Some Observations on Proportions of Uncounted Votes in the 2000 Presidential Election

In the wake of the 2000 presidential election, Americans were made aware of possible inequities in the voting system in the United States. While the election results in Florida were the immediate focus, all over the country, methods of tabulating votes were scrutinized. If some polling methods such as punch card ballots result in a higher proportion of uncounted votes than other methods such as optical scan, then voters in some polling areas may have a higher probability of their vote ``counting.'' This paper presents some interesting observations on the proportions of uncounted votes, by county, in the state of Georgia. The voting method is found to be a significant predictor of the proportion of uncounted votes, with the punch-card method producing higher proportions. Further analyses using publicly available data are presented, to see if other county-level variables can explain differences in proportions of uncounted votes. An economic variable is also found to be a strong predictor. Several examples of aggregation bias in these data are discussed, including an example of Simpson's paradox.

The paper -- pdf format

Georgia voting data

Florida voting data

More Florida voting data

---

mmeyer@stat.uga.edu