Retire to where? A discrete choice model of residential location

Many studies of migration attempt to identify attributes of places that influence residential location decisions. Most such studies relate an area's migration rates to its own attributes, or the flow of migrants between two areas to the attributes of both places. We focus on retirement migration within the US, and extend the literature by using a discrete choice framework in which a person's location decision depends on the attributes of all potential locations. Using a multinomial logit specification and 1990 Census county-to-county migration data, we model the location choice of individuals aged 65 to 74 years from among more than 3000 counties. In order to make the estimation problem manageable, we employ the sampling-of-alternatives approach of McFadden, and use aggregated migratory-flow data for which existing estimation software is applicable. Our results indicate that, in general, persons around retirement age avoid high taxes and housing costs, while they are drawn to areas with relatively high spending on such services as fire, police and recreation. Amenities such as coastline and warm weather are also valued.