With the election in 3 days, we are seeing a continued narrative of a tied or very close race. If you read Media Matters, the explanation for this narrative is not that it is true, but that the media requires and in some sense produces a close race in order to , well, frankly, in order to increase sales.
Into this fray comes a new debate about whether in fact the race is close, and whether polling data is showing that it is close. Specifically, the reports of Five Thirty-Eight, run by Nate Silver, and now hosted at the New York Times. Silver’s data show that the race is heavily favoring Obama, who he says has a 4-in-5 chance of winning on Tuesday.
On the other hand are a bunch of commentators who are either saying that this is a close race or that Romney will win. Not just partisan observers like Dick Morris and Karl Rove, but journalists too. Perhaps even many ordinary people on the right also expect Romney to win. Obviously they can’t both be right, so the question is, will the right conclusion be reached by those who were wrong about why they were wrong? Furthermore, if they are right, will they reach the correct conclusion about why the other guy was wrong? Going on past evidence, I’d have to say no!
–Like it or not, polls are usually correct, and they are even corrector when you’re dealing in the big picture and multiple polls. (They may miss specific local results or even a whole state. If I recall correctly, FiveThirtyEight missed correctly calling one state by 1% in the last election.)
–The success of polling doesn’t mean that good journalism can play no role. As always, but perhaps especially during an election, investigative journalism is needed to get behind the sound bites and information that campaigns release.
–I wish people would ask, who do you think will win, rather than who do you want to win. It’s probably more accurate about the outcome, especially if you are asked to put real money on it. At Intrade, they do exactly that. Right now it gives Obama a 67.1% chance of winning. That’s lower than FiveThirtyEight’s current 83.7% chance, but still a bet I’d want to take.
–What about the “black swan” effect my friends in the intel community are always worried about? Yes, there could be some startling last minute crisis or disclosure. But it’s getting awfully late. Here’s an election advantage that the intel guys don’t have: there’s a definite deadline by which the estimates are good for. Presumably in the intel world, estimates have to be made over indefinite time periods. FiveThirtyEight only has to be good through 7PM on Tuesday.
–What about bias? Ah, yes, voter fraud right? No, I doubt it. Rather the polls themselves might be biased, because they have to make assumptions. One important assumption is how to model the sample against the larger electorate. Do you take likely voters or registered voters? Do you ensure that Dems are <50% of the sample the Republican areas, and if so by how much? Without these assumptions the polls will be less representative. The same logic goes into the decennial census: sampling rather than brute force-attempt at headcount is going to be more accurate. (The well-known effect of the “undercount” in the Census means that certain populations are always undercounted, however hard you try to count everybody.)
Another form of bias, again more likely than organised voter fraud, is trouble with the electronic voting machines such as Diebold. There’s no doubt this goes on (Ohio being the most well-known case) but it’s not as clear how widespread it is.
So if you want to make some money, bet on Obama being re-elected president. Good news no doubt for liberals who’ve bought into the “close race” narrative. But will the conservatives and the journalists with a vested interest in a close race make the correct conclusion? Here’s Silver’s parting shot:
Nevertheless, these arguments are potentially more intellectually coherent than the ones that propose that the leader in the race is “too close to call.” It isn’t. If the state polls are right, then Mr. Obama will win the Electoral College. If you can’t acknowledge that after a day when Mr. Obama leads 19 out of 20 swing-state polls, then you should abandon the pretense that your goal is to inform rather than entertain the public.