Why Wikipedia has too many voting intention polls

Welcome to this week’s edition, taking a look at why Wikipedia has more voting intention polls listed for this Parliament than (in my view) it should.

But first, a little non-polling bonus: here are the cumulative principal authority council by-election results for the first quarter of the year:

As ever, if you have any feedback or questions prompted by what follows, or spotted some other recent polling you’d like to see covered, just hit reply. I personally read every response.

---

Been forwarded this email by someone else? Sign up to get your own copy here.

If you’d like more news about the Lib Dems specifically, sign up for my monthly Liberal Democrat Newswire.

---

My privacy policy and related legal information is available here. Please note that if you are subscribed to other email lists of mine, unsubscribing from this list will not automatically remove you from the other lists. If you wish to be removed from all lists, simply hit reply and let me know.

---

Wikipedia has a problem with the polls it lists

I usually describe my PollBase archive as the largest collection of British national voting intention polls. There are more polls out there waiting to be rediscovered from before 1979 but it’s the biggest dataset.

Except… if you look at data for this Parliament, you will find more polls listed by Wikipedia than I have included in PollBase. What’s up?

Having discovered the discrepancy I’ve been doing some digging, and the answer is informative of political polling more generally.

That’s because there are two types of polls which include questions about how people would vote.

The most obvious are those which ask such questions with a view to producing national vote results. But to do so, they usually don’t only ask people how they would vote. They also have to deal with people who say ‘don’t know’ and also how to cater for the fact that not everyone votes. The techniques for dealing with these two issues vary between pollsters, with varying degrees of modelling involved. But what this almost always means is that more questions, and more analytical work, are required to get from the simple voting question to the headline national vote share figures that they publish.

Then there’s the other type of poll. Suppose, say, you are polling about crime. You don’t want to know national vote share figures, but you do want to know how answers vary by political affiliation. So you ask the initial voting question, but don’t ask those other questions and don’t do that other analytical work. Your published results therefore don’t have headline voting intention questions, but do have crosstabs which show the results to your crime questions broken down by political affiliation.

But… because we have pretty good transparency rules for political polling in this country, the full details of the second category of polls are published. And that includes the voting intention question, complete with answers including flavours of don’t knows, wouldn’t says and wouldn’t vote.

So perhaps you can just take that question, recalculate the percentages excluding those categories and bingo, you’ve got national voting intention figures that those pesky pollsters omitted to publish?

Except… you haven’t.

Because those extra steps that the pollsters would have done, such asking about propensity to vote and adjusting the results based on the answers, haven’t been done. You’ve skipped a set of steps and so you don’t have a proper set of headline voting intention results from the poll.

That’s what one or more editors on Wikipedia look to have done: digging into some published tables, doing that first calculation but then added these incompletely calculated figures into the tables on Wikipedia.

That’s why you’ll find more apparent polls from focaldata and JL Partners in Wikipedia than in PollBase.

You will also find more polls from Omnisis, but as Omnisis, unusually (though not necessarily wrongly - we’ll have to see who has the best methodology come the election results), does not do those sorts of extra calculations, the figures extracted from their tables are legit, and so some older polls of theirs will appear in the next PollBase update.

Thank you to all three firms mentioned for responding so quickly and helpfully when I was writing this.

---

National voting intention polls

Here’s the latest from each currently active pollster:

For more details and updates through the week, see my daily updated table here.

Last week’s edition

Are things looking up for the Conservatives in the polls?

Know other people interested in political polling?

The problem with averaging polls and other insights from this week’s polling…