Subsidized Housing, Emergency Shelters, and Homelessness: An Empirical Investigation Using Data from the 1990 Census

Dirk W. Early and Edgar O. Olsen

July 2001

This paper uses data on the only systematic count of the homeless throughout the United States to estimate the effect on the rate of homelessness of a wide variety of potentially important determinants, including several major policy responses to homelessness that have not been included in previous studies. It improves upon estimates of the effect of previously studied determinants by using measures that correspond more closely to underlying theoretical constructs, especially by accounting for geographical price differences. It also conducts numerous sensitivity analyses and analyzes the consequences of the undercount of the homeless for point estimates and hypothesis tests. The paper's most important finding from a policy perspective is that targeting the current budget authority for housing assistance on the poorest eligible households will essentially eliminate homelessness among those who apply for assistance. Achieving this goal without concentrating the poorest households in housing projects and without spending more money requires vouchering out project-based assistance. The primary methodological finding of the paper is that the 1990 Decennial Census did not produce sufficiently accurate counts, especially of the street homeless, to permit very precise estimates of the effects of many factors which surely affect the rate of homelessness. The main exceptions are the price of housing and average March temperature. Plausible models of the undercount imply that in regressions with a rate of homelessness as the dependent variable estimators of the coefficients of explanatory variables will be biased towards zero. In regressions with the logarithm of a rate of homelessness as the dependent variable, only the estimator of the constant term will be biased downwards. The unknown magnitude of the undercount precludes predicting the effects of policy interventions on the number of homeless based on the results in this paper and previous studies.

Go to Study

// Crazy Egg // MouseFlow // Google Analytics // HEAP DEMO