People are surprisingly good at predicting local election results

People are surprisingly good at predicting local election results

I have previously looked at how British Election Study respondents view local crime rates (spoiler: people are very bad at judging local crime). But I had not realised there are also questions, last asked just before the 2019 election, asking respondents which party they thought would win locally.

Being honest, I expected the predictions to be terrible, like on crime. Instead, BES respondents are pretty good psephologists. The BES asks two questions: one on each party’s likelihood of winning in the respondent’s constituency (on a scale from 0 to 100) and another asking them to predict each party’s vote share. On the first question, around 7 in 10 people correctly gave their highest likelihood rating to the party which went on to win.

On a side note, very few people chose to make their answers (each on a 0-100 scale) add together to 100… but in fairness the question never specified probability (“How likely is it that each of these parties will win the General Election in your local constituency?”).

This number varies a lot from constituency to constituency. With small sample sizes in each constituency, it is wrong to say too much about individual results, but it seems as though respondents in safe Conservative seats were by far the best at predicting Conservative victories. Meanwhile, the worst predictions came in former Labour safe seats gained by the Conservatives in 2019. Plotted below are the 23 constituencies where fewest respondents correctly predicted a Conservative victory. They are all in the North, Midlands or North Wales.

Some might say it almost looks like a sort of… wall… in re… *gunshot*

Added together, the wisdom of the crowd means respondents collectively predicted the correct result in 91% of constituencies. These are plotted below, with incorrect predictions highlighted:

Of the 59 constituencies that respondents incorrectly predicted, 42 were won by the Conservatives, 7 by Labour, 5 by the SNP and 3 by the Liberal Democrats. Respondents correctly predicted every single constituency with a majority over 16,000, while, as you would expect, the predictions were closer to 50/50 for very marginal constituencies.

On the vote share question, the median prediction had a 13.1% error for the Conservatives, 14.3% for Labour, and 16.0% for the Liberal Democrats. This might not sound great, but a significant proportion of people came very close to predicting the correct results (21% came within 5% of the true Conservative vote share and 41% came within 10%).

This is especially interesting because BES also asked respondents to predict how the parties would fare nationwide, and there is little or no correlation to local constituency results. This suggests respondents are differentiating between nationwide indicators (polling, news coverage, etc) and local (past results, social networks) to make informed predictions about how the parties will perform.

None of this sounds massively consequential, but it does slightly change how I think about issues such as tactical voting. If voters have a fairly good sense of local political winds, perhaps that changes their own behaviour more than I would have guessed.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Create a website or blog at

%d bloggers like this: