Thread by @chrishanretty

Towns are sexy in UK polisci, aren't they? Let's talk a little bit about the super-sexy Towns Fund gov.uk/government/pub… (1/n)

As @estwebber and @georgegrylls reported this morning, the Perm. Sec at MHCLG gave evidence to the Public Accounts Committee this week thetimes.co.uk/article/robert… (2/n)

Jeremy Pocklington said that ministers had ignored civil service advice concerning the £3.9bn Towns Fund, and had instead applied [ahem!] "their own qualitative assessment" (3/n)

The Towns Fund has previously been the subject of a National Audit Office report (nao.org.uk/report/review-…), which sets out the very detailed scores that civil servants gave to different towns and rank-ordered recommendations (4/n)

(That data's in a horrible, horrible PDF, but I tidied it and put it here: docs.google.com/spreadsheets/d…) (5/n)

Briefly: 541 towns were considered, civil servants made recommendations as to how many towns should be picked in each region, & ranked towns within regions. Each region had a list of high-priority towns, but ministers had discretion over the 60 remaining towns (6/n)

Ministers could have followed the detailed civil service rankings. Here's what the data would have looked like if that had happened. 38 out of 60 funded areas are wholly or partly in Cons held seats (7/n)

But ministers exercised their discretion, and so we get this pattern, with 53 of 61 funded areas in Con-held seats (8/n)

(Technical note: a town (=built up area or subdivision) is in a Conservative seat if the town area overlaps, in whole or in part, with a Westminster seat. Some towns straddle Westminster seats) (9/n)

This association is robust, and highly significant across several different specifications. Here's a logistic regression table for those of you who like that (10/n)

Indeed, because I know how to have a good time, I went so far as to estimate all 4095 models possible using the variables in the NAO data. In *all* models the "Conservative seat" variable is significant (11/n)

If you want, you can estimate a fancier model, where a smoothed spline connects (average) Conservative majority in the seat(s) to funding decisions, controlling for rank within region. For towns on the cut-off, you get a peak at a Conservative majority of +5 to +7% (12/n)

In my judgement, no reasonable analyst of the NAO data could fail to conclude that Conservative areas were advantaged by the process (13/n)

What does that mean politically and legally? (14/n)

Politically, I am not sure how much it means. MPs can and ought to ask more questions about the "qualitative assessment" carried out by ministers... (15/n)

... but the recent attention to standards in public life suggests (cspl.blog.gov.uk/2020/09/22/why…, that the (political/norm-based) maintenance of those standards is slipping (16/n)

Legally, is there a case that Jenrick and Berry wilfully misconducted themselves to such a degree as to amount to an abuse of the public's trust, without reasonable excuse or justification? (17/n)

I don't know enough about the status of the duty to secure value for money in public expenditure, and I don't know what kind of "reasonable excuse or justification" Jenrick and Berry could offer for this pattern (18/n)

But I challenge anyone to look at the data or my code (gist.github.com/chrishanretty/…) and conclude that partisan advantage wasn't a factor here (19/19)

(p.s. I have left out the part of the code that involved wrangling with shapefiles; DM me if you need it)