Why don’t the Ohio early voting numbers line up? I don’t know!

I appreciate being the go-to person for early voting statistics and information, and I try to help reporters as best I can.  It can be hard, however, when a reporter challenges a piece of information that you gave them, which was drawn from official sources, with a number taken from a campaign.  I have no idea where campaigns get their figures.

The most recent set of inquiries come from Ohio. Someone in the Obama campaign believes that 28% of Ohioans voted in-person early or absentee.

The campaign may believe that, but the best information I have at my fingertips come from the Ohio Secretary of State’s website, the AP Elections Unit, and the EAC’s EAVS survey.  I show below why I discount the EAC information, so my best information is that approximately 30% of Ohioans cast an early ballot.

Ohio’s reported turnout November 4, 2008: 5,773,777

Ohio’s official Absentee ballots cast: 1,733,579

Absentee / Total = Percent Domestic Absentee = 30.025%

From the AP Elections Unit:

Ballot Cast: 5,773,777

Total “Advance” (Early + Absentee): 1,719,795

    Percent “Advance”: 29.8%

From the EAC’s Election Administration and Voting Survey:

(1) Total Voters Participating (Table 29A):   5,671,438
(2) Total Polling Place Voters (Table 29A): 3,771,194
(3) Total Early In Person (Table 29A): 386,390
(4) Total Domestic Absentee (Table 29B): 1,288,451

But you cannot just sum up (3) and (4) and divide by (1) because there are missing counties.  Even “at the polls” is based on 87 counties while there are 88 in the state.  Only 60 counties reported early voting totals, and only 84 reported absentee.

It is, to be frank, extremely frustrating that the EAC continues to require is contractors report badly misleading statistics in the report.  The percent voting domestic absentee is not 22.7% (as shown on page 25) because the divisors are not identical!

I know the agency knows this because we told them this in 2008 but they would not let us calculate the statistic correctly, based on an equal number of cases.  You’d think they’ve have learned their lesson, but check out the 2010 report, pg. 20, South Dakota “at the polls.”  The percent is based on 57 counties in the numerator and 63 in the denominator.

I’d advise anyone working with these data–or asking a scholar about them–to work with the raw figures.  The response rates have improved enough that the numbers in the report are roughly accurate, but there are still anomalies.