An Analysis of Voting Method by Race and Ethnicity May 2006 i
Question: Does the rate of voter turnout differ by voting method and race/ethnicity of the voter? Data: Georgia residents are asked to report their race/ethnicity when they register to vote as one of the following: white, black, Hispanic, Asian, or other. Because we know the race/ethnicity for most registrants using the voter registration database, there is no need to estimate this characteristic. Likewise it is not necessary to derive estimates for the type of voting method. Prior to the 2004 election, the voter history database reports whether someone voted in-person on election-day or by absentee mail ballot. 1 [Prior to the 2004 general election the state legislature passed a provision whereby voters could go to specified polling sites prior to election-day and cast a ballot in-person. This has been termed in-person absentee voting. Unfortunately, the state did not mandate a code in the voter history database whereby in-person and mail absentee voters could be separately denoted. As such, after 2002 it is not possible to separate absentee voters by whether they voted in-person or by mail. Therefore, the 2002 general election offers the best possibility for calculating the difference between mail absentee versus in-person election-day turnout rates by race.] Using the 2002 voter registration and history databases, it is possible to calculate turnout rates by voting method and race/ethnicity. Analysis: Table 1 reports turnout rates by race/ethnicity and voting method for the 2002 general election in Georgia. The table reports that 5.29% of the white turnout in this election was comprised of mail absentee voting. The comparable figure for black voters is 2.80%, producing a 2.49% difference between the two groups. Rates and difference measures for other racial/ethnic groups are also reported in Table 1. 2 A logistic regression model was also employed in order to estimate the probability of voting method (mail absentee ballot=1; in-person election day=0) by race/ethnicity. Dummy variables for each racial/ethnic category were included in the equation with white voters used as the excluded comparison category. 3 Table 2 presents a series of probability estimates by voting method and race of voter. 4 As the number of cases in the analysis exceeds 1.5 million, it is not surprising that these estimates are very similar to those presented in Table 1. Again, the model predicts that 5.29% of white voters and 2.80% of black voters cast an absentee mail ballot in the 2002 election (identical estimates to those figures in Table 1). The differences in mail absentee voting rates between whites and other racial/ethnic groups were also calculated and presented in the last column of Table 2. The results indicate that these differences varied from a low of.47% for the other category to a high of 2.63% for Hispanics, with all differences statistically significant at the.05 confidence level. Finally, a 95% confidence interval for each estimated 1 While we have data that encompasses the entire population of voters in the 2002 general election, we do not know the voting method (in-person or mail absentee) used by 516,398 voters due to incomplete information in the voter history file. 2 The raw frequency data used to construct Table 1 is found in Table A in the Appendix. 3 Model results are located in Table B in the Appendix. 4 Predicted probabilities calculated using Clarify 2.1 (Tomz, Wittenberg, and King 2003). ii
probability was also included in Table 2. As an example, the estimated absentee mail turnout rate for white voters is 5.29%, with a confidence interval ranging from 5.25% to 5.33%. Interpretation: We can be 95% confident that the actual (true) measure of absentee mail turnout for whites is +/-.04 percentage points from the model estimate of 5.29% (a fairly narrow interval again due to the size of the sample employed for analysis). iii
Table 1. Turnout Rates by Race and Voting Method, 2002 General Election Race/Ethnicity % Absentee % In-Person Difference between White Absentee Rate White 5.29% 94.71% ---- Black 2.80% 97.20% 2.49% Asian 3.73% 96.27% 2.65% Hispanic 2.63% 97.37% 1.56% Other 4.80% 95.20%.48% Race Unknown 4.44% 95.56%.85% Minority 2.88% 97.12% 2.43% Notes: Minority combines all categories except White and Unknown. Percentages derived from frequency data presented in Table A (see Appendix). Table 2. Estimated Probability by Race and Voting Method, 2002 General Election Race/Ethnicity Probability of Voting Absentee Ninety-Five Percent Confidence Interval Difference between White Absentee Rate White 5.29% 5.25% 5.33% ---- Black 2.80% 2.74% 2.86% 2.49% ** Asian 3.76% 2.99% 4.65% 1.53% ** Hispanic 2.65% 2.03% 3.40% 2.63% ** Other 4.81% 4.36% 5.27%.47% ** Unknown 4.44% 3.93% 5.02%.84% ** Notes: ** Denotes difference in probability is significant at the.05 level. Estimates derived from Table B (see Appendix). iv
Appendix: Table A. Turnout by Race and Voting Method, 2002 General Election Race/Ethnicity Absentee In-Person White 62,729 1,124,011 Black 8,729 303,321 Asian 78 2,014 Hispanic 55 2,035 Other 433 8,581 Race Unknown 245 5,279 Total 72,269 1,445,241 Minority 9,540 321,230 Notes: Minority combines all categories except White and Unknown. Due to incomplete data in the voter history database, it was not possible to classify 516,398 voters according to their method of voting (absentee or in-person). v
Table B. Logistic Regression Model Predicting Absentee Turnout Rates by Race/Ethnicity, 2002 Coefficient Black -.6623 *** (.0116) Asian -.3653 *** (.1155) Hispanic -.7251 *** (.1367) Other -.1007 * (.0494) Unknown -.1844 *** (.0655) Constant -2.8858 *** (.0041) N 1,517,516 Notes: * p<.05; ** p<.01; *** p<.001 Entries are logistic regression coefficients with standard errors in parentheses. White voters are utilized as the comparison group. vi