Comments

Abstract

Joseph Tracy: This paper adds to our understanding of the economic impact of immigrants on native residents in U.S. metropolitan areas. There is a growing empirical literature in labor economics that has examined the immigrant impact on wages of native workers. However, to understand the full economic effect of immigration on domestic households it is important to examine the impact in the housing market. And measuring immigrant impacts on wages and rents should provide a more complete test of the prevailing theoretical arguments in the literature. As discussed in Borjas, Freeman, and Katz, the wage impacts of immigrants on domestic workers tend to be small when measured at the locality and only begin to emerge as significant as the researcher expands the geographic scope of the analysis. An explanation for this finding is that immigrants into a city tend to displace native workers of similar skills.31 That is, an influx of immigrants into a city does not necessarily lead to an equivalent outward shift in the local labor supply since native workers affected may choose to migrate to another locality. The displacement hypothesis would also imply little impact on rents by an influx of immigrants into a city. For the same reason that an influx of immigrants into a city need not lead to an outward shift in the local labor supply, it may not lead to an outward shift in local housing demand. A finding of no significant immigrant impact on rents, when combined with the lack of any significant local wage impacts documented in the labor literature, would provide additional support to the displacement [End Page 188] hypothesis. However, a finding of a significant positive immigrant impact on rents would present a challenge to the displacement hypothesis. In a recent paper, Saiz finds that on average a 1 percent inflow of immigrants into a city is associated with a 1 percent increase in rents.32 Greulich, Quigley, and Raphael (hereafter GQR), in contrast, find no significant impact of the share of immigrants in a city on the median rent-to-income ratio (table 5). Both studies look at census data and arrive at different answers.33 How do we reconcile the two sets of empirical findings? Although it is beyond the scope of this comment to answer this question, I would like to raise a few issues that may point to a reconciliation of the results. Note first that the immigrant control measures differ between Saiz and GQR. Saiz uses the change in the number of immigrants living in a city over a period of time relative to the city population at the outset of that time interval (Iit - Iit-1)/Pit, where Iit is the number of immigrants in city i in year t, and Pitis the population of city i in year t. In contrast, GQR use the immigrant population share in the city or the change in the immigrant population share: Iit/Pit or (Iit/Pit) - (Iit-1/Pit-1). Which immigrant control variable is more appropriate? The answer may depend on the underlying theoretical model that the researcher is trying to test. Saiz is interested in testing whether a presumed outward shift in the demand for housing resulting from an inflow of immigrants affects rents paid by domestic households. He explicitly has a demand and supply of housing framework in mind and presents a formal model along these lines. Seeing the net change in the stock of immigrants into a city as a fraction of the city population would seem to be an appropriate control variable to measure shifts in the immigrant demand for housing. GQR do not discuss any theoretical framework in their paper. I believe, though, they also have in mind a model of the demand and supply of housing. I would argue, however, that the GQR immigrant control variable—the immigrant share—may be more appropriate for a quality-of-life (QOL) model.34 Suppose...