Our budget under attack:
estimating social security fraud
Beating Fraud is Everyones Business  is the governments Green Paper on Social Security Fraud. It makes the apparently authoritative statement, which has been widely reported, that social security fraud is now estimated at about £7 billion. Much of what follows was written in response to an e-mail circulated to the Social Policy discussion group on Mailbase, which suggested that this was an underestimate. That belief is probably widely shared, but there is very little basis for it.
The concern with fraud in the benefit system is long-standing. The political right picked up the issue after the publication of an anonymous exposé by a benefits officer, Robin Page, who was subsequently identified and sacked. Pages subsequent book, The Benefits Racket , crystallised concern, and the press latched on to the issue. A government inquiry into the subject in 1973 focused the attention of benefits administrators on the issue . The press kept the issue in public view throughout the 1970s , and the Conservative Government which came into power in 1979 greatly increased both the numbers of investigators and the estimated levels of fraud, raising the official estimate from £4 million to £200 million . The Governments justification for this increase was that the previous figure was based on detected, irrecoverable fraud; the £200 million represented what might be happening .A study in 1981 by Sir Derek Rayner upped the stakes further; it suggested that 8% of unemployed people (in one location) were working while claiming benefit . The claim was described by the permanent head of the DHSS as speculation in the absence of hard evidence , but it encouraged progressively increasing estimates of the level and extent of fraud. In the 1990s, Benefit Reviews have suggested levels of fraud, varying according to benefit, of up to 12.2%.
During this time, expenditure on social security has been increasing, rising from £49 billion in 1979 to £98 billion. The argument has been made not that fraud has risen proportionately, but that it has mushroomed, outstripping the growth in benefits. That is certainly the view of Peter Mathison, the Chief Executive of the Benefits Agency, in his evidence to the Public Accounts Committee: We always took the view that fraud was increasing through social security . The Green Paper is more graphic: our budget is under attack .
The extent of fraud
The Green Paper begins with the statement that a figure of 2% has been established; that there are strong suspicions of another 3%; and that there is a low suspicion of a further 2%.
The starting figure of 2%, though hardened by repetition over eighteen years, has no rational basis. The reason for settling on this figure was explained in parliament in 1980 by Reg Prentice, the Minister responsible at the time, when he upped the figure from its previous estimate. Prentice went further:
There is no need to ring around department stores to know that they, and other large commercial organisations, assume a loss through fraud of 1 or 2 per cent in their operations. Applying that to the DHSS, with its expenditure of £20 billion a year, leads to an estimated figure of £200 million 
There are some differences between the Benefits Agency (as it now is) and department stores; for example, the agency does not lay the money out on display and invite people to walk round. This set the trend, though, for making large estimates of fraud on the basis that theres a lot of it about.
The basis that the DSS has for talking about a higher level of fraud now is the strong suspicion reported in a number of surveys. Table 1 below reproduces the survey evidence summarised in the Green Paper.
The Benefit Reviews apply a distinctive standard of confirmation and suspicion. By comparison with the figures in the table, relatively few cases of fraud are directly confirmed in the sense that they are identified as fraudulent; they are dealt with by prosecution, disallowing benefit or encouraging voluntary withdrawal from claiming.  For the purposes of the Benefit Reviews, fraud is apparently taken as confirmed where overpayment is identified (though overpayment is not the same thing as fraud). On that basis the figure for Disability Living Allowance has been represented as confirmed fraud,  although many of the cases consist of disputed classifications and changes in gradings.
TO SEE TABLE 1 PLEASE CLICK HERE
Illustrative examples of suspicion from the surveys are given in the DSS evidence to the House of Commons Standing Committee on Public Accounts: they refer to levels of suspicion (identifying Level 3 and 4 but not, unfortunately, other levels). This is one of the examples of Level 3 suspicion:
A lone parent customer admitted to the interviewing officer that she was working part-time but that she had only worked for one day recently, and earned less than £15.00. A check with the employer revealed that the customer worked more frequently than this but that she did not earn more than £15.00 weekly (the earnings limit for a lone parent). However, the employer paid the customer cash in hand and kept no wage records. The employer was suspected to be collusive. Benefit remained in payment.
This is an example of Level 4 suspicion:
Following an ineffective visit to the house of an unemployed single male, the customer signed off, stating he would be starting work with his father. The investigator made a follow up visit to the customer and was told by his mother that the customer had been working for his father unofficially for some time but without pay! The whereabouts of the father was not disclosed.
Intriguingly, both cases seem to indicate that the information about changed circumstances came from the claimant - which is not strongly suggestive of an intent to deceive. Pressed by the Public Accounts Committee as to whether he believed the survey results were robust, Peter Mathison avoided the question:
I am not quite sure what you mean by robust. In terms of DLA [Disability Living Allowance], it is extremely difficult to identify quite whether it is fraud I do think it is about correctness and we are sure that there is a high level of incorrectness there.
The report of the committee concludes, oddly: we note their assurance that the methodology used was robust. As the reply makes clear, no such assurance was given.
The Green Paper estimates that up to 7% of claims are fraudulent, but the surveys do not show this; the benefits which are being discussed are only a part of a larger system. A high proportion of expenditure is directed towards circumstances in which fraud is generally considered to be marginal. Forty four percent of expenditure goes on elderly people while over 6% goes on Child Benefit. Fraud is marginal in these cases, not because parents or pensioners are saints, but because benefits based on age and entitlement in stable circumstances are much more difficult to muddle than benefits which depend on work status, domestic status or income.
The cost of fraud
Current spending on social security is estimated in the Green Paper at £100 billion. There are reasons to question this figure. For many years, the DSS, the Treasury and the Office for National Statistics have published inconsistent estimates of social security expenditure. For 1996-97, the last recorded outturn, the Treasury has £90.7 billion (a control total of £77 billion  plus £13.7 billion on cyclical expenditure); the Office for National Statistics has £91.6 billion; the DSS has £93 billion. For the current year, 1998-99, the Treasury has £92.4 billion (£79.6 control plus £12.8 billion cyclical,  and the DSS has issued estimates of £95.1 billion,  £97.9 billion (in their web page) and £100 billion (in the Green Paper). Partly, this inconsistency relates to the definition of the subject, and partly to variability in the accounts. Social security expenditure is not recorded month by month as it happens, like expenditure on a bank statement; expenditure is estimated on the basis of surveys until final accounts are settled.
On the basis that social security spending is £100 billion (which it isn't), the Green Paper sets the level of fraud in the region of £2 billion to £7 billion. The £2 billion figure constitutes confirmed fraud, though the basis on which it is confirmed is not clear. The next 3% represents high suspicion; the remaining 2% is based on low suspicion. The upper figure of 7% assumes, as far as I can tell, that all suspicions are justified; that where fraud exists the claimant is not entitled, and all the benefit claimed is claimed fraudulently; and that people are not entitled to any other benefit.
It seems unreasonable to suppose that every suspicion is justified, but there is at least some rationale for using it as the basis for a high estimate. The assumptions about entitlement have to be wrong because many of the principal types of fraud - working while receiving benefit, concealing extra income or capital, or living together - imply a reduced entitlement, not the absence of entitlement. Inappropriate claims for support while sick or disabled may be claims for higher rates of benefit when lower rates should be awarded, or (in the case of Incapacity Benefit) an alternative to unemployment benefits. The Benefit Integrity Project, which examined the payment of disability benefits, identified far more cases where people were receiving the wrong grade than where people were not entitled at all: of 54,839 cases, 3,926 claims were disallowed, 6,884 were reduced, and 1,299 increased. Overall, then, a reasonable estimate of fraud relating to these benefits should indicate a value which is lower in proportion to total expenditure than the number of claims they represent. (This used to be true of the statistics kept in the 1970s). 
If the DSS assumptions and figures were to be accepted as they stand, over £3,300 million might be said to be claimed fraudulently. This is based on reviews covering 84% of expenditure. The Green Paper, however, claims that the level of confirmed fraud and high suspicion comes to 5% of expenditure. There is an outstanding discrepancy of nearly £1,600 million which has still to be explained. Some of the discrepancy seems to be based on assumptions about fraud in the remaining benefits, but this can only account for part. I have a sinking feeling that the figure may come from adding Prentice's 2% confirmed plus a round 3% suspicion from surveys, in which case the suspicions would have been double counted. It may be that there are other explanations, but they are not given in the report.
Is fraud increasing?
Any explanation as to why fraud might have increased as a proportion of total expenditure has to take into account the changing pattern of expenditure. A large part of the increase in total expenditure is attributable to the growth of pensions. One of the biggest departures from Beveridge's principles was the state earnings-related pension; at the time it was agreed by both the main parties. People who retire now have more generous pensions than those who retired ten years ago. Women are more likely to be entitled to pensions in their own right, and the period of dependency on pensions has increased. However, there is no assumption that fraud by pensioners has mushroomed, and most of the growth is attributed to three other areas of spending.
First is the growth of unemployment. That statement is likely to cause some confusion, as official statistics on unemployment - based on the claimant count - have pointed to a steady fall, which should imply a proportionate reduction in fraud related to unemployment. The position is more complex. The National Insurance system as it applied to unemployment has effectively collapsed; the growth of long-term unemployment, the lack of previous work records and the tightening of contribution conditions meant that unemployed people were increasingly unlikely to be covered. The Job Seeker's Allowance, introduced in October 1996, still has an element relating to insurance, nominally covering the first six months, but by the time of its introduction only 8% of unemployed males received any National Insurance. The problem is not simply that more people are unemployed than in the past, but that those who are unemployed are treated differently by the benefit system.
Prior to the last general election, some of the fall in the numbers of people who were described as unemployed was accounted for by increases in the number of people receiving Incapacity Benefit because they were not employed and were long-term sick. There were 1.58 million such claimants in 1991 and 1.89 million in 1994-95, an increase of 310,000. The general observation appears to be that when unemployment increases, the numbers of people in related categories, like long-term sickness or single parenthood, rises proportionately. This happens because people who are unemployed in these categories claim under different rules. However, when unemployment falls, these categories do not fall proportionately, because the people in them have greater difficulty re-establishing themselves in the labour market.
Second, overlapping with the last category, there is the growth of benefits for sickness and disability. The main reason for increasing numbers of people recorded as disabled is unemployment, though there are other contributory factors, such as transferring responsibility from hospitals to community care, which has led to greater demand for disability benefits. Particular concern has been expressed about the Disability Living Allowance, where the Benefits Agency has set confirmed fraud at £499 millions (plus a further £79m of mild suspicion). The Public Accounts Committee expressed some perplexity about that figure, which seemed to them to confuse cases where people were receiving benefits to which they were not entitled from cases in which the claim was fraudulent:
the initial results from the Benefit Integrity Project suggest a high level of error rather than fraud. 
Third, there is Housing Benefit. Housing Benefit has its origins in the withdrawal of general subsidy from public housing in the 1970s, which made it necessary to introduce benefits to enable people to pay rents. The Conservative Government effectively ended most subsidies to public housing during the 1980s, and used Housing Benefit instead.  The removal of subsidies to public housing was followed by pressure to increase rents to market levels, and an attempt to revive the private rented sector by raising the ceilings on the rents which could be paid. Concern with Housing Benefit fraud comes not because there is great evidence of such fraud, but because of the lack of investigation; the local authorities which administer Housing Benefit have not pursued fraud with the vigour that central government would like, and sanctions have been introduced to whip the local authorities into line. (The problem is not that local authorities are necessarily disinclined to pursue fraud: it is that Housing Benefit is a monster, described in its day as the greatest administrative fiasco in the history of the Welfare State  - though that was before the Child Support Act - and that local authorities can barely keep track of what they have to do now.)
If the figures can be believed, over twenty-five years of crackdowns on fraud seem to have had no effect on fraud, or even the reverse effect from that intended. It may be that fraud has always been at the same level, though there is no sensible basis for a comparison. Benefit fraud used to be assessed in relation to the number of prosecutions and confirmed cases: this year the Benefits Agency told the Public Accounts Committee that they could not tell how many prosecutions there had been, as statistics were not kept in this form.
If fraud has increased, it is partly because the benefits rules create their own problems. Overall, the proportion of means-tested benefits has increased from 16% of expenditure in 1979 to 35% now. The marginalisation of large parts of the workforce has put them into ephemeral, precarious situations. When people have changing circumstances, rules relating to income, part-time earnings and domestic arrangements inevitably catch some people out. Partly, I would suggest, it may be because the obsessive pursuit of fraud has poisoned the water; the constant harping on fraud generates secretiveness, mistrust and, for some, a sense that the system is fair game. It seems clear, from the examples given by the DSS, that a claimant who does make an open disclosure of changed circumstances is liable to be treated as suspect; it is hardly surprising if some people are guarded about what they say.
The question of whether fraud really has increased must, however, be open. The surprise of the Public Accounts Committee - how can fraud on Income Support go up by over £300 million in a year? - is easy to share: the increase is not an increase in fraud, but in the estimates. It looks as if the Benefits Agency has taken a bicycle pump to the figures, inflating them at every stage of the process - in the initial assumptions, the acceptance of suspicion as fact, the estimate of value, the inclusion of further material on low suspicion, the extension of assumptions to benefits which have not been investigated, and a liberal rounding up of all the totals.
Chalker, L. (1979), Reply to Field, New Statesman, 7-12-79
Cm 390 (1998), Public Expenditure Statistical Analyses 1998-99, London: The Stationery Office
Cm. 4012 (1998), Beating Fraud is Everyones Business, London: Department of Security
Cmnd 5228, (1973), Report of the Committee on Abuse of Social Security Benefits, London: HMSO
Department of Social Security (1998), Welfare Reform Focus File 1, London: Department of Social Security
Field, F. (1979), The myth of the social security scrounger, New Statesman, 16-11-79
Golding, P. & Middleton, S. (1978), Why is the Press so obsessed with welfare scroungers?, New Society, 26-10-78; (1982), Images of Welfare, Basingstoke: Macmillan
Hills, J. (1998), Housing, in H. Glennerster, J. Hills (eds), The State of Welfare, Oxford: Oxford University Press
House of Commons (1998a), Select Committee on Public Accounts: 58th Report, Minutes of evidence, HC 570
House of Commons (1998b), Select Committee on Public Accounts: 58th Report, HC 570
Office for National Statistics, Annual Abstract of Statistics 1998, London: The Stationery Office
Page, R. (1971), The Benefits Racket, London: Temple Smith
Rayner, D. (1981), Payment of Benefits to Unemployed People, London: Department of Employment, DHSS
Smith, R. (1985), Who's fiddling? Fraud and abuse, in S. Ward (ed), DHSS in Crisis, London: Child Poverty Action Group
The Times, cited Walker, R. (1986), 'Aspects of administration', in P. Kemp (ed.), The Future of Housing Benefits, Glasgow: Centre for Housing Research, p 39