01101994 Regular Meeting - Laserfiche WebLink

Home Browse Search

;;.6/V Demographic and Statistical Methodology / Page 2 for candidate Jones, we would first compute, for each voting precinct, the percent of the precinct Place 7 vote going to Jones and the percent of the precinct voting age population in each racial/ethnic group (Hispanic, Black, and Anglo-plus-Other). Applying the ecological regression procedure to these sets of percents provides a way to statistically determine the overall (citywide) vote distribution for candidate Jones by racial/ethnic group. While we cannot observe this outcome for anyone voting precinct, since we have only votes by candidate and voting age population by race/ethnicity for each precinct, we can average this information across voting precincts (weighting each precinct by its total voting age population) to determine what percent of the overall Hispanic, Black and Anglo-plus-Other vote Jones {and any of his opponents} received. Ecological regression is, in essence, an averaging scheme that allows us to determine demographic support by candidate for a group of voting precincts, combined, when for each precinct, we have only total votes by candidate and total persons of voting age by race/ethnicity. Despite the smaller number of voting precincts available for analysis in the City, we were able to obtain statistically reliable estimates of Hispanic and Anglo support for candidates in virtually all the elections we examined. In the case of Black support, however, we were able to obtain statistically reliable estimates in only about 60% of the elections we examined. The problem results from the combination of fewer precincts available for analysis and the relatively small Black population of the City. Under these circumstances, where there is not that much information for analysis, the ability of ecological regression to discern statistically reliable voting patterns is diminished. We have inserted the label "NSR" (Not Statistically Reliable) in the ac- companying Appendix Table for those {predominantly Black} vote shares which did not meet statistical standards and could not be converted to a percent basis. Our election analysis results are reported for three racial/ethnic groups: 0 Hispanic 0 Non-Hispanic Black (referred to simply as Black in this report) 0 Non-Hispanic White (Anglo), non-Hispanic Native American, non-Hispanic Asian American and non-Hispanic other races -- all combined into a single group in this report and labeled Anglo-plus-Other {see explanation below}. The third category was necessary for the application of the ecological regression framework to San Marcos, because the three non-Hispanic populations combined with non-Hispanic White were all too small to be analyzed separately. Since this "composite" (Anglo-plus-Other) category is dominated statistically by non-Hispanic White population, voting patterns attributed to this group by ecological regression are effectively, and for all practical purposes, those of Anglo residents. We have noted this qualification on the accompanying Appendix Table.