The federal Election Assistance Commission -- the agency created after the 2000 presidential debacle that is tasked with overseeing voting machine testing and serving as a clearinghouse for election administration information -- published a survey of the 2006 election today that reveals some interesting stats.
The information, collected from election administrators nationwide, covers the number of registered voters per jurisdiction, voter turnout, types of voting systems used, percentage of votes cast by absentee and provisional ballots, etc.
One interesting nugget concerns the number of ballots cast vs. ballots counted in the election.
According to the report, about 82 million ballots were "cast or counted" in the 2006 election (the number isn't exact because not every jurisdiction responded to the survey). But some 3.2 million ballots that were cast never got counted. [I should note here that it's really confusing that the EAC refers to the 82 million ballots as "cast or counted" since it isn't possible for a ballot to be counted if it wasn't cast -- at least not a legal ballot. It would have been better for the report to just say "82 million ballots were cast."]
The report provides a table showing the number of ballots that went uncounted by state (see the middle column in the table at right). For example, in Florida 122,759 ballots went uncounted in 2006. In California, the number was 416,260 ballots. Illinois held the record, however, with a whopping 889,012 uncounted ballots.
Some of these figures seem less severe when you look at them as percentages (in California, for example, the 400,000 uncounted ballots amount to a little less than 5 percent of the ballots that were cast). But this isn't likely to be of consolation to voters who made the effort to cast a ballot but never had their vote counted or to candidates who may have lost their races by narrow margins and could have used the extra push from uncounted ballots. And the number of uncounted ballots in Illinois isn't helped by converting it to a percentage (nearly 25 percent of ballots cast in that state went uncounted).
So why do ballots go uncounted? It's not always clear.
Take provisional ballots for example. Although a little more than 794,000 provisional ballots were cast in polling places, only 79.5 percent of them were counted. Provisional ballots are given to voters who arrive at polls but whose name (for whatever reason) doesn't appear on the voter registration list (perhaps through clerical error or the voter showed up at the wrong precinct). The ballots are usually rejected if it turns out that the voter was ineligible to vote or had already voted by absentee ballot or at another polling location, but the report doesn't really specify.
With regard to absentee ballots that were cast in 2006, 347,000 of these never got counted. In some cases voters didn't return the ballots on time. In other cases, voters failed to sign the ballot envelope. But more than 52,000 ballots were rejected for "other" unspecified reasons.
To see what else the report covers, you can read it here (PDF). It would be great if all voting jurisdictions were required to participate in these post-election surveys so that stats like these could be more precise and more easily compared over time.
UPDATE: A reader has pointed out some problems with the EAC's survey numbers, particularly with regard to the table I highlighted showing the number of ballots counted in states as a percentage of the total number of ballots cast in those states. As reader Bob Richards correctly points out in the comments to this post, the percentages are based in some cases on uneven figures due to the fact that the EAC was unable to collect complete sets of data from every jurisdiction in a state.
Take Illinois as an example. According to the EAC's table 28a, it appears that 25 percent of all ballots cast in that state didn't get counted. But if you look at tables 26 and 27 that break out the numbers of ballots cast and counted, you can see that not all jurisdictions responded to both questions in the survey. Illinois has 110 voting jurisdictions (the first column in each table lists the number of jurisdictions, or "Jur" in each state), but only 3 of those jurisdictions responded to the question asking how many absentee ballots were cast. Then on table 27, it shows that 5 jurisdictions responded to the question asking how many absentee ballots were actually counted.
With regard to provisional ballots, 77 Illinois jurisdictions responded to the question asking how many provisional ballots were cast, but on table 27 we see that only 62 jurisdictions responded to the question asking how many of those ballots were actually counted. Since we don't have complete or equal reporting on both questions, the percentages factored from the responses to those questions aren't trustworthy.
I pointed out in my original post that the EAC didn't have complete numbers from all jurisdictions but I didn't realize the extent to which this made the EAC's report so unreliable. On page 17 of the report, the authors discuss this problem with the data, but noted only that "States that reported figures below 95 percent (e.g., Arkansas, Illinois, Indiana, New Mexico, Pennsylvania, Texas, and West Virginia) have inconsistencies in the manner in which data were collected and reported."
But per my conversation with Kim Brace, one of the authors of the EAC report and the founder of Election Data Services, the problems with the figures are even more extensive than the EAC report suggests. Brace told me that there's a huge problem with trying to collect reliable data from election administrators in general because jurisdictions across the country collect and define data in different ways.
For example, some jurisdictions define the number of ballots cast at precincts as the number of signatures in a pollbook, not the number of ballots that were actually cast in the precinct. So if a voter signs the pollbook but gets frustrated with the long line and leaves before casting a ballot, some jurisdictions would still count this as a cast ballot.