MPs Expenses and safe seats - is it happening again?

When I did my blogposts back in 2009 that found an apparent link between the safety of an MPs seat and the likelihood of them having been involved in the expenses scandal of that year it caused a bit of a stir. It got a fair bit of coverage and there were some who thought (hoped?) it might make a difference to the campaign for electoral reform.

There were however numerous people who dismissed what I had found claiming it was not statistically significant and/or there were problems with the methodology I used.

I never claimed to be a statistician (although my work was reviewed by someone who is better qualified on that score than me and he certainly thought there was something in it) but I did think it was striking when split into quartiles how the safer an MP's seat, the higher their apparent chances of being involved in the scandal.

Obviously the fuss from the original research has long since died down and I hope I have proved in the intervening years that I am not some sort of safe seat correlation obsessed one trick pony but when the latest expenses scandal broke I couldn't help but wonder if I was to apply my original analysis to it whether we would see anything different. I actually expected that we would not see the same sort of thing, after all this is a quite significantly changed parliament and as various people were queueing up to tell me before, my original findings apparently were just coincidental.

So surely this time around we would find no link? Certainly not a similar shaped graph to last time anyway?

Well I only have 14 names out of the 27 supposed transgressors so far but....

Er....

I've been FactChecked!

Channel 4 News' fact-checking department did a piece yesterday which analyses my safe seats and expenses scandal correlation work from back in May.

It has been prompted by an article that Nick Clegg wrote in The Daily Telegraph on Wednesday about expenses where in the final paragraph he says:

And of course, if we’re to sort out politics fully, we need to think about the future, not just the past. Britain needs voting reform to abolish the safe seats that make MPs complacent – those in safe seats were three times as likely to be identified as having abused their expenses by The Daily Telegraph. If we want to rebuild faith in politics, there can be no half measures. Only fundamental reform will be enough

Nick Clegg's office confirmed to Channel 4 that these figures were based on my research.

A very nice lady named Alice from Channel 4 called me on Wednesday to discuss all of this and to request the source data. At the time I didn't realise that they were going to do a full article where they analyse the work in detail and essentially try to pull it apart and see how it stands up. You can find the results of their analysis here. It is worth reading the whole thing to see the points they make for and against the research.

Addressing one of the counter-arguments to the usefulness of the source data they say the following:

...How important is the minister-factor in the make-up of the Telegraph's list?

It's not the case that everyone included is a big name. Thompson told us there were a large number of MPs he hadn't heard of - and he's a self-confessed political anorak.

Phantom mortgage claimants Elliot Morley and David Chaytor, both of whom have now been banned from standing again as Labour MPs, weren't exactly big household names. Neither was Margaret Moran, she of the £22,500 dry rot bill, or Fabian Hamilton, or Stephen Crabb - all mentioned in the early Telegraph investigations.

Their verdict is:

Clegg's claim is based on some fairly detailed analysis of a big tranche of MPs which made up the Telegraph's early expenses revelations, and found those in the top quartile of safe seats were three times as likely to feature as those in the bottom quartile.

Re-doing the analysis at a later date, when the paper had covered more MPs, showed those with the biggest majorities were twice, rather than three times, as likely to feature as those with the smallest.

Clegg implies that one factor causes the other, using the finding as a basis to call for reform of the first-past-the-post system which gives MPs such differing majorities.

But doesn't necessarily stack up. A correlation between two factors does not mean that one causes the other, and it's also worth bearing in mind that top frontbenchers are more likely to have safe seats than MPs as a whole.

Still, it's an intriguing piece of research - though perhaps one that we should be careful about stating as political fact without question.

I think their conclusion is fair here although I also think it is fair for Nick Clegg to highlight this situation in articles and bring it to people's wider attention.

I am pleased to see on their scale they give Nick's claim a 2 which on their scale of 0 - 5 seems pretty good (the lower the number better).


PS: As I am currently on holiday, I may struggle to respond to any comments on this quickly.

Mark Reckons writes for Guardian datablog

I have written a piece for the Guardian Datablog today.

It describes the details of my safe seats/MPs expenses correlation work that I have posted about here previously.

Update to MP expenses scandal and safe seats correlation

Further to my previous posts from a week and a half ago here and here which looked at the MPs who had been implicated in the expenses scandal by the Telegraph and seemed to find a correlation between the safety of the seat the MP held and the likelihood of being involved, I have gone through the data again using the latest available list from the Telegraph. The link to this is here although it changes daily and may well be out of date by the time you read this (although I have saved it locally myself). The source data about the MPs is from the 2005 election and from Keele University here.

The last time I did the work a week and a half ago only 94 MPs had been implicated. As of today there are 243 so there is more data to work with.

I have posted the results below but first I just want to again caveat all of this. I am not a statistician and I know that we don't have all the data yet. If you read through the comments on the previous threads you will see some people say that there are problems with the approach or method and that things like confounding variables are not taken account of. There are also some comments from people who think the approach is OK. I accept that my approach is fairly blunt and is by no means to final word, however I have had a number of queries from people asking if I have done an updated version of this as the previous one is now well out of date, hence the new post.

Right, here are the results:

The median number is now 287.7. It has drifted back towards the median from 262.4 previously.

The graph now looks like this:

There is still an increase in each quartile but it is less pronounced than previously. There is still quite a stark difference though between the bottom and the top quartile with the top having almost twice as many as the bottom and now with more source data to work from.

One of the objections previously was that the Telegraph were bound to focus on the most newsworthy MPs first, but almost 3 weeks on that effect must surely have dissipated as they ran out of newsworthy ones. I certainly found myself doing this update wondering who a lot of the newly added MPs are. I hadn't heard of lots of them!

Right, I think this is the last time I will post on this now. I think I have found an interesting effect albeit with potential problems. I have had a few e-mails from people asking for the source data and I will respond to these soon but with this latest data (I held off responding because it was quite out of date by the time the requests started). Anyone else who wants it, please e-mail me at the e-mail address at the top. I would like to see others take this forward and also for someone to come up with a decent way of analysing this effect once all the data had been released. I will be interested to see what other people can find about this.