In April, when Prime Minister Theresa May announced that Britain would hold a snap election on June 8, it looked like her Conservative Party would claim a landslide victory. However, recent voter intention polls suggest that the race is tightening — and that the Conservatives could lose seats.
Our forecasting model, based on voter expectations of who will win, predicts that the Conservative Party actually will increase its majority, winning about 361 of the 650 seats in Parliament.
British pollsters have two big misses. They were off target on the 2015 parliamentary elections and the 2016 Brexit referendum. This demonstrates the challenges of voter intention polls: identifying who will turn out to vote, how “undecided voters” will vote and what constitutes a representative survey sample. When predictive models rely on inaccurate voter intention polls, inevitably, the projections will be wrong, sometimes dramatically so.
Can citizens forecast the outcome? We think so.
Our seat share model takes a different approach — we use a survey question that asks respondents who they think will win the election. With such “citizen forecasting,” it does not matter whether the respondents turn out on Election Day or if they change their vote choice. Instead, this approach aggregates information on what citizens expect will happen. In essence, we draw on the “wisdom of crowds” — the idea that large groups make better decisions than any single individual.
In a current paper, we test the accuracy of our voter expectations model vs. the leading models based on British voter intentions from 1950 to 2015. The citizen forecasting model has a higher chance of correctly predicting which party will win — and higher accuracy, on average, in estimating parties’ share of seats in Parliament.
To predict Thursday’s election result, we estimate models for Conservative and Labour seat shares from monthly voter expectations data (1950-2015) compiled from Gallup, the British Election Study and the Essex Continuous Monitoring Survey. Our model also includes a control on previous seat shares. Thus, to make the prediction, we use current voter expectations, along with previous seat shares, to forecast what the next Parliament will look like. More details on our methodology are here.
The YouGov survey suggests the Conservative Party is going strong.
We use the May 30-31 survey by YouGov/the Times. The voter expectation question asks, “Regardless of what you would like to see happen at the general election, what do you think WILL be the result of a general election on 8th June?”
The survey results show that the British public is quite confident that May will continue to lead the government, with 62 percent of respondents expecting the Conservatives to win a majority and an additional 7 percent believing that the Conservatives will lead a hung Parliament — when no party claims a majority. Just 12 percent of respondents say Labour will garner a majority or lead a hung Parliament. The remaining respondents either “don’t know” or think another party will win.
In total, 69 percent of Britons say the Conservatives will be the largest party in Parliament. When we enter this into our statistical model, the prediction is that the Conservatives will obtain 361 seats (55.6 percent of the seats) — see the figure below. This would be a gain of 31 seats — a result that would validate May’s decision to call a snap election so early in the parliamentary term.
What does this mean for the Labour Party, with just 12 percent of the public expecting it to lead Parliament? Our prediction is that it will secure 236 seats (36.3 percent of the total). This would increase Labour’s seats by seven but still represents its fourth–worst election performance in postwar history.
The figure depicts the estimated seat share probability distributions for the Labour Party (red) and Conservative Party (blue) based on our citizen forecasting models. The centers of the two distributions represent our estimates of the parties’ seat shares — our best guesses — that correspond to the Conservative and Labour seat share percentages above. However, as with any statistical model, other lower probability outcomes are also possible. (*We are grateful to Joe Greenwood for pointing us to the May 2017 YouGov/the Times survey.)
DATA: Gallup, the British Election Study, the Essex Continuous Monitoring Survey and YouGov/the Times. NOTE: The data was collected before the terrorist-linked attacks in London on June 3. FIGURE: Murr, Stegmaier and Lewis-Beck, 2017.
These two seat share distributions are virtually separate from each other — they overlap only slightly in the center. Based on this, we are 97 percent confident that the Conservative Party will win more seats than Labour. There’s a 77 percent chance that the Conservatives will win an outright majority of seats and an additional 20 percent chance that they will lead a hung Parliament.
Our forecast of a 20 percent chance of a hung Parliament led by the Conservatives stands in contrast to the recent YouGov projection of a hung Parliament. That model uses voter intention polls and other factors to predict the results in each constituency. The Times reported May 31 that “YouGov acknowledged that the predictions were controversial and pointed to significant ‘churn’ in voting intentions.”
While the media likes to focus on the “horse race” aspect of campaigns by tracking voter intention polls, this type of poll relies on assumptions that can negatively affect its predictive power. This was what we saw in the 2015 general election and again with the 2016 Brexit vote. Based on our research, we think citizen forecasts have more to say about what happens this week.
Andreas Murr is an assistant professor of quantitative political science in the department of politics and international studies at the University of Warwick. His research focuses on election forecasting, the voting behavior of immigrants and the selection of party leaders.
Mary Stegmaier is an assistant professor in the Truman School of Public Affairs at the University of Missouri. Her research focuses on voting behavior, elections, forecasting, and political representation in the United States and abroad.
Michael S. Lewis-Beck is the F. Wendell Miller distinguished professor of political science at the University of Iowa. He has written or co-written more than 270 articles and books, including “Economics & Elections,” “The American Voter,” “French Presidential Elections,” “Forecasting Elections,” “The Austrian Voter” and “Applied Regression.”