In conjunction with my New York Times Magazine article on President Obama’s re-election prospects, my colleagues Matt Ericson and Shan Carter have designed an interactive feature that allows you to handicap Mr. Obama’s re-election odds based on the assumptions of your choosing.
The forecasts are based on a relatively simple three-variable model that I developed in conjunction with the article and that reflect the most important things that we should be thinking about at this stage. The variables are:
– Economic growth (as measured by G.D.P.) in 2012.
– Mr. Obama’s approval ratings a year in advance of the election.
– The ideology of Mr. Obama’s Republican opponent — is he relatively moderate or very conservative?
The first variable, economic growth, is of course very important. By default, the feature will show you three scenarios — zero percent G.D.P. growth (tantamount to a mild recession), 2.3 percent G.D.P. growth (the average prediction in the most recent Wall Street Journal forecasting panel), and 4 percent G.D.P. growth (better than the forecast and slightly above-average historically). However, you can use the slider to see the effects on the race if G.D.P. were to grow at virtually any other rate instead. (The Fed expects G.D.P. to grow at about 2.7 percent next year, for instance.)
The second factor, Mr. Obama’s approval rating through November 2011, is something that you should probably think of as being locked in rather than variable. Approval ratings a year in advance of the election do have some predictive power, and they are the best way to evaluate how the public regards a president’s performance during his first three years. (From a forecasting standpoint, it is redundant to consider what economic performance was like in the first three years once you’ve accounted for a president’s approval rating.)
By default, the model uses a 43 percent approval rating for Mr. Obama, which was the average as of late last month. If you prefer a different number — for instance, his approval rating rose to 45 percent in the Real Clear Politics average as of Wednesday — you can change that assumption under ‘advanced settings’. Keep in mind, though, that this variable should reflect Mr. Obama’s approval rating as of now — a year in advance of the election — and not what it will be on Election Day 2012. (If we knew what Mr. Obama’s approval rating was next year, his fate would be fairly easy to determine and there wouldn’t be much need for all this effort to handicap his odds.)
The final parameter, opponent ideology, is not something you can modify directly but instead is reflected in a series of match-ups against five potential Republican opponents who range from quite moderate (Jon M. Huntsman Jr.) to extremely conservative (Michele Bachmann). This factor reflects a candidate’s ideology — where she falls on the spectrum from moderate to conservative — and ideology alone, since we have decent ways to measure that and since it has been an important predictor of election outcomes historically. It does not reflect other variables related to “candidate quality” like favorability ratings, his qualifications for the Oval Office, and so forth. (Those qualities may be important but they are hard to measure and little data exists on them from past races). The estimates also do not reflect Mr. Obama’s head-to-head polling against the various opponents, which are not highly reliable at this stage of the campaign.
The five candidates included in the feature are Mr. Huntsman, Mrs. Bachmann, Mitt Romney, Rick Perry and Herman Cain. If you think Newt Gingrich or Rick Santorum might win the Republican nomination instead, you can use Mr. Perry as a stand-in for them since his ideology rating is virtually identical to theirs. If you think Ron Paul will win, you should probably go ahead and make your own assumptions about how he might play to the electorate; a one-dimensional ideology spectrum will not do a very good job of measuring the views of a libertarian.
I will have more detail on how the ideology scores are calculated in a subsequent article, but they are based on a combination of three statistical systems: (i) DW-Nominate scores for candidates like Mrs. Bachmann who have been in Congress; (ii) CFscores, developed by the political scientist Adam Bonica, which estimate a candidate’s ideology based on his fund-raising; and (iii) surveys, which have asked voters to assess the ideology of the candidates on a five-point spectrum from very liberal to very conservative.
These variables allow us to compare the candidates against the ideological ratings for past ones as developed by the political scientist Martin Cohen and his colleagues. Although ideology may not be as easy to quantify as something like G.D.P. growth, I have tested Mr. Cohen’s system along with a variety of other ways of measuring ideology and they are highly statistically significant predictors of election outcomes.
There is one last advanced setting that you are allowed to tweak. By default, the scenarios you see displayed in the feature reflect our handicapping of the race if we knew exactly what G.D.P. was going to be next year. If instead you want to use a forecast of G.D.P. — you think it will grow by 2.5 percent next year, but it could be higher or lower — you can click the box for “use forecasted rather than actual G.D.P.” and the model will modify its assumptions accordingly. The difference is simply that this will introduce more uncertainty into the forecast: predictions made of G.D.P. growth a year in advance are often quite poor. However, this is probably a more realistic assumption.
Note that the choice to use G.D.P., the most comprehensive overall indicator of economic performance, is important in my view. I am wary of models that use more obscure variables or combinations of variables that happen to “fit” the past data better but that will produce overconfident forecasts when applied predicatively.
Instead, our model is happy to acknowledge that there is considerable uncertainty in a forecast made this early out — the candidates will debate between various measures of the economic health, with voter perceptions fungible to some degree. And and there are all sorts of non-economic factors that we cannot account for.
As you will see when you play around with different assumptions, while the model can tell us something about the most likely outcomes, there is a chance that the election will not go to type and that Mr. Obama will lose decisively despite robust economic growth next year, win under circumstances that look bad for him, and so forth. In the Magazine article, I provide a lot of detail about all the things we do not and cannot know about next year’s election, making a devil’s-advocate case for some of these possible-if-not-likely scenarios.
Finally, note that everything in the model refers to the popular vote. We will go into much more explicit detail on the Electoral College once more reliable data is available on that next year, but it is premature to do that for now.