52150 World Bank Safety Nets Primer Notes 2007 No. 26 Reducing Error, Fraud and Corruption (EFC) in Social Protection Programs Because Social Protection (SP) and Social Safety Net (SSN) · We refer to intentional abuses of the program by claimants programs channel a large amount of public resources, it is as fraud and by staff as corruption. Fraud occurs when a important to make sure that these reach the intended claimant deliberately makes a false statement, conceals or beneficiaries. Error, fraud or corruption (EFC) reduces the distorts relevant information regarding program eligibility economic efficiency of these interventions by decreasing the or level of benefits. Corruption commonly involves amount of money that goes to the intended beneficiaries, and manipulation of beneficiary rosters (e.g. registering erodes the political support for the program. While no program potential beneficiaries for clientelistic purposes to garner is immune to EFC, evidence from developed countries political support); staff accepting illegal payments from demonstrates that such leakage can be brought to negligible legal or illegal beneficiaries; or diversion of funds to ghost levels. In five OECD countries this fraction is between 2-5% or illegal channels. for the SP sector as a whole. For SSN programs, which use more complex eligibility criteria and hence are more prone to EFC, this fraction is 10%. To achieve these results, programs Opportunity cost of fraud and error in the UK social have implemented a number of measures reviewed in this note. protection system In contrast, efforts to combat or even measure EFC are quite rare in developing countries, although some programs are In 2005, social protection spending represented a quarter of plagued by it. the total public spending in the United Kingdom. About $220 billion in benefits have been paid to 17 million beneficiaries. Definitions. Although most SP programs strive to transfer About 2.5% of this amount, or $5 billion, was lost to fraud and all their resources to the right beneficiaries, in the right amount and error. If saved, these resources could have been used to provide 32,000 more doctors, or 80,000 nurses, or 72,000 at the right time, a fraction is lost to EFC (Figure 1): teachers, or 62,500 police officers. · Error is the unintentional violation of program or benefit Source: DWP (2005) rules, resulting in the wrong benefit amount or payments to an ineligible applicant. One can distinguish between official error due to staff mistakes, and customer error which occurs when a customer inadvertently provides incorrect Why do we care about EFC in SP programs? SP programs information. channel a large amount of public resources to targeted beneficiaries, and even a small fraction of misappropriated benefits may add up to large sums of money with high Figure 1. What are error, fraud and corruption? opportunity costs. On average, SP spending represents 15.7% of GDP in developed countries, 7.4% in middle-income countries (MICs) and 3.8% in low-income countries (LICs). Within SP, this topic is more relevant for programs with complex eligibility requirements which make them more prone to EFC, such as social safety nets, disability insurance or unemployment supports. Spending for these programs averages 6% of GDP in developed countries, 3% in middle- income countries and about one percent of GDP in low-income countries. Ensuring that these resources reach the intended beneficiaries will determine the efficiency of public resources as well as the reputation and credibility of the program, and ultimately the support for the program. Emil Daniel Tesliuc and Annamaria Milazzo prepared this note based on the key references listed at the end of the note as part of the material used in the World Bank training course "Fraud & Error Control in Safety Net Programs". How much is lost to EFC? There is very limited evidence unemployment benefit and disability pension programs (1- in this respect, most of it from developed countries. A recent 2%); old-age pensions had the lowest rates (0.1-1%). These study (NAO, 2006) found that fraud and error rates are figures should be interpreted as good-practice, or as lower positive even in well-run programs operating in high-capacity bounds for the amount of funds affected by EFC, as they countries. In five OECD countries ­ the UK, Canada, Ireland, come from a small sample of countries and programs with New Zealand and the USA ­ fraud and error rates ranged high administrative capacity, with adequate procedures to between 2% and 5% for the whole SP system (Table 1). minimize EFC, including procedures for estimating the level Within the SP system, means-tested SSN programs had the of fraud and error with reasonable accuracy. highest fraud and error rates (5-10%), followed by Table 1. A Comparison of Fraud and Error Rates in 5 OECD Countries 1 New As % of expenditure/payments UK Canada Ireland Zealand USA (2004-2005) (1994, 2003) (2004) (2001) (2003) Total Fraud and Error 2.3% 3-5% ('94) - 2.7% - Fraud and Error in Housing Benefit 5.2% - - - 3.6% Fraud and Error in Income Support 5.3% 3.5% ('03) - - 5.07% Fraud and Error Old Age, Disability 4.9%, 0.1%, 1.9% 2 - 7% - 0.53% Source: NAO (2006) Notes: 1 Comparisons are difficult as benefit types differ between countries. This table shows rates for the following benefits. The Income Support benefit for Canada and the US is Unemployment Insurance. In the category, Old Age and Disability, the benefit in Ireland is Disability Allowance and in t the US Old Age, Survivors and Disability. Also measurements vary. Rates for the UK refer to fraud and error as % of total expenditure. Rates for Canada (`03) refer to the total savings identified as a percentage of total payments. Ireland measures fraud and error as a percentage of expenditure for specific benefit types. New Zealand estimates this percentage as the number of cases sampled that contained errors, which led to benefits being paid incorrectly. The USA measures the total value of improper payments per benefit type. 2 These rates are for Pension Credit, Retirement Pension, and Disability Living Allowance respectively. The information from developing countries is scarcer, as government. All these factors increase the risk that only few programs and countries have taken measures to benefits do not reach intended beneficiaries in time or in combat and/or measure the incidence of EFC. However, the correct amount. Having the right instruments and it is plausible to expect the share of funds affected by EFC tools to minimize EFC is thus more important for SSN be larger compared to developed countries. For example, programs, although there are economies of scale when in India a Government audit in 2003 found that 31.6% of the efforts to combat EFC cover the whole set of SP IAY (housing programs) and 53.5% of SGSY (credit programs. program) funds did not reach the intended beneficiaries. In the case of the SGRY, an employment program, (2) The level of EFC is likely correlated with the institutional administrative data report that 31.3 million man-days of capacity of the country and implementing agency. One work were generated in 2001-02, while the National Sample would expect to find more EFC in LICs and MICs than Survey data for 2002-03 estimate this number as less than 3 developed countries. However, evidence is scarce as very million man-days, only 10% of the official figure. few programs or countries measure the level of EFC using comprehensive, unbiased measurement tools (see Three points can be made about EFC in developed and Box 1). developing countries: (3) In developed countries there is a much greater concern (1) Across SP programs, the share of funds affected by about error and fraud, while developing countries worry EFC is larger in SSN programs, due to their inherent more about corruption. For example, none of the 9 design features. Eligibility for these programs is often OECD countries reviewed in NAO (2006) explicitly based on the economic position of the household, which reported issues around corruption of departmental staff. is more difficult to assess and harder to verify than, for It seems a small problem for administrators in these example, for an old-age pension where all you need is countries. Reasons could include a strict separation of proof of past contribution and age. Because SSN functions (e.g. between payments and assessments, programs are more narrowly targeted to poorer between decision-making and investigations and households and poverty status changes over time, reviews), staff training and management, effectiveness eligibility also changes over time. The responsibilities of internal audit and the integrity of systems processing for implementing SSN programs are often shared across payments. In contrast, corruption is often the lead different departments, agencies, and levels of concern in developing countries. Box 1. How can programs measure the level of error and fraud? Lessons from developed countries Measuring fraud and error in the SP system is difficult, but not impossible. There are different measurement approaches, used primarily in developed countries: o Total fraud and error in the SP system. Although complex, this methodology generates an unbiased representative estimate of the level of fraud and error in the program. This level may be expressed as % of program funds affected by error and fraud, or the number of cases, or both. To arrive at this estimate, a representative sample of cases (client files) is drawn from the active caseload. The program will provide the (electronic and written) records to a review team that would check each case for mistakes or inaccuracies in the data and, where appropriate, would visit and re-interview the applicant. For each case, the review team will estimate the amount of money under and over paid, and whether the difference is due to error or fraud. These figures are summed up to estimate the total level of error and fraud. These estimates are qualified by confidence intervals that take into account the fact that a relatively small sample is used in relation to the size of the actual caseload. Few countries measure the level of total fraud and error across SP programs. From the nine countries reviewed in the NAO study (2006), only Australia, Ireland and the UK measure overall level of fraud and error on the basis of `rolling measurement' and/or `snapshots'. This is considered as gold standard. o Level of fraud and error in specific programs. Some countries measure the level of fraud and error based on a thorough review of a random sample of beneficiaries files (Sweden, US, and New Zealand); others link their sampling and reviews to risk-based assessments (Ireland and US). o In many cases measuring focuses on other criteria than the total number of cases, and the total value of fraud and error, leading to a partial and non-comparable estimate of the level of error and fraud in the system. These criteria include: (i) Minimal accuracy targets (New Zealand, Sweden, Australia): These measurements focus on the proportion of accurate decisions to pay benefits and the proportion of accurate payments being made. (ii) Totality of improper payments per benefit type (USA): This is a measurement of the total amount of improper payments per benefit type, which lumps together fraud and error. There is no system-wide aggregation of the total value of improper payments. (iii) Savings and prosecutions achieved (Australia, Ireland, the Netherlands, and New Zealand): This measurement focuses on savings, performance and also prosecutions, based on observations and outcomes of reviewed cases. (iv) Sampling in specific benefit schemes (Ireland and USA): Measurements are targeted on specific benefit schemes, which are perceived at high risk from fraudulent claims. Source: based on NAO (2006) What can be done to reduce EFC? No program is immune for a participant with respect to the salary of the intake from error or even some degree of fraud and/or corruption, officers to make less attractive the possibility that an but having solid mechanisms, procedures and institutions to ineligible applicant may offer a cut of her/his benefit in prevent, detect, deter and measure EFC helps. Some of these return for entrance into the program; defining mechanisms or institutions are not specific to the SP sector conditionalities such as requirements for recipients to or the particular program. For example, implementing a get health care, attend school, or work to guard against broad government-wide governance agenda of due process, `ghost' beneficiaries; and using payment mechanisms that transparency and accountability will reduce the level of funds move benefits from the treasury to the individual recipient affected by EFC. A good practice, common in many with as few intermediaries as possible, to reduce the countries, is to subject SP programs to the regular oversight potential for diversion of funds. and control of the supreme audit institution and/or the public 2. Setting up adequate administrative procedures (see prosecutor's office. These institutions also act in case of Table 2). These include: ensuring that administrative allegations. processes are clearly defined and that staff and other resources are adequate to carry them out; instituting a At program level, the level of EFC can be reduced through range of quality control procedures to ensure that careful program design and consideration of the specific eligibility criteria are respected, payments are audited, institutional context and administrative setting. Some of the information systems have appropriate safeguards, etc; measures are: establishing sensible tolerances in the quality control 1. Using program design to minimize incentives and procedures to facilitate respect for rules and make efforts opportunities for EFC. The specific design features to enforce them more cost-effective; setting up adequate include: making program budgets consistent with grievance, appeal and "whistle-blowing" procedures for eligibility criteria to eliminate opportunities for bribery; applicants who believe they are eligible but were denied centralizing the determination of eligibility (when entry, for beneficiaries who are receiving incorrect program implementation is decentralized) and defining payments or are requested to pay kickbacks, for program simple eligibility criteria to reduce discretion and workers who suspect fraud by other program workers, opportunity for corruption; considering the benefit level for the general public who suspect irregularities of any Table 2. Administrative Procedures to Minimize EFC Outcome Actions taken Prevention · Tighter checks at outset of claim. · Make payments directly to bank accounts instead of by check. · Staff training. · Fraud-proof the new policy and operational developments. Detection · Data matching. · Investigation: random or time-based reviews (controls of the accuracy of the information provided to the program and of the decisions made by staff); · Intelligence: gathering information from the public (tip-offs); telephone hotline for reporting fraud; automatic flags from Management Information System (MIS). · Risk profiling: risk-based assessments to organise reviews (e.g. select client investigations) using scoring; profiling; · Joint operations with other agencies. Deterrence · Sanctions. · Prosecutions. · Confiscation of assets. · Measures to influence public attitudes, such as public awareness campaigns. Monitoring · Measure regularly the level of EFC, to assess whether the strategy works. sort; and taking action against miscreants with 4. Using transparency and communication well to meaningful penalties. Box 2 presents some innovative ensure that eligibility criteria, benefit levels, and rules are instruments used in OECD countries to reduce the clear to both the public and beneficiaries. This reduces incidence of EFC. EFC as well as unwarranted appeals or complaints. 3. Aligning institutional responsibilities well. The delivery involves actions by and interactions among To assure successful delivery of social protection benefits to different entities. The system of incentives must be the intended beneficiaries, specific recommendations to aligned with roles to facilitate the role of and rely on minimize the risk of EFC must be considered at each critical third parties (local government, health and school stage of program design and implementation (Table 3). officials) and civil society organizations. A clear definition of responsibilities is necessary. Box 2. Innovative approaches in the fight against EFC in OECD Countries Developed countries' experience shows that reducing EFC is feasible and cost-effective. For example, the US Food Stamps Program reduced its overpayment rate from 10% a decade ago to about 4%. From 1997 to 2006, the UK halved the level of fraud in the social protection sector. To obtain these results, programs invested in developing appropriate procedures, staff and IT support. NAO (2006) summarizes a number of innovative approaches used in SP programs in 9 OECD countries. It is by no means a comprehensive list of what countries are doing in the field of anti-fraud and error. The anti-fraud and error activities identified are organized in three categories: prevention, detection, and deterrence. These categories overlap and some activities or initiatives will cover more than one thematic category: · In prevention, Canada holds claimant information sessions (CIS) with groups at high risk of committing benefit fraud. Since 1999, Canada claims it has saved about US$600 million in overpayments in the benefit system. The Netherlands also uses contracts and campaigns to inform claimants of their rights as a preventative measure. Claimants sign a contract that explains rights and obligations when claiming benefits. Similarly, the New Zealand government aims to inform about 70% of claimants of rights and obligations. These measures aim to inform and remind claimants of their obligations. The underlying idea being that a better informed clientele is less likely to commit fraud. For error, preventative measures include staff training and introducing results-based management. In Canada, error rates have fallen by making line managers more accountable for the errors of their staff. Ireland aims to half its error rate through staff training. · In the detection category, data-matching is used in all countries. Data-matching can be highly cost-effective and target ratios are set in Australia and New Zealand for overpayments identified to cost of the program. A normal ratio would be four to one. So, data-matching can be highly cost-effective. Unique identifiers for claimants are also often part of data-matching strategies. The use of identifiers across different databases and departments allows analysts to more effectively match data on claimants held in different databases. Review activities are also instrumental to detection efforts. Different strategies are used in most countries ranging from risk-based reviews (e.g. reviews of certain professions in Canada, scoring in the UK, and profiling in Canada) to random and time- based reviews. In New Zealand, claimants have to re-establish core eligibility after a certain time period. · In deterrence, the main levers reported are increased sanctions and increases in the rates of prosecutions. This is done in Ireland, the US, and the UK. Source: based on NAO (2006) Table 3. Use EFC Lenses on Critical Phases of the Program Phase Recommendation Define lines of authority and accountability, including administration and service Institutional Design delivery. Align incentives and financing, particularly for intergovernmental responsibilities. Identify risks and remedies up front. Eligibility, Recertification Minimize inclusion and exclusion errors. Conditionality Balance accuracy and the burden of verification. Examine reasons for non-compliance. Payment processes Improve monitoring of cash flows and control procedures. Management Information Strengthen framework to analyze and design MIS systems. Promote the Systems (MIS) crosschecking of enrollments with other databases. Internal quality control Expand process evaluation, feed-back loops, and correction mechanisms within programs, particularly through spot checks. Appeals and complaints Handle complaints and appeals appropriately and in a timely manner, and publish program data. Financial management Define payment accounting and reporting requirements. Focus audits on risks. and auditing Include supreme audit institutions. Procurement Closely supervise implementation of large service contracts. Adapted from: World Bank (2007) Even though developing countries seem to Stepping up the effort to combat EFC will likely depend "underutilize" some of the instruments used OECD on the level of development of the country, and the countries to control errors and fraud, some positive institutional maturity of the program. Such systems take trends are emerging, such as greater use of technology, at some time to develop and so are more feasible or effective in least in MICs; more explicit thought on transparency; and permanent programs. As MICs are moving away from stop- greater attention to implementation issues (see Box 3 for start safety net programs around economic crisis or transitions examples). Better practices in collecting and using program of government, more sophisticated experience is developing information in developing countries can provide an there. We believe that many MICs are ripe to absorb many of indication of the scope and nature of the problem, and may the techniques and innovations used in OECD countries. LICs even prevent EFC from happening. A number of MICs are should probably start by investing in solid MISs. making large investment in MISs of the SP program, a key tool for the development of an effective policy to combat Some of the measures against EFC involve trade-offs with EFC. other desirable features of safety net policy. While reducing EFC is important, one should not overdo it at the expense of There are a number of lessons to be learned from the other program objectives. For example: experience of developed countries: (1) Fully funding the safety nets should eliminate a source of (1) Although it is virtually impossible to operate an error- corruption, but it is not always feasible given budget free system, it is important to place the reduction of constraints. EFC as one of the priorities, and to institute appropriate (2) Keeping eligibility criteria and payment structures simple mechanisms. will reduce the probabilities of EFC, but will make (2) There is ample evidence that some instruments to combat programs less precise in their targeting and lower the EFC are cost effective and represent a wise investment, impact on poverty per dollar spent on legitimate although it will add to its administrative cost. beneficiaries. Thus, policymakers will find themselves (3) Most measures to combat EFC require good foregoing some of the design options for minimizing EFC, management of program information, a good MIS. and thus will need to develop correspondingly more (4) The use of multiple instruments strengthens the system. sophisticated administrative procedures as alternative (5) Adequate program control and the introduction of means of keeping problems in check. incentive-based systems (with appropriate rewards and (3) Avoid EFC activities that deter honest clients. The main penalties) are particularly important, especially when purpose of a social program is to transfer resources to its implementation is decentralized. target group. Likely, most of the applicants are honest, (6) There is no ready-made solution: each EFC policy and should not be deterred from accessing the program should be tailored to the specific program and the by EFC measures that are too inquisitive, aggressive or characteristics of the country; programs are not isolated insensitive. within a country, they reflect character of overall (4) Activities to combat EFC may be at odds with other government-wide governance agenda. important objectives of the program, such as client satisfaction (which may require processing claims faster, but that will squeeze the verification efforts, hence the level of EFC). Box 3. Good practice of oversight and control mechanisms from the Latin America Region A number of middle-income countries have made significant steps forward in implementing mechanisms to control the use of funds in SSN programs: a) Brazil's Bolsa Familia Program (BFP) is a large conditional cash transfer (CCT) program implemented in a federal context. At federal level, the program is managed by the Ministry of Social Development (MDS). Most of its implementation, however, is done by the 5,564 municipalities. Consequently, the main pressure points for EFC in Brazil's decentralized context include political interference, political biases, and administrative errors in the registration process. To manage, monitor and reduce the risk of EFT, the Government has instituted several mechanisms, including: (a) Oversight & Control by Supreme Audit Agencies that undertakes regular random- sample operational audits (also known as quality control reviews); implementation evaluations; annual financial audits, and investigation of at-risk cases; (b) MDS' own program controls, including internal and external cross- checks of the registry; monitoring municipal implementation quality using the new system of financial incentives (performance-based administrative cost subsidies) and the "Decentralized Management Index", and; monitoring of the activities of the operating agent for registry and payments, via the new performance-based contract and associated financial penalties, and; (c) Municipal controls (social controls, local contacts). Municipalities serve as first point of contact for complaints or appeals for the beneficiaries. Formal Social Controls Councils have been established in all municipalities, and their roles include: overseeing the registration process; periodically evaluating beneficiary list; monitoring reporting on, and compliance with, conditionalities. Centralized hotlines are used to handle inquiries from general public, beneficiaries, and local program managers. b) The Heads of Household Program in Argentina is a large-scale emergency workfare program set up quickly during 2001-2 economic crisis. The Ministry of Labor is the responsible national agency and registration is decentralized through municipalities and civil society/political organizations. A set of complaint resolution tools has been established, including: (a) Toll free hotlines manned by call centers addressing questions on payment dates, eligibility and for reporting ineligible beneficiaries; (b) a Commission in the Ministry of Labor to handle allegations of program abuse or complaints; (c) Criminal offenses referred to Federal Prosecutor of the Social Security System. c) Colombia's Familias en Accion is a CCT program under Presidential Executive Agency (Social Action) implemented at municipal level. Spot checks reviewing and monitoring the operational procedures for program implementation in different localities are carried out on a random basis by external firm every 6 months. These spot checks serve several objectives, such as: checking if operations and procedures of program met objectives in practice; verifying how procedures were being interpreted; validating the accuracy of the data relevant for the program (i.e. verification of fulfillment of conditionalities, inscription of eligible families, payments, etc.). Source: Lindert et al (2007) and World Bank (2007) Key References: Barr, David, Presentation on "Reducing fraud and error in the UK: What it takes?", World Bank Workshop on Reducing Fraud and Error in SP Programs, May 16-17, 2007, Washington DC. UK Department for Work and Pensions. 2005. "Reducing fraud in the benefit system. Achievements and ambitions." October, 2005. Lindert, Kathy, Anja Linder, Jason Hobbs and Bénédicte de la Brière, 2007. The Nuts and Bolts of Brazil's Bolsa Família Program: Implementing Conditional Cash Transfers in a Decentralized Context. Social Protection Discussion paper No. 0709. The World Bank. UK National Audit Office (NAO). 2006. "International benchmark of fraud and error in social security systems." Report by the Comptroller and Auditor General, London: NAO. van Stolk, Christian. 2007. "Note on occurrence of fraud and error in international social security systems." RAND Europe, June 2007. The World Bank. 2007. "Control and Accountability Mechanisms in Conditional Cash Transfer Programs. A Review of Programs in Latin America and the Caribbean." World Bank, Latin America and the Caribbean Region, March 2007. and Various materials posted on the "Six Nations Benefit Fraud Conference: website: http://www.ssa.gov/sixnations/home.htm