Voters in South Carolina’s 1st district head to the polls today for a special election. What follows is my perspective on the election with links to worthwhile readings on the race. The Tale of the Tape
PPP polled the 1st district on March 24th, April 22nd, and May 5th. In head-to-head match-ups with Sanford, Elizabeth Colbert Busch was +2, +9, and -1 (respectively). Similarly, the Cook Political Report changed the race yesterday from “lean Democratic” to “Republican toss up.” It’s altogether fitting that one of the more intriguing races in recent political history will come down to the wire...
But despite the PPP polls, I’m skeptical of the prevailing narrative that a Sanford victory constitutes a “major comeback” from where he stood on April. This race has always been Sanford’s to lose in my view. How, then, can we explain the April 22nd PPP poll? For starters, given the timing of the poll—corresponding with Sanford’s trespassing charge—conservative respondents may have been reluctant to reveal their true vote intentions. This form of social desirability bias would have suppressed Sanford’s numbers in the April poll. Moreover, it's possible that Sanford voters were less likely to participate in the poll for the same reason. Indeed, people associated with the campaigns have consistently maintained that their internal polls show a tight race.
Along these lines, don’t be surprised to see Sanford outperform his recent polling. I’ve speculated that there’s a kind of "Bradley effect" in the numbers. Though it’s difficult to quantify, my sense is that some voters—in particular social conservatives—are wary of admitting to strangers over the telephone (i.e. pollsters) that they’d vote for someone with Sanford’s checkered past. We could classify this as yet another form of social desirability bias artificially suppressing Sanford's numbers. I’ll offer one data point supporting this contention: Sanford outperformed the March 24th PPP poll by 3 points (53% to 56%).
The most common question about the special election is: Can Mark Sanford win despite his sex scandal? It’s a topic Gibbs Knotts and I addressed in a recent editorial (but see also Danny Hayes here). Aided in part by published research on the electoral penalty of scandal, we think the answer is most certainly “yes.” A 2012 paper by Scott Basinger (University of Houston) finds the following with respect to political scandals: (1) 60 percent of House incumbents from 1973 to 2010 survived a scandal and were successfully elected, (2) sex scandals cost incumbents about 5.3 percent of the vote, (3) financial scandals, by comparison cost lawmakers about 7.8 percent of the vote. Basinger also finds, not surprisingly, that the most significant predictor of an incumbent’s vote share in a general election is their prior election vote share. In fact, there’s almost a 1:1 relationship between the two.
What does this mean for Sanford? Well, in his 2006 gubernatorial race, Sanford received over 60 percent of the vote in Berkeley, Dorchester, and Beaufort counties, and 57 percent in Charleston County. So if the past is any predictor, with a 5 percent penalty for his sex scandal, Sanford has a great shot of winning.
Addition by Subtraction is Bad Math
If Sanford wins today, one of the resulting narratives will be that this actually benefits national Democrats. A few people have made this point already (see for example this piece in the Washington Post by Chris Cillizza), but the argument misses the mark in my view. First, I’m skeptical of any argument whereby “losing is really wining.” Yes, Sanford would be something of an embarrassment to national Republicans. And yes, political commentators will continually bring up his checkered past. But I seriously doubt this outcome either (a) helps national Democrats fundraise against Republicans or (b) hurts Republican congressional candidates in other states. A Sanford victory does one thing above all else: put Democrats one seat further from regaining the House. Besides, if Democrats were truly "better off" losing the seat, why is the DCC spending almost a half million dollars on the race? Further, consider for the sake of argument the narrative resulting from a Sanford defeat. Democrats would almost certainly argue that Republican primary voters once again nominated someone too conservative for general election voters (e.g. another Sharron Angle or Christine O’Donnell). The spin cuts both ways in this case, so having an extra House seat trumps either narrative.
Voter Identification Laws
One of the interesting subplots of this race is the fact that South Carolina's new voter id law is in effect. If the polls are right and Elizabeth Colbert Busch loses by a narrow margin, we may hear claims that the voter id requirement was the culprit. Per the Voting Rights Act of 1965, the Justice Department will be monitoring the election (along with a handful of others). But do voter id laws supress turnout? The evidence is mixed and the effects are statistically difficult to discern, but answer seems to be "yes." John Sides reviews the evidence here. For example, a 2007 study by Alvarez, Bailey, and Katz finds that strict voter id laws depress turnout to a greater extent than "weak" identification requirements. Moreover, the effect is is greatest for citizens with less education and low income (but importantly the authors find no racial differences). However, a 2009 paper by Ansolabehere finds that while there is racial bias in requests for proof of identification, only 2/10ths of 1% of registered voters cite lack of identification as a barrier to casting a ballot.
Elizabeth Colbert Busch: Long Odds and Longer Odds
South Carolina’s 1st district has been represented by a Republican for thirty years. The defining feature of the district in this race is that Mitt Romney won here by 18 points just a few months ago. There are two implications worth highlighting. First, a victory by Elizabeth Colbert Busch would represent one of the biggest upsets in recent political history. Consider for example that only two Democrats in the 112th Congress represented districts where Romney did better. But second, as Nate Silver and others have pointed out, even if Colbert Busch wins today, she faces long odds holding onto the seat. Only 13 districts out of 59 special elections since 1997 (22%) have switched parties. And of those 13 seats, only 1—Arizona’s Ron Barber—is still in Congress today.
Predictions Based on Inadequate Evidence?
The most recent PPP poll has Sanford receiving 47% of the vote, eking out a 1 point victory over Colbert Busch. Given the threat of social desirability bias in these estimates, I think Sanford's headed for a larger victory. PPP also has a known Democratic “house effect.” So my guess--emphasis on guess--is that Sanford receives an even 50% of today’s vote. Though still a massive under-performance in a district Romney carried by 18 points, it nonetheless returns him to the House of Representatives (and likely for a while; the average tenure in the House is 10 terms). There are obvious caveats to this (insert hand-waiving here). For example, while predicting presidential elections is fairly straightforward (see here), predicting special-elections with low turnout is somewhat difficult.