'The Gamble' and the Rise of Data-Driven Journalism
For decades, the journalistic campaign narrative has been the standard means for the average political junkie to understand presidential elections. These books have a grand tradition, beginning with Theodore White’s The Making of the President series, covering the 1960 through 1972 elections, moving through Richard Ben Cramer’s What It Takes, perhaps the best political book ever written (if not one of the best nonfiction books ever written), and culminating in the Holy Trinity of the 2008 elections: Richard Wolffe’s semi-hagiographical Renegade: The Making of a President, Dan Balz and Haynes Johnson’s The Battle for America 2008: The Story of an Extraordinary Election, and of course, the granddaddy of them all, Mark Halperin and John Heilemann’s blockbuster, Game Change: Obama and the Clintons, McCain and Palin, and the Election of a Lifetime.
In the past few years, a new sort of campaign narrative has sprouted up: what we might call the political science campaign narrative. This rising form mirrors the broader ascendency of data-driven journalism, where the standard narrative is supplemented by statistical interpretations of the increasingly voluminous amounts of data available to the average reader. In truth, this form actually predates the journalistic campaign narrative, arguably beginning with Samuel Lubell. A journalist, Lubell figured out in the 1930s that he could use door-to-door interviews with voters in key areas of the country to predict elections. His books are more qualitative than quantitative, but then so too are many of the older classics of political science, such as Richard Fenno’s Home Style: House Members in Their Districts.
Lubell notwithstanding, for many years these political science campaign narratives took the form of collections of papers on the elections, often published in series such as Gerald Pomper’s The Election of 1984 (or 1988/1992/etc), or Paul Abramson, John Aldrich and David Rohde’s Change and Continuity series. These books were wonderful in explaining the nuts-and-bolts of elections, but their form made them disjointed, and they were often written in a voice that went over the heads of many readers.
But in 2000, the year that RealClearPolitics was founded, Richard Johnson, Michael Hagen and Kathleen Hall Jamieson drew upon voluminous polling data from the Annenberg Center to create an integrated, data-driven narrative of that election. Eight years later, Kate Kenski, Bruce Hardy and Jamieson revived the genre with The Obama Victory: How Media, Money and Message Shaped the 2008 Election. It remains, to my reading, the most useful narrative of that election.
Now, George Washington University’s John Sides (founder of the political science blog “The Monkey Cage”) and UCLA’s Lynn Vavreck are out with probably the most successful attempt to integrate political science and narrative to date: The Gamble: Choice and Chance in the 2012 Presidential Election. Their hypothesis is simple (and widely shared by data-driven election analysts): Journalists spend far too much time covering every “game changer” in election cycles and far too little time examining the underlying stability of these elections.
Sides and Vavreck are literally anti-“game change.” They lead off by noting that journalists identified 68 “game-changing” incidents during the 2012 election, and that the term had been used by journalists more than 20,000 times during the campaign. They proceed, over the course of 240 pages, to demolish many of these supposed game changers.
In Sides and Vavreck’s telling, the most famous “turning points” -- President Obama’s “You didn’t build that” comment (around which Republicans inexplicably organized their convention), Obama’s first debate performance, and Mitt Romney’s “47 percent” comment -- did move the polls briefly, but things quickly regressed to a mean of Obama holding a narrow, steady lead over his opponent. Elections -- both primary and general -- are, in this view, governed by fundamentals: the state of the economy, presidential approval, and incumbency. The rest comes out in the wash.
Perhaps a more accurate take on Sides and Vavreck -- this nuance is sometimes lost in the book -- is that they think game changers do matter, but such events are so frequent and random that they cancel out one another. If one candidate committed no gaffes, raised all the money, and ran all of the effective ads, he or she would win in a 70-point blowout. But both parties nominate quality candidates who mostly avoid gaffes, raise money well past the point of diminishing returns, and generate their share of compelling ads.
Think of it as a professional football game. Both sides make spectacular catches, create momentum sapping drives, and celebrate devastating tackles. But actual game changers -- points on the board that never go away -- are relatively rare, and while we like to say that any team can win on any given Sunday, the best teams tend to come out on top, and almost always do over the course of a season. If you just look at the box scores, the games tend to look pretty darned boring.
This stability is present, according to Sides and Vavreck, not only in general elections but also in primaries. The polling for the 2011-12 primary campaign looks ugly, but when one peers closer, a pattern emerges. The authors explain that all of the candidates endured a cycle of “discovery, scrutiny, and decline,” with Romney’s support being the only real constant. Viewed in this light, this chart of primary polling suddenly makes sense.
The book is at its best when using data to destroy widely held shibboleths about the 2012 campaign; it functions as a sort of “mythbusting for elections.” Romney wasn’t the candidate of so-called Republicans in Name Only; he actually was viewed favorably by wide swaths of the party. Romney didn’t lose because he was ideologically out of step; voters actually viewed him as closer ideologically to themselves than they did President Obama. Obama’s vaunted get-out-the-vote operation didn’t matter much; there’s only a weak correlation between the placement of his field offices and his vote share. All told, these offices probably contributed less than a half point to his electoral total.
Finally, it’s worth noting that the book is accessible to the lay reader. One of the most unfortunate things about modern political science is how bogged down in jargon and formula it has become. If my father wanted to know about the politics of Southern states in the 1940s, I could hand him a copy of V.O. Key’s Southern Politics in State and Nation, and he could read and digest virtually everything in the book. If he wanted to know how representatives related to their districts, he could trudge through Home Style.
But if I handed him a copy of American Political Science Review to help him understand a current concept, he probably couldn’t get past the fifth page. This is no slam on my father, who is a very intelligent guy; even a couple of semesters in graduate statistics don’t translate to the fluency needed to read a good chunk of what is published in technical journals.
The Gamble serves as a helpful reminders that political scientists really can make useful contributions to scholarly literature that the lay reader can understand. Hopefully the increased interaction between political scientists and such readers can help pull the discipline back toward a more accessible writing style.
Does this mean that books like Game Change and its recent successor, Double Down: Game Change 2012, are useless? Not at all. Such books still serve two essential functions.
First, there are years where the fundamentals simply don’t work that well. The 2000 election is a classic example. Every model predicted a popular vote win for Al Gore and Joe Lieberman (which happened), but only one, the much-maligned FairModel, predicted a particularly close race. Everyone else had the Democratic ticket winning by at least five points, with some going as high as 19 points.
Every model has an error term that functions effectively, like the error margins we’re familiar with from polling. The journalistic narrative can help fill in blanks when, as was the case in 2000, we end up on one end of the error margin or the other. In other words, the big plays don’t always cancel out, and if a team puts up more big plays than the other team, they can overcome poor fundamentals. The journalistic narrative can help explain elections that political science has difficulty with.
It's useful to recall as well that if 2 percent of Michigan primary voters had changed their mind, or if 0.5 percent of Ohio primary voters had done likewise, Rick Santorum might have been the nominee. While the election wasn't as chaotic as Heilemann and Halperin suggest, neither was it as predictable as Sides and Vavreck sometimes seem to imply.
Second, the statement “game changers do matter, but are so frequent and random that they cancel out” is awfully unsatisfying (especially since they sometimes do not). Heilemann and Halperin help us understand the ins and outs of how the campaigns manage to respond to moves from the other side in a way that Sides and Vavreck can’t. If you want to understand a football game, you have to understand the quality of the team’s quarterback, offensive line, running game, and so forth. But the play-by-play commentary really does illuminate the game by explaining how the teams are attempting to compensate for their own weaknesses and maximize their own strengths.
In fact, Double Down and The Gamble often tell a similar story, though usually in a different language. For example, Sides and Vavreck refer to the Republican primary as a case of “discovery, scrutiny, and decline,” while Heilemann and Halperin have a chapter called “Dating.”
The overall point is that if you want to understand the 2012 election, you really need to read both books. It’s important to read The Gamble to understand the effects of the two conventions, and the meta-narrative is closer to reality than Heilemann and Halperin’s. Most economic models suggested that Obama should win by three points, and he won by four. That’s significant.
But at the same time, it isn’t enough to know how the conventions affected the election. Heilemann and Halperin help you understand that Bill Clinton didn’t just show up and give a great speech that energized the Democratic base; it took work by a variety of players. And if you want to understand the down-home charm of Clinton and his appeal to a group of voters that weren’t readily available to Barack Obama, you don’t need a model, you just need the scene the authors relate where Clinton describes Obama as being “luckier than a dog with two dicks.”
What this comes down to is the point made by Nate Silver’s The Signal and the Noise: Why So Many Predictions Fail -- But Some Don’t. Silver himself has moved somewhat over the past five years away from heavy reliance on models and toward a more hybrid approach that still leans toward the political science narrative, but incorporates the techniques of the journalist and historian. The bottom line of Silver’s book is that statistical modeling revolutionized baseball (and other fields), but that teams still need old-fashioned scouts to explain things that models don’t handle well.
The same is true of election coverage. The political science approach will tend to be more accurate and important over time, but the journalistic narrative still provides critical context for understanding the numbers that political scientists rely upon. If you really want to understand the 2012 elections, you should rely on The Gamble -- but you should read Double Down too.