Friday 28 June 2024

The battle of the dodgy bar charts

 

The Lib Dems are the undisputed champions of the dodgy bar graph, complete with manipulated scales, which always somehow seem to show that the Lib Dems are either in the lead, or else are the leading challenger to whoever seems to be in the lead in any given constituency. Imitation, they say, is the sincerest form of flattery, and at least three parties (Labour, the Tories, and Plaid) are at it here in the Caerfyrddin constituency. Maybe the Lib Dems are too, but since we’ve had zero information from them to date (please don’t bother now, I’ve already voted), I can’t be certain whether even they would have the chutzpah to claim that they are ‘Winning Here’.

Unsurprisingly, the three bar charts I’ve seen don’t all show the same thing. Choosing which poll to use is a bit like the way that parties hold up different economists to justify their costings – they always choose to invest authority in the one that presents figures closest to their own views. Plaid are using a poll which shows Plaid as the clear front-runner with Labour some way behind and the Tories in third place. Labour and the Tories are both using polls which show Labour in the lead and the Tories in second place. You pays your money and you takes your choice: the accusation by Labour that Plaid are lying just because they are using a different poll was somewhat over the top, though. All are based on those large MRP polls which tend to be more accurate in the round than the smaller polls, but which are still based on models and assumptions. Even a poll of 13,000 is likely to include only around 20 voters per constituency, nowhere near enough to give an accurate local picture in itself. Accuracy depends instead on an analysis of demographics and an assumption that people in similar categories will vote in similar ways in different constituencies. It’s a fair assumption at a high level, but can never pick up on the hyper-local factors.

As to what’s actually happening in Caerfyrddin, who knows? In its various guises, the seat has swung back and fore between Plaid and Labour for decades; any poll showing those two parties in contention is potentially accurate, and either could be in the lead. The Tories have a solid base of support, and came second in the old Carmarthen East constituency in 2019, well ahead of Labour. They might well have a shout on the new boundaries in a good year for the party. But nobody with eyes to see would think of 2024 as a good year for them; more a year of total collapse. Any poll which shows Labour and the Tories in a close two-horse race thus fails the reasonableness test – it just ain’t credible. But then, credibility isn’t the aim. The aim, generally, is to persuade anyone planning to vote for a candidate not in the top two for the constituency (according to the selected poll) to choose instead whichever of the top two they dislike least. It’s a product of an unrepresentative voting system, but since the top two in most constituencies are Tory and Labour, neither of those parties sees any incentive to change the system. That means we’re stuck with a plague of dodgy bar charts assailing us from all sides, like some sort of graphical arms race, to increasingly peripheral effect.

3 comments:

Anonymous said...


Go with the bookies. Plaid are hot favourites.

Anonymous said...

I think the poll used by the Tories is an all Wales poll and not a specific Caerfyrddin poll. It is interesting there is no title for the poll in both the election communications I've received from thema and if I'm right in my assertion the claim by Simon Hart that he is the only challenger to Labour in the new Caerfyrddin seat is clearly wrong.

John Dixon said...

Anon2,

The leaflet that I've had from Hart specifically refers to it being from the Survation MRP UK-wide poll released on 4th June, so they're not claiming that it's a specific Caerfyrddin poll. Nor are Plaid making any such claim, by the way - they're using the IPSOS UK-wide MRP poll from 18th June. In either case, the number of electors from within the constituency will be tiny, and the accuracy of any constituency-level projection depends on the accuracy of the modelling based on demographics, which inevitably ignores any strong local factors. I don't know which poll Labour are using - their bar chart popped up on social media rather than in any leaflet that I've seen, and I can't now find it. I suspect, though, that it's the same Survation poll as the Tories are using.

Anon1,

Suggesting 'going with the bookies' is more than a little naughty in the light of the gambling scandal!