Big-issue model

The big-issue model, developed by PollyVote team members Andreas Graefe and Scott Armstrong, predicts the outcome of U.S. Presidential Elections based on polling data on how voters expect the candidates to handle the most important issue facing the country.

The big-issue model is simple to use and easy to understand, and it has decision-making implications. Political candidates can take action upon the forecast and develop campaign strategies. Two different strategies seem conceivable for allocating campaign efforts. First, candidates can try to be seen as the best candidate for dealing with the most important issue. Second, candidates can try to change voters’ perceptions about which issue is most important.

Data

The model relies on two sources of polls

Polls that ask people which issue they regard as most important, such as ‘‘What do you think is the most important problem facing this country today?’’

Polls that ask people which candidate / party will do the better job in handling that issue, such as “Regardless of which (2016) presidential (election) candidate you support, please tell me if you think Hillary Clinton or Donald Trump would better handle each of the following issues. How about..the economy?’’

Method

The model uses a simple decision rule, which performs three basic steps, in order to predict the election winner:

Identify the issue seen as most important by voters,
Calculate the two-party shares of voter support for the candidates on this issue, and
Predict the candidate with the higher voter support to win the popular vote.

Simple linear regression is then used to relate the incumbent party candidate’s rolling average (calculated by using exponential smoothing) of voter support on the most important issue (BI) to the dependent variable, which is the actual two-party popular vote share received by the candidate of the incumbent party (V). Using data from the last forecasts prior to Election Day in the thirteen elections from 1972 to 2020 leads to the following vote equation:

V = 25.4 + 54.5 * BI

That is, the big-issue model predicts that an incumbent would start with 25.4% of the vote, plus a share depending on BI. For example, if the incumbent’s voter support on the most important issue went up by 10 percentage points, the incumbent’s vote share would go up by roughly 5.5 percentage points.

Past performance

When the model was first proposed, an ex post analysis for the ten elections from 1972 to 2008 showed that the big-issue model provided forecasts that were competitive compared to forecasts from econometric models and the Iowa Electronic Markets (IEM) prediction markets (Graefe & Armstrong, 2012). An analysis of the model’s first ex ante forecasts for the 2012 election provided similar results. The big-issue model was again more accurate than the IEM and the typical econometric model (Graefe & Armstrong, 2014).

To this day, the model provided ex ante forecasts for the three elections from 2012 to 2020. The chart below shows the average absolute error one would have achieved when relying on the Big-issue model from each day until Election Day across (and in each) of the three elections from 2012 to 2020. For example, across the last four weeks (or 28 days) across the last three elections, the Big-issue model’s average error was 0.6 percentage points.

The forecasts of this simple model are highly accurate, especially when compared to benchmark approaches. As can be seen below, across the last 100 days before the three elections from 2012 to 2020, the Big-issue model would have consistently outperformed both the forecasts from the combined PollyVote and FiveThirtyEight. For example, assume that, four weeks (or 28 days) before the respective elections, you would have decided to rely on the Big-Issue model instead of the two benchmark forecasts. Then, you would have cut the forecast error of the PollyVote roughly in half. Compared to FiveThirtyEight, error reduction would habe been even higher, at 60%.

References

Graefe, A., & Armstrong, J. S. (2014). Forecasts of the 2012 U.S. presidential election based on candidates’ perceived competence in handling the most important issue. Political Science Research and Methods, 2(1), 141-149.
Graefe, A., & Armstrong, J. S. (2012). Predicting elections from the most important issue: A test of the take‐the‐best heuristic. Journal of Behavioral Decision Making, 25(1), 41-48.