This same problem crops up in reporting voter turnout. In the 2004 presidential election, Brace said that 902 counties failed to keep track of the number of voters who showed up at the polls. Instead, they counted the number of votes that were cast in the presidential race to determine their voter turnout. If a voter left the presidential race blank, he wasn't included in the voter turnout numbers.
Problems also crop up with the voter registration numbers in the EAC report. Some counties, in counting their voter registration numbers, only include active voters in that count. Other counties count active and inactive voters. (A voter is defined as inactive when he's registered to vote but has failed to turn out for two of the last federal elections and mail sent to him from election administrators is either unanswered or returned as undeliverable.) This might seem like a minor issue but Brace points out that in California there are 15 million active voters and 6 million inactive ones. It can really distort the registration numbers if counties don't have a single definition for what constitutes a registered voter and the figures are off by a couple million.
So why has the EAC bothered to put so much effort into a survey that is filled with incomplete and distorted data? Brace says it's the best that anyone has managed to do so far. The EAC is working on improving its survey and getting counties to respond (this is only the second election survey it's conducted). Brace says hopefully the 2008 survey will be better. But it appears that since the EAC can't force counties to respond to the survey or collect and track data in a uniform manner, there's little hope that will happen.
http://blog.wired.com/27bstroke6/2007/12/national-electi.html
(Note: You can view every article as one long page if you sign up as an Advocate Member, or higher).