Information Being Released

(1)

Reference: 20150330

10 September 2015

Thank you for your Official Information Act request, received on 4 August 2015. You requested the following:

“Copies of all reports or summaries from Treasury on their work using MSD’s integrated data set to develop a picture of the needs of children within Rotorua, and any correspondence or updates on plans to expand this nationwide.”

Information Being Released

Please find enclosed the following documents:

Item Date Document Description Decision

1. 17 November 2014

Children in Rotorua TSY example for Social Sector Board/Forum

Release in part

2. 17 November 2014

OIA SSPM Meeting 17 Nov - Minutes

Release in part

3. 4 December 2014 Email from Andrew Hunter with the following attachments:

- “TA_0_17_Date” A NZ Map locating areas with percentage of at least 2 factors

- “TLA analysis of 3 risk factors” A briefing of Children in NZ between 0-17 years and their interactions with core social services

- “Rotorua_AreaUnit_Data” A Rotorua area unit map with percentage of at least two factors

“Children aged 0-17 years by TLA – 3 factor analysis for SSF.XLSX” Spreadsheet – Analysis by Territorial Local Authority outlining

Release in part Release in full

Release in full

Released in full

(2)

2

Characteristics of children aged 0 to 17 years in July 2012 sorted by percent with at least 2 out of three risk factors.

Note: Spreadsheet is not included separately as it is replicated in attachment 2

4. 10 July 2015 Presentation to the Australian Human Development Conference Wellington, 10 July 2015

Release in part

I have decided to release the relevant parts of the documents listed above, subject to information being withheld under one or more of the following sections of the Official Information Act, as applicable:

•

personal contact details of officials, under section 9(2)(a) – to protect the privacy of natural persons, including deceased people.

This information was based on analysis undertaken by Treasury’s Analytics and Insights team using integrated research data held by the Ministry of Social Development. The goal of this analysis was to improve our understanding of children at risk of poor outcomes as young adults.

A report summarising this, and subsequent analysis, that includes a discussion of the data used, populations considered, methodology, and main findings will be published on Treasury’s website (http://www.treasury.govt.nz/) in mid September.

This report will include useful contextual information relevant to the information released.

In making my decision, I have considered the public interest considerations in section 9(1) of the Official Information Act.

Please note that this letter (with your personal details removed) and enclosed documents may be published on the Treasury website.

This fully covers the information you requested. You have the right to ask the Ombudsman to investigate and review my decision.

Yours sincerely

Andrew Hunter

(3)

Information for release

20150330

1. Children in Rotorua TSY example for Social Sector Board/Forum 1 2. OIA SSPM Meeting 17 Nov - Minutes 8 3. Email 4 December 2014 10 4. Presentation to the Australian Human Development Conference Wellington, 10

July 2015

(4)

IN-CONFIDENCE

Treasury:3064414v1 IN-CONFIDENCE 1

Children in Rotorua

There were 19,753 children aged between 0-17 years in Rotorua at the end of June 2012.

We know the key administratively recorded predictors or characteristics of children who are vulnerable or at risk of poor outcomes as youth/young adults from the various modelling research undertaken to date.

(Vulnerable defined as at being at high risk of neglect or abuse, poor outcomes defined as being in receipt of benefit for more than 2 years or having received a custodial sentence before age 22.)

Of the 19,753 children aged 0-17 years:

• 22% (N=4245) have been supported by benefits for more than 75% of the time since birth.

• 10% (N=1920) have had a CYF finding of neglect or abuse

• 24% (N=4750) have had a CYF report of concern (notification), finding of neglect or abuse, or period in state care

• 9% (N=1682) have had a parent/caregiver who have received a custodial sentence

We are able to estimate the overlap between markers of vulnerability and poor outcomes as youth or young adults for children aged 0-17 in Rotorua:

(Caveat: All the population count estimates in this report are based on integrated administrative data. Linkage error means that counts and overlaps between groups will be underestimated by as much as 10 to 15%.)

TOIA 20150330 Final Binder for

release

Doc 1

Page 1 of 46

(5)

IN-CONFIDENCE

3,218 children aged 15-17 years:

• 5% (N=151) have contact with CYF youth justice services

• 3% (N=104) have received benefits as an youth/adult

• 22% (N=721) have gained no school qualifications one year later**

• 12% (N=375) have a CYF finding of neglect or abuse

• 26% (N=837) have had a CYF report of concern (notification), finding of neglect or abuse, or period in state care

• 11% (N=361) have a parent/caregiver who have received a custodial sentence

• 26% (N=504) have been supported by benefits for more than 75% of the time since birth.

We are able to estimate the overlap between markers of vulnerability and poor outcomes as youth or young adults for children aged 15-17 in Rotorua:

In addition to describing the whole population of children aged 0-17 years, we have looked further at one particular cohort, children born between July 1990 and June 1991, who would have been aged 21 or 22 at the end of 2012. Of these 1,039 children:

• 148 were supported by benefits for more than 5 years before 13 and were either known to CYF or had a parent/caregiver who had received a custodial sentence before 13. 30% of these children went on to receive benefits for more than 2 years or receive a sentence before age 22

TOIA 20150330 Final Binder for

release

Doc 1

Page 2 of 46

(6)

IN-CONFIDENCE

• 499 had none of these risk factors before age 13 (i.e. were never supported by benefits, were not known to CYF, and didn’t have a parent who had received a custodial sentence). Only 4% of this group went on to receive benefits for more than 2 years or to receive a sentence before age 22.

This preliminary analysis illustrates there is a strong relationship between a small number of key risk factors and outcomes.

Can we identify these children?

MSD are able to identify the great majority (80-90%) of children identified by the various research models as high risk from within their own databases which link data from W&I on beneficiaries with information about children known to CYF. MSD knows their names and addresses.

However:

• Ethical reviews and feasibility studies highlight the potential for

misclassification. This potential applies whether predictive models or simple criteria are used. Administrative data can help identify some but not all of the people at risk of poor outcomes.

• Linkage errors mean the W&I – CYF data is not perfect and some children who are high risk will appear low risk as a result.

• There are limits to the information about risk and poor outcomes that can be obtained from administrative data. We need to carefully consider the possibility that the information that can be obtained might partly reflect the way that administrative processes are structured (and potentially biases in reporting) and overstate real differences in the burden of risk and poor outcomes between groups. Some groups of children may appear more at risk relative to others than they really are, and some groups of children who are at risk may not be able to be identified.

• Potential for misclassification and linkage error cautions against using administrative data as the sole means of determining access to services and highlights the importance of maintaining the exercise of professional judgement in any process.

Integrated (operational) data from across agencies could enable better targeting in some cases, i.e. in cases where key predictive variables for a given outcome are not contained in the MSD data. However, they also broaden the potential for linkage error. Alternatives to lists/targets generated from MSD data alone, or lists/targets generated from integrated data with the necessary permissions (e.g. Youth Service) include consent based risk screening or geographic targeting...)

TOIA 20150330 Final Binder for

release

Doc 1

Page 3 of 46

(7)

IN-CONFIDENCE

So we can identify some of the people who need assistance using administrative data, but do we know how to help them?

How can we use these findings?

Integrated research data enables us to identify key administratively recorded characteristics (or risk factors) associated with various poor outcomes. Individual agencies can use these findings to identify the individuals or families who are predicted to be most at risk of having poor outcomes in the following ways:

• use the identified risk factors to screen of their own clients (consent based risk-profiling) to target or refer to services or providers.

• use the risk factors observable in the MSD data to generate lists of clients to target or refer to services/providers in addition to other referral pathways without specific consent

• use models or the characteristics observed in the integrated data to identify geographic regions where high proportions of the target group resides, e.g. targeting particular communities or area units (suburbs).

• use the characteristics observed in the integrated data to generate lists of clients to target/refer clients to services or providers without specific consent, only if they have the necessary legal permissions to do so, e.g. Youth Service.

Legal mechanisms to support these different options would need to be tested.

TOIA 20150330 Final Binder for

release

Doc 1

Page 4 of 46

(8)

IN-CONFIDENCE

Table 1: Selected risk factors and outcomes for a particular birth cohort in Rotorua [Detail withheld under s9(2)(a)]

TOIA 20150330 Final Binder for

release

Doc 1

Page 5 of 46

(9)

Treasury:30644

Table 3: Population

IN-CONFID

4414v1 IN-CONFID

n profile of risk factors and o

IDENCE

IDENCE

d outcomes for children in Ro

7

Rotorua by year of age

[Detail below withheld under s9(2)(a)]

TOIA 20150330 Final Binder for release

Doc 1

(10)

IN-CONFIDENCE

Figure 1: Overlap between selected population groups aged 0-17 years in Rotorua, N=19,753

Figure 2: Overlap between selected population groups aged 15-17 years in Rotorua, N=3,218

TOIA 20150330 Final Binder for

release

Doc 1

Page 7 of 46

(11)

Social Sector Priorities Ministers Meeting

Minute

Monday 17 November 4:00pm – 5.00pm

Cabinet Committee room (8.5 EW)

Attendees: Hon English Hon Bennett Hon Coleman Hon Adams Hon Parata Hon Tolley Hon Woodhouse Hon Lotu-Iiga Hon Foss

Social Sector Forum (SSF) Chief Executives and staff Ministerial office staff

[Deleted - Not Relevant to Request]

TOIA 20150330 Final Binder for

release

Doc 2

Page 8 of 46

(12)

4. Replicate the Rotorua analysis for the entire country, and determine the best avenue for making this analysis public.

SSF lead: date tbc [Deleted - Not Relevant to Request]

release

Doc 2

Page 9 of 46

(13)

1

From: Andrew Hunter [TSY] [mailto:Andrew.Hunter@treasury.govt.nz]

Sent: Thursday, 4 December 2014 8:54 a.m.

Subject: Further information generated for SSF officials in support of response to Ministers [IN-CONFIDENCE]

Good morning, some of you will have seen this already but in the interests of ensuring you all have an opportunity to see this input and provide comment. I have attached information A&I have generated to support the response to Ministers actions point 4.

[Withheld under

s9(2)(g)(i)]

TOIA 20150330 Final Binder for release

Doc 3

(14)

2

We have produced the information in the overlapping circle (venn) diagram for each TLA. This demonstrates that for any given TLA (or area unit within a TLA) we know how many children have one, two or three of the three particular risk factors.

We have geo-mapped the data to show which TLA’s have relatively low and high proportions of children with two or three risk factors. However we don’t see this as particularly useful as it’s not telling us anything new - you would get the same picture at TLA or AU if used any number of measure alone or in combination (e.g. NZDep, number of children supported by benefit currently, school decile, etc). We have done this in response to a suggestion that presenting the information in this way may be useful.

Re the first (LHS) action, can this be changed to: ”We will provide you with national and TLA level information in the form of tables showing the number of children who have multiple risk factors which we know are associated with a range of poor outcomes”

An important point is that MSD can identify the vast majority of these children as they are currently or were previously supported by benefits or in contact with CYF. MSD can identify parents who have/are serving a custodial sentence, as people released from prison receive a one-off payment/allowance from MSD on release, in addition exit reason from benefit (and entry reason) is also captured which includes “started/ceased serving a prison sentence”.

Kind regards. Andrew

Andrew Hunter | Manager - Analytics & Insights | The Treasury

Tel: +64 4 917 6054 | Andrew.Hunter@treasury.govt.nz

CONFIDENTIALITY NOTICE

The information in this email is confidential to the Treasury, intended only for the addressee(s), and may also be legally privileged. If you are not an intended addressee: a. please immediately delete this email and notify the Treasury by return email or telephone (64 4 472 2733);

b. any use, dissemination or copying of this email is strictly prohibited and may be unlawful.

TOIA 20150330 Final Binder for release

Doc 3

(15)

Legend

Percent at least 2 factors

10%+

Between 5% and 10% 5% or less

Data sourced from Treasury, ESRI, Stats NZ, LINZ

Map produced December 2014 By Client Business & Intelligence

By Territorial Authorities

±

0

75

150

225

Document Name:

release

Doc 3

Page 12 of 46

(16)

Legend

Percent at least 2 factors 10%+

Between 5% and 10% 5% or less

Data sourced from Treasury, ESRI, Stats NZ, LINZ

Map produced December 2014 By Client Business & Intelligence

Percent at least 2 factors

By Area Unit, Rotorua

±

0

3.5

7

10.5

Km

Rotorua_AreaUnit_DataDocument Name:

release

Doc 3

Page 13 of 46

(17)

IN-CONFIDENCE

Children at risk of poor outcomes

There were 1,095,000 children aged between 0-17 years in New Zealand at the end of June 2012. We have identified children in the Integrated Child Dataset (ICD) based on information on registered births and school enrolments.

We have described the children’s interactions with core social services identifiable in the ICD. In particular we have identified the groups of children who have had a report of concern, suffered abuse or neglect, have spent long periods supported by benefit as a child, have poor secondary school (NCEA) achievement, contact with CYF youth justice services, benefit receipt and contact with corrections as a youth (Table 1). We know that many of these (administratively recorded) characteristics are correlated with (or predictive of) poor outcomes for children and young adults from the various predictive modelling exercises undertaken to date. The vulnerable children predictive risk model identified that parents’ benefit history, whether they have had contact with CYF, or had other children who were known to CYF were the strongest predictors of new-borns likelihood of being abused or neglected before age 2.

There are other characteristics that could be included in the analysis, in particular demographic and family characteristics, and siblings’ interactions with core social services, and parents benefit duration. (Note: that some of these characteristics are only known for children if they have been supported by benefits as a child at some stage). We will include these dimensions in future analysis.

We have defined poor early adult outcomes as being in receipt of benefit for more than 2 years or having received a custodial sentence before the age of 22.

Of the 1,095,000 aged 0-17 years:

• 14% had been supported by benefits for more than 75% of the time since birth.

• 7% had a CYF finding of neglect or abuse

• 18% had a CYF report of concern (notification), finding of neglect or abuse, or period in state care

• 5% had a parent/caregiver who have received a custodial sentence

TOIA 20150330 Final Binder for

release

Doc 3

Page 14 of 46

(18)

IN-CONFIDENCE

We are able to examine the overlap between the various characteristics associated with vulnerability and poor outcomes as youth or young adults for children aged 0-17 years. Table 2 shows the numbers and percentage of child with none, one, two or three of the selected risk factors. In the diagram below (not to scale) we show the overlap between the three risk factors. About 81% of all children aged 0-17 had none of the three risk factors:

(Caveat: All the population count estimates in this report are based on integrated administrative data. Linkage error means that counts and overlaps between groups will be underestimated by as much as 10%.)

TOIA 20150330 Final Binder for

release

Doc 3

Page 15 of 46

(19)

IN-CONFIDENCE

Of the 61,600 children aged 17 years:

• 4% had contact with CYF youth justice services

• 7% had received benefits as an youth/adult

• 19% had gained no NCEA school qualifications by age 18

• 8% had a CYF finding of neglect or abuse

• 20% had a CYF report of concern (notification), finding of neglect or abuse, or period in state care

• 7% had a parent/caregiver who have received a custodial sentence

• 10% had been supported by benefits for more than 75% of the time since birth.

We are able to estimate the overlap between the various characteristics associated with vulnerability and poor outcomes for children aged 17. Table 2 shows the numbers and percentage of child aged 17 with none, one, two or three of the selected risk factors. n the diagram below (not to scale) we show the overlap between three particular risk factors:

(To be added)

TOIA 20150330 Final Binder for

release

Doc 3

Page 16 of 46

(20)

Table 1: Selected risk factors and

IN-CONFIDENCE

nd outcomes for a particular birth cohort int in Rotorua [Below details withheld under s9(2)(a)]

release

Doc 3

Page 17 of 46

(21)

IN-CONF

Treasury:3064414v1 IN-CONF

Table 2: Number of children with selected risk

NFIDENCE

NFIDENCE 6 isk factors by Territorial Local Authority (TLA)

TOIA 20150330 Final Binder for release

Doc 3

(22)

IN-CONF

Treasury:3064414v1 IN-CONF

NFIDENCE

NFIDENCE 7

Doc 3

(23)

Use of integrated data to

understand children at risk of poor

outcomes as young adults

[Withheld under s9(2)(g)(i)]

Doc 4

(24)

Aims of the study

•

To better understand how many children are in contact

with different government agencies through their

childhood years (with a focus on those who have come

into contact with Child, Youth and Family services.)

•

To understand how common it is that the same

•

To understand how common it is that the same

children are in contact with multiple social agencies.

•

To better understand what happens to these children

as they grow up.

•

Provide information that could help identify which

groups to prioritise for possible future interventions.

2

Doc 4

(25)

Study design

•

A cohort analysis – children born in 1990/91

and observed up to age 21.

•

Study population

: Children enrolled as

domestic students in secondary school in 2006

y

(n=62,418)

•

Descriptive analysis of contact rates and

outcomes as young adults.

•

Use administrative data from the benefit, CYF

services, corrections and education systems.

Doc 4

(26)

Main datasets we use from MSD’s

Integrated Child Data

Doc 4

(27)

Contact with Child, Youth and Family

services

2%

5%

5% 2%

Notification only

Notification and

investigation

Substantiated finding of abuse or neglect

Care/Placement

86%

No contact with Child,

Youth and Family

TOIA 20150330 Final Binder for release

Doc 4

(28)

Doc 4

Page 25 of 46

(29)

Outcomes - females

Doc 4

(30)

Outcomes - males

Doc 4

(31)

Fiscal costs – females (to age 35)

Extent of contact with CYF

TOIA 20150330 Final Binder for release

Doc 4

(32)

Fiscal costs – males (to age 35)

Extent of contact with CYF

Doc 4

(33)

Percent of time included in main

benefit as supported child

7.8%

9.5%

75% or more

50% to 75%

54.4%

19.0%

9.3%

No welfare

support as child

Less than 25%

25% to 50%

Doc 4

(34)

Caregiver with Corrections history

12.7%

Caregiver has

Corrections History

54%

32.9%

Unknown

Caregiver has no

Corrections history

Doc 4

(35)

Overlap between groups

Included in main benefit

as supported child at least

75% of the time

5,901 (9.5%)

CYF contact

8,761 (14%)

1,758

3,872

892

Parent/caregiver

received community or

custodial sentence

7,923 (12.7%)

2,075

1,922

1,176

2,750

Doc 4

(36)

Overlap between groups

by age 13

75% of the time

8,109 (13%)

CYF contact

6,225 (10%)

3,074

2,315

1,011

Parent/caregiver

custodial sentence

7,242 (11.6%)

1,948

951

2,076

2,267

Doc 4

(37)

Overlap between groups

by age 5

75% of the time

10,804 (21%)

CYF contact

1,643 (3%)

6,508

435

489

Parent/caregiver

custodial sentence

5,123 (10%)

141

3,229

1,175

578

Doc 4

(38)

Did not achieve NCEA level 2

Females

Males

Number of characteristics at age 5

TOIA 20150330 Final Binder for release

Doc 4

(39)

Referral to CYF Youth Justice Services

Females

Males

Doc 4

(40)

On benefit with dependent child by 21

Females

Males

Doc 4

(41)

On benefit for more than 2 years by 21

Females

Males

Number of agencies in contact with by age 5

Doc 4

(42)

Corrections sentences served as adult

Females

Males

Doc 4

(43)

Served custodial sentence by 21

Females

Males

Doc 4

(44)

Predicted future fiscal costs (to age 35)

Doc 4

(45)

Summary of results

The data shows that compared to others, a greater proportion

of individuals whose parents/caregivers had served custodial or

community sentences or who had contact with CYF or with

MSD as a dependent child :

•

did not achieve basic school qualifications

••

were early entrants to the benefit system (sometimes with

their own children) and

•

served custodial or community sentence as an adult.

•

These outcomes are associated with significant fiscal costs

•

This analysis gives a basis for assessing intervention proposals

with a view to better understanding their longer term benefits.

Doc 4

(46)

Strengths/Limitations



Population cohort , 20+ years of data



Identification of ‘hard to reach’ populations/

sensitive topics (not often well addressed in more

direct surveys/studies)



Under-estimation of contact rates and outcomes

•

Incompleteness in data sources

•

Linkage not 100%



Limited content of administrative data

Doc 4

(47)

Next steps

•

Looking at outcomes for other groups of

vulnerable children (e.g. being supported by

benefit at birth using 1993 cohort).

•

Moving to use SNZ’s Integrated Data

•

g

Infrastructure (IDI)

–

Wider set of outcomes (earnings, tertiary

education, health)

–

More up-to-date data (up to 2014 at this stage)

–

Better population definition

Doc 4

(48)

Publication

•

‘Using integrated data to understand children

at risk of poor outcomes as young adults’,

. . . later this year.

Doc 4

(49)

Acknowledgements

•

Insights MSD –

providing access to ICD data and

their data-lab whilst on secondment to MSD

Doc 4