Number publication[1]
Table 1 3
Table 2 16
Table 3 18
Table 4 30
Table 5 30
Annex B 40
Annex C 43
Annex E 47
Annex F 55
Annex G 60
[1]
MOJ (2013). Analysis of the impact of employment on re-offending following release from custody, using Prope
https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/162375/impact-employment-reoffending.p
Table Name
Table 1: Re-offending rates by P45 employment status in the year after release from custody in 2008
Table 2: Descriptive statistics for offenders sentenced to a custodial sentence of less than one year
Table 4: Impact of removing P45 employment spells with randomly allocated start dates
Table 5: Removing offenders with a P45 employment spell after re-offence from comparison
Data underlying Figure 4: One year proven re-offending rates after matching.
Data underlying Figure 5: Survival curves showing time to first re-offence for offenders released from custody after a sente
Data underlying Figure 6: Survival curves showing time to first re-offence for offenders released from custody after a sente
Data underlying Annex D: Distributions of propensity scores before and after matching
Annex G: Benefit and P45 employment status of all offenders released from custody and offenders serving sentences in th
mpact of employment on re-offending following release from custody, using Propensity Score Matching
/uploads/system/uploads/attachment_data/file/162375/impact-employment-reoffending.pdf.pdf
Notes
These tables accompany MoJ's publication 'Analysis of the impact of employment on re-offending following rele
The analysis compares the re-offending rates for offenders who get P45 employment in the year following their
of offenders with no P45 employment. Matching enables us to more confidently estimate the impact of P45 em
differences between the offenders who did and did not get P45 employment on other characteristics.
All tables, except for Annex G, are based on the sample used in the Propensity Score Matching; so are a samp
Only those with valid OASys assessments were used in the analysis, so results cannot be generalised to the w
Annex G uses a different sample of offenders to the other tables. It tracks offenders released from custody and
employment status over time. This is given to provide extra context for the Employment and Re-offending analy
Tables 1 through to Table 3 are before matching, Table 4 through to Annex F are after matching.
Employment:
Any P45 employment spell lasting more than one day is considered an employment spell. Only P45 employme
re-offence are included in the analysis, so we can isolate the impact of P45 employment on the likelihood of re-
Re-offending:
Our outcome measure is the proven one year re-offending rate - any offence committed in the 12 months aft
caution, reprimand or warning in the 12 month period or within a further six month waiting period.
ent on re-offending following release from custody, using Propensity Score Matching'
yment in the year following their release from custody with a matched comparison group
y estimate the impact of P45 employment post-custody on re-offending by minimising the
other characteristics.
nders released from custody and offenders given sentences in the community and looks at their
ployment and Re-offending analysis.
e after matching.
ment spell. Only P45 employment spells which occur prior to the first
ployment on the likelihood of re-offending.
committed in the 12 months after release from custody which receives a court conviction,
nth waiting period.
Table 1: Re-offending rates by P45 employment status in the year after release from cu
P45 employment
after release
A
The mean is the average, a measure of the "typical" value of a data set and the standard deviation (SD) measures how spr
A high standard deviation means that the values are widely spread.
to a custodial sentence of under 12 months (before matching)
No P45 employment
after release
12,190
69%
88%
41%
14%
62%
64%
Mean (SD)A:
29.8 ( 9.4)
38.5 (37.8)
4.3 (12.6)
24.8 (21.9)
the standard deviation (SD) measures how spread out the values are.
Table 3: Descriptive statistics for custodial sentences of one year or more (before matc
P45 employment
after release
A
The mean is the average, a measure of the "typical" value of a data set and the standard deviation (SD) measures how spr
A high standard deviation means that the values are widely spread.
of one year or more (before matching)
No P45 employment
after release
12,320
43%
94%
35%
10%
54%
63%
Mean (SD)A:
31.4 (10.4)
32.1 (34.9)
4.9 (13.4)
18.7 (20.4)
the standard deviation (SD) measures how spread out the values are.
Table 4: Impact of removing P45 employment spells with randomly allocated start date
A
ppt = percentage point
allocated start dates (sensitivity analysis)
39.6% 7.8
ppt changeA
(1,869 pairs)
22.7% 4.3
ppt changeA
(3,102 pairs)
Table 5: Removing offenders with a P45 employment spell after re-offence from compa
A
ppt = percentage point
-offence from comparison (sensitivity analysis)
39.1% 6.7
ppt changeA
(2,274 pairs)
22.8% 3.9
ppt changeA
(3,609 pairs)
Data underlying Figure 4: Re-offending rates by P45 employment status in the year aft
[1]
Based on the sample used in the analysis
y P45 employment status in the year after release from custody in 2008, (after matching) [1]
41.7% 9.5%
24.4% 5.6%
matching) [1]
Data underlying Figure 5: Survival curves showing time to first re-offence for offenders
less than one year. Offenders who enter P45 employment and matched comparison gr
Variable Type
Personal/Demographic information
Gender Categorical
Age at date of index offence
(and age squared) Numerical
[1] The Copas rate controls for the rate at which an offender has built up convictions throughout their criminal c
has in a given amount of time, and the more likely it is that an offender will re-offend within one year. The Copa
[2] All previous offence variables exclude Penalty Notices for Disorder.
parison group
Values
Male; Female
Integer values
12 UK regions. Series of binary variables. (NB data only covers Prisons and
Probation Services
in England and Wales but some Scottish offenders included)
Series of binary variables: White, Black, Asian, Other, Unknown
Integer values
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Values of 0-52
Values of 0-52
Values of 0-52
Values of 0-52
Values of 0-52
Values of 0-52
Values of 0-52
Values of 0-52
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
Continuous variable
2 binary vars, reflecting No Problems / Some Problems / Significant Problems
2 binary vars, reflecting No Problems / Some Problems / Significant Problems
2 binary vars, reflecting No Problems / Some Problems / Significant Problems
2 binary vars, reflecting No Problems / Some Problems / Significant Problems
2 binary vars, reflecting No Problems / Some Problems / Significant Problems
Continuous variable
Continuous variable
s built up convictions throughout their criminal career. The higher the rate, the more convictions an offender
offender will re-offend within one year. The Copas rate formula is
Annex C:
Logistic regression model predicting P45 employment after release for offenders with custodial senten
Standard
Parameter Co-efficient Error
Intercept -0.0643 0.2846
Age at date of first contact with CJS 0.0214 0.00474
Age at date of index offence -0.0609 0.0155
Age at date of index offence, squared 0.00033 0.000215
Previous offences: Burglary (non-domestic) 0.0323 0.0116
Previous offences: Theft -0.0104 0.00526
Previous offences: Fraud and Forgery 0.0367 0.00907
Previous offences: Drink driving offences 0.0856 0.0251
Total number of previous convictions -0.0051 0.0021
Total number of convictions in year prior to index offence -0.0347 0.00788
Copas rate -0.2729 0.0576
Number of weeks receiving Jobseeker's Allowance in year prior to custody 0.0248 0.0104
Number of weeks in P45 employment in year prior to custody 0.0221 0.00156
Number of weeks spent on a DWP employment programme 0.00589 0.00248
Number of weeks receiving Incapacity Benefit in year prior to custody 0.00725 0.0105
Number of weeks receiving Income Support in year prior to custody 0.00766 0.0107
Number of weeks receiving Out Of Work benefits in year prior to custody -0.0181 0.0104
Whether received an intervention while in custody -0.3294 0.151
Whether attended the General Offending Behaviour Programme -0.4429 0.3046
No problems with financial management 0.225 0.0697
Serious problems with work skills -0.1173 0.0897
No problems with employment history -0.1183 0.0814
No problems with attitude to employment 0.1606 0.0752
Female -0.2761 0.1049
Burglary -0.2983 0.1287
Index Offence: Other Indictable offence -0.1386 0.0735
Ethnicity: Asian -0.2118 0.1142
GOR: London -0.212 0.0893
GOR: South East 0.1507 0.076
GOR South West 0.2312 0.1084
Criminogenic Needs Score: Accommodation -0.0146 0.00882
Criminogenic Needs Score: Education, Training and Employability -0.146 0.0242
Criminogenic Needs Score: Drug Misuse -0.036 0.0114
Logistic regression model predicting P45 employment after release for offenders with sentences of one
Standard
Parameter Co-efficient Error
Intercept 0.7154 0.2335
Age at date of first contact with CJS, squared 0.000165 0.000051
Age at date of index offence -0.0427 0.0107
Age at date of index offence, squared 0.000231 0.000145
Previous offences: Sexual (child) 0.0303 0.018
Previous offences: Fraud and Forgery 0.0131 0.0073
Previous offences: Drink driving offences 0.0386 0.024
Previous offences: Criminal damage -0.0286 0.0119
Total number of previous cautions 0.0181 0.0138
Total number of previous convictions -0.00028 0.00122
Total number of convictions in year prior to index offence -0.0113 0.00697
Copas rate -0.1323 0.044
Job density 0.0499 0.0204
Number of weeks receiving Jobseeker's Allowance in year prior to custody 0.00234 0.00169
Weeks in P45 employment in year prior to custody 0.019 0.00124
Number of weeks spent on a DWP employment programme 0.00726 0.0022
Number of weeks receiving Income Support in year prior to custody -0.00862 0.00259
Attended Drug Treatment Programme while in custody -0.1779 0.081
No problems with financial management 0.094 0.0578
No problems with work skills -0.105 0.0634
Illegal income from criminal activity: Some problems -0.1506 0.0526
No problems with unemployment -0.2676 0.0672
Index offence: Drug offences 0.1633 0.0638
Index offence: Fraud and Forgery 0.311 0.129
Index offence: Robbery 0.218 0.0664
Index offence: Sexual offences -0.1323 0.0915
Index offence: Violence 0.1287 0.0553
Ethnicity: Black -0.1183 0.0724
Ethnicity: Other (Not White, Black or Asian) 0.4522 0.23
GOR: London -0.2337 0.069
GOR: North East -0.2154 0.0922
GOR: North West -0.1102 0.0571
GOR: South West 0.1775 0.0913
GOR: Wales -0.1586 0.0924
GOR: West Midlands -0.1585 0.0685
Total criminogenic needs score -0.0436 0.0183
Criminogenic Needs Score: Education, Training and Employability -0.1953 0.0207
Criminogenic Needs Score: Drug Misuse -0.0237 0.0104
Criminogenic Needs Score: Alcohol Misuse 0.0326 0.0107
Criminogenic Needs Score: Atittudes -0.0396 0.0123
These tables show the variables that were used in the logistic regression model used in creating the matched c
(see annex B), but these are the ones that were important in predicting P45 employment after release. From th
characteristics affect whether an offender enters P45 employment on release from custody. In general, a positi
characteristic increases the likelihood of entering P45 employment, and conversely, a negative coefficient mean
likelihood of entering P45 employment.
Wald Chi-
Square Pr > ChiSq
0.0511 0.8212
20.425 <.0001
15.4447 <.0001
2.3591 0.1246
7.7119 0.0055
3.9455 0.047
16.3427 <.0001
11.5824 0.0007
5.9625 0.0146
19.4545 <.0001
22.4396 <.0001
5.6524 0.0174
200.0271 <.0001
5.6243 0.0177
0.4795 0.4886
0.514 0.4734
3.0529 0.0806
4.7602 0.0291
2.1144 0.1459
10.409 0.0013
1.7113 0.1908
2.1132 0.146
4.563 0.0327
6.9255 0.0085
5.371 0.0205
3.5533 0.0594
3.4408 0.0636
5.633 0.0176
3.9342 0.0473
4.5511 0.0329
2.732 0.0984
36.3054 <.0001
9.9739 0.0016
Wald
Chi-Square Pr > ChiSq
9.389 0.0022
10.3973 0.0013
15.8545 <.0001
2.5364 0.1113
2.829 0.0926
3.2228 0.0726
2.5829 0.108
5.8049 0.016
1.7298 0.1884
0.0527 0.8185
2.6355 0.1045
9.0413 0.0026
5.9914 0.0144
1.931 0.1647
233.9851 <.0001
10.8695 0.001
11.1059 0.0009
4.829 0.028
2.6415 0.1041
2.7388 0.0979
8.2133 0.0042
15.855 <.0001
6.5519 0.0105
5.8177 0.0159
10.7858 0.001
2.0923 0.148
5.4245 0.0199
2.6741 0.102
3.8661 0.0493
11.4799 0.0007
5.4603 0.0195
3.7269 0.0535
3.7771 0.052
2.9472 0.086
5.3483 0.0207
5.6555 0.0174
89.2392 <.0001
5.2379 0.0221
9.3022 0.0023
10.3005 0.0013
in creating the matched control group. Many more variables were available for use
ent after release. From the model outputs we can see how different offender and offence
stody. In general, a positive coefficient means that that offender or offence
negative coefficient means that that offender or offence characteristic decreases the
www.ats.ucla.edu/stat/sas/dae/logit.htm
Data underlying Annex D: Distributions of propensity scores before and after matching
Propensity Score
0.24-0.28 0.28-0.32 0.32-0.36 0.36-0.40 0.4-0.44 0.44-0.48 0.48-0.52 0.52-0.56 0.56-0.60
7% 8% 7% 6% 5% 4% 4% 3% 3%
4% 3% 2% 2% 1% 1% 1% 1% 0%
Propensity Score
0.24-0.28 0.28-0.32 0.32-0.36 0.36-0.40 0.4-0.44 0.44-0.48 0.48-0.52 0.52-0.56 0.56-0.60
7% 8% 7% 6% 6% 4% 4% 3% 2%
7% 8% 7% 7% 5% 4% 4% 3% 2%
Propensity Score
0.24-0.28 0.28-0.32 0.32-0.36 0.36-0.40 0.4-0.44 0.44-0.48 0.48-0.52 0.52-0.56 0.56-0.60
10% 9% 8% 7% 6% 5% 4% 4% 3%
9% 6% 5% 4% 2% 2% 1% 1% 1%
Propensity Score
0.24-0.28 0.28-0.32 0.32-0.36 0.36-0.40 0.4-0.44 0.44-0.48 0.48-0.52 0.52-0.56 0.56-0.60
10% 9% 8% 7% 7% 5% 4% 3% 3%
10% 9% 8% 7% 7% 5% 4% 3% 3%
0.60-0.64 0.64-0.68 0.68-0.72 0.72-0.76 0.76-0.80 0.80-0.84
3% 2% 1% 1% 0% 0%
0% 0% 0% 0% 0% 0%
2% 2% 1% 0% 0% 0%
2% 2% 1% 0% 0% 0%
2% 2% 1% 0% 0% 0%
1% 0% 0% 0% 0% 0%
2% 2% 1% 0% 0% 0%
2% 2% 0% 0% 0% 0%
Annex E: Assessing balance after matching
Before Matching
No P45 P45
Employment employment
(Control group) (Treatment group)
N: 12,190 2,360
Female 11.6% 6.4%
Criminal History:
Age at first contact with the CJS 17.5 19.5
Age at first contact with the CJS, squared 348.7 454.7
Age at date of index offence 29.8 28.4
Age at date of index offence, squared 978.6 901.5
Interventions:
Received the Drug Treatment Programme in custody 6.5% 2.5%
Received the General Offending Behaviour Programme in custody 2.0% 0.7%
Received any Prison Accredited Intervention 8.0% 3.0%
Labour Market:
Standardised Job Density in Local Authority 0.0 0.0
Weeks in P45 employment in year prior to custody 4.3 15.4
Weeks receiving Jobseeker's Allowance in year prior to custody 9.7 8.5
Weeks in DWP Programme in year prior to custody 2.6 2.7
Weeks receiving Incapacity Benefit in year prior to custody 10.5 3.5
Weeks receiving Employment and Support Allowance in year prior to c 0.2 0.0
Weeks receiving Income Support in year prior to custody 4.2 1.6
Weeks receiving out-of-work benefits in year prior to custody 24.8 13.6
OASys Assessment:
OASys re-offending predictor 60.9 44.9
OASys violence predictor 39.9 34.5
Illegal income - some problems 21% 14%
Illegal income - Serious problems 24% 11%
Financial Management - No Problems 32% 47%
Financial Management - Some Problems 41% 35%
Financial Management - Serious Problems 27% 17%
Employment History - No Problems 15% 37%
Employment History Some Problems 46% 46%
Employment History - Serious Problems 39% 17%
Attitude to employment No Problems 45% 72%
Attitude to employment - Some Problems 41% 23%
Attitude to employment - Serious Problems 14% 5%
Financial Situation - No Problems 27% 41%
Financial Situation - Some Problems 42% 38%
Financial Situation Serious Problems 31% 20%
Accommodation Criminogenic Score 3.1 2.1
Education, Training and Employability Criminogenic Needs Score 4.5 2.8
Relationships Criminogenic Needs Score 2.6 2.0
Lifestyle and Associates Criminogenic Needs Score 3.0 2.3
Drug Misuse Criminogenic Needs Score 3.3 1.7
Alcohol Misuse Criminogenic Needs Score 3.0 2.8
Attitudes Criminogenic Needs Score 4.6 4.1
Health and Other Considerations Criminogenic Needs Score 3.1 2.2
Total Criminogenic Needs Score 5.3 4.1
Before Matching
No P45 P45
Employment employment
(Control group) (Treatment group)
N: 12,320 3,753
Female 6% 6%
Black 11% 9%
Asian 5% 6%
Other 0% 1%
Criminal History:
Age at first contact with the CJS 17.9 20
Age at first contact with the CJS, squared 385.1 500.1
Age at date of index offence 31.4 30.6
Age at date of index offence, squared 1097.3 1053.9
Interventions:
Received the Drug Treatment Programme in custody 10% 6%
Received the General Offending Behaviour Programme in custody 6% 5%
Received the Sex Offender Treatment Programme 0% 0%
Received any Prison Accredited Intervention 14% 10%
Labour Market:
Standardised Job Density in Local Authority 0 0
Weeks in P45 employment in year prior to custody 4.9 13.2
Weeks receiving Jobseeker's Allowance in year prior to custody 7.8 6.4
Weeks in DWP Programme in year prior to custody 1.9 2
Weeks receiving Incapacity Benefit in year prior to custody 7.9 5
Weeks receiving Employment and Support Allowance in year prior to c 0.1 0.1
Weeks receiving Income Support in year prior to custody 2.6 1.3
Weeks receiving out-of-work benefits in year prior to custody 18.7 12.8
OASys Assessment:
OASys re-offending predictor 52.2 39.5
OASys violence predictor 34.8 29.0
Illegal income - some problems 22% 17%
Illegal income - Serious problems 31% 20%
Financial Management - No Problems 38% 50%
Financial Management - Some Problems 40% 35%
Financial Management - Serious Problems 22% 15%
Employment History - No Problems 20% 40%
Employment History - Some Problems 43% 42%
Employment History - Serious Problems 37% 18%
Attitude to employment - No Problems 54% 75%
Attitude to employment - Some Problems 35% 22%
Attitude to employment - Serious Problems 10% 4%
Financial Situation - No Problems 36% 46%
Financial Situation - Some Problems 40% 36%
Financial Situation - Serious Problems 24% 18%
Accommodation Criminogenic Score 3 2.1
Education, Training and Employability Criminogenic Needs Score 4.2 2.9
Relationships Criminogenic Needs Score 2.4 1.9
Lifestyle and Associates Criminogenic Needs Score 3.1 2.5
Drug Misuse Criminogenic Needs Score 2.9 1.8
Alcohol Misuse Criminogenic Needs Score 2.1 1.9
Attitudes Criminogenic Needs Score 4.3 3.6
Health and Other Considerations Criminogenic Needs Score 2.8 2
Total Criminogenic Needs Score 4.9 3.8
After Matching
Absolute Absolute
standardised No P45 P45 standardised
difference Employment employment difference
(Control group) (Treatment group)
2,298 2,298
18% 6.6% 6.4% 1%
0% 7.9% 7.7% 1%
5% 5.7% 5.4% 2%
3% 1.0% 0.7% 3%
7% 10.0% 11.0% 2%
5% 9.0% 10.0% 3%
7% 10.0% 9.0% 4%
6% 4.0% 5.0% 1%
4% 15.0% 14.0% 2%
6% 13.0% 13.0% 1%
4% 7.0% 7.0% 1%
2% 7.0% 6.0% 3%
2% 12.0% 11.0% 3%
4% 9.0% 10.0% 2%
3% 0 0 4%
65% 14.2 14.6 2%
8% 8.4 8.7 3%
2% 2.6 2.8 2%
43% 4.3 3.6 6%
7% 0.1 0 3%
25% 2 1.6 5%
55% 14.8 14 4%
3,622 3,622
0% 6% 6% 1%
7% 10% 9% 2%
7% 6% 6% 1%
5% 1% 1% 3%
29% 2 2 3%
17% 0.2 0.2 1%
17% 0.9 0.9 5%
5% 0.1 0.1 2%
0% 0.1 0.1 2%
27% 1 1 0%
25% 1 1 1%
29% 2.6 3 7%
25% 0.6 0.6 2%
8% 0.6 0.7 4%
7% 0.3 0.3 3%
27% 1.1 1.1 2%
6% 0.1 0.1 1%
22% 0.9 1 2%
29% 1.3 1.4 3%
5% 1 1.1 4%
42% 18.3 19.6 5%
21% 1.5 1.7 6%
44% -1.2 -1.2 6%
14% 6% 6% 1%
5% 4% 5% 3%
0% 1% 0% 4%
13% 10% 10% 2%
4% 9% 9% 1%
4% 8% 8% 2%
4% 13% 12% 4%
6% 5% 5% 1%
2% 16% 17% 2%
4% 11% 12% 1%
6% 6% 6% 1%
2% 5% 5% 1%
3% 10% 10% 1%
1% 11% 11% 1%
3% 0 0 0%
49% 12.2 12.1 1%
11% 6.5 6.6 1%
2% 2 2.1 0%
19% 5.6 5.1 3%
3% 0.1 0.1 1%
15% 1.5 1.4 2%
31% 13.9 13.2 4%
The table below show whether matched pairs are concordant (both offenders within the pair do re-offend, or bo
of P45 employment. Where the pairs are discordant, there are more pairs where comparison group member re
than the reverse. This suggests that P45 employment reduces the likelihood of re-offending.
P45 Employment
Comparison Group No proven re-offence Proven re-offence
within one year within one year
McNemars Test
Statistic (S) 50.148
DF 1
Asymptotic Pr > S <0.0001
Exact 1.4 x10-12
The upper bound of the p value can tell us whether the effect of P45 employment would be significant, even if t
of P45 employment by 5 per cent, 10 per cent, 15 per cent etc. Even if we increased the odds of employment b
effect on re-offending.
P45 Employment
Comparison Group No proven re-offence Proven re-offence
within one year within one year
No re-offence 2,262 476
within one year
Re-offence within one 680 204
year
McNemars Test
Statistic (S) 36
DF 1
Asymptotic Pr > S <.0001
Exact 2.2 x10-9
[1]
For further information about these tests, see p55 of the Employment and Re-offending publication.
within the pair do re-offend, or both do not re-offend) or discordant, and the direction of the effect
re comparison group member re-offends but the offender with a P45 employment spell does not,
re-offending.
ent would be significant, even if there was an unobserved variable which increased the odds
eased the odds of employment by 25 per cent, P45 employment would still have a significant effect on re-offending.
p (upper bound)
8.9 x10-13
8.9 x10-13
0.000000012
0.000000523
0.000012337
0.000174846
p (upper bound)
0
0
0.000009
0.000235
0.003083
0.023152
e-offending publication.
Annex G: Benefit and P45 employment status of all offenders released from custody a
sentences in the community in 2008[1]
Proportion of offenders
claiming benefits
4 weeks before entering sentence date 42%
Sentence Date 44%
2 weeks after sentence date 46%
2 years after sentence date 51%
Proportion of offenders released from custody or starting community sentences in 2008 who claimed b
or were in P45 employment at some point in the two years following release / sentence date
Proportion of offenders
who claimed benefits
Proportion of offenders in P45 employment or receiving DWP benefits each week before prison recepti
(data underlying chart in Annex G)
Proportion of offenders in
P45 employment
26%
25%
25%
30%
Proportion of offenders
who were in P45
employment
30%
34%
31%
51%
fits each week before prison reception and after prison release
g. It is used here to give extra context to the main Employment and Re-offending analysis.