| Why are the earnings of black households so low compared to those of white households? This question is addressed here within a parsimonious equilibrium model of residential location and intergenerational transmission of human capital. This model features two residential locations, which can be interpreted as neighborhoods. In order to obtain a seamless connection of model and data, Chapter 1 applies off-the-shelf clustering methods to summarize the complexity of US neighborhood-level data into a two-neighborhood characterization. Census tracts from large US Metropolitan Statistical Areas are clustered based jointly on racial configuration (R), average human capital ( H) and price of housing services (P). The chapter establishes that metropolitan US can be suitably characterized as a single city with a small number of representative neighborhoods. Two representative neighborhoods summarize around one third of the joint variation in (H, R, P) data while three representative neighborhoods summarize around half of the joint variation. Furthermore, census tracts clustered into each abstract representative neighborhood tend to be located close to each other, forming large areas of relatively homogeneous (H, R, P ) within an MSA. The two-neighborhood characterization is employed in Chapter 2 as an empirical counterpart of the theoretical model. In Chapter 2 the model is internally calibrated so that model generated data matches neighborhood segregation by color and earnings, neighborhood population sizes, intergenerational mobility (in earnings and consumption) and the ratio of human capital investment to output in the US. The Chapter's main finding is that, under this calibration, the proposed model produces nearly three quarters of the observed black-white percent difference in household earnings. Permanent racial inequality arises from residential racial segregation coupled with neighborhood human capital externalities. It is established that households' preferences over the racial composition of their neighborhood are at the base of the result. An interesting additional result is that, in contrast with a common line of thought, strong neighborhood color preferences are required to match US data, and halving the importance of racial preferences is enough to eliminate racial segregation and racial inequality from the benchmark equilibrium. |
