Using Integrated Data for Social Sector Policy and Practice 22 February 2016 Symposium at MSD

Transcription

1 Using Integrated Data for Social Sector Policy and Practice 22 February 2016 Symposium at MSD Justice & the IDI Initial uses, future plans, and the importance of collaboration Charles Sullivan, Sector Group

2 Overview Use of the IDI so far: 3 illustrative analyses What s next? Next analyses planned Enhancing collaboration 1

3 Illustrative analyses 2

4 1. Employment after release from prison 40% 35% Received no benefit for 3 consecutive months in the 6 months after release 30% 25% 20% Received wages/salary for 3 consecutive months in the 6 months after release 15% 10% 5% 0% Why is this oh-so-simple description exciting? 3

5 2. People interacting with justice system more likely to use mental health services (incl. substance abuse) 11% 35% 40% 51% General public People proceeded against by Police People charged in court People starting community sentence People using mental health services 364,000 46,000 38,000 20,000 Why is such non-news newsworthy? 4

6 3. Impact of community vs monetary sentences Main finding: Comparable offenders sentenced to community work rather than fines are more likely to re-offend 4 7 percentage points increase in reoffending (excluding breaches) after 2 years Example of more complex statistical analysis (propensity score matching) Why is this exciting? Clear policy implications Estimating causal impact of sentencing (albeit imperfectly); important precedent 5

7 Percentage on benefit Percentage employed Percentage on benefit Percentage employed Percentage on benefit Percentage employed.(3 continued) 50% 40% Aggravated drink-driving (third or subsequent offence) About 5,900 convictions in 2008 and 2009 where sentence was community work or fines Before matching More likely to be on benefit Before matching 70% 60% Before matching Sentenced to fine Sentenced to community work 30% 50% 20% 40% 10% 0% Month before and after conviction After matching 30% 20% Less likely to be employed Month before and after conviction After matching 50% 40% 30% Same benefit rates before conviction Differential impact of community work versus fines 70% 60% 50% 20% 40% Same employment rates before and after conviction 10% 30% 0% 20% Month before and after conviction Month before and after conviction After matching 6

8 Cumulative reconviction rate Cumulative reconviction rate (3 continued) 40% 30% Re-offending rates after matching More likely to be reconvicted for new offending Differences in new offending 20% 10% More likely to be reconvicted for breach of sentence Differences in breaches 0% Conviction in 2008/9 Month after conviction 7

9 Disclaimer Access to the data presented was managed by Statistics New Zealand under strict micro-data access protocols and in accordance with the security and confidentiality provisions of the Statistic Act These findings are not Official Statistics. The opinions, findings, recommendations, and conclusions expressed are those of the authors, not Statistics NZ or the Ministry of Justice. 8

10 What s next? 9

11 Next analyses Follow-up on mental health analysis Investment approach to justice 10

12 Collaboration: making a step change in the IDI in

13 Collaboration: why? (1) Productivity isn t everything, but in the long run it is almost everything. Paul Krugman,

14 Collaboration: why? (2) 1.To dramatically increase productivity MoJ has recently seen how code-sharing can halve time needed for analysis (and improve quality as well). Thanks to Health & Treasury A&I sharing code 2.Together we are co-creating the IDI working environment for years to come 13

15 Collaboration: what? Personal contact Joint analysis Expanded metadata Code sharing Chat in datalab. QA code. Review draft reports. E.g. A&I work with MoJ on sentencing impact. Continuum: writing a little code joint project. Better data dictionaries: e.g. tips on use of lookup tables; strengths and weaknesses of linking. Beyond data dictionaries: e.g. broader document about justice sector data to pre-empt common misunderstandings. Treasury/A&I code is great start. MoJ adding proven offending, re-offending, death/migration indicator macro. Share derived variables for others to use. What are you sharing? 14

16 Give to get 15