United States Elections Project
   
 

2008 Current Population Survey Voting and Registration Supplement

Last updated: May 17, 2010

Overview

The Census Bureau's Current Population Survey, Voting and Registration Supplement confirms that African-American and youth voter turnout increased between 2004 and 2008. The statistics from the survey indicate that the 2008 electorate became more representative of the American citizenry. Disparities in turnout rates among various demographic categories decreased between 2004 and 2008.

The CPS survey indicates that the national turnout rate and registration rate declined a slight 0.2 percentage points between 2004 and 2008. Using estimates of the voting-eligible population and the number of people who voted, I previously reported that the turnout rate increased by 1.6 percentage points between 2004 and 2008. I thus believe that the CPS decline is more apparent than real. The CPS is a survey and is prone to both statistical and non-statistical survey methodology errors, just like any other survey. When there are small changes in statistics from one election to the next, these survey errors are more likely to produce a discrepancy like the one here whereby the actual turnout rate increased slightly and the CPS turnout rate decreased slightly.

The Current Population Survey, Voter and Registration Supplement

Among the most reliable sources for a demographic profile of voters and registered voters is the Census Bureau's Current Population Survey Voting and Registration Supplement. In a November of a federal election year, the CPS asks a limited number of questions about voting and registration. These data may be accessed by the Census Bureau's DataFerret. Documentation for these data are available here. A Stata data file of some CPS variables is available here. These data were released in advance of the 2008 Voting and Registration Report.

It is always possible that I have made errors and I encourage interested persons to replicate the statistics I report. Since this is my own analysis, these numbers should not be considered an official census product. Some historical CPS statistics reported here were compiled as part of the Trends to Watch project funded by the Pew Center for the States. Many customizable election statistics at the state level, 1980-2006, are available through the Pew website.

The National 2008 Turnout Rate

To start, I define a voter as a citizen of voting-age (CVAP) who responded affirmatively that they voted in the 2008 election. Following convention from past census reports, respondents who reported that they did not vote, did not know, refused, or provided no response are defined as non-voters.

Response to CPS Vote Question
Yes 63.6%
No 22.6%
No Response 11.2%
Refused 1.3%
Don't Know 1.2%

The overall self-reported citizen-voting age turnout rate for the 2008 CPS is 63.6% with a margin of error of +/- 0.4 percentage points. This compares to my turnout rate estimate for those eligible to vote of 61.7%. The numerator for this number is the vote for president. My turnout rate estimate for the total ballots counted (excluding rejected absentee and provisional ballots because there is no way to estimate these numbers nationally at this time) is 62.3%, which is closer to the CPS turnout rate. Also note that the CPS sample frame is the non-institutional domestic population of the United States, while I try to account as best as possible all persons eligible and ineligible to vote, both domestic and abroad.

As I explain here, while the different sample frame of the CPS might explain some of the discrepancy between these statistics, the higher CPS turnout rate is most likely explained by what is known as "over-report bias" whereby a greater number of respondents report voting than aggregate statistics indicate. The 2008 CPS is no exception, as the graph presented below of the American National Election Study turnout rates (the premier academic election survey), the CPS, and my voting-eligible vote for president turnout rates demonstrates. Respondents' misreporting and survey non-response bias are two of the most likely explanations for over-report bias.

What is perhaps most interesting from this graph is that the 2008 CPS citizen voting-age population turnout rate of 63.6% is slightly lower than the 63.8% reported in 2004, while my voting-eligible turnout rates indicate that the turnout rate increased from 60.1% in 2004 to 61.7% in 2008. To put it in another way, the vote tallies from official state election websites show that 9.0 million more people voted for president in 2008 than in 2004. According to Census Bureau's population estimates, the voting-age population increased 10.1 million between 2004 and 2008 (neglecting any other changes of eligibility of the population, such as citizenship or felony status). It thus appears almost certain that the turnout rate did not decline between 2004 and 2008 despite these CPS statistics.

One other interesting piece of this puzzle is that the CPS estimate of the total number of voters is similar to the aggregate statistics from election officials. The CPS estimate is that 131,143,947 persons voted and the aggregate statistics indicate that 132,608,519 persons voted (this is an estimate of the total ballots cast, including those from overseas voters who are not in the CPS sample frame).

There are two plausible explanations for the turnout rate decline evident in the CPS. One is the familar "margin of error" statistical sampling error endemic to surveys. This sort of directional error would be more likely to happen when the turnout rate increased slightly, as it likely did between 2004 and 2008. The second may be that the CPS improved its survey methodology and has a lower over-report bias in 2008 compared to 2004. There are indications from the graph that the CPS over-report bias has been decreasing in recent presidential elections, so this is not an implausible source of the seemingly-surprising CPS turnout rate decline.

Irregardless, the Census Bureau's CPS voting and registration supplement is an important report that always receives major media coverage. This CPS turnout rate decline will therefore likely be uncritically reported by the media when the 2008 voting and registration report is released later in 2009. I do not believe that a turnout rate decline occurred between 2004 and 2008. While this may cast doubt on the accuracy of the 2008 CPS Voting and Registration Supplement, I would add that I believe that the data are still useful to illuminate patterns in voting and registration. Indeed, voter turnout statistics by race and age, discussed below, are consistent with patterns speculated to be present in the 2008 election.

overreport bias
This chart is adapted from Michael P. McDonald. 2003. "On the Over-Report Bias of the National Election Study." Political Analysis 11(2): 180-186 and Michael P. McDonald.  2005.  "Reporting Bias" in Polling in America: An Encyclopedia of Public Opinion.  Benjamin Radcliff and Samuel Best, eds.  Westport, CT: Greenwood Press.

The National 2008 Voter Registration Rate

The Census Bureau asks persons who did not vote if they are registered. Those who are registered to vote is thus a combination of all persons who report voting and those who did not vote who report being registered. Those who did not vote and did not report being registered and did not respond, refused to respond, or did not know are considered to be not registered to vote.

Response to CPS Voter Registration Question
Yes 71.0%
No 14.8%
No Response 11.4%
Refused 1.2%
Don't Know 1.6%

The 2008 citizen voting-age CPS registration rate of 71.0% is lower than the 2004 registration rate of 72.1%. To place this into the context of recent elections, the registration rate for presidential elections since 1980 is graphed below. This decline, like that of the 2008 voter turnout rate, is more likely apparent than real given that in 2008 overall voter registration increased 9.5 million or 5.4% over 2004, according to statistics compiled by the Associated Press. It may be that the same factors that produce the apparent lower CPS turnout rate in 2008 compared to 2004 is also confounding the CPS registration rate.

CPS Registration Rate

The Census Bureau further asks citizens of voting age that were not registered to vote why they were not registered. The most frequent response is "not interested" with 38.9% and in a similar vein, 3.9% report their vote will not make a difference. Consistent with a number of studies that find registration deadlines do lower turnout rates, 14.0% report that they missed the registration deadline in their state. If these persons who missed a deadline had voted, the national turnout rate would have been 4.1 percentage points higher. This is consistent with numerous studies, such as one conducted by myself published in Political Behavior ($) arguing that permitting persons to move their voter registration with them (aka portable registration) would increase American turnout rates. Portable registration as implemented in some states has robust fraud protections since it is easy to detect if a person votes more than once by checking their registration at their previous address.

Reason Not Registered to Vote
No Answer 0.1%
Refused 0.7%
Don't Know 4.0%
Missed Deadline 14.0%
Did Not Know Where or How to Register 3.6%
Did Not Meet Residency Requirement 3.5%
Permanent Illness of Disability 5.1%
Difficulty with English 1.4%
Not Interested 38.9%
My Vote Will Not Make a Difference 3.9%
Not Eligible 7.6%
Other 17.2%

Race/Ethnicity

Barack Obama's race was one of the major themes of the 2008 presidential election. The CPS 2008 CVAP turnout rates by race indicate that the Black, non-Hispanic turnout rate of 65.2% just barely fell shy of the White, non-Hispanic turnout rate of 66.1%. Still, Black turnout increased 4.9 percentage points from 2004, while White turnout decreased by 1.1 percentage points. This increasing Black turnout rate is consistent with the media's exit polls that showed that Blacks were an increasing share of the electorate, increasing from 11% in 2004 to 13% in 2008. The Hispanic CPS turnout rate also increased by 2.7 percentage points, but still lags far behind other racial groups.

These statistics appear consistent with greater excitement among African-Americans towards President Barack Obama. The increasing turnout of Hispanics might also be attributed to enthusiasm towards Obama. The apparent 2004 to 2008 CPS national turnout rate decline is clearly located entirely among White, non-Hispanics. The lower turnout among Whites might be reflective of lowered enthusiasm towards Senator John McCain, particularly among Republicans.

Pre-election, I forecasted that the national voting-eligible turnout rate may reach 64%. Such a turnout rate would have been the highest turnout rate since 1908, literally a century. This decline in White, non-Hispanic turnout, if real, may explain why American turnout rates fell short of my high end estimate. It is surprising to me that White turnout would decline in a high stimulus election such as 2008. Even if it did not, and we need to shift the 2008 CPS turnout rate upwards, these numbers suggest that White turnout did not keep pace with other groups, including Hispanics and the residual "other" racial category.

Methodology: The Census Bureau asks two questions to probe an individual's race (White, Black, etc.) and ethnicity (Hispanic/non-Hispanic). I present four categories that are a combination of these two questions: non-Hispanic White, non-Hispanic Black, Hispanic, and all other residual categories. I know this is not politically correct language, but these are the Census Bureau's designations and I want to be consistent with the Census Bureau's terminology to aid those who might wish to replicate these numbers. I also understand that I am collapsing the Asian-American and Native American populations into the same category, but the number of these respondents is often very low.

I include the margin of error (for a 95% confidence interval) for each cell in parentheses below the point estimates. For those unfamilar with margin of error calculations, the margin of error varies for the size of a sub-population and the reported percentage. Most survey organizations only report the margin of error for the entire sample, not for sub-populations.

Racial Category
2008 Turnout Rate (MoE)
2004 Turnout Rate (MoE)
Change
White, Non-Hispanic 66.1% 67.2% -1.1%
(0.4%) (0.4%)
Black, Non-Hispanic 65.2% 60.3% 4.9%
(1.1%) (1.1%)
Hispanic 49.9% 47.2% 2.7%
(1.3%) (1.4%)
Other 49.5% 48.0% 1.5%
(1.6%) (1.7%)

CPS Turnout Rates by Race

Age

Another major theme arising out of the 2008 election was the surge in youth voting. There is indeed evidence that youth turnout increased relative to older persons. The CPS turnout rate for citizens age 18-29 increased 2.1 percentage points between 2004 and 2008 while all other age categories experienced a decline. Despite this relative increase, youth turnout continues to lag behind all other age categories by sizable margins. The good news is that there is some evidence that youth turnout is catching up in the last two presidential elections.

Methodology: I sort respondents' age into four categories: citizens age 18-29, 30-44, 45-59 and 60+.

I include the margin of error (for a 95% confidence interval) for each cell in parentheses below the point estimates. For those unfamilar with margin of error calculations, the margin of error varies for the size of a sub-population and the reported percentage. Most survey organizations only report the margin of error for the entire sample, not for sub-populations.

Age
2008 Turnout Rate (MoE)
2004 Turnout Rate (MoE)
Change
18-29 51.1% 49.0% 2.1%
(0.8%) (0.8%)
30-44 61.8% 62.4% -0.6%
(0.7%) (0.7%)
45-59 68.5% 70.0% -1.5%
(0.6%) (0.6%)
60+ 70.8% 71.4% -0.7%
(0.6%) (0.7%)

CPS Turnout Rates by Age

Education

The numbers suggest that those with a high school diploma and those with the highest levels of education did not keep pace with the other two education categories. How is it possible that the simple average across the education categories does not correspond to the apparent 0.2 percentage point decline in the national CPS turnout rate from 2004 to 2008? The turnout rates by education are a good example of confounding effects. American citizens are also becoming better educated. Thus, people were moving up between education categories between 2004 and 2008 even as there were apparent sizable declines in turnout rates between the two elections among three of four education categories.

Methodology: I sort respondents' education levels into four categories: those who have not recevied a high school diploma, those who have received a high school diploma or a G.E.D., those who received some college education up to a bachelor's degree, and those who have had at least some post-graduate education.

I include the margin of error (for a 95% confidence interval) for each cell in parentheses below the point estimates. For those unfamilar with margin of error calculations, the margin of error varies for the size of a sub-population and the reported percentage. Most survey organizations only report the margin of error for the entire sample, not for sub-populations.

Education
2008 Turnout Rate (MoE)
2004 Turnout Rate (MoE)
Change
Less than H.S. Diploma 39.4% 39.5% -0.1%
(1.1%) (1.0%)
High School Diploma 54.9% 56.4% -1.5%
(0.6%) (0.6%)
Some College to Bachelor's Degree 71.5% 72.2% -0.7%
(0.5%) (0.5%)
Post-Graduate Education 82.7% 84.2% -1.5%
(0.9%) (0.9%)

Early Voting

The CPS reports that 29.7% of self-reported voters cast their ballot prior to Election Day. The Associated Press and I report that 30% of all votes were cast early, with a few states still not reporting. A plausible reason for the slightly smaller CPS number is that persons who returned their mail ballot on Election Day often cannot be disentangled from persons who cast their mail ballot earlier.

The remarkable increase in early voting since 1972 is apparent in the graph of CPS early voting rates in presidential elections, below. Note that for some surveys, mostly in the 1980s, the CPS Voting and Registration Supplement did not include a time of voting question.

CPS Early Voting