NVRA report highlights data problems, changes in how voters register

True election geeks like us can’t wait to get their hands on new sources of data, and the Election Assistance Commission’s (EAC) release earlier this summer of “The Impact of the National Voter Registration Act of 1993 on the Administration of Elections for Federal Office, 2003-2004” is no exception. Students of elections know that these reports to Congress are important sources of data on how states have implemented “Motor Voter” provisions.

There are some interesting things to note about this most recent report.

One is that this is the first NVRA implementation report issued by the EAC; in the past, the reports have been issued by the Federal Election Commission (FEC). Those who have worked in this area know that try as it might, the FEC was never adequately resourced to produce detailed reports; the quality of the FEC reports going back to the 1997-1998 reporting cycle (they are now archived on the EAC website) just provides current proof of that problem. The good news is that this most recent report is a well-produced document, available in html and pdf formats, well-written and relatively easy to read — a step in the right direction (when will they make the underlying data available with the reports in a useful format, though?).

Second, the report presents data that gives as mixed portrait of the current status of voter registration in the United States. To quote from the report’s summary:

Overall, voter registration increased in the 2004 general election, compared to the 2000 presidential election. Responses from 48 states showed total registration at more than 174.8 million voters in 2004, compared to the 162.4 million reported in 2000 from all 50 states and the District of Columbia … While the actual number of registered voters increased in the past four years, the rate of growth did not keep up with the growth rate of the voting age population. As a result, the percent of the voting age population that is registered to vote decreased from 78.9 percent in 2000 to 78.5 percent in the 48 states that reported data to the EAC.

Third, the report presents detailed state-by-state and national data on the origins of voter registration applications. Interestingly, in the 2003-2004 survey, 32.8 percent of applications came from motor vehicle offices, 32.4 percent by mail, and 25.4 percent in person. Unfortunately, the 2003-2004 survey was the first to ask states to report the number of in person applicants, making comparisons to previous report a bit tricky. But it is worth note that in the 2001-2002 reporting cycle, states claimed 42.77 percent of applicants came from motor vehicle offices and 27.64 percent by mail. Thus, the percentage of new registrants coming through the mail appears to have increased considerably, while the percentage coming from motor vehicle offices seems to have fallen dramatically. The available data from the 1997-1998 report indicates that these trends might be a recent phenomenon, as then registration by mail “accounted for nearly one quarter of all voter registration applications from 1997 through 1998”, while registrations from motor vehicle agencies represented 42.9 percent of applications then.

Fourth, there are some methodological issues that this most recent report documents well. One is that the EAC survey was unable to achieve complete coverage; two states and one territory did not provide data in 2004; several states and territories provide partial responses to the survey; and some states provided data that did not cover all local jurisdictions. That states either don’t or can’t report relatively important facts about how elections are being administered makes it difficult for policymakers and researchers, and also makes it hard to see how we can implement the type of end-to-end election auditing that many (like the the Caltech/MIT VTP) are advocating.

But here is the other methodological issue, lifted directly from page 5 of the report:

The 2004 Election Day Survey found that states report voter registration totals in different ways. Twenty-six states provided reports that included active voters only; reports from 20 states include both active and inactive voters. In four states, determination of whether to include active and inactive voters in reports of registered voters was at the discretion of local election jurisdictions.

This has obvious implications for those who use this data — comparability across states might be more problematic than we had realized before.

This last methodological issue also presents a dilemma that will need to be overcome as states start thinking of how to make their voter registration systems interoperable across state lines. We clearly need to make sure that states have identical definitions for registration status; to borrow a term that folks who do statistics will find familiar, states will need some sort of “data dictionary” before trying to compare files across state lines — and such “data dictionaries” might need to be part of data transaction standards for voter registration systems.