SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Individual Decision to Become an Entrepreneur and Neighborhood Entrepreneurial Activity
  5. 3. Data Description and Identification Strategy
  6. 4. Results
  7. 5. Robustness and Alternative Explanations
  8. 6. Conclusion
  9. Appendices
  10. References

We show that individuals residing in highly entrepreneurial neighborhoods are more likely to become entrepreneurs and invest more into their own businesses, even though their entrepreneurial profits are lower and their alternative job opportunities more attractive. Our results suggest that peer effects create nonpecuniary benefits from entrepreneurial activity and play an important role in the decision to become an entrepreneur. Alternative explanations, such as entry costs, social learning, and informal credit markets, are not supported by the data.


1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Individual Decision to Become an Entrepreneur and Neighborhood Entrepreneurial Activity
  5. 3. Data Description and Identification Strategy
  6. 4. Results
  7. 5. Robustness and Alternative Explanations
  8. 6. Conclusion
  9. Appendices
  10. References

Entrepreneurial activity and new firm formation are considered engines of economic growth and innovation (Schumpeter, 1934; Baumol, 1990; Murphy et al., 1991). Nevertheless, the field of economics is still far from having a complete understanding of what drives an individual to start a business. Research thus far has stressed the role of individual characteristics, access to initial capital (Evans and Jovanovic, 1989; Evans and Leighton, 1989), and, more recently, institutions and entry regulation (Desai et al., 2005; Klapper et al., 2006).

Social interactions may also play a role because, as shown by a growing body of literature, they affect the payoffs from a variety of economic decisions.1 There are several reasons why social interactions may matter for entrepreneurship. Individuals may learn how to run a business by observing their peers. Consistent with this view, Gompers et al. (2005) find that entrepreneurial learning and networks appear important in creating venture-backed firms. Social interactions may also create nonpecuniary benefits from entrepreneurial activity (e.g., social norms affecting whether entrepreneurs are highly regarded). Nonpecuniary benefits may have a major influence on occupational choice. Social status, prestige, or simply, the desire to conform increase the utility from being an entrepreneur (Cole et al., 1992; Bernheim, 1994).

Social scientists have long recognized the importance of peer effects on the creation of new firms and have surmised that this may lead to geographic concentrations of entrepreneurial activity.2 However, the mechanisms through which peer effects operate are unclear. In this paper, we attempt to fill this gap.

If peer effects are relevant, the probability of an individual becoming an entrepreneur is always expected to increase with the rate of entrepreneurship in her social group. We focus on other aspects of entrepreneurial activity to determine the mechanisms through which peer effects operate. If peer effects increase entrepreneurial nonpecuniary benefits, relatively less productive individuals may choose to become entrepreneurs. Hence, we should observe that entrepreneurial profits (investment) are lower (higher) for individuals belonging to highly entrepreneurial social groups. These predictions contrast with what social learning, knowledge spillovers, and other forms of agglomeration economies (Saxenian, 1994) would imply, as entrepreneurial profits would be higher if these factors enhanced entrepreneurial productivity (Glaeser et al., 1992).

The empirical identification of peer effects is challenging for two reasons. First, it is generally difficult to define an individual's peer group. While we are unable to observe an individual's contacts, we have information on the neighborhood (municipality) in which an individual lives. We thus identify an individual's social group with her neighborhood. We are confident that we capture at least some of an individual's contacts because urban sociology studies (Wellman, 1996) show that a surprisingly high fraction of interactions take place among individuals who live in the same neighborhood.

Second, the correlation between individual and aggregate outcomes may be driven by the simultaneity of choices, unobserved neighborhood characteristics, or economic conditions (Manski, 1993). We attempt to overcome these econometric challenges by using an approach pioneered by Bertrand et al. (2000). We analyze entry in the entrepreneurial activity within local labor markets (henceforth, LLMs). Within a LLM, economic incentives and opportunities are homogeneous because individuals can easily commute. Still, individuals living in different municipalities (within the LLM) have closer interactions with their neighbors. If peer effects exist, entrepreneurial activity in the municipality where an individual lives should affect the probability of becoming an entrepreneur, entrepreneurial profits, and investment, after controlling for LLM fixed effects (which capture omitted economic factors), dummies for the wealthiest and poorest municipalities within a LLM (which capture the fact that some social groups segregate in different areas and may have unobservable characteristics affecting their propensity to the entrepreneurial activity), and extensive individual and municipality level controls.

To further reduce concerns that our estimates are affected by omitted neighborhood characteristics, we instrument our proxies for entrepreneurial activity by using predetermined municipality cultural values.3 Our instruments help to explain differences in entrepreneurial activity across neighborhoods, but should not directly affect an individual's decision to become an entrepreneur, because we control for the individual's cultural values.

Our analysis, relying on a collage of tests and empirical evidence, suggests that social interactions are important. We find that individuals belonging to highly entrepreneurial neighborhoods are more likely to become entrepreneurs. They invest more but have lower entrepreneurial profits than entrepreneurs residing in neighborhoods with lower levels of entrepreneurial activity. Strikingly, nonentrepreneurs and prospective entrepreneurs (individuals who become entrepreneurs during the sample period before entering in the entrepreneurial activity) have higher wages if they reside in highly entrepreneurial neighborhoods. This suggests that individuals belonging to highly entrepreneurial social groups have, if anything, better outside options than other individuals in the same LLM. Nevertheless, they are more likely to pursue entrepreneurial activity and earn lower profits than other entrepreneurs in the same LLM. Peer effects seem to matter because they increase entrepreneurial private benefits, rather than increase the productivity of entrepreneurs.

Our findings are robust to the use of different proxies for entrepreneurial activity, the introduction of additional controls, and the use of different subsamples. They are also confirmed by several robustness checks. First, by looking at the behavior of movers, we do not find any evidence that the correlation between individual and aggregate occupational choices is due to sorting of entrepreneurs in some municipalities. We also find no evidence that individuals move to highly entrepreneurial neighborhoods right before starting a business.

Second, to increase the confidence in our identification strategy, we check that our results hold for young individuals who cannot have affected the cultural values we use as instruments. Most importantly, in this subsample, we are able to control for the initial capital that may be informally available to potential entrepreneurs and the fact that children of entrepreneurs are more prone to start a business. Although the sample is dramatically reduced, our results remain qualitatively invariant.

Finally, we test whether social networks in high entrepreneurship municipalities ease access to credit. We find that small firms in high entrepreneurship municipalities do not rely on informal loans or financial debt to a larger extent than companies in other municipalities. This suggests that credit markets are unlikely to foster entry of entrepreneurs with lower expected profits. Hence, the mechanism relating social interactions and entrepreneurial activity appears to be nonpecuniary.

This paper contributes to two strands of literature. First, it helps understanding the private benefits from running or controlling a business. In particular, it suggests that the amenity potential from running a business pertains not only some specific industries, such as sports or media (Demsetz and Lehn, 1985), but also some social groups that attribute high value to the entrepreneurial activity. Second, our findings are consistent with recent papers showing that entrepreneurial activity may involve substantial nonpecuniary benefits (Hamilton, 2000 and Moskowitz and Vissing-Jorgensen, 2002).4 Our contribution is to suggest that peer effects help explaining these nonpecuniary benefits.

The rest of the paper is organized as follows. Section 2 describes how individual characteristics, social norms, and economic conditions are expected to affect entrepreneurial entry, profits, and investment. Section 3 describes these data and the empirical methodology. The results and the robustness checks are presented in Sections 4 and 5, respectively. Section 6 concludes.

2. Individual Decision to Become an Entrepreneur and Neighborhood Entrepreneurial Activity

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Individual Decision to Become an Entrepreneur and Neighborhood Entrepreneurial Activity
  5. 3. Data Description and Identification Strategy
  6. 4. Results
  7. 5. Robustness and Alternative Explanations
  8. 6. Conclusion
  9. Appendices
  10. References

An individual's decision to become an entrepreneur and her profits may be correlated with entrepreneurial activity in the individual's neighborhood for several reasons. First, individual characteristics may differ across neighborhoods. Individual characteristics, such as initial wealth, wages, labor market status, age, and education have been shown to be related to the decision to start a business (Evans and Leighton, 1989). For instance, by relaxing liquidity constraints, individual wealth is known to affect positively the decision to become an entrepreneur (Evans and Jovanovic, 1989). Thus, there may be more entry in the entrepreneurial activity in neighborhoods with wealthier residents. Similarly, individuals without jobs and lower wages have a lower opportunity cost of starting their own business (Evans and Leighton, 1989). Consequently, they may not only be more inclined to become entrepreneurs, but also run less profitable firms. To the extent that individual characteristics differ systematically across neighborhoods, a correlation between neighborhood entrepreneurial activity and individual decision or profits may arise.

Second, the neighborhood's economic environment affects incentives to start a business. It may affect entrepreneurial productivity by favoring access to infrastructure, intermediate goods, and skilled labor (Krugman, 1991) or by offering propitious institutions (Desai et al., 2005; Klapper et al., 2006). A favorable economic environment should then increase the overall level of entrepreneurial activity and entrepreneurial income. Furthermore, tax rates may differ across neighborhoods. The higher personal taxes are in comparison to corporate tax rates, the stronger the incentives to become entrepreneurs because business owners can more easily underreport their taxable income (Cullen and Gordon, 2002).

If the above factors are held constant—in the next section, we explain to what extent our empirical strategy accomplishes this—then a positive relation between an individual's decision to start a business and entrepreneurial activity in the neighborhood as well as an effect of the latter on entrepreneurial profits indicates that social interaction matters. In work contemporaneous to ours, Nanda and Sorensen (2007) show that an individual is more likely to become an entrepreneur if her coworkers—admittedly, a definition of peer group less crude than the one based on neighborhoods—have been entrepreneurs before. This evidence suggests that social interactions are important. There are, however, different mechanisms through which social interactions may affect occupational choice. These mechanisms have different implications for entrepreneurial income and other aspects of entrepreneurial activity, such as entrepreneurial investment and wages of prospective entrepreneurs.

First, highly entrepreneurial neighborhoods may favor social learning and knowledge spillovers (Saxenian, 1994). A possible mechanism allowing the sharing of knowledge is the mobility of technically sophisticated employees between firms in the neighborhood (Fallick et al., 2006). If this is the case, then not only entry into entrepreneurial activity but also entrepreneurial profits should be higher in highly entrepreneurial neighborhoods. Sorenson and Audia (2000), however, show that US firms in the shoe industry are more likely to fail if they operate in areas with high concentrations of similar firms. This suggests that the positive correlation between entry and existing level of entrepreneurial activity is not driven by new entrants' higher productivity.

Second, peer effects may be particularly important in shaping individual career aspirations and attitude toward entrepreneurship. There is mounting evidence that the average entrepreneur is not an innovator with a great idea, but simply somebody who tries to make a living and does not want to work for someone else (Shane, 2007). Peer effects may thus create nonpecuniary benefits, by attributing prestige to the entrepreneurial activity or generating a desire to conform. Also, the value of autonomy may differ across social groups.

If peer effects drive nonpecuniary benefits from entrepreneurial activity, the selection in the entrepreneurial activity is not only driven by the expected entrepreneurial profits, but also by the extent to which individuals value entrepreneurship. The latter in turn depends on the social norm prevailing in an individual's peer group. Individuals belonging to social groups that value the entrepreneurial activity highly prefer to be entrepreneurs, even if they expect to earn relatively lower profits or have relatively higher wages. Additionally, they may overinvest in the entrepreneurial activity if they derive utility from doing so. In Appendix A, we present a simple model illustrating the effect of social norms on entrepreneurial activity.

The importance of nonpecuniary benefits in explaining upsurges in entrepreneurial entry is consistent with the results of Barnett et al. (2003) showing that boom-time entrants have higher failure rates. This suggests that a decline in the average fitness of entrepreneurs accompanies increases in entry. Under the assumption that the average fitness for survival does not change over time, these results suggest that peer effects cause a decline in the selection threshold for entry.

In what follows, we provide more direct evidence by exploring the effects on entry, profits, and investment.5 The two main hypotheses we test are the following:

  • Hypothesis 1: Entry into entrepreneurship is higher for individuals in neighborhoods with a high level of entrepreneurial activity.

  • Hypothesis 2: If peer effects increase non-pecuniary benefits from the entrepreneurial activity, entrepreneurial profits are likely to be lower in those neighborhoods with a higher level of entrepreneurship.

3. Data Description and Identification Strategy

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Individual Decision to Become an Entrepreneur and Neighborhood Entrepreneurial Activity
  5. 3. Data Description and Identification Strategy
  6. 4. Results
  7. 5. Robustness and Alternative Explanations
  8. 6. Conclusion
  9. Appendices
  10. References

3.1 Data Sources

Our main data source is Linda, a register-based longitudinal data set for Sweden, providing information about household organization, labor status, sources of income, wealth, housing, and other socioeconomic characteristics. We match the individual data from Linda with information about the 109 Swedish LLMs—which in turn include 289 Swedish municipalities6—provided by Statistics Sweden. In addition, we use Market Manager, a data set containing the financial statements of all (private and public) companies incorporated in Sweden to obtain information about firm capital structure.

Linda is a representative sample including some 300,000 households, or approximately 4% of the Swedish population. In contrast, the US National Longitudinal Survey, used by Evans and Jovanovic (1989) and Evans and Leighton (1989), includes only 5,225 individuals. Analogously, the Survey of Consumer Finances used by Moskowitz and Vissing-Jorgensen (2002) includes only 4,000 US households. Even if these data could be merged with information about households' neighborhoods, it would be difficult to draw any conclusions about how neighborhood characteristics influence individual occupational choice, because the data for individuals belonging to the same neighborhood is limited. This is not the case with Linda. Our sample includes on average 5% of the population (1,584 individuals) for all Sweden's 289 municipalities and never less than 3.9% of the population (106 individuals).

Finally, starting from 1995, Linda provides detailed information on whether an individual reports to the tax authority any capital income she has received from a company she controls and for which she works at least part-time. This enables us to define entrepreneurial activity using individual tax returns, as did Holtz-Eakin et al. (1994). For this reason, we limit our sample to the years 1995–2000.

3.2 Entrepreneurs, Neighborhoods, and Measures of Entrepreneurial Activity

Our definition of entrepreneur includes all individuals who receive salaries or capital income from a business that they control and in which they work at least part-time. Most individuals we define as entrepreneurs run nonincorporated companies. Similarly to Holtz-Eakin et al. (1994), we also consider individuals who run their own business as a second job to be entrepreneurs. We include these individuals because most businesses, even the successful ones, are generally started with small investments as part-time jobs (Shane, 2007). In the empirical analysis, we explore whether our results hold if we restrict the sample to full-time entrepreneurs.

We include individuals between ages 18 and 60, because individuals who are too young or too old are unlikely to set up an entrepreneurial activity. Additionally, we exclude individuals involved in agriculture, farming, and forestry, which are concentrated in rural areas and could bias our results toward finding a correlation between individual and aggregate occupational choices. According to our definition, approximately 5% of the population is involved in entrepreneurial activities, slightly less than in previous studies (Blanchflower, 2000; Blanchflower et al., 2001), which also include farmers.

We focus on the decision to become self-employed. In comparison to the decision to be self-employed, it is less likely to be determined by the same, possibly omitted, factors of the level of entrepreneurial activity. For this reason, we look at individuals who, according to our definition, can be classified as entrepreneurs in year t but not in year t−1. They represent approximately 1% of the working-age population each year.7

To analyze peer effects, we need to define entrepreneurial activity within an individual's social group. Similar to most of the existing literature, we assume that an individual's peer group is defined by administrative boundaries (Bertrand et al., 2000). This assumption is supported by the empirical evidence. For instance, Wellman (1996) shows that about 38% of Toronto inhabitants' yearly contacts take place between individuals who live less than 1 mile apart. Similarly, Connerly (1985) finds that 41% of the respondents to a Toronto inhabitants' survey had at least one-third of their friends residing within 1 mile. Because in Sweden organizations, such as churches, schools, daycare centers, local sports, cultural associations, and the administration of apartment buildings, foster social interactions among individuals who reside within the municipality, the smallest and most relevant unit to define neighbor effects is the municipality.

We use several proxies for entrepreneurial activity in the municipality: First, the fraction of entrepreneurs relative to all individuals under the age of 60, included in Linda, in a given municipality. The second definition takes into account that interaction may be closer between individuals with similar educational achievement. Hence, for any individual, we construct a peer group of individuals in the municipality with a similar education level. We rely on three educational groups: individual with less than a high school diploma, those with a high school diploma, and those with a university degree. We define the variable of interest as the proportion of entrepreneurs with a given educational achievement in the municipality's population with comparable educational achievement. Because individuals with different levels of education may have different propensities to become entrepreneurs, we include education dummies.

Finally, we use a measure based on economic outcome, namely the proportion of entrepreneurs in the top quartile of the income distribution in each municipality. This proxy captures the fact that in municipalities where the richest individuals are entrepreneurs, this profession may be considered highly prestigious. Hence, this variable may be a more direct proxy for the existence of role models.

We use the proxy for entrepreneurship at t – 1 to explain the probability of an individual becoming an entrepreneur at time t. For this reason, we lose 1 year. Our final sample consists of 469,504 individuals with a total of 1,684,596 individual-year observations from 1996 to 2000.

Table I shows that there is substantial variation across municipalities both in the proportion of individuals who are entrepreneurs and in the proportion of individuals who become entrepreneurs. In Table I we combine cross-sectional and time-series variation. However, over time the variance between municipalities is nearly 10 times larger than the variance within municipalities. This suggests that some municipalities indeed have persistently higher rates of entrepreneurship. It is consistent with Swedish empirical studies showing that regional levels of entrepreneurship are related to highly persistent cultural values (Davidsson, 1995; Davidsson and Wiklund, 1997).8

Table I.  Differences in Entrepreneurial Activity across Municipalities
 MeanMedianMinimumMaximumStandard DeviationInterquartile Range
  1. Notes: The variable ENTREPRENEURSHIP is the ratio of entrepreneurs relative to the population of the municipality; ENTRY RATE is the ratio of new entrepreneurs relative to the number of entrepreneurs in the municipality; EXIT RATE is the ratio of entrepreneurs abandoning entrepreneurial activity relative to the number of entrepreneurs in the municipality. The subsequent three rows present the ratio of the entrepreneurs with a given education level relative to the municipality's population with the same educational achievement. The education level is indicated in the first column. PROPORTION OF ENTREPRENEURS IN THE TOP QUARTILE is the proportion of entrepreneurs in the top quartile of the municipality's income distribution. ENTREPRENEURIAL INCOME (INVESTMENT) is the average income (investment) per entrepreneur in the municipality. The statistics employ observations for all municipalities from 1996 to 2000 (288 from 1996 to 1998, 289 from 1999 to 2000). All individuals between the ages of 18 and 60 who have entrepreneurial income are classified as entrepreneurs. The population includes 469,504 individuals between 18 and 60. Farmers have been excluded from the sample.

ENTREPRENEURSHIP0.0570.0520.0150.1830.0230.029
ENTRY RATE0.2410.2390.0000.7500.1450.124
EXIT RATE0.2240.2040.0000.6000.1440.149
ENTREPRENEURSHIP BY EDUCATION LEVEL
 <HIGH SCHOOL DIPLOMA0.0580.0500.0000.2420.0370.039
 HIGH SCHOOL DIPLOMA0.0520.0470.0000.2010.0260.031
 UNIVERSITY EDUCATION0.0770.0640.0000.6210.0630.055
 PROPORTION OF ENTREPRENEURS IN THE TOP QUARTILE0.0530.0500.0000.1880.0270.030
 ENTREPRENEURIAL INCOME (SEK 000)103.8199.4314.48444.3529.8124.72
 ENTREPRENEURIAL INVESTMENT (SEK 000)1.460.650.0037.502.731.63
INSTRUMENTS
 PROPORTION OF INDIVIDUALS OLDER THAN 60 & MEMBER OF THE STATE CHURCH0.9360.9470.7281.0000.0400.045
 PROPORTION OF VOTES FOR RIGHT-WING PARTIES IN 1982 ELECTIONS0.4600.4610.1510.7470.1140.146

Table I shows that there is also large variation in our proxy for entrepreneurial profits (income). Note that the entrepreneurial income we observe is always positive as eventual losses cannot be detracted from personal taxable income. This may lead us to overestimate the profits from entrepreneurial activity, which will be evident later, makes our results even more striking.

3.3 Identification

Interpreting the effect of alternative proxies for entrepreneurial activity on the probability of an individual becoming an entrepreneur, entrepreneurial income, and investment is problematic. Entrepreneurial activity could be correlated with unobservable individual or neighborhood characteristics that have an independent impact on the dependent variable. Luckily, our data set presents features that allow the use of methodologies similar to the ones successfully used in different contexts. In what follows, we discuss how we overcome identification problems arising from the simultaneity of decisions, unobserved heterogeneity in the economic environment, and unobserved heterogeneity between neighborhoods.

3.3.1 Simultaneity of Decisions

When we study an individual's decision to become an entrepreneur, we may incur in the reflection problem (Manski, 1993): The econometrician may be unable to identify whether individuals simultaneously respond to a shock or if some of them are influenced by the others, because individual and aggregate occupational choices are simultaneously determined.

We can overcome the reflection problem thanks to the time-series variation in our data set. It is reasonable to postulate that, if social interactions matter, the current individual decision to become an entrepreneur depends on past occupational choices in the neighborhood. Because the decision to become an entrepreneur depends not only on peer effects, but also on other factors, we can identify exogenous variables (i.e., past and current shocks to entry and exit) that affect the individual decision but not the predetermined level of entrepreneurial activity and vice versa.9 As Moffitt (2001) argues, this allows us to achieve identification.

3.3.2 Unobserved Heterogeneity in the Economic Environment

Statistics Sweden. constructs LLM areas according to observed commuting patterns. Within a LLM, individuals face the same work opportunities and markets for goods and services because they can easily commute. Thus, the economic environment can be considered homogeneous within the LLM (Vlachos, 2004).

Figure 1 shows the 109 LLMs and the municipalities within each LLM. To put things in perspective, Sweden has a population of nearly 9 million and comprises 109 LLMs. The average (median) population of a LLM is 81,200 (26,700). The average (median) area is 3,770 (2,318) sq km. The most densely populated LLM is Stockholm with 1,862,000 inhabitants in an area of 8,036 sq km. Stockholm includes 30 municipalities, Göteborg includes 16, while 61 LLMs—the less populated ones—include only one municipality.

image

Figure 1. LOCAL LABOR MARKETS AND MUNICIPALITIES

Download figure to PowerPoint

Following Bertrand et al. (2000) who examine peer effects in welfare participation, we include LLM fixed effects to control for economic factors. The LLM fixed effects should capture differences in economic incentives to undertake the entrepreneurial activity, such as differences in entry costs, financial development, market structure, or labor market conditions.

3.3.3 Unobserved Heterogeneity Between Neighborhoods

Assuming that LLM fixed effects capture all of the heterogeneity in the economic environment, it is still difficult to prove that peer effects have a significant effect on profits or on the decision to become an entrepreneur. Individuals in different neighborhoods could differ in some unobserved characteristics, such as having diverse work experience, that affect the propensity to start a business. A further concern is that LLM fixed effects may not fully account for differences in the economic environment, for instance because of differences in municipality income tax rates.

To account for unobserved neighborhood heterogeneity, we include extensive municipality and individual level controls that we describe in Appendix B and dummy variables for the municipalities with highest and lowest wealth per capita within the LLM. We recognize that our controls may not fully capture municipality heterogeneity. We attempt to further mitigate concerns that unobserved heterogeneity across municipalities drives our results by demonstrating that the estimates are robust to econometric approaches that focus on cross-sectional variation as well as those that rely exclusively on time-series heterogeneity.

First, we focus on time-series variation and include municipality fixed effects. These capture any time-invariant differences in economic environment unaccounted by the LLM fixed effects.10 Second, we instrument our proxies for entrepreneurial activity. We look for instruments that generate exogenous between-group variation, similar to Case and Katz (1991), Cutler and Glaeser (1997), and Duflo and Saez (2002), and neglect any changes in entrepreneurial activity over time (which may depend on some municipalities becoming more favorable than others to the entrepreneurial activity). Our instruments are the proportion of pensioners who are members of the state church11 and the proportion of individuals who voted for right-wing parties in the early 1980s. Because municipality culture, and therefore religious beliefs and political orientation, could be affected by entrepreneurial activity, we use predetermined proxies for cultural values. Even if we use current pensioners' religious beliefs, these are most often lifetime beliefs, and can be considered predetermined.

Our instruments are unlikely to be jointly determined with the current level of economic activity (for instance, because they are both related to entrepreneurial activity in the early 1980s). During the 1980s and early 1990s, Sweden underwent profound economic transformation. It experienced a banking crisis followed by widespread bankruptcies, a major tax reform in 1990–1991, and the dissolution of the centralized wage-setting arrangements, which significantly modified the Swedish industrial structure (Davis and Henrekson, 2005). Much of the entrepreneurial activity we observe in our sample was initiated because of this process. Cultural values in the early 1980s can thus be considered exogenous with respect to the current level of entrepreneurial activity.

Our conjecture is that our instruments can explain the current level of entrepreneurial activity. First, in Sweden, left-wing parties have generally favored the expansion of the public sector and large established companies (Hogfeldt, 2004). A high fraction of votes for right-wing parties in the early 1980s may be related to an aversion toward large companies and the public sector. This may have affected the attitude toward entrepreneurial activity once institutions became more favorable to entrepreneurs. Second, as Weber (1905) first argued, religious beliefs are associated with different economic attitudes. More recently, Barro and McCleary (2003) and Guiso et al. (2003) find that religion is positively associated with economic performance and attitudes that are conducive to market-oriented institutions. Religion may thus create a positive environment for entrepreneurial activity.

In Table III, the first-stage regression (which uses the proportion of entrepreneurs in the population as the dependent variable) confirms that our instruments have high explanatory power for entrepreneurial activity12 and enter with the expected sign in the first stage.

Table III.  The Effects of Cultural Values on Municipality Characteristics
 Proportion of Entrepreneurs (1)Education Level (2)Log(Total Wealth) (3)Log(Labor Income) (4)
  1. Notes: The dependent variables are the proportion of entrepreneurs, the average educational attainment, the average per capita wealth, and the average per capita income in the municipality. All explanatory variables are defined in Tables I and II. All the equations include 4-year dummies and 108 LLM fixed effects. The standard errors are corrected for heteroskedasticity and take into account that observations for the same municipality may be correlated. T-statistics are reported in parentheses. We also report the adjusted R2 and the Joint F-test of the instruments (the proportion of individuals older than 60 who are members of the state church and the proportion of votes for right-wing parties in the 1982 elections). We use 1,442 observations.

DIVERSITY0.0192(2.25)0.3895(2.96)0.0436(0.56)0.0519(3.23)
COMPETITION0.0015(1.11)0.0227(1.25)0.0076(0.72)−0.0034(−1.53)
UNEMPLOYMENT RATE−0.0026(−5.15)−0.0059(−0.78)0.0009(0.26)0.0027(2.94)
PROPORTION OF UNEMPLOYED IN ENTREPRENEURSHIP PROGRAMS0.0117(2.67)−0.0792(−1.21)−0.0426(−1.06)0.0008(0.10)
SHARE OF EMPLOYED IN FIN SECTOR0.1443(2.85)−1.2649(−1.59)−0.2373(−0.41)0.1866(1.92)
PROPORTION OF PUBLIC EMPLOYEES−0.1200(−1.56)−2.1079(−1.77)−0.5800(−0.96)−0.1910(−1.31)
INCOME PER CAPITA0.0855(12.75)−0.1247(−1.20)1.5396(24.71)0.0756(5.96)
WEALTH TAX PER CAPITA−0.3024(−18.73)−3.0133(−12.54)−0.2122(−1.54)0.5993(20.41)
PROPORTION OF IMMIGRANTS−0.0034(−0.21)0.4900(1.92)0.3338(1.88)0.0923(2.95)
ENTRY RATE−0.0657(−12.79)0.0871(1.14)−0.0005(−0.01)−0.0070(−0.74)
EXIT RATE0.0047(0.69)0.2427(2.44)−0.0303(−0.46)−0.0029(−0.24)
WEALTHIEST MUNICIPALITY IN LLM−0.0091(−6.53)−0.1409(−6.70)0.0169(1.59)−0.0024(−0.93)
POOREST MUNICIPALITY IN LLM−0.0004(−0.31)0.0363(1.75)−0.0047(−0.43)−0.0027(−1.06)
PROPORTION OF INDIVIDUALS OLDER THAN 60 & MEMBER OF THE STATE CHURCH0.0014(18.38)−0.0009(−0.80)0.0004(0.68)−0.0002(−1.06)
PROPORTION OF VOTES FOR RIGHT-WING PARTIES IN THE 1982 ELECTIONS0.0399(2.53)0.0170(0.07)−0.1635(−1.30)−0.0552(−1.90)
LLM fixed effectsYes Yes Yes Yes 
Adjusted R20.718 0.359 0.688 0.789 
 Statisticp-ValueStatisticp-ValueStatisticp-ValueStatisticp-Value
  
Joint F-test of instruments337.160.0000.630.4261.010.3662.370.094

These instruments are valid only if they are unrelated to omitted factors that directly affect the individual decision to become an entrepreneur (e.g., having a business degree). We attempt to mitigate concerns regarding the validity of the instruments as follows. First, we roughly control for the corresponding individual cultural traits. We do observe whether an individual is part of the state church, and therefore we can control for the fact that individual religious beliefs can directly affect the choice to become an entrepreneur. We also control for individual income and wealth, which are correlated with the decision to vote for right-wing parties, and for factors affecting labor demand, which could be affected by a rightist local administration, such as the proportion of individuals employed in the public sector and the rate of unemployment.

Second, we perform tests of overidentifying restrictions. The Hansen's J statistics never allow us to reject the null that our instruments do not have a direct effect on the dependent variable. Third, and perhaps most importantly, our instruments do not appear to be related to observable measures of social group economic outcomes other than the current level of entrepreneurial activity. In Table III, we find that our instruments are unrelated to municipality income per capita, individual wealth, and average years of schooling. To the extent that our instruments are not related to observable municipality characteristics, they are unlikely to capture unobservable characteristics that affect the decision to become an entrepreneur.

4. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Individual Decision to Become an Entrepreneur and Neighborhood Entrepreneurial Activity
  5. 3. Data Description and Identification Strategy
  6. 4. Results
  7. 5. Robustness and Alternative Explanations
  8. 6. Conclusion
  9. Appendices
  10. References

4.1 The Decision to Become an Entrepreneur

We estimate the probability that an individual becomes an entrepreneur using a probit model, a linear probability model, and two-stage least squares (2SLS).13 In most specifications, we include 108 LLM fixed effects, but we also explore the robustness of our results to the inclusion of municipality fixed effects. When using the probit model, which notoriously creates problems with fixed effects, we include only seven regional dummies.

Table IV presents the parameter estimates. Results show that individual and aggregate occupational choices are correlated. Entrepreneurship in the municipality has a positive and significant effect on the probability of an individual becoming an entrepreneur. The marginal effect of the proxy for entrepreneurial activity is similar in the probit model (Model 1)—when LLM fixed effects are not included—and in the linear probability model (Model 2). Neither the inclusion of municipality fixed effects (Model 3) nor the instrumentation strategy (Model 4)14 yield sizable changes in the magnitude of the coefficient of our variable of interest. It thus appears irrelevant whether we exploit the cross-sectional variation of these data—as we do in the 2SLS estimates, given that our instrument are basically time-invariant15—or the limited time-series variation of the entrepreneurship proxy, as we do, when we use municipality fixed effects.

Table IV.  The Decision to Become an Entrepreneur
 Without Education GroupsEnt. in Top QuartileWith Education Groups
Probit (1)OLS (2)OLS/ Municipality FE (3)2SLS (4)2SLS (5)2SLS (6)2SLS (7)
  1. Notes: The dependent variable is a dummy variable that takes value one if individual i becomes an entrepreneur at time t and zero otherwise. Individuals who were already entrepreneurs at time t−1 are excluded. In the specification “without education groups”, ENTREPRENEURSHIP is defined as the proportion of individuals who are entrepreneurs in the municipality; in the specification “with education groups”, ENTREPRENEURSHIP is the proportion of entrepreneurs among the individuals with a given education level in the municipality. ENTREPRENEURSHIP IN THE TOP QUARTILE is the proportion of entrepreneurs among individuals in the top quartile of the income distribution of the municipality. ENTREPRENEURSHIP OUTSIDE THE SOCIAL GROUP is the proportion of entrepreneurs among individuals with different education level residing in the same municipality. All other explanatory variables are defined in Tables I and II. All the equations include 4-year dummies, two education dummies for individuals with high school and university degrees, and 11 dummies that refer to the sector in which an individual is employed. The equation is estimated using a probit and a linear probability model. The latter is estimated by ordinary least squares (OLS) or two-stage least squares (2SLS). In the linear probability models, we have included 108 LLM fixed effects, while in the probit Model 7 regional fixed effects. In Model 3, we also include 289 municipality fixed effects. In the 2SLS estimates, the instruments are the proportion of individuals older than 60 who are members of the state church and the proportion of votes for right-wing parties in the 1982 elections. The standard errors are corrected for heteroskedasticity and take into account that observations for the same municipality may be correlated. T-statistics are reported in parentheses. For the probit estimates, we report marginal effects instead of parameter estimates; the marginal effects have been calculated setting the independent variables equal to the mean. We also report the adjusted R2 (Pseudo R2 for probit estimates). For the 2SLS estimates, we report the Bound–Jaeger–Baker's (1995) test for the quality of instruments and the Hansen's J statistics for overidentifying restrictions. Estimates are multiplied by 100. In the specification where the proxy for entrepreneurial activity is entrepreneurship in the top quartile, we include only observations relative to individuals in the three lowest quartiles of the income distribution.

ENTREPRENEURSHIP14.22(26.22)15.09(17.97)14.93(5.54)11.90(7.53)26.71(3.62)12.03(7.57)4.43(5.49)
ENTREPRENEURSHIP OUTSIDE THE SOCIAL GROUP            0.03(0.46)
Individual Level Controls
 INCOME OF OTHER MEMBERS−0.06(−10.39)−0.06(−10.96)−0.05(−7.24)−0.06(−10.97)−0.06(−10.95)−0.06(−10.50)−0.03(−4.42)
 LABOR INCOME−0.08(−6.91)−0.09(−6.84)0.06(5.65)−0.09(−6.83)−0.09(−6.83)−0.10(−7.00)−0.17(−6.58)
 WEALTH0.06(12.50)−0.43(−10.06)0.03(4.00)−0.43(−10.07)−0.43(−10.00)−0.43(−10.27)0.03(3.85)
 WEALTH20.00(−5.87)0.09(13.06)0.00(0.32)0.09(13.08)0.09(12.97)0.09(13.21)0.00(0.30)
 SHARE OF LIQUID ASSETS0.01(1.19)0.05(7.71)0.00(0.32)0.05(7.69)0.05(7.42)0.05(7.72)−0.01(−0.76)
 AGE0.20(22.37)0.17(21.60)0.14(17.24)0.17(21.61)0.17(21.66)0.16(20.91)0.13(17.52)
 AGE2/1000.00(−19.99)−0.20(−19.21)0.00(−15.48)−0.20(−19.21)−0.20(−19.25)−0.20(−18.52)−0.15(−15.44)
 MALE0.87(39.54)0.87(37.08)0.74(29.98)0.87(37.11)0.87(37.25)0.87(36.97)0.68(25.01)
 MOVER0.01(0.12)0.03(0.49)−0.01(−0.22)0.03(0.48)0.03(0.46)0.02(0.42)0.00(−0.03)
 CHANGES IN FAMILY STRUCTURE0.05(1.65)0.06(1.95)0.01(0.36)0.06(1.96)0.06(1.96)0.06(1.94)0.01(0.30)
 NUMBER OF CHILDREN0.05(4.35)0.10(7.96)0.06(4.69)0.10(7.98)0.10(8.01)0.10(8.19)0.06(4.07)
 MARRIED0.25(8.70)0.24(7.92)0.23(7.87)0.24(7.91)0.24(8.10)0.24(7.97)0.22(5.34)
 DIVORCED0.05(1.06)0.07(1.82)0.09(1.96)0.07(1.81)0.08(1.95)0.08(2.05)0.13(2.84)
 UNEMPLOYED−0.01(−0.16)0.01(0.14)0.06(0.77)0.01(0.07)−0.01(−0.07)0.00(0.02)0.17(2.56)
 IMMIGRANT−0.16(−1.99)−0.08(−1.07)0.04(0.43)−0.08(−1.02)−0.05(−0.58)−0.10(−1.30)0.15(1.35)
 WAGE PREMIUM−0.20(−11.07)−0.22(−9.23)−0.35(−10.73)−0.22(−9.24)−0.22(−9.21)−0.23(−9.25)−0.43(−14.80)
 CHURCH−0.05(−1.55)−0.09(−2.91)−0.06(−1.71)−0.09(−2.86)−0.09(−2.66)−0.10(−3.02)−0.09(−3.00)
Municipality Level Controls
 DIVERSITY−0.05(−0.23)0.27(1.55)0.03(0.93)0.19(1.00)0.56(1.43)0.20(1.06)−0.04(−0.15)
 COMPETITION0.08(−2.77)0.03(0.89)1.02(1.06)0.04(1.42)0.06(1.11)0.04(1.43)0.11(3.02)
 UNEMPLOYMENT RATE−0.01(−1.47)0.00(0.00)0.00(0.12)−0.02(−1.28)−0.02(−0.79)−0.02(−1.47)−0.02(−1.22)
 PROPORTION OF UNEMPLOYED IN ENT. PROG.0.08(0.63)0.09(0.88)0.02(0.17)0.13(1.27)0.14(0.81)0.12(1.17)−0.02(−0.19)
 SHARE OF EMPLOYED IN FINANCIAL SECTOR1.09(1.11)1.45(1.91)2.39(0.39)2.03(2.22)−1.78(−0.86)1.87(2.05)0.30(0.32)
 PROPORTION OF PUBLIC EMPLOYEES0.67(0.56)0.37(0.30)−1.69(−0.79)−0.24(−0.19)−2.55(−1.36)−0.33(−0.27)1.39(0.71)
 INCOME PER CAPITA0.00(0.79)−0.20(−1.26)−0.43(−0.97)0.15(0.73)−0.19(−0.35)0.15(0.72)−0.18(−0.59)
 WEALTH TAX PER CAPITA0.00(0.58)0.18(0.49)−0.64(−0.39)−0.71(−1.37)−2.81(−3.59)−0.85(−1.68)0.22(0.25)
 UNEMPLOYMENT RATE * UNEMPLOYED−0.02(−2.40)−0.02(−2.63)−0.02(−2.54)−0.02(−2.56)−0.02(−2.33)−0.02(−2.49)−0.02(−3.08)
 PROPORTION OF IMMIGRANTS * IMMIGRANT−1.21(−1.25)−1.52(−1.78)−1.88(−1.43)−1.56(−1.81)−2.11(−2.17)−1.29(−1.60)−2.58(−2.13)
 ENTRY RATE5.66(31.57)5.68(20.75)5.94(16.52)5.44(19.84)5.56(13.43)5.43(19.91)5.61(18.88)
 EXIT RATE0.13(0.61)−0.20(−1.24)−0.41(−1.50)−0.25(−1.57)−2.18(−3.80)−0.27(−1.66)−0.35(−1.41)
LLM fixed effectsNo Yes No Yes Yes Yes Yes 
Municipality fixed effectsNo No Yes No No No No 
N1,493,927 1,493,927 1,493,927 1,493,927 1,120,445 1,493,927 1,493,927 
Adjusted R20.040 0.007 0.008 0.007 0.006 0.007 0.007 
Log-likelihood−100,585             
       Statisticp-ValueStatisticp-ValueStatisticp-ValueStatisticp-Value
        
Test of overidentifying restrictions      0.160.692.740.100.530.470.000.98
Bound–Jaeger–Baker F-test      28.800.003.490.0229.540.0022.070.00
Test of ENTREPRENEURSHIP = ENTREPRENEURSHIP OUTSIDE THE SOCIAL GROUP            31.350.00

In the remaining specifications reported in Table IV, we exploit our alternative proxies for entrepreneurial activity in the municipality. First, we check whether, for individuals in the three lowest quartiles of the income distribution, the decision to become an entrepreneur is affected by the proportion of entrepreneurs in the highest quartile (Model 5). We find that this is indeed the case. The coefficient of our variable of interest doubles with respect to the earlier specifications. Because the occupational choices of the most successful individuals should have different determinants from the occupational choices of the rest of the population, our estimates are less likely to be affected by omitted variable bias in this specification. Thus, this result is comforting.

Finally, because it seems reasonable to assume that an individual interacts more closely with individuals residing in the same neighborhood and of similar levels of education, we define an individual's peer group as individuals with similar educational achievement in the municipality (Model 6). Once again, our estimates appear unaffected.

Peer effects also appear significant from an economic point of view. A one standard-deviation increase in the proportion of entrepreneurs in the municipality increases the probability of an individual becoming an entrepreneur by approximately 0.27 percentage points in Model 6, approximately two-thirds of the effect of a one-standard deviation change in the individual's wealth, calculated at the mean.16 The economic magnitude is larger when we use the proportion of entrepreneurs in the top quartile of the income distribution. In this case, a one-standard deviation increase in the proxy for entrepreneurial activity increases the probability that an individual becomes an entrepreneur by 0.72 percentage points.17 Our specification tests for the quality of instruments suggest that the more conservative estimates based on the proportion of entrepreneurs are more reliable. In fact, when we use the proportion of entrepreneurs in the top quartile of the income distribution, our instruments are weak according to the Bound–Jaeger–Baker test (Bound et al., 1995). The parameter estimate is one fifth of the reported coefficient if we estimate Model 5 by ordinary least squares (OLS).

Because we control for LLM (or municipality) fixed effects, the positive correlation between individual and aggregate occupational choice is unlikely to depend on differences in entry costs, labor market conditions, or other omitted economic factors. We can interpret our estimates of the coefficient of the variable measuring the municipality's entrepreneurial activity as evidence of peer effects if we believe that our instruments are not correlated with omitted individual characteristics and municipality time-varying factors that could have an independent effect on occupational choice. Because we control for a long list of individual and municipality characteristics, this condition is likely to be satisfied. We test this hypothesis formally using Hansen's J-test of overidentifying restrictions. We can never reject the null hypothesis that the instruments do not have a direct impact on the choice of becoming an entrepreneur with approximately a 10% confidence level. The probability that the null hypothesis is true is even higher in our favored specifications in which the instruments are strong according to the Bound–Jaeger–Baker test.

Our identification assumptions are also supported by a more direct test of the mechanism of social interactions. In Model 7, besides our proxy for entrepreneurship among individuals with similar educational achievement in the municipality, we include the level of entrepreneurship among individuals with different educational achievement than the individual under consideration. This variable is unlikely to capture social interactions; a positive and significant coefficient would suggest that our estimates are biased by omitted factors. The estimates support our interpretation that social interactions matter. Not only is the level of entrepreneurship in the individual's social group always positive and significant,18 but this variable also has a larger effect than the level of entrepreneurship among individuals with different educational achievement (which is not statistically significant).19

4.2 Entrepreneurial Profits

Our goal is to evaluate whether entrepreneurial profits are higher in municipalities with higher levels of entrepreneurial activity. In order to do so, we have to consider that individuals are more likely to start entrepreneurial activity in municipalities with higher levels of entrepreneurship. A lower threshold for an individual's selection in entrepreneurial activity implies that individuals with relatively lower skills become entrepreneurs. Entrepreneurs may thus earn lower profits because they have, on average, lower (unobservable) skills in municipalities with high entrepreneurship. We use the Heckman selection model to evaluate the direct effect of the municipality's entrepreneurial activity on profits. First, we estimate the probability of individual i being an entrepreneur, using a specification similar to the one in Table IV. We use these estimates to compute the Mills' ratio. Then we include the inverse Mills' ratio in the equation for individual profits with our main variable of interest and the control variables.20 We estimate the profits equation both by OLS and 2SLS.

The results are presented in Table V. The selection equation holds no surprises. The probability of an individual being an entrepreneur has largely the same determinants as the probability of an individual becoming an entrepreneur. The only difference is that this time we also include a dummy that takes value 1 if the individual was also an entrepreneur during the previous year

Table V.  Profits from Entrepreneurial Activity
 Selection Equation (1)Without Education GroupsEntrepreneurship in Top QuartileWith Education Groups
OLS (2)2SLS (3)OLS (4)2SLS (5)2SLS (6)2SLS (7)
  1. Notes: In the selection equation, the dependent variable is a dummy variable that takes value 1 if an individual is an entrepreneur at time t and zero otherwise. In the second stage, the dependent variable is the logarithm of the income from the entrepreneurial activity plus 0.01. In the specification “without education groups”, ENTREPRENEURSHIP is defined as the proportion of individuals who are entrepreneurs in the municipality; in the specification “with education groups”, ENTREPRENEURSHIP is the proportion of entrepreneurs among the individuals with a given education level in the municipality. ENTREPRENEURSHIP IN THE TOP QUARTILE is the proportion of entrepreneurs among individuals in the top quartile of the income distribution of the municipality. ENTREPRENEURSHIP OUTSIDE THE SOCIAL GROUP is the proportion of entrepreneurs among individuals with different education levels residing in the same municipality. LAMBDA is the inverse Mills' ratio. All other explanatory variables are defined in Tables I and II. All equations include 4-year dummies, two education dummies for individuals with high school and university degrees and 11 dummies that refer to the sector in which an individual is employed. In the selection equation we also include seven regional dummies. In the second stage we include 108 LLM fixed effects. The selection equation has been estimated using a probit model. The second-stage equation has been estimated using OLS or 2SLS. In the 2SLS estimates, the instruments are the proportion of individuals older than 60 who are members of the state church and the proportion of votes for right-wing parties in the 1982 elections. The standard errors are corrected for heteroskedasticity and take into account that observations for the same municipality may be correlated. T-statistics are reported in parentheses. We also report adjusted R2 (Pseudo R2 for the selection equation). For the 2SLS estimates we report the Bound–Jaeger–Baker's (1995) test for the quality of instruments and the Hansen's J statistics for overidentifying restrictions. In the specification where the proxy for entrepreneurial activity is entrepreneurship in the top quartile, we include only observations relative to individuals in the three lowest quartiles of the income distribution of the municipality.

ENTREPRENEURSHIP2.301(23.82)2.041(4.72)2.757(3.01)1.826(5.84)6.987(2.63)2.197(2.22)5.114(2.47)
ENTREPRENEURSHIP OUTSIDE THE SOCIAL GROUP            −0.445(−1.50)
Individual Level Controls
LABOR INCOME−0.115(−67.61)            
INCOME OF OTHER MEMBERS−0.014(−14.18)            
EXPERIENCE  −0.023(−2.25)−0.023(−2.26)−0.022(−2.21)−0.021(−2.14)−0.023(−2.15)−0.071(−4.71)
PART-TIME  −0.508(−6.79)−0.372(−6.57)−0.491(−1.99)0.082(0.53)−0.293(−2.17)0.814(1.21)
WEALTH−0.025(−3.40)−0.034(−2.43)−0.031(−2.12)−0.034(−2.48)−0.019(−1.05)−0.025(−1.57)−0.048(−2.86)
WEALTH20.012(9.12)0.003(1.16)0.002(0.64)0.002(1.09)−0.005(−1.01)−0.002(−0.46)0.001(3.01)
SHARE OF LIQUID ASSETS0.016(15.16)−0.011(−4.19)−0.013(−3.88)−0.012(−4.33)−0.022(−3.35)−0.017(−3.75)−0.020(−2.29)
AGE0.048(40.24)−0.010(−2.08)−0.014(−1.95)−0.011(−2.29)−0.040(−2.29)−0.026(−2.31)−0.088(−2.23)
AGE20.000(−35.94)0.000(1.70)0.000(1.74)0.000(1.88)0.000(2.22)0.000(2.18)0.001(2.28)
MALE0.136(55.46)0.039(3.61)0.029(1.77)0.036(3.35)−0.035(−0.85)−0.004(−0.14)−0.215(−1.81)
MOVER−0.254(−27.75)−0.173(−8.70)−0.153(−4.26)−0.165(−8.33)−0.014(−0.16)−0.084(−1.43)0.024(0.68)
CHANGES IN FAMILY STRUCTURE−0.024(−7.82)−0.026(−2.21)−0.024(−1.95)−0.026(−2.18)−0.009(−0.66)−0.016(−1.22)0.003(0.23)
NUMBER OF CHILDREN0.008(5.20)0.013(2.62)0.013(2.51)0.012(2.48)0.009(1.58)0.010(1.96)0.022(3.42)
MARRIED0.047(11.47)0.020(1.35)0.017(1.16)0.019(1.25)−0.003(−0.17)0.008(0.47)−0.040(−1.30)
DIVORCED0.004(0.74)0.147(7.11)0.145(6.93)0.146(7.08)0.136(6.17)0.141(6.66)0.111(3.26)
UNEMPLOYED−0.168(−16.05)−0.079(−1.92)−0.068(−1.48)−0.071(−1.72)0.029(0.39)−0.018(−0.32)0.325(1.50)
IMMIGRANT−0.096(−9.58)0.168(5.12)0.175(5.09)0.161(4.86)0.201(5.15)0.194(5.14)0.279(2.91)
WAGE PREMIUM0.073(29.71)0.048(7.77)0.043(4.13)0.047(7.67)0.015(0.78)0.029(2.01)−0.056(−0.93)
CHURCH0.009(2.21)−0.050(−3.70)−0.049(−3.71)−0.051(−3.81)−0.055(−3.99)−0.052(−3.89)−0.095(−3.94)
Municipality Level Controls
DIVERSITY−0.050(−1.95)−0.064(−0.52)−0.067(−0.54)−0.100(−0.83)−0.192(−1.32)−0.081(−0.65)0.232(1.20)
COMPETITION0.005(1.22)−0.032(−1.93)−0.030(−1.84)−0.040(−2.35)−0.043(−2.19)−0.040(−2.41)−0.083(−3.30)
UNEMPLOYMENT RATE−0.002(−1.23)0.024(3.04)0.021(2.41)0.032(4.20)0.026(2.91)0.031(3.80)0.035(3.54)
PROPORTION OF UNEMPLOYED IN ENTR. PROGRAMS0.005(0.29)−0.005(−0.09)0.004(0.07)−0.038(−0.67)−0.047(−0.73)−0.036(−0.64)0.053(0.70)
SHARE OF EMPLOYED IN FINANCIAL SECTOR1.053(7.86)−0.716(−1.75)−0.692(−1.48)−0.491(−1.04)0.073(0.09)−0.878(−1.84)−1.661(−1.79)
PROPORTION OF PUBLIC EMPLOYEES−0.082(−0.43)−0.126(−0.17)−0.191(−0.26)0.065(0.09)0.217(0.35)−0.100(−0.14)−2.687(−2.42)
INCOME PER CAPITA−0.046(−2.10)0.221(2.34)0.292(2.46)0.076(0.85)0.251(1.77)0.137(1.32)0.476(3.04)
WEALTH TAX PER CAPITA0.067(1.21)−0.083(−0.31)−0.259(−0.80)0.483(1.85)0.582(1.72)0.239(0.86)−0.577(−1.47)
UNEMPLOYMENT RATE * UNEMPLOYED0.000(−0.08)0.005(1.38)0.006(1.48)0.005(1.29)0.005(1.33)0.005(1.35)0.007(1.60)
PROPORTION OF IMMIGRANTS * IMMIGRANT0.171(1.46)−1.490(−4.39)−1.503(−4.50)−1.326(−3.71)−1.240(−3.61)−1.485(−4.36)−1.306(−2.57)
ENTRY RATE0.327(17.95)−0.274(−3.64)−0.342(−2.93)−0.143(−1.99)−0.401(−2.62)−0.306(−2.36)−1.172(−2.72)
EXIT RATE−0.067(−3.15)−0.016(−0.16)−0.013(−0.13)0.211(2.17)0.724(2.59)0.067(0.69)0.276(1.68)
LAMBDA  −0.318(−17.42)−0.368(−5.16)−0.333(−18.91)−0.678(−3.45)−0.523(−4.08)−1.556(−2.52)
LLM fixed effectsNo Yes Yes Yes Yes Yes Yes 
N1,684,596 79,356 79,356 56,642 56,642 79,356 79,356 
Log-likelihood−446,016             
Adjusted R20.470 0.170 0.170 0.170 0.165 0.169 0.170 
     Statisticp-Value  Statisticp-ValueStatisticp-ValueStatisticp-Value
               
Test of overidentifying restrictions    0.880.35  4.630.0991.030.312.0450.15
Bound–Jaeger–Baker F-test    29.210.00  2.120.0830.230.0016.240.00
Test of ENTREPRENEURSHIP = ENTREPRENEURSHIP OUTSIDE THE SOCIAL GROUP            6.070.01

In the second stage, the inverse Mills' ratio is negative and significant. This indicates that entrepreneurs have lower unobservable skills than nonentrepreneurs. More importantly, our results suggest that individuals who belong to more entrepreneurial municipalities are willing to be entrepreneurs even if on average they earn lower profits. Strikingly, the income of entrepreneurs in the three lowest quartiles of the income distribution is lower in social groups where a larger proportion of the richest individuals are entrepreneurs (Models 4 and 5).

The results are not only statistically but also economically significant. A one-standard deviation increase in the municipality's proportion of entrepreneurs (or entrepreneurs with a given educational achievement) decreases entrepreneurial profits by slightly more than 10%. Also in this case, the effect is more pronounced when we use the proportion of entrepreneurs in the top quartile of the income distribution as a proxy for entrepreneurial activity. In Model 5, a one-standard deviation increase in the proxy is associated with a 35% decrease in the entrepreneurial profits of the individuals in the three lowest quartiles of the income distribution.21 Also in this case our specification tests suggest that the more conservative estimates are more reliable. The specification tests for the other two proxies for entrepreneurial activity indicate that our instruments are strong. Most importantly, we cannot reject the null that the instruments have no direct impact on entrepreneurial profits. When we use the proportion of entrepreneurs in the top quartile of income distribution, however, our instruments are weak. Additionally, the test for overidentifying restrictions suggests that the probability that the instruments have no direct impact on entrepreneurial profits is slightly lower than 10%. Indeed, the magnitude of the effect is only one-third and comparable to one of the other proxies for entrepreneurship when we do not instrument the proportion of entrepreneurs in the top quartile of the income distribution (Model 4).

Our identification assumptions are supported not only by statistical tests but also by a test of own and cross-group influences (Model 7), similar to the one we perform in Subsection 4.1. The coefficient of the entrepreneurship in an individual's social group is always negative and significant, and the coefficient of the entrepreneurship among individuals with different educational achievement is not statistically significant being almost 10 times smaller.

Also note that our results are unlikely to depend on the fact that there may be more part-time entrepreneurs in some municipalities, as the coefficient of our variable of interest remains negative and significant when we control for the share of individual income earned in the entrepreneurial activity (by including a dummy for part-time entrepreneurs or the share of income earned from the entrepreneurial activity). Similarly, results are qualitatively invariant if we control for tenure by including a dummy variable capturing whether the individual has been self-employed for more than 1 year.

The result that entrepreneurial profits are lower in municipalities where the entrepreneurship rate is higher is not completely surprising in light of recent papers suggesting that the nonpecuniary benefits of self-employment are substantial. Moskowitz and Vissing-Jorgensen (2002), for instance, show that entrepreneurs largely under-diversify their portfolios investing in their own businesses. The returns they enjoy on their entrepreneurial activities are too low to justify their behavior. As a consequence, they expect entrepreneurs to enjoy large nonpecuniary benefits. Hamilton (2000) finds that entrepreneurs enter and persist in business despite the fact that they have both lower initial earnings and lower earning growth than they would as employees. Our contribution is to show that the importance of nonpecuniary benefits may vary substantially across social groups.22

5. Robustness and Alternative Explanations

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Individual Decision to Become an Entrepreneur and Neighborhood Entrepreneurial Activity
  5. 3. Data Description and Identification Strategy
  6. 4. Results
  7. 5. Robustness and Alternative Explanations
  8. 6. Conclusion
  9. Appendices
  10. References

5.1 Entrepreneurial Investment and Entry Costs

If peer effects increase the utility from entrepreneurial activity, then individuals in municipalities with high entrepreneurship should invest more, even if ceteris paribus they earn lower profits than entrepreneurs in other municipalities. We control for the return to entrepreneurial activity by including entrepreneurial profits. As before, we take into account that individuals self-select into the entrepreneurial activity by using a two-stage selection model.

Our conjecture is confirmed in Table VI (Model 1). Overall, it appears that entrepreneurial investment is larger when peer effects are stronger. A one-standard deviation increase in the proxy for entrepreneurial activity is associated with an almost 30% increase in investment.

Table VI.  Entrepreneurs' Alternative Job Opportunities
VariableInvestment (1)Nonentrepreneurs' Labor Income (2)Prospective Entrepreneurs' Labor Income (3)
  1. Notes: The dependent variables are the logarithm of entrepreneurial investment (investment) in Model 1, and the logarithm of non-entrepreneurial labor income (labor income) in Models 2 and 3. In Model 2, we consider only individuals who are not entrepreneurs; in Model 3, we consider individuals who start an entrepreneurial activity during the sample period when they are not yet entrepreneurs (prospective entrepreneurs). The investment equation includes only observations referring to individuals who are entrepreneurs. ENTREPRENEURSHIP is defined as the proportion of individuals who are entrepreneurs in the municipality. All equations include the same controls as the equations in Table V with the exception of Models 2 and 3 where we exclude EXPERIENCE and ENTREPRENEURIAL INCOME. Additionally, in Model 1 we control for entrepreneurial income. Parameter estimates of the control variables are not reported for brevity. All models have been estimated using a Heckman selection model. In all models we account for selection into the entrepreneurial activity including the inverse Mills ratio LAMDA as in Table V. The selection equations have been estimated using a probit model. The second-stage equations have been estimated using 2SLS. The instruments are the proportion of individuals older than 60 who are members of the state church and the proportion of votes for right-wing parties in the 1982 elections. The standard errors are corrected for heteroskedasticity and take into account that observations for the same municipality may be correlated. T-statistics are reported in parentheses. We also report the adjusted R2, the Bound–Jaeger–Baker's (1995) test for the quality of instruments and the Hansen's J statistics for overidentifying restrictions.

ENTREPRENEURSHIP 4.875(2.16)   0.650(3.50)  1.709(2.06)
N33,192 1,605,240  71,878 
Adjusted R2 0.024    0.283   0.165 
 Statisticp-ValueStatisticp-ValueStatisticp-Value
       
Test of overidentifying restrictions1.8720.392 3.3430.187 4.7030.095
Bound–Jaeger–Baker F-test18.010.00011,9000.0005046.72 0.000

These results not only confirm our interpretation of previous findings but also allow us to exclude a potential alternative explanation. Entry costs could potentially differ across municipalities. Even though differences in entry costs are unlikely to be generated by differences in regulations, which are homogeneous within the country, social interactions within the municipality could generate valuable information on how to start a business. This information could lower the (fixed) entry cost without increasing the productivity in the entrepreneurial activity. Differences in entry costs are consistent with higher entry rates and lower entrepreneurial incomes in high entrepreneurship municipalities (Klapper et al., 2006). However, they cannot explain why individuals invest more if they are surrounded by entrepreneurs. Less productive entrepreneurs who start a business in supposedly low-entry cost municipalities should invest less, not more than similar entrepreneurs. Nonpecuniary benefits from entrepreneurial activity can instead explain our findings on both entrepreneurial profits and investment.

5.2 Differences in Job Opportunities

A concern with the interpretation of our results could be that LLMs are not the relevant economic unit of analysis. If the labor market were segmented within the LLM, different economic environments could explain why individuals invest more in the entrepreneurial activity and at the same time have lower profits. As we argue in Subsection 3.3, we think that this is unlikely because LLMs have relatively small areas and are defined on the basis of observed commuting patterns. This implies that even though entrepreneurs live in high entrepreneurship municipalities, they work in other municipalities. Additionally, our results are robust to the inclusion of municipality fixed effects.

Another possible explanation for our findings is that individuals living in different municipalities may have unobservable time-varying characteristics that affect their job opportunities. In Table VI (Model 2) we explore this possibility by looking at the labor income of individuals who are not entrepreneurs. We estimate a selection model similar to the model for entrepreneurial income. We find that the labor income of nonentrepreneurs in high entrepreneurship municipalities is higher suggesting that job opportunities are stronger in these municipalities. This result holds not only for nonentrepreneurs, but also for individuals who become entrepreneurs during the sample period before they start the entrepreneurial activity (Model 3), suggesting that the lower entrepreneurial income in higher entrepreneurship municipalities is unlikely to depend on poor skills. Overall, individuals in higher entrepreneurship municipalities appear to have better outside options. Thus, their behavior supports the existence of nonpecuniary benefits from entrepreneurial activity deriving from peer effects.

5.3 Sorting

If entrepreneurs or potential entrepreneurs move to municipalities that are more favorable to entrepreneurial activity, the positive correlation between individual and aggregate occupational choices could depend on sorting instead of peer effects. We test whether sorting of individuals more prone to entrepreneurial activity can explain our results by looking at movers. We find that individuals who move to municipalities with more entrepreneurs are no more likely to become entrepreneurs than individuals who already reside in those municipalities (as suggested by the insignificant coefficient of the variable obtained by interacting the mover dummy with the difference between the proportion of entrepreneurs in the destination and the original municipalities in Panel A of Table VII). The result does not depend on the time horizon we choose for identifying movers. The coefficient of the interaction variable continues to be insignificant if we define individuals who moved between 1 and 3 years earlier as movers.

Table VII.  Sorting
Panel A. The Decision to Become an Entrepreneur and Moving
 
 Movers within 1 Year (1)Movers within 3 Years (2)
ENTREPRENEURSHIP  15.458(16.39) 15.882(8.22)
MOVER * DIFF(ENTREPRENEURSHIP)   0.335(0.20)  3.094(1.41)
N1,179,314 497,308 
Adjusted R2   0.007   0.007 
Panel B. Mobility of Entrepreneurs and Nonentrepreneurs
 
 NonentrepreneursEntrepreneurs
1 Year3 Years1 Year3 Years
  1. Notes: In Panel A, the dependent variable is a dummy variable with a value equal to one if individual i becomes an entrepreneur at time t and equal to zero otherwise. Individuals who are already entrepreneurs at time t−1 are excluded. ENTREPRENEURSHIP is defined as the proportion of individuals who are entrepreneurs in a municipality. DIFF(ENTREPRENEURSHIP) is the difference between level of entrepreneurial activity in destination and origin municipalities. MOVER is a dummy that is equal to one if an individual moved within 1 or 3 years (Models 1 and 2, respectively) and is equal to zero otherwise. All equations include the same controls as the equations in Table IV. The equation has been estimated using OLS. The standard errors are corrected for heteroskedasticity and take into account that observations for the same municipality may be correlated. T-statistics are reported in parentheses. We also report the adjusted R2. Estimates are multiplied by 100.

Did not move96.44%92.60%97.87%94.78%
Move to higher entrepreneurship area 1.78% 3.23% 1.03% 2.09%
Move to lower entrepreneurship area 1.78% 4.17% 1.10% 3.13%

Furthermore, in Panel B of Table VII, we analyze individuals' moving decisions. We find that entrepreneurs are always less likely to move. Most importantly, nonentrepreneurs are more likely to move to higher entrepreneurship municipalities than are entrepreneurs. Overall, the evidence does not support the possibility that individuals who are inclined to become entrepreneurs or the self-employed sort in municipalities where there are more entrepreneurs.

5.4 Young People's Entrepreneurial Choices

To further understand why the level of entrepreneurial activity affects individual occupational choice, we analyze the occupational choice of young people (ranging from 18 to 35) in 1995. For this subsample, we are able to observe the capital profits of the parents. This is a good proxy for family wealth and an imperfect proxy for whether an individual's parents were involved in an entrepreneurial activity. We can thus test whether social interactions are still significant after controlling for the fact that an individual may have inherited a business or received loans and transfers from parents.

The young people subsample also allows us to further check the quality of our instruments. Our identification strategy so far has relied on the fact that, thanks to the large institutional changes that affected Sweden during the 1980s, differences in culture in the early 1980s are predetermined with respect to the current level of entrepreneurial activity. Entrepreneurial activity was a much less common phenomenon in the early 1980s. However, there may have been some preexisting differences in entrepreneurial activity that affected the local culture. Also, if individuals persist in entrepreneurial activity, some of the activity we observe may have affected the local culture in the early 1980s.

Our instruments, however, are exogenous with respect to the occupational choices of young individuals who were not part of the labor force in the early 1980s. We reestimate both our equations of interest considering only the level of entrepreneurial activity among young people and how this affects other young people's occupational choices and entrepreneurial profits.

Table VIII shows that our main results are qualitatively unchanged: Entrepreneurial activity has a positive (negative) effect on the probability of becoming an entrepreneur (entrepreneurial profits) as we find in Section 4.23 Family wealth, as proxied by the capital income of the parents, does not have a significant effect on the decision to become an entrepreneur. It is, however, negatively related to entrepreneurial profits, suggesting that wealthy individuals may enjoy higher utility from the entrepreneurial activity. This finding is consistent with Hurst and Lusardi (2004) who argue that being an entrepreneur may be a luxury good.

Table VIII.  Young People's Entrepreneurial Choice
 Decision to Become an EntrepreneurEntrepreneurial Profits
OLS (1)2SLS (2)OLS (3)2SLS (4)
  1. Notes: Only individuals younger than 18 years in 1983 are included. We report the estimates for the linear probability model (Table IV) and entrepreneurial income (Table V). The equations include the same controls as Table IV and Table V, respectively. In addition, here we control for the PARENTAL CAPITAL INCOME IN 1970. Parameter estimates of most of controls are omitted for brevity. The standard errors are corrected for heteroskedasticity and take into account that observations for the same municipality may be correlated. T-statistics are reported in parentheses. We also report the adjusted R2. Estimates in Models 1 and 2 are multiplied by 100; the coefficient for PARENTAL CAPITAL INCOME IN 1970 in the income equation is multiplied by 1,000.

ENTREPRENEURSHIP12.787(3.54)8.391(1.60)4.569(2.25)9.955(1.98)
PARENTS' CAPITAL INCOME IN 1970 0.356(1.46)0.355(1.46)0.199(−3.99) 0.203(−7.96)
N81,185 81,185 1,946 1,946 
Adjusted R2 0.006 0.005 0.188  0.184  

5.5 Informal Credit Markets

So far our results allow us to exclude that peer effects increase entrepreneurial profits. However, peer effects might not necessarily matter for nonpecuniary reasons. In high entrepreneurship municipalities, social networks might ease access to credit and facilitate entry and investment for less profitable entrepreneurs.

To explore this possibility, we examine the capital structure of firms that have less than 50 employees and less than SEK 1,000,000 (approximately USD 130,000) in assets. If our results were driven by the availability of cheap informal credit, small firms, incorporated in municipalities where entrepreneurial activity is higher, should use more trade credit and other informal loans. Similarly, if access to formal credit markets were easier, small firms should have higher leverage.

We investigate how firms finance their assets distinguishing among loans from financial institutions, trade credit, and other loans. We also look at how much trade credit firms grant. In Table IX, our proxy for entrepreneurial activity is related to firms' financial ratios neither in OLS nor in 2SLS regressions. If anything, firms in high entrepreneurship municipalities grant less trade credit. This casts doubts on the importance of credit markets and supports the interpretation that peer effects matter as a result of nonpecuniary benefits.

Table IX.  The Capital Structure of Small Firms
 Informal DebtTrade Credit
OLS (1)2SLS (2)OLS (3)2SLS (4)
ENTREPRENEURSHIP1.8149(1.44)1.7699(0.56)0.1206(0.84)0.6315(1.57)
ROA−0.7557(−5.16)−0.7557(−5.16)−0.0076(−7.54)−0.0075(−7.54)
TANGIBLE ASSETS−0.0617(−0.54)−0.0615(−0.53)−0.0270(−2.17)−0.0250(−1.96)
TOTAL ASSETS−1.9985(−3.02)−2.0006(−3.01)−0.2045(−5.77)−0.2049(−5.78)
Adjusted R20.7868 0.7868 0.0513  0.0513 
 Financial LeverageAccounts Receivable
OLS (5)2SLS (6)OLS (7)2SLS (8)
  1. Notes: We report estimates for informal debt (defined as the sum of long and short term liabilities not granted by financial institutions or firms in the same industrial group over total assets), account payables (defined as the ratio of account payables to total assets), financial leverage (defined as the ratio of financial loans to total assets), and account receivables (defined as the ratio of account receivables to sales) as function of the proportion of entrepreneurs in the firm's municipality of incorporation, returns on assets (ROA), the ratio of fixed assets to total assets (TANGIBLE ASSETS), and the logarithm of total assets. We also include LLM fixed effects and industry fixed effects at the four-digit level. The standard errors are corrected for heteroskedasticity and take into account that observations for the same municipality may be correlated. T-statistics are reported in parentheses. We also report the adjusted R2. The sample consists of firms with less than 50 employees and less than SEK 1,000,000 assets. The sample includes 185,294 firms and is based on year 2000 financial statements. Coefficient estimates for ROA in the Financial Leverage and Accounts Receivable regressions are multiplied by 10,000.

ENTREPRENEURSHIP0.0646(0.74)0.1403(0.39)0.0972(2.91)0.1693(2.33)
ROA−1.7126(−0.84)−1.7119(−0.84) 0.0220 (2.28) 0.0225  (2.28)
TANGIBLE ASSETS 0.0408 (4.39) 0.0401 (4.12)−0.0885(−21.19)−0.0881(−21.32)
TOTAL ASSETS−0.1304(−2.77)−0.1307(−2.77) 0.0401 (17.31) 0.0401 (17.34)
Adjusted R2 0.0016  0.0016  0.1575  0.1574 

5.6 Overconfidence

In principle, overconfidence could explain why individuals with lower profits invest more in their business. In this case, our instruments would capture an omitted individual characteristic, overconfidence. The tests of overidentifying restrictions and the fact that we control for roughly the corresponding individual cultural traits suggests that religiosity and political orientation enter into the equation through the level of entrepreneurial activity. The very fact that some neighbors are entrepreneurs seem to affect the individual's occupational choice, making omitted individual characteristics, such as overconfidence, unlikely.

The level of entrepreneurial activity, however, could create herding (Bernardo and Welch, 2001). Some individuals observing a high level of entrepreneurial activity may revise upward (downward) their expectations on entrepreneurial profits (effort) and thus be more likely to become entrepreneurs. In this respect, a high level of entrepreneurial activity would endogenously generate overconfidence. Social interactions would increase the utility from entrepreneurial activity without affecting the actual profitability (effort).

We view this as a nonpecuniary effect of social interactions, similar the desire for conformity or prestige. We are unable to fully distinguish between these effects. We can provide, however, time-series evidence suggesting that social interactions are unlikely to generate overconfidence. Individuals should have become more overconfident in the late 1990s, during the high-tech boom, when the number of initial public offering (IPOs) dramatically increased and many entrepreneurs made fortunes. Some individuals may have overestimated their expected profits based on the entrepreneurial successes of others and decided to become entrepreneurs in areas with high entrepreneurial activity. In this case, our results should be driven by the correlation between individual and aggregate occupational choices in the second part of the sample (1998–2000). In fact, the results for the subperiods 1996–1997 and 1998–2000 are similar to those we report. This suggests that herding phenomena are unlikely to explain our findings.

5.7 Further Robustness Checks

Our results are robust to a number of modifications of the equations we present. For example, we check whether the definition of entrepreneurship is a key determinant of our results. In fact, it is not. We reestimate all equations excluding individuals who run incorporated businesses. These individuals can save through their company. It is thus difficult to evaluate the income of entrepreneurial activity because we have no measure of capital gains for unlisted companies. All our results, however, are qualitatively invariant if we include only individuals with nonincorporated businesses who, not having the option of saving through their companies, cannot benefit from capital gains.

To gauge a better understanding of why entrepreneurial activity matters, we explore the mechanism through which social interactions should operate. Community attachment is generally much stronger in rural areas. Hence, if the effect of entrepreneurship in the municipality is due to social interactions, we expect our results to be stronger in rural LLMs. If instead our results depend on the failure to control for municipality heterogeneity, we expect the results to be stronger for urban LLMs (Stockholm, Göteborg, Malmö, and Uppsala), where approximately one-third of the individuals in our sample live. In this case, a LLM includes many more municipalities and our controls may capture municipality heterogeneity to a lesser extent.

We reestimate the two equations by interacting our proxy for entrepreneurship with two dummy variables, the first takes the value of 1 for urban areas and the second takes the value of 1 for nonurban areas. We find that the effect of the variable proxying for social interactions on the probability to become an entrepreneur is positive and significant for both urban and nonurban areas. However, when we use 2SLS, entrepreneurial profits are only negatively related to the proxy for entrepreneurial activity in nonurban municipalities. Hence, entrepreneurs seem to enjoy non-pecuniary benefits by conforming to other individuals only where we expect peer effects to be stronger.

Finally, we remove subsets of the control variables, and, in particular, the education and LLM fixed effects. None of these robustness checks produces results that are significantly different from those we report. This increases our confidence that our results are not due to omitted variable bias. If unobservable characteristics about individuals or municipalities drove our results, one would expect that increasing the set of unobservable characteristics by treating observable characteristics as unobservable would have a large impact on the estimate of the social interaction term. In fact, the estimates are almost invariant.

6. Conclusion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Individual Decision to Become an Entrepreneur and Neighborhood Entrepreneurial Activity
  5. 3. Data Description and Identification Strategy
  6. 4. Results
  7. 5. Robustness and Alternative Explanations
  8. 6. Conclusion
  9. Appendices
  10. References

This paper suggests that peer effects play a significant role in the individual decision to become an entrepreneur. Individuals who belong to more entrepreneurial social groups earn lower entrepreneurial profits and invest more in their own business, even though their alternative job opportunities appear more attractive. This suggests that peer effects may increase the nonpecuniary benefits of running one's own business.

Our results indicate that cultural values may have an indirect effect on economic decisions. We find that in religious communities, there are more entrepreneurs and, for this reason, even nonreligious individuals are more likely to choose an entrepreneurial activity. This suggests that culture may be related to economic performance (Knack and Keefer, 1997) not only because of a direct effect on individual choices, but also because it generates peer effects. In this respect, small differences in cultural values may lead to dramatic differences in economic outcomes. We believe that this is an interesting topic for future research.

Appendices

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Individual Decision to Become an Entrepreneur and Neighborhood Entrepreneurial Activity
  5. 3. Data Description and Identification Strategy
  6. 4. Results
  7. 5. Robustness and Alternative Explanations
  8. 6. Conclusion
  9. Appendices
  10. References

Appendix A

We sketch a simple model of occupational choice. Within this model we illustrate how the decision to become an entrepreneur, entrepreneurial profits, and investment may differ between individuals belonging to different social groups in a homogeneous economic environment (the context on which we focus in the empirical analysis).

We model the effect of social interactions in two ways. First, in some social groups being an entrepreneur may be considered valuable. Hence, we assume that the utility of entrepreneurs increases not only in entrepreneurial profits, but also in the output from entrepreneurial activity. The extent to which entrepreneurial output is weighted in the utility function depends on the social norm prevailing in an individual's peer group. Second, belonging to a high entrepreneurial social group may increase profits from entrepreneurial activity as a result of social learning and knowledge spillovers.

Formally, the utility of an individual i, involved in the entrepreneurial activity is

  • image

Here, Yig is the output from the entrepreneurial activity of individual i belonging to social group g; Ii is the entrepreneurial investment of individual i; σg is the value attributed to entrepreneurial investment by the social group g, to which individual i belongs. In other words, σg is the parameter capturing nonpecuniary benefits. We assume that σg≥ 0. If σg= 0, only entrepreneurial profits, πig≡ Yig − Ii, affect the utility. If σg > 0, entrepreneurs derive utility from being able to invest in their own business.24

Entrepreneurs who are penniless, borrow Ii at zero interest rate, and can perfectly commit to reimburse the loan.

Entrepreneurial profits depend on individual productivity, indexed by i, and the individual's social group, g. The output from the entrepreneurial activity (Yig) increases in entrepreneurial investment (Ii) and productivity (Aig) as follows: Yig = AigIγi. The entrepreneurial productivity of individual i, in turn, depends positively on her entrepreneurial skills, si, and the stock of entrepreneurial knowledge in the individual's social group (social learning), ag

  • image

We also assume that inline image25

Entrepreneurs choose the level of investment maximizing their utility. Thus, the optimal investment of an individual who has chosen the entrepreneurial activity solves the program

  • image

The first-order condition is

  • image

If there are no nonpecuniary benefits (σg= 0), entrepreneurs maximize profits and equate the marginal utility of investment to its marginal cost: γAigIγ−1i= 1. The term 1 +σg indicates the extent of overinvestment, which is increasing in the strength of the non-pecuniary benefits (σg).

The utility of a worker depends merely on her wage, wi. An individual chooses to become an entrepreneur if UEi(I*i) ≥ wi .

Because inline image, the skills of the marginal entrepreneur are lower. Thus, individuals belonging to social groups with higher stock of entrepreneurial knowledge are more likely to choose to become entrepreneurs. Similarly, because inline image, more individuals become entrepreneurs in social groups that highly value entrepreneurial activity.

Because the average entrepreneur is less skilled and the extent of overinvestment is larger in social groups that highly value entrepreneurial activity, average entrepreneurial profits are lower. Conversely, when entrepreneurs benefit from entrepreneurial knowledge in their social group, they have higher productivity in their own business and, for given skills, they earn higher profits if they belong to groups with high levels of entrepreneurial activity.

Appendix B: Control Variables

We control for individual and municipality characteristics.

Panel A of Table II summarizes the individual characteristics. These are:

Table II.  Descriptive Statistics
Panel A. Individual Characteristics
 
 MeanMedianMinimumMaximumStandard DeviationInterquartile Range
LABOR INCOME 5.220 5.211 0.000 7.309 5.148 5.193
INCOME OF OTHER MEMBERS 5.191 5.131 0.000 7.309 5.241 5.377
EXPERIENCE 0.056 0.000 0.000 1.000 0.229 0.000
PART-TIME 0.035 0.000 0.000 1.000 0.184 0.000
WEALTH 5.876 5.912 0.000 9.013 0.431 0.429
SHARE OF LIQUID ASSETS−1.209−0.561−6.695 0.777 1.796 1.261
AGE 40.488 40.000 18.00060.00011.77617.000
MALE 0.494 0.000 0.000 1.000 0.500 1.000
MOVER 0.093 0.000 0.000 1.000 0.290 0.000
CHANGES IN FAMILY STRUCTURE 0.173 0.000 0.000 1.000 0.378 0.000
NUMBER OF CHILDREN 1.111 1.000 0.00012.000 1.184 2.000
MARRIED 0.551 1.000 0.000 1.000 0.497 1.000
DIVORCED 0.072 0.000 0.000 1.000 0.259 0.000
UNEMPLOYED 0.135 0.000 0.000 1.000 0.342 0.000
IMMIGRANT 0.096 0.000 0.000 1.000 0.295 0.000
WAGE PREMIUM 0.011 0.035−6.663 4.175 0.569 0.390
CHURCH 0.824 1.000 0.000 1.000 0.381 0.000
Panel B. Municipality Characteristics
 
VariableMeanMedianMinimumMaximumStandard DeviationInterquartile Range
  1. Notes: Panel A: LABOR INCOME is the logarithm of an individual's salary plus 0.01. INCOME OF OTHER MEMBERS is the logarithm of the income of the other household members plus 0.01. EXPERIENCE is a dummy variable that takes value one for individuals who are entrepreneurs both at time t and t−1 and zero otherwise. PART-TIME is a dummy variable that takes value one if the individual earns less than 80% of her income from the entrepreneurial activity and zero otherwise. WEALTH is the logarithm of household's total wealth plus 0.01. SHARE OF LIQUID ASSETS is the logarithm of the proportion of wealth invested in bank accounts or securities plus 0.01. AGE is the individual's age. MALE is a dummy variable that takes value one for males and zero otherwise. MOVER is a dummy variable that takes value one if an individual changed municipality during the previous year and zero otherwise. CHANGES IN FAMILY STRUCTURE is a dummy variable that takes value one if there have been any changes in family structure during the previous year and zero otherwise. NUMBER OF CHILDREN is the number of children in the household. MARRIED is a dummy variable that takes value one if an individual is married and zero otherwise. DIVORCED is a dummy variable that takes value one if an individual is divorced and zero otherwise. UNEMPLOYED is a dummy variable that takes value one if an individual is unemployed and zero otherwise. IMMIGRANT is a dummy variable that takes value one if an individual is born outside Sweden and zero otherwise. The WAGE PREMIUM is the residual of the regression of the individual labor income on age and its square, the family status variables mentioned before, the immigrant dummy, a dummy equal to one for individuals with a handicap, the unemployed dummy, and dummy variables controlling for an individual's education level, industry of employment, occupation, and seven regional dummies. CHURCH is a dummy variable that takes value one if the individual is a member of the state church and zero otherwise. All observations from 1995 to 2000 are included.

  2. Panel B: PROPORTION OF UNEMPLOYED IN ENTREPRENEURSHIP PROGRAMS is the number of unemployed individuals enrolled in the municipality's entrepreneurship programs during the year over the number of those unemployed at the end of the year; UNEMPLOYMENT RATE is the rate of unemployment; PROPORTION OF FINANCIAL SECTOR EMPLOYEES is the share of employment in the financial sector; SPECIALIZATION is the share of employment in the five most important industries; COMPETITION is the number of firms per employee in the municipality relative to the number of firms per employee in Sweden; INCOME PER CAPITA is the income per capita; WEALTH TAX PER CAPITA is the wealth tax per capita.

PROPORTION OF UNEMPLOYED IN ENTREPRENEURSHIP PROGRAMS0.2520.2190.027 1.1470.1350.136
UNEMPLOYMENT RATE6.3356.1291.21613.7892.1723.148
PROPORTION OF FINANCIAL SECTOR EMPLOYEES0.0100.0090.003 0.1170.0070.003
SPECIALIZATION0.4510.4380.341 0.7060.0660.078
COMPETITION1.9601.9440.703 3.5610.5030.644
PROPORTION OF PUBLIC EMPLOYEES0.0680.0690.038 0.1040.0100.014
INCOME PER CAPITA156,662  154,339  122,718  340,051 21,467 22,817  
WEALTH TAX PER CAPITA1,1371,030   300 6,986  594  757
  • 1
    The logarithm of the salary received by an individual plus 0.01 (LABOR INCOME) and the logarithm of the income of the other members of the household plus 0.01 (INCOME OF OTHER MEMBERS), both measured the year before the occupational choice. While an increase in non-entrepreneurial income is expected to decrease the probability of an individual becoming an entrepreneur, the expected effect of the income of the other household members is ambiguous. This variable may have a positive effect because more resources are available to set up a new business. Additionally, this variable may have a negative effect if individuals are more likely to become entrepreneurs when they need extra resources to make a living.
  • 2
    EXPERIENCE is a dummy variable for individuals who have been self-employed for more than 1 year. It is included to capture the fact that tenure can affect entrepreneurial profits.
  • 3
    A dummy variable for individuals who earn less than 80% of the income from the entrepreneurial activity (PART-TIME). This variable takes into account that an individual may have lower entrepreneurial income because she is involved part-time in the entrepreneurial activity.
  • 4
    The logarithm of wealth (WEALTH) and the square of the logarithm of wealth (WEALTH2) of an individual's household. Wealth includes all financial and real assets of the household that are subject to a wealth tax plus 0.01. These variables have been included because wealthy individuals are less likely to be subject to liquidity constraints that keep them from starting a business (Evans and Jovanovic, 1989; Holtz-Eakin et al., 1994).
  • 5
    The logarithm of the ratio of liquid assets, including securities and bank accounts plus 0.01 to total wealth (SHARE OF LIQUID ASSETS), which takes into account that only the most liquid assets may be available to fund a new business.
  • 6
    The individual age (AGE) and its square (AGE2).
  • 7
    A dummy variable for men (MALE), to account for gender differences.
  • 8
    A dummy variable for married individuals (MARRIED), a dummy variable for divorced individuals (DIVORCED), the logarithm of the number of children in the household (NUMBER OF CHILDREN), a dummy variable equal to 1 if either the number of children or the marital status changed during the last year and equal to zero otherwise (CHANGES IN FAMILY STRUCTURE). These variables may be related to an individual's risk aversion because they proxy for the responsibility an individual has toward the household (Evans and Leighton, 1989). Moreover, individuals whose status recently changed may have a stronger need for extra resources. This may affect their willingness to start a new business.
  • 9
    A dummy variable for individuals born abroad (IMMIGRANT); a dummy variable for individuals that changed municipality during the last year (MOVER); and a dummy variable for individuals who were unemployed the year before starting the entrepreneurial activity (UNEMPLOYED).
  • 10
    The wage premium or discount (WAGE PREMIUM). This variable has been computed as the residual of the regression of the individual labor income on individual age and its square, the family status variables mentioned before, a dummy equal to 1 for individuals with a handicap, and dummy variables controlling for immigrant status, education level, industry of employment, occupation, and the seven Swedish macroregions.
  • 11
    A dummy variable equal to 1 for individuals who are members of the state church (CHURCH).

Panel B of Table II summarizes the municipality characteristics from Market Manager and Statistics Sweden:

  • 1
    The level of unemployment in a municipality (UNEMPLOYMENT RATE) and the proportion of public sector employees in the population (PROPORTION OF PUBLIC EMPLOYEES).
  • 2
    The number of unemployed individuals enrolled in the municipality's entrepreneurship programs during the year over the number of unemployed at the end of the year26 (PROPORTION OF UNEMPLOYED IN ENTREPRENEURSHIP PROGRAMS).
  • 3
    The proportion of employment in the financial sector (PROPORTION OF FINANCIAL SECTOR EMPLOYEES), which captures financial development. Although this variable is endogenous, and a more developed financial sector may certainly depend on higher demand for financial services in municipalities with more entrepreneurs, we include it as a control because firm creation is positively affected by financial development. If we did not control for this variable, the stock of entrepreneurs in a municipality may help explain occupational choices because of the greater availability of financial services.
  • 4
    The share of the top five industries in local employment to the share of the top five industries in national employment (SPECIALIZATION), which provides a measure of specialization. This variable proxies for the existence of dynamic externalities (Glaeser et al., 1992), which may increase productivity in areas that specialize in few sectors, and determine a concentration of entrepreneurs in some municipalities.
  • 5
    The number of firms per employee incorporated in a municipality relative to the number of firms per employee in Sweden (COMPETITION).27
  • 6
    Per capita income (INCOME PER CAPITA) and per capita wealth tax (WEALTH TAX PER CAPITA).
  • 7
    Entrepreneurial entry (ENTRY RATE) and exit rates (EXIT RATE), which capture temporary shocks that may favor entry into the entrepreneurial activity. They also control for differences in firm dynamics that can influence our results. If in municipalities with a high proportion of entrepreneurs, more firms go out of business and are replaced by new ones, we could observe a positive correlation between the individual decision to become an entrepreneur and the proportion of entrepreneurs. This, however, would not indicate either social learning or social norms, but would simply be related to firm dynamics. By controlling for firm entry and exit rates, we overcome this problem.

In addition, we include two education-group fixed effects (for individuals with high school diplomas and university education, respectively), 11 industry dummies, which refer to the sector in which an individual is employed, 4-year dummies, 108 LLM dummies, and dummy variables for the wealthiest and poorest municipalities within a LLM.28

Footnotes
  • 1
  • 2

    See Aldrich (2005) and Licht and Siegel (2006) for comprehensive surveys.

  • 3

    In some alternative specifications, we include municipality fixed effects.

  • 4

    The low average returns of individuals who engage in independent inventive activities can be explained similarly (Astebro, 2003).

  • 5

    We do not explicitly look at exit from the entrepreneurial activity because both social norms and social learning imply that individuals belonging to highly entrepreneurial social groups are less likely to abandon the entrepreneurial activity. In unreported regressions, we find that indeed this is the case.

  • 6

    We have 288 municipalities until 1999, when one of the municipalities split in two.

  • 7

    Individuals who already are entrepreneurs at t−1 are excluded from the sample. Results for the decision to be an entrepreneur are similar to the ones for the decision to become an entrepreneur and are also reported.

  • 8

    These studies explore only a small subset of the geographical areas we study.

  • 9

    We use aggregate entry in (exit from) the entrepreneurial activity.

  • 10

    In unreported specifications, we directly control for municipality tax rates. These, as well as most of the other municipality level controls, are never statistically significant.

  • 11

    Swedish individuals who are members of evangelical churches are generally also members of the state church.

  • 12

    As shown in Tables IV to VI, the F-test of the regression of our variable of interest on the instruments is always strongly statistically significant even after controlling for all the independent variables that we include in the second stage. Hence, we do not have to worry about possible inconsistency problems arising when the correlation between the instruments and the endogenous explanatory variable is weak (Bound et al., 1995).

  • 13

    In all specifications, we cluster errors at the municipality level because some of the control variables do not vary within the municipality. As pointed out by Moulton (1990), if we did not cluster errors at the municipality level, measurement errors could bias the standard errors of our estimates. Standard errors do not differ much from the ones we report if we cluster errors at the LLM level.

  • 14

    Results (nonreported) are also qualitatively similar when we use only one of the two instruments.

  • 15

    Only the proportion of pensioners who are part of the state church has some limited variation over time.

  • 16

    The effect of our variable of interest is twice as large as the effect of the wage premium perceived by the individual.

  • 17

    The magnitude of the effect is similar if we use the whole sample rather than the individuals in the lowest three quartiles of the income distribution. This suggests that the larger impact is due to the different proxy for entrepreneurial activity and not due to the different sample.

  • 18

    The decrease in the magnitude of the coefficient is likely to depend on the high correlation between our variable of interest and the variable included to run the falsification test.

  • 19

    The statistical significance and the sign of the control variables are consistent with the findings in the literature. This is comforting. Individuals appear more likely to become entrepreneurs if they have a low opportunity cost (lower salary and wage premium) or a stronger need for extra resources (they have more children and household members with low incomes). Furthermore, household wealth has a positive impact on the probability of starting a business only if the wealth is relatively large, as Hurst and Lusardi (2004) find for the United States. Finally, municipality characteristics, other than those proxying for entrepreneurial activity, have a marginal impact suggesting that individual level controls capture most of the heterogeneity in the population within the LLM.

  • 20

    The system is identified not only by the distributional assumptions but also by the following exclusion restrictions: in the second stage we do not include the individual's salary and the income of other household members, which should be unrelated to the entrepreneur's productivity.

  • 21

    The magnitude of the effect is similar if we use the whole sample rather than the individuals in the lowest three quartiles of the income distribution.

  • 22

    A possible criticism to our results is that entrepreneurial income is often underreported for tax reasons. Because we compare entrepreneurial income in different locations, not entrepreneurial income with employees' wages, our results should be immune to this criticism as long as social interactions do not make individuals more efficient in avoiding taxes. Nevertheless, in unreported results, we explore this possibility and consider that entrepreneurs could camouflage consumption expenditures by logging them as investment in the entrepreneurial activity to avoid taxes. We thus use a measure of gross entrepreneurial income as a dependent variable instead of entrepreneurial profits. Our estimates appear unchanged. Another concern is that our results may be driven by individuals who set up a business as a part-time activity. If part-time entrepreneurs were crucial, our results could again indicate that peer effects matter for tax avoidance behavior rather than for occupational choice. If we consider as entrepreneurs only individuals who obtain at least 90% of their income from the entrepreneurial activity (full-time entrepreneurs), the estimates confirm our earlier findings. This again suggests that our findings are driven by peer effects and that entrepreneurs do not simply underreport income for tax purposes.

  • 23

    The economic effect of the proxy for entrepreneurial activity on the probability of becoming an entrepreneur is 0.19 percentage points. The effect on the income is, however, larger in this subsample. A one-standard deviation increase in entrepreneurial activity is associated with a 40% decrease in income.

  • 24

    The way in which we model the individual behavioral bias is similar to Malmendier and Tate (2005). It would be straightforward to re-parameterize the utility function to consider that the marginal disutility of effort (or investment) could be lower for individuals belonging to social groups that value entrepreneurial activity to a larger extent.

  • 25

    An individual's location affects entrepreneurial productivity. Some locations offer good infrastructure, easy access to intermediate goods, or other advantages that may enhance entrepreneurial productivity. Here we abstract from these sources of heterogeneity in entrepreneurial productivity because the empirical analysis focuses on individuals belonging to different social groups in a homogeneous economic environment.

  • 26

    Note that, given the definition, this variable can take values larger than 1.

  • 27

    The number of firms incorporated in a municipality differs from the number of self-employed individuals, because in many cases firms are not incorporated.

  • 28

    In other specifications, we also control for other municipality characteristics including proxies for firm performance and population density and a variable ranging from 1 to 6 that measures the individual educational level more precisely than the fixed effects. Because these variables are not statistically significant and do not affect the coefficients of the other explanatory variables, we do not report the results.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Individual Decision to Become an Entrepreneur and Neighborhood Entrepreneurial Activity
  5. 3. Data Description and Identification Strategy
  6. 4. Results
  7. 5. Robustness and Alternative Explanations
  8. 6. Conclusion
  9. Appendices
  10. References
  • Aldrich, H.E., 2005, “Entrepreneurship,” in N.Smelser and R.Swedberg, eds., Handbook of Economic Sociology, Princeton , NJ : Princeton University Press, 451476.
  • Astebro, T., 2003, “The Return to Independent Invention: Evidence of Unrealistic Optimism, Risk Seeking or Skewness Loving? Economic Journal, 113, 226239.
  • Barnett, W.P., A. Swanson, and O. Sorenson, 2003, “Asymmetric Selection Among Organizations,” Industrial and Corporate Change, 12, 673695.
  • Barro, R.J. and R.M. McCleary, 2003, “Religion and Economic Growth.” NBER Working Paper 9682.
  • Baumol, W., 1990, “Entrepreneurship: Productive, Unproductive and Destructive,” Journal of Political Economy, 98, 893921.
  • Bernardo, A.E. and I. Welch, 2001, “On the Evolution of Overconfidence and Entrepreneurs,” Journal of Economics and Management Strategy, 10, 301330.
  • Bernheim, B.D., 1994, “A Theory of Conformity,” Journal of Political Economy, 102, 841877.
  • Bertrand, M., E.F.P. Luttmer, and S. Mullainathan, 2000, “Network Effects and Welfare Cultures,” Quarterly Journal of Economics, 115, 10191055.
  • Blanchflower, D.G., 2000, “Self-Employment in OECD Countries,” Labour Economics, 7, 471505.
  • Blanchflower, D.G., A. Oswald, and A. Stutzer, 2001, “Latent Entrepreneurship across Nations,” European Economic Review, 45, 680691.
  • Borjas, G.J. and L. Hilton, 1996, “Immigration and the Welfare State: Immigrant Participation in Means-Tested Entitlement Programs,” Quarterly Journal of Economics, 111, 575604.
  • Bound, J., D.A. Jaeger, and R.M. Baker, 1995, “Problems with Instrumental Variables Estimation When the Correlation between the Instruments and the Endogenous Explanatory Variable is Weak,” Journal of the American Statistical Association, 90, 443450.
  • Case, A.C. and L.F. Katz, 1991, “The Company You Keep: The Effects of Family and Neighborhood on Disadvantaged Youth,” NBER Working Paper 3705.
  • Cole, H.L., G.J. Mailath, and A. Postlewaite, 1992, “Social Norms, Saving Behavior, and Growth,” Journal of Political Economy, 100, 10921125.
  • Connerly, C.E., 1985, “The Community Question,” Urban Affairs Quarterly, 20, 537556.
  • Cullen, J.B. and R.H. Gordon, 2002, “Taxes and Entrepreneurial Activity: Theory and Evidence for the US,” NBER Working Paper 9015.
  • Cutler, D.M. and E.L. Glaeser, 1997, “Are Ghettos Good or Bad? Quarterly Journal of Economics, 112, 827872.
  • Davidsson, P. 1995, “Culture, Structure and Regional Levels of Entrepreneurship,” Entrepreneurship and Regional Development, 7, 4162.
  • Davidsson, P. and J. Wiklund, 1997, “Cultural Values and Regional Variations in New Firm Formation,” Journal of Economic Psychology, 18, 153340.
  • Davis, S.J. and M. Henrekson, 2005, “Wage-Setting Institutions as Industrial Policy,” Labour Economics, 12, 345377.
  • Demsetz, H. and K. Lehn, 1985, “The Structure of Corporate Ownership: Causes and Consequences,” Journal of Political Economy, 93, 11551177.
  • Desai, M.A., P.A. Gompers, and J. Lerner, 2005, “Institutions, Capital Constraints and Entrepreneurial Firm Dynamics: Evidence from Europe,” Mimeo, Harvard University, Cambridge , MA .
  • Duflo, E. and E. Saez, 2002, “Participation and Investment Decisions in a Retirement Plan: The Influence of Colleagues' Choices,” Journal of Public Economics, 85, 121148.
  • Evans, D.S. and B. Jovanovic, 1989, “An Estimated Model of Entrepreneurial Choice under Liquidity Constraints,” Journal of Political Economy, 97, 808827.
  • Evans, D.S. and L.S. Leighton, 1989, “Some Empirical Aspects of Entrepreneurship,” American Economic Review, 79, 519535.
  • Fallick, B., C.A. Fleischman, and J.B. Rebitzer, 2006, “Job Hopping in Silicon Valley: Some Evidence Concerning the Microfoundations of a High-Technology Cluster,” Review of Economics and Statistics, 88, 472481.
  • Glaeser, E.L., H.D. Kallal, J.A. Scheinkman, and A. Shleifer, 1992, “Growth in Cities,” Quarterly Journal of Economics, 100, 11261152.
  • Glaeser, E.L., B. Sacerdote, and J.A. Scheinkman, 1996, “Crime and Social Interactions,” Quarterly Journal of Economics, 114, 502548.
  • Gompers, P.A., J. Lerner, and D.S. Scharfstein, 2005, “Entrepreneurial Spawning: Public Corporations and the Formation of New Ventures, 1986–1999,” Journal of Finance, 60, 577614.
  • Guiso, L., P. Sapienza, and L. Zingales, 2003, “People's Opium? Religion and Economic Attitudes,” Journal of Monetary Economics, 50, 225252.
  • Hamilton, B.H., 2000, “Does Entrepreneurship Pay? An Empirical Analysis of the Returns to Self-Employment,” Journal of Political Economy, 108, 604631.
  • Hogfeldt, P., 2004, “The History and Politics of Corporate Ownership in Sweden,” NBER Working Paper 10641.
  • Holtz-Eakin, D., D. Joulfaian, and H.S. Rosen, 1994, “Sticking It Out: Entrepreneurial Survival and Liquidity Constraints,” Journal of Political Economy, 102, 5375.
  • Hong, H., J.D. Kubik, and J.C. Stein, 2004, “Social Interaction and Stock Market Participation,” Journal of Finance, 59, 137163.
  • Hurst, E. and A. Lusardi, 2004, “Liquidity Constraints, Household Wealth and Entrepreneurship,” Journal of Political Economy, 112, 319347.
  • Klapper, L., L. Laeven, and R. Rajan, 2006, “Entry Regulation as a Barrier to Entrepreneurship,” Journal of Financial Economics, 82, 591629.
  • Knack, S. and P. Keefer, 1997, “Does Social Capital Have an Economic Payoff? A Cross-Country Investigation,” Quarterly Journal of Economics, 112, 12511288.
  • Krugman, P., 1991, Geography and Trade, Cambridge , MA : The MIT Press.
  • Licht, A. and J.I. Siegel, 2006, “Social Dimensions of Entrepreneurship,” in Casson  and B.Yeung, eds., Oxford Handbook of Entrepreneurship, Oxford , UK : Oxford University Press, 511538.
  • Malmendier, U. and G. Tate, 2005, “CEO Overconfidence and Corporate Investment,” Journal of Finance, 60, 26612700.
  • Manski, C.F., 1993, “Identification of Endogenous Social Effects: The Reflection Problem,” Review of Economic Studies, 60, 531542.
  • Moffitt, R.A., 2001, “Policy Interventions, Low-Level Equilibria, and Social Interactions,” in S.N.Durlauf and H.P.Young, eds., Social Dynamics, Cambridge , MA : The MIT Press, 5482.
  • Moskowitz, T.J. and A. Vissing-Jorgensen, 2002, “The Private Equity Premium Puzzle,” American Economic Review, 92, 745778.
  • Moulton, B.R., 1990, “An Illustration of a Pitfall in Estimating the Effects of Aggregate Variables on Micro Units,” Review of Economics and Statistics, 72, 334338.
  • Murphy, K.M., A. Shleifer, and R.W. Vishny, 1991, “The Allocation of Talent: Implications for Growth,” Quarterly Journal of Economics, 106, 503530.
  • Nanda, R. and J.B. Sorensen, 2007, “Peer Effects and Entrepreneurship,” Mimeo, Stanford Graduate School of Business.
  • Saxenian, A., 1994, Regional Advantage: Culture and Competition in Silicon Valley and Route 128, Cambridge , MA : Harvard University Press.
  • Schumpeter, J., 1934, The Theory of Economic Development, Cambridge , MA : Harvard University Press.
  • Shane, S., 2007, The Illusions of Entrepreneurship, New Haven , CT : Yale University Press.
  • Sorenson, O. and P.G. Audia, 2000, “The Social Structure of Entrepreneurial Activity: Geographic Concentration of Footwear Production in the United States, 1940–1989,” American Journal of Sociology, 106, 424461.
  • Vlachos, J., 2004, “Who Wants Political Integration? Evidence from the Swedish EU-Membership Referendum,” Journal of Public Economics, 88, 15891604.
  • Weber, M., 1905, The Protestant Ethic and the Spirit of Capitalism, London , Unwin.
  • Wellman, B., 1996, “Are Personal Communities Local? A Dumptarian Reconsideration,” Social Networks, 18, 347354.