Gavin Hales
@gmhales.bsky.social
Researching/discussing policing, crime and the criminal justice system. Senior Associate Fellow @policefoundationuk.bsky.social but my own views.
Based in London, UK.
Based in London, UK.
Reposted by Gavin Hales
I'm tending towards the view that gaps in police self-defined ethnicity data need not limit the analysis of ethnic disproportionality, as long as ethnic appearance data are well completed. That's because SDE and EA generally align well at the record-level.
If you cross-tabulate the self-defined ethnicity (SDE) and ethnic appearance (EA) variables, you'll see that only 3.3% of s60 records have no EA recorded, and the gaps in SDE are almost all 'not stated', which means SDE was requested but not provided.
December 8, 2025 at 7:35 PM
I'm tending towards the view that gaps in police self-defined ethnicity data need not limit the analysis of ethnic disproportionality, as long as ethnic appearance data are well completed. That's because SDE and EA generally align well at the record-level.
I'm tending towards the view that gaps in police self-defined ethnicity data need not limit the analysis of ethnic disproportionality, as long as ethnic appearance data are well completed. That's because SDE and EA generally align well at the record-level.
If you cross-tabulate the self-defined ethnicity (SDE) and ethnic appearance (EA) variables, you'll see that only 3.3% of s60 records have no EA recorded, and the gaps in SDE are almost all 'not stated', which means SDE was requested but not provided.
December 8, 2025 at 7:35 PM
I'm tending towards the view that gaps in police self-defined ethnicity data need not limit the analysis of ethnic disproportionality, as long as ethnic appearance data are well completed. That's because SDE and EA generally align well at the record-level.
Reposted by Gavin Hales
Around 4/5 s60 stops over the last 2 years (83%) took place in the month of August, and 90% of those took place in Kensington & Chelsea or Westminster. That points to the use of s60 primarily being concerned with Carnival.
December 8, 2025 at 6:02 PM
Around 4/5 s60 stops over the last 2 years (83%) took place in the month of August, and 90% of those took place in Kensington & Chelsea or Westminster. That points to the use of s60 primarily being concerned with Carnival.
Reposted by Gavin Hales
As I noted here, there is a generally high degree of agreement between SDE and EA data which suggests that at least for broad ethnic groups EA can be used fairly reliably in combination with SDE or even on its own to examine Qs like disproportionality. bsky.app/profile/gmha...
December 8, 2025 at 5:41 PM
As I noted here, there is a generally high degree of agreement between SDE and EA data which suggests that at least for broad ethnic groups EA can be used fairly reliably in combination with SDE or even on its own to examine Qs like disproportionality. bsky.app/profile/gmha...
Reposted by Gavin Hales
If you cross-tabulate the self-defined ethnicity (SDE) and ethnic appearance (EA) variables, you'll see that only 3.3% of s60 records have no EA recorded, and the gaps in SDE are almost all 'not stated', which means SDE was requested but not provided.
December 8, 2025 at 5:40 PM
If you cross-tabulate the self-defined ethnicity (SDE) and ethnic appearance (EA) variables, you'll see that only 3.3% of s60 records have no EA recorded, and the gaps in SDE are almost all 'not stated', which means SDE was requested but not provided.
Reposted by Gavin Hales
I wasn't aware until just now that the Home Office recently published the findings of the 2023 Police Activity Survey www.gov.uk/government/p...
December 7, 2025 at 10:54 AM
I wasn't aware until just now that the Home Office recently published the findings of the 2023 Police Activity Survey www.gov.uk/government/p...
I wasn't aware until just now that the Home Office recently published the findings of the 2023 Police Activity Survey www.gov.uk/government/p...
December 7, 2025 at 10:54 AM
I wasn't aware until just now that the Home Office recently published the findings of the 2023 Police Activity Survey www.gov.uk/government/p...
This is well worth your time
Anyway it's an honour the Guardian has licensed a version of a London Centric piece as today's Long Read. The pics are amazing. The fee will pay for me to hire a freelancer for a week. But mainly it means I'm freed from the alpaca... which is replaced by a snail.
www.theguardian.com/news/ng-inte...
www.theguardian.com/news/ng-inte...
The snail farm don: is this the most brazen tax avoidance scheme of all time?
The long read: Terry Ball – renowned shoe salesman, friend to former mafiosi – has resolved to spend his last years finding ways to cheat authorities he feels have cheated him. His greatest ruse? A ta...
www.theguardian.com
December 6, 2025 at 3:51 PM
This is well worth your time
"The future of autonomous vehicles and policing" www.policeforum.org/trending6dec25
New PERF report helps police prepare for autonomous vehicles
www.policeforum.org
December 6, 2025 at 3:46 PM
"The future of autonomous vehicles and policing" www.policeforum.org/trending6dec25
@syalrajeev.bsky.social your article on police retrospective facial recognition has been illustrated with a photo of a live facial recognition operation. The latter tech has been shown to not be biased when used with the correct settings. www.theguardian.com/technology/2...
Home Office admits facial recognition tech issue with black and Asian subjects
Calls for review after technology found to return more false positives for ‘some demographic groups’ on certain settings
www.theguardian.com
December 6, 2025 at 3:05 PM
@syalrajeev.bsky.social your article on police retrospective facial recognition has been illustrated with a photo of a live facial recognition operation. The latter tech has been shown to not be biased when used with the correct settings. www.theguardian.com/technology/2...
New evaluations of two retrospective #facialrecognition algorithms www.gov.uk/government/p...
Facial recognition technology tests: National Physical Laboratory
Independent testing of facial recognition algorithms commissioned by the Home Office.
www.gov.uk
December 5, 2025 at 8:19 PM
New evaluations of two retrospective #facialrecognition algorithms www.gov.uk/government/p...
Reposted by Gavin Hales
My latest insights on stop and search. #7 is probably the most interesting.
I've been having a rummage in the most recent Home Office stop and search data, published earlier this month. The first thing to note is that Merseyside Police are out on their own in terms of rates per 1,000 population. #stopsearch.
1/
1/
November 30, 2025 at 8:35 PM
My latest insights on stop and search. #7 is probably the most interesting.
My latest insights on stop and search. #7 is probably the most interesting.
I've been having a rummage in the most recent Home Office stop and search data, published earlier this month. The first thing to note is that Merseyside Police are out on their own in terms of rates per 1,000 population. #stopsearch.
1/
1/
November 30, 2025 at 8:35 PM
My latest insights on stop and search. #7 is probably the most interesting.
Reposted by Gavin Hales
The implications of that include (1) SDE & EA can be used in combination to minimise gaps in ethnicity data (i.e. use EA to fill gaps in SDE); and (2) a wider Black grouping should include people with White/Black heritage in the Census denominator (ditto White/Asian for consistency).
November 29, 2025 at 3:52 PM
The implications of that include (1) SDE & EA can be used in combination to minimise gaps in ethnicity data (i.e. use EA to fill gaps in SDE); and (2) a wider Black grouping should include people with White/Black heritage in the Census denominator (ditto White/Asian for consistency).
Reposted by Gavin Hales
Importantly, there is generally a high degree of alignment between SDE and officer-perceived ethnic appearance, especially for the main ethnic groups. Notably, at least 4/5 people who self-identify as White/Black mixed heritage are recorded as having a Black appearance.
November 29, 2025 at 3:52 PM
Importantly, there is generally a high degree of alignment between SDE and officer-perceived ethnic appearance, especially for the main ethnic groups. Notably, at least 4/5 people who self-identify as White/Black mixed heritage are recorded as having a Black appearance.
Reposted by Gavin Hales
It may be of interest that in stop and search data, Black people are more likely than White and Asian to have their self-defined ethnicity recorded as 'not stated'. That means that if SDE is used for calculating disproportionality, Black disproportionality will be under-stated.
November 29, 2025 at 3:52 PM
It may be of interest that in stop and search data, Black people are more likely than White and Asian to have their self-defined ethnicity recorded as 'not stated'. That means that if SDE is used for calculating disproportionality, Black disproportionality will be under-stated.
Reposted by Gavin Hales
Note the way only a very small % of stops in Merseyside are of non-White people, reflecting the demographics of the resident population. I'm sure it's why there is little interest in (or controversy about) Merseyside Police's use of S&S.
10/
10/
November 29, 2025 at 2:58 PM
Note the way only a very small % of stops in Merseyside are of non-White people, reflecting the demographics of the resident population. I'm sure it's why there is little interest in (or controversy about) Merseyside Police's use of S&S.
10/
10/
Reposted by Gavin Hales
This is also interesting: there's quite a consistent pattern where forces with a higher % of non-White stops have a higher % of self-defined ethnicity (SDE) recorded as 'not stated'. West Mids, W Yorks and CoLP appear to be outliers.
9/
9/
November 29, 2025 at 2:58 PM
This is also interesting: there's quite a consistent pattern where forces with a higher % of non-White stops have a higher % of self-defined ethnicity (SDE) recorded as 'not stated'. West Mids, W Yorks and CoLP appear to be outliers.
9/
9/
Reposted by Gavin Hales
I think that suggests any efforts to address disproportionality that focus on individual officer decision making are unlikely to have more than a marginal impact.
8/
8/
November 29, 2025 at 1:49 PM
I think that suggests any efforts to address disproportionality that focus on individual officer decision making are unlikely to have more than a marginal impact.
8/
8/
Reposted by Gavin Hales
One addition, which I think is very important - comparing stop and search criminality detected/positive outcome rates by subject ethnicity and police force.
Very little difference at force-level suggests the bar for suspicion is applied quite consistently.
7/
Very little difference at force-level suggests the bar for suspicion is applied quite consistently.
7/
November 29, 2025 at 1:25 PM
One addition, which I think is very important - comparing stop and search criminality detected/positive outcome rates by subject ethnicity and police force.
Very little difference at force-level suggests the bar for suspicion is applied quite consistently.
7/
Very little difference at force-level suggests the bar for suspicion is applied quite consistently.
7/
Reposted by Gavin Hales
And looking at the data for 2024/25, it is notable that the Met is the only force to have both above average rates of stop and search (per 1,000) and resulting arrest rates. Lincs' arrest rate is conspicuously low.
5/
5/
November 29, 2025 at 12:57 PM
And looking at the data for 2024/25, it is notable that the Met is the only force to have both above average rates of stop and search (per 1,000) and resulting arrest rates. Lincs' arrest rate is conspicuously low.
5/
5/
Reposted by Gavin Hales
Looking at arrest rates by police force over time, I was interested to see that arrest rates in the Met are now higher than those in comparative forces. That was also the case around 2013-16, when S&S volumes/rates in the MPS were also low by historic standards.
4/
4/
November 29, 2025 at 12:57 PM
Looking at arrest rates by police force over time, I was interested to see that arrest rates in the Met are now higher than those in comparative forces. That was also the case around 2013-16, when S&S volumes/rates in the MPS were also low by historic standards.
4/
4/
Reposted by Gavin Hales
I need to do some more digging on why that might be. One possibility is a shift in the offence/powers context, e.g. reflecting the rise in shoplifting in recent years. Another is a change in emphasis on arrests rather than other positive outcomes where offences are found.
3/
3/
November 29, 2025 at 12:57 PM
I need to do some more digging on why that might be. One possibility is a shift in the offence/powers context, e.g. reflecting the rise in shoplifting in recent years. Another is a change in emphasis on arrests rather than other positive outcomes where offences are found.
3/
3/
Reposted by Gavin Hales
Next, I was interested to see what has happened to S&S arrest rates. It's interesting to see that arrest rates have increased in recent years, despite overall S&S volumes not changing much - which is not consistent with the long-term picture.
2/
2/
November 29, 2025 at 12:57 PM
Next, I was interested to see what has happened to S&S arrest rates. It's interesting to see that arrest rates have increased in recent years, despite overall S&S volumes not changing much - which is not consistent with the long-term picture.
2/
2/