Inferences from the DNC Provisional Ballot Voter Survey

The survey conducted in Cuyahoga County, Ohio,1 shows that the single most important cause of voters casting a provisional ballot in the county in the November 2004 election was residential mobility. About 60 percent of the provisional ballots were cast by those who either were voting in Ohio for the first time or who had previously voted in Ohio but had since moved. Among those who had previously voted in Ohio and not moved since doing so, voters younger than 55 years of age were much more likely to cast a provisional ballot than older voters were. Among those who had previously voted in Ohio but since moved, African American voters were more likely than white voters were to cast a provisional ballot Before considering the detailed results from the survey, note that the matched, case control design of the Cuyahoga survey means that when the appropriate sampling weights are used, the survey exactly estimates the overall frequency of casting a provisional ballot. From administrative records we know that the proportion of provisional ballots among all ballots cast is 0.03518, a value the survey estimate is constrained to reproduce exactly.2 Estimates for the proportion of provisional ballots cast by groups of voters in Cuyahoga County will not be exact but will be subject to sampling error. A technical appendix to this memo describes the methods used to compute estimates from the survey data. Despite the enforced accuracy of the overall proportion, the sample was implemented using a zipcode-level matching design that may introduce bias in the estimates of other quantities when making inferences about all voters in Cuyahoga County. A nonprovisional voter has a positive probability of being included in the second sample only if the voter lives in the same zipcode as a provisional voter. If there are zipcodes in which there are nonprovisional voters but no provisionals, then the nonprovisional voters in those zipcodes have zero probability of being included in the nonprovisional survey.3 In this case estimates using the nonprovisional survey data are biased. Given the way the estimators for the survey proportions are derived, it may be reasonable to say the survey estimates are biased only if there are zipcodes in Cuyahoga County where it is impossible that any provisional ballots were cast.4 Because the personal experiences and adminstrative problems that cause a provisional ballot may affect almost any voter (e.g., moving residences, misdirected absentee ballots, record keeping errors), it may be reasonable to rule out this source of bias. The estimates discussed in this report ignore this potential bias. Tables 1, 2 and 3 show how often Cuyahoga voters cast provisional ballots given various personal attributes and election-day experiences they had. The first column of each table lists