The Inevitable Election?
Or Why Louisiana ain’t Kentucky
Since the November 21st runoff election for Governor, gallons of ink have been spilled attempting to explain the election’s “surprising” result (one good example here). Louisianians elected John Bel Edwards, a previously little-known State House Democrat, over heavily-favored Republican Senator David Vitter. A political earthquake, right?
But what if I told you that this wasn’t really much of a surprise at all?
For those of you who poured over the polling numbers from as far back as Labor Day, the election’s result shouldn’t have been a shock at all. In fact, it was really a fait accompli.
Why? From September 23rd to November 21st, John Bel Edwards never trailed David Vitter in any publicly-released polling.
Perhaps you were swept up in the drama of the inconceivable. During the early days of the campaign, the pundits told us David Vitter probably couldn’t lose. In fact, some said he was focusing his campaign on leveraging his “inevitability.”
But a funny thing happened on the way to the ballot box. Vitter did lose and he lost big. Yes, this would have been surprising result to many even through the summer of 2015. But since the campaign began in earnest around Labor Day, the outcome that perhaps should have been the labeled “inevitable” was in fact Vitter’s defeat.
Why? That old killjoy: polling data.
It is now clear that Louisiana voters soured on Vitter early in the fall. When Public Policy Polling (PPP) released their first statewide survey on September 23rd, the result stunned many: John Bel Edwards was up 50 to 38 over David Vitter in a hypothetical runoff match-up. The campaign was still in its early days, although salvos had been fired. Republican competitors Jay Dardenne and Scott Angelle were swinging away at the Senator and Vitter ferociously hit back through his PAC and campaign. If there were cracks in Vitter’s armor at that point, there were few public indicators of those fissures. Vitter’s fundraising remained epic in scale and the political class rarely published any lede paragraph on the Governor’s race without the words “front-runner Senator David Vitter.”
To me, one particular aspect of this first poll is important . Despite the fact that voters did not know much about John Bel Edwards at the time (38% of respondents said they “did not know enough about JBE” to render an opinion), they knew enough to know they preferred him over David Vitter as Governor. Surely, all caveats: one poll does not a trend make.
Yet, this pattern would continue and intensify in all subsequent polling.
While the rancor of a Louisiana election played in the background, the data quietly sent a monochromatic signal: John Bel Edwards would win. Campaigns touted endorsements, candidates (sometimes) attended debates; massive amounts were spent on slash-and-burn television ads; all the while, something strangely boring was happening, or not happening as it were.
Watching the rolling poll average above (compiled by continuously averaging every three surveys) you’ll see that over the span of the campaign, the trend-lines barely budged. Poll after poll, from different pollsters using different techniques, produced a rather clear trend that remained unbroken. Even partisan pollsters (for example, Republican activists “Red Racing Horses” and “The Hayride” that might have been incentivized to concoct data to reinforce a counter-narrative) failed to dispel the “Edwards maintains a secure lead” narrative.
Edwards’ double-digit lead held from the outset, wandering up or down but never truly exhibiting a dramatic variance. Despite this trend-line, without knowing the outcome in advance, how could we know that we should trust this data? After the “polling fail” in Kentucky, in which the candidate leading with a small, single digit lead lost the election by a large margin, would our Governor’s race follow suit? Three main clues told a different story in Louisiana: Majority, Consistency, and Repetition.
Majority: The Magic 50%
It is dangerous to apply any hard-and-fast rules to the interpretation of political polls: too many factors can influence the predictive nature of survey data. There is one clue, however anecdotal, that helps indicate the true nature of a political environment: majority. If a candidate, in a binary match-up, earns a majority of the predicted vote share, one can project a couple of insights. First, if a majority of a statistically-significant sample of voters believes something, a candidate may have confidence that somewhere, somehow, she may be able to win an election. Surely, the election is not over — events can impact this majority’s preferences, they may not turnout on election day, etc. But still, achieving a majority is a significant achievement in an electorate. Second, enough voters have made a decision to prefer one candidate over another. This, too, is profound. Voter preferences, especially if the survey is taken far in advance of an election, can often be (at best) agnostic.
So what did the “majority” clue tell us in the runoff for Governor? Of the 14 separate polls I tracked from the primary election to the runoff, John Bel Edwards only earned less than 50% in four. In the 18 total head-to-head polls I tracked since September, Edwards was below 50% against Vitter five times. In terms of “majority,” the data was screaming: Edwards is significantly ahead.
Consistency: Never trailed In Any Poll
Now that we’ve listened to the clues sent by “majority,” we can delve into its corollary: “consistency.”
What does this mean? In a binary match-up, who is ahead and who is behind? When watching the polling trend-lines of a race, are you seeing a bouncing ball, or a straight line? This is the difference between seeing “conflicting” data and “consistent” data.
When you have polls that conflict upon their top-line outcome, the details of the poll (methodology, partisanship, and date) become much more important. Looking at the details may produce insights. It also might lead to more confusion.
In the gubernatorial runoff, we saw the opposite. Polls conducted from September on indicated a startling amount of consistency: All showed John Bel Edwards ahead. None showed David Vitter ahead. Once again, the polls were screaming a message. Another clue to an Edwards win.
Repetition: Ask Enough People And You’ll Get Your Answer
Finally, did “repetition” give us any clues? Like “majority” is to “consistency,” “repetition” is a corroborating clue. Once you see a consistent majority, the begged question becomes “is this repeatable.”
I tracked eighteen polls that accumulated over 12,720 survey responses over a six-week time-span. John Bel Edwards led in every single poll, an overwhelming number of his individual polling leads (72%+) were with a majority (50%+) of those surveyed. This amount of consistency, over this many surveys, begets “repetition.” And this clue, too, made it clear that the next Governor would be John Bel Edwards.
Following Clues to A Clear Truth: Predicting An Edwards Victory
To many political observers, this clear truth of the race may come as a surprise.
If you read the headlines, it might have seemed as if the race took many dramatic twists and turns. There was the early revival of the Vitter’s prostitution scandals; the two-barreled assault by Vitter and his PAC on his Republican rivals; Vitter’s debate dodges; a bizarre WDSU debate that dwelt on national social issues; a new interview with a former Vitter courtesan claiming coercion of abortion; Vitter’s use of a private investigator revealed in an peculiar coffee house confrontation; Prostitutes over Patriots; Syrian Refugees; and so, so much more. Surely, as colorful of a campaign and worthy to be regarded among any of Louisiana’s historically technicolor contests.
Yet, throughout this turbulence, the polling averages basically held firm. Edwards’ double-digit leads never really closed and Vitter never, ever led. And through repeated experiments, this truth never really changed.
Still, early in November, many wanted to cast this race as a twin of Kentucky’s Governor’s race. After-all, the Democrat leading narrowly in the polls there lost big in the end. Yet, searching for similar clues, we are left mostly empty-handed. The polls showed almost no majority, little consistency, and lacked repetition. The Democrat, Conway, led in many surveys, but always by very low single-digits (often within margins of error), never with a majority, and far fewer polls were conducted (only six polls were conducted between August and November’s election day). Without many clues to the contrary (along with the fact that Kentucky’s was a three-way race with left-leaning independent Drew Curtis), the comparison should have fallen flat on its face.
On the other hand, the Louisiana result bares witness to the magnetic pull of fundamentals — although not along typically partisan bedrock upon which many analyses were built.
The powerful truth is that Senator David Vitter never really had a chance against the profoundly genuine and consistent campaign of John Bel Edwards. This analysis does not imply that campaigns and events did not matter: quite the opposite, in fact.
The data reveals that the Edwards team’s decision to enforce a disciplined campaign message, which highlighted a strong character contract and promoted Edwards as a change candidate opposed to the highly unpopular outgoing Governor Jindal, allowed the powerful fundamental forces at play to work their “inevitable” magic.
As voters revealed their preference for “anybody but Vitter or Jindal,” Edwards employed his powerful biography and consistent message to define his own candidacy and narrative before Vitter’s team was able to build a case against them. By framing the election on his terms, as opposed to Vitter’s, Edwards paved the way for his ascent — and Vitter’s fall.
The clues revealed by the polling reviewed above told this story in clear terms. Through this particular lens the contest for Louisiana’s race surely was “inevitable” all the way back in September — and perhaps earlier. The contours of this inevitability story are the same — only the roles were reversed. Those watching the clues above knew the end of the story even before intermission.