Download 2nd Paper

Non Pecuniary Benefits of Small Business Ownership Erik G. Hurst University of Chicago [email protected]

Benjamin Wild Pugsley University of Chicago [email protected]

July 2010

PRELIMINARY Abstract We find non pecuniary benefits of business ownership to be an important consideration in explaining small business formation. Using a variety of data sources, we show that the overwhelming majority of small businesses do not match the theoretical concept of an entrepreneur. Specifically, nearly all small businesses do not grow, do not want to grow, do not innovate, and do not want to innovate. Second, using new data from the Panel Study of Entrepreneurial Dynamics, we show that a majority of small business owners report that non pecuniary benefits (e.g., being one’s own boss, having flexibility over one’s schedule) were the primary reason that they started their business. Third, we embed non pecuniary benefits into a standard model of occupational choice that abstracts from other traditional channels of risk, ability, and liquidity constraints. This simple model generates predictions consistent with many aspects of the data: we find that the propensity to start a small business is still positively correlated with individual wealth despite the omission of binding capital constraints; small businesses will be concentrated in industries with lower returns to scale; and we show a wage firm-size premium even without assumed differences in productivity among firms or workers. We also show that the existence of non pecuniary benefits reduce the aggregate welfare gains from subsidizing the formation of small businesses, but that the aggregate measure obscures a stark redistribution—subsidizing small businesses in a world with non pecuniary benefits and lump sum taxes disproportionately helps the rich at the expense of the poor. In the final section of the paper, we quantify the importance of non pecuniary benefits by calibrating the model to micro data. Collectively, our results show that one needs to take seriously non pecuniary benefits when crafting models of small business formation.

1

Introduction

Economists and policy makers alike have have long been interested in the effects of various economic policies on business ownership. For example, academic work (see Cullen and Gordon (2007) and Cagetti and De Nardi (2009)) evaluates the implications of various tax regimes on business formation. For policy makers, a key consideration of any potential legislation is the impact on small business owners. Just recently, policy makers advocating health care reform justified reform as promoting entrepreneurial activity by “reducing the [health care] burden on small firms and their workers.”1 In fact, the U.S. Small Business Administration is a federally funded agency whose purpose is to help Americans “start, build, and grow businesses.” Researchers and policy makers (often implicitly) equate all small business owners with “entrepreneurs.” While some may view this as a tautology, economic models have a very specific notion of an entrepreneur. The theory considers entrepreneurs as individuals who (1) innovate and render aging technologies obsolete (Schumpeter, 1942), (2) take economic risks (Knight (1921), Kihlstrom and Laffont (1979), Kanbur (1979), and Jovanovic (1979)), or (3) are considered jacksof-all-trades in the sense that they have a broad skill set (Lazear, 2005). Most of the empirical work on entrepreneurship measures entrepreneurs by sampling small business owners or the selfemployed.2 Then it is important to assess whether the universe of small business owners behaves consistently with the theoretical concept of an entrepreneur. Our goals in this paper are fourfold. First, using a variety of different data sets which track the growth and innovation behavior of a large representative sample of small businesses, we show that most small businesses do not match the theoretical concepts of an entrepreneur outlined above. Our results show that the vast majority of surviving small business owners never grow in any substantive way nor do they want to grow in any substantive way. For example, we find that only roughly ten percent of newly established surviving small businesses added more than five employees during their first four years of business—only three percent added more than ten employees. Even when asked (at the time of inception) about their desired firm size in five years, more than 75 percent of new business owners prefer a small number of employees. We also use a variety of data sources to show that very few small business owners innovate along any observable dimensions. Again most new business owners also report not wanting to innovate along a variety of observable measures. To put some structure on these results, we also show that most small businesses are concentrated in a few industries. The overwhelming majority of small businesses in the U.S. are either skilled craftsmen (plumbers, electricians, etc.), professional service providers (doctors, lawyers, accountants, etc.), real estate agents, insurance agents, small shop keepers (hardware store owners, convenient store owners, etc.), or small service provider (barbers, dry cleaners, auto repair mechanics, etc.). Nearly all of these businesses start small and remain small throughout their entire 1

See http://www.whitehouse.gov/administration/eop/cea/Health-Care-Reform-and-Small-Businesses/. See, for example, Evans and Jovanovic (1989), Evans and Leighton (1989), Quadrini (1999), Hurst and Lusardi (2004), Gentry and Hubbard (2004), Cagetti and De Nardi (2006), Fairlie and Woodruff (2007). Some other offers equate entrepreneurs with those individuals who file Schedule C tax forms. See, for example, Holtz-Eakin, Joulfaian, and Rosen (1994). 2

1

lifecycle. When people equate entrepreneurs with small business owners, on average, they are equating entrepreneurs with one of the small scale businesses listed above. When viewed through this lens, it is not surprising that most small businesses do not grow, do not want to grow, do not innovate and do not want to innovate. If most small businesses do not want to grow or do not want to innovate, why do they start? The second goal of the paper is to explore this question. To do so, we use new data from the Panel Study of Entrepreneurial Dynamics (PSED), which surveys nascent business owners about their decision to start their business. Of these new business owners, over 50 percent of them reported that non pecuniary benefits were the primary reason as to why they started their business.3 By comparison, only 34 percent of respondents reported that they were starting the business to generate income, only 40 percent indicated that they were starting a business because they wanted to create a new product or because they had a good business idea, and only 4 percent reported starting a business because they had no other employment options.4 These percentages were constant across business owners of varying degree of business activity.5 Using the panel nature of the PSED, we show that those small businesses that started for non pecuniary reasons were also much less likely to subsequently grow, were much less likely to report wanting to grow, were much less likely to subsequently innovate, and were much less likely to report wanting to innovate. These results suggest that non pecuniary benefits are an important driver of a significant fraction of small business activity. Such preferences can reconcile why small businesses are formed by many individuals despite the lack of interest in subsequent growth or innovation. Additionally, the importance of non pecuniary benefits associated with small business ownership is consistent with two recent empirical studies. Hamilton (2000), using panel data from the Survey of Income and Program Participants (SIPP), finds that lifetime earnings fall substantially as individuals transition from wage employment into self employment. Moskowitz and Vissing-Jørgensen (2002) aggregate a variety of different data sets to construct an index of private equity investment and find that even with the undiversifiable risk small business owners face, they command no expected return premium over investing in a well diversified public equity portfolio. Both papers suggest that the non pecuniary benefits of owning a business could explain their puzzling results.6 3

Such non pecuniary included: “wanted to be my own boss”, “wanted to have a flexible schedule”, “wanted to work from home”, or “wanted to make my hobby a career”. 4 Unlike many developing labor markets, business ownership out of necessity explains a relatively small proportion of small businesses in the U.S. 5 Note, the percentages exceed 100 percent because respondents could report up to two reasons as to why they started their business. The percentages sum to less than 200 percent because some respondents did not report multiple responses and of those that did report multiple responses, they were often in the same broad category. 6 Hamilton (2000) treats the reported wage earnings of small business owners as being the actual wage earnings of small business owners. This assumption, however, is not appropriate if small business owners under report their earnings when responding to household surveys. Recent work by Hurst, Li, and Pugsley (2010) estimate consumption demand systems for respondents in the Panel Study of Income Dynamics and the Consumer Expenditure Survey and find that small business owners under report their incomes by roughly 20 - 30 percent. However, Hamilton also ignores the part of compensation due to fringe benefits for traditional wage and salary workers. Estimates of such benefits range of 20 - 30 percent of income. The two effects—ignoring the mis-reporting of income by the self employed and the receipt of fringe benefits by wage workers—likely roughly offset. Taken together, the bias in the Hamilton results from simultaneously ignoring potential income under-reporting and fringe benefits is small. Moskowitz and Vissing-Jørgensen (2002), on the other hand, incorporate a significant under-reporting of income by the self employed

2

The third goal of our paper is the most important contribution. We theoretically explore the implications of embedding non pecuniary benefits of small business ownership into a general equilibrium model of entrepreneurial choice. To our knowledge, no one has ever put forth a model of small business formation where households care about the non pecuniary benefits of owning a business. Within the model, many households differ in their initial endowments of wealth as well as in their preference for small business ownership. All households are equally capable workers and inelastically supply their labor to firms or managing their own businesses. The households generate demand for many differentiated goods, which differ by efficient scale implied by their technological requirements. Free entry ensures that each good may be produced by a firm or a household run business. Each household must decide whether to start a business producing a particular good or seek employment at an existing firm. We evaluate the equilibrium consequences of this configuration and its sensitivity to the distribution of preferences and wealth. When crafting the model, we omit the traditional channels of innovation, risk taking, and productivity advantages highlighted by others to explain small business formation. The only reason that individuals work in the small business sector is because they have a high utility flow from being a small business owner or because they have enough initial wealth to make the marginal utility of consumption of goods lower. We do this for two reasons. First, we do so to highlight the implication of our mechanism by showing results when all the other mechanisms are turned off. Second, given the results documented in the first two parts of the paper, we feel that these other mechanisms are less important for explaining the entry decisions for most actual small businesses within the United States. Despite its parsimonious nature, the model yields many interesting predictions which calls into question the interpretation of many established empirical results within the literature. First, in a model in which there are no financial frictions, we can generate a positive relationship between household wealth (or exogenous changes in household wealth) and business entry. The empirical relationship between wealth (or changes in wealth) and business entry has commonly been cited as evidence that small businesses face liquidity constraints.7 However, in our model similar relationship between wealth and business entry is generated purely from income effects. As households become richer, they want to purchase more of the goods that increase utility where one such good is the utility flow from owning a business. The results in this paper suggest that a positive relationship between exogenous changes in wealth and small business formation is not, by itself, prima facie evidence of the existence of liquidity constraints facing small businesses. Second, the model generates differences in small business propensities across different types of goods and services. The key cost associated with starting a business in our model is the lost productivity that results from producing the good or service in a small firm as opposed to producing the good or service at lower average cost in a larger firm. In equilibrium, goods whose production process displays increasing returns to scale over a larger region will be less likely to be produced into their analysis. 7 See, for example, Evans and Leighton (1989), Evans and Jovanovic (1989), Quadrini (1999), Gentry and Hubbard (2004), Cagetti and De Nardi (2006), Fairlie and Woodruff (2007), Fairlie and Krashinsky (2006). See Hurst and Lusardi (2004) for a critique of the empirical results in this literature.

3

by small firms compared to goods whose production process quickly reach diminishing returns to scale. In other words, it is easier to get the utility gains from running small business as a carpenter than as an auto manufacturer. Third, the model yields a skewed distribution of firms toward really small firms and away from other smaller and medium sized firms. In the absence of non pecuniary business owners, all goods would be produced by firms operating at their their natural efficient scales. In our model, this varies directly with the type of good produced. Introducing non pecuniary benefits truncates the distribution of firms in the sense that small and medium scale business cannot compete against the non pecuniary entrepreneur who has a firm size of one. Though potentially less productive, the entrepreneur is willing to work at a lower wage by taking some of total compensation in the form of direct utility flow. This effectively lowers his marginal costs below his competitors. Smaller and medium scale firms cannot compete and shut down. Fourth, the model yields a relationship between firms sizes and firm wages. It has been well documented that larger firms pay higher wages.8 One potential explanation for this finding is that larger firms are more productive and more capable workers match with those firms. Our model can also generate this result. However, even if we made household-operated firms more productive than labor employed at firms producing the same goods, the implied small business wages would still be lower than the larger firms. With many households competing against each other to run small businesses, they would drive down the pecuniary return of business ownership to equate the total compensation across the two sectors. As a result, the self employed worker in a really small firm will earn lower wages, all else equal, than he would have earned had he been employed in a larger firm. Fifth, the model shows that measured labor productivity in the economy is linked to the distribution of non pecuniary benefits in the economy. The higher the level of non pecuniary benefits, the lower the level of measured labor productivity. An increase in the non pecuniary benefits of working for one’s self will shift people away from the firm sector towards the self employed sector. While such a reallocation will be utility maximizing, it will lower labor productivity in the sense that individuals will forgo the benefits of working in firms that can exploit potential returns to scale. Finally, we show that with such preferences, subsides to small businesses can be welfare reducing. In our model where the only reason for small business formation is because of the non pecuniary benefits, subsidies to promote small businesses are strictly welfare reducing even if they are funded by lump sum taxes. This is not surprising given there are no frictions in which the subsidy is trying to overcome. However, it does suggest that any benefits of a subsidy designed to stimulate small business formation will have to be weighed against the cost of stimulating small business activity for those who are only interested in the non pecuniary benefits. Of greater interest, however, is the fact that the model finds that subsidies to small business can be regressive in the sense that they can make richer households better off at the expense of making poorer households worse off. This comes directly from the income effect associated with the non pecuniary desire to start a small 8

See Brown and Medoff (1989).

4

business. In the model, richer households will start businesses even in a world with no subsidy. For them the subsidy is a pure transfer. The poorest households are the least likely to start a business even with strong subsidies. Such redistributive costs need to be considered when thinking about the decision to subsidize small business formation in a world where a market imperfection may stifle entrepreneurial activity. The key point we wish to stress is that the existence of non pecuniary benefits driving the decision to start a small businesses changes could potentially alter the perceived benefits of subsidizing small business broadly. Our fourth and final goal of the paper is to use a variety of micro data sets to calibrate the model and quantify the importance of non pecuniary benefits in explaining small business formation (this section under construction). Taken as a whole, our results highlight the fact that the non pecuniary benefits associated with small business formation are both theoretically and empirically important. However, before proceeding, we just want to stress that we do not believe that non pecuniary benefits are the primary driver of all business start-ups. We know for a fact that nearly all large firms started out as small firms. It is just that most small firms will never end up having more than a handful of employees (nor do they want to have more than a handful of employees). To test the concepts of entrepreneurs that are usually spelled out in our theoretical models, one should use more specialized data sets as opposed to using the universe of small businesses. As shown by the work of Kaplan and Lerner (2009) and Hall (2010) firms that receive venture capital funding or those that progress to the IPO stage better match the concept of an “entrepreneur” set out in economic models. These firms, however, represent less than one-half of one percent of all new businesses within the U.S. Additionally, we want to stress that studying entrepreneurship (as usually spelled out in our theoretical models), or innovation more broadly, is a very worthy venture. Again, the point we are making in this paper is that empirically most business owners do not match the concepts of entrepreneurs that we usually spell out in theoretical models. Moreover, as public policy, we tend to promote entrepreneurship by subsidizing small business owners, for example, by linking regulatory exemptions or loan guarantees to firm size. Understanding the forces that drive business ownership, we argue, will only help to make public policies with respect to promoting innovation and economic growth more efficient. Finally, to understand the importance of financial frictions as a deterrent to innovation, the nature of the distribution of firms size (overall and by industry), and the relationship between small business subsidies and household well being, one needs to think hard about why all small businesses form.

2

Background Facts about Small Businesses

In this section, we will show that small businesses do not grow, do not want to grow, do not innovate and do not want to innovate. Throughout this section, we bolster our claims using data from the 1989-1997 Statistics of U.S. Businesses, the 2006-2008 Panel Study of Entrepreneurial Dynamics, the 2004-2008 Kauffman Firm Survey, the 2003 National Survey of Small Business Finances, and 2005 U.S. Business Dynamic Statistics. When we first use these data sources, we will briefly describe 5

the data. In the Data Appendix, we discuss these data sets (and all data sources used in the paper) in much greater detail.

2.1

Most Small Businesses are Concentrated in a Few Industries

Before discussing the innovation and growth behavior of most small businesses, we start by showing the distribution of small businesses in total and by industry. Many of the results in this subsection will be used to motivate theoretical choices we make in Section 3. Additionally, it provides background on the population of small firms. Such information can be used to inform priors about the extent to which these firms should be expected to either grow or innovate. Panel A of Table 1 shows the distribution of small businesses within the U.S. from the 1989-1997 Statistics of U.S. Businesses (SUSB) in the aggregate and for various broad industry groupings.9 In particular, we segment businesses as being in the following industries: general services (nonprofessional services, non-health services, non-arts/entertainment services); construction (general contractors, plumbers, electricians, painters, wall papers); professional services (lawyers, accountants, architects); retail trade (convenient stores, hardware stores, book stores); FIRE (insurance agents, real estate agents, mortgage brokers); wholesale trade, entertainment services (movie theaters, transportation, warehouse and communication (trucking companies, freight companies), health care services (doctors, dentists, social workers), accommodation and food services (restaurants, hotels), manufacturing, and agriculture and mining (landscaping companies, veterinary services).10 The first column of Panel A of Table 1 shows the fraction of small businesses in total and by industry. Throughout the paper, we define small businesses as those businesses with employment less than 20 employees.11 For example, 71.2 percent of all firms in the U.S. are small businesses. There are big differences in such percentages across industries. 85.5 percent of all firms in the construction industry have less than 20 employees while only 50.0 percent of all firms in the manufacturing have less than 20 employees. The second column shows the amount of employment within each industry that is in small businesses. Across all industries, 20 percent of all employment in the U.S. is in firms with less than 20 employees. Nearly 40 percent of the employment within the construction and general service industries is in firms with less than 20 employees while only 7.0 percent of all employment within the manufacturing industry is in firms with less than 20 employees. These numbers are lower bounds in that they exclude non-employer firms. The third column of Table 1 shows the fraction of all businesses with less than 20 employees within a given industry relative to the total amount of businesses with less than 20 employees in the entire economy. Of firms with less 9

The SUSB covers most economic activity within the U.S. although it does exclude data from non-employer businesses, private households, railroads, agricultural production, and most government activities. The data are collected by the U.S. Census Bureau, in cooperation with the Office of Advocacy of the U.S. Small Business Administration (SBA). Our choice of the 1989-1997 period is not important to our results. Similar results are found using data from earlier or later periods. All the SUBS data can be found at http://www.census.gov/econ/susb. 10 For a full description of how the specific 2-digit SIC codes that map into our industry classifications, see Appendix Table A1. 11 Unless otherwise noted, we use the terms “firms with less than 20 employees” and “small businesses” interchangeably.

6

than 20 employees, nearly 75 percent of these firms are in the general service, professional service, health care services, retail trade, construction, or the finance, insurance, and real estate (FIRE) industries. Only 3.6 percent of small businesses are in the manufacturing industry.12 Appendix Table A3 expands on this latter point in much greater detail. In particular, Appendix Table A3 shows the 25 three-digit industries with the highest share of firms with less than 20 employees out of all firms with less than 20 employees.13 Between 1989 and 1997, 50 percent of all small businesses in the U.S. are in one of these 25 three-digit SIC industries. As seen from this table, when one talks about the median small business in the United States they are essentially talking about doctors, dentists, lawyers, restaurants, accountants, insurance agents, real estate agents, plumbers, electricians, general contractors, grocery store owners, beauticians, or mechanics.14 Do the patterns shown for established businesses in the SUSB hold for newly established firms? In Panel B of Table 1, we explore the answer to this question. To do this, we use data from the 2006-2009 Panel Study of Entrepreneurial Dynamics (PSED).15 Using a nationally representative sample of 34,000 individuals during the fall of 2005 and the early winter of 2006, 1,214 “nascent entrepreneurs” were identified. To be considered a nascent entrepreneur, individuals had to meet the following four criteria. First, the individual had to currently consider themselves as involved in the firm creation process. Second, they had to have engaged in some start up activity in the past twelve months. Third, they had to expect to own all or part of the new firm. Finally, the initiative, at the time of the initial screening survey, could not have progressed to the point that it could have been considered an operating business. The goal was to sample individuals who were in the process of establishing a new business. In the winter of 2006, after the initial screening interview, the 1,214 respondents that had been initially identified as being in the process of starting a business were surveyed about a wide variety of the activities associated with their business start up. As part of the first real interview, respondents were asked detailed questions about their motivation for starting the business, the current activities undertaken as part of the start up process, the competitive environment in which the business would take place, and their expectation of desired future business size and activities. Follow up interviews were done for the survivors in the winter 12

For reference, Appendix Table A1 shows the share of firms and the share of employment in each industry regardless of firm size. 13 To do this, we compute the share of firms with less than 20 employees in the SUSB data described above within each three-digit industry relative to all firms with less than 20 employees. We then rank the three-digit industries by this shares. There are 379 distinct three-digit SIC industries in the SUSB data. 14 An additional 30 3-digit SIC industries contains an additional 25 percent of all small businesses (implying that only 55 out of the 379 three-digit SIC industries contain 75 percent of all businesses with less than 25 employees). These additional 3 digit industries primarily include those in the construction industry (e.g., carpenters, stone masons, painters, wallpaper hangers, and roofers), the general service industry (e.g., janitorial services, laundry and dry cleaners, other repair shops, and copy and mail stores), and the retail industry (e.g., furniture and home furnishing stores, gas stations, liquor stores, auto and home supply stores, and drug stores). 15 An earlier wave of the PSED (PSED I) is also available for analysis. The PSED I interviewed nascent business owners during the 1998-2000 period. This earlier wave had a much smaller sample. The PSED documentation recommends using the PSED II over the PSED I given that they made many substantive improvements to the survey design between the first and second waves. As a result, all of our analysis is done with the second wave of data (PSED II). For simplicity, we will drop the notation that indicates what wave we are using and just refer to the data for our analysis being from the PSED. All the documentation and data for the PSED can be found a http://www.psed.isr.umich.edu/psed/home.

7

of 2007 and the winter of 2008. In Panel B of Table 1, we summarize the industries in which these firms were starting their business. As seen from Panel B of Table 1, nearly 75 percent of all new small businesses were started in the general service, professional service, health care services, retail trade, construction, or the finance, insurance, and real estate (FIRE) industries. Only 6 percent were started in the manufacturing industries. These numbers are similar to the data on established employer firms within the SUSB. Taken together, the results in Table 1 and Appendix Table A3 illustrate two facts about small businesses. First, some industries have a much higher concentration of small businesses than other industries. This applies for both established small businesses as well as for newly started small businesses. Second, when we talk about small businesses, we are primarily talking about health professionals (doctors and dentists), non-health professionals (lawyers, accountants and architects), real estate and insurance agents, shop keepers, skilled craftsmen (plumbers, electricians, carpenters), restaurateurs, and general service providers (auto repair shops, beauty shops, dry cleaners). Such a reference point will prove useful when we show in the subsequent sections that most small businesses do not grow, do not want to grow, do not innovate, and do not want to innovate.

2.2

Most Small Businesses Do Not Grow

Tables 2a and 2b show data from the 2005 Business Dynamic Statistics (BDS). Like the SUSB data discussed above, the BDS is also produced by the U.S. Census Bureau. The BDS is designed to provide summary statistics about the distribution of firm size for firms of different age and within different industries. Again, like the SUSB, the data only tracks the employment patterns of employer firms.16 Table 2a shows the percent of establishments within different firm age categories that are firms with less than 20 employees. We do this for the entire economy (top row) and then separately within different one digit industry classifications.17 The way the table should be read is as follows. Of all firms within the economy that have been in existence between 0 and 10 years, 89.9 percent of those firms have less than 20 employees. Additionally, of all firms within the construction industry that have been in existence for less than 10 years, 93.6 percent of them have less than 20 employees. Table 2b shows similar patterns except showing the share of employment in firms with less than 20 employees as opposed to the share of establishments. For example, within firms in the construction industry that have been in existence for less than 10 years, 57.2 of employment in such firms is in firms with less than 20 employees. There are two things we wish to point out from Tables 2a and 2b. First, among mature firms (firms in existence between 10 and 25 years), most firms and much of the employment are in small firms. For example, across the economy as whole, nearly 74 percent of all firms that have been in existence between 10 and 25 years are firms that have less than 20 employees. Even well into their lifecycle, the overwhelming majority of firms remain small. A substantial share of employment in mature firms is in firms with less than 20 employees. Second, as we highlighted above, there 16 17

The BDS data can be found at http://www.ces.census.gov/index.php/bds/bds_database_list. The one digit categories shown in Table 2 are the lowest level of aggregation provided by the BDS.

8

is substantial variation among industries. Relative to construction, very little of employment of mature firms is in small businesses within the manufacturing industry. Additional industries that include a high concentration of the employment of mature firms being in small businesses include the FIRE, wholesale trade, retail trade and service industries. This is consistent with the fact that most small businesses are skilled craftsmen, professionals such as lawyers and doctors, small shop keepers, restaurants, and real estate and insurance agents. Tables 2a and 2b focus on static cross sections and do not directly explore the employment dynamics of small firms. To shed light on this, we use data from the 2003 National Survey of Small Business Finances (NSSBF). The NSSBF is a random sample of businesses with less than 500 employees and is conducted by the Board of Governors of the U.S. Federal Reserve. The survey is designed to measure the financial position of these businesses. However, the survey also contains other background questions. In 2003, firms were asked to state whether the total employees within their business grew, remained the same, or contracted since the previous year (2002). Likewise, firms were also asked to state whether their total employees within the business grew, remained the same, or contracted since three years earlier (2000). The responses to these questions by small firms are shown in Table 3a. Like above, we define small firms as those firms with less than 20 employees. We break down the responses by firm age to try to highlight differences between newer businesses and more established businesses. The NSSBF asks businesses to report how long the business has been in existence. As seen from the table, the overwhelming majority of small firms do not grow year to year or even over three year periods. Not conditioning on firm age, only 14.0 percent of surviving small businesses grew between 2002 and 2003 and only 21.3 percent grew 2000 and 2003. Taking the converse, roughly 80 percent of surviving small firms did not grow at all over a relatively long periods. The percentages are slightly higher among newer firms. However, even among small firms which have been in existence between 1 and 10 years, only 18.9 percent grew between 2002 and 2003 and only 27.6 percent grew between 2000 and 2004. The NSSBF data does show that some firms did grow (even though most did not). But among the growers, the NSSBF data does not tell us how much they grew. To assess this question, we turn to the Kauffman Firm Survey (KFS). The KFS is a panel study of 4,928 businesses that were newly founded in 2004. These firms are currently being followed through time to see how these firms evolved since their inception. Currently, data is available on these firms up through 2008. For the work below, we only focus on those firms that have survived up through 2008. There were 2,617 such firms in the data.18 Because the KFS is a four year panel, we can assess the four year growth rate of businesses within the KFS. In each wave of the survey, the KFS asks firms to report the number of their employees. Between 2004 and 2008, 41.9 percent of the surviving firms in the KFS reported growing the total number of employees within their firm. In columns 2 and 3 of Table 3b, we show the fraction of firms who added more than 5 employees between 2004 and 2008 and the fraction of firms who added more than 10 employees between 2004 and 2008. While many of the surviving new firms with the 18

All data for the KFS can be found at http://www.kauffman.org/kfs/.

9

KFS added employees, very few added more than one or two employees. For example, only 10 percent of all surviving firms added more than 5 employees during this time period and only 3.6 percent of all surviving firms added more than 10 employees. While many of these small firms in the KFS did add at least one employee, hardly any of them increased the number of employees within their firm by more than a few employees. The results in Tables 2 and 3 suggest two things. First, most small businesses do not grow at all over long periods. Second, for most of those that do add employees, the number of employees they add is relatively small. While it is almost certain that all big firms started as small firms. The converse is not true. Very few of small firms end up big. The statement is even stronger. Very few small firms grow in any significant way at all. The results above are almost certainly lower bounds given that all of our analysis selects on firms that have survived.19

2.3

Most Small Businesses Report Not Wanting To Grow

Is the reason that most small businesses do not grow because they are constrained from growing? Table 4 casts doubt on this potential claim. Most businesses owners report that they do not want to grow. Respondents within the PSED were asked about their desired future firm size. Specifically, new firms were asked the following: “Which of the following two statements best describe your preference for the future size of this new business: ‘I want this new business to be as large as possible’ or ‘I want a size I can manage myself or with a few key employees”’. The top row of Table 4 shows the response to this question by PSED respondents. We look at two samples with the PSED. The first is a sample of all PSED respondents. This is the same as the sample used in Panel B of Table 1. The second sample is those respondents who actually had positive revenues at the time of the first PSED interview. This latter sample distinguishes people who only said that they were planning to start a business from those who actually followed through and engaged in some market business activity. Nearly three quarters of all respondents—regardless of sample—reported they wanted to keep their business small. This is also consistent with their expectations about their future business size. In a separate part of the survey, the respondents were asked to provide their expectation as to the number of employees that the firm would employ when the firm was 5 years old. The median number of employees was between 3 or 4, depending on the sample. Even the 75th percentile of responses was small as respondents only expected to employ between 8 and 10 employees. Not only do very few small businesses grow, most of them do not want or expect to grow when they form their new 19

The fact that very few small businesses grow is consistent with recent research by Puri et al. (2010). In that paper, the authors show that the small businesses that tend to grow are the ones that receive venture capital (VC) funding. According to Kaplan and Lerner (2009), roughly 1,000 small businesses receive their first VC funding during a typical year. Given that roughly 600,000 new businesses that employ others are started each year, Kaplan and Lerner (2009) conclude that less that 1/6th of 1 percent of new businesses obtain VC funding. If it is truly the VC back companies are the small businesses that better match the theoretical conception of an entrepreneur, it would be hard to identify such companies in standard data sets that either include the universe or a random sample of all firms. These truly entrepreneurial firms represent such a small fraction of all firms that researchers would learn little about ”entrepreneurs” from studying small businesses broadly.

10

business.20

2.4

Most Small Businesses Do Not Innovate

Additionally, the results in Tables 5a and 5b show that most firms do not innovate, at least according to easily observable measures. The KFS survey asks respondents to report separately whether they have already applied or are in the process of applying for any patents, copyrights, or trademarks. Respondents are asked to report their patent, copyright, and trademark application process since the business has been in existence. These results are shown in Table 5a. Within the first four years of business, only 2.7 percent of businesses have applied or were in the process of applying for patents. Copyright and trademark behavior is slightly higher. Yet, still the overwhelming fraction of new businesses does not apply for patents, copyrights or trademarks. The PSED asks similar question of its respondents. These results are shown in Table 5b. However, instead of asking about the application process for patents, trademarks, and copyrights separately, respondents are asked about all three in the same question. As seen in Table 5b, only 5 to 6 percent of PSED respondents reported applying for a patent, trademark, or copyright, depending on the sample. The numbers in the PSED are likely smaller than the KFS results because the KFS results come from firms who have been in existence for four years. The PSED results are for firms who only have been in business for less than one year. In a separate question, the PSED asked businesses whether they have developed any proprietary technology, processes, or procedures. This is a slightly broader measure of innovation. Yet, only 6.5 percent of small businesses report than they have developed any proprietary business practices or technology. Additionally, PSED respondents were asked the following question: “Right now, are there many, few, or no other businesses offering the same products or services to your customers?” Respondents provided were allowed to give the following answers: many, few, or no other.21 For the full sample, 36 percent of respondents reported that there were many existing firms already providing the same products or service to their expected customer base. Less than 20 percent of respondents reported that no one other business was provided their expected product or service to their expected customer base. For the sample of respondents who actually started production, 43 percent reported that there were many existing firms already providing the same product or service to their expected customer base and only 13 percent reported that they we either offer a novel product or service or targeting a new customer base. Collectively, the results in Tables 5a and 5b show that very few small businesses innovate along any of a variety of dimensions. The vast majority of small businesses (upwards of 80 or 90 percent) report no innovative activity at all. 20 As we will caveat often throughout the paper, we are not saying that there are no small businesses that want to grow or that no small business will actually grow substantially. We are just saying that the overwhelming majority of small businesses will not grow substantially, nor do they want to grow substantially. This paper is focusing on these businesses. 21 Respondents were also allowed to answer “do not know” or “not applicable”. Only 12 of the 1,214 responded that they did not know and there were no responses of “not applicable”.

11

2.5

Most Small Businesses Report Not Wanting To Innovate

The above results are even stronger in that most new small businesses report that they do not plan to do any innovative activity at all. The PSED also asks about expected innovative activity. Table 5b shows these results. For example, only roughly 10 percent of all new businesses reported that they plan to develop proprietary technology, processes, or procedures in the future. The numbers are slightly higher with respect to business’s patent, copyright and trademark behavior. This is likely because many firms trademark the name of their business. The results in Table 5b suggest that not only do new businesses not innovate; they do not plan to innovate in the future.

2.6

The Importance of Non Pecuniary Benefits in Small Business Formation

The vast majority of U.S. small businesses do not grow, do not want to grow, do not innovate, and do not want to innovate. These results are consistent with casual observations about the innovative and growth patterns of plumbers, real estate agents, book store owners, lawyers, and doctors. As shown above, these occupations make up the bulk of small businesses within the U.S. These results, however, beg the question as to why small businesses form. Before attempting to answer that question, one needs to combine the results in the previous section with the results discussed in the introduction. Recent empirical research has shown that the median small business owner earns less as a small business owner than they would have as a wage worker (see Hamilton (2000)) and that the mean risk adjusted return to business ownership is less than the mean risk adjusted return to investing in public equities (see Moskowitz and VissingJørgensen (2002)). The fact that the median small business owner is taking a loss in pecuniary benefits by becoming a small business owner suggests the importance of non pecuniary benefits in small business decisions. The existence of non pecuniary benefits could explain why many individuals enter small businesses despite the fact they they will not grow, do not plan to grow, will not innovate and do not plan to innovate. To explore the importance of non pecuniary benefits in business entry decisions, we again turn to the PSED data. As part of the initial survey of the PSED, the business owners were asked “Why did you want to start this new business?”. The respondents provided unstructured answers and the PSED staff coded the answers into 44 specific categories. All the categories are listed in Appendix Table A4, along with the number of all PSED respondents who provided the reason on either their first report (in the first parentheses) or on their second report (in the second parentheses). We took the raw responses to the question ”Why did you start your business” and created five broad categories of our own.22 The five categories were: (1) non pecuniary reasons, (2) reasons related to the generation of income, (3) reasons related to the desire to develop a new product or 22 The PSED grouped the responses into 6 broad categories. Those categories were: ”income reasons”, “business opportunity reasons”, “employment reasons”, “personal reasons”, “lifestyle reasons”, and “other reasons”. For highlighting the importance of non pecuniary benefits with respect to starting a business, the broad classification put forth by the PSED staff was not ideal. For example, the “personal reason” category creating by the PSED staff included both the responses “have lots of experience at this type of work” (which may affect the pecuniary returns to starting the business) and the response “enjoy this type of work, its my hobby” (which may affect the non pecuniary returns to starting the business).

12

because they had a good business idea, (4) reasons related to the fact the respondent has no better job options, and (5) all other reasons. The main responses in the non pecuniary category include: “want to be my own boss”, “want flexibility over my schedule”, “want to work from home”, “enjoy the work/it is my hobby”. The main responses in the generating income category include: “to make money” or “need extra income”. The main responses in the new product/business idea category include: “satisfy a business need”, “there is high demand for this product/business”, “untapped market”, and “lots of experience at this type of work”. A full breakdown of our classification of the raw responses into these five broad categories can also be found in Appendix Table A4. Table 6 provides the distribution of first responses by category (column A) and the distribution of either the first or second response by category (column B) for the two PSED samples discussed above. Three preliminary things should be noted about the results in Table 6. First, only 60 percent of respondents provided a second response. Second, given that the respondent could provide any answer they wanted, the first and second response often fell into the same broad category (e.g., answer 1 was ”be own boss” and answer 2 was “have flexibility over schedule”, both of which we count as being a non pecuniary benefit of starting a business). Third, summing down column A exactly equals 100 percent while summing down column B exceeds 100 percent given that respondents could report a second answer. The main result from Table 6 is that those starting a small business report that non pecuniary benefits are an important driver of their business entry decision. These results are consistent across both PSED samples. For example, 35.3 percent of all the first reports from respondents in the PSED referred to non pecuniary reasons being an important driver of their decision to start a business. Only 32.2 percent of respondents reported that they were starting their business because they had a good business idea or that they wanted to create a new product. Aggregating over their two responses, half of all PSED respondents reported that a non pecuniary reason was the primary reason that they started the business. The similar numbers for ”income reasons” or ”good business idea reasons” were 34 percent and 41 percent, respectively. In a separate part of the survey, respondents were also asked to assess the importance of various factors in their decision to start a business. Answers ranged on a five point scale from “a very great extent” to “no extent” with “a great extent”, “some extent”, and “a little extent” being the middle categories. We coded a factor as being important if the business owner responded that the factor was important to a “very great extent” or “a great extent”. Table 7 summarizes the responses to these questions for our two PSED samples. Like the results in Table 6, we grouped the questions into the following broad categories: “non pecuniary motivations”, “status motivations”, “income motivations” and “other motivations”. The results in Table 7 reinforce the results in Table 6. When asked, nearly all respondents conveyed that non pecuniary benefits were important to a great extent or a very great extent when making their business start up decisions. For example, 72 percent of the business owners reported that it was important or very important to start the business because it provided “greater flexibility for personal and family life” and 75 percent of business owners reported it was important to start a business because “it gave them considerable freedom to adapt their own approach to how they 13

worked”. For comparison, only 26 percent reported that it was important to start the business to “develop an idea for a product”. Again, the take-away from this table is that non pecuniary benefits are an important factor in an individual’s decision to start a small business. Table 8 shows the innovation and growth behavior between new businesses that do and do not report that non pecuniary benefits were an important driver of their business formation. When segmenting new firms by whether or not non pecuniary benefits were an important driver of small business formation, we use whether or not the firm reported that non pecuniary benefits were an important driver on either their first or second report. As seen from Table 8, firms that started for non pecuniary reasons expected to have smaller businesses in the future, were less likely to want to be big, and were less likely to have an observable measure of innovation. Collectively, the results in Tables 6–8 show that households report that non pecuniary benefits are an important driver of their small business decisions. These non pecuniary benefits can reconcile the fact that small business owners earn less in their small business than they would in wage work. It can also explain the facts that these businesses remain in business without any observable growth or innovation or any desire to grow or innovate.

3

A Simple Model of the Small Business Sector

In this section, we describe a simple equilibrium model of the small business sector that matches the key features of the data we describe in section 2. In particular, we build in the existence of non pecuniary benefits for small business ownership into a general equilibrium model of occupational choice. As shown above, most business owners report the existence of non pecuniary benefits as an important reason as to why they started their business. In the model below, the small business sector will only be populated by people who start their business for non pecuniary reasons. The reason for this is that we want to highlight how a simple model of occupational choice with only non pecuniary benefits of small business ownership can generate many of the patterns in the data described above as well as generating a number of additional data patterns highlighted by others in the literature. We realize that some small businesses do start because of traditional entrepreneurial reasons. These firms, however, comprise only a small percentage of all small businesses. The goal of the model is to explain the behavior of these other small businesses. To keep things simple we make a number of additional abstractions. First, we will completely ignore the dynamics of small business formation and growth. As we explain in 2.2 and 2.3, most small businesses just do not grow or have any intention to grow.23 Second, we ignore financial market frictions. Hurst and Lusardi (2004) find that liquidity constraints do not appear to bind and initial capital requirements for most businesses are quite low. Even without financial frictions, it will become clear that the normality of business ownership will produce an increasing relationship between wealth and probability of business ownership. Finally, our last innocuous simplification treats all workers as equally capable as employees or proprietors of their own business. We view 23

Eliminating dynamics and risk excludes pursuing a number of interesting questions, some of which Pugsley (2010) takes up in a dynamic model of entrepreneurship.

14

this is a stepping off point to see how far we can go before needing to confront the more complex issues of skill sorting in the labor market. In our model, households will differ only in their initial wealth and their preference (if any) for running a business. They will have to decide whether to use their labor to own and operate a business or work in a distinct firm sector. If they decide to run a business, they also have to decide what goods to sell among the many types of goods. Firms can produce anything small businesses can produce, but are unconstrained in their ability to hire additional labor and may exploit their scale. We study an equilibrium where firms and small businesses compete to sell each good and where in equilibrium each good is supplied by the manufacturer offering the lowest price.

3.1

Final Good Sector

We assume that a competitive “final good” sector combines intermediate inputs xb to produce a final good



σ−1 σ

C= B

xb

 σ σ−1 db ,

(1)

of the type described by Dixit and Stiglitz (1977) and Spence (1976) where σ represents the elasticity of substitution between inputs. We use b to refer to a specific intermediate good and label the set of intermediate good types B. As we discuss below, the different intermediate inputs, b, will be distinguished by the technologies used by the firms and small businesses to manufacturer them. A cost-minimizing final good sector, discussed below, implies conditional input demand functions for each intermediate good, b, such that: xb (pb ) = P C

 p −σ b

P

,

(2)

where pb represents the price of good b and P is the price (or marginal cost) of one unit of the final good

 P ≡ B



p1−σ db b

1 1−σ

.

Henceforth, we set the final good as the numeraire by setting P equal to 1.

3.2

Households and Non Pecuniary Benefits

Households have preferences over consumption of a final good and whether or not they allocate their labor to running a business. They will supply labor inelastically so the only question is whether to allocate their time to the firm labor market where it will earn the market wage or to operating a small business where it earns any sales revenues from production. We refer to the latter as proprietor’s income. We will assume each household has preferences ordered by a utility function of the form U = log C + γE ,

15

(3)

where C represents consumption of the Spence-Dixit-Stiglitz final good and E is an indicator that is 1 if the household runs a business and 0 otherwise. Each household may differ in the value of   its parameter γ. We assume that γ ∈ γ, γ where γ ≥ 0 and let the marginal distribution F (γ) characterize the distribution of γ over this support. Households make only one important decision—whether or not to run a business. Since this is a static model, the household will consume its initial endowment and any additional proceeds from wage or proprietor’s income. Each household faces a market wage w for employment at a firm and a schedule of market prices pb if it were to instead operate a business producing type b goods using the available technology. If it runs a business the household must choose which b ∈ B it wants to produce. In equilibrium, each self employed household must be indifferent among the set of goods produced by small businesses. In anticipation of this later outcome we will label the proprietor’s income z, which does not depend on b. Each household’s labor supply is indivisible and equal to 1. Rogerson (1988) shows how the non-convexity associated with indivisible labor supply produces equilibrium allocations that are not Pareto optimal. To restore optimality, he introduces lotteries over the labor supply decision with complete insurance markets so that households may equalize consumption over either idiosyncratic outcome. We employ the same procedure here so that households of type γ may choose the value of E, where E represents both the probability of starting a business and the state-contingent price of consumption should the business start. Then 1 − E represents the probability of the business not starting and also the price of consumption in that event. As in Rogerson (1988), optimizing households will equalize consumption and the problem is iso-morphic to maximizing (3) choosing C and E ∈ [0, 1] subject to P C + (w − z) E = w + P y ,

(4)

where y represents the households wealth as an initial endowment of the final good. Households differ in their initial endowments y. We assume that y ∈ [0, y] and let the marginal distribution F(y) characterize the distribution of y over the support. We assume that y and γ are independent so that household heterogeneity is characterized by the joint distribution F (γ, y) = F (γ) F (y) . We write the above equation so w on the right hand side has the interpretation of the full value of the household’s time, and w − z represents the pecuniary opportunity cost (if any) of running a small business. We will later show that w − z is strictly positive in equilibrium.

3.3

Production of Intermediates by Firms and Small Businesses

We previously mentioned the various intermediate goods within the set B. Here we make this product space explicit. Each type of good b is characterized by the technology used to produce it, where b serves both as a label and the value of a parameter governing the efficient scale of   production. We assume that b ∈ b, b = B where b ≥ 0. The minimum efficient scale of the technology increases with the value of b. Labor is the exclusive factor of production so the scale of the firm may also be measured by its employment.

16

We assume good b may be manufactured by either a firm or a small business using the technology fb (h) = Ahθ − b

(5)

where h represents the units of labor. This technology exhibits increasing and then decreasing returns to scale. The quantity of h at which returns to scale are exactly constant will coincide with the minimum efficient scale of the firm. This can be quickly found by solving for the value of h that makes the elasticity of scale

hfb0 (h) fb (h)

exactly equal to 1. We label this quantity h∗b for each type

b and can show for this technology that h∗b

 =

b A (1 − θ)

1 θ

.

(6)

If a firm were operate at scale h∗b , then given wage w its marginal cost of production would be 1  θ b1−θ would be w . We will show that in a competitive equilibrium, free entry forces firms 1−θ θ Aθ (1−θ)

to operate at minimum efficient scale h∗b . Firms are unconstrained in their choice of h ≥ 0, however if a household runs a small business it must set h = 1. This prevents the small business from reaching the efficient scale for most values of b.24 We let κb represent the output of a single household producing b κb = fb (1 − ρ) = A (1 − ρ)θ − b ,

(7)

where we let the effective quantity of household labor possibly differ from 1 by exogenous factor ρ. If an entrepreneurial household produces good b, then the household would earn pb κb as proprietor’s income, which we have already hinted will in equilibrium be constant across types and equal to z. Additionally, the household would receive utility flow γ as non pecuniary compensation.

3.4

A Competitive Two-Sector Equilibrium

We construct a decentralized equilibrium where entrepreneurial households compete with firms to supply each good b, and worker households provide the labor required by the firms. We show that we can partition the set of goods B into non-overlapping subsets B e and B f which contain the set of goods produced by entrepreneur households and the set of goods produced by firms respectively. We define the equilibrium we will study. Definition 1. With P = 1, given a distribution of households F (γ, y) characterized by preference parameter γ and initial wealth y, and production technologies described by (1), (5), and (7), a two-sector competitive equilibrium consists of the following: 24 The way we have set up preferences is that individuals get the non pecuniary benefit from running the business themselves. This is consistent with the fact that most of the small businesses, described in Section 2, have very few employees at all. The extreme form of the non pecuniary benefits - that it turns on if the firm has no employees and turns off it has positive employees - is made for simplicity. We could imagine a more flexible specification that let the non pecuniary benefits decay as the number of employees increase. Doing so would not alter the main implications of the model.

17

1. Wage w and intermediate good prices pb 2. Allocations Cγy and Eγy that given prices w and pb maximize (3) subject to (4) for households of type γ, y 3. Wealth cutoffs y1γ and y2γ that depend on γ such that

Eyγ

   {0} if y ≤ y1γ   ∈ [0, 1] if y1γ < y ≤ y2γ    {1} otherwise

4. Allocations hb that maximize firm profits given w and pb for firms producing good b 5. A density of firms nb producing b that may freely enter or exit the market 6. A partition of B so that B e ∪ B f = B and B e ∩ B f = ∅, where if b ∈ B f then nb > 0 and nb = 0 otherwise 7. Final good market clears  

 Cγy dF (y|γ) dF (γ) = B

σ−1 σ

xb

 σ σ−1 db

8. Intermediate markets clear 
 n f (h ) db if b ∈ B f b b b xb =

 Eγyb dF (y|γ) dF (γ) if b ∈ B e 9. And labor markets clear  

1−



Eyγ dF (y|γ) dF (γ) =

nb hb db .

(8)

B

Standard arguments ensure that such an equilibrium exists, although we have been unsuccessful so far in showing that it is unique. To solve for this equilibrium, we first address the marginal households, i.e., suppose y ∈ (y1γ , y2γ ) for some household y, γ. The following proposition establishes the household’s indifference between b ∈ Be. Proposition 1. For any household γ, y where Eγy > 0, proprietor’s income pb κb does not depend on b. Proof. This follows immediately from the assumption of access to the same technology. Suppose to the contrary that there exists b0 such that pb0 κb0 > pb κb for all b ∈ B e . Then all households that run a small business would prefer to produce b0 and drive the price pb0 down until the returns are equalized. 18

Letting z = pb κb , an optimal choice of Eγy requires γ = λ (w − z) where λ =

1 Cγy

(9)

is the marginal utility of income. For these marginal entrepreneurial households

w − z represents the opportunity cost of running a business. Proposition 3 establishes that w − z is strictly positive. w−z γ ,

For the marginal households then Cγy = that Eγy =

w+y−Cγy . w−z

and satisfying the budget constraint (4) requires

Using the solution from Eγy we can locate the wealth thresholds easily, by

finding the values of y that make Eγy exactly equal to 0 or 1. Then y1γ =

w−z γ −w

and y2γ =

w−z γ −z.

Consumption for households outside of these thresholds will be equal to their endowment y and any earned income, w or z. It is useful to define two aggregate quantities. We let E without any subscripts represent the total supply of labor allocated to small business E ≡


γ
y γ

y1

Eγy dF (y|γ) dF (γ). If both y and γ

have independent uniform distributions, then:

E=

1 2

 
(w + 2y + z) γ − γ − (w − z) log γγ

, y−y γ−γ

when y1γ and y2γ are inside the support of y for all γ. In the same manner, we can define C to represent the total net demand for the final good C ≡


γ
y γ

y1

(Cγy − y) dF (y|γ) dF (γ). Again, if

both y and γ have independent uniform distributions, then: (w − z)2 log C=

  γ γ

−

1 2

w2 − z 2 + 2 wy − zy

y−y γ−γ





γ−γ


.

In the firm sector (when b ∈ B f ) free entry ensures that hb = h∗b . Given the technology represented by (5), this is the only value of hb at which profits are exactly to zero. With price equal to marginal cost then pb = w

!1 θ

b1−θ

.

Aθθ (1 − θ)1−θ

Using this firm level labor demand and the required price of b, intermediate good market clearing pins down the density of firms nb . If both y and γ have independent uniform distributions, then: nb =

Cp−σ b



1−θ bθ

 .

Recall that we have normalized P = 1 so all prices are in units of the final good. Next we determine the small business sector and firm sector partitions. In a competitive market with free entry, each good b will be supplied by the producer offering the lowest price. A good b will belong to the small business sector B e if and only if its price when sold by a small business

19

z κω

is less than the marginal cost of a firm at its efficient scale w good

b∗ that



b1−θ Aθθ (1−θ)1−θ

1 θ

. We locate a cutoff

equates the marginal cost of firms with the price charged by small businesses.   e 6= ∅, then B e = [b, b∗ ] and B f = b∗ , b , where b∗ is the root within the Proposition 2. If B h i interval b, min{A (1 − ρ)θ , b} of the following equation

w

!1 θ

b1−θ Aθθ

(1 − θ)

=

1−θ

z A (1 − ρ)θ − b

.

(10)

Proof. See Appendix With all equilibrium objects expressed in terms of the market wage w and equilibrium proprietor’s income z, it only remains to identify these prices by clearing the labor and intermediates markets. Since the intermediates markets for b ∈ B f has already been cleared, we focus on b ∈ B e . Market clearing requires that κb Eγyb = Cp−σ b , and since we have established that entrepreneur households are indifferent over b ∈ B e we need only check that this holds for aggregate small busi1−σ Eγyb = Cz −σ , which we ness production. Multiplying market clearing through by κ−σ b , then κb

integrated over all B e requires Cz −σ




Be

db = E. If both y and γ have independent uniform κσ−1 b

distributions, then: Cz

−σ



κσb∗ − κσb σ

 −E =0

(11)

where b∗ is the root defined by proposition 2. Likewise, labor market clearing requires 1 − E =


nb h∗b db. We omit this express from the main text because of its length, but include it in the

Bf

technical appendix. Unfortunately it is not possible to obtain algebraic solutions for w, z, and b∗ even when making simplified assumptions for both the distributions of y and γ. However given parameter values, we can numerically solve for the roots of the 3 simultaneous equations constituted by labor market clearing (8), (11), and (10), where the last equation must be solved for the appropriate root.

4

The Importance of Non Pecuniary Benefits

In this section, we show that the the introduction of non pecuniary motives into our simple equilibrium model generates sharp implications for the relationships between earnings, productivity, wealth, and firm size that are consistent with the evidence we present in section 2 as well as many additional established empirical regularities highlighted in the broader literature. As we highlight throughout, the inferences drawn from these empirical regularities can be altered significantly if one fails to account for the potential of non pecuniary benefits to small business formation.

4.1

Earnings Gaps and Aggregate Productivity

First, consistent with the empirical findings of Hamilton (2000) and Moskowitz and VissingJørgensen (2002), the model generates a gap in earnings between wage workers and business owners. 20

The self employed are willing to produce the good at a wage lower than they could have earned in the firm sector because they receive some of their compensation in the form of non pecuniary benefits. The following proposition establishes that the pecuniary opportunity cost of running a small business is always positive in an equilibrium with a small business sector. Proposition 3. If B f 6= ∅ and γ > 0 then w − z > 0. Proof. Since B f is non empty, at least some household type must be willing to work as an employee. That household is either marginal or an inframarginal employee. If the household is marginal then it satisfies (9) with equality. Since γ > 0 and λ > 0 then w − z must also be positive. If the household is inframarginal and Eyγ = 0 then γ < λ (w − z) and again w − z must be positive. Notice that this result does not rely on ρ > 0 such that labor is somehow less effective when operating a business instead of employed at a firm. As long as r is constant for all households, having ρ < 0 would support an equilibrium with w − z > 0. The existence of non pecuniary benefits also informs the well documented relationship between wages and firm size. Many researchers have documented that workers in smaller firms earn less than workers in larger firms (see Brown and Medoff, 1989). In Figure 1, we plot the equilibrium wage gap, normalized by total value added C, over alternative parametrizations of the distribution of γ. We show how the wage gap increases with the average strength of the non pecuniary benefit. Non pecuniary compensating differentials for running a business are a key aspect of understanding the relationship between wages and firm size at least on the low end of the firm size distribution. The wage gap is also tied to measured aggregate productivity. If there were no non pecuniary motives and every household worked in the firm sector so B f = B, average labor productivity AP (total value added / total hours) would equal w. We will continue to refer to this case as the “zero gamma” economy. With a small business sector: AP = w − (w − z) E . To see this we just integrate over all the households budget constraints. We can think of AP as a weighted average of income from either sector, or as the wage w adjusted for the wage gap w − z, as we have written here. Figure 2 plots how measured aggregate productivity also declines with the mean of the distribution of γ.25 For reference we plot aggregate productivity of the zero gamma economy as the dot on the vertical axis.26 As non pecuniary motives become more important, the wage gap and the size of the small business sector E both grow, lowering AP . It is true that w also grows as wages adjust for a small firm labor supply, but this effect is always offset by the losses from (w − z) E, as we establish in the following proposition. 25 26

We omit the plot for small values of E [γ] to avoid confronting difficult corner solutions for the wealth thresholds. With γ = 0, the equilibrium wage w0 is easy to work out since C = w, you can show that 

w0 = A (1 − θ)

1−θ

θ

θ

 1  θ

21

b

(1−θ)(1−σ) θ

 db

1 σ−1

.

Proposition 4. If B e 6= ∅, and γ > 0, then

∂AP ∂E[γ]

< 0.

The proof of this claim relies on a careful application of the implicit function theorem. The resulting algebra is extremely tedious and lengthy and we omit it from the main text. In summary, the simple model shows that the a model with non pecuniary benefits will result in individuals in the firm sector earning higher pecuniary returns than workers in the self employed sector. This results in a very discrete relationship that implies a positive firm size/wage relationship. Finally, the extent of non pecuniary benefits will affect measured labor productivity within the economy. Even though no technology parameters will change, differences in the distribution of non pecuniary benefits across locations or across time will result in differences in measured labor productivity.

4.2

Wealth and Business Ownership

The second important implication of our model is that without any financial frictions, the model produces an increasing relationship between initial wealth y and the probability of owning a business E. Proposition 5. If B e 6= ∅ then

∂Eγy ∂y

≥0

Proof. If the household is a worker, then Eγy = 0 and then

∂Eγy ∂y

=

1 w−z

∂Eγy ∂y

= 0. If the household is marginal,

> 0 by the previous proposition, and when the household is an inframarginal

entrepreneur, then Eγy = 1 and

∂Eγy ∂y

= 0.

An increasing relationship between wealth and entry is often interpreted as evidence of binding liquidity constraints for small business owners. The presence of non pecuniary benefits raises questions about relying on such an identification strategy. Figure 3 plots the probability of business ownership (Eyγ )over the wealth distribution. For each y we average over the conditional distribution F (γ|y). For a particular value of γ the wealth cutoffs are relatively close together and the probability of entry is increasing linearly in y. However heterogeneity in γ makes Ey a smooth non linear function of y as these thresholds evolve over the entire distribution of γ. The shape of this relationship is consistent with Probit estimations of entry on wealth, see for example Hurst and Lusardi (2004). In our model, the probability is flat over a segment of the population that is not liquidity constrained. At low levels of initial income, the marginal utility of consumption is large relative to the marginal utility of the non pecuniary benefits of business ownership. Likewise the wealthy pay an opportunity cost to run the business in the form lost wages because they enjoy running a business relative to other forms of consumption. Again, this result undermines much of the empirical strategy performed by Evans and Leighton (1989), Evans and Jovanovic (1989), Quadrini (1999), Gentry and Hubbard (2004), Cagetti and De Nardi (2006), Fairlie and Woodruff (2007), Fairlie and Krashinsky (2006). In these models, the relationship between wealth and the probability of starting a business (or even exogenous changes in wealth and the probability of starting a business) are evidence that liquidity constraints bind.

22

Our model yields the same predictions in a world with no financial frictions. If one takes the non pecuniary benefits of owning a small business seriously, using the relationship between exogenous changes and wealth and the probability of starting a business as being de-facto evidence of liquidity constraints is invalid.

4.3

What Do Small Businesses Produce?

Third, the model of non pecuniary benefits informs the type of goods we should observe a high concentration of small business owners. In our model, self employed households only produce goods that would have been produced by small to medium scale firms. Recall that the interval [b, b∗ ] defines the small business sector B e . Then any factor that enlarges the size of the small business sector does does by increasing the equilibrium cutoff b∗ . This tells us that if any b ∈ B e were to be produced by a firm in a competitive market, the firm would have a smaller efficient scale than any other firm producing in the firm sector b0 ∈ B f . This is consistent with the sorting we document in section 2.1 where most household owned businesses start in a very narrow set of industries that operate at a small scale in the long run. This results suggest that using the concentration of small businesses within a sector can inform researchers about the average returns to scale in that sector. To our knowledge, this approach has never been pursued to estimate the returns to scale across various industries. Additionally, the magnitude of the distribution of non pecuniary benefits has a direct impact on the size of the small business sector. Proposition 6. The size of B e increases with E [γ] Proof. See Appendix To see how the small business sector B e depends on the distribution of γ, Figure 4 plots the equilibrium cutoff b∗ for various E [γ] holding all other moments and parameters fixed. As non pecuniary motives become more important, the small business sector grows by successfully competing with higher b firms. The firms costs are higher because of the tighter labor market, and entrepreneur households are willing to bear the additional cost in lost wages in return for the non pecuniary compensation.

4.4

Distribution of Firm Size

Finally, the distribution of γ has important implications for the equilibrium cross sectional distribution of firms. Entrepreneur households draw business away from the small to medium size firms. This is the flip side of the previous point about b∗ . Here instead of thinking about b we use a change of variables to express the density of firms as a function of size h. After a change of variables the density n may be written as

nh = C


1−θ ! −σ θ hθ A (1 − θ) Aθθ

(1 − θ) 23

1−θ

1 . θAhθ

In figure 5 we plot this distribution of firm sizes measured by employment h. For reference, we also include a dashed line representing the distribution of firms in a zero gamma economy. Perhaps in this picture it is especially clear that entrepreneur households specialize in the types of goods that would have been produced by smaller scale firms.

5

A Regressive Small Business Subsidy

In this section, we think about how a model of non pecuniary benefits could inform the costs and benefits of subsidizing small business ownership. Despite their political appeal, the welfare calculus of a small business subsidy is not at all obvious. The importance of non pecuniary benefits in the decision to become a small business owner makes this especially difficult. To make this point we introduce a very simple subsidy into our model funded by a lump sum tax levied equally across all households. We show that the redistributive role of this subsidy could actually benefit the wealthy at the expense of the poor. We want to stress that our model offers no reason for policy makers to want to subsidize small businesses. Our goal is to highlight (1) the potential costs of subsidies to small business owners and (2) the distributional effects of subsidizing small business owners. We realize that any costs must be weighed against potential benefits. Most of the literature focuses only on the benefit. We feel our model is well suited to highlight some of the costs. To begin, we introduce a simple proportional subsidy to the model. An unsubsidized small business household producing b earns pb per unit sold. We let s represent a proportional subsidy to small business households so that small business owners will instead earn pb (1 + s) per unit sold.27 We augment the earlier equilibrium definition to include the subsidy and a new requirement that the government balance its budget through a lump sum tax levied across all households. Definition 2. With P = 1 and small business subsidy s > 0, given a distribution of households F (γ, y) characterized by preference parameter γ and initial wealth y, and production technologies described by (1), (5), and (7), a two-sector subsidized competitive equilibrium consists of the following: 1. A lump sum tax T , paid by all households 2. Wage w and intermediate good prices pb 3. Allocations Cγy and Eγy that given prices w and pb maximize (3) subject to (4) for households of type γ, y   This subsidy may be interpreted as a s A (1 − ρ)θ − b reduction in fixed operating costs b for each small business of type b. 27

24

4. Wealth cutoffs y1γ and y2γ that depend on γ such that    {0} if y ≤ y1γ   ∈ [0, 1] if y1γ < y ≤ y2γ    {1} otherwise

Eyγ

5. Allocations hb that maximize firm profits given w and pb for firms producing good b 6. A density of firms nb producing b that may freely enter or exit the market 7. A partition of B so that B e ∪ B f = B and B e ∩ B f = ∅, where if b ∈ B f then nb > 0 and nb = 0 otherwise 8. Final good market clears  

 Cγy dF (y|γ) dF (γ) = B

σ−1 σ

xb

 σ σ−1 db

9. Intermediate markets clear 
 n f (h ) db if b ∈ B f b b b xb =

 Eγyb dF (y|γ) dF (γ) if b ∈ B e  

10. Labor markets clear 1−



Eyγ dF (y|γ) dF (γ) =

nb hb db B

11. The government balances its budget 

T =

κb pb (1 + s) Eb db .

(12)

We repeat the steps from section 3.4 to compute the equilibrium with a subsidy. In this case we must replace proprietor’s income z with (1 + s) z in equations (8) and (11), leaving (10) (where z/κb represents the selling price pb ) unchanged. Since E is linear in y, the government budget balance equation may be solved analytically for T as a function of w, z (1 + s) , and b∗ . With all endogenous quantities as a function of w, z (1 + s) , and b∗ , then given parameter values, these may be recovered as the roots of the 3 simultaneous equations using numerical solutions. We take two approaches to quantity the welfare gains or losses from the subsidy. First we consider aggregate welfare, as measured by a utilitarian planner. Second, because the aggregate measure obscures some interesting redistribution, we look at the households’ individual burdens computing an equivalent variation measure of the subsidy’s cost. Using the first approach, the model implies that small business subsidies reduce aggregate welfare. To see this, we define a utilitarian measure of aggregate welfare Ws as the equally weighted 25

sum of each household’s utility in equilibrium under subsidy s ≥ 0.  

Ws =

(log Cyγ + γEyγ ) dF (y|γ) dF (γ) .

Figure 6 plots Ws as a function of s . The overall reduction in welfare is not surprising. In our example there are no market failures that would provide a beneficial role for a subsidy, and the unsubsidized competitive outcome is first best. With equal Pareto weights, the s = 0 allocation can be supported as a solution to a planning problem where increasing s > 0 simply distorts the allocation of labor across the two sectors. Holding Var [γ] fixed, varying E [γ] does not change the rate at which the subsidy trades off aggregate welfare. The more interesting result is the redistribution hidden behind the aggregate measure. The existence of non pecuniary motives makes the individual welfare effects of the subsidy highly non-linear. To study the household level effects of the subsidy we introduce a measure of equivalent variation. We compute EVyγ as EVyγ (s) = C (us ; w, z) − (w + y) where us is household y, γ equilibrium utility under subsidy s, and C (us ; w, z) is the minimum expenditures required at the unsubsidized equilibrium prices w and z in order achieve us and (w + y) is the unsubsidized equilibrium expenditures (or total wealth). We normalize this measure by w + y and express equivalent variation EVyγ /(w + y) as a fraction of the households total wealth. Using the subsidized and unsubsidized equilibrium allocations we can compute this measure over the entire joint distribution of households to study the household level welfare costs of the subsidy. Using this measure we find this simple small business subsidy to be regressive, actually benefiting wealthy business owners at the expense of wage employees. Figure 7 plots this welfare measure for the baseline case. The left panel plots the normalized EV measure over the entire joint distribution F (y, γ) for a small subsidy policy s = 0.05. It is a little difficult to read the surface plot, but it is evident that for some households (with EV /(w + y) > 0) the subsidy is a net benefit. In the right hand plot we integrate over γ to recover 

EVy =

EVyγ dF (γ|y)

the total welfare gain or loss for all households with wealth y and plot this measure over the wealth distribution F (y). We plot several policies ranging from a small subsidy s = 0.05 to a large subsidy s = 0.25. From this graph it is evident that even when summing across high and low γ households, wealthy households stand to benefit from a subsidy. Figure 8 makes this point more apparent by considering the three distributions of γ we have studied under a low subsidy in the left hand panel and a high subsidy in the right hand panel. Part of the large welfare cost to the poorer households is driven by the lump sum taxation assumption. This is an extreme example where all households equally share the tax burden regard-

26

less of their total wealth w + y. A more progressive policy where tax rates are based on wealth could reverse this policy. In fact, poorer households wage income actually increases. To see why consider the effect of a subsidy. It makes entrepreneurship more lucrative to all households. Many would have run businesses anyway, but some will switch from wage employment to business ownership constricting the labor supply. The downward sloping aggregate labor demand curve implies a higher equilibrium wage. In the lump sum taxation example, the modest increase in wages for poorer worker households is dominated by the additional tax burden needed to fund the subsidy. These mechanics also give some intuition for the result that wealthy entrepreneur households stand to benefit from the subsidy. While the subsidy entices some worker households into a higher probability of business ownership, the effect on this margin is relatively small. However, all business owning households stand to benefit from the subsidy, and the wealthy business owners who would have started their businesses anyway especially so. The best case scenario for them is a subsidy with small group of existing business owners, this way the individual benefit of the subsidy is not diluted by a larger tax needed to pay for a subsidy across a larger small business sector.

6

Calibrating the Distribution of Non Pecuniary Benefits γ

In this section we describe a method to calibrate the moments of the distribution of γ from observable characteristics of the cross sectional distributions of households and firms. [The remainder of this section is under construction]

7

Conclusion

This paper makes four main points. First, small business owners observed in the U.S. economy do not fit the entrepreneur of standard economic models. We show that most small businesses are confined to a very narrow range of products and services. Moreover, over time most do not grow, or even have plans to grow at the time of business inception. Additionally, most of these business owners fail to innovate along measurable dimensions such as patents or trademarks. And as we show, most do not have any plans to innovate. These observations compete with various models of entrepreneurship that assign to entrepreneurs a unique ability to manage, take risks, or develop new ideas. Second we show that in the absence of these standard motives, non pecuniary benefits of business ownership are an important consideration in the business formation decision. We utilize the newly released PSED to follow panels of nascent business owners and document their motives and successes over time. Using these survey data we find that just over half of new businesses cite non pecuniary benefits as a primary reason to start their business and further that these businesses are less likely to grow or report wanting to grow over time. The significance of these preferences goes a long way in resolving the differentials observed in proprietor’s income relative to firm sector income, small versus large firm wages, and private equity investment relative to public equity investment. Third we embed these preferences into a simple general equilibrium model of occupational 27

choice, allowing for heterogeneity in these tastes as well as wealth. This modest change helps us understand several of our observations. We explain an increasing relationship between wealth and business ownership, previously oft-attributed to liquidity constraints. We understand the gaps in earnings and investment returns as compensating differentials between sectors. We understand why the typical small business owner is a contractor, shop keeper, lawyer, and so on. Further it helps us understand why there are so many very small firms relative to small and medium size companies. Using the model, we examine the implications of a very simple small business subsidy, which could be thought of as reduction in the fixed entry costs of small business owners. We then show that the distortion is productivity decreases and when weighting all households equally, the subsidy is aggregate welfare reducing. On its own this is unremarkable given since we do not incorporate any market imperfections. However we show how the aggregate welfare measure obscures a great deal of redistribution. Wealthy households that would have become business owners even without the subsidy are made better off at the expense of poorer worker households who are made considerably worse off. This simple example raises concern that small business subsidies have the potential to act as a regressive tax. Lastly we use the cross sectional implications of the model to estimate moments of the distribution of non pecuniary benefits in the population. (under construction) Taken together, these results imply that when thinking about small businesses, one cannot safely ignore the role of non pecuniary benefits. And as a corollary, we show that using small business owners as a proxy for an innovative or risk taking entrepreneur can be misleading. We make a number of abstractions in order to add clarity to these two points. We exclude the risk and dynamics present in business ownership as well as (perhaps least innocuously) the heterogeneity in ability found within the labor force, both across individuals and sectors. We see thinking hard about these omissions along either dimension as a worthy future goal.

References Brown, C. and J. Medoff. 1989. “The employer size-wage effect.” The Journal of Political Economy 97 (5):1027–1059. Cagetti, M. and M. De Nardi. 2006. “Entrepreneurship, frictions, and wealth.” Journal of Political Economy 114 (5):835–870. Cagetti, Marco and Mariacristina De Nardi. 2009.

“Estate Taxation, Entrepreneurship, and

Wealth.” American Economic Review 99 (1):85–111. Cullen, J.B. and R.H. Gordon. 2007. “Taxes and Entrepreneurial Risk-taking: Theory and Evidence for the US.” Journal of Public Economics 91 (7-8):1479–1505. Dixit, A.K. and J.E. Stiglitz. 1977. “Monopolistic competition and optimum product diversity.” The American Economic Review :297–308.

28

Evans, D.S. and B. Jovanovic. 1989. “An estimated model of entrepreneurial choice under liquidity constraints.” The Journal of Political Economy 97 (4):808. Evans, D.S. and L.S. Leighton. 1989. “Some empirical aspects of entrepreneurship.” The American Economic Review 79 (3):519–535. Fairlie, R.W. and H.A. Krashinsky. 2006.

“Liquidity constraints, household wealth, and en-

trepreneurship revisited.” manuscript, UC Santa Cruz . Fairlie, R.W. and C.M. Woodruff. 2007.

“Mexican entrepreneurship: A comparison of self-

employment in mexico and the united states.” . Gentry, W.M. and R.G. Hubbard. 2004. “Success Taxes,” Entrepreneurial Entry, and Innovation. National Bureau of Economic Research Cambridge, Mass., USA. Hall, Robert E. 2010. “The Burden of the Nondiversifiable Risk of Entrepreneurship.” American Economic Review 100 (3):1163–94. Hamilton, B.H. 2000. “Does entrepreneurship pay? An empirical analysis of the returns to selfemployment.” Journal of Political Economy 108 (3):604–631. Holtz-Eakin, D., D. Joulfaian, and H.S. Rosen. 1994. “Entrepreneurial decisions and liquidity constraints.” The RAND Journal of Economics 25 (2):334–347. Hurst, E. and A. Lusardi. 2004. “Liquidity constraints, household wealth, and entrepreneurship.” Journal of Political Economy 112 (2):319–347. Hurst, Erik G., Geng Li, and Benjamin Pugsley. 2010. “What can Engle Curves Tell Us About the Self-Employed?” . Jovanovic, B. 1979. “Job matching and the theory of turnover.” The Journal of Political Economy 87 (5):972. Kanbur, SM. 1979. “Of risk taking and the personal distribution of income.” The Journal of Political Economy 87 (4):769–797. Kaplan, S.N. and J. Lerner. 2009. “It Ain’t Broke: The Past, Present, and Future of Venture Capital.” . Kihlstrom, Richard E. and Jean-Jaques Laffont. 1979. “A general equilibrium entrepreneurial theory of firm formation based on risk aversion.” The Journal of Political Economy :719–748. Knight, Frank. 1921. Risk, Uncertainty, and Profit. Houghton Mifflin. Lazear, Edward P. 2005. “Entrepreneurship.” Journal of Labor Economics 23 (4):649–680. Moskowitz, T.J. and A. Vissing-Jørgensen. 2002. “The returns to entrepreneurial investment: A private equity premium puzzle?” American Economic Review 92 (4):745–778. 29

Puri, M., R. Zarutskie, J. Driessen, and L. Phalippou. 2010. “On the Lifecycle Dynamics of Venture-Capital- and Non-Venture-Capital-Financed Firms.” . Quadrini, V. 1999. “The importance of entrepreneurship for wealth concentration and mobility.” Review of Income and Wealth 45 (1):1–19. Rogerson, R. 1988. “Indivisible labor, lotteries and equilibrium.” Journal of monetary Economics 21 (1):3–16. Schumpeter, Joseph. 1942. Capitalism, Socialism, and Democracy. Harper. Spence, M. 1976. “Product selection, fixed costs, and monopolistic competition.” The Review of Economic Studies 43 (2):217–235.

30

w-z C

0.25

0.20

0.15

E@ΓD 0.04

0.06

0.08

0.10

Figure 1: Wage gap (as a fraction of aggregate output) ρ = 0, θ = 0.75, b = 1, b = 5, σ = 2, y = 0, y = 30, γ − γ = 0.02.

AP

8.6

8.5

8.4

E@ΓD 0.02

0.04

0.06

0.08

0.10

Figure 2: Average Product of Labor (AP), ρ = 0, θ = 0.75, b = 1, b = 5, σ = 2, y = 0, y = 30, γ − γ = 0.02.

Prob 1.0

0.8

0.6

0.4

0.2

y 5

10

15

20

25

30

Figure 3: Probability of business ownership for y households, ρ = 0, θ = 0.75, b = 1, b = 5, σ = 2, y = 0, y = 30, γ − γ = 0.02, and with E [γ] = 0.05 (red), 0.10 (green), 0.15 (blue).

31

b*

2.2

2.1

2.0

E@ΓD 0.05

0.06

0.07

0.08

0.09

0.10

Figure 4: Small business cutoff b∗ , ρ = 0, θ = 0.75, b = 1, b = 5, σ = 2, y = 0, y = 30, γ − γ = 0.02.

f HhL 0.35 0.30 0.25 0.20 0.15 0.10 0.05 h 4

6

8

10

Figure 5: Distribution of firm sizes, ρ = 0, θ = 0.75, b = 1, b = 5, σ = 2, y = 0, y = 30, γ −γ = 0.02, and with E [γ] = 0.05 (red), 0.10 (green), 0.15 (blue). The dashed line represents the distribution of firms in the zero gamma economy.

Ws

3.08

3.07

3.06

0.1

0.2

0.3

0.4

0.5

s

Figure 6: Aggregate welfare effect of small business subsidy s ≥ 0, ρ = 0, θ = 0.75, b = 1, b = 5, σ = 2, y = 0, y = 30, γ − γ = 0.02, and with E [γ] = 0.05 (red), 0.10 (green), 0.15 (blue)

32

30

y

EV

20 10

y+w

0 y 5 0.000

10

15

20

25

30

-0.0001 -0.0002

-0.002 -0.0003 -0.004

-0.0004

0.040 0.045

-0.0005

0.050 Γ

0.055

-0.0006

0.060

(a)

EV w+y

EV (b) w+y over F (y) for s = 0.05, 0.10, 0.15, 0.20, and 0.25

over F (y, γ) for s = 0.05

Figure 7: Equivalent variation as a fraction of full income (w + y) of subsidy s policies, ρ = 0, θ = 0.75, b = 1, b = 5, σ = 2, y = 0, y = 30, γ − γ = 0.02, and with E [γ] = 0.05.

EV

EV

y+w

y+w 0.00005

y 5

10

15

20

25

30

y 5

10

15

20

25

30

-0.0002

-0.00005

-0.0004

-0.0006

-0.0001

(a)

EV w+y

over F (y) for s = 0.05

(b)

EV w+y

over F (y) for s = 0.25

Figure 8: Equivalent variation as a fraction of full income (w +y) of high and low subsidy s policies, ρ = 0, θ = 0.75, b = 1, b = 5, σ = 2, y = 0, y = 30, γ − γ = 0.02, and with E [γ] = 0.05 (red), 0.10 (green), 0.15 (blue)

33

Table 1: Small Business (Firms with Less than 20 Employees) and New Business Concentration by Industry A. Data from SUSB

B. PSED Data

34

Industry

Percent of Small Businesses in Each Industry (Out of All Firms in Industry)

Percent of Employees in Small Businesses in Each Industry (Out of All Employees in Industry)

Percent of Small Businesses in Each Industry (Out of All Small Businesses)

Percent of New Businesses in Each Industry (Out of All New Businesses)

All Firms

71.2

19.9

—

—

General Services Construction Professional Services Retail Trade Wholesale Trade Agriculture/Mining Entertainment Services FIRE Health Care Services Accomm./Food Service Transportation/Warehouse Manufacturing Information

83.9 85.5 77.7 56.6 65.2 67.1 72.8 52.6 70.5 61.1 48.5 50.0 —

43.5 38.3 27.2 24.9 24.3 22.8 20.8 20.4 19.0 15.5 13.3 7.0 —

18.7 11.4 8.1 15.4 7.4 2.2 2.0 8.5 10.7 6.3 3.6 5.2

16.2 10.9 13.9 18.7 4.5 2.7 3.5 7.7 4.6 5.8 2.4 6.0 4.0

Notes: Data in Panel A come from the 1987-1997 Statistics of U.S. Businesses (SUSB). Data in Panel B come from Wave A of the 2006 PSED. Data in Panel C come from the 2004-2008 Kaufman Firm Survey (KFS). Data from the SUSB only includes information on firms with at least one employee. The PSED and KFS only include information on new businesses. See text for descriptions of all samples. For the PSED sample, we use all small businesses sampled. The PSED and KFS data are weighted using the survey weights. The SUSB data come with SIC industry classifications. The KSF data and PSED data come with NAICS industry classifications. Aside from the “Information” industry, we were able to broadly match with SIC and NAICS classifications. We just kept “Information” as a separate industry for the latter two.

Table 2: Firm Age Categories That Are Firms with Less Than 20 Employees, by Industry (Data from 2005 Business Dynamics Statistics) (a) Percent of Establishments within Firm Age Categories

Firm Age Industry

0-10 Years Old

10-25 Years Old

All Firm Ages

All

89.9

73.8

69.4

Agriculture Construction FIRE Wholesale Trade Retail Services Manufacturing TCU

94.5 93.6 91.5 90.7 86.2 90.3 84.7 89.4

87.8 85.5 75.7 74.2 62.7 77.2 67.4 56.5

89.6 87.8 61.8 64.1 57.7 76.3 60.1 56.8

(b) Percent of Employment within Firm Age Categories

Firm Age Industry

0-10 Years Old

10-25 Years Old

All Firm Ages

All

44.1

24.7

18.9

Agriculture Construction FIRE Wholesale Trade Retail Services Manufacturing TCU

60.0 57.2 49.9 52.1 47.8 39.9 34.2 42.1

46.0 37.1 31.8 30.2 24.3 23.4 16.0 14.9

44.1 37.5 18.5 20.9 18.8 20.2 8.0 11.4

Notes: All data in Tables 2a and 2b can be found at http://www.ces.census.gov/index.php/bds/bds_database_ list. The industry classifications are the ones provided by the Business Dynamic Statistics (BDS). Like the SUSB data, the BDS data only includes information on establishments with paid employees.

35

Table 3: Change in Employment by Firm Age (a) Existing Small Businesses (Emps < 20)(Data from 2003 National Survey of Small Business Finances)

Percent Changing Employment over Last 1 Year Employment Change Increase No Change Decrease Sample Size

1-10 Years

Age of Firm 11-20 Years 20+ Years

18.9 74.1 6.9 1,175

10.8 79.4 9.8 828

9.1 84.0 7.0 739

Percent Changing Employment over Last 3 Years

All

1-10 Years

14.0 78.2 7.8 2,742

27.6 61.0 11.4 855

Age of Firm 11-20 Years 20+ Years 19.6 64.7 15.7 825

15.4 72.4 12.3 737

All 21.3 65.6 13.1 2,417

36

(b) New Businesses (Data from 2004-2008 Kauffman Firm Survey)

Percent Changing Employment over Last 4 Years

All New Firms Sample Size

Percent with ∆Employment > 1 Employee

Percent with ∆ Employment > 5 Employees

Percent with ∆ Employment > 10 Employees

41.9 2,617

10.8 2,617

3.6 2,617

Notes: See text for a description of both the 2003 NSSBF sample (used in Table 3a) and the KFS sample (used in Table 3b). We restricted the NSSBF to focus only on those firms with less than 20 employees. For the KFS sample, we looked at all new firms regardless of firm size. However, we did restrict the KFS sample to only those firms that remained in business for the four years since the survey started. The median and mean number of employees for the firms in the KFS sample was 1 and 3, respectively. The 90th percentile of number of employees for the firms in the KFS sample was 14. All data was weighted using the sample weights provided within the survey.

Table 4: Desired and Expected Firm Size of New Business Owners (Data from 2004-2007 PSED)

Percent of New Firms That Report They Want to Be “Big”a

37

Expected Number of Employees Working in Firm When it is 5 Years Old (25th percentile) Expected Number of Employees Working in Firm When it is 5 Years Old (Median) Expected Number of Employees Working in Firm When it is 5 Years Old (75th percentile) a

Sample: All PSED Respondents

Sample: PSED Respondents with Positive Revenues in Year 1

24.3

23.0

1 4 10

0 3 8

The PSED question reads: “Which of the following two statements best describe your preference for the future size of this new business: I want this new business to be as large as possible, or I want a size I can manage myself or with a few key employees?” We define big as those who report that they want their business to be as large as possible. Notes: See text for a description of both the 2003 NSSBF sample (used in Table 3a) and the KFS sample (used in Table 3b). We restricted the NSSBF to focus only on those firms with less than 20 employees. For the KFS sample, we looked at all new firms regardless of firm size. However, we did restrict the KFS sample to only those firms that remained in business for the four years since the survey started. The median and mean number of employees for the firms in the KFS sample was 1 and 3, respectively. The 90th percentile of number of employees for the firms in the KFS sample was 14. All data was weighted using the sample weights provided within the survey.

Table 5: Innovation Behavior of New Businesses (a) Actual Innovation Behavior of New Businesses (Data from 2004-2008 Kauffman Firm Survey)

Measure of Innovative Activity by Year

All New Firms Sample Size

Percent of Firms Who Have or Are Applying for a Patent

Percent of Firms Who Have or Are Applying for a Copyright

Percent of Firms Who Have or Are Applying for a Trademark

Percent of Firms Who Have or Are Applying for a Copyright or Trademark

2.7 2,581

8.9 2,550

12.3 2,546

17.3 2,510

Notes: Data are from the 2004-2008 Kaufman Firm Survey (KFS). Sample sizes differ slightly over the responses to the different questions depending on response rates. See the text for a full discussion of the KFS. All data are weighted using the provided survey weights. (b) Actual and Expected Innovative Activities of New Business Owners (Data from 2004-2007 PSED)

38 Percent Percent Percent Percent Percent Percent

of Firms that Already Developed Proprietary Technology, Processes, or Procedures of Firms that Expect To Develop Proprietary Technology, Processes, or Procedures in Future of Firms that Already Applied for Patent, Copyright, or Trademark of Firms that Expect to Apply for Patent, Copyright, or Trademark in Future of Firms that Expect R&D Spending Will Be a Major Priority for Business of Firms Stating That Many Existing Firms Already Offer Same Product/Service to Expected Customer Base Percent of Firms Stating That No Existing Firms Offers Same Product/Service to Expected Customers Sample Size

Sample: All PSED Respondents

Sample: PSED Respondents with Positive Revenues in Year 1

6.5 14.6 4.9 26.0 25.7 35.7

8.3 9.2 6.0 17.9 19.5 43.3

19.2 1,214

13.3 602

Notes: This table summarizes the responses to the questions asked of the nascent small business owners in the PSED about their actual and expected innovative activities. See text for the details. We focus on two samples. The first sample is all PSED respondents of nascent small business owners. The second sample is set of all nascent entrepreneurs who actually had positive revenues during Wave A (first follow up wave) of the survey. All data are weighted using the PSED sample weights.

Table 6: Importance of Non Pecuniary Reasons for Starting a Business by Nascent Entrepreneurs in the PSED

Reasons for Starting Business (Up to Two Reasons Provided)

39

Non Pecuniary Reasons To Generate Income Had a Good Business Idea / Create New Product Lack of Other Employment Options Other

I. Sample: All PSED Respondents (1,214 obs.)

II. Sample: PSED Respondents with Positive Revenues in Year 1 (602 obs.)

A. First Report

B. Any Report

A. First Report

B. Any Report

35.3 19.5 32.2 2.2 10.8

50.5 34.1 40.6 3.8 15.7

37.6 21.4 28.3 2.6 10.2

53.9 36.6 34.9 4.0 15.5

Notes: This table summarizes the responses to the questions “Why did you start this new business?” for respondents to the University of Michigan’s Panel Study of Entrepreneurial Dynamics (PSED). The PSED was a national sample of individuals who were in the process of starting their own business. The initial wave of the sample included 1214 nascent entrepreneurs. See text for a full discussion. We classified the responses to the open ended question of “Why did you start this new business?” into five broad categories of response: non-pecuniary responses, income reasons, having a good business idea, lack of other employment options, and other. For a complete discussion of our classification, see appendix Table A4. Respondents were allowed to provide up to two reasons for why they started the business. In columns A, we report the fraction of respondents who provided the specific reason on their first report. In columns B, we report the fraction of respondents who provided the specific reason on either their first or second report. The numbers in column B sum to less than 200% because about one-quarter of respondents did not provide a second report and, of those that did, some provided a report that was classified in the same broad category. Within the table, we focus on two samples. The first sample is all PSED respondents of nascent small business owners. The second sample is set of all nascent entrepreneurs who actually had positive revenues during Wave A (first follow up wave) of the survey. All data are weighted using the PSED sample weights.

Table 7: Percent of Nascent Entrepreneurs in the PSED Reporting Whether Various Motives for Starting a Business Was of Great or Very Great Importance

Sample: All PSED Respondents (1,214 obs.)

Sample: PSED Respondents With Positive Revenues InYear 1 (602 obs.)

Non Pecuniary Motivation Have greater flexibility for personal/family life Have freedom to adapt own approach to work To fulfill a personal vision Any of the above non pecuniary reasons

71.6 75.4 64.4 92.0

73.3 78.3 61.7 91.9

Status Motivation To be respected by friends To achieve a higher position in society To achieve something and get recognition for it Any of the above status reasons

15.2 17.3 33.2 40.1

13.1 13.8 30.0 36.6

Income Motivation Give self/family financial security To earn larger personal income To have a chance to build great wealth/income Any of the above income reasons

73.2 66.1 44.9 80.3

72.3 65.7 43.4 79.1

Other Motivation To develop an idea for a product To continue a family tradition To follow example of a person you admire To build a business children can inherit To have power to influence an organization

28.6 16.8 24.8 37.4 23.7

22.3 17.1 22.9 29.3 18.0

Motivation

Notes: This table summarizes the responses to the questions about what motivated the nascent business owners in the PSED to start the business. For each of the above motivations, respondents were asked to what extent that it was an important part of their decision. Respondents replied with a five point scale with 1 being “no extent”, 2 being “a little”, 3 being “some”, 4 being “a great”, and 5 being “a very great extent”. The table reports the percentage of respondents that replied by answering either 4 or 5 with respect to the motivation’s importance. We focus on two samples. The first sample is all PSED respondents of nascent small business owners. The second sample is set of all nascent entrepreneurs who actually had positive revenues during Wave A (first follow up wave) of the survey. All data are weighted using the PSED sample weights.

40

Table 8: Difference in Non Pecuniary-Motivated Business Characteristics Sample: All PSED Respondents (1,214 obs.)

41

Variable Expected Number of Employees Working in Firm When it is 5 Years Old (Median) Percent of New Firms That Report That They Want to Be “Big” Percent of Firms that Already Developed Proprietary Technology, Processes, or Procedures Percent of Firms that Expect To Develop Proprietary Technology, Processes, or Procedures in Future Percent of Firms Stating That Many Existing Firms Already Offer Same Product/Service to Expected Customer Base Percent of Firms Stating That No Existing Firms Offers Same Product/Service to Expected Customers

Difference

p-value of Difference

-1.00 -0.06 -0.04 -0.005

0.09 0.02 < 0.01 0.81

0.08 -0.06