American Politics Research

This article revises, updates, and examines the background for a highly accurate model for forecasting the national two-party popular vote in presidential elections. The model provides a vote prediction in early September based on Gallup trial-heat or presidential preference polls and the (nonannualized) rate of economic growth in the second quarter of the election year. It is estimated over the 12 presidential elections from 1948 to 1992. The mean absolute error of the model's out-of-sample postdictions is less than 1 1/3 percentage point, and its actual error in predicting the 1992 vote was about half a percentage point. The article also assesses the reasons for confidence in the model, as well as an approach to gauging uncertainty in any specific forecast. The reasons presidential elections can be forecast at or before the beginning of the general election campaign also are explored. Finally, the forecasting model is applied to the 1996 presidential campaign between Clinton and Dole.

In this article I descnbe a model for forecasting the outcomes of U.S. presidential elections. The model uses three predictors: the incumbent president's approval rating at midyear, the annual rate of growth of real gross domestic product during the first half of the election year, and the length of time that the president's party has held the White House. The time factor most clearly distinguishes this model from other presidential forecasting models. Using this model, it is possible to forecast the outcome of the presidential race in early August with greater accuracy than most final preelection polls. Based on the president's approval rating in mid-May (55%) and the rate of economic growth during the first quarter of 1996 (2.8%), the model yields a conditional forecast of a decisive victory for Bill Clinton in November with approximately 56% of the major party vote.

To forecast presidential elections, I explore the dynamic of the vote ("time") and introduce a measure of candidate support that covers both the incumbent and the challenger. Stochastic models help identify the dynamic of the presidential vote as second-order autoregressive. The strength of the candidates is gauged by an index of electoral success in presidential primaries—in particular, whether the nominee won the first pnmary. Also included as a vote predictor is the economy, as measured by gross national product (GNP) growth and inflation in the election year. The forecasting equation predicts victory for Bill Clinton, with 57.1 % of the major party vote in November 1996. Time is on his side, in the sense that the autoregressive dynamic favors election of a presidential candidate whose party just captured the White House. But what predicts a comfortable margin is Clinton's edge in the candidate comparison, with the economy exerting little electoral pull this year.

Presidential election forecasting models may miss the mark, sometimes grossly, as the 1992 contest demonstrated. The reason for this, we argue, is specification error. The models include irrelevant variables and exclude relevant ones. In particular, prospective voting variables have been ignored. When prospective economic and political evaluations are added, alongside traditional retrospective evaluations, forecasting quality improves sharply. These full-time forecasting models that tap voter onentations toward the future, as well as toward the past, promise long-run accuracy gains.

In this article we present a simple forecasting model that has been successful at predicting past presidential elections. The two variables included in the model are cumulative per capita income growth and presidential approval. These "fundamental" variables predict the vote especially well when measured shortly in advance of the election, when the outcome is already becoming clear in the polls. Their predictive power drops quite quickly as one steps back from the election, however; readings of income growth and approval taken early in the election year only modestly predict readings of the same variables just prior to the election. Thus we turn to leading economic indicators that allow us to forecast presidential elections by giving advance indication of changes in the economy and approval during the election year. For 1996, our model offers a cautious prediction of a Clinton victory.

Building on the work of previous forecasters, I develop a model of presidential elections that deviates from earlier work by including a measure of aggregate personal finances. The results of the analysis indicate a highly accurate model and predict a Democratic victory in 1996. The discussion of findings emphasizes that, although the model predicts a Democratic victory, caution should be exercised before concluding that the outcome is cast in stone or that the campaign cannot make a difference.

We use the distribution of responses to the Gallup poll's "generic" House vote question (i.e., "If the elections for Congress were being held today, which party's candidate would you like to see win in your congressional district?") to forecast the Republican-Democratic vote split in presidential year congressional elections. Although these responses consistently offer an overly optimistic picture of the Democrats' prospects, they nonetheless serve as the basis of accurate forecasts.