Detecting and preventing welfare fraud

Foreword | Another Trends & Issues paper (Prenzler 2011) identified key dimensions of welfare fraud in Australia and some of the social issues associated with efforts to reduce fraud. The present paper makes a more detailed examination of detection and prevention strategies, reporting on a range of initiatives that have been introduced progressively over the last three decades. These initiatives appear to be consistent with trends in comparable jurisdictions. Similarly, policymakers and anti-fraud practitioners in Australia are faced with challenges apparent in other welfare systems internationally. Strict eligibility tests, close monitoring of welfare recipients and a robust prosecution policy allegedly deter genuine applicants and are overly intrusive and punitive. At the same time, if all available detection and prevention techniques are not utilised, fraud can become prevalent and public confidence in the integrity of the welfare system may be reduced. This paper finds that the current suite of anti-fraud measures is successful in detecting numerous frauds and reducing losses through recoveries. At the same time, the enlargement of initial prevention measures is likely to see increasing interest and commitment from government.

Adam Tomison

Director

This paper examines anti-fraud measures currently operating in Australia’s welfare system, administered by the government agency, Centrelink. Using official data, an examination is made of the operations and rationales of different strategies and their impacts, including estimated savings. The paper covers nine strategies, including data-matching, public tip-offs, media campaigns, stepped-up investigations and recovery action. The findings indicate that Centrelink has adopted international best practice measures to combat fraud and appears to be particularly successful at detecting and stopping fraud. At the same time, the main challenge appears to lie in the area of finding and demonstrating more effective primary prevention measures.

The challenge of welfare fraud

In Australia and many other countries, the development of welfare systems during the post-World War II boom followed a similar pattern. Systems were enlarged on the back of rising prosperity in a process that extended an increasing number of types of benefits to an increasing number of claimants. During the initial stages of expansion, there were only limited controls on eligibility and systems were highly vulnerable to fraud (Reeve 2006). There were a number of reasons for this (Webb 2001):

• it was often difficult to check on the bona fides of applicants or to monitor recipients for changed circumstances;
• welfare providers were expected to be sensitive to the needs and circumstances of applicants; and
• anti-fraud measures generally entailed procedures that could deter and stigmatise recipients and delay urgently needed benefits.

The relatively open nature of welfare came under attack in the 1980s and 1990s following oil crises, ‘stagflation’ and the rise of ‘economic rationalism’ and ‘user pays’ philosophies (Bradbury 1988). There was also a growing interest across government in reducing losses from fraud, evidenced, for example, in the publication of the Commonwealth Fraud Control Policy in 1994 (AGD 2002). Public opinion was also set firmly against ‘dole cheats’ and others allegedly defrauding social security. Many governments made increasing use of means testing to reduce the costs of welfare and target it more efficiently to the most deserving.

In Australia, the Fraser Government (1975–1983) and Hawke–Keating Government (1983–1996) tightened compliance measures. The Howard Government (1996–2007) made welfare fraud a major election campaign issue and boosted resources to combat fraud (Dunlevy & Hannon 1997; Kingston 1996). The Rudd Government, which took office in November 2007, also committed itself to enlarging anti-fraud programs. In January 2008, Human Services Minister Joe Ludwig declared that:

The Rudd Labor Government is committed to ensuring people in need have access to adequate assistance. But we won’t tolerate people who abuse the community’s willingness to give them a hand up (Ludwig cited in Karvelas 2008: 4).

The government’s inaugural budget included an additional $138m to fight welfare fraud, with projected net savings from improved compliance of $600m over four years (Ludwig 2008b). In early 2009, Centrelink announced the allocation of $43.2m to ‘boost the anti-fraud, call centre and online infrastructure capacity’ (LeMay 2009: np). The Minister also signalled a greater focus on primary prevention through better initial compliance checks, with associated avoidance of debt (Ludwig 2008a). These moves were motivated in part by rising inflation, a commitment to reduce expenditures and a range of strong future demand pressures—including from the growing longevity of Australians and increasing tertiary education (Centrelink, 2007; Green 2008; Karvelis 2008).

Welfare fraud prevention research

A good deal of the independent and academic literature on welfare fraud is highly critical of anti-fraud measures as Draconian and punitive, and comes from the social work and civil libertarian areas (eg Marston & Walsh 2008; Mosher & Hermer 2005). Anti-fraud measures in welfare are, however, under-researched, especially in criminology and in terms of ‘what works’. One of the main exceptions to this is Kuhlhorn’s 1982 study of the implementation of data-matching in Sweden (Kuhlhorn 1997). The study was centred on the impact of a computer-based system for cross-referencing two databases. One consisted of income estimates for recipients of subsidised housing (the target group); the other contained estimates for recipients of sickness insurance. In the first case, applicants were motivated to understate their income in order to increase their housing subsidy. In the second case, they were motivated to overstate their income to increase their sickness benefit. In 1979, the first year of operation, 64,710 households were identified where housing subsidy income estimates were 1,000 Swedish Crowns or more below the medical insurance claim. This was 19 percent of households checked. These households were sent letters requesting explanations. The final outcome was that 9,170—or 2.7 percent—of ‘checked households’ lost all or part of their subsidy. A further 30,238 households lost all or part of their subsidy as a result of voluntarily reporting changes in their income. In total, 6.1 percent of households whose details were cross-referenced had their subsidy removed or downgraded.

Kuhlhorn (1997: 238) described the cross-referencing capacity of the new Swedish system as a ‘crime prevention Eldorado’. He did not describe any prosecutions resulting from the cases that appeared to be deliberately falsified. The study did, however, report the findings of a national opinion poll on the legitimacy of the program, which found that 94 percent of people thought the checks were appropriate and 94 percent thought they would motivate people to more accurately report their income. Of recipients of the housing subsidy in the sample, 87 percent supported the checks.

An early US study on the introduction in 1982 of computer-based ‘wage-matching’ systems for clients on a food stamp and a family support program estimated total savings from halted payments and restitution in four counties at US$4.3m. The total systems costs were US$2.2m. Net savings were therefore substantial, varying between 56 percent and 100 percent in the individual sites (Greenberg, Wolf & Pfiester 1986).

Reeve (2006) compared anti-welfare fraud initiatives in Australia and the United Kingdom. From 2000 in the United Kingdom, the Department for Work and Pensions introduced centralised service delivery, enlarged risk assessments, more follow up investigations of suspicious claims and more aggressive prosecution and recovery actions. Between 2001 and 2005, it estimated benefit fraud was reduced by 50 percent. However, Reeve (2006) argued that this estimate was highly speculative and that there was no real evidence of a deterrent effect from prosecutions. Reeve (2006) was supportive of similar Australian initiatives as responsible cost-effective measures to protect public money, but argued there was scope for improved administrative procedures to maximise primary prevention. He advocated moving beyond a ‘detection-investigation-prosecution model’ to an ‘intelligence-led model’ based on an ‘IT-enabled-transformation of administrative services’ (Reeve 2006: 44). This would include wider data-matching processes between more government agencies and improved software that allows ‘real time transaction monitoring’ of applications across government and bank databases (Reeve 2006: 44).

Increased efforts to combat welfare fraud can be more effective in detecting and stopping existing fraud (secondary prevention) than preventing the initial onset of fraud (primary prevention). In theory, prosecutions might increase for a period after detection and prevention initiatives are introduced but then the number should start to reduce as the system matures.

This is a key implication of Reeve’s (2006) study. However, enhanced detection mechanisms can mean that offenders taken out of the system are simply replaced by new entrants, creating an expensive ‘revolving door’ phenomenon, rather than an effective prevention system.

Source: Centrelink unpublished data 2010

Australian initiatives

Australia has followed overseas developments in the area of welfare fraud detection and prevention in areas such as data-matching, surveillance, identification checks, forensic accounting and communicating rules (Prenzler forthcoming, Reeve 2006). The last 30 years have seen a great deal of innovation. All government chief executive officers have obligations to minimise fraud under the Commonwealth Financial Management and Accountability Act 1997 and the Commonwealth Fraud Control Guidelines (AGD 2002). The establishment of Centrelink in 1997 as the main service delivery agency for welfare benefits allowed for the centralisation and standardisation of anti-fraud methods—although Centrelink’s partner agencies also conduct their own anti-fraud programs.

The following section reviews anti-fraud strategies utilised by Centrelink (located within the Department of Human Services portfolio). It outlines the background and operation of each strategy and reports available data on impacts and cost effectiveness. All estimated savings are gross except where otherwise indicated. The information was supplied by Centrelink.

Data-matching program

The Data-Matching Agency was established in 1991. Its governing legislation included mandatory tax file declarations to facilitate data-matching between government agencies. The main agencies were the then Department of Social Security and the Australian Taxation Office (Centrelink & The Data-matching Agency 2006). The purpose of data matching is to ensure that information about a customer held by an income support agency is consistent with that held by other agencies. Centrelink’s current data-matching program involves the Australian Taxation Office (ATO), the Department of Veterans’ Affairs and Centrelink. There are three main types of matching.

• Payment matching is designed to ensure customers are not ‘double-dipping’ or receiving a payment that might be precluded by another payment.
• Income matching allows customers’ declared income to be checked against income data held elsewhere, such as with the ATO. This also applies to relevant partners and parents.
• Personal Identity Discrepancies works by comparing personal identity information held by Centrelink for a specific Tax File Number with the personal identity details held by the ATO.

In 2008–09, Centrelink conducted four data-matching cycles, involving 53,643 reviews. The corrections resulted in fortnightly savings of $786,057, which totalled $112.5m in debt recovery actions against Centrelink customers for the year.

Other data-matching initiatives

Centrelink is also involved in numerous data-matching exercises outside the legislated requirements of the data-matching program. Information on the operations of these projects is shown in Table 1. Most of the detected errors are managed administratively; others involve fraud investigations. Checks are frequently conducted on a weekly basis. Partner agencies include Australian Government departments, state and territory corrective services departments, higher education institutions and financial institutions.

Fraud tip-off line

The Centrelink electronic Tip-off Recording System was introduced in 1998 and a toll free telephone number was introduced in 2005. Disclosures can also be made by mail and email. Public tip-offs received in 2008–09 triggered 50,277 Centrelink reviews and debts and savings of $119.3m. An Australian National Audit Office (ANAO) review of the system reported that in 2007–08 Centrelink received 101,595 tip-offs, which were related to seven percent of all investigations into non-compliance and fraud; with 16.2 percent of tip-offs resulting in an alteration to a customer’s payment and/or a debt being raised (ANAO 2008a). Centrelink agreed to better protect the privacy of customers and the privacy and safety of informants.

Media campaign

In 2005, Centrelink began a four year media campaign—Support the System that Supports You—which encouraged customers to report changes in their circumstances that might affect their entitlements. By 2008, the campaign resulted in 294,000 reports and a further 29,000 tip-offs.

Stepped up investigations

In 2008–09, tip-offs, data-matching and other triggers led to 26,084 formal investigations of possible fraud. Outcomes included $113.4m in savings and debt; with 5,082 matters referred to the Commonwealth Director of Public Prosecutions. Of the latter, 3,388 cases were prosecuted with 3,354 convictions—a conviction rate of 98.9 percent. The average saving per investigation was calculated at $4,347. In 2008–09, Centrelink adopted an ‘intelligence-led model’ of investigations which makes early assessments of the probity of cases and prioritises cases most likely to yield adequate evidence for prosecution. This contributed to a reduced number of investigations.

Investigation capabilities have been enhanced by the creation of specialist intelligence analyst positions and through shared intelligence between Centrelink and law enforcement agencies. In 2008–09, there were 10 Australian Federal Police agents posted to Centrelink Fraud Investigation Teams and two Centrelink intelligence officers posted to the Australian Crime Commission. In 2008–09, the Australia Transaction Reports and Analysis Centre supplied information to Centrelink in 2,251 cases, resulting in an estimated $8.1m of annualised savings (AUSTRAC 2009). Overseas investigations are enhanced by Australia’s participation in the Windsor Agreement on intelligence sharing with New Zealand, the United Kingdom, Canada and the United States. Centrelink also reports on a number of sub-types of investigations, as outlined below.

Cash economy investigations

These fraud investigations are targeted at welfare clients who receive ‘cash-in-hand’ payments. The investigations are targeted at types of work and locations associated with the cash economy, including harvesting (‘fruit picking’) and hospitality. Joint field operations were conducted with a variety of agencies including the ATO, Australian Federal Police and the Department of Immigration and Citizenship. Centrelink is also a member of the Joint Agency Strategic Cash Economy Working Group that includes the ATO, Department of Education, Employment and Workplace Relations and the Department of Immigration and Citizenship. In 2008–09, intelligence processes directed attention to private security, labour hire, fishing and restaurants and cafes. The year saw 7,925 cash economy investigations, including the investigation of 124 cash economy operations, with $15.4m in savings and debts. In 2008, field operations in Sydney and Melbourne airports and the Rocks in Sydney identified 75 taxi drivers who were receiving unemployment benefits (Karvelas 2008).

Centrelink has recently implemented a new cash economy business model broadly based on an intelligence-driven model operating across compliance and fraud investigation activities. The model includes a greater focus on working with industry to improve procedures in collecting information, educating and reminding businesses of obligations in relation to fraud prevention in a manner that is the least intrusive to business. This is expected to result in a significant reduction in the need to undertake cash economy operations in the field.

Identity-related fraud investigations

Identity fraud involves offenders stealing, borrowing, fabricating or altering identities to obtain illegitimate payments. Offenders may make use of highly sophisticated forgery tools. Centrelink deploys a specialist Identity Fraud Detection Team whose capacities include advanced computer equipment and skills. In 2008–09, 3,873 investigations were conducted into possible identity fraud, with 166 referrals for prosecution in the same year and $15.1m in debts and savings.

Optical surveillance

Covert or ‘optical’ surveillance was adopted as an ‘Enhanced Investigation Initiative’ by Centrelink in 1999. Cases of suspected benefit fraud amenable to this type of examination are outsourced to a panel of private investigation firms. Typical cases under this initiative would involve a disability payment recipient who is suspected of overstating their disability or having no disability. Another scenario would be that of a person working for cash and receiving unemployment benefits. In the first year of operation, 1,063 cases were finalised with 70 percent leading to almost $4m in payments targeted for recovery. In 2008–09, 1,023 surveillance operations were completed; 589 or 57.5 percent were considered ‘actionable’, leading to annualised reductions in payments of $5.5m and debt recovery actions of $21.2m against persons owing money to Centrelink. Total savings were estimated at $26.7m or $26,126 per investigation, with an ‘effectiveness indicator’ rating of 72.5 percent.

Professionalisation

Centrelink investigator training is compliant with the Commonwealth Fraud Control Guidelines 2002, which require certificate level competencies for operatives and diploma level competencies for managers (AGD 2002). Centrelink is also currently positioning all investigators at the APS 5 level, with team leaders that manage the casework of up to eight investigators.

Stepped up recovery action

Centrelink upgraded its debt collection processes following ANAO reviews (ANAO 2008b, 2005). The second audit in 2007 found Centrelink had implemented better recovery strategies, which included increased resources devoted to debt recovery and enlarged options to facilitate repayment (eg Australia Post, BPay, telephone, mail, internet and direct debit). Centrelink also outsourced selected difficult cases to private debt recovery agencies. At the same time, debt had increased ‘rapidly’ from $967m in 2003 to $1.3b in 2007, with 650,000 customers in debt in 2007 (ANAO 2008b: 15). In 2008–09, Centrelink met a target of 70 percent recoveries, or $1.9b for the year.

Stepped-up identity verification checks

A 2007 audit of a sample of Centrelink customer records by the ANAO (2007) noted significant improvements to proof of identity (POI) arrangements since the introduction of the current tiered POI model in 2001. Centrelink further strengthened its POI arrangements during 2009 while retaining alignment with the policy introduced in 2001 as part of the ‘Whole of Government Identity Framework’. Original documents continue to be sighted by Centrelink Customer Service Advisors and details of identity documentation are captured as part of the process. Centrelink’s current ‘New Claim’ POI requirements generally exceed those of banks and financial institutions (Centrelink, personal communication 2010). For example, Centrelink does not generally accept certified copies of identity documents.

Business integrity process redesign

In 2008–09, Centrelink began a roll out of digital holdings for all documentation to allow staff anywhere in Australia to access documents without the need to locate paper copies. This enables more efficient and comprehensive compliance reviews and ‘real time’ verification at the time of application.

Discussion

Delivery of welfare payments and the prevention of fraud involve a difficult balancing act. On the one hand, there are obligations related to the protection of customers’ privacy, the speedy delivery of benefits and the avoidance of additional hardship to customers through investigation and debt recovery action. Alternatively, there is a legal and ethical duty to ensure taxpayers’ dollars go to genuine recipients.

In that regard, accountability occurs through a number of mechanisms. All Centrelink customers can have their claims reviewed internally as well as appeal to the Social Security Appeals Tribunal and the Administrative Appeals Tribunal. Centrelink’s complaints handling system had been commended by the ANAO for ‘accessibility, responsiveness and objectivity’ (ANAO 2009: 15). Customers can also complain to the Commonwealth Ombudsman, who received 7,226 complaints and approaches about Centrelink in 2008–09. While ‘complaint themes’ included ‘debt management’, they did not specifically include fraud-related matters (Commonwealth Ombudsman 2009: 59). The Ombudsman observed that Centrelink is ‘generally very responsive’ to his interventions and often resolves matters within 24 hours (Commonwealth Ombudsman 2009: 60). Centrelink is also subject to audits by the ANAO and is subject to privacy and freedom of information laws.

Centrelink appears to be in step with best practice in fraud management internationally, including by giving increasing attention to prevention through diverse initiatives such as better initial applicant identification checks and facilitating communication about clients’ changed circumstances (cf ANAO 2006; Reeve 2006; SNCCP 2008). However, the ongoing detection of fraud indicates the need for better methods of primary prevention (Ludwig 2008b; Prenzler forthcoming). Recent initiatives in data-matching with financial institutions and in ‘real time’ identity verification procedures offer considerable potential. A number of other options are available. It is possible that more could be done in the area of diagnostics, by categorising fraud cases in more detail and analysing offender methods. Resources could then be targeted at the most frequent or most expensive types of frauds to close off opportunities. A study of the motivations of offenders would also be useful to identify the extent to which lower end benefit and salary levels, and levels of personal debt, provide incentives for fraud (Marston & Walsh 2008). Questions of the legitimacy and effectiveness of the system would also be well served by more surveys, including surveys of customers (Kuhlhorn 1997). The costs of detection and prevention systems could also be set against estimated gross savings to identify the more effective strategies that might be enlarged (cf Greenberg, Wolf & Pfiester 1986).

Conclusion

An examination of Centrelink’s detection and prevention strategies indicate the current approach is showing considerable impact in secondary prevention—detecting fraud and stopping its continuation—and in recovery orders. There also appear to have been some promising impacts in primary prevention—including through enhanced identification verification procedures and the advertising and communication of rules. This is likely to be the area where fraud management will be concentrated in the future, with a view to increasing primary prevention and reducing the ‘downstream’ costs of enforcement.

References

All URLs correct at April 2011

• Attorney-General’s Department (AGD) 2008. Prosecution policy of the Commonwealth. Canberra: AGD
• Attorney-General’s Department (AGD) 2002. Commonwealth fraud control guidelines. Canberra: AGD
• AUSTRAC 2009. Annual report. Canberra: AUSTRAC
• Australian National Audit Office (ANAO) 2009. Centrelink’s complaint handling system. Canberra: ANAO
• Australian National Audit Office (ANAO) 2008a. Centrelink’s tip-off system. Canberra: ANAO
• Australian National Audit Office (ANAO) 2008b. Management of customer debt—Follow-up audit (Centrelink). Canberra: ANAO
• Australian National Audit Office (ANAO) 2007. Proof of identity for accessing Centrelink payments. Canberra: ANAO
• Australian National Audit Office (ANAO) 2006. Assuring Centrelink payments—The role of the random sample survey programme. Canberra: ANAO
• Australian National Audit Office (ANAO) 2005. Management of customer debt (Centrelink). Canberra: ANAO
• Bradbury B 1988. Welfare fraud, work incentives and income support for the unemployed. Sydney: Social Welfare Research Centre, University of New South Wales
• Centrelink 2007. Annual report 2006–07. Canberra: Centrelink
• Centrelink & The Data-matching Agency 2006. Data-matching program: Report on progress 2005–06. Canberra: Centrelink
• Commonwealth Ombudsman 2009. Annual report 2008–09. Canberra: Commonwealth Ombudsman
• Dunlevy S & Hannon K 1997. Red handed—330,000 welfare payments stopped. Daily Telegraph 30 October: 1
• Green K 2008. Child poverty and welfare reform. Presentation to Brotherhood of St. Laurence: Melbourne: April 7
• Greenberg D, Wolf D & Pfiester J 1986. Using computers to combat welfare fraud: The operation and effectiveness of wage matching. Westport, CT: Greenwood Publishing
• Karvelis P 2008. Welfare fraud on Rudd’s hit list. The Australian January 25: 4
• Kingston M 1996. $19m in savings show little room for welfare crackdown. Sydney Morning Herald 21 August: 6
• Kuhlhorn E 1997. Housing allowances in a welfare society: Reducing the temptation to cheat, in Clarke R (ed), Situational crime prevention: Successful case studies. Monsey, NY: Criminal Justice Press
• LeMay R 2009. Budget ’09: Centrelink’s tech boost. www.zdnet.com.au/news/security/print.htm?TYPE=story&AT=339296400-130  
• Ludwig J 2008a. Address to smart cards summit. Ministerial speech 11 June. http://www.dhs.gov.au/media/archives/ludwig/speeches/080611-Ministerial-...
• Ludwig J 2008b. Combating welfare fraud to keep the system fair. Media release 13 May. http://www.dhs.gov.au/media/archives/ludwig/2008/05/080513-combating-welfare-fraud-to-keep-system-fair.php  
• Marston G & Walsh T 2008. A case of misrepresentation: Social security fraud and the criminal justice system in Australia. Griffith Law Review 17(1): 285–300
• Mosher J & Hermer J 2005. Welfare fraud: The constitution of social assistance as crime. Ottawa: Law Commission of Canada
• Prenzler T 2011. Welfare fraud in Australia: Dimensions and issues. Trends & Issues in Crime and Criminal Justice. Canberra: Australian Institute of Criminology
• Reeve G 2006. Human services administration: A more intelligent approach to reducing benefit fraud. Public Administration Today Jan/Mar: 37–45
• Swedish National Council for Crime Prevention (SNCCP) 2008. Unemployment benefit fraud: Motives, extent and preventive measures. Stockholm: SNCCP
• Webb J 2001. Prosecuting social security fraud: Protecting the trust fund for present and future generations. High Tech and Investment Fraud 49(6): 1–5

About the Authors

Tim Prenzler is a Chief Investigator at the Griffith University Brisbane node of the Australian Research Council Centre of Excellence in Policing and Security.