Why Hillary Clinton Isn't the Favorite After All

Ask around: Washington is pretty certain Hillary Clinton is the favorite to win the White House. Democrats have a natural turnout advantage in presidential years, seasoned political operatives reason. Five of the past six popular-vote tallies have gone to the Democratic candidate. And early polls that show Clinton sporting a big lead, especially among women, have strategists wondering how the Republican nominee could ever catch up.

But outside of the capital, from Georgia to New York to California, there's another set of political professionals watching this race: academics and model-makers. And based on the data they track, Democrats have little reason to be so bullish about Clinton's chances.

"Viewing her as a prohibitive favorite at this point is misplaced, definitely," says Alan Abramowitz.

Abramowitz isn't a Republican pollster or a professional Clinton-hater. He's a political science professor at Emory University in Atlanta. And he and his ilk—the wonky academics who research in anonymity while pundits predict races on TV—offer the most compelling case for reconsidering Clinton as the likely winner.

"I would feel comfortable saying that it's a 50-50 race right now," says Drew Linzer, a political scientist who is an independent analyst in Berkeley, California. "But I don't think anyone would be wise going far past 60-40 in either direction."

Veteran political operatives regard these predictions as nothing more than musings from the Ivory Tower. But political scientists who specialize in presidential-race forecasts aren't relying on their guts. They've built statistical models that draw on the history of modern presidential campaigns (since Harry Truman's reelection in 1948) to determine with startling accuracy the outcome of the next White House contest.

The best-known forecasting tool of the bunch—and one that plainly spells out Clinton's looming trouble—is Abramowitz's "Time for Change" model. He first built it before George H.W. Bush's 1988 election, and he has used it to predict the winner of the popular vote in the seven White House races since. (The model predicted that Al Gore would win the presidency in 2000, when he became the first person since Grover Cleveland to earn the majority of the popular vote nationally but lose the Electoral College.)

The model uses just three variables to determine the winner: the incumbent's approval rating, economic growth in the second quarter of the election year, and the number of terms the candidate's party has held the White House. Official forecasts aren't made until the summer before the presidential election. But reasonable estimates rooted in current political and economic conditions demonstrate Clinton's vulnerability.

Consider this scenario: President Obama retains equal levels of approval and disapproval, better than he has had most of his second term; and gross domestic product growth in the second quarter of 2016 holds at 2.4 percent, the same as last year's rate of growth. Under this scenario, the "Time for Change" model projects that Clinton will secure just 48.7 percent of the popular vote.

In other words, she loses.

Slight increases in Obama's approval rating and economic growth aren't enough to change the outcome for Clinton. Every 10-point improvement in the president's approval—if, for example, 55 percent of voters approved of Obama while 45 percent didn't—earns Clinton only an additional 1 percentage point of the popular vote. It takes an extra 1 percent year-over-year GDP growth to give Clinton an extra half percentage point of the popular vote.

For Clinton to reach 50 percent of the popular vote, under this model's rules, the president would need to see a 5-point increase in his approval rating and GDP growth would have to hit 3.5 percent. It's certainly possible, but it's fair to call that a best-case scenario for Obama in his final year as president.

So while Democrats see the recent gains in both Obama's approval and economic growth as signs that Clinton enters the race as the favorite, the academic modeling suggests that assessment is far too sunny. In fact, the recent uptick is the only thing keeping her from being a prohibitive underdog.

The reason Clinton struggles under seemingly decent conditions is obvious. After one party holds the presidency for two terms, voters want change. In the model, this desire for a new direction manifests itself as a 4-point reduction in the candidate's take of the popular vote compared with what candidates could expect had their party held the White House for just one term.

"One of the regularities you'll find for all presidential elections since World War II is, after a party has been in power eight years and is trying to hold on to the White House for a third consecutive term, it gets harder," Abramowitz says. "Another way of looking at it: In the first election after a party takes over the White House, you have a significant advantage. And the next time, after you've held another term, you lose that advantage."

Campaign operatives love to hate this academic assessment of politics, much like Wall Street belittles the technical analysts who use past performance to predict stock-market moves.

The tension between the strategists and the scientists speaks to the distinct approaches they employ: Political professionals (including journalists) study strategy, tactics, the day-to-day activities of a campaign, while political scientists see fundamentals shaping every election, almost no matter the strength of a candidate.

In 2012, for example, most strategists think Obama won because he ran one of the best presidential campaigns in American history while Mitt Romney ran one of the worst. According to political scientists, however, Obama's victory was a product of favorable conditions, such as an improving economy, decent approval ratings, and his incumbency. The unemployment rate was high, yes, but the state of the economy matters little compared with the direction it's headed.

In an era of hyper-professionalized, financially flush campaigns, it is this set of fundamentals that will make the difference between winning and losing, the scientists argue.

"The notion the campaign doesn't matter, it's not that simple," says Michael Lewis-Beck, a political science professor at the University of Iowa. "It doesn't matter so much because everyone is campaigning so hard that they cancel each other out."

Lewis-Beck showcased his own presidential model—one of many that now dot the political landscape—on the political science blog Monkey Cage. Academics began developing statistics-based predictions as early as the 1970s, but they have become more popular and mainstream since Nate Silver correctly forecast the 2008 and 2012 presidential elections.

In an era of hyper-professionalized, financially flush campaigns, it is this set of fundamentals that will make the difference between winning and losing, the scientists argue.

Silver is a controversial figure in the political science world, where he's seen as a practitioner who went mainstream and came to define the entire forecast-model genre. As Romney supporters can attest, Silver's forecasts have been accurate, but they also depend on polls—many of which are not yet available or are of little use this far from Election Day. This means that Silver's forecasts might not be accurate untila couple of months before an election, and to political scientists who develop models, the goal is not just to be accurate but to be accurate long enough before an election to make a true forecast.

Other methods abound. Political scientists have used job growth and state-based economic indicators in their models, for example, while Abramowitz tried to update his own to account for increased polarization among voters. (He plans to scrap the update after it was less accurate about the 2012 election than his old model.) Lewis-Beck says the public's expectations for how a presidential race will turn out are predictive, while a historian at American University has a checklist of conditions that must be met for an incumbent party to win reelection.

One political scientist, Helmut Norpoth at Stony Brook University in New York, bases his model entirely on which party holds the White House, and for how long. It lets him make predictions years in advance: He has already forecast that Republicans have a 65 percent chance at winning the presidency next year.

"There's a cyclical pattern in the elections," Norpoth said. "It swings back and forth. And you can see it in the time lines since 1828."

Models can be wrong, of course. Norpoth says his 2008 prediction missed in part because he made it before the onset of a financial crisis that tanked the economy. And even models with a better track record aren't perfectly calibrated.

The biggest assumption that all the models make is that Republicans and Democrats will nominate someone from the mainstream of their party—and that might amount to a fatal flaw in predicting 2016, when the GOP could pick a candidate, such as Ted Cruz or Rand Paul, who's not favored by the establishment.

And other potential problems lurk: Models that suggest Clinton would earn 49 percent of the vote come with a margin of error that might make the difference; Abramowitz worries that because there have been only six modern-day presidential elections in which no incumbent from either party is running, his model's sample size might be too small; and in a race between Clinton and an equally talented, outsized personality, such as Jeb Bush, the qualities of the individual candidates might matter more.

But conceding that the models aren't perfect isn't the same as saying they're not effective. When I talked with Linzer, I argued that Clinton has an advantage. It comes down to women, I said, especially educated white women who, early polling shows, have a special affinity for the former senator and first lady. How can the GOP hope to persuade enough members of this group to break away to win swing states such as Colorado and Pennsylvania?

"It's just way more complicated than that," Linzer said. "For every argument that you can pick out of the cross-tabs, I can pick a counterargument. Off the top of my head: She's not going to earn the same enthusiasm that Obama did among nonwhite voters."

As he put it, our brains trick us into believing things that seem plausible but don't hold up to scrutiny. It might seem plausible that Clinton is a favorite, but the historical record simply says otherwise.

"I'm sorry," Linzer said, "to rain on your thought parade."

Because, yes, to the scientists, it's not our thoughts about this election that count. It's the data.