segunda-feira, outubro 01, 2012

Before the debates: polls and forecasts

Two days before the first debate, where do we stand in terms of election forecasts for the US presidential election?

1. Polls: polls are not forecasts, but they are of course informative. One obvious argument is that, the closest we get to the election, the less likely it is that any factor will emerge that may change the balance of forces. But the other - actually more plausible - argument is that, the more time elapses and the campaign progresses, the more likely it is that voters' intentions reflect the "fundamentals" of the election and the latent preferences of voters. At the moment, all poll aggregators place Obama ahead by about 4 points.

2. Models: econometric models use data from past elections, at the national or state level, and a number of contextual variables (mostly from the economy - such as change in national or per capita income, unemployment, and others - and sometimes from approval polls) to predict outcomes. The latest issue of PS collects almost all of them. The average is 50.4 for Obama, an extremely narrow advantage.  An article using ensemble Bayesian model averaging of those models estimates that "between 46.4% and 52.5% of the US voters will support the incumbent in 2012, and there is a 0.60 probability that the vote for Obama will be greater than 50%." Very close indeed.

3. Prediction markets: these are markets where investors buy and sell shares of political outcomes. Their performance has been extremely good, although there is a debate on whether they are "superior" to polls when the latter are properly treated. In any case, this is what the markets have been telling us until now:

Iowa Vote Share:

Iowa probability of winning:

Intrade Obama's probability of winning:


Enviar um comentário