The data desert grows

Why are we becoming less likely to record the ethnicity of those convicted of crimes?

Regular readers will know I have been complaining about the data desert on migration (and even ethnicity) over the last couple of years.

As well as the government blocking the release of data that used to be published (on welfare claims, tax paid etc), vital data sources like the Labour Force Survey / Annual Population Survey have collapsed.

I have helped liberate a bit of data by putting in Freedom of Information requests to HMRC and digging things out of the census and DFE datasets.

But the data desert on migration (or even ethnicity) is massive and growing - and gets in the way of a sensible conversation about migration policy.

The data desert gets bigger

This post is about another surprising way the data desert is growing.

You might have thought that following 2020 and the huge public discussion surrounding the death of George Floyd there would be improvements to statistics on criminal justice and ethnicity.

But the reverse is true - we are becoming less and less likely to record the ethnicity of people convicted of crimes.

The chart below, extracted from MOJ data, shows the proportion of those convicted for whom no ethnicity information was recorded for different types of offences.

Though the rates vary by offence group, more and more offences see the ethnicity of those convicted recorded as “unknown”.

The chart above looks at the shares of those convicted, but the same trends appear if you look at those proceeded against or those sentenced.

The same is true when we look at the level of individual offences.

A particularly salient example at the moment is child sex offences - the share where ethnicity is not recorded has grown from one in twenty, to over a quarter.

This dramatic increase in the “unknown” category effectively trashes the statistics as a whole - because if you look at the numbers where ethnicity was recorded, you can’t know if changes are being driven by real-world changes, or just differential changes to non-recording rates by ethnic group.

The meaningfulness of the data is further trashed by massive variations in recording practices between areas. Picking on the most topical example again, here is the share of child sex offences where ethnicity was not recorded by police force areas. How are we supposed to have a meaningful conversation about what’s really going on in different places, given this variation?

It is not just sex offences: similar variation is apparent if we look at all non-summary offences:

A similar trend towards the “unknown” category is visible for statistics on arrests too. The share where ethnicity is not stated has gone up - although it is still a much lower share than for those convicted of non-summary offences:

What is going on here?

The trend towards more people being registered as “unknown” is not true for the age of those convicted - the share where the age of the person was “unknown” stayed below 0.2%

There has been a (much smaller) change in the share of those convicted for whom their sex was “not known” - from 1% to 3% of all those convicted of non-summary offences. I assume this is something to do with the rise of trans identification. But it is not on the same scale as the rise in unknown ethnicity.

The official notes to the data tool say that, “The 'Unknown' category includes all others for whom ethnicity information is not available, either because they have chosen not to state their ethnicity or because no information has been recorded.”

We don’t see this massive rise in “unknown” ethnicity in other policy areas. For example, the share of school pupils of “unclassified” ethnicity has remained roughly the same, going from 1% to 1.7% over the last nine years. In other statistical sets the share unknown is falling.

It seems to me that policy - including local procedures - is clearly making a big difference. If this were just about people’s willingness to identify collapsing over time, we wouldn’t see such massive differences between similar places in the maps above.

What does the data we do have show?

One of the points I make again and again is that there is no such thing as a typical migrant - there are vast differences between groups on criminal justice just as there are on employment and wages and family structure and education and everything else.

We can see there are massive variations in arrest rates by ethnicity, but with some groups much higher than average, and some much lower.

To be clear - these are ethnicity groups, not nationality groups. Some of the groups consist mainly of people born in the UK, and others mainly of people born overseas.

I present ethnicity data here because no official data by nationality or migration status is published. It exists, but is not published.

I think that’s really poor. And sadly, given the growing share where ethnicity is not recorded, it’s getting harder to even say anything about ethnicity trends over time.

Conclusion

When you look at the maps above and see the massive inconsistency in recording practices, it’s hard to feel the criminal justice system has learned everything it should from the grooming gangs scandal about gripping the data. The maps further show why just leaving all this to the local level to investigate is not a good idea.

What is going on with these statistics is just one of many issues a national inquiry into the grooming gangs could get to the bottom of.

More generally, it often feels like government and officials are doing everything they can to stop us from having a sensible and data-driven discussion about migration.

Even the data that is released has been getting worse.

I was pleased to see Kemi Badenoch’s commitment the other day to start publishing the data the government are currently sitting on.

God knows we need some light on the subject.