Putting Your Money Where Your Model Is
There is nothing I like better than an academic wager. I fear that this time I’m over-matched: I’ve bet on politics against Columbia’s professor, Robert Erikson, and Bob is one of the leading political scientists in the U.S. But we had to bet, because his models of politics currently give a very different reading of 2008 than mine.
Let me start with some background data, drawing from my latest W.S.J. column:
A straightforward reading of recent polling data makes the presidential election look like a foregone conclusion … Polls from the past two weeks put Sen. Obama somewhere between 4 and 12 percentage points ahead of Republican John McCain. In the poll-of-polls aggregate on RealClearPolitics.com, Sen. Obama’s lead is a massive 5.7 percentage points — as large as it has ever been.
Senator Obama’s lead is larger than the stated margin of error …. Moreover, Sen. Obama is leading in all of the major battleground states, except Virginia and Florida which remain incredibly close.
How should we interpret polling data four months prior to an election?
Political scientists Robert Erikson (of Columbia) and Christopher Wlezien (of Temple) have recently mined daily polling reports from the last half-century of elections, mapping the relationship between early polling numbers and final election returns. At this point in the race, they find that around half of any lead should be discounted, as early advantages tend to dissipate. (You can read the full paper here, or an ungated version here.)
Profs. Erikson and Wlezien point to another reason to be wary of Sen. Obama’s early polling lead: On average, the voting public tends to be more strongly anti-incumbent three-and-a-half years into an administration than they are on Election Day. Based on patterns in previous cycles, the professors suggest that this exaggerated anti-incumbent feeling is boosting Sen. Obama’s lead by around three percentage points.
So if you first halve Obama’s six point lead, then subtract a three point anti-incumbency bias, you are left with a dead heat.
A naïve reading of polls suggests an Obama landslide; a sophisticated reading points to a dead heat. Prediction markets are somewhere in the middle, suggesting a two-in-three probability that our next president will be a Democrat.
Yet it is the naïve reading of the polls … that dominates media headlines.
There’s a broader intellectual debate here, too. Erikson and Wlezien note that most comparisons of polls vs. prediction markets are comparisons of the naïve reading of the polls versus the prediction markets. That’s a pretty weak test, and so when they ran comparisons of their adjusted-polls versus markets, the markets no longer looked so good.
In fact, over the past four election cycles, the poll-adjusted forecasts did slightly better than the markets. Bob scores this as a victory for the polls over the markets, while I think that four election cycles are too few to overturn the fact that in nearly every domain I’ve seen, prediction markets yield the most accurate forecasts. (You can read more on our debate here and here.)
Given Bob’s model, he should be willing to accept an even money bet: say a bottle of wine for a bottle of wine. Given the markets, I should be willing to bet my two bottles against his one. And so we compromised: I’ve bet three bottles that Obama will win against his two bottles on McCain.
Bob tells me I’ve got the upper hand in this bet: in ongoing research he is finding that polls are slow to reflect economic conditions. Adding in a third adjustment to account for the effects of a weak economy, he concludes that Obama should be the betting favorite.
But he tells me he is willing to accept the bet for another reason: “If McCain wins, I could use that wine.”
The full column, comparing polls and markets through the election cycle, is available here.