Privacy Law and Policy Reporter
The Data-matching Program (Assistance and Tax) Act 1990 (Cth) authorised a very large program, involving seven agencies, to be conducted by a segment of the Department of Social Security. The Australian National Audit Office (ANAO) conducted an audit of the program between January and June 1993. It tabled a report on that audit on 27 October 1993, containing 16 recommendations, the key findings of which are summarised below.
The report is about ''the efficiency and effectiveness with which the Department of Social Security (DSS) conducts its data-matching activities under the authority of [the Act]'. The audit did not, and could not, include an evaluation of whether or not the program was worth undertaking, or worth continuing, because it is the role of the ANAO to audit government operation.
Other significant findings include:
3 per cent of all social social security reviews were the result of discrepancies discovered by the DMP.
85 per cent of these cases reviewed under the program failed to uncover an incorrect payment.
The government predicted that data matching would lead to reductions in 0.9 per cent of social security payments; in fact it has only lead to reductions in 0.1 per cent of payments.
The use of data-matching with the Tax File Numer has recovered only $0.8m of overpayments.
On the Department's own figures, the savings from data-matching have already peaked and will fall to under $50m by 1995-1996.
The Department has twice dramatically reduced its estimate of savings from data-matching, most recently cutting its estimates of the next two financial years by more than two-thirds (from $173m back to $54m for 1994/1995, and from $193m down to $48m for 1995-1996.
Data-matching has not yet uncovered one case of identity fraud.
Data-matching does not out-perform random selections drawn from specific client groups in relation to cancellations and downward variations."
On average, only 9% of data- matching reviews lead to a change in payments. This is only slightly higher than the average 7 per cent success rate for the 2.5 million reviews conducted each year. Further, the average payment variation for all reviews is $15, considerably more than the average payment variation for data-maching at just $9.00. And while data-matching does on average pick up a much higher overpayment (at $306, it is more than seven times the average) this amounts to less than $500,000 per fortnight for the entire country (assuming DSS's own figures).
''[W]ith respect to data-matching, DSS and the Data-Matching Agency effectively fulfil their obligations under the Act and other guidelines with regard to privacy considerations' (p xiv).
Roger Clarke and Tim Dixon
ANAO's Report was prepared prior to the progress report for 1992-1993. That report shows some improvements in the comprehensiveness of the cost-benefit analyses. It also shows that ANAO's observation that ''savings from the program have not been as significant as initially predicted' is a substantial understatement. Prior to the commencement of the program, a prediction of savings was made for only one segment (cancellations of benefits). This was originally estimated at $300 million per annum, but in the program's second full year (1992-1993) resulted in savings of only $17 million (pp xi, 17). This was largely because the predicted 60,000-70,000 cancellations turned into 3,682, and the predicted 40,000 downward variations turned into 5,613, as shown below:
ANAO's comments concerning where data-matching is superior to other DSS review mechanisms represent only a small proportion of the exercises performed (pp 48-49), and the total of all reviews resulting from the data-matching program only represents 3 per cent of all reviews undertaken by DSS (p 42). ANAO states ''no identity frauds had been detected from [the personal identity discrepancy reviews]' (pp xiv, 37); ''the Age Pension and Family Payment programs are not subject to any comprehensive regime of risk-based reviews' (p13); and ''data-matching does not outperform random selections drawn from specific client groups in relation to cancellations and downward variations' (p20). The appropriate conclusion therefore is that, in a few specific areas, the program is better than nothing; but it may not be better than conventional reviews based on analyses of high-risk groups.
In relation to ANAO's view that a strength of the program is identification of overpayments, there is nothing in the report to suggest that the estimate given by DMP was audited. No information is provided as to the extent to which these ''identified' overpayments may be collectible. The average amount of overpayment discovered does not appear to be markedly higher than the level at which the Department appears to routinely waive the debt (due to the costs of collection). Indeed, for Family Allowance (where it is supposed to cover the absence of effective risk-based audits), the average is below the $200 level.