Summary of Tenure Choice and Labor Market Outcomes

This paper tested at the individual level the argument that regional homeownership rates are positively correlated with regional unemployment rates.

The null hypotheses were based on search theory which argued that renters have a higher matching rates to firms while home owners, due to fixed cost in property acquisition and immobility, suffer more from regional demand shocks. Specifically, the hypotheses contends that renters, compared to homeowners, have lower unemployment probabilities, shorter duration of unemployment and higher wages.

Some previous research papers showed that at aggregate level house ownership has a negative impact on labor market outcomes. However, these papers failed to present evidence at individual level. Besides, much of the current evidence is bivariate, which means there may be omitted variables involved. To test hypotheses at individual level, Coulson and Fisher used two data sets, namely the March 2000 wave of Current Population Survey (CPS) and 1993 wave of the Panel Survey of Income Dynamics (PSID), with a multivariate model.

The CPS data included information on 29753 individuals’ housing, labor market experiences and demographic characteristics. The sample reflected the relatively wealthier portion of the country. Coulson and Fisher first used probit model, in which the dependent variable is a dummy variable, to test the hypothesis that homeowners are more likely to be unemployed. The univariate test turned out to reject the hypothesis. To tease out the impact of omitted variables, Coulson and Fisher then added some other significant variables such as age, races, education levels, marital status, and locations, but refined regression still rejected the hypothesis.

Coulson and Fisher then used CPS data and OLS method to test the hypothesis that renters have more wages. Both the unconditional and conditional tests showed that home owners have more wages than renters, which rejected the hypothesis.  

Having rejected the first and third hypothesis using CPS data, Coulson and Fisher used PSID to test the three hypotheses, particularly the second that renters have shorter duration of unemployment. Compared to CPS, which only included a simple cross-section, PSID had longitudinal data, which could be better used to test spell length of the sample. The sample consisted of 5125 individuals, which had on average lower incomes, lower education levels and younger ages compared to those in the CPS data. The univariate and multivariate regression tests both showed that home ownership is a significant negative indicator for unemployment and home owners earn more than renters. Assuming that the length of spells satisfies Weibull distribution, Coulson and Fisher ran the regression and found that home ownership exerts a negative influence on the length of the unemployment. The second hypothesis was rejected.

Coulson and Fisher rejected the null hypotheses that homeowners have lower unemployment probabilities, longer unemployment spells and lower wages than renters. The possible reasons why the hypotheses didn’t hold may be that mobility of renters tend to equalize unemployment rates across areas regardless of the homeowners’ behavior, or firms don’t observe status of homeownership and tend to pool renters and owners into a single labor market.

However, this paper was criticized by Munch, Rosholm, and Svarer. They pointed out that Coulson and Fisher didn’t address the potential endogeneity of the home owner variable, which made it unclear if the positive labor market outcomes for home owners are causal or spurious.