Trends in associations between ethnicbackground and being sentenced to custodyfor young offenders in England and Walesbetween 2009 and 2016Harvinder Singh Bhamra and Noah Uhrig, Ministry of JusticeAnalytical ServicesPublished 30 November 2017Amended 1 December 2017 to clarify technical presentation of results in Table 4

IntroductionOverall volumes of young people, aged 10-17 years, in custody have dropped 70% since the 2007peak in youth proven offending from 2,909 to 877 in 20161. Falls have been seen across all ethnicgroups, however the volumes of Black, Asian, and other minority ethnic (BAME) young people havedecreased at a slower rate of 47% from 726 in 2007 to 385 in 2016. Therefore, the proportion ofBAME young people in custody has increased from 25% to 44% between 2007 and 2016, althoughthe actual number fell.This study builds on descriptive analysis commissioned by the Lammy Review2 that exploredcustodial outcomes for BAME young people3. It aims to show whether BAME young people were anymore or less likely than their White counterparts to be sentenced to custody and whether thisassociation varied over time. In so doing, it aims to address the role of custodial sentencing as apossible driver of historical change in volumes of young White and BAME offenders aged 10-17 incustody.Unlike previous work3, this analysis controls for several factors that can influence sentencingdecisions (i.e. offender age, gender, offence group and criminal history) allowing an assessment tobe made between ethnicity and custodial sentencing under similar circumstances. Not all factorswere, however, considered in the analysis (e.g. specific offence committed, offender plea or anyassociated mitigating or aggravating issues) and this must be taken into account when interpretingfindings.Summary of key resultsThe analysis of data from 2009 to 2016 found: Compared to White young people, their Black, Asian and Mixed ethnic equivalents were morelikely to be sentenced to custody in 2016 (by 71%, 86%, and 35% respectively); Black young people were consistently more likely than their White counterparts to be sentencedto custody over time and for a range of specific offence groups, though the magnitude ofdisproportionality remained constant over time; and The pattern of association between ethnic background and custodial sentencing for other ethnicgroups was not consistent over time.While demonstrating consistent disproportionality in the likelihood of custody for Black and Asianyoung people, this alone does not appear to account for the change in BAME custodial volumes orproportions. Other factors such as custodial sentence lengths and upstream justice system driverssuch as police activity, charging, prosecutions, offender plea and convictions may also play a role inaccounting for these historical changes.1Youth Justice Board (2017). Youth Custody Report: September uploads/attachment 7.xls2Ministry of Justice (2017). Lammy Review: Final report. An independent review into the treatment of, andoutcomes for, Black, Asian and Minority Ethnic individuals in the Criminal Justice lammy-review-final-report3 Ministry of Justice, Analytical Services (2017). Exploratory analysis of 10-17 year olds in the youth secureestate by black and other minority ethnic groups. ps2

ApproachThe analysis sample was constructed using data drawn from government management informationsources. An extract was obtained from the Police National Computer (PNC) recording all sentencesgiven to each offender aged 10 to 17 convicted or cautioned4 along with their criminal history. Selfidentified ethnicity was merged onto the PNC data from the Ministry of Justice Court ProceedingsDatabase. There were 306,111 records from the PNC. Merging on Court Proceedings data usingSurname, First Initial, Date of Birth and Court Appearance (sentencing) date yielded a match of 77%on these criteria. A bias analysis was then conducted by comparing the distribution of the selfidentified ethnicity across both matched and unmatched data which showed a comparable pattern inthe distribution of the observed variable. Any biasing effect from unmatched data was considerednegligible.Data were obtained for the years 2007, 2009, 2011, 2013, 2015 and 2016. The 2016 data allowedthe analysis of the most recent published calendar year, while the two-year intervals between 2007and 2015 allowed for comparing the effects to previous years5. While the youth custodial populationpeaked between 2007 and 2008, the self-identified ethnicity data recorded in youth and magistrates’courts was not of sufficient quality or completeness for inclusion in this analysis until the introductionof the LIBRA case management system in 2009. Therefore, the analyses presented here begins in2009. Missing ethnicity for any matched individual with records from multiple years was directlycoded from any year where it was not missing.Multivariate logistic regression models were built to account for the associations between selfidentified ethnicity, sex, age, offence group, previous criminal history, and being sentenced tocustody, compared with being sentenced to another type of disposal including community orders,fines, out of court disposals, suspended sentence orders, or discharge. This allowed the associationsbetween ethnicity and custody rates to be assessed under similar criminal circumstances.Additional logistic models, with interactions between ethnicity and year, were built. These allowed forthe examination of whether the probability of being sentenced to custody varied over time withinethnic group.The ethnicity grouping in this analysis is the five point classification – White, Black, Asian, Mixed, andChinese or Other. This is derived from the 16-category classification used within the 2001 census,which remains the standard in courts management information systems. This categorisation wasused to remain consistent with the ethnicity classification used in other Ministry of Justice statisticalpublications.While a number of associations were observed between the likelihood of custodial sentencing and arange of offence / offender characteristics, it is important to note that the current analysis did not takeinto account all factors which can be involved in making sentencing decisions. For example, theanalysis controlled for seventeen broad offence groups, allowing for comparisons between offendersfrom different ethnic backgrounds within these groups6. However, there remains a range of offence4Where an offender was convicted of more than one offence at the same court hearing, the offence attractingthe most severe sentence was considered to be the main offence.5 Two year intervals were selected to provide a snapshot of the time trend.6 These aggregated offence groups were developed drawing on the Police National Legal Database (PNLD)classification that underpins the Youth Offender Group Reconviction Scale (YOGRS).3

seriousness within the offence groups7. For example, murder and common assault are both violenceagainst the person offences. To partially control for this issue, the analysis separates indictable andsummary violence against the person offences, though the range of seriousness in other offencegroupings are not included in the modelling and there remains a range of seriousness even withinindictable and summary offending.The offence groups used in this analysis were used to summarise the range of offences youngpeople commit. They are not directly comparable with those recorded in the court appearance data,which are used when custody rates are published in the Criminal Justice Statistics Quarterly and forother related statistics throughout Race and the Criminal Justice System 2016.LimitationsIn the merged dataset, records were dropped where the PNC data on sex8, or offence group wasmissing, or where the self-identified ethnicity from the Courts Proceeding data was not matched witha PNC record9. Findings are generalisable only to young people, aged 10-17, sentenced atmagistrates’ courts or the Crown Court in England and Wales in the years investigated.Furthermore, while the logistic regression models allowed the associations between ethnicity andcustodial sentences to be examined under similar criminal circumstances, these models cannot takeinto account all factors involved in sentencing. For example, they do not include the specific offencecommitted, offender pleas10, or any associated mitigating and aggravating factors11, and this needs tobe taken into account when interpreting results.Descriptive StatisticsWithout controlling for any other variables or conditions (i.e. unadjusted), custody rates variedaccording to a number of offender characteristics. Table 1 shows the characteristics of sentencedoffenders and the unadjusted custody rate by key characteristics across the analysed years.In 2016, 74% of young offenders self-identified as White, while 14% identified as Black, 5% as Asian,6% as Mixed, and 1% as Chinese or Other ethnicity. The percentage of offenders identifying as7In Statistics on Race and the Criminal Justice System 2016, patterns of offending and outcomes varied byethnicity at the level of specific offences. Variations in the custody rates could therefore potentially reflectvariations in the mix of offences across ethnicity.8 ‘Sex’ is considered here to refer to whether someone is male or female based on their physiology, with‘gender’ representing a social construct or sense of self that takes a wider range of forms. Sex rather thangender is used in this paper because this classification better reflects how individuals are generally reported indata related to the Criminal Justice System.9 The proportion of records dropped for this reason in each year is as follows: 2009 18%, 2011 17%, 2013 16%,2015 16%, 2016 20%. Additional checks for bias were conducted by comparing the distributions of offendercharacteristics across samples with and without missing ethnicity, sex, and offence group. This showedcomparable patterns in the distributions of the observed variables, thus any biasing effect from missing datawas judged to be negligible. For this reason, findings reported in this paper are robust to this missing data.10 Plea data is largely not included in data extracts obtained from youth and magistrate court records, where themajority of youth cases are heard.11 The Youth Justice Board collects data on offending-related risks and needs; however, it is not yet technicallypossible to link to this information to Ministry of Justice data.4

White decreased over the observed years from 84% in 2009. The percentage of Black and Asianyoung offenders increased over the observed years (7% to 14% and 3% to 5% respectively).The overall rate of immediate custodial sentencing (custody rate) for the cohort of offenders in 2016was 9%12. The remaining 91% received another youth justice disposal such as a caution, fine,community sentence, or discharge. In the data across the observed years, custody rates remainedbetween 8-11%, while offence volumes decreased from 2009 onwards.Without controlling for any other variables or conditions, higher custody rates were observed forBAME young people compared with their White counterparts across all years. The highest rate ofcustody in 2016 was observed for young people who were Black (13%), followed by Asian (12%).The lowest was observed for those self-reporting as White (8%). Over the same years, custody ratesfor Black young people remained between 11-15%, while custody rates for their Asian young peopleremained between 10-13%. Custody rates increased between 2009 and 2011 for all ethnic groups,apart from Chinese or Other. Among the White group custody rates remained relatively stablebetween 8-10%.Young people aged 10 to 14 years made up 17% of the offenders in the data for 2016, with theremaining 83% being aged 15 to 17 years. The custody rate for those aged 10 to 14 (4%) was lessthan half than that for 15 to 17 year olds (10%) in 2016.Those aged 10 to 14 made up 20% of the offenders in 2009, with this figure decreasing to between15-17% in the observed years following 2011. The custody rates remained fairly steady across all ofthe observed years, between 4-6% for offenders aged 10 to 14, and 9-12% for offenders aged 15 to17.Between 2009 and 2016, the proportion of offenders who were males increased from 84% to 88%.Custody rates in 2016 for males were five times higher than for females (10% versus 2%), while ratesseen in 2009 (9% for males and 4% for females) were broadly similar in the other years.The most common offences for offenders in the data for 2016 were summary violence against theperson offences, at 17%. Possession and supply of drugs offences made up 11%, up from 8%observed in 2009. Possession of weapons also saw an increase over the observed years, from 3% in2009 to 7% in 2016. Custody rates were highest for Acquisitive violence (30%), Burglary (domestic)(26%), and lowest for Criminal damage (2%). This was true of all the observed years in the data.More than three-quarters (76%) of the offenders in 2016 had previous convictions or cautions. Thisdecreased from 91% in 200913. The custody rate for this group in 2016 was 11%, while the rate was3% for those with no prior convictions or cautions. This was similar in all observed years.12Custody rates stated in this report are calculated from full-year PNC data and may differ slightly from otherpublished statistics because of variation in time frame, data source, or analytical approach.13The reductions in volumes in the youth justice system has been driven by falls in number of 10-17 year oldentering the criminal justice system for the first time. For an overview of trends and drivers, see Sut