The Corner

Scientism in 2012

One of the more curious aspects of this election is the appearance that almost more passion has gone into arguments over who will win than into arguments over who should win. In a New York Times blog yesterday, for instance, Paul Krugman claims that “you’re stupid” if you think this election is even “close.” He went on to rebuke the media for not reporting that Romney had already lost when a little cogitation would easily demonstrate the fact. It was, he argued, “lazy” reporting.  He might have added that they were missing a terrific scoop.

Also yesterday, Daniel Hodges went one better on his Daily Telegraph blog, and proclaimed confidently that even Romney’s own campaign is “finally admitting” that the election has “slipped away.” The evidence for this claim was somewhat thin, consisting of Mrs. Romney’s looking a bit down at one point on the campaign plane, and one Romney staff member who, when asked what he was doing in final stretch, replied “praying.” But Mr. Hodges presumably doesn’t need much evidence, since he maintained on the morrow of Romney’s knock-down victory in the first debate (and ever after) that Romney had already lost and the rest of the campaign could be canceled without any consequences to anyone.   

It’s understandable that observers with such well-founded knowledge get a little impatient when confronted by irrational skeptics. So when critics such as Josh Jordan and Dan Krumm produced skeptical analyses of Nate Silver’s statistical predictions of an Obama victory, Professor Krugman accused them of waging a war on science. Well, perhaps they will turn out to be mistaken. But their analyses were not different in form from Mr. Silver’s own work (which is impressive but not flawless) and scientific findings are always provisional. What is unscientific is proclaiming a new scientific finding two days before the actual experiment.

What is going on here? The two seemingly obvious explanations are confirmation bias and the bandwagon effect. Professor Krugman and Mr. Hodges want an Obama victory and so they look for polling results that tend to confirm their bias (and their hopes). Once they get poll results that fit this requirement, they broadcast them loudly in order to encourage others to join them on the bandwagon. These factors play a part, but they seem a little inadequate as an explanation of their certainty. We are all influenced by these two factors, but we don’t all enlist science to instruct the future to obey our wishes with such firm precision.

#more#I’ve expressed before my admiration for Mr. Hodges’s devil-may-care prediction that the election had already been decided several weeks in advance and for his fighting stoic refusal to amend this in the light of Romney’s slow surge prior to Hurricane Sandy. He’s a plunger, by Jingo, and if that single number on which he’s placed all his chips comes up on the roulette wheel, he deserves his winnings. But Professor Krugman is quite another case. His conviction of an Obama victory rests, as far as I can see, on an exaggerated respect for statistical forecasts above all other indicators of relative political support. How reasonable or even “scientific” is this?

Let’s begin by assuming that the polls are always accurate instead of, as I think, usually accurate. Even so, some indicators of electoral support are not easily captured by polling numbers — for instance, enthusiasm. If the supporters of one side are more enthusiastic, they are more likely to volunteer for the campaign and to vote. Most observers report larger and more enthusiastic crowds for Romney than for Obama. I spoke last week at UVA, where I was told (by a moderately enthusiastic Obama supporter, as it happens) that support for Obama was much reduced from last time whereas the smaller level of Republican support had held up well. Good organization would compensate somewhat for this enthusiasm differential, of course. But it’s still a factor that the polls may not capture. Attempting to express such factors in numerical terms is likely both to fail and to give a false impression of certainty. To ignore such factors because they resist such neat formulation is to compound the falsity. It is a mark not of science but of scientism, as Hayek points out. And in general it is better to be roughly right than precisely wrong. 

Another difficulty with statistical forecasts is that they are forecasts. They measure people’s opinions, attitudes, and voting intentions yesterday and not on Election Day. Professor Krugman was assuming that opinions would not change in the four or five days between the polls he saw and the actual casting of votes. When the time scale is short, this may seem reasonable; but when the gap between the candidates is narrow (as the polls suggest it is), then one candidate may very well overtake another — either because he was already enjoying a surge of support and needed only another day or two to put him over the top, or because events in the real world sharply and unexpectedly increase his support.

In the second case such “exogenous” events can produce a dramatic turnaround. Let me give two examples. Three Spanish elections back, the governing conservative Popular Party was expected to win an easy victory. All the polls showed this. When a bomb in the Madrid railway station caused a massive loss of life, ministers stated on the advice of their security experts that it was the work of domestic ETA terrorists. On the eve of the election, evidence emerged that al-Qaeda was responsible, allowing the socialist opposition to charge Ministers with lying. The Socialists won an easy victory. The second example: With one week to go in the 1970 British election, Labour led the Tories in all the polls by margins as high as twelve points. Three days before the election, trade figures were published showing a large deterioration in the balance of payments. This undermined Labour’s central election argument that it had successfully led Britain out of economic crisis, and confirmed Tory charges of economic failure. On Election Day, the Tories surprised everyone, including themselves, by winning a clear majority.  

Neither bombs nor economic indicators are likely to figure in this campaign. But Victor Davis Hanson senses that the long slow surge to Romney that was halted and reversed by Hurricane Sandy may now have resumed. Some polls seem to support him. And though unexpected events are by definition impossible to forecast, more news may emerge about Benghazi that will sap the president’s standing and reputation on foreign policy still further. Finally, Saturday’s off-the-cuff retort from the president to the effect that “voting is the best revenge” — picked up and thrown back at him by the Romney campaign in a brilliant quickie ad — may prove to be what Ben Wattenberg calls “a moment of political nakedness” (i.e., an unscripted event that tells the voters what someone is “really like”). Such moments — arguably, President Obama’s first debate performance was such a moment — can radically alter the electorate’s view of someone for good or ill. And they can happen on Election Day itself.

Am I suggesting that these things will happen? Not at all. Any one such event is most unlikely. The point is that even the most accurate polls cannot measure all the factors that produce an election result, and most obviously, the impact of those factors that occur after the poll has been taken. That is one reason, incidentally, why early voting is undesirable; it means that some people cast votes before all the relevant facts known to Election Day voters are known to them.

On top of all that, my initial concession for the sake of argument — that the polls are always accurate — is plainly wrong. Indeed, since the polls differ in every election, there are always some polls that get it wrong. These pollsters then seek to correct the blemishes that have produced their false predictions. But since the voters are constantly changing their behavior in ways unknown to the statisticians, they unintentionally introduce new obstacles to the poll’s accuracy. And the more that the polls themselves become major topics of interest and concern, the more likely voters are to alter their behavior in ways that distort poll findings. For instance, if voters sense that certain opinions are not really respectable, they may falsely deny holding them or simply refuse to answer questions at all. If such voters cluster in one party, that will depress its favorability ratings. In this campaign, only 9 percent of people contacted, one out of eleven, have responded to pollsters. Which party and whose supporters are likely to suffer most from this discouragement? I think we know.

Does that mean Romney is likely to win? Again, not at all. It means that his defeat is not certain, contrary to what Krugman and Hodges assert, or even very likely. For what it’s worth, my view all along has been that the likely result lay between the two extremes of a 1–2 percent victory for the president and a 5–6 percent win for the challenger. (The difference is explained by the fact that the voters already knew Obama well and Romney hardly at all — which gave the former a lower ceiling of support and the latter a better chance of improving their opinion of him.) My final guesstimate, which is rooted in several judgments, including the latest polls, is that Romney will emerge as the winner by a smaller margin than 5–6 percent, probably more like 3–4 percent. But I don’t know this for certain, since the future is unknowable, and I would not be amazed by an Obama victory. It would confirm that demographic changes unfavorable to the Republicans have reached the point where they have tipped the electorate into the Democratic column. Equally a Romney victory would show that the Obama team has been fighting a brilliant campaign to win the election of 2020.  Both are possibilities until Thursday night.

Krugman and Hodges, however, see only certainties. So does the Irish bookie, Paddy Power, who has started paying out on an Obama victory. He has already paid out $650,000. It’s an advertising gambit, of course, and given the modest sums being wagered, not a very expensive one. Still, the correct response for those who believe in uncertainty, free will, and the sovereign independence of the United States is to place a bet with Paddy on Governor Romney. At 10/3, the odds are excellent.

Let’s hope Paddy has sufficient funds left to pay out on Wednesday.          


The Latest