2021v1, Saturday: What's measured matters.

When measures become targets, they're useless as measures. But when something isn't measured at all, it's invisible...

Short thought: a rare day off yesterday with spouse. So nothing written. Catching up today, I spotted one of the terrifying number of tabs I currently have open and awaiting attention, which dealt with Goodhart’s law. And that took my mind to fraud.

Yes, I know: for me, that’s a fairly short leap from practically anywhere. But bear with me for a moment. Goodhart’s Law, named after economist Charles Goodhart, holds that “Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.” Or, more simply and pithily: measures become useless as measures once you start using them as targets.

Why is this? Because measures are diagnostic: things that help you understand a situation. They’re therefore only any use as long as they’re objective. And as soon as people start being graded according to that measure, it will be gamed. Think hospital waiting lists, or A&E waiting times, or paying teachers (or funding schools) by kids’ school grades. Or stack ranking in workplaces.

Which means I have mixed feelings about this story in the Times (£, sorry). It talks of six metrics by which police forces will be ranked. None of them, predictably, explicitly deals with fraud - the single largest crime problem the UK faces, yet the one with probably the least amount of focused resource and attention.

This brings back memories. In 2007-8 I spent a short amount of time working with people at what was then the (short-lived) National Fraud Strategic Authority. One thing that passed across the NFSA’s desk was a proposed set of police performance indicates. There were 147 of them, I recall - and not a single one dealt with fraud. Most were obviously chosen because they were things that were easy to count, rather than things that would genuinely make a difference in the effectiveness of policing.

So on the one hand, having just six - and high-level at that (homicide, serious violence, drug supply, neighbourhood violence, cybercrime and “victim satisfaction”) - is a step up. As is a Home Office source’s comment that “I wouldn’t classify them as targets.”

So not a Goodhart problem, then?

Hardly. Because the next sentence is: “It’s about tracking progress — we’re giving forces extra officers and now we want to see outcomes.”

So it’s directly a Goodhart problem. As has been the case before, forces will game their resource spend to make sure the Home Office is happy.

And fraud, as ever, will be forgotten. As will its millions of victims. Because, to be cynical, their problems cost too much to fix.

(Yes, yes. “Cybercrime” could be interpreted to include a lot of fraud. But it’s not the same thing. And how is “dealing with cybercrime” to be measured? I have no idea, and I’m pretty sure the Home Office doesn’t either.)


Someone is right on the internet: This is the third piece in a row in which I’ve mentioned the Horizon scandal. I won’t apologise. This is disgraceful, and it deserves a lot of noise.

So today I’ll simply point to someone who, if we had Pulitzers in this country, would deserve one for his coverage of this: Nick Wallis. This piece by him in Private Eye describes the scandal in great detail. Read every word. And then get very angry. And then decide how the issues that arise herein might arise elsewhere - and start thinking about who should be held to account for not making this a priority.

One quick sideline: I should have mentioned when writing about Horizon that the Post Office - shortly before Mr Justice Fraser tore Horizon to shreds - agreed to a £58m settlement of many postmasters’ claims. You can make your own mind up as to whether sociopathy on this level can be satisfied by that, or whether individuals should also be held to account.