ABSTRACT OF TIIE THESIS



Housing Preferences; A Multivariate Analysis

by RAPHAEL J. CAPRIO

Thesis director: Professor George W. Carey



This dissertation examines the association of social attributes and housing preferences. A sample of 317 households which represent 25 percent of all new homeowners in four suburbs of Newark, New Jersey (Belleville, Bloomfield, Montclair, and Nutley) were interviewed to determine whether the evaluation of housing can be inferred from various social and economic characteristics of households. Basing the selection of variables on earlier work in social area analysis and factoral ecology, twenty-three measures of social structure are examined utilizing analysis of variance. The analysis indicates that the four municipal samples differ significantly from each other with respect to key social measures such as income, education, household size, etc. Dimensions of social structure are then identified by use of a principal components factor analysis. The results, in every case, are consistent with findings of earlier factorial ecologies. The degree of congruence between individual factor matrices is then analyzed using techniques of vector and matrix comparison. In spite of the socio-economic heterogeneity of the four municipal samples, the factor analysis yielded an eight factor model that is found to be quite consistent from one sample to another. This indicates considerable uniformity in patterns of social structure.

Each household was also asked to evaluate ten housing amenities. An evaluation instrument utilizing the paired comparisons method is administered, and aggregate evaluation scores for each housing amenity is obtained for every household. When compared between samples, only three amenities-race of neighbors, education and occupation of neighbors, and size of lot-are found to differ significantly. The ten evaluation measures are also factor analyzed to determine empirical evaluation dimensions. In contrast to the analysis of social dimensions, the evaluation factor structures prove to be quite heterogeneous between samples. 

Factor scores are then obtained for each household for each of eight social dimensions, and five evaluation dimensions. Multivariate regression techniques are used to determine if the locations of households in a multidimensional social space matches locations in a multivariate evaluation space. The multivariate linear regression model was tested at two scales. The first, or municipal scale, yields incongruency. Specifically, evaluation patterns cannot be predicted given household socio-economic attributes. The second, or regional scale, shows congruency. Due to the significant heterogeneity of the larger sample we discern certain associations between socio-economic attributes and housing evaluations. Lower economic status households, for example, express considerable interest in having neighbors of the same race. Higher economic status households, on the other hand, are more interested in buying homes in areas of larger lots, and with neighbors of a similar educational and occupational background.