120 Week 8 F /For WIZARD KIM

House Prices, Home Equity Borrowing, and Entrepreneurship

Stefano Corradin European Central Bank

Alexander Popov European Central Bank

This paper shows that housing wealth helps alleviate credit constraints for potential entrepreneurs by enabling home owners to extract equity from their property and invest it in their business. Using a large U.S. individual-level survey dataset for the 1996–2006 period, we find that a 10% increase in home equity raises the share of individuals who transition into self-employment each year from 1% to 1.07%. Our results persist when we use proxies for aggregate housing demand shocks and for the topological elasticity of housing supply to generate variation in home equity that is orthogonal to entrepreneurial choice. (JEL G21, L26)

In this paper, we identify the effect of housing wealth on new business creation by exploiting the exogenous variation in home equity induced by the U.S. housing boom of the late 1990s and early to mid-2000s. Using a uniquely suited individual-level dataset of U.S. home owners and renters, we investigate how changes in home equity affect the probability of becoming an entrepreneur. This allows us to study whether nascent entrepreneurs are deterred by credit constraints and whether the ability to extract housing wealth by increasing their mortgage helps them to overcome these constraints.

A number of influential papers have suggested that potential entrepreneurs can be discouraged from starting a business if borrowing constraints prevent them from raising sufficient capital.1 Black and Strahan (2002), Cetorelli and

We thank Geert Bekaert, Patrick Bolton, Luigi Guiso, Harry Huizinga, David Hirshleifer (the editor), Dwight Jaffee, Augustine Landier, Ross Levine, Deborah Lucas, Alexander Michaelides, Virgiliu Midrigan, Marco Pagano, Jonathan Parker, Maria Fabiana Penas, Monika Piazzesi, Robert Shiller, Ken Singleton, Phil Strahan, Amir Sufi, Roine Vestman, Gianluca Violante, Nancy Wallace, four anonymous referees, and seminar participants at the ECB, the Third Entrepreneurial Finance and Innovation Conference in Boston, the 2013 ASSA meeting in San Diego, Carlos III University of Madrid, the 2013 SED meeting in Seoul, Tilburg University, Cass Business School, and the 2014 EFA meeting in Lugano for valuable comments. Roberto Felici, Carlos Garcia de Andoain Hidalgo, and Thomas Kostka provided outstanding research assistance. The opinions expressed herein are those of the authors and do not necessarily reflect those of the ECB or the Eurosystem. Supplementary data can be found on The Review of Financial Studies web site. Send correspondence to Alexander Popov, European Central Bank, Financial Research Division, Sonnemannstrasse 22, D-60314 Frankfurt, Germany. E-mail: Alexander.Popov@ecb.int.

1 For important contributions, see Evans and Jovanovic (1989), Evans and Leighton (1989), Holtz-Eakin, Joulfain, and Rosen (1994), Blanchflower and Oswald (1998), and Wang, Wang, and Yang (2012), among others. See Kerr and Nanda (2009b) for a thorough review of the literature.

© The Author 2015. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com. doi:10.1093/rfs/hhv020 Advance Access publication March 5, 2015

The Review of Financial Studies / v 28 n 8 2015

Strahan (2006), and Kerr and Nanda (2009a) document a strong impact of positive shocks to access to external finance, following banking deregulation in the United States, on the rates of new business creation. A house price boom can provide another source of exogenous shocks to credit constraints whereby start-ups can extract the additional home equity in their property and invest it in their business. The effect is potentially significant given that residential property represents 60% of all personal wealth in the United States (U.S. Census 2010) and that prior research has already documented the households’ propensity to extract equity from their home, to various ends. For example, Hurst and Stafford (2004) highlight home equity borrowing as a mechanism whereby households smooth their consumption over time. Mian and Sufi (2011) empirically investigate how existing homeowners responded to the rising value of their home equity between 2002 and 2006. They provide evidence that this home equity-based borrowing channel, which may have been fueled by increasing availability of mortgage credit, was an important cause for the rapid rise in household leverage in the run-up to the financial crisis of 2008–2009.

The magnitude of the U.S. house price boom provides for an ideal identification of the home equity borrowing channel of entrepreneurship. Nationally, real home prices rose by 86% between the fourth quarter of 1996 and the first quarter of 2006 (Shiller 2007), but there were large regional differences. For example, over this period home prices almost tripled in Miami but declined by 10% in Detroit. Figure 1 plots the change in establishments births between 2003 and 2006 (the peak of the housing boom) against the change in the state-level Federal Housing Finance Agency house price indices over the same period, inflation-adjusted, for the fifty U.S. states, plus the District of Columbia. A positive relationship is readily available, suggesting higher entrepreneurial activity in states with a house price boom.

We investigate the link between housing and new business creation using microdata on home equity and business ownership from the Survey of Income and Program Participation (SIPP) of the U.S. Census Bureau, from 1997 to 2006. In each survey year, the respondents are asked whether they run and/or operate a business and the value of their business equity. They are also asked questions about their residential property. We construct an empirical proxy for home equity as the difference between the value of the household’s property and the size of the household’s mortgage. The survey also contains household- level data on a variety of other relevant characteristics, such as the household’s nonhousing wealth and labor income and the head’s age, race, education, and marital status. We control for housing debt, too, to account for the possibility that more levered households may harbor higher entrepreneurial tendencies. To identify variation across households residing in the same metropolitan state area (MSA) at the same point in time, we control for MSA×year fixed effects. The three SIPP panels put together contain information on 78,793 unique households, for 1996–2000, 2001–2003, and 2004–2006, for a total of 217,014 possible observations.

2400

House Prices, Home Equity Borrowing, and Entrepreneurship

Figure 1 Changes in rates of new business creation and changes in house prices This figure plots the change in establishments births between 2003 and 2006 (the peak of the U.S. housing boom) against the change in the state-level Federal Housing Finance Agency house price indices over the same period, for the fifty states, plus the District of Columbia. House price data are inflation adjusted.

The central part of the paper deals with the identification of the collateral channel. There are nontrivial endogeneity concerns related to the impact of home equity on the transition into entrepreneurship. Individuals who ultimately become entrepreneurs can be different from the rest in ways that matter for entrepreneurship. For example, they may disproportionately reside in areas with booming local economies, where the propensity to start a business is higher. In fact, it is highly likely that the areas that experienced the largest house price booms during the early to mid-2000s were also intrinsically entrepreneurial areas (such as large coastal cities). Alternatively, an outward shift in the supply of credit which accelerated house price growth (see Mian and Sufi 2009) may have also relaxed constraints on business loans, leading to higher levels of self-employment. And even in the absence of such omitted variable bias, rising house prices can feed back into local demand booms, raising the return to entrepreneurship.

We implement a number of empirical strategies to differentiate exogenous shocks to housing wealth from local economic effects (such as demand booms),

2401

The Review of Financial Studies / v 28 n 8 2015

which may be correlated with changes in house prices and in small business creation. First, we employ a version of the identification strategy suggested by Chetty and Szeidl (2010). Namely, we use the change in average U.S.-wide house prices between the year when the house was bought and the current year, divided by the local (MSA-level) topological elasticity of housing supply from Saiz (2010), as an instrument for home equity. The idea behind this approach is that an increase in the economy-wide demand for housing will increase house prices, and this effect should be stronger in MSAs with less elastic housing supply, where the adjustment in response to aggregate demand shocks takes place on the price margin. Second, we repeat our analysis only on the subsample of individuals who live in the MSAs with the most elastic housing supply. By doing so, we address the concern that the propensity to start a business can be positively correlated with the local price response to economy- wide changes in housing demand. Third, we seek to alleviate concerns that the house price boom is driving entrepreneurship directly, by excluding business start-ups in construction, finance, and real estate from the analysis. We also exclude business start-ups in health and education to account for the fact that the collateral channel we identify may be contaminated by a housing boom-driven increase in the demand for services with high-income elasticities. Fourth, we compare homeowners to renters, hypothesizing that the same increase in house prices should affect entrepreneurial propensity relatively more for homeowners as renters lack entirely the collateral channel. Finally, we compare the change in mortgage debt of new business owners and of nonbusiness owners around the same point in time in the same geographic locality. If the home equity borrowing channel is active, mortgage debt should increase more for new business owners, year-on-year, as they are expected to draw down their home equity to increase their business investment.

We find that the probability of starting a new business is strongly, positively correlated with the value of the home equity. A 10% increase in home equity raises the probability that a non-business-owning household will switch to entrepreneurship in the next period by up to 7%. This effect translates into an increase in the share of households in the sample who switch to self- employment from around 1% to 1.07%, from one year to the other. The results remain statistically robust to model specification and to accounting for a number of data features and alternative explanations. In particular, the positive effect of lagged home equity on the probability of starting a business is robust to controlling for a wide range of demographic and income characteristics, for the local business cycle, and for the bankruptcy code, and it is not driven by the propensity of individuals to start a new business when more levered. It is positive and significant regardless of whether we define business ownership in terms of owning and operating a business, in terms of holding nonzero business equity, or in terms of either. It is still present in the data when we correct for potential incidental variables problem and when we account for the fact that the SIPP oversamples low-income areas. Crucially, it survives when we use

2402

House Prices, Home Equity Borrowing, and Entrepreneurship

global housing demand shocks and the local elasticity of housing supply to extract the exogenous element of the change in housing wealth, when we look at areas in which the propensity to start a business is unlikely to be correlated with changes in house prices, when we exclude new businesses likely driven by local demand booms, and when we compare owners to renters. We also find a strong positive correlation between new business ownership and the change in mortgage debt. This implies that once they switch away from fixed income to entrepreneurship, individuals tend to draw down their home equity to finance their business investment, confirming that one’s house can serve as efficient collateral in business financing.

Our empirical results differ from Hurst and Lusardi (2004), who use microlevel data from the Panel Study on Income Dynamics (PSID) to argue that liquidity constraints do not matter for entrepreneurship for most of the wealth distribution. Furthermore, they find that households that lived in regions in which house prices appreciated strongly were no more likely to start a business than were households in other regions. Their analysis is mainly conducted on data from the 1998–1994 period when house prices in the United States were relatively flat. Conversely, we are able to exploit the effect of the large increase in house prices during the early to mid-2000s, when in some states (such as California) house prices doubled over the span of five years.

Our reduced-form estimates imply potentially important linkages between housing and real economic activity. For example, our results suggest that the housing boom before the Great Recession may have resulted both in higher rates of new business creation and in higher investment. This conjecture is corroborated in aggregate industry-level data by Adelino, Schoar, and Severino (Forthcoming) who document that areas with a bigger increase in house prices between 2002 and 2007 experienced a strong increase in small business employment relative to large business employment. Our findings also inform the work of Loutskina and Strahan (2015), who find that positive shocks to local house prices exert a strong positive effect on local economic growth. Related to our work, Chaney, Sraer, and Thesmar (2012) examine listed U.S. firms over the 1993–2007 period and provide evidence that when the value of a firm’s real estate appreciates by $1, its investment increases approximately by $0.06. The evidence presented in this paper also informs the work of Robb and Robinson (2014), who examine the capital structure of start-ups and find that new firms extensively rely on formal credit, with many start-ups receiving debt financed through the personal balance sheets of the entrepreneur. Our results also have important potential implications for economic activity during downturns, suggesting that high levels of mortgage debt after 2006 may have depressed self-employment, denting the positive effect on new business creation of higher unemployment rates. Our work is thus related to the evidence of Midrigan and Philippon (2011), who find that both output and employment declined more after 2007 in regions in which household leverage increased more between 2001 and 2007.

2403

The Review of Financial Studies / v 28 n 8 2015

The paper closest to ours is that by Schmalz, Sraer, and Thesmar (2013), who compare French homeowners and renters and find that homeowners are more likely to start a business in areas in which house prices appreciated more because higher collateral values increase borrowing capacity. Our paper is different in two important ways. First, we observe the actual housing wealth that homeowners have in their property. Therefore, we can estimate the effect of a change in home equity on the probability of transition into entrepreneurship within the sample of homeowners, in addition to comparing owners and renters, and so we do not have to deal with the concern that homeowners may harbor higher latent entrepreneurial tendencies than renters. Second, because Schmalz, Sraer, and Thesmar (2013) neither observe bank lending nor levels of home equity, they do not know whether the collateral channel operates by allowing homeowners to secure a larger business loan against the rising value of their property, or by enabling them to extract the additional home equity, thus bypassing the project screening function of the banking sector. Because we observe changes in mortgage levels, we find that nascent U.S. entrepreneurs indeed extract housing wealth from their residential property, by increasing their mortgage debt. Our paper thus directly identifies the home equity borrowing channel of entrepreneurship.

1. Data

We estimate the effect of changes in housing wealth on the propensity to start a business using household level survey data from the Survey of Income and Program Participation (SIPP) of the U.S. Census Bureau from 1996 to 2006. In each survey year, the respondents are asked questions related to business ownership. The survey also contains questions on the value of the house and on the size of the mortgage, allowing us to construct a proxy for home equity by taking the difference of the two.

The survey contains household-level data on a variety of additional individual characteristics. In particular, it has a detailed inventory of the household’s financial assets, in addition to demographic characteristics, which are theoretically related to entrepreneurial choice and business equity ownership, such as age, gender, race, education, and marital status. At each moment, SIPP tracks approximately 30,000 households. During the period considered, information was collected from three consecutive groups of households that were interviewed during the years 1996–2000 (four times), 2001–2003 (three times), and 2004–2006 (two times), respectively. The three SIPP panels put together contain information on 78,793 unique households, for a total of 217,014 possible observations.2

2 The SIPP oversamples individuals from areas with high poverty concentrations. We perform robustness checks to make sure that our main results are not affected by this feature of the survey.

2404

House Prices, Home Equity Borrowing, and Entrepreneurship

During its active period, each panel is interviewed every year, and panels of households do not overlap across periods. The longitudinal nature of the survey enables us to analyze dynamic characteristics, such as changes in employment status and income, changes in household and family composition, or changes in housing dynamics. The survey’s cross-sectional features allow us to keep track of housing wealth. We focus on the identification that arises when the value of the property, and consequently of the home equity, changes.

Theory provides little guidance on how to classify “entrepreneurs” (see Hurst and Lusardi 2004). The SIPP allows us to distinguish between direct ownership of business and ownership of business equity, which may or may not be ownership of equity in the household’s own business. Specifically, we utilize responses in the survey to the question “Did the household own and operate a business in the previous year?” to define business ownership, as well as responses to the question “What is the value in dollars of the household’s total business equity?” to define ownership of business equity. This strategy is somewhat richer than previous studies utilizing household data on entrepreneur- ship. For example, Hurst and Lusardi (2004) define entrepreneurship from a question in the PSID, which asks households whether they “[...] own a business [...] or have a financial interest in any business enterprise,” so they are unable to distinguish between direct and indirect ownership.

In terms of the household’s assets, we calculate net wealth as total wealth minus total debt. Total debt includes any mortgage on the household’s current home. We later distinguish between mortgage and nonmortgage debt. Total net wealth excludes the value of equity in the house.

To examine the role of home equity on the transition into entrepreneurship, we create a pooled sample of nonbusiness owners from the three survey waves. A household is defined as entering entrepreneurship if the household head or the spouse owns and operates a business in the very next period of the same survey wave. In robustness checks, we define entrepreneurship as ownership and operation of a business in any of the subsequent periods of the same survey wave. Consistent with Hurst and Lusardi (2004), we eliminate households in which the head is still in school or is close to retirement and focus on nonretired household heads between the ages of 22 and 60.

In addition to household information, we include data on gross state product (GSP) growth, state unemployment, and a proxy for homestead exemptions. The economic rationale for including the former two is that a more vibrant economy and/or a more depressed local labor market can raise the returns to self-employment (see Fairlie 2013). Regarding the latter, homestead exemption enables a filer for bankruptcy to retain home equity in her primary residence up to the exempted amount. A higher exemption may thus increase the rates of new business creation by imposing a lower cost on a potential business exit. At the same time, because the debts of the firm are personal liabilities of the firm’s owner, lending to the firm is legally equivalent to lending to its owner. Higher homestead exemptions can thus endogenously generate tighter

2405

The Review of Financial Studies / v 28 n 8 2015

Table 1 Variables summary statistics

Variable No. observations Mean SD Min. Max.

Business owner next period 46,679 0.01 0.10 0 1 Age

31–40 58,127 0.29 0.46 0 1 41–50 58,127 0.30 0.42 0 1 51–60 58,127 0.22 0.43 0 1

Education dummy High school 58,127 0.25 0.43 0 1 Some college 58,127 0.32 0.47 0 1 College or more 58,127 0.25 0.43 0 1

Dummy: African American 58,127 0.14 0.35 0 1 Dummy: Female 58,127 0.51 0.50 0 1 Dummy: Married 58,127 0.53 0.50 0 1 Labor income 58,127 $48,998.01 $52,144.90 $0 $1,212,060 Dummy: Unemployed 58,127 0.03 0.17 0 1 Household nonhousing wealth 54,626 $74,008.71 $1,063,061 $-1,013,580 $22,047,280 Home equity 58,127 $44,111.26 $80,412.79 $-299,446 $546,000 Home mortgage 58,127 $43,548.18 $67,443.40 $0 $330,002 Home property value 58,127 $87,659.44 $122,718.70 $0 $850,000 State unemployment 58,127 5.13 0.94 2.28 7.81 GSP growth 58,127 5.71 2.30 0.08 14.85 Homestead exemption/100,000 58,127 $2.11 $3.60 $0 unlimited

The sample includes all households in SIPP for the 1996–2000, 2001–2003, and 2004–2006 waves between the age of 22 and 60 that did not own a business the first time they were interviewed. “Business owner next period” is a dummy variable equal to one if the household owns or operates a business in the next period. “Age 31–40” is a dummy variable equal to one if the household head’s age is between 31 and 40 years. “Age 41–50” is a dummy variable equal to one if the household head’s age is between 41 and 50 years. “Age 51–60” is a dummy variable equal to one if the household head’s age is between 51 and 60 years. “High school” is a dummy variable equal to one if the household head has finished at most high school. “Some college” is a dummy variable equal to one if the household head is a college dropout. “College or more” is a dummy variable equal to one if the household head has at least a college degree. “African American” is a dummy variable equal to one if the household head is African American. “Female” is a dummy variable equal to one if the household head is a female. “Married” is a dummy variable equal to one if the household head is married. “Labor income” denotes the annual household income from supplied labor. “Unemployed” is a dummy variable equal to one if the household head is currently unemployed. “Household nonhousing wealth” denotes the total wealth of the household net of the amount of home equity. “Home equity” denotes the difference between the value of the household’s property and the value of the household’s mortgage. “Home mortgage” denotes the value of the household’s mortgage. “Home property value” denotes the value of the household’s property. “State unemployment” denotes the current rate of unemployment in the household’s state. “GSP growth” denotes the current rate of gross state product growth in the household’s state. “Homestead exemption” denotes the property value that can be exempted in bankruptcy procedures; we have assigned a value of $1,000,000 in states with unlimited homestead exemption. Omitted category in age is “Age 22–30.” Omitted category in education is “High school dropout.”

financial conditions. For example, Berkowitz and White (2004) show that small firms located in states with unlimited homestead exemptions are more likely to be denied credit. In our sample, the homestead exemption ranges from $0 in Maryland to an unlimited amount in eight U.S. states in 2006.

Table 1 presents summary statistics of the main variables in the data. There are a total of 58,127 households observed during the first year of each survey wave. Of the 46,679 households who answered the question “Did the household own and operate a business in the previous year?,” 453 answered “yes,” implying that about 1% of nonentrepreneurs become business owners each year. Fifty- nine percent of household heads are aged between 31 and 50 years; 57% have at least some college education; 14% are African American; 51% are female; 53% are married; and 3% are unemployed. In terms of financials, average

2406

House Prices, Home Equity Borrowing, and Entrepreneurship

Table 2 Descriptive statistics of new business owners and nonbusiness owners

Variable Nonbusiness Business p-value owner next period owner next period of difference

Age 31–40 0.288 0.322 0.07 41–50 0.303 0.325 0.03 51–60 0.234 0.216 0.08

Education dummy High school 0.254 0.210 <0.01 Some college 0.325 0.318 0.50 College or more 0.255 0.358 <0.01

Dummy: African American 0.137 0.064 <0.01 Dummy: Female 0.517 0.444 <0.01 Dummy: Married 0.539 0.673 <0.01 Labor income $49,368 $64,928 <0.01 Dummy: Unemployed 0.027 0.031 0.17 Household nonhousing wealth $76,455 $160,363 0.02 Home equity $45,736 $78,232 <0.01 Home mortgage $45,064 $76,567 <0.01 Home property value $90,800 $154,798 <0.01 State unemployment 0.051 0.052 0.34 GSP growth 0.057 0.059 0.40 Homestead exemption/100,000 $2.106 $2.378 0.01

The sample includes all households in SIPP for the 1996–2000, 2001–2003, and 2004–2006 waves between the age of 22 and 60 that did not own a business the first time they were interviewed. All statistics are means. For variable definitions, see Table 1. The unweighted percentage of households that became business owners in the next period after they were interviewed for the first time is 0.010.

labor income is around $49,000, and average nonhousing wealth (i.e., total wealth excluding home equity) is around $74,000. The average property is worth $87,659, with roughly equal size mortgage and home equity. Finally, average state-level unemployment over the period was 5.1%; average GSP growth was 5.7%; and on average $211,000 worth of home property value could be exempted from bankruptcy procedures.3

Table 2 reports descriptive statistics on the subsample of household that transitioned into entrepreneurship in the very next period. We compare those to descriptive statistics on the subsample of households that remained nonbusiness owners. In most cases, the differences between the two samples are statistically significant, and in some cases, the differences are economically large. For example, those transitioning into entrepreneurship are on average more educated and are more likely to be white, male, and married. They are also more likely to have a higher labor income and have higher nonhousing wealth. Importantly, those transitioning into entrepreneurship have more home equity, as well as a more valuable property and higher mortgage debt. This implies that when estimating the effect of home equity on the transition into entrepreneurship, we need to pay attention to differences in mortgage debt as otherwise our estimates can be contaminated by the independent effect of leverage on the probability of starting a business. Finally, Table 2 implies

3 We have assigned a value of $1,000,000 to the home exemption in states in which the homestead exemption is unlimited.

2407

The Review of Financial Studies / v 28 n 8 2015

Table 3 Characterizing new business owners

Variable Mean

Business equity new business owners $27,874.97 Business equity nonbusiness owners $3,378.97 Industry Agriculture, forestry, fishing, and hunting 0.023 Mining 0.000 Construction 0.155 Manufacturing 0.031 Wholesale trade 0.008 Retail trade 0.085 Transportation, warehousing, and utilities 0.070 Information 0.008 Finance, insurance, real estate, and rental and leasing 0.047 Professional, scientific, management, administrative, and waste management 0.264 Education, health, and social services 0.109 Arts, entertainment, recreation, accommodation, and food services 0.124 Other services (except public administration) 0.070 Public administration 0.008 Active military duty 0.000

The sample includes all households in SIPP for the 1996–2000, 2001–2003, and 2004–2006 waves between the age of 22 and 60. The share of each industry is determined based on the new business owners who reported the NAICS code of their business.

that individuals are more likely to transition into entrepreneurship if they live in a state with (marginally) lower unemployment and a higher homestead exemption.4

Table 3 summarizes the data on new business owners, that is, on individuals who transitioned into ownership and operation of a business between the previous and the current period. The average business equity of new business owners is $27,875. Although it is considerably higher than the average business equity of nonbusiness owners ($3,379), its relatively low level implies that entrepreneurs in the sample are on average not owners of large businesses or important job creators. This is confirmed by the sectoral breakup of new businesses reported in the table: most of the new business owners are in professional services, followed by sectors such as construction; arts, entertainment, recreation, accommodation, and food services; education, health, and social services; and retail trade. Only 3.1% of the new entrepreneurs are in manufacturing. These stylized facts caution against deriving implications in terms of job creation from observations on transition into self-employment.

2. Empirical Methodology and Identification

The main hypothesis that we test is that an increase in house prices will increase the probability that an individual will transition into entrepreneurship. The intuition for this result is that if household leverage remains the same (i.e., keeping the mortgage balance constant), a higher value of the property will

4 All of these state-level variables are time varying.

2408

House Prices, Home Equity Borrowing, and Entrepreneurship

increase the home equity, raising the value of potential business investment and reducing the effective cost to the agent of switching from fixed income to variable income.5

We test this prediction using the following nonlinear probability model:

Prob(new businessij t +1 = 1|Zij t ) = ϕ(Zij t ) = ϕ ( β1 + β2 ·ln(1 + home equityij t )

+β3 ·Xij t + β4 ·�j t + εij t ) , (1)

where “new business” is a dummy equal to one if household i in MSA j does not own or operate a business during the period when first interviewed (t ) but does so in the next period (t + 1). “home equityij t ” denotes the difference between the value of the household’s property and the size of the household’s mortgage debt at time t . Because the main explanatory variable is defined as “ln(1+home equityij t )”, we include in the sample those with zero home equity too (i.e., renters and homeowners whose mortgage equals 100%). However, we also run our regressions after excluding renters, in order to exploit further the richness of our sample, which allows us to compare the extensive margin of entrepreneurship within the sample of homeowners. Consistent with the hypothesis, we expect that β2 > 0.

Xij t denotes a vector of demographic and income proxies. These include age, education, race, gender, marital status, and employment status, for each household i in state j at time t . Variables related to liquidity constraints are also included in Xij t , in particular, the household’s current labor income and net wealth. Not accounting for wealth may bias our results upward because wealthier individuals may be simultaneously more likely to own a more expensive house and to become entrepreneurs. Theory predicts that the inability to acquire the capital necessary to start a business is one of the main theoretical obstacles faced by would-be entrepreneurs. A large literature has documented a positive relationship between initial wealth and subsequent business entry (e.g., Evans and Jovanovic 1989; Evans and Leighton 1989; Fairlie 1999; Quadrini 1999). However, Hurst and Lusardi (2004) challenge this view. Using detailed survey data, they show that personal wealth is important only for the richest households, and that once these are excluded from the sample, there

5 Corradin and Popov (2013) formulate a simple model of career choice with homeownership in the presence of borrowing constraints to motivate this hypothesis. In their model, agents start out as fixed-wage workers and can choose to become entrepreneurs in each following period. Their value function in each period depends on own wealth (as in Quadrini 2000; Cagetti and De Nardi 2006; and Buera 2009), but also on house prices through two channels. The first is that the current level of house prices affects the equity value they can extract from their home, increasing their mortgage balance. The housing stock thus has a collateral value component beyond the value derived from the direct consumption of housing services, due to prospects of limits on the amount of borrowing. The second channel results from the interaction between the first channel and the cost of external financing available to the potential entrepreneur. As in Quadrini (2000), the entrepreneur can take external financing through a business loan up to an amount that corresponds to the difference between the maximum permitted level of capital investment and total net wealth. The model thus predicts that agents are more likely to become entrepreneurs when the value of their home equity is high. It also predicts that on average, new entrepreneurs increase their mortgage balance and rely more on external financing.

2409

The Review of Financial Studies / v 28 n 8 2015

is no statistical relationship between wealth and entrepreneurial choices. If the positive relationship between wealth and entrepreneurship is confined to the top of the wealth distribution, then self-employment contains a “luxury good” component; namely, as households become wealthier, they are more likely to purchase the benefits associated with owning a business, such as prestige, power over decision-making, or flexible time schedule.6 We take these considerations into account by showing that our main results are robust to dropping the wealthiest 5% or 10% of the households from the sample.7

Finally, �j t is a matrix of MSA×year fixed effects. These control for any remaining time-varying factors that are common to all households in an MSA. Consequently, the estimate of β2 is driven off comparisons of households in the same MSA and year who have access to different levels of housing wealth.

While this empirical model allows us to evaluate the association between current home equity and future small business ownership, there are a number ways in which the probit estimates could be biased. The most obvious one is related to omitted variable bias. For example, credit standards were gradually relaxed throughout the United States in the wake of the dot-com bust. Easy credit may have simultaneously generated an increase in house values by raising local demand for housing and a decrease in borrowing limits for new firms. In other words, credit standards may have declined relatively more in areas with rising house prices, a possibility corroborated by the evidence of Mian and Sufi (2011) and Duca, Mullbauer, and Murphy (2011). Alternatively, individuals who ultimately become entrepreneurs may disproportionately reside in areas with local economies in which the propensity to start a business is naturally higher. If the areas that experienced the largest house price booms during the early to mid-2000s happen to be latently entrepreneurial too (such as California, Massachusetts, and New York), then a positive association between home equity and new business creation may again be largely driven by an omitted variable bias. And even if changes in home equity are uncorrelated with credit standards or with unobservable entrepreneurial tendencies, rising house prices can feed back into local demand booms, raising the return to entrepreneurship and, subsequently, the rate of new business creation. In this case, home equity will be related to entrepreneurship though a different channel than the home equity borrowing channel we seek to identify.

To isolate a clear causal effect of housing wealth on entrepreneurship, we need to identify exogenous shocks to home equity which are unrelated to local entrepreneurial propensity or economic conditions. To achieve this, we use an instrumental variable procedure to generate exogenous variation in home

6 In the same vein, Hamilton (2000) shows that most entrepreneurs enter and persist in business, although they have both lower initial earnings and lower income growth than paid employees. Moskowitz and Vissing-Jorgensen (2002) show that the returns on private equity are no higher than the returns on public equity, even though entrepreneurial investment is poorly diversified. Both papers thus suggest that there are important nonpecuniary benefits of entrepreneurship.

7 The results from these tests are reported in the Internet Appendix.

2410

House Prices, Home Equity Borrowing, and Entrepreneurship

equity. Building on the research design of Chetty and Szeidl (2010), we first calculate, for each homeowner in the sample, the change in national (U.S.- wide) house prices between the year when the house was bought and the current year. We also use information on the local topological elasticity of housing supply from Saiz (2010). The rationale is that local adjustment in response to changes in the demand for housing is more likely to take place on the quantity margin in places with highly elastic housing supply, and on the price margin in areas with inelastic housing supply.8 We calculate this elasticity both at the state and at the MSA level.9 The second approach is preferred as it yields a higher variation in the local elasticity of housing supply; however, for anonymity purposes, the MSA is not reported for 37% of households, and so in these cases, we apply the state-level elasticity.10 We then divide the change in national house prices between the year when the house was bought and the current year by the local elasticity of housing supply. The resulting ratio is our instrument for home equity. The idea is that following a positive aggregate U.S.-wide demand shock for housing, house prices will appreciate (and consequently home equity will increase, keeping the mortgage balance constant) considerably more in local areas with less elastic housing supply due to topological reasons, where adjustment in response to such shocks takes place on the price margin. It is therefore reasonable to expect that this instrument will help identify the variation in home equity that is unrelated to changes in local business conditions.

In addition to constructing an instrument for home equity, we implement a number of other strategies aimed at identifying the effect. First, we repeat our analysis only on the subsample of individuals who live in the MSAs with the most elastic housing supply, such as the ones in the Great Plains. In those, the propensity to start a business is less likely to be correlated with the local price response to economy-wide changes in housing demand. Second, by excluding from the analysis business start-ups in industries such as construction, finance, and real estate, we seek to alleviate concerns that the house price boom is driving entrepreneurship through a demand channel rather than through a home equity borrowing channel. Finally, we compare the change in mortgage debt of new business owners and of nonbusiness owners around the same point in time in the same geographic locality. If new business owners can indeed efficiently tap into housing wealth, one should observe a higher increase in mortgage debt, from one period to another, for new business owners relative to the rest.

8 Other recent papers to pursue identification of changes in collateral values using the local topological elasticity of housing supply include Chaney, Sraer, and Thesmar (2012), Adelino, Schoar, and Severino (Forthcoming), and Cvijanovic (2014), among others.

9 The original data in Saiz (2010) are at the MSA level. We calculate state-level elasticities by averaging the MSA-level elasticities for all MSAs in the state, weighted by the MSA’s population.

10 The SIPP stopped reporting the MSA field starting with the 2004–2006 wave.

2411

The Review of Financial Studies / v 28 n 8 2015

3. Empirical Results

We now present different sets of empirical estimates of model (1). In Section 3.1, we present the evidence from probit regressions on the association between housing wealth and new business creation, alongside a batter of robustness checks. In Section 3.2, we present the headline results in the paper, where we account for various sources of endogeneity in order to be able to interpret the positive effect of housing wealth on new business creation as a causal one.

3.1 Preliminary results: Probit estimation 3.1.1 Main result. In Table 4, we report the results from the base probit estimation of model (1). We start with all households who do not own or operate a business in the first period when they were interviewed (Columns (1) and (2)), and then we repeat the tests on the subsample of homeowners who do not own or operate a business in the first period when they were interviewed (Columns (3) and (4)). We control for the demographic characteristics of the household (age, education, race, gender, marital status, labor income, and employment status). To clearly separate the effect of housing wealth from the effect of the rest of the household’s financials, we report specifications without (Columns (1) and (3)) and with (Columns (2) and (4)) nonhousing wealth and the size of the household’s mortgage.

The estimates suggest that higher housing wealth significantly increases the probability that a household makes the transition to business ownership in the next period. Neither the coefficients nor their significance appears to be sensitive to the inclusion of nonhousing wealth and of mortgage debt. The intuition for the observed effect is that when house prices increase and raise the value of the property, holding the mortgage fixed, individuals can now extract more housing wealth from the house to buy working capital if they are to switch from employment to self-employment. In all cases, the null hypothesis that current home equity has no effect on future business ownership is rejected with p < 0.05. The effect is of a sizeable magnitude. Take the specification with individual and financial controls and MSA×year fixed effects in the case of homeowners (Column (4)). Going from the 25th to the 75th percentile of past house price increases (corresponding to a 17.70% increase) implies that the average house in the sample (valued at $87,659) appreciated by $15,516, or a 35% increase in home equity, holding the mortgage constant. The estimate of the marginal effect implies that for a household with the sample mean demographic and income characteristics, an 35% increase in home equity raises the probability that the household head will transition into entrepreneurship in the future by 0.35×0.0011=0.039 percentage points. Given a mean share of households who transition into entrepreneurship in the next period of 1.0%, this is equivalent to a 3.9% increase in the probability that the household will transition into entrepreneurship in the next period.

2412

House Prices, Home Equity Borrowing, and Entrepreneurship

Table 4 Business ownership and home equity: Probit results

Business owner next period

Owners and renters Owners

(1) (2) (3) (4)

Log (1 + Home equity) 0.0211∗∗∗ 0.0156∗∗∗ 0.0288∗∗∗ 0.0243∗∗ (0.0039) (0.0053) (0.0089) (0.0111)

Age 31–40 0.0492 0.0273 0.1014 0.0982 (0.0725) (0.0825) (0.0748) (0.0833)

Age 41–50 0.0122 −0.0451 0.0678 0.0312 (0.0757) (0.0842) (0.0834) (0.0916)

Age 51–60 −0.0710 −0.1413 −0.0258 −0.0724 (0.0738) (0.0917) (0.0900) (0.1013)

High school 0.0297 0.0814 0.0409 0.0485 (0.0545) (0.0663) (0.0788) (0.0877)

Some college 0.0526 0.1003 −0.0084 −0.0127 (0.0721) (0.0832) (0.0943) (0.1013)

College or more 0.1494∗ 0.1545∗ 0.1128 0.0561 (0.0779) (0.0927) (0.1027) (0.1116)

African American −0.1754∗∗ −0.1130 −0.1917∗ −0.1196 (0.0793) (0.0806) (0.1135) (0.1074)

Female −0.0782∗ −0.0892∗∗ −0.0461 −0.0560 (0.0421) (0.0417) (0.0553) (0.0538)

Married 0.0974∗∗∗ 0.0792∗∗ 0.1066∗∗ 0.0957∗ (0.0336) (0.0376) (0.0441) (0.0507)

Log (1 + Labor income) 0.4151 −0.2117 −0.1724 −0.9184 (0.3856) (0.4789) (0.5773) (0.5909)

Unemployed 0.1080 0.0697 −0.0348 −0.0167 (0.1151) (0.1314) (0.1166) (0.1160)

Log (1 + Household nonhousing wealth) 0.1894∗∗∗ 0.2048∗∗∗ (0.0424) (0.0414)

Log (1 + Home mortgage) 0.0047 0.0088 (0.0053) (0.0058)

MSA × year fixed effects Yes Yes Yes Yes Marginal effect of increasing home equity 0.0006 0.0006 0.0010 0.0011 Observations 38,421 36,473 23,123 22,301 Pseudo R-squared 0.05 0.06 0.04 0.05

The table reports probit estimates of transition into business ownership in the next period. The sample is composed of all households in SIPP for the 1996–2000, 2001–2003, and 2004–2006 waves between the age of 22 and 60 that did not operate a business the first time they were interviewed (Columns (1) and (2)) and of all households that owned a home and did not own and operate a business the first time they were interviewed (Columns (3) and (4)). “Home equity” denotes the difference between the value of the household’s property and the value of the household’s mortgage. “Age 31–40” is a dummy variable equal to one if the household head’s age is between 31 and 40 years. “Age 41–50” is a dummy variable equal to one if the household head’s age is between 41 and 50 years. “Age 51–60” is a dummy variable equal to one if the household head’s age is between 51 and 60 years. “High school” is a dummy variable equal to one if the household head has finished at most high school. “Some college” is a dummy variable equal to one if the household head is a college dropout. “College or more” is a dummy variable equal to one if the household head has at least a college degree. “African American” is a dummy variable equal to one if the household head is African American. “Female” is a dummy variable equal to one if the household head is a female. “Married” is a dummy variable equal to one if the household head is married. “Labor income” denotes the annual household income from supplied labor. “Unemployed” is a dummy variable equal to one if the household head is currently unemployed. “Household nonhousing wealth” denotes the total wealth of the household net of the amount of home equity. “Home mortgage” denotes the value of the household’s mortgage. Omitted category in age is “Age 22–30.” Omitted category in education is “High school dropout.” Standard errors clustered by state are reported in parentheses, where ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

We also find that individuals are more likely to become entrepreneurs if they have at least a college degree and if they are white, male, and married, which is consistent with the results of Hurst and Lusardi (2004). The effect

2413

The Review of Financial Studies / v 28 n 8 2015

Table 5 Business ownership and home equity: Robustness

Business Business Business owner owner or owner

Business next period, Business business next Business owner OLS equity equity period, owner

next period, + sample next next all in the OLS weights period period years future

(1) (2) (3) (4) (5) (6)

Log (1 + Home equity) 0.0003** 0.0004** 0.0121*** 0.0167*** 0.0129*** 0.0186*** (0.0001) (0.0002) (0.0037) (0.0040) (0.0049) (0.0046)

Individual controls Yes Yes Yes Yes Yes Yes MSA × year fixed effects Yes Yes Yes Yes Yes Yes Marginal effect of 0.0009 0.0014 0.0003 0.0006

increasing home equity Observations 42,464 42,464 50,881 51,469 46,840 37,895 (Pseudo) R-squared 0.01 0.01 0.04 0.04 0.05 0.06

The table reports OLS estimates (Columns (1) and (2)) and probit estimates (Columns (3)–(6)) of transition into business ownership. The independent variable is a dummy equal to one if the household head owns and operates a business next period (Columns (1) and (2) and (5)) or in any future period (Column (6)), a dummy equal to one if the households owns business equity next period (Column (3)), or a dummy equal to one if the household owns and operates a business or owns business equity next period (Column (4)). The sample is composed of all households in SIPP for the 1996–2000, 2001–2003, and 2004–2006 waves between the age of 22 and 60 that owned a home and did not own and operate a business (Columns (1) and (2) and (5) and (6)), did not own business equity (Column (3)), or did not own business equity or own and operate a business (Column (4)) the first time they were interviewed. In Column (5), all observations during each wave for each nonbusiness owning household are included. “Home equity” denotes the difference between the value of the house hold’s property and the value of the household’s mortgage. All individual controls from Table 4 are included. In Column (2), the survey-provided household weights are used to account for the fact that SIPP oversamples low-income households. Standard errors clustered by state are reported in parentheses, where ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

of nonhousing wealth is positive and significant, implying that other types of wealth that can be used to finance a business also increase the probability of transition into entrepreneurship. The point estimate implies that variations in home equity have a twice bigger effect on entrepreneurship than variations in nonhousing wealth, which may partially be due to the fact that nonhousing wealth includes the value of durables, such as vehicles and nonhome real estate. Finally, the effect of mortgage debt is positive but insignificant, implying that controlling for housing and nonhousing wealth, households with additional leverage are no more likely to start a business.

3.1.2 Robustness. Next, we perform, in Table 5, a battery of alternative tests to evaluate the robustness of the statistical association between housing wealth and new business creation, for the full sample of owners and renters. We start by recognizing that in parametric panel data models with fixed effects, where the independent variable is a dummy variable, the estimates will in general be inconsistent if the time dimension is small (Neyman and Scott 1948). As a solution to this “incidental variables” problem, we re-evaluate model (1) using a linear probability model instead of a nonlinear probability one. As Column (1) of Table 5 reports, the significant positive association between current home equity and future transition into entrepreneurship remains robust to this alternative specification.

2414

House Prices, Home Equity Borrowing, and Entrepreneurship

In Column (2) of Table 5, we address another possible concern related to our methodology, namely, the fact that our data may not be derived from a representative survey. The SIPP intentionally oversamples low-income households to maximize the coverage of households getting public services. Indeed, median labor income in the sample is $38,400 compared with mean labor income of $48,998, and median nonhousing wealth is five times smaller than the mean one, implying that the income and wealth distribution in our dataset is positively skewed. We address this issue by performing a version of the empirical test in Column (1) where we have used the reported survey weights to reweight our observations. As reported in Column (2), this alternative methodology does not yield results that are qualitatively or quantitatively very different from our baseline estimates.11

Another concern is that the positive association between housing wealth and entrepreneurship can be driven by our choice of proxy for entrepreneurship. We have defined an entrepreneur as an individual who owns and operates a business. While accurate, this definition of entrepreneurship by construction does not account for indirect business ownership. We now replicate our preferred specification after employing alternative definitions of business ownership that have been used in the literature. We first define transition into business ownership as a dummy variable equal to one if the household declares zero business equity in the current period but positive business equity in the next period. This definition accounts for the fact that individuals may become business owners not only by starting a business themselves but also by investing in other agents’ (such as family members’) business. Column (3) of Table 5 reports the estimates from this test. The null hypothesis that current home equity has no effect on future business equity is rejected at the 1% statistical level. In Column (4) we employ the definition of entrepreneurship used by Hurst and Lusardi (2004); namely, we define business owners as households who are either running and operating their own business or own business equity. We find that the positive association between home equity and the probability of transition into business ownership continues to be statistically strong (p < 0.01) for this broadest possible definition of entrepreneurship. We conclude that our results so far are not driven by our choice of proxy for new business creation.

A fourth concern is related to the timing of transition into entrepreneurship. In our main empirical tests we have focused on the moment when non-business- owning households were interviewed for the first time, and we have then looked at the probability that they would start a business in the next period. Given that in two of the three SIPP waves there are more than two observations per household, this leaves out a potentially substantial number of start-ups. We now address this issue in two ways. First, we look at the probability of

11 The estimates in Columns (1) and (2) are directly comparable to the marginal effects reported in Table 4.

2415

The Review of Financial Studies / v 28 n 8 2015

transitioning into entrepreneurship in the next period at all times, not just the first time when households were interviewed. This allows us to include multiple observations of the same household. The estimates reported in Column (5) of Table 5 suggest that our results so far have not been affected materially by the exclusion of a number of later start-ups. In Column (6), we address the same issue differently by defining entrepreneurship as the probability of a nonbusiness owner becoming an entrepreneur at any point in the future. There are 678 such households, compared with 453 who transition in the next period. This approach suffers from a purely mechanical problem in that different waves of the SIPP include a different number of years, and so households’s entrepreneurial choices are observed over different durations in each panel. At the same time, it allows us to capture a larger number of start-ups without including the same household in the regression multiple times. The results remain qualitatively unchanged: the strong positive association between home equity and transition to entrepreneurship survives this alternative definition of the switching horizon. Again, the null hypothesis that current home equity has no effect on future business ownership is rejected with p < 0.01. Quantitatively, the probit estimate increases to 0.0236 (from 0.0210 in Table 4, Column (2)). This is expected given that the probability of transitioning into entrepreneurship is higher over a longer horizon.

Importantly, all results in Table 5 are recorded after including our exhaustive set of variables, which proxy for demographic characteristics, nonhousing wealth, and mortgage debt. We conclude that the positive association between home equity and new business creation is not due to the high correlation between home equity and net wealth, disposable income, changes in the return to entrepreneurship, or to the potential impact of homeownership on the propensity for business ownership through the channel of higher leverage.

Our main results seem at odds with those of Hurst and Lusardi (2004), who use microlevel data from the PSID to argue that liquidity constraints do not matter for entrepreneurship for most of the wealth distribution. Furthermore, they find that households that live in regions in which house prices appreciated strongly are no more likely to start a business than are households in other regions. We believe that a number of factors can explain the difference between their work and ours. First, their sample consists of around 8,000 individuals (compared with around 79,000 in our sample), so they are not able to exploit as rich a regional variation in house price changes as we are. Second, their analysis is conducted on data from the 1989–1994 period when house prices in the United States were relatively flat. Conversely, we are able to exploit the effect of the large increase in house prices during the early to mid-2000s, when in some states (such as California) house prices doubled over the span of five years.

3.1.3 Interaction effects. Our results suggest that credit constraints serve as a barrier to new business creation and that a credit expansion based on an

2416

House Prices, Home Equity Borrowing, and Entrepreneurship

increase in home equity borrowing can help nascent entrepreneurs overcome this barrier. This raises a number of additional questions related to the link between house prices, the business cycle, and credit expansion. First, is the home equity borrowing channel of housing more potent for individuals that are more credit constrained? If households have sufficient nonhousing wealth, such as savings they can tap into or durable nonhousing goods they can liquidate, then the effect of additional home equity should not have a material effect on the propensity to start a business, and so the association we uncover might be spurious. Second, does the effect of an increase in housing wealth vary along the business cycle? In particular, the early to mid-2000s saw an unprecedented expansion in credit supply in general and in home equity borrowing, in particular (Mian and Sufi 2009), whereas home equity borrowing may not have been so prevalent in earlier decades. In fact, Hurst and Lusardi (2004) analyze the early 1990s, a period notable for a sharp credit crunch, which many hold accountable for the 1991 recession (e.g., Bernanke and Lown 1991). Finally, which role do characteristics of the local business environment, such as the local business cycle or legal treatment of housing assets, play in the association between housing wealth and new business creation? Because our specifications include MSA × year fixed effects, our tests cannot say anything on the level effect of those, but we could still study how they interact with housing wealth.

In Table 6, we address these questions by including interactions of home equity with demographic, financial, and state-specific variables, as well as year dummies, in our main model. We observe three sets of facts. First, while the coefficient on the interaction of home equity with nonhousing wealth is insignificant, the coefficient on the interaction of home equity with labor income is negative and significant, suggesting that housing wealth is more important for individuals with a lower labor income. The evidence thus implies that housing wealth stimulates new business creation relatively more for credit constrained households. Second, we find that the same level of housing wealth has a similar effect on new business creation in the 1990s and 2000s, as the coefficient on the interaction of home equity with the 2002 dummy and the 2005 dummy is insignificant.12 Finally, we also find that the interactions of state-level unemployment and of the level of the homestead exemption with home equity is positive and significant. The first piece of evidence confirms the intuition that for the same level of housing wealth, individuals are more likely to transition into self-employment when job opportunities are scarce. The second suggests that for homeowners with higher housing wealth, the direct effect of higher homestead exemptions in the case of personal bankruptcy through reduced business risk dominates the indirect effect through endogenously tighter credit constraints.

12 The omitted year category is the 1997 dummy.

2417

The Review of Financial Studies / v 28 n 8 2015

Table 6 Business ownership and home equity: Interaction with individual and state characteristics

Business owner next period

Age 31–40 × Log (1 + Home equity) 0.0176 (0.0125)

Age 41–50 × Log (1 + Home equity) 0.0194∗ (0.0105)

Age 51–60 × Log (1 + Home equity) 0.0304∗ (0.0172)

High school × Log (1 + Home equity) 0.0190 (0.0161)

Some college × Log (1 + Home equity) −0.0179 (0.0122)

College or more × Log (1 + Home equity) −0.0099 (0.0175)

African American × Log (1 + Home equity) −0.0017 (0.0141)

Female × Log (1 + Home equity) 0.0111 (0.0093)

Married × Log (1 + Home equity) 0.0007 (0.0071)

Log (1 + Labor income) × Log (1 + Home equity) −0.3210∗∗∗ (0.0990)

Unemployed × Log (1 + Home equity) −0.0204 (0.0232)

Log (1 + Household nonhousing wealth × Log (1 + Home equity) 0.0097 (0.0097)

Log (1 + Home mortgage) × Log (1 + Home equity) 0.0015 (0.0023)

State unemployment × Log (1 + Home equity) 0.0071∗ (0.0040)

GSP growth × Log (1 + Home equity) −0.0022 (0.0022)

Homestead exemption × Log (1 + Home equity) 0.0018∗ (0.0011)

Dummy 2002 × Log (1 + Home equity) −0.0180 (0.0144)

Dummy 2005 × Log (1 + Home equity) −0.0049 (0.0105)

Individual controls Yes MSA × year fixed effects Yes Observations 36,378 Pseudo R-squared 0.06

The table reports probit estimates of becoming a business owner in the next period. The sample is composed of all households in SIPP for the 1996–2000, 2001–2003, and 2004–2006 waves between the age of 22 and 60 that did not own and operate a business the first time they were interviewed. “Home equity” denotes the difference between the value of the household’s property and the value of the household’s mortgage. “Age 31-40” is a dummy variable equal to one if the household head’s age is between 31 and 40 years. “Age 41-50” is a dummy variable equal to one if the household head’s age is between 41 and 50 years. “Age 51-60” is a dummy variable equal to one if the household head’s age is between 51 and 60 years. “High school” is a dummy variable equal to one if the household head has finished at most high school. “Some college” is a dummy variable equal to one if the household head is a college dropout. “College or more” is a dummy variable equal to one if the household head has at least a college degree. “African American” is a dummy variable equal to one if the household head is African American. “Female” is a dummy variable equal to one if the household head is a female. “Married” is a dummy variable equal to one if the household head is married. “Labor income” denotes the annual household income from supplied labor. “Unemployed” is a dummy variable equal to one if the household head is currently unemployed. “Household nonhousing wealth” denotes the total wealth of the household net of the amount of home equity. “Home mortgage” denotes the value of the household’s mortgage. “State unemployment” denotes the current rate of unemployment in the household’s state. “GSP growth” denotes the current rate of gross state product growth in the household’s state. “Homestead exemption” denotes the current rate of exemption in bankruptcy of one’s property in the household’s state. “Dummy 2002” is a dummy variable equal to one if the household was first interviewed in 2002. “Dummy 2005” is a dummy variable equal to one if the household was first interviewed in 2005. Omitted category in age is “Age 22–30.” Omitted category in education is “High school dropout.” Omitted category in year is 1997. All individual controls from Table 4 are included. Standard errors clustered by state are reported in parentheses, where ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

2418

House Prices, Home Equity Borrowing, and Entrepreneurship

3.2 Headline results: Instrumental variables estimation 3.2.1 Ruling out omitted variable bias. While our probit estimates establish a robust positive association between current home equity and future new business creation, there are a number of ways in which the probit estimates could be biased. The most obvious is related to omitted variable bias. For example, credit standards were gradually relaxed throughout the United States in the wake of the dot-com bust. Easy credit may have simultaneously generated an increase in house values by raising local demand for housing and a decrease in borrowing limits for new firms. Mian and Sufi (2011) and Duca, Mullbauer, and Murphy (2011) document a larger decline in credit standards in local areas with a house price boom. Along the same lines, Corradin (2014) suggests that if mortgage lenders had optimistic views on house price dynamics in terms of low house price volatility and/or high expected house price growth rate, they may have loosened underwriting standards and offered low down payments contracts. Alternatively, individuals who ultimately become entrepreneurs may disproportionately reside in areas in which the propensity to start a business is naturally higher. If the areas that experienced the largest house price booms during the early to mid-2000s happen to be latently entrepreneurial, too (such as California, Massachusetts, and New York), then a positive association between home equity and new business creation again may be largely driven by an omitted variable bias. And even if changes in home equity are uncorrelated with credit standards or with unobservable entrepreneurial tendencies, rising house prices can feed back into local demand booms, raising the return to entrepreneurship and, subsequently, the rate of new business creation. In this case, home equity will be related to entrepreneurship through a different channel than the collateral channel we seek to identify.

To isolate a clear causal effect of home equity on entrepreneurship through the collateral channel, we need to identify exogenous shocks to home equity that are unrelated to local entrepreneurial propensity or economic conditions. To achieve this, we use an instrumental variable procedure to generate exogenous variation in home equity. Building on the research design in Chetty and Szeidl (2010), we first calculate, for each homeowner in the sample, the change in national (U.S.-wide) house prices between the year when the house was bought and the current year. Next, we use information on the local topological elasticity of housing supply from Saiz (2010). The idea is that local adjustment in response to changes in the demand for housing is more likely to take place on the quantity margin in places with a highly elastic housing supply (i.e., flat areas, such as Kansas) and on the price margin in areas with inelastic housing supply (i.e., coastal cities, such as Boston or San Francisco). We calculate this elasticity both at the state level and at the MSA level. The second approach is preferred as it yields a higher variation in the local elasticity of housing supply (there are ninety-six reported MSAs). The downside is that for about one-third of the households in SIPP, the MSA is not reported for confidentiality reasons, and so we assign the state-level value to these households. We then divide, by

2419

The Review of Financial Studies / v 28 n 8 2015

Table 7 Business ownership and home equity: IV results

Business owner next period: Owners

First stage Second stage Reduced form

(1) (2) (3) (4) (5) (6)

House price growth/MSA elasticity 17.882*** 2.278*** 0.2462 0.1663 (0.223) (0.197) (0.2384) (0.2481)

Log (1 + Home equity) 0.1304* 0.1617* 0.0377* 0.0285 (0.0886) (0.1064) (0.2233) (0.0222)

Wald F -statistics 6424.81 134.29 Individual controls Yes Yes Yes Yes Yes Yes Nonhousing wealth and Home No Yes No Yes No Yes

mortgage included MSA × year fixed effects Yes Yes Yes Yes Yes Yes Marginal effect of increasing 0.0053 0.0068

home equity Observations 38,771 22,946 19,353 19,346 19,202 19,137 Pseudo R-squared 0.72 0.22 0.03 0.03 0.04 0.05

The table reports estimates from OLS (Columns (1) and (2)) and from probit (Columns (3)–(6)) regressions. In Columns (1) and (2), the dependent variable is “Log (1 + home equity).” In Columns (3)–(6), the dependent variable is a dummy variable equal to one if the household owns or operates a business in the next period. The sample is composed of all households in SIPP for the 1996–2000, 2001–2003, and 2004–2006 waves between the age of 22 and 60 that owned a home and did not own and operate a business the first time they were interviewed. “Home equity” denotes the difference between the value of the household’s property and the value of the household’s mortgage. “House price growth” is the U.S.-wide change in house prices between the year when the property was bought and the current year. For renters, it is equal to zero. “MSA elasticity” is the MSA’s elasticity of housing supply (from Saiz 2010). In Columns (3) and (4), “Log (1 + home equity)” is instrumented using the U.S.-wide change in house prices between the year when the house was bought and the present year, divided by the MSA’s elasticity of housing supply. All individual controls from Table 4 are included. Standard errors clustered by state are reported in parentheses, where ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

the topological elasticity of housing supply, the change in national house prices between the year when the house was bought and the current year. The resulting ratio is our instrument for home equity. The main idea is that when there is an aggregate demand shock for housing at the national level, house prices will appreciate (and consequently home equity will increase) considerably more in MSAs with less elastic housing supply due to topological reasons, where adjustment in response to such shocks takes place on the price margin. It is therefore reasonable to expect that this instrument will help identify a variation in home equity that is unrelated to local business conditions. Furthermore, this instrument takes on different values for different individuals in the same MSA and year, and as a result, the housing wealth coefficient is driven solely off the comparison of people in the same MSA and year who have experienced different housing price appreciation based solely on differences in the year in which they purchased their home.

In Table 7, we report the estimates from model (1), where we use the instrumental variable just described to extract the exogenous component of home equity. Because for renters, we would have to arbitrarily set the value of the instrument to zero, we run our tests on the sample of homeowners only. We first report the estimates from the first-stage regression of the natural logarithm of home equity on the instrumental variable. We do so after

2420

House Prices, Home Equity Borrowing, and Entrepreneurship

excluding (Column (1)) and including (Column (2)) the natural logarithms of two potentially endogenous financial variables: nonhousing wealth and housing leverage. The reported coefficients imply that the change in national house prices between the year of purchase and the current year, divided by the local elasticity of housing supply, predicts a large share of the variation in individual housing wealth. The value of the first-stage Wald statistics, reported as “Wald F- statistics,” is strictly higher than the critical value for the IV regression to have no more than 10% of the bias of the probit estimate (see Stock and Yogo 2005).

Next, we report the estimates from the second-stage regression, again with and without proxies for nonhousing wealth and the size of the mortgage. Both variables are endogenous: individuals planning to start businesses in the future may save more today, thus paying down mortgage debt and amassing more financial assets. Because we do not have enough instruments for all endogenous variables, we report two versions of each regression, one that does not control for nonhousing wealth and for the size of the mortgage, and one that does.

The estimates in Columns (3) and (4) strongly suggest that the null hypothesis that current housing wealth has no effect on future business ownership is rejected. In both cases, the effect is significant at the 10% statistical level. Once again, the effect is of a sizeable magnitude. Take the specification with individual and financial controls and MSA × year fixed effects (Column (4)). Going from the 25th to the 75th percentile of past house price increases (corresponding to a 17.70% increase in property values) implies that the average house in the sample (valued at $87,659) appreciated by $15,516. Given an average home equity in the sample of $44,111, this corresponds to a 35% increase in home equity, holding the mortgage constant. The point estimate implies that for a household with the sample mean demographic and income characteristics, a 35% increase in home equity raises the probability that the household head will transition into entrepreneurship in the next period by 0.35×0.0068 = 0.24 percentage points. Given a mean share of households who transition into entrepreneurship in the next period of 1.0%, this is equivalent to a 24% increase in the probability that the household will transition into entrepreneurship in the next period.13

For an alternative interpretation, consider the following back-of-the- envelope calculation. The average difference in business equity between a new business owner and a nonbusiness owner is $24,500 (see Table 3), which corresponds to 55% of average home equity in the sample. Our estimate in Column (4) of Table 7 implies that the 55% increase in home equity necessary to generate the average difference in business equity between business owners and nonbusiness owners in the sample increases the probability of switching to self-employment by 38.5%. There were roughly 110 million households in the

13 This effect is directly comparable to Schmalz, Sraer, and Thesmar (2013), who find that going from the 25th to the 75th percentile of past house price growth increases the probability of firm creation by homeowners, relative to renters, by up to 13%.

2421

The Review of Financial Studies / v 28 n 8 2015

United States in 2006, around 1% of whom (or 1.1 million households) switch to self-employment each year. Therefore, the 55% increase in home equity required to generate the average difference in business equity between business owners and nonbusiness owners results in an additional 423,500 households who switch from fixed income to self-employment, year-on-year.

The validity of the identification strategy rests on the assumption that an exogenously generated change in house prices is a legitimate instrument for housing wealth in model (1). The first-stage estimates reported in Columns (1) and (2) of Table 7 suggest that the instrument is strongly related to housing wealth. However, for the IV estimate to be consistent, it must also be the case that the instrument is uncorrelated with the residual in model (1). If the instrument influences new business creation for reasons other than changes in individual housing wealth, our approach is called into question. As a way of addressing this concern, we note that if the only impact of the instrument is through changes in housing wealth, then the instrument should be insignificant if included in model (1). Columns (5) and (6) report the estimates from this reduced form model. The test fails to reject the null hypothesis that the effect of U.S.-wide changes in housing prices, normalized by the MSA’s elasticity of housing supply, equals zero. Furthermore, when we compare the top-half to the bottom-half of the MSAs in terms of elasticity, we find that this probability of transitioning into entrepreneurship in the next period is 1.01% in the former and 0.94% in the latter and that the difference between the two is not significant. We find the same when we compare the top tertile to the bottom tertile.

3.2.2 Ruling out local demand booms and feedback effects. Our identification strategy addresses head-on the concern that the positive association between home equity and entrepreneurship may be driven by local economic conditions that simultaneously raise house prices and increase new business formation. Nevertheless, one concern with identification still remains, namely, that (even exogenous) changes in home prices can feed back into local demand booms. Mian, Rao, and Sufi (2013) document that households consume a large share of any additional home equity. If the demand for entrepreneurial services (or a subset of these) goes up relatively more in areas with the highest increase in home prices, then our estimates of the size of the home equity borrowing channel can still be biased upwards.

We have so far partially dealt with this possibility by comparing owners and renters, whereby the latter entirely lack access to home equity borrowing, as well as by employing lagged explanatory variables whereby we map home equity at time t into the probability of transition into entrepreneurship at time t + 1. In Table 8, we address this concern in two additional ways. First, we repeat our tests on a subsample of MSAs with the highest elasticity of housing supply. In these, adjustment in the housing supply is relatively quick, and so it is unlikely that an aggregate demand shock would raise prices enough to generate an empirically substantial local demand boom. Second, we exclude

2422

House Prices, Home Equity Borrowing, and Entrepreneurship

Table 8 Business ownership and home equity: Ruling out demand booms and feedback effects

High-elasticity MSAs High-elasticity Excluding housing and excluding housing

MSAs boom-driven sectors boom-driven sectors

Probit IV Probit IV Probit IV

(1) (2) (3) (4) (5) (6)

Log (1 + Home equity) 0.0269*** 0.0447*** 0.0159*** 0.0223** 0.0280*** 0.0539*** (0.0059) (0.0146) (0.0051) (0.0121) (0.0053) (0.0168)

Individual controls Yes Yes Yes Yes Yes Yes MSA × year fixed effects Yes Yes Yes Yes Yes Yes Marginal effect of increasing 0.0008 0.0013 0.0004 0.0006 0.0008 0.0016 home equity Observations 19,292 18,245 36,308 34,848 19,276 18,221 Pseudo R-squared 0.06 0.05 0.05 0.01 0.06 0.05

The table reports probit estimates of becoming a business owner in the next period. The sample is composed of all households in SIPP for the 1996–2000, 2001–2003, and 2004–2006 waves between the age of 22 and 60 that did not own and operate a business the first time they were interviewed. “Home equity” denotes the difference between the value of the household’s property and the value of the household’s mortgage. In Columns (2), (4), and (6), “Log (1 + home equity)” is instrumented using the U.S.-wide change in house prices between the year when the house was bought and the present year, divided by the MSA’s elasticity of housing supply (from Saiz 2010). In Columns (1) and (2) and (5) and (6), only individuals living in the top 50% of the MSAs in terms of elasticity of housing supply are included. In Columns (3)–(6), the following sectors are excluded: Construction; Finance, insurance, real estate, and rental and leasing; and Education, health, and social services. All individual controls from Table 4 are included. Standard errors clustered by state are reported in parentheses, where ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

from the sample those entrepreneurs entering businesses that are directly linked to the housing boom. Two obvious such businesses are construction firms and mortgage brokerages. If all new small businesses in the sample are related to the building or selling of real estate property, then it will be hard to argue for a home equity borrowing channel of housing. This strategy is somewhat related to that of Adelino, Schoar, and Severino (Forthcoming), who separate the effect of housing on small business creation for tradeables and nontradeables, and show that the effect is there for tradeables, too, alleviating concerns that the effect is driven by local demand booms. However, our sample contains too few manufacturing start-ups for this strategy to be implemented.

In the first two columns of Table 8, we report probit (Column (1)) and IV (Columns (2)) estimates of model (1), where we focus on the subsample of households living in the top half of the MSAs in SIPP, in terms of local elasticity of housing supply. Our main result continues to hold, and in both cases, the effect of an increase in housing wealth on new business creation is significant at the 5% statistical level. The size of the numerical effect, however, declines relative to the full sample, suggesting that part of the increase in entrepreneurship in response to a boom in house prices may be due to a wealth effect that operates independently of from the home equity borrowing channel.

In Columns (3) and (4), we drop all entrepreneurs who start businesses for which demand may have increased in the wake of a housing boom. We exclude the following sectors: sector 3 (Construction), sector 9 (Finance, insurance, real estate, and rental and leasing), and sector 11 (Education, health, and social

2423

The Review of Financial Studies / v 28 n 8 2015

services). We exclude the first two because entry in those may be driven directly by an aggregate housing demand shock. As for education, health, and social services, these are traditionally the sectors with the highest income elasticity of demand. For example, if a housing boom makes people richer through a wealth effect and this raises their demand for health services, then a doctor’s decision to leave her job in a hospital and start a private practice cannot be directly linked to the home equity borrowing channel of housing. Our results survive the exclusion of these businesses, and the main effect is still significant at the 1% statistical level, suggesting that even though such effects are plausible, they are not the main driver of our results in the aggregate.

Finally, the main result of the paper still obtains in the data when we combine both strategies, that is, when we exclude households living MSAs with a very inelastic housing supply and drop businesses which are driven by the housing boom (Columns (5) and (6)).

3.2.3 Changes in mortgage debt after transition into entrepreneurship. Why does higher housing wealth induce entrepreneurship? While we have

argued for a home equity borrowing channel, whereby homeowners extract equity from their residential property to finance their business, they also simply could be collateralizing their property to obtain business loans. One direct way to evaluate whether housing wealth indeed drives new business creation through a home equity borrowing channel is to observe changes in mortgage debt after the household’s transition into entrepreneurship. If one’s house is indeed an efficient source of wealth, a new entrepreneur will partially convert her home equity into business investment by increasing her mortgage. If not, it would be difficult to argue for the existence of a home equity borrowing channel. Our data actually allow us to observe the behavior of new business owners, in terms of changes in total and in particular in mortgage debt.

We now proceed to explicitly test for this. In practice, we estimate the following equation:

� ln(1+mortgage debtij t +1) = β1 + β2 ·new business ownerij t + β3 · ln(1+mortgage debtij t ) + β4 ·Xij t + β5 ·�j t + εij t , (2)

where the dependent variable is the logarithm change in mortgage debt between period t and period t + 1 for household i in MSA. A new business owner is defined as a household head who owns or operates a business or owns business equity at time t + 1, but did not do so at time t . Consistent with the home equity borrowing channel hypothesis, we expect that β2 > 0. To account for natural convergence, we include last period’s stock of mortgage debt, too. Because we now include all individuals in the tests, and potentially do so more than once, the sample increases to a maximum of 117,660 observations.

2424

House Prices, Home Equity Borrowing, and Entrepreneurship

Table 9 Changes in mortgage finance: Comparing business owners to nonbusiness owners

Log (1 + Home mortgage_t + 1) − Log (1 + Home mortgage_t)

(1) (2)

New business owner 0.3092∗∗∗ 0.2793∗∗ (0.0687) (0.0883)

Log (1 + Home mortgage_t) −0.3147∗∗∗ −0.5533∗∗∗ (0.0102) (0.0126)

Individual controls Yes Yes MSA × year fixed effects Yes Yes Observations 117,660 77,547 R-squared 0.17 0.35

The table reports OLS estimates of the log change in mortgage debt from the previous to the current period, for the full sample (Column (1)) and for the subsample of homeowners (Column (2)). “New business owner” is a dummy equal to one if the household did not own business equity or operated a business in the previous period, but does in this period, and to zero if it does not. The sample is composed of all households in SIPP for the 1996–2000, 2001–2003, and 2004–2006 waves, between the age of 22 and 60. “Home mortgage” denotes the value of the household’s mortgage. All individual controls from Table 4 are included. Standard errors clustered by state are reported in parentheses, where ***, **, and * indicate significance at the 1%, 5%, and 10% level, respectively.

Table 9 reports the estimates from the regression of the change in mortgage debt on the new business owner dummy, controlling for the same demographic characteristics and time-varying state characteristics as in Table 4, as well as for state and year fixed effects. We observe smaller increases in mortgage debt for individuals whose mortgage debt is higher to begin with. Importantly, the estimate of β2 in model (2) implies that new business owners accumulate substantially more mortgage debt than otherwise identical individuals who did not become entrepreneurs over the same period. This is the case both in the full sample (Column (1)) and for the sub-sample of home owners only (Column (2)). Taking the latter case, a new entrepreneur tends to increase her mortgage by around 0.6 of a standard deviation more than an otherwise identical individuals who did not become a business owner. Our results thus broadly confirm the conjecture that financing of the business is facilitated by access to housing wealth, in that business owners can and do use their house as a source of equity.

4. Conclusion

In this paper, we evaluate the importance of credit constraints for new business creation using the exogenous variation in home equity for a large sample of U.S. homeowners over the period 1996–2006. We find that households with higher home equity today are significantly more likely to own and operate a business in the future. We record this result after implementing a number of strategies aimed at isolating a variation in home equity that is orthogonal to local economic effects. Numerically, a 10% increase in home equity increases the probability that a non-business-owning household will switch to entrepreneurship in the future by around 7%. This effect is robust to

2425

The Review of Financial Studies / v 28 n 8 2015

controlling for a wide range of demographic and income characteristics and for MSA×year fixed effects. Finally, we find a strong positive correlation between transition into entrepreneurship and the increase in mortgage debt. The intuition is that once individuals switch from a fixed-income job to entrepreneurship, they draw down their home equity to finance their business investment, confirming that real estate is indeed efficient source of funds.

Our reduced-form estimates imply potentially important linkages between housing and real economic activity. For example, our results suggest that the housing boom before the Great Recession may have resulted in higher rates of new business creation, while increasing levels of mortgage debt since 2006 may have depressed self-employment, denting the positive effect on new business creation of higher unemployment. There can be important economic spillovers from entrepreneurship, too, such as higher growth, lower crime, or higher rates of job creation. Regarding the latter, recent work has established that employment disproportionately expands in industries and areas most sensitive to changes in home prices (Adelino, Schoar, and Severino Forthcoming). In future work, it also would be instructive to incorporate an analysis of the implications for entrepreneurship of reducing transaction costs in the housing market and of various exemptions related to housing in the case of personal bankruptcy.

References

Adelino, M., A. Schoar, and F. Severino. Forthcoming. House prices, collateral, and self-employment. Journal of Financial Economics.

Berkowitz, J., and M. White. 2004. Bankruptcy and small firms’ access to credit. RAND Journal of Economics 35:69–84.

Bernanke, B., and C. Lown. 1991. The credit crunch. Brookings Papers on Economic Activity 22:205–48.

Black, S., and P. Strahan. 2002. Entrepreneurship and bank credit availability. Journal of Finance 57:2807–33.

Blanchflower, D., and A. Oswald. 1998. What makes an entrepreneur? Journal of Labor Economics 16:26–60.

Buera, F. 2009. A dynamic model of entrepreneurship with borrowing constraints: Theory and evidence. Annals of Finance 5:443–64.

Cagetti, M., and M. De Nardi. 2006. Entrepreneurship, frictions, and wealth. Journal of Political Economy 114:835–70.

Cetorelli, N., and P. Strahan. 2006. Finance as a barrier to entry: Bank competition and industry structure in local U.S. markets. Journal of Finance 61:437–61.

Chaney, T., D. Sraer, and D. Thesmar. 2012. The collateral channel: How real estate shocks affect corporate investment. American Economic Review 102:2381–409.

Chetty, R., and A. Szeidl. 2010. The Effect of housing on portfolio choice. NBER Working Papers 15998.

Corradin, S. 2014. Household leverage. Journal of Money, Credit, and Banking 46:567–613.

Corradin, S., and A. Popov. 2013. House prices, home equity, and entrepreneurship. ECB Working Paper 1544.

Cvijanovic, D. 2014. Real estate prices and firm capital structure. Review of Financial Studies 27:2690–735.

2426

House Prices, Home Equity Borrowing, and Entrepreneurship

Duca, J., J. Mullbauer, and A. Murphy. 2011. House prices and credit constraints: Making sense of the U.S. experience. Economic Journal 121:533–51.

Evans, D., and B. Jovanovic. 1989. An estimated model of entrepreneurial choice under liquidity constraints. Journal of Political Economy 97:808–27.

Evans, D., and L. Leighton. 1989. Some empirical aspects of entrepreneurship. American Economic Review 79:519–35.

Fairlie, R. 1999. The absence of the African-American owned business: An analysis of the dynamics of self- employment. Journal of Labor Economics 17:80–108.

———. 2013. Entrepreneurship, economic conditions, and the Great Recession. Journal of Economics and Management Strategy 22:207–31.

Hamilton, B. 2000. Does entrepreneurship pay? An empirical analysis of the returns to self-employment. Journal of Political Economy 108:604–31.

Holtz-Eakin, D., D. Joulfaian, and H. Rosen. 1994. Sticking it out: Entrepreneurial survival and liquidity constraints. Journal of Political Economy 102:53–75.

Hurst, E., and A. Lusardi. 2004. Liquidity constraints, household wealth, and entrepreneurship. Journal of Political Economy 112:319–47.

Hurst, E., and F. Stafford. 2004. Home is where the equity is: Liquidity constraints, refinancing and consumption. Journal of Money, Credit, and Banking 36:985–1014.

Kerr, W., and R. Nanda. 2009a. Democratizing entry: Banking deregulation, financing constraints, and entrepreneurship. Journal of Financial Economics 94:124–49.

———. 2009b. Financing constraints and entrepreneurship. NBER Working Paper 15498.

Loutskina, E., and P. Strahan. 2015. Financial integration, housing, and economic volatility. Journal of Financial Economics 115:25–41.

Mian, A., and A. Sufi. 2009. The consequences of mortgage credit expansion: Evidence from the U.S. mortgage default crisis. Quarterly Journal of Economics 124:1449–96.

———. 2011. House prices, home equity-based borrowing, and the U.S. household leverage crisis. American Economic Review 101:2132–56.

Mian, A., K. Rao, and A. Sufi. 2013. Household balance sheets, consumption, and the economic slump. Quarterly Journal of Economics 128:1687–726.

Midrigan, V., and T. Philippon. 2011. Household leverage and the recession. NBER Working Papers 16965.

Moskowitz, T., and A. Vissing-Jorgensen. 2002. The returns to entrepreneurial investment: A private equity premium puzzle? American Economic Review 92:745–78.

Neyman, J., and E. Scott. 1948. Consistent estimates based on partially consistent observations. Econometrica 16:1–32.

Quadrini, V. 1999. The importance of entrepreneurship for wealth concentration and mobility. Review of Income and Wealth 45:1–19.

———. 2000. Entrepreneurship, saving, and social mobility. Review of Economic Dynamics 3:1–40.

Robb, A., and D. Robinson. 2014. The capital structure decisions of new firms. Review of Financial Studies 27:153–79.

Saiz, A. 2010. The geographic determinants of housing supply. Quarterly Journal of Economics 125:1253–96.

Schmalz, M., D. Sraer, and D. Thesmar. 2013. Housing collateral and entrepreneurship. NBER Working Paper 19680.

2427

The Review of Financial Studies / v 28 n 8 2015

Shiller, R. 2007. Understanding recent trends in house prices and home ownership. Proceedings- Economic Policy Symposium - Jackson Hole, Federal Reserve Bank of Kansas City: 89.123.

Stock, J., and M. Yogo. 2005. Testing for weak instruments in IV regressions. In Identification and inference for econometric models: Essays in honor of Thomas Rothenberg, 80–108. Cambridge: Cambridge University Press.

U.S. Census. 2010. The United States Census Bureau.

Wang, C., N. Wang, and J. Yang. 2012. A unified model of entrepreneurship dynamics. Journal of Financial Economics 106:1–23.

2428

Copyright of Review of Financial Studies is the property of Oxford University Press / USA and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.