netrashetty
Netra Shetty
Iberdrola USA (formerly Energy East) is a Fortune 500 company that serves 2.9 million public utility customers with energy, primarily electricity and natural gas.
Females find niche communities more valuable than sites like Facebook
Women are well known for their high usage of social media sites, where they make an attractive target. They like to socialize with friends and family, share information and give advice.
But social media is about more than just traditional social networking sites, and according to research from women’s community site iVillage in conjunction with Burke Research, Facebook may not be the best place for marketers to find women.
Asked about their attitudes toward online communities designed for women, like iVillage itself (which has a vested interest in such research), CafeMom or BabyCenter, vs. web portals and social networking sites, female internet users preferred online communities for a variety of reasons.
They trusted the sites most, especially for information on brands and products, and they felt the people on women’s sites were most qualified to understand their needs.
Social networks like Facebook, by contrast, were seen as less useful, less trustworthy and more of a waste of time.
Women appear to be compartmentalizing their social media habits on different types of sites. Earlier research from iVillage and BlogHer found that women thought social networks were best for keeping up with friends and family, as well as entertainment.
Online communities did not serve these functions as well.
But they did help respondents find out about new products and make purchase decisions.
A March 2010 survey from Yahoo! found a similar pattern: Women were more likely to find and share brand information on women’s lifestyle or online community sites than social networks, and believed communities were more likely to provide them with good information on this front.
which respondents are experienced with eBay computer auctions (and thus similar to the
auction participants) and which respondents are inexperienced. Because survey measures
are prone to bias, I exploit a mixture of respondents, some experienced with eBay computer
auctions and some not, to correct for any bias between the mean and standard deviation of the
survey responses and the true common value and dispersion of information facing the auction
participants. The use of inexperienced respondents increases the pool of potential survey
respondents and permits the survey to be implemented quickly. The use of experienced
respondents allows me to correct for potential bias from using more noisy inexperienced
responses.
Section 2 reviews the motivation for a survey based measure of information dispersion
in auctions. Section 3 presents the auction data employed and the survey design. Section 4 analyzes the success of the survey as a correlated measure. Section 5 presents the
background data collected in the survey and its implications for correcting for survey bias.
Section 6 presents the bias correction procedure. Section 7 examines the di§erence between
results from the survey-based measures and alternative hedonic regression methods. Section
8 concludes this paper.
2 Motivation for Survey Data
There are several reasons why a researcher might want to collect survey data to augment data
from commercial markets. A researcher must often control for the value of the item when
determining the e§ect of other regressors on price. Empirical work in general has employed
hedonic regression of price on product characteristics to control for the value of the item.
A large number of hedonic characteristics will demand a large number of observations for
identiÖcation. A survey allows respondents to áexibly assess the value of a large number of
characteristics even in a small sample of items. Alternatively, empirical work has restricted
itself to examining identical items to control for item values. However, using identical items
may result in a sample that is either too small or exhibits too little variation in the regressors
of interest. Survey respondents can handle di§erences in item characteristics, allowing the
researcher to include more heterogeneous items in order to ensure a su¢ ciently large sample
and su¢ cient variation in the regressors of interest. The survey measure of value can be
constructed to be independent of the price, as long as survey respondents are not shown
price information. Thus, the survey data provides an exogenous regressor that controls for
the value of the item.
The survey measures can be designed to be independent of other regressors as well. For
example, in my survey design, details about the seller, the bidders, and the bids in the
auction are omitted. This creates several advantages for using survey data over methods
that recover the signals from the observed distribution of bids. My survey responses are
functions of the product description only. Reserve prices and opening bids that appear in
many online auctions would truncate the observed distribution of bids. My survey responses
are not ináuenced by the number of bidders in the auction nor by the reputation of the
seller. By construction, the survey data is not a function of bidding behavior in the actual
4auction. The independence of my survey data from the auction data allows me to test
between di§erent types of bidding behavior and separately identify the e§ect of reputation
from the dispersion of information or preferences and from other determinants of price.
Empirical work has also proxied for the common value using blue book values. However,
blue book values and hedonic methods cannot take into account any anomalies in the products. For example, a computer that was being sold on eBay was described as working but
locked: the password had been lost, so there was no way to logon to the computer. Hedonic
estimation or the use of a blue book value would treat this anomaly as unobservable to the
econometrician, but such anomalies may be important determinants of the price of the auctioned item. They might drive the number of bidders that enter the auction. The number
of bidders is often included in the price model as a regressor. This presents an endogeneity
problem for estimation, since the number of bidders is now correlated with the error term.
In contrast, the human readersíestimates do reáect values that are more closely tied to the
semantics of the product description than any hedonics-based measure or book value. By
having people read the auction descriptions and respond with their value for the item, I am
able to capture the idiosyncrasies of each item as well as its hedonic characteristics.
Variation in the survey responses also generates information that does not exist in onedimensional measures from hedonic analysis or book values. The standard deviations over
the responses in each auction serve as a measures of the dispersion of private information
signals in the auctions. They provide a measure of the survey respondentsícertainty about
their valuations.
This extra information about the unobservable private signals is particularly useful in
testing auction theory. Often, the only information available from auctions is the number of
bidders, observed bids, and product characteristics. In a limited number of cases, ex post
values of the auctioned item are available. The literature on nonparametric identiÖcation
has shown that given this observable data, the distribution of private signals is just identiÖed
assuming a private values setting but underidentiÖed in a common values setting without
further parametric assumptions.
1 As a result, tests of information structure in an auction
(whether auctions are private value or common value) and bidding behavior (whether or
not bidders play Nash equilibrium strategies) are rarely conducted jointly. By measuring
dispersion, identifying power is not expended on recovering the underlying distribution of
information signals, so a joint test of an auctionís information structure and bidding behavior
is possible.
In sum, the ability to design the independence of survey data from other regressors, the
ability to exploit human assessment of information, and information provided by the second
moment of survey data make it an appealing source of information to complement market
data, in particular for auctions
Females find niche communities more valuable than sites like Facebook
Women are well known for their high usage of social media sites, where they make an attractive target. They like to socialize with friends and family, share information and give advice.
But social media is about more than just traditional social networking sites, and according to research from women’s community site iVillage in conjunction with Burke Research, Facebook may not be the best place for marketers to find women.
Asked about their attitudes toward online communities designed for women, like iVillage itself (which has a vested interest in such research), CafeMom or BabyCenter, vs. web portals and social networking sites, female internet users preferred online communities for a variety of reasons.
They trusted the sites most, especially for information on brands and products, and they felt the people on women’s sites were most qualified to understand their needs.
Social networks like Facebook, by contrast, were seen as less useful, less trustworthy and more of a waste of time.
Women appear to be compartmentalizing their social media habits on different types of sites. Earlier research from iVillage and BlogHer found that women thought social networks were best for keeping up with friends and family, as well as entertainment.
Online communities did not serve these functions as well.
But they did help respondents find out about new products and make purchase decisions.
A March 2010 survey from Yahoo! found a similar pattern: Women were more likely to find and share brand information on women’s lifestyle or online community sites than social networks, and believed communities were more likely to provide them with good information on this front.
which respondents are experienced with eBay computer auctions (and thus similar to the
auction participants) and which respondents are inexperienced. Because survey measures
are prone to bias, I exploit a mixture of respondents, some experienced with eBay computer
auctions and some not, to correct for any bias between the mean and standard deviation of the
survey responses and the true common value and dispersion of information facing the auction
participants. The use of inexperienced respondents increases the pool of potential survey
respondents and permits the survey to be implemented quickly. The use of experienced
respondents allows me to correct for potential bias from using more noisy inexperienced
responses.
Section 2 reviews the motivation for a survey based measure of information dispersion
in auctions. Section 3 presents the auction data employed and the survey design. Section 4 analyzes the success of the survey as a correlated measure. Section 5 presents the
background data collected in the survey and its implications for correcting for survey bias.
Section 6 presents the bias correction procedure. Section 7 examines the di§erence between
results from the survey-based measures and alternative hedonic regression methods. Section
8 concludes this paper.
2 Motivation for Survey Data
There are several reasons why a researcher might want to collect survey data to augment data
from commercial markets. A researcher must often control for the value of the item when
determining the e§ect of other regressors on price. Empirical work in general has employed
hedonic regression of price on product characteristics to control for the value of the item.
A large number of hedonic characteristics will demand a large number of observations for
identiÖcation. A survey allows respondents to áexibly assess the value of a large number of
characteristics even in a small sample of items. Alternatively, empirical work has restricted
itself to examining identical items to control for item values. However, using identical items
may result in a sample that is either too small or exhibits too little variation in the regressors
of interest. Survey respondents can handle di§erences in item characteristics, allowing the
researcher to include more heterogeneous items in order to ensure a su¢ ciently large sample
and su¢ cient variation in the regressors of interest. The survey measure of value can be
constructed to be independent of the price, as long as survey respondents are not shown
price information. Thus, the survey data provides an exogenous regressor that controls for
the value of the item.
The survey measures can be designed to be independent of other regressors as well. For
example, in my survey design, details about the seller, the bidders, and the bids in the
auction are omitted. This creates several advantages for using survey data over methods
that recover the signals from the observed distribution of bids. My survey responses are
functions of the product description only. Reserve prices and opening bids that appear in
many online auctions would truncate the observed distribution of bids. My survey responses
are not ináuenced by the number of bidders in the auction nor by the reputation of the
seller. By construction, the survey data is not a function of bidding behavior in the actual
4auction. The independence of my survey data from the auction data allows me to test
between di§erent types of bidding behavior and separately identify the e§ect of reputation
from the dispersion of information or preferences and from other determinants of price.
Empirical work has also proxied for the common value using blue book values. However,
blue book values and hedonic methods cannot take into account any anomalies in the products. For example, a computer that was being sold on eBay was described as working but
locked: the password had been lost, so there was no way to logon to the computer. Hedonic
estimation or the use of a blue book value would treat this anomaly as unobservable to the
econometrician, but such anomalies may be important determinants of the price of the auctioned item. They might drive the number of bidders that enter the auction. The number
of bidders is often included in the price model as a regressor. This presents an endogeneity
problem for estimation, since the number of bidders is now correlated with the error term.
In contrast, the human readersíestimates do reáect values that are more closely tied to the
semantics of the product description than any hedonics-based measure or book value. By
having people read the auction descriptions and respond with their value for the item, I am
able to capture the idiosyncrasies of each item as well as its hedonic characteristics.
Variation in the survey responses also generates information that does not exist in onedimensional measures from hedonic analysis or book values. The standard deviations over
the responses in each auction serve as a measures of the dispersion of private information
signals in the auctions. They provide a measure of the survey respondentsícertainty about
their valuations.
This extra information about the unobservable private signals is particularly useful in
testing auction theory. Often, the only information available from auctions is the number of
bidders, observed bids, and product characteristics. In a limited number of cases, ex post
values of the auctioned item are available. The literature on nonparametric identiÖcation
has shown that given this observable data, the distribution of private signals is just identiÖed
assuming a private values setting but underidentiÖed in a common values setting without
further parametric assumptions.
1 As a result, tests of information structure in an auction
(whether auctions are private value or common value) and bidding behavior (whether or
not bidders play Nash equilibrium strategies) are rarely conducted jointly. By measuring
dispersion, identifying power is not expended on recovering the underlying distribution of
information signals, so a joint test of an auctionís information structure and bidding behavior
is possible.
In sum, the ability to design the independence of survey data from other regressors, the
ability to exploit human assessment of information, and information provided by the second
moment of survey data make it an appealing source of information to complement market
data, in particular for auctions
Last edited: