Abstract
This paper examines the impact of changes in housing affordability on regional entrepreneurship. Two-way fixed-effects estimates suggest an increase in the level of house prices in a commuting zone results in a decline in establishment openings as a share of existing establishments—consistent with a crowding-out effect. In contrast, an increase in the growth rate of house prices results in a small (although not always statistically significant) increase in establishment openings—consistent with a positive wealth effect from capital gains. To address endogeneity concerns, the authors adopt two alternative instruments for commuting zone house-price growth: a measure of local real estate lending and a geography-based measure of the elasticity of local housing supply. They extend the analysis using restricted-use establishment-level microdata from the Quarterly Census of Employment and Wages (QCEW) for the state of Colorado. Results from the QCEW data are consistent with those from the commuting zone sample.
Entrepreneurship is widely recognized as a crucial ingredient in economic growth. Schumpeter (1942) described entrepreneurial dynamism as a key driver of business cycles and ultimately of development; the connection between entrepreneurship and growth has found more recent empirical and theoretical support in Audretsch et al. (2006), Wennekers and Thurik (1999), and Baumol and Strom (2008). Entrepreneurship has also been shown to contribute to growth specifically at a regional scale, both through direct job creation and through linkages and information spillovers (e.g., Acs & Armington, 2006; Acs et al., 2009, 2013; Audretsch & Keilback, 2007; Bunten et al., 2015; Lee, 2017).
To the extent that entrepreneurship is growth enhancing at a local level, it is important to understand why some regions have higher rates of entrepreneurial activity. Fritsch and Wyrwich (2019) argued that history is a major factor: a region's “entrepreneurial culture”—the informal institutions that encourage or discourage risk- taking and innovative thinking—is highly persistent over time. Specific policies aimed at enticing new business formation, such as start-up incubators or tax breaks and other incentives for new or relocating firms, might also play a role, although empirical examinations generally have found such policies to be relatively inefficient (e.g., Bartik, 2019; Harper-Anderson & Lewis, 2017; Madaleno et al., 2018).
This paper focuses on the effects of regional housing markets on entrepreneurial activity. A substantial literature considers the effect of the local composition of demand on patterns of business locations and openings. Hotelling’s (1929) model of spatial competition, in which competing firms strategically co-locate centrally across an evenly spaced array of consumers, illustrates the “pull” that demand can exert on business-location decisions. Hotelling's original model assumed spatially homogeneous and perfectly inelastic demand; however, spatially differentiated and/or elastic demand may pull firms more strongly toward areas in which consumers have a higher willingness to pay (see Graitson, 1982 for a summary). In the context of home prices, this suggests that firms may prefer to open in high-income areas, which are typically characterized by high real estate prices. Neighborhood characteristics, including incomes and home values, have been shown to impact firm locations. Schuetz et al. (2012) found that retail employment density increased with neighborhood income, while Meltzer (2016) and Meltzer and Capperis (2017) found that increases in neighborhood home prices increase business churn, including an increase in establishment births.
In addition to shifting product demand, changes in home prices might affect entrepreneurial patterns through impacts on entrepreneurs themselves. Higher home prices may positively impact establishment openings on-net through a wealth effect. A substantial literature found that house-price shocks impact entrepreneurial decisions through liquidity channels, with rising home prices increasing the real net wealth and access to equity-financed capital of would-be entrepreneurs (e.g., Chaney et al., 2012; Evans & Jovanovic, 1989; Schmalz et al., 2017). Harding and Rosenthal (2017) have shown that a 20% real increase in home value over a 2-year period raises the likelihood of entry into self-employment by roughly 1.5 percentage points.
However, rising house prices may have an offsetting effect on entrepreneurship if they result in a decline in housing affordability. All else equal, households with lower residual income net of housing costs likely have fewer resources to devote toward entrepreneurial activities. In such cases, housing costs may crowd out latent entrepreneurial ability. Additionally, firms may have to pay workers higher nominal wages in more expensive housing markets (Saks, 2008), making these areas less appealing from the perspective of an existing firm considering opening a new branch. Therefore, it remains an important task to distinguish between the two effects of changes in house prices on new establishment openings: a positive wealth effect through the capital gains received by homeowners, and a negative effect on the consumption of nonhousing goods through the impact of positive house-price shocks on renters. (See the online Appendix for a more detailed exposition of this theoretical framework.)
To the extent that the decline in housing affordability in high productivity areas is a secular trend, prima facie evidence suggests the negative affordability effect may outweigh the positive capital gains effect. As a share of existing establishments, both establishment openings and closures have experienced a secular decline since the late 1970s (Figure 1), a trend exacerbated by the Great Recession.
Trends in business dynamism.
To assess the impact of the capital gains and affordability effects on entrepreneurship, we adopt a two-tiered approach. First, using commuting zone level data for the period 2005 through 2014 we examine the impact of house-price shocks on the number of establishment openings. We adopt two separate measures to track changes in house prices: (1) The Federal Housing Finance Agency (FHFA) House Price Index, a weighted repeat-sales index of single-family house prices, and (2) the median share of monthly income spent on rent by renter households in a given commuting zone, constructed using the Integrated Public Use Microdata Series (IPUMS)/American Community Survey (ACS) microdata (Ruggles et al., 2019). Our measures of establishment openings and counts of establishments with growing employment are drawn from the Census Bureau's Statistics of U.S. Businesses (SUSB).
Using several estimation strategies, including instrumental variables approaches based on local real estate lending and housing supply elasticity, we find that declines in housing affordability—measured by either the share of rent in monthly income or the level of house prices—are negatively associated with establishment openings and positively associated with closures, consistent with a crowding-out effect. However, the growth rate of house prices is usually positively associated with establishment openings (although with some deviation across specifications) and negatively associated with closures, consistent with a positive wealth effect. Importantly, the wealth effect varies with the rate of local homeownership: interacting the growth rate of house prices with the local homeownership share shows a positive marginal effect on the rate of house-price growth only in labor markets with the highest homeownership rates, and negative or statistically insignificant effects at the sample mean level of homeownership.
Second, because the impact of changes in housing affordability may differ across establishment types, we extend our results by leveraging unique restricted-use microdata from the Quarterly Census of Employment and Wages (QCEW) for the state of Colorado—a fast-growing state characterized by a dynamic urban corridor in which home values are rising, and with varied rural economies with specializations in tourism, agriculture, and extractive industries. The longitudinal QCEW data allow us to track the behavior of individual establishments over time and distinguish whether changes in home prices matter more for de novo establishments—newly created, single-unit establishments—or new establishments belonging to multiunit structures.
Overall, our results support the contention found elsewhere in the literature that declines in housing affordability—to the extent they are caused by restrictions on housing supply—may negatively impact labor market efficiency (here, through the entrepreneurship channel), and suggest that local policy makers who wish to take advantage of the potential positive effects of house-price growth would do well to implement policies that make homeownership accessible to a wider swath of the population. Analysis of the longitudinal microdata in Colorado suggests that rising home prices increase the growth rate of de novo establishments, as homeowners’ increased equity financing capacity boosts the rate of new start-ups. Meanwhile, longer-run upward trends in home prices put downward pressure on the share of multiunit businesses expanding into Colorado counties, perhaps due to firms’ reluctance to branch into areas where labor costs are rising.
Data and Summary Statistics
Commuting Zone Sample
Housing Affordability
Three main variables are used to summarize local housing market conditions. First, we use both the level and growth rate of the FHFA House Price Index. We aggregate the county-level FHFA measures to commuting zones using the crosswalks provided by Autor and Dorn (2013). Commuting zones are clusters of counties characterized by strong commuting ties and were initially defined by Tolbert and Sizer (1996) to construct a geographic unit that best captured the notion of a local labor market. We focus on commuting zones because of our interest in the relationship between housing affordability and local labor market conditions.
Second, we use IPUMS/ACS microdata (Ruggles et al., 2019) for 2005 through 2014 to construct an estimate of the median share of rent in monthly income paid by renter households at the commuting zone level. We focus on the share of rent in income for renter households in the bottom quintile of the household income distribution, as it is these households for which the budget constraint (and thus the crowding-out effect) is likely to bind the most; lower-income households consistently spend a larger share of their income on housing (Schuetz, 2019). The median share of rent in monthly income provides an alternative measure of affordability to the level of house prices—given by the FHFA Index—although they tend to be correlated in practice. Figures 2a through 2c present trends for all housing market variables over the sample period. Figure 3 maps the rental share variable across commuting zones. Because the FHFA data do not contain information on all counties, our final data set consists of 671 of the 722 commuting zones making up the contiguous United States for which we have information on housing market conditions.
Commuting zone house-price trends. (a) Share of rent in monthly income, (b) FHFA House Price Index, and (c) %Δ FHFA Index.
Commuting zone housing affordability map.Notes. Data from IPUMS/ACS. Commuting Zone Crosswalk from Autor and Dorn (2013). The map presents an average of the median share of rent in monthly income for renters in the bottom quintile of the income distribution for the 2005–2014 period.
The figures reveal four key features of the data. First, despite the decline in house prices during the Great Recession, housing affordability—measured by the share of rent in monthly income—has steadily declined for the poorest quintile of income earners. Second, house prices display a positive, long-run secular trend. The FHFA House- Price Index is normalized to 100 in 1986, meaning that prices in 2014 were approximately 2.4 times higher than the two previous decades. Third, the rate of house-price growth exhibited a relatively quick recovery after the Great Recession, returning to an annual growth rate of nearly 5% by 2014. Finally, the geographic variation in the rent-to-income ratio depicted in Figure 3 shows housing cost burdens concentrated in coastal areas, some inland areas in the western United States such as Colorado and the Las Vegas metro area, and much of the rural West and Mountain West.
Business Dynamism
Data on business dynamism are drawn from the SUSB. We aggregate the SUSB data from counties to commuting zones, again following Autor and Dorn (2013). The SUSB data allow us to track several main variables of interest. First, as separate measures of dynamism, we use the share of establishment openings and closures in existing establishments, respectively. Second, because openings and closures primarily represent changes on the external margin, we also look at expansions and contractions of existing establishments as measures of inframarginal change. 1
Several additional points about the dynamism data are worth making here. First, Figure 4 maps establishment openings as a share of existing establishments across commuting zones in our sample for 2005 through 2014 (effectively redrawing Figures 2a–2c for sample commuting zones). Figure 4 makes clear that the most dynamic areas have significant overlap with the areas that are least affordable, according to Figure 3. However, other commuting zone-specific factors may be correlated with both housing market and labor market outcomes. Thus, Figures 5a through 5c plot the relationship between measures of housing affordability and establishment openings, after controlling for a commuting zone-specific fixed effect. The figures indicate that housing affordability tends to be positively related to the share of openings in existing establishments, consistent with the presence of a crowding-out effect. The rate of house-price growth also appears correlated with openings, consistent with a positive wealth effect.
Establishment openings by commuting zone.Notes. Data from IPUMS/ACS. Commuting Zone Crosswalk from Autor and Dorn (2013). The map presents an average share of establishment openings in existing establishments for the 2005–2014 period.
Housing affordability and establishment openings. (a) Openings and rent, (b) openings and house prices, and (c) openings and price growth.Notes. Figures depict data from IPUMS/ACS, SUSB, and FHFA. Each figure residualizes the data on commuting zone fixed effects, (log) population, and (log) employment.
Additional Data
Controls for demographic characteristics—including average age, population shares by race/ethnicity, and educational attainment—and economic characteristics—including average household income, employment shares by industry, and the unemployment rate—come from the IPUMS/ACS microdata. Additional data on population and employment are obtained from the Bureau of Economic Analysis (BEA) Regional Economic Accounts. We construct controls for real estate lending volume—including separate controls for multi- and single-family lending, as well as commercial lending—at the commuting zone level using the Federal Deposit Insurance Corporation's (FDIC) Historical Bank Data database. We estimate commuting zone lending as the state-level lending volume from the FDIC data multiplied by the share of the state's population residing in a commuting zone. Finally, we make use of land-based housing supply elasticity measures from Saiz (2010). The Saiz (2010) measures were originally constructed at the metropolitan-area level. To convert the elasticities to the commuting zone level we take the average elasticity across metropolitan areas in a state and assign it to all commuting zones in that state. Sample means for all variables are presented in Table 1.
Sample Means.
Note. Sample means weighted by average commuting zone population.
SUSB = Statistics of U.S. Businesses; FHFA = Federal Housing Finance Agency; BEA = Bureau of Economic Analysis; FDIC = Federal Deposit Insurance Corporation.
QCEW Sample
Establishment Data
Restricted-use establishment-level microdata from the QCEW for the state of Colorado 2 is used to construct variables capturing trends in establishment growth by ownership structure. The QCEW provides quarterly counts of employment and wages for employers covering more than 95% of U.S. jobs. Our data cover almost the entire universe of establishments operating in the state of Colorado. We distinguish between new establishments belonging to multiestablishment employers already operating in the state of Colorado, and new single-unit (de novo) establishments. Importantly, in the Colorado QCEW, new single-unit establishments are inclusive of newly incorporated firms belonging to self-employed individuals, new employer establishments belonging to firms that previously employed no workers, and new establishments belonging to multiunit establishments that do not operate any other establishments within the state of Colorado (but may operate establishments in other states). We aggregate counts and growth rates of each type of establishment to the county level for all 64 Colorado counties, spanning 2004 Q1 to 2018 Q2. These data also allow us to construct measures of average wages and total employment at the county level.
Figure 6 presents trends in both de novo establishments and establishments belonging to multiestablishment employers over the sample period. The figure suggests that the effects of the initial housing market boom were sustained longer for establishments belonging to multiestablishment structures, but that new expansions into Colorado have done better in the recovery.
Establishment trends by type.Notes. Data from restricted-use QCEW data for the State of Colorado. Figure presents trends in the number of establishments by type, normalized to 100 in 2004Q1.
Figure 7 maps the share of de novo establishments by county in Colorado. The share of single-unit establishments is higher in the central and southern mountains and lowest in the eastern plains. Densely populated Front Range counties tend to have moderate de novo shares in the 80% to 90% range, in keeping with the state average. But the de novo share has grown most rapidly along the Front Range, as seen in Figure 8.
The average share of “de novo” (single-unit) establishments, 2012–2018.Notes. Data from restricted-use QCEW data for the State of Colorado. Figure presents the share of de novo establishments in total establishments for Colorado counties.
Percentage point growth in “de novo” (single-unit) establishment share, 2012–2018.
Housing Affordability
We again make use of the FHFA House Price Index to measure housing affordability. While the FHFA makes available a county-level version of its index, it does so only on an annual basis and is therefore not useful for comparison with the QCEW data. In contrast, the FHFA's three-digit zip-code index is published on a quarterly basis. Thus, we aggregate the three-digit zip-code index to the county level using the adjustment factors provided by the Missouri Census Data Center's geographic correspondence engine. 3 Figure 9 maps the average house-price growth in Colorado counties from 2012 Q1 to 2018 Q2. The figure indicates prices are growing fastest in the Denver metropolitan area and along the Front Range corridor.
Average quarterly house-price growth in Colorado counties, 2012–2018.Notes. Data aggregated to the county level from the FHFA's Quarterly Three-Digit Zip-Code Index.
To take advantage of the quarterly nature of the data and to further assess the role of house prices in driving changes in establishment openings, we separate the growth rate of the FHFA Index into its trend and cyclical components. 4 To illustrate, Figure 10 plots both the actual and trend growth rate for Denver County.
Actual and trend component of price growth, Denver County. Notes: Data aggregated to the county level from the FHFA's Quarterly Three-Digit Zip-Code Index. Trend component obtained using a Hodrick-Prescott filter (
Estimation Strategy
Commuting Zone Sample
To assess the impact of housing affordability on establishment openings and business dynamism, we adopt the following two-way fixed-effects approach:
We carefully consider concerns of reverse causality. Alonso’s (1964) model of location, and the substantial bid-rent literature that follows, posit that households and firms prefer to locate near hubs of economic activity. By extension, an increase in a place's economic activity would be expected to increase demand for land in booming regions, pushing up house prices. Recent empirical work (e.g., Chapple et al., 2010) found empirical support for this causal channel, particularly in regions with rapid growth in high-paying industries.
To address possible endogeneity between house-price movements and establishment openings and closures, we focus on instrumenting for the growth rate of house prices, because this variable appears positively related to openings in our initial analysis 5 and we are interested in testing whether the negative impacts of declining affordability may nonetheless be offset by the positive effects caused by additional establishment openings. We use two such instruments. First, we use data from the FDIC Historical Statistics on Banking to construct an estimate of real estate lending at the commuting zone level. Intuitively, increases in residential real estate lending should only be correlated with entrepreneurship—after controlling for the level of commercial lending—through their impact on house prices. Second, we make use of land-based housing supply elasticity measures constructed by Saiz (2010). We use exogenous variation in local house-price growth created by the interaction between the Great Recession and local housing supply elasticities. As long as the geographic factors determining the Saiz (2010) housing supply elasticity are uncorrelated with unobservables related to establishment openings and house-price growth, the variation in house prices caused by the interaction between the Great Recession and the elasticity of housing supply remains a valid instrument.
QCEW Sample
We use the Colorado restricted microdata from the QCEW to separate out the effect of house prices on the growth rate of new entrants at the state level versus the growth rate of establishments belonging to incumbent firms. De novo establishments, most of which are very small at the time of entry, are unlikely to exert any demand-push influence on housing prices. As a result, causality in this relationship is likely to run from home prices to de novo establishments rather than the reverse. We estimate this model using the following specification:
Results
Commuting Zone Results
Table 2 presents results from estimating Equation (1) on our commuting zone (CZ) sample using ordinary least squares (OLS) for openings, closures, and overall dynamism (openings plus closures) as a share of existing establishments.
Estimation Results 1.
Notes. Standard errors in parentheses, clustered at the commuting zone level. *p < .10, **p < .05, ***p < .01. Columns 1 through 4 dependent variable is establishment openings as a share of existing establishments. Column 5 dependent variable is establishment closures as a share of existing establishments. Column 6 dependent variable is the sum of openings and closures as a share of existing establishments. Control variables include industry employment shares, (log) population, employment, and average household income, population, and employment growth.
Beginning with the specification in Column 1, our results suggest an initial positive relationship between house-price growth and openings, but a negative relationship between openings and our measures of housing (un)affordability (although the coefficient on the level of the FHFA Index is statistically insignificant). In both cases, the estimated relationship is economically small in magnitude. Column 1 suggests that a 1 percentage point increase in the rate of house-price growth increases openings as a share of existing establishments by about one-tenth of 1 percentage point (or about 0.04 standard deviations). A 1 percentage point increase in the median share of rent in monthly income for households in the bottom quintile of the income distribution decreases openings by about one-third of one-tenth of a percentage point.
As additional controls and fixed effects are added in Columns 2 through 4, the impact of declining affordability on openings becomes statistically insignificant and the magnitude of the impact of house-price growth falls (although remains statistically significant). In contrast, even with a battery of controls and fixed effects, both the growth rate and level of house prices appear to have a statistically significant impact on establishment closures. A higher rate of house-price growth is associated with a lower share of closures in existing establishments, while a higher level of house prices is associated with a greater rate of closures (potentially from decreased demand at local establishments due to crowding out in the budget constraint). A 1 percentage point increase in the rate of house-price growth is associated with a decline in the share of closures in existing establishments by about 0.05 percentage points. A 1 log point increase in the level of the FHFA Index increases the share of closures in existing establishments by about 3.6 percentage points. This implies that a 1 standard deviation increase in the log of the FHFA Index (0.4) increases the share of closures in existing establishments by 1.44 percentage points, or about a 0.70 standard deviation—a rather large effect. The increase in closures from declining affordability appears to far outweigh the spillovers from the capital gains effect. Finally, Column 6 reveals one of the weaknesses of our measure of dynamism (openings plus closures), in that it suggests declining affordability (as measured by the level of the FHFA Index) is associated with greater overall dynamism. One might be inclined to interpret this as a positive outcome; however, our results in Columns 1 through 5 suggest the increase in dynamism is driven entirely by an increase in closures.
Tables 3 and 4 extend the estimates from Table 2 to account for inframarginal changes—expansions and contractions of employment at existing establishments, state-specific time trends, state-by-year fixed effects, and the interaction between homeownership and house-price growth. (Note that “expansions” in Table 3 refer to counts of “establishments that have positive first-quarter employment in both the initial and subsequent years and increase employment during the time period between the first quarter of the initial year and the first quarter of the subsequent year” [SUSB], 6 not to the creation of new establishments by existing firms.)
Estimation Results 2—Extended OLS Results for Openings and Expansions.
Notes. Standard errors in parentheses, clustered at the commuting zone level. *p < .10, ** p < .05, ***p < .01. Columns 1 through 3 dependent variable is openings as a share of existing establishments. Columns 4 and 5 dependent variable is expansions as a share of existing establishments. OLS = ordinary least square.
Estimation Results 3—Extended OLS Results for Closures, Contractions, and Dynamism.
Notes. Standard errors in parentheses, clustered at the commuting zone level. *p < .10, **p < .05, ***p < .01. Columns 1 and 2 dependent variable is closures as a share of existing establishments. Columns 3 and 4 dependent variable is contractions as a share of existing establishments. Columns 5 and 6 dependent variable is establishment dynamism (openings and closures as a share of existing establishments). OLS = ordinary least square.
The results in Tables 3 and 4 indicate a pattern similar to Table 2. The growth rate of the FHFA Index is positively associated with openings and expansions, and negatively associated with closures and contractions (albeit in an economically small fashion). In contrast, while the level of house prices has no statistically significant impact on openings, it has a statistically and economically significant negative effect on expansions, and a positive effect on closures and contractions. The inframarginal effects (on expansions and contractions) are consistent with crowding in the budget constraint as higher house prices result in lower expenditures on other goods, such as retail. Finally, interacting the rate of house-price growth with commuting zone homeownership reveals that the positive effect of house-price growth on openings is mitigated by the residential tenure. House-price growth has a negative effect on openings at low levels of homeownership but becomes increasingly positive as the share of homeowners increases. For ease of interpretation, Figure 11 plots the marginal effect of house-price growth on openings across levels of homeownership. The figure shows that house-price growth has a statistically insignificant impact on openings at the sample mean level of homeownership.
The marginal effect of price growth on openings, by homeownership share.Notes. Figure presents the marginal effect of house-price growth on openings, using the estimates from Table 3. Dashed line is plotted at the sample mean level of homeownership.
Instrumental Variables
Table 5 presents results from our banking instrumental variable (IV) specification. We use the level and growth rate of per-branch residential real estate lending, as well as the growth rate of per-branch multifamily real estate lending, as instruments for the rate of house-price growth. We leave commercial (nonreal estate) lending as a control in each specification.
Instrumental Variable Results 1—Banking IV.
Notes. Standard errors in parentheses, clustered at the commuting zone level. *p < .10, **p < .05, ***p < .01. Table presents instrumental variables results, where the level and growth rate of per-branch residential real estate lending, as well as the growth rate of per-branch multifamily real estate lending, are used to instrument for the growth rate of the FHFA price index. Columns 1 through 5 use openings, closures, openings plus closures, expansions, and contractions as the dependent variable, respectively. FHFA = Federal Housing Finance Agency.
For instrumental variables estimates to be valid, the instruments must be highly correlated with the endogenous variable and orthogonal to the regression error term. As a rule of thumb, Staiger and Stock (1997) suggested that a first-stage F-statistic of at least 10 is required to avoid problems associated with weak instruments. The orthogonality assumption can be tested using the Sargan-Hansen test. The Sargan-Hansen test assesses the null hypothesis that the instruments are uncorrelated with the errors, provided that the first-stage equation is overidentified (the number of instruments exceeds the number of endogenous variables). A rejection of the null hypothesis indicates that the orthogonality assumption is violated. We report both the first-stage F-statistic and the p-value from the Sargan-Hansen test. In each case, the first-stage F-statistic exceeds 10 and we are unable to reject the null hypothesis that the instruments are uncorrelated with the errors.
Given that our instruments satisfy the necessary restrictions, the results in Table 5 add to the body of results obtained above, suggesting that, when the endogeneity of house-price growth to openings is addressed using instrumental variables, the rate of house-price growth has an economically small, significant negative effect on openings. For each of the other measures of establishment activity, the signs remain the same as the OLS estimates, with statistically significant effects for dynamism (driven by the decline in openings) and expansions.
Table 6 provides an alternative set of IV estimates, instrumenting for house-price growth using the interaction between an indicator for the Great Recession and the Saiz (2010) housing supply elasticity. The first-stage equation is exactly identified, so the orthogonality assumption cannot be tested directly. However, Mian and Sufi (2009, 2010, 2011) conducted extensive tests of the exclusion restriction for the Saiz (2010) elasticity as an instrument and came to the conclusion that the primary channel through which it impacts other economic variables is through its impact on house-price growth. The first-stage F-statistic for the Saiz (2010) instrument far exceeds 10.
Instrumental Variable Results 2—Housing Supply Elasticity IV.
Notes. Standard errors in parentheses, clustered at the commuting zone level. *p < .10, **p < .05, ***p < .01. Table presents instrumental variables results, where the interaction between an indicator variable for the Great Recession and the Saiz (2010) elasticity is used to instrument for house-price growth.
Table 6 further confirms our main results. First, the effect of house-price growth on openings is close to zero and statistically insignificant—in line with previous results, which find small and/or negative effects (in particular, the prior IV results and the specification in which house-price growth is interacted with homeownership). Second, the sign of the effect of house-price growth on closures, expansions, contractions, and dynamism remains the same as all previous specifications (although in this case the dynamism effect is driven by the decline in closures associated with house-price growth).
QCEW Sample
In this section, we leverage the QCEW restricted-access Colorado data to test a more nuanced version of our hypothesis. Recall that theory suggests two opposing channels between home prices and establishment births: a liquidity channel, whereby rising prices increase equity and ease credit constraints for would-be entrepreneurs; and an affordability channel through which rising prices squeeze residual incomes but may also limit local labor supply or require firms to pay a compensating wage. The positive liquidity effect may be more impactful on the formation of new business while the affordability effect may discourage existing businesses from expanding into commuting zones with rising home prices.
We first estimate Equation (2), testing the relationship between changes in home prices on the growth rate and shares of de novo and multiunit establishments on an annual basis. The use of a 1-year, rather than quarterly, time period allows for the use of additional controls that are available on an annual basis: the ratio of nonemployer establishments to wage and salary workers, which proxies for the self-employment rate (U.S. Census Nonemployer Statistics); poverty rates and median household incomes (U.S. Census Small Area Income and Poverty Estimates); and a demand shock instrument following Bartik (1991) that estimates local employment growth by applying national industry growth rates to county-level industry employment compositions.
Results in Table 7 support the liquidity channel hypothesis in that rising home prices tend to increase the growth rate of de novo establishments increases. Estimates on the effect of home price growth on the growth of multiunit establishments are negative but not significant. 7
Annual Panel Results.
Notes. Standard errors in parentheses, clustered at the county level. *p < .10, **p < .05, ***p < .01.
Table 8 presents results from a quarterly panel applying one-quarter lags on all explanatory variables, with changes in the FHFA House Price Index disaggregated into a cyclical and trend component as detailed above. Columns 4 and 5 reinforce our finding that house-price growth, both cyclical and trend, significantly increases the growth rate of de novo establishments, supporting the hypothesis that home-equity gains are contributing to new entrepreneurial activity. Columns 1 and 2 show no significant effect of home prices on the growth of multiestablishment businesses in Colorado counties.
Quarterly Panel Results.
Notes. Standard errors in parentheses, clustered at the county level. One-quarter lags are applied to all explanatory variables. *p < .10, **p < .05, ***p < .01.
However, Column 3 shows that positive trends in home prices decrease the share of multiestablishment businesses in a county. Importantly, this effect suggests that existing businesses may shy away from branching into countries in which housing costs—and in turn labor costs—are rising. Much of Colorado, particularly metro Denver and the broader Front Range, is experiencing substantial and sustained increases in home prices. These results suggest that, while rising home values may give some potential entrepreneurs the equity needed to start a new business, the lack of affordable housing may deter existing firms looking to expand.
Discussion and Conclusion
Openings of new establishments, either de novo or branching from existing multiunit businesses, are an indispensable driver of employment growth. This paper seeks to shed light on the effects of home prices on local entrepreneurial activity. We find mixed results: rising home values have some positive impacts on local business growth, increasing establishment openings and expansions and marginally decreasing closures and contractions. These effects are strongest in areas with high rates of homeownership, suggesting that home price growth increases entrepreneurship through a liquidity channel whereby homeowners’ access to equity financing increases as home values rise. Using restricted-use data for Colorado, we further find that rising home prices boost the creation of de novo establishments in a county and have no positive effect on the proliferation of establishments from multiunit businesses, further supporting the hypothesis that home price growth drives local entrepreneurship by increasing access to credit.
However, increases in the level of the FHFA Index tend to decrease overall establishment dynamism—the gross rate of establishment openings and closures in a county. Dynamism, net of direct employment effects from openings and closures, has been found elsewhere to contribute to economic growth by generating information spillovers and improving labor market matching. Home price growth also decreases establishment openings in areas with lower homeownership rates. Especially in cities, where homeownership rates tend to be lower, rising home prices are unlikely to spark entrepreneurial activity, and may instead crowd out the creation of new businesses due to rising costs of living. As such, particularly in areas with low homeownership rates, our results lend support to policies aimed at increasing affordability.
This work illuminates an important distributional effect whereby homeowners (who are more than twice as likely as renters to have household incomes above $75,000; Joint Center for Housing Studies of Harvard University, 2017) benefit from increased equity when home prices rise, while renters face increased costs. For the real incomes of renters to rise as a result of home price growth, increased labor demand from equity-led entrepreneurship would need to increase wages by enough to offset rising housing costs.
While this work cannot specifically answer whether this condition is met, our findings shed light on important dimensions of housing policy. Most directly, our results lend support to policies aimed at increasing homeownership—in turn amplifying the positive effects of home price growth on entrepreneurship—and at translating establishment births and growing labor demand into wage gains. Data from the U.S. Census Housing Vacancy Survey reports a national homeownership rate of 65.1% in the fourth quarter of 2019, down roughly 5 percentage points from its peak in 2007, although prerecession rates were artificially inflated by subprime lending. Importantly from a distributional standpoint, the gap between White and non-White homeownership rates stands at 18 percentage points. Efforts to increase access to homeownership, especially among groups who have faced historical discrimination in housing markets (Rothstein, 2017) could promote both efficiency and equity in translating home price increase into entrepreneurship.
Our finding that higher levels of home prices tend to decrease the formation of new local branches of multiunit establishments also has implications for housing supply policy. Hilber and Robert-Nicoud (2013) and Ortalo-Magne and Prat (2014) have shown—in alternate modeling frameworks—that when building permits are issued by local zoning boards subject to lobbying by homeowners, housing will tend to be undersupplied in equilibrium (echoing the home voter hypothesis of Fischel, 2001). Desmond (2018) showed that approximately 52% of all poor working families in the United States spend over half their income on housing, and one in four of these families spend over 70% of their income on rent and utilities. This is due, at least in part, to policies that artificially restrict the supply of housing. Glaeser and Gyourko (2018) argued that the implicit tax on housing development created by regulations is higher in many areas than any reasonable estimate of the externalities associated with new housing construction. Using data on the Low-Income Housing Tax Credit (LIHTC), Diamond and McQuade (2019) showed that additional LIHTC development in high-income areas successfully lowers the price of housing and attracts lower-income residents, reversing some of the negative effects of supply restrictions.
Empirically, zoning practices such as density restrictions increase home prices and decrease affordability (see Quigley & Rosenthal, 2005, for a meta-analysis), which also tends to increase residential segregation (Lens & Monkkonen, 2015). While such outcomes may be in the narrow interest of those homeowners who gain from elevated property values, they have significant adverse effects: Ganong and Shoag (2017) found that home price increases in high productivity areas drive out low-income residents, contributing to geographic income disparities. And Hsieh and Moretti (2019) showed that restrictions on new housing supply in high productivity areas like San Francisco and New York limit workers’ access to increased productivity resulting from agglomeration economies. Using a spatial equilibrium model, the authors contend that supply constraints lowered aggregate U.S. growth by 36% between 1964 and 2009.
These issues have attracted national public press and policy attention. Between April and June of 2019, The New York Times published at least 13 articles and opinion pieces on affordable housing. 8 The 2020 presidential candidate Elizabeth Warren made her proposal for the federal government to expand its investment in affordable housing—to the tune of $500 billion—a central part of her campaign. A 2019 California measure—Senate Bill 50—attempted and failed to shift control of a development from local to state government for the purpose of expanding the supply of housing in high-cost areas. Oregon has taken steps toward eliminating single-family zoning, while President Biden's housing plan calls for “eliminat[ing] local and state housing regulations that limit affordable housing options and contribute to urban sprawl” (Joe Biden for President: Official Campaign Website, 2020). Our results suggest that—especially in concert with efforts to increase access to homeownership—policies that exert downward pressure on housing prices expanding supply could also stimulate entrepreneurial dynamism.
Our work also highlights the effects of factors such as homeownership rates on the connection between home prices and entrepreneurship. Moreover, we illustrate that entrepreneurial dynamism, homeownership rates, and housing cost burdens vary substantially across the country, making region-specific analyses, like the one we conduct here for Colorado, potentially informative tools for policy-makers. Several additional follow-on studies may be instructive. This paper differentiates between de novo and multiestablishment firms, but does not examine industry-specific entrepreneurship, which may be important in distinguishing locally serving business ventures from businesses producing tradable goods that may be less dependent on local demand. Future work might also evaluate the effects of home prices on self-employment in addition to establishments with one or more employees.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.