The unprecedented, if not entirely shocking, victory by Vance McAllister in the special election for the Fifth Congressional District demonstrates just how wacky elections of this nature can turn out, but also points out how such elections results can be produced.
An awful lot of somewhat unlikely things had to happen for McAllister to claim victory over state Sen. Neil Riser in this contest. Trailing Riser by 14 percent but only getting 18 percent of the vote in the general election, the numbers for McAllister to pull this off were daunting. Given the previous results and demographics involved, he simply couldn’t make an appeal to the “anybody but Riser” faction in the electorate, built upon the quaint notion that Riser was an “insider” and he an “outsider.” At a 17 percent turnout level, he would have to win about 70 percent of the defeated candidates’ votes. At 15 percent, it would be 80 percent. Therefore, he needed to boost turnout past these historical norms.
Also, he needed not just to get defeated candidates votes, he had to grab new ones. This would dilute the advantage Riser had by the creation of these new, so to speak, votes for him. Finally, Riser had to do the opposite; that is, his campaign could not expand the electorate much nor attract new voters in order to allow McAllister to eat into his natural advantage coming out of the primary.
The anecdotal evidence suggested that McAllister made an intentional play to expand his numbers by going after voters that were less attached to politics and by moving to the left on certain issues. In terms of quantifiable data, this would mean, of registered voters, he targeted those who would identify as other party (practically speaking, mostly “independents”) and Democrats but, more precisely, non-white Democrats who were more likely to be liberal and want to feel they could distinguish between two Republicans at the polls.
This can be tested by using voting and turnout data from both elections and registration data for the runoff. By performing a multiple regression on the aggregate data from the 23 parishes, it can be determined whether this strategy was effective. (More precisely for those interested, this meant a regression of McAllister’s proportion of the runoff vote on the proportion of other party registrants, proportion of white Democrats, and change in turnout percentage, with 23 cases using aggregate-level data for each, using because of the smaller sample size a confidence interval of 10 percent to determine statistical significance.)
Running this analysis, it turned out that the higher the proportion of other party members and the lower the proportion of whites in a parish, the higher the proportion of the vote McAllister received. Also significantly related with these was turnout differential, where the higher it was runoff to general election, the greater the proportion of the vote McAllister received. (For those interested in the gory details, all of the entered variables were significant at p<.047, the equation as a whole was significant at p=.006 and the adjusted R-squared was .370: this was the lowest p-value and highest R-squared value of any equation among the several tested under alternative hypotheses using different variables of race, partisanship, and turnout.)
In other words, these results seem to confirm that’s what happened. That is, McAllister disproportionately attracted non-whites and independents but, further, turnout increased where he had his best results in an election where turnout was down only to about 19 percent from 21.5 percent previously. Using bivariate correlations, changes in turnout were unrelated to any particular racial or demographic group, so the overall decline when the turnout statistics come out in a few days should show a fairly uniform change across all groups, yet in the parishes where he did the best, he had the greatest expansions of turnout, or attracting new voters much better than did Riser (Pearson r=.354, p=.09).
Whatever his campaign did, against heavy odds, it worked. Although the raw numbers also suggest that Riser assisted by doing little to help himself. His vote total increased only about 4,000 from general election to runoff, going up by only about 8 percent in gain of total electorate. This also is suggested by the fact that the greatest raw number vote gains made by McAllister came in the parishes he did the best. Therefore, it’s likely a combination of well-executed strategy that succeeded in disproportionately picking off the defeated candidates’ voters most of whom appeared to stay in the electorate and in expanding the electorate precisely where the odds of picking up favorable voters were best, and of an opponent’s strategy that did not adequately counter that to protect a numerical lead of a nature that seldom loses elections.
But it did happen, and in a way that produced not just a McAllister win, but one by around 20 points. Again, perhaps only in a special election runoff were the dynamics in place to produce this yet each side played it in a way where those dynamics enabled that result. As for moving beyond the mechanics of a result with no precedent in the history of Louisiana federal elections and that nobody outside the McAllister camp thought was possible to get to the meaning of it, that is the subject of a future post.