This paper examines the determinants of the decision of low-income renters to move out of their current dwelling. Maximum score estimation is shown t be superior to ordinary discrete choice estimation techniques (probit, logit) for this problem, ad for similar discrete choices that require revering a previously optimal decision. The estimation reveals psychological costs of moving for typical low income renters of at least 8 percent of their income; these costs are even higher for older, longer tenure, or minority households. Policies that displace low income renters will have large social costs. In addition, the estimation results are used to calculate implicit household willingness to pay (WTP) for neighborhood amenities. This WTP based on mobility behavior is much greater than WTP estimates derived using hedonic methods, and is argued to be more accurate. This paper uses a semiparametric empirical technique to estimate the determinants of the decision of low-income renters to move out of their dwelling. These estimates show that low-income residents highly value remaining in their dwelling. In addition, these estimates are used to illustrate an alternative method to measure willingness to pay for neighborhood amenities. Moving decisions are usually examined with standard discrete choice models such as probit or logit (e.g., Venti and Wise (1984), or Weinberg, Friedman, and Mayo (1981). But the moving decision presents econometric difficulties for standard discrete choice models. As will be explained in section 1, because the household decision about moving is conditional on having previously preferred the original location, the disturbance term in mobility models is unlikely to follow the simple distributional forms required for probit or logit estimation. Maximum score estimation is an alternative estimation technique for discrete choice models that is robust to unusual distributions of the disturbance term. Although the theoretical properties of maximum score estimation have been well-explored (see Manski (1975,1985)), our paper presents one of the first empirical applications of maximum score techniques. Maximum score estimation of our residential mobility model yields similar parameter estimates to probit estimation, but much smaller standard errors. This advantage of greater precision may prove attractive to other researchers. We use our estimates of the residential mobility model to calculate the value to households of remaining at their current dwelling rather than being forced to move out. We use the household's mobility response to rent changes to infer a monetary value of remaining in the current dwelling. Our calculations indicate that the typical low-income renter household is willing to pay at least 8 percent of its annual income to avoid being forced out of its current dwelling. These "psychological moving costs" increase greatly for older or longer tenure households. Large "psychological moving costs" have important implications for public policy towards low-income neighborhoods. Neighborhood improvement policies or private market forces may displace low-income renters. If the losses suffered by low-income renters due to being forced out of their current dwelling unit are significant, as indicated in this paper, then it is important to include these losses in any evaluation of the net benefits of a neighborhood improvement program. In addition, policy makers might want to consider policies to prevent or compensate for privately-induced displacement. Estimates of the monetary value of low-income renters' psychological moving costs are important to determining the effects of these policies, and deciding appropriate compensation. Finally, this paper uses the residential mobility estimates to infer the willingness t pay (WTP) of low-income renters for neighborhood amenities such as the physical condition of the neighborhood, neighborhood school quality, and the safety of the neighborhood from crime. The relative responsiveness of household mobility to changes in these neighborhood amenities, versus changes in rents, implicitly reveals households' monetary valuations of these amenities. The more common approach to measuring household WTP for neighborhood amenities is the hedonic price approach, which relies n the equilibrium relationship between housing prices and amenities. The calculations in this paper suggest that mobility-based WTP estimates for amenities may often be greater than hedonic based estimates of WTP. We consider which approach is more accurate. Section 1 of the paper presents our econometrics, specification, and data. Section 2 presents the results. Section 3 is the conclusion.
