What to make of the Mid Bedfordshire opinion poll?

Constituency opinion polls are hard to do well. Indeed, some of my posts which have aged the most badly, enough to make me cringe even when writing this sentence, were based on individual constituency polls in previous elections.

Parliamentary by-election opinion polls have the added twist of lower turnout, and hence turnout modelling carefully calibrated to try to get general elections right don’t always neatly transfer across. Plus also the dynamics of a by-election campaign mean that it is common to see much more dramatic movements during a by-election campaign than in a seat during a general election.

That said, the record of by-election polls judged by the final poll before polling day versus the result is pretty good. So they are in general worth paying attention to.

What, then, to make of Opinium’s online poll for the Labour Party in Mid Bedfordshire?

The poll passes various basic tests: it’s from a reputable pollster, with reasonable question wording and recent fieldwork (16-27 June). There’s a little bit of caution to make about the fieldwork as the poll started two and a half weeks ago, which can be a long time in a by-election. (I used to track the daily canvass data for the Lib Dems during a series of winning big swing Parliamentary by-elections and looking at some of my data from those days, quite a lot can change in such periods.)

But nothing obviously a major problem there.

However, look at those vote shares. As you may reasonably expect me to be – consciously or otherwise – biased towards the Lib Dems, let’s look at the combined Labour and Lib Dem vote share, leaving aside your or my personal preferences for how the split within that might play out.

Overall, the poll finds that the Conservative vote is down by 36 points, but that the Lab/Lib Dem share is only up by 8 points. In other words, over three quarters of the (net) decline in the Conservative vote in an English seat has going to people other than Labour or Lib Dem.

That would be a quite remarkable combined failure by Labour and the Lib Dems, at odds with the pattern in by-elections in England in this Parliament in Conservative seats or indeed in Conservative seats in previous Parliaments with a Conservative government. (I say England as the presence of nationalists in Scotland and Wales makes the situation in contests there understandably different.)

Where you do just occasionally see two or more of the main parties fail is when there’s a surge for an independent. Think George Galloway, for example. That indeed is what this poll shows too.

But is the 19% credible? It’s for independent councillor Gareth Mackey.

Is he really, thinking elections and not cats, a new George Galloway? Especially as, unlike George Galloway, he doesn’t have the advantage of being a former MP?

Of course, judging the strength of an independent from outside can be tricky but history can help. Looking at Parliamentary by-elections in England in which an independent candidate who was not a previous MP stood and saved their deposit, the best ever result was 28% and 22% in a 1946 combined universities seat by-election and then we’re down to 10% in the Hartlepool by-election of 2021, 8% in Wakefield in 2022 and 7% in Sedgefield in 2007.

In other words, 19% for Cllr Mackey would be a stand out historically brilliant result. It would be the best result for a pure independent in normal circumstances (i.e. excluding the oddities of those combined university seats) since the Second World War. (The same applies even if you relax the definition of ‘independent’ and include smaller fringe parties in the comparison too, though there was a 19% for one in 1963 which would make Mackey a joint record-breaker on this poll rather than a solo one.)

So while there have been some good results for the independents in council elections in the Mid Bedfordshire constituency, that alone is far short of what we’d need to explain a result like that.

But there is a mundane explanation available. Cllr Mackey has been a local Mayor three times, and it’s highly plausible given his political history that his support is very geographically concentrated. So perhaps the story here is not history in the making but problems with how geographically representative Opinium’s sample was?

There are also other reasons to doubt the sample when you dive into the crosstabs of the polls. As I have, ahem, written about in a certain book, you need to be careful not to over-interpret crosstabs based on small samples in particular.

But the poll shows Labour on 8% among the over 65s. Given the importance of older voters, especially in lower turnout by-elections, it’s hard to see how doing quite so poorly among the over 65s matches up with being on course to win. Lib Dem support in this poll is at double that level among the over 65s.

Then there’s the level of don’t knows: 30% in this poll, which seems reasonable for an early by-election poll. However, it also means that, starting with a sample size of 724, and with a further 9% refusing to answer, the voting intention figures come from the answers from just 441 people. As well as giving us a larger than normal margin of error, that also means that if just 10 people who said Labour had said Conservative, for example, the poll would have put the Conservatives ahead.

In other words, it’s a poll that looks to be particularly vulnerable to the luck of the draw on the sample, with it requiring only a little bit of bad luck – let along any geographic or turnout problems – to produce a poll with the wrong headline result.

Add to that the eyebrow rising 19% for Councillor Mackey, and it’s fair to conclude that it would be very surprising indeed if the poll turns out to be an accurate foreshadowing of the by-election result.

Get the essential polling book