From the August 2012 issue of Research Magazine • Subscribe!

Prediction Time

Who’s going to win the presidential race?

Forecasting election outcomes has taken on some of the trappings of a science in recent years. Polling data have become increasingly detailed. Prediction markets synthesize the bets placed by participants into probability figures about elections and other events. Various political, media, academic and financial experts use various quantitative and qualitative methods to predict political outcomes or at least sketch out likelihoods.

Unavoidably, gut feelings enter the analyses. The political scene is too complex, murky and turbulent to foster strong confidence in number-crunching. At the same time, even gut feelings need to be informed by data. In any case, some humility is a valuable asset in dealing with such a tricky field of hard-to-hit targets.

Now let us make some predictions. In so doing, let’s also say a few words about the track record of the column you are now reading, Political Monitor, which debuted in Research at the start of 2012. Too rarely do political analysts follow up on how their predictions fared.

In the column for April, I wrote about Americans Elect, the group seeking to offer an alternative presidential candidate:

“The organization’s potential to shake up the 2012 race is considerable. For a candidate nominated by this new group to win the presidency this year is very unlikely—though in this volatile political season, not something to be ruled out altogether. But even short of winning, that candidacy could have a significant impact, influencing the debate and perhaps tilting the balance of electoral support from one major-party candidate to the other.”

Wrong. Instead, Americans Elect shut down its Internet-based nomination process after no prospective candidate reached the threshold level of support required by the organization’s rules. On a brighter note, I was right to scoff at conspiracy-minded claims that Americans Elect was, as I put it, “designed to be a Wall Street vehicle.” What kind of conspiracy offers Internet voting and then folds its tent when not enough people click?

Generally, other forward-looking statements I have made in this column have held up decently, though to a large degree that is because the matters involved remain at least partly unresolved. In the January column, for instance, I expressed skepticism about the Occupy Wall Street movement’s future. So far that was right, though the upcoming political conventions—the Republicans in Tampa in late August, the Democrats in Charlotte in early September—will test Occupiers’ ability to regain the limelight. I expect they will have little success, having lost the advantages of surprise and novelty.

Elizabeth Warren “more likely than not” will win the Massachusetts U.S. Senate race, I wrote in February’s column. The subsequent “Fauxcahontas” controversy exemplifies the unpredictability of politics and likely weakened her momentum. A polling average monitored by the website Real Clear Politics showed Brown with 44.2% and Warren with 44% in May and June, versus a Warren lead of 42.5-41% in March and April. I hereby revise my Warren outlook to “slightly more likely to lose than win.”

The March column, titled “Post-Tea Party,” was written with a firm expectation that Romney would be the GOP nominee. That turned out to be correct, of course, but Rick Santorum’s surge shortly after the piece went to press was disconcertingly evocative of that “Dewey Defeats Truman” newspaper the re-elected president held up in 1948.

In the May column, I saw bright political prospects for New Jersey Gov. Chris Christie. However, I took a cautious tack in discussing whether he would be Romney’s vice presidential choice, noting both pros and cons to such a selection. Having not ventured far on the limb, I will have only modest bragging rights if Christie is the pick.

As it happens, Christie has declined steeply in Intrade’s prediction market for the GOP vice presidential nomination. The New Jersey governor traded below 3% in early July, compared to over 10% in early April. This decline may have less to do with Christie himself than with the rise of others as much-discussed contenders, notably Ohio Sen. Rob Portman and former Minnesota Gov. Tim Pawlenty.

Methodological Questions

How much stock to put in prediction markets is an intriguing question and the subject of considerable debate in academic and other circles. Some experts, such as University of Pennsylvania economist Justin Wolfers, contend the markets’ track record stacks up favorably against even the most sophisticated expert forecasts.

Others point to limitations of size and liquidity that may hinder the markets’ ability to synthesize available data. The debate thus mirrors the ongoing back-and-forth about the efficiency of financial markets. Suffice it to say that the last word has not been written on what types of predictive tools hold the most promise.

Presidential Prediction

Now let’s circle around to the big question of the political season: Who’s going to win the White House? Let’s look at some numbers.

Intrade’s contract for “Barack Obama to be re-elected President in 2012” generated a 56% chance on July 9. Trading had put it generally around 60% from February through April, probably not coincidentally the height of the raucous GOP primary season. The figure was below 50% for much of last autumn. Its all-time high was a brief peak at 69.9% in May 2011 on news of the death of Osama bin Laden.

At Real Clear Politics, a June average of 10 polls showed Obama at 47.6% to Romney’s 43.7%, a gap of 3.9 percentage points. This included a Bloomberg poll that was a much-noted outlier, which showed the president up 53-40%. The polls did not yet reflect public reaction to the Supreme Court’s upholding of the Obama health care law (an outcome the relevant Intrade market had given only a 30% probability, incidentally).

Statistician Nate Silver gave Obama’s re-election a 67.8% chance in his New York Times blog “Five Thirty Eight” on June 29, an upswing from his forecasting model’s previous 62.6%. Silver’s model uses polling data adjusted for “house effects” (the record of different polling firms in erring in favor of one side or the other in politics), along with economic variables including stock performance. An S&P 500 upswing following Europe’s bank bailout agreement was key to the improved Obama outlook.

Thus, some broad numerical pictures of the race have given a clear, but not overwhelming, edge to Obama. To be sure, those pictures are snapshots, and there could be sharp fluctuations in the months ahead in the polls, prediction markets and forecasting models. The sensitivity of Silver’s model to a shift in the S&P 500 illustrates the Obama campaign’s vulnerability to negative economic and financial-market data.

Moreover, quantitative methods carry the risk of obscuring or downplaying factors that are not readily quantifiable. In the presidential race, one such factor may be the presidential debates, scheduled for October. Both Obama and Romney are capable debaters, albeit ones with somewhat mixed records versus various opponents. In 2008, Obama seemed to win some, lose some versus Hillary Clinton; and to gain an edge in debating John McCain largely because of McCain’s limitations as a debater.

The 2012 Republican primaries featured an unprecedented number of debates. These encounters had major impacts on the race—undermining Rick Perry, for instance, and for a time aiding Newt Gingrich. Moreover, they gave Romney intensive practice at such encounters.

By contrast, President Obama goes into this fall’s debates without recent debating experience (except mock exercises with supporters) but now with an extensive White House record to defend. He also has a reputation for eloquence that will raise expectations that he will perform well in debates, and yet as president he also has had some tendency to seem thin-skinned in response to criticism.

It seems plausible that Romney will benefit considerably from this fall’s debates. Moreover, such a prospect—involving unique, personal interactions that have not taken place yet—is the sort of thing that likely would not get adequately taken into account in polls, prediction markets and forecasting models, showing up only after the fact.

Prediction: The debates will matter and push poll numbers Romney’s way. But let’s go out on a limb. Prediction: Romney is going to win the election.

Page 1 of 2
Single page view Reprints Discuss this story
This is where the comments go.