Homeownership, housing capital gains and self-employment

Abstract

This paper measures the impact of individual-level housing capital gains on transitions into and out of self-employment. Drawing on special features of the 1985–2013 American Housing Survey (AHS) panel, our most robust models control for recent expenditures on home maintenance, MSA-by-year fixed effects, lagged proxies for wealth and other household attributes. Net of home maintenance, a 20% real increase in home value over a two-year period raises the likelihood of entry into self-employment by roughly 1.5 percentage points; housing capital losses have little effect on exits. Controlling for house fixed effects, self-employed homeowners are also more likely to hold a HELOC, facilitating easy, low-cost access to home equity that could be used to cover business expenses. These and other estimates suggest that links between homeownership and self-employment are strong enough to be important when home prices are rising rapidly, but modest when housing capital gains are limited or negative.

Introduction

This paper demonstrates that housing capital gains encourage entry into self-employment while housing capital losses have little effect on exits. We also show that self-employed homeowners are more likely to hold a home equity line of credit (HELOC), facilitating easy, low-cost access to home equity that could be used to cover business expenses. In documenting these and other patterns, the paper highlights a link between self-employment and owning a home that until recently has received little attention. This is true for the vast literature on homeownership and also an extensive literature on self-employment and entrepreneurship. In the aftermath of the Great Recession, however, links between housing market volatility and labor market outcomes have taken on new importance, contributing to a burst of new papers that seek to assess the effect of homeownership and housing capital gains on self-employment.

Relative to other work in this area, our paper is the first to directly measure the magnitude of homeowner response to household-level housing capital gains net of home maintenance expenditures. This is possible because our data, the 1985–2013 American Housing Survey (AHS) panel, reports expenditures on home maintenance. Previous work has instead proxied for homeowner capital gains using aggregate measures of house price inflation (e.g. Hurst and Lusardi, 2004; Disney and Gathergood, 2009; Adelino et al., 2015; Corradin and Popov, 2015; and Kerr et al., 2015), and/or relied on policy events to confirm that housing capital gains increase self-employment (e.g. Jensen et al., 2015; Schmalz et al., 2017).¹ While informative, these approaches provide only indirect estimates of the extent to which a change in an individual's house value affects that individual's propensity for self-employment. Previous studies have also tended to focus on short time periods when home prices were mostly rising including 1985–1988 in Hurst and Lusardi (2004), 1997–2006 in Corradin and Popov (2015), 2002–2007 in Adelino et al. (2015), and 2000–2004 in Kerr et al. (2015). Our models are estimated using the full 1985–2013 AHS panel which visits homes in the U.S. every two years. The long sweep of time includes multiple home price boom-bust episodes and increases sample variation. That, along with other special features of the AHS, enhances opportunities to reveal different facets of the relationship between homeownership and self-employment.

Drawing on the AHS data, and conditioning on MSA-by-year fixed effects, proxies for lagged wealth, and other controls, we find that a 20% real increase in home value over a two-year period (net of maintenance) increases the probability that a homeowner transitions into self-employment by roughly 1.5 percentage points. Conditioning on year and house fixed effects we also obtain robust evidence that self-employed homeowners are roughly 2.5 percentage points more likely to hold a HELOC. These and other estimates suggest that links between homeownership and self-employment are strong enough to be important when home prices are rising rapidly, but modest when capital gains are limited or negative.²

There are several reasons to take note of the connection between homeownership and self-employment. Self-employment contributes to new small firm creation, it is sometimes associated with innovation and new product creation, and it can also provide opportunities for employment when wage-work is not accessible.³ For these reasons, self-employment plays a central role in the labor market and has been the subject of considerable study. It is also well known that U.S. home prices rose sharply 1996–2006, crashed 2007–2009, and have moved back up to their pre-crash peak as of fall 2016.⁴ Over this same period, U.S. homeownership rates jumped from 64% in the mid-1990s to 69% in 2006, and then fell to roughly 63.5% by fall 2016.⁵ Such extreme volatility amplifies possible spillovers that may affect self-employment decisions, adding to other previously established effects of homeownership on families and their local economies.⁶ Partly in response to this extensive volatility in housing markets, a burst of new papers has sought to highlight and measure different aspects of the impact of housing capital gains on self-employment. We discuss those and related studies below, beginning with different ways in which homeownership may affect self-employment outcomes.⁷

Several channels could account for a positive effect of homeownership and housing capital gains on self-employment. The first is that self-employment is a risky source of income and risk aversion declines with wealth. For these reasons, positive wealth shocks from housing capital gains could increase preferences for self-employment. This point has been stressed by Hurst and Wild Pugsley, 2011, Hurst and Pugsley, 2015) and Kerr et al. (2015).

A different mechanism, emphasized by Evans and Leighton (1989) and Evans and Jovanovic (1989), is that limited personal wealth and restricted access to credit may make it difficult for aspiring entrepreneurs to overcome startup costs, impeding access to self-employment for that reason.⁸ Evidence on this channel has been varied and controversial. Prior to 2000, much of the evidence in favor of this view drew on inheritances and other sorts of windfall financial gains (e.g. lottery winnings), treating these events as exogenous shocks to personal wealth.⁹ Hurst and Lusardi (2004), however, argued that in the U.S. personal wealth only affects self-employment for the top 5th percentile of the wealth distribution and also noted that low-wealth self-employed individuals are not disproportionately likely to concentrate in industries with low startup costs. For these and other reasons, they argue that personal wealth and access to credit have not been a significant impediment to self-employment for the typical entrepreneur.¹⁰

To further explore this issue both Hurst and Lusardi (2004) and Disney and Gathergood (2009) turned to house price inflation as an alternative exogenous driver of personal wealth; these are among the earliest studies we are aware of that considered a possible link between housing markets and self-employment.¹¹ Hurst and Lusardi (2004) drew on region-level measures of house price inflation (for the nine U.S. Census regions) based on the Federal Housing Finance Agency (FHFA) home price index, while Disney and Gathergood (2009) used county-level house price inflation in the UK. Hurst and Lusardi (2004) fail to find any evidence that house price inflation affects self-employment. Disney and Gathergood (2009) do find that rising house prices increase self-employment but with caveats.¹²

Several recent studies have revisited the role of access to credit as a driver of self-employment and business startups. Among these, Jensen et al. (2015) and Schmalz et al. (2017) consider government mortgage regulations that affect opportunities for homeowners to draw on home equity and mortgage debt to finance business activity. Jensen et al. (2015) consider the impact of a 1992 policy that allowed individuals in Denmark, for the first time, to use home equity and mortgage debt to finance non-housing expenditures including business expenses. They estimate that the policy change increased the number of small businesses in Denmark.¹³ Analogously, Schmalz et al. (2017) draw on the fact that mortgage policy in France does not allow homeowners with a mortgage to use their home as collateral against a small business loan. They then compare the impact of region-level house price inflation on owners without a mortgage, owners with a mortgage, and also renters where the latter is used primarily as a placebo and check on model specification. Results support expectations that house price inflation increases self-employment most for homeowners without a mortgage. Jensen et al. (2015) and Schmalz et al. (2017) both argue that their results are consistent with the view that collateral constraints and restricted access to credit impede business activity.

Corradin and Popov (2015) obtain analogous results using the U.S. SIPP panel. In their most robust model, they instrument for individual home equity using national-level home price appreciation since the time of home purchase, scaled by the local MSA housing supply elasticity (taken from Saiz, 2010). Based on that and other specifications, they estimate that rising individual home equity increases transitions into business ownership.¹⁴

Kerr et al. (2015) also consider the collateral channel using 2000–2004 US individual-level panel data from the Longitudinal Employer-Household Dynamics database (LEHD). They conduct a series of empirical exercises, including comparing owner-occupiers to renters (treating renters primarily as a placebo), examining differences across states with different levels of homeowner home equity protection in the event of a bankruptcy (e.g. Cao, 2014), splitting samples by industry based on those with and without high startup costs, and other modeling devices. They find robust evidence that housing capital gains encourage entrepreneurship, while attributing much of the estimated effect to factors other than collateral constraints including possibly wealth-induced shifts in preferences as noted above.

While debate about the extent to which credit access impedes self-employment persists, it is important to recognize that savvy small business owners are aware of the possibility that home equity and housing capital gains can provide a low cost, accessible source of funds and is sometimes an attractive substitute to a small business loan.¹⁵ A Google search on “using home equity to finance a business” yielded 6,860,000 hits as of August, 2016.¹⁶ Based on survey data from Barlow Research Inc., Shane, 2012, Shane, 2015) also reports that at the peak of the housing boom in 2006, 28% of small businesses had used home equity as a source of business finance, up from 18% in 2001. These anecdotes echo findings in Jones (1994) and Canner et al. (2002) who provide early evidence that homeowners often use mortgage debt to finance non-housing activities. Using the SIPP panel for U.S. households, Corradin and Popov (2015) also report that new small business owners take on additional home mortgage debt, consistent with using home equity to help cover business expenses. Estimates in our paper that self-employed homeowners are more likely to hold a HELOC is also consistent with this view.¹⁷

Different from above, Bracke et al. (2015) suggest a negative relationship between homeownership and self-employment. They argue that because homeownership and self-employment are both risky, homeowners may shy away from self-employment to limit their exposure to risk, similar in spirit to arguments in papers by Sinai and Soulelles (2005) and Davidoff (2006). While the qualitative nature of the portfolio-based argument is intuitive, the magnitude of the effect on tendencies for self-employment is less clear and could be dominated by other arguments described above.

A common feature of all of the studies above is the need to address various empirical challenges that complicate efforts to assess links between homeownership and self-employment. One such challenge is that higher wealth families typically occupy more expensive homes. This increases their potential for housing capital gains since capital gains increase with house value for a given rate of house price inflation. Because wealth may also increase taste and opportunities for self-employment, insufficient controls for personal wealth could cause estimates of the influence of housing capital gains on self-employment to be upward biased. A second challenge is that unobserved local labor demand shocks affect opportunities for wage work and related tendencies for self-employment. Those same shocks may also increase housing demand causing home prices to increase. This could affect estimates of the effect of capital gains although in this instance the direction of bias is less clear because positive labor demand shocks could push tendencies for self-employment in either direction.

We address these challenges using unique features of the 1985–2013 AHS panel in conjunction with a series of modeling strategies and identifying assumptions. Unique among panels, the AHS follows roughly 55,000 homes – not households – every two years. Each survey provides detailed information on the house and its current occupants, including self-employment status, income, whether the household owns or rents, mortgage attributes, and a battery of usual socio-demographic descriptors. The MSA in which a home is located is also reported for homes located in the 145 largest MSAs in the U.S.

In all of the estimation to follow we restrict the estimating samples to homes for which the MSA is identified in the data. In our simplest models, we then merge in annual measures of MSA-level house price inflation from the Federal Housing Finance Agency (FHFA) which are used to proxy for individual-level housing capital gains. In our more robust models, we replace MSA-level measures of house price inflation with MSA-by-year fixed effects that do much to difference away time varying unobserved labor demand shocks and other time varying MSA-level factors. Identification is then based on explicit and implicit interactions between MSA-level house price inflation and key indicators of the potential for housing capital gains, controls for which differ across specifications.

Our initial models explore the effect of MSA-level house price inflation on owner-occupiers relative to that of renters. These models treat renters as a control group since renters do not receive housing capital gains, mimicking some of the models in Corradin and Popov (2015), Kerr et al. (2015), and Schmalz et al. (2017). It is well known, however, that renters have less wealth than owner-occupiers and are more mobile. In addition, arguments in Bracke et al. (2015) suggest that renters may have different preferences for self-employment because of portfolio considerations. For these and other reasons highlighted later, our preferred models restrict the sample to just owner-occupiers.

Identification in the homeowner-only models relies on direct measures of individual housing capital gains and losses. This approach presents both opportunities and new challenges. One advantage is that individual-level housing capital gains vary extensively within a given metropolitan area, most importantly because capital gains are proportional to initial home value for a given level of house price inflation.¹⁸ That variation greatly increases power to identify the effect of housing capital gains on self-employment. Drawing directly on individual housing capital gains, however, requires that one subtract off recent expenditures on home maintenance and improvements. This is necessary to avoid overstating housing capital gains and also because home maintenance and improvements could be correlated with labor demand shocks that affect anticipated income. Unique among major surveys, the AHS panel includes detailed information on recent home maintenance and improvement expenditures (e.g. Harding et al., 2007). Our preferred models, therefore, are based on individual housing capital gains over the previous two years net of maintenance and improvement expenditures over that same period.¹⁹

Our primary challenges in using individual-level housing capital gains are encapsulated in two key assumptions that are necessary for identification. The first assumption is that the two-year lagged value of an individual's house is exogenous to subsequent transitions into or out of self-employment after conditioning on MSA-by-year fixed effects and the other controls. This allows us to use lagged house value as a proxy for individual wealth and also as a starting point from which subsequent housing capital gains are measured. A second key assumption is that within-MSA variation in house price appreciation is exogenous to self-employment transitions conditional on the model fixed effects and other controls. This is consistent with a view that labor demand shocks operate at the MSA level while within-MSA variation in home price growth is driven by primarily by shocks to neighborhood-level amenities.²⁰

We proceed as follows. Section 2 outlines our modeling strategy in more detail. Section 3 describes our data and summary measures. This includes comparisons of self-employment rates in the AHS to analogous measures in decennial Census and ACS data, datasets that are more commonly used for labor market studies. Section 4 presents estimates of the self-employment models while Section 5 considers the effect of self-employment on the tendency to hold a primary mortgage and HELOC. Section 6 compares estimates for younger versus older individuals since older individuals are wealthier and may exhibit different tendencies for self-employment for that reason. We conclude with Section 7.