Thursday, 3 November 2011

2011 England Riots: New Police Statistics and Ethnicity

Contact email:
Previous article: 2011 England Riots: Statistics of Ethnicity
Very short summary:
1. Black people are heavily overrepresented in all riots but one (Salford, 1.7% of arrests)
2. Over 50% arrested are black in all London riots (up to 82% black, 10% white in Lambeth)
3. All but two London areas (both central) with above average black population rioted
4. Poor and uneducated non-black areas did not riot or had small riots
5. Older black people are more likely to riot than whites of any age, incl. youths
6. Whites aged 10-17 are not overrepresented, blacks aged 10-17 are - heavily
7. Black Londoners are 2.87 times as unemployed as whites, but 9.65 times more likely to riot

(For the purposes of the research, I assume that people arrested for riots are representative of rioters)

The Ministry of Justice has released data on the rioters, including a detailed account of their ethnic composition. The data confirm the previous observations on the role of ethnicity in the riots. There is also information on Liverpool and Nottingham, though not on Bristol. I will visualise the new data on London, Greater Manchester, West Midlands, Nottingham, and Merseyside to show how it corresponds to the previous research.

The media still appear apathetic to the role of ethnicity - for instance, the BBC article on MoJ report is titled England rioters 'poorer, younger, less educated', as if those factors had a large predictive power. Moreover, it repeats the Ministry of Justice claim that "In some areas - such as Salford - this partially matched the ethnicity of the general population, in others - such as Nottingham - it was not representative of the general population". In fact, Salford is the only instance of the rioters matching the ethnic makeup, and it accounts for just 1.7% of all defendants. In all other cases black Britons were heavily overrepresented.

Majority of rioters are black: confirmed
46% of riot suspects across England and Wales are black, but the figure raises to 55% in the London riots, which account for 70% of the arrests:

Fig. 1. Source: MoJ report (Table 1.9) 

Since there were no official ethnic data released until late October, I had to use CCTV images to determine the approximate ethnic makeup of rioters. The new data prove the validity of this approach. Note that 16.7% of the defendants in London refused to state their ethnicity, which exceeds the number of unidentifiable people on CCTV, meaning its relative accuracy. The overall data for the largest riots are as follows (see the previous estimation for comparison):

No data are available on Bristol, but Nottingham demonstrates similarly great disproportions. Birmingham is included into West Midlands, Manchester into Greater Manchester. 33% rioters in Birmingham were white and 46% were black (17.3 times more likely to riot than whites).  67% rioters in Manchester were white and 25% were black (4.2 times more likely to riot than whites).

Fig. 3. Source: MoJ report (Table 1.13)

Overall, blacks are extremely overrepresented (13.4), Other (East Asian, Afghan et al.) are in the middle (2.1), whites and Asians are equally underrepresented (both 1.0).

Black adults rioted more than white youths
The rioters are most frequently described as youths, but black people aged 21 to 39 are by far more likely to riot than any white age group, including the most risk-prone group aged 18 to 20. The overrepresentation of black people aged 10 to 17 and 18 to 20 is possibly the most dramatic feature of the England riots. The former is somewhat understressed because the group is too broad, including very young people (aged under 14).

Fig. 4. Sources: MoJ report (Table 1.16)

The pertinence of ethnicity seems to overshadow any other aspect. Below is a comparison of ethnicity-based factors to the key factors chosen by the BBC. BBC focus on juveniles (aged 10 to 17) is interesting, but people aged 18 to 39 were more important and numerous. Moreover, the BBC claim about juveniles being unusually riot-prone does not hold water for white juveniles. 10.3% of all rioters were white juveniles, but they also form 8.0% of total UK population - no major overrepresentation is involved. Juveniles being black leads to much more striking results than overall special education or free meals.

Ethnicity more important than unemployment: London riots
Lower employment, wealth, education all correlate with the riots. But they also correlate with ethnicity. No major social group (unemployed, poor, undereducated) has such a small population as black Britons, yet is so unusually prominent in the riots. The social groups are spread across the UK, but the riots are concentrated in black areas. As shown in the previous article, many poor and unemployed areas had no riots: Northeast England, major cities in Wales, North Ireland, Scotland. In London, all major riots were predominantly black. Lambeth is particularly dramatic, with blacks constituting 82% of the arrested and whites only 10%. 

Fig. 6. Sources: MoJ report (Table 1.7)

My previous article provides a set of London riots mapped with various factors. Now that the approximate number of crimes per borough is known, it is possible to map the concentration of riots. Note that the map is best viewed with the table below, since areas like Bexley (East London) are not especially black. In some cases, the black population was low, but could compensate for this by extreme overrepresentation, such as in Wandsworth (7.8% total people are black, 5.2 crimes per 10 000, but 61% rioters were black). Names of boroughs can be found here.

Unemployment is still usually cited as the main correlate of the riots. Indeed, black people in all London boroughs are much more likely to be unemployed than whites. But their overrepresentation among rioters is even greater than their overrepresentation among the unemployed. That means unemployment and poverty do not entirely explain black rioting. Lambeth and Wandsworth did experience particularly intense rioting, but otherwise there is no correlation between inequality in unemployment and riots.

Albeit black Londoners have an increased chance of being unemployed, it is not as large as to make them the majority of all unemployed. Yet the increased chance to riot does turn them into the rioting majority. While the graph above deals with likelihood, the graph below shows total percentage of unemployed and rioters by ethnicity.

MoJ table 4.5 shows that the unemployed are unusually frequent among white rioters, but not among black rioters. In fact, the proportion of JSA claimants among black rioters (~20%)  is smaller than among white rioters (~25%). Still, even unemployed whites are underrepresented in the riots in comparison to average blacks.

London boroughs and potential riot factors
There is a variety of factors that have correlation with the riots, including ethnicity, education, poverty, and others. Below is a table summarising some of the key facts about the riots. Brent seems to be an exception from all correlates, since the police were deployed there preventively. There was also little or no rioting in central areas like Westminster, Hammersmith and Fulham, and Kensington and Chelsea. No ethnic data on Ealing and Tower Hamlets riots are available. [CSV version]
Fig. 10. Sources: see below

Socioeconomic factors prove meaningless to the riots when they are not a correlate of ethnicity. Using police CCTV sources proved to be a viable method of measuring the ethnic makeup of rioters. Consequently, the conclusions of the previous article remain unaltered. Even if 'chavs' formed a part of the rioters, they were a clear minority. The media should focus on ethnicity when discussing the riots, rather than obfuscating this factor by citing Salford, which is the only instance of black Britons not being overrepresented. Black Britons are dramatically more likely to take part in the riots than their compatriots of other ethnicities, including whites of any social status.

Riots data
Ministry of Justice data from 'Statistical bulletin on the public disorder of 6th-9th August 2011':
BBC 'England rioters 'poorer, younger, less educated'':

General data
ONS 'Region and Country Profiles - Key Statistics Tables - October 2011':
ONS 'Population Estimates by Ethnic Group Mid-2009 (experimental) - Current Estimates':
Maps of ethnicity, education, class and other factors in London:

Monday, 15 August 2011

2011 England Riots: Statistics of Ethnicity

Contact email:
Update, 20th August: sample size increased from 48 to 173 for London, from 98 to 247 for Birmingham, from 37 to 117 for Manchester, from 29 to 32 for Bristol. The ethnic pattern did not experience significant change (see below).

Very short summary:
1. Over 50% rioters are black,  below 30% are white;
2. Black people are over 10 times more likely to participate in riots than whites;
3. High unemployment in London -> riots?
4. Black areas -> riots?
5. Low social class != riots, low training !=riots;
6. Black areas = high unemployment;
7. High unemployment + black areas = riots;
8. High unemployment + non-black areas = no riots;

9. Average or low unemployment + black areas (rare) = riots, but fewer

1. CCTV in 2; 2. CCTV, data in 2; 3. map in 1D; 4. map in 1A; 5. maps in 1B, 1C; 
6. map in 1E-; 7. maps in 3E; 8. maps in 1E, map in 3B; 9. map in 1E-2


The media seem to focus exclusively on the social class of the August rioters, while claiming the absence of a large ethnocultural undercurrent. There is an excessive concentration on 'chavs' or 'yobs', who are understood to be white. However, chavs form but a minority of rioters and looters, while the disproportionate majority is Black African and Afro-Caribbean. This will be illustrated below using authoritative sources, in three arguments.

1. Distribution of ethnicity and other factors in London. I will show the distribution of riots and addresses of the arrested on the following maps of London: a. Black African and Afro-Caribbean population; b. Education, skills and training; c. Social class; d. Unemployment; e. Combined map of ethnicity and unemployment.

No significant correlation was found on B and C. D and especially A show a stronger correlation. However, that may be because of a large tie between A and D. The best result is with E, which predicts riots with formidable precision.

2. Ethnicity of riot suspects on police CCTV by percentage. I will examine all available photographs on the official police CCTV sources to determine the ethnic make-up of rioters. I will then compare the proportions of ethnicities to the total population.

Photographs show that most (~50%) rioters were black even in areas with a small black population, such as Bristol. There is a vast overrepresentation of Black British among rioters in comparison to their total percentage in a city. Their likelihood of participating in riots in comparison to whites is more than 10 times higher. Ethnicities are marked to ensure verifiability.

3. Black population and chance of riots in London boroughs and large settlements. I will substantiate the preceding arguments with additional data showing: a. London boroughs by black population and riots; b. 20 largest settlements, their black population and riots; c. Unemployment and riots alongside the list of the largest settlements.

This shows that black population is greatly indicative of the chances of riots both in London boroughs and in settlements. The total size of the settlement does seem to influence the chance of riots, as in the case of Nottingham or Glasgow. Albeit in London unemployment was found to have a fairly strong correlation with the riots, that is not the case with the UK. Many non-rioting areas had very high unemployment (parts of Scotland, NE England, S. Wales, N. Ireland, etc.), but all of them had a minor black population.

1. Race over unemployment over education and class
Maps of riots and addresseses of rioters, taken from The Guardian and The Daily Telegraph as of 13th August, are superimposed on maps from LondonProfiler by UCL.

a. Black Africans and Afro-Carribeans. No riots occurred in areas that were far away from large concentrations of black population. Most of the areas with a large black population witnessed the heaviest riots that included multiple arsons.

b. Not the underskilled. The primary characteristic of the underclass, specifically 'chavs', is the absence of education, skills and training. This map shows a clear lack of any correlation between riots and these factors. Large areas in the east and especially the west of London such as Hillingdon are undereducated, but saw no riots. The south central areas are more skilled, but had riots.

c. Not the lower classes. 'Chavs' are generally associated with the lower class - in fact, some believe the concept itself to be classist. Yet, as with education, there are large lower class areas such as Havering that saw no riots. While there are few riots in the more well-off areas, some of the most devastated large areas, such as Croydon, were not lower class. 

d. Unemployed, perhaps. The underclass is inevitably unemployed. The east, north-west and west extremes show reasonable employment, thus excluding the most extravagant outliers in the above cases. The south still does not fit well, however. Unemployment appears to be the best purely socioeconomic correlate of riots.

e. Reason: black and unemployed, or just black. Areas with a high black population and high unemployment (light green) dominate all other black areas (cyan and light gold). Almost all riot addresses fall within the vicinity of light green zones. Though black areas with normal employment are a minority, they also are covered by the rioters, if to a smaller degree. White/cyan areas experienced riots.

Most areas with a low black population have lower unemployment. However, it is important to note that isolated dark green areas - non-black, but with high unemployment - have no riots. A particularly vivid example of it is Tower Hamlets, whose poor non-black populace separates two large rioting zones. This map explains the absense of riots in the extreme east and their presence in the south.

It is impossible to predict the riots using the class and the skill maps, and even unemployment is highly imprecise. The ethnic map is the best predictor of riots.

2. CCTV shows the majority of rioters are black
Police CCTV pictures of suspects in the 3 largest riots: London (1,051 arrested), Birmingham (218 arrested, 445 for West Midlands), Manchester (47 arrested, 174 for Greater Manchester). Bristol (24 arrested) is an example of a riot in a smaller settlement.

The size of the samples is adequate: 173 for London, 218 for Birmingham, 117 for Manchester, 32 for Bristol. The ratio of ethnicities remains similar in images published by the police after 13th August. Nottinghamshire (106 arrested) and Merseyside (77 arrested) CCTV images are not clear enough to be assessed, but seem to show a large black contingent as well.

The ethnic map shows impressive predictive power, and it is possible to confirm this argument empirically. In London, Birmingham and Bristol whites form a minority of rioters: respectively 27%, 30%, and 25%. Only in Manchester do they constitute the majority - 45%. Manchester also has the largest percentage of non-white, non-black people participating. Black people are not only overrepresented, but simply outnumber everyone else by a large degree: 50%, 45%, and 47% - though only 28% in Manchester. Even if all white rioters were from the underclass, which is unlikely, 'chav' rioters would still be far fewer than black rioters. Indeed, it may be suggested that the 'chavs' joined the ongoing lootings, but did not instigate them, as in this video from Manchester: a black gang breaks into the shop, then mixed passers-by run in.

This table sums up the collections of all CCTV pictures available from police departments as of 20th August 2011.

CCTV images with race of the suspects marked:
(full-sized police pictures:
(full-sized police pictures:
(full-sized police pictures:
Bristol: (AF and AC seem to be the same person)
(full-sized police pictures:
All on one image:

London example (thumbnail):

3. UK and London areas with a small black population did not riot
An analysis of cities and London boroughs by black population and chance of rioting. A map of UK unemployment to illustrate the lack of correlation.

 a. Non-black London boroughs did not riot. This graph demonstrates that the likelihood of rioting increases dramatically with the percentage of black population. Rioting in Brent was expected, but was prevented by the community and large police presence. Rioting in Bromley appears mild or non-existent, while in Woolwich (Greenwich) it was underreported in the media. Westminster reflects the incident at Oxford Circus, which involved a mob of 50.

b. Non-black settlements UK-wide did not riot. The riot threshold of black population is lower for large settlements than for boroughs, as it means they include smaller areas with large black population (e.g., London has Haringey). Consequently, Nottingham's 4.7% matters. One of the theories was that all large cities would experience riots because they follow similar patterns, but that is not the case. Large cities with a small black population did not riot.

c. Highly unemployed non-black settlements did not riot. (see the figure for B) Unemployment was shown to be the single best socioeconomic correlate of riots in London, better than education or class, but that was because it was an aspect of black communities. The UK has large areas with high unemployment that have a small black population. There were no disorders there.

Both theoretical and empirical evidence shows that it is improper to disconnect the August 2011 riots from race and ethnic culture. It is, likewise, inappropriate to associate it with the uneducated, low-skilled white lower class ('chavs'). It was severely underrepresented in the riots, and did not produce riots in numerous poor non-black areas. None of the socioeconomic factors have the same predicting power as black ethnicity in this case. In fact, they are largely meaningless to the riots when not a correlate of black population.

Why the black community played such a prominent role in the riots may be disputed. But it is clear that role of the black community was greater than that of any other group, and it should be accordingly regarded in the media.

Maps of ethnicity, education, class and other factors in London:
Suspected addresses map:
Riots map:
UK map of unemployment by date:

Police CCTV images (updated 13th August)
Metropolitan Police suspects:
West Midlands Police suspects:
Greater Manchester Police suspects:
Avon & Somerset Constebulary suspects:

Data on ethnicity
London ethnic groups by borough:
Other cities: