Time to ditch constituency projections?
While polls might sometimes get things a bit wrong (although they’re usually more accurate than people think!), opinion polling is generally a positive thing in our democracy. It tells us (and politicians) what people are thinking and the direction of travel between elections. It would be hard to imagine life without it!
More information is usually a good thing, but when it’s obviously inaccurate, then ‘information’ can easily become misinformation.
That brings us to the many organisations that are offering projections on constituency level results. In this General Election, we’ve seen an explosion in websites and polling companies that claim to project results in individual Parliamentary Constituencies.
We should be clear that all the organisations involved in these projections are reputable and there is no suggestion that anyone is doing anything deliberately wrong.
Being able to accurately project constituency results should be a helpful tool, but can they really do it? Let’s look at one high profile constituency to see what that looks like from the TEN (by our count) different agencies and organisations offering seat level projections -
From this we can tell that either Reform or the Conservatives or Labour will win the seat. The Reform vote is somewhere between 17% and 52%, Conservatives between 16% and 44% and Labour between 18% and 33%.
Note the election was 10 days away when we wrote this!
Could you imagine if there was that level of variety in the national opinion polls - what would people be saying about the polling industry?
What are voters or journalists supposed to make of these?
We think that if projection information is going to be published, then it needs to be as accurate as it can be. Because clearly some of the numbers above are not!
In its current form, these kinds of projections are dangerous because people might make decisions based on these figures.
A news outlet might publish or broadcast a story based on Projection B saying that Reform is being smashed by the Conservatives in Clacton. Alternatively, they could tell the opposite story using Projection F. Choose the right projection, you can use it to support almost any narrative you want.
A tactical voting website might tell people to vote for one party over another party depending on which projection they choose.
A voter might decide on how to vote based on any one of the projections above, or indeed on what they read in the media or are told on a tactical voting website.
Party campaigners can choose which figures to publish in their campaigns - picking and choosing ones that back their story. Then watching it spread across social media.
None of this is good for democracy.
Clacton is not a unique example - below we’ve included projections from 10 more constituencies that also have a wide range of projected outcomes.
To reiterate, all ten of the projections above are from reputable organisations who are genuinely trying to do their best
But the fact is they can’t all be right. It may be that the sophisticated modelling and other techniques used to create these projections simply aren’t sufficiently developed.
We know that these types of projection systems work best in countries where there is a much more dominant two party system. If the national opinion polls are right, then we might see the lowest vote for the main two parties in more than 100 years. That might explain some of the volatility in the numbers.
But at present some of these projections are the equivalent of BBC weather saying that there’s a heatwave tomorrow, and ITV weather telling you that there’s going to be snow!
