This paper uses the S-Night counts of persons in homeless shelters and living on the streets from the 1990 Decennial census to study the causes of variation in rates of homelessness across metropolitan areas. The effects of several policies are explored. Specifically, we estimate the importance of emergency homeless shelters and federally subsidized housing as solutions to the problem of homelessness. The results suggest that the expansion of homeless shelters has led to an increase in the total number of homeless persons. Additional shelters induce some households to leave the worst traditional housing situations. We find no evidence in support of the notion that increasing the number of federally subsidized units would decrease the number of homeless. However, our estimates suggest that a city can lower the rate of homelessness by targeting subsidized housing toward the very poor. Considerable debate exists over which policies are the most successful at reducing the number of homeless. This study uses data from the 1990 decennial census to estimate the effectiveness of, among other things, the expansion of homeless shelters and the availability and targeting of subsidized housing. The results suggest that the introduction of emergency shelters for the homeless has led to an increase in the number of homeless in the U.S. No evidence is found to support the notion that a general increase in the number of subsidized units will decrease the total number of homeless. However, estimates suggest that federally subsidized housing targeted toward the poor reduces the rate of homelessness.