An integrated system for estimating sub-county household income distributions.

An integrated system implemented in the San Diego Region (San Diego County) for preparing census tract level household income estimates was discussed. The income estimates were for households as defined in the 1990 Census. The estimates were done first for he region then for 41 groups of census tracts known as subregional areas (SRAs) then for the regions 347 census tracts. The modified lognormal curve was employed with a different equation using the parameters of income class PARM (reflecting constant terms) and EXP (a nonlinear adjustment parameter). To estimate the household income distribution the parameters of median income PARM and EXP were needed. Using 1990 census median incomes and income distributions as inputs an iterative method was used for 9250 combinations of PARM and EXP to find the parameters that minimize the error between the observed and estimated income distributions. For the investigation of historical trends in the parameters 1980 census data from eight income ranges were also assembled. Estimating the parameters in the 1990 postcensal period required the values for median income (for the region estimated annually based on several data sources) PARM (the change was based on the annualized difference between the PARM values calibrated from the 1990 and 1980 census income distributions) and EXP (based on its nonlinear relationship to PARM). Inputs to the model included the base year data the estimated parameters control totals from an independent estimate of the number of households for the region SRAs and census tracts and adjustment factors calculated from a comparison of the estimated and actual 1990 household income distributions. The integrated household income estimation system was fully implemented by means of finding solutions to the issues of forming households from individual tax returns residual and adjustment factors the applicability of the modified lognormal curve and extrapolating parameters. The internal and external consistency of the estimates was also evaluated.