Posted on May 20, 2013

Skepticism About the Census Voter Turnout Finding

Paul Taylor and Mark Hugo Lopez, Pew Research, May 15, 2013

The Census Bureau made big news last week when it reported that the black voter turnout rate (66.2%) exceeded the white voter turnout rate (64.1%) for the first time ever in 2012. But a closer look at the numbers raises some intriguing questions.

It’s possible that the lines may have first crossed in 2008. But it’s also possible they may not have crossed at all.

Let’s start with the second scenario. It’s based on data that suggest that last year, blacks may have been more inclined than whites to report that they voted when in fact they didn’t. This is known as a “social desirability bias,” a familiar concern among survey researchers.

The Census Bureau bases its estimates of voter turnout on self-reports from a survey of a nationally representative sample of about 55,000 households. The survey is conducted in the two weeks after each federal election and is considered the best source of information on the demographics of the nation’s electorate. However, this self-report method typically produces a modest over-estimate of turnout, and 2012 was no exception. According to the Census Bureau’s estimates, 133 million Americans voted last year, but according to the official state-by-state tallies, just 129 million did.

Moreover, if you analyze the discrepancies by state, as the Pew Research Center has (download our Election Turnout Rates by State data in Excel), you find a pattern that casts some doubt on the Census Bureau’s announcement. It turns out that the phenomenon of over reporting tended to be most pronounced in states that have the highest share of blacks in their citizen-age electorate.

{snip} In Mississippi, the Census Bureau estimated a statewide turnout rate of 74.5%, while the actual state tallies showed a turnout rate of 60.4% — a gap of 14.1 percentage points. In Washington, the Census Bureau estimated a jurisdiction wide turnout rate of 75.9%, compared with an actual turnout rate of 63.7% — a gap of 12.2 percentage points. {snip}

Mississippi and Washington also happen to be the two jurisdictions in the country with the highest share of blacks in their voting age citizen eligible population — 35% and 49%, respectively. {snip}

Might this be because non-voting blacks were more eager than non-voting whites to tell survey takers that they voted for the first ever African-American president? While there’s no way of knowing for sure, the data are suggestive. When we plotted the state discrepancies in 2008 and 2004, we found a similar pattern, but we also found the racial skew was stronger in 2008 and 2012, the two elections in which Obama was on the ballot, than in 2004.

To better understand these patterns, we computed a “correlation coefficient,” which measures the relationship between two phenomena of interest — in this case, the over reporting of turnout in a state (the difference between the estimated and official voter turnout rates) and the share of a state’s adult population that is black. Our analysis finds a positive correlation of .52 (on a scale of -1 to 1) in 2012, .54 in 2008 and .41 in 2004. If we remove the two biggest outliers in the scatter plot analysis — Mississippi and Washington — the overall correlation remains positive, but it is only about half as large.


But wait, what about the first scenario — the possibility  that this milestone actually occurred in 2008, not 2012?

That assessment is based an analysis that removes from the pool of eligible voters all adults who have been disenfranchised as a result of felony convictions, something the Census Bureau does not (and cannot) do. According to the Sentencing Project, an advocacy group, nearly 6 million adults are ineligible to vote for that reason, a disproportionate share of whom are black. If you recalculate turnout rates after removing those disenfranchised voters, then 68.5% of eligible blacks voted in 2008, compared with 67% of eligible whites, according to Bernard L. Fraga, a political scientist studying at Harvard. The Census Bureau, by contrast, had 66.1% of whites voting that year, compared with 64.7% of blacks.