Principal, Xamax Consultancy Pty Ltd, Canberra
Visiting Fellow, Department of Computer Science, Australian National University
Version of 11 March 1994
Published in Privacy Law & Policy Reporter 1,1 (February 1994), and in Policy (Autumn 1994)
© Xamax Consultancy Pty Ltd, 1993, 1994
This document is at http://www.anu.edu.au/people/Roger.Clarke/DV/PaperMatchPDMP.html
The Data Matching Agency, part of the Department of Social Security, operates a particularly large matching program which it calls the 'Parallel Data Matching Program' (PDMP). This involves flows of privacy-sensitive data, including Tax File Number information, among seven agencies. When Parliament authorised the program in 1990, it made it subject to a 'sunset clause'. The program outcomes have fallen far short of the estimates used to justify its creation. But rather than cancelling the program, Parliament has twice extended its date of expiry.
This paper documents how the original benefits estimates, which enveigled the Cabinet and the Parliament into approving the PDMP, were enormously exaggerated. It shows that the Department has continued to abuse the technique of cost/benefit analysis in order to misrepresent the program's worth.
On the basis of the information available, plus reasonable assumptions, assessed using conventional cost/benefit analysis, the program has lost the taxpayer money during its first three years, is currently operating at only breakeven level, and may never provide a return on the investment. Leaving aside the privacy considerations, in sheer economic terms it is not worth doing.
Following its withdrawal of the Australia Card Bill in September 1987, the Government introduced legislation to enhance the Tax File Number (TFN) scheme. After considerable discussion and negotiation, this was passed into law in December 1988, together with the Privacy Act. For reviews of this important period of Australian social history, see Clarke (1987, 1992a).
The debates included iron-clad assurances from the then Treasurer, subsequently Prime Minister, that this would be the sole use of the enhanced TFN. In fact, throughout 1989, and by way of measures included in the 'mini-budget' of February 1990, the Government greatly extended use of the TFN. The Opposition was passing through a 'bone-dry' phase in which no government measure which had the appearance of fiscal responsibility was to be opposed. It therefore supported these measures. By the end of 1990, the TFN was authorised for use in programs relating to non-taxation matters administered by the Australian Taxation Office (in particular Child Support), and to all schemes providing financial benefits to citizens, including programs within the portfolios of social security, veterans' affairs, education and housing.
In order to ensure that taxpayers provide the TFN to the Taxation Office, economic sanctions had been considered to be sufficient; no compulsion existed. To encourage its provision to benefit-paying agencies, however, supply of the TFN was made a condition of application for, and continued receipt of, all benefits. For a detailed analysis of the sequence of events, see Clarke (1991).
As part of its Budget announcements in August 1990, the Government stated that it would extend the use of the Tax File Number still further. It was to be used as the basis for 'the parallel matching of data' from five major client-oriented agencies:
and two other agencies:
The measures were passed into law in early 1991, with a sunset clause limiting the authorisation for the program to 23 January 1993.
In late 1992, the Senate was not satisfied that sufficient information was available to make a decision whether to remove the sunset provision or let the PDMP lapse. It extended the clause until 22 January 1994. In late 1993, despite having before it the evidence outlined in this paper, the Senate again extended the provision.
The most recent formulation of the PDMP's aims (DSS 1992, pp.32-33 and 1993, pp.6-7) is as follows:
A general description of data matching is in Clarke (1992b). The PDMP process is specified in the Data-Matching Program (Assistance and Tax) Act 1990, and described in DSS (1991, 1992, 1993), Kelly (1992) and DPL (1993, pp.2-5). In brief, data is provided by the five source agencies to the Data Matching Agency (DMA). The DMA checks it for validity, then extracts the TFNs and provides them to the ATO. The ATO returns to the DMA identity and income details associated with each TFN. The DMA compares the data from the five source agencies and two assistance agencies, with the aim of identifying identification discrepancies, and anomalies in the patterns of benefit payments and declared income. The DMA provides information about discrepancies and anomalies back to the source agencies.
This 'cycle' is repeated between 5 and 9 times each year. Each cycle involves over 10 million attempted matches, producing about 200,000 apparent discrepancies and anomalies, of which about 15,000 are selected for review. About 10,000 of these survive a preliminary examination, and about 1,500 result in action being taken (indicative figures only; for current details, see DSS 1993, Tables 1-4, pp.109-114).
The original justification for the PDMP (DSS 1990) was severely criticised by this author at the time (see Clarke 1992b, pp.68-9), because the limited information available to the Cabinet and the Parliament contained a number of major errors.
The scheme fared far less well than the Department had predicted. In 1991-92, the gross savings were $25 million rather than the $290 million so confidently predicted. The reasons nominated for this large drop (DSS 1992) were:
A recent Report by the Australian National Audit Office (ANAO 1993) noted that, in relation to 1992, of the $300 million projected savings from 70,000 cancellations, only $17 million had been realised from only 3,682 cancellations. Another source of savings, downward variations in payments had been estimated at 40,000, but only 5,613 had resulted (p.xi, p.17).
As reality set in, the estimates were progressively reduced to about one-eighth of their original size (DSS 1991, pp.105-6; DSS 1992, p.viii, 106; DSS 1993, p.118). The figures in Exhibit 1 relate only to those benefits which arise from cancellations and downward variations, because that was the basis of the Department's original estimates.
91-92 92-93 93-94 94-95 95-96 DSS's Estimates - October 1990 290.0 >300 >300 >300 >300 DSS's Estimates - October 1992 - 107.3 114.1 81.2 99.1 DSS's Estimates - October 1993 - - 35.8 39.8 40.2 DSS's 'Actual' Outcomes 15.0 30.8 - - -
This dramatic example of wishful thinking attracted very little opprobrium (e.g. the Audit Report limited its criticism to "savings from the program have not been as significant as initially predicted" ANAO 1993, p.viii; and Parliament was unprepared to take any action on the matter). It must therefore be assumed that such appalling exaggeration is an accepted part of the game of getting programs approved by Cabinet and Parliament.
The over-enthusiasm of the Department for the program is of historical interest. Of ongoing concern, however, was the Department's failure to subject the PDMP to conventional cost/benefit analysis. Indeed, there were many aspects of the 1992 Report which indicated a failure to even understand the concepts involved (Clarke 1993a, pp.9-11); for example, no costs were imputed for the efforts of other agencies and clients, the basis on which savings were projected into the future was not stated, and net present value techniques were not applied. But the most glaring error was the complete omission of the staff costs involved in 137,000 manual examinations of files, 18,000 actual reviews, 10,000 actions against clients, 1,300 queries by clients, 150 formal appeals, 1,500 debt recovery actions (of which 700 involved negotiations with the debtor), and 100 briefings of the DPP. This omission was despite the statement that "the real cost has been in the time and effort of staff administering the program" (DSS 1992, p.13).
An audit report (ANAO 1993), summarised in Clarke (1993b), also severely criticised the quality of cost/benefit analysis undertaken, and pointed out that the Act "requires the tabling of a comprehensive report in both Houses of Parliament ... Sufficiently comprehensive cost/benefit information had not been included in either Report ..." (p.4). The DSS's 1993 Progress Report acknowledged that "The Privacy Commissioner and some other commentators suggested [the Department's previous reports] lacked sufficient accounting rigour and a more comprehensive cost-benefit analysis covering opportunity costs and costs absorbed by the Department should be conducted. This view was also shared by the Australian National Audit Office. The Department therefore decided to undertake such an analysis" (DSS 1993, p.79).
The information provided in the Department's Report of October 1993 was an improvement on the parlous attempts of previous years but still contained a number of significant errors (DSS 1993, pp.59-97, 118, 121-3, 129). These included the attribution to the data matching program of savings which resulted from another source, the imputation of gross and unjustifiable levels of deterrent effect, the use of unreasonably low opportunity cost figures, the omission of the costs of achieving overpayment recovery, and misapplication of the discounting rate and formulae.
Exhibit 2A shows the net benefits claimed by DSS. In Exhibit 2B, these figures have been adjusted by the author to dampen the Department's exaggeration of benefits and to reflect more reasonable assumptions about opportunity costs (for details, see Clarke 1993c). Exhibit 2B also shows the cumulative net value of the adjusted figures, expressed in 1990/91 dollars, after allowing for the time-value of money and using the standard government discount rate of 8% p.a.
90-91 91-92 92-93 93-94 94-95 95-96 Benefits - 15.0 30.8 35.8 39.8 40.2 Costs 8.0 13.3 13.7 14.7 13.8 13.3 Net Benefits -8.0 1.7 17.1 21.1 26.0 26.9
Benefits - 11.2 26.9 27.2 29.7 30.0 Costs 8.0 25.7 25.3 23.8 22.1 22.1 Net Benefits -8.0 -14.5 1.6 3.4 7.6 7.9 Discount Factor 1.0 0.926 0.857 0.794 0.735 0.681 Net Value -8.0 -13.4 1.4 2.7 5.6 5.4 Cumulative Net Value -8.0 -21.4 -20.0 -17.3 -11.7 -6.3
These figures relate only to the benefits and costs recognised by DSS. Few benefits and some significant costs are incurred by other agencies involved in the scheme, and hence an analysis from the perspective of Australian society as a whole would result in a less favourable outcome.
The original justification for the parallel data matching program was inaccurate to the point of either incompetence or fraudulence. Even after the Department has removed some of the most unreasonable of its misrepresentations, its current justification still contains significant exaggerations. Far from showing vast economic benefits to justify its grossly privacy-invasive nature, the program has lost money so far, will do little better than break even during 1993-94, and may well never pay back the original investment. It is not economically justifiable to continue with the program.
In December 1988, the Government tightened the provisions relating to the Tax File Number (a measure which, unlike its predecessor Australia Card proposal, enjoyed community support). The then Treasurer gave iron-clad assurances that use of the TFN would be restricted to tax purposes. Despite this, the Government extended use of the TFN to all schemes providing financial benefits to citizens, and made provision of the code a condition of receiving benefits. Matching based on the nascent national identification number is but the latest measure in the Government's campaign of social control based on data surveillance.
ANAO (1993) 'Audit Report No. 7 of 1993-94: Efficiency Audit: Department of Social Security - Data Matching' Australian National Audit Office, Canberra (October 1993)
Clarke R. (1987) 'Just Another Piece of Plastic for Your Wallet: The Australia Card' Prometheus 5,1 June 1987. Republished in Computers & Society 18,1 (January 1988), with an Addendum in Computers & Society 18,3 (July 1988)
Clarke R. (1988) 'Information Technology and Dataveillance' Commun. ACM 31,5 (May 1988). Re-published in C. Dunlop and R. Kling (Eds.) 'Controversies in Computing', Academic Press, 1991
Clarke R. (1991) 'The Tax File Number Scheme: A Case Study of Political Assurances and Function Creep' Policy 7,4 (Summer 1991)
Clarke R. (1992a) 'The Resistible Rise of the Australian National Personal Data System' Software L. J. 5,1 (January 1992)
Clarke R. (1992b) 'Computer Matching by Government Agencies: A Normative Regulatory Framework' Dept of Commerce, Australian National Uni. (August 1992)
Clarke R. (1993a) 'Comments on the DSS's October 1992 Report on Progress of the Data Matching Program (Assistance and Tax)' Dept of Commerce, Australian National Uni. (January 1993)
Clarke R. (1993b) 'ANAO's Audit Report on the DSS Parallel Matching Scheme' Dept of Commerce, Australian National Uni. (October 1993)
Clarke R. (1993c) 'The Parallel Data Matching Program: A Highly Privacy-Invasive Scheme That Is Economically Unjustifiable' Working Paper, Department of Commerce, Australian National University (November 1993)
DPL (1993) 'Data-Matching in Commonwealth Administration' Dept of the Parliamentary Library, Canberra, Issues Brief No. 8 (June 1993)
DSS (1990) '1990 Budget Decision - Computer Matching' Dept of Social Security, Canberra (October 1990)
DSS (1991) 'Data Matching Program (Assistance and Tax): Report on Progress' Dept of Social Security and Data Matching Agency, Dept of Social Security, Canberra (October 1991)
DSS (1992) 'Data Matching Program (Assistance and Tax): Report on Progress - October 1992' Dept of Social Security, Canberra (October 1992)
DSS (1993) 'Data Matching Program: Report on Progress - October 1993' Dept of Social Security, Canberra (October 1993)
Kelly P. (1992) 'Australian Federal Privacy Laws and the Role of the Privacy Commissioner in Monitoring Data Matching' in Clarke R. & Cameron J. (Eds.) 'Managing Information Technology's Organisational Impact II' Elsevier/North Holland, Amsterdam, 1992
The PDMP is an exercise in mass dataveillance (Clarke 1988): inconsistency between the data collections of different agencies is treated as prima facie evidence that an error has occurred, or that a misdemeanour or fraud has been committed. Various automated and manual steps are used to filter out spurious and relatively unimportant inconsistencies. The remainder are investigated. Selected cases result in enquiries being directed by any of the source agencies to the client, and ultimately to variation or cancellation of the benefit, and action to recover over-payments of benefit and to collect under-payment of tax.
The key elements of the PDMP's privacy-invasiveness are:
Go to Roger's Home Page.
Go to the contents-page for this segment.
Send an email to Roger
Last Amended: 13 October 1995
|These community service pages are a joint offering of the Australian National University (which provides the infrastructure), and Roger Clarke (who provides the content).|
| The Australian National University|
Visiting Fellow, Faculty of
Engineering and Information Technology,
Information Sciences Building Room 211
Pty Ltd, ACN: 002 360 456|
78 Sidaway St
Chapman ACT 2611 AUSTRALIA
Tel: +61 2 6288 1472, 6288 6916