Hall Steve – Serco Recidivism Measure

advertisement
A New Approach to Reducing Reoffending
Presented by
Steve Hall
Director Reducing Re-offending
Reintegration Puzzle Conference
1
Ever hear about desistance from crime?
August 2003 ,UK
Re-offending in New Zealand
And desistance...
No simple definition of recidivism
Ad hoc measurement.
Remarkable inconsistency of approach.
Different definitions applied in different contexts.
Measured in different ways.
Different measures being compared as if same.
(Maltz 1984).
Recidivism increasing
In 2011 in the UK:
More offenders have previous history of offending
(90%).
A third committed or linked to 15+ crimes.
46% of imprisoned had 15+ crimes.
2001- 30% imprisoned offenders.
“Ministry of Justice officials say the figures show a
"clear trend" of a rising re-offending rate”.
BBC News Report 24 May 2012
.
Recidivism decreasing
However of the 2011 cohort:
Proven re-offending rate - 25.5%.
Increased 0.6 percentage points in 12 months.
Fall of 0.7 percentage points since 2000.
More likely to re-offend than 2000 cohort.
After controlling for offender characteristics, this is a
decrease of 3.1 percentage points.
So why so much variation?
Context is critical:
The target group.
The event/events used to indicate the occurrence of
re-offending, and the source of the information.
The time period over which the indicator events are
observed and counted.
Recidivism in Australia Payne J, Australian institute of Criminology
Measurement is complex
Event inside
12 months
Waiting period
Up to 24 months
Proven conviction outside
24 months don't count
Only
convictions
count
Only first
conviction counts
Ministry of Justice UK: Proven re-offending statistics: definitions
and measurement. October 2012
Thinking differently?
Never confuse a single defeat with a final defeat.
~F. Scott Fitzgerald
It is a mistake to suppose that people succeed through
success; they often succeed through failures.
~Anon
Comparing (re-offending) rates
Scottish Centre for Crime and Justice Research 2012
Is change happening?
Two in every three prisoners will have been previously
imprisoned.
25% of prisoners reconvicted within three months of
being released.
25 -30% of prisoners reimprisoned within one year of
release.
Recidivism rates been consistent over time.
A (new) New Zealand approach
“The average number of days spent out of
custody in the 12 months following release
from prison.”
The Out of Custody Index (OCI)
A very different approach
Includes everyone in custody - “social impact”.
Uses IOMS data - providing “real time” information.
Incentive to work with all prisoners – no “failure point”.
Work starts prior to any conviction.
Alternative (out of the box) solutions.
Amount of “failures” less significant.
Positive measure encourages “strengths” based practice.
Positive improvements have (realisable) $ value.
Simple business rules
IOMS data.
Prison releases.
Minimum (initially 42 days) continuous custody.
Death, deportation and extradition only exclusions.
Multiple prison releases included.
Early development phases
Several other prison sites and combinations considered.
Normalising populations by risk, release type, age,
prison duration failed to produce a valid match.
Normalising by risk makes logical sense, but risk
measure is not yet complete.
Outcome of Development
Early findings:
No “equivalent” direct comparison group.
Significant regional influence causing increased remand
times for Auckland Courts.
ROC*ROI is not available nor accurate for all prisoners
on their date of release.
Possibility of comparing against “whole population”.
641
1
1
6
1
2
13
3
2
35
9
5
1
1
1
24
12
241
3
1
619
62
3
6
1
1
15
7
3
12
10
294
1
30
2
679
1
268
2
1
1
2
1
4
6
2
12
3
349
3
1
9
42
4
19
1
3
15
12
1
17
1
3
9
1
1
1
2
1
8
12
103
1
3
5
343 988
3
7
407 1044
4
156
1
470
1
34
23
7
142
8
3
6
221
14
20
2
1
6
2
8
6
3
15
4
11
3
1
37
183
372
209
8
10
1
6
5
2
684
50
6
611
163
1
6
3
3
240
213
1
3
296
4
3
1
2
382
3
1
369
34
6
483
5
290
216
679
8
1
364
3
5
1
305
82
1
464
25
Grand Total
WHANGANUI PRISON
WELLINGTON PRISON
WAIKERIA PRISON
TONGARIRO/RANGIPO
SHCF
ROLLESTON PRISON
RIMUTAKA PRISON
OCF
NRCF
NEW PLYMOUTH
MECF
MANAWATU PRISON
INVERCARGILL PRISON
7
3
5
2
HAWKES BAY PRISON
18
350
CHRISTCHURCH PRISON
174
7
AUCKLAND PRISON
ARWCF
Initial Prison
AROHATA PRISON
ARWCF
AUCKLAND PRISON
CHRISTCHURCH PRISON
CHRISTCHURCH WOMENS
DUNEDIN PRISON
HAWKES BAY PRISON
INVERCARGILL PRISON
MANAWATU PRISON
MECF
MT EDEN WOMENS
NEW PLYMOUTH
NRCF
OCF
RIMUTAKA PRISON
ROLLESTON PRISON
SHCF
WAIKERIA PRISON
WHANGANUI PRISON
Grand Total
AROHATA PRISON
Last Prison
CHRISTCHURCH WOMENS
New Zealand prison releases (sentenced
199
357
6
982
159
4
539
261
327
1311
3
357
346
318
451
2
2
1623
246
7493
7493
19
207
381
155
4
2 426
704
2
1
3
1
5
190
4
1
1
192
1 2258
1
1
5
8
5
86
325
1
1
2
2
1
5
1
5
7
4
8
2
1
14
2
2
4
4
1
1
3
1
7
1
2
1
1
7
3
2
1
38
1
9
95
2
371
157
2 524
11
161 432
2
1
1
4 723
7
1
38
1
1
3
1
89 337 197 197 2357
3
2357
1
1
1
1
4 1357
9
1
1
99 379 168 561
5
16
Grand Total
WHANGANUI PRISON
WAIKERIA PRISON
TONGARIRO/RANGIPO
SHCF
ROLLESTON PRISON
RIMUTAKA PRISON
OCF
NRCF
NEW PLYMOUTH
NATIONAL OFFICE
1
2
4
4
MECF
MANAWATU PRISON
INVERCARGILL PRISON
HAWKES BAY PRISON
CHRISTCHURCH PRISON
AUCKLAND PRISON
ARWCF
Initial Prison
AROHATA PRISON
ARWCF
AUCKLAND PRISON
CHRISTCHURCH PRISON
CHRISTCHURCH WOMENS
HAWKES BAY PRISON
INVERCARGILL PRISON
MANAWATU PRISON
MECF
NATIONAL OFFICE
NEW PLYMOUTH
NRCF
OCF
RIMUTAKA PRISON
SHCF
TONGARIRO/RANGIPO
WAIKERIA PRISON
WELLINGTON PRISON
WHANGANUI PRISON
Grand Total
AROHATA PRISON
Last Prison
CHRISTCHURCH WOMENS
New Zealand prison releases (remand)
160
430
6
732
92
345
195
203
2297
2
105
414
167
540
11
1
1425
1
133
7259
7259
3
1 124
6 1388 137
Final measurement design
Reference group will always be different but provide
benchmark.
Normalising results by risk banding ( ROC*ROI) will
improve over time.
The release population from MECF is significantly different
than any other prison and the whole population. Even after
scaling by risk and other factors still see a gap based
largely on remand duration.
RI and CI release populations
compared
2011/2012
All Releases (14773)
RI Population (7103)
2599
CI Population (6223)
Reference Group (988)
4117
387
601
1118
Maori 189
Non-Maori 198
Cohort size and impact (MECF)
All
1 Week 2 Weeks 3 Weeks 4 Weeks 5 Weeks 6 Weeks 7 Weeks 8 Weeks
3031
1817
1633
1497
1352
1234
1144
1051
996
The cohort group is reduced by a further 12.5% because
prisoners who are not continually in custody are excluded.
Timeframes
Measurement
period
July 2010 –
June 2011
releases
Serco year 1
measure
Measurement
period
July 2011 –
June 2012
Measurement
period
Jul 2012 – Jun
2013
Measurement
period
July 2013 –
June 2014
Serco started
management at
MECF from
September 2011
Serco begin
performance
measurement
from July 2013
Serco year 2
measure
Serco year 3
measure
Serco year 4
measure
Measurement
period
July 2014 –
June 2015
Measurement
period
July 2015 – Measurement
period
June 2016
July 2015 –
June 2016
Target 25% reduction expected
in 2016/17 Annual Report
Measurement validity
Designed to be a ”proxy” measure.
Re-imprisonment (or not) is an event that is determined
by the state’s actions, systems and processes.
Connected to recidivism or re-offending but not the
same.
Amount of time an individual is able to “avoid” custody
immediately post release is a very strong ‘indicator’ of
desistance.
OCI rates
OCI Total
OCI MECF Scaled
OCI Dept Scaled
325
320
315
310
305
300
295
290
285
April
20092010
July
20112012
Current RI rates in New Zealand
20
RI Reimp
22
RI MECF
RI Dept
24
26
28
30
32
34
April
2009-2010
April
2011-2012
RI rate 2011-12 27.1 RI rate 2012-13 27.0
36
Comparing OCI with RI
RI
OCI
July
2011-2012
Limit of RI data
Measuring performance improvement
Serco and Corrections baselines will be measured
from July 2011 – June 2012 release period.
330
Average Days out of Custody
325
July
2008
320
315
310
Reference Group
305
Comparison Group
300
CI Population
295
290
285
20
07
/0
7
20
-2
07
00
/1
8/
0
20
- 2 06
08
00
/0
8/
1
20
- 2 09
08
00
/0
8
4
20
- 2 /1 2
08
00
/0
9/
7
20
- 2 03
08
00
/1
9/
0
20
- 2 06
09
00
/0
9/
1
20
- 2 09
09
00
/0
9
4
20
- 2 /1 2
09
01
/0
0/
7
20
- 2 03
09
01
/1
0/
0
20
- 2 06
10
01
/0
0/
1
20
- 2 09
10
01
/0
0
4
20
- 2 /1 2
10
01
/0
1/
7
20
- 2 03
10
01
/1
1/
0
20
- 2 06
11
01
/0
1/
1
20
- 2 09
11
01
/0
1
4
20
- 2 /1 2
11
01
/0
2/
7
- 2 03
01
2/
06
280
Using results from 2010/2011 releases we can see MECF at 296, and the Department of
Corrections at 315 days out of custody.
Rolling 12-Month Reference Period
From 1 July 2013
Serco MECF started reporting performance based on the
OCI measure.
First time in the world that a “distance travelled” approach
has been used to measure recidivism.
MECF aims to increase the average post custody release
period by three days per annum for the next four years.
Twelve days increase in OCI is equivalent to a 25%
reduction in re-offending rates for the target group.
280
1/07/2017
285
1/04/2017
290
1/01/2017
295
1/10/2016
1/07/2016
1/04/2016
1/01/2016
1/10/2015
1/07/2015
1/04/2015
1/01/2015
1/10/2014
1/07/2014
1/04/2014
1/01/2014
1/10/2013
1/07/2013
1/04/2013
1/01/2013
1/10/2012
1/07/2012
1/04/2012
1/01/2012
1/10/2011
1/07/2011
1/04/2011
1/01/2011
1/10/2010
1/07/2010
1/04/2010
Outcomes – All prisoners
330
325
320
315
310
305
300
Dept Target
MECF Target
OCI MECF Scaled
OCI Dept Scaled
280
1/07/2017
1/04/2017
285
1/01/2017
290
1/10/2016
295
1/07/2016
1/04/2016
1/01/2016
1/10/2015
1/07/2015
1/04/2015
1/01/2015
1/10/2014
1/07/2014
1/04/2014
1/01/2014
1/10/2013
1/07/2013
1/04/2013
1/01/2013
1/10/2012
1/07/2012
1/04/2012
1/01/2012
1/10/2011
1/07/2011
1/04/2011
1/01/2011
1/10/2010
1/07/2010
1/04/2010
Outcomes - Māori
330
325
320
315
310
305
300
Dept (Maori) Target
MECF (Maori) Target
OCI MECF (Maori) Scaled
OCI Dept (Maori) Scaled
Outcomes - compared
All prisoners
Māori
Impact of cohort type on OCI
310
305
300
OCI MECF 0
OCI MECF 7
OCI MECF 14
OCI MECF 35
295
290
285
OCI MECF 42
Using the tool analyse trends
OCI compared 10 years
But over 2-3 years
Introducing court comparison
Conclusions (1)
Any measure of recidivism will have limitations – most
have lots.
Current measures of recidivism ignore remand prisoners
(half of all prison releases).
Fergus McNeil suggests we should not make choices on
the basis of re-offending alone, since this tells us little
about the long-term prospects – new offending might
actually be part of this longer term change.
Conclusions (2)
Measuring “success” (time not offending) might help
increase it.
Current measures of recidivism are just too slow to
provide information that we can usefully act upon.
The cost of crime and the visibility on this are
important drivers for change (and investment in better
solutions).
Measuring success is an important motivator for staff
and prisoners.
Download