electionlineWeekly--May 5, 2011

I. In Focus This Week

Data Dispatches
Exploring the Census’ Voting & Registration Supplement

By Charles Stewart and Andreas Westgaard

Consistent, reliable, and publicly available election data can be hard to come by. That’s why it’s important to highlight data sources that break the mold and inform the way we conduct elections.

A noteworthy data source is the Census Bureau’s Current Population Survey (CPS). Both the availability of the data and its longevity make it an invaluable resource for journalists, academics, and the public who wish to study a wide variety of economic and public policy issues.

The CPS provides rich demographic data on issues ranging from unemployment to school enrollment statistics. However, one feature of the CPS that is especially relevant to studying election administration and policy is its Voting and Registration Supplement (VRS).

The VRS is now a multi-decade study about the participation of citizens in elections, conducted immediately following each biennial federal election. Since 1964, Americans have been asked questions like how they register, what modes they use to vote (by mail or in person), and why non-voters fail to participate. In short, the VRS has a wealth of information that helps us understand some basic facts about voting and the problems potential voters encounter when they go to the polls or register.

The Census Bureau recently released the VRS data from the 2010 midterm election. While it appears there will not be a written report, the data are available now for those who wish to catch a glimpse at what the electorate looked like in 2010, and at the barriers faced by some who were unable to vote.

In one interesting highlight, the latest data shows voters continuing their long-term trend toward favoring non-precinct place voting alternatives. In 2010, 18.2 percent of respondents reported they voted absentee or by mail, 8.4 percent voted early in-person, and 73.4 percent voted on Election Day. This compares with the last midterm election in 2006, when 13.8 percent reported voting by mail, 5.8 percent early, and 80.4 percent on Election Day.

Because of the large sample size of the VRS (56,000 households, which yielded 44,802 individuals who stated they voted in 2010) and the nationwide scope of the study, it is possible to break down the data at the state level and ascertain where the fastest changes are occurring in where voters vote.

When we do that, we see that growth in the use of the mail to return ballots, compared to 2006, exceeded 10 percentage points in six states, Colorado, Montana, North Dakota, Washington, Arizona, and Ohio; in-person early voting grew by more than 10 percentage points in North Carolina, Nevada, New Mexico, and Georgia.

In addition to midterm elections, we can also assess claims made across consecutive election cycles. While recent political developments (Barack Obama’s election in 2008 and the rise of the Tea Party movement in 2010) have focused attention on the racial composition of the electorate, the VRS gives us the best access to data to track these trends over time.

Using estimates derived from the VRS, we see that the racial composition of the electorate in 2010 was similar to that of 2008 — 77.5 percent non-Hispanic white in 2010, compared to 76.3 percent in 2008; 11.4 percent African American compared to 12.1 percent; and 6.9 percent Hispanic compared to 7.4 percent.

There are cautions to the use of the VRS data, however; most notably, it is based on a survey, and respondents to surveys tend to over-report voting. Interestingly enough, however, recent research by the Census Bureau shows that the VRS over-reports turnout in some states and under-reports it in others. This means that on the whole, the VRS turnout estimates appear to be less prone to over-reporting bias than other large national surveys that explore voting. In addition, there is a margin-of-error with all the estimates produced by the VRS. The regular written reports on the VRS data have discussions of standard errors in the use of the estimates.

Finally, the raw data are available through the Census Bureau’s DataFerrett service. For those comfortable downloading government datasets, DataFerrett is capable of extracting the dozens of variables associated with the VRS study and exporting it in a variety of formats that are compatible with spreadsheet and statistical software. It is also possible to use DataFerrett to produce simple tables of frequencies and relationships.

The implications of data like the VRS and interactive government tools like DataFerret are clear: whether you’re an academic, election official, or simply an engaged citizen, the data for assessing elections can be right at your fingertips.