This is an HTML version of an attachment to the Freedom of Information request 'Documents relating to the monitoring, evaluation and governance arrangements of the Cashless Debit Card'.


139






142


143


144







148
s22

s
149
s22

150
s22

151
s22


152






155


156

157
Negotiating to share data – summary of engagement with other agencies 
Approach to stakeholder engagement: 
Some state and territory agencies may have concerns about sharing data. The department 
is engaging with state and territory agencies that are expected to be open to sharing data and able 
to commence sharing data within a relatively short period of time. The department is 
simultaneously engaging with agencies that may be more hesitant to share data. It is anticipated 
that securing some ‘quick wins’ will demonstrate capacity to share data safely and will 
demonstrate the benefits that data sharing may deliver.  
The department is working to establish head agreements with state and territory counterparts, 
with schedules covering specific data analytics projects to be added on a rolling basis. The 
department is seeking access to aggregated data whilst also establishing arrangements for supply 
of unit record level data. 
The department is engaging with other Commonwealth agencies about using data assets held 
by those agencies. This includes engagement with the Department of Education, Skills and 
Employment about labour market and employment services data. The department is also 
developing proposals to use integrated Multi-Agency Data Integration Project (MADIP) data, 
managed by the Australian Bureau of Statistics. Additionally, the department is exploring 
opportunities to share data with the Families Responsibilities Commission.  
Table 1: Key steps and estimated timeframes to access and use state and territory data 
Steps 
Estimated timeframe 
Data-sharing agreement 
 Historically, similar agreements have
 Negotiating a head agreement setting out
taken in excess of 12 months to finalise
general principles
 Dependent on state/territory agreement
 Negotiating individual schedules governing  May require an extended period of time
each data-sharing and analytics project
to negotiate
 Requires agreement from the custodian
 Schedules require agreement on a range
of each dataset covered by a schedule
of technical details: must specify how
data will be transferred, stored and used
and how privacy and risks will be
managed.
Data supply 
 Depends on state and territory agencies
 Arrangements for secure data transfer
 Requires state and territory agencies
 Data may need to be deidentified prior
to prioritise work to prepare and transfer
to supply (a technical process)
data

158
Steps 
Estimated timeframe 
Data linkage and integration 
 Data linkage is a complex and highly
 Aggregate datasets may be integrated
technical process.
to support analysis.
 Minimum estimated time required:
 If unit record level data is supplied it may
4 to 6 weeks
be possible to link data at the individual
 The estimated timeline depends on:
level so that CDC participants can
- quality of data provided by state and
be distinguished within external datasets.
territory agencies and extent of data
cleaning that is required to correct
errors in supplied data
- complexity of the datasets and
whether these have previously been
linked to Commonwealth data
- resource availability (note that the
Australian Bureau of Statistics have
advised that their capacity may
be constrained).
Data analysis 
 Minimum 5 to 6 weeks to generate and
validate findings for each data analytics
project
 Requires data cleaning, testing variables
to use for indicators, and development
and validation of models
Interpretation, data visualisation and 
 Additional time is required to interpret
reporting  
and present findings.
s47B

159
s47B







163
13. This initial analysis did not show differences in rates at which participants accessed Emergency Relief
after commencing on the CDC, compared to prior to commencing on the program. Further analysis
is required to understand factors affecting financial stress and use of services, especially after the
shocks associated with the COVID-19 pandemic in 2020.
14. The insights presented at Attachment A have been generated using a new data asset that the
department has created that draws together departmental administrative data and CDC transaction data.
This data asset is being expanded iteratively and is supporting a corresponding expansion of the
department’s capability to analyse the impact of the CDC program.
15. The department has commenced work to investigate program impact. This includes investigation into
crime data, further analysis on participant spending habits, and analysis of employment patterns.
16. As you have requested, this work is using robust methods, including comparisons of CDC participant
outcomes with other similar income support payment recipients. The department will continue testing
and refining models to produce the most robust findings possible, including the suitability and
of identified comparison groups. Findings from this next stage of work will be provided in the
31 March 2022 update.
Sensitivities: 
17. Some preliminary findings may seem to suggest limited impact of the CDC in some areas. Further
analysis is needed to determine whether any early trends represent actual impacts, and securing access
to state data will provide a broader evidence base to draw upon.
Risk Management: 
18. This is a complex project and requires specialist skills across a range of areas. The department has
procured specialist services from an external supplier (Deloitte) to support delivery.
Departmental Funding / Financial Implications: 
19. Nil. The 2021–22 Budget allocated up to $2 million for data collection and analysis of the
CDC program.
Regulatory Implications: Nil. 
Consultation: 
20. Performance and Evaluation Branch.

164
Attachments: 
Attachment A:  CDC data analytics – December 2021 update 
Attachment B:  MS21-000689 – Update on Cashless Debit Card (CDC) Data Infrastructure and Analytics 
Project 
Attachment C:  Data definitions 
Contact Officer:  Mike Websdane 
Cleared by: 
Patrick Burford 
Position: 
Branch Manager 
Position: 
A/g Group Manager 
Branch:  
Cashless Welfare Engagement and               
Group:  
Communities 
Support Services 
Phone/Mobile:  s47F
 
Phone/Mobile:  s47F
 
Cleared by: 
Liz Hefren-Webb 
Position: 
Deputy Secretary 
Phone/Mobile:     s47F
 
Signature: 
Date: 







167
s22

168
s22

169
s22

170
s22


171


172





175
Data definitions 
This document outlines definitions, assumptions and considerations that readers should 
be aware of when reading the December update on CDC data analytics.  
Participants who have exited the CDC program: 
 CDC exits are defined as participants who exit for the following reasons:
o
ceased receiving a trigger payment
o
deemed financially capable by the Department of Social Services
o
deemed financially capable by Community Panel
o
wellbeing assessment.
 Analysis includes customers who have exited the CDC because a trigger payment
has been cancelled for 13 or more weeks, regardless of the reason for that payment
being cancelled. Further analysis will be conducted to understand these payment
cancellations.
 Participants who exit the program for other reasons, such as where a customer has
passed away, have not been included in this analysis.
 The number of CDC exits is compared with the number of participants who have
participated in the CDC program for a total of 12 or more months in total.
 The time on the CDC program is calculated as the total amount of time
a participant has been active on the CDC. This excludes periods where
a participant has exited for a short time. The total time on the CDC program may
not be continuous.
Priority goods: 
 Priority goods data ceases at 1 March 2020 for this analysis.
 Analysis is limited to spending on Indue cards that occurs at merchants only.
BPAY, bank transfers and direct deposits are not able to categorised as priority
or non-priority.
 Spending can only be categorised by Visa merchant category codes which does not
take into account spending on non-priority goods at priority merchants.
 Details of non-quarantined funds spent are unknown.
Parents and children: 
 A ‘parent’ is someone who has at least one child under 18 at the start fortnight and
end fortnight of a six month window.

A ‘child’ is any person under 18 years old linked to a CDC program participant
who is eligible for Family Tax Benefit (FTB) or likely to be being cared for by
a participant with an FTB ineligibility code that indicates they may be caring for
a child.

176
Spending at utilities merchants: 
 Transactions are limited to merchant code ‘Utilities – Electric, Gas, Water, and
Sanitary’.
 BPAY and bank transfers are currently unable to be used to determine spending
at utility merchants.
Distinct merchants/merchant diversity: 
 Merchant diversity is defined as the number of unique merchant category codes
being used during a calendar year.
 Merchant codes that were used in previous years but were not used in the current
year are counted.
 Number of merchants available to BasicsCard users is current as at 31 July 2021.
Declined transactions: 
 Transactions declined due to restricted Merchant Category Code include a variety
of types of merchants, such as drinking places, betting and casino gaming venues
and packaged alcohol stores. Transactions may be declined for a number of other
reasons.
Ceduna crime statistics: 
 Ceduna crime statistics have been sourced from Data SA.1 SA crime statistics are
recorded by postcode. Not all postcodes were represented in the crime data.
 Suburbs for inclusion in each LGA were sourced from the Australian Statistical
Geography Standard Geographic Correspondences (2016).2
 Oak Valley (Maralinga Tjarutja) and Yalata were also included in the analysis of
the Ceduna LGA as the CDC operates in these communities.
 Total population for this analysis was calculated using the population estimates for
the LGAs of Ceduna, Maralinga Tjarutja, Tumby Bay, Kangaroo Island, Gawler
and Coober Pedy, as well as the population estimates for Yalata State Suburb.
Population numbers have been sourced from the ABS Census (2011 and 2016) and
Regional Development Australia Eyre Peninsula Inc Population Estimates (2020).3
1 https://data.sa.gov.au/data/dataset/crime-statistics 
2 https://data.gov.au 
3 https://profile.id.com.au/rda-eyre-peninsula/population-estimate?WebID=100 


177


178





s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22

s22


213


214






217




219



s22



s22

s22

s22

s22

227
Explainer – about the data 
This document provides more detail about the data and analytical methods used to generate 
findings for the March update of the CDC Data Infrastructure and Analytics Project. Findings 
are presented at Attachment A
What data have been used? 

Social security data: data related to social security payments, including income
support payment data and other administrative data recorded by Services Australia.

Data reported by service providers through the Data Exchange (DEX): service
delivery data reported by providers funded by the department.

Regional data on community outcomes and crime.
o
Some community outcomes analysis has been undertaken using
publicly-available regional data. Different datasets, at differing levels
of aggregation, are available in each jurisdiction: these determine the types
of analysis that are possible. Not all analysis can be undertaken for all CDC
regions.

Data presented are as at 28 January 2022, unless otherwise stated.

The period after the commencement of the COVID-19 pandemic has been excluded
from analysis unless otherwise stated. Some graphs include the period affected by the
pandemic.
o
This period includes economic and other shocks. Additional policy measures,
including stimulus payments, were also implemented.

Data relating to small numbers of people are suppressed in accordance with department
policy on confidentialisation. Approximations are used where it is not appropriate
or not possible to be precise, such as <5 rather than a precise number, or n.p., meaning
‘not provided’.

Generally, data are rounded to the nearest whole number, or one percent for clarity.
Comparative methods: 

Comparative analysis: analysis comparing different groups or units. This update
includes comparative analysis undertaken at the regional level and comparisons
between CDC participants and other income support payment recipients. Consistent
time periods and variables are used for comparisons.

Comparison groups of similar income support payment recipients were constructed
based on several variables:
o
address, and a minimum threshold of time at that address
o
age filters: these corresponded with CDC eligibility in each region
o
a minimum threshold of continuous days receiving an income support payment
o
length of time that social security data about an individual are available
o
exclusion of Income Management participants and previous CDC participants.

228

Comparison regions were selected on the basis of multiple factors: remoteness,
socioeconomic status and disadvantage, land area, population, and population receiving
income support payments.
o
Indices used: Accessibility / Remoteness Index of Australia (ARIA+), the Index
of Relative Socio-Economic Disadvantage (IRSD), the Index of Relative
Socio-Economic Advantage and Disadvantage (IRSAD), the Index of Education
and Occupation (IEO), and the Index of Economic Resources (IER).
Longitudinal methods: 

Longitudinal analysis: analysis of the same variables over a period of time: In this
report, this includes ‘pre-post’ analysis and analysis of trends over multiple years.
o
Some pre-post analysis compares data prior to CDC implementation with data
after implementation commenced to identify any changes.
o
Some pre-post analysis explores changes after commencement of the
COVID-19 pandemic. This may identify impacts of the pandemic, associated
shocks, or impacts of associated policy measures.
o
Stimulus payments introduced as part of the pandemic response, including the
Coronavirus Supplement Payment, significantly increased the amount of cash
available in some communities. Impacts on community or participant outcomes
may be identifiable.
Demographic data: 

CDC demographic data (page 3, Attachment A) are as at 28 January 2022.
o
The ‘original assessment community’ — the region where a participant
commenced on the CDC — has been used for most analysis. The demographic
tables on page 3, however, break down CDC participant numbers by current
‘assessment community’ to more accurately reflect participants’ current
location.

Indigenous status is based on self-reported categories.

For CDC participants who move to the Northern Territory (NT) or Cape York region,
prior location history is overwritten in the administrative data.
o
Most analysis uses the ‘original assessment community’ indicator is used
to determine a participant’s location. In these instances, CDC participants who
have moved to the NT or Cape York will be included with the NT or Cape York
region. This affects only a small number of participants.
Domestic violence (DV) crisis payment requests (page 4, Attachment A): see crisis 
payments, page 5 below. 
Crime analysis (pages 5–8 and 22–33, Attachment A): 

Crime analysis uses data at a community level. Data are not limited to CDC participants
only.

229

Pre-post analysis was designed to keep analysis periods balanced.
o
The sample size for each piece of analysis was determined based on the period
of time after implementation of the CDC for which data were available.
o
The period after the commencement of the COVID-19 pandemic has been
excluded unless otherwise noted.

Changes over time are calculated by conducting standard statistical tests, where
samples are large enough, to determine if there is a statistically significant difference
between them. Where the difference is significant, the difference between the pre and
post data period is reported as the percentage change between the means.

Analysis of data during the COVID-19 pandemic is included in the Appendix. All
COVID-19 related analysis must be interpreted with a high level of caution due
to increasing variability in data. These tests are only statistically significant when
analysing the average rates across the period of analysis.

Policing practices may affect data. Proactive policing strategies to encourage the
reporting of certain offences, or proactive targeting of specific offences, may increase
the number of offences recorded.

There is a very high level of variability in all crime datasets and seasonality is typical
for crime data. Seasonality cannot be accounted for in the modelling used in this
analysis as it is irregular; there is also a limited number of years suitable for analysis.

Some graphs presenting crime analysis are aggregated by quarter, and some by month.
This varies throughout the community outcomes section and Appendix. Aggregating
by quarter may make some graphs easier to interpret.
Offences related to domestic violence (DV) (pages 5, 23–24, Attachment A): 

South Australian offences linked to domestic violence are reported at a postcode
level. A small number of suburbs that do not fall into the defined CDC or comparison
regions are included in the postcodes used for analysis.

Analysis of domestic violence offences could not be undertaken for Bundaberg and
Hervey Bay. Queensland offences are aggregated at a level that groups all assault
charges, and does not identify domestic violence (DV) related charges. As such, this
area is not deemed suitable for DV analysis.
Western Australia (WA) crime data (pages 7, 23, 25, 29–31, Attachment A): 

Regions: WA offence data used are at a district level. Districts are large, with
populations of between 35,000 and 65,000 people.
o
Crime analysis could not be undertaken for East Kimberley. Data are
aggregated by regions that do not align with the East Kimberley CDC region.
o
Goldfields: data for the Goldfields-Esperance district are used. 67 per cent
of the Goldfields-Esperance population falls within the Goldfields CDC region.
Findings should be interpreted with caution.
o
Mid West-Gascoyne has been used as the comparison district for
Goldfields-Esperance. Mid West-Gascoyne includes Carnarvon, which was
selected as a comparison region for the Goldfields CDC region. It also includes
greater Geraldton, which is similar in socioeconomic terms but varies in other
ways. Caution must be taken when interpreting results.

230

Changes to recording and reporting practices implemented between May and
November 2017 affected crime data used in this analysis. Trends over this period
should be interpreted with caution.

Breach of Family Violence Restraint Order data for WA were not included in this
analysis due to inconsistencies in reporting.
Prison crisis payments and prison-related benefit cancellations (page 9, Attachment A): 

This analysis includes approved crisis payments or a benefit cancellation with relevant
cancellation codes.

These payments and cancellations do not directly correlate with when an offence was
committed. An offender may imprisoned at a later date and imprisonment periods may
vary.
o
When multiple prison release payments were made after commencing on the
CDC it was assumed that at least one offence occurred after commencing on the
CDC. A single prison release payment occurring after commencing on the CDC
was not used to indicate that an offence that was committed after commencing
on the CDC.
Gambling and school attendance (slide 10, Attachment A): 

Gambling data are available by Local Government Area (LGA). LGAs do not
completely align with the boundaries of the Bundaberg and Hervey Bay CDC region.
To account for this, weighted averages were applied to the LGAs of Fraser Coast and
Bundaberg to approximate the boundaries of the CDC region.

School attendance and gambling data could not be analysed for all CDC regions.
Adequate data were not available at the appropriate level of disaggregation for all
jurisdictions.
Leaving income support payments (slides 12–13, Attachment A): 

CDC participants leaving payment: participants that had a cancellation of all CDC
trigger payments, after their first CDC payment, lasting for more than 3 months.

Income support payment recipients in comparison regions leaving payment:
payment recipients that had a cancellation that occurred after the date of
implementation of the CDC in the CDC region.

A temporary pause on placing new eligible participants on the CDC was implemented
on 26 March 2020 in the context of the COVID-19 pandemic. CDC participants and
comparison income support payment recipients were not included in the analysis
if their first trigger payment occurred during the pause period.
o
CDC participants and comparison income support payment recipients were
included if the cancellation occurred after the pause period — after they would
have commenced on the CDC.
o
After the pause was lifted, a staggered approach was taken to commence new
CDC participants. An approximate date for the end of the pause period was used
to account for this.

231

Leaving payment due to employment: employment was determined using payment
cancellation codes related to employment or likely to relate to employment.
o
Some cancellation codes were deemed likely to relate to employment.
The number of participants leaving income support with these codes were
weighted by 0.8 to account for the uncertainty of the categorisation.
o
This approach is consistent with the method used for Try, Test and Learn Fund
outcomes analysis. The method was informed by Department of Education,
Skills and Employment survey data on employment outcomes.
Crisis and urgent payments (pages 5, 16–18, Attachment A): 

Only participants who had requested a crisis or urgent payment prior to commencing
on the CDC were included in the analysis.

Payments requested after commencing on the CDC: this only includes payment
requests where the participant is active on the CDC at the time of the request.

The pre and post CDC period is measured in relation to the date of each participant’s
first CDC payment.

Requesting fewer payments: this refers to a participant requesting on average fewer
payments per quarter.
o
Data may be available for differing periods of time pre and post CDC.
To account for this, the analysis compares the average number of payment
requests that a participant made per quarter before and after commencing on the
CDC.

Analysis of urgent payment requests includes all types of requests, except for approved
requests for funeral expenses. For cultural reasons these requests may be more common
in remote locations and regions with a high Indigenous population. Only approved
requests are assigned a reason code.

Domestic violence (DV) related crisis payment requests are a subset of all crisis
payment requests. The same approach to analysing change after commencing on the
CDC was used.

A DV crisis payment request is defined by a set of codes specifically describing
DV crisis payments. Perpetrator-related DV requests were not separated from all DV
crisis payment requests.
Crisis and urgent payments - aggregate analysis over time (pages 17–18, Attachment A): 

The average rate of crisis payment requests represented the number of requests that are
made, divided by the number of people, each quarter.
Rent Deduction Scheme (RDS) (page 19, Attachment A): 

Codes relating to RDS deductions and arrears were used for this analysis.
Centrepay (page 20, Attachment A): 

Codes existing under Centrepay deductions were used to identify deductions for
accommodation.


232







236

The department is continuing bilateral discussions in parallel to minimise the risk
of delay. Bilateral data-sharing arrangements may progress more quickly than
a multilateral agreement.

Once formal agreements are in place, some time may be needed to commence data
supply and prepare data for analysis.
Data priorities: 

Priority is being given to:
o
crime and police data
o
health data, including hospital data and use of alcohol and drug services
o
alcohol sales and consumption data
o
data on gambling expenditure and behaviour.

The department is seeking access to unit record level data (deidentified individual data)
or disaggregated data at the lowest level available.
o
Where possible, the priority is to integrate data — to link data from different
sources at the unit record level.
o
This will allow the department to distinguish between CDC participants and
other records in other datasets, and to combine social security data with other
data to better analyse a wide range of participant outcomes.

The department is requesting access to aggregate data (at a lower level of aggregation
than is publicly available) whilst progressing proposals to access unit record level data.

In the longer term, the department is seeking a wide range of data, including data
related to housing, child protection, financial capability and financial support services,
and education, so that evidence can be generated across a range of measures
on an ongoing basis.
Northern Territory (NT): 

A senior executive meeting with the Northern Territory Police Force has led to
agreement to commence sharing data as soon as possible.
o
A formal Memorandum of Understanding is being drafted.
o
Access to unit record level police data is expected to be secured relatively
quickly.

Engagement with the NT Department of Corporate and Digital Development regarding
wider data sharing is continuing, along with aggregate data requests.
Queensland: 

The Queensland Department of Communities, Housing and Digital Economy is now
facilitating aggregate data requests, including the following at the Statistical Area 2
(SA2) level:
o
hospital emergency department data
o
data on use of alcohol and other drug services
o
crime data.

237

s47B
It is anticipated that access to unit record level data may require additional
engagement and time.
South Australia (SA): 

The SA Department of Premier and Cabinet Engagement is facilitating aggregate data
requests, commencing with crime and police data.

The Department of Human Services has also expressed interest in sharing data
related to financial capability and financial counselling and support services. A
project proposal is being drafted.
Western Australia (WA): 

In response to your letter of 18 January 2022, the Hon Simone McGurk MLA,
Minister for Child Protection; Women’s Interests; Prevention of Family and
Domestic Violence; Community Services, has agreed to expedite data sharing with
the WA Department of Communities.
o
Minister McGurk has also referred whole-of-government considerations
related to sharing unit record level data to the Hon Stephen Dawson MLC,
Minister for Innovation and ICT; Medical Research; Volunteering.
o
This is consistent with the whole-of-government approach being taken under
the Intergovernmental Agreement on data sharing.

The department is working with the WA Department of Communities to progress
data-sharing arrangements.
o
Letters of exchange are being drafted.
o
Aggregate data requests are being workshopped. These cover crime and
police data and education data including school attendance.
 A project to share unit record level child protection and housing data is also being
planned with the Department of Communities.