Employment Density Does Not Decline as a Negative Exponential Function of Distance from the Center

Abstract

The negative exponential model has been used to describe the distribution of employment in urban areas without systematic evidence that employment density does decline as a negative exponential function of distance from the center. Census tract data for employment in 43 large urban areas in 2010 from the Longitudinal Employer-Household Dynamics program are used to explore this question. Traditional linear estimation of the log-transformed model is rejected, and nonlinear regression is used to estimate the model. The initial regressions produced parameter estimates that were unreasonably large for many areas. The problem was found to be high-density tracts near but not at the center. An ad hoc procedure was employed in which distance to the center was set to zero for tracts identified as being part of the central business district (CBD). This produced parameter estimates that were more reasonable, though the fit seemed better for the higher density tracts closer to the center than the larger number of more distant, lower density tracts. The model was then estimated using only those tracts outside of the CBDs. Parameter estimates were completely different from and inconsistent with the estimates made using all tracts. The conclusion is that employment density does not decline as a negative exponential function of distance from the center.