Dissertation report on Volatility Change and Private and Government Investment

Description
In economics, investment is related to saving and deferring consumption. Investment is involved in many areas of the economy, such as business management and finance whether for households, firms, or governments.

ABSTRACT

Title:

THREE ESSAYS IN VOLAITLITY CHANGE AND PRIVATE AND GOVERNMENT INVESTMENT Namsuk Kim, Doctor of Philosophy, 2005

Directed By:

Prof. John Haltiwanger, Department of Economics

Studies of the volatility of the U.S. economy suggest a noticeable change in mid 1980s. There is some empirical evidence that the aggregate volatility of the U.S. economy has been decreasing over time. The response of firms to the change of economic volatility and economic fluctuation has been studied in terms of many margins a firm can adjust –capital, labor, capacity, material, etc. However, we have not studied the most important margin – the product. My dissertation studies the effect of profit volatility on the firm/plant level product diversification. Chapter 2 profiles diversification and shows that there is a downward trend of aggregate diversification in many industries. Cyclicality of diversification is not clear at the aggregate or industry level. Firms change their diversification very frequently and very differently from one another. Chapter 3 verifies the trend of volatility at the aggregate, sectoral, and firm level and studies the relationship between diversification and volatility at the firm level. Firm level diversification decreases as the aggregate, sectoral and idiosyncratic volatility decreases. Research on the volatility change is concentrated on recent U.S. history. However, new data allow us to study events of significant volatility change in early 19th century. Chapter 4 of my dissertation studies the causes and effects of the volatility change in early 19th century in U.S. and U.K.

THREE ESSAYS IN VOLAITLITY CHANGE AND PRIVATE AND GOVERNMENT INVESTMENT

By

Namsuk Kim

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 John Haltiwanger, Chair Professor John Shea Professor John Wallis Professor Gordon Phillips Doctor Ronald Jarmin

Dedication

To my love, Tai Hwa

ii

Acknowledgements
The research in Chapter 2 and Chapter 3 in this paper was conducted while the author was a research associate at the Center for Economic Studies (CES) in Bureau of the Census. The research project is approved and sponsored by CES, Research Data Center Project #0265. I thank Lucia Foster, Ron Jarmin, and CES staffs for supporting my research. This paper has been screened by CES to ensure that no confidential data are revealed. I thank Richard Sylla, Jack Wilson and Robert Wright for the data used in Chapter 4. I was involved in their project gathering and constructing a financial market database and compiled the raw data collected by them. Professor Harry Kelejian, Imgmar Prucha, John Chao at University of Maryland, and Joon Y. Park at Rice University gave me helpful comments on developing ARCH model with missing data used in Chapter 4. I thank the anonymous referees for Economic History Review for their comments. Professor John Haltiwanger and John Wallis helped me not only as advisers but as great mentors. Professor John Haltiwanger granted a research opportunity at CES and spent his valuable time serving as the adviser and chair of the committee. Professor John Wallis guided me throughout my time in University of Maryland and showed me the way to a good economist and a good teacher. In every stage of my research, Professor John Shea gave me a great deal of helpful comments on the direction and the detail of my research. Assistant Director Ron Jarmin at CES and Professor Gordon Phillips made tremendous contribution to my dissertation by reviewing drafts of my thesis. iii

I appreciate the financial support from the World Bank. Jaime Saavedra, Headuck Lee, and Gilberto Moncada provided me with a great research environment. I also thank seminar participants at University of Maryland, US Bureau of Census, Economic History Association, Cliometric Society, World Bank, Econometric Society, and Seoul National University. Research results and conclusions expressed are those of the author and do not necessarily indicate concurrence by the Bureau of the Census, the CES, or the World Bank.

iv

Table of Contents
Dedication ..................................................................................................................... ii Acknowledgements...................................................................................................... iii Table of Contents.......................................................................................................... v List of Tables ............................................................................................................... vi List of Figures ............................................................................................................ viii Chapter 1: Introduction ................................................................................................. 1 Chapter 2: Product Diversification: Profile .................................................................. 4 Section 1. Introduction.............................................................................................. 4 Section 2. Data .......................................................................................................... 5 Section 3. Why Diversification? ............................................................................... 9 Section 4. Stylized Facts of Diversification............................................................ 12 4-1. Long-term Trend.......................................................................................... 13 4-2. Short-term dynamics.................................................................................... 17 Section 5. Conclusion ............................................................................................. 31 Chapter 3: Volatility Change and Diversification ...................................................... 34 Section 1. Introduction............................................................................................ 34 Section 2. Volatility and Diversification ................................................................ 35 Section 3. Stylized facts: Volatility ........................................................................ 37 Section 4. Diversification and Volatility: Estimation ............................................. 40 Section 5. Conclusion ............................................................................................. 47 Chapter 4: Volatility Change and Government Investment........................................ 49 Section 1. Introduction............................................................................................ 49 Section 2. The History ............................................................................................ 57 Section 3. Data sources and Market Integration Tests............................................ 62 Section 4. American State Bonds in London and the United States....................... 68 Pennsylvania: ...................................................................................................... 73 The Crises of 1837 and 1839: ............................................................................. 75 The Collapse of 1842:......................................................................................... 79 Section 5. Conclusion ............................................................................................. 84 Chapter 5: Conclusion................................................................................................. 86 Appendix A: Data in Chapter 2 and Chapter 3 ........................................................... 89 Appendix B: Multivariate ARCH with missing data in Chapter 4 ............................. 92 Appendix C: Additional Statistics of Annual Firm-level Product Diversification ..... 94 Reference .................................................................................................................. 113

v

List of Tables
Chapter 2 Table 1: Percentage Change of Diversification (1967-1997)…………………... Table 2: Average Establishment Level Diversification Index and Correlation with Growth of Real Shipment by 2 digit SIC Industry……………… Table 3: Average Firm Level Diversification Index and Correlation with Growth of Real Shipment by 2 digit SIC Industry …………………... Table 4: Production Diversification ………………………….………………… Table 5: Distribution (percentiles) of Firms' correlation coefficients between new/lost industry and the primary industry…………………………… Chapter 3 Table 1: Profit Volatility by Industry ………………………………………...... Table 2: Left-censored Tobit Estimation (Firm Level) ……………………....... Chapter 4 Table 1: Average Bond Yields for New York and Ohio Bonds in London and New York, and Difference in Yields ………………………………… Table 2: Amount of State Debt Outstanding on Sept 1st, 1841, Percentage of Debt authorized between 1836 and 1841, and amount authorized between 1839 and 1841……………………………………………..... Table 3: Default, Resumption and Repudiation Dates…………………………. Table 4: Bond Market Integration Test (Multivariate ARCH with Two Equations)………………………………………………………. Table 5: Stock Market Integration Test (Multivariate ARCH with Two Equations)………………………………………………………. Table 6: Bond Yield of New York and Ohio Bonds in the United States and London……………………………………………………………....... Table 7: Bond Yield in the United States and London Pennsylvania Bonds, and New York/Ohio Average Yield……………………………………..... Table 8: Bond Yields in the United States and London Illinois, Massachusetts, and Indiana Bonds…………………………………………………….. 16 20 24 21 29

38 45

56

60 61 66 67 69 71 72

Appendix Table AA1: Description of SIC Codes………………………………………..... 91 Table AC1: Average Firm Level Diversification Index………………………... 94 Table AC2: Average Firm Level Diversification Index of Single Units……….. 95 Table AC3: Average Firm Level Diversification Index of Multi Units………… 96 Table AC4: Share of Diversified Production (Rpd) …………………………… 97 Table AC5: Share of Within-plant Factor in Firm Level Diversification (Rwp).. 98 Table AC6: Annual Diversification Index Change Decomposition by POSC, NEGC, POSB and NEGD …………………………………………. 99 Table AC7: Firm Level Diversification Change Decomposed by Diversified/Specialized Plants (Multi Unit Firms) ………………... 100

vi

Table AC8: Firm Level Diversification Change Decomposed by Intensive/Extensive Components (Continuing MU Firms) ……….. 101 Table AC9: Firm Level Diversification Change Decomposition (Continuing MU Firms) ………………………………………………………… 102 Table AC10: Average Diversification Index by Firm Size Quartile (Using Total Employment) ………………………………………………... 103 Table AC11: Average Diversification Index by Firm Age Quartile…………… 104 Table AC12: Average Diversification Index by Region……………………….. 105 Table AC13: Average Diversification Index by Quintile of Share of Interplant Transfer (IPT/TVS) ……………………………………………..... 106 Table AC14: Average Diversification Index by Quartile of Share of Labor Cost (Wage/Total variable Cost) ……………………………………...... 107 Table AC15: Average Diversification Index by Quartile of Share of Nonproduction Worker Labor Cost (Non-production worker wage/Total labor cost) ……………………………………………. 108 Table AC16: Average Diversification Index by Quartile of Share of Exported Good (Value of exported good/Total value of shipment) ………… 109 Table AC17: Decade Average of Number of Industries by Number of Products (5-digit SIC) ……………………………………………………….. 110 Table AC18: Decade Average of Number of Counties Where Plants Are Located by Number of Plants……………………………………… 111 Table AC19: Evolution of Aggregate and Idiosyncratic Volatility…………….. 112

vii

List of Figures
Chapter 2 Figure 1: Diversification Indexes (1963-1982) ……………………………………. 5 Figure 2: Average Diversification Index (1967-1997, CM) ………………………. 15 Figure 3: Establishment Level Diversification Index……………………………… 18 Figure 4: Share of Single Unit Establishments (Firms) …………………………… 18 Figure 5: Firm Level Diversification Index………………………………………... 21 Figure 6: Firm Level Average Diversification Index (D1, D2, D3 and D4)………. 22 Figure 7: Changes in the Number of Plants and Products…………………………. 25 Figure 8: Share of Diversified Production (Rpd) and Share of Within-plant Diversification (Rwp) ………………………………………………….. 28 Figure 9: Positive, Negative and Net Change of Annual Firm Level Diversification………………………………………………………….. 29 Chapter 3 Figure 1: Volatility of Average Firm Level Profit Rates………………………….. 39 Figure 2: Volatility of Average Firm Level Profit Rates by Industry…………….. 39 Figure 3: Mean and Standard Deviation of Firm Level Idiosyncratic Volatility….. 39 Figure 4: Average Idiosyncratic Volatility by Size of Firm……………………….. 40 Chapter 4 Figure 1: Volatility of Average State Bond Yields…………………………………50 Figure 2: All State Bond Yields (Average in US and in London) ………………… 53 Figure 3: Difference in Ohio Bond Yields (New York minus London) …………... 53 Figure 4: Difference in Yields, NY Bonds (New York minus London) ………….. 56 Figure 5: Interest Rates (America and Britain, scaled) ……………………………. 58 Figure 6: NY, OH, and PA Bond Yields (In London and New York, scaled) ……. 70 Figure 7: Difference in Ohio Bond Yields (New York minus London) …………... 80

viii

Chapter 1: Introduction
Studies of the volatility of the U.S. economy suggest a noticeable change in the mid 1980s. There is some empirical evidence that the aggregate volatility of the U.S. economy has decreased over time.1 The volatility of real GDP growth in the United States has fallen by half since the early 1980s relative to the prior postwar experience. Not only output, but many other economic indicators show less volatility. Inflation also stabilized after the mid 1980s. Some studies have argued that an improvement in U.S. monetary policy can explain both the lower output and inflation volatility. 2 Others have attributed the decreased volatility of GDP to a reduction in the size of shocks hitting the U.S. economy-in other words, 'good luck'.3 Recent studies argue that both policy and good-luck played a role and that changes in inventory behavior stemming from improvements in information technology have played a role in reducing real output volatility.4 The causes of change in volatility have been studied, although a consensus has yet to be reached. Research on the effects of the volatility change has accumulated as well. The response of firms to changing economic volatility or economic fluctuations has been studied along many margins –capital, labor, capacity, material, etc. 5 However, the most important margin – the product – has not been studied thoroughly. Throughout the history of 20th century U.S. business, diversification was a strategic option pursued by corporate entities. High diversification was a virtue, and big conglomerates were regarded as the engine of fast growing economies. In the late 20th century, many big companies were split either by antitrust lawsuits or for strategic purposes, but we still observe
1 2

Blanchard and Simon (2001), McConnell and Perez-Quiros (2000) Clarida, Gali, and Gertler (2000) 3 Ahmed, Levin, and Wilson (2001), Blanchard and Simon (2001) 4 Stock and Watson (2002), Kahn et al (2002) 5 See Sakellaris (2000) for a survey.

1

massive mergers and acquisitions toward horizontal and/or vertical integration in many industries, such as petroleum, telecommunications, printing, and so on.6 Economists have followed the trend of multi-output production of manufacturing plants and firms, but despite theoretical advances, the variation in diversification across industry and time still remains a mystery. Although there are thousands of papers on corporate diversification, most of them focus on the diversification in the financial portfolio of the firm and its effect on productivity or the value of the firm.7 A comprehensive empirical study on product diversification is long overdue. Except for some anecdotal evidence, there are few publicly available statistics measuring the extent of establishment, firm, or industry diversification at a short-term frequency over the long run. Because of this lack of data, it was not possible to study diversification along with business activity, although product diversification is one of the most important aspects of a firm's behavior over time. Research on volatility changes is concentrated on recent U.S. history. However, new data allow us to study events of significant volatility change in the early 19th century. I developed a new econometric technique, an ARCH model that deals with sparse datasets with missing observations, and did a complete analysis of financial market fluctuations. A study on the radical volatility change in the early 19th century in the U.S. and U.K sheds light on the causes and effects of the volatility change. In my thesis, I study the cause and effect of the volatility change related to the firm level product diversification and government investment. First, I establish a detailed profile of firm/plant level product diversification in manufacturing sector. Second, I study the relationship

Federal Trade Commission(2004, 1999), Samli (2004), Kirkpatrick (2002), and Wilcox et al (2001) 7 3,190 papers are found by a key work "firm-level diversification" in Google-scholar. See Schoar(2002) and Lins and Servaes(1999) for example.

6

2

between the firm level diversification and volatility of the U.S. manufacturing sector. Third, I investigate volatility change in early 19th century U.S. and U.K. Chapter 2 discusses the quality and limitations of the datasets as well as the measure of diversification, provides a conceptual discussion of diversification and describes stylized facts regarding the long-term and short-term dynamics of diversification. Chapter 3 provides a conceptual discussion of the relationship between diversification and profit volatility, describes stylized facts regarding volatility change, and estimates the relationship between firm level diversification and aggregate, industrial and idiosyncratic profit volatility. Chapter 4 briefly explains the history of government financing in the early 19th century, and explains the causes and effects of the volatility change during the crises of the 1840s.

3

Chapter 2: Product Diversification: Profile
Section 1. Introduction
Although diversification is one of the big issues in business history for a long time, most of studies concentrate on large conglomerates or certain industries for a relatively short time period. This chapter develops a thorough longitudinal analysis of diversification for the manufacturing sector. Gollop and Monahan(1991) is the one of a few existing studies of micro level diversification for the whole manufacturing sector in the long run. They showed that manufacturing firms specialized within plants, while they diversified among plants, until 1982 (see Figure 1). However, there are no empirical studies on diversification covering the last two decades, and it still remains unclear why firms change their product portfolios and how they change diversification over time. There are quite a few researches on the cyclicality of product diversification. Many of them suggest the diversification moves pro-cyclically, while a few others suggest counter-cyclical diversification.8 One of the goals of this chapter is to verify whether the diversification is pro- or counter-cyclical over the long time period. In this chapter, a detailed profile of diversification is described. I construct diversification index using 5-digit and 7-digit SIC product codes to see long term trends, short term cyclicality, and the evolution of diversification by firm/plant level characteristics. Diversification at the firm and establishment level will be analyzed from various perspectives. Section 2 discusses the quality and limitations of the datasets as well as the measure of diversification. Section 3 provides a conceptual discussion of diversification. Section 4 describes
8

See Axaloglou (2003), Keuschnigg (2001), Jovanovic (1993), for instance.

4

Figure 1 Diversification Indexes
Establishment Level
0.23 0.22 0.21 0.2 0.19 1963 1967 1972 1977 1982
0.5 0.48 0.46 0.44 0.42 0.4 1963 1967 1972 1977 1982

Firm Level

Note: Shipment weighted aggregate series Source: Gollop and Monahan(1991), Table 4, pp.328

stylized facts regarding the long-term and short-term dynamics of diversification. Section 5 summarizes the facts and analyses.

Section 2. Data
The three datasets I use are the Census of Manufactures (CM), Annual Survey of Manufactures (ASM) and Longitudinal Business Database (LBD) from 1974 to 1998.9 CM and ASM compose the Longitudinal Research Database (LRD). LRD is a time series of economic variables collected from manufacturing establishments in CM and ASM programs. LRD contains establishment level identifying information; information on the factors of production (inputs, such as levels of capital, labor, energy and materials) and the products produced (outputs); as well as other basic economic information used to define the operations of a manufacturing plant.10 LBD provides longitudinally linked data for all employer establishments (i.e., those with paid employees) contained in the Census Bureau's business register, the Standard Statistical Establishment List (SSEL). Basic data items, such as payroll, employment, location, industrial

9

CM is available in 1977, 1982, 1987, 1992, and 1997. ASM is available annually, 1973-76, 1978-81, 1983-86, 1988-91, 1993-96, and 1998-2000. LBD is currently available 1975 to 1999. ASMs in 1999 and 2000 are not used because the product codification was changed from SIC to NAICS in 1998. See Appendix A for a discussion. 10 Some product data that are imputed by Census Bureau are excluded from the sample.

5

activity and firm affiliation are included in LBD. LBD is used to get data on firm age, total employment, and the number of plants of multi-unit firms. Using LRD product files, I use a Herfindahl-type index as a measure of establishment and firm level diversification.11 My diversification index satisfies the following requirements: it varies directly with the number of different products produced; it varies inversely with the increasingly unequal distribution of products across product lines; and it is bounded between zero and unity.

D1 ? 1 ? ? si2 ,

where si = share of producti that is identifiedby 5 digit SIC code

? ? ij + 1 ? 2 D2 ? 1 ? ? ? ? 2 ? ?s j , where s j = share of product j that is identifiedby 5 digit SIC code ? ? n of shipmentsbetweenproduct j and Firm i' s primaryproducti by 4 digit SIC ? ij = correlatio D4 ? 1 ? ? si2 , D3 ? 1 ? ? si2 , where si = share of producti that is identifiedby 3 digit SIC code where si = share of producti that is identifiedby 4 digit SIC code

D1 is the simplest diversification index which incorporates the number of products and share of the products' shipments. Since it is simple, we can apply this method to any years in LRD data. D1 can show a very consistent time series of diversification and accounts the most detailed product information collected in ASM. One disadvantage of D1 is that it equally accounts products no matter how different they are: how they are related in terms of sales or production. In order to include the information on how the industries in which the firm diversifies are related, D2 uses the correlation coefficient of shipments of the industries as a distance weight between diversified products. Example 2 shows the difference between D1 and D2.

11

This measure has been widely used in the literature. See Gollop and Monahan(1991)

6

Example 1: Firm I produces A and B (5 digit SIC) in two different industries (4 digit SIC) with equal share Diversification Measure Case Index D1 D1=1-(.25+.25)=.5 corr(A,B)=1 D2=1-(.25+.25)=.5 D2 corr(A,B)=0 D2=1-(.25+.5*.25)=.625 corr(A,B)=-1 D2=1-(.25+0*.25)=.75

Example 2: Firm I produces A, B, C and D (5-digit SIC) with equal share Product A Product B Product C Product D 5-digit SIC 28124 28331 28332 28343 4-digit SIC 2812 2833 2833 2834 3-digit SIC 281 283 283 283 Shipment share .25 .25 .25 .25

D1=1-(.0625-.0625-.0625-.0625)=.75 D4=1-(.0625-.25-.0625)=.625 D3=1-(.0625-.5625)=.375

Therefore, D2 will be generally higher than D1 unless the firm diversifies all its products in same 4 digit industry. A comparison of D1 and D2 will shed light on how different industries firms diversify with their products. One may ask a question: Is diversification in different 5-digit SIC products a real diversification? There are cases where those products are so similar and ordinary people would not distinguish them easily. In such cases, it is better to use less detailed product classification to construct diversification index. D3 and D4 are additional measures of diversification to show only across-industry not within-industry diversification. Example 2 shows the difference across D1, D3 and D4. In this example, D4 is 17% lower than D1, suggesting that 17% of this firm's diversification came from within-4-digit-industry diversification. The fact that D3 is 50% lower than D1 shows a half of its diversification is due to within-industry diversification by 3-digit SIC. By showing D3 and D4 along with D1, we can see that Firm I is a highly diversified firm by 5-

7

digit SIC, but a rather specialized firm by 3-digit. By comparing these indexes with D1, it will be clear how much the within-industry diversification contributes to the total diversification. For the long-term trend analysis, I focus on the quinquennial CMs. The number of observations in CM is quite stable around 300,000 establishments. For a multi-unit firm level diversification index, the value of shipments of seven or five digit Standard Industrial Classification System (SIC) products is aggregated across the establishments of the firm and divided by the total value of shipments of the firm to get the share of each product. 12 The detailed calculation method is described in Appendix A.13 For the short-term analysis, I produce annual diversification indices at the establishment and firm levels using ASM and CM. The annual number of observations is stable around 70,000 establishments. I can use up to 5-digit SIC product codes to construct the annual diversification index because only 5-digit product codes are consistently available in ASM. With the same logic behind D1, D3 and D4, it is not clear which of the 5-digit or 7-digit SIC product code is better for the analysis of diversification. When the 7-digit code is used to construct the diversification index, I get higher index values, and we can study product variety in detail. However, the 7-digit code is very detailed and 7-digit products in the same 5-digit product

12

A single-unit firm is defined as a firm with only one location. A multi-unit firm is defined as a firm that owns multiple establishments. See Appendix A for detail of SIC. 13 Gollop and Monahan(1991) included a product heterogeneity component in their index construction, available only in CM. Their index is as follows:
? ? Diversification Index ( D) ? 1 / 2 ?1 ? si2 ? si sk ( zik ? ? ik ) ? ? ? i i k ?i ? ? where si = shareof a seven ? digit product ?1 if the i th and k th products are identical zik = ? th th ? 0 if the i and k products are not identical

?

??

? | wkj ? wij | ? ? ? ik ? ? ? ? 2 ? ? j wkj ? input cost shareof the j th input inthe k th product

1/ 2

?

8

code are often very similar to the each other.14 If we are interested in product diversification across a variety of "different" products, the 5-digit, 4-digit or even 3-digit code would be better. In this chapter, all 7-digit, 5-digit, 4-digit and 3-digit are reported, if available.

Section 3. Why Diversification?
Diversification has been treated as a firm characteristic in numerous studies. Many empirical studies on Total Factor Productivity (TFP) include the firm's diversification level as a control variable.15 Studies of the performance of the q-theory of investment also include multiproduct dummy variables. 16 Multi-product dummy variables are also used to proxy financial constraints in some studies. 17 Some have conjectured that diversified firms have different investment and entry/exit decisions, yet these empirical studies did not examine the firm's diversification directly. 18 Many studies find that multi-product firms behave differently from single-product firms, but the diversification decision has not been incorporated endogenously in the empirical literature.19 The adjustment of a firm's product portfolio has only recently drawn attention from researchers.20 There are many important studies on diversification in the area of strategic behavior studies and corporate finances. Campa and Kedia (2002) focuses on the relationship between the
14 15

An example of product classification in the chemical industry is given in Appendix A. Giandrea(2002), Gemba and Kodama(2001) 16 Bond and Cummins(2000), Fazzari, Hubbard and Petersen(1988), Abel and Blanchard(1986), Hayashi and Inoue(1991), Dwyer(2001). 17 Abel and Eberly(2001a and 2001b), Barnett and Sakellaris(1999), Gilchrist and Himmelberg(1995), Gross(1994) 18 Caballero, Engel and Haltiwanger(1995), Chatterjee and Cooper(1993), Dunne, Roberts and Samuelson(1989). Firm's exit and investment decisions are combined with financial constraints in Whited(1992) and Winter(1999). 19 The product portfolio decision has been considered in I/O literature in terms of business management. For example, Anderson, de Palma and Nesterov(1995), Ottaviano and Thisse(1999), Pepall and Norman(2001). 20 Cooper and Haltiwanger(2000), Sakellaris(2000)

9

decision of diversification and firm value. When they use panel data and instrumental variables to control for the exogenous characteristics that predict the decision to diversity, the evidence in favor of the assertion that diversification destroys value is weaker. When they jointly estimate the decision of a firm to diversify and its firm value, diversification seems a value-enhancing strategy. The diversification discount is more likely to be a premium in this case. They also find that firms that refocus their operations would have suffered a significant decreased in value if they had remained diversified, suggesting that the observed correlation between diversification and firm value is rather the outcome of actions by profit-maximizing firms reacting to shocks in their environments. In their estimation, they include a dummy variable for diversification, firm size, proxied by the log of total assets, profitability, investment, lagged variables and organizational aspects of industry (fraction of firms that are conglomerates, fraction of industry sales accounted for by conglomerates), economic environment (number of M&A, GDP, business cycle) and other firm publicity (listed on Nasdaq, NYSE, AMEX or part of S&P index, incorporated outside US). Villalonga (2004) also estimates the value effect of diversification by matching diversifying and single-segment firms on their propensity score – the predicted values from a probit model of the propensity to diversify. He also finds that on average, diversification does not destroy value. These papers suggest that the decision of diversification is consistent with profitmaximization and that it is a reaction to exogenous environment. Maksimovic and Phillips (2002) develop a model where the firm optimally chooses the number of segments in which it operates depending on its comparative advantage and industry demand shocks. Their model predicts firm-size distributions and investment and growth decisions of focused single-industry and multiple-segment firms. Plants of conglomerates are found less productive than plants of single-segment firms of a similar size, but this is consistent with the fact that conglomerates are value-maximizing, supporting the hypothesis that firms invest in industries in which they have a comparative advantage. Conglomerate firms also grow less in an industry if

10

their other plants in other industries are more productive and if their other industries have a larger positive demand shock. My dissertation extends these studies to build a more detailed profile of diversification and to examine its relationship to exogenous environment. The segment, the traditional definition of industry in which firms diversify, is 3-digit SIC in most of the papers mentioned above. The decision of diversification is often captured by dummy variable that takes value of 1 when firms diversify into multiple segments. Summary statistics in my paper will show diversification indexes measured by various definitions, including 2-digit, 3-digit, 4-digit, 5-digit and 7-digit SIC and also distance measure between industries. These results will shed light on various aspects of diversification, depending on how we define "diversification". Papers mentioned above focus on the relationship between diversification and firm performance (value or productivity). They show that diversification is a rational choice of profit maximization as a reaction to the exogenous environment, including GDP, demand shock by industry, other firms' performance. My paper will focus on the effect of exogenous factor, that is, what affects the decision of diversification, especially changes in economic volatility at the aggregate, industry and firm level. The variables in my estimation are similar to those in previous studies, including firm size proxy, profitability, age or organizational aspects. However, because I explicitly use various measures of degree of diversification and volatility, it will show not only whether to diversify or how many segments to diversify, but also how much to diversify as a response to economic volatility. Only a few papers pay attention to the short-term dynamics of product diversification. Chatterjee and Cooper(1993) link product diversity with the business cycle, but only at the aggregate level.21 Product choice is determined by the production technology and technology is

21

Chatterjee and Cooper (1993) analyzed the product diversity fluctuation with a firm exit/entry model.

11

usually regarded as something that changes only in the long run. This explains why short-run dynamics of diversification have seldom been studied in the short run. There are several potential motives for diversification. Jovanovic(1993) lists: (1) Gaining Market Power: A firm with market power in two substitute products can have higher profits than two single-product monopolies. (2) Avoiding Risk: With liquidity constraints, firms' investment, especially for small firms, depends on cash flows. Firms may diversify over the products to smooth their sales. (3) Having Access to Funds: In an imperfect capital market, funds tend to go to the large firms, not necessarily to the efficient ones. Firms may want to diversify across products to keep their size big. (4) Making Products Compatible: A set of products may be produced more efficiently together than individually. The optimal set of products is determined by the technology. (5) Reaping Efficiency Gains: By making several products, a firm can exploit cost synergies in producing, selling, promoting, and advertising. The diversified firm can also have a richer internal labor market to meet the demand of various production tasks. (6) Pursuing Managerial Goals: The manager may have a motive other than profit maximization. A diversified firm can reduce unemployment fluctuations, increase the volume of sales (though not necessarily profit), and discourage shareholder monitoring through complicated financial statements. Among these potential motives, risk-avoidance dominates the literature. To verify the effect of risk on diversification, stylized facts of diversification are profiled in the next section. The relationship between risk and diversification is analyzed in Chapter 3.

Section 4. Stylized Facts of Diversification
This section describes the trend and cyclicality of diversification from various viewpoints. Section 4-1 is about the long-term trend, and Section 4-2 is about the short-term dynamics. Section 4-2 also includes a description of annual diversification changes.

12

4-1. Long-term Trend
The Census of Manufactures surveys all establishments in the US manufacturing sector every five years. This allows us to study the long run behavior of diversification at the firm and establishment level. In CM, basic data obtained for all establishments include kind of business, geographic location, type of ownership, total revenue, annual and first quarter payroll, and number of employees in the pay period. For some establishments, much less data detail is requested and no information on materials consumed is collected.22 Product diversification is regarded as a firm level decision. However, it is important to study the establishment level diversification because a multi-unit firm can diversify not only within the firm but also within plants. The analysis allows us to see the trend of diversification within plants. Average diversification indexes are generated by 5 digit and 7 digit SIC product codes. Figure 2 plots the trend of the average diversification index at the establishment level. Diversification has steadily decreased at the establishment level since 1967 as measured using by either 5 or 7 digit SIC product codes. At the firm level, diversification stayed high until 1982 and then started to decrease. As Gollop and Monahan(1990) argued, until early 1980s, firms were diversifying while plants were specializing. Since then, however, both firm and plant level diversification has declined. The downward trend of aggregate diversification is surprising because many researchers have conjectured that firms should diversify more and more for various reasons. However, it is

22

CM is widely used in economic analysis and forecasting by many organizations, such as the Bureau of Economic Analysis, the Bureau of Labor Statistics, the Federal Reserve Board, state and local agencies, trade associations, companies, researchers, national and local news media.

13

premature to conclude that every firm decreased its diversification, because there is heterogeneity in firm level diversification. First, the trend is different for multi unit (MU) and single unit firms (SU). The diversification of MU firms and establishments is the driving force of the aggregate trend. Diversification decreased for the MU establishments, but increased for MU firms, up to 1982. Beginning in 1982, diversification decreased both at the establishment and firm level for multiunit firms. Diversification of SU establishments (or firms) has had a completely different trend, decreasing until 1987 and then increasing.

14

Figure 2 Average Diversification Index(D1) 1967-1997, CM
Total
Establishment level 0.5 0.4 0.3 0.2 0.1 0 1967 1972 1977 1982 1987 1992 1997
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1967 1972 1977 1982 1987 1992 1997 Firm level

Multi-units
Establishment level 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 1967 1972 1977 1982 1987 1992 1997 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1967 1972 1977 1982 1987 1992 1997 Firm level

Single-units
0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 1967 1972 1977 1982 1987 1992 1997 5 digit SIC 7 digit SIC

Note: Diversification index is shipment weighted series Source: Author's calculation

15

Table 1 Percentage change of Diversification (1967-1997)
Industry 20 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 Food Textile Apparel Lumber Furniture Paper Printing Chemical Petroleum Rubber Leather Stone Metal Fabricated Metal Machinery Electronic Transportation Instruments Miscellaneous Total -0.07 -0.27 0.36 -0.12 -0.19 -0.30 -0.23 -0.23 -0.02 -0.42 -0.19 -0.16 -0.37 -0.29 -0.17 -0.57 -0.41 -0.49 -0.19 Establishment SU -0.24 -0.18 0.54 -0.20 -0.12 -0.13 -0.22 -0.43 -0.55 -0.13 0.21 -0.01 -0.12 -0.03 -0.09 -0.49 -0.43 -0.41 -0.09 MU -0.10 -0.31 0.15 -0.12 -0.29 -0.31 -0.26 -0.22 -0.01 -0.46 -0.33 -0.20 -0.38 -0.35 -0.21 -0.56 -0.41 -0.48 -0.25 Total -0.00 -0.25 -0.21 0.01 -0.16 0.03 0.06 -0.05 -0.08 -0.28 1.61 0.14 -0.17 -0.36 -0.03 -0.28 -0.06 -0.09 0.00 Firm SU -0.24 -0.18 0.54 -0.20 -0.12 -0.13 -0.22 -0.43 -0.55 -0.13 0.21 -0.01 -0.12 -0.03 -0.09 -0.49 -0.43 -0.41 -0.09 MU -0.00 -0.26 -0.18 0.04 -0.24 0.00 0.06 -0.06 -0.09 -0.21 1.00 0.00 -0.15 -0.23 -0.09 -0.25 -0.07 -0.10 -0.02

Note: Food (Industry 20) includes Tobacco due to the disclosure issue.

Figure 4 explains why the overall trend of aggregate diversification is dominated by the movements due to multi-unit establishments or firms. Single-units firms comprise 30-50% of all establishments, but the share of economic activity attributable to single-units is a mere 57%. At the firm level, the non-weighted share of single-units is 50-70%. The trend of diversification is also different by industry. In Table 1, eighteen of nineteen industries exhibit declining diversification at the establishment level. The one exception is Apparel (36% increase). The rate of increase in the SU index in Apparel (54%) far exceeds the corresponding rate for the MU index (15%). For the eighteen industries with declining diversification, the decline is more severe in MU establishments in thirteen cases. At the firm level, twelve of nineteen industries exhibit declining diversification. The seven exceptions are Food, Lumber, Paper, Printing, Leather, Stone and Miscellaneous Manufacturing. The increase in firm-level diversification in these seven industries is driven

16

largely by multi-unit firms. For the twelve industries with declining diversification, the decline is more severe in SU firms in six industries. To summarize, the aggregate diversification index declined both at the establishment and firm level. However, there is great heterogeneity across MU/SU and by industry at the establishment and firm level. Diversification declined in the majority of industries both at the establishment and firm level. The decline is most severe in establishments that are part of MU firms. The evidence suggests that within-plant diversification of MU firms is decreasing. This will be verified in Section 4-2.

4-2. Short-term dynamics
This section investigates short-run dynamics using the ASM sample. The number of observations decreases when we use only ASM plants in CM, and we lose some analytical power when we focus on ASM data. However, ASM enables us to construct an annual diversification index to study short-term variations, which has never been attempted in the literature.

Establishment level Analysis: Aggregate level
Figure 2 and 3 show the same aggregate trend of diversification at the establishment level. The difference between them is the sample size and frequency; Figure 2 uses quinquennial CM data with roughly 300,000 establishments, while Figure 3 uses annual ASM data with roughly 70,000 establishments. At the aggregate level, the annual diversification index has a downward trend. The trend is mostly explained by the movement of MU establishments.23

23

By design, the average diversification index is higher for establishments that produce more products. Roughly 50% of sample establishments produce only one product each year. About 25% produce 2 products, 10% produce three, 5% produce four, and 5% produce five or more products. See Figure 6 for analysis of changes in the number of products.

17

Figure 3 Establishment Level Diversification Index(D1)
Annual Average Diversification Establishment Level 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 74 77 80 83 86 89 92 95 98 Total SU MU

Cyclical Movement of Establishment Level Diversification 0.02 0.01 0 -0.01 -0.02 74 79 84 89 94 Diversification Index Growth of Real Shipment 0.2 0.15 0.1 0.05 0 -0.05

Single Units 0.04 0.03 0.02 0.01 0 -0.01 -0.02 -0.03 -0.04 74 79 84 89 94 Diversification Index Growth of Real Shipment 0.2 0.15 0.1 0.05 0 -0.05 0.02 0.01 0 -0.01 -0.02 74 79 84 89 94 Multi Units 0.2 0.15 0.1 0.05 0 -0.05 Diversification Index Growth of Real Shipment

Figure 4 Share of Single Unit Establishments (Firms)
Establishment Level 100% 80% 60% 40% 20% 0% 74 SU share 79 84 89 94
100% 80% 60% 40% 20% 0% 74 77 80 83 86 89 92 95 98 SU share SU Shipment share Firm Level

SU Shipment share

18

Is aggregate diversification is pro-cyclical? In the second graph of Figure 3, the diversification index seems to move pro-cyclically until 1990, then starts diverging. The diversification index is linearly detrended and the growth of real shipment in the manufacturing sector is obtained from ASM statistics published by Census.24 The aggregate cyclical behavior is driven by multi-unit establishments (fourth graph of Figure 3), while single-unit firms show clear counter-cyclical movements. It seems that single-unit firms diversify more in recessions than in booms. On the other hand, multi-unit establishments diversify more in booms than in recessions.

Establishment level Analysis: Industry level
The average diversification index shows large variations across industry, as shown in Table 2.25 Overall, establishments in Food, Printing, Chemical, Petroleum, and Metal industries have higher diversification on average over the period 1974-1998. For single-unit establishments, Food, Lumber, Printing, Chemical and Petroleum have high diversification. For multi-unit establishments, Printing, Chemical, Petroleum, Metal and Machinery have high diversification. To summarize, Printing, Chemical, and Petroleum industries have high diversification both in single-unit and multi-unit establishments. The high diversification of Food and Lumber is driven by their highly diversified single-unit establishments. Multi-unit establishments in Metal and Machinery crank up the average diversification level in those industries. Cyclicality is also heterogeneous by industry. The sign of the correlation between the sectoral diversification index and the sectoral growth rate of real shipments is mixed across industries. Out of nineteen industries, eleven have positive correlations. Among the significant six correlations, four industries have positive signs. For single-unit establishments, thirteen industries have positive signs and four of five significant correlations are positive. For multi-unit
24 25

See Appendix A Industry is classified by 1987 basis SIC. See Appendix A for detail.

19

Table 2 Average Establishment Level Diversification Index and Correlation with Growth of Real Shipment by 2 digit SIC Industry
Total Industry Mean Corr. coeff. Mean Corr. coeff. Mean Corr. coeff. SU MU

20 Food 0.26 0.09 0.19 -0.28 0.27 0.14 22 Textile 0.19 0.37 0.13 0.26 0.20 0.36 23 Apparel 0.19 0.71* 0.17 0.53* 0.22 0.64* 24 Lumber 0.19 -0.51* 0.18 -0.34 0.20 -0.57* 25 Furniture 0.23 -0.21 0.16 -0.03 0.27 -0.14 26 Paper 0.21 0.10 0.12 0.07 0.22 0.19 27 Printing 0.30 0.08 0.24 0.45* 0.33 0.11 28 Chemical 0.35 -0.09 0.18 -0.42* 0.37 -0.45* 29 Petroleum 0.58 0.75* 0.27 0.36 0.59 0.65* 30 Rubber 0.19 0.61* 0.15 0.25 0.20 0.66* 31 Leather 0.15 0.39 0.13 0.55* 0.16 0.41* 32 Stone 0.11 -0.03 0.10 0.17 0.12 0.00 33 Metal 0.29 0.50* 0.16 -0.11 0.30 0.52* 34 Fabricated Metal 0.16 0.15 0.14 0.03 0.17 0.39 35 Machinery 0.28 -0.31 0.15 0.30 0.31 -0.18 36 Electronic 0.20 -0.24 0.12 -0.25 0.21 -0.17 37 Transportation 0.22 -0.42* 0.13 0.25 0.22 -0.44* 38 Instruments 0.21 -0.16 0.11 0.09 0.23 -0.33 39 Miscellaneous 0.15 0.20 0.09 0.56* 0.18 0.20 *significance at the 95% level Note: Industry 20 includes industry 21 (Tobacco) due to the private information disclosure issue. Source: Diversification index from author's calculation, Growth rate of real shipments from Census Bureau

establishments, eleven industries in total and five out of eight significant correlations have positive signs.

Firm level analysis: Aggregate level
At the aggregate level, the annual diversification index computed at the firm level shows a downward trend from 1974 to 1998 in the first graph of Figure 5. The level of firm diversification is higher than the establishment level index, mainly because of the high diversification of multi-unit firms. Single-unit firms comprise 60% of the sample but account for less than 10% of total shipments in average in Figure 4.

20

Figure 5 Firm Level Average Diversification Index (D1)
Cyclical Movement of Firm Level Diversification

Firm Level Average Diversification Index
0.02 0.2 0.15 0.1 0.05 0 -0.05 74 79 84 89 94

1 0.8 0.6 0.4 0.2 0 74 76 78 80 82 84 86 88 90 92 94 96 98

0.01 0.01 0.00 -0.01 -0.01 -0.02 -0.02 -0.03

Diversification Index

All Firm

SU

MU

Growth Rate of Real Shipments

Single Units 0.04 0.03 0.02 0.01 0 -0.01 -0.02 -0.03 -0.04 74 79 84 89 94 0.2 0.15 0.1 0.05 0 -0.05
0.02 0.015 0.01 0.005 0 -0.005 -0.01 -0.015 -0.02 -0.025 -0.03 74

Multi Unit Firms 0.2 0.15 0.1 0.05 0 -0.05 79 84 89 94

Diversification Index Growth Rate of Real Shipments

Diversification Index Growth Rate of Real Shipments

There is not a clear cyclicality of diversification at the aggregate level. The second graph of Figure 5 plots the average diversification index and the growth rate of real shipments in the manufacturing sector. 26 SU firms seem to have countercyclical diversification, i.e., firms specialize in booms and diversify in recessions. There is no clear co-movement for MU firms.

26

I use the linearly detrened diversification index the growth rate of real value of shipment.

21

Figure 6 Firm Level Average Diversification Index (D1, D2, D3 and D4)
Firm level diversification index 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 74 77 80 83 86 89 92 95 98 d1 d2
Single Units 0.25 0.20 0.15 0.10 0.05 0.00 74 76 78 80 82 84 86 88 90 92 94 96 98 su1 su2 su3 su4 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 74 76 78 80 82 84 86 88 90 92 94 96 98 mu1 mu2 mu3 mu4

d3

d4
Multi Units

Figure 6 compares different measures of diversification. Since D2 uses distance weights to incorporate the relationship across industries, D2 is higher than D1. However, the gap between D1 and D2 is not so big. Recalling that D1 is close to D2 for a firm if all the industries are positively correlated to the primary industry of the firm, it means that firms do not diversify into really different industries. For single unit firms, the gap between D1 and D2 becomes smaller in

22

the 1990s. This suggests that specialization in closely related industries is more prevalent in single units. Since D3 and D4 use less detailed product classification, they are much lower than D1. Using the 4-digit SIC, D4 is about 20% lower than D1 which means the intra-industry diversification contributes about 20% of the total 5-digit product level diversification. Using 3digit SIC, D3 is about 30% lower than D1. However, the proportion of D3 to D1 or D4 to D1 doesn't change much over time. This suggests the composition of inter- or intra-industry diversification remains stable in my sample period. Different measures of diversification show different levels of index but the trends and cyclicality look remarkably similar to one another. Since almost all aspects of diversification analyses share similar trends across different measures, I will use D1 to explain trends of diversification for the rest of the paper. However, I'll also show other measures of diversification if discussions about magnitude of different indexes are needed.

Firm level analysis: Industry level
In Table 3, the average diversification index shows great variation by industry.27 Overall, firms in Paper, Chemical, Petroleum, Transportation, and Instruments have a higher mean diversification index in 1974-1998. Single-unit firms in Food, Lumber, Printing, Chemical and Petroleum industries have high diversification. For multi-unit firms, Paper, Chemical, Machinery, Transportation Equipment and Instruments have high diversification. The Chemical industry has high diversification both in single-unit and multi-unit firms. The high diversification of Paper, Transportation and Instruments is driven by their highly diversified multi-unit firms. Table 3 also displays the estimates of correlation coefficients between the diversification index and the value of shipments by sector. Out of nineteen industries, twelve have negative
27

SIC is based on 1987 changes. See Appendix A for detail.

23

Table 3 Firm Level Average Diversification Index(D1) and Correlation with Growth of Real Shipment by 2 digit SIC Industry
Total Industry mean Corr. Coeff. SU Corr. mean Coeff. MU Corr. mean Coeff.

20 Food 0.19 -0.28 0.51 -0.30 0.53 -0.20 22 Textile 0.13 0.26 0.57 -0.02 0.55 -0.17 23 Apparel 0.54 0.44 * 0.17 0.53 * 0.61 -0.08 24 Lumber 0.47 -0.34 0.18 -0.34 0.63 -0.53 * 25 Furniture 0.16 -0.03 0.56 -0.44 0.50 0.07 26 Paper 0.12 0.07 0.75 0.10 0.73 0.27 27 Printing 0.56 -0.57 * 0.24 0.45 * 0.58 -0.67 * 28 Chemical 0.74 -0.64 * 0.18 -0.42 * 0.77 -0.71 * 29 Petroleum 0.27 0.36 0.71 0.00 0.70 0.07 30 Rubber 0.66 0.52 * 0.15 0.25 0.71 0.53 * 31 Leather 0.49 -0.13 0.13 0.55 * 0.55 0.04 32 Stone 0.10 0.17 0.68 0.10 0.63 0.35 33 Metal 0.63 -0.10 0.16 -0.11 0.70 -0.26 34 Fabricated Metal 0.65 -0.02 0.14 0.03 0.72 0.34 35 Machinery 0.73 0.45 0.69 -0.49 * 0.15 0.30 36 Electronic 0.69 -0.52 * 0.12 -0.25 0.70 -0.28 37 Transportation 0.73 0.08 0.13 0.25 0.75 -0.01 38 Instruments 0.11 0.09 0.80 0.07 0.77 -0.18 39 Miscellaneous 0.78 -0.30 0.09 0.56 * 0.81 -0.30 *significance at the 95% level Note: Industry 20 includes industry 21 (Tobacco) due to the private information disclosure issue. Source: Diversification index from author's calculation, growth rate of real shipments from Census Bureau

correlation coefficients. Among the six significant correlations, four have negative signs. As discussed in Section 4-1, thirteen industries have positive signs and four of five significant coefficients are positive for SU firms. For MU firms, twelve industries in total and three of four significant correlations have negative signs. To summarize, it is very difficult to draw a clear conclusion regarding the cyclicality of aggregate or industry level diversification from the data. It is necessary to see the distribution of firms and to study diversification directly at the firm level to see how firms change their diversification.

24

Figure 7 Change in the Number of Plants and Products
Change in Number of Products 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1970s -2 or less 0 (N>1) -1 +1 1980s 1990s 0 (N=1) +2 or more 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1970s -2 or less 0 (N>1) -1 +1 1980s 1990s 0 (N=1) +2 or more Change in Number of Plants

Firm level analysis: Change of diversification
Since the choice of number of products is discrete, it is interesting to see how firms adjust their number of products over time. The first graph of Figure 6 displays the distribution of firms over year-to-year changes in the number of products. Firms are classified into six groups every year: Firms that discontinue producing two or more products compared to the previous year (-2); Firms that discontinue producing one product (-1); Firms that produce a single product in both years (0 with N=1); Firms that produce the same number of multiple products (0 with N>1); Firms that produce one more product than previous year (+1); Firms that produce 2 or more products than previous year (+2). We get an annual distribution of firms by this classification. Figure 6 shows the distribution of the annual series averaged by decade, showing that the number of single-product producers decreased in the 1990s (white-colored block).28 Firms with no change in the number of products (black-colored block) also decreased, while there was an increase in the

28

From the bottom, N(t)-N(t-1)<=-2, -1, 0 given that N(t)-N(t-1)=1, 0 given that N(t-1)>1, +1, and +2 are displayed in Figure 6.

25

share of firms that increased or decreased one product (slashed blocks). These firms are "productswitchers" that adjust their product portfolio with one marginal product. Multi-unit firms make a discrete choice regarding the number of plants that operate. The second graph in Figure 7 shows the decade average of the distribution of firms by the annual change in number of plants, with the same categories as the first graph of Figure 7. The shares of single plant firms (white-colored block) and of firms with the same number of plants in any two consecutive years (black-colored block) decreased in the 1990s. There are many multi-unit firms that adjust the number of plants up or down by one. Even when two multi-unit firms produce identical products, they can be different in terms of how they allocate production. For example, in Table 4, Firm I produces product X in plant A and product Y in plant B. Firm II produces both X and Y in plant A and only X in B. Firm I owns two specialized plants while Firm II has one diversified plant and one specialized plant, although they have the same firm level diversification index. The diversification index can be decomposed to distinguish these two firms. Equation 1 groups the products into two categories: those produced in multiple plants or in a single plant. The share of production diversification factor (rpd) reflects the diversification of production, not the diversification of products. rpd is 0 for Firm 1 and 0.5 for Firm II. Equation 2 investigates further the link between establishment and firm diversification. Since a firm is defined as the sum of its establishments, a firm's diversification must be a function of diversification within and among its plants. Consider adding and subtracting a shipmentsweighted average of diversification indexes for a firm's establishments to the right-hand side of an identity equating the firm's diversification index with itself. The within-plant factor reflects the contribution of within-establishment diversification to overall firm level diversification. The among-plant factor recognizes that differences in product mix across plants are captured in the firm measure but not in the individual plant measure. It quantifies the contribution of

26

diversification among a firm's plants. In the example of Table 4, the within plant factor is .375 for Firm II.

Equation 1 Production Diversification

d = 1? (

?S

Diversified Production

i?A 1 2 3

2 i

+

?S

Specialized Production

i?B 1 2 3

2 i

) = (rpd + rps )d

where,

rpd = ? S i2 / ? S i2 , rps = 1 ? rpd
i?A i

i ? A product i produced in multiple plants i ? B product i produced only in one plant

Equation 2 Within/Among-plant Diversification
f est f d f = ? a j d est j + ( d ? ? a j d j ) = ( rwp + rap ) d j j 1 4 24 3 144 2 44 3 Within -plant Among-plant

where,
a j = shipment share of the jth plant d f = firm level diversification, rwp = ? a j d est d f , rap = 1 ? rwp j
j

d est = plant level diversification

The first graph in Figure 8 plots the share of diversified production (rpd) from 1974 to 1998. The production diversification factor increased in 1990s but is below 2% for the whole sample period. Therefore, specialized production is much more common. The second graph in Figure 8 plots the share of within plant diversification in overall diversification (rwp). Within-plant

27

Figure 8 Share of Diversified Production (rpd) and Share of Within-plant diversification (rwp)
Share of Diversified Production (rpd) 2.5% 2.0% 1.5% 1.0% 0.5% 0.0% 74 79 84 89 94

Share of Within-plant factor (rwp)

50% 40% 30% 20% 10% 0% 74 79 84 89 94

diversification declined over the last three decades. The two graphs in Figure 8 imply that firms are specializing productions more and more. The aggregate statistics suggest that the average firm doesn't change its diversification much in short time period. Figure 9 plots the average net change of firms' diversification in two consecutive years, that is, NET=avg(d(t)-(d(t-1)) for firms that are operating in both years. NET is very small throughout the sample period. One might be tempted to conclude that fluctuations in diversification do not matter much because of the small annual changes. However, we see much bigger fluctuation when we break down the net changes into two components, the positive changes (POS=avg(d(t)-(d(t-1)) for the firms with d(t)>(d(t-1)) and negative

changes(NEG=avg(d(t)-(d(t-1)) for the firms with d(t)<(d(t-1)). NET is equal to POS minus NEG (NET=POS-NEG). Figure 9 suggests there are many firms that increase or decrease their diversification keeping the overall average change small.

28

Figure 9 Positive, Negative and Net change of diversification
Change of Diversification for Continuers, Net, POS and NEG 0.080 0.060 0.040 0.020 0.000 -0.020 -0.040 -0.060 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 Net Change for Continuers POS NEG

Table 5 Distribution (percentiles) of Firms' correlation coefficients between new/lost industry and the primary industry 5th
Firms with increasing number of industries Firms with decreasing number of industries .02 -.1

10th
.32 .2

Median
.87 .86

90th
.97 .97

95th
.98 .98

It is important to know which industry the firm diversifies. Does the firm diversify into industries which have positive correlation coefficients with current primary industry? We can shed a little bit of light on this issue by looking at the distribution of firms' correlation coefficients between new/lost industry and the primary industry when the firm increases/decreases number of products. Table 5 shows percentiles of correlation coefficients of firms. Among the group of firms with increasing number of industries in two consecutive years, the median firm's correlation coefficient between the new industry and its primary industry is .87. Even the 5th percentile of firms has positive correlation (.02). This suggests that when the firm increases its product portfolio, it usually diversifies into similar industries with positive correlation with its primary

29

product. Likewise, among the group of firms with decreasing number of industries, the median firm's correlation coefficient is .86. The correlation is .2 for the 10th percentile firm and -.1 for 5th percentile firm. This means many firms shut down products that have positively correlated industries but some firms withdraw from negatively correlated industries. This suggests that the avoiding-risk factor is less important and the trend of specialization is more important in the firm's decision of diversification.

Firm level analysis: average diversification index by firm characteristics
Appendix C includes average diversification indexes by various firm characteristics. Single unit firms have lower diversification than multi unit firms (Table AC-2 and AC-3). Big firms have higher diversification (Table AC-10), as old firms (Table AC-11). There is no clear regional difference in diversification (Table AC-12). If the firm is vertically integrated, the firm will diversify into the products that are consumed within the firm to produce the final product. The share of Interplant Product Transfers to the total value of shipments of the firm (IPT) is used as an indicator for vertical integration. Table AC-13 shows that diversification increases with IPT but starts to decrease if IPT is too high, suggesting that a firm with very high vertical integration diversifies less and specializes more. Table AC-14 shows that diversification is higher for firms with lower labor cost share. Labor intensive firms tend to specialize. A high ratio of organizational workers may be needed facilitate the complicated process of multi-product production. Diversification increases with the share of non-production worker wage to the total wage cost but starts to decrease when the share gets very high (Table AC-15). The relationship between diversification and exporting is not clear in Table AC-16, although non-exporting firms tend to have lower diversification because they are relatively small firms. With the limited information from ASM product data, we can see how heterogeneous

30

products are by looking at the number of industries (at the 2-digit SIC level) spanned by the number of products of the firm. In Table AC-17, for example, firms which produce 10 products diversify across 2.7 industries in the 1970s, while they diversify across 2.3 industries in the 1990s (row 10). The number of industries declines for firms that produce many products. The second panel of Table AC-17 shows the number of industries by 3-digit SIC. Table AC-18 shows the geographical dispersion of plants within firms by displaying the number of different counties where plants are located as a function of the number of plants. For example, firms with 10 plants locate them in 8.3 counties in the 1970s and 8.7 counties in 1990s (row 10). In general, firms diversify more geographically in 1990s than in 1970s. In summary, firm-level diversification is very heterogeneous by firm characteristics, but most of the statistics confirm our conjecture about what types of firms have high diversification: Big firms, old firms, capital-intensive firms, firms with many organizational workers, etc. Furthermore, the trend of diversification is common across regions in US. It is worth a notice that firms seem less diversified horizontally but more diversified geographically: Even the highly diversified firm specializes in a couple of 2-digit industries, but firms have operated their plants in more diversified locations over time.

Section 5. Conclusion
There have been a lot of studies on firm level diversification but none have covered the whole manufacturing sector over a long period of time. In this chapter, I studied firm level diversification of the manufacturing sector between 1974 and 1998 and described the trend and cyclicality of diversification in detail. From the rich description in this chapter, the new findings of firm level product diversification can be summarized as follows:

31

(1) Aggregate diversification declined both at the establishment and firm level since the early 1980s. The downward trend is common across many industries. The declining

diversification is quite surprising because diversification has been regarded as a virtue of
firms in last several decades. (2) Whether diversification is pro-cyclical or counter-cyclical is not clear at the aggregate or industry level. Many studies have argued on the pro- or counter-cyclicality of diversification, but this chpater shows that there is heterogeity in cyclicality of diversification. (3) A large fraction of firms change the number of products and plants annually. The declining diversification measure suggests that firms have become more specialized, but it is clear that the number of products is not fixed for firms even in the short run.

I constructed different indexes of diversification to capture different aspects of diversificaiton. Diversification index using across-indsutry correlation as distance weights shows that firms, especially single unit firms, have not diversified into remotely related indsutries. Diversificaion indexes using less detailed product classification show that within-industry diversification (at 3 or 4-digit SIC level) contributes 20%-30% of total diversification. From anecdotal evidence, it is widely known that product diversification is a decision variable for firms, which is contrary to assumptions of fixed diversification in many theoretical models in the literature. This chapter shows that firms actively change their product diversification at a short-term frequency. The number of products and plants behaves like an adjustment margin. These stylized facts can be a benchmark against which firm production models should be verified. The fact that firms can change the number of products frequently sheds light on studies of flexible capital. If firms adopt more flexible capital to produce their outputs, the degree of

32

diversification doesn't need to be fixed over time. Firms will have more margins to respond to the business cycle if they have more flexible capital as well as more flexible labor contracts. The new findings in this chapter can be added to the set of the evidence that firms have flexible capital in US manufacturing sector. A question naturally arising from the evidence of declining diversification is whether this fact is unique in U.S. Especially in East Asia, big conglomerates that diversify across a variety of industries have been regarded as the engine of fast growth. A cross-country study on diversification trends will be one of the next research topics in this field.

33

Chapter 3: Volatility Change and Diversification
Section 1. Introduction
Some of the studies on economic volatility in US were introduced in Chapter 1. The significant change in mid-1980s was not restricted to any one sector, level or indicator. Many economic indicators show less volatility. Stock and Watson (2002) show that the moderation in volatility is widespread and appears in both nominal and real series. The decline in volatility is most pronounced for residential investment, output of durable goods and output of structures. The decline in volatility appears both in measures of real economic activity and in broad measures of wage and price inflation. The decline in aggregate volatility is pervasive. Recent studies show that volatility has decreased not only at the aggregate level but also at sectoral level. They find that the decrease is not confined to any one sector, but is common to many sectors. Kim et al (2004) shows that the volatility reduction in aggregate output is visible in more sectors of output than simply durable goods production. Specifically, there is an evidence of a volatility reduction in the production of structures and non-durable goods. Comin and Mulani (2003) investigate the evolution of volatility at the firm level. They find that while the growth rate of aggregate sales has become more stable over time at the firm level, the volatility of the growth rate of sales at the firm level has increased. They argue that idiosyncratic firm-level volatility diverges from the aggregate trend. But they use the data only for only public firms. 29 It has not been confirmed whether idiosyncratic volatility has been increasing for all firms, including small non-public firms. This chapter verifies these findings on volatility with ASM and CM data. Then I study the effect of volatility on the firm level diversification decision. Among the suggested motives for
29

They use COMPUSTAT data.

34

diversification, risk-avoidance dominates the literature: With liquidity constraints, firm investment depends on cash flows.30 If firms diversify over products to smooth their profits, then they should respond to the volatility of profit shocks on every level. In particular, aggregate, sectoral and idiosyncratic profit shocks can affect firm level diversification. My main findings confirm the decrease in aggregate, sectoral and idiosyncratic volatility of the profit rate, and show that a less volatile profit rate leads to less diversification. Section 2 provides a conceptual discussion of diversification and examines the relationship between diversification and profit volatility. Section 3 describes stylized facts regarding the volatility change. Section 4 tests the relationship between firm level diversification and the aggregate, industrial and idiosyncratic profit volatility. Section 5 summarizes the facts and analyses.

Section 2. Volatility and Diversification
Changes in volatility can affect diversification at different levels. More formally,

(1) ( 2)

d it = f (? ( Ait )) , where i = 1,2,..., N , t = 1,2,...T Ait = where
Aggregate factor

At {

+ ( Ast ? At ) + ( Ait ? Ast ) 1 4 24 3 1 4 24 3
Industrial factor

Idiosyncratic factor

At =

1 N

?A
i

it

, Ast =

1 Ns

?A
i?s t +5

it

(3)

? ( Ait ) =

?

t +5 j =t ? 4

( Aij ? A it ) 2 10

, A it

? =

j =t ? 4

Aij

10

30

Jovanovic(1993)

35

where the diversification for firm i ( d i ) is a function of the volatility of the profit rate ( Ait ). In Equation (2), the profit rate consists of three factors, aggregate, industrial, and idiosyncratic factors. There are profit shocks at three levels ( At , Ast ? At , Ait ? Ast ) and the equation holds as an identity. So the industrial and idiosyncratic components are defined as deviations from the average industry or firm profit shocks.31 Equation (3) defines the volatility of the time series for firm level profits as ? ( Ait ) by computing the series of standard deviations of 10-year rolling windows of Ait .32 Profit shocks at the aggregate, industrial and idiosyncratic level are assumed to be orthogonal to one another by construction. Since the shocks are orthogonal, the standard deviations of the shocks over time (volatility) are orthogonal to one another. Therefore, orthogonality is preserved for the volatility of observed profit rates at the aggregate ( ? ( At ) ), industry ( ? ( Ast ? At ) ) and firm level ( ? ( Ait ? Ast ) ). We can test the following hypotheses regarding the partial effect of profit shocks on the firm level diversification:

(1) H 0 :

?f >0 ?? ( At ) ?f >0 ?? ( Ast ? At ) ?f >0 ?? ( Ait ? Ast ) , s = Two ? digit SIC Industry

(2) H 0 :

(3) H 0 :

, i = 1,2,..., N

31 32

This is similar to a Cholesky decomposition. The standard deviation of a 10-year window is used as the measure of volatility in Comin and Mulani (2003). Stock and Watson (2002) uses the standard deviation by decade. Kahn et al(2002) uses the standard deviation in three sample periods (1953-1968, 1968-1983, and 1984-2000).

36

It is very intuitive that the sectoral and idiosyncratic volatility affect the diversification decision. Firms can insure themselves against bad profit shocks by diversifying into different industries and products. However, firms cannot avoid the aggregate shock because no matter how many products they produce, the aggregate shock will hit them equally. The aggregate shock in this analysis includes not only aggregate profit fluctuations of manufacturing sector but any disturbance that is not captured by sectoral or idiosyncratic volatility in the economy. For example, fluctuations in the service sector or financial sectors will show up as aggregate volatility change. A time trend is not identified separately from this aggregate component.

Section 3. Stylized facts: Volatility
Figure 1 shows that the aggregate profit volatility ( ? ( At ) ) has constantly decreased over my sample period. Since I use a rolling standard deviation across 10 years as the measure of volatility, the volatility measure for the first 4 years is only forward looking, and volatility fro the last 5 years is backward looking. Therefore, only the data between 1978 and 1993 are appropriate. Profit rates are measured as sales minus variable costs, divided by the capital stock.33 Table 1 shows the volatility of the average firm level profit rate by industry. Almost all industries had lower profit volatility in 1993 than in 1978. The first graph of Figure 2 displays industries that had low volatility in the 1980s. The second graph of Figure 2 shows industries with high volatility in the 1980s – Rubber, Leather, Machinery and Instruments. The volatility of industries not shown in Figure 2 is constant or slightly decreasing over time. The downward trend

33

See Appendix A for detail.

37

Table 1 Profit Volatility by Industry ( ? ( Ast ? At ) )
Industry 20 Food 22 Textile 23 Apparel 24 Lumber 25 Furniture 26 Paper 27 Printing 28 Chemical 29 Petroleum 30 Rubber 31 Leather 32 Stone 33 Metal 34 Fabricated Metal 35 Machinery 36 Electronic 37 Transportation 38 Instruments 39 Miscellaneous 1978 0.19 0.14 0.52 1.07 0.35 0.20 0.30 0.21 0.82 0.17 1.33 0.10 1.22 0.07 0.58 0.94 1.11 0.44 1.21 1983 0.13 0.10 0.36 0.06 0.32 0.22 0.14 0.28 0.72 0.56 1.31 0.13 0.30 0.18 1.20 0.73 0.81 0.28 1.08 1988 0.13 0.12 0.42 0.11 0.29 0.18 0.17 0.31 0.33 0.65 1.52 0.16 0.39 0.20 1.25 0.31 0.55 0.52 0.36 1993 0.11 0.15 0.38 0.12 0.22 0.09 0.09 0.13 0.41 0.24 1.28 0.20 0.37 0.04 0.52 0.54 0.47 0.47 0.21

Note: Food (Industry 20) includes Tobacco (Industry 21) due to the private information disclosure policy of the Bureau of Census

of volatility is widespread across industries, but not universal. This is consistent with evidence in the literature. At the idiosyncratic level, some firms have higher volatility, and other firms have lower volatility in the 1990s than in the 1970s. I calculate the volatility for each firm, then take the mean ( avg (? ( Ait ? Ast )) ) and cross-sectioned standard deviation ( std (? ( Ait ? Ast )) ) in every
i

i

year. Figure 3 shows the evolution of firm level volatility. The mean of idiosyncratic volatility increased in the early 1980s but fell in the late 1980s as shown in the first graph of Figure 3. Although there is an increase in the late 1990s, the standard deviation of firm level volatility remained the same or slightly increased between 1979 and 1994 as shown in the second graph of Figure 3.

38

Figure 1 Mean and Average Volatility of Firm Level Profit Rates
Aggregate Volatility 0.3 0.25 0.2 0.15 0.1 0.05 0 78 81 84 87 90 93
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 74 76 78 80 82 84 86 88 90 92 94 96 98 Aggregate Profit Rate

Figure 2 Volatility of average firm level profit rates by industry
Industries with decreasing volatility in 1980s 1.4 1.2 1 0.8 0.6 0.4 0.2 0 78 23 36 81 84 24 37 87 29 39 90 93 33

Industries with increasing volatility in 1980s 2 1.5 1 0.5 0 78 81 30 84 31 87 35 90 93 38

Figure 3 Mean and standard deviation of firm level idiosyncratic volatility
MEAN of Firm level volatility 0.6 0.5 0.4 0.3 0.2 0.1 0.0 78 81 84 87 90 93
0.014 0.012 0.01 0.008 0.006 0.004 0.002 0 78 81 84 87 90 93 STD of firm level volatility

39

Figure 4 Average Idiosyncratic Volatility by Size of Firm
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 small(emp<500) big(emp>=500)

The downward trend of idiosyncratic volatility is different from evidences in the literature. Comin and Mulani (2003) showed an upward trend of idiosyncratic volatility for relatively big firms in COMPUSTAT data. Although they showed the result is not coming from the sample bias in the paper, I verified the trend of idiosyncratic volatility for big and small firms, separately. Figure 4 shows the idiosyncratic volatility by the two size group of firms. 34 An increasing volatility is not observed even for big firms. Unlike the downward trend of aggregate or sectoral volatility, the trend of idiosyncratic volatility is not unarguable.

Section 4. Diversification and Volatility: Estimation
The three key stylized facts of diversification from Chapter 2 are (1) a strong downward trend of diversification, (2) industrial variation in the cyclicality of diversification and (3) heterogeneous movement of firm level diversification. And the three key empirical results of volatility in Chapter 3 are (1) decreased aggregate volatility, (2) decreased volatility in many
34

I used total employment as the size measure.

40

industries and (3) decrease of firm level volatility. Risk-avoidance is an incentive to diversity which links these sets of findings. The hypothesis is that the change of volatility at the aggregate, industrial and firm level can affect a firm's diversification. The firm level diversification index is regressed on the volatility of aggregate, industrial and firm level profit rate.

d it = ? 0 + ? 1 AGGVOLt + ? 2 INDVOLst + ? 3 IDIVOLit + ? 4 X it + ? it , where i = firm i s = 2 - digit industry of firm i X = firm level characteristics

d is one of the firm level diversification measures.. AGGVOL is the volatility of the average of firm level profit rates( ? ( At ) ). INDVOL is the volatility of the industry level average of the deviation from aggregate profit rates ( ? ( Ast ? At ) ). IDIVOL is the volatility of the deviation of firm level profit rates from the industry average( ? ( Ait ? Ast ) ). Firm level characteristics(X) include Firm Size(SIZE), Firm Age (AGE), and the Share of Organizational Workers to the total employment(FOE). By using four measures of diversification as dependent variable, we can capture different effects of volatility and firm characteristics on diversification. D1 uses 5-digit SIC which is the most detailed information available on products in ASM firms and it is the benchmark case of estimations. D2 adds distance measure to D1 using correlation of industries in which the firm diversifies. D2 is bigger than D1 when the firm diversifies in uncorrelated or negatively correlated industries. Therefore, effects of right hand side variables will be magnified for firms with D2 higher than D1. When we use 3-digit or 4-digit SIC (D3 and D4), we only consider

41

across-industry diversification. The same amount of change in right hand side variable has different effects on these different measures of diversification and a comparison of coefficients shows whether the firm reacts most sensitively with its diversification across 3-digit, 4-digit or 5digit industry. Table 2 shows the results of firm level regressions using the left-censored Tobit estimation method. By definition, single-product producers have a diversification index equal to 0. Therefore, the left-censored Tobit model is appropriate because we have a mass point at 0 for the dependent variable. I use 10 year rolling window to get volatility, but volatility in 1974-1977 and 1994-1998 can use less than ten years of observation. Therefore, I showed the estimation results for the total sample period (1974-1998) and the period of 1978-1993 to check the robustness of estimation. I repeated the regression using Diversification Index(D1), Index with distance weight(D2), Index using 3-digit SIC(D3) and Index using 4-digit SIC(D4) as the left-hand side variable. Time trend(YEAR) and location(REGION) are controlled as fixed effects. In the sample period of 1974-1998, the coefficient estimates for volatility (AGGVOL, INDVOL, IDIOVOL) are in all cases and they are statistically significant for most cases. Coefficients for AGGVOL are different by the specification, but coefficients for INDVOL and IDIOVOL are relatively stable and robust. Coefficients for AGGVOL, INDVOL and IDIOVOL are different depending which measure of diversification is used as the left-hand side variable. However, the sign of the estimates remains positive and the order of magnitude (AGGVOL>IDIOVOL>INDVOL) are the same with D1 and D3 as the dependent variable. Estimation result for time period 1978-1993 is very similar to the result for 1974-1998. The sign and order of magnitude are not affected by the choice of left-hand side variable or specification. The result shows that diversification responds to aggregate volatility, industry volatility, and the idiosyncratic volatility of firm performance relative to those of other firms in the sector. When

42

other idiosyncratic firm level characteristics (SIZE, AGE, FOE) are included in the estimation, they reduce the level of IDIOVOL and INDVOL coefficients. Decreased aggregate volatility can reduce diversification by a great amount. In the specification IV for sample period 1978-1994, on average, 1% change in aggregate volatility (AGGVOL) will reduce diversification by .9% in 3-digit (D3), .93% in 4-digit (D4), and 1.4% in 5-digit (D1). When the aggregate volatility falls, firms reduce diversification at all levels, 3, 4 or 5-digit industries, but the biggest decrease occurs at the 5-digit SIC level diversification (D1). It suggests that firms specialize within (3 or 4 digit) industries but relatively diversify across multiple industries when volatility declines, which is consistent with other results in the previous chapter. The decrease of diversification is even bigger when we consider the distance between diversified industries (D2). If the aggregate volatility decreases by 1%, diversification decreases by 1.73% for D2. Estimates for coefficients of INDVOL or IDIOVOL do not show much difference among one another. On average, 1% change in the industry volatility will reduce diversification measures by 0.01-0.05%. Likewise, 1% change in the idiosyncratic volatility will reduce an average firm's diversification measures by 0.01-0.02%. The results in Table 2 show that the effect of aggregate volatility change on diversification is sensitive to the measurement of diversification. D3 uses 3-digit SIC and it is most closely linked to the "segment" which is widely used in diversification literature as the definition of industry. From the fact the aggregate volatility have decreased in last three decades in U.S., I find that the decrease in aggregate volatility can have contributed to the decrease in diversification. There is little difference in magnitude of this effect on D3 and D4, which suggests that the average firm has changed its diversification by changing the product portfolio across 3digit industries, not 4-digit when aggregate volatility falls. However, the fact that the magnitude of this effect is much bigger for D1 suggests that the average firm has reduced its diversification

43

across 5-digit industries by a lot. The coefficient is biggest for D2 where the firm diversifies across non-correlated or negatively correlated industries. A firm that has diversified across many 5-digit industries and across non-correlated or negatively correlated industries will have biggest decreased in its diversification with the same amount of changes in volatility if we measure diversification as D2. It predicts that we will observe much bigger decrease in diversification by industry (5-digit) than diversification by segments (3-digit) when the aggregate volatility declines. If we study diversification only using segments of firms, we may not be able to capture this high underlying degree of specialization at 5-digit industry level. Regression results suggest that firm diversification responds positively to the volatility of aggregate, industrial, and idiosyncratic profit shocks. As the aggregate volatility has decreased in the U.S. manufacturing sector, firms have had less incentive to diversify against bad aggregate shocks. Industrial volatility has the same effect on firm level diversification. Idiosyncratic volatility decreased in the late 1980s, suggesting that firms have less incentive to diversify to hedge against idiosyncratic shocks. Aggregate volatility plays a big role in explaining the change of diversification. Although firms cannot hedge themselves against aggregate volatility by diversification, they still adjust diversification in response to the aggregate shocks, which might include business trend, changes in the financial environment, or business regulation changes.

44

Table 2 Left-censored Tobit Estimation (Firm Level)
Dependent Variable=D1, D2, D3, and D4 (Firm level diversification index) Fixed Effects= YEAR, REGION Name of Distribution=Normal Sample Period: 1974-1998 Number of Observations=561 565 Non-censored Values=234 490
I d1 Intercept Aggvol Indvol Idiovol Aggprof Indrpof Idioprof Size Age Foe Intercept Aggvol Indvol Idiovol Aggprof Indrpof Idioprof Size Age Foe Intercept Aggvol Indvol Idiovol Aggprof Indrpof Idioprof Size Age Foe Intercept Aggvol Indvol Idiovol Aggprof Indrpof Idioprof Size Age Foe coeff -0.37 0.61 0.04 0.05 0.18 0.04 0.05 ** ** ** ** ** ** ** std 0.006 0.150 0.020 0.010 0.040 0.010 0.010 coeff -0.71 0.40 0.02 0.01 0.02 -0.03 -0.01 0.13 II ** ** ** ** ** ** ** ** std 0.006 0.013 0.002 0.001 0.003 0.001 0.001 0.001 coeff -0.80 0.85 0.02 0.01 0.02 -0.02 0.01 0.11 0.01 III ** ** ** ** ** ** ** ** std 0.006 0.014 0.002 0.001 0.003 0.001 0.001 0.000 0.000 coeff -0.84 0.87 0.02 0.01 0.02 -0.02 -0.01 0.11 0.01 0.17 -0.79 1.12 0.01 0.02 -0.01 -0.02 -0.01 0.11 0.01 0.16 -1.22 0.57 0.05 0.02 -0.05 -0.06 -0.04 0.14 0.01 0.17 -1.15 0.69 0.01 0.02 0.00 -0.03 -0.01 0.13 0.01 0.15 IV ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** std 0.006 0.014 0.002 0.001 0.003 0.001 0.001 0.000 0.000 0.005 0.006 0.014 0.002 0.001 0.004 0.001 0.001 0.000 0.000 0.006 0.008 0.019 0.003 0.002 0.005 0.001 0.001 0.001 0.000 0.008 0.008 0.010 0.003 0.001 0.004 0.001 0.001 0.001 0.000 0.007

d2

-0.32 0.85 0.03 0.05 0.14 0.04 0.05

** ** ** ** ** ** **

0.006 0.150 0.020 0.010 0.040 0.010 0.010

-0.66 0.67 0.02 0.02 -0.01 -0.02 -0.01 0.13

** ** ** ** * ** ** **

0.006 0.013 0.002 0.001 0.003 0.001 0.001 0.001

-0.75 1.10 0.01 0.02 -0.01 -0.02 -0.01 0.11 0.01

** ** ** ** ** ** ** **

0.006 0.014 0.002 0.001 0.004 0.001 0.001 0.000 0.000

d3

-0.68 0.43 0.04 0.06 0.12 0.02 0.04

** ** ** ** ** ** **

0.009 0.021 0.003 0.002 0.006 0.002 0.001

-1.08 0.13 0.04 0.03 -0.04 -0.07 -0.04 0.16

** * ** ** ** ** ** **

0.008 0.010 0.003 0.002 0.005 0.001 0.001 0.001

-1.18 0.55 0.05 0.02 -0.04 -0.06 -0.04 0.14 0.01

** ** ** ** ** ** ** **

0.008 0.019 0.003 0.002 0.005 0.001 0.001 0.001 0.000

d4

-0.60 0.44 0.00 0.06 0.16 0.04 0.05

** ** ** ** ** **

0.008 0.010 0.003 0.002 0.005 0.002 0.001

-1.00 0.16 0.01 0.02 0.00 -0.03 -0.01 0.15

** ** ** ** ** ** **

0.007 0.010 0.003 0.001 0.004 0.001 0.001 0.001

-1.11 0.67 0.01 0.02 0.00 -0.03 -0.01 0.13 0.01

** ** ** ** ** ** **

0.007 0.010 0.003 0.001 0.004 0.001 0.001 0.001 0.000

Note: * significance at the 95% level, ** significance at the 99% level REGION: represents 9 different geographical locations in the data. See Appendix C.

45

Table 2 Left-censored Tobit Estimation (continued)
Dependent Variable=D1, D2, D3 and D4 (Firm level diversification index) Fixed Effects= YEAR, REGION Name of Distribution=Normal Sample Period: 1978-1993 Number of Observations=359 177 Non-censored Values=156 234
I d1 Intercept Aggvol Indvol Idiovol Aggprof Indrpof Idioprof Size Age Foe Intercept Aggvol Indvol Idiovol Aggprof Indrpof Idioprof Size Age Foe Intercept Aggvol Indvol Idiovol Aggprof Indrpof Idioprof Size Age Foe Intercept Aggvol Indvol Idiovol Aggprof Indrpof Idioprof Size Age Foe coeff -0.26 -0.07 0.04 0.07 0.18 0.02 0.05 ** ** ** ** ** ** std 0.007 0.040 0.003 0.002 0.006 0.002 0.001 coeff -0.63 0.04 0.02 0.01 -0.03 -0.02 -0.01 0.13 ** ** ** ** ** ** ** ** II std 0.006 0.030 0.002 0.002 0.005 0.001 0.001 0.001 coeff -0.74 1.36 0.02 0.01 -0.06 -0.01 0.00 0.11 0.01 III ** ** ** ** ** ** ** ** std 0.006 0.030 0.002 0.002 0.005 0.001 0.001 0.001 0.000 coeff -0.78 1.40 0.02 0.01 -0.07 -0.01 0.00 0.11 0.01 0.16 -0.73 1.73 0.00 0.01 -0.13 -0.01 -0.01 0.11 0.01 0.15 -1.15 0.90 0.05 0.02 -0.14 -0.06 -0.04 0.14 0.01 0.16 -1.07 0.93 0.01 0.02 -0.08 -0.02 -0.01 0.13 0.01 0.14 IV ** ** ** ** ** ** * ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** ** std 0.007 0.030 0.002 0.002 0.005 0.001 0.001 0.001 0.000 0.007 0.007 0.030 0.002 0.002 0.005 0.001 0.001 0.001 0.000 0.007 0.010 0.049 0.003 0.002 0.007 0.002 0.001 0.001 0.000 0.010 0.009 0.040 0.003 0.002 0.006 0.002 0.001 0.001 0.000 0.009

d2

-0.21 0.27 0.02 0.07 0.12 0.03 0.05

** ** ** ** ** ** **

0.007 0.040 0.003 0.002 0.006 0.002 0.001

-0.59 1.00 0.01 0.01 -0.10 -0.02 -0.01 0.13

** ** ** ** ** ** **

0.006 0.030 0.002 0.002 0.005 0.001 0.001 0.001

-0.69 1.69 0.00 0.01 -0.13 -0.01 0.00 0.11 0.01

** ** ** ** ** ** **

0.006 0.030 0.002 0.002 0.005 0.001 0.001 0.001 0.000

d3

-0.56 0.57 0.04 0.10 0.14 0.00 0.03

** ** ** ** ** * **

0.010 0.050 0.004 0.003 0.008 0.002 0.001

-0.98 0.20 0.05 0.03 -0.12 -0.07 -0.04 0.16

** * ** ** ** ** ** **

0.009 0.047 0.003 0.002 0.007 0.002 0.001 0.001

-1.11 0.87 0.05 0.02 -0.14 -0.06 -0.04 0.14 0.01

** ** ** ** ** ** ** ** **

0.008 0.019 0.003 0.002 0.005 0.001 0.001 0.001 0.000

d4

-0.47 0.67 0.00 0.09 0.20 0.02 0.04

** ** ** ** ** **

0.009 0.050 0.004 0.002 0.008 0.002 0.001

-0.89 0.11 0.01 0.02 -0.05 -0.03 -0.01 0.15

** * ** ** ** ** ** **

0.008 0.040 0.003 0.002 0.006 0.002 0.001 0.001

-1.03 0.90 0.01 0.02 -0.08 -0.02 -0.01 0.13 0.01

** ** ** ** ** ** ** **

0.009 0.040 0.003 0.002 0.006 0.002 0.001 0.001 0.000

Note: * significance at the 95% level, ** significance at the 99% level REGION: represents 9 different geographical locations in the data. See Appendix C.

46

Section 5. Conclusion
This chapter shows a new empirical relationship between the diversification and profit volatility. Micro level data for the manufacturing sector allow us to verify some of the stylized facts about volatility which are discussed in the literature. Using firm level profit rates, I find:

(1) Aggregate volatility declined over my sample period. (2) Volatility decreased since the 1980s for most industries. (3) The mean of firm level idiosyncratic volatility decreased in the late 1980s and the crosssectional standard deviation of volatility did not change much.

Findings (1) and (2) are consistent with trends of volatility that have been established with aggregate data in the literature. Finding (3) is contrary to the upward trend of idiosyncratic volatility which has been found in other studies with large firm data. Although I cannot find evidence of a discontinuous drop in the middle of the 1980s as has been argued in the literature, the volatility of aggregate, industrial and idiosyncratic profit shocks has been falling in the manufacturing sector since 1980s. The left-censored Tobit regression shows that firm level diversification is positively affected by aggregate, industrial and idiosyncratic profit volatility. Therefore, the decrease of volatility in US manufacturing has contributed to the decrease of diversification. As we have seen in the text, the overall volatility decreased, industry level volatility also decreased in many industries, and finally, idiosyncratic volatility has been reduced over my sample period. Firms have less incentive to diversify and the diversification index clearly shows a downward trend. I

47

have not settled the arguments about the cause of the decreased volatility, but the effect of the volatility change on firm level diversification is clearly shown in this chapter. The comparison of estimation results across different measures of diversification shows us that changes in volatility strongly affect diversification across 5-digit SIC industries rather than 4 or 3-digit industries. With volatility decreasing over time, firms specialize within (3 or 4-digit) industries, and diversify across multiple industries. This is consistent with other results in the paper and this trend of diversification may continue if the volatility keeps decreasing. This chapter is a stepping stone for empirical analysis of the motives for diversification. If risk-avoidance is the biggest motive for diversification, firms don't need to diversify as much when volatility decreases. But I have not examined other factors that could cause changes of diversification. For example, firms may specialize to enhance productivity. One of the next research topics would be whether the trend toward specialization indeed increased productivity at the firm level. Another research topic would be a counterfactual analysis. If volatility did not decrease, then would diversification have increased? It is not easy to answer this counterfactual question in U.S., but it can be answered by looking at other countries, especially developing countries. Many countries in Asia and Latin America have been struggling with high economic volatility over the last several decades. If firm level diversification trends are analyzed for these countries, we will be able to clarify the relationship between the diversification and economic volatility.

48

Chapter 4: Volatility Change and Government Investment35

Section 1. Introduction
During the 1830s and 1840s, the British and American economies experienced a series of shared macroeconomic fluctuations. A sharp financial crisis in May of 1837 was followed by a brief recovery in 1838 and 1839. A second financial crisis in October of 1839, while less severe than the panic in 1837, nonetheless produced a recession and deflation that lasted until 1843. A third financial crisis, in the winter of 1842, affected primarily the United States, although conditions continued to deteriorate in Britain through 1842 as well. These fluctuations lead to a violent change in the volatility of economy. Figure 1 plots the volatility which is measured by the standard deviation of bond price in 12 months.36 It shows a sharp increase of the volatility both in US and UK after 1839. The US and UK economies were closely linked by trade and finance, leading historians to speculate about the role of each country in provoking the crises. Temin's Jacksonian Economy attributes the Panic of 1837 and the Crisis of 1839 to the Bank of England and international factors, absolving the Bank of the United States, Nicholas Biddle (the bank's president from 1823 to 1839), and President Andrew Jackson. Biddle himself criticized the Bank of England for its policies in 1839, as did Jenks and Hammond.37 On the other side, in A Study in Trade-Cycle

35

This chapter is co-authored with John Joseph Wallis in University of Maryland. A revised version of this chapter, "The Market for American State Government Bonds in Britain and the United States, 1830 to 1843," will be published in Economic History Review, November 2005 (forthcoming). 36 See Chapter 3 for a formal definition. 37 Leland Jenks, Migration of British Capital, pp. 90-95, and Bray Hammond, Banks and Politics in America, pp. 500-513, stress the importance of British capital markets and international forces in bringing on the crises. Nicholas Biddle, in a letter to John Clayton dated April 9, 1841, in which he defended his actions at the Bank of the United States and attempted to

49

Figure 1 Volatility of average state bond yield to maturity
Volatility of average State Bond YTM in London and American Market 0.018 0.016 0.014 0.012 0.010 0.008 0.006 0.004 0.002 0.000 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 volatility (us) volatility (uk)

History, Matthews concluded that '... it is in the nature of things futile to try and draw any hardand-fast line assigning to either country causal primacy in the cycle as a whole or in its individual phases. But enough has been said in the present chapter to indicate the powerful nature of forces making for instability from within the United States in this period.'38 The market for American state debts played a central role in financial relationships

explain why the bank had failed after his departure as President, Biddle wrote: 'I have just stated that the winter of 1838-'39 was a season of great abundance and ease in moneyed concerns, both in England and this country; but England was soon after startled by the discovery that the grain crop was deficient, and a demand arose for specie to export for grain, combined with some continental loans, that changed the whole surface of affairs. The Bank of England itself, after borrowing ten millions of dollars from the Bank of France, was still so much drained for coin that it was forced into very severe restrictive measures, which raised the interest of money to twice or three times its usual rate. The most injurious effect was on the stocks of this country [the U.S.], which were no longer convertible in England, accept at great sacrifices. These causes immediately reacted on this country, producing the usual effects of embarrassment in the community and alarm among the banks.' In House Document #226, 29th Congress, First Session, p. 488. 38 Matthews, Trade-Cycle, p. 69.

50

between Britain and the United States. In the late 1830s American states embarked on an internal improvement boom, raising the amount of state debt outstanding from $81 million in 1835 to $198 million in 1841. American states authorized and issued bonds worth $13 million in 1836, $21 million in 1837, $35 million in 1838, $22 million in 1839, $19 million in 1840, and $6 million in 1841. By 1841, estimates are that half of the $200 million in state debt was held abroad, primarily in Britain.39 State bonds provided a critical link between financial markets in the two countries. By 1836, state bonds were the only long-term American debt instrument traded in Britain. The United States federal government retired all its debt in 1835. The single American corporation whose stock traded regularly in London was the Second Bank of the United States, which lost its national charter in 1836. Millions of dollars of identical state bonds traded in London, New York, and Philadelphia. Movements in bond prices give us a window into the connections between British and American financial markets. The boom in state transportation and banking projects, and the associated wave of new state bond issues, also play a critical role in our understanding of macroeconomic events. Temin attributed the quick recovery of the American economy from the Panic of 1837 to state expenditures for canals and railroads, financed largely by British lending. 'The recession of 183738... was brought to a speedy end by the restoration of the capital flow from Britain to the United States and by the expansion of demand stemming from the rise in state government expenditures.' Temin attributed the 1839 crisis to credit tightening by the Bank of England and the long recession that followed to tightening markets in Britain for American state debts: 'The state projects initiated in the late 1830s had been started in the expectation of external [British] financing.... Unfortunately, the new inflow of foreign capital did not continue [in 1839]... and the

39

See, for example, Scheiber's Ohio Canal Era estimates of foreign holders of Ohio bonds, Ratchford's American State Debts, and McGrane Foreign Bondholders.

51

manifold projects of the states were abandoned.'40 By the summer of 1842, eight states and the Territory of Florida were in default on their debts, and Mississippi and Florida had repudiated their bonds outright. The collapse of state credit was the most serious consequence of the depression that began in 1839.41 The purpose here is to determine whether credit markets for American state bonds in and between the three major financial crisis were tighter in the United States or in Britain, and whether shocks to the volatility of bond markets originated in the United States or in London.42 Although there are some subtleties of interpretation, the major questions are straightforward and their answers are quantitative. First, were British and American financial markets well integrated? Not surprisingly, we find that they were. Figure 2 gives the average bond yield for state bonds in London and for state bonds in New York.43 Financial markets effectively arbitraged the prices of American state bonds in London and the U.S. within a band of plus or minus roughly 1 percent (100 basis points), attributable to the high transaction costs of trans-Atlantic commerce in this period, with a lag of roughly two months.

40

The first quote is from Temin, The Jacksonian Economy, p. 151 and the second quote from p. 153. 41 Temin, Jacksonian Economy, p, 157, citing Gallman's unpublished estimates of annual GNP, argues that the crises in the United States had a much larger effect on prices than on output. Also see Temin, 'The Anglo-American Business Cycle' where he shows that the American economy experienced greater price fluctuations over these business cycles, while the British economy experienced greater fluctuations in real economic activity. 42 By tighter we mean simply that bond yields were higher, as we have no information on credit rationing in either market. It appears, however, that states willing to pay market rates could issue bonds in New York and London up to 1842. 43 Source: Sylla, Wilson, and Wright, Price Quotations. Figure 2 shows average yields to maturity in the London market between 1831 and 1843 for the bonds of New York, Pennsylvania, Ohio, Indiana, Illinois, and Massachusetts; and average yield to maturity in the United States for the bonds of New York, Pennsylvania, Ohio, and Illinois.

52

Figure 2
0.70 0.60 0.50 London 0.40 0.30 0.20 New York 0.10 0.00 Jan-31

All State Bond Yields
Average in US and in London

Jan-33

Jan-35

Jan-37

Jan-39

Jan-41

Jan-43

Jan-45

Figure 3
0.1 0.08 0.06 0.04 difference 0.02 0

Difference in Ohio Bond Yields
New York minus London

-0.02 -0.04 -0.06 Jan-31

Jan-33

Jan-35

Jan-37

Jan-39

Jan-41

Jan-43

Jan-45

US - London Difference

53

Because markets were integrated we can ask whether the pattern of bond price movements in the Crisis of 1839 and the Collapse of 1842 were consistent with shocks originating in the United States, in Britain, or neither country. We find clear evidence that shocks in both crises originated in the United States. Figure 3 shows the difference in contemporaneous bond yields for Ohio bonds in New York and London, and Figure 4 shows the contemporaneous bond yields for New York state bonds in New York and London.44 Because of the lag with which price shocks were transmitted from one market to the other, we can see where shocks originate. In October of 1839 and again in the winter of 1842, bond prices moved sharply higher in New York, two months before prices moved in London. This is clear evidence that the "shocks" of 1839 and 1842 originated in the United States. Finally, as Temin suggests, we ask whether bond price movements show whether British investors were more willing to lend to American states, relative to American investors, during the state borrowing boom from 1837 to 1839, and then became less willing to lend to states after October of 1839. Table 1 presents average bond yields for New York and Ohio bonds in New York and London, as well as the average difference in bond yields for both states, for the three relevant time periods. Before the Panic of 1837 (January 1831 to March 1837), yields on New York and Ohio bonds were close to the same in both markets. After the Panic (July 1837 to July 1839), when states began borrowing in earnest, bond yields were distinctly higher in London for bonds from both states. New York state bonds paid yields of 4.49 percent in New York and 5.01 percent in London, while Ohio bonds paid yields of 4.16 percent in New York and 5.06 percent in London. Conversely, after the Crisis in October of 1839 (January 1840 to October 1841), yields on American state bonds were generally lower in London than they were in US. New York state bonds paid yields of 6.92 percent in New York and only 6.06 percent in London, while Ohio

44

Source: Sylla, Wilson, and Wright, Price Quotations. All bond yields are calculated yields to maturity, averaged over all the bonds for an individual state.

54

bonds paid 6.53 percent in New York and 6.51 percent in London. There is nothing in these numbers to suggest that British lenders cut off credit to US, after the Crisis of 1839. The crises of 1839 and 1842 clearly began in the United States. Between the Panic of 1837 and the Crisis of 1839 credit markets for state bonds were distinctly tighter in London than in the United States. Between the Crisis of 1839 and the Collapse of 1842, credit markets for state debt in the United States were tighter than markets in London. We find little evidence that state borrowing and the market for state bonds collapsed because of pressures emanating from Britain after 1839. The next section provides a brief history of state borrowing and the macro-economy in the 1830s. The second section looks at the data sources on state bonds and the question of market integration. The final section of the paper examines the pattern of state borrowing in the 1830s and identifies the forces operating in America that moved the market for state bonds. Economic historians have focused on the Panic of 1837, paid some attention to the Crisis of 1839, and ignored the Collapse of 1842. Figure 2 suggests that as far as the market for state bonds was concerned, 1839 and 1842 are more interesting years to study, and that by overlooking the Collapse of 1842, a crisis neglected in virtually all accounts of this era, we may have missed the biggest crisis of them all. What happened in 1842?

55

Figure 4
0.1 0.08 0.06 difference 0.04 0.02 0 -0.02 Jan-31

Difference in Yields, NY Bonds
New York minus London

Jan-33

Jan-35

Jan-37

Jan-39

Jan-41

Jan-43

Jan-45

US - London Difference

Table 1 Average Bond Yields for New York and Ohio Bonds In London and New York, and difference in yields Average Bond Yield New York Bonds in London NY Average Bond Yield Ohio Bonds in London NY NY-London Yield Difference New York Ohio

Average Yield 3.28% 3.34% 3.76% 3.76% 0.10% -0.08% 1/31 to 3/37 Standard 0.52% 0.55% 0.31% 0.59% 0.43% 0.48% Deviation Average Yield 5.01% 4.49% 5.06% 4.16% -0.56% -0.88% 7/37 to 7/39 Standard 0.35% 0.36% 0.28% 0.50% 0.72% 0.62% Deviation Average Yield 6.06% 6.92% 6.51% 6.53% 0.76% 0.06% 1/40 to 10/41 Standard 0.44% 0.84% 0.43% 1.02% 0.64% 1.07% Deviation Calculated from Sylla, Wilson, and Wright, Price Quotations. Note that average yields are for all dates with an observation. The difference in yields is calculated only for dates with observations in both markets.

56

Section 2. The History
The early 1830s were a period of general economic expansion in both Britain and the United States, marred by a brief recession in 1834. The expansion turned into a boom in 1835, driven by a rapid increase in public land sales in the United States. The boom was reflected in rising prices in both countries, an increase in international trade, and an increase in the flow of capital from Britain to the United States. Prices stopped rising in early 1837, and a sharp break in cotton prices combined with tight credit conditions in Britain and the United States to produce a financial panic in May of 1837. In the United States the panic resulted in the suspension of specie payments by banks throughout the country, and in Britain the failure or near failure of several large commercial houses engaged in the American trade. The Bank of England did its part to bring about the panic by raising the Bank Rate from 4 to 5 percent.45 Figure 5 shows the Bank Rate, short term interest rates in London and New York, and the New York price of 60 day bills payable in London.46

45 46

Clapham, Bank of England, p. 153, and Hidy, House of Baring, pp. 205-24. Interest rates in the New York and Boston are the average of the high and low rates reported in Smith and Cole, Fluctuations, Table 74, pp. 192-3. Interest rates in London: From National Bureau of Economic Research. Bank Rate: Clapham, Bank of England, vol II, Appendix B, p. 199. Exchange Rates on 60 day bills, Smith and Cole, Fluctuations, p. 190 and Officer, "Integration in the American Foreign Exchange Market," p. 563.

57

Figure 5

The effects of financial tightening were compounded in the United States by the decision of the federal government to distribute the federal fiscal surplus of $36 million to the states in 1837, and by President Jackson's specie circular requiring that all public land purchases be redeemed in specie. The two measures together disrupted the normal allocation of gold reserves within the banking system, further exacerbating the liquidity problems of New York banks brought on by tightening international markets. Whether the Panic of 1837 in the United States was caused primarily by international or domestic forces is a question with a long pedigree that we do not attempt to answer.47 The Panic of 1837 was followed by a year of bank specie suspensions in the United
47

See Rousseau, 'Jacksonian Monetary Policy;' Temin Jacksonian Economy and 'The AngloAmerican Business Cycle, 1820-60;' and Timberlake, 'The Specie Circular and the Distribution of the Surplus' and 'The Specie Standard and Central Banking in the United States Before 1860;' and Macesich 'Sources of Monetary Disturbances in the United States, 18341845.'

58

States, financial distress in Britain, deflation in both countries, and a sharp decline in the volume of trade in 1838. But the recession was short lived. By the fall of 1838 land sales, international trade, prices, and capital flows had all turned up again. Banks in the United States resumed specie payments in the summer of 1838. As Temin stressed, the quick recovery in the United States was partly the result of fiscal stimulus created by the rapid expansion of state borrowing to build canals, railroads, and banks. Mid-Atlantic states had been borrowing since the 1820s to build canal networks, beginning with New York's Erie Canal in 1817. In 1836 a second wave of borrowing began, both in the older states - New York, Pennsylvania, Maryland, and Ohio - and in a new group of states in the west - Indiana, Illinois, Michigan, Arkansas, and Mississippi. Table 2 provides debt outstanding by state in1841, the share of debt outstanding authorized after 1836, and the total amount authorized in 1839, 1840, and 1841.48 This was peacetime fiscal expansion on a scale never witnessed in the young United States.

48.

Table 1 reports debt outstanding on September 1, 1841. The information is taken from the William Cost Johnson Report, Report of Committees, House of Representatives, 27th Congress, 3d session, Report No. 296.. The Report gives debt outstanding by year of authorization, not by year of issue. So debt authorized in 1836 was issued in or after 1836.

59

Table 2 Amount of State Debt Outstanding on Sept.1, 1841, Percentage of Debt authorized between 1836 and 1841, and amount authorized between 1839 and 1841. State Total Debt Outstanding Share Authorized 1836 and later 64% 100% 3% 100% 100% 83% 94% 7% 100% 68% 100% 98% 71% 100% 71% 59% 36% 74% 84% 45% Debt Authorized 1839, 1840, 1841 $0 $0 $100,000 $0 $9,862,293 $1,363,000 $1,445,500 $1,185,000 $1,465,085 $994,854 $1,869,137 $40,000 $0 $410,261 $8,049,755 $3,994,123 $13,202,084 $600,000 $0 $2,416,729

Alabama $15,400,000 Arkansas $2,676,000 Florida $4,000,000 Georgia $1,324,550 Illinois $13,527,293 Indiana $12,751,000 Kentucky $3,085,500 Louisiana $23,985,000 Maine $1,734,861 Maryland $15,214,761 Massachusetts $5,969,137 Michigan $5,611,000 Mississippi $7,000,000 Missouri $842,261 New York $21,796,768 Ohio $10,924,123 Pennsylvania $36,336,043 South Carolina $3,691,234 Tennessee $3,416,166 Virginia $8,744,308 Total $198,030,005 59% $46,997,820 Outstanding Source: "The William Cost Johnson Report." House Report, 296, 27th Congress, 3rd Session, 1843. The numbers for Ohio in the Johnson report are unreliable for the later years. We include Scheiber's, Ohio Canals, estimates of borrowing for 1840 and 1841, pp. 143-151, and the $20 million figure cited in the Census of 1880.

The transportation boom, however, died quickly in the Northwest. Indiana, Illinois, and Michigan all sold bonds on credit to eastern investment banks. These new states issued bonds for which they were liable for interest payments immediately, but for which they would receive payments only in installments.49 In July of 1839, the Morris Canal and Banking Company of New Jersey defaulted on Indiana, and the state quickly was forced to curtail construction on its network of canals and railroads. By the fall, Illinois and Michigan were forced to slow or stop

49

The installments were fixed in time and amount. The states were not paid when the banks sold the bonds, these were not consignment or commission sales.

60

Table 3 Default, Resumption, and Repudiation Dates State Date Resumed or Repudiated Date Indiana January 1841 Resumed July 1847 Florida January 1841 Repudiated February 1842 Mississippi March 1841 Repudiated February 1842 Arkansas July 1841 Resumed July 1869 Repudiated July 1884, Holford Bonds Michigan July 1841 Resumed January 1846 Repudiated Partially Part paid bonds, July 1849 Illinois January 1842 Resumed July 1846 Maryland January 1842 Resumed July 1848 Pennsylvania August 1842 Resumed February 1845 Louisiana February 1843 Resumed 1844 Repudiated ? Source: English, "Sovereign Default" Note: Louisiana never formally repudiated any bonds, thus the uncertainty of the date of Louisiana's repudiation. See English for a discussion of Louisiana's repayment of these bonds. construction when investment banks defaulted on their obligations to the states. Land sales and land values in these northwestern states had been rising steadily through the 1830s. When transportation construction stopped, land values and property tax revenues began falling and, by late 1839, it was apparent that these states would soon have trouble servicing their debts.50 In January of 1841, Indiana was the first state to default on interest payments. Table 3 lists the states that defaulted on interest payments, the date of default, whether the state resumed payments or repudiated their debts, and if they resumed, the date of resumption. The collapse of internal improvement projects in the Northwest was not the only economic problem in 1839. The Bank of England, again facing drains on its specie reserves, began raising the Bank Rate in the summer (Figure 5). A crisis broke out in the United States
50

For detailed consideration of land values and property tax revenues in Indiana in these years see Wallis 'The Property Tax.' Only Illinois continued to borrow, at extremely high rates, in an attempt to maintain its credit and to continue construction. The state was not successful at either goal. Heavy borrowing in 1840 saddled the state with debts the state would struggle to pay into the 1850s, without any significant physical accomplishments. The best overall history of state investments in transportation is Goodrich Government Promotion. Goodrich has recently been supplemented by Larson Internal Improvements. Details about Indiana can be found in Fatout Indiana Canals and Illinois in Krenkel Illinois Internal Improvements. The default crisis is discussed at length in Wallis, Sylla, and Grinath, 'Soveriegn Default and Repudiation.'

61

when the Bank of the United States of Pennsylvania (the BUSP) suspended specie payments in October. This was followed by suspensions throughout the western and southern parts of the United States, but not in New York and New England. As Hammond and Smith emphasize, the BUSP's immediate problem in 1839 was domestic, not foreign. Pressure from New York and Boston banks forced the BUSP to suspend.51 Although 1839 marked the end of the Northwestern transportation boom, New York, Ohio, and Pennsylvania continued to authorize new debt issues for their canals (Table 2). Despite rising interest rates, $47 million in debts were authorized and issued in 1839, 1840, and 1841. We can test Temin's conjecture that the end of British willingness to lend to American states after 1839 brought on the crisis and contributed to the depression that followed.

Section 3. Data sources and Market Integration Tests
Figure 2 is constructed from data collected by Richard Sylla, Jack Wilson, and Robert Wright (SWW). 52 They gathered quotations on debt and equity prices from contemporary newspapers in London, New York, Philadelphia, Boston, Baltimore, and other American cities.53 Prices are available for American markets from the early 1790s up to the 1850s. Prices in London

51

By 1839 the BUSP had an extensive operation in Britain headed by Samuel Jaudon, so attributing the causes of the bank's demise to domestic and international forces is complicated. But the causes of the suspension in October of 1839 were a run on the Philadelphia bank by banks in New York and Boston. Hammond, 'Chestnut Raid on Wall Street;' Smith, Economic Aspects. 52 Their database will soon be available at ICPSR: 'Price Quotations in Early U.S. Securities Markets.' 53 Price quotations were typically reported weekly, recorded by the date of the newspaper issue. Prices were not quoted on the same day in each market, and in several cases quotes were provided for more than one day in each week. Our analysis focuses on weekly quotations, except where noted. The Boston market data are not available from June 1841 to September 1843, and the New York series on Massachusetts bonds is incomplete. The Baltimore market data include complete data on generic 'Maryland 5s' and 'Maryland 6s' without maturity dates, and the prices for specific Maryland bonds is spotty.

62

are only available from 1811 to 1843. State government bonds typically traded in New York, Philadelphia, and London, as well as in the regional market of issue (for example, Maryland bonds in Baltimore and Massachusetts bonds in Boston). SWW list over 100 bonds from 18 states trading at some point in London. Trading occurred in new issues and the secondary market. Bonds traded actively for a few months after they were issued, but perhaps because bonds were held mostly by long-term investors, relatively few bonds continued to trade regularly in the secondary market. The most consistent series are available for New York, Ohio, and Pennsylvania. New issues were marketed by the states themselves and through the agency of investment banking intermediaries. Legislation authorizing bond issues typically required that bonds be sold at par or better. The par restriction clearly applied to new issues marketed by states, sometimes applied to issues by intermediaries, and never applied to the secondary market. 54 When prices in the secondary market dropped below par, states and their agents could not sell new bonds at par. States could accurately claim that new bonds could not be sold in New York or London, even when simultaneously there was an active secondary market in state bonds in both markets. What states often failed to say is that there was no market for bonds with par sales restrictions when the market price fell below par. The inability of states to market their bonds was usually a function of their unwillingness (or their agents' inability) to borrow at higher interest

54

The restrictions states placed on intermediaries are difficult to track. When the state appointed a state official to sell bonds in New York or London, the official was clearly bound by the par restriction. When states used investment bankers the situation was less clear. One would think that once investment bankers had paid for the bonds, they would no longer be bound by the par restriction. Investment bankers who took consignment of bonds would be bound by the par restriction. Yet, for example, Nicholas Biddle and the BUSP took almost all of Mississippi's 1838 issue of $5 million, paid for it on credit over the following year, and then failed to sell the bonds in London. The BUSP used $3,008,000 in Mississippi bonds as collateral for European loans, Smith, Economic Aspects, p. 218. Altogether, the BUSP used almost $13,000,000 in state bonds as collateral for loans in the fall of 1839 and winter of 1840. It is not clear why Biddle didn't sell the bonds, unless, perhaps, he could not because of concerns about par restrictions.

63

rates. States that were willing to borrow at market rates could always borrow. Fortunately, most state bonds are reported with their yield and maturity, e.g., 'New York 5's 1854.' This enables us to calculate, for each individual bond, its yield to maturity.55 For each state we calculate the average yield to maturity for all the bonds traded in each market, e.g., 'Ohio Bonds trading in New York.' This is a simple average because there is no information available on trading volumes to provide us with weights. There are often significant gaps in the series, and some of the short-term fluctuation results from changes in the bonds reported in a particular week. The 'United States' yields we quote for Ohio, New York, and Illinois bonds come from the New York market, and for Pennsylvania bonds the yields are from the Philadelphia market. Visual examination of the bond yields in Figure 2 suggests that the market for state bonds in London and in the United States were closely related. To investigate the relationship more formally, we ran a series of ARCH tests. Yit is log of the average bond yield in country 'i' on date 't', ai is the constant term for country i, and ,it is the market specific disturbance term:

(1) Yit = a i + ? it (2) E (? it ) = aij + bij E (? i ,t ?1? j ,t ?1 )

The errors in Equation (1) follow a multivariate normal distribution with auto regressive conditional heteroscedasticity (ARCH) as in Equation (2). 56 The dependent variable, Yi is average yield to maturity in market i, of bonds that were commonly traded in both markets. The ai are constant terms measuring the log of the average

55

For simplicity, we assumed that all bonds matured on January 1 and paid a single annual premium. Bond yields for the last two years proceeding maturity were dropped from the calculated averages. Yields were calculated for bid and ask prices, and both bid and ask yields were included in the market averages. 56 The ARCH estimator is more fully described in Appendix B.

64

bond yield in each market. The aij estimate the constant element of the covariance between yields in the two markets. The bij estimate the effect of the lagged disturbances on the covariance between the prices in the two markets. The bij measure whether the covariance of the disturbances are related to lagged disturbances. That is, for example, whether last period's errors in London affects this period's errors in New York. In an integrated market the bij should be close to one. If the errors in the two markets were related, and they were, this is evidence that the two markets were integrated. The calculation of yield to maturity for a bond requires the maturity, coupon rate and the weekly prices of the bond, and there are missing observations on Y when not all of these data are available. Missing observations can be dealt with in several ways. First, we linearly interpolate for the unobserved data, and then run regressions using weekly and monthly data. 57 Second, monthly data contain far fewer missing observations and give us a check on the robustness of the results using weekly data, but at a loss of significant number of observations. The monthly data are realistic, however, given the time lags involved in the flow of information between the U.S. and Britain in the early 19th century. Finally, we estimate a regression using only observed weekly data, using our own method of analysis as ARCH with Missing Observations (described in the Appendix B) to account for missing observations in the data. The regression results are provided in Table 4. The first column uses the weekly sample of linearly interpolated bond yields, the second column the monthly sample of linearly interpolated yields, and the third column is based only on the observed monthly data using our adjustment for missing observations. The results indicate strong evidence for integration. The constant covariance of returns in the two markets, the aij , are very close to zero. The effect of lagged disturbances on the covariance of returns, the bij, are positive and very close to one,

57

This is the first step of an EM algorithm which is a popular tool for finding maximum likelihood estimates in incomplete data problems. See Meng and Rubin (1993).

65

Table 4 Bond market integration test (Multivariate ARCH with 2 equations) Y=Only observed Y=Observed and interpolated data data Parameter Weekly Monthly (Monthly) .051 .039 a1 .039 (1400.7)** (74.2)** .047 .039 a2 .04 (1648.4)** (68.5)** .000 .000 a11 .001 (7.26)** (8.35)** 1.01 .86 b11 .70 (3.12)** (5.53)** .000 .000 a21 .001 (4.80)** (6.54)** 1.01 .65 b21 .68 (3.11)** (4.41)** .000 .000 a22 .001 (10.81)** (13.19)** 1.01 .85 .68 b22 (3.12)** (5.83)** Number of 1230 366 289 Observations Time Period 1829-1843 Technical Note: (1) 1: London market, 2: American market (2) The first dependent variable is the average YTM of NY, PA and OH bonds in London. The second dependent variable is the average YTM of NY and OH bonds in NYC and PA in Philadelphia. (3) ** denotes significance at 99% indicating that shocks to one market are quickly reflected in yields in the other market. These results are unaffected by the use of linearly interpolated weekly or monthly data, or controlling for the presence of missing observation in the design of the estimator.

For comparison, Table 5 performs a similar exercise on stock price indexes in London and in American markets. Missing observations are not a problem with the market indices. We have run regressions on weekly prices, the change in weekly prices, and monthly prices. Unlike the bond markets, where the underlying securities are the same in both markets, the equities traded in the London market are different from the equities in the American markets. As with the

66

Table 5 Stock market integration test (Multivariate ARCH with 2 equations) Weekly data Parameter a1 a2 a11 b11 a21 b21 a22 b22 Y=Log of Stock Price=log(pt) 4.72 (6078.15)** 4.60 (4585.44)** .001 (22.96)** .95 (6.88)** -.000 (-.21) .94 (6.69)** .000 (10.5)** 1.03 (7.04)** Y=Capital Gain=log(pt)-log(pt-1) -.00 (-.5) -.00 (-.1) .00 (40.7)** .47 (9.5)** -.00 (-.1) -.13 (-1.6) .00 (52.2)** .31 (6.1)** Monthly data Y=Capital Gain =log(pt)-log(pt1) -.00 (-.0) -.00 (-.7) .00 (10.4)** .62 (5.5)** .00 (.1) .11 (.6) .00 (11.9)** .49 (2.9)**

Number of 711 177 Observations Time Period 1821-1836 Technical Note: (1) 1: American market (Baltimore, Boston, New York and Philadelphia), 2: London market (2) The average price in London is indexed by the average of the first years stock prices, because the stock prices denoted in sterling and stayed around 25, where American prices stayed around 100. (3) ** denotes significance at 99% bond market, however, there is substantial evidence of market integration.

The results clearly show that the market for state bonds was well integrated. Transacting between the two markets, however, was not costless. Differences in interest rates between London and New York of one half to a whole percentage point in yields (100 basis points) were not uncommon. The transaction cost wedge between the markets is not surprising. In the 1830s bank notes of Philadelphia banks typically traded at 1 percent discount in New York in times

67

when there was no default risk. The discount merely reflected the time and effort involved in presenting the bank note to the issuing bank for redemption. It was possible for bond yields to be higher in New York than in London, but not too much higher. We find episodes when contemporaneous prices in the United States and Britain diverge, but they always return to the transaction cost bounds within a few months. There is no evidence that the trans-Atlantic market for state bonds ever became disintegrated.

Section 4. American State Bonds in London and the United States
Figure 2 shows the average yields of state bonds traded in London and the United States, but disentangling what happened in the three financial crises requires examining states individually. The five largest state borrowers were Pennsylvania, Louisiana, New York, Ohio, and Maryland (Table 2).58 Louisiana and Maryland were not steady borrowers and we do not have consistent records on their bond yields. New York began borrowing in 1817 and Ohio in 1825. Both states completed their major canal projects in the early 1830s. Both states resumed borrowing in 1837, and borrowed heavily and regularly through 1842. Hence, there are long and fairly complete bond yields for those states before 1834 and after 1837, but only sporadic information in 1835, 1836, and 1837. Pennsylvania began borrowing in the 1820s and continued to borrow through the 1830s, so there are long and fairly complete records for Pennsylvania. Illinois, Indiana, and Massachusetts did not begin borrowing heavily until 1837. We have only sporadic quotations for those three states. We focus, therefore, our analysis on the bonds of New York, Ohio, and Pennsylvania.

58

In table 2, Ohio is eighth in total debts. By 1841, Ohio had already paid back a substantial amount of its debts issued in the 1820s. Ohio would also continue to borrow in 1841, 1842, and 1843.

68

Table 6 Bond Yields of New York and Ohio Bonds in the United States and London
NY Bond London 1831 US US-London Diff 0.03% London Ohio Bond US US-London Diff -0.29%

mean 3.46% 3.49% 4.04% 3.75% st. dev. 0.18% 0.13% 0.25% 0.26% 1832 mean 3.16% 3.03% -0.12% 3.74% 3.35% -0.38% st. dev. 0.18% 0.17% 0.26% 0.37% 1833 mean 2.97% 2.84% -0.13% 3.53% 3.69% 0.16% st. dev. 0.14% 0.12% 0.08% 0.34% 1834 mean 2.90% 3.60% 0.70% 3.55% 3.86% 0.30% st. dev. 0.04% 0.34% 0.07% 0.61% 1835 mean 3.20% 3.06% -0.14% 3.61% 3.50% -0.11% st. dev. 0.08% 0.48% 0.19% 0.54% 1836 mean 4.57% 4.69% 0.12% 4.13% 4.48% 0.35% st. dev. 0.46% 0.39% 0.11% 0.78% 1837 Q1 4.71% 4.87% 0.16% --4.63% --Q2 --4.80% --5.09% 4.89% -0.20% Q3 5.46% 3.94% -1.52% 5.28% 3.54% -1.74% Q4 5.53% 4.44% -1.08% 4.97% 3.95% -1.02% st. dev. 0.40% 0.39% 0.27% 0.60% 1838 Q1 5.03% 4.34% -0.69% 5.02% 4.34% -0.68% Q2 4.76% 4.82% 0.05% 4.74% 4.79% 0.04% Q3 4.66% 4.83% 0.17% 4.91% 3.76% -1.15% Q4 4.69% ----4.81% ----st. dev. 0.19% 0.25% 0.15% 0.36% 1839 Q1 4.96% 5.00% 0.04% 5.22% 4.51% -0.71% Q2 4.95% 4.94% -0.01% 5.30% 4.32% -0.98% Q3 5.04% 5.19% 0.15% 5.96% 4.83% -1.12% Q4 6.19% 7.76% 1.57% 6.62% 7.46% 0.84% st. dev. 0.47% 1.67% 0.61% 1.69% 1840 Q1 5.62% 6.83% 1.21% 6.25% 6.51% 0.26% Q2 5.55% 6.23% 0.68% 6.01% 6.30% 0.29% Q3 5.50% 6.27% 0.77% 6.08% 6.73% 0.65% Q4 5.70% 6.22% 0.52% 6.12% 6.26% 0.14% st. dev. 5.60% 6.41% 6.12% 6.50% 1841 Q1 5.87% 6.89% 1.02% 6.30% 6.45% 0.15% Q2 6.47% 7.46% 0.99% 6.92% 7.01% 0.09% Q3 6.53% 7.16% 0.62% 7.01% 6.20% -0.82% Q4 6.75% 7.99% 1.24% 6.88% 7.31% 0.43% st. dev. 0.34% 0.90% 0.36% 1.03% 1842 Q1 6.62% 9.89% 3.27% 8.86% 12.78% 3.93% Q2 7.70% 8.53% 0.83% 11.98% 11.70% -0.28% Q3 7.37% 7.45% 0.09% 10.71% 10.94% 0.23% Q4 7.32% 7.72% 0.40% 9.78% 11.86% 2.08% st. dev. 0.50% 1.67% 1.51% 2.53% 1843 Q1 6.64% 6.53% -0.11% 10.43% 12.08% 1.66% Q2 5.85% 5.80% -0.05% 11.56% 9.62% -1.93% Q3 5.51% 4.92% -0.59% 8.77% 7.23% -1.54% Q4 4.86% 4.74% -0.12% 6.61% 5.76% -0.85% st. dev. 0.74% 0.81% 2.17% 2.71% Notes to Table 6, 7, and 8: All bond yields are taken from Sylla, Wilson, and Wright. Each weekly observation is converted to yield to maturity. Differences between US and London yields for a year or quarter are the simple differences in the annual or quarterly average in the table. Pennsylvania prices in the United States are those quoted in Philadelphia. New York and Ohio in the United States are those quoted in New York.

69

Figure 6
0.7 0.6 0.5 Yield 0.4 0.3 0.2 0.1 0 Jan-31

NY, OH, and PA Bond Yields (Scaled)
In London and New York

Pennsylvania

New York

Jan-33

Jan-35

Jan-37 Jan-39 Date

Jan-41

Jan-43

Jan-45

London

US

Table 6 gives average bond yields to maturity for New York and Ohio bonds in both London and the United States by year from 1831 to 1836 and by quarter from 1837 to 1843; the standard deviation of the yields in each year; and the average difference in yields in the two markets. 59 Table 7 presents the same information for Pennsylvania bonds, as well as the difference between the yield of Pennsylvania bonds in Philadelphia and the average yield of New York and Ohio bonds in New York.60 Table 8 presents the infrequently reported yields we have for Illinois, Massachusetts, and Indiana. These three tables provide the detailed statistics underlying the summary findings in Table 1.61 Figure 6 graphs weekly bond yields for New York, Ohio, and Pennsylvania bonds in both London and in the US.

59 60

The difference is the arithmetic difference in the average prices for the year or quarter. Sylla, Wilson, and Wright, Price Quotations. 61 Sylla, Wilson, and Wright, Price Quotations.

70

Table 7 Bond Yields in the United States and London Pennsylvania Bonds, and New York/Ohio average yield
PA London 1831 1832 1833 1834 1835 1836 1837 mean st. dev. mean st. dev. mean st. dev. mean st. dev. mean st. dev. mean st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. 3.92% 0.11% 3.59% 0.16% 3.61% 0.10% 3.89% 0.08% 3.97% 0.06% 4.12% 0.17% 4.56% 4.94% 5.06% 4.78% 0.29% 4.87% 4.86% 4.83% 4.72% 0.07% 4.96% 5.02% 5.31% 6.91% 0.62% 6.73% 6.99% 6.30% 6.03% 6.45% 5.78% ------0.15% 14.87% 15.09% 22.72% 13.98% 3.51% 15.89% 19.26% 16.21% 10.91% 3.67% US 3.95% 0.03% 3.74% 0.08% 3.96% 0.20% 4.39% 0.27% 4.14% 0.18% 4.55% 0.26% 5.02% 5.06% 4.40% 4.42% 0.35% 4.56% 4.55% 4.56% 4.55% 0.02% 4.55% 4.87% 5.38% 6.14% 0.58% 5.87% 5.96% 5.42% 5.80% 5.76% 8.05% 8.34% 8.43% 11.42% 1.90% 17.99% 21.86% 23.55% 18.85% 2.99% 21.02% 19.20% 15.68% 12.53% 3.51% US-London Diff 0.03% 0.16% 0.34% 0.50% 0.16% 0.43% 0.47% 0.13% -0.66% -0.37% -0.31% -0.31% -0.27% -0.17% -0.41% -0.14% 0.07% -0.77% -0.86% -1.04% -0.87% -0.23% 2.27% 8.34% 8.43% 11.42% 3.12% 6.77% 0.83% 4.88% 5.13% -0.06% -0.53% 1.62% NY & Ohio US Avg 3.62% 3.19% 3.26% 3.73% 3.28% 4.59% 4.75% 4.84% 3.74% 4.19% 4.34% 4.80% 4.29% --4.75% 4.63% 5.01% 7.61% 6.67% 6.27% 6.50% 6.24% 6.67% 7.23% 6.68% 7.65% 11.34% 10.12% 9.20% 9.79% 9.31% 7.71% 6.08% 5.25% PA - NY & Ohio Diff 0.33% 0.55% 0.69% 0.66% 0.85% -0.04% 0.28% 0.22% 0.66% 0.22% 0.22% -0.25% 0.27% ---0.20% 0.24% 0.37% -1.47% -0.79% -0.31% -1.07% -0.44% 1.38% 1.11% 1.75% 3.77% 6.65% 11.75% 14.36% 9.07% 11.71% 11.49% 9.61% 7.28%

1838

1839

1840

1841

1842

1843

See notes in Table 6.

71

Table 8 Bond Yields in the United States and London Illinois, Massachusetts, and Indiana Bonds
1831 1832 1833 1834 1835 1836 1837 mean st. dev. mean st. dev. mean st. dev. mean st. dev. mean st. dev. mean st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. Q1 Q2 Q3 Q4 st. dev. Illinois London --------4.35% 0.00% ----5.28% 0.00% --------------5.65% 5.64% --0.02% 5.83% 5.89% ----0.06% 7.10% 7.32% 7.67% 8.05% 7.36% 8.06% ------0.00% 26.78% 29.81% ----4.89% --26.98% 26.99% --0.01% Illinois US ----------------------------------4.89% ----0.03% 5.69% --5.92% 10.81% 2.69% 10.88% 8.89% 8.80% 7.97% 8.76% 12.65% 13.60% 12.68% 23.53% 6.37% 43.74% 53.83% 46.85% 47.91% 5.21% 43.37% 31.52% 23.24% 20.14% 9.82% MA London --------------------------------4.05% --4.39% 4.41% 0.07% 4.42% 4.68% 4.06% 5.95% 0.66% ----4.84% 4.91% 4.88% 4.94% ------0.00% ------5.03% 0.00% 5.11% 4.93% ----0.09% Indiana London ------------4.27% 0.01% 5.11% 0.45% 5.97% 0.18% 6.19% --7.07% 6.93% 0.38% 6.37% 6.45% 6.79% 5.71% 0.48% 5.92% 5.92% --8.76% 0.96% --8.41% 8.10% 8.22% 8.16% 8.36% ------0.09% 26.00% 35.68% 39.02% 37.95% 4.28% 35.41% --26.75% 24.80% 3.03%

1838

1839

1840

1841

1842

1843

See notes in Table 6.

72

The individual state series show the same pattern as the aggregate series: bond yields rise gradually in 1837, rise sharply in the fall of 1839, and rise and fluctuate wildy in the winter of 1842. The spread in bond yields between the U.S. and London, however, moved differently in each crisis. The bond yield spreads reflect how expectations and information differed in the United States and London. To exploit the yield spreads, however, we first need to appreciate the situation in Pennsylvania.

Pennsylvania:
New York began the Erie Canal in 1817 and completed it in 1825; Ohio began construction on two canals in 1825 and completed them in 1832; and Pennsylvania began work on its canal system in 1826 and completed the Main Line in 1835. By 1836, the New York and Ohio canals were returning revenues to the state Treasury in excess of operating costs and interest payments, while the Pennsylvania canals were a financial disaster. Financial markets priced the bonds of the three states accordingly. In the early 1830s, yields on Pennsylvania bonds were consistently higher than the yields on New York and Ohio bonds, usually 0.5 percent or more (Table 7, column 5).62 Pennsylvania's situation changed in 1836. When Nicholas Biddle lost the Bank War to Andrew Jackson, the Bank of the United States sought a charter from the state of Pennsylvania. In 1836, the BUS was rechartered as the Bank of the United States of Pennsylvania. The charter was very generous to the state, including a promise by the BUSP to underwrite $6 to $8 million in state bond issues:

62

The near equivalence of New York, Ohio, and Pennsylvania yields in 1836 is misleading, since there were only 4 observations on New York bond yields in New York that year. Most of the New York bonds were paid off in 1836.

73

The Bank committed to pay an additional twenty installments of $100,000 each, beginning June 1, 1836 and continuing for the next nineteen years, to pay $500,000 on March 3, 1837, to subscribe for various specifically designated public improvement stocks amounting to $675,000, to make long term loans to the state up to $6,000,000 for which the state agreed to turn over to the Bank bonds redeemable in 1868 (at par if they were 4 per cent bonds at one hundred and ten if they were 5's) and to make temporary loans up to a maximum of $1,000,000 in any one year at 4 per cent interest. (Smith, Economic Aspects, p. 179)

In 1837, the yields on Pennsylvania bonds suddenly became fixed within narrow limits. Between November 1837 and April 1839 the maximum yield on Pennsylvania bonds in Philadelphia was 4.56 percent, the minimum yield was 4.42 percent (Table 8). The standard deviation on the Pennsylvania yield in 1838 was .02 percent, the lowest standard deviation for any state's bonds in any year in Tables 4, 5, and 6. Deliberately or not, the BUSP pegged the price of Pennsylvania bonds as a result of its obligations to purchase state bonds over this 18 month period. Other lenders were not so optimistic about Pennsylvania, however. From 1837 to 1840 yields on Pennsylvania bonds in London remained considerably higher than yields in Philadelphia.63 The BUSP's condition worsened in 1839, when its extensive operations in the state bond market, the cotton market, and the market for international and domestic exchange went sour.64 In October of 1839, the BUSP was forced to suspend convertibility of its demand liabilities into specie because of a run by New York and Boston Banks. The suspension of payments precipitated a banking crisis in the United States; with banks in the south and west suspended until 1842. But the suspension did not release the BUSP from its obligations to the state of Pennsylvania. Until early 1841, although the BUSP no longer pegged the yield, Pennsylvania bonds continued to enjoy lower yields in Philadelphia, despite steady borrowing by the state, than
63

The BUSP's willingness to purchase state bonds kept the yields on Pennsylvania bonds in Philadelphia lower (prices higher) than either New York or London after late 1837. The use of the Pennsylvania bond yields for cross country comparisons is, as a result, problematic. 64 Details of the bank's demise can be found in Smith, Economic Aspects, Hammond, Banks and Politics in America, and Govan, Nicholas Biddle.

74

did the bonds of Ohio and New York in New York or London. In February 1841, the state attempted to force the BUSP to resume specie payments, whereupon the bank closed its doors and went out of business. With BUSP out of the market, yields on Pennsylvania bonds in Philadelphia jumped immediately, from 6.01 percent on January 2 to 9.5 percent on March 7. For the remainder of 1841, Pennsylvania bond yields were above 8 percent in Philadelphia, and prices on Pennsylvania bonds were no longer quoted in London. Yields on Pennsylvania bonds were now 2 percentage points higher than yields on New York and Ohio bonds. The BUSP's artificial support of state credit between 1837 to 1840 makes problematic the use of Pennsylvania bond prices as a indicator of market conditions in those years. Pennsylvania was in deep financial trouble in 1841. The state's credit returned to a level consistent with its financial situation when the failure of the BUSP forced the state back into regular credit markets. In late 1841, yields on Pennsylvania bonds in Philadelphia began rising rapidly. Pennsylvania defaulted on its bond obligations in 1842, with devastating consequences for the state bond market on both sides of the Atlantic.

The Crises of 1837 and 1839:
The Panic of 1837 occurred in a window of time where bond price data are hard to come by. First, New York completed work on its canals in the 1820s and Ohio in the early 1830s. New York paid off most of its debt by 1835. Although both New York and Ohio started new projects in 1836 (New York authorized a $2,000,000 bond issue in 1836), neither state borrowed heavily until later in 1837. As a result, there are gaps in the quotation series for both states in 1837, reflecting the absence of marketable bonds in both New York and London. Second, western states had just begun issuing bonds when the Panic hit, and we do not have usable price series for

75

Indiana or Illinois in 1837. With the exception of Pennsylvania, not many state bonds traded in the spring of 1837. Between 1831 and 1836, the yield differential between the United States and London was small on average: .0013, only13 basis points. Yields were slightly lower in London than New York, consistent with the general idea that credit markets were deeper and interest rates lower in London, as well as with higher transaction costs of marketing American bonds in Britain.65 The average difference for Pennsylvania bonds was .003, for New York bonds .001, and for Ohio bonds -.0008 (on average Ohio bonds had slightly higher yields in London than in New York). As we saw earlier, these markets were well integrated. Bond yields began rising in 1836, a full percentage point in the U.S. and almost 3/4 of a percentage point in London. Credit markets tightened everywhere in 1836 (Figure 5). Yields continued to rise through early 1837 in both London and the U.S., but more quickly in the U.S.. When the Panic broke out in May, however, yields moved in opposite directions in the U.S. and in Britain. In the third quarter of 1837, yields on New York bonds in London rose to 5.46 percent, while in the U.S. they fell to 3.94 percent (Table 6, columns 1 and 2). For the remainder of 1837, all of 1838, and the first three quarters of 1839, it was more expensive for state governments to borrow in London than in New York (Table 1). Both 1838 and 1839 were years of heavy new state borrowing and there were frequent quotations in every market. In the aftermath of the Panic of 1837, credit markets for American state bonds were significantly tighter in London than in the United States. The summer of 1839 was a turning point for the transportation boom in the northwest. The Morris Bank defaulted on its July installment to Indiana. As the year progressed it became

65

The idea that British credit markets were deeper than American goes back at least to Callender, 'Early Transportation and Banking Enterprises,' whose essay lays out the importance of British capital for American development and the role of American states in tapping foreign and domestic credit markets to support internal improvements.

76

clear that Indiana, Michigan, and Illinois were in serious trouble, a concern immediately reflected in yields on their bonds. Indiana bond yields in London rose from 5.92 percent in June, to 8.76 percent in November (Table 8, there are no quotes in between those dates in London, and there are no quotes for Indiana bonds in New York before 1840).66 Illinois bonds in New York went from a yield of 4.9 percent in July to 11.1 percent in November and 13 percent in December, while in London Illinois bonds went from a yield of 5.98 percent in July to 7.32 percent in January. Financial markets acknowledged that it was primarily the western states whose credit was threatened. Yields on eastern state bonds rose in 1839, but not nearly to the extent of yields on western bonds. The BUSP once again suspended specie payments in October, 1839. This crisis hit U.S. markets for state bonds much harder than it hit the London market. In the third quarter of 1839, the average yield on New York bonds was 5.19 percent in the U.S. and 5.04 percent in London. In the fourth quarter of 1839, the average yield on New York bonds was 7.76 percent in the U.S. and 6.19 in London. For the remainder of 1840 and 1841, average yields stayed higher in the U.S. than in London. The New York-London differential on New York bonds was over 0.5 percent throughout both years (Figure 4); for Ohio bonds, between 0.1 and 0.5 percent (Figure 3); and for Illinois bonds over a full percentage point or more. Unlike the aftermath of the Panic of 1837, when markets for state bonds were tighter in London than they were in the United States, after the Crisis of 1839 yields on state bonds were higher in the United States than in London.

We are now in a position to examine the origin of the shock to bond markets. Figures 2

66

This may be because the Indiana bonds sold on credit to the Morris Bank were sent to London, and from there to Amsterdam. The Morris Canal and Banking Company took the Indiana bonds it purchased on credit and used them to pay off the mortgage held on the canal by Dutch creditors. By the summer of 1839, the Morris Bank did not hold any Indiana bonds, it had already sold or hypothecated all of them. The story is not told anywhere, but can be tracked through the Company minutes at the New Jersey State Archives.

77

and 3, the difference in Ohio and New York bond yields in the United States and London, show distinct spikes in the U.S.-London yield differentials at the end of 1839.67 This is the first spike in the bond differentials, small compared to what was to come in 1842, but telling nonetheless. Up to 1839, markets in New York, Philadelphia, and London shared the same information. In the fall of 1839, news hit the markets in America first. Bond prices dropped and yields rose in New York about two months before yields rose in London. Unlike 1837, when credit conditions tightened on both sides of the Atlantic and the news about the Panic of 1837 did not originate in either country, in 1839 the event that shocked bond markets clearly originated in the United States.68 Temin suggested that American states were forced to abandon their internal improvement projects after the Crisis of 1839 because British capital dried up. His conjecture finds no support in the financial market data. After the Panic of 1837, it was consistently more expensive for states to borrow in London than in New York and Philadelphia. After the Crisis of 1839, it was consistently cheaper for states to borrow in London than in New York and Philadelphia, and this was true for all states. States, of course, found it harder to borrow everywhere in 1840 and 1841, when yields on New York and Ohio bonds reached 7 percent, and yields on Illinois and Indiana bonds went to 8 percent and higher. But it was not relatively harder to borrow in London than it was in America. The idea that the depression that developed in the United States after October 1839 was due to the tightening of British capital markets is not supported by the bond yields. Although yields were more favorable to borrowers in London than in the U.S., states found it difficult to borrow in both the U.S. and London in 1840. States with par restrictions on their bonds could not market any bonds at prevailing rates. But they could sell bonds if they were

67

This is in the time interval when Pennsylvania bond prices are supported by the BUSP, so the yield differential between Philadelphia and London becomes more negative. This, however, is a function of the BUSP support. 68 There is no possibility that a positive shock hit the London markets in October of 1839. Equity prices in London were falling, not rising in the late 1839, see Gayer, Rostow, and Schwartz, Growth and Fluctuation.

78

willing to pay market rates. For example, Illinois had issued bonds to state contractors in lieu of cash payments, bonds the contractors had accepted at par. When state agents went to London in July of 1840, they took both new state bonds with par restrictions and contractor bonds. The contractor bonds were identical to the new bonds in every respect except the par restriction. 'None of the state bonds were sold, but an agreement was made to sell $1,000,000 of the contractors' bonds to Magniac, Smith and Company of London at a rate of eighty-three.'69 Ohio continued to borrow through 1843, authorizing new bond issues at less than par. The state was able to sell $400,000 in bonds in July of 1840 to Barings at a price of 95 and an additional $400,000 in bonds in May of 1842 at 'the distressingly low price of 60.'70 States could borrow, but not if they insisted on selling 5 or 6 percent bonds at par.

The Collapse of 1842:
Financial historians have paid little attention to the Collapse of 1842, but big things were happening that year in the market for state bonds. The collapse in state debt markets is traditionally attributed to state defaults (Table 3). The timing of defaults and bond yield movements shows that the onset of the default crisis cannot account entirely for the collapse of state debt markets. In 1841, Indiana and Florida defaulted in January, Mississippi in March, and Michigan and Arkansas in July. Yet yields on New York and Ohio bonds were not noticeably higher in the first quarter of 1841 than they had been for most of 1840. Although yields rose in the second and third quarter of 1841, the increase was small compared to the jump that occurred
69 70

Krenkel, Illinois Internal Improvement, p. 122. Scheiber, Ohio Canals,pp. 140-158, quote from p. 152. At a price of 60, the yield on a 6 per cent bond was roughly 10 percent. Ohio did not include a par restriction in the legislation authorizing bond issues in 1836, so most Ohio bonds could be sold at any price. Ohio did have problems with price restrictions, however. One issue of bonds had been sold for less than the legislated minimum and at one point markets in New York believed, erroneously, that the state was about to default on bonds that had been sold in violation the legislated price.

79

Figure 7
0.1 0.08 0.06 0.04 difference 0.02 0

Difference in Ohio Bond Yields
New York minus London

-0.02 -0.04 -0.06 Jan-41

Jul-41

Jan-42

Jul-42

Jan-43

Jul-43

US - London Difference

in the fourth quarter of 1841 and the first quarter of 1842. Something happened in the winter of 1842 that shook American credit markets. And it wasn't just the defaulting states that experienced a crisis in the winter of 1842: yields for issues of Ohio and New York bonds, states that avoided default, spiked in the U.S. market as well. The crisis in the winter of 1842 originated in the United States. The news hit American markets first, American markets quickly increased the risk premium placed on American state bonds, and London did not digest the news from American markets for several months. In the first quarter of 1842, the yield on New York bonds was 9.89 percent in New York and 6.62 percent in London; on Ohio bonds, 12.78 percent in New York and 8.86 percent in London; on Illinois bonds, 43.74 percent in New York and 26.78 percent in London. Figure 7 focuses on bond yield differences for Ohio between January 1841 and December 1843 (this figure expands the time scale of Figure 3; a graph for New York is similar). As the default crisis unfolded in 1841,

80

yields in Ohio stayed close to yields in London. But in December of 1841 and January of 1842, yields moved sharply higher in the United States, peaking in late March.71 Yields on New York and Ohio bonds were at least 8 percentage points higher in New York than the contemporaneous prices in London. The shock was not transitory. Bond yields remained higher in both markets through 1842. But the disjunction between bond yields in the United States and in London was transitory. By April 17, 1842, New York and Ohio bonds were again trading for the same prices in London, New York, and Philadelphia.72 For the second quarter as a whole, yields were only 0.0028 (28 basis points) higher in New York than in London. Markets were well integrated, but the shock hit America first and, given the time it took information to propagate to Britain, London did not react for two months. What happened? Pennsylvania was the locus of the crisis. As early as 1839, Pennsylvania was in deep financial trouble, but BUSP loans masked the state's weakness until the state was forced back into regular credit markets. After the BUSP failed in February 1841, the yields on Pennsylvania bonds in Philadelphia began rising: from 5.76 percent in the last quarter of 1840, to 8.05 percent in the first quarter of 1841, 8.43 percent in the third quarter, 11.42 percent in the fourth quarter, and 17.99 percent in the first quarter of 1842. State chartered banks were not at liberty to refuse loans to the state government that chartered them. Throughout 1841, Pennsylvania leaned on its banks. In November of 1841, Pennsylvania announced that it would require a loan from all banks in the state equal to 5 percent of their capital. The news that hit American markets in December 1841 and January 1842, as the
71

Yields on Ohio and New York bonds began moving higher in December, but Figure 7 shows a widening gap in January. We do not have London prices for Ohio bond between December 5, 1841 and January 9, 1842. The difference in yields began widening in early December. 72 Bond prices moved sharply upward in London that week, New York bonds went from yields of 6.38 percent to 8.05 percent and Ohio bonds from yields of 9.34 percent to 12.5 percent. At the same time, yields moved down in New York, bring yields in the two markets back into parity. By this time, Pennsylvania bonds were no longer trading in London.

81

state began gathering loans from its banks, was that Pennsylvania was actually carrying out its threat to make the banks sustain the state credit through forced loans. In February of 1842, the state precipitated a banking panic in Philadelphia, when it attempted to withdraw funds from the Bank of Philadelphia necessary to cover the interest payments due that month.73 At that point, the state had not yet decided whether it would rescue the state credit by extorting more money from state chartered banks. When Pennsylvania made it clear that it would not force more loans from state banks in April of 1842, the crisis was over. As a result of the state's decision not to press its banks it became inevitable that Pennsylvania would default on its August 1842 interest payment, and Pennsylvania bond yields accordingly rose steadily until the third quarter. Yields on Pennsylvania bonds would not fall back below 10 percent until April of 1844, and the state resumed interest payments in February 1845. Conditions were similar in New York, where the state pressed state chartered banks to purchase state bonds. New York bank holdings of state 'stock' rose from nothing in 1839 to almost $7 million in 1842. The New York state legislature met in emergency session in March to consider how to deal with the impending state default. It responded with the 'Stop and Tax' law of 1842, stopping borrowing, stopping construction on canals, and re-instituting the state property tax. These measures ended the crisis in New York bonds. Ohio relied heavily on its banks for funds. Ohio raised $900,000 in 1842, $500,000 from state chartered banks and the $400,000 borrowed through Barings in London.74 As long as Ohio

73

The funds Pennsylvania withdrew were the funds they had borrowed from other banks in the state. Pennsylvania did not default on its bonds until 1842, but they were several days late on an interest payment in February because of the banking crisis. The crisis in Pennsylvania and the state's interaction with its banks is described in Kettell 'Debts and Finances' and the Pennsylvania Report in House Document 226, 29th Congress, First Session. 74 Scheiber, Ohio Canals, pp. 140-158. 'During the remainder of 1842, the fund board sustained installment payments on the three-year loans by issuing bonds to Ohio banks at prices of 70 to 75 [yields of roughly 8 percent]. In this manner, nearly $700,000 of bonds were sold for cash payments of only $500,000.' p. 152. As noted earlier, Ohio borrowed sold its bonds to Barings for a price of 60.

82

continued its internal improvement projects, financial markets continued to purchase Ohio bonds at prices significantly below par. Yields on Ohio bonds in New York and London remained over 10 percent until the second quarter of 1843. Because Ohio could pressure its banks in a way that it could not pressure financial markets, as interest rates rose the state issued more bonds to its banks and fewer directly to financial markets. When it did place new bonds, it placed them in London, not in New York. Rising yields on Ohio and New York bonds in late 1841 were not a response to the default crisis in Mississippi, Florida, Arkansas, Indiana, and Michigan. The Governor of Mississippi announced that he supported repudiation in early 1841. When Mississippi and Florida repudiated their bonds by legislative act in February of 1842 this was old news. The news in the winter of 1842 was the threat that New York, Ohio, and Pennsylvania would cannibalize their banks to keep state finances afloat. Pennsylvania's announcement of the forced loan program in November 1841 gave concrete expression to the threat. Throughout the winter of 1842 it was not clear what additional steps states would take to deal with the crisis. Fundamental uncertainty drove bond prices down, bond yields up, and brought an end to any hopes that states would be able to raise large amounts of capital at reasonable rates to continue their internal improvement projects. News of the threat and an appreciation of its magnitude took several months to reach Britain. The Collapse of 1842 was not brought on by tight credit markets in Britain after the Crisis of 1839, but by a political crisis in the United States in the winter of 1842. Even so, financial markets sorted themselves out quickly. Interest rates on all state debts were higher in April of 1842 than in October of 1841, but markets in London and New York paid the sample yields on Ohio and New York bonds.

83

Section 5. Conclusion
In 1841, Nicholas Biddle argued that European conditions played an important role in the economic crisis in 1839, and one need go no further than Leland Jenks's Migration of British Capital or Bray Hammond's Banks and Politics in America to see how much economic historians have laid the blame for the depression of 1839 to 1842 at the feet of international conditions. Peter Temin made no bones about the centrality of British credit in bringing on the Crisis of 1839 and the collapse of state internal improvement projects: 'The state projects initiated in the late 1830s had been started in the expectation of external [British] financing.... Unfortunately, the new inflow of foreign capital did not continue [in 1839]... and the manifold projects of the states were abandoned.' Three clear findings of this paper challenge this traditional interpretation. First, the conditions that brought on the Panic of 1837 could not have anything to do with the crisis in American state debts after 1839. The majority of state debt outstanding in 1841 was incurred after Panic, not before. The vast majority of debt in New York, Ohio, Massachusetts, Indiana, Illinois, Michigan, Arkansas, Maryland, and Mississippi involved in the default crisis was authorized in 1837 or later and issued long after the Panic of 1837 was over. New York, Ohio, and Pennsylvania continued to issue and market bonds in 1840 and 1841. Second, before 1837 state bonds had marketed for slightly higher prices (lower yields) in London than New York and Philadelphia. After the Panic of 1837, London markets for American state bonds became noticeably tighter than American markets, during the three years of the heaviest borrowing: 1837, 1838, and 1839. Yet, when the BUSP suspended payments in October 1839, the economic crisis set in, and bond yields rose sharply in both countries, yields in London became significantly lower than yields in the U.S.. States such as Ohio and Illinois could borrow at lower cost in London than in New York, and so they borrowed in London. There is no evidence that British credit markets dried up relative to American markets after 1839. States had more

84

trouble borrowing in both markets, of course, and states were forced out of the market entirely if they insisted on borrowing at par. The collapse of state transportation projects in Indiana, Illinois, and Michigan in 1839 had nothing to do with credit markets in London, and everything to do with the defalcation of American banks such as the Morris Canal and Banking Company, which failed to pay states for bonds they had already accepted and on which the states were liable to pay interest immediately. Third, the movement of bond yields during the Crisis of 1839 and the Collapse of 1842 indisputably show that the shocks to financial markets originated in the United States and spread to London, not the other way around. As the economic crisis deepened in 1840 and 1841, New York, Ohio, and Pennsylvania put greater pressure on their own state chartered banks to buy state bonds. This was not because the states could not sell bonds in London, but because the yields on those bonds in London, New York, and Philadelphia were justifiably rising. Pennsylvania's forced loan policy, beginning in November of 1841, tipped American markets into crisis. The Panic of 1837 has received the lion's share of attention from economic historians, but, as the macroeconomic situation deteriorated, the North Atlantic economy was hit by two more crises in 1839 and 1842. It seems clear that the impetus for these crises came from the United States and was intimately tied to the market for American state government bonds and the failing efforts of American states, particularly in the west, to finance public investments in finance and transportation. The Crisis of 1839 and the Collapse of 1842 were not caused by the same forces as the Panic of 1837. The collapse of American state finances in the 1840s was predicated on events that occurred after 1837.

85

Chapter 5: Conclusion
One of the achievements of my thesis is to discover the trend and cyclicality of diversification in the entire U.S. manufacturing sector in the last 30 years. Findings are summarized as follows: (1) Aggregate diversification declined both at the establishment and firm level since the early 1980s. The downward trend is common in many industries. The declining diversification is quite contrary to the conjecture that the diversification has been increasing in the last three decades. (2) Whether the diversification is pro-cyclical or counter-cyclical is not clear at the aggregate or industry level. The conjecture that the diversification is pro-cyclical cannot be confirmed by the data. (3) A large fraction of firms change the number of products and plants annually. The declining diversification measure suggest that firms becomes more specialyzed, but it is certain that the number of product is not fixed for firms even in the short run. It is shown that product diversification is a decision variable for firms, which is contrary to assumptions of fixed diversification in many theoretical models in literature. I show that firms actively change their product diversification at a short-term frequency. More Firms specialize in one product and the number of products and plants behaves like an adjustment margin. Trend of volatility is verified by the micro level data and new empirical relationship between diversification and volatility is found. Using the firm level profit rates, I find: (1) the aggregate volatility declined. (2) The volatility decreased since the 1980s for most industries. (3) The mean of firm level idiosyncratic volatilities decreased in late 1980s and the standard deviation doesn't change much. The left-censored Tobit regression shows that the firm level diversification is positively affected by the aggregate, industrial and idiosyncratic profit volatility. Therefore, the decrease of volatility, in other words, the reduced risks in US manufacturing sector contributes to the decrease of diversification.

86

In summary, firms specialize more in the past 30 years in U.S. manufacturing sector. It is because the profit volatility decreased at the aggregate, industrial and firm level. Therefore, firms have less incentive to diversify over different products to insure themselves against profit shocks. However, a large fraction of firms adopt flexible production lines which allow them to adjust the number of products at the short term frequency responding to the economic fluctuation. A lot of questions about diversification have been raised and partially answered. But it was not easy to see the whole picture of evolution of diversification because there hadn't been enough data. With rich description and analysis in my thesis, we now better understand diversification of firms and the role of volatility on diversification. It is now possible to move on to next questions on diversification and specialization: whether specialization enhanced productivity, whether diversification increased the profits by reducing idiosyncratic risks, whether the high volatility played a role in high diversification in developing countries. Historically, the volatility change gravely affects the economy in many ways in U.S. The event of 1840s shows an example of volatility change caused by the investment of U.S. government. The bond markets in U.K and U.S were integrated across Atlantic Ocean in early 19th century. The movement of bond yields during the Crisis of 1839 and the Collapse of 1842 indisputably show that the shocks to financial markets originated in the United States and spread to London, not the other way around. The Panic of 1837 has received the lion's share of attention from economic historians, but, as the macroeconomic situation deteriorated, the North Atlantic economy was hit by two more crises in 1839 and 1842. It seems clear that the impetus for these crises came from the United States and was intimately tied to the market for American state government bonds and the failing efforts of American states, particularly in the west, to finance public investments in finance and transportation. The Crisis of 1839 and the Collapse of 1842 were not caused by the same forces as the Panic of 1837. The collapse of American state finances in the 1840s was

87

predicated on events that occurred after 1837.

88

Appendix A: Data in Chapter 2 and Chapter 3
Primary Data source: LRD and LBD LRD: LRD provides a company-level database containing detailed statistics on research and development activities; and supports research on the issues of productivity, profitability, and the use of research and development. The database contains detailed company-level research and development information compiled from the annual Industrial Research and Development survey for survey years 1972 through 2001. Over the 30 year period, the total sample for the survey size has varied considerably. Since 1992, the total sample size has been fairly stable at approximately 25,000 companies. The sample design strategy has evolved over the years. The company has been defined as both the sample unit and the data collection unit since inception. Prior to 1992, a given sample would be used for a number of years before being replaced. The probability of selection was a direct function of total company employment; companies with more than 500 employees were included with certainty. LBD: LBD is a research dataset constructed at the Census Bureau's Center for Economic Studies. LBD is an establishment based file created by linking the annual snapshot files from Census Bureau's Business Register over time. It contains high quality longitudinal establishment linkages. Firm level linkages are currently under development at CES. Currently, LBD contains the universe of all U.S. business establishments with paid employees from 1976 to present. LBD covers almost 24 million unique establishments from 1975 to present.

Supplementary Data source: NBER R&D and Productivity file from NBER, and statistics from ASM: Annual Survey of Manufactures published by Census Bureau.

89

Diversification Index: I measure 5-digit product diversification using LRD as described in the text. 5-digit product shares are calculated by TVPS/TVS where TVS (Total Value of shipments) is the sum of TVPS (Total Value of Product Shipment) at the establishment level. For a firm level index, the product shares are calculated by FTVPS/FTVS, where FTVPS is the sum of TVPS of a product produced in every plant of the firm and FTVS is the sum of TVS across plants. Some product data are imputed and they are eliminated from the sample. ASM sample base is the establishment rather than the firm, some establishments of a multi-unit firm may not be selected in ASM sample. This can distort the firm-level diversification measure of multi-unit firms. In most cases, however, all the establishments of a multi-unit firm are included in ASM sample. All the plants of a company, so-called Certainty Companies, are included in ASM for certain, but many of the non-certainty multi-unit firms also have all of their plants in ASM. 75 Matching ASM and LBD enables us to find the establishments of a multi-unit firm which are not selected for the ASM sample. After shipment weighted, the share of those establishments is negligible. The aggregated diversification index is not sensitive to this sample problem. See LRD documentation for detail.

Industry Classification: LRD classifies establishments by industry using the Standard Industrial Classification System (SIC). The structure of SIC makes it possible to tabulate, analyze, and publish establishment data on a 2-digit, a 3-digit, or a 4-digit industry code basis, according the level of industrial detail considered most appropriate. In addition to industry, the Census Bureau also collects and publishes information on product classes and individual products produced by manufacturing establishments. Product classes (5-digit codes) and products (7-digit codes) of manufacturing industries are assigned codes based on the industry from which they originate.
75

Those companies are usually big firms with no less than 250 employees. The establishments that had been dropped out of sample were added with zero statistical weight and called 'McGuckin Adds'.

90

Beginning in 1997 the US, Canada, and Mexico began publishing and collecting statistics under the new North American Industrial Classification Systems (NAICS). NAICS is based on a consistent, economic concept: Establishments that use the same or similar processes to produce goods or services are grouped together. The SIC, developed in the 1930s and revised periodically over the past 50 years, was not based on a consistent economic concept. A major change in SIC occurred in 1987. Some industries are demand based while others are production based. From 1998 ASM, the product class is coded by NAICS.
Table AA-1 Description of SIC codes SIC Code Level 2-digit 28 Major industry group 4-digit 2834 Industry 5-digit 7-digit 28347 2834711 Product class Product

Description Chemicals and allied products Pharmaceutical Preparations Vitamin, nutrient, and hematinic preparations, for human use Multivitamins, plain and with minerals (except B complex vitamins and fish livers oils)

Establishment and Firm Identifier: Permanent Plant Number (PPN) assigned to each establishment by Census is used as the establishment identifier. For the single-unit firms/establishments, PPN begins with 0. For multi-units, the first six digits of the ten-digit PPN identify the firm.

Profit rate: Profit rate is measured by the nominal sales (TVS) minus the variable costs, divided by the capital stock. The variable costs are composed of total wage cost (SW) and the total material costs (CM). Profit is deflated by GDP deflator. Book value of capital stock(MA and BA) is collected in ASM and CM and it is deflated by the 2-digit industry level deflator. The Bureau of Economic Analysis publishes 2-digit industry capital stock both in nominal and real values. I use the ratio of the nominal capital stock to real capital stock as the 2-digit industry level capital deflator. The base year for the deflator is 1996. The growth rate of real shipment (RTVS) is the symmetric growth measure: Growth of RTVS at time t = (RTVSt-RTVSt-1)/ [(RTVSt+RTVSt-1)/2]

91

Appendix B: Multivariate ARCH with missing data in Chapter 4
The multivariate ARCH model for stock/bond market integration test in Chapter 4 has the following feature:

Yit = X it Bi + ? it

? t | ? t ?1 ~ N (0, ? ) E (? it | ? t ?1 ) = 0 E (? it ? jt | ? t ?1 ) = ? ij = aij + bij ? it ?1? jt ?1
We are maximizing the log likelihood function to estimate parameters, (a,b,B):

, i = 1,2,L , n

log L = ? ?

T 1 T 1 n log(2? ) ? ? log| ? t | ? ? ? t ? ? t?1? t t=2 2 t=2 2 t=2 2 T

where ? t = ? t (a , b, ? t ?1 ) and ? ijt = aij + bij ? it ?1? jt ?1

? t = Yt ? X t B
? log L = log L(a , b, B| data )
Since the bond prices are not observed sometimes, the YTM series that are necessary for bond market integration test have some missing data. To get a bench mark result, we linearly interpolated the surrounding observations for these missing observations and used this imputed series for the ARCH estimation. For example, the data are available in week 3 and week 5 in London, the price interpolated from these two prices is used for week 4. This imputation may cause a consistency problem in our estimates, since the imputation is not likely to be the true data generating process. We develop a new method to handle the missing observations directly. Suppose an observation is missing at time t-1. For example, the data are not available in London in week 4, while the data are available in week 3 and week 5 in both markets. We take

92

the covariance matrix of time t, conditional on time t-2 disturbance terms, instead of time t-1 disturbance terms. That is, the covariance of week 5 is conditional on week 3, not week 4:

E (? it ? jt | ? t ? 2 ) = aij + bij E (? it ?1? jt ?1 | ? t ? 2 ) E (? it ?1? jt ?1 | ? t ? 2 ) = aij + bij ? it ? 2 ? jt ? 2
2 ? E (? it ? jt | ? t ? 2 ) = aij (1 + bij ) + bij ? it ? 2 ? jt ? 2

In this fashion, we can represent the conditional covariance of time t with the time t-2 error terms. Generalizing this method to calculate recursively to the case where time (t-s) data are the latest observations before t, we get
* * E (? it ? jt | ? t ? s ) = aij + bij ? it ? s ? jt ? s

where
* s ?1 ? ? aij = aij (1 + bij + L + bij ) ? * s ? ? bij = bij

Therefore, (a*, b*) has one-to-one nonlinear relationship to (a, b). Using this relationship, we can adjust the likelihood function without changing the number of parameters we are estimating:

log L = ? ? where

1 1 n log(2? ) ? ? log| ? t | ? ? ? t ? ? t?1? t t ?A 2 t ?A 2 t ?A 2

? t = ? t (a , b, ? t ? s ) and ? ijt = aij (1 + bij + L + bijs?1 ) + bijs ? it ? s ? jt ? s A = time periods when the data are observed ? log L = log L(a , b, B| observed data )
The covariance matrix is conditional on the latest observations available. The likelihood function is still determined by a, b and B conditional on the actual data. There is a computational problem because the likelihood function is highly nonlinear on a and b, but the consistency of the estimate is preserved.

93

Appendix C: Additional Statistics of Annual Firm-level Product Diversification
Table AC-1 Average Firm level Diversification Index year 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 frequency 21018 20605 20913 24889 23973 25306 25024 23277 22770 20079 20055 18089 17914 19085 16835 21211 29814 29975 30728 30744 40298 38547 36961 42766 25359 Not weighted 0.17204 0.17327 0.16792 0.17545 0.17235 0.15597 0.15612 0.15839 0.19000 0.18967 0.16184 0.16478 0.16314 0.20029 0.20881 0.16033 0.10996 0.12077 0.16033 0.13381 0.09048 0.09355 0.09561 0.11777 0.13921 Shipment weighted 0.72984 0.72827 0.73121 0.72929 0.72479 0.71227 0.70771 0.71600 0.70695 0.71414 0.71090 0.70861 0.70209 0.69528 0.69410 0.68754 0.68208 0.67950 0.66403 0.66211 0.65296 0.64973 0.64735 0.62857 0.64273

94

Table AC-2 Average Firm level Diversification Index of Single Unit firms year frequency Not weighted 74 17495 0.10282 75 17228 0.10550 76 17655 0.10258 77 20925 0.11449 78 20061 0.11005 79 21503 0.10384 80 21465 0.10621 81 19897 0.10692 82 18864 0.13812 83 16364 0.13223 84 16533 0.10144 85 14717 0.10125 86 14760 0.10396 87 15141 0.13981 88 12818 0.14349 89 17434 0.10485 90 26218 0.06756 91 26592 0.08218 92 26532 0.12504 93 26704 0.09106 94 36439 0.05728 95 34846 0.06069 96 33426 0.06310 97 38635 0.08861 98 22226 0.09980

Shipment weighted 0.17855 0.18238 0.16741 0.18504 0.17599 0.15623 0.16029 0.16702 0.18113 0.17185 0.14605 0.14244 0.15206 0.18516 0.18218 0.15249 0.13617 0.13718 0.17138 0.16005 0.13421 0.13091 0.13068 0.14643 0.14477

95

Table AC-3 Average Firm level Diversification Index of Multi Unit firms year 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 frequency 3523 3377 3258 3964 3912 3803 3559 3380 3906 3715 3522 3372 3154 3944 4017 3777 3596 3383 4196 4040 3859 3701 3535 4131 3133 Not weighted 0.51578 0.51901 0.52204 0.49729 0.49183 0.45077 0.45714 0.46143 0.44055 0.44266 0.44536 0.44204 0.44010 0.43247 0.41725 0.41640 0.41907 0.42409 0.42149 0.41641 0.40397 0.40291 0.40299 0.39044 0.41877 Shipment weighted 0.77045 0.76725 0.76902 0.76322 0.75931 0.75342 0.74905 0.75328 0.74102 0.74617 0.75148 0.74545 0.74465 0.72989 0.7284 0.7313 0.72715 0.72277 0.70572 0.70243 0.70013 0.69309 0.68796 0.6645 0.67831

96

Table AC-4 Share of Diversified Production (rpd) year 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98

Share of Diversified Production (rpd)
0.003744 0.003589 0.003430 0.004498 0.006080 0.006118 0.005320 0.005597 0.006016 0.006107 0.007289 0.007863 0.009350 0.007971 0.009129 0.009075 0.008278 0.008905 0.010446 0.011226 0.013564 0.016755 0.020226 0.018314 0.010821

Note:

d = 1? (

?S
i?A

2 i

+

Diversified Production

1 2 3

?S
i?B

2 i

) = (rpd + rps )d

Specialized Production

1 2 3

where,

rpd = ? S i2 / ? S i2 , rps = 1 ? rpd
i?A i

i ? A product i produced in multiple plants i ? B product i produced only in one plant

97

Table AC-5 Share of Within-plant Factor in Firm Level Diversification(rwp) year Share of Within-plant factor(rwp) 74 0.422879 75 0.421189 76 0.416774 77 0.402075 78 0.414824 79 0.436601 80 0.451316 81 0.448953 82 0.429708 83 0.430263 84 0.408616 85 0.402632 86 0.384211 87 0.394595 88 0.393574 89 0.387701 90 0.39031 91 0.37973 92 0.385892 93 0.378976 94 0.375346 95 0.366295 96 0.367318 97 0.375907 98 0.353448

Note:
f est f d f = ? a j d est j + ( d ? ? a j d j ) = ( rwp + rap ) d j j 1 4 24 3 144 2 44 3 Within -plant Among-plant

where, a j = shipment share of the jth plant d f = firm level diversification, rwp = ? a j d est d f , rap = 1 ? rwp j
j

d est = plant level diversification

98

Table AC-6 Annual Diversification Index Change Decomposed by POSC, NEGC, POSB and NEGD year Net D change POSC NEGC POSB NEGD 75 0.003 0.025 -0.022 0.421 0.448 76 0.002 0.020 -0.018 0.391 0.447 77 0.006 0.033 -0.027 0.540 0.564 78 -0.003 0.017 -0.020 0.342 0.393 79 -0.007 0.023 -0.030 0.294 0.511 80 -0.004 0.020 -0.024 0.405 0.463 81 0.007 0.026 -0.019 0.361 0.509 82 0.002 0.035 -0.033 0.518 0.580 83 0.000 0.032 -0.032 0.297 0.514 84 -0.011 0.027 -0.038 0.342 0.395 85 0.003 0.025 -0.022 0.479 0.607 86 0.009 0.033 -0.024 0.372 0.673 87 0.009 0.044 -0.035 0.558 0.569 88 -0.002 0.030 -0.032 0.389 0.498 89 0.014 0.037 -0.023 0.303 0.564 90 -0.002 0.024 -0.026 0.398 0.568 91 0.009 0.026 -0.017 0.334 0.500 92 0.007 0.041 -0.034 0.533 0.509 93 -0.003 0.025 -0.028 0.336 0.459 94 -0.005 0.029 -0.034 0.324 0.379 95 -0.005 0.028 -0.033 0.450 0.567 96 0.000 0.028 -0.028 0.413 0.532 97 0.014 0.057 -0.043 0.514 0.570

Note: POSC=average diversification change of continuing firms with positive change NEGC=average diversification change of continuing firms with negative change POSB=average diversification change of starting firms NEGD=average diversification change of shutting-down firms

99

Table AC-7 Firm level Diversification Index Change Decomposed by Diversified/Specialized Plants (MU Firms) Diversified Specialized Net Entry Production Production Plant year D(t-1) D(t) D(t)-D(t-1) 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 0.77045 0.76725 0.76902 0.76322 0.75931 0.75342 0.74905 0.75328 0.74102 0.74617 0.75148 0.74545 0.74465 0.72989 0.7284 0.7313 0.72715 0.72277 0.70572 0.70243 0.70013 0.69309 0.68796 0.6645 0.76725 0.76902 0.76322 0.75931 0.75342 0.74905 0.75328 0.74102 0.74617 0.75148 0.74545 0.74465 0.72989 0.7284 0.7313 0.72715 0.72277 0.70572 0.70243 0.70013 0.69309 0.68796 0.6645 0.67831 -0.0032 0.001771 -0.00581 -0.00391 -0.00589 -0.00437 0.004227 -0.01226 0.00515 0.005311 -0.00602 -0.0008 -0.01476 -0.00149 0.002899 -0.00415 -0.00438 -0.01705 -0.00329 -0.00231 -0.00704 -0.00513 -0.02346 0.013819 0.001 0 0.002 0.001 0.006 0 0 -0.002 0 0.006 0.001 0.004 -0.011 0.008 0.004 0 0 -0.002 0 0.006 0.001 0.003 -0.003 0.003 0 0.006 -0.008 0.017 0.072 0.004 -0.007 -0.017 0.004 0.001 0.009 0.006 -0.019 0.004 0.005 -0.004 0.001 -0.03 0.015 0.035 0 -0.002 -0.01 -0.006 0.002 -0.001 0 -0.009 -0.039 0 0.007 0.007 0.001 -0.001 -0.002 -0.003 0.012 -0.001 -0.003 0.005 0 0.018 -0.007 -0.025 0.003 0.003 0.007 0.005

Note:

? ? ? ? ? ? 2 2 2 2 2 2 ? S S ?d t = ? ? ? ? ? , , 1 i t i t ? ?? ? ? S i ,t ? ? S i ,t ?1 ?? ? ? S i ,t ? ? S i ,t ?1 ? ? i?PN , PC i?PC , PX i?PC i?PC i?PN i?PX ?4 ?1 ?4 ? ?1 ?4 4424 443 4424 443 1 44 4 24 444 3
Diversified Production Factor Specialized Production Factor Plant Net Enry Factor

Where,
i ? PC,PN product i which is produced both in plant PC and PN i ? PC product i which is produced only in plant PC PC = Continuously operating plant at time t-1 and t PN = new plant in time t PX = exiting plant in time t

100

Table AC-8 Firm Level Diversification Change Decomposed by Intensive/Extensive Components (Continuing MU Firms) year D(t-1) D(t) D(t)-D(t-1) intensive extensive 75 0.72984 0.72827 -0.00157 0.014 -0.065 76 0.72827 0.73121 0.002944 0.013 -0.063 77 0.73121 0.72929 -0.00192 -0.008 -0.326 78 0.72929 0.72479 -0.0045 0.022 -0.054 79 0.72479 0.71227 -0.01253 0.087 -0.078 80 0.71227 0.70771 -0.00456 0.008 -0.084 81 0.70771 0.716 0.008293 0 -0.072 82 0.716 0.70695 -0.00905 -0.032 -0.425 83 0.70695 0.71414 0.007188 0.008 -0.35 84 0.71414 0.7109 -0.00324 0.008 -0.218 85 0.7109 0.70861 -0.00229 0.02 -0.083 86 0.70861 0.70209 -0.00652 0.021 -0.167 87 0.70209 0.69528 -0.00681 -0.016 -0.659 88 0.69528 0.6941 -0.00118 0.033 -0.119 89 0.6941 0.68754 -0.00656 0.009 -0.114 90 0.68754 0.68208 -0.00546 0 -0.077 91 0.68208 0.6795 -0.00258 0.006 -0.104 92 0.6795 0.66403 -0.01547 -0.034 -0.464 93 0.66403 0.66211 -0.00192 0.026 -0.103 94 0.66211 0.65296 -0.00915 0.037 -0.089 95 0.65296 0.64973 -0.00323 0.011 -0.067 96 0.64973 0.64735 -0.00237 0.012 -0.078 97 0.64735 0.62857 -0.01878 -0.017 -0.412

Note:

? 2 2 2 2 ? ?d t = ? ? S it ? S it ?1 ? ? ? S it ? ? S it ?1 ? i?NC i?NN i?NX ?4 ? 1 44 244 31 44 24 44 3

(

)

Intensive Component

Extensive Component

where, NC = Products which are continuously produced at time t and t-1 NN = New products at time t NX = Exiting products at time t

101

Table AC-9 Firm Level Diversification Change Decomposition (Continuing MU Firms) year D(t-1) D(t) D(t)-D(t-1) I II III IV V 75 0.729 0.728 -0.002 -0.001 0.000 -0.003 0.002 -0.001 76 0.728 0.731 0.003 0.000 0.000 -0.009 0.001 0.002 77 0.731 0.729 -0.002 -0.001 -0.001 0.008 0.003 -0.003 78 0.729 0.725 -0.005 -0.001 0.000 -0.017 0.001 0.007 79 0.724 0.712 -0.013 -0.005 0.000 -0.071 0.001 0.033 80 0.712 0.708 -0.005 0.000 0.000 -0.004 0.000 0.000 81 0.707 0.716 0.008 0.000 0.000 0.006 0.001 -0.006 82 0.716 0.707 -0.009 0.002 0.000 0.018 -0.002 -0.005 83 0.706 0.714 0.007 0.000 0.000 -0.003 0.000 0.000 84 0.714 0.711 -0.003 -0.006 0.000 -0.001 0.001 0.003 85 0.710 0.709 -0.002 -0.001 0.000 -0.010 0.002 0.002 86 0.708 0.702 -0.007 -0.003 0.000 -0.008 0.004 0.001 87 0.702 0.695 -0.007 0.005 0.005 0.007 0.009 -0.006 88 0.695 0.694 -0.001 -0.008 0.000 -0.005 0.006 -0.003 89 0.694 0.688 -0.007 -0.004 0.000 -0.002 0.000 0.000 90 0.687 0.682 -0.005 0.001 0.000 0.001 0.002 -0.002 91 0.682 0.680 -0.003 0.000 0.000 -0.003 0.001 0.000 92 0.679 0.664 -0.015 0.002 0.000 0.031 0.001 -0.017 93 0.664 0.662 -0.002 0.000 0.000 -0.020 0.004 0.007 94 0.662 0.653 -0.009 -0.006 0.000 -0.033 0.000 0.021 95 0.652 0.650 -0.003 -0.001 0.000 -0.001 0.001 -0.003 96 0.649 0.647 -0.002 -0.003 0.000 0.002 0.002 -0.005 97 0.647 0.629 -0.019 0.003 0.000 0.009 -0.001 -0.004

VI 0.000 0.000 -0.001 0.000 0.004 0.000 -0.001 0.000 -0.002 -0.002 0.000 0.000 -0.003 -0.001 0.000 -0.001 0.000 -0.002 0.000 0.001 0.000 0.000 -0.001

Note:
2 2 2 2 2 2 ? ? ( ? S it ? ? S it ?1 ) ? ( ? S it ? ? S it ?1 ) ? ( ? S it ? ? S it ?1 ) ? i?PN , PC i?PC i?PC i?PC , PX i?PN , PC i?PC , PX ? 14 i?NN i?NX i?NC i?NC i?NC i?NC 4424 443 14 4424 443 14 4 4 24 44 3 ? III I II ?d t = ? 2 2 2 2 2 2 ? ? ( ? S it ? ? S it ?1 )? ( ? S it ? ? S it ?1 )? ( ? S it ? ? S it ?1 ) PC i?PC i?PN i?PX i?PN i?PX ? ii? ?NN i?NX i?NC i?NC i?NN i?NX 4 24 44 3 14 4 4 24 44 3 14 4 4 24 44 3 ? 144 IV V VI ?

where, i ? PC,PN product i which is produced both in plant PC and PN i ? PC product i which is produced only in plant PC PC = Continuously operating plant at time t-1 and t PN = new plant in time t PX = exiting plant in time t NC = Products which are continuously produced at time t and t-1 NN = New products at time t NX = Exiting products at time t

102

Table AC-10 Average Diversification index by Firm Size Quartile (using Total Employment) year Q1 Q2 Q3 Q4 74 0.07024 0.11382 0.14729 0.75515 75 0.08011 0.10555 0.15428 0.75116 76 0.06637 0.08019 0.1505 0.75283 77 0.07045 0.1081 0.16285 0.74781 78 0.07293 0.10058 0.15412 0.74418 79 0.07987 0.09921 0.13989 0.73631 80 0.07995 0.1018 0.14419 0.73015 81 0.08231 0.10548 0.14073 0.73629 82 0.11481 0.13247 0.16729 0.72735 83 0.10931 0.12754 0.17919 0.73425 84 0.07759 0.10823 0.15079 0.73811 85 0.07221 0.10397 0.15143 0.73414 86 0.07345 0.10293 0.15335 0.72898 87 0.10895 0.1343 0.1843 0.72023 88 0.12096 0.1465 0.19893 0.72184 89 0.06612 0.10534 0.14494 0.71398 90 0.0256 0.05889 0.10398 0.7002 91 0.06967 0.07131 0.09999 0.69605 92 0.09566 0.11621 0.14918 0.67995 93 0.04253 0.08192 0.13201 0.67592 94 0.02035 0.04344 0.07967 0.66692 95 0.03054 0.05187 0.08986 0.66235 96 0.02431 0.05761 0.08899 0.65909 97 0.04765 0.08467 0.11322 0.6391 98 0.06216 0.08942 0.12658 0.65602

103

Table AC-11 Average Diversification index by Firm Age Quartile year Q1 Q2 Q3 74 0.10838 0.73304 75 0.09881 0.73066 76 0.10149 0.7334 77 0.11194 0.74246 0.20676 78 0.10493 0.7389 0.21242 79 0.11248 0.73188 0.18803 80 0.1119 0.7268 0.18362 81 0.13044 0.73308 0.19815 82 0.18572 0.72829 0.23278 83 0.21439 0.73565 0.22702 84 0.16314 0.74176 0.21376 85 0.16239 0.73685 0.22128 86 0.15152 0.73123 0.21645 87 0.20567 0.72757 0.24972 88 0.2022 0.72996 0.25988 89 0.1482 0.71243 0.21835 90 0.01702 0.72989 0.13726 91 0.06007 0.71629 0.10023 92 0.17508 0.71894 0.17977 93 0.14875 0.70952 0.16359 94 0.00988 0.6992 0.15489 95 0.00819 0.70008 0.14159 96 0.03105 0.69624 0.13328 97 0.033 0.67854 0.09753 98 0.0651 0.69799 0.18122

Q4

0.25671 0.27856 0.2963 0.26939 0.24911 0.22522 0.23635 0.25665 0.28749

Note: The firm age variable is very limited in the data and it is not easy to determine the exact age if the firm is already old in early years of the panel. For example, most of firms are one year old or eleven years old in 1974. Therefore, we get only Q1 and Q2 in 1974.

104

Table AC-12 Average Diversification Index by Region year 1 2 3 4 5 74 0.77 0.73 0.75 0.68 0.75 75 0.76 0.72 0.75 0.68 0.75 76 0.73 0.72 0.73 0.71 0.72 77 0.71 0.75 0.75 0.68 0.70 78 0.72 0.76 0.74 0.71 0.71 79 0.74 0.73 0.70 0.68 0.74 80 0.68 0.69 0.71 0.75 0.74 81 0.69 0.72 0.70 0.73 0.72 82 0.66 0.69 0.70 0.70 0.69 83 0.64 0.71 0.73 0.72 0.72 84 0.67 0.70 0.72 0.71 0.71 85 0.65 0.66 0.72 0.72 0.75 86 0.64 0.68 0.74 0.68 0.70 87 0.61 0.70 0.73 0.69 0.66 88 0.66 0.70 0.71 0.70 0.68 89 0.55 0.70 0.68 0.67 0.73 90 0.69 0.63 0.69 0.64 0.71 91 0.69 0.64 0.70 0.68 0.68 92 0.58 0.70 0.65 0.63 0.69 93 0.56 0.67 0.69 0.55 0.68 94 0.53 0.61 0.66 0.62 0.67 95 0.54 0.66 0.68 0.62 0.68 96 0.60 0.64 0.63 0.61 0.68 97 0.67 0.59 0.59 0.56 0.67 98 0.53 0.63 0.63 0.60 0.66 -0.30 -0.14 -0.16 -0.11 -0.12 growth 0.65 0.68 0.70 0.67 0.70 avg

6 0.70 0.69 0.75 0.71 0.71 0.69 0.68 0.73 0.75 0.71 0.77 0.74 0.68 0.72 0.77 0.69 0.66 0.73 0.68 0.70 0.68 0.66 0.68 0.67 0.69 -0.01 0.71

7 0.68 0.67 0.67 0.70 0.67 0.69 0.72 0.73 0.70 0.70 0.70 0.70 0.70 0.70 0.70 0.68 0.72 0.69 0.68 0.67 0.69 0.65 0.67 0.72 0.64 -0.07 0.69

8 0.63 0.69 0.81 0.73 0.72 0.67 0.67 0.72 0.69 0.72 0.70 0.77 0.76 0.71 0.76 0.75 0.69 0.69 0.62 0.62 0.65 0.67 0.69 0.56 0.71 0.12 0.70

9 0.70 0.72 0.75 0.74 0.71 0.70 0.69 0.72 0.74 0.72 0.69 0.66 0.68 0.68 0.61 0.62 0.66 0.65 0.64 0.63 0.66 0.56 0.62 0.53 0.64 -0.08 0.67

Region: Census divides the survey coverage area into nine regions 1- New England 2- Middle Atlantic 3- East North Central 4- West North Central 5- South Atlantic 6- East South Central 7- West South Central 8- Mountain 9- Pacific

105

Table AC-13 Average Diversification Index by Quintile of Share of Interplant Transfer (IPT/TVS) year Q1 Q2 Q3 Q4 Q5 76 0.42182 0.7558 0.79679 0.82443 0.79593 80 0.3851 0.73711 0.7877 0.80484 0.79789 81 0.37644 0.71383 0.80569 0.81557 0.78551 83 0.37729 0.72743 0.79797 0.81067 0.79484 84 0.35734 0.74381 0.78309 0.82224 0.80036 85 0.69318 0.88635 0.84501 0.84354 0.82585 86 0.64833 0.87164 0.85129 0.8146 0.76701 87 0.35649 0.69586 0.77471 0.80134 0.79128 88 0.3628 0.71831 0.79106 0.80681 0.77924 89 0.34989 0.7215 0.77604 0.81465 0.78827 90 0.33929 0.7307 0.76133 0.79746 0.78621 91 0.3516 0.7363 0.77439 0.80103 0.76788 92 0.36359 0.69286 0.75283 0.80838 0.76168 93 0.3722 0.72572 0.801 0.8011 0.75524 94 0.3642 0.75497 0.78134 0.80233 0.76531 95 0.37081 0.74259 0.77943 0.79182 0.768 96 0.38238 0.72306 0.78326 0.79478 0.74925 97 0.35028 0.70035 0.75996 0.77148 0.73377 98 0.3979 0.71243 0.75979 0.77466 0.72153 avg 0.401102 0.741612 0.787509 0.805354 0.775529 growth -0.05671 -0.05738 -0.04644 -0.06037 -0.09348

Note: If the firm is vertically integrated, the firm will diversify into the products that are consumed within the firm to produce the final product. The share of Interplant Product Transfer (IPT) to the total value of shipment of the firm (TVS) can be used as an indicator for vertical integration. IPT is not available in 1974, 1975, 1977-1979 and 1982. IPT is imputed by Census in 1985 and 1986.

106

Table AC-14 Average Diversification Index by Quartile of Share of Labor Cost (Wage/Total variable cost) year Q1 Q2 Q3 Q4 74 0.73195 0.77103 0.70465 0.47082 75 0.73183 0.77237 0.68039 0.42186 76 0.74156 0.75945 0.68572 0.39776 77 0.74148 0.75151 0.68137 0.4805 78 0.73501 0.75453 0.66543 0.46308 79 0.72745 0.72381 0.70646 0.42038 80 0.7181 0.739 0.67844 0.408 81 0.72583 0.74206 0.68887 0.42501 82 0.71783 0.73063 0.68983 0.39111 83 0.72828 0.72869 0.69865 0.40078 84 0.72607 0.73752 0.68569 0.4363 85 0.72517 0.72687 0.68406 0.48524 86 0.71799 0.72963 0.66074 0.46496 87 0.70771 0.72373 0.66228 0.50532 88 0.71597 0.70859 0.64846 0.42453 89 0.71136 0.69815 0.61738 0.45333 90 0.70265 0.70381 0.60421 0.42153 91 0.69841 0.69354 0.60926 0.32964 92 0.67997 0.68176 0.63447 0.36963 93 0.69016 0.65971 0.57821 0.36207 94 0.68736 0.64837 0.54104 0.41022 95 0.68482 0.61582 0.55917 0.38524 96 0.68435 0.59943 0.55329 0.33004 97 0.66233 0.56502 0.52385 0.32556 98 0.6649 0.63352 0.56906 0.31085 average 0.710342 0.703942 0.640439 0.41175 growth -0.0916 -0.17835 -0.19242 -0.33977

107

Table AC-15 Average Diversification Index by Quartile of Share of Non-production Worker Labor Cost (Non-production worker wage/Total labor cost) year Q1 Q2 Q3 Q4 74 0.55187 0.75475 0.76206 0.70286 75 0.56427 0.7471 0.76438 0.70123 76 0.57483 0.75197 0.75958 0.70622 77 0.59947 0.7513 0.76723 0.69581 78 0.69377 0.74991 0.74476 0.70855 79 0.69686 0.72402 0.74055 0.68786 80 0.66241 0.72518 0.73253 0.69428 81 0.6767 0.72763 0.73707 0.7065 82 0.54637 0.72389 0.73332 0.70329 83 0.69501 0.71472 0.74301 0.69907 84 0.69692 0.7192 0.74217 0.68599 85 0.68971 0.7229 0.73595 0.68724 86 0.63953 0.73491 0.73584 0.67633 87 0.45808 0.6935 0.73113 0.71745 88 0.69406 0.70099 0.73571 0.64684 89 0.68519 0.70169 0.71826 0.64999 90 0.66614 0.70691 0.70374 0.66552 91 0.60434 0.71988 0.72284 0.63657 92 0.37518 0.65645 0.68251 0.68811 93 0.65794 0.68016 0.69868 0.61852 94 0.64731 0.67238 0.68941 0.60972 95 0.65646 0.68186 0.65506 0.60829 96 0.61303 0.71463 0.63529 0.60625 97 0.6275 0.66142 0.68409 0.56785 98 0.58693 0.68624 0.66646 0.58956 average 0.622395 0.712944 0.720865 0.666396 growth 0.063529 -0.09077 -0.12545 -0.1612

108

Table AC-16 Average Diversification Index by Quartile of Share of Exported Good (Vale of exported good/Total value of shipment) year zero Q1 Q2 Q3 Q4 76 0.39333 0.73924 0.782 0.79292 0.79241 80 0.37168 0.715 0.77888 0.79276 0.74343 81 0.35137 0.72602 0.78731 0.77326 0.77691 83 0.35646 0.71234 0.76977 0.7855 0.76804 84 0.35952 0.71027 0.78391 0.78461 0.74787 85 0.69156 0.84352 0.53336 0.16977 0.1389 86 0.63121 0.83046 0.73142 0.62301 0.15634 87 0.33951 0.67881 0.75178 0.77366 0.75517 88 0.34016 0.65229 0.7524 0.77518 0.74045 89 0.31167 0.68086 0.73958 0.77748 0.73457 90 0.30347 0.65271 0.75553 0.76801 0.71493 91 0.29289 0.66419 0.7519 0.7608 0.69294 92 0.29181 0.60629 0.73559 0.7554 0.66064 93 0.32628 0.68021 0.73716 0.75447 0.63553 94 0.31695 0.65814 0.7278 0.76065 0.65382 95 0.30225 0.67236 0.72279 0.74663 0.65976 96 0.29083 0.65728 0.73251 0.73671 0.64809 97 0.29153 0.57485 0.71318 0.72691 0.61513 98 0.33604 0.71924 0.61496 0.72717 0.67017 Average 0.36308 0.693373 0.731675 0.725521 0.647637 growth -0.14565 -0.02705 -0.21361 -0.08292 -0.15426

109

Table AC-17 Decade Average of Number of Industries of Firms by Number of Products (5-digit SIC) of Firms Number of 2-digit SIC industry Number of Products 1 2 3 4 5 6 7 8 9 10+ 1970s 1 1.14 1.29 1.45 1.61 1.78 2.04 2.23 2.54 2.73 1980s 1 1.17 1.34 1.47 1.61 1.75 1.99 2.12 2.41 2.53 1990s 1 1.16 1.31 1.43 1.52 1.69 1.84 1.99 2.12 2.37

Number of 3-digit SIC industry Number of Products 1 2 3 4 5 6 7 8 9 10+ 1970s 1 1.25 1.58 1.93 2.24 2.66 3.11 3.58 4.12 4.46 1980s 1 1.28 1.63 1.95 2.30 2.64 3.11 3.44 3.96 4.23 1990s 1 1.27 1.59 1.86 2.15 2.47 2.89 3.16 3.49 3.93

110

Table AC-18 Decade Average of Number of Counties Where Plants Are Located by Number of Plants Number of 1970s 1980s 1990s Plants 1 1 1 1 2 1.7767 1.8325 1.82546 3 2.5164 2.64582 2.68428 4 3.3936 3.52944 3.51556 5 4.19915 4.33805 4.43135 6 5.02278 5.25762 5.0979 7 5.96204 6.0732 6.11653 8 6.47792 6.91824 6.88664 9 7.30809 7.66521 7.63317 10+ 8.398 8.1173 8.7455

111

Table AC-19 Evolution of Aggregate and Idiosyncratic Volatility Idiosyncratic Volatility Year Aggregate Volatility Mean Standard deviation 74 0.2829 0.45160 0.53751 75 0.25907 0.47257 0.55326 76 0.24071 0.47965 0.56108 77 0.23216 0.44162 0.54740 78 0.22878 0.44330 0.54637 79 0.20681 0.36748 0.45144 80 0.165 0.36301 0.44863 81 0.16504 0.34636 0.42912 82 0.16587 0.36942 0.46405 83 0.16741 0.36309 0.47956 84 0.13238 0.46554 0.54460 85 0.13134 0.47380 0.55781 86 0.15054 0.49389 0.56556 87 0.14018 0.50964 0.59999 88 0.14104 0.50290 0.61146 89 0.14093 0.35073 0.51840 90 0.1411 0.28708 0.44155 91 0.14162 0.26169 0.42493 92 0.14266 0.26486 0.47525 93 0.12227 0.27937 0.56882 94 0.10086 0.28571 0.54273 95 0.10776 0.27957 0.52498 96 0.09622 0.26822 0.49589 97 0.09772 0.27511 0.51267 98 0.09138 0.27344 0.50223

112

Reference
Abel, A. and J. Eberly, "Investment and q with Fixed Costs: An Empirical Analysis," mimeo (2001b) Abel, A. and J. Eberly, "Q Theory without Adjustment Costs and Cash Flow Effects without Financing Constraints," mimeo (2001a) Abel, A. and O. Blanchard, "Present Value of Profits and Cyclical Movements in Investment," Econometrica 54, Issue 2 (1986), pp.249-73 Anderson, S., A. de Palma and Y. Nesterov, "Multiproduct Firms: A Nested Logit Approach," Journal of Industrial Economics, Vol. 40 (1992), pp.261-276 Anderson, S., A. de Palma and Y. Nesterov, "Oligopolistic Competition and the Optimal Provision of Products," Econometrica, Vol. 63 (1995), pp.1281-1301 Axarloglou, K, "The Cyclicality of New Product Introductions," Journal of Business (2003) Barnett, S. and P. Sakellaris, "A New Look at Firm Market Value, Investment, and Adjustment Costs," Review of Economics and Statistics, Vol. 81:2 (May 1999), pp.250-260 Blanchard, O. and J. Simon, "The Long and Large Decline in U.S. Output Volatility," Brookings Papers on Economic Activity (2001) Bond, S. and J. Cummins, "The Stock Market and Investment in the New Economy: Some Tangible Facts and Intangible Fictions," Brookings Paper on Economic Activity (2000), pp.61-124 Bourne, Edward G. The History of the Surplus Revenue Act of 1837. Original published in 1885. Reprinted by Burt Franklin: New York, 1968. Cabellero, R., E. Engel and J. Haltiwanger, "Plant Level Adjustment and Aggregate Investment Dynamics," Brookings Paper on Economic Activity (1995)

113

Callender, Guy Stevens. "The Early Transportation and Banking Enterprises of the States."

Quarterly Journ

Campa, J., and S. Kedia, "Explaining the Diversification Discount," Journal of Finance (Aug 2002) Chatterjee, S. and R. Cooper, "Entry and Exit, Product Variety And The Business Cycle," NBER Working Paper 4562 (1993) Clapham, John. The Bank of England. Cambridge: Cambridge University Press, 1945. Comin, D and S. Mulani, "Diverging Trends in Macro and Micro Volatility: Facts," Economic Research Reports, New York University (Sep 2003) Cooper, R. and J. Haltiwanger, "On the Nature of Capital Adjustment Costs," NBER Working Paper 7925 (2000) Dunne, T., M. Roberts and L. Samuelson, "The Growth and Failure of U.S. Manufacturing Plants," Quarterly Journal of Economics, Volume 104, Issue 4 (Nov. 1989), pp.671698 English, William B. "Understanding the Costs of Sovereign Default: American State Debts in the 1840's." American Economic Review, 86, March 1996, 259-75. Fatout, Paul. Indiana Canals. West Lafayette, Indiana: Purdue University Studies, 1972. Fazzari, S., R. Hubbard and B. Petersen, "Financing Constraints and Corporate Investment," Brookings Papers on economic Activity, No. 1 (1998), pp.141-95 Federal Trade Commission, "The Petroleum Industry: Mergers, Structural Change, and Antitrust Enforcement," Staff Report (Aug 2004) Federal Trade Commission, "U.S. Carbonated Soft Drink Bottling Industry and Antitrust Policy Since 1980," Staff Report (Dec 1999) Gayer, Arthur D., W.W. Rostow, and Anna Jacobson Schwartz. The Growth and Fluctuation of the British Economy, 1790-1850. Oxford: Clarendon Press, 1953.

114

Gemba, K., F. Kodama, "Diversification Dynamics of the Japanese Industry," Research Policy, v30, n8 (Oct 2001), pp1165-84 Giandrea, M., "The Effect of Industry Evolution on Total Factor Productivity Growth," memeo (2002) Gilchrist, S. and C. Himmelberg, "Evidence on the Role of Cash Flow for Investment," Journal of Monetary Economics, Vol. 36 (1995), pp.541-572 Gollop, F. and J. Monahan, "A Generalized Index of Diversification: Trends in U.S. Manufacturing," The Review of Economics and Statistics 73 (1991), pp.318-330 Goodrich, Carter. Government Promotion of American Canals and Railroads. New York: Columbia University Press, 1960. Govan, Thomas Payne. Nicholas Biddle. Chicago: University of Chicago Press, 1959. Gross, D., "Investment and Financing Decisions of Liquidity Constrained Firms," (Nov. 1994) Hammond, Bray. "The Chestnut Raid on Wall Street, 1839." Quarterly Journal of Economics, 61, 4, August 1947, 1-23. Hammond, Bray. Banks and Politics in America: From the Revolution to the Civil War. Princeton: Princeton University Press, 1957. Hidy, Ralph W. The House of Baring in American Trade and Finance. Cambridge: Harvard University Press, 1949. Jenks, Leland Hamilton. The Migration of British Capital to 1875. New York: Knopf, 1938. Jorgensen, S. and P. Kort, "Optimal Investment and Finance in Renewable Resource Harvesting," Journal of Economic dynamics and Control, Vol. 21 (1997), pp.603-630 Jovanovic, V., "The Diversification of Production," Brookings Paper: Microeconomics (1993) Kahn, J. M. McConnell and G. Perez-Quiros, "On the Causes of the Increased Stability of the U.S. Economy," Economic Policy Review, Federal Reserve Bank of New York (May 2002) Kettell, Thomas Prentice, "Debts and Finances of the States of the Union: With Reference to the

115

General Condition and Character; Chapter IV, Middle States B Pennsylvania." The Merchant Magazine, vol. XX, no III, (March, 1849), 256-68. Keuschnigg, C., "Business Formation and Aggregate Investment," German Economic Review (2001) Kim, C., C. Nelson, and J. Piger, "The Less-Volatile U.S. Economy: A Bayesian Investigation of Timing, Breadth, and Potential Explanations," Journal of Business and Economic Statistics (Jan 2004) Kirkpatrick, David D., "Independent Publisher is Diversifying," New York Times (Mar 2002) Krenkel, John H, Illinois Internal Improvements 1818-1848. Cedar Rapids: The Torch Press, 1958. Larson, John Lauritz. Internal Improvements. University of North Carolina Press (2001) Lins K. and H. Servaes, "International Evidence on the Value of Corporate Diversification," The Journal of Finance, vol. 54, no. 6 (1999) Macesich, George. "Sources of Monetary Disturbances in the United States, 1834-1845." Journal of Economic History, 20 3, September 1960, pp. 407-34. Maksimovic, V. and G. Phillips, "Do Conglomerate Firms Allocate Resources Inefficiently?" Journal of Finance (2002) Maksimovic, V. and G. Phillips, "The Market for Corporate Assets: Who Engages in Mergers and Asset Sales and Are There Efficiency Gains?" Journal of Finance (2001) Matthews, R.C.O. A Study in Trade-Cycle History. Cambridge: Cambridge University Press, 1954. McConnell, M. and G. Perez-Quiros, "Output Fluctuations in the United States: What Has Changed Since the Early 1980's?" American Economic Review, Vol. 90, No. 5 (Dec 2000), pp1464-1476 McGrane, Reginald C. Foreign Bondholders and American State Debts. New York: Macmillan

116

Company, 1935. Meng, Xiao-Li and Donald Rubin, "Maximum Likelihood Estimation via the ECM Algorithm: A General Framework." Biometrica, June 1993, pp. 267-98. Officer, Lawrence H., "Integration in the American Foreign Exchange Market, 1791-1900," Journal of Economic History, Vol. 45, 3, September, 1985, 557-85. Ottaviano, G. and J. Thisse, Monopolistic competition, Mutiproduct Firms and Optimum Product Diversity (1999) Pepall, L and G. Norman, "Product Differentiation and Upstream-Downstream Relations," Journal of Economics and Management Strategy, Vol. 10 (Summer 2001), pp.201-33 Ratchford, B. U. American State Debts. Durham: Duke University Press, 1941. Rousseau, Peter L. "Jacksonian Monetary Policy, Specie Flows, and the Panic of 1837." NBER Working Paper 7528, February 2000. Sakellaris, P., "Patterns of Plant Adjustment," Working Paper, University of Maryland, 2000 Samli, A. Coskun, "Corporate Search for Market Power: Seven Strategic Megatrends of the Twentieth Century," Business and Society Review 109 (Summer 2004), pp. 263-280 Scheiber, Harry N., Ohio Canal Era: A Case Study of Government and the Economy, 1820-1861. Athens, Ohio: The Ohio State University Press, 1969. Schoar, A., "Effects of Corporate Diversification on Productivity," The Journal of Finance, Volume 57 Issue 6 (2002) Smith, Walter B. and Arthur Harrison Cole. Fluctuations in American Business, 1790-1860. Harvard University Press: Cambridge, 1935. Smith, Walter B. Economic Aspects of the Second Bank of the United States. Cambridge: Harvard University Press, 1953. Stock, J., and M. Watson, "Has the Business Cycle Changed and Why?" NBER Working Paper (Sep 2002)

117

Sylla, Richard E., Wilson, Jack, and Wright, Robert. Price Quotations in Early U.S. Securities Markets. Database archived at ICPSR. Temin, Peter. "The Anglo-American Business Cycle, 1820-60," The Economic History Review Temin, Peter. The Jacksonian Economy. New York: W. W. Norton and Company, 1969. Timberlake, Richard H. "The Specie Circular and the Distribution of the Surplus." Journal of Political Economy, 68, 2, April 1960, 109-117. Timberlake, Richard H. "The Specie Standard and Central Banking in the United States Before 1860." Journal of Economic History, 21, 3, (Sept 1961) pp. 318-341. Villalonga, B., "Does Diversification Cause the Diversification Discount?" Financial Management (2004) Wallis, John Joseph, Richard E. Sylla, Arthur Grinath III, "Sovereign Repudiation and Default: The Emerging-Market Debt Crisis in U.S. States, 1839-1843," Mimeo. Wallis, John Joseph. "The Property Tax as a Coordinating Device: Financing Indiana"s Mammoth System of Internal Improvements." Explorations in Economic History, July 2003. Whited, T., "Debt, Liquidity Constraints and Corporate Investment: Evidence from Panel Data," Journal of Finance 47 (September 1992), pp.1425-60 Wilcox, H Dixon, Kuo-Chung Chang, and Varun Grover, "Valuation of Mergers and Acquisitions in the Telecommunications Industry: A study on Diversification and Firm Size," Information and Management 38 (Aug 2001) Winter, J., "Does Firms' Financial Status Affect Plant-Level Investment and Exit Decisions?" CES Working Paper, Bureau of Census (1999)

118



doc_305858315.pdf