Official Statistics

Changing Tunes report (HTML version)

Published 22 October 2026

This analysis looked at the reoffending behaviour of 113 men participating in the Changing Tunes programme. It covers those who began the programme in prison between November 2018 and October 2023.

Changing Tunes delivers music-based rehabilitation programmes in prisons and communities across 21 sites in England and Wales. It uses long-term engagement with professional musicians, with the option for continued support after release, to help participants build positive futures. Males and females were analysed separately due to the known differences in reoffending behaviour. However, this analysis only looked at males as there were too few females within the cohort to analyse separately, leading to insufficient matching quality to report reliable findings.

1. Headline results - male

The overall results do not show a statistically significant effect on a person’s reoffending behaviour.

The headline analysis in this report measured proven reoffences in a one-year period for a ‘treatment group’ of 113 male offenders who received support some time between November 2018 and October 2023 and for a much larger ‘comparison group’ of similar offenders who did not receive it. The analysis estimates the impact of receiving support from Changing Tunes on reoffending behaviour.

For 100 typical men in the treatment group, the equivalent of:		For 100 typical men in the comparison group, the equivalent of:
24 of the 100 men committed a proven reoffence within a one-year period (a rate of 24%), 1 man more than in the comparison group.		23 of the 100 men committed a proven reoffence within a one-year period (a rate of 23%).
143 proven reoffences were committed by these 100 men during the year (a frequency of 1.4 offences per person), 32 offences more than in the comparison group.		111 proven reoffences were committed by these 100 men during the year (a frequency of 1.1 offences per person).

Time to first reoffence has not been included as a headline result due to low numbers of reoffenders, which could give misleading results.

Please note totals may not appear to equal the sum of the component parts due to rounding.

	For 100 typical men who receive support, compared with 100 similar men who do not:
	The number of men who commit a proven reoffence within one year after release could be lower by as many as 7 men, or higher by as many as 9 men. This is not a statistically significant result.
	The number of proven reoffences committed during the year could be lower by as many as 43 offences, or higher by as many as 107 offences. This is not a statistically significant result.

✔	What you can say about the one-year reoffending rate:
	“This analysis does not provide clear evidence on whether support from Changing Tunes increases or decreases the number of participants who commit a proven reoffence in a one-year period.”
✖	What you cannot say about the one-year reoffending rate:
	“This analysis provides evidence that support from Changing Tunes increases/decreases/has no effect on the reoffending rate of its participants.”
✔	What you can say about the one-year reoffending frequency:
	“This analysis does not provide clear evidence on whether support from Changing Tunes increases or decreases the number of proven reoffences during a one-year period.”
✖	What you cannot say about the one-year reoffending frequency:
	“This analysis provides evidence that support from Changing Tunes increases/decreases/has no effect on the number of proven reoffences committed during a one-year period by its participants.”

The figures in this section present the key measures of reoffending for the treatment and comparison groups. Figure 1 shows the one-year proven reoffending rate and figure 2 shows the one-year proven reoffending frequency.

Figure 1: One-year proven reoffending rate after support from Changing Tunes

Figure 2: One-year proven reoffending frequency after support from Changing Tunes

2. Results in detail

The headline result in this report refers to the following:

1. Male analysis: treatment group matched to offenders across England and Wales using demographics, criminal history and individual risks and needs.

This headline result controlled for offender demographics and criminal history and the following risks and needs: accommodation, employment history, financial history, relationships, lifestyle, mental health, thinking skills, drug and alcohol use, and attitudes towards offending.

The sizes of the treatment and comparison groups for reoffending rate and frequency analyses are provided below. To create a comparison group that is as similar as possible to the treatment group, each person within the comparison group is given a weighting proportionate to how closely they match the characteristics of individuals in the treatment group. The calculated reoffending rate uses the weighted values for each person and therefore does not necessarily correspond to the unweighted figures.

Analysis	Treatment group size	Comparison group size	Reoffenders in treatment group	Reoffenders in comparison group (weighted number)
Male	113	109,091	27	39,961 (25,382)

Two measures of one-year reoffending were analysed (see results in Tables 1-2):

1. Rate of reoffending
2. Frequency of reoffending

2.1 Significant results

No results were found to be statistically significant.

2.2 Tables of all reoffending measures

Tables 1-2 show the overall measures of reoffending. Rates are expressed as percentages and frequencies expressed per person.

Table 1: Proportion of men who committed a proven reoffence in a one-year period (reoffending rate) after support from Changing Tunes compared with a matched comparison group

Number in treatment group	Number in comparison group	Treatment group rate (%)	Comparison group rate (%)	Estimated difference (% points)	Significant difference? (p-value)
113	109,091	24	23	-7 to 9	No (0.88)

Table 2: Number of proven reoffences committed in a one-year period (reoffending frequency - offences per person) by men who received support from Changing Tunes compared with a matched comparison group

Number in treatment group	Number in comparison group	Treatment group frequency	Comparison group frequency	Estimated difference	Significant difference? (p-value)
113	109,091	1.43	1.11	-0.43 to 1.07	No (0.40)

The standard acceptable level of statistical significance to demonstrate impact is 0.05. This means that for the difference between the treatment and comparison groups to be considered statistically significant or impactful, the p-value in the tables above must be 0.05 or lower, indicating that the probability of the result occurring by chance is 5% or less.

3. Profile of the treatment group

Changing Tunes delivers music-based learning and mentoring sessions in prison settings, led by professional musicians. Activities include learning musical instruments, songwriting, music recording, and preparing performances and concerts through both group and individual sessions. The intervention is delivered on an ongoing, open-access basis rather than as a fixed-length course, with participants able to engage for varying periods depending on their attendance, ranging from a few weeks to several years.

	Participants included in analysis (113)	Participants not included in analysis (452 with available data)
Sex
Male	100%	100%
Ethnicity
White	81%	80%
Black	14%	12%
Asian	2%	3%
Unknown	2%	5%
Other	1%	1%
Nationality
UK nationality	93%	87%
Foreign nationality	5%	8%
Unknown nationality	2%	5%
Index disposal
Prison	100%

Please note totals may not appear to equal the sum of the component parts due to rounding.

The individuals in the treatment group were aged 19 to 70 years at the beginning of their one-year period (average age 35).

Information on index offences for the 452 participants not included in the analysis is not available, as they could not be linked to a suitable sentence.

For 81 people, no personal information is available as they could not be identified in our databases or did not have the relevant adjudication result.

Information on individual risks and needs was available for 97 males in the treatment group (86% of males), recorded near to the time of their original conviction. This information is not complete for all males across all risks considered for this analysis. For those where information is known for specific risks, some key findings are shown below. For more information, see the Excel annex published with this evaluation.

• 86% of the male participants had some or significant problems with impulsivity.
• 78% of the male participants had some or significant problems relating to participating in regular activities that encourage offending.
• 74% of the male participants had some or significant difficulties with coping.

Note: These percentages represent the proportion of individuals with a recorded risk or need, excluding those where a record could not be found or the risk or need was recorded as ‘NA’.

4. Matching the treatment and comparison groups

The analyses matched the treatment group to a comparison group. A large number of variables were identified and tested for inclusion in the regression models. The matching quality of each variable can be assessed with reference to the standardised differences in means between the matched treatment and comparison groups (see standardised differences annex). Over 95% of variables are categorised as green on JDL’s traffic light scale, indicating that the matching quality achieved on the observed variables was very good.

Further details of group characteristics and matching quality, including risks and needs recorded by the Offender Assessment System (OASys), can be found in the Excel annex accompanying this report.

This report is also supplemented by a general annex, which answers frequently asked questions about Justice Data Lab analyses and explains the caveats associated with them.

5. Additional information on the dataset and terminology

Index dates

The index date is the date at which the follow up period for measuring reoffending begins.

• For those with custodial sentences, the index date is the date they are released from custody.

Other considerations

Part of the cohort within this publication overlaps with the COVID-19 pandemic including lockdowns and operational restrictions. It will therefore be affected by the continued recovery of the courts system. Particularly, continued delays in the processing of cases mean that increased numbers of reoffence convictions may fall outside of the six-month waiting period and therefore not be counted in these statistics. Programme delivery in prisons may also have been impacted by the COVID-19 pandemic.

Participants excluded from the analysis

Participants were excluded if they were previously convicted of a sexual offence or were under the age of 18 on their index date. In addition, their intervention must have started within a prison sentence. Female participants were also excluded because the resulting matching quality was not sufficient to report reliable findings.

6. Limitations and caveats to our findings

Propensity Score Matching (PSM)

This evaluation uses ‘Propensity Score Matching’ (PSM), which is a quasi-experimental evaluation methodology. It is considered a level 4 on the SMS (Scientific Methods Scale). However, PSM is not as robust as randomised control trials or prospectively matched evaluations. For more information on the strengths and limitations of propensity score matching, see the following published reports:

• Mews, Hillier, McHugh, & Coxon (2013) (HTML)
• Eaton & Mews (2019) (HTML).

A key limitation of PSM is that it can only reduce bias based on information about participants that is recorded and available (called ‘observed factors’). Crucially, this information is required for both the treatment and comparison groups. There is a risk that unobserved or unavailable information could influence the results of a PSM evaluation. To mitigate this risk, the JDL strives to include all observed factors expected to be predictive of selection onto the intervention evaluated and of reoffending risk. For more information, see the JDL Methodology Paper (HTML) which outlines the methodology applied in more detail.

Reoffending outcomes and analysis

This evaluation uses the same definition of ‘reoffending’ as used in the Ministry of Justice’s Proven Reoffending National Statistics. More information on this methodology is available on the Proven reoffending statistics collection (HTML). Importantly, this definition does not include crimes that are committed but not recorded by the police or do not lead to a caution or conviction. It also does not adjust for any periods of time individuals are returned to custody, either due to recall or additional sentences, or time spent outside the UK.

There are other rehabilitation outcomes, in addition to reoffending used in this evaluation, which may be important to consider for specific interventions. Examples could include employability, educational or skills attainment, physical or mental health, health of relationships, or attitudes.

7. Numbers of people in the treatment and comparison groups

[1] Adjudication results must be guilty to be considered for analysis, as an individual must have committed an initial offence and have been convicted for it in order for the reoffending rate to be measured.

[2] People were excluded from the analysis if they were still in prison or serving life sentences during the period that reoffending was measured or were too recently released and therefore not included in the latest proven reoffending data available.

[3] Inclusion criteria include being aged 18 or over at the time of release from prison.

8. Further information

Official Statistics

Our statistical practice is regulated by the Office for Statistics Regulation (OSR).

OSR sets the standards of trustworthiness, quality and value in the Code of Practice for Statistics that all producers of official statistics should adhere to.

You are welcome to contact us directly with any comments about how we meet these standards.

Alternatively, you can contact OSR by emailing regulation@statistics.gov.uk or via the OSR website.

Contact

Press enquiries should be directed to the Ministry of Justice press office.

https://www.gov.uk/government/organisations/ministry-of-justice/about/media-enquiries

Other enquiries about the analysis should be directed to:

Justice Data Lab team

Email: justice.datalab@justice.gov.uk

© Crown copyright 2026

Produced by the Ministry of Justice

Alternative formats are available on request from justice.datalab@justice.gov.uk

This document is released under the Open Government Licence