I have some sympathy for the UK election pollsters who now face the prospect of an independent inquiry, alongside a tidal wave of critical comment.
Given the ‘systematic overstatement’ of the Labour share there is undoubtedly an issue of accuracy. But this will be true of most research that relies on claimed and anticipated behaviour. The problem for polls is they have the responsibility of predicting who governs the country and can’t escape a forensic comparison of actual Vs predicted behaviour!
A more credible cause for sympathy, however, is my suspicion that the divergence between actual and claimed intentions reflects a number of ‘behavioural influences’ and the relatively complex choice faced by voters. To improve accuracy pollsters will need to understand the significance of potential influences and how the choices voters faced impacted these.
To illustrate I have highlighted 5 influences below that may have contributed to the difference between the polls and the election result. These may not explain much (or any) of the discrepancy but should highlight the complexity that needs to be considered when criticising the pollsters.
1. ‘Shy’ Tory bias.
The phrase shy Tories dates back from the 1992 election where again the polls did not predict the result. There does appear to be a deep-rooted reticence to admit you are voting Tory relative to Labour. That reticence may also vary by age of voter too. Do people feel that supporting a more socialist orientated party will meet with more social acceptance?
2. Risk aversion
We know from behavioural economics that we value what we now relatively highly when compared to potential future gains. For this election was the risk of losing stability and progress seen as particularly high? Firstly, the economy has recovered creating a gain for many of us to hang on to. Secondly, the poll results continuously reminded us that no party would win a majority and that the status quo was at risk.
3. Research methods and media failed to replicate the voting experience
Casting a vote, for most was a form filling activity that included local council elections and named candidates. Certainly the walk to a polling station and crossing of ballot papers will have exceed the one or two minutes given to an on-line survey. The structure, presentation and effort involved for on-line surveys may have contributed to a different choice from some. Put another way, were polls getting more ‘System 1’ responses than the real event?
4. Voters embraced the easier choices
Was the choice faced by voters more complex than in previous elections? (Just picture the election debate with the rows of party leaders). Ukip, SNP and others were given a platform alongside the two main parties potentially making it more difficult for voters to identify with one single political party. Significantly, the outcome of choices (various coalitions) was also uncertain.
The two winning parties appear to have represented a simpler choice. For SNP it was more autonomy for Scots and for the Tories it was don’t discard the progress.
5. The Conservative campaign was more effective
With around a third of voters apparently undecided the activity in the last few weeks would have been vitally important. People have a tendency to remember what is most recent. Were the Conservatives (who were better funded) more focused and more effective in influencing behaviour in the final days or weeks?
As a voter in a marginal constituency the communication I received was overwhelmingly Conservative. Addressed to me personally, the choices were simplified to “competence or chaos” with the risk of progress being lost emphasised. In addition, it was clearly explained that my marginal vote was especially important – together with visual evidence that it could impact the national result.
So 5 potential influences to explore alongside others to understand the discrepancy between the election results and the polls prediction of behaviour. Given such influences and the complex choice faced by voters we need to understand more before we wholeheartedly condemn the pollsters performance.