effect on the bundle evaluation. Giving a price discount for the bundle can help overcome the negative evaluations and improve the evaluation and purchase of the bundle. In the following sections we present the results of two studies designed to test the hypotheses outlined above.
Study 1 The main objective of this study is to test Hypothesis 1. Additionally, we introduce a bundle discount variable and investigate whether the level of bundle discount affects the purchase likelihood for the product form bundle. Specifically, we are interested in whether the bundle discount per se motivates consumers to buy the bundle. Method Subjects. Subjects were 80 undergraduate students enrolled in an introductory marketing course. They were awarded extra course credit for their participation. Product categories and materials. We used two information product categories and two conventional product categories, specifically books (print and electronic) and journal subscriptions (print and electronic), and margarine (stick and tub) and soap (bar
25
and liquid). Advertisements for fictitious brands were developed to describe these products. Design and procedure. We employed a 2 (level of discount: 0%, 25%) by 2 (category type: information, conventional) between subjects design with 2 product replicates. The order of presentation of the product categories was counterbalanced within each category type. The prices were presented in absolute values (price of form 1, price of form 2, bundle price). Every subject evaluated two product advertisements for either book and newspaper, or margarine and soap. Subjects reviewed each ad and stated their likelihood of purchasing the three options (the bundle and the two individual product forms). Purchase likelihood was measured in two different ways: by allocating 100 points between the three options10, and by indicating how likely they would be to buy the three options (1 = “very unlikely,” 9 = “very likely”). Next the subjects were asked to rate the perceived complementarity of the two product forms using a 3-item Likert scale11 (1 = “strongly disagree,” 7 = “strongly agree”). Finally, several control measures were taken including attitude towards the ad (bad/good, unpleasant/pleasant, worthless/valuable, and unfavorable/ favorable; MacKenzie and Lutz 1989; Mick 1992) and demographics (e.g. gender and computer usage).
Results
10
We do not include the “buy nothing” option in the 100-point allocation among the different product forms because we are interested in which form(s) our subjects are going to choose after they have decided to buy the product. Thus we try to control for other factors that may influence the product purchase decision and are not of specific interest in this study. 11 “It would be more useful to have both the paperback book and the electronic book than just the paperback book”, ”There is additional value in having both the paperback book and the electronic book as compared to having only one of them” and “It would be more useful to have both the paperback book and the electronic book than just the electronic book.”
26
The reliabilities of the perceived complementarity scales are between 0.81 and 0.92 in the four categories (N = 40 per category). The mean levels of perceived complementarity (after averaging the three items of the scale) between the product forms are as follows: XBOOK = 3.21 (St. Dev. =1.67), XNEWS = 3.17 (1.62), XMARG = 3.56 (1.81) and XSOAP = 4.19 (2.10). To test Hypothesis 1 we use linear regression in which the independent variable is the perceived complementarity between the product forms in the bundle, and the dependent variables are the points allocated to the bundle and the likelihood of buying the bundle (separate model for each product category and each dependent variable). Our regression results indicate that in the book, margarine and soap categories there is a positive relationship between the degree of perceived complementarity and the purchase likelihood for the bundle (Likelihood of buying the bundle: bBOOK = 0.535, bMARG = 0.746, bSOAP = 0.731, all p’s<0.01; Points allocated the bundle: bBOOK = 0.466, bMARG = 0.675, bSOAP = 0.626, all p’s<0.01), thus providing support for H1. In the newspaper category the relationship between perceived complementarity and the likelihood of buying the bundle is also positive (bNEWS=0.297, p=0.063), and the relationship between perceived complementarity and points allocated to the bundle is positive but not significant (bNEWS=0.178, p>0.1). In sum, our results generally provide support for Hypothesis 1. We use one-way ANOVA to investigate the effect of bundle discount on the purchase likelihood for the product form bundle. In the book and newspaper categories, there is no significant difference between the points allocated to the bundle and the likelihood of buying the bundle in the discount and no discount conditions (all p’s>0.44).
27
A discount of 25% does not make consumers more likely to buy information bundles. The perceived incremental value of the second form must therefore be less than 50% of the value of the first form. This is to be expected since both forms have the same content with its utility derived in one form or the other, but not in both. In the margarine and soap categories the situation is different – in both categories respondents are more likely to buy the bundle and allocate more points to the bundle in the discount versus the no discount conditions (Likelihood of buying the bundle: XMARG/HIGH = 4.90 versus XMARG/LOW = 2.74, F1,38 =6.20, p<0.05; XSOAP/HIGH = 6.62 versus XSOAP/LOW = 3.42, F1,38 =13.39, p<0.01; Points allocated the bundle: XMARG/HIGH = 32.33 versus XMARG/LOW = 14.47, F1,38 =4.14, p<0.05; XSOAP/HIGH = 43.75 versus XSOAP/LOW = 23.42, F1,38 =4.58, p<0.05). This indicates that, because consuming one form does not reduce the value of the other, consumers perceive the bundling of conventional product forms as a quantity discount and are more willing to buy the bundle if a discount is offered. In general, consumers were significantly more wiling to buy both forms of the conventional products than the information products, supporting Hypothesis 1a.
Discussion On the basis of these results, it appears that the purchase behavior of the consumers is related to the level of perceived complementarity between the different forms in the bundle, and can be positively affected by offering a nominal discount of 25% in the conventional but not in the information product cases. In Study 1 we do not communicate any usage situations to the consumers. We only present them with a short product description and with the item and bundle prices. We use these results as a
28
baseline and design a second study that investigates what happens when we communicate various usage situations to the consumers. Our goal is to show that although the bundle discount itself does not motivate consumers to buy information bundles, it can be effective when different usage situations for the forms are presented.
Study 2 Study 2 tests hypotheses 1a, 2, 3 and 4. In the usage situation manipulation we present consumers with possible usages for the two product forms in the following way: in the same usage situation condition we advertise situations for which the two forms are equally appropriate, while in the different usage situation condition we suggest usage situations for which the two forms are differentially appropriate. Below we describe how the stimuli were developed. Stimuli Development Product categories. As in Study 1, we use two information and two conventional product categories - book and newspaper, and margarine and coffee, respectively. Usage situations. We employed extant substitution-in-use (SIU) methods to generate usage situations appropriate for each product form (see Srivastava et al. 1984). The SIU approach (Stefflre 1971) is an iterative procedure for constructing product specific usage-situational taxonomies. Using the SIU approach, the perceived similarity between the different alternatives for the usage situations of interest and the resulting product-market structures can be investigated (Ratneshwar and Shocker 1991; Srivastava, Leone and Shocker 1981). In our study, first, a sample of consumers (N = 15) generated a set of usage
29
situations for the two forms in each product category. Then, a second sample (N = 10) evaluated the appropriateness of each product to each usage situation on a yes/no/don’t know scale. Finally, a structured questionnaire with product forms and usages was developed and administered (N = 67). The subjects judged the appropriateness of each product form for each usage situation on a 5-point Likert scale (not appropriate to very appropriate) including a “don’t know” option. << Insert Table 2 about here >> The perceived appropriateness of the generated usage situations for each product form is presented in Table 2. For book, newspaper, and coffee product categories, one of the two forms is perceived as more appropriate for some usage situations and the other product form is perceived as more appropriate for other usage situations. For example, in the book product category, a print book is perceived as more appropriate than an electronic book for giving as a present (XPRINT – XELECT = 2.24, p<0.05), reading for pleasure (MD = 1.76, p<0.05), reading while traveling (MD = 1.74, p<0.05) and reading it to other people (MD = 1.45, p<0.05). On the other hand, an electronic book is perceived as more appropriate than a print book for e-mailing pages/chapters (MD = 2.43, p<0.05), searching (MD = -0.95, p<0.05) and copying citations/paragraphs (MD = 0.60, p<0.05). Both product forms of the above three categories are perceived as equally appropriate for several usage situations. For example, in the book product category, the perceived appropriateness of getting them on a short notice and using them for archival purposes are not significantly different across product forms. Materials. Advertisements for fictitious brands were used to describe these products and manipulate the usage situations. We selected the usage situations in which
30
one product form is perceived as significantly more appropriate than the other for designing the different usage situation manipulation, and the situations in which the product forms are equally appropriate for designing the same usage situation manipulation. The number of usage situations is balanced across conditions and everything else in the advertisements is kept constant. Thus, we presented four usage situations in total in each advertisement: in the same usage situation condition we listed four situations in which the two forms are equally appropriate, while in the different usage situation condition Form 1 is more appropriate for two of the usage situations and Form 2 is more appropriate for the other two usage situations. With respect to the margarine product category, there was no significant difference in how the two product forms were perceived for most of the situations. Therefore, we used the same four usage situations to manipulate the same/different conditions, just changing the wording – stating that form 1 is more appropriate for two of the usage situations and form 2 is more appropriate for the other two usage situations in the different usage situation condition, while stating that the forms are appropriate for all four situations in the same usage situation condition. We consider this a more conservative manipulation and elaborate on this in the results section.
Pretest 1 We pretested the two advertisements for each of the four categories using a between-subjects design in which each subject saw only one advertisement (N=160). Based on MacKenzie and Lutz (1989) and Mick (1992), we use eight semantic differential scales to measure attitude towards the ad (bad/good, unpleasant/pleasant,
31
worthless/valuable, and unfavorable/favorable) and ad credibility (unconvincing/convincing, biased/unbiased, unbelievable/believable and noninformative/informative). Across the categories, the alphas varied between 0.85 and 0.93 for attitude towards the ad, and between 0.74 and 0.85 for ad credibility. The results showed that the attitudes towards the ads and the ad credibility were not significantly different across the manipulated same/different usage situation conditions. Therefore, differences in responses to the advertisements across conditions cannot be attributed to differences in attitudes towards the ad/ad credibility.
Pretest 2 We performed a second pretest of our stimuli (N=75) to assess the extent to which the product forms were perceived as complements using the same complementarity scale as in Study 1. The results indicated that our subjects viewed the product forms as more complementary in the different usage situation condition (vs. the same usage situation condition) in the book, newspaper, coffee and margarine product categories (XBOOK/DIFF = 4.32 versus XBOOK/SAME = 3.00, F1,35 =5.21, p<0.05; XNEWS/DIFF = 4.25 versus XNEWS/SAME = 2.96, F1,35 =5.33, p<0.05; XMARG/DIFF = 4.77 versus XMARG/SAME = 3.48, F1,36 =4.33, p<0.05; XCOFF/DIFF = 5.46 versus XCOFF/SAME = 4.25, F1,36 =5.31, p<0.05). Consequently, we decided to proceed with the study.
Method Subjects. 406 undergraduate students enrolled in several marketing courses participated in the experiment (N = 240 in the information bundle categories, and N=166
32
in the conventional categories)12. They were awarded extra course credit for their participation. Design and procedure. We employed a 2 (usage situations: same, different) by 2 (level of discount: 0%, 25%) by 2 (category type: information, conventional) between subjects design with 2 product replicates within each category type (book and newspaper, and margarine and coffee). Additionally, for the information category type, we introduce another between subjects factor, relative price, which was varied so that the print form price was either equal to the electronic form price or higher than the electronic form price. The prices were presented in dollar amounts. We used the same experimental procedure and measures as in Study 1.
Summary of Study 2 Results As in study 1, respondents allocated 100 points between the individual items and the bundle to indicate their likelihood of purchasing each alternative. Table 3 presents the average share of the points that were allocated to the bundle in each of the experimental conditions; these shares can be interpreted as average probabilities of choosing the bundle. In accord with Hypothesis 1a, the table shows that the average probability of choosing conventional bundles is considerably higher than the average probability of choosing information bundles in all conditions. Consistent with Hypothesis 2, the discount and different usage conditions only have a major effect on the probability of choosing information product bundles when they are combined. On the other hand, consistent with Hypothesis 3, the discount and different usage conditions are both associated with higher choice probabilities for conventional bundles across all conditions.
Our use of the additional relative price manipulation necessitated a larger sample for the information products.
12
33
Finally, the high price for print and low price for electronic items is associated with a higher incidence of preference for the bundle, as predicted by Hypothesis 4, only in the no discount case. We present a more complete analysis of these results, including formal hypothesis tests, in the following section. << Insert Table 3 about here >>
Detailed Results We first discuss the manipulation check, and then present results of hypothesis tests for conventional products, followed by information products. Manipulation check. The reliabilities of the perceived complementarity scales were between 0.82 and 0.92. We ran one-way ANOVAs for each product category with USAGE as the between-subjects factor and perceived complementarity as the dependent variable. In all four categories the perceived complementarity in the different usage situations condition was significantly higher than the perceived complementarity in the same usage situations condition (USAGE: XBOOK/DIFF = 3.84 versus XBOOK/SAME = 3.32, F1,239 =4.33, p<0.05; XNEWS/DIFF = 3.99 versus XNEWS/SAME = 3.46, F1,238 =4.46, p<0.05; XMARG/DIFF = 4.65 versus XMARG/SAME = 3.86, F1,78 =4.27, p<0.05; XCOFF/DIFF = 4.56 versus XCOFF/SAME = 3.81, F1,84 =5.16, p<0.05). Thus, we successfully manipulated the same/different usage situations variable and could proceed with the analyses. The results of the analyses employed in testing Study 2 hypotheses are presented in Table 4 (GLM procedure results) and Figure 1 (the interactions in a graphical form). Since it is bounded at zero and 100, the measure of points allocated to the bundle employed as our dependent variable may not be normally distributed. Accordingly we
34
applied a logit transform in which the dependent measure is defined as ln ((points + .5)/(100 – points + .5)) in conducting hypothesis tests.13 << Insert Table 4 and Figure 1 about here >> Conventional products. According to Hypothesis 3 we expect a main effect of both usage situation and bundle discount on likelihood of buying the bundle. As anticipated, two significant main effects were found in the analysis of the margarine and coffee data - USAGE (margarine: F1,76 =6.01, p<0.05; coffee: F1,82 =4.09, p<0.05) and DISCOUNT (margarine: F1,76 =9.24, p<0.01; coffee: F1,82 =23.65, p<0.05). The two-way interaction between USAGE and DISCOUNT is not significant in both product categories (p>0.1). Consequently, our results in the margarine and coffee categories provide support for Hypotheses 3a: consumers are more likely to buy the product form bundle when they are suggested different usage situations for the two product forms as compared to when they are suggested the same usage situations for the two product forms, and for Hypothesis 3b: consumers are more likely to buy the product form bundle when they are offered a bundle discount as compared to when they are not offered a discount. In the margarine case just stating that one form was more appropriate than the other for a given situation was sufficient to produce the desired manipulation. In the sense that it is not reinforced by a general perception that the different forms are more appropriate for different usage situations, this manipulation can be regarded as conservative. Evidently it is possible to create a perception of appropriateness for a given usage situation by simply presenting information that this is the case.
13
The addition of .5 makes it permissible to take logs in cases where points = 0 or 100. While we employed the logit transform, models using the raw points variables gave very similar results.
35
While not shown in Table 4, we also tested the hypothesis of equal choice probabilities for electronic and information products. As one might expect from the large differences in average choice probabilities between conventional and information products in Table 3, this hypothesis was rejected (F1,669 = 76.69, p<0.001). Information products. According to Hypothesis 2 we expect the usage situations considered when making a purchase decision to moderate the effect of bundle discount on purchase likelihood. To test this hypothesis, we use the GLM procedure in SAS in which the dependent measure is the points allocated to the bundle, and USAGE, DISCOUNT and RELATIVE PRICE are the between subjects factors. As anticipated, a significant USAGE by DISCOUNT interaction was found in both categories (F1,232 =6.03, p<0.05 for book, and F1,232 =3.91, p<0.05 for newspaper). The first planned contrast revealed that when the consumers are presented with different usage situations, they are significantly more likely to buy the bundle in the discount versus no discount condition (Fcontrast1,232 =13.28, p<0.01 for book, and Fcontrast1,232 =13.64, p<0.01 for newspaper), and the second planned contrast revealed that when consumers are suggested the same usage situations there is no significant difference in their purchase likelihood in the discount as compared to the no discount condition (p’s>0.37). Consequently, our results provide support for Hypotheses 2a and 2b. Finally, the last planned contrast (testing Hypothesis 4) showed that there is no significant difference between the likelihood of buying the bundle when the traditional print form is priced at a premium and the electronic form is priced at a discount, and the likelihood of buying the bundle when the traditional and the electronic forms are priced at the same level (p’s>0.42). This is consistent with the data in Table 3, which show similar
36
choice probabilities for both conditions. A post hoc contrast comparison revealed that our prediction is valid in the case of no bundle discount in the book category (Fcontrast1,232 =3.78, p=0.05): as shown in Table 3, the average probability of buying both print and electronic books is .1195 when the electronic form has a lower price, but only .0610 when prices are equal. However a similar comparison was insignificant in the other cases. The results for Hypothesis 4 can be explained as follows. Recall that our test of Hypothesis 4 was based on the assumption that the print form is valued significantly more than the electronic version. Specifically we assumed VFi - VFj> kPA, where k is increase in price of the print version (decrease on price of the electronic) as a proportion of price. In our study k = .25, so the print version must be worth at least 25 percent more than the electronic form as a proportion of the average price to make the bundle more attractive. However, a substantial proportion of our respondents would likely choose the electronic version at equal prices for print and electronic: at equal prices the average number of points allocated to the electronic book was 37 of 100; the average for the electronic newspaper was 40 of 100 (these averages did not differ much between discount conditions). In the no discount condition, the corresponding averages for print were 56 and 53 respectively.14 These results suggest that VFi and VFj (the utilities derived from the form per se) are not very different for many consumers. If so, since content utility can be obtained from any of the alternate forms, these consumers are likely to be better off buying just the electronic version when it has a low price rather than buying the bundle, which is what happened in our study. At a low price for the electronic version, the proportion buying only electronic form increased by about 10 percent for books, and 13 percent for newspaper. These consumers likely did not choose the bundle because the
14
In the discount condition, about ten percent more respondents chose the bundle rather than print only.
37
print version became too expensive relative to its incremental value. Thus our test of Hypothesis 4 appears to fail because the electronic form is valued more than we anticipated. Additionally, an interaction between the content and form utility may explain the results associated with Hypothesis 4.
Discussion The results of our study provide useful insights into how consumers react to product form bundles. In the context of information products, since consumers obtain content utility only once in buying the bundle (content utility is fixed going from either of the form to the bundle), the product forms tend to be considered overall more as substitutes. Even with differing degrees of perceived complementarity/contingency across our subjects, the purchase probabilities of the bundle do not increase with just price discounts on the bundle. The bundle purchase probabilities increase only when price discounts are accompanied by communication about different usage situations to impact the consumers’ perceived complementarities between the two forms in the bundle. These results may be more specific to the valuations of our student respondents, who we observed not to value form utilities very much in comparison to content utility, but they also point to the fact that different usages/usage situations of the forms (and form utilities) have to be emphasized clearly in order to highlight the value of the information product bundle, no matter which consumer group we market the bundle to. The important take-away is that such communication strategies help significantly in the case of information product bundles and, thus, marketers have a viable strategy to harvest the potential extra revenues in selling bundles.
38
In the case of conventional product form bundles on the other hand, the absolute probabilities of purchasing bundles are much higher as compared to information product bundles, even though the forms are perceived as strong substitutes. This is not surprising as consumers obtain content utilities from both forms in the bundle (content utility increases in going from either of the form to the bundle), and hence, in some sense, it is not a proper comparison. Price discounts for conventional product bundles increase the purchase probabilities of the bundles as they work as quantity discounts, and communication about different usages also seem have to a positive impact. An interesting observation from our study is that, although the product forms seem to be moderate/strong substitutes, bundle purchase probabilities are significantly high enough to make mixed bundling a viable and profitable strategy in both categories, albeit for different reasons. While in the case of conventional product form bundles, mixed strategies could be arguably optimal in the presence of quantity discounts, in the case of information product form bundles, mixed strategies can be very profitable given the very low marginal cost of the additional form. In both categories we observe that the communication strategies play an important role in supporting mixed bundling strategies. In this regard, our results are somewhat different from the traditional bundling results, which suggest a pure component strategy in the presence of strong substitutes and low marginal costs (Venkatesh and Kamakura 2003), and provide additional insights for the case of product form bundles. Our results also show some support for differentially pricing the information product forms. While we show that this is better than pricing the forms equally if consumers have a relatively high preference for one item as compared to the other, in our
39
experiment we could not enforce this condition. Nevertheless, the hypothesis was supported in the case of books under the no discount case. An interesting interpretation of the result is that the prices of the individual product forms could play an important role in bundle evaluation. If the price of the less preferred form is more than the consumers’ valuation for the form, there could be a magnified negative effect on the bundle evaluation. This implies that price information can have similar effects as non-price information studied in extant literature (e.g., Yadav and Monroe 1993). While this notion needs to be tested more formally, it does highlight the importance of pricing the individual product forms appropriately in a mixed bundling strategy.
Managerial Implications How should companies market information bundles? In practice, online publishers use various product and pricing strategies. For example, Wall Street Journal offers its online edition at 40% of the price of the print edition, and the discount for choosing the bundle is 17%. Business Week and Fortune offer free online content for their print subscribers only, and Newsweek and Washington Post offer their online editions for free to any interested readers. Finally, National Academies Press, a publisher selling books online, offers electronic copies for most of its books at about 75% of the print book price, and an average bundle discount of 40%. In general, the prevailing marketing practice in the case of information products has been to price differentially the various forms and to offer bundle discounts. With the exception of WSJ, none of these organizations stresses different usage situations for the print and electronic forms. However, our results indicate that information providers should try to
40
communicate different usages for their product forms to make them to be perceived as more complementary forms rather than just substitutes. We provide evidence that discounts commonly used in practice work only when different usage situations for the product forms are communicated. The message that we would like to convey is that stressing different usages for the product forms may be needed to motivate consumers to buy the bundle even when a discount is present. We find that communication of different usage situations, or other attempts to enhance the value of alternative forms of information products can increase the salience of the form utilities and may be needed because of the inherent duplication of content that is provided by the different forms of information products. If consumers do not perceive the form utilities to be very significant, then the content utilities can render the forms to be very strong substitutes. Something has to be done to induce buyers to purchase a second form in the face of this perception of substitutes due to content. One strategy is to make the functionalities of the forms very different and educate the consumers of the relative usage situations of the different forms. This will make the form utilities much more significant. The other option is to change the content of the two forms – for example, online newspapers are updated often during the day and thus the content is more dynamic as compared to the static content of the print form. Similarly, electronic versions of books can be updated and distributed to consumers for free. Electronic versions can also provide links to other relevant content that print versions cannot provide. However, this will increase the cost of the electronic form, and the cost side should be taken into account before considering such a strategy. Our results also suggest that if one form is significantly preferred much more than
41
the other, charging a high price for the more preferred form, and a low price for the less preferred form may be an effective strategy. This might well be as effective as giving discounts while providing higher revenues to the marketer. These results have particular relevance in the bundling of online content such streaming video and DVDs, online music, CDs, and DVD-Audio, where it is not clear how consumers’ valuation for the different forms are distributed. They also suggest as the online channel matures and quality of digital content improves, differential pricing strategies and discounting strategies should change significantly.
Conclusion The study contributes to the marketing literature in several ways. To our knowledge, this is the first attempt to study the effects of product form bundling on consumer preferences and consumption behavior. We provide a useful approach using the perspective of content utility and form utility, and we model the effect of both price and non-price information in bundle evaluation process and draw conclusions about their impact. We also introduce a usage situational perspective in studying bundling issues. The SIU approach has been proven successful in investigating the influence of usage situational variables on consumers’ behavior and is helpful in the present setting as well. Our study complements the findings in extant literature on bundling by providing insights into conditions under which those findings are directly applicable to product forms and instances where those results may not hold. We show that the perceived complementarity among the bundle components is extremely important and suggest how product form bundles might be marketed. Finally, we provide evidence that the conventional and
42
information product form bundles are different from a conceptual point of view and discuss a mechanism that can explain this difference. Several possible ways for extending the present study are worth mentioning. First, explaining the distinction between conventional and information product form bundles is worth future effort. Consumers may see bundling of conventional product forms as a quantity discount, while bundling of information product forms as bundling of complementary items. Second, it will be interesting to see how exactly the consumers evaluate product form bundles. Some authors suggest an anchoring and adjustment model for bundle evaluation (i.e. Yadav 1994), while others argue for an averaging model whereby component ratings are balanced into an overall evaluation (i.e. Gaeth et al. 1990). This issue is especially important in designing and pricing of electronic products and bundling them with traditional products. Different pricing and price presentation strategies should be employed in the above cases of bundle evaluation. A third interesting research issue comes from the fact that the digital form offers the potential of an augmented version of the traditional form. Consequently, companies may be able to increase the level of complementarity of the forms in information bundles by somehow changing or improving the electronic form, and thus positively influence bundle sales. Finally, it is also important to support/verify our findings using empirical data from online content providers. Future research should address and clarify such strategies.
43
CHAPTER III: ESSAY TWO. BUNDLING AND PRICING STRATEGIES FOR DIGITAL PRODUCTS Abstract Content providers such as publishers of books, newspapers and magazines have started to offer products in electronic form, and even individual sections, in addition to their print products. For example, some book publishers offer print and PDF books, while others offer print books and individual PDF chapters. Which strategy is more attractive for a company to pursue? We model the profit under these two strategies and outline conditions in which one or the other leads to higher profit. We apply our analytical framework to data from a field experiment implemented on the website of a book publisher.
44
Introduction
“U.S. consumer spending for online content grew 14% in 2004 reaching an all-time annual high of $1.8 billion…While there is no doubt that the market remains strong, with only 11.6% of the Internet population purchasing content online in Q4 2004, there is still significant room for growth .”
(Michael Zimbalist, President, Online Publishers Association15)
The emerging technological solutions and the Internet channel unequivocally change the marketing of information. In the music category, for example, tapes and CDs give way to music downloads and online subscription services, and new formats as those introduced by dualdisc.com allow content providers to sell music in CD format plus DVD-Audio or surround sound as a bundle on the same disk. A similar trend is observed in the publishing industry where print products such as newspapers and magazines are offered in electronic form too, and consumers are even encouraged to buy both forms (e.g., Wall Street Journal). The electronic products can be unbundled at no extra cost into separate sections to be sold individually or re-bundled with the traditional print products. Although not yet common, publishers do offer full and unbundled electronic content in addition to print content, allowing customers to buy, for example, individual PDF chapters or entire PDF books and reports separately or together with the print editions (e.g., US Government Printing Office). While appealing because of low marginal costs and the likely catering to emerging consumer needs, selling electronic products may lead to additional revenue generation but can also result in product cannibalization. The National Academies Press (NAP), for instance, publishes over 200 books a year on a wide range of topics in science, engineering, education and health, including full length books, reports and reference materials. In 1996 NAP started posting most of
15
Online Paid Content U.S. Market Spending Report 2004, Online Publishers Association, March 2005, www.online-publishers.org
45
its titles online as low (fax) quality content free for anyone to browse, search and sample. Although majority of the consumers continued buying print books, many website visitors began utilizing the online capability to download chapters, some even demanding higher quality electronic content at a price. Thus, NAP decided to consider two alternative product line strategies – on one hand, introducing books in PDF form in addition to the current print books; on the other hand, marketing individual PDF chapters with the print books. It is in this context that we execute our study. In this paper, we attempt to model the profits under two feasible strategies a content provider may employ: traditional mixed bundling, in which the product line consists of print book, PDF book and the bundle of the two, and content unbundled mixed bundling, which is selling print book, PDF chapters and the combination of the two. Although traditional mixed bundling is a special case of content unbundled mixed bundling in which consumers are constrained to buy all the chapters (instead of one or more chapters), we model them separately in order to provide better insights on what determines the optimal prices and profits. Our goal is to provide insights on which strategy is more beneficial for a company to pursue and how the electronic content should be priced to extract optimal profit. Further, we empirically test our analytical predictions with data from an online experiment executed on the website of NAP. Our study contributes to the marketing theory and practice in the following manner. While there is an extensive body of research suggesting that mixed bundling is the optimal strategy when there is asymmetry in the reservation prices for the bundle components (e.g. Adams and Yallen 1976; Schmalensee 1984; McAfee, McMillan and Whinston 1989), few studies focus on cases in which different mixed bundling strategies
46
are feasible and evaluate their attractiveness. Also, we look at form bundling and content unbundling simultaneously, which makes our approach unique and adds to the extant literature. Finally, while previous studies often treat bundle components as either complements or substitutes or unrelated products, we allow heterogeneous contingent valuations of one component given the other in the population. From a substantive point of view, we show empirically that the content unbundled mixed bundling strategy is more profitable than the traditional mixed bundling strategy under certain conditions. Our findings have significant practical implications, recommending how to profitably design and price electronic content. The paper is structured as follows. First, we review related research and position our study. Second, we model the profits under traditional and content unbundled mixed bundling strategies, and derive which one is better under various conditions. Third, we report the empirical results and conclude with the implications of the study to the extant literature and practice.
Research Background Bundling is marketing two or more products and/or services as a package at a special price (Guiltinan 1987). Demand side incentives favoring bundling include negative correlations in reservation prices (Stigler 1963; Schmalensee 1984), complementarity in consumption (Telser 1979), and uncertainty in the valuations of the quality of the goods (Kenney and Klein 1983). With respect to supply side incentives, large scale bundling of information goods, for example, can be profitable because it creates economies of aggregation when their marginal costs are low (Bakos and
47
Brynjolfsson 1999; 2000). Companies can employ a pure bundling strategy whereby only the bundle is offered, or a mixed bundling strategy, in which the bundled items are also sold separately. Extant research has looked at product and pricing strategies for bundles, including optimal bundle price and composition (Hanson and Martin 1990; Venkatesh and Mahajan 1993), how the degree of heterogeneity in the reservation prices affects optimal bundle pricing (Jedidi et al. 2003) and conditions favoring bundling/unbundling of industrial systems (Wilson, Weiss and John 1990). In terms of contingent valuations in bundling decisions, studies have compared interrelated with independently valued products in a bundle. For example, Venkatesh and Kamakura (2003) model the optimal bundling strategies for interrelated products under monopoly and suggest that moderate or strong substitutes should be offered separately; the same is applicable for complements if the marginal costs are moderate relative to the market’s maximum willingness to pay. A seller would gain by mixed bundling for weak substitutes/complements if the variable costs are not too high. Regarding information goods (e.g., music, weather forecasts, websites), large scale bundling is approximately optimal if consumers’ values of subsequent goods do not decrease too quickly; otherwise pure bundling is optimal even when there are strong negative or positive correlations of values across goods (Geng et al. 2005). To summarize, the literature on bundling identifies the conditions for profitable bundling and specifies optimal bundling strategies in various situations. What is missing in this stream of research is how to handle situations in which different mixed bundling strategies are feasible. With respect to information products, companies can implement
48
both traditional and content unbundled mixed bundling strategies, and which one is more profitable does not have a straight-forward answer. Different consumers may have different contingent valuations for the form bundles, and different contingent valuations for the unbundled electronic units (e.g., chapters within a book). Consequently, the existing premises on optimal bundling strategy and pricing can not be applied directly. Our predictions and empirical findings are relevant and insightful from both theoretical and managerial point of view.
Bundling Model In our model the seller is a monopolist who offers a wide selection of book titles as hardcover or paperback books. Subject areas are very diverse, ranging from agriculture and education to medicine and engineering. The product and pricing strategy the publisher is employing at the moment is segment-based and reflects the needs, buying power and quality expectations of the consumers in the different subject domains. The publisher is currently choosing between two possible product line strategies: 1) Content Unbundled Mixed Bundling - selling PDF book chapters individually and together with the print books, and 2) Traditional Mixed Bundling - selling full PDF books individually and together with the print books. Next, we express optimal pricing of electronic content and profit under each strategy, and then compare the two profits.
Content Unbundled Mixed Bundling Model basics
49
We assume that a book has J chapters and PC is the price per chapter
(P1 = K = PJ
= PC ) . V ji is the valuation of chapter j
( j = 1K J ) by consumer i
(i = 1K I ) . The chapters are rank-ordered in terms of their valuation starting from the
highest, e.g. V1i ? V2i ? ... ? V Ji . The number of chapters a consumer buys depends on the valuation of the chapters, namely consumer i buys chapter j if V ji ? PC and V j +1,i < PC . Thus, consumer i will buy exactly QC rank-ordered chapters if VQi ? PC and VQ +1,i < PC .
Revenue from selling electronic chapters In this section we first express the revenue from selling electronic chapters at individual level and then the aggregate the electronic chapters revenues across consumers. At this point we only consider offering electronic chapters to consumers; the print book sales and the associated revenue and costs are introduced in the next section. The optimal price per chapter and the total and marginal revenues from chapters at individual level are graphically displayed in Figure 2. << Insert Figure 2 about here >> We assume that the demand for chapters is linear. The valuation for the chapters, VC , is presented on the y-axis, and the cumulative number of chapters at every valuation point, QC , is presented on the x-axis. N C is the number of chapters with positive valuation (e.g., the consumer will buy N C chapters if the price is zero). The optimal price per chapter PC is equal to the valuation of the marginal chapter by the consumer. Let V Hi be the valuation of the chapter with the highest valuation for a consumer. The quantity of chapters bought and the marginal revenue from chapters can be expressed as follows:
50
(1)
QC = N C ? PC
NC VHi
(2)
MRC =
? (QC PC ) = ? PC
? ?? ? N C ? PC
??
??
NC VHi
? PC
? ? ? ? PC ? N ? ? = N C ? 2 PC C = 0 VHi
1 Solving equation 2 for PC , the optimal price per chapter is PC* = V Hi and the optimal 2
1 * revenue is RC = V Hi N C . Thus, the optimal price depends on the valuation of the highly 4
valued chapter, while the optimal revenue depends on both the valuation of the highly valued chapter and the number of chapters with positive valuation for the consumer. Aggregated across consumers, the number of chapters bought and the marginal revenue from chapters are expressed below: (3)
I I I ? N Ci N Ci ? ? ? = ? = ? N P Q N P ? ? ? Ci C? Ci C ? Ci ? VHi ? i =1 i =1 VHi i =1 ? i =1 I
(4)
? i =1 MRCi = ?
I
? ?? ? QCi ? PC ?
?
??
I
?
? ?=
I
? ?? ? ? N Ci ? PC ?
?? i =1
i =1
??
I
I
N Ci VHi
? PC
? PC
? ? ? ? PC ? I I N ? ? = ? N Ci ? 2 PC ? Ci = 0 i =1 i =1 V Hi
Let N = ? N Ci k F and S = ?
i =1
i =1
N Ci k F , where k is the number of potential VHi
consumers on the market and F is the fraction of consumers for which VHi ? PC . Then
N is the average demand for chapters at zero price for the customers who buy at least one
chapter, and S is the average slope of the demand for the customers who buy at least one chapter. Thus, the optimal price per chapter is PC* =
1 N S and the optimal revenue 2
51
* is RC =
1 (N )2 S . Thus, the higher the average demand for chapters at zero price for the 4
customers who buy at least one chapter, and the lower the average slope of the demand for the customers who buy at least one chapter, the higher the optimal price and revenue. This means that companies can charge a relatively higher price per chapter when consumers value positively relatively more chapters, and when the decrease in valuation from the most preferred to the least preferred chapter is relatively slow.
Revenue from selling electronic chapters and the print book Here we add the revenue from selling the print book and the associated costs. The price of the print book is assumed exogenous to the model as it was set when the book was initially introduced to the market. At present the company is deciding on whether to offer the electronic chapters in addition to the print book and at what price. Consumers do not receive an extra discount for buying the print book plus electronic chapters bundle and the bundle price is simply the sum of the two individual prices, PP and QC PC respectively. Consumers can buy the print book, one or more electronic chapters, or the print book plus one or more chapters. The profit function under content unbundled mixed bundling can be expressed as follows: (5)
I ?? I N Ci CUMB N P ? CUMB = PP Q P + ?? ? C? ? ? Ci i =1 V Hi ?? i =1
? ? CUMB ? ? FC , ? PC ? ? cQ P ? ?
CUMB where QP is the number of print books sold under CUMB, c is the print book variable
cost and FC is the fixed cost. We assume that the variable cost of electronic chapters is equal to zero. Differentiating the profit with respect to the chapter price, the optimal chapter price and profit are as follows: 52
(6)
CUMB I I ? ? CUMB N ? QP = ( PP ? c) + ? N Ci ? 2 PC ? Ci = 0 ? PC ? PC i =1 i =1 V Hi
(7)
* = PC
1 2 kFS
CUMB ? ? ? QP + ? ( ) k F N P c ? ? P ? PC ? ?
(8)
?
* CUMB
CUMB ? ? QP 1 ?? I ? ? = ??? N Ci ? ? ?( PP ? c) ? 4 ?? i =1 ? PC ? ? ? ? 2
2
? ? ? ?
?V
i =1
I
N Ci
Hi
CUMB + ( PP ? c) QP ? FC
CUMB 2 ? ? ? ? QP 1 ? 2 CUMB ?(k F N ) ? ?( PP ? c) = ? FC ? ? + ( PP ? c) QP 4k F S ? ? PC ? ? ? ? ?
The optimal price per chapter and profit under content unbundled mixed bundling is affected by several parameters. First, the higher (lower) the average demand for chapters at zero price for the consumers who buy and the smaller (bigger) the average slope of demand for the consumers who buy, the higher (lower) the optimal price and profit. From the consumers who enter the market for chapters, if more consumers have positive valuation for the chapters in a book and if the decrease in valuation from the most preferred to the less preferred chapter is relatively small across these consumers, publishers can charge higher price per chapter and achieve higher profit level. Second, the optimal price per chapter and the optimal profit also depend on how much the demand for print books is affected by a change in the price of the electronic chapters. If an increase in the chapter price leads to an increase (decrease) in the print book sales, the optimal chapter price is higher (lower) as compared to the case when a change in the chapter price does not affect print book sales. In terms of optimal profit, it is the magnitude of the rate of change of the print book demand function given a change in the chapter price that is important, not the direction – the more a change in the chapter price
53
affects the print book sales, the lower the profit as compared to the case when a change in the chapter price does not affect print book sales.
Traditional Mixed Bundling Strategy Model basics If the electronic chapters in a book are bundled and only the full electronic book is offered, consumer i will buy the full electronic book only if the valuation of the electronic book Vi exceeds its price PEB ( Vi = ? V ji ? PEB ). The aggregate electronic
j =1 J
book revenue is equal to PEB k [1 ? F ( PEB )] , where F ( PEB ) is the fraction of consumers for whom the valuation for the electronic book is smaller than the price of the electronic book ( Vi < PEB ) and k is the number of potential customers on the market.
Revenue from selling the full electronic book In this section we express the revenue from selling the full electronic book. Selling the print book is not yet considered in the model. The optimal price for the electronic book and the total and marginal revenues are graphically presented in Figure 3. << Insert Figure 3 about here >> The valuation of the electronic book by the different potential consumers, VEB , is presented on the y-axis, and the number of potential consumers, k , is presented on the xaxis. The optimal price, PEB , is equal to the valuation of the electronic book by the marginal consumer. The quantity of electronic books bought and the marginal revenue from electronic books can be expressed as follows:
54
(9)
QEB = k F (VEB ? PEB )
(10) MREB =
? [k F (VEB ? PEB ) PEB ] ? [1 ? F (VEB < PEB )] = k [1 ? F (VEB ? PEB )]+ k PEB =0 ? PEB ? PEB
Without loss of generality, we assume that the valuations of the consumers for the full electronic book are uniformly distributed between VL and VH, and that VL= 0. Consequently, 1 ? F (VEB ? PEB ) = 1 ? F ( PEB ) = 1 ?
PEB VH ? PEB = and VH VH
? [1 ? F ( PEB )] 1 =? . Solving equation 10, the optimal price of the full electronic book ? PEB VH
k VH 1 * * is PEB = . Thus, the higher the upper bound = V H and the optimal revenue is REB 4 2 of the electronic book valuation in the population, the higher the optimal price and revenue. The revenue also depends on the number of potential consumers.
Revenue from selling the full electronic book and the print book Again the price of the print book is assumed exogenous to the model as it was set when the book was initially introduced to the market. At present the company is considering whether to offer the electronic book in addition to the print book and at what price. Consumers do not receive extra discount for buying the print book plus electronic book bundle and the bundle price is just the sum of the two individual prices, PP and PEB respectively. Consumers can buy the print book, the electronic book or the bundle of the two books. The profit function under traditional mixed bundling can be expressed as follows: (11)
TMB TMB ? TMB = PP QP + k [1 ? F ( PEB )] PEB ? c QP ? FC ,
55
TMB is the number of print books sold under TMB, c is the print book variable where QP
cost and FC is the fixed cost. We assume that the variable cost of the electronic book is equal to zero. Differentiating the profit with respect to the electronic book price, the optimal electronic book price and profit are as follows: (12)
TMB ? ? TMB ? QP ? [1 ? F ( PEB )] = ( PP ? c) + k [1 ? F ( PEB )]+ k PEB =0 ? PEB ? PEB ? PEB
(13)
* = PEB
VH 2k
TMB TMB ? ? VH VH ? QP ? QP + ? ( ) = + ( ? ) k P c P c ? ? P P ? PEB ? 2 2 k ? PEB ?
(14)
?
* TMB
TMB 2 ? ? VH ? 2 ? ? QP TMB ?k ? ?( PP ? c) = ? ? + ( PP ? c) QP ? FC = 4k ? ? PEB ? ? ? ? ?
2
TMB ? ? QP k VH VH ? TMB = ? ( ? ) P c ? P ? + ( PP ? c) Q P ? FC 4 4k ? ? PEB ?
Similar to the content unbundled mixed bundling case, the optimal electronic book price and profit under traditional mixed bundling is affected by several parameters. First, the higher (lower) the upper bound of the electronic book valuations distribution, the higher (lower) the optimal price and profit. If consumers have relatively high valuation for the electronic book, publishers may charge higher electronic book price, resulting in higher profit. Second, the optimal electronic book price and the optimal profit also depend on how much the demand for print books is affected by a change in the electronic book price. If an increase in the electronic book price leads to an increase (decrease) in the print book sales, the optimal electronic book price is higher (lower) as compared to the case when a change in the electronic book price does not affect print book sales. In terms of optimal profit, similar to the optimal profit under CUMB, it is the
56
magnitude of the rate of change of the print book demand function given a change in the electronic book price that is important, not the direction – the more a change in the electronic book price affects the print book sales, the lower the profit as compared to the case when a change in the electronic book price does not affect print book sales.
Comparing TMB with CUMB
In this section we compare the optimal profit under the two bundling strategies. The fixed costs and the variable costs per print book are assumed to be equal across the two conditions. After subtracting (12) from (8) and doing some regrouping we get: (15)
?
* CUMB
??
* TMB
CUMB 2 ? ? ? ? QP 1 ? 2 CUMB ?(k F N ) ? ?( PP ? c) = ? FC ? ? + ( PP ? c) QP 4k F S ? ? PC ? ? ? ? ?
TMB 2 ? ? VH ? 2 ? ? QP TMB ?k ? ?( PP ? c) ? ? ? ? ( PP ? c) QP + FC = 4k ? ? PEB ? ? ? ? ? CUMB 2 ? ? ? ? QP 1 ? 2 ?(k F N ) ? ?( PP ? c) )+ ? ? 4k F S ? ? PC ? ? ? ? ?
= ( PP ? c) ( Q
CUMB P
?Q
TMB P
TMB 2 ? ? VH ? 2 ? ? QP ?k ? ?( PP ? c) ? ? ? 4k ? ? PEB ? ? ? ? ?
* ?* CUMB > ? TMB when:
(16)
CUMB TMB ? QP ? QP ? ? PP ? c ?
? 1 ? ? + 4k F S ?
CUMB 2 ? ? ? ? ? QP VH 2 ?(k F N ) ? ?( PP ? c) ? ?? ? PC ? ? 4 k ? ? ? ?
?? k ? 2 ? ? Q TMB P ?? ? ?? ? P ?c? ? ?P ? EB ? ? ?? P
? ? ? ?
2
? ?>0 ? ?
57
When CUMB is more profitable than TMB? This section outlines the conditions in which Content Unbundled Mixed Bundling is more profitable than the Tradition Mixed Bundling strategy. We look at three expressions in order to predict whether CUMB is more profitable than TMB (17a – 17c): (17a)
CUMB TMB CUMB TMB QP ? QP > 0 which is true when QP > QP
CUMB TMB QP and QP are the print sales under CUMB and TMB, respectively. These are
purchases by customers who have a strong preference for print; that is, customers who receive a higher value from the print form as compared to the value of the electronic form (accounting for their prices).
CUMB CUMB ??kFN kFN ? ? ? k F N ? ? ? QP ? QP (17b) ? ?. ? P ?c ; P ?c? ? P ?c? ? ?? ? ?P ? ? > 0 which is true when ? P ? ? C P ? ? P C ? P ? ? ?
2 2
CUMB ? QP is the rate of change of the print book demand given a change in the electronic ? PC
chapter price. We consider whether the absolute value of the rate of change is bigger or smaller than the average demand for chapters at zero price divided by the contribution margin of the print book. (17c)
TMB ? k ? ? ? QP ? ? ? ? ?P ?c? ? ?P EB ? P ? ? 2
TMB ? ?k ? QP k ? ? ? ? ? ; ? which is true when > 0 ? ? ? ? P P ? c P ? c EB P P ? ? ?
2
TMB ? QP is the rate of change of the print book demand given a change in the electronic ? PEB
book price. We look at whether the magnitude of the rate of change is bigger or smaller than the number of potential consumers on the market divided by the contribution margin of the print book. << Insert Table 5 about here >> 58
In Table 5 we present a comparison of the profits under content unbundled and traditional mixed bundling under all possible combinations of conditions 17a, 17b and 17c. For example, if print book sales under CUMB (vs. TMB) are higher, the rate of change of the print books sales given a change in the electronic chapter price is relatively small, and the rate of change of the print books sales given a change in the electronic book price is relatively high (line b, Table 5), CUMB is always more profitable. We expect the two rates of change of print book sales given changes in the PDF chapters price and electronic book price to be relatively small and equations 17b and 17c to be positive, and consequently - to fall within profit comparisons d or h (lines d and h, Table 5). Thus, CUMB will be more profitable than TMB if profit comparison equation d or profit comparison equation h in Table 5 is true. In sum, we predict that CUMB will be more profitable than TMB when equation 16 has a positive sign. The conditions in which equation 16 has a positive sign are outlined in Table 5. Overall, which strategy is more profitable depends on the extent to which it differentially affects the print book sales, combined with the extent to which it can extract higher profit given the optimal electronic product price that minimizes print sales cannibalization to electronic sales. If the publisher markets more than one book, or more than one item in general, the individual item profit functions can be aggregated across items and then compared on an aggregate level. This will not change the nature and the predictions of the model.
Data and Estimation Procedure Data
59
To empirically test our predictions, we use data provided by NAP. These are actual purchases in an online experiment involving intercepting customers on the publisher’s website. Although NAP had been offering free browsing of its books for years, customers could buy only print (paperback or hardcover) books before the experiment. During the experiment, customers who already had a print copy of a title (for which a companion PDF version was available) in their shopping carts, were intercepted as they clicked on the check-out button and assigned to one of two conditions. About 500 titles were used in both conditions. Experimental Condition 1: Content Unbundled Mixed Bundling Strategy Details of the PDF book including its price per chapter (in dollars) were presented to the consumers, and they had the option to check out a sample PDF before making a decision. In this condition respondents could buy one or more PDF chapters, the print book, or a combination of the two forms. Prices of PDF chapters for the participating titles were set at different levels relative to the pro-rated price of the print books (at 50%, 75%, 100% and 125%). For example, a print book containing 10 chapters priced at $50 would have a pro-rated price per chapter equal to $5. Then, the PDF chapter price could be set at 50%, 75%, 100% or 125% of the pro-rated price per chapter. The books were randomly assigned to the electronic price conditions; for example, all print books priced at $50 were randomly assigned to the four electronic price conditions. The reason for assigning books and not customers to the electronic chapters price conditions was to treat all the participants equally and avoid situations in which the same book would be offered as PDF chapters at different price to different consumers. We have data on 3256 customers who participated in the experiment.
60
Experimental Condition 2: Traditional Mixed Bundling Strategy Details of the PDF book including its price (in dollars) were presented to the consumers, and they had the option to check out a sample PDF as well. Participants could buy the PDF book, the print book, or the bundle of the two. The prices of the PDF books were set at different levels relative to the price of the print books (at 25%, 50%, 75%, 100%, and 110%). As in condition 1, books and not consumers were randomly assigned to the electronic book price levels. The dataset contains 950 customers.
Estimation Procedure
In this section we discuss the estimation procedure we employ to estimate the parameters of interest. We use the estimated coefficients to perform a simulation and investigate what will the probability of buying each alternative be when the PDF price as percentage of print varies. The probabilities are the predicted market shares of the various alternatives under different PDF pricing scenarios, which allows us to compute revenues and profits. We use a discrete choice framework to empirically investigate consumer choice. Specifically, we fit Random Parameters Logit Model (RPLM) to the experimental data (separately for the two conditions) and estimate the parameters of interest. This model is appealing for several reasons. First, the logit models in general are conceptually attractive as they are grounded in economic theory and have excellent empirical performance (Guadagni and Little 1983). Second, the RPLM accounts for heterogeneity across consumers in both brand preferences and responses to marketing variables (Jain, Vilcassim and Chintagunta 1994). Finally, it is empirically tractable and can be computed
61
using readily available computer software (e.g. NLOGIT 3.0 version of LIMDEP). The utility function of consumer i for alternative j in the choice set is equal to: (18)
U ji = ?1 ji CAT1 + ... + ? 5 ji CAT5 + ? 6 ji PRC PRINT + ? 7 ji PCTPDF + ? ji ,
where CAT1 to CAT5 are dummy variables for the five subject categories containing more than 5% of the observations in the sample, PRC PRINT is the price of the print book in dollars and PCTPDF is the price of the PDF item expressed as a percentage of the price of the print book (25% to 125%). The probability of consumer i choosing alternative j is equal to:
(19)
P ( j | vi ) =
exp(? j f ji + ? ji x ji )
? j =1 exp(? j f ji + ? ji x ji )
J
,
where j = 1...J alternatives in the choice set of consumer i (i = 1...I ) ,
? j is a vector of nonrandom (fixed) coefficients, ? ji is a coefficient vector that is randomly distributed across individuals, vi enters ? ji ,
f ji is a vector of choice varying attributes of choices, multiplied by ? j , x ji is a vector of choice varying attributes of choices, multiplied by ? ji ,
? ji is assumed to be distributed iid extreme values, and
vi is a random term with mean vector zero and covariance matrix I. We use the RPL procedure in NLOGIT 3.0 (LIMDEP) to estimate the models. The procedure works as follows: first, the mean vector and the covariance matrix of the coefficient vector ? ji are estimated on a sample of consumers drawn from the population using maximum simulated likelihood estimation; then the likelihood function of a
62
consumer’s sequence of choices is approximated by simulation given the mean vector and the covariance matrix of the coefficient vector. Consequently, the expected parameters for each individual customer are computed conditional on this customer’s observed choices. We account for heterogeneity in two ways. First, under both CUMB and TMB strategies, we estimate separate alternative specific constants for print, PDF and bundle for the consumers in the five book categories containing more than 5% of the observations in the sample – agriculture, behavioral science, education, general interest and medicine. Second, under CUMB, we use random parameters for the effect of
PCTPDF (PDF price as percentage of print price) on print and bundle sales, assuming that
they are normally distributed in the population. This allows us to get insights into the category-specific effects and control for them when estimating the price effects. We use fixed coefficients for the effect of print price on choice, and for the effect of PCTPDF on PDF sales16. Under TMB, we use fixed parameters for all print price and PCTPDF effects on choice17. Our smaller sample size is the likely reason for the non-significant standard deviation of the parameters distributions.
Simulation
Next, we use the RPL model parameters to perform a simulation in which we vary the price of the PDF books and chapters and compare the CUMB and TMB profits under various pricing scenarios. Separately for each strategy and pricing scenario, we compute
16
We estimate another model allowing these coefficients to vary across consumers as well. The model in which they are fixed fits the data better. 17 We estimate another model allowing these coefficients to vary across consumers. The model in which they are fixed fits the data better.
63
the utility and the predicted probability of the alternatives in the choice set for every consumer in our sample using equations 18 and 19. Then, using variable print book cost information provided by the publisher, the profit is computed as follows: (20) ? = (PPRINT + PBUNDLE )(PRC PRINT ? VC PRINT ) + (PPDF + PBUNDLE ) PRC PDF
Finally, we average the profits across the sample. We then compare the profits under CUMB and TMB at various price levels, and suggest optimal product line and pricing.
Results
Descriptive Statistics The actual purchase rates under the two mixed bundling strategies are presented in Table 6. Overall, under content unbundling mixed bundling 59.57% of the customers held on to their print book purchases, 10.08% switched to PDF, 4.13% bought both forms, while the rest 26.22% abandoned their carts. Under traditional mixed bundling 47.37% of the customers held on to their print book purchases, 12.32% switched to PDF, 5.89% bought both forms, while the rest 34.42% abandoned their carts. The exact breakdown under the specific price conditions is presented in Table 6 as well. Note that in the second part of the table for each condition the bundle percentages are added to the print and PDF percentages in order to see the total print and PDF sales for each condition. The number of observations for each pricing level is displayed as well. << Insert Table 6 about here >>
Estimation Results – Content Unbundled Mixed Bundling The results under Content Unbundled Mixed Bundling are presented in Table 7.
64
Consumers buying books in the behavioral science and education categories have a significantly higher preference for the print form and are more likely to buy it as compared to consumers buying books in the other categories ( ? = 0.39 and ? = 0.52 respectively, p’s < .05). They are also less likely to buy PDF chapters, although the coefficients are only marginally significant ( ? = -0.34, p = .09, and ? = -0.31, p = .18). The other investigated category specific effects are not significantly different. The print price is significantly affecting consumers’ probability of buying PDF chapters ( ? = -0.02, p < .05), suggesting that consumers are less likely to buy PDF chapters from the more expensive books (books with relatively higher print price). Finally, PCTPDF is affecting the choice probabilities in the following manner: when the PDF price increases, consumers are significantly more likely to buy print books ( ? = 0.90, p < .05), significantly less likely to buy PDF chapters ( ? = -0.55, p < .05), and marginally less likely to buy the bundle ( ? = -5.80, p = .09). The derived standard deviations of the PDF percentage parameters for the print and bundle choices are significant as well (1.75 and 3.89 respectively, p < .05), suggesting that the effects vary across consumers. The RPL model allows us to estimate individual specific PCTPDF parameters and account for heterogeneity when making predictions. << Insert Table 7 about here >>
Estimation Results – Traditional Mixed Bundling The estimation results under Traditional Mixed Bundling are presented in Table 8. Similar to the results under CUMB, consumers buying books in the behavioral science and education categories are more likely to choose the print book ( ? = 0.50, p = .08, and
65
? = 0.97, p < .05) as compared to the customers buying books in the other categories.
Consumers buying books in the behavioral science category are less likely to buy the PDF book ( ? = -2.34, p < .05) as compared to the customers buying books in the other categories. The other investigated category specific effects are not significantly different. Additionally, the print price is significantly affecting the probability of buying print ( ? = 0.01, p = .07) and the bundle ( ? = -0.04, p < .05), suggesting that consumers are more likely to buy the print book and less likely to buy the bundle regarding the relatively more expensive print books. Finally, as the PDF book price increases, the probability of buying the PDF book decreases ( ? = -1.25, p < .05). << Insert Table 8 about here >> In sum, the CUMB and TMB estimation results seem intuitive. The models that we use fit the data well and allow us to estimate parameters that can be further used to perform simulations and make predictions.
Simulation Results Using the estimated coefficients from the CUMB and TMB models, we perform a simulation by choosing different PDF price levels (PDF price as percentage of print price) and computing the individual utilities, purchase probabilities and profits using equations 18, 19 and 20, and average the profit across individuals. The simulation results for the average profit per consumer under various price levels are presented in Table 9. The simulation results suggest that in the case of Traditional Mixed Bundling the PDF book should be priced the same as the print book in order to get the highest profit ($12.74), while in the case of Content Unbundled Mixed Bundling the PDF chapters
66
should be priced at 125% of the price of the print book ($15.21). Content Unbundled Mixed Bundling seems to generate a higher profit as compared to Traditional Mixed Bundling when priced at levels above 75% of the price of the print book. Under Content Unbundled Mixed Bundling consumers are more likely to buy the print book and less likely to choose the “no choice” option, which helps generate higher profit. Although more consumers buy the bundle under TMB, this extra revenue from the current print customers do not compensate for the customers who decide not to make a purchase. We explore this issue more in discussion section. << Insert Table 9 about here >>
CUMB/TMB vs. Print Books Only Here we use as a baseline the case under traditional mixed bundling where the price of the print book is the same as the price of the PDF book. In this case consumers willing to buy the book will buy it in the form they prefer as the prices are the same. As the baseline strategy is in fact the optimal strategy under TMB, this implies that introducing PDF books with the same price as the print books will not change the overall profit for the company. Introducing PDF chapters priced at 125% of the price of the print books will result in higher profit as compared to offering only books in print form (19% increase). Consequently, under the conditions considered in our empirical investigation, offering individual chapters is more beneficial for the company then either offering print books only or print and full electronic books.
67
Relating the Empirical Results to the Predictions of the Bundling Model Here we compare the optimal conditions under CUMB and TMB with the predictions of the model. First, which strategy is more profitable depends on which one leads to higher print sales (equation 17a). The predicted print book market share under CUMB when the PDF chapters are priced at 125% of the print book price is 71%, while the predicted print book market share under TMB when the PDF book is priced at 100% of the print book price is 48.1%. Consequently, equation 17a holds. Second, which strategy is more profitable depends on whether the absolute value of the rate of change of the print book sales with respect to changes in the PDF chapter price is smaller than the average demand for chapters at zero price divided by the contribution margin of the print book (equation 17b). The rate of change is 0.90, the average contribution margin is about $18, and the average demand for chapters at zero
CUMB price is 3.591. Consequently, ? Q P ? PC ? [? k F N (PP ? c ); k F N (PP ? c )] and
equation 17b holds. Finally, which strategy is more profitable depends on whether the absolute value of the rate of change of the print book sales with respect to changes in the PDF book price is smaller than the number of potential consumers on the market divided by the contribution margin of the print book. The rate of change is equal to -0.245, and the average contribution margin is $18. As there are 1,000 potential customers on the market according to our assumption (which is an underestimation relative to the number of respondents in our experiment), equation 17c holds and
TMB ? QP ? PEB ? [? k (PP ? c ); k (PP ? c )].
Consequently, our results suggest that we have to expect the profit under CUMB
68
to be higher than the profit under TMB if
CUMB TMB ? QP ? QP ? ? PP ? c ?
? 1 ? ? + 4 kFS ?
?? k F N ? 2 ? ? Q CUMB P ?? ? P ?c? ? ?? ? ? ? ? ? PC ?? P
? ? ? ?
2
? V ?? k ? 2 ? ? Q TMB P ? ? > H ?? ?? ? ? ? 4 k ?? PP ? c ? ? ? PEB ? ? ?
? ? ? ?
2
? ? (line ? ?
d in Table 1). If there are 1,000 potential customers on the market, the first term of the above equation is equal to 12.7218. The second term of the equation is equal to 62.0919 (average print book price = $31.33). Finally, the third term is equal to 71.6020. Thus, the sum of the first two terms (74.81) is bigger than the third one and the predicted profit under CUMB is bigger (vs. TMB). This is in line with our empirical results.
Discussion
Our study highlights several important issues. First, there are significant subject category effects suggesting that consumers buying books in education and behavioral sciences have a stronger preference for print as compared to consumers buying in the other subject domains, and are less likely to either switch from print to PDF or buy the bundle. On the one hand, these consumers may not have extensive experience with electronic products, or may not use computers as often as other consumers. On the other hand, the print form may be more appropriate for their specific usage situations. Thus, publishers interested in selling electronic products to this customer segment may study why these consumers are more likely to stick with the print form and less likely to buy the PDF form.
18 19
Term 1 = [(0.71-0.481)*1,000]/18=12.72 Term 2 = [1/639.89]*[(3,591/18)*(3,591/18)-(0.9*0.9)]=62.09; S=(1/(2*31.33*1.25/14)*(3,591+18*0.9)=639.89 20 Term 3 = [(62.93/4000)*[(1000/18)*(1000/18)-(-0.234)(-0.234)]=71.60; VH=(2*31.33*1000)/[1000+18*(-.234)]=62.93
69
Second, our study reveals that for relatively more expensive books consumers are more likely to stick with the print form and less likely to switch to the PDF form or buy the bundle of the two forms. Although quite unintuitive at first, there may be good reasons for such a result. On the one hand, consumers may have relatively higher valuation for the more expensive print books because they are likely to be hardcover books with pictures and graphics, and the print form may be perceived as a better choice with regard to these attributes than the PDF form; customers may even be likely to display such books in their offices or at home (e.g., coffee table books) or have other reasons to prefer the print form. On the other hand, consumers may be less willing to pay a high price for PDF books as they know that their marginal costs are negligible. For example, keeping everything else constant, if a print book costs $25.00 and the PDF book price is set at $18.75 (75% of the price of the print book), consumers may be more likely to buy the PDF book vs. the print book as compared to the case where the print book costs $75 and the PDF book price is set at $56.25 (75% of the price of the print book). Although the price of the electronic book is the same percentage of the price of the print book in both cases, consumers are likely to be aware that the marginal costs of the electronic books are negligible and they would not expect to be charged a big dollar amount even for expensive print books. Thus, owning electronic products may be unappealing to consumers in certain subject domains, and there may be an upper bound on how much consumers are willing to pay for electronic content in terms of a dollar amount. It may be unreasonable to determine the electronic form price as a percentage of the print form price, and other pricing mechanisms may be more appropriate in some cases.
70
Third, our simulation predicts that, under traditional mixed bundling, as the price of the PDF book approaches and exceeds the price of the print book, the share of the nochoice option increases and the share of the print option does not change. This is not the case under content unbundled mixed bundling where as the chapter price increases, the share of the no-choice option decreases while the share of the print book option increases. Recent studies in the marketing domain may provide insight into this finding. Dhar (1997) argues that consumers tend to focus more on the comparative characteristics among the alternatives provided than on their own utilities. On the one hand, expanding the choice set by adding an attractive alternative increases the preference for the nochoice option when respondents can choose only one alternative (Dhar 1997). On the other hand, respondents are less likely to defer choice when both attractive alternatives could be selected (Dhar 1997). If they have to make a choice between print and full PDF books, consumers may focus on the comparative characteristics of the two forms and evaluate which form is better on various attributes such as layout, browsing, image quality, etc. Because both forms have their advantages and disadvantages, and because price is often perceived as a signal of product quality, consumers may be more likely to find the two alternatives equally attractive and defer choice when the prices are similar as compared to when the PDF price is significantly lower than the print price. When the full PDF book price is significantly lower than the print price consumers may be more likely to buy the PDF book as it becomes relatively more attractive because of its low price, or buy the cheaper bundle and benefit from the advantages of both forms. It seems that under content unbundled mixed bundling it is easier for consumers to make a choice as they are less likely to select the no-choice option. Thus, content unbundled mixed
71
bundling may be more profitable than traditional mixed bundling also because it is easier for consumers to make a choice and more consumers enter the market. Finally, there is a self-selection bias in our sample because consumers are intercepted in the experiment after they have already decided to buy the print book. Consequently, we are offering a conservative test with respect to the quantity of electronic chapters being sold at the optimal price level under CUMB, and CUMB may be even much more profitable than TMB. The reason is that consumers for whom the reservation price for the print book is lower than the price of the print book and will not buy the print book, may be attracted to buy one or several chapters. We are not able to capture these purchases in our data because we intercept consumers after they have decided to buy the print book. Our third essay accounts for the revenue from such consumers.
Managerial Implications
Out study provides valuable insights to content providers on how to successfully design and price information products in addition to their print products, assuming that the providers have reasons not to or do not want to change the price of the print products. Content unbundled mixed bundling is a viable strategy to pursue in certain market conditions as it leads to higher profit. The unbundled electronic units should be priced at a premium compared to the pro-rated price of the print book in order to achieve the optimal level of profit. The demand for electronic versions of products such as books is sizeable and it justifies the extension of the existing product lines. Although our study does not look at this issue, introducing electronic products may attract new consumers
72
and expand the market as well. It may be more beneficial for the publishers to consider different pricing schemes for the different book categories. Our results highlight that consumers buying books in certain subject domains are less likely to buy electronic products, and more likely to stick with the traditional print form. Thus, in order to penetrate these segments, content providers may focus on educating the consumers on the benefits the electronic form provides, and employ strategies that encourage first trial. Publishers may find it also interesting that the more expensive print books are less likely to be substituted with electronic books. As these are likely to be high margin books that bring substantial income, introducing electronic version of such books will not necessarily hurt this revenue stream.
Conclusion
In this paper, we model the profits under two strategies a content provider may employ: traditional mixed bundling, in which the product line consists of print book, PDF book and the bundle of the two, and content unbundled mixed bundling, which is selling print book, PDF chapters and the combination of the two. Our study suggests that offering unbundled PDF chapters is more profitable than offering only the full PDF books when the print books prices do not change when PDF content is introduced. We empirically support our predictions with actual data gathered in an online experiment executed on the website of a publisher of electronic books. We offer important insights on how customers from different subject domains react to electronic products and how they are likely to behave when different product and pricing schemes are employed.
73
Our study contributes to the literature on bundling as it offers guidelines on the optimal strategy when different mixed bundling strategies are feasible. Further, it looks at both form bundling and content unbundling, which is different than the literature so far and provides additional insights. Last but not least, our data includes purchases of books in a wide range of subject domains, thus allowing us to generalize our findings and recommendations (while controlling for some category specific effects). Our findings highlight important questions and unresolved issues that future research may address. It will be interesting to discover why consumers from different subject domains have distinctive preferences and inclination with regard to electronic products. Further, future studies may clarify why consumers are less likely to make a choice when they are offered print and electronic products with similar prices. Providing further insights may help companies successfully design and implement penetration strategies regarding electronic content, resulting in higher customer satisfaction and repeated purchases. Third, it will be interesting to investigate whether offering individual book chapters in print form is a viable strategy for a publisher to pursue. Although such a product strategy may be too expensive because of the costs associated with unbundling print content, it may be appealing under certain conditions.
74
CHAPTER IV: ESSAY THREE. BUNDLING AND UNBUNDLING OF ELECTRONIC CONTENT
Abstract
This paper investigates the attractiveness of product lines of items such as books and newspapers available in conventional and electronic media. For example, the publisher of Wall Street Journal can offer the following product line of subscription options: print WSJ, WSJ online, separate sections of WSJ online (e.g. Money & Investment, Technology), print WSJ and online WSJ bundle, and print WSJ and online section bundle. We employ a choice experiment in which a sample of consumers is presented with hypothetical product offerings at various prices and asked to make a choice. The data is used to develop a profit-maximizing configuration of products and prices. Similar approaches to the product line pricing problem have been employed for conventional products, but not when bundling of different forms of a product is an option, and not when the different products may be complements rather than substitutes.
75
Introduction
The Internet as an information channel has facilitated the development and dissemination of electronic forms of traditional print products such as books and newspapers. Electronic information products are unique in that the marginal cost of reproducing and distributing them is often much lower than the cost of the print products. Hence, content providers may be interested in selling the electronic form by itself or together with the print form to get higher revenue. Further, the electronic products can be unbundled at no extra cost, and the unbundled components can be re-bundled with the traditional print products. For instance, Wall Street Journal is currently offered as print WSJ, WSJ online, and the bundle of the two subscription options. Theoretically, the publisher can also sell individual sections of WSJ in electronic form, and even re-bundle the electronic units with the print WSJ (e.g., offer a subscription to the print WSJ and the financial section in electronic form as a package). Thus, if consumers switch from the traditional print products to the cheaper electronic products or units, the publisher may encounter product cannibalization and lost revenues. On the contrary, if consumers subscribe to the print products and electronic products/units bundles, the publisher may extract extra revenue. The focus of the present study is on product lines of items available in conventional and electronic media. Specifically, we discuss a method for determining optimal product line design and pricing for an item available in print and electronic form. In addition, we examine the effect of both price and non-price attributes on consumers’ preferences and choice decisions. In our study we use choice experiment and choicebased conjoint analysis to obtain estimates of customer valuations of the price and non-
76
price attributes of interest. We present respondents with hypothetical choice situations in which the attributes are varied to determine how choice fluctuates with changes in the levels of attributes. Then, accounting for response and preference heterogeneity, we estimate random parameters logit models and use the results to derive optimal product mix and pricing strategy. The output of the study is a general methodology for setting product lines and prices for items available in conventional and electronic media. The contribution of the study is as follows. It extends the research on bundling by examining bundling of interrelated products in the context of product forms, which is a new and managerially relevant application. There are three unique characteristic of the above problem that makes it different from what have been studied in the literature so far. First, the product forms can range from being perceived as substitutes to being perceived as complements to each other, and these contingent valuations may vary across the consumers. Second, the forms are not only interrelated but are likely to contain the same information; thus, by choosing the bundle, consumers acquire two items with the same content. Third, there are various bundles that can be offered – print newspaper and online newspaper bundle, print newspaper and online newspaper section(s) bundle, and online sections bundle. Consequently, the predictions of the literature in terms of optimal bundle composition and pricing may not be applied directly. We study how consumers perceive electronic and conventional print products, and provide insight on how demand fluctuates with changes in price and other attributes. Similar approaches to the product line pricing problem have been traditionally employed for conventional products, but not when bundling of different forms of a product is an option, and not when the different products may be complements rather than substitutes.
77
The third essay extent the second essay in several ways. First, we consider all feasible print and electronic options simultaneously, thus being able to investigate the market expansion when full electronic products and electronic units are offered. Second, we incorporate in our design different price levels for the print products and different levels of bundle discount in addition to the electronic products price levels, which results in more realistic and generalizable conclusions. Finally, the choice-based conjoint experiment we execute allows us to better control the data collection process and generates panel data with many observations per consumer to better model the consumer choice process. The paper is structured as follows. First, we review the relevant marketing and economics literature and position the study. Second, we present the theoretical foundations and the methodology of our study. Third, we discuss the experimental design and data collection procedure. Fourth, we report our empirical findings and their practical implications.
Research Background
Bundling is the strategy of marketing two or more products and/or services as a package at a special price (Guiltinan 1987). Two common approaches are pure bundling whereby only the product bundle is offered, and mixed bundling, in which the bundled items are also sold separately. Hanson and Martin (1990) provide an overview on why the sellers are motivated to bundle, discussing both demand and supply side incentives. The demand side incentives that make bundling profitable include negative correlations in reservation
78
prices (Stigler 1963; Schmalensee 1984), complementarity in consumption (Telser 1979), and uncertainty in the valuations of the quality of the goods (Kenney and Klein 1983). Mixed bundling is more beneficial than pure components or pure bundling, and it is the optimum bundling strategy when there is asymmetry in the reservation prices for the bundle components (Adams and Yallen 1976; Schmalensee 1984; McAfee, McMillan and Whinston 1989). In terms of supply side incentives for bundling, Bakos and Brynjolfsson (1999, 2000) demonstrate that large scale bundling of information goods can be very profitable because it creates economies of aggregation when their marginal costs are very low. In the marketing domain, empirical studies focus on issues such as bundling and unbundling of industrial systems (Wilson, Weiss and John 1990), determining bundle prices and composition (Hanson and Martin 1990), and optimal bundle pricing (Venkatesh and Mahajan 1993). Additionally, several analytical articles discuss contingent valuations in bundling decisions, contrasting interrelated from independently valued products in a bundle (e.g. Venkatesh and Kamakura 2003). In sum, the above-presented research identifies the conditions for profitable bundling and specifies the optimal bundling strategies in various situations. Sellers are motivated to bundle because, on the demand side, it can help them extract consumer surplus, and, on the supply side, it can increase producer surplus or lead to cost savings.
Positioning Our Study
Most of the research so far suggests that mixed bundling is the optimal solution for producers of various goods (Adams and Yallen 1976, Schmalensee 1984, McAfee,
79
McMillan and Whinston 1989, etc.). This policy enables the seller to reduce effective heterogeneity among those buyers with high reservation prices for both goods, while still selling at a high markup to those buyers willing to pay a high price for only one of the goods (Schmalensee 1984). With regard to bundling of information goods such as magazines and journals, recent articles argue that pure bundling is an optimal strategy for a monopolist marketing a large number of information goods given negligible marginal costs and bulk sales (Bakos and Brynjolfsson 1999, 2000). Geng et al. (2005) further refine their results and suggest that mixed bundling is approximately optimal if consumers’ values to subsequent goods do not decrease too quickly, otherwise pure bundling is optimal even when there are strong negative or positive correlations of values across goods. Our study contributes to this stream of research in the following manner. First, we investigate the attractiveness of various bundled and unbundled forms of print and electronic content, which is an important area with significant managerial implications. Second, we address the problem both from a seller’s and a buyer’s point of view. We not only study how consumers are likely to behave when offered bundled/unbundled content in different forms, but also use a methodology that companies may apply in designing their product lines and pricing them optimally after accounting for consumer preferences. Third, the application in the book and newspaper categories that we discuss can be easily extended to other categories such as magazines and music. Finally, our results have interesting implications in the area of personalization of product offering and customization in the digital goods domain. There are three recent papers that focus on issues similar to the ones in our study.
80
Next we briefly outline these studies in order to provide comparisons with ours. First, Venkatesh and Kamakura (2003) model the optimal bundling and pricing strategies for interrelated products under monopoly. Our study is different from theirs because we investigate optimal bundling strategies in the case of different forms of the same product and various combinations of these forms, while they consider bundling of distinctive products that are interrelated (e.g. computer and printer). Moreover, we use choice-based conjoint analysis with stated choices to estimate the parameters of interest and further include them in our optimization problem while Venkatesh and Kamakura (2003) rely on simulation-based analytical solution. In the second study, in a context similar to ours, Venkatesh and Chatterjee (2003) focus on bundling of print and electronic magazines, including separate sections of these magazines. Our study complements their analytical findings by including offering more options to the consumers. While they consider print magazine, online magazine and separate modules of the online magazine and all possible combinations of these in the product line but not bundling them (for example, they do not include a bundle of print magazine with an online magazine at a price lower than the sum of the component prices), we allow bundling of the print item with the online item, and bundling of the print item with the online modules at a price lower than the sum of the prices of the bundle components. The third paper by Jedidi et al. (2003) suggests a model for capturing continuous heterogeneity in the joint distribution of reservation prices for products and bundles. Their results suggest that the product line pricing policy depends on the degree of heterogeneity in the reservation prices. Our study further extends theirs by considering non-price attributes in addition to product and bundle price. Additionally, Jedidi et al (2003) study bundles of conventional products such as the Times Magazine
81
and investigate the optimal price for the magazines and the bundle of the two, while we study the same information content under different product forms and include all feasible bundled, unbundled and re-bundled options in the choice set. In sum, our paper extends the bundling research by investigating bundling of different product forms containing the same information. Additionally, we provide optimal product line design and pricing strategies while incorporating consumer preference for bundled and individual print and electronic forms of a product, and after adding different combinations of re-bundled electronic units with print products to the choice set.
Theoretical Background
Next we discuss our expectations about the attractiveness of bundling of electronic content with traditional print content, and of unbundling of electronic content and/or re-bundling it with traditional print content.
Product Form Complementarity in Choice Complementarity has been traditionally characterized by referring to the sign of the cross-price elasticity of demand - if the sign is positive (negative), products are substitutes (complements). In marketing, complementarity is often defined relative to product-specific utilities and the corresponding consumer needs (Chernev 2005). Products are substitutes if both can satisfy the same need to the consumer, they are complements if they are consumed jointly (but not necessarily simultaneously) in order to satisfy a need (Lattin and McAlister 1985; Henderson and Quandt 1958). Previous
82
research suggests that complementarity between products can cause bundling to be profitable (Telser 1979). Bundles composed of complements have higher purchase intent versus bundles of similar or unrelated products (Harlam et al. 1995), and complementarity positively affects bundle reservation prices (Gaeth et. al 1990). From consumers’ point of view, if the product forms are substitutes, the purchase of one form lowers the value of the alternative form, and therefore makes bundling the items less attractive; if the forms are complements, bundling becomes more attractive. Consumers may perceive a high degree of complementarity based on search economies, improved satisfaction because of the bundle, and/or improved total image of the bundle (Guiltinan 1987; Simonin and Ruth 1995). By and large, we expect a positive relationship between the degree of perceived complementarity between the bundle components and the purchase likelihood for the bundle. On one hand, if consumers believe that the individual product forms can be used interchangeably, they will buy only one of the forms and not the product form bundle. On the other hand, if consumers perceive additional utility in having both product forms versus just one of them, they will choose the product form bundle.
Attractiveness of Bundling of Electronic Content with Traditional Print Content Analytical results suggest that in general moderate or strong substitutes should be offered separately; the same is applicable for complements if the marginal costs are moderate relative to the market’s maximum willing to pay (Venkatesh and Kamakura 2003). Also, a seller gains by mixed bundling for weak substitutes/complements if the variable costs are not too high (Venkatesh and Kamakura 2003). Finally, although the
83
product line pricing policy depends on the degree of heterogeneity in the reservation prices for the items, firms would benefit from using a mixed bundling strategy (Jedidi et al. 2003). Building upon these findings, we would expect mixed bundling strategy be the optimal solution. This statement is supported by both analytical solutions and related empirical results. However, there are three unique features that characterize our problem: consumers are likely to be heterogeneous in the way they see the different product forms on the substitute-complement continuum, the marginal costs of the electronic products are almost negligible, and the market’s willingness to pay for some electronic products may be significantly lower than the one for conventional print products. Consequently, our results may suggest a different optimal solution. In terms of optimal prices, Jedidi et al. (2003) advocate that a uniformly high price strategy for all products and bundles is optimal when the heterogeneity in the reservation prices is high; otherwise, a hybrid strategy is optimal. Thus, it is an empirical question on how to price the same print and electronic content as individual products and as a bundle. In sum, the findings of the extant literature do not offer guidelines on to how to design and optimally price product lines consisting of different forms of the same product. Our study aims at providing insights into this question with important implications for practitioners.
Attractiveness of Unbundling of Electronic Content and Re-Bundling with Print Content In a different context (industrial systems), Wilson, Weiss and John (1990) identify the following cases favoring unbundling: larger unit margins from unbundling, market growth from unbundling, new market segments from unbundling, and inferior but cheaper systems from unbundling. With respect to unbundling of electronic content (e.g.
84
offering individual sections of WSJ online), it would be an appealing strategy if the companies could attract new customers who are more price-sensitive and otherwise would not buy the full product, thus leading to market growth. If unbundling causes cannibalization of print sales to unbundled content, the unbundling strategy could result in lower profits. Consequently, attracting new market segments and market growth would favor unbundling of electronic content. In terms of pricing of unbundled units, contrary to conventional wisdom on mixed bundling, Venkatesh and Chatterjee (2003) claim that low-priced components (e.g. individual sections) should be targeted at consumers who have low value for the information content. Thus, it might be optimal to pursue such a pricing strategy: on one hand, trying to attract new customers by offering unbundled content at a low price, and on the other hand, extracting extra profits by offering full print content plus individual sections in electronic form to the consumers who see the different forms as complementing each other.
Product Form Attractiveness in Choice Dhar (1997) argues that consumers tend to focus more on the comparative characteristics among the alternatives provided than on their own utilities. When the choice task is to choose one alternative from a choice set, consumers are more likely to defer choice when the alternatives are equally attractive. When the task allows the choice of more than one alternative from a choice set of two attractive alternatives, the preference for a no-choice option decreases. Thus, if the consumers perceive one product form better then the other on most of the attributes (e.g. form 1 is superior to form 2), we would expect them to be more likely to choose the superior form. On the other hand, if the consumers perceive form 1 better then form 2 on some of the attributes, while form 2 85
better than form 1 on the rest of the attributes (form 1 and form 2 are equally attractive), we would expect them to be more likely to defer choice when they have to select just one alternative. Finally, if they are given the option to choose more than one form (e.g. buy a bundle), consumers may decide to buy both form if they are equally attractive. Thus, we would expect the relative attractiveness of the product forms to influence choice in the following manner: on one hand, the more differentially attractive the two forms are, the higher the probability to defer choice; on the other hand, the higher the difference between the attractiveness of the superior product form and the bundle, the more likely consumers will be to select the bundle.
Methodology
We use a choice-based conjoint framework to study our questions of interest. Since the early 1970s, conjoint analysis has received considerable academic and industry attention (Green and Srinivasan, 1990). Conjoint analysis and the related technique of experimental choice analysis represent the most widely applied methodologies for measuring and analyzing consumer preferences (Carroll and Green 1995). The choicebased conjoint is a relatively new type of conjoint analysis and is considered to better approximate actual decision processes as compared to the traditional ratings or rankingsbased conjoint analyses21. Consequently, more realistic aggregate level estimates are expected. In the next sections we describe in detail our experimental design, the data collection procedure and our modeling approach.
For an excellent review and empirical comparison between ratings-based and choice-based conjoint models see Elrod, Louviere and Davey (1992).
21
86
Study Design and Data Collection
We use two product categories in our study – a book (“Vault Guide – insider information on industries, careers, and employers”) and a newspaper (Wall Street Journal). We have selected these two products because they are very popular among our sample, MBA students, and our respondents are well-experienced with them. Additionally, current MBA students are an important target audience for both publications and this increases the external validity and applicability of our results. In the conjoint experiment the respondents evaluated various scenarios for each product and made a choice. We vary the following factors in the scenarios: print (absent, present), full electronic (absent, present), electronic unit/section (absent, present), bundle (absent, present) and price level (low, medium, high). The conjoint factors and the price levels are presented in Table 10, panels A and B. We took the actual print and electronic/PDF form prices of the Vault Guide and WSJ for the medium price level; the electronic unit price was computed by dividing the electronic/PDF price by three22 and then adding a 15% premium. The price of the print product and electronic product bundle is the sum of the two individual product prices discounted 15%, while the price of the print product and electronic section bundle is the sum of the two individual product prices discounted 7%23. Finally, for all product offers, the low and the high price levels represent a 15% discount/premium compared to the medium price level. We chose a 15% gap between the different pricing levels to assure that the respondents are able to register such price differences while keeping the face validity of the prices of the various items.
22 23
There are three sections in the product. We decreased the bundle price by only 7% in this case to preserve the face validity of our prices and avoid situations in which the print product costs much more than the print product/electronic section bundle (but still employ three different price levels).
87
<< Insert Table 10 about here >> For each of the two product categories the participants were presented with a choice situation (e.g. that they are considering subscribing to WSJ) and asked to consider sixteen independent choice scenarios with three options in each (e.g. Option 1: Subscribe to the Online WSJ for $39.50; Option 2: Subscribe to the Print WSJ for $114.50; Option 3: Will not subscribe to either of the two). The options in each scenario were randomly selected from a full factorial design in which product offer and price were varied. Participants were asked to select the option that they were most likely to pursue in such a situation. After evaluating all the scenarios for the first category, the same was repeated in the second product category. At the end the participants answered question related to the relative attractiveness of the product forms on five attributes (image quality, browsing, layout, convenience of use and archival quality) and demographics. Eighty seven full and part time MBA students (60% part time students; 37% female) enrolled in graduate level marketing courses participated in our study. Each participant evaluated sixteen choice scenarios in the book category and sixteen choice scenarios in the newspaper category (the order was counterbalanced).
Modeling Consumer Preferences We use a discrete choice framework to model consumer choice. Specifically, we fit Random Parameters Logit Model (RPLM) to the conjoint data and to estimate the parameters. This model is appealing for several reasons. First, the logit models in general are conceptually attractive as they are grounded in economic theory and have excellent empirical performance (Guadagni and Little 1983). Second, when desired, the RPLM can
88
account for heterogeneity across consumers in both brand preferences and responses to marketing variables (Jain, Vilcassim and Chintagunta 1994). Finally, it is empirically tractable and can be computed using readily available computer software (e.g. NLOGIT 3.0 version of LIMDEP). Accounting for heterogeneity across consumers is an important consideration when analyzing consumer purchase behavior from panel data - consumers with the same demographic and socioeconomic characteristics (e.g. income, family size) when confronted with a given set of covariates (e.g. price, feature advertisements) may exhibit different choice behavior due to differences in overall brand preferences (intercept/preference heterogeneity) and/or variations in their responses to these covariates (slope/response heterogeneity; Jain, Vilcassim and Chintagunta 1994). The above argument applies to our data as well. For example, we may observe different choice behavior by consumers with the same age, education and income level because of differences in overall preference for print vs. electronic content, and because of variations in their responses to marketing mix variables such as price. Thus, the RPLM allows us to control for the unobserved heterogeneity among individual consumers when estimating various parameters and using them to make predictions. The RPLM is a one level multinomial logit for individuals i = 1, …, N. P ( j | vi ) = exp(? j f ji + ? ji x ji ) ,
?m=1 exp(?m f mi + ? mi xmi )
J
where U ( j, i) = ? j f ji + ? ji x ji + ? ji , j = 1, …, J alternatives in the i’s choice set,
? j is a vector of nonrandom (fixed) coefficients, ? ji is a coefficient vector that is randomly distributed across individuals, vi enters ? ji ,
89
f ji is a vector of choice varying attributes of choices, multiplied by ? j , x ji is a vector of choice varying attributes of choices, multiplied by ? ji ,
? ji is assumed to be distributed iid extreme values,
vi is a random term with mean vector zero and covariance matrix I. We use the RPL procedure in NLOGIT in LIMDEP to estimate the models. The procedure works as follows: first, the mean vector and the covariance matrix of the coefficient vector ? ji are estimated on a sample of consumers drawn from the population using maximum simulated likelihood estimation; then the likelihood function of a consumer’s sequence of choices is approximated by simulation given the mean vector and the covariance matrix of the coefficient vector. Consequently, the expected parameters for each individual customer are computed conditional on this customer’s observed choices.
Results
Conjoint Experiment << Insert Table 11 about here >> The estimation results are presented in Table 11. We include as attributes dummy variables for the following conjoint factors (1-present, 0-absent): print, electronic full, electronic unit and bundle; the price variable contains the actual prices the respondents evaluated with each product offer. We estimate a Random Parameters Logit model for each product category in which the preference parameters are assumed to be normally distributed across the sample. The results are consistent across the two categories: consumers value very positively the print attribute (the part worth is 8.485 in the book
90
category and 11.333 in the newspaper category) and the full electronic attribute (7.306 and 6.793 respectively), while the part worth of the electronic unit attribute is relatively smaller (3.063 and 1.452 respectively). The bundle attribute part worth is negative and relatively high in both categories (-6.647 and -3.384 respectively), and the price parameter is negative (-0.261 and -0.133)24. Interestingly, the derived standard deviations of the means of all five parameters are significant in both categories (varying from 0.107 and 0.025 for the means of the price parameters to 5.249 and 6.222 for the means of the print parameters). This suggests that that there is significant heterogeneity in how much people value the different product forms and their combinations, and how much price changes affect their choices. In terms of the price response heterogeneity, at least 95% of the consumers have negative responses to a price increase in the book category (their price parameters fall within two standard deviations from the parameter means and are below zero) and at least 99% of the consumers have negative responses to a price increase in the newspaper category (their price parameters fall within three standard deviations from the parameter means and are below zero). The part worth of the print attribute is positive for 68% to 95% of the consumers (their print parameters fall within one to two standard deviations from the parameter mean), and the part worth of the full electronic attribute is positive for 95% to 99% of the consumers (their full electronic parameters fall within two to three standard deviations from the mean). About 68% to 95% of the consumers have positive electronic unit part worth in the book category while the same is true for less than 68% of the consumers in the newspaper category. Finally, with respect to how much consumers value the bundle factor, the situation is very
We also estimate another model in each category in which the price coefficients are not constrained to be the same across the alternatives. The model we report fits the data better.
24
91
different across the two categories: the bundle part worth is negative for more than 99% of the respondents in the book category, while the same is true for less than 68% of the respondents in the newspaper category. Note that we can not directly compare the preference parameters across the two product categories because of the scaling of the parameters in the estimated models. We can compute what price makes consumers indifferent between buying one product offer or the other. For example, in the book category, the average consumer will be indifferent between buying the print book and the full electronic book when the print book price is $29.00 and the electronic book price is $24.36 [exp(8.485–0.261*29.00) = exp(7.3060.261*24.36)]. Also, he/she will be indifferent between buying the print book and the print book plus electronic book bundle when the print book price is $29.00 and the bundle price is $31.49 [exp(8.485–0.261*29.00) = exp(8.485+7.306-6.6470.261*31.49)]. In the newspaper category, the average consumer will be indifferent between subscribing to the print WSJ and the online WSJ when the print WSJ price is $99.50 and the online WSJ price is $65.67 [exp(11.333–0.133*99.50) = exp(6.7930.133*65.67)]. Also, he/she will be indifferent between subscribing to the print WSJ and the print WSJ plus online WSJ bundle when the print WSJ price is $99.50 and the bundle price is $125.14 [exp(11.333–0.133*99.50) = exp(11.333+6.793-3.384-0.133*125.14)]. In sum, the conjoint experiment results are very consistent across the two investigated product categories. Because there is significant attribute preference and price response heterogeneity in our sample, the random parameters logit model allows us to estimate the part worth of the conjoint factors and their distributions. Next we use the above results to evaluate all possible market scenarios (combinations of different product
92
offers at different price levels), compute the market share of the various product offers and compare overall profits.
Market Shares and Profits In this section we present the market shares and profit under various market scenarios. We first look at offering a complete product line (print product, electronic product, electronic unit, print product and electronic product bundle, print product and electronic unit bundle), and then compare the results with the case where incomplete product line is offered (print product, electronic product, print product and electronic product bundle) with or without bundle discount. The goal is to empirically compare the various approaches a company may pursue and discover the most profitable product line configuration and pricing strategy.
Complete product line If the content provider decides to go with all available product offers, we need to compare 243 (five product offers with three price levels each) market scenarios. We use the estimated individual level part worths for the five conjoint factors to compute the utility of the available profiles for every consumer under each market scenario. Then we compute the individual probability of choosing each product profile and average the results across consumers for every market scenario. Finally we calculate the total revenue, costs and profit for each of the 243 market scenarios. We assume that the variable costs for the electronic products are equal to zero, and we consider different levels of print variable costs (from print variable cost equal to 40% of the medium price
93
level of the print products to print variable cost equal to 90% of the medium price level of the print products). We use the medium price level as a base as this is the actual market price the publishers of WSJ and Vault Reports are charging. We do not consider fixed costs in our analysis as they just shift the profit function and do not change its slope. The most profitable market scenarios in case of complete product line under the different variable cost levels are presented in Tables 12 and 13. << Insert Table 12 about here >> The book category market simulation results are displayed in Table 12. The top three market scenarios with respect to overall profit are listed within each considered cost level (print variable cost is 40%, 50%, 60%, 70%, 80% and 90% of print price). Together with the three most profitable scenarios, we also present the top two market scenarios in which the print book price is at its medium price level (which is the current price the publisher of Vault Report is charging for the book). For example, when the variable cost of the print book is assumed to be 40% of the price of the print book (first five lines in Table 4), our results show that the most profitable market scenarios is #181 (market shares is brackets): print book at high price (31.1%), electronic book at low price (23.4%), electronic unit at high price (14.1%), print book and electronic book bundle at low price (9.4%), print book and electronic unit bundle at low price (3.9%), and no choice option (18.1%); the total revenue is equal to $20.94 (calculated if there is only one consumer on the market), cost is $5.15, and the profit is $15.79. From the scenarios in which the print book is at medium price, the market scenario that brings the highest profit is #100 (it is ranked number 12 within all market scenarios): print book at medium price (42.0%), electronic book at low price (20.1%), electronic unit at high price (12.5%), print
94
book and electronic book bundle at low price (6.5%), print book and electronic unit bundle at low price (2.5%), and no choice option (16.3%); the total revenue is equal to $20.70 (assuming that the total market consists of one consumer), cost is $5.93, and the profit is $14.77. When comparing the top three most profitable market scenarios and the top two most profitable market scenarios in which the print book is at medium price across the considered print variable cost levels, we notice that the results are quite consistent – market scenarios #181, #182 and #172 are always the top three overall, and market scenarios #100 and #91 are always the top two in which print is at medium price. Thus, in the book category, the publisher generates the highest profit when the print book and the electronic unit are offered at high price, and the electronic book and the two bundles are offered at low price. The reason is that, although the publisher is forgoing print book revenue when the print book price is high, the company is compensating this loss with the increased revenues from the electronic book and the two bundle options which are priced at a low level and attract more buyers. As the electronic products have zero variable costs, this scenario results in lower overall revenue but higher profit as compared to the scenario when the print book price is at medium. The higher electronic unit price allows the company to price discriminate against consumers who value only a portion of the print book, and get the maximum revenue from this segment. << Insert Table 13 about here >> The newspaper category market simulation results are displayed in Table 13. As before, we list the top three market scenarios with respect to overall profit within each considered cost level, and the top two market scenarios in which the print newspaper
95
price is at its medium price level (which is the current price the publisher of Wall Street Journal is charging). For example, when the variable cost of the print newspaper is assumed to be 40% of the price of the print newspaper (first five lines of Table 13), we can see that the most profitable market scenarios is again #181 (market shares is brackets): print newspaper at high price (2.5%), electronic newspaper at low price (51.7%), electronic unit at high price (10.3%), print newspaper and electronic newspaper bundle at low price (25.9%), print newspaper and electronic unit bundle at low price (2.1%), and no choice option (7.5%); the total revenue is equal to $51.26 (assuming that the total market consists of one consumer), cost is $12.14, and the profit is $39.12. Regarding the scenarios in which the print newspaper is at medium price, the market scenario that brings the highest profit is also #100 (it is ranked number 4 within all market scenarios): print newspaper at medium price (5.5%), electronic newspaper at low price (51.3%), electronic unit at high price (10.1%), print newspaper and electronic newspaper bundle at low price (24.2%), print newspaper and electronic unit bundle at low price (1.7%), and no choice option (7.1%); the total revenue is equal to $51.21 (assuming that the total market consists of one consumer), the cost is $12.53, and the profit is $38.68. Comparing the most profitable market scenarios overall and most profitable market scenarios in which the print newspaper is at medium price across the considered print variable cost levels, the results are generally consistent – market scenarios #181, #184, #187, #172, #178 and #208 are among the top overall, and market scenarios #100 and #91 are the top two in which print is at medium price. Thus, in the newspaper category, the publisher generates the highest profit when the print newspaper and the
96
electronic unit are offered at high price, the electronic newspaper is offered at low price, and the two bundles are offered at low, medium or high price (depending on the cost structure). As in the book category, although the publisher is forgoing print newspaper revenue because of its high price, the company is compensating this with the increased revenues from the two bundle options which are priced low and attract more buyers. Because the electronic products have zero variable cost, this scenario results in lower overall revenue but higher profit as compared to the scenario when the print newspaper price is medium. The higher electronic unit price enables the company to price discriminate against consumers who value only a portion of the print newspaper.
Incomplete product line (traditional mixed bundling with bundle discount) In this case the publisher offers the print product, the electronic product and the bundle of the two. As each of the three product offers has three price levels, we compare 27 possible market scenarios. The results are presented in Table 14 (book category) and Table 15 (newspaper category). Again the results are consistent across the two product categories – the publisher generates the highest profit when the print product is offered at high price, the electronic product at low price, and the bundle is at low price. In most of the cases one of the top three profiles in terms of overall profit is the one in which the print product is at medium price, and the electronic product and the bundle are at low price, which is very similar to what the publisher of WSJ is pursuing currently (print at medium price, online at low price and bundle at medium price, as in market scenario #85). The publisher of Vault Reports is offering the print and electronic books at medium price each, and no discount on the bundle is given.
97
<< Insert Table 14 and Table 15 about here >>
Incomplete product line (traditional mixed bundling with no bundle discount) This case is similar to the previous one with the only difference being that consumers do not get additional discount for choosing the bundle and the price they are charged for the bundle is the sum of the two individual prices. We consider nine possible market situations in total. The results are presented in tables 16 and 17 (book and newspaper categories respectively). In the book category, the market scenario in which the print book is at high price and the electronic book is at low price is consistently the most profitable scenario, followed by the scenario in which the print price is at medium level and the electronic price is at low level. With regard to the newspaper category, the electronic product should be offered at low price, while the price of the print product depends on the variable print costs as percentage of the print price – when the costs are relatively low, the print product should be offered at low price, but when the costs are relatively high, the print product should be offered at high price. << Insert Table 16 and Table 17 about here >>
Comparing the complete and incomplete product line profits Looking at the overall profits of the most appealing market scenarios in the book category listed in Tables 12, 14 and 16, we can conclude that the publisher generates the highest profit when the incomplete product line with bundle discount is chosen (Table 14); this holds regardless of the print variable costs. For example, at the 70% cost level, the overall profits for market scenario #181 are as follows: $12.12 (incomplete product
98
line with bundle discount, Table 14), $11.93 (complete product line, Table 12) and $10.82 (incomplete product line with no bundle discount, Table 16). The results are consistent for all most profitable market scenarios and cost levels. With regard to the newspaper category, the results are very similar with one exception – the complete product line is the most profitable strategy when the print variable costs are relatively high (e.g. 80%-90% of the medium print price). For example, at the 80% cost level, the overall profits for market scenario #184 are as follows: $27.36 (complete product line, Table 13), $27.04 (incomplete product line with bundle discount, Table 15) and $26.75 (incomplete product line with no bundle discount, Table 17). When the variable print costs are relatively low or medium level (e.g. 40%-50%-60%-70%), the results are as in the book category: incomplete product line with bundle discount is the best option, followed by the complete product line, and then by the incomplete product line with no bundle discount. Thus, the incomplete product line with discount leads to the highest profit in both product categories. We further elaborate on this result in the discussion section.
Relative Form Attractiveness and Choice Recall that the respondents evaluated the relative attractiveness of print book as compared to electronic (PDF) book and print newspaper as compared to electronic (online) newspaper on five attributes – image quality, browsing, layout, archival quality and convenience of use (7-point scale, 1-electronic better than print, 7-print better than electronic). They also provided information on the importance of each of these attributes when making a choice within the specific product category (7-point scale, 1-not at all
99
important, 7-very important). We mean-centered the five scales and computed four index variables for each of the two categories: print form index, electronic form index, difference index (bundle vs. superior form) and difference index (print form vs. electronic form). The example that follows is an illustration how the index variables were computed. BOOK Image Quality (IQ) Browsing (B) Layout (L) Archival Quality (AQ) Convenience of Use (CU) 7 3 2
Attribute 1 2 4 4 Perception Mean Centered (a) -3 -2 0 0 Importance (b) 2 5 5 2 Print Index (PI) = IQ(a*b) + B(a*b) + L(a*b) + AQ(a*b) + CU(a*b) Print Index (PI) = (-3)*2 + (-2)*5 + 0*5 + 0*2 + 3*2 = - 10 Electronic Index (EI) = - Print Index Electronic Index (EI) = 3*2 + 2*5 + 0*5 + 0*2 + (-3)*2 = 10 Bundle Index (BI)= Sum of Max (Print, Electronic) for IQ, B, L, AQ, CU Bundle Index (BI)= 3*2 + 2*5 + 0*5 + 0*2 + 3*2 = 22 Difference Index 1 (DI1) = Bundle Index – Max (Print Index, Electronic Index) Difference Index 1 (DI1) = 22 – 10 = 12 Difference Index 2 (DI2) = |Print Index – Electronic Index| = |10 – (-10)| = 20
We estimate a random parameters logit model for each product category using the following utility functions: U of Print Product (PP) = bPP + bPRC *Price + bPI*PI U of Electronic Product (EP) = bEP + bPRC * Price + bEI_1*EI U of Electronic Unit (EU) = bEU + bPRC * Price + bEI_2 * EI U of Print Product & Electronic Product (PPEP) = bPPEP + bPRC * Price + bDI1_1*DI1 U of Print Product & Electronic Unit (PPEU) = bPPEU + bPRC * Price + b DI1_2* DI1 U of No-Choice = bPRC* Price + bDI2* DI2 << Insert Table 18 and Table 19 about here >>
100
The results are displayed in Table 18 (book category) and Table 19 (newspaper category)25. Our goal is to discover whether the relative product form attractiveness influences choice over and above the preference for the different product offers and their price. In the book category the probability of choosing the print product is significantly positively related to the Print Index (bPI = 0.029, p < .05), for both categories the probability of choosing the electronic product is significantly positively related to the Electronic Index (bEI_1 = 0.029 for the book category and bEI_1 = 0.011 for the newspaper category, p’s < .05), and the probability of choosing the electronic unit is significantly positively related to the Electronic Index (bEI_2 = 0.013 for the book category and bEI_1 = 0.007 for the newspaper category, p’s < .05). With respect to the two difference indices, the results reveal that the difference in attractiveness between the bundle and the more attractive product form significantly affect the probability of buying the print product/electronic product bundle in the book category only (bDI1_1 = 0.107, , p < .05); the difference in attractiveness of the two forms significantly affects the probability to defer choice also in the book category only (bDI2 = 0.023, p < .05). For the book category we estimate individual level parameters for the intrinsic preference for the print form, the PDF form and the bundle. We then regress the differences in the constants for pairs of alternatives for each individual for differences in the indexes. Specifically, we compute the difference between the print and the PDF book alternative specific constants, and then regress the difference on the Difference Index 2; also, we compute the difference between the alternative specific constants of the bundle
25
In the book category we estimate a model in which the means of the price parameter and the alternative specific constants for print book, PDF book and print book/PDF book bundle are allowed to vary and are assumed to be normally distributed, while in the newspaper category only the price parameter is allowed to vary. These are the models that provide the best fit to the data.
101
and the individual form, and then regress this difference on the Difference Index 1. In both cases the regression coefficients are negative but marginal or non-significant (0.007, p>.10, and -0.022, p = 0.099). Nevertheless, these results provide weak evidence that the indexes are related to the brand specific constants: the higher the value of the indexes, the lower the difference between the brand specific constants. This suggests that the relative attractiveness of the product forms explains at least a portion of the difference in the intrinsic preference for these product forms. In sum, controlling for the preference for the various product offers and their prices, the more consumers find the specific form attractive and superior to the other form, the more likely they are to choose it. Furthermore, the probability of buying the bundle in the book category is positively associated with the difference in the attractiveness between the bundle and the more attractive product form.
Discussion
Several issues are important from both theoretical and managerial point of view. Why is offering the incomplete product line with discount more profitable than offering the complete product line? Our book category simulation results suggest that when the electronic units are introduced, customers who would otherwise buy the print book or the full electronic book switch to the electronic units option. Further, the no choice option market share decreases only slightly as the unbundled electronic units do not attract a significant number of new customers. Consequently, there is cannibalization of print and full electronic books sales to unbundled electronic units sales, and this loss in revenue is not compensated by the revenue from the new customers who enter the market when the
102
electronic units are offered. In the newspaper category the results follow a similar pattern. When the electronic units are introduced, consumers who would otherwise buy the full electronic product choose a single unit instead. Again, the no-choice option share decreases only slightly. This results in loss of revenue because of the cannibalization of full electronic product sales to electronic units sales, and because of the limited market expansion from selling electronic units. Why is the incomplete product line with no discount worse than the other two strategies? In the book category most of the consumers who buy the bundle in the incomplete with discount case buy print only when there is no bundle discount. Because of this, the company is better off charging a discounted bundle price as it brings extra revenue without incurring extra cost. Regarding the newspaper purchases, most of the consumers who buy the bundle in the incomplete product line with discount now switch to print newspaper or electronic newspaper only. As offering the electronic newspaper as a part of the bundle does not involve extra costs, it is more profitable for the company to sell the bundle. Thus, our simulation results show that when the unbundling of electronic content does not lead to attracting a sizeable amount of new customers or getting extra revenue from the current customers, it is more profitable for the company not to pursue such an approach. The optimal solution is to offer the print products at a high price to extract more revenue from the customers who have a high valuation for this form, and to offer the electronic products and the bundle of the two forms at low price. This way the company can still persuade the more price sensitive consumers to enter the market and buy the cheaper electronic products, and extract additional revenue from those consumers
103
who value the bundle and buy both forms when the bundle is offered at a discount. The optimal pricing strategy is as follows: book category – print book at $33.40, electronic book at $25.20, and bundle at $43.80 (33% effective bundle discount); newspaper category – print newspaper at $114.50, electronic newspaper at $34.50, and bundle at $105.20 (42% effective bundle discount). The conclusions of Essay 2 and Essay 3 are complementing each other. Recall that in Essay 2 consumers do not receive discount for buying the bundle. Therefore, it is more profitable for the company to offer unbundled electronic units (content unbundled mixed bundling) as compared to full electronic books (traditional mixed bundling). In Essay 3 we also compare these two strategies, although they are not exactly the same as in Essay 2: offering the complete product line (which is similar to CUMB but not the same, see explanation bellow) vs. offering the incomplete product line without bundle discount (which is TMB in Essay 2). The results in Essay 3 are similar to those in Essay 2 – the profit is higher when unbundled electronic units are offered (but lower than TMB with discount), while TMB with no discount is the least profitable. How do our findings relate to the extant literature on bundling? Previous studies suggest that mixed bundling is more beneficial than pure bundling and pure components is the optimal bundling strategy when there is asymmetry in the reservation prices for the bundle components (Adams and Yallen 1976; Schmalensee 1984; McAfee, McMillan and Whinston 1989). Further, with respect to interrelated products and when the marginal costs are not too high, Venkatesh and Kamakura (2003) argue that while pure components strategy is optimal for moderate to strong substitutes and complements, mixed bundling is optimal for weak substitutes/complements. We extend the above
104
findings by further clarifying which bundling strategy is more appealing when different mixed bundling strategies are feasible. We show that, applied to product forms, mixed bundling is the optimal strategy when full print and electronic products are offered; in the case of separate electronic units companies are better off pursuing pure bundling (which is offering the full electronic product). Also, similar to Wilson et al. (1990) we show that when there is no market growth from unbundling and new market segments are not attracted by the unbundled components, product bundles should not be unbundled. In terms of optimal pricing, Jedidi et al. (2003) argue that when the reservation prices of the bundle components are highly heterogeneous, a uniform high price for all products and bundles is optimal; otherwise a hybrid strategy may work better. In addition, Venkatesh and Chatterjee (2003) claim that low-priced components should be targeted at consumers who have low value for the information content. Our findings suggest that, in the case of mixed bundling of different forms of information products, it is preferable to charge high price for the print product but low price for the full electronic product and the bundle. As the costs associated with the electronic products are negligible, it is beneficial for the company to sell them at a relatively low price to consumers who have a low value for the information content and attract them to make a purchase. Thus, we show that it is not always optimal to charge a high price for the individual products under mixed bundling, and the cost structure may play a role as well. In sum, we show that with respect to product forms, the mixed bundling strategy of offering the individual product forms and the bundle is the optimal product strategy for a content provider to pursue. Adding to the existing literature on bundling of interrelated products, we show that mixed bundling is optimal when the valuations of the bundle
105
components vary significantly across consumers, as well as the contingent valuation for the bundle. We also show that the cost structure may play an important role in bundling decisions, especially when determining the optimal pricing strategy – it may allow the seller to generate higher profit in the case of mixed bundling versus pure components (incomplete product line with discount versus incomplete product line with no discount) even when the bundle discount is sizeable. Specifically, companies are better off selling the bundle at a discounted price as compared to selling only the individual forms – from a seller’s point of view the bundle does not involve extra cost over and above the cost of the print form but can generate extra revenue over and above the print form revenue.
Managerial Implications
Our study has important managerial implications in the areas of product line design and pricing. We offer guidelines to publishers and other content providers on how to profitably market digital content together with traditional print content. Companies should implement mixed bundling regarding lines of information products, charging a high price for the print products and offering the electronic products and the bundle at a low price. Additionally, we outline a choice-based conjoint methodology that can be applied to a wide range of products including magazines, journals, music, etc. Second, our study has implications in the area of market segmentation and targeting. Based on the estimates that can be obtained about the valuations of the different product forms, companies can segment their customers and target their product offering and pricing accordingly. For example, consumers who have low valuation for the information content may be encouraged to try and buy lower priced electronic products,
106
while customers with high valuation for the information content may be encouraged to buy the bundle. Further, targeting can be applied at individual level as well resulting in personalization of product offering and customization of digital goods.
Conclusion
In this essay we use choice experiments and choice-based conjoint analysis to obtain estimates of customer valuations of the price and non-price attributes of interest. We present respondents with hypothetical choice situations in which the attributes are varied to determine how choice fluctuates with attribute changes. Then, controlling for consumer heterogeneity, we estimate random effects logit models and use the results of the conjoint analysis to derive the optimal product mix and pricing. Consequently, we are able to come up with recommendations for optimal product line design and pricing strategy. Our empirical investigation and conclusions are conditional on the content provider being a monopolist. It will be interesting to investigate what the optimal product line and pricing strategies will be when competition exists on the marketplace. In some conditions it may be optimal to offer unbundled electronic units, and complete product line may be the market equilibrium. The rational behind such an equilibrium solution may be that offering more product options results in higher customer satisfaction and likelihood of repeat purchase. Another interesting question is how the market norms affect the expectations of the consumers on how the print content, the electronic content and the bundle are priced. Although in our study these norms are taken into account when modeling the valuations of the consumers for the different product offers and
107
consequently in the market simulations, it will be beneficial for the marketing theory and practice to further investigate the long term effects of offering low priced electronic products and form bundles on the valuation of information products, as well as the unbundling of electronic content.
108
TABLES Table 1: Purchase Probabilities for Bundle: Product, Discount and Usage Conditions
Panel A. Product, Discount and Usage Conditions Discount No Discount d% Discount Predictions Bundle INFO CONV INFO CONV Same Usage (1a) Pr (VFj|Fi > Pj) (1b) Pr (VC+VFj|Fi> Pj) (2a) Pr (VFj|Fi + d(Pi + Pj) > Pj) (2b) Pr (VC+VFj|Fi+d(Pi + Pj) > Pj) INFO: 1a=3a, 1a=2a, 2a<4a, 3a<4a CONV > INFO, all conditions Panel B. Discount and Relative Price Conditions – Information Products Only Price of each item same (PA) No Discount d % Discount Predictions (5) Pr (VFj|Fi > PA | VFi >VFj) + Pr (VFi|Fj > PA | VFj > VFi) (6) Pr (VFj|Fi > (1 - 2d)PA | VFi > VFj) + Pr (VFi|Fj > (1 - 2d)PA | VFj > VFi) 7>5, 8 > 6, 6>5, 8>7 Item i higher priced: Pi = (1+k)PA, Pj = (1-k)PA (7) Pr (VFj|Fi > (1-k) PA | VFi-2k PA >VFj) + Pr (VFi|Fj > (1+k)PA | VFi-2k PA < VFj) (8) Pr (VFj|Fi > (1 – k - 2d)PA | VFi - 2kPA >VFj) + Pr (VFi|Fj > (1 + k - 2d)PA | VFi - 2kPA < VFj) Different Usage (3a) Pr (V*Fj|Fi > Pj) (3b) Pr (VC+V*Fj|Fi> Pj) (4a) Pr (V*Fj|Fi + d(Pi + Pj) > Pj) (4b) Pr (VC+V*Fj|Fi+d(Pi + Pj) > Pj) CONV: 1b<3b, 2b<4b, 1b<2b, 3b<4b
109
Table 2: Perceived Appropriateness of Generated Usage Situations
Usage Situation MD** Rank e-mail pages/chapters -2.425 1 give as a present 2.238 2 Form 1: read for pleasure 1.762 3 PAPER BOOK read while traveling 1.738 4 Form 2: read to others 1.452 5 ELECTRONIC search -0.952 6 BOOK copy citations/paragraphs -0.600 7 need on a short notice 0.167 8 archive -0.024 9 read the news -2.548 1 Form 1: read old articles -1.902 2 PAPER read for pleasure -1.690 3 NEWSPAPER read during lunchtime/breaks 1.476 4 Form 2: archive -1.171 5 ONLINE read at home 0.805 6 NEWSPAPER search 0.756 7 read to others 0.463 8 follow the stock market -0.375 9 e-mail articles 0.310 10 for cooking 1.071 1 Form 1: for baking 0.976 2 STICK of shortening in cakes 0.921 3 MARGARINE take on a picnic -0.350 4 Form 2: eat with pancakes 0.262 5 TUB of spread on a tray -0.093 6 MARGARINE spread on bread for breakfast -0.048 7 for frying fish/meat -0.025 8 wash dishes 3.000 1 Form 1: wash your hands in public places 2.372 2 LIQUID SOAP wash delicate laundry 1.930 3 Form 2: take on a picnic 0.767 4 BAR of SOAP take a shower at home -0.628 5 take a shower while traveling -0.349 6 wash your hands at home 0.302 7 take a shower in the gym 0.140 8 when looking for a good cup of coffee -2.770 1 offer to your guests at home -2.580 2 Form 1: while traveling 1.130 3 INSTANT COFFEE when in a hurry 1.040 4 Form 2: on a picnic -0.480 5 GROUND COFFEE during a meeting -0.420 6 while in class -0.190 7 in the middle of the day -0.170 8 *p<0.05 **The mean difference calculated as mean appropriateness of Form 1 minus mean appropriateness of Form 2. Product t stat -12.825 10.408 7.638 6.581 6.203 -4.421 -2.399 0.510 -0.100 -11.790 -7.963 -7.275 6.173 -4.566 3.778 4.036 2.649 -1.684 1.915 4.079 3.992 3.441 -1.236 1.097 -0.371 -0.184 -0.097 17.037 11.715 8.908 2.760 -2.071 -1.044 2.383 0.374 -9.222 -9.259 3.191 2.982 -1.297 -1.305 -0.548 -0.848 Rank 1 2 3 4 5 6 7 8 9 1 2 3 4 5 7 6 8 10 9 1 2 3 4 5 6 7 8 1 2 3 4 6 7 5 8 1 2 3 4 6 5 8 7 * * * * * * *
* * * * * * * *
* * *
* * * * * * * * * *
110
Table 3: Study 2 - Mean Percentage of Points Allocated to Bundle Panel A: Bundle Discount and Usage Conditions Discount Bundle No Discount INFO CONV 25 % Discount INFO CONV Predictions Product Book Newspaper Margarine Coffee Book Newspaper Margarine Coffee Same Usage (1a) 9.77% (1a) 9.17% (1b) 15.10% (1b) 12.30% (2a) 9.18% (2a) 11.36% (2b) 26.85% (2b) 31.00% INFO: 1a=3a, 1a=2a, 2a<4a, 3a<4a CONV > INFO, all conditions Different Usage (3a) 8.28% (3a) 6.35% (3b) 23.75% (3b) 17.61% (4a) 20.08% (4a) 18.90% (4b) 49.50% (4b) 48.19% CONV: 1b<3b, 2b<4b, 1b<2b, 3b<4b
Panel B: Bundle Discount and Relative Price Conditions Discount No Discount INFO 25 % Discount Predictions
a
INFO
Booka Newspaperb Bookc Newspaperd
Equal Prices (5) 6.10% (5) 6.88% (6) 15.68% (6) 16.23% 7>5, 8 > 6, 6>5, 8>7
Print High. Electronic Low (7) 11.95% (7) 8.63% (8) 13.58% (8) 14.07%
Prices for book in equal condition: print = electronic= $29.99, bundle = $59.98; in highlow condition: print = $37.49, electronic = $22.49, bundle = $59.98. b Prices for newspaper in equal condition: print = electronic = $38.99, bundle = $77.97; in high-low condition: print = $48.79, electronic = $29.19, bundle = $77.97. c Prices are same as a except bundle = $44.99. d Prices are same as b except bundle = $58.49.
111
Table 4: Study 2 ANOVA Results
Source USAGE (USE) DISCOUNT (DISC) REL. PRICE (RPRC) USE * DISC USE * RPRC DISC * RPRC USE * DISC * RPRC Error Total
SS 26.67 40.99 0.72
df
MARGARINE MS F 1 26.67 6.01 1 40.99 9.24 1 0.72 0.16
Sig. SS 0.02 22.53 0.00 130.31 0.69 9.76
df
COFFEE MS F 1 22.53 4.09 1 130.31 23.65 1 9.76 1.77
Sig. 0.05 0.00 0.19
337.24 616.25
76 80
4.44
451.87 928.23
82 86
5.51
Source SS USAGE (USE) 16.92 DISCOUNT (DISC) 32.68 REL. PRICE (RPRC) 2.97 USE * DISC 27.11 USE * RPRC 5.24 DISC * RPRC 16.84 USE * DISC * RPRC 0.91 Error 1042.35 Total 1145.02
DV: Ln of points allocated to the bundle
df 1 1 1 1 1 1 1 232 240
BOOK MS 16.92 32.68 2.97 27.11 5.24 16.84 0.91 4.49
F 3.77 7.27 0.66 6.03 1.17 3.75 0.20
Sig. 0.05 0.01 0.42 0.01 0.28 0.05 0.65
NEWSPAPER SS df MS F 1.49 1 1.49 0.33 48.46 1 48.46 10.54 1.02 1 1.02 0.22 17.96 1 17.96 3.91 2.17 1 2.17 0.47 9.89 1 9.89 2.15 0.94 1 0.94 0.20 1066.50 232 4.60 1148.43 240
Sig. 0.57 0.00 0.64 0.05 0.49 0.14 0.65
112
Table 5: Profit Comparison under CUMB and TMB Strategies.
Print Sales a.
CUMB TMB ? QP > QP ? ? ?;
CUMB ? QP ? ? PC
TMB ? QP ? ? PEB
* ?* CUMB > ? TMB
IF
? ? ? Q CUMB P ?? ? ?P ? C ?? ? ? kFN ? ? ? ?? ? ? ( P ? c) ? ? ? P ?
2 2
? ?
? ? kFN ? ? kFN ;+?? ?U? ? PP ? c ? ? PP ? c ?
? ? ?k ? ? k ? ? ? ? ?; P ? c ? U ? P ? c ;+?? P ? ? ? ? P
CUMB TMB ? QP ? QP ? ? PP ? c ?
? VH ?+ ? 4k ?
?? ? Q TMB P ?? ? ?P ? EB ??
? ? k ? ? ?? ? ? ? P ?c? ? ? ? P
2
2
? 1 ?> 4k F S ? ?
? ? ? ?
b.
CUMB TMB ? QP > QP ?
? ? kFN kFN ? ? ; ? ? PP ? c PP ? c ?
? ? ?k ? ? k ? * ? Always ? * ? ? ?; P ? c ? U ? P ? c ;+?? CUMB > ? TMB P ? ? ? ? P
c.
Q
CUMB P
>Q
TMB P
? ? ? kFN ? ? kFN ? ? ? ? ?; P ? c ? U ? P ? c ;+?? P ? ? ? P ?
? ?k k ? ? ; ? ?P ?c P ?c? P ? ? P
? ?k k ? ? ? ? P ?c;P ?c? P ? ? P
CUMB TMB ? QP ? QP ? ? PP ? c ?
? 1 ?> ? 4k F S ?
? 1 ?+ ? 4k F S ?
? ? ? Q CUMB P ?? ? ?P ? C ? ?
? ? kFN ? ? ? ?? ? ? ( P ? c) ? ? ? P ?
? ? ? ?
2
2
? V ?+ H 4k ? ?
?? k ? 2 ? ? Q TMB P ? ?? ? P ?c? ? ?? ? P ? P EB ? ? ? ?
?? k ? 2 ? ? Q TMB P ? ?? ? P ?c? ? ?? ? P ? EB ? ? ?? P
2
? ? ? ?
? ? ? ?
2
? ? ? ?
? ? ? ?
? ? ? ?
? ? ? ?
d.
Q
CUMB P
>Q
TMB P
? ? kFN kFN ? ? ? ? P ?c ;P ?c? P ? ? P
? ? ? kFN ? ? kFN ? ? ? ? ?; P ? c ? U ? P ? c ;+?? P ? ? ? P ?
CUMB TMB ? QP ? QP ? ? PP ? c ?
?? k F N ? 2 ? ? Q CUMB P ? ? ?? ? ( P ? c) ? ? ? ? P ? C ? ? ?? P
2
2
? V ?> H 4k ? ?
2
e.
Q
CUMB P
?Q
TMB P
TMB 2 ? V H ?? ? Q P ? ?k ? ? k ? ? ? U +? ? ? ; ; ? ? ? ? ? ? ? ? ? P ? c P c ? PEB ? P ? 4k ? ? ? ? P ? ? ?
? k ? ?? ? P ?c? ? ? P ?
? ? Q TMB ? Q CUMB P ?>? P ? PP ? c ? ? ?
? ? ? ?
2
? 1 ?+ ? 4k F S ?
2
? ? ? Q CUMB P ?? ? ?P ? C ? ?
2
? ? kFN ? ? ? ?? ? ? ( P ? c) ? ? ? P ?
2
f.
Q
CUMB P
?Q
TMB P
? ? kFN kFN ? ? ? ? P ?c ;P ?c? P ? ? P
? ? ? kFN ? ? kFN ? ? ? ? ?; P ? c ? U ? P ? c ;+?? P ? ? ? P ?
? ? ?k ? ? k 1 ? ? ? ?; P ? c ? U ? P ? c ;+?? ? P ? ? ? P ? 4k F S
? ? k F N ? 2 ? ? Q CUMB P ? ? ?? ? ( P ? c) ? ? ? ? P ? C ? ? ?? P
? V ?+ H 4k ? ?
?? ? Q TMB P ?? ? ?P ? EB ??
? ? k ? ? ?? ? ? ? P ?c? ? ? ? P
? ? Q TMB ? Q CUMB P ?>? P ? P c ? ? P ? ?
g.
CUMB TMB QP ? QP
? ?k k ? ? ? ? P ?c;P ?c? P ? ? P
? ?k k ? ? ; ? ?P ?c P ?c? P ? ? P
* Always ? * CUMB < ? TMB
h.
Q
CUMB P
?Q
TMB P
? ? kFN kFN ? ? ? ? P ?c ;P ?c? P ? ? P
1 4kF S
?? k F N ? 2 ? ? Q CUMB P ? ? ?? ? ( P ? c) ? ? ? ? P ? C P ? ? ? ?
? ? ? ?
2
? ? Q TMB ? Q CUMB P ?>? P ? PP ? c ? ? ?
? VH ?+ ? 4k ?
?? k ? 2 ? ? Q TMB P ? ?? ? P ?c? ? ?? ? P ? P EB ? ? ? ?
? ? ? ?
2
? ? ? ?
113
Table 6: Actual Purchase Rates under the Two Mixed Bundling Strategies
Content Unbundled Mixed Bundling PDF Price 0.50 0.75 1.00 1.25 Total Print 57.21% 54.11% 63.55% 62.88% 59.57% PDF 11.21% 13.59% 9.47% 6.67% 10.08% Bundle 4.95% 3.93% 4.87% 3.13% 4.13% Nothing 26.64% 28.37% 22.11% 27.32% 26.22% Total 21.17% 25.86% 23.42% 29.55% 100.00%
Traditional Mixed Bundling PDF Price 0.25 0.50 0.75 1.00 1.10 Total Print 43.57% 45.20% 51.60% 48.57% 40.00% 47.37% PDF 16.43% 17.65% 8.33% 5.00% 11.43% 12.32% Bundle 7.14% 4.95% 6.73% 5.00% 5.71% 5.89% Nothing 32.86% 32.20% 33.33% 41.43% 42.86% 34.42% Total 14.74% 34.00% 32.84% 14.74% 3.68% 100.00%
114
Table 7: Estimation Results - Content Unbundled Mixed Bundling
Variable Agriculture Choice Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Coefficient 0.486 0.232 -0.860 0.391 -0.341 -0.873 0.515 -0.311 -0.672 0.262 -0.209 0.739 0.129 0.057 0.235 0.000 -0.017 -0.006 0.902 -0.553 -5.802 SE 0.350 0.412 1.146 0.173 0.202 0.590 0.192 0.229 0.569 0.306 0.358 0.907 0.142 0.163 0.384 0.006 0.006 0.021 0.285 0.179 3.364 b/St.Er. 1.389 0.565 -0.750 2.262 -1.691 -1.479 2.684 -1.357 -1.181 0.855 -0.584 0.814 0.907 0.353 0.612 0.038 -2.672 -0.279 3.169 -3.092 -1.725 2.142 1.985 3245 43 -3290.938 -4498.525 2415.175 23 0.000 P[|Z|>z] 0.165 0.572 0.453 0.024 0.091 0.139 0.007 0.175 0.238 0.393 0.559 0.416 0.364 0.724 0.541 0.970 0.008 0.780 0.002 0.002 0.085 0.032 0.047
Behavioral Science
Education
General Interest
Medicine
Print Price
PDF Price
PDF Price
Derived SD of parameter distributions Print 0.819 1.753 Bundle 1.958 3.887 Number of observations Iterations completed Log likelihood function Restricted log likelihood Chi squared Degrees of freedom Prob[ChiSqd > value] =
115
Table 8: Estimation Results - Traditional Mixed Bundling
Variable Agriculture Choice Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Print PDF Bundle Coefficient -0.317 -0.732 -0.487 0.504 -2.339 -0.235 0.974 -0.003 0.321 0.252 -0.266 0.085 -0.099 -0.285 -0.661 0.011 -0.001 -0.035 -0.245 -1.252 -0.815 SE 0.337 0.551 0.812 0.283 1.036 0.542 0.267 0.424 0.481 0.290 0.454 0.480 0.187 0.258 0.384 0.006 0.008 0.012 0.262 0.397 0.508 b/St.Er. -0.940 -1.329 -0.599 1.780 -2.257 -0.434 3.649 -0.007 0.668 0.869 -0.587 0.177 -0.531 -1.102 -1.723 1.802 -0.151 -2.992 -0.934 -3.156 -1.605 950 7 -1056.12 -1316.98 -1088.57 P[|Z|>z] 0.347 0.184 0.549 0.075 0.024 0.665 0.000 0.995 0.504 0.385 0.557 0.860 0.596 0.270 0.085 0.072 0.880 0.003 0.351 0.002 0.109
Behavioral Science
Education
General Interest
Medicine
Print Price
PDF Price
Number of observations Iterations completed Log likelihood function No coefficients Constants only
116
Table 9: Simulation Results – Profit under the Two Mixed Bundling Strategies. A. Content Unbundled Mixed Bundling
PDF Price as % Print Price 0% 25% 50% 75% 100% 110% 125% Actual Market Share PDF Bundle No Choice 0.166 0.214 0.273 0.161 0.061 0.301 0.133 0.021 0.284 0.105 0.014 0.257 0.082 0.014 0.231 0.075 0.014 0.222 0.065 0.015 0.210 0.094 0.017 0.243 Profit PDF 0.000 0.833 1.376 1.631 1.708 1.709 1.690 1.609
Print 0.347 0.477 0.561 0.624 0.672 0.689 0.710 0.646
Print 6.295 8.655 10.188 11.310 12.185 12.477 12.862 11.653
Bundle 3.910 1.435 0.607 0.471 0.536 0.581 0.655 0.629
Total 10.205 10.923 12.170 13.412 14.429 14.767 15.207 13.890
B. Traditional Mixed Bundling
PDF Price as % Print Price 0% 25% 50% 75% 100% 110% 125% Actual Market Share PDF Bundle No Choice 0.203 0.079 0.277 0.165 0.072 0.306 0.132 0.064 0.335 0.105 0.057 0.362 0.082 0.050 0.387 0.075 0.047 0.397 0.064 0.043 0.412 0.120 0.059 0.349 Profit PDF 0.000 1.302 2.084 2.475 2.588 2.577 2.517 2.141
Print 0.441 0.457 0.469 0.477 0.481 0.481 0.481 0.472
Print 7.921 8.219 8.431 8.562 8.618 8.623 8.611 8.455
Bundle 1.231 1.390 1.490 1.536 1.537 1.527 1.502 1.492
Total 9.152 10.911 12.005 12.572 12.743 12.727 12.630 12.088
117
Table 10: Conjoint Design and Price Levels
A. Conjoint Factors Product offer
Print 1 0 0 1 1
Print Product (PP) Electronic Product (EP) Electronic Unit/Section (EU) Print Product & Electronic Product Bundle (PPEP) Print Product & Electronic Unit Bundle (PPEU) B. Price Levels WSJ Subscription (26 weeks) Print WSJ Electronic WSJ Electronic WSJ Section Print WSJ & Electronic WSJ Print WSJ & Electronic WSJ Section Vault Guide Print Book Electronic Book Electronic Section Print Book & Electronic Book Print Book & Electronic Section
Full Electronic 0 1 0 1 0
Electronic Bundle Price Unit Level 0 0 L, M, H 0 0 L, M, H 1 0 L, M, H 0 1 L, M, H 1 1 L, M, H
Low 86.50 34.50 13.20 105.20 94.90 Low 25.20 25.20 9.70 43.80 33.20
Medium 99.50 39.50 15.20 120.90 109.20 Medium 29.00 29.00 11.10 50.40 38.20
High 114.50 45.50 17.40 139.10 126.60 High 33.40 33.40 12.80 58.00 44.00
118
Table 11: Random Parameters Logit Model Results (conjoint choice experiment)
Book Category Coefficient SE b/SE Random parametes in utility functions Bundle -6.647 0.534 -12.457 Print 8.485 0.878 9.666 Price -0.261 0.028 -9.378 Full Electronic 7.306 0.793 9.210 Electronic Unit 3.063 0.419 7.314 Bundle Print Price Full Electronic Electronic Unit Variable P[|Z|>z] 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Derived standard deviations of prameter distributions 2.236 0.365 6.127 5.249 0.515 10.193 0.107 0.013 8.494 1.621 0.296 5.469 2.316 0.230 10.068
Number of observations Log likelihood function Restricted log likelihood Chi squared (df=10) R-sq Adjusted R-sq Newspaper Category
4176 (87 respondents x 16 choices x 3 options) -904.67 -2486.96 3164.58 (p=0.000) 0.636 0.635
Coefficient SE b/SE Random parametes in utility functions Bundle -3.384 0.572 -5.921 Print 11.333 1.376 8.235 Price -0.133 0.014 -9.575 Full Electronic 6.793 0.619 10.968 Electronic Unit 1.452 0.380 3.819 Bundle Print Price Full Electronic Electronic Unit
Variable
P[|Z|>z] 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Derived standard deviations of prameter distributions 4.005 0.565 7.083 6.222 0.720 8.646 0.025 0.004 6.157 3.253 0.502 6.475 3.515 0.357 9.852
Number of observations Log likelihood function Restricted log likelihood Chi squared (df=10) R-sq Adjusted R-sq
4176 (87 respondents x 16 choices x 3 options) -783.75 -2490.55 3413.59 (p=0.000) 0.685 0.684
119
Table 12: Book Category – Market Simulation Results
Market Share PDF Print Book Print Book No Revenue Unit and and Choice PDF Book PDF Unit 0.141 0.144 0.164 0.125 0.146 0.141 0.144 0.164 0.125 0.146 0.141 0.144 0.164 0.125 0.146 0.141 0.144 0.164 0.125 0.146 0.141 0.144 0.164 0.125 0.146 0.141 0.144 0.164 0.146 0.125 0.094 0.099 0.092 0.065 0.064 0.094 0.099 0.092 0.065 0.064 0.094 0.099 0.092 0.065 0.064 0.094 0.099 0.092 0.065 0.064 0.094 0.099 0.092 0.065 0.064 0.094 0.099 0.092 0.064 0.065 0.039 0.016 0.037 0.025 0.024 0.039 0.016 0.037 0.025 0.024 0.039 0.016 0.037 0.025 0.024 0.039 0.016 0.037 0.025 0.024 0.039 0.016 0.037 0.025 0.024 0.039 0.016 0.037 0.024 0.025 0.181 0.182 0.175 0.163 0.159 0.181 0.182 0.175 0.163 0.159 0.181 0.182 0.175 0.163 0.159 0.181 0.182 0.175 0.163 0.159 0.181 0.182 0.175 0.163 0.159 0.181 0.182 0.175 0.159 0.163 20.939 20.502 20.442 20.699 20.446 20.939 20.502 20.442 20.699 20.446 20.939 20.502 20.442 20.699 20.446
Rank
Market Scenario # 181 182 172 100 91 181 182 172 100 91 181 182 172 100 91 181 182 172 100 91 181 182 172 100 91 181 182 172 91 100
Print Book
PDF Book
Cost
Cost % Profit of Print Price
1 2 3 12 15 1 2 3 15 18 1 2 3 22 23 1 2 3 30 33 1 2 3 37 38 1 2 3 37 38
0.311 0.322 0.306 0.420 0.413 0.311 0.322 0.306 0.420 0.413 0.311 0.322 0.306 0.420 0.413 0.311 0.322 0.306 0.420 0.413 0.311 0.322 0.306 0.420 0.413 0.311 0.322 0.306 0.413 0.420
0.234 0.236 0.225 0.201 0.194 0.234 0.236 0.225 0.201 0.194 0.234 0.236 0.225 0.201 0.194 0.234 0.236 0.225 0.201 0.194 0.234 0.236 0.225 0.201 0.194 0.234 0.236 0.225 0.194 0.201
5.149 15.790 5.069 15.433 5.051 40.00% 15.390 5.926 14.773 5.813 14.634 6.436 14.503 6.337 14.165 6.314 50.00% 14.127 7.408 13.292 7.266 13.181 7.723 13.216 7.604 12.898 7.577 60.00% 12.865 8.889 11.810 8.719 11.727
20.939 9.011 11.929 20.502 8.872 11.630 20.442 8.840 70.00% 11.602 20.699 10.371 10.328 20.446 10.172 10.274 20.939 20.502 20.442 20.699 20.446 20.939 20.502 20.442 20.446 20.699 10.298 10.641 10.139 10.363 10.103 80.00% 10.339 11.852 8.847 11.625 8.821 11.585 9.354 11.406 9.096 11.366 90.00% 9.076 13.078 7.368 13.334 7.365
Market Scenario Print Book PDF Book PDF Unit Print Book Print Book # PDF Book PDF Unit 91 Medium 100 Medium 172 High 181 High 182 High Low Low Low Low Low Medium High Medium High High Low Low Low Low Low Low Low Low Low Medium
120
Table 13: Newspaper Category – Market Simulation Results
Rank Market Scenario # 181 208 19 100 91 181 100 208 172 91 181 172 182 100 91 181 184 182 100 91 187 178 184 100 101 187 178 184 106 97 Print WSJ Low Medium Medium Medium Medium Medium High Print WSJ Online WSJ Market Share Online Print WSJ Print WSJ No Revenue Unit and and Choice Online WSJ Online Unit 0.103 0.123 0.096 0.101 0.115 0.103 0.101 0.123 0.116 0.115 0.103 0.116 0.105 0.101 0.115 0.103 0.108 0.105 0.101 0.115 0.109 0.124 0.108 0.101 0.103 0.109 0.124 0.108 0.107 0.122 0.259 0.284 0.206 0.242 0.241 0.259 0.242 0.284 0.258 0.241 0.259 0.258 0.269 0.242 0.241 0.259 0.142 0.269 0.242 0.241 0.067 0.067 0.142 0.242 0.249 0.067 0.067 0.142 0.060 0.060 0.021 0.021 0.011 0.017 0.017 0.021 0.017 0.021 0.020 0.017 0.021 0.020 0.003 0.017 0.017 0.021 0.044 0.003 0.017 0.017 0.062 0.061 0.044 0.017 0.003 0.062 0.061 0.044 0.042 0.041 Market Scenario 178 181 182 184 187 208 0.075 0.095 0.056 0.071 0.069 0.075 0.071 0.095 0.073 0.069 0.075 0.073 0.078 0.071 0.069 0.075 0.092 0.078 0.071 0.069 0.098 0.096 0.092 0.071 0.073 0.098 0.096 0.092 0.086 0.084 Print WSJ High High High High High High 51.256 51.985 52.621 51.213 50.834 51.256 51.213 51.985 50.539 50.834 51.256 50.539 50.228 51.213 50.834 51.256 44.963 50.228 51.213 50.834 41.559 41.200 44.963 51.213 50.571 41.559 41.200 44.963 41.794 41.435 Online WSJ Cost Cost % of Print Price Profit
1 2 3 4 11 1 2 3 4 7 1 2 3 4 8 1 2 3 8 13 1 2 3 16 23 1 2 3 21 24 Market Scenario 19 91 97 100 101 106 172
0.025 0.026 0.133 0.055 0.055 0.025 0.055 0.026 0.025 0.055 0.025 0.025 0.026 0.055 0.055 0.025 0.035 0.026 0.055 0.055 0.044 0.044 0.035 0.055 0.058 0.044 0.044 0.035 0.100 0.099 Online WSJ Low Low Low Low Low Low Low
0.517 0.451 0.498 0.513 0.504 0.517 0.513 0.451 0.508 0.504 0.517 0.508 0.519 0.513 0.504 0.517 0.579 0.519 0.513 0.504 0.619 0.607 0.579 0.513 0.515 0.619 0.607 0.579 0.605 0.594
12.137 13.174 13.913 12.534 12.442 15.171 15.667 16.468 15.071 15.553 18.205 18.085 17.796 18.801 18.663 21.239 15.401 20.762 21.934 21.774 13.821 13.712 17.601 25.068 24.609 15.548 15.426 19.801 18.044 17.890
40.00%
39.120 38.811 38.708 38.679 38.392 36.086 35.546 35.518 35.468 35.281 33.051 32.454 32.432 32.412 32.171 30.017 29.561 29.466 29.279 29.060 27.738 27.487 27.361 26.145 25.961 26.010 25.773 25.161 23.750 23.544
50.00%
60.00%
70.00%
80.00%
90.00%
Online Print WSJ Print WSJ Unit Online WSJ Online Unit High Medium Medium High High High Medium Low Low High Low Low High Low Low Low Low Low Medium Low Low
Online Print WSJ Print WSJ Unit Online WSJ Online Unit Low Low Medium Low Low Low
Low Medium High Low High Low Low High Low Low High Medium Low High High Medium High Low
121
Table 14: Book Category – Market Simulation Results (Incomplete Product Line)
Market Share PDF Print Book Book and PDF Book 0.308 0.259 0.328 0.228 0.269 0.308 0.328 0.259 0.228 0.269 0.308 0.328 0.259 0.337 0.269 0.308 0.328 0.337 0.259 0.269 0.308 0.328 0.337 0.259 0.269 0.308 0.328 0.337 0.259 0.269 0.109 0.072 0.041 0.122 0.025 0.109 0.041 0.072 0.122 0.025 0.109 0.041 0.072 0.011 0.025 0.109 0.041 0.011 0.072 0.025 0.109 0.041 0.011 0.072 0.025 0.109 0.041 0.011 0.072 0.025
Rank
Market Scenario # 172/181/182 91/100/101 184 208 85 172/181/182 184 91/100/101 208 85 172/181/182 184 91/100/101 169/187 85 172/181/182 184 169/187 91/100/101 85 172/181/182 185 169/187 91/100/101 85 172/181/182 184 169/187 91/100/101 85
Print Book
No Choice
Revenue
Cost
Cost % of Print Price
Profit
1 2 3 4 7 1 2 3 4 6 1 2 3 4 6 1 2 3 4 6 1 2 3 5 6 1 2 3 5 6
0.365 0.479 0.407 0.400 0.514 0.365 0.407 0.479 0.400 0.514 0.365 0.407 0.479 0.426 0.514 0.365 0.407 0.426 0.479 0.514 0.365 0.407 0.426 0.479 0.514 0.365 0.407 0.426 0.479 0.514
0.217 0.190 0.224 0.250 0.192 0.217 0.224 0.190 0.250 0.192 0.217 0.224 0.190 0.226 0.192 0.217 0.224 0.226 0.190 0.192 0.217 0.224 0.226 0.190 0.192 0.217 0.224 0.226 0.190 0.192
21.748 21.764 20.311 21.153 20.818 21.748 20.311 21.764 21.153 20.818 21.748 20.311 21.764 19.707 20.818 21.748 20.311 19.707 21.764 20.818 21.748 20.311 19.707 21.764 20.818 21.748 20.311 19.707 21.764 20.818
5.503 6.395 5.197 6.055 6.252 6.879 6.497 7.994 7.568 7.815 8.255 7.796 9.593 7.606 9.378 9.631 9.095 8.873 11.191 10.941 11.007 10.394 10.141 12.790 12.504 12.38 11.69 11.41 14.389 14.07
40.00%
16.244 15.369 15.113 15.099 14.566 14.868 13.814 13.770 13.585 13.003 13.493 12.515 12.172 12.101 11.440 12.117 11.215 10.833 10.573 9.877 10.741 9.916 9.566 8.974 8.314 9.365 8.617 8.298 7.375 6.751
50.00%
60.00%
70.00%
80.00%
90.00%
Market Scenario #
Print Book PDF Book Print Book PDF Book Low Low Low Low Low Medium Medium Low High Low Medium Low
85 Medium 91/100/101 Medium 169/187 High 172/181/182 High 184 High 208 High
122
Table 15: Newspaper Category – Market Simulation Results (Incomplete Product Line)
Market Share Online Print WSJ No WSJ and Choice Online WSJ 0.557 0.498 0.576 0.423 0.515 0.581 0.576 0.557 0.521 0.515 0.581 0.576 0.557 0.521 0.515 0.581 0.576 0.521 0.557 0.515 0.581 0.655 0.576 0.704 0.643 0.704 0.655 0.581 0.685 0.643 0.214 0.237 0.256 0.263 0.283 0.277 0.256 0.214 0.306 0.283 0.277 0.256 0.214 0.306 0.283 0.277 0.256 0.306 0.214 0.283 0.277 0.162 0.256 0.078 0.134 0.078 0.162 0.277 0.067 0.134 0.085 0.111 0.104 0.147 0.134 0.114 0.104 0.085 0.145 0.134 0.114 0.104 0.085 0.145 0.134 0.114 0.104 0.145 0.085 0.134 0.114 0.138 0.104 0.150 0.120 0.150 0.138 0.114 0.125 0.120
Rank
Market Scenario # 19 28 91/100/101 55 127 172/181/182 91/100/101 19 208 127 172/181/182 91/100/101 19 208 127 172/181/182 91/100/101 208 19 127 172/181/182 184 91/100/101 178 85 178 184 172/181/182 97/106 85
Print WSJ
Revenue
Cost
Cost % of Print Price
Profit
1 2 3 4 6 1 2 3 4 5 1 2 3 4 5 1 2 3 4 6 1 2 3 4 5 1 2 3 4 5
0.143 0.154 0.064 0.167 0.068 0.028 0.064 0.143 0.028 0.068 0.028 0.064 0.143 0.028 0.068 0.028 0.064 0.028 0.143 0.068 0.028 0.046 0.064 0.068 0.102 0.068 0.046 0.028 0.123 0.102 Print WSJ
54.129 55.407 52.320 56.666 53.445 51.609 52.320 54.129 52.633 53.445 51.609 52.320 54.129 52.633 53.445 51.609 52.320 52.633 54.129 53.445 51.609 43.523 52.320 38.369 45.145 38.369 43.523 51.609 41.307 45.145
14.216 15.560 12.737 17.099 13.975 15.170 15.921 17.771 16.646 17.469 18.204 19.105 21.325 19.976 20.963 21.238 22.290 23.305 24.879 24.456 24.272 16.484 25.474 11.615 18.805 13.07 18.54 27.31 17.01 21.16 Print WSJ
40.00%
39.912 39.847 39.583 39.567 39.469 36.439 36.399 36.358 35.987 35.976 33.405 33.215 32.804 32.657 32.482 30.371 30.030 29.328 29.250 28.988 27.337 27.039 26.846 26.754 26.340 25.30 24.98 24.30 24.29 23.99 Print WSJ Online WSJ Low Low High Medium Low
50.00%
60.00%
70.00%
80.00%
90.00%
Market Scenario #
Online Print WSJ WSJ Online WSJ
Market Scenario #
Online WSJ
19 Low Low Low 28 Low Medium Low 55 Low High Low 85 Medium Low Medium 91/100/101 Medium Low Low 97/106 Medium Low High
127 Medium Medium 172/181/182 High Low 178 High Low 184 High Low 208 High Medium
123
Table 16: Book Category – Market Simulation Results (Incomplete Product Line, No Bundle Discount)
Rank Market Scenario # 163-189 82-108 1-27 28-54 109-135 163-189 82-108 1-27 190-216 109-135 163-189 82-108 1-27 190-216 109-135 163-189 82-108 190-216 1-27 109-135 163-189 82-108 190-216 1-27 109-135 163-189 82-108 190-216 1-27 109-135 Print Book Market Share PDF Print Book No Book and Choice PDF Book 0.337 0.272 0.218 0.163 0.200 0.337 0.272 0.218 0.252 0.200 0.337 0.272 0.218 0.252 0.200 0.337 0.272 0.252 0.218 0.200 0.337 0.272 0.252 0.218 0.200 0.337 0.272 0.252 0.218 0.200 0.010 0.013 0.015 0.010 0.008 0.010 0.013 0.015 0.007 0.008 0.010 0.013 0.015 0.007 0.008 0.010 0.013 0.007 0.015 0.008 0.010 0.013 0.007 0.015 0.008 0.010 0.013 0.007 0.015 0.008 0.226 0.192 0.167 0.183 0.216 0.226 0.192 0.167 0.263 0.216 0.226 0.192 0.167 0.263 0.216 0.226 0.192 0.263 0.167 0.216 0.226 0.192 0.263 0.167 0.216 0.226 0.192 0.263 0.167 0.216 Revenue Cost Cost % of Print Price Profit
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
0.427 0.523 0.600 0.645 0.575 0.427 0.523 0.600 0.478 0.575 0.427 0.523 0.600 0.478 0.575 0.427 0.523 0.478 0.600 0.575 0.427 0.523 0.478 0.600 0.575 0.427 0.523 0.478 0.600 0.575
19.686 20.593 21.281 20.776 19.913 19.686 20.593 21.281 18.708 19.913 19.686 20.593 21.281 18.708 19.913 19.686 20.593 18.708 21.281 19.913 19.686 20.593 18.708 21.281 19.913 19.686 20.593 18.708 21.281 19.913
5.066 6.215 7.138 7.591 6.770 6.332 7.769 8.923 7.036 8.463 7.599 9.323 10.707 8.443 10.156 8.865 10.876 9.850 12.492 11.848 10.132 12.430 11.257 14.276 13.541 11.40 13.98 12.66 16.061 15.23
40.00%
14.620 14.378 14.143 13.185 13.142 13.354 12.824 12.359 11.672 11.450 12.087 11.270 10.574 10.265 9.757 10.821 9.716 8.858 8.790 8.064 9.554 8.162 7.451 7.005 6.372 8.288 6.609 6.044 5.221 4.679
50.00%
60.00%
70.00%
80.00%
90.00%
Market Scenario #
Print Book PDF Book Print Book PDF Book Low Medium Low Medium Low Medium Sum Sum Sum Sum Sum Sum
1-27 Low 28-54 Low 82-108 Medium 109-135 Medium 163-189 High 190-216 High
124
Table 17: Newspaper Category – Market Simulation Results (Incomplete Product Line, No Bundle Discount)
Rank Market Scenario # 1-27 28-54 55-81 82-108 109-135 1-27 28-54 82-108 55-81 109-135 1-27 82-108 28-54 163-189 109-135 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 82-108 1-27 163-189 109-135 28-54 163-189 82-108 1-27 190-216 109-135 163-189 82-108 190-216 109-135 1-27 Print WSJ Market Share Online Print WSJ No WSJ and Choice Online WSJ 0.611 0.562 0.494 0.676 0.627 0.611 0.562 0.676 0.494 0.627 0.611 0.676 0.562 0.719 0.627 0.676 0.611 0.719 0.627 0.562 0.719 0.676 0.611 0.671 0.627 0.719 0.676 0.671 0.627 0.611 0.112 0.102 0.090 0.081 0.076 0.112 0.102 0.081 0.090 0.076 0.112 0.081 0.102 0.056 0.076 0.081 0.112 0.056 0.076 0.102 0.056 0.081 0.112 0.053 0.076 0.056 0.081 0.053 0.076 0.112 0.093 0.125 0.169 0.125 0.164 0.093 0.125 0.125 0.169 0.164 0.093 0.125 0.125 0.151 0.164 0.125 0.093 0.151 0.164 0.125 0.151 0.125 0.093 0.196 0.164 0.151 0.125 0.196 0.164 0.093 Revenue Cost Cost % of Print Price Profit
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
0.184 0.212 0.247 0.119 0.134 0.184 0.212 0.119 0.247 0.134 0.184 0.119 0.212 0.074 0.134 0.119 0.184 0.074 0.134 0.212 0.074 0.119 0.184 0.081 0.134 0.074 0.119 0.081 0.134 0.184 Print WSJ
48.721 48.400 47.887 42.070 41.150 48.721 48.400 42.070 47.887 41.150 48.721 42.070 48.400 37.078 41.150 42.070 48.721 37.078 41.150 48.400 37.078 42.070 48.721 35.696 41.150 37.078 42.070 35.696 41.150 48.721
11.754 12.478 13.413 7.934 8.330 14.693 15.598 9.918 16.766 10.413 17.631 11.902 18.717 7.750 12.496 13.885 20.570 9.042 14.578 21.837 10.334 15.869 23.508 10.644 16.661 11.63 17.85 11.97 18.74 26.45 Print WSJ
40.00%
36.967 35.922 34.475 34.136 32.820 34.029 32.802 32.152 31.121 30.737 31.090 30.169 29.683 29.328 28.655 28.185 28.152 28.037 26.572 26.563 26.745 26.201 25.213 25.052 24.490 25.45 24.22 23.72 22.41 22.27 Print WSJ Online WSJ Sum Sum Sum
50.00%
60.00%
70.00%
80.00%
90.00%
Market Scenario # 1-27 28-54 55-81 82-108
Online Print WSJ WSJ Online WSJ Sum Sum Sum Sum
Market Scenario #
Online WSJ
Low Low Low Medium Low High Medium Low
109-135 Medium Medium 163-189 High Low 190-216 High Medium
125
Table 18: Book Category - Forms’ Attribute Perceptions.
Variable Price Print Book (PP) Electronic Book (EP) Print Book & Electronic Book Bundle (PPEP) Electronic Book Unit (EU) Print Book & Electronic Book Unit Bundle (PPEU) Print Form Index (PI) PDF Form Index (EI_1) PDF Form Index (EI_2) Difference Index 1 (DI1_1): Bundle vs. Superior Difference Index 1 (DI1_2): Bundle vs. Superior Difference Index 2 (DI2): Print vs. Electronic Price Print Book (PP) Electronic Book (EP) Print Book & Electronic Book Bundle (PPEP) Number of observations Iterations completed Log likelihood function Restricted log likelihood Chi squared Degrees of freedom Prob[ChiSqd > value] Coefficient SE b/St.Er. P[|Z|>z] Random parameters in utility functions -0.058 0.017 -3.392 0.001 4.339 0.615 7.055 0.000 3.743 0.579 6.464 0.000 -5.861 1.775 -3.302 0.001 Nonrandom parameters in utility functions 1.861 0.291 6.392 0.000 1.768 0.671 2.635 0.008 0.029 0.010 2.840 0.005 0.029 0.009 3.357 0.001 0.013 0.003 4.036 0.000 0.107 0.033 3.232 0.001 0.018 0.013 1.319 0.187 0.023 0.006 3.612 0.000 Derived st. dev. of parameter distributions 0.091 0.010 9.107 0.000 2.289 0.254 8.998 0.000 2.291 0.253 9.068 0.000 3.461 0.852 4.063 0.000 4176 (87 respondents x 16 choices) 68 -917.41 -2429.63 3024.44 16 0.000
126
Table 19: Newspaper Category - Forms’ Attribute Perceptions.
Variable Price Print Newspaper (PP) Electronic Newspaper (EP) Print & Electronic Newspaper Bundle (PPEP) Electronic Newspaper Unit (EU) Print Newspaper & Electronic N. Unit Bundle (PPEU) Print Form Index (PI) PDF Form Index (EI_1) PDF Form Index (EI_2) Difference Index 1 (DI1_1): Bundle vs. Superior Difference Index 1 (DI1_2): Bundle vs. Superior Difference Index 2 (DI2): Print vs. Electronic Price Number of observations Iterations completed Log likelihood function Restricted log likelihood Chi squared Degrees of freedom Prob[ChiSqd > value]
Coefficient
SE
b/St.Er.
P[|Z|>z] 0.000 0.000 0.000 0.000 0.819 0.007 0.166 0.000 0.005 0.402 0.855 0.783
Random parameters in utility functions -0.042 0.007 -5.868 Nonrandom parameters in utility functions 2.747 0.661 4.159 2.586 0.292 8.867 3.148 0.847 3.718 0.042 0.186 0.229 1.994 0.744 2.680 0.008 0.006 1.386 0.011 0.003 3.645 0.007 0.003 2.822 -0.011 0.013 -0.837 0.003 0.014 0.183 -0.001 0.004 -0.276
Derived st. dev. of parameter distributions 0.029 0.003 8.666 0.000 4176 (87 respondents x 16 choices) 18 -1068.52 -2433.21 2729.37 13 0.000
127
FIGURES Figure 1: Study 2 Results (significant effects only) A. USAGE by DISCOUNT
Mean LN of Points Allocated to the Bundle
Mean LN of Points Allocated to the Bundle
Category: MARGARINE/COFFEE
0.0
Category: BOOK/NEWSPAPER
-2.4 -2.6 -2.8 -3.0 -3.2 -3.4 -3.6 -3.8 -4.0 0 1
-.5
DIFFERENT USAGES
-1.0
DIFFERENT USAGES
-1.5
-2.0
-2.5
SAME USAGES
SAME USAGES
-3.0 0 1
NO DISCOUNT
DISCOUNT
NO DISCOUNT
DISCOUNT
B. DISCOUNT by RELATIVE PRICE
Mean LN of Points Allocated to the Bundle
Category: BOOK
-2.6 -2.8 -3.0 -3.2 -3.4 -3.6 -3.8
DISCOUNT
NO DISCOUNT
-4.0 -4.2 0 1
PRICE PRINT = PRICE ELECT
PRICE PRINT > PRICE ELECT
128
Figure 2: Optimal Price and Total and Marginal Revenue from Electronic Book Chapters.
VC VH PC MR VL ½ NC NC QC TR
129
Figure 3: Optimal Price and Total and Marginal Revenue from Full Electronic Book.
VEB VH PB MR VL VPB k TR
130
REFERENCES March 2005 Online Paid Content U.S. Market Spending Report (2005), http://www.online-publishers.org/pdf/paid_content_report_030905.pdf. Aaker, David and Kevin Keller (1990), “Consumer Evaluations of Brand Extensions,” Journal of Marketing, 54 (January): 27-41. Adams, William J. and Janet L. Yellen (1976), “Commodity Bundling and the Burden of Monopoly,” Quarterly Journal of Economics, 90 (August): 475-98. Bakos, Yannis and Erik Brynjolfsson (2000), “Bundling and Competition on the Internet: Aggregation Strategies for Information Goods,” Marketing Science, 19 (1): 63-82. Bakos, Yannis and Erik Brynjolfsson (1999), “Bundling Information Goods: Pricing, Profits and Efficiency”, Management Science, 45 (12), 1613-1630. Betancourt, Roger R. (2005), The Economics of Retailing and Distribution, Edward Elgar Publishing, Chapter 6. Carroll, J. Douglas and Paul Green (1995), “Psychometric Methods in Marketing Research: Part I, Conjoint Analysis”, Journal of Marketing Research, 32 (November), 385-391. Chakravarti, Dipankar, Rajan Krish, Paul Pallab and Joydeep Srivastava (2002), “Partitioned Presentation of Multi-Component Bundle Prices: Evaluation, Choice and Underlying Processing Effects,” Journal of Consumer Psychology, 12 (3): 215 - 29. Chernev, Alexander (2005), “Feature Complementarity and Assortment in Choice”, Working paper, Kellogg School of Management, Northwestern University, Evanston, IL. Dhar, Ravi (1997), “Consumer Preference for a No-Choice Option,” Journal of Consumer Research, 24 (September), 215-231. Estelami, Hooman (1990), “Consumer Savings in Complementary Product Bundles,” Journal of Marketing Theory and Practice, Summer: 107-14. Fennell, Geraldine G. (1978), “Consumers’ Perceptions of the Product-Use Situations,” Journal of Marketing, 42 (April): 39-47. Gaeth, Gary, Irwin Levin, Goutam Chakraborty and Aron M. Levin (1990), “Consumer Evaluation of Multi-Product Bundles: An Information Integration Analysis,” Marketing Letters, 2: 47-57. Geng, Xianjun, Maxwell B. Stinchcombe and Andrew B. Whinston (2005), “Bundling Information Goods of Decreasing Value,” ? forthcoming in Management Science. Green, Paul and V. Srinivasan (1990), “Conjoint Analysis in Marketing: New 131
Developments with Implications for Research and Practice”, Journal of Marketing, 54 (4), 3-19. Guadagni, Peter M. and John D.C. Little (1983), “A Logit Model of Brand Choice Calibrated on Scanner Panel Data, Marketing Science, 2 (3), 203-239. Guiltinan, Joseph P. (1987), “The Price Bundling of Services: A Normative Framework,” Journal of Marketing, 51 (April): 74-85. Hanson, Ward and R. Kipp Martin (1990), “Optimal Price Bundling”, Management Science, 36 (2), 155-174. Harlam, Bari A., Aradna Krishna, Donald R. Lehmann and Carl Mela (1995), “Impact of Bundle Type, Price Framing and Familiarity on Purchase Intention for the Bundle”, Journal of Business Research, 33, 57-66. Henderson, James M. and Richard E. Quandt (1958), Microeconomic Theory: A Mathematical Approach, New York: McGraw-Hill Book Company, p. 29. Jain, Dipak C, Vilcassim, Naufel J, and Pradeep K. Chintagunta (1994), “A RandomCoefficients Logit Brand-Choice Model Applied to Panel Data,” Journal of Business & Economic Statistics, 12 (July), 317-328. Jedidi, Kamel, Sharan Jagpal and Puneet Manchanda (2003), “Measuring Heterogeneous Reservation Prices for Product Bundles”, Marketing Science, 22 (1), 107-130. Johnson, Michael D., Andreas Herrmann and Hans H. Bauer (1999), “The Effects of Price Bundling on Consumer Evaluations of Product Offerings,” International Journal of Research in Marketing, 16: 129-42. Kahneman, Daniel and Amos Tversky (1979), “Prospect Theory: An Analysis of Decision Under Risk,” Econometrica, 47: 163-91. Kenney, Roy W. and Benjamin Klein (1983), “The Economics of Block Booking,” Journal of Law and Economics, 26: 497-540. Lattin, James M. and Leigh McAlister (1985), “Using a Variety-Seeking Model to Identify Substitute and Complementary Relationships Among Competing Products,” Journal of Marketing Research, 22 (August), 330-339. MacKenzie, Scott B. and Richard J. Lutz (1989) “An Empirical Examination of the Structural Antecedents of Attitude Toward the Ad in an Advertising Pretesting Context,” Journal of Marketing, 53 (April): 48-65. Macklin, Theodore (1922), Efficient Marketing for Agriculture, The Mackmillian Co. New York: 25.
132
McAfee, R. Preston, John McMillan and Michael Whinston (1989), “Multiproduct Monopoly, Commodity Bundling, and Correlations of Values,” Quarterly Journal of Economics, 104 (2): 371-83. Mick, David Glen (1992), “Levels of Subjective Comprehension in Advertising Processing and their Relations to Ad Perceptions, Attitudes and Memory,” Journal of Consumer Research, 18 (March): 411-24. Ratneshwar, S. and Allan D. Shocker (1991), “Substitution in Use and the Role of Usage Context in Product Category Structures,” Journal of Marketing Research, 28 (August): 281–95. Sarvary, Miklos and Philip M. Parker (1997), “Marketing Information: A Competitive Analysis,” Marketing Science, 16 (1): 24-38. Schmalensee, Richard (1984), “Gaussian Demand and Commodity Bundling,” Journal of Business, 57 (1): S211-30. Simonin, Bernard L. and Julie A. Ruth (1995), “Bundling as a Strategy for New Product Introduction: Effects on Consumers’ Reservation Prices for the Bundle, the New Product, and Its Tie-in,” Journal of Business Research, 33: 219-230. Srivastava, Rajendra K., Mark I. Alpert and Allan D. Shocker (1984), “A CustomerOriented Approach for Determining Market Structures,” Journal of Marketing, 48 (Spring): 32-45. Srivastava, Rajendra K., Robert P. Leone and Allan D. Shocker (1981), “Market Structure Analysis: Hierarchical Clustering of Products Based on Substitution-in-Use,” Journal of Marketing, 45 (Summer): 38-48. Stefflre, Volney R. (1971), New Products and New Enterprises: A Report of an Experiment in Applied Social Science, Irvine, CA: University of California. Stigler, George J. (1963), “United States v. Loew’s, Inc: A Note on Block Booking,” Supreme Court Review, 153 (1963): 152-57. Stremersch, Stefan and Gerard J. Tellis (2002), “Strategic Bundling of Products and Prices: A New Synthesis for Marketing,” Journal of Marketing, 66 (January): 55-72. Telser, L. G. (1979), “A Theory of Monopoly of Complementary Goods,” Journal of Business, 52 (2): 211-30. Thaler, Richard (1985), “Mental Accounting and Consumer Choice,” Marketing Science, 4 (3): 199-214. Venkatesh, R. and Rabikar Chatterjee (2005), “Bundling, Unbundling and Pricing of Hybrid Products: The Case of Magazine Content,” Working Paper, University of 133
Pittsburg, PA. Venkatesh, R. and Vijay Mahajan (1993), “A Probabilistic Approach to Pricing a Bundle of Products or Services”, Journal of Marketing Research, 30 (November), 494-508. Venkatesh, R. and Wagner Kamakura (2003), “Optimal Bundling and Pricing under a Monopoly: Contrasting Complements and Substitutes from Independently Valued Products”, Journal of Business, 76 (2), 211-231. Wilson, Lynn O., Allen M. Weiss and George John (1990), “Unbundling of Industrial Systems”, Journal of Marketing Research, 27 (May), 123-138. Yadav, Manjit S. (1995), “Bundle Evaluation in Different Marketing Segments: The Effects of Discount Framing and Buyers’ Preference Heterogeneity,” Journal of the Academy of Marketing Science, 23 (3): 206-15. Yadav, Manjit S. (1994), “How Buyers Evaluate Product Bundles: A Model of Anchoring and Adjustment,” Journal of Consumer Research, 21 (September): 342-53. Yadav, Manjit S. and Kent B. Monroe (1993), “How Buyers Perceive Savings in a Bundle Price: An Examination of a Bundle’s Transaction Value,” Journal of Marketing Research, 30 (August): 350-58. Wansink, Brian and Michael L. Ray (1996), “Advertising Strategies to Increase Usage Frequency,” Journal of Marketing, 60 (January): 31-46. www.businesswire.com (2004), “U.S. Consumer Spending for Online Content Totals $853 Million in the First Half of 2004, According to Online Publishers Association Report”, Nov. 15.
134
doc_327658533.pdf