Even if AI does help, don't get your hopes up about making the state more productive

The UK government employs more than 6 million people. Paying them will cost about £300bn this year. The total cost of providing public services - the NHS, prisons, the police, schools, the army, and so on - is more than double that. As taxes rise (again), voters might reasonably ask whether their money is being spent efficiently. To get a grip on that question, we need to measure what the government gets in return for all that spending, which means we need to measure what those 6 million public workers are producing. This is harder than it might sound.

- Ben Zaranko

Let's start with a relatively simple example. How do you measure the output of a surgeon working for the NHS? You might start by counting how many operations they carry out. Perhaps you'd make some sort of adjustment for the complexity of each operation, or their success rate in keeping patients alive. Fiddly, perhaps, but doable.

What about a police officer? Do you measure their output as the number of arrests made? Or should it be the number of crimes they prevent from happening in the first place by patrolling the streets? What about a soldier? Or a judge? It can become very difficult very quickly.

These are the sorts of conundrums that the folks at the Office for National Statistics (ONS) have to confront when building their estimates of public service productivity - the latest set was published last week. Because it's an unusual exercise, requiring all sorts of assumptions, we should interpret the figures with caution. They nonetheless provide some interesting insights.

There was some good news for sceptical taxpayers: the ONS has revised up its previous estimates, and now thinks that public service productivity increased for the fourth consecutive year in 2024.