ELECTIONATE JULY 16, 2012
The possibility that new voter-ID laws could disenfranchise thousands of Democratic- voters in pivotal swing states has received considerable attention recently. After all, 9.2 percent of registered voters in Pennsylvania lack photo identification, including 18 percent of registered voters in heavily Democratic Philadelphia. But these flashy numbers might be misleading. If voter-ID laws have consequences for voter turnout, they’re difficult to detect.
Several studies conducted in the wake of the 2006 midterms showed a weak correlation between tougher voter-ID laws and reduced turnout. Nate Silver has usefully summarized the data and explained the difficulty of interpreting it well, but I thought I’d flag a chart from a study by Robert Erikson and Lorraine Minnite that has stuck with me:
As you can see, the correlation depends on only a few data points, which undermines the practical significance of the findings if states are influenced by other factors. For instance, states implementing voter-ID laws were relatively Republican; that's significant, since 2006 was a decidedly Democratic year. The entire correlation turns on South Dakota (imagine the best fit line if turnout had increased by 6 percent in South Dakota), but the overwhelmingly white state had a hotly contested Senate race in 2002 between Tim Johnson and John Thune, drawing multiple visits from President Bush. The race was decided by just 524 votes. In 2006, the closest race was the gubernatorial contest—the governor won by 16 points. If the contests were flipped between 2002 and 2006, would turnout have decreased by six percent in lily-white South Dakota due to voter ID laws? I doubt it, but it can’t be proven. The bottom line: If voter-ID laws do meaningful reduce turnout, the sample is not yet good enough to withstand scrutiny.
Perhaps more importantly, studies conducted based on the 2006 midterms miss the two most relevant data points. Between 2004 and 2008, Indiana and Georgia became the first two states to require voters to provide photo identification—precisely the type of law at issue in Pennsylvania and elsewhere. 2004 and 2008 were presidential election years, which would tend to offer better intrastate comparisons than midterm election years, when state turnout is primarily shaped by local elections. That’s especially true in Georgia, where the Obama campaign made a modest effort to compete in a state quite similar to North Carolina and Virginia, two other solid-Bush states that Obama contested due to a large and untapped African American voting bloc. The other southern states with large African American populations represent points for comparison in the other direction.
So did laws requiring photo identification suppress African American turnout in Georgia in 2008? If it did, any consequences were overwhelmed by enthusiasm for Obama's historic candidacy. According to the exit polls, African American turnout increased more in Georgia than any other competitive state, including North Carolina and Virginia. According to the CNN exit polls, African Americans represented 30 percent of the Georgia electorate, up from 25 percent in 2004. In absolute terms, more than 350,000 additional African Americans voted—a 42 percent increase. The exit polls also suggest that the increase in African American turnout was more modest in Virginia and North Carolina, although it is worth noting that Georgia’s black population grew more over the last decade, giving Obama a larger pool of potential new voters. The exit poll numbers are imperfect, but it is difficult to contest that Georgia saw African American voter turnout rise to an extent greater than or equal to similar states.
On the other hand, overall turnout grew at a slower pace in Georgia than Virginia or North Carolina after adjusting for population growth. The increase in turnout was more comparable to Alabama, Mississippi, and South Carolina than the two Mid-Atlantic battlegrounds. Which is a better analogue to semi-competitive Georgia—the uncompetitive Deep South states, or the extremely competitive Mid-Atlantic states? Unfortunately, there isn't a definitive answer. The Obama campaign made a concerted effort to register voters in Georgia, unlike the uncompetitive states in the Deep South. But the Obama campaign largely suspended advertising by September, and I suspect that Georgia voters didn't think that their votes might decide the contest.
Someone could squint at the numbers and argue that Georgia turnout looks a little less than it ought to be, and that same person might also observe that a 2-to-3 percent drop-off among registered voters predicted by most studies would cover the gap. That might seem simple enough were it not for the surge in African American turnout. If African American turnout in Georgia increased to the extent suggested by the exit polls, then the slower increase in overall turnout is almost entirely attributable to white voters. That’s just not the effect predicted by opponents of voter-ID, and it suggests that other forces were at play, as was the case in lily-white South Dakota.
Even if photo ID did reduce turnout among Democratic-leaning groups, Obama performed as expected in Georgia and Indiana: The final polls were dead-on. If voters were turned away at the polls, it wasn’t enough to sway the results. Potential voters might have dissuaded due to their lack of identification, but since the polls were accurate, they must not have been included in polls of likely voters. For that same reason, the pre-election polls should remain accurate in 2012.
If voter identification requirements suppressed Democratic turnout in Georgia or Indiana, the effects were not readily observable in the final results. African American turnout surged in Georgia, the polls were dead-on, and Obama improved more over Kerry in Indiana than any other state of the continental forty-eight. None of this demonstrates that voter ID requirements do not have a disparate impact on minority voters, nor can it preclude the possibility that Obama would have done better without photo identification requirements. But it does suggest that voter ID laws are unlikely to sway the results of a national election. And if they do, we might not know for sure.
Follow me on twitter @electionate or at the Electionate blog.