Liquidity constraints, household wealth, and entrepreneurship revisited

Abstract

A large body research shows a positive relationship between wealth and entrepreneurship and interprets the relationship as providing evidence of liquidity constraints. Recently, however, the liquidity constraint interpretation has been challenged because of the finding that the relationship between business entry rates and assets is flat throughout most of the asset distribution and only rises dramatically after this point (Hurst and Lusardi 2004). We reexamine the liquidity constraint hypothesis in three ways. First, we demonstrate that examining the relationship separately for those who experience a job loss and those who do not reveals generally increasing entry rates through the wealth distribution for both groups. Based on the entrepreneurial choice model of Evans and Jovanovic (1989), these two groups face different incentives, and thus have different solutions to the entrepreneurial decision. We also find evidence of a stronger relationship between entrepreneurship and a different measure of wealth - net housing equity - for the two groups. Second, we examine the liquidity constraint hypothesis using a two-period simulation model that extends the Evans and Jovanovic (1989) model. The model shows how exogenous wealth shocks can be used to accurately identify the presence of liquidity constraints even allowing for endogenous saving and correlated abilities. Third, we provide new evidence from matched Current Population Survey (1993-2004) data to study whether changes in housing prices affect self-employment entry. We find that housing appreciation measured at the MSA-level is a significantly positive determinant of entry into self-employment after controlling for changes in local economic conditions and other factors. Our estimates indicate that a 10 percent annual increase in housing equity increases the mean probability of entrepreneurship by 17 percent and that the effect is not concentrated at the upper tail of the distribution. These estimates are not sensitive to controlling for pre-existing trends in housing prices suggesting that the results are not being driven by expected local economic growth.