What will be the impact of the late surge in electoral registration?

What will the impact be of the surge in new electoral registrations in the run-up to the deadline to get registered for the general election? Here’s some data.

Curiously, although the surge has got much coverage, it has rarely attempted to put the number into a proper context, particularly how many votes this might alter per constituency.

The Electoral Reform Society deserves an honourable mention for their detailed press release, though it doesn’t go into the details of the possible partisan effect which I follow below and the coverage I’ve seen citing the ERS figures often stop short even of how far the ERS went.

So step along with me as I figure out whether the number is big enough to have a likely chance of altering the result of the general election.

How many registrations were there? From the calling of the election until the deadline for registering in time to be able to vote in this general election, there were 3,850,859 applications to register.

But there was a similar surge of applications last time too. So if we want to know whether the surge this time is likely to change the election result, we need to look at the change on last time. (For more on this point, see my tweets.)

Last time there were 2,938,291 in the corresponding period, so the increase in the number of registration applications was 912,568.

Next – how many of these were actually new registrations, rather than duplicate registrations, i.e. where even without the application that person would have still been on the register at that address? The level of duplicate applications is sufficiently large as to often be a cause of complaint or spur to calls for changes in the system from those involved in the electoral registration system.

Let’s take the 2017 duplicate number and assume that things were the same this time round. The Electoral Commission’s estimate was that 36.9% of applications were duplicates in 2017. For 2019 that would reduce the 912,568 figure to 575,830 people actually being added to the electoral register.

Next up, we have to estimate how many of these people will vote. Just because you register specially for an election does not mean you will vote, just as not everyone who applies for a postal vote then actually casts it.

I’m not aware of any data on what the turnout level of late registrants is, though we do know the age profile of them skews young (as it did last time) and younger people are less likely to vote. On the other hand, perhaps those who register late are more likely than average to vote. Given the turnout last time overall was 68.8%, let’s go for 70% turnout amongst this group.

That brings the figure down from 575,830 people to 403,081 votes. That is 620 votes per constituency.

But not all those 620 people will vote the same way. To gauge the possible impact, let’s focus on Labour for the moment as most of the media coverage has been about whether Labour will benefit.

Perhaps late registrants skew heavily Labour, especially given their overall skew towards being younger. Labour currently is around 10 percentage points behind the Conservatives in the polls. Rather than being 10 points behind, let’s go for saying that Labour might be, say, 20 points ahead with the late registrants. (That 30 point difference may well be on the generous side, but as with all the estimates above you can rework the numbers with your own estimates.)

That would mean that the net effect of the 620 votes in a constituency would be to give Labour 124 more votes than the Conservatives.

If everything else were the same as in 2017, that margin would give Labour one extra seat. Only one.

Another way of looking at is to expand the net slightly, and take into account that it’s not only Labour who might net benefit from late registrants. Plus, of course, vote shares have changed in seats since 2017.

So another way of looking at it is that last time there were 12 seats where an extra net 124 votes for the party in second place would have made them the winner. But not everywhere will see the party in second being the one that is a net gainer from late registrants so that’s a maximum theoretical figure rather than a likely one.

The conclusion then? The net impact of the late registrants is likely to be, at most, in the high single-digit number of constituencies.

You can re-run the numbers above and come up with a lower figure with a bit of basic caution – such as if turnout amongst late registrants is lower than average and/or if there isn’t such a big variation from the national average levels of support among them. You can also re-run the figures to get a higher total, though it’s hard to see how credible numbers in the calculations can still give you anything other than the conclusion that in a super-tight election, late registrants might tip the result.

But of course, in a super-tight election, a myriad of other factors can also be pointed to as the one that tipped it. Including the type of building and floor level for polling stations.


Update: in response to this post, several people have made comments along the lines of ‘but what about X factor…?’, such as what if the number of new registrations is concentrated in marginal seats. Two things are worth noting about such ideas. First, that there isn’t evidence, so far at least, that such other suggested factors are true (and new data since I wrote that suggests that if anything the new registrations are concentrated in the seats where they are least rather than most likely to make a difference). They are good questions to ask but, in the absence of evidence, are not therefore good reasons to be sure of something different. Second, given how small the final impact is in the calculations above, even if there is some other factor that greatly increases it, a big increase on a small number can still end up small. Doubling that bottom-line likely impact still leaves a pretty small impact.

Update #2: I’ve updated the figures used from the Electoral Reform Society in line with the data here instead of using those in the ERS press release linked above. They add an extra 25 people per constituency to the calculations but this is not sufficient to alter the numbers of seats that might be influenced. Thank you to Justin DeVito for pointing out this improvement to the data.ย 



Keep up with news about elections by email

If you’d like to be notified by email when new posts about how are how our elections are run appear on this blog, just sign up here. (Note: if you’re already signed up for a daily email alert with all my new blog posts, then there’s no need to sign up for these alerts too as the stories will also be in the full daily digest.)

  • If you submit this form, your data will be used in line with the privacy policy here to update you on the topic(s) selected. This may including using this data to contact you via a variety of digital channels.
  • This field is for validation purposes and should be left unchanged.

Interested in more stories about how our elections are run? Follow my dedicated election law channels onย Facebook or Twitter which include bonus extra stories that don’t appear on this blog.


There are 8 comments Share your views

Share your views

Your email address will not be published. Required fields are marked *

All comments and data you submit with them will be handled in line with the privacy and moderation policies.