Fundamentals models for presidential elections

Academic political scientists tend to discount the value of opinion polls as predictors of presidential elections and tend to look at the so-called ‘fundamentals’. They construct models that correlate vote percentages with data that can be quantified. The ‘Bread and Peace’ model of Douglass Hibbs that I have written about recently is one such model that uses disposable income and wartime casualties as the independent variables. Of course, there exist a whole range of independent variables that one can choose to use in one’s regression calculations and they each predict different outcomes.

While the Hibbs and some other models seem to predict that president Obama should lose handily to Mitt Romney, opinion polls show him clinging to a small but persistent lead.

On the surface, the 2000 election seemed to provide another anomalous result for the fundamentals models. On the basis of some models (such as Hibbs’s), it seemed as if Al Gore should have won easily but the final result gave him just a wafer thin victory in the total vote. As a result of that election, people speculated about the various non-fundamentals such as style and political missteps as explanations. But Larry M. Bartels and John Zaller published an article in the March 2001 issue of the journal PS: Political Science and Politics found that when one examined the aggregate of all the different models, the result did predict a close outcome.

In any case, our primary point is that specific explanations of the 2000 result are quite superfluous. Given the systematic, predictable effects of the economic and political “fundamentals” embodied in our 48 regression models, there is simply nothing special about the 2000 election outcome to be explained, Indeed, to a greater extent than most elections, 2000’s really did come out the way it “should” have, at least with respect to the national popular vote.

The idea that one should look at the average predictions of all the models seems to have become the new conventional wisdom. A political scientist colleague of mine who follows these things closely says that currently the average gives Obama a very narrow and unstable lead and that the Hibbs model is at the low end of the range for his vote total.

The October issue of PS: Political Science and Politics will feature the predictions of these models for the coming election and I will report on them as soon as they are released.


  1. left0ver1under says

    There’s an old political joke: “We hold elections to prove if the polls are right.”

    I doubt nearly every pre-election “prediction” because most are partisan. Is anyone biased towards a particular party really going to admit that the other guy is ahead? Not likely. Personally, I’d ban polling within a fixed time before an election, 30 or even 60 days.

    Pollsters are trying to emulate the 1984 US presidential election:,_1984

    Early east coast returns on election night favoured Ronald Reagan so heavily that people in the western US who were undecided or might have voted democrat decided to vote for Reagan. What might have been a close election turned into the biggest landslide in US history, with Reagan taking 49 of 50 states. The problem was so big that new rules were enacted preventing the media from publishing early returns.

    Many people who would have voted for Mondale gave up, believing they couldn’t win, and that is exactly the intent of pollsters and predictors today: to fool the public into believing a specific candidate will win, that there is no point in voting against him. It’s a form of propaganda and voter intimidation.

  2. jamessweet says

    The problem with the “it’s only the fundamentals” hypothesis is that, at least downticket, we see that it is demonstrably not true. Akin’s most recent poll numbers are presumably not the result of a sudden shift in fundamentals, eh? I find it difficult to believe that similar factors could not powerfully influence a presidential election.

    That said, I do think that it’s clear the fundamentals matter a whole lot more than any sort of silliness about “the issues” in determining who wins the election.

Leave a Reply

Your email address will not be published. Required fields are marked *