Abstract:

One of the most salient developments in the geography of US labor market is the increasing geographic concentration of college graduates. The literature has focused on demand driven hypotheses such as skill complementarity. Although a few attempts to study supply shift hypotheses have been made through examining wage premium and wage growth, the evidence is more elusive. We reckon wage premium itself does not depict the full picture of sorting because individuals’ ability to sort themselves to a location depends on their bid rent, which is determined by the location wage premium they expect and the amenity premium they perceive. Thus, we propose a structural framework to estimate bid rent gradient, which comprises a marginal willingness to pay (MWTP) gradient and a wage gradient, for each urban attribute by education and age. The framework is built on a random utility model with migration and is closely related to Roback. Comparing to the traditional literature of internal migration, our estimates of location choice have a structural interpretation. That is the willingness to pay. We borrow a novel idea of iterated estimation from the literature to estimate location choice, to obtain fixed point estimates, and to instrument endogenous variables. The quality of the framework in predicting sorting is shown by a regression on city net-migration rate against the change in total bid rent derived from the estimates of MWTP and wage gradients. The estimation result shows interesting variation of MWTP for city size and college graduate concentration. The MWTP decreases with age but increases with education attainment. The pattern lends support to the hypothesis of learning in cities. A comparison of MWTP and wage gradients indicates that the demand for urban amenities and for learning in cities can be major impact on education sorting.