Anyone who writes or comments about politics will make predictions that fail to come true. I certainly have.
Some forecasting errors are preventable. Using a combination of valid surveys and applying the lessons of past electoral outcomes will keep you from making egregious errors. But to paraphrase Sherlock Holmes, when you have eliminated the impossible from your election prognostications, whatever remains, however improbable, could happen.
None of the “experts,” for example, thought that former Republican National Committee Chairman Ed Gillespie had a shot of unseating Virginia’s Democratic Sen. Mark Warner last year. Warner, a former governor who had won his seat in 2008 by crushing former governor (and current GOP presidential candidate) Jim Gilmore with 65 percent of the vote, was supposed to be a shoo-in for reelection in 2014. Few analysts had the Warner-Gillespie race on their radar screens. Little outside money flowed into Virginia to help Gillespie.
This proved to be the Republicans’ biggest mistake in an otherwise triumphant year. Just as in North Carolina and most other battleground states, the undecided vote in Virginia broke against Warner. He only edged Gillespie by 49.1 percent to 48.3 percent. With a little extra help, the Republican may well have prevailed.
Here in North Carolina, the victory of Thom Tillis over Kay Hagan didn’t qualify as the improbable becoming fact. Analysts who set aside their personal biases and simply tracked the last several weeks of polling could see that the momentum was shifting against Hagan, and that a Tillis win was quite possible. (After the election, the Tillis campaign confirmed that its own polling tracked with the public surveys: he trailed most of the race but surpassed Hagan in the end when the undecideds broke his way.)
A clearer case of getting it wrong in our state was the widespread prediction that 2013 changes to North Carolina election laws would suppress voter turnout, particularly among black voters. It didn’t happen. The 2014 turnout was higher than in North Carolina’s previous midterm election, 2010. And black voters made up a higher share of the electorate in 2014 (21.4 percent) than in 2010 (20.1 percent).
Liberals who opposed the legislature’s decision to eliminate out-of-precinct voting, end same-day voter registration during early voting, and modify the early-voting calendar insist that higher turnout in 2014 didn’t invalidate their predictions. Instead, they insist that the higher competitiveness of last year’s U.S. Senate race compared with Richard Burr’s easy reelection in 2010 offset what would otherwise have been significant impediments to voter participation.
This is not a particularly convincing argument. It grants that whatever you might think of the election-law changes, they can hardly be described as “voter suppression” if the vast majority of voters can cast their ballots without incident. Moreover, the argument fails to grapple with the real flaw in the original prediction: it lacked an empirical basis.
Most political science research shows that early voting has a tenuous relationship to voter turnout. Making it more convenient to cast a ballot before Election Day can, in theory, induce more people to vote than otherwise would. But early voting also reduces the immediacy of Election Day, reducing voter attention and intensity. The net effect, if any, appears pretty small.
One recently published study may offer a clue as to why turnout in North Carolina went up rather than down in 2014. Elliott Fullmer, a political scientist at Randolph-Macon College, looked not at the number of days of early voting but instead at how many voting sites were available, a statistic he called “early voting density.” Fullmer concluded in his Election Law Journal study that it “has a significant and positive effect on voter turnout.” As it happens, the General Assembly reduced North Carolina’s early voting period from 17 days to 10 days but mandated that the same number of hours of early voting be available. Counties responded by creating more early voting sites and keeping them open longer each day.
Predicting political events is hard enough. It is impossible if you allow your assumptions to supplant the evidence.
John Hood is chairman of the John Locke Foundation. Follow him @JohnHoodNC.