Abstract
The dissertation examines the spatial patterns of employment and worker residences with three urban density functions: monocentric, polycentric, and omnicentric. Analysis of the 1980 journey-to-work census data for the Los Angeles region reveals that the polycentric density functions statistically predict the actual distributions better than the monocentric density functions. It further shows that the omnicentric density function best predicts the distribution of worker residences. These findings suggest that polycentricity of spatial structure exists in large urban areas, and implies that accessibility to the general employment opportunities is the primary determinant in the residential location choices. The research also investigates urban commuting behavior by estimating the minimum average commute required by these three models. The results show that different urban forms require different amounts of minimum commuting. The standard monocentric model requires a small amount of commute–about one-tenth of the actual commute. The polycentric model predicts the actual commute much better than the monocentric model. Its required commute is about two-fifths of the actual commute, indicating that polycentricity has a positive effect on the estimate of required commute. The omnicentric model best explains the actual commuting patterns among the three models. Its required commute accounts for about 45 percent of the actual commute. These empirical results lead to the conclusion that an urban model better predicting the actual spatial patterns also better explains the actual commuting behavior. This conclusion helps to preserve the assumption that urban workers make attempts to economize on commuting in their location choices. This assumption is implicit in all the three models and implies a positive relationship between the fit of an urban model in explaining the actual distributions and the ability to predict the actual commute. The finding, the standard monocentric model is very poor at explaining the observed urban commute in a major metropolitan area, is more an indictment of the monocentricity assumption than a rejection of the assumption on the commuting behavior. The standard monocentric model greatly underpredicts the actual urban commute because it inadequately represents the actual spatial structure in large metropolitan areas. Relaxing the monocentricity assumption yields better prediction of actual commuting behavior.