How The Tories Could Have Nobbled The Election

There has been a lot of talk about election fraud on Twitter which I am assuming is related in some way to the Tories' plans to bring in ID requirements for voting.   This is of course a form of voter suppression.   So there's a perfectly adequate scandal that is unfolding right before our eyes.  But I have become distracted by the possibility of an even bigger one.  I don't have any evidence for this.  Indeed I don't actually believe it.   But it is intriguing and as things stand at the time of writing it is still just about conceivable.


One of the arguments against voter ID is that there is little voter fraud to be protected against and that the British voting system is in fact pretty robust.  Votes are recorded in writing, and counted by hand.  Electronic manipulation can therefore be discounted.  A great many people are involved in the counting the process which is done in the open.  Counting is spread up and down the country and carried out simultaneously.  To try and rig the vote would need a very large team of people all of whom would have to work with great skill and dexterity to avoid being caught red handed.  Even political parties don't have that many people who are simultaneously trustworthy enough to be given details of a nefarious plot, unethical enough to be prepared to take part and artful enough to pull it off.

And the results once published are pored over by politicians, journalists, academics and the people who bet on the outcomes of elections.   All but the most minor and subtle of interventions would be picked up - probably sooner but certainly later. 

So on the whole the British voting system strikes me as pretty much fit for its purpose of counting the preferences of those voting.

And the broad outlines of the 2019 election result are consistent with the polls leading up to it, so on the face of it there is nothing much to worry about.   But all elections bring some surprises, and the extent of the Conservative penetration into northern seats was a bit bigger than I was expecting.  I don't know the north of England at all well so that is probably enough to explain my surprise.  But what if there was something going on? 

Because there was one rather remarkable statistic in this post election poll and analysis.

https://lordashcroftpolls.com/2019/12/how-britain-voted-and-why-my-2019-general-election-post-vote-poll/

This was quite a big survey and showed 38% as having voted by post.  This is over twice the proportion in the last election.  Scaled up this equates to 14 million postal votes being cast.  Only 7 million were cast in 2017.  This is a massive difference and shouts out for an explanation.   The most likely explanation is simply that Ashcroft has got it wrong.  His track record is not especially good.  But what if it is right?  Here's a possibility.

Suppose the Tories were looking for a way of boosting their vote.  One way would be to send unsolicited postal votes to the kinds of addresses that are more likely to vote Conservative.  They could identify hot prospects from their canvassing data, or they could buy market research data or they could commission their own polling.  It would have to be done with some care, because if done too blatantly it would look suspicious.   But so long as non-target seats had roughly the same activity as target ones it would be quite difficult to detect.  The results would just look like Conservative voters were a bit more motivated to actually vote than Labour ones.

Would it leave any detectable pattern in the voting data?  Maybe.  We know the numbers of postal votes cast in 2017, and also their distribution.  Every set of data has a characteristic distribution.  You can tell the difference between a football score and a rugby score without being told which sport for example.  Statisticians take note of the way data is distributed, it increases the inferences you can take from it.  There are quite a few distributions that are known and whose properties have been thoroughly evaluated.

 I don't know what the typical distribution for postal voting is but I'd hazard a guess it is the most common one.  It's so common it's called the Normal Distribution - it's the bell shaped graph you see if you go anywhere near any statistical data anywhere.  If you plot a graph of how many postal votes per constituency you will probably get this typical bell shape.  The number of people casting postal votes might have doubled for some reason unrelated to politics - fear of bad weather or not wanting to go out in the dark for example.  If so the shape of the bell curve will still be the same.   But what if a particular kind of voter has been sent a postal vote they didn't ask for while others were left to their own devices.  If so then the distribution of the number of postal votes will have changed and this would be possible to detect without having to be Einstein.  A-level statistics would suffice.

This plan would be possible to carry out with relatively few people, but would certainly need to hatched pretty high up in the Conservative Party hierarchy.   It might not be necessary for the leader to be aware of it though.   And although it would be possible to keep it pretty undetectable there would be the unavoidable risk that enough voters would query that they had been sent postal votes when their friends and relatives hadn't to bring the whole thing to light.  So on the whole I'd say that the likelihood is that this whole thing is no more than an interesting idea and that depressing as the result was for the future of the country it was probably fair.   But I'll run the numbers when the data gets into the public domain anyway.