Reading Shattered, an account of Hilary Clinton’s failed election campaign by Jonathan Allen and Amie Parnes, I’ve been struck by how limited political modelling has proved in recent elections. This had been in the case in the 2008 primary contest with Obama, in which the unprecedented character of his candidacy (as well as the candidate himself) repudiated the assumptions built into the campaign’s models. From pg 132:
By the time in 2008 that she realized Obama had a better strategy for racking up delegates by dominating her in low-turnout caucus states and among African American voters, it was way too late for her to reverse the cold mathematical reality of her defeat. In that year, African Americans had voted as a bloc in southern primaries, delivering massive delegate hauls to Obama.
We can see similar tendencies throughout the 2016 campaign. The primary challenge provided by Bernie Sanders confounded expectations, as can be in his mobilisation of first-time primary voters in Iowa. From pg 116:
Reading the data in the boiler room, members of the analytics team were surprised by the reports on new registrants. The overall number was a little more than they had expected. But they had also underestimated the margins for Bernie. The first-timers were breaking 90 percent to 10 percent in his favor. Running the data through their models, they could see why the race was so tight.
The subsequent developments are a good example of the practical implications of such changes. The campaign can overcome such failures but, through doing so, might fail to learn the lessons. From pg 116:
Hillary’s get-out-the-vote team on the ground, bolstered by a handful of talented veteran organizers, had been built with the expectation that Bernie wouldn’t do as well as he did. They overperformed, and their work had bailed out the analytics squad. That was good news in that Hillary had eluded defeat, but the outcome served to obscure flaws in Elan Kriegel’s modeling—namely, that it hadn’t correctly accounted for the number of new registrants or the degree to which they would break for Hillary—and Mook’s corresponding allocation of resources for in-person contact with caucus-goers. “The seeds of what we see across the campaign were present there,” said one person familiar with the campaign’s strategy and tactics. “It was a warning sign that they just barely scraped by, and I don’t think they took that seriously.”
If we rely on past and present data to predict future events, the weakness of the model we use will reside in its capacity to cope with genuine novelty. One response to this might be to account for such novelty as once-in-a-lifetime chance occurance. But one of the conclusions we might draw from the Centre for Social Ontology’s Social Morphogenesis project is that social novelty is being generated at an ever-increasing rate. In large part this is because novelty breeds more novelty: the unprecedented character of Obama’s candidacy generated novelty in ideological form, political constituency, electoral methodology and communications strategy. This novel campaign then provides the backdrop for Hilary’s failed campaign, transforming the inherited context to a much greater degree than any campaign did prior to Bill’s own.
This might seem like a unnecessarily abstract way of saying politics is becoming more unpredictable. But I think it’s important that we attempt to account for that unpredictability, its origins, character and consequences. The question which really fascinates me is who will be empowered if, as seems likely, these failures trend towards ubiquity. In light of this, it’s interesting to observe how closely Donald Trump’s instincts converged with Bill Clinton’s. From pg 128-129:
Neither a traditional poll nor Mook’s preferred analytics—voter-behavior models based on surveys and demographic data—were as finely tuned as his own sense of political winds, Bill thought. They were an important part of a modern campaign but not the only part. “You couldn’t place all of your eggs in the data/polling basket,” one of Bill’s confidants said of his thinking. “He had the ability to sort of figure out what’s going on around him, to sort of take everyone’s feedback and synthesize it and measure [it] along with his experience and then report back.” Bill had done this thing twice. His handle on politics was as natural as Jimi Hendrix’s feel for the guitar. Hillary couldn’t grasp the sentiment of the electorate, the resentfulness white working-and middle-class Americans felt watching the wealthy rebound quickly from the 2008 economic crisis while their families struggled through a slow recovery. Her team didn’t really get it, either.
And from pg 130-131:
Bill’s time on the ground only encouraged his skepticism of Mook’s reluctance to send him outside population centers. Having grown up in Arkansas, Bill understood that a major political player—a senator, a governor, or a former president—could bridge ideological divides by just showing up in small towns that never got much attention from elected leaders. He liked to go to small towns in northern New Hampshire, Appalachia, and rural Florida because he believed, from experience, that going to them and acknowledging he knew how they lived their lives, and the way they made decisions, put points on the board. Mook wanted Bill in places where the most Hillary-inclined voters would see him. That meant talking to white liberals and minorities in cities and their close-in suburbs. That was one fault line of a massive generational divide between Bill and Mook that separated old-time political hustling from modern data-driven vote collecting. Bill was like the old manager putting in a pinch hitter he believed would come through in the clutch while the eggheaded general manager in the owner’s box furiously dialed the dugout phone to let him know there was an 82 percent chance that the batter would make an out this time. It’s not that Bill resisted data—he loved poring over political numbers—but he thought of it as both necessary and insufficient for understanding electoral politics.