|Table 1: National popular two-party vote forecasts|
A common explanation of electoral behavior is that elections are referenda on the incumbent party’s performance during the term that is ending. For more than three decades, scholars have amplified and tested this theory, most commonly by developing econometric models, usually to predict the outcomes of US presidential elections. Most models include between two and five variables, and they typically combine indicators of economic conditions and of public opinion to measure the incumbent’s performance.
Since the 1990s, the forecasts of competing models have been being published regularly at around Labor Day of the election year. For the past five elections, the forecasts of leading models have been published in American Politics Research, 24(4), and PS: Political Science and Politics, 34(1), 37(4), 41(4) and 45(4). Most of the models predicting presidential elections have produced forecasts using data available near the end of July in the election year.
One study analyzed the track record of eight models that were that were available for each of the five elections from 1996 to 2012. The models’ track record in predicting the winner is impressive. Out of a total of forty forecasts, there were only four cases in which a model’s forecast predicted the wrong election winner. For the models’ vote share forecasts, the picture is somewhat mixed, however. The models’ best year was 2012, when the average error was only 1.9 percentage points. In contrast, the models had their worst year in 2000, when the average error was at its highest value (5.1 percentage points). Across all five elections, the average error was 3.1 percentage points. In comparison, a naïve forecast of predicting that the votes are split equally among the two major parties (i.e., 50%) in each of the past five elections would have yielded an error of 2.4 percentage points, which is 24% lower than the average error of the forecasting models. Also, there was no clear pattern of which model is most accurate. In each of the five elections, a different model ranked first (Graefe, 2013). Furthermore, the structures of some of the models have changed over time, so it is difficult to identify the most accurate models.
Use of econometric models in the PollyVote
Table 1 shows the forecasts from the econometric models that are currently available as well as the combined econometric model forecast. Most of these models have been developed by economists and political scientists, who provide their forecasts around Labor Day of the respective Election Year. However, some models release their first forecast as early as two years before Election Day. Polly continuously keeps an eye on updated and new model forecasts and incorporates these into her prediction.