The danger in the latest British Election Study data

The news that the British Election Study (BES) has added a set of constituency spending figures to its hugely useful dataset of constituency results from the 2015 general election may seem a somewhat unremarkable incremental improvement that will make the lives of researchers a little easier.

Yes, but.

There is a huge magnetic pull from the data included in any data set which often trumps its actual validity or usefulness. Quite simply, put data in a file and people are strongly drawn towards using it without first questioning closely its validity or relevance.

The problem with constituency spending data is that it is still the norm in the political science community to use it as a measure of the intensity of campaigning by that side in that constituency. That is what in theory it should measure. But it doesn’t.

This in part is a matter of the current legal cases involving the Conservative Party and whether all its constituency campaigning was properly recorded.

Only in part, though. Because even if every last hotel room and bus mileage is accurately reported, constituency expense limits do not capture the national campaigning that can be, and is, quite legally concentrated in a small number of marginal seats.

Constituency expense limits do not measure constituency campaign. Run Facebook ads featuring your party leader in just your marginal seats, and that’s a national expense, not a constituency expense. Ditto for billboard posters. Ditto for direct mail.* Ditto for hired parrots reciting your election slogans.**

Yet the constituency expense limit data still have its magic pull on political scientists who just can’t quite resist temptation and use it as if does. They use it to analyse campaigning and, even worse, they use it to ‘adjust’ for campaigning to investigate other phenomena.***

Fellow attendees at the annual conference for political scientists – EPOP – will over the years have heard me raise this question frequently. And the answers are always the same – yes, you have a point… but it’s the only data we have so it’s the data we use and now back to talking about the conclusions based on assuming it is accurate…

The magnetic pull is so strong there isn’t even usually a passing footnote in academic articles making a passing reference to how the authors know the data is not perfect.

So whilst I can understand why the BES has added the data. It should really come with a warning. Whatever you are about to use it for is almost certain to mislead you.


* I simplify slightly, but only slightly. Parties can legally spend many times the limit in constituencies on campaigning aimed to sway the voters in them and without counting against the local limits.

** A tactic previously used in Indian elections.

*** If the amount of national expenditure which goes on local campaigning was proportionate across all seats and all parties to the constituency expenditure totals, then these would still act as a useful proxy. That is, however, a huge assumption and one that goes unstated, not to mention untested.

One response to “The danger in the latest British Election Study data”

  1. Ref Indian Election use of parrots … as long as they are not Norwegian Blue Parrots as in Monty Python 🙂

Leave a Reply

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.