Aggregation Bias

Abstract

In order to assess the extent of the bias that exists in ecological correlations census data from the Public Use Sample and Census Summary. Tapes for the Chicago Standard Consolidated Area were used and a comparison was made of correlation and regression coefficients at the individual versus tract level. The assumption of larger aggregate level coefficients was not consistently upheld Although it was most frequently the case that aggiegate coefficients were larger than individual level coefficients, at both the tract and county levels aggregate coefficients were observed that were smaller or identical to individual level coefficients An accept able two-variable aggregate regression equation was estimated for the prediction of current fertility at the tract level The equation was acceptable in the sense that the aggregate regression coefficients were relatively close to the individual level coefficients