01101994 Regular Meeting - Laserfiche WebLink

Home Browse Search

;)/:,M DEMOGRAPHIC AND STATISTICAL METHODOLOGY The analysis we have prepared involves two basic sets of information: 0 Demographic results from the 1980 and 1990 decennial censuses 0 Official returns from City elections between 1980 and 1993. - The demographic data represent counts of population by race and Hispanic origin for all City residents and residents of voting age. These data were tabulated first by census block, the smallest geographic unit for which census data are available, and then carefully aggregated to form voting precincts which prevailed in the City during each of the elections examined. In a few situations, we had to approximate precinct boundaries with census block boundaries, because voting precincts do not always conform precisely to census geography. Since census blocks are quite a bit smaller than precincts, the extent to which we could aggregate blocks to construct precincts proved quite satisfactory for the City, well within the tolerance required for our analysis. The statistical method we employed to examine City voting patterns requires data on voting age population by precinct and racial/ethnic group. These data were not a problem for 1990 because they were produced as part of the redistricting data files distributed by the Census Bureau. The same data were not routinely produced for 1980, however, and would have been both difficult and costly to obtain. For 1980, we decided to investigate an alternative approach that has worked well in a number of redistricting applications. This is the substitution of total population for voting age population. If the statistical correlation between these two demographic measures is high, then one can be used in place of the other with no effect on the quality of analytical results. For San Marcos in the 1990 census, the correlation between voting age population and total population across precincts was quite high (0.998), indicating that the two measures are almost perfect substitutes. 0 n the extremely plausible assumption that the correlation for 1980 was also on this order of magnitude, we decided to use total population in those segments of our analysis requiring 1980 census results. We were pleased with this alternative approach. , City election returns between 1980 and 1993 were drawn from official canvasses provided by the office of the City Attorney. We used only those returns reported by individual precinct for each candidate whose racial/ethnic support we analyzed. Prior to 1990, this meant not using absentee returns, which were not assigned to individual precincts during this period. Thus, the analysis before 1990 is based on election-day returns; that since 1990 is based on total returns. The statistical technique that permits us to estimate the racial/ethnic composition of election returns is called ecological regression. This method is commonly used in the analysis of voting patterns, where one seeks to determine the support given a particular candidate by different demographic groups. The ecological regression model employs two sets of data -- voting age population by racial/ethnic category, tabulated by voting precinct, and the number of votes by candidate for the contest in question, also tabulated by voting precinct. We can use a hypothetical example to illustrate the application of the method. Suppose we are interested in the case of John Jones, a candidate for Council Place 7. To determine demographic support