238
COMMERCIAL IN CONFIDENCE
F I N A L R E P O R T
Cost-benefit analysis of the Cashless Debit
Card
Evaluation of the program from 2015/16 to 2019/20
Prepared for
Department of Social Services
20 September 2021
THE CENTRE FOR INTERNATIONAL ECONOMICS
www.TheCIE.com.au
239
The Centre for International Economics is a private economic research agency that
provides professional, independent and timely analysis of international and domestic
events and policies.
The CIE’s professional staff arrange, undertake and publish commissioned economic
research and analysis for industry, corporations, governments, international agencies
and individuals.
© Centre for International Economics 2021
This work is copyright. Individuals, agencies and corporations wishing to reproduce
this material should contact the Centre for International Economics at one of the
following addresses.
C A N B E R R A
S Y D N E Y
Centre for International Economics
Centre for International Economics
Ground Floor, 11 Lancaster Place
Level 7, 8 Spring Street
Canberra Airport ACT 2609
Sydney NSW 2000
Telephone
+61 2 6245 7800
Telephone
+61 2 9250 0800
Facsimile
+61 2 6245 7888
Email
xxxxxx@xxxxxx.xxx.xx
Email
xxx@xxxxxx.xxx.xx
Website
www.TheCIE.com.au
Website
www.TheCIE.com.au
DISCLAIMER
While the CIE endeavours to provide reliable analysis and believes the material
it presents is accurate, it will not be liable for any party acting on such information.
240
Contents
Summary
1
1 Overview and evolution of the CDC
9
Principles of the Cashless Debit Card program
9
Program locations and uptake
10
Program logic of the CDC
14
This report
20
2 Previous evaluations of the CDC
21
Estimated impacts of the CDC to date
21
3 Methodology
29
Building from the second impact evaluation’s findings
29
Cost-Benefit Analysis approach
32
Benefits from improved outcomes and from changed consumption
33
Consultations
36
4 Changes in expenditure patterns
38
Overview of aggregate CDC transaction data
38
Outgoing transactions using the CDC
39
Declined transactions using the CDC
47
5 Benefits from improved outcomes
51
Summary of benefits from the second impact evaluation
51
Social and community benefits from less problem gambling
52
Education and child wellbeing benefits
59
Economic benefits associated with improved employment outcomes
66
6 Benefits from a change in consumption patterns
74
Benefits associated with reduced alcohol consumption
74
Benefits from reduced cash availability
85
7 Estimated costs of the CDC Program
87
Costs to the Australian Government
87
Costs to participants
91
8 Cost-benefit analysis results
99
Results summary
99
Benefits by program site compared to costs
101
Benefits and costs per participant
102
Benefits and costs over time
105
241
242
4.15 Share of people with declined transactions vs time since account was
opened
50
4.16 Comments from stakeholder consultations – Paying rent can be
challenging
50
5.1 Total benefits associated with improved outcomes
52
5.2 Proportion for whom the CDC helped reduced gambling problems
53
5.3 Costs from gambling addiction (2014/15 values)
55
5.4 Inputs to estimation of gambling benefits
57
5.5 Benefits from a reduction in problem gambling
58
5.6 Gambling benefits by stakeholder
59
5.7 Summary of baseline data regarding child wellbeing
60
5.8 Net impact on child wellbeing
61
5.9 Summary of quantified child wellbeing benefits
63
5.10 Total child welfare and family benefits across the full program duration
65
5.11 Comments from stakeholder consultations – students receiving breakfast 66
5.12 Kaplan-Meier survival curves for CDC participants and non-participants 70
5.13 Comments from stakeholder consultations – Motivating people to find
employment
71
5.14 Studies linking economic outcomes with health outcomes
72
6.1 Total societal cost of alcohol misuse each year in 2010 and 2020
77
6.2 Inflators to project the total societal cost of alcohol misuse in 2020
77
6.3 The SEV as a projector of risk associated with alcohol misuse
78
6.4 Comparison of AUDIT scores for Western Australian Program sites to
benchmarks
79
6.5 Comparison of AUDIT scores for Ceduna to benchmarks
79
6.6 Drink driving in Bundaberg and Hervey Bay compared to Rest of
Queensland
80
6.7 Frequency and amount of drinking in program sites (ex. Bundaberg and
Hervey Bay)
80
6.8 Ratio of alcohol misuse cost per person between program sites and rest of
Australia
81
6.9 Costs of alcohol misuse by participants under the CDC case
82
6.10 Perceived changes in consumption as a result of the CDC
82
6.11 Changes in consumption due to the CDC, by AUDIT score level
83
6.12 Reduction in drinking risk among CDC cohort relative to the base case
84
6.13 Costs of alcohol misuse by participants under the base case
85
6.14 Benefit of avoided costs from alcohol misuse, relative to the base case
85
6.15 Estimates from the literature about cash availability affecting crime rates 86
7.1 Costs of the CDC Program borne by government until 2019/20
87
7.2 Engagement with Support Services by CDC participants and non-
participants
89
7.3 Support service attendance of eventual CDC participants
90
243
7.4 Support Services expenditure (DSS component)
91
7.5 Summary of participant’s feelings while on CDC
92
7.6 Comments from stakeholder consultations – Availability of EFTPOS a
concern
94
7.7 Estimated payment method costs per transaction
94
7.8 Number of potential cash payments by size
95
7.9 Cost from restricting cash payments between 2015/16 and 2019/20
96
7.10 Comments from stakeholder consultations – The Card adds to the stigma of
being on income support
97
7.11 Comments from stakeholder consultations – The ‘white card’
98
8.1 Cost-benefit analysis results (2015/16 to 2019/20)
99
8.2 Breakdown of net cost (up to 2019/20)
101
8.3 Total discounted benefits by Program site, compared to costs
101
8.4 Total discounted benefits by Program site
102
8.5 Breakdown of net cost per person
102
8.6 Discounted benefits per person by Program site, compared to costs
103
8.7 Discounted benefits per person by Program site
103
8.8 Costs and benefits over time
105
8.9 Cost-benefit analysis results for each year
105
8.10 Sensitivity analysis results
107
A.1 Expenditure shares for Australians with government pensions and
allowances as the main source of income
111
A.2 Selected Living Cost Indexes
112
A.3 Alignment of SLCI and CDC categories
112
B.1 Detailed statistical model output for regression predicting spending shares114
B.2 Survival Analysis
117
B.3 Cox proportional hazards model results for unemployment spells
118
C.1 Reasons for a declined transaction related to restrict item types
119
D.1 Estimation of the relative risk reduction in each program site
122
D.2 Relative risk for moderate-or-higher risk consumption
124
E.1 Relationship of total costs and number of participants
125
E.2 Cost per participant
126
E.3 Costs of the CDC Program borne by government including projection for
2020/21
126
244
Summary
The Cashless Debit Card (CDC) sets aside 80 per cent of participant’s welfare payments
to a restricted access bank account that blocks transactions that contain alcohol,
gambling products, and some gift cards, and prevents cash from being withdrawn.
Participants have access to the remaining 20 per cent of their welfare payments to use at
their discretion.1
The primary objective of the program is to reduce social harms caused by excessive
consumption of alcohol, illicit drugs, and gambling.
Since 2016, the Department of Social Services (DSS) has rolled out the program to six
regions, Ceduna and Surrounds, East Kimberley, Goldfields, Bundaberg and Hervey
Bay, the Northern Territory, and Cape York. These locations were selected based on a
range of factors, including community interest, support, readiness and willingness, high
levels of disadvantage and welfare dependence, and high levels of social harm caused by
drug and alcohol misuse and problem gambling.
The Centre for International Economics (CIE) has been commissioned by DSS to
undertake a cost-benefit analysis (CBA) of the first four CDC program regions: Ceduna,
East Kimberley, Goldfields, and Bundaberg and Hervey Bay.
Our approach to conducting the CBA
CBA is a tool to enable systematic and evidence-driven evaluation of the benefits and
costs of a program.
In collaboration with DSS, and a review of the literature, the full range of potential
impacts were identified for inclusion in the CBA. These impacts were categorised across
domains of economic, health, housing, safety, family and child wellbeing, and social and
community benefits, as summarised in chart 1 below. Both positive and negative impacts
were considered in the analysis.
Although a wide range of potential impacts were identified, not all could be assessed
given the evidence base. Chart 1 indicates those that were able to be valued and included
within the CBA (those with a ).
1 We note that this is not always the case, such as participants in the Northern Territory. For
more information on the operation of the CDC program, see
https://www.dss.gov.au/families-and-children/programmes-services/welfare-
conditionality/cashless-debit-card-overview
245
1 Summary of the CDC impacts that were investigated
Data source: CIE in collaboration with DSS
The primary source of evidence informing the impacts of the program is the second
impact evaluation.2 We have quantified the value of impacts identified in this evaluation,
and used the Data Over Multiple Individual Occurrences (DOMINO) dataset, CDC
transaction data, and Support Services data to extend this analysis where appropriate.
To confirm the impacts and specific modelling assumptions, a consultation process was
undertaken with a selected group of jobactive, community advisory groups, and
community services across each of the four regions in scope. This was an important
process to test these inputs with the lived experiences of stakeholders who interact at a
personal level with the program and participants.
A reduction in alcohol misuse is the biggest benefit
Based on self-reported changes in alcohol consumption measured in the second impact
evaluation, we estimate that
the costs of alcohol misuse reduced by 15-20 per cent
across the program locations.
This reduction is largest in East Kimberley, where the per
2 Mavromaras K., Moskos M., Mahuteau S., Isherwood L. 2021 ‘
Evaluation of the cashless debit
card in Ceduna, East Kimberley and the Goldfields region: Consolidated report’, Future of employment
and skill research centre, The University of Adelaide
246
person costs of alcohol misuse are estimated to be almost three times higher than the
Australian average.
Between 2015/16 to 2019/20,
the value of the reduced alcohol misuse associated with
the CDC program is estimated to be $8.5 million in present value terms.3 These
benefits are seen through improved productivity, reduced traffic accidents, and reduced
interactions with the criminal justice system and the health system.
Key assumptions relating to alcohol misuse costs (chart 2) have been informed by the
academic literature and then tested in sensitivity analysis.
2 Approaches to deal with uncertainty in alcohol-related harms
Source of uncertainty
Assumption applied in this study
Whether self-reported reductions in alcohol consumption Self-reported reductions in consumption are equal to
among CDC participants are an accurate reflection of the actual reductions
magnitude of actual impacts
The extent to which reduced consumption leads to
For CDC participants with moderate-or-higher drinking
reduced alcohol misuse costs, such as reduced drink
risk, if their consumption reduces it partially reduces but
driving incidents
doesn’t completely avoid their cost of alcohol-related
harms
Whether the pattern of alcohol misuse costs among
CDC participants have the same alcohol misuse costs as
welfare recipients is similar to those of the broader
the broader population within the program sites
population in the program sites
Source: CIE.
One important uncertainty related to estimated benefit of reduced alcohol misuse is the
extent to which it is attributable to the CDC. The second impact evaluation stated that ‘it
is not possible to attribute these changes to the CDC alone’, but rather ‘to the full
complement of relevant policies in the trial areas’.
For instance, DSS also funds a range of Support Services alongside CDC, such drug and
alcohol counselling, which provide additional support to participants. Although these
additional services are not directly in scope for this analysis, they are likely to drive some
of the reduction in alcohol consumption, such as in the case of alcohol treatment
services. CDC participants had an average of 160 per cent more attendances at Support
Services per person per year, compared to non-participants, and community members
saw these services as a significant benefit.
CDC transaction data suggests that
participants continue to attempt to buy alcohol even after the initial period of having received the CDC. Even after attempted alcohol
transactions have been blocked a few times and the participant becomes familiar with the
features of the Card, there is no change in the frequency of transactions that are blocked
because they relate to restricted items/merchants.
This suggests there is little evidence of a ‘learning by doing’ effect, whereby participants
might attempt less alcohol-related or similarly restricted transactions after they have been
3 The present value of past and future cash flows are calculated using a discount rate. By
discounting future cash flows to today’s value, the CBA accounts for the opportunity cost of
the cash flows and is able to report on the ‘present value’ of the benefits and costs.
247
on the Card for a while. This would be expected for example, if such transactions are
associated with attempts to purchase alcohol while intoxicated, and if alcohol
consumption is falling among participants. with CDC, there is no evidence to support a
learning by doing effect.
Gambling and child welfare benefits are relatively small
Between 2015/16 to 2019/20, the other quantified benefits of the CDC program were:
■
$2.3 million in benefits from a reduction in gambling in present value terms. The
second impact evaluation found that CDC has helped to reduce gambling related
harms, especially for family related harms. The benefits we have measured include
improvements to family relationships, improved mental health, and a reduction in
crime.
■
There was a small net benefit from improved child welfare. Although the second
impact evaluation found there were improvements in children’s access to healthy food
and health outcomes, there was a worsening of safety and school attendance
outcomes. However, the total benefit value of improvements outweighed the cost of
negative changes.
There is insufficient evidence to substantiate other benefits
A wide range of other potential impacts were investigated, but there was insufficient
evidence to conclude there was a measurable net improvement or deterioration as a result
of the CDC. These impacts include:
■
The impact on safety appears to be modest, and perhaps negative. Two prior
evaluations found no reliable evidence that crime/safety outcomes improved.
Although there is some evidence that safety might be improving in some regions,
there is also evidence that some safety outcomes worsened, such as the frequency of
domestic violence and drug offences in the Goldfields and stealing in East Kimberley.
■
The evidence on illicit drug use is mixed. Because of the clandestine nature of the
illicit drug market and the likelihood that individuals will under report their drug
consumption, identifying impacts from the program is difficult. The evidence on the
use of illicit drugs is mixed, and the harms associated with drug use vary considerably
by the incidence of use, the frequency of use, and the drug type. The evidence is not
strong enough to validate there was a positive or negative impact on illicit drug use.
■
There was no noticeable impact on employment outcomes. Statistical modelling
found no improvement in employment prospects for the CDC participants compared
to surrounding areas, after controlling for a range of factors.
■
Negative mental health outcomes were evident, but the additional impact from
CDC is unclear. Although there is evidence of stigmatisation of participants, it is
difficult to isolate this from the negative mental health impacts from being
unemployed and on welfare payments in general. However, the CDC program does
248
249
250
The cost per participant was $3 379 (chart 4), mainly consisting of costs associated with
the card provider and Departmental costs. This is significantly higher than the benefit per
person of $540.
4 Comparison of cost and benefit per participant (2015/16 to 2019/20)
Avoided cost of alcohol misuse
Card provider
Reduced social/community costs of gambling
Departmental
Net child wellbeing benefits
Evaluation and other
Cost
Benefit
0
500
1 000
1 500
2 000
2 500
3 000
3 500
4 000
Cost and bBenefits per person ($/person, npv)
Data source: Cost and CDC transaction data supplied by DSS to CIE, calculations by CIE.
When considering the total financial costs in 2019/20 only, such as the Departmental
costs, Card provider costs, evaluation costs and other operational costs, the cost per non-
zero transaction (i.e. excluding balance enquiries) was $6.64.
Benefit estimates are uncertain due to attribution and survey bias
issues
Many of the benefits calculations rely on self-reported changes in outcomes and
consumption behaviours. This creates a potential bias in the benefit estimates, as
participants may in inclined to not report or under report certain harms and consumption
behaviours, especially for socially unacceptable outcomes or illicit substance use.
These limitations, and the impact on this analysis, are summarised in table 5 below.
Because of these limitations, interpreting the CBA results should be undertaken with
care.
Cost estimates are based on historical data and not subject to uncertainty.
5 Summary of limitations
Limitation
Impact on analysis
Possible survey bias. We heard through community
The survey results from previous evaluations might have
stakeholders that CDC participants are far more likely to a bias towards negative impacts.
report negative news, rather than positive news.
This would underestimate the total benefits generated.
251
Limitation
Impact on analysis
Survey data available is not necessarily reflective of the
The analysis needed to combine evidence sources, such
actual impact or outcome. Although survey responses
as survey data and literature sources to estimate the
may indicate a change in an outcome, such as perceived impact.
improvements in health, actual outcomes, such as
Self-reported reductions in alcohol consumption may
emergency department presentations may differ. For
overstate actual reductions, which would result in total
example, respondents to the Household Expenditure
benefits being overestimated.
Survey run by the Australian Bureau of Statistics to tend
underestimate their alcohol spending.
These assumptions are subject to scenario testing.
There appears to be social pressure for people to not
Although the extent to which this is occurring cannot be
support the Card. Community stakeholders suggested
determined, this may create a bias in the second impact
that when a participant benefited from the program, they evaluation survey and qualitative results.
are unlikely to say so.
This would underestimate the total benefits generated.
Many of the results from previous evaluations are
To accommodate for these mixed impacts, in places, this
inconclusive. Previous evaluations suggest that some
evaluation has taken the net impact, i.e. the difference
participants benefit from the Card, some receive no
between the proportion of respondents who experienced
change in outcomes, and some experience negative
a positive impact and those that experienced a negative
impacts. The second impact evaluation is the primary
impact.
source of evidence for impacts of the CDC that we have
By doing so, we have assumed that the benefit received
valued in this study.
is of equal magnitude or value to the negative impact.
This limitation in part reflects an attribution issue, in that
the CDC has been implemented in a context of many
other policy changes and the COVID-19 pandemic in
2020/21. As a result, it is sometimes difficult to
disentangle the effects of the CDC, yielding inconclusive
findings.
There are many concurrent polices and initiative
The extensive list of other program and policies in place
operating in the CDC sites.
creates an issue of attributing outcomes to the CDC
program.
The analysis has taken a conservative approach and only
included impacts where is evidence base is strong.
Nonetheless, benefit estimates may be overstated to the
extent we cannot distinguish the effects of the CDC from
the effects of concurrent policies.
Source: CIE.
Despite uncertainty around benefit estimates, the core conclusion that the benefits of
the CDC are outweighed by the costs appears to be robust.
The final benefit-cost ratio of 0.16 indicates that benefits would have to be more than six
times higher than estimated to result in a positive net benefit (i.e. a benefit-cost ratio
above 1). Similarly, for the program to have broken even between 2015/16 and 2019/20,
the cost per participant would need to have been 84 per cent lower at $540 per person.
Sensitivity testing showed that under a range of plausible alternative assumptions relating
to benefit estimation and the discount rate, the benefit-cost ratio remained between 0.11
and 0.21. This highlights that the high program costs consistently outweigh the benefits
under a range of difference scenarios and assumptions.
252
1 Overview and evolution of the CDC
The Cashless Debit Card (CDC) sets aside an individual’s welfare payments to a
restricted access account for meeting priority needs. It ensures the responsible use of
welfare payments by reducing expenditure on (and consumption of) alcohol, illicit
drugs and gambling.
Additional Support Services were also funded alongside the CDC’s implementation,
including financial management counselling, drug and alcohol counselling, and other
services.
As the program expands into new regions, the number of participants increases. In
June 2020, around 13 000 people used the Card to make payments across the East
Kimberley, Ceduna, Goldfields and Bundaberg and Hervey Bay regions.
A program logic has been developed in collaboration with DSS, and identifies diverse
benefit streams for the CDC across health, safety, housing, and economic outcomes.
Principles of the Cashless Debit Card program
The CDC reduces access to welfare support payments in cash to restrict spending on
drugs, alcohol and gambling (table 1.1). It is predicated on the principle that welfare
should provide a social safety net, and not facilitate alcohol and drug misuse that
contributes to high rates of violence and abuse, and entrenches individuals and
communities in a cycle of poverty.
Under the CDC, 80 per cent of welfare payments are placed onto a debit card for
purchasing necessary goods and making housing and related services’ payments, and
directs 20 per cent to a nominated bank account where it can be withdrawn as cash.4
CDC can be used at most merchants that accept EFTPOS that have not been blocked by
the Card provider, Indue Ltd.5
4 We note that this is not always the case, such as participants in the Northern Territory. For
more information on the operation of the CDC program, see
https://www.dss.gov.au/families-and-children/programmes-services/welfare-
conditionality/cashless-debit-card-overview
5 Although the Cape York and the Northern Territory regions are out of scope for this analysis,
we note that the restriction rate can vary in these newer sites.
253
1.1 How the CDC operates
Feature
CDC conditions
Card accepting
Any merchant that is able to accept Visa or eftpos unless their primary business is the sale
merchants
or provision of restricted items, where CDC payments are automatically blocked based on
Merchant Category Codes (MCC) and/or specifically blocked by the merchant’s Card
Acceptor Identification Code (CAID).
Merchant
No direct conditions. Most merchants who accept the CDC have no agreement with the
Responsibilities
card provider
Welfare payments
80% of the welfare payment
allocated to the Card
Unrestricted welfare
20% of the welfare payment, placed in the participant’s nominated bank account
Implementation
Automated identification of restricted items and blocking of CDC purchases, known as
Product Level Blocking, has been enabled at over 7 000 merchant stores. Mixed
Merchants (merchants that sell both restricted and unrestricted items) may implement
Product Level Blocking or enter into an agreement to not allow the purchase of restricted
items with a CDC.
Restricted goods and
Alcohol, Drugs, gambling, and some gift cards.
services
Source: Parliament of Australia, 2019 https://www.aph.gov.au/About_Parliament/Parliamentary_Departments/
Parliamentary_Library/FlagPost/2019/April/Replacing-the-BasicsCard, and the Department of Social Services, Cashless Debit Card,
https://www.dss.gov.au/families-and-children/programmes-services/welfare-conditionality/cashless-debit-card-overview
Program locations and uptake
The CDC program began in 2016 and currently operates in six distinct regions.
The regions were selected based on a range of factors, including community interest,
support, readiness and willingness, high levels of disadvantage and welfare dependence,
and high levels of social harm caused by drug and alcohol misuse and problem gambling.
The first four regions are in scope for this evaluation, and summarised in table 1.2 below.
Since March 2021, the program has expanded into the Cape York region of Queensland
and across the Northern Territory. Program participants in these regions are out of scope
for this evaluation.
1.2 Rollout of the CDC program by 2019/20
Regions
Year
Program site characteristic
started
Ceduna
2016
Approximately 1 000 people in 2020 who are recipients of working age payments (such
Region, SA
as Newstart Allowance/JobSeeker a and Youth Allowance) are using the Card. Age
Pension and Veterans’ Pension recipients are not included, however, they can volunteer
to be part of the program.
East Kimberly, 2016
Approximately 1 700 people who are recipients of working age payments are using the
WA
Card. As above, Age Pension and Veterans’ Pension recipients are not included but
volunteering for the program is available.
Goldfields
2018
Approximately 3 700 people who are recipients of working age payments (such as
Region, WA
Newstart and Youth Allowance) are using the Card. Pension recipients are excluded but
can volunteer to participate.
254
Regions
Year
Program site characteristic
started
Bundaberg
2019
Approximately 6 500 people, aged 35 years and under who receive Newstart Allowance,
and Hervey
Youth Allowance (Job seeker), Parenting Payment (Single) or Parenting Payment
Bay, QLD
(Partnered) are using the Card. This age group was chosen to address high youth
unemployment and intergenerational welfare dependence. A person can volunteer to
remain on the program once they turn 36 years of age. All people within the region can
volunteer to join the program if they receive an eligible income support payment.
a Newstart Allowance has been replaced by JobSeeker as of 20 March 2020.
Note: The number of people ‘using the Card’ refers to the number of people making a transaction using a CDC in June 2020 (the last
month of 2019/20, which is the end of the evaluation period under the central case for this analysis).
Source: Indue transaction data supplied by DSS to CIE was used to calculate the number of people making transactions in each month
of 2019/20, with the remainder of information sourced from the DSS website, se
e https://www.dss.gov.au/families-and-
children/programmes-services/welfare-conditionality/cashless-debit-card-overview
The Ceduna region was the first implementation site, and the smallest of the four (chart
1.3). It includes the relatively remote town of Ceduna and surrounding areas.
The East Kimberley region is within Western Australia and borders with the Northern
Territory. The largest town in the region is Kununurra.
The CDC sites within the Goldfields region cover a large area of Western Australia. The
land area of the Goldfields region is considerable, with a number of regional towns
throughout the region.
The Bundaberg and Hervey Bay Region is geographically defined as covering the same
area as the Federal Electoral Division of Hinkler.6 This region covers around half of the
area of the Bundaberg Local Government Area (LGA). This area includes the city of
Bundaberg, which is the main population centre containing more than 70 000 residents.
The Hinkler electorate also covers part of the Fraser Coast LGA.
6 Department of Social Services, 2019,
Cashless Debit Card: Bundaberg and Hervey Bay region
Queensland, Australian Government, see
https://www.dss.gov.au/sites/default/files/documents/09_2019/cdc-factsheet-bundaberg-
and-hervey-baycsb-edits-15jul19-1_0.pdf
255
1.3 Locations of the first four program sites
Data source: Chart supplied by DSS and labels modified by the CIE.
Program locations are designated regional and remote sites in Australia that typically
have significant economic and social challenges.
The degree of social harms and challenges within each location is determined by factors
such as the number of people on income support in the area, alcohol related morbidity,
domestic violence reports, prevalence of substance abuse, unemployment rates, crime
rates, and the like. For example, in 2016 when the CDC Program was introduced to the
region, the Shire of Wyndham-East Kimberly had an unemployment rate of
10.8 per cent, compared to 5.7 per cent for regional Western Australia.7
Through discussions with DSS, it was outlined that community support was also a key
factor in determining program site locations. Community members were consulted about
the implementation of the program, and provided input into critical issues and support
needs in each community.
7 .id, 2021,
Economic profile: Shire of Wyndham-East Kimberley, available at:
https://economy.id.com.au/rda-kimberley/unemployment?WebID=140
256
The CDC applies to anyone that meets the relevant trigger payment criteria in a location
specified in legislation. While Indigenous community leaders have been keen to be early
adopters of the program, later locations have involved significant numbers of non-
Indigenous Australians becoming participants of the program, such that most people
currently on the program do not identify as being an Aboriginal or Torres Strait Islander
person.
Around 30 per cent of the population in the program sites identify as Aboriginal or
Torres Strait Islander (compared to fewer than three per cent of the total Australian
population). However, Indigenous participants make up more than three-quarters of all
CDC participants in Ceduna and East Kimberley.8
The subsequent program location set up in Goldfields, Western Australian is unique and
had a vastly different demographic to the earlier two locations. It has a relatively small
Indigenous Australian population, with Aboriginal and Torres Strait Islanders making up
12.3 per cent of the 12 995 people that live there.9 The working age group made up 56.2
per cent of the population, with a median age of 50 in 2016.10 However, like Ceduna and
East Kimberly, the composition of CDC participants includes a higher proportion of
Indigenous Australians given the proportion of Aboriginal and Torres Strait Islanders
living in Goldfields (57 per cent non-Indigenous and 43 per cent Indigenous).11
It is important to note that the CDC program is not purely an income management
program, with the more recent locations having a stronger focus on employment
outcomes. For instance, when the CDC legislation for Hinkler and Goldfields was
introduced to Parliament in 2018, the scheme was broadened to include moving people
off welfare and into the workforce. The Explanatory Memorandum stated, ‘The
community has significant issues regarding youth unemployment, intergenerational
welfare dependency and families who require assistance in meeting the needs of their
children’12, and it was suggested that the program be modified to address these issues.
8 ANAO, 2018,
The implementation and performance of the cashless debit card trial, Auditor-General
Report No.1 2018-1, p16.
9 Australian Bureau of Statistics, 2016,
Census, extracted via Census TableBuilder, available at:
https://www.abs.gov.au/websitedbs/censushome.nsf/home/tablebuilder
10 Australian Bureau of Statistics, 2016,
Census, extracted via Census TableBuilder.
11 Mavromaras, K., Moskos, M., Isherwood, L., and Mahuteau, S., 2019,
Cashless Debit Card
Baseline Data Collation in the Goldfields Region: Qualitative Findings, prepared by the Future of
Employment and Skills Research Centre, University of Adelaide for the Department of Social
Services, p18, available at:
https://www.dss.gov.au/sites/default/files/documents/04_2019/cdc-baseline-data-collection-
qualitative-findings-29-march-2019.pdf
12 Hunt, J., 2020, ‘Evaluating the Cashless Debit Card: How will it solve poverty and
unemployment?,
CAEPR Topical Issue, 2/2020, available at:
https://caepr.cass.anu.edu.au/sites/default/files/docs/2020/6/Hunt_TI_2_2020_Final.pdf
257
Program logic of the CDC
The program logic for the CDC steps through how activities and outputs of the program
are clearly attributed to the program’s impacts. This also helps provide the evidence-
based link between changes in behaviours and measured benefits (chart 1.4).
1.4 Summary of the program logic for CDC
D
N
S
Australian Government investment into the CDC
A
Change in legislation
TS
Establishment of local community partnership and collaboration
U
IVITIE
P
CT
Investment in additional Support Services
IN
A
Contracting of the Card service provider
TS
U
Practical changes to how a proportion of participants income support payment can be
TP
expensed
OU
N
IN
TIO
P
GE
Amount of cash available in the
Reduced consumption of alcohol,
N
M
A
U
S
community reduced
illegal drugs, and gambling
CH
CON
S
E
M
Improved
Reduced substance
ORT
R
Social Security safety net supports
financial
misuse and
H
vulnerable Australians
S
TE
TCOM
management
gambling
OU
S
M
E
IU
M
Improved housing
Increased economic and
Reduced addiction and
D
R
E
TE
TCOM
security
community participation
dependency
M
OU
M
S
R
E
E
Improved health
Improved economic
Improved neighbourhood
G T
outcomes for the
TCOM
outcomes
community
safety
LON
OU
Note: This program logic was developed by the CIE in collaboration with DSS as part of the development of the CDC CBA Economic
Framework.
Source: Developed by CIE in collaboration with DSS.
Inputs
As at June 2020, the total cost of the program in Ceduna, East Kimberley, Goldfields,
and the Bundaberg and Hervey Bay regions was $79.8 million, with $39.0 million paid to
the private company card provider, Indue Ltd, to cover all operational aspects of the
Card. DSS is responsible for coordinating governance arrangements for the CDC and for
CDC policy, administration and delivery of the CDC.
258
A card provider procurement processes was conducted for the Ceduna, East Kimberley,
Goldfields, and Bundaberg and Hervey Bay regions in which Indue was selected as the
most suitable provider for the CDC. Major banks in Australia were not interested in
delivering the initial CDC trial, and thus Indue Limited was selected on grounds of its
experience delivering the BasicsCard.13 DSS has advised that Indue demonstrated its
ability and experience in dealing with Government payments, its value for money, and it
is an Authorised Deposit-Taking Institution that is regulated under the
Banking Act 1959.
Indue was consigned to deliver the CDC (with DSS), as well as the IT build for the
program, banking services, and local customer support for CDC participants through
contracting Local Partners. Local Partners continue to provide general support, including
facilitating initial Card set up, account balance checking, bill payments, temporary and
replacement cards and assisting participants to address issues as they arise. Local
Partners are tasked with providing information to participants on the Community Panels,
and the application process to decrease the proportion of their restricted funds below
80 per cent.
In addition to 20 per cent of participants’ income support payment being unrestricted,
participants in the first four program sites can transfer up to $200 from their Indue
account into a personal unrestricted account every 28 days.14 There is no requirement for
these external transfers for other expenses to be approved by DSS. If rent or other large
denomination payments cannot be made with the cash available, the participant can
apply for an exception. However, stakeholders from the consultation process described
this exception process as a complex procedure, especially regarding ad-hoc
accommodation payments.
The CDC also uses an EMV chip, whereas the earlier BasicsCard uses a magnetic strip
(making the CDC inherently more secure and harder to counterfeit).
Activities
The Australian Government has made additional investment into Support Services in
each region where the CDC program is in place. Feedback received through the
consultation process highlighted these Support Services as a core component of the CDC.
Before commencing each new program site, DSS worked collaboratively with local
leaders and existing service providers to identify critical issues and support needs in the
community. The Support Services provided are reviewed every year to evaluate their
effectiveness, and to adapt the services to the community’s immediate needs.
13 ANAO, 2018,
The implementation and performance of the cashless debit card trial, Auditor-General
Report No.1 2018/19.
14 Department of Social Services,
Cashless Debit Card – Frequently Asked Questions, Australian
Government, s
ee https://www.dss.gov.au/families-and-children-programs-services-welfare-
reform-cashless-debit-card/cashless-debit-card-frequently-asked-questions
259
The Australian Government has invested more than $2.1 million in Ceduna and
$4.6 million in East Kimberly15 to build a system to deliver services including drug and
alcohol rehabilitation, mental health services, additional family Support Services,
targeted youth activities, and financial counselling services.16
Outputs
Outputs are the direct result of investments and activities, which in this case is practical
changes in how a portion of a participant’s income support payment can be spent as a
result of changes in the design of the income support payment delivery:
■
20 per cent of the welfare payment is transferred to their bank account for cash
dependent expenditure such as school excursions, tuck shop, garage sale, etc.
■
80 per cent welfare payment is allocated to the Card, and it cannot be used to buy
alcohol, illicit drugs, gambling products or withdrawn as cash.
In addition to this, outputs also include:
■
financial Support Services to provide general support like Card set-up, account
balance checking, bill payments, temporary and replacement cards and assistance
with other issues related to the CDC program, and
■
additional Support Services such as drug and alcohol counselling, improved access to
rehabilitation services, family Support Services, etc.
Outcomes
The outcomes are the direct effects of the outputs that occur because of the program,
which in this case include a reduction in the amount of cash, lower expenditure of
alcohol and illicit drugs and gambling related products, and increased use of Support
Services.
According to a survey undertaken by the Reserve Bank of Australia (RBA), cash use in
Australia is relatively more intensive in regional areas due to an older age demographic,
or insufficient presence of businesses accepting cards due to poor internet access. The
survey also reported that 27 per cent of all consumer payments were made with cash in
2019, and that participants with lower household incomes were more likely to be higher
cash users than others.17 The Centre for Aboriginal Economic Policy Research
interviewed 51 participants of CDC in the East Kimberley Region, who reported key
items purchased using cash before the CDC Card. Although those interviewed were not a
15 We understand that Goldfields and Bundaberg and Hervey Bay had investments in Support
Services, but data about the value of these investments has not been provided by DSS or
otherwise identified.
16 DSS, 2021,
Cashless Debit Card Overview, available a
t: https://www.dss.gov.au/families-and-
children/programmes-services/welfare-conditionality/cashless-debit-card-overview
17 Delaney, L., McClure, N. and Finlay, R., 2020, ‘Cash use in Australia: Results from the 2019
Consumer Payments Survey,
RBA Bulletin – June 2020, available at:
https://www.rba.gov.au/publications/bulletin/2020/jun/cash-use-in-australia-results-from-
the-2019-consumer-payments-survey.html
260
representative sample by community, gender or age, the data does confirm the prior use
of cash for transactions in the program region (chart 1.5).
1.5 Key items purchased using cash before the CDC program
Item purchased using cash
Percentage of respondents
Number of respondents (n=51)
Present to give someone
29.4
15
Social events (e.g. Kimberley moon)
13.7
7
Eating out
25.5
13
Big item for the home (e.g. fridge)
49
25
Medicine from the chemist
21.6
11
Transport costs (e.g. for taxis and buses)
64.7
33
Lunch money for children
21.6
11
Bills
23.5
12
Rent
29.4
15
Fuel
51
26
Small grocery shopping
62.7
32
Big grocery shopping
58.8
30
Source: Klein, E. and Razi, S. (2017) The Cashless Debit Card Trial in the East Kimberley. CAEPR Working Paper no. 121/2017
A reduction in the cash economy is an expected outcome of the CDC program.
Lower expenditure on (and consumption of) alcohol and illicit drugs and gambling
related products results from restricted access to cash and vendors selling such products.
Increased use of Support Services results from concurrent access to financial
management counselling, drug and alcohol counselling, and other Support Services
provided alongside the CDC program.
Impacts
The objective of the CDC program is to reduce the social harm caused by income support
fuelled alcohol and drug misuse, and problem gambling. The reduction in social harm or
benefit streams can be categorised in the following domains:
■
Education and child welfare impacts: include those pertaining to educational
opportunities and outcomes such as enrolment, participation rate, and
consequentially, performance in educational programmes of participants and others in
the family or the community.
■
Safety impacts: include changes in perception of safety in the community and losses
associated with crimes such as property theft and damage, loss of life, domestic
violence and child abuse, and crimes within the illegal drug market. This domain also
includes the cost of tackling crime such as policing, criminal courts, prisons, cost
related to insurances against property theft and damage, family and counselling
services, etc. It is to be noted that the impact on crime is not only associated with
lower consumption of alcohol and drugs but also with reduced level of cash in the
community. Besides crime related events and costs, the safety domain also
261
encompasses protection against financial harassment such as unreasonable requests
from family for money, etc.
■
Social and community impacts: include increased community cohesion, engagement
and belonging as a result of improved physical and mental wellbeing, increased
economic participation, and use of public facilities for social activities due to reduced
crime level in the community.
■
Housing impacts: focuses on potential alternative use of income being saved from
reduced expenditure on drug, alcohol, and gambling consumptions on housing costs
such as private rental markets.
■
Health impacts: include reduction in costs associated with morbidity and mortality
due to diseases and mental illness, and workplace and traffic accidents linked to
excessive alcohol and illicit drug use and gambling addiction. This includes tangible
costs such as hospital, medical, ambulance, nursing home, pharmaceutical, and other
support service costs and intangible costs such as value of human life, and pain and
suffering.
■
Economic impacts: include changes in paid and unpaid production costs associated
with workforce participation, productivity at work, and loss of labour due to
mortality. This domain also includes impacts on businesses as a result of redirection
of income support payment to non- alcohol, drug, and gambling related purchases.
The pathways of these benefits are presented in chart 1.6.
262
1.6 CDC Impact pathways
Source: CIE in collaboration with DSS.
263
This report
The CIE has been commissioned by the DSS to conduct a CBA of the CDC program.
The analysis is limited to the first four regions, including Ceduna, the East Kimberley,
the Goldfields, and Bundaberg and Hervey Bay regions.
Cost-benefit analysis is a tool to assess government policy decisions, with a focus on
estimating the monetary value of costs and benefits relative to the state of the world
without the program or policy.18 Not all costs and benefits may be amendable to
monetary valuation, and qualitative impacts can often provide useful context in
interpreting CBA results.
Key objectives of this study are to provide quantitative evidence on overall performance,
key drivers of costs and benefits, and what is required to achieve objectives for
communities participating in the program. This cost-benefit analysis is an
ex post (i.e.
backward-looking) analysis, meaning it will focus on the costs and benefits of the CDC
program to-date rather than the impact of continued or future use of the CDC.
The remainder of this report is structured as follows:
■
Chapter 2 summarises key findings of the previous evaluations of the CDC,
■
Chapter 3 provides an overview of the methodology for this study,
■
Chapter 4 examines data about spending using the CDC to understand patterns in
consumption,
■
Chapter 5 measures the value of improved outcomes including social and community
outcomes and child wellbeing outcomes,
■
Chapter 6 measures the avoided costs of alcohol misuse associated with reduced
consumption of alcohol by CDC participants,
■
Chapter 7 reports costs to government of the CDC, together with estimation and
discussion of costs not borne by government such as mental health costs or
inconvenience from not being able to pay with cash,
■
Chapter 8 combines the cost and benefit estimates from the preceding chapters, and
assesses the net benefit of the CDC program under a range of alternative assumptions,
■
Appendices provide greater detail about the assumptions, methodology and detailed
results.
18 For more general information about Cost-Benefit Analysis to support government decision-
making, see: Department of the Prime Minister and Cabinet, 2020,
Cost-benefit analysis —
guidance note, March 2020, available at:
https://pmc.gov.au/sites/default/files/publications/cost-benefit-analysis_0.pdf
264
2 Previous evaluations of the CDC
Since the CDC commenced in 2015/16, there have been various data collection
activities and evaluations undertaken.
The most recent and substantive is the CDC’s second evaluation, which provided
quantitative and qualitative analysis and data that can be used to inform this CBA.
The previous CDC evaluations identified some consistent outcomes and impacts.
However, not all findings were consistent across the studies, and some outcomes had
inconclusive results.
Estimated impacts of the CDC to date
The CDC program has undertaken two large-scale evaluations, the first published in
2017, and the second in 2021.
Both evaluations considered the extent to which the program supported participants,
families, and communities. In particular, they have observed how the Card has reduced
the harms caused by welfare funded alcohol, gambling, and drug misuse.
The initial evaluation (2017) focused on the first two program sites — Ceduna and East
Kimberley regions. The evaluation undertook primary data collection within these sites,
including face-to-face surveys with CDC participants and their family members, face-to-
face surveys of community members that were not CDC participants or family members,
and qualitative research interviews and focus groups with community leaders,
stakeholders and merchants.
The evaluation also drew from administrative data source from Indue, State and
Australian Governments. The evaluation found that the Card had a ‘considerable
positive impact’, with a large degree of support from stakeholders and community leaders
based on observed positive changes.
The second CDC evaluation (2021) considered the first three program trial sites: Ceduna,
East Kimberly and the Goldfields. The evaluation methodology combined qualitative
and quantitative approaches to examine the anticipated, and actual, outcomes from the
program. It included a quantitative survey of CDC participants with nearly 2 000 valid
responses and over 340 in-depth interviews of stakeholders and CDC participants.
The second evaluation also included a range of statistical analysis of community-level
and administrative datasets, such as crime data, using robust methods such as
multivariate analysis to understand the determinants of which CDC program participants
experienced benefits.
265
This evaluation had mixed results, and often concluded that outcomes achieved in the
program regions could not be directly attributed to the CDC. This was because of the
wide range of other programs and policies that were operating within the regions. The
evaluation did not state which programs and policies were making the biggest impact.
For the purposes of the current study, the second CDC evaluation provided useful insight
into the proportion of participants that experienced a positive or negative change, and
some of the drivers of outcomes, both positive and negative. Modelling adjustments have
been made in the CBA where necessary to account for any uncertainty.
Key findings from these two evaluations are summarised in table 2.1.
266
2.1 Summary of CDC evaluation findings
Impact
First evaluation
Second evaluation
Summary of evidence
category
Alcohol
Reduced consumption
Reduced consumption
Clear evidence of reduced consumption
■ Of participants who reported that they do drink ■ Although alcohol consumption reduced, it was not possible The second evaluation provides an estimate for the
alcohol, 41 per cent reported drinking alcohol
to attribute these changes to CDC alone
reduction in alcohol consumption by level of consumption
less frequently
risk for each program site. The First evaluation supports the
overall findings that alcohol consumption has decreased.
■ 37 per cent of participants reported binge
drinking less frequently (6 of more drinks)
Illicit drug use Reduced consumption
Inconclusive results
Inconclusive evidence indicating a change in harms
■ Of participants who reported using illegal drugs ■ No conclusions could be made about whether CDC
Both evaluations reported a decrease in illicit drugs use for
before the program commenced, 48 per cent
influenced personal or social harms caused by illicit drug
a small population.
reported using illegal drugs less often
consumption
However, the change in harms across the CDC population
■ The qualitative research identified some
■ Attempts to find evidence from other community-level data as a whole from illicit drug use cannot be determined. The
anecdotal evidence of possible reduced illicit
sources were not successful
harms associated with drug use vary considerably by the
drug use
incidence of use, the frequency of use, and the drug type.
■ However, it appears that the CDC is supporting a decline in
■ Caution is recommended when using these
illicit drug use, while noting that the evidence cannot
The incidence of personal drug use increased from 6 per
results, due to the small sample size
definite attribute this change to the program
cent in the 12 months before being on CDC to 11 per cent
since being on CDC. Of the participants who reported using
illicit drugs since being on the CDC, about 80 per cent were
not using before the CDC. However, this is considered
circumstantial as drug consumption behaviour is driven by a
range of other factors, such as the illicit drug market supply
and population factors. There is no comparator that
examines illicit drug uptake for those not on the CDC.
The second evaluation noted that 22 per cent of
participants that use illicit drugs reported daily or almost
daily use. This indicates that harms may still be occurring,
However, it does not measure changes in drug use intensity
(only frequency) or the type of drug consumed. It is possible
that the amount of use per occasion may be influenced by
the CDC.
The consultation process undertaken through this
evaluation was not able to confirm if illicit drug use had
been impacted by the program. One stakeholder noted that
people will “find a way” to source illicit drugs regardless of
the CDC program, indicating that findings in relation to drug
use are not necessarily a failure of the program.
267
Impact
First evaluation
Second evaluation
Summary of evidence
category
Gambling
Reduced consumption
Reduced consumption
Evidence indicates a reduction in gambling
■ Of participants that gamble, 48 per cent
■ There was short-term evidence that CDC helps to reduce
Both evaluations were consistent in that the CDC has
reported gambling less
consumption
helped to reduce gambling. The second evaluation limited
their findings to the “short-term”.
Through the consultation process undertaken, this finding
was supported.
Safety
Limited evidence of an improvement
Inconclusive results tending towards a worsening of safety
Insufficient evidence of a change in outcome
■ Administrative data related to levels of crime
■ Most CDC participants did not feel safer since the
Both evaluations found no reliable evidence that
generally did not show evidence of reduction
introduction of the CDC. But this finding differed
crime/safety outcomes improved.
since the implementation of the CDC. Except
substantially across sites, between men and women, and
There was some positive evidence from the second
for decreased drug driving in Ceduna, and an
between indigenous and non-indigenous people.
evaluation, but the weakness of the statistical evidence
increase in criminal incidents in East Kimberley ■ Although inconclusive, crime data suggests that domestic suggests this is unreliable for the purpose of estimating
■ There was no change in perceptions of safety
violence increased in East Kimberley and the Goldfields,
benefits in a cost-benefit analysis.
■ Community leaders, stakeholders and
drug offences increased in the Goldfields and stealing
The first evaluation reported mixed results, with some
merchants reported that violence and crime fell
increased in East Kimberley. These results are subject to
positive outcomes in Ceduna, but other worsening
strong caveats and limitations.
outcomes in East Kimberley.
Given the mixed and often statistically insignificant results,
the evidence of safety impacts is insufficient to enable
quantification.
Healthy eating Improved
Was not a primary focus
Insufficient evidence of a change in outcome
and
■ Merchant reports of increased purchases of
■ The evaluation focused on the “wellbeing” of participants,
Although there is some anecdotal evidence of people buying
purchasing
baby items, food, clothing, shoes, toys and
rather than directly on health food choices
more food, there is no evidence indicating this impact was
behaviours
other goods for children
experienced across the CDC participant population as a
■ The wellbeing assessment included a survey of self-
whole, or in a statistically significant way.
assessed wellbeing, and a question if the CDC had
impacted life quality, rather than consumption behaviours
An increase in food expenditure does not necessary indicate
an increase in healthy consumption behaviours.
■ Anecdotal evidence from stakeholders indicated that a
reduction in alcohol consumption led to participants
The second evaluation outlined that further long-term
evidence on nutrition will emerge as the program continues
spending more on food
and additional data is collected, which will be important to
develop this line of inquiry.
268
Impact
First evaluation
Second evaluation
Summary of evidence
category
Relationship
Mixed impact
Mixed impact
Both positive and negative outcomes recorded
and family
■ The participants with children reported that:
■ Most participants reported no major change regarding
Both evaluations considered multiple measures, spanning
–
from general wellbeing, school attendance, access to food,
40 per cent were better able to look after
aspects of children’s welfare. However;
etc.
their children post implementation, and 39
– a minority reported an overall positive view
per cent were more involved with their
The results were mixed across these measures. Some
– another larger minority reported an overall negative
children’s homework
participants reported improvements, while others reported
view, and
outcomes worsening. This is not unexpected given the
– 24 per cent reported that they were worse
– qualitative interviews were more positive compared to
myriad of factors that influence these outcomes, and is not
off, as they could not buy goods for their
quantitative surveys
considered to be a failure of the program.
children with cash, and
Individual analysis against these measures is needed to
– 17 per cent felt better off, as there were
determine the net impact to relationships and families.
better able to meet basic needs
Employment
Nil impact
No evidence of positive or negative impacts
■ The evaluation found no change in employment outcomes.
It presented information about employment outcomes
among the CDC cohort and barriers to unemployment.
However, there was little evidence of changes in job seeker
activity.
Financial
Improved savings and limited evidence of other
Mixed impact
No evidence of overall positive or negative impact
situation and
outcomes worsening
■ Most participants (75 per cent) reported no change in their
Across the broad range of financial metrics there was little
literacy
■ Community leaders, stakeholders and
financial situation. Among those that experienced a
consistency in terms of outcomes improving or worsening.
merchants felt that the CDC had positive
change, more participant experienced outcomes
There were some participants that experience benefits, and
financial impacts
worsening, compared to those that experienced outcomes
others experience additional financial barriers. Again, this is
not unexpected given the myriad of factors that influence
■ 45 per cent of participants reported being
improving
these outcomes, and is not considered to be a failure of the
better able to save money
■ The most vulnerable participants benefited from financial
program.
■ Across a range of financial indicators there was
planning and money management. That is, finances were
little difference between Wave 1 (a few months
made easier for those most affected by harmful behaviour
post-CDC implementation) and Wave 2 (9
of others and those in the severest forms of financial
months later), but there were statistically
hardship
significant increases in the share of
participants that had run out of money to pay
for school supplies or non-food essentials for
children
269
Impact
First evaluation
Second evaluation
Summary of evidence
category
Wellbeing
Mixed impact
Negative impact
Participants reported a negative impact on their wellbeing
■ 32 per cent of participants reported that the
■ A large proportion of participants reported a negative
Across both evaluations, participants reported that the CDC
program had made their lives worse
impact on quality of life
program “made their lives worse”.
■ 23 per cent felt that the program made their
■ Non-indigenous participants were more likely to report that Although some participants stated that their lives improved,
lives better
the CDC made their life worse (69.2 per cent, compared to
these responses were outnumbered by those experiencing
negative impacts.
■ 4 per cent raised stigma or shame as a
48.5 per cent for indigenous participants)
negative impact from the Card
■ 14.5 per cent of participants reported an overall positive
This is not unexpected given the purpose of the CDC is to
impact
restrict welfare recipient’s access to cash, and the response
to this question is not sufficient to determine impacts on
■ Most participants highlighted feelings of discrimination,
participant wellbeing.
embarrassment, shame, and unfairness because of being
on the Card. Only a small minority of CDC participants did
not report any of these negative feelings about the CDC.
Data source: Mavromaras K., Moskos M., Mahuteau S., Isherwood L. 2021 ‘
Evaluation of the cashless debit card in Ceduna, East Kimberley and the Goldfields region: Consolidated report’, Future of employment and skill
research centre, The University of Adelaide, and ORIMA research 2017, Cashless Debit Card Trial Evaluation: Final Evaluation Report.
270
Other studies have been more qualitative in nature, with key findings summarised in
table 2.2.
2.2 Other evaluations and reviews of the CDC
Report
Positive impacts
Negative impacts
Mavromaras, K.,
■ Technology advances such as ability to ■ Not possible to distinguish between impacts
Moskos, M.,
use as a normal bankcard, reducing
of concurrent policing and alcohol
Isherwood, L., and
stigma
interventions, as well as seasonal
Mahuteau, S., 2019,
■
influences in the region and the CDC roll out
Cashless Debit Card
Key groups of people with known drug
Baseline Data Collation
and alcohol problems who commonly
■ Not suitable for people with disability and
in the Goldfields
reported increasing their spending on
their carers
Region: Qualitative
food, clothes, and essentials
■ Exemption process is available but
Findings
■ Small cash component of the CDC was
complicated
perceived to provide incentive for
■ Loopholes and evasive tactics were
participants to seek work
inevitable (most commonly gift card
■ Those participants that had positive
purchase)
impacts were people with previous
experience on income management
and those who were already
technologically literate and already
good with finances
■ Positive responses were ideally about
financial management
■ Having a positive effect on the
prevalence and severity of crime, family
violence and anti-social behaviour
■ Reduced “humbugging” as everyone on
CDC have restricted access to cash
■ Impact on alcohol consumption
reduction more pronounced than
substance abuse
Marston, G. et al.,
■ Difficulty providing for children in family due
2020, Hidden Costs:
to reduced Cash
An Independent Study
■
into Income
Inability to partake in second-hand market
Management in
and cash economy reduced consumer
Australia
choices, and various examples were
provided by participants
■ Difficulty paying rent and bills as exemption
process complicated
■ Stigma and Shame around welfare cards
■ A lower level of self-control among IM
participants can make transitions from
welfare to work more difficult
■ Resisting IM legally through exemptions or
finding loopholes
271
Report
Positive impacts
Negative impacts
ANAO, 2018, The
■ The ANAO noted that the CDC was rolled out
implementation and
more widely based on learnings from the
performance of the
program despite the program not being
cashless debit card
designed to test scalability
trial. Auditor-General
Report No.1 2018-1
■ The monitoring and evaluation were
inadequate and as such it was not possible
to conclude whether there has been a
reduction in social harm, and whether the
Card was a lower cost welfare quarantining
approach
Hunt, J., 2017,
The
■ Kununurra data around reported assaults
cashless debit card
rising sharply in line with the CDC program,
trial evaluation: Does it
with no explanation of the impact of the
really prove success?
CDC or other contributing factors
CAEPR Topical Issue
No. 2/2017
Hunt, J., 2020,
■ Need for Support Services to
■ Structural poverty appears to be a feature
Evaluating the
complement CDC reiterated
of the region, focus should be on promoting
Cashless Debit Card:
job opportunities not punishing the poor
How will it solve
■
poverty and
Would not prevent people looking for
unemployment
loopholes especially for the severely
addicted
■ CDC was poorly targeted as many on
income support are not gambling or using
drugs
■ Some people responding well to the Card in
Bundaberg and Hervey Bay while others
struggling
Source: CIE and other studies as noted.
The Methodology chapter below outlines how these previous evaluations have been used
to inform the CBA.
272
3 Methodology
This study is the first Cost Benefit Analysis (CBA) of the CDC.
To guide the CBA, an economic framework was developed by the CIE in 2020 in close
collaboration with DSS.19 This framework has been an important document in
directing the key areas for analysis and investigation (base case, benefit selection,
impact measurement, etc).
While the framework identified the full range of potential benefits, the CBA has been
partially informed by the second CDC impact evaluation. This evaluation provided
some of the evidence base on the impact from the program up to June 2020.
Additional analysis was also undertaken on the change in consumption and
expenditure behaviours and employment outcomes.
All impacts identified in the economic framework were analysed, although not all had
a sufficient evidence base and conclusive results to enable quantification in the CBA.
Where benefits cannot be quantified, they are discussed qualitatively.
Building from the second impact evaluation’s findings
With the release of the CDC program’s second impact evaluation in 2021, there was an
opportunity to apply the reported findings and impacts as inputs into this CBA.
The second evaluation is the most recent evidence base, and considered a wider range of
sites and participants, analysed a wide range of community-level and administrative
datasets, and provided findings from consultations with a substantial number of
participants and other stakeholders. It has therefore been the key source of evidence used
to support this CBA.
Not all potential impacts identified within the Economic Framework were considered in
the second impact evaluation, such as employment outcomes and impacts from a change
in consumption. To fill this evidence gap, this analysis draws on DSS Data Over
Multiple Individual Occurrences (DOMINO) dataset and CDC program data to evaluate
if the CDC has led to these additional benefits.
19 The CIE, 2020,
Economic framework for cost-benefit analysis of the Cashless Debit Card Trial, Final
Report, November 2020.
273
Second impact evaluation approach
To determine the impact of the CDC program in the first three program sites (Ceduna,
East Kimberley, and the Goldfields), the second impact evaluation analysed both
qualitative and quantitative data.
Qualitative data was gathered from in-depth interviews with stakeholders (178) and CDC
participants (231).
■
Interviews with stakeholders were used to gather perceptions about the perceived
impacts of the CDC and the perceptions regarding the future of the CDC.
■
Interviews with participants gained information about people’s views about the CDC
program and perceptions of its impact on their lives and their community. People
interested in participating in an interview either contacted the research team directly
or consented to have their contact information provided to the research team by their
stakeholder organisation. 20 per cent of these interviews were with family members of
another CDC participant.
Quantitative data was sourced through a large-scale survey of CDC participants in the
three trial sites (1 963 valid responses). The evaluation also sourced Australian
Government and state government administrative data.
■
Survey data was the main quantitative source of information on outcomes gathered by
the evaluation. The survey collected data on the participant’s demographic
information, employment status, financial position, behaviour and attitudes towards
alcohol and drugs, health, feelings about being on the Card and about the community.
■
The administrative data provided by the Australian Government included CDC
program data provided by the DSS, and administrative data from the Card provider.
■
Although the evaluation considered state government data, this was found to be
mostly unsuitable for the purposes of the evaluation, with the exception of Police data
provided by Western Australia and South Australia.
Limitations from evidence base within this CBA
The data collected through the second impact evaluation is a robust source of data on the
impact of the CDC. However, its methodology was designed with a different purpose to
this CBA. Because of this, there are some specific limitations that need to be noted:
■
Through the consultation process undertaken as part of this CBA, some stakeholders
mentioned that their customers/CDC participants are far more likely to report
negative news, rather than positive news. This is expected to be true when talking
about their experience on CDC, with participants more likely to report issues, rather
than any positive benefits they have experienced while on the Card. This may have
also been the case with the stakeholder consultations undertaken through the second
impact evaluation, potentially creating a bias in the results, and potentially
understating the benefits created.
■
The evaluation relied on the survey results to drive much of the quantitative analysis
on outcomes. Some of these outcomes are not necessarily reflective of the actual
impact. For example, the survey asked participants if they felt there had been changes
to safety and general health. Although they may have reported a change in these
274
outcomes, measures for safety (such as instances of domestic and family violence,
thefts, etc.) or actual health outcomes may differ from the self-reported responses.
■
The participant survey provided data about the change in outcomes, such as if an
outcome got worse, the same or better. However, it did not provide a measure or a
value of the change. For example, if a participant stated that school attendance for
their children had improved, the survey did not indicate how many days school
attendance changed (i.e. a one day improvement or a 10 day improvement). To
inform the CBA, other evidence sources were combined to the second impact
evaluation, such as other survey data and literature sources to estimate the impact.
■
Through the consultation process undertaken as part of this CBA, some stakeholders
mentioned that there is significant social pressure for people to not support the Card.
If a CDC participant were to publicly state that they benefited from the Card or
supported it, then there may be social, community, or family backlash. These
stakeholders suggested that even through a participant may have benefited from the
Card, they are unlikely to say so. However, the extent to which this is occurring
cannot be determined. If prevalent, this may create a bias in the second impact
evaluation results, which may have understated the benefits created or over stated
negative impacts.
■
Many of the results are inconclusive. The second impact evaluation suggests that
some participants benefit from the Card, some receive no change, and some
experience negative impacts. To accommodate for these mixed impacts, in places, this
CBA has taken the net impact, i.e. the difference between the proportion of
respondents who experienced a positive impact and those that experienced a negative
impact. By doing so, we have assumed that the benefit received is of equal magnitude
or value to the negative impact. However, this may not be true if the benefit generated
are more significant than the negative impacts felt by others.
■
The second impact evaluation noted that there are many concurrent polices and
initiative operating in the CDC sites. This creates an issue of attributing the impacts to
the CDC program.
■
The second impact evaluation only covers the first three CDC program sites, which
are Ceduna, East Kimberley, and the Goldfields. Accordingly, neither the survey nor
the analysis of community-level and administrative data covers Bundaberg and
Hervey Bay. Because of this, we have assumed that the impacts to Bundaberg and
Hervey Bay as consistent with the average impact across the other three regions.
275
Cost-Benefit Analysis approach
The key steps in a cost benefit analysis are:
■
establishing the base case
■
quantifying the changes from the base case
■
placing values on the changes
■
generating the Net Present Value (NPV) of the future net benefits stream, and
■
undertaking sensitivity analysis to test key assumptions and inputs.
This type of analysis measures the costs and benefits to a range of stakeholders, including
Government, community, participants, and families.
A range of overarching assumptions are relevant for conducting cost-benefit analysis,
which are specified in the Cost-Benefit Analysis Guidance Note published by PMC.20
Choices relating to these overarching assumptions are summarised below:
■
Definition of the base case: This cost-benefit analysis of the CDC program is an
ex-
post analysis, meaning that it is backward-looking. In general, the base case for cost-
benefit analysis should be a ‘do nothing’ or ‘business as usual’ option. For the purpose
of this analysis, the base-case is a scenario where the CDC Program was not
conducted in the Goldfields, East Kimberley, Ceduna, and Bundaberg and Hervey
Bay.
– The CDC program was implemented at a time during which other policy
interventions were taking place. Further, the CDC program is accompanied by an
expansion of Support Services provided to welfare recipients. Because the scope of
this cost-benefit analysis is the CDC program, and not the concurrent policy
interventions, these other policies and the expansion of Support Services are
considered to occur under the base case. However, it is difficult to disentangle
some of the impacts of the CDC from potential impacts of these concurrent
interventions, which is discussed throughout this report. Nonetheless, the objective
is to measure the incremental costs and benefits of the CDC program only.
■
Defining the range of options: Only one option is considered, which is the option
that was taken to have the CDC program in each location. While generally cost-
benefit should consider multiple options, for this purpose of this
ex post analysis to
assess the merits of the chosen policy, we will only consider the CDC program as-
implemented.
■
Over what period do we measure impacts: We only consider the use of the CDC
program in the initial four regions until the end of 2019/20, because this is the period
for which CDC program cost data has been provided by DSS. However, the CDC
program is expected to have impacts for participants and others over a longer period.
Many of the benefits included within the CBA estimate the lifetime impact from the
20 Department of the Prime Minister and Cabinet, 2020,
Cost-benefit analysis — guidance note,
March 2020, available a
t: https://pmc.gov.au/sites/default/files/publications/cost-benefit-
analysis_0.pdf
276
program.21 For example, we measure the avoided loss of productivity over the
lifetime of people that die due to alcohol misuse, and the life-long impacts for children
from improved health outcomes, healthier food consumption, and increased school
attendance. These life-long impacts are discounted to present values.
■
Whose costs and benefits count: For the purpose of this analysis, measuring national
costs and benefits is appropriate, and there are unlikely to be any relevant
international impacts. Costs and benefits to all people residing in Australia will be
included if they can be estimated.
■
How do we discount costs and benefits: To compare costs and benefits occurring at
different points in time, it is necessary to convert the value of future costs and benefits
to an equivalent value received immediately. This is referred to as ‘discounting’, and a
discounted value is referred to as the present value of a future cash flow. To estimate
the present value, future values are multiplied by a factor reflecting a specified rate of
return over time, in this case, the social discount rate.22 The higher the social discount
rate, the more the future cash flows will be discounted, resulting in a lower present
value.
– The value of costs and benefits in each past and future year are discounted to a
base year of 2015/16, which is the year that the first trials in Ceduna and East
Kimberley commenced.
– A real discount rate of 7 per cent is used for the analysis, with sensitivity testing of
3 and 10 per cent. These rates are consistent with guidance from the Department of
Prime Minister and Cabinet about discounting in cost-benefit analysis.23
– The nominal value of costs and benefits has been converted to real values using a
price year of 2020 (the most recent year for which GDP deflators are available
from the ABS).
Benefits from improved outcomes and from changed consumption
We split potential benefits of the CDC into two categories:
■
benefits where there is evidence of an improvement in outcomes, such as an
improvement in school attendance or reduction in social problems associated with
gambling, and
■
benefits where there is evidence of a change in consumption patterns, such as a fall
in consumption of alcohol, drugs or gambling.
21 We have not excluded any quantifiable benefit categories on the basis of them accruing after
the end of 2019/20.
22 The discount rate is the rate used to determine the present value of future cash flows. By
discounting future cash flows to today’s value, the CBA accounts for the opportunity cost of
the cash flows. I.e. the consumer preference, consumption benefit, and financial benefit from
receiving a dollar today rather than a dollar in the future. Discounting future cash flows also
allows a true comparison of current and future cash flows.
23 Department of the Prime Minister and Cabinet, 2020,
Cost-benefit analysis — guidance note,
March 2020, available a
t: https://pmc.gov.au/sites/default/files/publications/cost-benefit-
analysis_0.pdf
277
The reason to segment these benefit types is that, where there is evidence of an
improvement in outcomes, we can value the change in outcomes directly. However,
where there is only evidence of changed consumption patterns, we must rely on evidence
from the literature about the relationship between consumption patterns and outcomes,
and, in turn, benefits.
The key example of this is that the second impact evaluation provides evidence of a
change in alcohol consumption, however it does not provide evidence on the magnitude
of the change, nor does it indicate if the participant has experienced a change in alcohol
related harms. In this example, we have relied on other literature sources to estimate the
change in harms and expected benefits achieved.
Measuring impacts of the CDC on outcomes
To assess the outcomes of program participants relative to the base case, we must
compare their realised outcomes with the CDC to a comparator. The following
comparator groups are variously available, broadly ordered in terms of the robustness of
the comparison for inferring the impacts of the CDC:
■
welfare recipients not participating in the program but in a comparable location
■
CDC participants during the period prior to their participation in the program
■
welfare recipients not participating in the program across any location, or
■
other Australians.
Data about outcomes for program participants and other welfare recipients are more
readily available than data about spending. The relevant data sources are:
■
the DSS Data Over Multiple Individual Occurrences (DOMINO) dataset24
■
Mavromaras K., Moskos M., Mahuteau S., Isherwood L., (2021)
Evaluation of the
Cashless Debit Card in Ceduna, East Kimberley and the Goldfields Region, prepared by
University of Adelaide, and
■
other evaluations of the CDC program, such as the ORIMA evaluation.
The differences in outcomes, such as the incidence of alcohol or drug-related illness
between participants and non-participants, will reflect the impact of changed spending
patterns and other concurrent policies, such as increased provision of counselling and
financial literacy classes.
A key issue for understanding the benefits of the CDC is identifying correlation versus
causation. Participation in the CDC program may be correlated with better outcomes,
such as lower crime rates, but this may be due to selection of program locations or other
factors unrelated to the activities of the CDC program.25 The Australian Government
24 Key aspects of this dataset are described at
: https://www.aihw.gov.au/about-our-data/our-
data-collections/department-of-social-services-data-over-multiple-i
25 For example, the CDC program may result in a higher degree of outcomes monitoring,
engagement by the local community and other outcomes that are not related to the CDC itself
but can lead to benefits. It is difficult to disentangle such benefits from the benefits of the CDC
itself as a compulsory income management tool.
278
selected locations for the CDC on the basis of them having high levels of antisocial
behaviour. Further, the community at the program sites had to request the program after
a consultation period, so a site could only be in response to a recommendation by an
inquiry or inquest.26
Statistical modelling approaches such as multiple regression modelling27 are useful to
disentangle the impacts of the CDC from impacts due to demographic or other
differences between participants and non-participants. These statistical modelling
approaches are used in the second impact evaluation to support causal identification.
However, in a range of cases, statistical modelling approaches produce inconclusive
results. In these cases, survey statistics relating to perceptions of various CDC impacts
are sometimes the best available evidence. To inform the benefit calculations, we have
drawn from the second impact evaluation’s statistical modelling and survey statistics.
Flow-on impacts
As shown in the program logic for this analysis, the benefits of the CDC are often
interrelated. For example, health benefits would lead to economic benefits, since
healthier people may be more likely to participate in the labour force. Similarly, reduced
welfare dependency may result in long-term improvements in educational participation
of children, reduced crime and improved housing security.
Many of these impacts may be very long term, and thus not observable in the data for the
CDC program to-date. For example, improved education of the children of participants
would take many years to result in greater productivity within the community. Children
may also move out of the community at a later stage, making it difficult to measure
changes in their outcomes.
Where relevant, we have specified the ‘flow-on’ benefits that accrue for each benefit
stream and where flow-on benefits are likely but have not been quantified. For example,
the avoided costs of alcohol misuse are often ‘flow-on’ costs, such as traffic accidents as a
result of drink driving.
Estimating changes in spending on restricted items
The key difficulty with estimating changes in spending patterns for program communities
in that we do not observe spending patterns under the base case. This is because CDC
spending data is only available for participants.
We can compare the CDC spending patterns to spending from other data sources for
similar groups, but because this will rely on different datasets the comparison will not
26 Department of Social Services, 2015,
Cashless Debit Card Final Assessment Regulation Impact
Statement, p2.
27 Multiple regression modelling is a type of statistical modelling that aims to predict the value of
a variable based on the value of two or more other variables. For example, it may try and
predict the level of crime in local areas based on the number of CDC program participants in
the region, the number of non-participants, and demographic characteristics of the population
in each area.
279
involve formal statistical testing, which would give a false sense of accuracy to such a
comparison.
Appendix A summarises the data available about spending of non-participants, and
limitations in comparing this to spending data from the CDC. Given the limitations to
identifying changes in spending on alcohol and other restricted items, the ability to draw
conclusions from this data about spending patterns is limited.
Hence, we have placed greater reliance on evidence from the second impact evaluation
about changes in spending patterns, which relied on a survey of participants. The survey
instrument called for respondents to report current spending patterns (i.e. after the
implementation of the CDC) and what changes they believe occurred since the
implementation of the CDC, such as a decrease in frequency or amount of alcohol
consumption.
Estimates of the benefits from changes in consumption patterns are based on evidence
from the academic literature, since it has not been possible to estimate the relationship
between individual spending patterns (based on Card data) and outcomes. An example of
such a relationship would be a link between the amount an individual spends at
supermarkets and their employment outcomes. Even if such a relationship was confirmed
by the data, it would likely reflect correlation rather than causation, and would be very
weak evidence of such a benefit.
The academic literature includes studies that estimate the societal cost of alcohol misuse,
gambling, and illicit drugs for Australia. Applying benefit estimates from other
studies/contexts/areas to a different situation is referred to as ‘benefit transfer’. The
accuracy of results estimated using benefit transfer will depend on the closeness of the
context, time period, demographic characteristics, and a range of other factors between
the source study and the current study. We discuss the appropriateness of applying
benefit estimates from the literature in this report, including the uncertainty associated
with application of these estimates.
Consultations
Consultations with a range of stakeholders were necessary to capture evidence of what
happens ‘on the ground’. These consultations helped align the CBA modelling inputs
gathered through data analysis and literature review to the lived experiences of those who
interact at a personal level with CDC program participants.
Organisations were identified to participate in the consultation process by:
■
an initial review of the list of CDC stakeholders, provided to CIE from DSS
■
a comparison of each stakeholder to the number of CDC participants and the volume
of activity for each organisation. Organisations were only considered for consultation
if they have sufficient CDC activity
■
the distribution of these organisations across each program region was then
considered to ensure that there was a wide geographical spread, and
280
■
when there were multiple potential organisations, only five star providers were
shortlisted.28
The shortlisted organisations were then invited to a discussion with CIE. The discussions
were intentionally flexible and focused on key topics of interest.
These discussion helped to:
■
to gather further evidence on the qualitative benefits of CDC, building from the
analysis already undertaken through the impact assessments
■
to validate our CBA framework and the finding in the data analysis against the
experiences ‘on the ground’, and
■
to refine and test our CBA modelling assumptions.
Some high level comments and examples from these case studies have been highlighted
throughout the report. Although not statistically significant, and cannot be relied upon to
make population level findings, these examples help to build on the narrative and provide
some specific examples of the impact of the CDC program.
They are not representative
and cannot be used for making general statements about the impact to the CDC
participant population.
28 The Department of Education, Skills and Employment calculates the relative performance of
jobactive providers and rates each service out of five stars. Five star providers are 30 per cent or
more above the national average performance (after accounting for differences in participants
and labour market characteristics).
281
4 Changes in expenditure patterns
The CDC directly influences expenditure patterns of participants, in particular
expenditure is directed way from alcohol and gambling. This is supported by the
Card’s transaction data, and through the reported impacts from program participants
in the second impact evaluation.
Our analysis of spending using the CDC transaction data suggests that attempted
purchases of restricted items such as alcohol using the CDC are found to occur in a
volatile fashion. Declined transactions remain frequent even among those who have
used the CDC for an extended period. In CDC program locations, there is an upward
trend in the share of transactions that are declined due to attempted alcohol
purchases.
This partly reflects the design of the CDC, which does not prevent participants from
purchasing alcohol, but it does suggest there has not been a ‘learning by doing
effect’. In other words, participants still attempt to buy alcohol products after their
attempted alcohol transactions have been blocked a few times, and after the
participant is more familiar with the features of the Card. This was supported by one
stakeholder who stated that participants are likely to continue to ‘test’ the Card at
multiple vendors to see if a transaction is approved.
Overview of aggregate CDC transaction data
DSS have supplied monthly aggregate CDC Program data. The data supplied to the CIE
consists of the following for each CDC program site:
■
outgoing transactions
– counts of ongoing transactions and the number of unique payers, and
– total value of transactions by Merchant Category Group (defined by DSS)
■
incoming transactions
– counts of outgoing transactions and the number of unique recipients, and
– total value of incoming transactions
■
declined transactions
– number of declined transactions by reason, and
– total value of declined transactions by reason.
This section provides an overview of trends in outgoing and declined transactions, and
compares spending shares by type of good/service to data from the ABS.
282
Number of payers and recipients
The number of cards for which there was an application to make a payment in each
month is referred to as the number of applicants (chart 4.1). The number of applicants
tends to rise gradually from the point that the CDC program is introduced in each region,
with a higher rate of take-up during the early months of 2020. This is likely associated
with the COVID-19 pandemic and associated lockdowns, which is likely to have
increased the reliance on income support of people within the program regions.
In response to the COVID-19 pandemic, a ‘paywave’ functionality was included on the
CDC at the end of July 2020. This meant that a number of new cards were issued to
CDC participants, resulting in the number of ‘applicants’ (i.e. number of unique cards
with an attempted transaction) spiking in July 2020. The spike therefore shows an
increase in the number of cards, rather than the number of people.
4.1 Number of applicants for payments
18
Ceduna
East-Kimberley
Goldfields
Bundaberg and Hervey Bay
16
's) 14
0
0
12
10
icants (0
pl
8
r of ap
6
be
um
4
N
2
0
r
n
g
r
n
g
r
n
g
r
n
g
r
n
g
Ap
Ju
ec
ec
ec
ec
ec
Au
Oct
D
Feb
Ap
Ju
Au
Oct
D
Feb
Ap
Ju
Au
Oct
D
Feb
Ap
Ju
Au
Oct
D
Feb
Ap
Ju
Au
Oct
D
2016
2017
2018
2019
2020
Data source: CDC Program Data, CIE.
We consider applicants rather than recipients because recipients that do not issue
payments would not be considered to be ‘using’ the Card and driving the benefits from its
use. However, the number of applicants and recipients in each month is very similar (e.g.
in Ceduna since April 2016, there has been an average of 843 payers and 844 recipients in
each month).
Outgoing transactions using the CDC
The number of outgoing transactions per unique payer is somewhat seasonal, and
gradually increasing (table 4.2). There was also a step change in use in 2020, likely
associated with COVID-19 and potentially some substitution from cash to card
transactions during lockdowns.
There may have also been impacts on transactions because of additional income support
related to COVID-19, translating into higher spending. For instance, during COVID-19,
283
JobSeeker payments received a supplement amount of $550 per fortnight, before being
scaled back to an increase of $250 in September 2020.
The number of outgoing transactions per unique payer is higher for sites that initiated the
CDC program later, with Ceduna being the lowest and Bundaberg and Hervey Bay being
the highest.
4.2 Number of outgoing transactions per CDC payer
Ceduna
East-Kimberley
Goldfields
Bundaberg and Hervey Bay
45
er 40
pay
35
per 30
ions
25
sact
20
tran 15
going 10
y out
5
hl
0
ont
r
n
g
b
r
n
g
b
r
n
g
b
r
n
g
b
r
n
M
Ap
Ju
ec
ec
ec
ec
Au
Oct
D
Fe
Ap
Ju
Au
Oct
D
Fe
Ap
Ju
Au
Oct
D
Fe
Ap
Ju
Au
Oct
D
Fe
Ap
Ju
2016
2017
2018
2019
2020
Data source: CDC Program Data, CIE.
Spending shares compared to ABS data
A key impact of the CDC is that it is expected to change spending patterns. Most
obviously, spending by participants on restricted types of goods and services is expected
to reduce relative to spending by these participants if they were not participating in the
CDC program. Spending on other goods and services would experience a corresponding
increase, and this may be larger for some types of goods and services (e.g. fresh food)
than other types. Alcohol expenditure tends to be a substitute for expenditure on
necessary goods and services (Pu et al., 2008)29, such as utilities, food or health services.
Spending by participants must be compared to the counterfactual, which is spending by
participants if they had not been in the CDC program. The main counterfactual for this
CBA is the price weight series from the Selected Living Cost Indexes (SLCIs) publication
by ABS.30 The SLCIs provide a measure of the cost of living for each of four types of
households. To do this, they need a separate set of weights for each household. The
'other government transfer recipient' household category refers to all households whose
principal source of income is a government pension or benefit other than the age pension
or veterans affairs pension.
29 Pu, C., Lan, V., Chou, Y. and Lan., C., 2008, ‘The crowding-out effects of tobacco and alcohol
where expenditure shares are low: analysing expenditure data for Taiwan,
Social Science and
Medicine, 66(9), pp.1979-1989, available a
t: https://pubmed.ncbi.nlm.nih.gov/18313191/
30 This is further discussed in the Economic Framework report.
284
However, this comparison of CDC expenditure to the SLCIs accounts only for the CDC
participant expenditure that is on the Card. Since 20 per cent of participant’s income
support payments can be deposited into the participant’s nominated account, this is not
quite a complete list of all transactions. This has the impact of potentially understating
some of the transaction groups for the CDC participants.
Comparing the spending shares of CDC program participants and those for ‘other
government transfer recipients’ from the SLCIs reveals the following (chart 4.3):
■
Food spending is significantly higher among CDC participants. The food category in
the CDC spending dataset represents around 30 per cent of total spending across the
entire time period of the program across the initial four regions. In contrast, the food
category accounts for slightly more than 15 per cent of spending in the comparison
group.
■
If non card-based transactions in the CDC spending data are assumed to be entirely
housing-related payments, then
spending on housing is very similar between the
CDC data and SLCI weights. Through discussions with stakeholders, it was
confirmed that non-card-based payments include some direct debit transactions, such
as rental payments, but also other direct debit transactions that may not relate to
housing.
■
Transport and recreation spending (which includes eating out in the CDC dataset)
are relatively similar between the SLCIs and the CDC data.
■
Clothing and footwear spending is higher in the CDC data than the SLCIs. This
may suggest that on average, access to these goods and services is worse than in other
areas such as urban locations, and that prices are higher as a result for this type of
good.
■
All other remaining categories are difficult to align between the CDC data and SLCIs,
or are very different in magnitude. For example, health and education spending are
fare lower in the CDC data, which may reflect lower availability of these services in
non-urban area such as the program locations. Services and other spending in the
CDC datasets may be components of the insurance, financial services and
communication categories, but these cannot be easily aligned between the CDC and
SLCIs to facilitate accurate comparison.
285
4.3 Comparison of CDC spending to spending by welfare recipients across Australia
CDC recipient data
Cost of living weights
Housing (+DSS non Card-based)
Transport
Food and non-alcoholic beverages
Recreation and culture (inc. eating out)
Clothing and footwear
Health
Education
Furnishings, household equipment and services
Insurance and financial services
Alcohol and tobacco
Communication
Other
Services
0
5
10
15
20
25
30
35
Share of spending (per cent)
Data source: ABS SLCIs, CDC program data, CIE.
We cannot draw a conclusion from this data about whether spending on alcohol, other
drugs or gambling has fallen as a result of the CDC. This is because 20 per cent of
income support payments are unrestricted for participants, and there are other means of
avoiding the restrictions such as asking non-participants to purchase restricted items in
exchange for other goods/services.
However, the SLCI weights suggest that spending on alcohol and tobacco is around 8 per
cent of total spending, which suggests that average spending on alcohol and tobacco
could be maintained by a CDC program participant. Low income households sometimes
spend more than 20 per cent of income on alcohol (chart 4.4), and these households may
be more likely to have health or other costs associated with problematic consumption of
alcohol. Therefore, the restriction of spending to, at most, 20 per cent may still be
effectively reducing alcohol-related harm.
286
4.4 Distribution of household alcohol spending
Data source: Jiang, Livingston and Room (2015).
4.5 Comments from stakeholder consultations – Reduced need for Christmas
hampers and return to school support
Some community services provide disadvantaged community members with hampers
during Christmas and additional financial support at the start of the year when
children are returning to school.
One stakeholder mentioned that since the CDC program was implemented in their
region, the number of community members needing this support has dramatically
decreased, to the point where they no longer need to provide this additional support.
This stakeholder attributed this to CDC participants now having the financial capacity
to better provide for their families during festive times and for school expenses.
Trends in spending shares
Consultations with DSS staff suggest that a potential impact from the CDC to investigate
is a shift away from spending on types of goods and services that are unhealthy towards
goods and services that are more healthy (based on the substitution effects like those
discussed in Pu et al. 2008). For example, participants may be spending more on the
‘food’ category, which we expect primarily includes food from supermarkets, and less on
other categories such as eating out.
This potential impact was further explored through the stakeholder consultation process.
Although stakeholder could not definitely say if CDC participants were spending more
on healthy food, but there were some specific examples of where this was the case.
Such changes in spending patterns may occur immediately once someone commences
the CDC program, since they are immediately unable to purchase restricted items.
However, there may also be more gradual changes in spending if behaviour changes
287
more gradually. It is worthwhile to assess if the data suggests any gradual shifts in
behaviour that would be expected impacts of the CDC.
There is little evidence in the data of increases in the share of spending on food. Spending
on food on the Card has represented a steadily declining share of spending over time
(chart 4.6). In other words, a lower proportion of CDC program participant funds is
being spent on food, and instead is spent on other types of goods and services.
The decline in spending on food could be attributable to a broader shift away from food
spending due to factors affected both CDC participants and non-participants, such as an
increase in prices of fresh food.
4.6 Share of spending on food
Ceduna
East-Kimberley
Goldfields
Bundaberg and Hervey Bay
60
t)
r cen 50
pe
od ( 40
30
ng on fo
ndi 20
pe
f s
10
hare o
S
0
r
n
g
r
n
g
r
n
g
r
n
g
r
n
Ap
Ju
ec
ec
ec
ec
Au
Oct
D
Feb
Ap
Ju
Au
Oct
D
Feb
Ap
Ju
Au
Oct
D
Feb
Ap
Ju
Au
Oct
D
Feb
Ap
Ju
2016
2017
2018
2019
2020
Data source: CDC Program Data, CIE.
All four regions saw a decrease in the share of spending on food from March 2020. This
was an impact of the additional COVID-19 supplement payments, which increased the
total income received by participants.
4.7 Comments from stakeholder consultations – Increased food expenditure
The stakeholder consultations provided anecdotal evidence that some CDC
participants have been spending more on food.
Some families who previously purchased only a few food items have now been seen
with ‘full trollies’. The example highlights that there are instances where the CDC is
making a significant impact for families. Stakeholders saw this as a significant benefit
from the CDC.
Although evidence was provided on specific instances/examples, stakeholders were
unsure if this impact was widespread, or experienced by a few.
The reduction in spending on food is driven by an increasing share of spending on ‘non-
card-based transactions’ (chart 4.8). Bundaberg and Hervey Bay did not experience the
288
same increase in non-card-based transactions. If these non-card-based transactions are
primarily rent payments, this suggests that participants may be spending an increasing
share of income on rent, which may result in greater housing benefits such as more
secure tenancies.
4.8 Share of spending on non card-based transactions
Ceduna
East-Kimberley
Goldfields
Bundaberg and Hervey Bay
45
40
-based
) 35
cent 30
(per 25
ng on non-card
ions 20
ndi
sact 15
pe
f s
tran 10
5
hare o
S
0
r
n
g
r
n
g
r
n
g
r
n
g
r
n
Ap
Ju
ec
ec
ec
ec
Au
Oct
D
Feb
Ap
Ju
Au
Oct
D
Feb
Ap
Ju
Au
Oct
D
Feb
Ap
Ju
Au
Oct
D
Feb
Ap
Ju
2016
2017
2018
2019
2020
Data source: CDC Program Data, CIE.
In Ceduna, spending shares for other merchant types fluctuate and exhibit some
seasonality, but remain relatively consistent over the period (chart 4.9). The key source of
variation is in the ratio of food to non-card-based transactions.
4.9 Spending shares for Ceduna
Food
Non Card-Based Transactions
Transport - Private
Services
Recreation - Eating out
Department, Discount and Variety Stores
Housing
Other
100%
90%
t)
80%
r cen
70%
ng (pe 60%
ndi
50%
40%
onthly spe
30%
f m
20%
hare o
S 10%
0%
n
g
n
g
n
g
n
g
n
Apr
Ju
ec
ec
ec
ec
Au
Oct
D
Feb
Apr
Ju
Au
Oct
D
Feb
Apr
Ju
Au
Oct
D
Feb
Apr
Ju
Au
Oct
D
Feb
Apr
Ju
2016
2017
2018
2019
2020
Data source: CDC Program Data, CIE.
289
Similar trends as those in Ceduna occurred in the Goldfields and East Kimberley regions.
However, Bundaberg and Hervey Bay exhibit a different trend, with the decrease in the
share of spending on food attributable to a rise in spending on the ‘other’ category of
merchants (chart 4.10).
4.10 Spending shares for Bundaberg and Hervey Bay
Non Card-Based Transactions
Food
Department, Discount and Variety Stores
Recreation - Eating out
Services
Recreation - Goods and equipment
Housing
Other
100%
90%
t)
80%
r cen
70%
ng (pe 60%
ndi
50%
40%
onthly spe
f m 30%
20%
hare o
S 10%
0%
l
ar
r
r
ay
n
g
v
ec
n
ar
ay
n
Feb
M
Ap
Ju
M
Ju
Au
Sep
Oct
No
D
Ja
Feb
M
Ap
M
Ju
2019
2020
Data source: CDC Program Data, CIE.
Statistical modelling of spending shares
We have estimated a statistical model of the share of spending on each product category.
We estimate a multiple regression model, which is an approach to disentangle multiple
influences on a variable of interest. In this case, we aim to predict spending on each
category of good and service (e.g. food) over time, based on seasonal patterns and trends
over time for each category of spending.
This modelling confirms that there is a downward trend in food spending, and upward
trend in the value of non-card-based transactions (table 4.11). The model estimates an
annual trend in spending share for each type of good and service, which can be positive
or negative. It also estimates a confidence interval for each estimate, which gives a lower
and upper bound representing the uncertainty associated with the estimate. Some of the
trends are not statistically significant, which means there is insufficient evidence to
conclude they are different from zero. If the ‘p-value’ estimated for each trend is less than
the conventional threshold of 0.05, then it suggests that the trend is statistically
significant.
These trends are similar even if the period since March 2020 onwards is excluded, during
which COVID-19 appears to have had a predominant influence on spending patterns
290
because of the associated increase in welfare payments and social distancing
requirements.
4.11 Statistical results for annual trend in spending share
Spending category
Annual trend in Signif.
P>t
95% Confidence interval
spending share
(per cent)
Lower bound Upper bound
Childcare/Education/Training/Employment
0.02
0.829
-0.16
0.20
Clothing and footwear
-0.25
**
0.007
-0.43
-0.07
Department, Discount and Variety Stores
-0.06
0.544
-0.23
0.12
Food
-3.34
***
0.000
-3.52
-3.16
Holidays and travel
0.01
0.894
-0.17
0.19
Housing
-0.44
***
0.000
-0.62
-0.26
Medical
-0.04
0.626
-0.22
0.13
Non Card-Based Transactions
3.34
***
0.000
3.16
3.51
Other
0.64
***
0.000
0.46
0.82
Pets
0.00
0.991
-0.18
0.18
Recreation - Activities and memberships
0.08
0.374
-0.10
0.26
Recreation - Eating out
0.29
**
0.002
0.11
0.47
Recreation - Goods and equipment
0.04
0.654
-0.14
0.22
Services
-0.21
*
0.024
-0.38
-0.03
Transport - Private
-0.24
**
0.008
-0.42
-0.06
Transport - Public
-0.03
0.757
-0.21
0.15
Transport - Rental Car
0.00
0.995
-0.18
0.18
Utilities
0.18
*
0.043
0.01
0.36
Note: P < 0.05 = *, P < 0.01 = **, P < 0.001 = ***. A p-value is a result from a statistical test. It indicates whether the estimated
coefficient is statistically significant (i.e. different from zero). More specifically, it shows whether an estimated effect as large as the
estimated coefficient (in this case the annual trend) is likely to have been produced by the model if the true coefficient is zero. A low p-
value means that the estimated effect is very unlikely to have been produced if there truly was no effect. Typically, a p-value of less
than 0.05 is taken to suggest that a result is statistically significant, meaning that it is not spurious.
Source: CDC Program Data, CIE.
Full statistical modelling output for this estimation is presented in Appendix B.
Declined transactions using the CDC
The CDC restricts purchases of certain restricted items, such as alcohol purchases. When
a participant attempts such a transaction, the transaction is declined. The Indue data
provided by DSS for this project identifies the reason for declined transactions, with
declined transactions occurring in relation to attempted purchase of restricted items or
unrelated reasons such as having an insufficient account balance. Appendix C
summarises our categorisation of declined transaction reasons.
The share of declined transactions that relate to attempted purchase of restricted items (or
at restricted merchants) is quite volatile (chart 4.12), mainly because the number of
declined transactions per person is volatile (chart 4.13). However, there is a general
upward trend in the share of transactions that are declined, which is most apparent for
291
the locations where the CDC has been present for some time (i.e. Ceduna and
Goldfields).
4.12 Ratio of declined to successful transactions
Ceduna
East-Kimberley
Goldfields
Bundaberg and Hervey Bay
Linear (Ceduna)
0.09
0.08
0.07
l transactions 0.06
sfu 0.05
ces
0.04
to suc 0.03
0.02
clined
de 0.01
tio fo 0.00
r
n
g
b
r
n
g
b
r
n
g
b
r
n
g
b
r
n
Ra
Ap
Ju
ec
ec
ec
ec
Au
Oct
D
Fe
Ap
Ju
Au
Oct
D
Fe
Ap
Ju
Au
Oct
D
Fe
Ap
Ju
Au
Oct
D
Fe
Ap
Ju
2016
2017
2018
2019
2020
Data source: CDC Program Data, CIE.
4.13 Number of declined transactions related to restricted items per payer
Ceduna
East-Kimberley
Goldfields
Bundaberg and Hervey Bay
Linear (Ceduna)
2.50
yer
r pa 2.00
1.50
transactions pe 1.00
lined
0.50
r of dec
be
0.00
um
r
N
n
g
b
r
n
g
b
r
n
g
b
r
n
g
b
r
n
Ap
Ju
ec
ec
ec
ec
Au
Oct
D
Fe
Ap
Ju
Au
Oct
D
Fe
Ap
Ju
Au
Oct
D
Fe
Ap
Ju
Au
Oct
D
Fe
Ap
Ju
2016
2017
2018
2019
2020
Data source: CDC Program Data, CIE.
292
4.14 Comments from stakeholder consultations – Concerns when travelling
outside the program site
One of the limitations of the Card is that once a participant leaves the program region,
they may encounter businesses that do not accept the Card.
This was highlighted as a concern for participants, with multiple stakeholders
receiving feedback that their customers were not able to pay for accommodation and
other items outside of the program site.
This ‘dis-benefit’ is especially apparent for transient participants that might frequently
visit areas outside of the program location.
Number of declined transactions over time
Participants tend to have a similar number of transactions being declined irrespective of
the amount of time that has passed since they received their CDC.
This means there is little evidence of a ‘learning by doing’ effect, whereby participants
might attempt less alcohol-related or similarly restricted transactions after they have been
on the Card for a while.
This would be expected for example, if such transactions are associated with attempts to
purchase alcohol while intoxicated, and if alcohol consumption is falling among
participants. That is, as participants consume less alcohol, we would expect them to
attempt less restricted item transactions.
However, because there is no change in the number of declined transactions from when a
participant commences and then becomes familiar with CDC, there is no evidence to
support a learning by doing effect. The share of people with declined transactions since
their commencement onto the Card is show in 4.15 below.
293
4.15 Share of people with declined transactions vs time since account was opened
Data source: CDC Program Data, CIE.
This may suggest that participants are continuing to consume alcohol, as they are
continuing to make these purchases.
4.16 Comments from stakeholder consultations – Paying rent can be challenging
We heard that some participants have ad-hoc and informal housing arrangements.
For example, they might be paying board to a family member, or they might have
been informally subletting.
Stakeholders stated that a common complaint about the Card is that some ad-hoc
rental arrangements are blocked.
However, stakeholders also mentioned that there are processes in place to ‘set-up’ the
CDC Card to support these rental arrangements. However, we heard that this process
can lead to delays in rental payments, which can damage relationships between the
participant and their landlord.
294
5 Benefits from improved outcomes
The CDC has generated $2.3 million in benefits from reducing gambling related social
harms. The benefits of reduced gambling apply to all gambling activity, not just
problem gambling activity. Most of the quantified benefits are attributable to
Indigenous CDC participants.
The CDC also has generated a small net benefit for children’s wellbeing. These
benefits are associated with improved child health and nutrition. The net benefits to
children are small because of worsening outcomes for school attendance and safety.
Summary of benefits from the second impact evaluation
The second impact evaluation found clear evidence of improvements in outcomes in a
limited set of areas. Categorising these outcomes into the six benefit domains in our
economic framework, the second impact evaluation found:
■
Economic:
no discernible change in employment outcomes, or any other economic
benefit categories (such as welfare dependence)
■
Health and wellbeing:
no evidence of an overall improvement, with a larger
proportion experiencing negative quality of life impacts than those experiencing an
improvement
■
Social and community:
positive findings, with evidence of a short term improvement
in family and social life
■
Education and child wellbeing:
positive, albeit mixed, findings
■
Safety, crime and family violence:
mixed evidence, with survey evidence showing a
small improvement and statistical evidence showing a small deterioration, and
without enough evidence to attribute the change to the CDC alone
■
Housing and related services:
some evidence of worsening outcomes, with twice as
many participants having a change for the worse than those having a change for the
better, but ultimately, little effect overall, and
■
Individual stability:
negative impacts, discussed further in the cost chapter of this
report.
Based on the changes in outcomes measured by the second impact evaluation where
there was a clear conclusion about the direction of the impacts, we estimate a set of
benefits associated with social and community benefits from reduced gambling, and child
wellbeing benefits and disbenefits (chart 5.1).
295
5.1 Total benefits associated with improved outcomes
Cost/benefit item
Undiscounted
Discounted
$m
$m, NPV a
Social and community benefits of reduced gambling in the short term
2.8
2.3
Child wellbeing – health
0.6
0.5
Child wellbeing – food
0.1
0.1
Child wellbeing – safety
-0.4
-0.3
Child wellbeing – education
-0.2
-0.1
Total benefits associated with improved outcomes
2.9
2.5
a
The net present value is calculated by taking the present value of all cash inflows over the analysis period
Data source: CIE.
Social and community benefits from less problem gambling
Gambling represents a significant social issue for each of the CDC program regions.
Gambling can be an addiction, and it is often correlated with alcohol and drug-related
social issues. For instance, some people may be drawn to gambling to fund alcohol and
drug use.
The total level of gambling across the program regions is difficult to measure. This is
because there are a wide range of modes, and there is limited data collected and reported
on each. For instance, gambling could take place in legal and illegal card games, poker
machines within a TAB, and online.
However, from the CDC baseline data collected, poker machines have been identified as
a key mode of gambling within CDC communities.
Within the second impact evaluation, it was identified that the proportion of CDC
participants who gamble differs by program site.31 For instance the Ceduna region had
the highest incidence, with 22 per cent of participants reporting that they gambled,
followed by the Goldfields and East Kimberley regions (11 and 6 per cent respectively).
The CDC directly aims to reduce gambling across all the program sites by limiting the
amount of cash available to participants and prohibiting the use of the Card towards
gambling activities.
We note that there were some workarounds that have been applied by some participants
to continue gambling activities, such as utilisation of the cash component of the Card, the
utilisation of other income sources and seeking cash through other means. However, the
extent to which these workarounds have taken place is undetermined.
31 Mavromaras K., Moskos M., Mahuteau S., Isherwood L., 2021,
Evaluation of the cashless debit
card in Ceduna, East Kimberley and the Goldfields region: Consolidated report, Future of employment
and skill research centre, The University of Adelaide
296
Evidence from the second impact evaluation
The second impact evaluation found that the frequency of gambling for CDC participants
fell 3.5 percentage points since being on the CDC.32
For all program regions in scope, 21 per cent of participants reported that the CDC has
helped reduce gambling problems. This benefit applied to the participant, their family,
their friends, and the wider community (‘where you live’). The results from the second
impact evaluation’s survey are summarised below in table 5.2.
5.2 Proportion for whom the CDC helped reduced gambling problems
All sites
East
Goldfields Goldfields Non-
Ceduna and
Kimberley
Indigenous
Indigenous
surrounds
Per cent
Per cent
Per cent
Per cent
Per cent
Full sample
Has CDC made a positive
21.0
22.9
27.3
12.2
23.8
difference
Of those that reported a positive difference, who experienced the positive difference:
Participant
34.8
53.3
28.0
20.7
32.4
Participant’s family
43.0
56.5
44.4
10.3
50.8
Participant’s friends
38.4
57.5
38.7
10.5
34.8
Where the participant lives
59.7
63.4
52.6
75.8
52.0
Source: Mavromaras K., Moskos M., Mahuteau S., Isherwood L., 2021, Evaluation of the cashless debit card in Ceduna, East
Kimberley and the Goldfields region: Consolidated report, Future of employment and skill research centre, The University of Adelaide
When considering how to apply these results within the CBA, there are a few important
distributional impacts that are visible, for instance:
■
Not all participants experience a benefit — Although some participants have
reported a positive difference, 79.1 per cent of the sample reported ‘no difference’ or
‘Don’t know/missing’. It is important that benefits are attributed only to those
populations that reported a positive impact, and not the whole CDC population.
■
Indigenous participants appear to benefit the most — Indigenous participants, along
with their family and friends appear to benefit more than non-Indigenous people.
When comparing the survey results across Goldfields Indigenous and Goldfields non-
Indigenous responses, Indigenous participants report significantly higher ‘positive
differences’, apart from responses to the wider community (i.e. ‘where the participant
lives’). For instance, across the whole sample, more than double the proportion of
Indigenous responses reported a positive difference (27.3 per cent compared to 12.2
per cent). However, without the Indigenous survey data for the other regions, this
cannot be confirmed with certainty.
■
Benefits are consistent across regions — When comparing the responses for the full
sample, there is not a substantial difference in responses across the regions, except for
32 Mavromaras K., Moskos M., Mahuteau S., Isherwood L., 2021,
Evaluation of the cashless debit
card in Ceduna, East Kimberley and the Goldfields region: Consolidated report, Future of employment
and skill research centre, The University of Adelaide
297
the more detailed breakdown of the Goldfields Indigenous and Goldfields non-
Indigenous responses. 21.0 of all responses reported a positive difference, compared to
22.9 per cent in East Kimberley and 23.8 per cent in Ceduna and surrounds.
■
Lack of data on the value of gambling activities — Although the evaluation sought
information on the frequency of gambling, it did not seek information on the amount
spent on gambling. For instance, the evaluation found that more than 80 per cent of
participants reported gambling once a month or less, and less than 20 per cent report
gambling more regularly (e.g. weekly or daily/almost daily). However, it is unknown
if the total value of gambling was different between these two groups.
Overall, the evaluation reported that the CDC has been helping to reduce gambling
related harms, especially in the context of family and broader social life. Qualitative
evidence also suggest that the trial has helped to redirect funds away from problem
gambling towards essential spending such as food. However, the impact from the CDC
was reported to be small, and applying to a small part of the CDC population.
Estimating the social cost from gambling
There are many previous studies considering the personal and social impacts of
gambling.33 34 35 36 37 38 39 Much of these relate to the costs to employment and
productivity, individual emotional health, family and relationships, crime, and financial
costs. However, many of these studies focus on the impact of problem gamblers, without
consideration for lesser degrees of gambling activity.
33 Walker, D., 2014, ‘The Social Costs of Gambling’,
International Centre for Youth Gambling
Problems and High-Risk Behaviours, Spring 2014 14(1), available at:
http://youthgambling.mcgill.ca/en/PDF/Newsletter/Spring2014.pdf
34 Mavromaras K., Moskos M., Mahuteau S., Isherwood L., 2021,
Evaluation of the cashless debit
card in Ceduna, East Kimberley and the Goldfields region: Consolidated report, Future of employment
and skill research centre, The University of Adelaide
35 Grinols, E., 2011, ‘The Hidden Social Costs of Gambling’,
The Gambling Culture, available at:
https://www.baylor.edu/content/services/document.php/144584.pdf
36 Victorian Responsible Gambling Foundation, 2017,
The social cost of gambling to Victoria:
Research report, available a
t: https://responsiblegambling.vic.gov.au/documents/121/research-
social-cost-of-gambling.pdf
37 Livingstone, C., Francis, L. and Johnson, M., 2017,
Community benefits claimed by licensed clubs
operating poker machines in the ACT, available at
: https://fare.org.au/wp-
content/uploads/Community-benefits-claimed-by-licensed-clubs-operating-poker-machines-in-
the-ACT-FINAL.pdf
38 The SA Centre for Economic Studies, 2009,
Social Impacts of Gambling: A Comparative Study,
available a
t: https://www.cbs.sa.gov.au/sites/default/files/resource-
files/social_impacts_of_gambling_-_a_comparative_study_-
_april_2009.pdf?timestamp=1607644800065
39 The Select Committee on Gambling 1999,
The Social and Economic Impacts of Gambling in the
ACT, Legislative Assembly for the Australian Capital Territory, available at:
https://www.parliament.act.gov.au/__data/assets/pdf_file/0008/381878/3finalgamblingrepo
rt.pdf
298
The second impact evaluation found that of the CDC participants that gamble, more
than 80 per cent gamble once a month or less, and less than 20 per cent report gambling
more regularly (e.g. weekly or daily/almost daily). Although the total value of gambling
activities was not captured, we can assume that those that gamble less frequently may
still experience harms, but these harms might be less severe than those that gamble more
frequently.
This assumption is consistent with previous reports, such as the Productivity
Commission’s Inquiry into Gambling (2010). This inquiry found that not all people who
experience harms from gambling are considered to be problematic gamblers, with strong
evidence that gambling can have adverse health, emotional and financial impacts on
many more people than those categorised as ‘problem gamblers’.40
Studies such as ‘the social cost of gambling to Victoria’ (2017) have estimated the social
cost of gambling by low-risk, medium-risk and problem gamblers, based on the Problem
Gambling Severity Index (PGSI).41 This index is a standardised measure of at risk
behaviour in problem gambling. People who ‘sometimes’ experience two of these nine
questions are considered low-risk gamblers, with the risk increasing depending on the
frequency of impacts experienced.
A summary of the estimated social costs identified in this study are included in the table
below (values in 2014/15 dollars). The costs in this table represent the annual cost
incurred in 2014. Because of this, some of the costs are the average annual costs, and
some are a one-off cost. For example, the cost of fatality by suicide was calculated by
dividing the average years of life lost to the average total cost of fatality by suicide, while
the cost from divorce was estimated to be the average amount of financial assistance
awarded to victims of crime in 2014/15.
5.3 Costs from gambling addiction (2014/15 values)
Cost
Description
Stakeholder
Low risk
Moderate
Problem
impacted
risk
gamblers
$
$
$
Reduced
Cost to the individual include
Business
165
1 591
9 549
employment
loss of income, job search
and the
productivity and
activities.
individual
Income lost from
Costs to businesses included
missed work
lost productivity and other
employer costs (such as
retraining workers or searching
for replacement workers)
40 Productivity Commission, 2010,
Gambling, Australia Government, available at
https://www.pc.gov.au/inquiries/completed/gambling-2010/report
41 Victorian Responsible Gambling Foundation, 2017,
The social cost of gambling to Victoria:
Research report, available a
t: https://responsiblegambling.vic.gov.au/documents/121/research-
social-cost-of-gambling.pdf
299
Cost
Description
Stakeholder
Low risk
Moderate
Problem
impacted
risk
gamblers
$
$
$
financial problems,
Bankruptcy imposes costs on
Individual
807
2 751
13 536
bad debts and
society in the form of legal and
bankruptcies
other resources expended.
There are also significant
opportunity costs for money
that would have been better
spent on other products and
activities
Committing crimes
The additional cost of crime
Governments
138
509
2 371
to get money for
can relate to police resources,
gambling
apprehension, adjudication,
and incarceration expenditure.
Strain on family and Include divorce, separation,
Families
579
4 169
23 640
relationships
child abuse and neglect.
Domestic violence is also
related to gambling disorders
Strain on family and
Individual
181
323
2 054
relationships
Mental and physical has been reported to include
Individual
1 581
3 700
6 529
health issues related stress related sickness,
to stress
cardiovascular disorders,
anxiety, depression, and
cognitive disorders
increased suicide
ending the life of despondent
Individual
502
190
1 959
attempts
gamblers.
increased suicide
imposes costs on families and Families
748
283
2 916
attempts
the wider society as well as
Costs of Health and
Governments
1 621
2 244
5 190
Human Services
Support Services
Total costs
6 322
15 761
67 745
Source: Victorian Responsible Gambling Foundation, 2017,
The social cost of gambling to Victoria: Research report, see
https://responsiblegambling.vic.gov.au/documents/121/research-social-cost-of-gambling.pdf
Not all of these impacts will be relevant or as severe for the CDC population. For
example, given the CDC population is already receiving employment services and
income support, the cost to the individual from loss of income, job search activities, lost
productivity and other employer costs are already included in the baseline, and not
directly related to gambling activities. For this reason, the costs associated with reduced
employment productivity and income lost from missed work are not included within this
analysis.
It is important to note that there are other personal and social issues that can also
contribute to these negative personal and social impacts, such as addiction to drugs or
alcohol, and mental health conditions. Attributing the full cost to an individual’s
gambling activities is often difficult. For example, many of the negative social impacts
300
listed above could also be linked to these comorbidities, attempting to proportion the cost
attributed to the gambling and other drives costs is difficult.
Benefits related to reduced gambling because of CDC
The CDC is estimated to have created
$2.3 million in benefits (in present value terms)
associated with reduced gambling, between 2015/16 and 2019/20, most of which accrue
to participants in the Bundaberg and Hervey Bay program site, based on assumptions set
out in table 5.4.
5.4 Inputs to estimation of gambling benefits
Modelling input
Value
Source
CDC population that
Ceduna - 22 per cent
Mavromaras K., Moskos M.,
gambles
Goldfields - 11 per cent
Mahuteau S., Isherwood L. 2021
‘Evaluation of the cashless debit card
East Kimberley - 6 per cent
in Ceduna, East Kimberley and the
Bundaberg and Hervey – 13 per cent (average Goldfields region: Consolidated
of other locations)a
report’
Assumed distribution
Low risk - 80 per cent
Mavromaras K., Moskos M.,
across low risk, moderate
Moderate risk - 10 per cent
Mahuteau S., Isherwood L. 2021
risk, and problem gamblers
‘Evaluation of the cashless debit card
Problem gamblers - 10 per cent
in Ceduna, East Kimberley and the
Goldfields region: Consolidated
report’
Drawing from the findings:
■ 80 per cent of participants gamble
monthly or less, and
■ 20 percent weekly or daily/almost
daily – evenly split between
moderate risk and problem
gamblers
Reduction in harms from
Associated costs relating to the following cost
Victorian Responsible Gambling
reduced gambling
categories:
Foundation 2017 ‘The social cost of
gambling to Victoria: Research
■ Committing crimes to get money for
report’
gambling
(Values inflated to 2021-22 values)
■ Strain on family and relationships
■ Mental and physical health issues related to
stress
■ increased suicide attempts
■ Costs of Health and Human Services
Support Services
Participants that have
The net proportion of people who experience a Mavromaras K., Moskos M.,
benefited from CDC
“positive difference”.
Mahuteau S., Isherwood L. 2021
Ceduna - 24 per cent
‘Evaluation of the cashless debit card
in Ceduna, East Kimberley and the
Goldfields - 20 per cent
Goldfields region: Consolidated
East Kimberley - 23 per cent
report’
Bundaberg and Hervey – 21 per cent (average
across all regions)a
301
Modelling input
Value
Source
Change in benefit value
Of those that reported a positive difference,
Mavromaras K., Moskos M.,
the stakeholder that benefits will move to a
Mahuteau S., Isherwood L. 2021
lower risk category. For example, a participant ‘Evaluation of the cashless debit card
with moderate risk would move to low risk.
in Ceduna, East Kimberley and the
Ceduna - 32 per cent for participant, 51% for
Goldfields region: Consolidated
families
report’
Goldfields - 24 per cent for participant, 27% for This assumption is based off the
families
proportion of participants and family
member who benefited from this
East Kimberley - 57 per cent for participant,
evaluation.
51% for families
Bundaberg and Hervey – 35 per cent for
participant, 43% for families (average across
all regions)a
a
Without survey results for Bundaberg and Hervey, the average of the other three program sites has been applied.
Source: CIE and other sources as noted.
5.5 Benefits from a reduction in problem gambling
1 400 000
)
1 200 000
ing ($
bl
1 000 000
gam
em
800 000
obl
s pr
600 000
les
400 000
om
it fr
ef
200 000
en
B
0
2016
2017
2018
2019
2020
Ceduna
East Kimberley
Goldfields
Bundaberg & Hervey Bay
Note: Benefits are presented in undiscounted terms.
Data source: CIE.
When considering benefit values by stakeholder group, both participants and families
benefit by the same proportion (chart 5.6). Governments also benefit through a reduction
in Support Services.
302
5.6 Gambling benefits by stakeholder
Government
18%
Participants
Participants
41%
Families
Government
Families
41%
Data source: CIE.
Education and child wellbeing benefits
One of the objectives of the CDC is to encourage socially responsible behaviour.42 An
important component of this objective relates to the improvement of the welfare of
participants’ children and family members.
There is a strong argument for supporting and investing in child welfare and family
wellbeing. One recent study estimated that the life-long economic cost of abuse and
violence against children and young people cost a total of $11.2 billion.43 This included
costs directly related to abuse and assaults against children and young people, including
life-long financial impacts on productivity, premature mortality, quality of life, burden of
disease, health services, justice system, child protection services, education, and housing
and homelessness.
Some of the expected benefits from improved child welfare and wellbeing include
improved enhanced human capital productivity and participation, greater social
42 Australian Government, 2021,
Guides to Social Policy Law: Social Security Guide, Version 1.282,
available a
t: https://guides.dss.gov.au/guide-social-security-law/8/7/1/20
43 Deloitte Access Economics, 2019,
The economic cost of violence against children and young people,
prepared for the Office of the Advocate for Children and Youth People, available at:
https://www2.deloitte.com/au/en/pages/economics/articles/economic-cost-violence-
against-children-young-people.html
303
inclusion, improvements in health outcomes, and reduced crime.44 45 46 There is also
potential for wider benefits beyond the individual, family or community to society more
broadly.
Within the baseline data collected for Bundaberg and Hervey Bay and Goldfields regions,
a wide range of concerns regarding the welfare of participants’ children and family
members were raised within the regions. Many of these concerns were consistent across
the two program locations, as summarised in table 5.7.
5.7 Summary of baseline data regarding child wellbeing
Negative experience
Bundaberg and
Goldfields
Hervey Bay
■ Concern of the impact from alcohol and drug misuse and gambling on
children’s wellbeing
■ AOD misuse linked to family violence and lack of adequate food, clothes
and shelter
■ Feelings of safety were negatively impacted
■ Lack of appropriate supervision resulting in unsafe environments
■ Grandparents needing to take on care responsibilities
Source: Mavromaras K., Moskos M., Mahuteau S., Isherwood L. 2019
Cashless Debit Card Baseline Data Collection in the Bundaberg
and Hervey Bay Region: Qualitative Findings
Unfortunately baseline data collection was not undertaken in Ceduna and East
Kimberley. However, many of these impacts were discussed in the first CDC impact
evaluation, indicating that it is likely these concerns were also apparent in these
locations.
Evidence from previous evaluations
The quantitative evidence supporting an impact to child wellbeing because of CDC is
mixed. For instance:
■
The first impact evaluation found that of the participants with children, 40 per cent
were better able to look after their children post implementation, and 39 per cent were
more involved with their children’s homework. However, 24 per cent reported that
they were worse off, as they could not buy goods for their children with cash, opposed
to 17 per cent that felt better off, as there were better able to meet basic needs.
44 Council of Australian Governments, 2009
, Investing in the Early Years – A National Early
Childhood Development Strategy, Commonwealth of Australia, available at:
https://www.startingblocks.gov.au/media/1104/national_ecd_strategy.pdf
45 Kilburn, M. and Karoly, L., 2008,
The Economics of Early Childhood Policy: What the Dismal
Science Has to Say About Investing in Children, available at:
http://www.rand.org/pubs/occasional_papers/OP227/
46 Heckman, J., 2006,
The Economics of Investing in Early Childhood, in The Niftey Conference,
University of New South Wales, Sydney.
304
■
The second impact evaluation found that most participants experienced no major
change regarding their children’s welfare. However, across Ceduna, East Kimberley
and the Goldfields, 17.8 per cent of survey respondents reported an overall positive
change, compared to 22.5 per cent who reported an overall negative change.
The second impact evaluation asked CDC participants if change had been experienced
since the start of the CDC across several domains—children’s health, the amount of food
children had access to, children’s safety, school attendance, children’s happiness, and
children’s participation in cultural and social activities. Program participants were asked
to respond as either worse, same, or better. Table 5.8 outlines the net result from this
survey, by subtracting the percentage of worse responses from the percentage of better
responses.
However, there are limitations to this approach. For instance, participants were simply
asked if they perceived the CDC to make each measure worse, the same, or better. No
data was collected on the severity of the change. We have assumed that the distribution
of the impact or quantum of change within each of these responses to be the same.
However, this may not be true if the benefit generated are more significant than the
negative impacts felt by others, although this cannot be determined.
5.8 Net impact on child wellbeing
Outcome measure
All
Ceduna
Goldfields
Goldfields
Goldfields
East
evaluation
indigenous
non-
total
Kimberley
sites
indigenous
Per cent
Per cent
Per cent
Per cent
Per cent
Per cent
Overall health
1.7
6.6
8.3
-5.7
1.4
-0.6
Access to healthy food
2.6
9.5
7.6
-7.3
0.3
3.1
Safety
-5.2
4.5
3.9
-23.2
-9.5
-2.1
School attendance
-3.3
8.3
5.3
-12.2
-3.3
-9.6
Happiness
-7.1
7.7
-2.7
-18.3
-10.2
-9.7
Participation in cultural
activities
-10.7
0.3
-3.3
-18.1
-10.4
-17.7
Participation in social
activities
-10.6
-4.8
-3.7
-18.3
-10.7
-13.7
Source: Mavromaras K., Moskos M., Mahuteau S., Isherwood L. 2019
Cashless Debit Card Baseline Data Collection in the Bundaberg
and Hervey Bay Region: Qualitative Findings, and Mavromaras K., Moskos M., Mahuteau S., Isherwood L., 2021,
Evaluation of the
Cashless Debit Card in Ceduna, East Kimberley and the Goldfields region – Quantitative Supplementary Report.
The impact across the seven domains are mixed. There is an overall positive impact
(albeit small) for overall health and access to healthy food, however, the remaining
domains have had a net negative impact.
The qualitative evidence from the evaluation describes a much more positive response to
the CDC. For instance, the most suggested positive impact of the CDC was that it
increased the amount of money left over for food and clothing and an improved
availability of money to do family activities on the weekends. However, other responses
suggested that restrictions to cashflows were a negative aspect, with families needing to
find the lowest cost avenue for family spending.
305
Given the difference in the second impact evaluation’s findings regarding child welfare, it
is useful to consider the findings from the first impact evaluation. Some of the key
findings include:
■
Many stakeholders reported that the CDC had a positive impact on parenting and
family wellbeing, particularly in relation to parental responsibility, school attendance,
and parent engagement with school and childcare.
■
Merchant reports and observations from stakeholders and community leaders found
an increase in purchases of baby items, food, clothing, shoes, toys and other goods for
children.
■
In addition to an increase school attendance, other positive education impacts
included more children arriving to school with packed lunches, and an increase in the
number of families paying for school excursions and other school-related costs.
■
Family members reported there being a decreased reliance on grandparents to
financially provide and care for their grandchildren.
This feedback from stakeholders was generally supported through the quantitative survey
undertaken. For instance, 40 per cent of participants who had caring responsibilities
reported that they had been better able to care for their children since participating in the
CDC program. However, when asked about the impact of the program on their
child/children’s lives overall, the results were mixed.
When considering how to apply these results within the CBA, there are a few important
impacts that will need to be considered, for instance:
■
The evidence is mixed, with both positive and negative impacts — Given there are
differences in the reported outcomes, from both qualitative and quantitative sources,
the CBA will need to consider the net impact across the domains. The lack of
consistent evidence highlights that the CDC appears to have improved the welfare of
some children, but not for all children.
■
Indigenous participants are likely to benefit the most — The Goldfields survey
results were the only provide results spilt by Indigenous and non-Indigenous
participants. The results highlight that the experience of Indigenous participants is
significantly more positive than non-Indigenous participants across several of the
outcomes. However further analysis of the second impact evaluation survey results
would be needed to confirm this.
■
The quantum of change is not estimated — Although the quantitative results
highlight the net improvement of deterioration for outcome domains, i.e. the number
of participants that reported a net increase or decrease, data on the degree to which an
outcome has been changed was not requested. For example, across all evaluation
regions, 3.3 per cent of participants reported that school attendance has gone down.
However, this does not indicate if overall attendance has gone down 1 day or 5 days,
as an example.
Estimating the social cost of child welfare
Given there are a range of outcomes discussed above under the umbrella of ‘child
welfare’, the CBA will consider each of these in turn to calculate the net impact.
306
Drawing from the baseline data and the qualitative impacts heard from the evaluation
survey, costs have been extracted for when a child’s basic needs are not met.
Table 5.9 summarises the estimated value and approach for each impact.
5.9 Summary of quantified child wellbeing benefits
Impact domain
Economic impact of neglect
Source
Overall health
AIHW has estimated that, 2.2% of the disease burden in Australia AIHW, 2015,
Australian
was due to child abuse and neglect. This estimate attributes the
Burden of Disease Study:
mental health and injury outcomes experienced at all ages
Impact and causes of illness
attributable to exposure during childhood.
and death in Australia, Risk
The study found that child abuse and neglect were causally
factor estimates for Australia:
linked to anxiety disorders, depressive disorders and suicide and Supplementary tables, Table
self-inflicted injuries.
S3
The economic cost of this health burden has been estimated as
the baseline cost from negative health impacts from childhood
neglect within each region.
Access to
Childhood abuse has lifelong impacts, including contributing
Hemmingsson, E. Johansson,
healthy Food
towards obesity in adulthood. This link is driven by a range of
and K., Reynisdottir, S.,
factors, such as the development of low self-esteem, food
2014, ‘Effects of childhood
insecurity, disturbed sleep patterns and elevated response to
abuse on adult obesity: a
stress.
systematic review and meta-
The cost from a lack of healthy food has been estimated through analysis’,
Obesity Review the health cost of obesity.
Safety
To estimate the costs of household safety issues, we have drawn Deloitte Access Economics,
from previous studies that have estimated the cost of childhood
2019,
The economic cost of
neglect on family Support Services. This approach avoids double
violence against children and
counting across other quantified impacts.
young people
School
The benefits from education include personal returns from future Price, J., 2020,
Worlds apart:
attendance
labour productivity, increased participation in the workforce, and
Remote Indigenous
improved health outcomes. However, these benefits are highly
disadvantage in the context
depended on high attendance rates in school years.
of wider Australia, The Centre
There are various negative impacts from low school attendance,
for Independent Studies,
including:
Policy paper: No 34
■ increased social isolation, including alienation and lack of
engagement with the school community and peers, leading to
emotional and behavioural difficulties
■ an increased likelihood of drop-out. Students that are
chronically absent between grades 8 to 12 are seven times
more likely to drop out of school, and
■ the relationship between absence and achievement is
consistently negative and declines in achievement are evident
with any level of absence.
The estimated benefit from an additional day of education draws
from a 2018 study undertaken by the World Bank that found that
each additional year of education produces a private return
between 8.2 per cent to 9.3 per cent, given the income level of
the country. Considering that the Centre for Independent Studies
have compared remote and very remote Aboriginal communities
to “third world countries”, we have adopted the World Bank’s
middle rate of return, being 9.2 per cent uplift for each additional
year of education.
307
Impact domain
Economic impact of neglect
Source
If this rate of return was applied to the full time adult average
ordinary time annual earnings, there is an estimated $44.06 per
annum in additional earnings created from each school day
attended once the student commences in employment. Over an
individual’s career, this is estimated to be $425 in personal
benefits created for each additional day of school attended
(present value).
Happiness
Unable to be quantified
Participation in
Unable to be quantified
cultural activities
Participation in
Unable to be quantified
social activities
Source: CIE.
It is important to note that these costs are considered relevant to only a small subset of
the child population. Not all children of CDC participants are anticipated to experience
these cost.
To estimate the proportion of children that may experience these harms, we have drawn
from a study of harms attributable to child maltreatment in Australia.47 This study
identified the prevalence of a range of childhood harms, and estimated that 2.4 per cent
of children suffer from ‘neglect’. Because of the strong connection between neglect in this
study and the impacts outlined above, we have applied this prevalence rate to the CDC
population to estimate the number of children experiencing similar harms. This is a
conservative assumption, since the CDC sites encounter more social harms than the
average across Australia.
Quantified benefits of improved child welfare and family wellbeing
While evidence is mixed, on balance we conclude that of the benefits that can be
quantified, there is a small overall net benefit.
■
Health impacts — Ceduna, East-Kimberley, and Bundaberg and Hervey Bay all
experienced a positive impact on overall health. Ceduna saw the greatest benefit per
child (estimated to be over $7 916 per impacted child), however, because of the
relatively small population in Ceduna, this region achieved the smallest total benefit
value. East-Kimberley was the only region to experience a decline in health impacts.
The decline experienced in East-Kimberley was not large enough to completely offset
the positive health impacts from the other regions.
■
Improved access to healthy food — Ceduna had a significantly higher per person
benefit, compared to the other three program sites, being approximately four times the
impact. Although positive, the impact to Goldfields and Bundaberg and Hervey Bay
regions was low, with benefits estimated to be between $340 and $415 per impacted
47 Moore S. Scott, J., Ferrari, A., Mills, R., Dunne, M., Erskine, H., Devries, M., Degenhardt,
L., Vox, T., Whiteford, H., McCarthy, M. and Norman, R., 2015, ‘Burden attributable to child
maltreatment in Australia’,
Child Abuse and Neglect, available at:
https://pubmed.ncbi.nlm.nih.gov/26056058/
308
child respectively. East-Kimberley was the only region to have a decline in access to
healthy food, however this negative impact was estimated to be low (approximately
$147 per impacted child)
■
Safety impacts — This analysis has estimated safety impacts as the cost to provide
family Support Services for families where a child is suffering from neglect. Without a
clear definition of what is a safety impact within the second evaluation, this approach
was seen to be most appropriate. Based on the evaluation findings, all regions
reported a negative impact to child safety (i.e. a decrease in child safety). The
estimated per person cost ranges between $483 in East-Kimberley to nearly $2 200 in
the Goldfields region.
■
School Attendance — The impact on school attendance was inconsistent across the
regions, with Ceduna experiencing a significant positive impact (positive 8.3 per cent),
East-Kimberley experiencing a significant negative impact (negative 9.6%), and
Goldfields and Bundaberg and Hervey Bay having smaller negative impacts. Overall,
the impact on school attendance was negative.
Overall, the two positive impacts (heath impacts and improved access to healthy food)
were sufficient to offset the negative impacts (net reductions in safety and school
attendance). This is shown in chart 5.10 below.
5.10 Total child welfare and family benefits across the full program duration
800 000
700 000
600 000
)
($ 500 000
400 000
it value
ef
en 300 000
B
200 000
100 000
0
Overall health
Healthy food
Safety
School attendance
Net benefit
Data source: CIE.
Although there is a net benefit, the value of this benefit is small. The average net benefit
for each impacted child between 2015/16 to 2019/20 is approximately $447. This
includes a mix of one-off impacts/benefits (such as family Support Services) and lifelong
impacts/benefits (such as improved health, obesity, and productivity uplift from school
attendance).
The second impact evaluation also considered various other impacts that cannot be
quantified, such as, happiness, participation in cultural activities, and participation in
social activities. The reported change against these impacts were mostly (and
significantly) negative across the program regions. It is unclear if the estimated net child
309
welfare benefit of $447 per child would be sufficient to offset these other negative
qualitative impacts.
To calculate these impacts, the following general assumptions were used:
■
There are 2.32 children per indigenous participant, and 1.66 children per non-
indigenous participant.48
■
The prevalence of childhood neglect is 2.4 per cent.49
■
The value of a statistical life year is $217 000.50
5.11 Comments from stakeholder consultations – students receiving breakfast
The consultation process identified examples where the CDC made a considerable
impact to individual families.
For example, in two separate consultations, stakeholders mentioned that there has
been a noticeable increase in children having eaten breakfast before school. One
stakeholder stated that
“the best benefit from the CDC is that kids are turning up to
school having had breakfast and with a packed lunch”.
This stakeholder reflected that the community was aware of some struggling families
in the community, and that the CDC has made a substantial impact on their ability to
provide food for their children.
This highlights that there are specific examples of where the CDC is making a
significant impact for families.
Unfortunately, this benefit cannot be incorporated because there is no systemic
evidence indicating how many families experienced this benefit.
Economic benefits associated with improved employment outcomes
The second impact evaluation found ‘no discernible change in employment outcomes
since the introduction of the CDC … within all three trial sites’.51
One modelling approach not explored by the second impact evaluation is survival
analysis, which is a statistical modelling type used to analyse the expected duration until
48 Australian Bureau of Statistics, Births: Australia, Australian Government, available at:
https://www.abs.gov.au/statistics/people/population/births-australia/latest-release
49 Moore S. Scott, J., Ferrari, A., Mills, R., Dunne, M., Erskine, H., Devries, M., Degenhardt,
L., Vox, T., Whiteford, H., McCarthy, M. and Norman, R., 2015, ‘Burden attributable to child
maltreatment in Australia’,
Child Abuse and Neglect, available at:
https://pubmed.ncbi.nlm.nih.gov/26056058/
50 Department of the Prime Minister and Cabinet, 2021,
Best Practice Regulation Guidance Note:
Value of Statistical Life, March 2021, available a
t: https://pmc.gov.au/resource-
centre/regulation/best-practice-regulation-guidance-note-value-statistical-life
51 Mavromaras K., Moskos M., Mahuteau S., Isherwood L. 2021 ‘
Evaluation of the cashless debit
card in Ceduna, East Kimberley and the Goldfields region: Consolidated report’, Future of employment
and skill research centre, The University of Adelaide
310
an event occurs. In this context, survival analysis of unemployment spells can provide
evidence as to whether CDC participants find employment more quickly once they are
on the CDC. The implication of welfare recipients in the program sites finding
employment more quickly would be to reduce welfare costs to government, improve
wellbeing of individuals, and improve the economic welfare of the sites as a whole.
This section examines the impact of poor employment outcomes for communities,
whether survival analysis provides evidence of superior employment outcomes for CDC
participants compared to non-participants, and what value such an improvement would
have.
Impacts of poor employment outcomes for a community
Low employment and labour force participation rates are a significant problem in the
program sites.
These problems are especially acute for Indigenous Australians. According to the NSW
Ombudsman, the Aboriginal unemployment rate is approximately three times greater
than for the rest of the population.52 When considering remote and very remote areas,
the divide becomes greater. For instance, very remote areas, the unemployment rate for
Indigenous people is 29 per cent, compared to 3 per cent for non-Indigenous people
(nationally).53
The employment rate has been described as a key indicator of a stable community,
providing income, fulfilment, and a sense of self-esteem.54 Communities with high
unemployment rates have been directly linked to a wide range of negative social
outcomes, such as poor school attendance and higher crime rates.55
In a CBA, employment is generally considered a cost to the program, not in itself a
benefit. This is because employment would represent a displacement of resources, rather
than a net increase in employment. For example, this is the case when employing a
person is at the expense of employing someone else (no additional job creation).
However, employment benefits do exist if the labour resources employed by the project
were previously unemployed or underemployed, or if the actual wage increased above
the reservation wage.56
By supporting residents of the program sites into employment there is a benefit from:
■
helping families to break the poverty cycle and address intergenerational
unemployment. In 2018, the Inquiry into Intergenerational Welfare Dependence
52 NSW Ombudsman, 2011,
Addressing Aboriginal disadvantage: the need to do things differently,
October, p.3.
53 Price, J., 2020, ‘
Worlds apart: Remote Indigenous disadvantage in the context of wider Australia’, The
Centre for Independent Studies, Policy paper: No 34
54 Price, J., 2020, ‘
Worlds apart: Remote Indigenous disadvantage in the context of wider Australia’, The
Centre for Independent Studies, Policy paper: No 34
55 Price, J., 2020, ‘
Worlds apart: Remote Indigenous disadvantage in the context of wider Australia’, The
Centre for Independent Studies, Policy paper: No 34
56 The difference between a worker’s actual wages and what they would be willing to accept.
311
reported on welfare dependence of families and children. The Inquiry identified
various barriers to employment. These factors include location, transportation,
appropriate and flexible employment opportunities, support to maintain employment,
and parenting responsibilities.57 These factors are compounded in regional and
remote areas. In addition to this, Aboriginal people were identified as a group at
greater risk of entrenched disadvantage,58 and
■
overcoming barriers for people that are long term unemployed. After a long period of
unemployment, people face additional barriers to employment from loss of skills, loss
of confidence, and resistance from employers.59 Because of these impacts, on average,
people who become long term unemployed are less than half as likely to gain
employment within a month as someone who has been short-term unemployed.60
This can have significant impacts to individuals and the wider community. For
instance, being unemployed for more than six months is associated with lower
wellbeing, poorer health, having children with worse academic performance, and
communities have a higher rates of crime and violence.61
Impacts of the CDC on employment outcomes
There are a range of studies applying survival analysis approaches to analysis of
unemployment spells, with the most relevant for this analysis being RBA (2020).62
Hazard ratios for a range of characteristics were estimated in a Cox regression, which is
statistical model that predicts the expected time of an event of interest based on a range of
explanatory variables. In their modelling, the duration of employment spells was
predicted on the basis of sex, age, country of birth, relationship in the household, social
marital status, number of children, whether the person if looking for full-time or part-time
work, and employment history.
57 Parliament of the Commonwealth of Australia 2018 ‘
Living on the edge: Inquiry into
Intergenerational Welfare Dependence’, House of Representatives Select Committee on
Intergenerational Welfare Dependence
58 Parliament of the Commonwealth of Australia 2018 ‘
Living on the edge: Inquiry into
Intergenerational Welfare Dependence’, House of Representatives Select Committee on
Intergenerational Welfare Dependence
59 Parliament of the Commonwealth of Australia 2018 ‘
Living on the edge: Inquiry into
Intergenerational Welfare Dependence’, House of Representatives Select Committee on
Intergenerational Welfare Dependence
60 Cassidy, N., Chan, I., Gao., A and Penrose, G., 2020, ‘Long-term unemployment in
Australia’,
RBA Bulletin, December 2020, available at: Reserve Bank of Australia
https://www.rba.gov.au/publications/bulletin/2020/dec/long-term-unemployment-in-
australia.html
61 Nichols, A., Mitchell, J., Lindner, S., 2013, ‘
Consequences of Long-Term Unemployment’, Urban
Institute
62 Cassidy, N., Chan, I., Gao., A and Penrose, G., 2020, ‘Long-term unemployment in
Australia’,
RBA Bulletin, December 2020, available at: Reserve Bank of Australia
https://www.rba.gov.au/publications/bulletin/2020/dec/long-term-unemployment-in-
australia.html
312
We estimate a similar model using data from DOMINO for Newstart/JobSeeker
recipients only. For this cohort of welfare recipients we observe:
■
demographic characteristics such as age, sex and country of birth
■
location (which we aggregate to the Statistical Area 3 level based on the Australian
Bureau of Statistics region classification) at any period of time
■
previous and future spells of receiving welfare, and
■
the reason why a payment entitlement was suspended/cancelled, which we have
categorised into reasons relating to obtaining employment and unrelated reasons.
We merge this data with Indue data about CDC participants, including the date that each
participant opens their CDC account and the data of the last transaction.
Based on this data, we divide each welfare spell into segments defined by their CDC
status (whether they are currently on the CDC) and their location (which changes over
time for many participants). This provides a dataset of periods of time, with some periods
of time ending in suspension or cancellation of the payment due to the recipient
obtaining employment.
This forms the estimation dataset used to analyse the relative duration of unemployment
spells for CDC recipients and non-recipients. For the purpose of this modelling, we
include in the estimation dataset only spells of unemployment where the payment
recipient is in the:
■
Eyre Peninsula and South West SA3 (corresponds to Ceduna),
■
Kimberley SA3,
■
Goldfields SA3, or
■
Wide Bay SA4 (corresponds to Bundaberg and Hervey Bay).
By comparing CDC participants to the wider SA3 and SA4 areas, we are ensuring that
the counterfactual has the same labour market and employment conditions as the CDC
population group. For some of the program sites, the SA3 or SA4 is closely aligned to the
program location. In this case, the modelling is like a ‘before and after’ statistical
approach, that compares the employment outcomes of the population before and after
implementation of the CDC program.
We find insufficient evidence that CDC participants have shorter unemployment spells
than non-participants, however the modelling cannot conclusively rule out that such
differences do not exist. That is, there is no evidence of an improvement in employment
prospects for the program sites collectively compared to surrounding areas after
controlling for differences in demographic and other factors. There is some weak
evidence for superior outcomes among CDC participants in East Kimberley but the
difference in employment outcomes is not statistically significant (see Appendix B). This
aligns with the finding from the second impact evaluation that East Kimberley had the
highest alcohol consumption, and therefore may be most likely to have superior
employment prospects as a result.
However, the model results are inconclusive because the relationship between likelihood
of a welfare recipient remaining employed is not constant over time between CDC
participants and non-participants (chart 5.12). The Kaplan-Meier curves estimate and
313
visualise survival functions. These curves show the share of Newstart/JobSeeker
recipients in the vicinity of the program areas that are unemployed after the indicated
number of days since they began receiving the payment.
The curves for CDC and non-CDC participants are close together, with the CDC curve
mostly but not always above the non-CDC curve. This provides suggestive evidence that
the CDC does not improve employment prospects, since a smaller share of Newstart
recipients on the CDC (compared to non-participants) become employed after 1000, 2000
or 3000 days of receiving the payment.
5.12 Kaplan-Meier survival curves for CDC participants and non-participants
Note: K
aplan-Meier curves estimate and visualise the probability of an event (e.g. employment) occurring beyond a specified time. In
this case, each curve measures the probability that an individual has not obtained employment since the time they first receive
welfare payments. In this case, the proportion of Newstart/JobSeeker recipients continuing to receive welfare payment due to not
obtaining employment declines as they have received payments for a longer period of time.
Data source: CIE.
However, these survival curves do not account for differences in characteristics such as
age between the CDC and non-CDC cohorts. To control for these differences and test the
statistical significance of any differences, we estimate a Cox proportional hazards
regression, with the results of this modelling reported in Appendix B. The conclusion
drawn from this statistical modelling is that there is no clear evidence of an impact of the
CDC on employment prospects. This is consistent with the findings of the second impact
evaluation.
314
5.13 Comments from stakeholder consultations – Motivating people to find
employment
One of the consequences of the Card reported by stakeholders is that it increases the
motivation for welfare payment recipients to find employment and thereby not need
to use the CDC.
One stakeholder mentioned that the thought of commencing on the Card provided the
motivation for some of their clients to increase job search activities. There had been
instances where job seekers moved into employment just to avoid the Card.
However, when analysing the data about when welfare payments stop because the
recipient obtains employment, this does not appear to be a wide spread impact.
Value if employment outcomes were to improve
The above analysis found no evidence of an improvement in employment outcomes
associated with or caused by the CDC program.
However, if the CDC program was to drive improved employment outcomes in the
future, the value of these benefits would be significant.
As a financial proxy to estimate the value of this benefit we have applied the following
logic:
■
The economic benefit of an increase in the participation rate is the sum of consumer
(employer) and producer (worker) surplus changes. If the employer’s and worker’s
benefits and costs net to a positive value, then there have been economic benefits
created.
■
The workers in this context would obtain a surplus equal to the difference between the
wage they receive and their opportunity cost. The opportunity cost for these workers
is their ‘reservation wage’, which we have assumed is equal to the value from not
working and receiving welfare, which is the next best alternative.
■
The difference in the financial return from working (wage) versus not working
(welfare) is the surplus obtained from working. We have assumed that all workers are
eligible for JobSeeker payments (around $550 per fortnight at mid-202063) and would
gain employment at an entry level on the Building and Construction General On-site
Award ($844.86 per week64). This assumption was tested through the consultation
process. Since stakeholders across multiple CDC sites stated that there were many
entry-level positions available within mining and construction industries, and one of
the barriers to gaining employment in these positions was a lack of motivation.
Because of this feedback, this entry level wage was seen to be an appropriate
assumption. This comes to an estimated net surplus of $570 per week.
63 This rate has been provided by DSS to the CIE in generalisation. JobSeeker payments vary
according to a person’s individual circumstances, and it is possible to receive other payments
alongside JobSeeker payments.
64 Fair Work Ombudsman, 2020,
Pay Guide – Building and Construction General On-site Award,
Australian Government
315
■
This assumes that these workers would be willing to accept a wage equal to their
welfare payments (making them financial indifferent between working and not
working).
This logic implies that the net benefit for workers is the difference between their wage
and potential welfare payments.
However, there are other benefits from the attainment of sustainable employment. For
instance, there are many studies drawing a link between personal or household income
and health outcomes. Some of these have been summarised in table 5.14.
5.14 Studies linking economic outcomes with health outcomes
Study citation
Type of benefit
Findings
Australian studies
Australian Institute of Health
Higher income
AIHW reported that the higher a person’s income, education
and Welfare (2016) ‘Australia’s could lead to better or occupation level, the healthier they tend to be.
health 2016)
health outcomes
However, one of the health risk factors is inadequate fruit
and vegetable consumption, which there is no significant
difference between people in the lowest and highest
socioeconomic groups. Those in the lowest socioeconomic
group were more likely to smoke, have insufficient daily
activity and suffer from some chronic diseases.
Friel, Denniss ‘Unfair economic Higher household
This study found that health related outcomes are strongly
arrangements make us sick’
income could lead
correlated to a household's income, with a social gradient for
to improved health health being observed for life expectancy and a range of
outcomes
chronic diseases.
Isaacs, Enticott, Meadows,
Increased
This study compared psychological distress across
Inder (2018) ‘Lower Income
socioeconomic
socioeconomic groups. The researchers found that lower
Levels in Australia Are Strongly status decreases
socioeconomic status was associated with elevated distress
Associated With Elevated
psychological
in all areas of Australia, and elevated distress was more
Psychological Distress:
distress
likely in those with lower household incomes. For instance,
Implications for Healthcare and
more than 1-in-4 people making up the poorest one-fifth of
Other Policy Areas’
Australians have current psychological distress at a
high/very-high level, and this compares to about 1-in-20 in
the richest one-fifth of Australians
International studies
Chen, Liu, Binkey (2012), ‘An
Increased income
This study compared consumption pattern data of milk and
exploration of the relationship
can reduce annual soft drinks. These two products allowed the consumer to
between income and eating
calories
easily choose healthier (or less unhealthy) varieties as
behavior’
minimal nutrition knowledge was needed, with no price
differences between products. Such as choosing low fat milk
over high fat milk.
The study found that for every $10,000 in income (2005-
2006 dollars) the calories income per year was 2 932
calories (milk and soft drink results combined). This is the
equivalent of 2/3 of a pound a year weight loss. As this
study only considered milk and soft drink, it is expected that
the impact across all food consumed is much higher.
316
Study citation
Type of benefit
Findings
French, Tangney, Crane, Wang, Increased income
This study considered the associations between household
Appelhans (2019) ‘Nutrition
can increase
income and the diet quality of household food purchases.
quality of food purchases
consumption of
The study found that lower-income households purchase
varies by household income:
quality food
less healthful foods overall, fewer fruits and vegetables and
the SHoPPER study’
more sugary beverages compared to households with higher
income. However, no significant differences were observed
between low- and medium-income households after
adjustment for education, marital status and race.
Schiller, Lucas, Peregoy (2012) Lower income
This study found that 22.8 per cent of families with income
‘Summary health statistics for
families have worse less than $35 000 (US dollars in 2011) self-reported fair or
u.s. Adults: national health
health outcomes
poor health. This reduced to 12.9 per cent for families
interview survey, 2011’
earning $35 000 to 49 999. For these families, there was
lower life expectancy, and higher prevalence of coronary
heart disease, stroke, emphysema, chronic bronchitis,
diabetes, dental illnesses, and many others.
Woolf, Aron, Dubay, Simon,
Improved economic This reported considered various research articles on health
Zinnerman, Luk (2015) ‘How
conditions lead to
and income. The report concluded that improving economic
are income and wealth linked
lower health care
conditions, for those who are poor and those in the middle
to health and longevity’
costs
class, could improve health and help control the rising costs
of health care.
Source: CIE and other studies as noted.
Displacement of employment of non-participants
If the CDC improves employment outcomes for CDC participants, but at the expense of
non-participants, this would mitigate the benefits of such an improvement.
317
6 Benefits from a change in consumption patterns
The CDC cannot be used to purchase alcohol, gambling, and illegal drugs or cash like
products such as some types of gift cards. Reduced consumption of restricted items is
anticipated to generate benefits for some participants.
The key benefit of the CDC in relation to changes in consumption relates to reduced
alcohol spending, and the associated reduction in the cost of alcohol misuse. The cost
of alcohol misuse in the CDC program sites is estimated to be $21 million in
2019/20, and $43.3 million (discounted) over the period (since 2015/16).
The benefit of
reduced alcohol misuse as a result of the CDC program is estimated at
$4 million in 2019/20, and $8.5 million (discounted) over the period from 2015/16 to
2019/20.
Stakeholders within the program sites confirmed that the consumption of alcohol
appears to have reduced. However, these consultations also confirmed that
participants can ‘get around the Card’ in creative ways to access alcohol. For this
reason, stakeholders suggest that the biggest benefit from a change alcohol
consumption has been seen in low and moderate users, and less so in high risk or
dependent users.
We do not find any net quantifiable benefits associated with reduced cash availability,
principally associated with the mixed evidence in this regard from the second impact
evaluation.
Benefits associated with reduced alcohol consumption
There are three stages to estimating the benefit of reduced alcohol consumption/misuse:
■
estimate the societal costs of alcohol misuse in Australia using the best available
estimate from the academic literature
■
attribute an amount of these costs to each program site based on the evidence in the
second impact evaluation about relative consumption patterns, and
■
estimate the difference in costs of alcohol misuse in the program sites associated with
the CDC based on the evidence from the second impact evaluation about how alcohol
consumption changed.
The following sections step through these stages of the analysis, including discussion of
data, assumptions and calculation approaches.
318
Societal costs of alcohol misuse in Australia
The most recent study comprehensively estimating the societal costs of alcohol misuse in
Australia is Manning, Smith and Mazerolle (2013). This study uses a mixed-methods
approach to conduct bottom-up estimation of the total societal cost at an Australia-wide
level, without any disaggregation by region or demographic characteristics. It estimated
there to be $14.352 billion of costs associated with alcohol misuse in 2010.65 This is
comprised of:
■
productivity costs (42.1 per cent), which is the sum of reduced workforce and
household labour due to premature mortality and sickness, and reduced workforce
participation due to absenteeism
■
traffic accident costs (25.5 per cent), which includes human costs from fatalities and
serious injuries, vehicle and property damage, and other general costs
■
criminal justice system costs (20.6 per cent), including police attending and
investigating alcohol-related incidents, child protection and Support Services, out-of-
home care for family members affected by alcohol-related incidents, costs to
government and lost productivity associated with imprisonment, loss of life and
wellbeing associated with alcohol-related violence, and court costs, and
■
health system costs (11.7 per cent), including hospital costs, nursing home costs,
pharmaceutical expenses and ambulance costs.
This study is the only estimate of societal costs of alcohol misuse reported by the
Australian Institute of Health and Welfare (AIHW) in their consolidation of the most
recent information on the impacts of consumption of alcohol and other drugs.66 A
previous study, Collins and Lapsley (2008), estimated there to be $10.8 billion of
intangible costs (e.g. labour and health costs) and $4.5 billion of intangible costs such as
loss of life through violence.67 Manning, Smith and Mazerolle (2013) updates and
expands the estimates from Collins and Lapsley (2008), and is the preferred estimate due
to its recency.
Some key exclusions that will tend to make the cost estimates from Manning, Smith and
Mazerolle (2013) an underestimate include the following, as noted by the authors:
■
Alcohol-attributable presenteeism, which relates to poor health leading to a reduction
in a worker’s capacity to perform. Sullivan (2019) estimates that presenteeism has a
total societal cost approximately four times that of absenteeism in New Zealand.
Manning, Smith and Mazerolle (2013) do not report the estimated cost of
65 Manning, M., Smith, C. and Mazerolle, P., 2013, ‘The societal costs of alcohol misuse in
Australia,
Trends & Issues in crime and criminal justice, no.454, Canberra: Australian Institute of
Criminology
, https://www.aic.gov.au/publications/tandi/tandi454
66 AIHW, 2021,
Alcohol, tobacco & other drugs in Austraia, last updated 16 April 2021, available at:
https://www.aihw.gov.au/reports/alcohol/alcohol-tobacco-other-drugs-
australia/contents/impacts/economic-impacts
67 Collins, D. and Lapsley, H, 2008,
The costs of tobacco, alcohol and illicit drug abuse to Australian
society in 2004/05, available at:
https://nadk.flinders.edu.au/files/3013/8551/1279/Collins__Lapsley_Report.pdf
319
absenteeism, which would be necessary to enable applying this ratio to estimate
presenteeism costs.
■
Negative impacts on others associated with someone else’s drinking are partially
accounted for in their direct cost estimates. For example, traffic accident costs will
include costs to others associated with someone else’s drinking. Direct inclusion of all
costs associated with someone else’s drinking would involve some extent of double-
counting with the cost categories already quantified.
Projecting the total societal costs to 2020
The total societal cost of alcohol misuse will change over time due to a range of factors.
Three of the key factors are:
■
how the number of people at risk changes
■
how the cost of resources changes, and
■
how the risk level of the population changes.
We project the total societal cost of alcohol misuse to be equal to $21.273 billion in 2020
(table 6.1), based on the combination of these three uplift factors (table 6.2), as
summarised in the equation below:
𝑇𝑜𝑡𝑎𝑙 𝑠𝑜𝑐𝑖𝑒𝑡𝑎𝑙 𝑐𝑜𝑠𝑡 𝑜𝑓 𝑎𝑙𝑐𝑜ℎ𝑜𝑙2020
= 𝑇𝑜𝑡𝑎𝑙 𝑠𝑜𝑐𝑖𝑒𝑡𝑎𝑙 𝑐𝑜𝑠𝑡 𝑜𝑓 𝑎𝑙𝑐𝑜ℎ𝑜𝑙2010 × (1 + 𝑃𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑔𝑟𝑜𝑤𝑡ℎ)
× (1 + 𝑃𝑟𝑖𝑐𝑒 𝑔𝑟𝑜𝑤𝑡ℎ) × (1 + 𝑆𝐸𝑉 𝑔𝑟𝑜𝑤𝑡ℎ)
The growth of each component is the growth between 2010 and 2020, the total societal
cost of alcohol is obtained from Manning, Smith and Mazerolle (2013)68, and each
growth factor is as defined below:
■
Population growth: We assume that the growth in the adult population will be the
most relevant driver for growth in the population at risk.69
■
Price growth: We use the GDP deflator from for general government at the national
level to inflate prices for justice and health system costs, which are primarily costs to
government. For traffic accident costs we use the Consumer Price Index (CPI)70 and
for productivity costs we use the Wage Price Index.71
68 Manning, M., Smith, C. and Mazerolle, P., 2013, ‘The societal costs of alcohol misuse in
Australia,
Trends & Issues in crime and criminal justice, no.454, Canberra: Australian Institute of
Criminology
, https://www.aic.gov.au/publications/tandi/tandi454
69 Australian Bureau of Statistics, 2020,
National state and territory population, September 2020,
available a
t: https://www.abs.gov.au/statistics/people/population/national-state-and-
territory-population/latest-release#data-download
70 Australian Bureau of Statistics, 2021,
Consumer Price Index, Australia, March 2021, available at:
https://www.abs.gov.au/statistics/economy/price-indexes-and-inflation/consumer-price-
index-australia/latest-release
71 Australian Bureau of Statistics, 2021,
Wage Price Index, Australia, March 2021, available at
https://www.abs.gov.au/statistics/economy/price-indexes-and-inflation/wage-price-index-
australia/latest-release
320
■
Growth in the Summary Exposure Value (SEV) for Australia: SEV is obtained from
the GBD Compare data tool.72. The GBD Compare tool states that “SEV, or
summary exposure value, is a measure of a population’s exposure to a risk factor that
takes into account the extent of exposure by risk level and the severity of that risk’s
contribution to disease burden.”. We extract the SEV data series from GBD Compare
between 2010 and 2019, and apply an average growth rate over this period to further
project SEV to 2020.
6.1 Total societal cost of alcohol misuse each year in 2010 and 2020
Year
Justice
Health system
Productivity
Traffic accidents
Total
$ billion
$ billion
$ billion
$ billion
$ billion
2010
2.958
1.686
6.046
3.662
14.352
2020
4.217
2.404
9.312
5.340
21.273
Note: Estimates are shown to 3 decimal places to remain consistent with the precision of results reported by Manning, Smith and
Mazerolle (2013).
Source: Manning, M., Smith, C. and Mazerolle, P., 2013, ‘The societal costs of alcohol misuse in Australia, Trends & Issues in crime
and criminal justice, no.454, Canberra: Australian Institute of Criminology, CIE.
6.2 Inflators to project the total societal cost of alcohol misuse in 2020
Inflator
Justice
Health
Productivity
Traffic
system
accidents
Ratio
Ratio
Ratio
Ratio
Population inflator
1.18
1.18
1.18
1.18
Price inflator
1.20
1.20
1.30
1.23
SEV inflator
1.01
1.01
1.01
1.01
Source: CIE and sources as noted above table.
Use of the SEV to project the level of risk for a given population is the most uncertain
aspect of this approach. Two key concerns are summarised in box 6.3. Noting these
issues, SEV is an appropriate measure because it captures the complexity of changes in
alcohol consumption behaviour via a single metric. As this discussion makes clear, there
is a complex relationship between alcohol consumption behaviour and the costs of
alcohol misuse.
72
https://vizhub.healthdata.org/gbd-compare/
321
6.3 The SEV as a projector of risk associated with alcohol misuse
Firstly, using the average growth in SEV from 2010 to 2019 in order to project 2020
will not account for any step change in growth associated with the COVID-19
pandemic and associated lockdowns. The pandemic affected patterns of alcohol
consumption in a range of ways, such as increasing typical volumes consumed, a shift
in locations of consumption away from licenced premises towards home
consumption, and other factors. However, these impacts may have been less in the
program locations compared to capital cities due to shorter/fewer lockdowns and less
disruption in general. Additionally, some of these changes in alcohol consumption
may have been associated with new unemployment, and given that the cohort which
is the focus of this analysis is welfare recipients on the CDC program, more of whom
are not employed, these impacts may be less than reported for other areas/cohorts.
Secondly, we assume that the risk level associated with alcohol misuse relating to
other cost burdens (e.g. justice costs), increases in proportion with the SEV. This is
appropriate if risk exposure for disease will be similar to risk exposure for other types
of costs. However, for example, risk exposure for drink driving incidents may also
change if vehicle ownership patterns change. Alternative measures, such as the
number of people exceeding the lifetime risk guideline or single occasion risk
guideline,73 but these measures ultimately suffer from the same issue that they may be
better predictors of health risk than risk of other cost types (such as imprisonment risk
or risk of presenteeism costs).
Source: CIE.
Estimated costs of alcohol misuse in the communities under the CDC case
We estimate the base case costs of alcohol misuse among CDC participants by
multiplying the number of participants by a societal cost of alcohol misuse per adult,
which varies across the program locations.
Based on the current adult population of Australia of 20.03 million, this implies a societal
cost of alcohol misuse of $1 062/person in 2020. We adjust this by a set of factors
reflecting the relative risk of adults in each program location compared to the Australia-
wide average.
The Alcohol Use Disorders Identification Test (AUDIT) score provides an indication of
the level of risky drinking behaviour.
East Kimberley has higher proportions of risky drinking behaviour, with higher shares of
participants being in the ‘very high’, ‘high’ and ‘moderate’ risk levels than all benchmarks
73 See: Australian Bureau of Statistics, 2018,
National Health Survey: First Results, 2017-18 —
Australia, table 10.1 ‘Alcohol consumption — Lifetime Risk(a), Persons’ and table 11.1
‘Alcohol consumption — Single occasion risk(a), Persons’, which are discussed at:
https://www.abs.gov.au/statistics/health/health-conditions-and-risks/alcohol-
consumption/2017-18
322
(chart 6.4). However, the Goldfields region has similar or slightly lower risk levels than
benchmarks. Comparison of Ceduna to benchmarks is mixed (chart 6.5), with higher
levels in the low risk category and less people at moderate risk, but approximately twice
as many people in the ‘very high’ category compared to the Australia-wide (or other)
averages.
6.4 Comparison of AUDIT scores for Western Australian Program sites to
benchmarks
East Kimberley
Goldfields
Perth
Rest of WA
WA
Australia
Low (0-7)
Moderate (8-15)
High (16-19)
Very High (20 or more)
0
10
20
30
40
50
60
70
80
90
Share of respondents (per cent)
Data source: Mavromaras K., Moskos M., Mahuteau S., Isherwood L., (2021) Evaluation of the Cashless Debit Card in Ceduna, East
Kimberley and the Goldfields Region, prepared by University of Adelaide, CIE.
6.5 Comparison of AUDIT scores for Ceduna to benchmarks
Ceduna
Adelaide
Rest of SA
SA
Australia
Low (0-7)
Moderate (8-15)
High (16-19)
Very High (20 or more)
0
10
20
30
40
50
60
70
80
90
Share of respondents (per cent)
Data source: Mavromaras K., Moskos M., Mahuteau S., Isherwood L., (2021) Evaluation of the Cashless Debit Card in Ceduna, East
Kimberley and the Goldfields Region, prepared by University of Adelaide, CIE.
The second impact evaluation did not collect AUDIT data for Bundaberg and Hervey
Bay. However, the rates of reported drink driving offences are higher in Bundaberg and
lower in Hervey Bay (contained within Fraser Coast LGA) than the Rest of Queensland
(chart 6.6). The combined Bundaberg and Fraser Coast LGAs have similar levels of
323
reported drink driving offences compared to the Rest of Queensland. This suggests that
costs of alcohol misuse may be similar in Bundaberg and Hervey Bay compared to the
Queensland and likely Australian average.
There are limitations in linking reported drink driving offences to the cost of alcohol
misuse. As seen in table 6.1, costs of traffic accidents represent 25 per cent of the total
costs of alcohol misuse. However, we do not have data about alcohol-related
presenteeism or absenteeism for Bundaberg and Hervey Bay, which is the largest
component of costs. Alcohol-related health costs and criminal justice costs were relatively
smaller components of costs than traffic accidents, and thus we have focussed on drink-
driving offences as a measure of relative costs in Bundaberg and Hervey Bay.
6.6 Drink driving in Bundaberg and Hervey Bay compared to Rest of Queensland
Bundaberg Regional Council
Fraser Coast Regional Council
Combined Bundaberg & Fraser Coast
Rest of QLD
8
ces
7
en
r)
yea 6
e/
iving off
opl 5
ink dr
pe
0 4
dr
0
0
r/1 3
rted
po
be 2
of re
(num
te
1
Ra
0
2015
2016
2017
2018
2019
2020
Data source: CIE.
6.7 Frequency and amount of drinking in program sites (ex. Bundaberg and Hervey
Bay)
Measure
Units
East Goldfields
Ceduna
All
Kimberly
and
three
surrounds
sites
Frequency of drinking days per month
No./month
3.0
1.7
1.5
2.1
Amount of alcohol consumed on a usual drinking day
No./day
8.1
4.9
5.5
6.1
Average drinks per week
No./week
5.5
2.0
1.9
2.9
Note: Responses to questions about frequency and amount of consumption were in bands. Responses to the frequency question were
‘Never’, ‘Monthly or less’, ‘2-4 times per month’, ‘2-3 times per week’, and ‘4 or more times per week’, which were assumed to
correspond to 0, 1, 3, 10.83, and 17.33 drinks respectively. Responses to the amount question were ‘1-2’, ‘3-4’, 5-6’, 7-9’, and ‘10 or
more’, which were assumed to correspond to 1.5, 3.5, 5.5, 8 and 11 drinks respectively. The average drinks per week is the product of
the average frequency of drinking days and amount of alcohol consumed on a usual drinking day.
We multiply the cost per person by an assumed ratio of alcohol misuse costs per person
for each program site compared to the Australian average (chart 6.8).
■
For Bundaberg and Hervey Bay, on the basis of there being little systematic difference
in reported drinking driving incidents per person compared to the Rest of Queensland,
324
we have assumed that the cost of alcohol misuse is the same per person as the rest of
Australia (i.e. $1 062/person).
■
For the remaining sites, we calculate the ratio of the share of people in the very high
and high AUDIT score categories between each site and the Australia-wide average.
– For example, East Kimberley has a total of 17.6 per cent of participants in the very
high or high band, while the Australia-wide average is 5.95 per cent of people
being in this tier. Accordingly, we apply a factor of 296 per cent to the cost per
person for Australia to estimate the cost per person for East Kimberley, which is
$3 141/person.
– The relationships between consumption measures, risk and costs are complex. For
example, it is unclear what proportion of total costs of misuse are associated with
the proportion of people in the ‘very high’ AUDIT score category. This method is
an approximation to adjust for risk levels using the most timely and comprehensive
data available about consumption by CDC participants (the Second Evaluation
Report data). There are limitations with this approach, in that the cost has not
been built up via a bottoms-up approach, and that using the AUDIT score in this
manner is not an established approach. We have not identified a preferred
approach in the literature for mapping AUDIT score results to costs.
– This approach implicitly assumes that people with high and very high AUDIT
scores are entirely responsible for the costs of alcohol misuse, which is unlikely to
be true. As a result of this assumption, we will overestimate the cost per person in
the program sites.74
Note, importantly, that these adjustment factors are based on the AUDIT score results
for CDC participants, who would already have experienced the reduction in alcohol
consumption they reported was associated with the CDC. Therefore, the cost per person
implied by these factors will be the total cost of alcohol misuse with the CDC.
6.8 Ratio of alcohol misuse cost per person between program sites and rest of
Australia
Factor
Ceduna
East
Goldfields Bundaberg
Kimberley
and Hervey
Bay
Per cent
Per cent
Per cent
Per cent
Ratio of alcohol misuse costs in each community relative to
138
296
106
100
Australian average
Source: CIE.
Applying these factors to the cost per person of $1 026, and multiplying by the total
number of CDC participants in each year produces an estimate of the costs of alcohol
misuse with the CDC, totalling $43.3 million for the first four program sites (table 6.9).
74 As discussed below, this overestimation is counteracted by the effect of this assumption in
causing underestimation of the impact of the CDC on the cost of alcohol misuse.
325
6.9 Costs of alcohol misuse by participants under the CDC case
Site
2015/16 2016/17 2017/18 2018/19 2019/20 Total undisc.
Total disc.
$million
$million
$million
$million
$million
$million $million, NPV
Ceduna
1.0
1.3
1.4
1.6
1.7
7.0
6.7
East Kimberley
2.5
4.9
5.3
5.7
6.4
24.7
22.3
Goldfields
0.0
0.0
0.7
4.3
4.8
9.8
7.3
Bundaberg and Hervey
0.0
0.0
0.0
1.9
7.8
9.7
7.0
Bay
All sites
3.5
6.2
7.4
13.5
20.8
51.3
43.3
Note: ‘Undisc’ and ‘disc’ refer to undiscounted and discounted respectively.
Source: CIE.
Estimated costs of alcohol misuse relative to the base case
The estimated costs of alcohol misuse in the base case are calculated as:
1
𝐶𝑜𝑠𝑡𝑏𝑎𝑠𝑒 𝑐𝑎𝑠𝑒,𝑝 = 𝐶𝑜𝑠𝑡𝐶𝐷𝐶 𝑐𝑎𝑠𝑒,𝑝 × (1 − 𝑖𝑚𝑝𝑎𝑐𝑡 𝑜𝑓 𝐶𝐷𝐶𝑝)
where the impact of the CDC for program site
p is a percentage point change in alcohol
misuse costs based on evidence from the second impact evaluation.
It is not straightforward to estimate the change in alcohol misuse costs based on the
changes in consumption reported in the second impact evaluation. The frequency and
amount of consumption reduced for 20-30 per cent of participants, while around 40-50
per cent of participants had at least some change in consumption. Importantly, the
proportional change in drinking is very similar across the program sites.
6.10 Perceived changes in consumption as a result of the CDC
East Kimberley
Goldfields
Ceduna and surrounds
Reduced amount of alcohol
Reduced frequency of drinking
Consumed more low-alcohol drinks
Stopped drinking altogether
None of the above
0
10
20
30
40
50
60
Share of participants (per cent)
Data source: Mavromaras K., Moskos M., Mahuteau S., Isherwood L., (2021) Evaluation of the Cashless Debit Card in Ceduna, East
Kimberley and the Goldfields Region, prepared by University of Adelaide, CIE.
326
However, it is crucial whether it is high or low risk/cost individuals that are decreasing
consumption. The impacts of all low risk participants stopping drinking would be much
lower or negligible in comparison to the impact of all moderate or higher risk participants
stopping drinking.
Data from the second impact evaluation suggests that participants with more risky
consumption habits were disproportionately represented among those that reduced
drinking (chart 6.11). For example, 15 per cent of participants that reduced the amount of
alcohol at any one time were in the very high risk category, but this category only
represented 10 per cent of the participant population.
Data about the changes in consumption by AUDIT score are presented separately by
program site in the second impact evaluation,75 but not replicated here for brevity.
6.11 Changes in consumption due to the CDC, by AUDIT score level
Data source: Mavromaras K., Moskos M., Mahuteau S., Isherwood L., (2021) Evaluation of the Cashless Debit Card in Ceduna, East
Kimberley and the Goldfields Region, prepared by University of Adelaide.
We estimate the proportional reduction in alcohol consumption in each community
(chart 6.12) as follows:
75 This is presented at Figure A 4-7, 4-8 and 4-9 in section 6 of the Quantitative Supplementary
Report: Mavromaras K., Moskos M., Mahuteau S., Isherwood L.,, 2021,
Evaluation of the
Cashless Debit Card in Ceduna, East Kimberley and the Goldfields Region — Quantitative
Supplementary Report, p.336, available at:
https://www.dss.gov.au/sites/default/files/documents/02_2021/fac_evaluation-cdc-ceduna-
east-kimberley-and-goldfields-quantitative-supplementary-report_012021.pdf
327
■
The share of the CDC population that reduced drinking: The share of the CDC
population that reduced drinking is the average among the share that reduced the
amount of alcohol consumed, share that reduced the frequency of drinking and share
that consumed more low-alcohol-drinks.76
■
Relative reduction of moderate-or-higher risk cohort compared to average
reduction across entire cohort: Based on the ratio between the share of moderate,
high and very high risk participants that report reductions in consumption to the share
of participants that report reductions in consumption.
■
Share of moderate or higher risk cohort that reduced drinking: This is the product of
the share of the population that reduced drinking and the relative reduction of the
moderate-or-higher risk cohort.
■
Reduction in drinking risk: This is calculated according to the following formula
1
𝑅𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑖𝑛 𝑟𝑖𝑠𝑘 = 𝑆ℎ𝑎𝑟𝑒 𝑟𝑒𝑑𝑢𝑐𝑖𝑛𝑔 𝑐𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 × (1 −
)
𝑅𝑅
where
– 𝑆ℎ𝑎𝑟𝑒 𝑟𝑒𝑑𝑢𝑐𝑖𝑛𝑔 𝑐𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 is the share of moderate-or-higher risk cohort that
reduced drinking
– 𝑅𝑅 is the relative risk of people with moderate-or-higher AUDIT score to those
with low but non-zero AUDIT scores. We assume this is equal to 2.7, which is the
average relative risk factor for a range of alcohol-related health issues, social
problems related to alcohol, and hospital admission.77
Appendix D provides the data underlying these calculations, and discusses the
limitations of the approach chosen to estimate these values.
6.12 Reduction in drinking risk among CDC cohort relative to the base case
Measure
East
Goldfields
Ceduna and
Average
Kimberley
surrounds
(applied to
Bundaberg
and Hervey
Bay)
Per cent
Per cent
Per cent
Share of CDC population that reduced drinking
20.7
17.0
16.7
Relative reduction of moderate-or-higher risk cohort
163.0
170.4
149.6
compared to average reduction across entire cohort.
Share of moderate-or-higher risk cohort that reduced
33.7
29.0
24.9
29.2
drinking
Reduction in drinking risk
21.4
18.4
15.8
18.5
Source: CIE.
76 We do not include the share that stopped drinking altogether in this calculation based on the
less than proportional share of people with a moderate, high or very high that indicated they
stopped drinking altogether.
77 See Appendix E, which explains CIE calculations to derive 2.7 from: Conigrave, K., Saunders,
J. and Reznik, R., ‘Predictive capacity of the AUDIT questionnaire for alcohol-related harm’,
Addiction, 1995(90), 1479-1485
328
Assuming these reductions in costs of alcohol misuse across the program sites implies a
total cost (without the CDC program) of $43.3 million (table 6.13).
6.13 Costs of alcohol misuse by participants under the base case
Site
2015/16 2016/17 2017/18 2018/19 2019/20 Total undisc.
Total disc.
$million
$million
$million
$million
$million
$million $million, NPV
Ceduna
1.0
1.3
1.4
1.6
1.7
7.0
6.7
East Kimberley
2.5
4.9
5.3
5.7
6.4
24.7
22.3
Goldfields
0.0
0.0
0.7
4.3
4.8
9.8
7.3
Bundaberg and Hervey
0.0
0.0
0.0
1.9
7.8
9.7
7.0
Bay
All sites
3.5
6.2
7.4
13.5
20.8
51.3
43.3
Note: ‘Undisc’ and ‘disc’ refer to undiscounted and discounted respectively.
Source: CIE.
The benefit of avoided costs of alcohol misuse relative to the base case are shown in table
6.14. Consistent with the original estimation of Manning, Smith and Mazerolle (2013),
these are split among productivity costs (42.1 per cent), traffic accident costs (25.5 per
cent), criminal justice system costs (20.6 per cent) and health system costs (11.7 per cent)
6.14 Benefit of avoided costs from alcohol misuse, relative to the base case
Site
2015/16 2016/17 2017/18 2018/19 2019/20 Total undisc.
Total disc.
$million
$million
$million
$million
$million
$million $million, NPV
Ceduna
0.2
0.2
0.2
0.2
0.3
1.1
1.1
East Kimberley
0.5
1.0
1.1
1.2
1.4
5.3
4.8
Goldfields
0.0
0.0
0.1
0.8
0.9
1.8
1.3
Bundaberg and Hervey
0.0
0.0
0.0
0.4
1.5
1.8
1.3
Bay
All sites
0.7
1.3
1.5
2.6
4.0
10.0
8.5
Note: ‘Undisc’ and ‘disc’ refer to undiscounted and discounted respectively.
Source: CIE.
Benefits from reduced cash availability
Cash availability is a distinct outcome from reduced spending on restricted goods and
services. A range of studies have estimated the relationship between cash availability and
crime and found robust evidence of a positive relationship. That is, less cash availability
is causally linked to less crime in a region (table 6.15).
There might be site specific factors that make benefit realisations more difficult. A study
conducted on mobility based on ethnographic research by Vincent and Klein in Ceduna
and East Kimberly, as well as regression analysis by Vincent, Markham and Klein in
329
Ceduna, provided evidence of displacement of local populations coinciding with the
introduction of the CDC scheme in the areas. The narrative accounts by Vincent and
Klein in Ceduna and East Kimberly were partly substantiated through a statistical
analysis that found evidence of net migration being 9.3 per cent lower in Ceduna,
Wyndham and Kununurra in comparison to similar towns without CDC and 5.2 per cent
lower compared to Australia as a whole.78
6.15 Estimates from the literature about cash availability affecting crime rates
Study name
Key findings
Quantitative outputs
Wright et al (2017)
■ Transition from check-based welfare payments to
Introduction of EBT led to
‘Less Cash, Less
Electronic Benefit Transfer (EBT) is associated with a
falls in crime of 9-13 per cent
Crime: Evidence from
decrease in street crime, including burglary and larceny
depending on the region
the Electronic Benefit
where EBT was introduced
■
Transfer Program
Likely explanation is that EBT reduced the amount of
cash on the streets available to be stolen or used for
illegal purposes.
Mai, H. , Cash,
■ While the abolition of cash will not eliminate shadow
Abolition of cash is likely to
freedom, and crime:
economy, it might shrink the size raise the cost of illegal reduce the size of the
Use and impact of cash
payments and reduce the size of the shadow economy.
shadow economy by an
in world going digital,
estimated 2 to 3 per cent.
■
EU Monitor
The shift from cash to electronic payments in Sweden
led to a significant decline in the number of bank
robberies and security van robberies and therefore less
cash could lead to fewer crimes related to cash
stealing.
Muyiwa et al, Impact of ■ The implementation of a cashless policy using
This study was based on the
cashless economy in
electronic-based transaction is expected to increase
survey participants
Nigeria
employment, reduces cash related robbery thereby
perception of benefits and
reducing the risk of carrying cash, reduces cash related therefore no quantitative
corruption and attracts foreign investment to the
outputs were reported.
country.
Data source: CIE and other studies as noted.
We have not included any benefits from reduced cash availability because evidence about
how safety outcomes have changed for the communities is thoroughly assessed in the
second impact evaluation, with a finding of mixed impacts. It may be that there are
benefits from reduced cash availability that are reducing crime, but that there are
counteracting impacts from the CDC increasing crime (such as decreased quality of life,
less autonomy, added stigma, or thefts associated with obtaining stolen goods to sell for
cash as a means of purchasing restricted items).
78 Vincent, E. Markham, F. and Klein, E. 2019, ‘“Moved on”? An exploratory study of the
Cashless Debit Card and Indigenous mobility’,
Aust Journal of Social Issues, 55, 27-39, available
at: https://onlinelibrary.wiley.com/doi/abs/10.1002/ajs4.84
330
7 Estimated costs of the CDC Program
The Australian Government is estimated to have spent $67.4 million (discounted) on
the CDC program since its inception.
This excludes the costs of other Support Services available to CDC participants, which
are intertwined with the program but separate from the impact of the CDC itself.
While participant access to wrap around services positively contributes to final
outcomes, it does come at a cost to the Australian Government. Across all program
sites, CDC participants had a weighted average of 160 per cent more attendances at
Support Services per person per year, compared to non-participants.
Costs to participants include the inconvenience of less available cash, and the costs
to mental health, essentially related to the association of being ‘on the Card’. The
costs of reduced cash are calculated to be small, and relatively negligible per person.
The costs to mental health are deemed to be inseparable to the mental health costs of
being unemployed, and no such costs are considered to be specifically related to the
Card.
Costs to the Australian Government
The costs of delivering the CDC program have been provided by DSS (table 7.1). The
cost categories presented are those that were provided by DSS. We have not included the
cost of Support Services, as described below.
7.1 Costs of the CDC Program borne by government until 2019/20
Cost item
2015/16 2016/17 2017/18 2018/19 2019/20
Total
Total
undiscounted discounted
$million
$million
$million
$million
$million
$million
$million
Card provider
5.8
4.0
5.0
9.3
14.9
39.0
32.9
Evaluation
0.3
1.0
0.5
1.1
1.9
4.8
4.0
Other (communications,
0.6
0.0
0.4
0.9
0.4
2.3
2.0
legal, consultancy)
Departmental
4.3
3.1
9.1
4.9
12.3
33.6
28.5
Total
11.0
8.1
15.1
16.2
29.4
79.8
67.4
Data source: CIE.
Extrapolation of costs to 2020/21 is presented in Appendix E, which is used in sensitivity
analysis shown in Chapter 8.
331
Engagement with Support Services
This cost-benefit analysis is limited to the Cashless Debit Card program, but a range of
other policy initiatives in the program areas were pursued concurrently. A key example
of such a policy initiative is the increased funding for and intended increase in take-up of
Support Services.
This additional funding was used to commission services such as drug and alcohol
rehabilitation services, financial management services, and family violence services.
These services were provided in addition to the existing services provided through
jobactive and the Community Development Program (CDP).
Various reviews have highlighted the additional need for such wraparound services,
particularly in regional and remove areas, and the significant benefits associated with
them. The need for additional local Support Services (such as drug and alcohol
rehabilitation services, financial management services, and family violence services as
mentioned above) was also one of the key reasons why some Indigenous leaders
supported the program within the community.79
However, it is important to note that the response to these additional wrap around
services has been mixed. Through the qualitative evidence gathered within the second
impact evaluation, respondents stated a lack of awareness of additional Support Services
that had been funded under the umbrella of the CDC in their locations. Although some
respondents were aware of these additional services, concerns were expressed that these
funds had not been targeted well. A local Indigenous leader who previously supported
the CDC program stated that the needed Support Services were introduced late and was
not appropriate.80
Support Services provided to welfare payment recipients in the CDC program areas
before the program are still considered in the base case.
Higher rate of support service engagement among CDC participants
We observe that take-up in 2019 of Support Services by CDC Program Participants is
higher than take-up by non-participants (chart 7.2). This includes all Support Services,
such as those provided through jobactive, CDP and the additional services funded
through the roll-out of the CDC in each region.
This comparison is shown for a selection of SA3s with the most CDC participants in
2019. For example, while non-participants each had around 0.6 support service
attendances in 2019, CDC program participants had on average more than 1.2
attendances.
79 Davey M., 2017,
Aboriginal leader withdraws support for cashless welfare card and says he feels used,
the Guardian, available a
t: https://www.theguardian.com/australia-
news/2017/aug/23/aboriginal-leader-withdraws-support-for-cashless-welfare-card-and-says-
he-feels-used
80 Davey M., 2017,
Aboriginal leader withdraws support for cashless welfare card and says he feels used,
the Guardian.
332
7.2 Engagement with Support Services by CDC participants and non-participants
2.0
r
Not CDC
CDC
1.8
pe
9
r)
1
0
ea 1.6
in 2
1.4
son/y
er
nces
/p 1.2
da
1.0
ten
nces
da 0.8
vice at
0.6
rt ser
0.4
rson (atten
po
pe 0.2
up
S
0.0
Goldfields
Bundaberg
Kimberley
Hervey Bay
Eyre Peninsula and
South West
Note: Support services include drug and alcohol rehabilitation services, financial management services, and family violence services.
Support services in this chart does not include jobactive and the CDP.
Data source: Data extracted from the Data Exchange by DSS.
Across all regions, CDC participants had a weighted average of 160 per cent more
attendances per person per year, compared to non-participants.81
There are a range of potential explanations for this higher rate of participation amongst
CDC participants:
■
This higher rate of engagement may partially reflect increased engagement as a direct
result of having the CDC. The second impact evaluation reported that many
stakeholders felt there was increased workload for local organisations, particularly to
assist participants with practical issues associated with the CDC.
■
The uplift in engagement may also reflect greater funding for local organisations
providing Support Services. This may translate into greater availability of Support
Services, and thus greater take-up. For instance, as part of the CDC rollout, each
region received additional funding for local Support Services.
■
Support service engagement may be higher among CDC participants because of their
other characteristics, rather than directly because they receive the CDC. Looking at a
selection of areas where the CDC program was implemented, there is a consistently
higher rate of support service attendance among people who eventually or current
have the CDC compared to those who never do (chart 7.3). This is more consistent
with CDC participants having higher engagement because of their characteristics,
rather than as a direct result of the CDC or an uplift in funding.
– Further, there is little evidence of a consistent change in support service
engagement correlated with timing of implementation of the CDC. Take-up by
CDC participants has been roughly similar or slightly falling since implementation
in Ceduna, while Bundaberg has experienced a steady increase during the past 5
81 This is a weighted average, with the weighting of the uplift in each SA3 being the number of
people in that SA3 in 2019 with the CDC.
333
years. This weakens the evidence for the CDC implementation being associated
with an uplift in engagement.
7.3 Support service attendance of eventual CDC participants
Bundaberg
Kimberley
Goldfields
Eyre Peninsula and South West
C
4.0
on
D
pers
a C
3.5
get
per
do 3.0
nces
tually
never 2.5
even
ho
w 2.0
ho
l attenda
w
ua
hose 1.5
those
and t 1.0
io of ann
een
w
0.5
Rat
bet
0.0
2015
2016
2017
2018
2019
2020
Note: Support services include drug and alcohol rehabilitation services, financial management services, and family violence services.
Support services in this chart does not include jobactive and the CDP.
Data source: CIE.
Implication of increased support service engagement for costs and benefits
Ideally, if the CDC drives uplift in support service attendances relative to the base case, it
would be ideal to measure the costs and benefits of this uplift. We have not done so for
the following reasons:
■
The DEX dataset provides individual-level data about the number and provider of
support service episodes, but the dataset does not categorise each support service. For
example, while we can identify that a support service attendance occurred at a
particular primary school or community centre, we cannot identify whether the
support service was related to financial support, alcohol or other drug counselling, or
something else. The type of support service is a key determinant of both costs and
benefits of the service.
■
We cannot estimate the share of uplift in Support Services attributable to the CDC
directly. Any concurrent but separate policy change to increase funding for Support
Services in the CDC communities would occur both in the base case and the policy
case.
■
The benefits of changed support service engagement may already be captured through
benefits we measure associated with reduced alcohol misuse and other benefits. That
is, as the second impact evaluation noted, the reduction in alcohol consumption could
not be attributed specifically to the CDC, but rather to the range of policies
implemented concurrently.
The choice not to add the costs of an uplift in Support Services to the other costs of the
CDC will result in the incremental costs of the CDC relative to the base case being
underestimated. However, our approach to estimation of benefits means that most of the
334
benefits of an uplift in Support Services are likely already being captured in other benefit
estimates included. The net effect of this assumption is that the net cost (benefit) of the
CDC Program will be underestimated (overestimated).
DSS have supplied data about the costs borne by DSS associated with the additional
Support Services funded alongside the CDC program (such as additional drug and
alcohol rehabilitation services, financial management services, and family violence
services as mentioned above). The total costs of these additional services was $6.8 million
between 2015/16 and 2019/20.82 However, almost none of this expenditure is associated
with Goldfields, despite it having more participants in 2018/19 than East Kimberley and
Ceduna combined, and no estimate has been provided of this expenditure in Bundaberg
and Hervey Bay.
7.4 Support Services expenditure (DSS component)
Goldfields
East Kimberley
Ceduna
2.0
1.8
1.6
-DSS
e
1.4
tur
ndi
illion) 1.2
m
pe
ex
t ($ 1.0
nen 0.8
vices
po
0.6
rt ser
com
po
0.4
up
S
0.2
0.0
2016/17
2017/18
2018/19
2019/20
Note: Support services include drug and alcohol rehabilitation services, financial management services, and family violence services.
Support services in this chart does not include jobactive and the CDP.
Data source: CIE.
Costs to participants
Previous CDC evaluations have explored the perceptions, views, and overall wellbeing of
participants after participating in the program.
Through these evaluations, participants raised various unintended consequences and
social concerns. These relate to feelings of discrimination, stigma, and embarrassment
from being on the Card. Table 7.5 summarises some of the key impacts reported by
participants, from the second impact evaluation.
82 This is based on a cost estimate supplied by DSS for ‘Support services expenditure - DSS
component’ over 2015/16 to 2019/20, and our extrapolation for 2020/21 based on the ratio of
Program participants in Sep-20 and Dec-20 to the number of Program participants in 2019/20.
335
7.5 Summary of participant’s feelings while on CDC
Impact from participants
All participants
Indigenous status
Gender
Feelings of discrimination
■ 57% of participants felt
■ Indigenous people were
■ There was no
discriminated against
no more or less likely to
significant difference
most or all the time
feel discriminated
between female and
compared to non-
male participants
indigenous people
Feelings of embarrassment ■ 58% of participants felt
■ There was a very small
■ There was no
embarrassed most or
increase in the
significant difference
all the time
proportion of non-
between female and
Indigenous people
male participants
feeling embarrassed
sometimes, most, or all
the time (77% compared
to 71%)
Feelings of unfair
■ 61% of participants felt
■ There was a very small
■ There was no
treatment
that being on the CDC
increase in the
significant difference
was unfair most or all
proportion of non-
between female and
the time
Indigenous people that
male participants
felt the CDC was unfair
sometimes, most, or all
of the time (77%
compared to 73%)
Data source: Mavromaras K., Moskos M., Mahuteau S., Isherwood L. 2021 ‘Evaluation of the cashless debit card in Ceduna, East
Kimberley and the Goldfields region: Consolidated report’.
The second impact evaluation concluded that approximately 75 per cent of participants
had negative feelings of discrimination, embarrassment, and unfairness from being on the
Card. These quantitative results were consistent with the qualitative evidence collected
through the second impact evaluation’s discussions with participants.
Based on the findings from the second impact evaluation, the convenience, social and
mental health costs have been summarised into two groups:
■
cost to participants from having limited access to cash, and
■
mental distress associated with participation.
Cost from limited access to cash
Throughout various sections of the second impact evaluation, the reduced ability to use
cash to purchase goods and services has been raised as a cost for participants. For
instance:
■
many expenses that are cash-dependent like school excursions, some bills and rents,
and some large denomination purchases like buying car and furniture, and
■
there is a limited ability to buy products in the second-hand market.
This limited ability to use cash as the preferred method of payment has been considered a
cost for participants.
336
Proportion of the population that use cash
A proportion of the national population continues to prefer to make payments using
cash, and cash payments make up a significant share of lower-value payments. For
instance, the Reserve Bank’s 2019 Consumer Payments Survey found that cash payments
make up 27 per cent of the total number of payment transactions in 2019.83
There are a wide range of reasons why people choose to use cash. However, one of the
main reasons is to assist in budgeting or to spend using their own (rather than borrowed)
funds.84 For high cash users,85 this benefit is true for nearly 50 per cent, and
approximately 15 per cent of all consumers.
This may be because when using cash, it is easier to recognise the financial impact by
physically taking the cash out of your pocket and giving it to someone else. However,
with electronic payments, it is easy to ‘tap’ a card without appreciation for the amount of
money that has been spent.
Within the Consumer Payments Survey, respondents were asked if they would be
affected if shops stopped accepting cash or if it became difficult to withdraw cash. From
this question, the majority of high cash users reported that they would experience a
“major inconvenience or genuine hardship” if cash were no longer available, compared
to approximately 25 per cent of all respondents.
However, there is a trend of people shifting away from cash payments.
■
27 per cent of all consumer payments were made with cash in 2019, compared with 37
per cent in 2016 and 69 per cent in 2007, and
■
Although cash payments still account for a significant share of small value
transactions, the introduction of credit and debit cards that can ‘tap and go’ has
shifted consumers preferences more towards using cards even for small transactions.
For instance, the share of transactions of $10 or less made in cash has reduced by 18
percentage points since 2016.86
Although Australian consumers in general are increasingly preferring to use electronic
payment methods, surveys such as the Consumer Payments Survey indicate that there is
still a preference for some consumers to continue to use cash. This is particularly true in
regional areas, which have a greater proportion of people in both ‘high cash user’ and
‘intermediate cash user’ categories.
83 Caddy J., Delaney l., Fisher C., Noone C., 2020, ‘Consumer Payment Behaviour in Australia’,
RBA Bulletin March 2020, available at:
https://www.rba.gov.au/publications/bulletin/2020/mar/consumer-payment-behaviour-in-
australia.html
84 Caddy J., Delaney l., Fisher C., Noone C., 2020, ‘Consumer Payment Behaviour in Australia’,
RBA Bulletin March 2020.
85 Those that use cash for over 80 per cent of transactions
86 Caddy J., Delaney l., Fisher C., Noone C., 2020, ‘Consumer Payment Behaviour in Australia’,
RBA Bulletin March 2020.
337
7.6 Comments from stakeholder consultations – Availability of EFTPOS a
concern
Some CDC communities prefer cash payments, partly because of the cost of operating
EFTPOS machines, and partly because of the reliability of cash payments.
For instance, one stakeholder mentioned that it is common for EFTPOS machines to
go offline in their community. When this happens, it can take several days before the
system is back online.
While EFTPOS is offline, businesses often rely on ‘IOUs’, for customers that cannot
provide cash payments, such as CDC participants. However, this creates a risk for the
business that the customer may not return to pay off their debt.
Cost per transaction method
There are a few Australian studies that considered the benefits and costs of payment
methods. For example, a 2005 study found that for a transaction of $50, the cost of
payment for cash was $1.64, compared to $0.80 for a debit card and 0.99 for a credit
card. However, for a $10 transaction, the cost of a cash-based payment type decreased to
0.96, while the costs for debit cards and credit cards remained the same.87
A more recent study undertaken by the Reserve Bank of Australia (RBA) in 2007
attempted to measure the long-run incremental resource cost of using cash, EFTPOS and
credit cards as payment methods. The analysis included costs such as communications
technology, producing cash, issuing cards, and withdrawing cash from ATMs. The study
found that cash is the lowest cost payment method for small transaction sizes, which
generally cash is most commonly used. However, the cost of a cash payments rises with
the value of the transaction, to the point where cash becomes more costly than EFTPOS
for payments above $50 in value. Table 7.7 below outlines the findings from this study.
7.7 Estimated payment method costs per transaction
Transaction size
Credit Card
EFTPOS
Cash a
($)
($, 2007/08)
($, 2007/08)
($, 2007/08)
10
0.80
0.50
0.31
20
0.82
0.50
0.35
50
0.86
0.52
0.66
100
0.94
0.54
0.70
200
1.10
0.59
0.75
500
1.57
0.73
1.42
a The source document applied two different approaches to calculate this cost per transaction. This table has taken the average of the
two approaches.
Source: Schwartz, C., Fabo, J., Bailey, O. and Carter, L. 2007
, Payment Costs in Australia, Table 14., see
https://www.rba.gov.au/payments-and-infrastructure/resources/publications/payments-au/paymts-sys-rev-conf/2007/7-payment-
costs.pdf
87 Simes R., Lancy A. and Harper, I., 2006,
Costs and Benefits of Alternative Payments Instruments in
Australia, Melbourne Business School Working Paper No 2006-08.
338
One limitation of this study is that is does not attempt to measure the benefits associated
with various payment methods.
Estimating the cost to participants
When estimating the cost for participants of having limited access to cash, it is important
to note that 20 per cent of a participant’s income support payment remains unrestricted
and that up to $200 could be externally transferred by a participant out of their Indue
account to their personal unrestricted account every 28 days. These funds could still be
withdrawn as cash.
Because of this, any cost imposed by CDC will need to consider that this portion of
income could still be used as cash.
When estimating the number of transactions that could have been made with cash, we
have considered that cash is used more often for small transactions. Based on
calculations from the RBA, the proportion of cash sales and the resulting estimated
number of CDC transactions that would have been cash are outlined in table 7.8.
7.8 Number of potential cash payments by size
Size of
Per cent of
Cash
Cash
Cash
Cash
Cash
transaction
transactions
payment
payment
payment
payment
payment
that used now on Card
now on Card
now on Card
now on Card
now on Card
cash (based
FY16
FY17
FY18
FY19
FY20
on 2019
values)
$
Per cent
Number
Number
Number
Number
Number
1-10
45
129 511
248 500
571 243
1 485 270
2 251 805
11-20
32
6 160
92 045
195 257
476 073
756 447
21-50
22
122 153
140 261
304 176
741 871
1 336 195
51-100
13
52 185
64 358
140 361
335 832
645 159
Source: Consumer Payment Behaviour in Australia, Bulletin March 2020, and CIE
To estimate the cost to participants from restricting the amount of cash available, we
have considered the total cost of transactions that would have previously been paid by
cash to the cost of transactions that would now be made through the Card.
As a proxy for the benefit received by consumers from using cash over a card payment,
we have assumed that the difference in the transaction costs between the two methods
represents the consumer benefit from using cash. This would be an underestimate of the
true cost, as non-financial costs have not been captured (such as the benefits from
budgeting).
Based on this approach, the cost for CDC participants from having 80 per cent of their
income not able to be withdrawn in cash is shown in table 7.9, and are considered to be
small, and comprise a relatively negligible impost per person.
339
7.9 Cost from restricting cash payments between 2015/16 and 2019/20
Transaction
2015/16
2016/17
2017/18
2018/19
2019/20
amount
$
$
$
$
$
$
10
26 220
50 310
115 651
300 700
455 888
20
809
12 082
25 630
62 492
99 295
50
1 067
1 225
2 656
6 479
11 669
100
359
443
966
2 311
4 439
Total cost
28 455
64 060
144 903
371 981
571 292
Data source: CIE.
Mental distress associated with participation
The second evaluation found that around 75 per cent of CDC participants felt
embarrassed, stigmatised and unfairly targeted by the program. They reported that
feelings of stigmatisation led to some CDC participants trying to hide the fact they were
on the Card and avoided their usual local shops.
This provides strong evidence that participants experience mental distress while being on
the Card. However, it is difficult to separate the feelings of stigma and disconnection with
the wider community from CDC and from being on income support and unemployed.
Many studies show that being unemployed has a negative health and social impact. For
instance:
■
after becoming unemployed, men experienced significantly greater symptoms of
depression and anxiety than employed men88
■
social stigma around unemployment had a highly corrosive negative impact on
people’s social and emotional wellbeing, such as symptoms of anxiety, depression and
feeling worthless89
■
unemployed people are stigmatised in the labour force and experience disadvantages
when applying for job vacancies.90 This is particularly true for long-term unemployed
people
88 Linn, M., Sandifer, R., and Stein, S., 1985, Effects of unemployment on mental and physical
health,
American Journal of Public Health 75(5), pp:502-506, available at:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1646287/
89 Anti-poverty Week, 2019,
The Stigma of Unemployment, available at:
https://antipovertyweek.org.au/2019/08/the-stigma-of-unemployment/
90 Krug, G., Drasch, K., Jungbauer-Gans, M. 2019, ‘The social stigma of unemployment:
Consequences of stigma consciousness on job search attitudes, behaviour and success’,
Journal
for Labour Market Research 53, available at:
https://labourmarketresearch.springeropen.com/articles/10.1186/s12651-019-0261-4
340
■
being unemployed leads to a drop in ‘status among friends, family and the
community, which can lead to loss of self-esteem91
■
becoming unemployed can affect mental and physical health, relationships and sense
of identity92, and
■
people experiencing unemployment are more than twice as likely to have feelings of
worthlessness, and not feel “reasonably happy”, and three times more likely to not
feel useful.93
Paying for items through the Card makes it more visible that an individual is receiving
income support. It is likely that this increased visibility strengthens the negative
associations from being on income support, further contributing to feelings of stigma and
embarrassment. However, we note that the Card provided has made changes to the
design of the card to limit this impact.
With the evidence available, the additional mental distress associated with the CDC
program cannot be separated from the negative impacts of being unemployed.
7.10 Comments from stakeholder consultations – The Card adds to the stigma of
being on income support
Stakeholders agreed that being unemployed and on income support can be
stigmatising and can lead to negative mental health impacts (such as anxiety,
depression, loss of confidence, disconnection with community, etc.). This is not an
impact of the CDC.
However, stakeholders stated that being on the Card makes it more obvious that an
individual is on income support, and adds to these negative mental issues. We heard
that this impact can be especially concerning for those participants that feel like they
do not suffer from drug or alcohol harms. These participants feel like they are being
socially judged and grouped together with people that are experiencing drug or
alcohol addiction.
Even though the Card provider has taken steps to make the Card not stand out from
other bank cards, many of the towns within the CDC program areas are small, and we
heard that ‘everyone knows what the card looks like’ and ‘everyone knows who is on
the Card’.
91 Institute for Work and Health, 2009,
Unemployment and mental health, available at:
https://www.iwh.on.ca/sites/iwh/files/iwh/reports/iwh_issue_briefing_mental_health_2009.
pdf
92 Beyond Blue, 2021,
Unemployment and mental health, available at:
https://www.beyondblue.org.au/the-facts/unemployment
93 Farre, L., Fasani, F., Mueller, H., 2018, ‘Feelings useless: the effect of unemployment on
mental health in the Great Recession’,
IZA Journal of Labor Economics, 7(8), available at:
https://izajole.springeropen.com/articles/10.1186/s40172-018-0068-5
341
7.11 Comments from stakeholder consultations – The ‘white card’
One stakeholder reported that CDC participants feel that the Card was ‘imposed upon
them’ and see it as a way of being controlled.
For example, some Indigenous communities refer to the Card as the “white card”-
referring feeling that the Card is a “top down” measure that is being imposed on them
by non-Indigenous people.
342
8 Cost-benefit analysis results
The CDC program is associated with a net cost of $57.4 million in present value
terms.
The main benefit category is reduced costs of alcohol misuse ($8.5 million), followed
by the social and community benefits of reduced gambling ($2.3 million). However,
these benefits are relatively small in comparison to the total costs of $68.8 million.
Total benefits were highest in East Kimberley, which has the highest costs of alcohol
misuse. Ceduna has relatively high benefits per person, due largely to gambling
reduction benefits.
A number of non-financial costs were supported by evidence but could not be
quantified, and were therefore excluded from the CBA model, including an uplift in
Support Services expenditure directly associated with the CDC, mental distress, and
disempowerment/lack of autonomy.
Similarly, there are likely to be benefits
associated with the CDC that have not been
valued because of weak evidence of attribution in the previous evaluations.
Results summary
The costs and benefits of the CDC program in the first four sites are shown in table 8.1.
These costs and benefits mostly accrue over the period from 2015/16 to 2019/20, except
costs associated with reduced alcohol misuse and child wellbeing, which accrue over a
longer period. For example, reduced loss of life costs from reduced drink driving
accidents includes the lost productivity over the remainder of an individual’s life.
8.1 Cost-benefit analysis results (2015/16 to 2019/20)
Cost/benefit item
Evidence of a Source of Quantified
Total
Total
clear change
evidence
undisc.
disc.
due to the
CDC?
$m $m, NPV
Costs
Costs of system administration by Indue
UoA
39.0
32.9
Cost of policy evaluation
UoA
4.8
4.0
Other communications, legal and consultancy costs
UoA
2.3
2.0
Other departmental costs attributable to the CDC
UoA
33.6
28.5
Program
Support Services expenditure - DSS component
CIE
0.0
0.0
343
Cost/benefit item
Evidence of a Source of Quantified
Total
Total
clear change
evidence
undisc.
disc.
due to the
CDC?
$m $m, NPV
Support Services expenditure - non-DSS component
CIE
0.0
0.0
Inconvenience to participants who prefer cash
CIE
1.2
1.0
Mental distress associated with participation
UoA
N/A
N/A
Disempowerment of vulnerable groups
UoA
N/A
N/A
Impact on small businesses
UoA
N/A
N/A
N/A
Total costs
80.9
68.3
Benefits
Alcohol misuse — criminal justice
UoA
2.0
1.7
Alcohol misuse — health system
UoA
1.1
1.0
Alcohol misuse — productivity
UoA
4.4
3.7
Alcohol misuse — traffic accidents
UoA
2.5
2.1
Reduced gambling
UoA
2.8
2.3
Child wellbeing — health
UoA
0.6
0.5
Child wellbeing — food
UoA
0.1
0.1
Child wellbeing — safety
UoA
-0.4
-0.3
Child wellbeing — education
UoA
-0.2
-0.1
Improved employment prospects
UoA/CIE
N/A
N/A
N/A
Improved health eating
CIE
N/A
N/A
N/A
Safety, crime and family violence
UoA
N/A
N/A
N/A
Housing and related services
UoA
N/A
N/A
N/A
Total benefits
13.0
10.9
Net results
Net benefit
-57.4
Benefit-cost ratio
0.16
Source: CIE.
344
8.2 Breakdown of net cost (up to 2019/20)
20
10
0
-10
illion, npv)
m -20
-30
-40
its or costs ($
ef -50
en
B -60
-70
Avoided cost of
Reduced social and Net child wellbeing
Costs
Net benefit or cost
alcohol misuse
community costs of
benefits
gambling
Data source: CIE.
Benefits by program site compared to costs
The program site with the greatest benefits is East Kimberley (chart 8.3 and 8.4), mainly
because it has the highest costs of alcohol misuse in the base case, and because
Bundaberg and Hervey Bay has only been recently implemented and have had less time
for benefits to be realised. Despite Bundaberg and Hervey Bay only having been recently
implemented, it has the same total benefit as Ceduna due to its higher count of
participants. Goldfields has a slightly smaller total benefit amount compared to these two
sites.
Note that while we have estimated benefits for each program site separately, cost data
provided by DSS have not been disaggregated by program site.
8.3 Total discounted benefits by Program site, compared to costs
Benefit category
Avoided
Reduced
Net child Total benefit
Costs Net benefit
cost of
social and
wellbeing
alcohol
community
benefits
misuse
costs of
gambling
$million, npv $million, npv $million, npv $million, npv $million, npv $million, npv
Ceduna
1.1
0.8
0.5
2.3
N/A
N/A
East Kimberley
4.8
0.8
0.1
5.6
N/A
N/A
Goldfields
1.3
0.2
-0.3
1.2
N/A
N/A
Bundaberg and Hervey Bay
1.3
0.6
-0.1
1.8
N/A
N/A
Total
8.5
2.3
0.1
10.9
68.3
-57.4
Source: CIE.
345
8.4 Total discounted benefits by Program site
Net child wellbeing benefits
Reduced social and community costs of gambling
Avoided cost of alcohol misuse
6.0
5.0
4.0
3.0
illion, npv)
m 2.0
it ($
ef
en 1.0
B
0.0
-1.0
Ceduna
East Kimberley
Goldfields
Bundaberg & Hervey Bay
Data source: CIE.
Benefits and costs per participant
As with total costs and benefits, the majority of benefits per person are avoided costs of
alcohol misuse, which are $419 per person (chart 8.5).
8.5 Breakdown of net cost per person
1000
pv)
500
son, n
0
er
/p
-500
-1000
rson ($
r pe -1500
-2000
-2500
its or costs pe -3000
ef
en
B -3500
Avoided cost of
Reduced social and Net child wellbeing
Costs
Net benefit or cost
alcohol misuse
community costs of
benefits
gambling
Data source: CIE.
However, the benefits per CDC participant94 (table 8.6 and chart 8.7) is quite different
across program sites. While East Kimberley still has the highest benefit in total, Ceduna
94 The denominator in the calculation of benefit per participant is the present value of the number
of total participants in the program. That is, it is the sum of the discounted number of
participants over 2015/16 to 2019/20. This results in a ‘discounted benefit per participant’.
346
has a high benefit per person with a particularly significant contribution from gambling
reduction benefits.
8.6 Discounted benefits per person by Program site, compared to costs
Benefit per participant
Avoided cost
Reduced
Net child Total benefit
Costs
Net benefit
of alcohol
social and
wellbeing
misuse
community
benefits
costs of
gambling
$/person,
$/person,
$/person,
$/person,
$/person,
$/person,
npv
npv
npv
npv
npv
npv
Ceduna
304
216
140
660
N/A
N/A
East Kimberley
900
135
19
1 053
N/A
N/A
Goldfields
237
110
- 57
289
N/A
N/A
Bundaberg and Hervey
226
108
- 25
309
N/A
N/A
Bay
Total
419
134
6
559
3 401
- 2 842
Note: The number of participants used to calculate benefits per person is also discounted.
Source: CIE.
The findings in 8.6 represent the average benefit and cost across the CDC participant
population.
8.7 Discounted benefits per person by Program site
Net child wellbeing benefits
Reduced social and community costs of gambling
Avoided cost of alcohol misuse
1 200
pv)1 000
son, n 800
er
/p 600
rson ($ 400
r pe
200
its pe
ef
0
en
B
- 200
Ceduna
East Kimberley
Goldfields
Bundaberg & Hervey Bay
Note: The number of participants used to calculate benefits per person is also discounted.
Data source: CIE.
Throughout all consultations, stakeholders were able to identify both benefits and costs
for CDC participants. Many of the benefits reported by stakeholders are hearsay or
anecdotal in nature. However, through these discussions, it was clear that not all
participants experience the same impact.
Stakeholder agreed that the impact for CDC participants would differ among three
general groups:
347
1 Those experiencing drug or alcohol related harms and benefit from Card –
Stakeholders described specific instances where participants were experiencing harms
from their alcohol, drug, or gambling consumption, and because of the Card, their
consumption and harms were reduced. Examples were provided of family members
who participated in Support Services funded by CDC, and they had a noticeable
change in spending behaviours and started to spend more on food for their families.
Other examples were provided of school children who came to school having eaten
breakfast and with a packed lunch.
a) However, some stakeholders mentioned that participants facing harms from
addiction and dependence still find ways to buy drugs and alcohol. These
participants may be experiencing the highest harm, but through creative means are
continuing to fund their alcohol or drug consumption.
2 Those who do not experience harms, but still benefit – There were some
participants who did not experience alcohol or drug related harms, but benefited from
the Card. For example, one stakeholder stated that “older CDC participants” in their
region have benefited greatly from the Card, however, not because of the reduced
expenditure on alcohol, but because of a reduction in elder abuse. For instance, before
the CDC program, family members previously stole cash from elderly family
members to fuel their own drug or alcohol use. With the CDC program, even if the
Card is stolen by family members, the Card could not be used to withdraw cash or
used to buy alcohol. There were other examples of participants who for the first time
had sufficient funds to pay for school excursions.
3 Those who do not experience harms, but do not benefit – Some participants felt like
they did not benefit from the restrictions imposed onto them from the Card, and
overall, the feelings of stigma from the Card outweighed any benefit they may have
received. For example, we heard examples of people who had successful careers and
were financially adapt, but after moving into a carer’s role, they commenced on the
Card and felt their behaviours was being unfairly restricted. Some of these participants
attempted to get off the Card without success. Stakeholders mentioned that many
CDC participants initially feel concerned when moving onto the Card. Although
some concerns are reduced over time for many, concerns regarding social
discrimination and stigma remained.
There is no systemic evidence available to categorise participants into these three groups
or to validate that these impacts are experienced by other participants. Further, there is a
complex relationship between alcohol consumption and costs of alcohol misuse
(discussed in Chapter 6). However, the second impact evaluation outlined on how
specific impacts relate to demographic characteristics.
For example, the Evaluation found that female participants were more likely to report a
reduction in alcohol consumption than male participants, and indigenous participants
were similarly more likely to report a reduction.95
95 Mavromaras K., Moskos M., Mahuteau S., Isherwood L., 2021, ‘
Evaluation of the cashless debit
card in Ceduna, East Kimberley and the Goldfields region: Consolidated report’, Future of employment
and skill research centre, The University of Adelaide, p.60.
348
Benefits and costs over time
The CDC program is associated with a net cost in all years (chart 8.8), but the benefit-
cost ratio fluctuates over time (table 8.9). The interaction between the growing scale of
the program and different patterns of benefits across sites results in a changing pattern of
overall net benefits, with a particularly large cost in 2015/16 relative to benefits
suggesting that start-up costs for the program were relatively higher.
8.8 Costs and benefits over time
40.0
Benefits
Costs
Net benefit (undiscounted)
) 30.0
m
20.0
it or cost ($ 10.0
nef
be
0.0
ted
-10.0
iscoun
Und -20.0
-30.0
2015/16
2016/17
2017/18
2018/19
2019/20
Note: All values are undiscounted.
Data source: CIE.
8.9 Cost-benefit analysis results for each year
Measure
2015/16
2016/17
2017/18
2018/19
2019/20
Total
$million, $million, npv
$million,
$million,
$million,
$million,
npv
npv
npv
npv
npv
Benefits
1.0
1.6
1.9
3.3
5.2
13.0
Costs
11.0
8.1
15.1
16.2
29.4
79.8
Net benefit (undiscounted)
-10.0
-6.5
-13.3
-12.9
-24.1
-66.8
Benefits
1.0
1.5
1.6
2.7
4.0
10.8
Costs
11.0
7.5
13.2
13.2
22.4
67.4
Net benefit (discounted)
-10.0
-6.1
-11.6
-10.5
-18.4
-56.6
Benefit-cost ratio
0.09
0.19
0.12
0.20
0.18
0.16
Source: CIE.
Sensitivity analysis
We test the sensitivity of the results to key assumptions by varying assumed parameters
one-at-a-time and measuring the net benefits under these alternative parameter values.
For example, we test how the results for the CBA are altered by assuming a lower
discount rate of 3 per cent. This also includes alternative approaches to the analysis to
349
understand the sensitivity of the results to the methodological approach in addition to the
parameter values.
The results for the sensitivity analysis show that the overall CBA results for the program
are not highly sensitive to the chosen range of assumptions:
■
A lower or higher discount rate has a negligible impact on the net benefit or benefit-
cost ratio, because the evaluation period of the analysis is relatively short. A key
caveat to this sensitivity analysis is that we cannot vary the discount rate for alcohol
misuse costs, because Manning, Smith and Mazerolle (2013) do not provide their
estimates of the societal costs of alcohol misuse in Australia under different assumed
discount rates.
■
The productivity costs of alcohol misuse may not be applicable to the cohort of CDC
participants, because these participants are often not employed. If the costs associated
with lost productivity (presenteeism and absenteeism) are excluded, the benefit-cost
ratio of the CDC program falls to 0.11. These costs are included in the central case
because there are costs to lost productivity associated with CDC participants that lose
future prospects of work and alcohol may be the cause of unemployment for CDC
participants.
■
Inclusion of costs and benefits for 2020/21 has little effect on the results, because the
benefits are similar in per person terms in 2020/21, and costs are projected for this
year to remain the same in per person terms as 2019/20 (see Appendix E). Data about
fixed and recurrent costs of the CDC program has not been provided by DSS, but if a
large share of costs are fixed, then the cost per person in 2020/21 may be lower than
previous years. This would result in an improved benefit-cost ratio for this sensitivity
test.
■
A higher relative risk of moderate-or-higher drinking risk (compared to low risk
drinking) would increase the benefits of alcohol misuse from $8.5 million to $12.0
million in present value terms. Varying this assumption doesn’t affect the conclusion
of the analysis that the costs of the CDC program outweigh the benefits. This
sensitivity test assumes a relative risk factor of 5.9 instead of 2.7 (as discussed in
Appendix D). This alternative parameter value implicitly assumes that the costs of
alcohol misuse are more closely related to prevalence of social problems from alcohol
rather than alcohol-related health problems, which we believe is a less plausible
assumption that the central case.
■
A larger base case cost of alcohol misuse in Bundaberg and Hervey Bay has little
impact on the overall results. This is mainly because the CDC was only more recently
implemented in Bundaberg and Hervey Bay, and benefits from this program site
accrue over fewer years as a result. For this sensitivity test we assume that the societal
cost of misuse of alcohol is $1 909 per person rather than $1 062, which is based on
the average cost per person across the other three program sites.
The benefit-cost ratio of 0.16 under the central case indicates that benefits would have to
be more than six times higher than estimated to result in a positive net benefit (i.e. a
benefit-cost ratio above 1). Similarly, for the program to have broken even between
2015/16 and 2019/20, the cost per participant would need to have been 84 per cent lower
at $540 per person.
350
8.10 Sensitivity analysis results
Sensitivity analysis case
Total
Alcohol Gambling
Child
Net
Benefit-
costs
misuse
wellbeing
benefit cost ratio
$m, NPV $m, NPV $m, NPV
$m, NPV $m, NPV
Ratio
Central case
68.8
8.5
2.3
0.1
-57.4
0.16
Low discount rate (3%)
75.1
9.7
2.6
0.1
-62.7
0.16
High discount rate (10%)
63.9
7.7
2.2
0.1
-53.9
0.16
Exclude productivity costs of alcohol misuse
68.3
4.8
2.3
0.1
-61.1
0.11
Include costs and benefits for 2020/21
92.0
11.5
3.3
0.1
-77.1
0.16
Higher relative risk of problem drinking
68.3
12.0
2.3
0.1
-53.9
0.21
Larger base case cost of alcohol misuse in
68.3
9.5
2.3
0.1
-56.4
0.18
Bundaberg and Hervey Bay
Source: CIE.
351
9 Conclusion
By setting aside 80 per cent of participant’s welfare payments to a restricted access bank
account, the CDC aims to reduce social harms caused by excessive consumption of
alcohol, illicit drugs, and gambling.
Within the first four CDC regions, the largest benefit was from a reduction in alcohol
related harms, with an estimated benefit value of $8.5 million between 2015/16 to
2019/20. These benefits are seen through improved productivity, reduced traffic
accidents, and reduced interactions with the criminal justice system and the health
system.
Other benefits include a reduction in gambling related harms ($2.3 million) and a small
net benefit for child welfare.
Total benefits and benefits per person were highest in East Kimberley, which has the
highest costs of alcohol misuse.
There are likely to be other impacts and benefits associated with the CDC that could not
been valued because of weak evidence of attribution in the comprehensive evaluations.
For instance:
■
There have been significant benefits for some individual participants and families.
However, the evidence does not indicate that these impacts have been experienced by
a statistically significant proportion of the population. For example, anecdotal
evidence highlighted that some families are spending more on food, and there are
more children attending school having eaten breakfast and with a packed lunch. This
points to the heterogenous nature of the participant population, with some
participants experiencing significant benefits, and others that are not.
■
There are some impacts that have inconclusive results and further data collection is
required to value the impact. For example, the impact of community safety and
consumption of illicit drugs is unclear.
■
Although there is evidence of stigmatisation of participants, it is difficult to isolate this
from the negative mental health impacts from being unemployed and on welfare
payments in general. However, the CDC program does make it more visible when
someone is on welfare payments, especially within small communities, which does
escalate these mental health concerns.
With the total program costs being $68.3 million in present value terms, the benefits were
not sufficient to outweigh the costs. Other non-financial costs were supported by
evidence but could not be quantified, and were therefore excluded from the CBA model,
including an uplift in Support Services expenditure directly associated with the CDC,
mental distress, and disempowerment/lack of autonomy.
352
This analysis found that the program has generated a net cost of $57.4 million in present
value terms, representing a benefit-cost ratio of 0.16. This indicates that the benefits
would have to be more than six times higher than estimated to result in a positive net
benefit.
This analysis draws from a wide range of data sources, including previous evaluations
and new analysis of DSS’s administrative data. Interpreting the CBA results should be
undertaken with care, and in consideration of the limitations outlined within this report.
353
A Data available about spending patterns
This appendix summarises the two most relevant data sources for spending patterns
across types of goods and services. These are the
Household Expenditure Survey (HES)
publication96 and the Selected Living Cost Indexes (SLCIs)97, both published by the
Australian Bureau of Statistics (ABS). The SLCIs are preferred because they are split by
type of household, as discussed below.
Household Expenditure Survey
The HES reports estimates of average weekly expenditure on different broad expenditure
groups, such as Current housing costs, Alcoholic beverages, and Clothing and footwear.
The ABS produces these estimates by asking a large sample of people to keep a diary
recording their expenditures over a short period (a week or so, from memory). They also
ask respondents to recall some big infrequent expenditures like motor vehicle purchases
outside of this short period, so they can be 'spread' over a longer period.
ABS reports estimated average weekly expenditure of Australians with their main source
of income being government pensions and allowances, with a breakdown into spending
categories (chart C.1). Similarly, the HES publication contains estimates income decile,
wealth decile, and for low economic resource households.
There are a range of limitations to drawing inferences from comparison between the
expenditure shares in the HES and expenditure shares from the CDC data:
■
Spending cannot be categorised by the type of good and service in the CDC dataset,
where only the merchant can be identified.
■
Spending shares for alcohol and tobacco are identified in the HES and the CDC data,
but because 20 per cent of income is provided to CDC program participants in cash,
this is not a full picture of CDC participant spending. Further, the HES data suggests
expenditure on alcohol and tobacco is significantly less than 20 per cent of weekly
expenditure, meaning that this average level of spending is still possible for CDC
participants.
96 Information about the HES is available at:
https://www.abs.gov.au/statistics/economy/finance/household-expenditure-survey-australia-
summary-results/latest-release
97 Australian Bureau of Statistics, 2017,
Selected Living Cost Indexes: 17th Series Weighting Pattern, Sep
2017, available at:
https://www.abs.gov.au/ausstats/abs@.nsf/PrimaryMainFeatures/6474.0?OpenDocument
354
■
Spending on gambling is not identified in the HES, but understood to be included in
the ‘recreation’ component.
■
Spending on illegal drugs cannot be identified in the HES.
■
HES data is not spatially disaggregated, so differences between regions will not be
visible to enable a more relevant comparison for each program location.
■
HES data for the ‘government pensions and allowances’ category of ‘main source of
income’ will include aged pension recipients, who may have very different spending
characteristics to recipients of other payments due to being older.
A.1 Expenditure shares for Australians with government pensions and allowances
as the main source of income
Medical care and
Miscellaneous goods
health expenses
and services
7%
5%
Recreation
9%
Transport
12%
Other
22%
Food and non-
alcoholic beverages
19%
Alcoholic beverages
2%
Tobacco products
Current housing
2%
costs (selected
dwelling)
22%
Data source: ABS
Household Expenditure Survey 2015-16.
Another key issue with the HES is that HES respondents tend to understate their
expenditure on alcohol and tobacco.98 The ABS produces the Consumer Price Index
(CPI) by adjusting the HES data to account for this underreporting (among other
adjustments) to obtain new CPI weights. However, the CPI weights are published by
capital city, but not by type of household, so they are not a useful comparator to CDC
program participants.
Selected Living Cost Indexes
The SLCIs provide a measure of the cost of living for each of four types of households.
To do this, they need a separate set of weights for each household. They produce the
weights in almost the same way as the CPI. The 'other government transfer recipient'
column (table A.2) refers to all households whose principal source of income is a
98 Australian Bureau of Statistics, 2019,
Consumer Price Index: Concepts, Sources and Methods, 2018,
available at:
https://www.abs.gov.au/AUSSTATS/abs@.nsf/Latestproducts/6461.0Main%20Features620
18?opendocument&tabname=Summary&prodno=6461.0&issue=2018&num=&view=
355
government pension or benefit other than the age pension or veterans affairs pension'.
The SLCIs are a more useful comparator because they are split by type of household.
A.2 Selected Living Cost Indexes
Commodity group
Pensioner
Employee
Age
Other
Self-
CPI
and
LCI
pensioner
government
funded
beneficiary
LCI
transfer
retiree
Living Cost
recipient LCI
LCI
Index (LCI)
Per cent
Per cent
Per cent
Per cent
Per cent
Per cent
Food and non-alcoholic beverages
18.28
16.45
19.21
17.59
15.54
15.75
Alcohol and tobacco
8.87
8.17
6.25
11.11
7.40
7.71
Clothing and footwear
3.16
3.41
2.92
3.35
2.65
3.23
Housing
23.33
14.82
20.32
26.00
12.40
22.93
Furnishings, household equipment
8.02
8.93
8.73
7.42
9.23
8.56
and services
Health
7.17
5.81
10.69
4.09
10.99
5.88
Transport
9.45
10.62
9.94
9.32
10.74
10.68
Communication
3.31
2.44
3.13
3.28
2.70
2.41
Recreation and culture
9.52
13.14
11.24
8.10
20.98
12.81
Education
1.88
4.45
0.15
3.35
1.24
4.44
Insurance and financial services
7.02
11.79
7.41
6.35
6.12
5.59
All groups
100.00
100.00
100.00
100.00
100.00
100.00
Note: Based on 2015-16 Household Expenditure Survey (HES) data and 2017-18 Household Final Consumption Expenditure (HFCE)
data. Figures may not add up due to rounding.
Source: ABS, CIE.
Alignment of SLCI and CDC spending categories
There is not a straightforward alignment between the SLCI weight categories and CDC
spending categories (table A.3). There are multiple categories of spending in the CDC
dataset that correspond to individual SLCI categories, and then a few categories from
each that do not have a corresponding category in the other dataset (e.g. insurance and
financial services, and ‘other’ in the CDC dataset). The alignment between furnishing,
household equipment and services in the SLCIs and the sum of utilities, pets,
department, discount and variety stores in the CDC dataset is particularly prone to error,
given that department, discount and variety stores sell many products apart from
furnishings and household equipment.
A.3 Alignment of SLCI and CDC categories
SLCI category
CDC category 1
CDC category 2
CDC category 3
CDC category 4
Housing
Housing
Non Card-Based
Transactions
Transport
Transport - Private
Transport - Public
Transport - Rental
Car
356
SLCI category
CDC category 1
CDC category 2
CDC category 3
CDC category 4
Food and non-
Food
alcoholic beverages
Recreation and
Recreation - Eating
Recreation - Goods
Recreation - Activities Holidays and travel
culture
out
and equipment
and memberships
Clothing and
Clothing and
footwear
footwear
Health
Medical
Education
Childcare/
Education/ Training/
Employment
Furnishings,
Utilities
Pets
Department,
household
Discount and Variety
equipment and
Stores
services
Insurance and
financial services
Alcohol and tobacco
Communication
Other
Other
Services
Services
Source: ABS Selected Living Cost Indexes, DSS, CIE.
357
B Detailed statistical modelling output
Modelling of spending shares
Full statistical output including tests of the significance of coefficients is presented in
table B.1. Note that a dummy variable is included for each combination of location and
spending category. For example, the average spending share for Bundaberg and Hervey
Bay clothing and footwear purchases is 9.22 per cent, and there is an trend decrease in
spending on clothing and footwear of 0.25 per cent per annum. The annual spending
trends are the key coefficients of interest, with other variables included as controls to
yield an accurate estimate of the common trend across locations for each spending
category.
B.1 Detailed statistical model output for regression predicting spending shares
Variable
Coef.
Std.
t
P>t
[95% Interva
name
Err.
Conf.
l]
Lower
Upper
Clothing and footwear
9.22
4.86
1.90
0.058
-0.31
18.74
Bundaberg
and Hervey Department, Discount and Variety Stores
63.39
4.86
13.06
0.000
53.87
72.91
Bay
Food
319.98
4.86
65.90
0.000 310.46 329.50
Holidays and travel
-0.67
4.86
-0.14
0.891 -10.19
8.85
Housing
12.62
4.86
2.60
0.009
3.09
22.14
Medical
9.23
4.86
1.90
0.057
-0.29
18.75
Non Card-Based Transactions
289.80
4.86
59.69
0.000 280.27 299.32
Other
111.15
4.86
22.89
0.000 101.63 120.67
Pets
-6.82
4.86
-1.40
0.160 -16.34
2.70
Recreation - Activities and memberships
10.38
4.86
2.14
0.033
0.85
19.90
Recreation - Eating out
62.91
4.86
12.96
0.000
53.39
72.44
Recreation - Goods and equipment
26.12
4.86
5.38
0.000
16.60
35.64
Services
38.48
4.86
7.92
0.000
28.95
48.00
Transport - Private
84.16
4.86
17.33
0.000
74.64
93.68
Transport - Public
-2.00
4.86
-0.41
0.680 -11.52
7.52
Transport - Rental Car
-10.04
4.86
-2.07
0.039 -19.56
-0.51
Utilities
-6.36
4.86
-1.31
0.190 -15.89
3.16
Childcare/Education/Training/Employment
-8.87
4.13
-2.15
0.032 -16.97
-0.77
Ceduna
Clothing and footwear
5.53
4.40
1.26
0.209
-3.10
14.16
Department, Discount and Variety Stores
30.67
4.40
6.97
0.000
22.04
39.30
Food
516.96
4.40 117.47
0.000 508.33 525.59
Holidays and travel
6.68
4.40
1.52
0.129
-1.95
15.31
358
Variable
Coef.
Std.
t
P>t
[95% Interva
name
Err.
Conf.
l]
Lower
Upper
Housing
28.83
4.40
6.55
0.000
20.20
37.46
Medical
5.55
4.40
1.26
0.208
-3.08
14.18
Non Card-Based Transactions
139.68
4.40
31.74
0.000 131.05 148.31
Other
22.96
4.40
5.22
0.000
14.33
31.59
Pets
-9.73
4.40
-2.21
0.027 -18.36
-1.10
Recreation - Activities and memberships
-6.50
4.40
-1.48
0.140 -15.13
2.13
Recreation - Eating out
30.16
4.40
6.85
0.000
21.53
38.79
Recreation - Goods and equipment
13.02
4.40
2.96
0.003
4.39
21.65
Services
46.74
4.40
10.62
0.000
38.11
55.37
Transport - Private
199.23
4.40
45.27
0.000 190.60 207.86
Transport - Public
4.72
4.40
1.07
0.284
-3.91
13.35
Transport - Rental Car
-10.26
4.40
-2.33
0.020 -18.89
-1.63
Utilities
-3.83
4.40
-0.87
0.385 -12.46
4.80
Childcare/Education/Training/Employment
-8.04
4.12
-1.95
0.051 -16.12
0.05
East
Kimberley
Clothing and footwear
11.57
4.39
2.64
0.008
2.97
20.18
Department, Discount and Variety Stores
38.37
4.39
8.75
0.000
29.77
46.98
Food
543.52
4.39 123.89
0.000 534.92 552.12
Holidays and travel
14.87
4.39
3.39
0.001
6.26
23.47
Housing
19.91
4.39
4.54
0.000
11.31
28.52
Medical
-1.26
4.39
-0.29
0.775
-9.86
7.35
Non Card-Based Transactions
111.36
4.39
25.39
0.000 102.76 119.97
Other
13.26
4.39
3.02
0.003
4.66
21.86
Pets
-9.50
4.39
-2.17
0.030 -18.10
-0.90
Recreation - Activities and memberships
-6.37
4.39
-1.45
0.146 -14.98
2.23
Recreation - Eating out
24.68
4.39
5.63
0.000
16.08
33.29
Recreation - Goods and equipment
13.55
4.39
3.09
0.002
4.94
22.15
Services
48.42
4.39
11.04
0.000
39.81
57.02
Transport - Private
186.85
4.39
42.59
0.000 178.25 195.46
Transport - Public
31.78
4.39
7.24
0.000
23.17
40.38
Transport - Rental Car
-10.15
4.39
-2.31
0.021 -18.76
-1.55
Utilities
-11.30
4.39
-2.58
0.010 -19.91
-2.70
Childcare/Education/Training/Employment
-7.14
4.26
-1.67
0.094 -15.50
1.22
Goldfields
Clothing and footwear
7.57
4.39
1.72
0.085
-1.04
16.18
Department, Discount and Variety Stores
53.83
4.39
12.26
0.000
45.22
62.45
Food
359.47
4.39
81.85
0.000 350.85 368.08
Holidays and travel
8.91
4.39
2.03
0.043
0.30
17.52
Housing
16.52
4.39
3.76
0.000
7.91
25.14
Medical
9.06
4.39
2.06
0.039
0.45
17.68
Non Card-Based Transactions
269.20
4.39
61.30
0.000 260.59 277.81
Other
53.77
4.39
12.24
0.000
45.16
62.38
359
Variable
Coef.
Std.
t
P>t
[95% Interva
name
Err.
Conf.
l]
Lower
Upper
Pets
-7.83
4.39
-1.78
0.075 -16.45
0.78
Recreation - Activities and memberships
0.64
4.39
0.14
0.885
-7.98
9.25
Recreation - Eating out
53.62
4.39
12.21
0.000
45.01
62.23
Recreation - Goods and equipment
25.66
4.39
5.84
0.000
17.05
34.27
Services
48.30
4.39
11.00
0.000
39.69
56.91
Transport - Private
126.53
4.39
28.81
0.000 117.92 135.15
Transport - Public
8.75
4.39
1.99
0.046
0.14
17.36
Transport - Rental Car
-10.02
4.39
-2.28
0.023 -18.63
-1.41
Utilities
-5.31
4.39
-1.21
0.227 -13.92
3.30
Childcare/Education/Training/Employment
0.02
0.09
0.22
0.829
-0.16
0.20
Annual
trend
Clothing and footwear
-0.25
0.09
-2.71
0.007
-0.43
-0.07
Department, Discount and Variety Stores
-0.06
0.09
-0.61
0.544
-0.23
0.12
Food
-3.34
0.09 -36.66
0.000
-3.52
-3.16
Holidays and travel
0.01
0.09
0.13
0.894
-0.17
0.19
Housing
-0.44
0.09
-4.79
0.000
-0.62
-0.26
Medical
-0.04
0.09
-0.49
0.626
-0.22
0.13
Non Card-Based Transactions
3.34
0.09
36.57
0.000
3.16
3.51
Other
0.64
0.09
7.04
0.000
0.46
0.82
Pets
0.00
0.09
-0.01
0.991
-0.18
0.18
Recreation - Activities and memberships
0.08
0.09
0.89
0.374
-0.10
0.26
Recreation - Eating out
0.29
0.09
3.16
0.002
0.11
0.47
Recreation - Goods and equipment
0.04
0.09
0.45
0.654
-0.14
0.22
Services
-0.21
0.09
-2.25
0.024
-0.38
-0.03
Transport - Private
-0.24
0.09
-2.65
0.008
-0.42
-0.06
Transport - Public
-0.03
0.09
-0.31
0.757
-0.21
0.15
Transport - Rental Car
0.00
0.09
-0.01
0.995
-0.18
0.18
Utilities
0.18
0.09
2.02
0.043
0.01
0.36
Constant
10.47
3.43
3.05
0.002
3.74
17.20
Data source: CIE.
Approach to modelling survival benefits
Survival analysis is an established method for analysing the determinants of duration for
unemployment spells. This enables us to construct a time to event dataset (in this case a
time to death) and to conduct a high-level survival analysis.
For this analysis, we have used a Cox proportional hazard regression model as the
preferred method to conduct the survival analysis, which allows for multiple coexisting
effects in one model (as opposed to the more popular Kaplan-Meier Curves) (box 6.7).
360
B.2 Survival Analysis
Survival Analysis is a set of statistical methods to estimate expected durations until
one and more events happen.99 Survival analysis requires time-to-event data and
special techniques as the data violates crucial assumptions for standard linear
regression models (for example non-normality or censoring, i.e., the event of interest
does not fall into the time span that we observe).
Common techniques for survival analysis include:
■
Life tables: (or actuarial tables) often used by insurance companies to derive the
probability to survive to a particular age or the remaining life expectancy
■
Survival functions: estimate the probability of surviving any past a point in time
■
Kaplan-Meier curves: estimate and visualise survival functions, and
■
Cox proportional hazard regression model: describe effects of categorical and
quantitative variables on survival.
For any statistical model, some assumptions are necessary and biases distorting
results are possible. In particular, models can suffer under endogeneity problems and
selection bias, which arise from the nature of treatments and health programs.
Broadly, endogeneity arises when variables are excluded, which have a potential
effect on both the independent and dependent variable. Selection bias occurs when a
selection of participants or patients is not random.
The DOMINO dataset allows us to estimate the time to an unemployment spell ending,
whether due to the person obtaining employment or some other reason.
Multivariate Cox regression
The model results suggest that there is not a statistically significant difference in the
probability of obtaining employment for a CDC participant compared to a non-
participant, after controlling for the characteristics of the participant such as their age
(table B.3). This is shown by the p-values on the CDC-coefficient well-exceeding 0.05, a
standard threshold for p-values to identify whether a result is statistically significant. This
means that being in the CDC cohort for any of the sites does not result in your hazard for
becoming employed (and ending the unemployment spell) deviating from the average.
That is, there is no statistically significant difference between whether someone becomes
employed depending on whether they are a CDC participant in any of the sites. The
result for CDC participation in East Kimberley has a positive point estimate that is
materially above 1, which provides some suggestive evidence that there may be a benefit
99 See Cassidy et al (2020) for an example of a recent application to the Australian unemployment
context: Cassidy, N., Chan, I., Gao, A. and Penrose, G., 2020, ‘Long-term Unemployment in
Australia’,
RBA Bulletin, December 2020, available at:
https://www.rba.gov.au/publications/bulletin/2020/dec/long-term-unemployment-in-
australia.html
361
of increased employment probability for participants at that site. Given the result is not
statistically significant, we do not rely on this as evidence of such benefits.
The sample size for this analysis is 75 181 individuals, among which 6 921 individuals
become employed at the end of their unemployment spell. The remainder have other
reasons for the end of the unemployment spell, such as death or moving overseas.
B.3 Cox proportional hazards model results for unemployment spells
Variable
Coefficient
Standard
z-
P-
95 per cent confidence
error score
value
interval
Lower bound Upper bound
CDC (Ceduna)
0.87
0.14
-0.88 0.380
0.63
1.19
CDC (Bundaberg and Hervey Bay)
0.97
0.07
-0.45 0.655
0.83
1.12
CDC (Goldfields)
1.05
0.11
0.44 0.657
0.85
1.30
CDC (East Kimberley)
1.23
0.17
1.51 0.132
0.94
1.60
Age
1.01
0.00
2.25 0.025
1.00
1.01
Age (18-29)
1.21
0.07
3.34 0.001
1.08
1.35
Age (30-44)
1.02
0.05
0.31 0.754
0.92
1.12
Age (45-64)
0.82
0.07
-2.33 0.020
0.69
0.97
Age (65+)
0.68
0.06
-4.16 0.000
0.57
0.82
Male
1.00
0.02
0.03 0.975
0.95
1.05
Born overseas (English-speaking
0.88
0.06
-1.76 0.079
0.76
1.01
country)
Born overseas (non-English-speaking
1.22
0.07
3.32 0.001
1.08
1.37
country)
Burnett
0.71
0.03
-7.84 0.000
0.65
0.77
Eyre Peninsula and South West
0.73
0.03
-7.02 0.000
0.67
0.80
Goldfields
0.57
0.04
-9.15 0.000
0.50
0.64
Gympie
0.74
0.03
-6.90 0.000
0.68
0.81
Hervey Bay
0.92
0.03
-2.28 0.023
0.85
0.99
Kimberley
0.40
0.02 -16.18 0.000
0.36
0.45
Maryborough
0.77
0.03
-6.11 0.000
0.71
0.84
Time trend based on start-date of spell
1.00
0.00
-0.72 0.471
1.00
1.00
Time trend based on start-year of spell
1.05
0.04
1.12 0.264
0.97
1.13
Source: CIE.
However, if the ‘hazard’ of being a CDC participant is not proportional to duration to the
event, the Cox proportional hazards model will not produce accurate coefficient
estimates. Visually, on survival curves, hazards will likely not be proportional if the
survival curves cross over. Testing for proportionality of hazards, we find we cannot
reject the assumption that hazards are proportional for Ceduna (p-value 0.43), Goldfields
(p-value 0.936) and East Kimberley (0.31), but not for Bundaberg and Hervey Bay
(0.0018).
362
C Declined transaction reasons
This study only includes declined transactions where the reason is associated with the
product/merchant type being disallowed. That is, we excluded declined transactions for
reasons such as having insufficient funds. Table C.1 provides details of which reasons for
declined transactions are assumed to be related to restricted item or merchant types.100
C.1 Reasons for a declined transaction related to restrict item types
Reason transaction declined
Reason relates to restricted
item types?
Card Not Present not allowed
No
Declined – Advised to Reject
Yes
Declined – Terminal in Excluded List
Yes
Declined – Terminal not in Approved List
Yes
Direct Debit Insufficient funds
No
Direct Debits have been stopped for merchant
Yes
Direct debits not allowed for this merchant
Yes
Transaction declined due to Card elapsing its expiry date
No
Transaction declined due to exceeding withdrawal limit
Yes
Transaction declined due to incorrect PIN entry
No
Transaction declined due to insufficient funds
No
Transaction declined due to restricted Merchant Category Code
Yes
Transaction declined due to terminal not on whitelist
Yes
Transaction declined due to terminal on blacklist
Yes
Transaction declined due to the Card being listed as lost
No
Transaction declined due to the Card being listed as restricted
No
100 One notable exclusion of a reason assumed to be related to restricted items is where a
transaction is declined because the merchant is not on the whitelist. In the early stages of the
CDC rollout, all merchants had to be whitelisted to be considered an approved merchant type,
but some (perhaps smaller) merchants weren’t whitelisted yet despite not selling restricted item
categories. We have excluded these transactions from counts of declined transactions related to
restricted item types, which will underestimate the number of declined transactions in early
periods that were related to restricted items. Including these transactions in counts of declined
transactions has little effect on the overall results, but does suggest a slightly weaker trend
increase in declined transactions related to restricted items.
363
Reason transaction declined
Reason relates to restricted
item types?
Transaction declined due to the Card being listed as stolen
No
Transaction declined due to the Card not being issued yet however it is embossed No
Source: CIE.
364
D Calculation of changes in alcohol consumption by
program site
Relative risk reduction among moderate-or-higher risk drinkers
Table D.1 presents the data underlying the calculation of relative risk reduction among
the cohort with an AUDIT score greater than or equal to 8 (i.e. Moderate, High and Very
High). This is a key input to estimating the reduction in costs of alcohol misuse
associated with the CDC and discussed in Chapter 6.
This approach uses proportions of the CDC population that report changes in various
measures of consumption, and maps this to a single change in costs. This faces the
following limitations:
■
Reported changes in consumption may not be an accurate estimate of actual changes
in consumption.
■
Reported changes in consumption cannot be attributed to the CDC alone, but rather
to the CDC along with concurrent policy changes such as the increase in provision of
Support Services.
■
Changes in each measure of consumption may have different impacts on cost of
alcohol misuse for that respondent.
– For example, someone who reduced the amount of alcohol at any one time may
reduce their consumption sufficiently to move to the low risk category if they drink
infrequently and the reduction in consumption was large. Alternatively, if they
drink frequently and made only a small reduction in alcohol consumption, this
may imply a negligible reduction in risk.
– Ultimately, the intention in using the approach of taking the average share across
reduced amount, frequency and alcohol concentration is to factor in the responses
to these questions and obtain a single estimate of the reduction in cost, which is
necessary because we have only a single estimate of the cost of alcohol misuse by
program site (rather than an estimate of alcohol misuse by AUDIT score, or cost
by amount/frequency/concentration of alcohol consumption). We use a simple
average across these three measures because:
… some combination of reduction in these three factors is likely to be associated
with a material reduction in risk, rather than merely a change in one of these
variables,
… academic literature often finds that those who do not drink at all sometimes
have less of a reduction in alcohol risk than those who are low drinkers.101
101 See, for example, Kuitunen-Paul and Roerecke (2018) which states “Compared to past year
abstainers (AUDIT=0), moderate drinkers had a lower mortality risk”: Kuitunen-Paul, S. and
365
… it avoids a misleading impression of precision in this estimate,
… it is consistent with other parts of the analysis.
■
Changes in only moderate, high and very high risk participants are counted, based on
the most relevant literature measuring differences in relative risk between people with
AUDIT scores greater than or equal to 8 (i.e. moderate, high or very high risk) and
those less than 8 (i.e. low risk).102
D.1 Estimation of the relative risk reduction in each program site
Measure
Low Moderate
High Very high
Average
Per cent Per cent Per cent Per cent Per cent
East Kimberley
Reduced amount of alcohol at any one time
14
42
22
22
Reduced frequency of drinking
12
47
15
26
Consumed more low-alcohol drinks
5
44
19
31
Stopped drinking altogether
50
10
22
17
Proportion in the CDC population
39
38
11
13
Average share across reduced amount, frequency and
10
44
19
26
alcohol concentration
Relative risk reduction of cohort
26
117
170
203
163
Goldfields
Reduced amount of alcohol at any one time
39
42
10
10
Reduced frequency of drinking
33
44
9
14
Consumed more low-alcohol drinks
41
39
8
12
Stopped drinking altogether
75
22
2
1
Proportion in the CDC population
63
23
5
8
Average share across reduced amount, frequency and
38
42
9
12
alcohol concentration
Relative risk reduction of cohort
60
181
180
150
170
Ceduna
Reduced amount of alcohol at any one time
49
38
6
7
Reduced frequency of drinking
34
21
16
30
Consumed more low-alcohol drinks
33
57
0
10
Stopped drinking altogether
65
24
11
0
Proportion in the CDC population
61
21
6
11
Average share across reduced amount, frequency and
39
39
7
16
alcohol concentration
Relative risk reduction of cohort
63
184
122
142
150
Source: Data from the Mavromaras K., Moskos M., Mahuteau S., Isherwood L., (2021)
Quantative Supplementary Report, CIE
calculations.
Roerecke, M., 2018, ‘Alcohol Use Disorders Identification Test (AUDIT) and mortality risk: a
systematic review and meta-analysis’,
Journal of Epidemiology & Community Health, available at:
https://pubmed.ncbi.nlm.nih.gov/29921648/
102 For example: Kuitunen-Paul, S. and Roerecke, M., 2018, ‘Alcohol Use Disorders
Identification Test (AUDIT) and mortality risk: a systematic review and meta-analysis’,
Journal
of Epidemiology & Community Health, available a
t: https://pubmed.ncbi.nlm.nih.gov/29921648/
366
Relative risk of moderate-or-higher drinking
The relationship between more risky drinking as measured by the AUDIT score and the
costs of alcohol misuse is complex. For example, the relationship between consumption
and absenteeism is likely to be different than the relationship between consumption and
liver disease.
However, for the purpose of this study, we estimate a single relative risk factor between
moderate-or-higher risk drinking and alcohol misuse, which is 2.7 (table D.2). This is the
average across multiple relative risk factors for the end-points shown in table D.2.
These relative risk factors are obtained from Conigrave, Saunders and Reznik (1995).103
It is a single study of 330 ambulatory care patients in Sydney.
This is not a recent study, and the literature on alcohol-related harms is developing,
notably with a greater understanding of the risks associated with lower levels of
consumption, and the magnitude of any benefits from low consumption relative to zero
consumption.104
However, it is the only study able to be identified that estimates the relationship between
AUDIT score and social harms, which based on this study have a much stronger
relationship with risky alcohol consumption than health-related harms. While there are
many studies examining the relationship between alcohol consumption and health issues,
a much smaller proportion specifically analyse the relationship between AUDIT score105
and these outcomes. Also, most studies identified that measure relationships between
AUDIT score and health outcomes were studies in the US context, which would have a
different relationship between drinking risk and costs. Therefore, we have preferred to
use Conigrave, Saunders and Reznik (1995), but test a higher relative risk factor in
sensitivity analysis. Some other studies identified are summarised below for comparison:
■
Kuitunen-Paul and Roerecke (2018):106 An AUDIT score of greater than or equal 8
was associated with elevated mortality risk after 2-10 years of follow-up, with a
relative risk factor of 1.24. This study was a comprehensive meta-analysis. It also
found that moderate drinkers had a similar or lower mortality risk compared to past-
year abstainers (relative risk of 0.75 in US Veterans studies and relative risk of 0.99 in
population-based studies).
103 Conigrave, K., Saunders, J. and Reznik, R., ‘Predictive capacity of the AUDIT
questionnaire for alcohol-related harm’,
Addiction, 1995(90), 1479-1485: Table 1
104 See the conclusions of: Iranpour, A. and Nakhaee, N., 2019, ‘A Review of Alcohol-Related
Harms: A Recent Update’,
Addict Health, 11(2): 129-137, available at:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6633071/
105 We use AUDIT score as the primary measure of alcohol use risk because it is measured
robustly in the second impact evaluation.
106 Kuitunen-Paul, S. and Roerecke, M., 2018, ‘Alcohol Use Disorders Identification Test
(AUDIT) and mortality risk: a systematic review and meta-analysis’,
Journal of Epidemiology &
Community Health, available at
: https://pubmed.ncbi.nlm.nih.gov/29921648/
367
■
Jia et al (2013):107 The average AUDIT score for general drivers in Guangzhou was
7.4, while for convicted drunk driving offenders it was 11.1, suggesting a moderate
level of alcohol problems and potentially a causal relationship.
■
Bradley et al (2016):108 A single cohort study of 24 Veterans Affairs systems in the
United States, which found positive relationships between higher AUDIT scores and
higher gastrointestinal hospitalisations and physical trauma. There were inconsistent
findings with respect to outcomes for patients who decreased to lower AUDIT score
risk groups at follow-up.
We have used a simple average in the absence of a reliable way to apply weightings to
these different endpoints. Health outcomes will likely be drivers of other categories of
alcohol costs. For example, hospital admission would be associated with the prevalence
of absenteeism, as would social problems related to alcohol.
In sensitivity analysis we have assumed that the relative risk factor is 5.9, based on only
the relative risk of social problems related to alcohol. This alternative assumption would
reflect the majority of costs of alcohol misuse being more closely associated with
prevalence of social problems rather than the range of health issues associated with
drinking. As discussed above, while health system costs are relatively small, much of the
remainder of societal costs of alcohol misuse are related to health outcomes, such as
traffic accidents (related to hospital admissions and trauma) and productivity (with
absenteeism perhaps being more closely related to health status rather than prevalence of
social problems).
D.2 Relative risk for moderate-or-higher risk consumption
Endpoint
Relative risk factor
Ratio
Death
2.4
Liver disease or gastrointestinal bleed
4.0
Elevated blood pressure
1.8
Trauma
1.8
One or more medical disorders which could be related to
1.8
alcohol
Social problems related to alcohol
5.9
Hospital admission
1.5
Simple average
2.7
Source: Conigrave, K., Saunders, J. and Reznik, R., ‘Predictive capacity of the AUDIT questionnaire for alcohol-related harm’,
Addiction,
1995(90), 1479-1485: Table 1, CIE calculation of average.
107 Jia, G., King, M., Sheehan, M., Fleiter, J., Ma, W., & Zhang, J., 2013, ‘Baseline study of
alcohol dependence among general drivers and drunk driving offenders in Guangzhou, China’,
Proceedings of the 2013 Australasian Road Safety Research, Policing and Education Conference, pp.1-13,
available a
t: https://eprints.qut.edu.au/63058/16/Paper_187_-_Jia_-
_Alcohol_and_Driving.pdf
108 Bradley, K., Rubinsky, A., Lapham, G., Berger, D., Bryson, C., Achtmeyer, C., Hawkins,
E., Chavez, L., Williams, E., and Kivlahan, D., 2016, ‘Predictive validity of clinical AUDIT-C
alcohol screening scores and changes in scores for three objective alcohol-related outcomes in a
Veterans Affairs population’,
Addiction 2016 Nov, 111(11), available at:
https://pubmed.ncbi.nlm.nih.gov/27349855/
368
E Extrapolation of costs to 2020/21
The most recently available cost data for the CDC is up to 30 June 2020. However, much
of the analysis undertaken throughout this CBA made use of participant data and benefits
up to 30 June 2021.
We have limited the analysis to 30 June 2020 so that it covers the period during which
both cost and benefit data is available. However, we estimate the total cost of the
program up to financial year 2020/21 for the purpose of sensitivity analysis in Chapter 8.
To estimate the cost, the relationship between the number of participants and the total
costs was considered. As shown below, as the number of participants increased over
time, the total costs increased (chart E.1).
E.1
Relationship of total costs and number of participants
Total costs (departmental and card provider)
35
30
2019/20
illion)
m 25
20
2017/18
2018/19
tal costs ($ 15
en
2015/16
rtm 10
2016/17
Depa 5
0
0
2 000
4 000
6 000
8 000
10 000
12 000
14 000
Number of payment recipients
Note: Each data point, including the number of participants, refers to a point in time.
Data source: CDC Program Data, CIE.
When considering the cost per participant, there is a declining trend, as shown below in
chart E.2.
The cost data for 2020/21 was not able to be provided by DSS because the data is
commercial in confidence and not yet publicly available. To accommodate this, the
analysis has estimated the 2020/21 costs.
To estimate the 2020/21 costs, we have applied the cost per participant value in 2019/20
to the number of payment recipients. This approach ensure that the additional year of
costs is proportional to the preceding year, rather than attempted to project based off the
previous financial years. For instance, when a new program site is being established,
369
there are a range of additional establishment costs. These establishment costs are
generally a one-off expense and are not expected to be ongoing.
However, there are limitations in this approach. For example, the marginal cost has been
falling over time since 2015/16 (i.e. the cost of each additional participant). By referring
only to the previous year, this approach does not allow for any efficiency improvements
that may have been made in 2020/21.
E.2
Cost per participant
2015/16
2016/17
2017/18
2018/19
2019/20
2020/21 projected
10 000
9 000
8 000
erson)
/p 7 000
6 000
5 000
r participant ($ 4 000
st pe 3 000
2 000
Total co 1 000
0
2015/16
2016/17
2017/18
2018/19
2019/20
2020/21
projected
Data source: CIE.
By applying the estimated cost per participant to the number of participants, we estimate
that the total cost in 2020/21 is $33.2 million (table E.3).
E.3
Costs of the CDC Program borne by government including projection for
2020/21
Cost item
15/16
16/17
17/18
18/19
19/20
20/21
Total
Total
(proj.)
undisc.
disc.
$million
$million
$million
$million
$million $million
$million
$million
Card provider
5.8
4.0
5.0
9.3
14.9
16.9
55.8
44.9
Evaluation
0.3
1.0
0.5
1.1
1.9
2.1
6.9
5.6
Other (communications,
0.6
0.0
0.4
0.9
0.4
0.4
2.7
2.3
legal, consultancy)
Departmental
4.3
3.1
9.1
4.9
12.3
13.9
47.5
38.3
Data source: CIE.
370
371
THE CENTRE FOR INTERNATIONAL ECONOMICS
www.TheCIE.com.au
372
26 November 2021
Mike Websdane
Cashless Welfare Engagement and Support Services Branch
Department of Social Services
71 Athllon Drive
Greenway ACT 2900
Dear Mr Websdane
RE: NOTED DATA LIMITATIONS IN OUR COST BENEFIT
ANALYSIS OF THE CASHLESS DEBIT CARD (CDC)
Thank you again for the opportunity to undertake a cost benefit analysis
(CBA) of the first four CDC program regions: Ceduna, East Kimberley,
Goldfields, and Bundaberg and Hervey Bay.
This note is provided to assist DSS interpret and communicate the CBA results
and qualifications.
During the project, collaboration with the DSS team identified a wide range of
benefits and costs for this analysis and potential metrics. This led to a wide
range of data sources being identified.
Key inputs were sourced from previous evaluations and reviews of the CDC
program. Where evidence gaps existed, we undertook statistical analysis of the
Data Over Multiple Individual Occurrences (DOMINO), Data Exchange
(DEX) and transactional data sets, and tested modelling assumptions and
inputs with service providers within each region.
Although we drew from the most recent data available, there were limitations
in the data. For instance:
■
existing data collections are not designed to measure economic impacts,
and focus on outcomes that are not necessarily mutually exclusive and/or
linked to specific individuals. As such, not all of the available evidence was
well suited to an economic analysis
■
the evidence base was typically limited to ‘averages’. This prevented the
separate measurement of impacts for those that benefited, and those that
did not, and
■
previous evidence and evaluations did not include all of the regions
considered in the CBA.
373
We are confident that the analysis contained in the CBA is robust. The depth of analysis and
qualifications made reflect the evidence base available. We accept that results may differ if data
limitations are addressed in future. We also note that the program has expanded into other regions
since this CBA was undertaken, and the applicability of our findings to these other regions is
unknown.
We would welcome the opportunity further discuss.
Yours sincerely
Sarina Lacey
Principal, Health Economics and Policy
