Citizen forecasts

Vote expectation surveys ask people who they expect to win the election. For example:

“Who do you think will win the US presidential election?”

The aggregate responses of these citizen forecasts are then used as a forecast of who will win the election. If data on historical elections are available, one can convert the aggregate responses to popular vote-share forecasts. Although vote expectation surveys have been around at least as long as scientific polling, they have long been overlooked as a method for forecasting election outcomes.

Accuracy of citizen forecasts

One study compared the accuracy of citizen forecasts to forecasts from polls, prediction markets, quantitative models, and expert judgment. Across the last 100 days prior to the seven US presidential elections from 1988 to 2012, citizens’ vote expectations provided more accurate forecasts of election winners and vote shares than each of the four established methods. Compared with polls, vote expectations reduced the error of vote-share predictions by 51%. Compared with prediction markets, error was reduced by 6%. In other words, citizen forecasts appear to be the most accurate individual method for forecasting US presidential elections available to date.

Citizen forecasts in the PollyVote

PollyVote added citizen forecasts as a new component method for forecasting the 2016 election. An ex post analysis of previous elections showed that citizen forecasts would have further increased the accuracy of the combined PollyVote forecast. Across the last 90 days prior to the six elections from 1992 to 2012, adding vote expectations to the PollyVote would have reduced the error of the original Pollyvote by 7%.

Responses to the vote expectation questions cannot be directly used as vote-share forecasts. In order to translate the results to forecasts of the vote shares, one study estimated the following vote equation based on 217 surveys conducted between 1932 and 2012:

V = 41.0 + 17.1 * E,

where V is the actual two-party popular vote share of the incumbent party and E is the two-party percentage of survey respondents that expect the incumbent party candidate to win. That is, the incumbent starts out with 41% of the vote, plus a share that depends on people’s expectations. If respondents expect the incumbent to win increases by 10 percent, the incumbent’s vote share increases by 1.7 percentage points.

The model also reveals the strong degree of partisanship among US voters. As noted by Campbell, “no matter how bad the campaign goes for a party, it can count on receiving about 40 percent of the two-party vote; no matter how well a campaign goes for a party, it will receive no more than about 60 percent of the two-party vote.” The vote equation reflects this as it predicts that the incumbent would receive at least 41 percent of the vote, even if nobody expected him to win the election (E = 0). Conversely, if all survey respondents expected the incumbent to win (E = 1), the model would predict the incumbent to receive at maximum 58.1 percent of the vote.

For example, a Washington Post- ABC News survey conducted from March 3-6, 2016, found that % of respondents expected the incumbent party’s candidate Clinton to win, whereas % expected  the Republican Trump to win. Entering these figures into the vote equation above leads to a forecast for the Democratic two-party vote share of %. Note that you can adjust the figures in the green highlighted cells to see how Polly’s citizen forecast would change depending on the survey results.