User:Aithne Arnette/National Fraud Initiative

Since 1996 the Audit Commission has run the National Fraud Initiative, a UK-wide anti-fraud programme. Between 1996 and 2009 it traced £664m in fraud, including £215m in 2008-9, as more councils provided data.

The NFI works by subjecting data drawn from a range of sources to automated processing. Output is sent to participants, together with guidance provided by the NFI.

Audit Commission figures for the amounts of fraud, overpayment and error identified through the data processing and subsequent investigations have been criticised for their use of 'multipliers' to inflate the outcomes.

Some sources have commented critically on the Audit Commission's powers to obtain data, asserting that these are, for example, "privacy invasive".

What is data matching as carried out by the NFI? Does it identify actual anomalies or merely potential anomalies?

In its 2007 memorandum to the House of Commons Select Committee on Home Affairs, the Audit Commission stated that in respect of cases identified as 'matches', "clear guidance is given to the relevant bodies and their auditors that they should treat all matches as anomalies to be checked, rather than being proof that fraud has occurred"

Consistently with this account, it has been asserted that "by its function" data matching will flag "any apparent contradiction between one set of data and another about a household as anomalous".

Thus, for example, The City of York council asserts that "Every council in England takes part, providing databases of users of its various services. The Audit Commission matches the databases together looking for suspicious matches."

Coventry City Council makes the following assertion: "Where a match is found it indicates that there is an inconsistency that requires further investigation. No assumption can be made as to whether there is fraud, error or other explanation until an investigation is carried out."

The Audit Commission has, however, argued that "matches" (ie cases identified as 'hits' using the algorithms applied during the computing processes) can be of at least three sorts:

• a potential inconsistency which is not an actual inconsistency;

• an actual inconsistency that is the result of error; and

• an actual inconsistency that is the result of fraud.

There is also a category of "co-incidental" match, a match which arises because some of the data used was incorrect or confused, such as the case when two different individuals with the same name are confused, or when multiple occupancy buildings "give rise to "data quality issues".

The claim that the Commission may lawfully use computer processing of personal data to indicate that there is a "potential inconsistency" rather than an "actual" inconsistency is clear. This links to Audit Commission claims that the justification for any exercise is not to be sought in the significance of the computer output but in whether it serves the purpose of assisting in the prevention and detection of crime. On this analysis, "data matching" is any computing process whose output assists with the purposes in question, irrespective of whether or not the "hits" produced show any real inconsistencies per se.

There is also a theoretical possibility that a case might be identified as a "match" because of some flaw in the computer algorithms used to process the data, or because of flawed understanding of the two data sets whence the data originated, or because of some flaw in the "meta data" which is provided or created as part of the exercise. On this basis, some have asserted that when sharing data, whether in exercises of this sort, or as part of economy drives within Government, it is important that all "meta data" is checked for fairness and accuracy.

The Human Rights organisation Liberty produced a briefing on the Serious Crime Bill, which introduced wider powers for the NFI to obtain and process personal information.

In that briefing,Liberty argued that the Government might have been "disingenuous" in suggesting that people with nothing to hide had nothing to fear and in giving a misleadingly limited idea of the situations in which personal data would be shared between different bodies. For example, Baroness Scotland had said that the Government would ensure that the provisions were used to target suspected fraudsters, rather than simply those who are potential fraudsters. Liberty doubted that this was a fair account of the matter.

Liberty was also concerned that the proposals were being said to comply with the Data Protection Act when in fact some of them changed it. Liberty expressed further concerns that the definition of data matching was being widened so that it might be seen as encompassing the use of algorithms designed to "'profile’ innocuous mass data in order to identify patterns or characteristics that might (emphasis added) indicate some sort of unusual behaviour or impropriety. This is essentially a fishing expedition, which is not based on any suspicion or intelligence that a particular person or company has done anything wrong." Liberty described such algorithms as 'data mining' and gave an example (page 11). They put forward a number of proposed amendments which were not accepted. It would appear that that which Liberty feared has come to pass, as the two SPD exercises carried out as part of the NFI show.

Such data processing and subsequent investigations do cause distress and complaints. One pilot exercise carried out in Lancashire provoked 90 complaints. Another provoked so much public concern, with front page newspaper coverage, that a public apology was issued and individual letters of apology sent to those who were entitled to the discount.

In the case of the primary Section 11 NFI match, the situation identified by the processing i.e. the presence of a second adult on the electoral register where a Section 11(1) discount of "the appropriate amount" has been deducted, is, as the Elmbridge report cited in this article demonstrates, a perfectly legal and proper situation. Yet the Audit Commission's Code of Practice and its SPD Audit Guidelines insists that these cases are all of equal importance and that: "It is for participants to investigate matches in accordance with their usual practices for investigation of fraud and error."

In the case of the secondary council tax match, which identifies young adults coming up to the age of 18 in a dwelling where a Section 11(1) discount was deducted from that year's council tax bill, the presence of a 17 year old person cannot affect entitlement as under the Act (cited elsewhere in this article) a person has to have attained adulthood to count as a "resident" for CT purposes.

These examples explain, perhaps, why such exercises provoke a degree of public disquiet.

Changes in Fair Processing Notification

Allegations of the sort made above by the City of York, which are termed "fair processing notification" or "privacy notices" to use the more modern term, are quotations from the Audit Commission's Code of Data Matching Practice, which says (Page 11)

"2.2 What is data matching?

2.2.1 The Audit Commission Act 1998 defines data matching as the comparison of sets of data to determine how far they match. The purpose of data matching is to identify inconsistencies that may indicate fraud. "

Page 7 of the same document has the following assertion:

"Data matching in the NFI involves comparing sets of data, such as the payroll or benefits records of a body, against other records held by the same or another body to see how far they match. This allows potentially fraudulent claims and payments to be identified. Where no match is found, the data matching process will have no material impact on those concerned. Where a match is found, it indicates that there is an inconsistency that requires further investigation."

The Statutory Code of Data Matching Practice, laid before Parliament by a minister, includes model FPNs with the same wording.

Following complaints, the Audit Commission contacted participants asking them to modify their fair processing notification. It made the following assertion:

38. Having reviewed the sample level 2 and level 3 fair processing notices in the Code when considering this element of your complaint, I note that they contain a sentence which states that a match ‘indicates that there is an inconsistency that requires further investigation’. The sentence that immediately follows this makes it clear that no ‘assumption can be made as to whether there is fraud, error or other explanation until an investigation is carried out’. However, the first sentence could be more precise if it said that an individual match may indicate an inconsistency. The Commission intends to update the level 3 notice on its website accordingly in the near future and also notify NFI participants of the same amendment to its recommended level 2 notice.

Some councils have failed to modify their FPNs in line with the request of the Audit Commission. This decision has been justified by some councils on the basis that the Code in which the original definition is provided is a statutory document and they must comply with the model notices provided therein.

The so-called single person discount exercises.

Council Tax Law does not use the phrase 'single person discount', and nor does the law provide for any discount to which all persons literally living alone are entitled. Entitlement arises on any day when the conditions set out under Section 11(1) of the Local Government Act apply. The Act uses the word "or". Clearly, one cannot be entitled on any one day under both heads.

The Audit Commission has agreed that the nickname can be misleading. "The entitlement under section 11 is also commonly referred to as the "single person discount" in both the local authority and voluntary advice sectors, although we agree with you that this term does not fully describe the circumstances in which it is available."

The "single person discount" data processing exercises carried out by the NFI, both of which which involve automated processing using the full electoral register, have been the subject of a number of legal disagreements.

A report produced for Elmbridge Borough Council shows the sort of issue faced by participants attempting to follow up the cases identified as "hits" by the NFI and also clarifies some of the legal issues, especially at pages 68(2) and 68(3).

This report clarifies two things:

Firstly it explains clearly that "people can quite properly receive an SPD yet have more than one adult occupier."

Secondly, it explains that the definitions of "residence" in the two legal frameworks from which data is drawn differ, so that the two sets of data are not by their very nature a clear match for each other.

The leading case on "sole or main residence", the definition used in the Local Government Finance Act, is Williams v Horsham District Council. In that case it was said "We think that it is probably impossible to produce a definition of 'main residence' that will provide the appropriate test in all circumstances. Usually, however, a person's main residence will be the dwelling that a reasonable onlooker, with knowledge of the material facts, would regard as that person's home at the material time. That test may not always be an easy one to apply, but we have no doubt as to the conclusion to which it leads in the present case."

On the basis of this case, it is inappropriate for a council to use the full electoral register as a guide to where, for council tax purposes, a person resides. The council should take into account all the material facts and take the perspective of a reasonable onlooker.

A number of local councils and/or electoral registration officers refused to provide their electoral registers to the Audit Commission.

These refusals were motivated by concerns about the law relating to the full electoral register and by concerns relating to the data processing in question, including the definitions of residence referred to above.

Timothy Pitt Payne has stated that

As part of the 2007/08 NFI, the Audit Commission matched Council Tax (CT) data to other datasets, with a view to detecting fraudulent applications for Council Tax Single Persons Discount (CT SPD). As part of this exercise, the Audit Commission sought to collect Council Tax and Electoral Register data from all local authorities. The thinking behind the exercise was that if the Electoral Register showed more than one person registered at an address, but the CT data showed that there is a claim for CT SPD from one of those persons, then this was prima facie evidence that the claim is fraudulent

As explained below, the Audit Commission legal department fully accepted in 2009 when responding to a complaint about allegedly inaccurate information in the NFI Annual Report for 2006/7, that this comparison provides no evidence of fraud.

A legal briefing circulated by the Audit Commission advances the Audit Commission argument that the electoral register was being used for a legitimate audit purpose. It says "The provision of the electoral roll and council tax records for a particular local authority will enable the appointed auditor for that authority to match or compare the two sources of data. The electoral register may indicate, for example, that there is more than one person living in a particular property but the council tax records may show that the person liable for paying council tax is claiming the single occupant discount (emphasis added)."

The Audit Commission denies that this opinion states auditors can by this comparison evaluate the honesty of claims made by taxpayer and it denies that the opinion states that they can ascertain entitlement to a discount by this comparison. On the quotation above, which includes the word 'but', it has said:

"The paragraph does not state that the electoral register can determine entitlement to single person discount. It simply states that the electoral register may show more than one person living at a particular property and the council tax records may show that an occupant of that property is receiving the single person discount as a result of a claim to be entitled. Neither statement is incorrect. The Commission agrees that the electoral register cannot determine entitlement, and has never argued that it could."

A similar assertion was made by the same lawyer later in a second opinion: "The provision of the electoral roll and council tax records for a particular local authority will enable the appointed auditor for that authority to match or compare the two sources of data. The electoral register may indicate, for example, that there is more than one person living in a particular property but the council tax records may show that the person liable for paying council tax is claiming the single occupant discount. That may indicate, amongst other things, that fraudulent claims for single person discounts are being made. Where potential problems arise, those matters are referred back to the local authority concerned to investigate further,"

Later in the same briefing he stated that "possible cases of fraud which are identified when the electoral register is compared with council tax records are referred back to the local authority for investigation". On that basis he describes the exercise as one which is a "mechanism for detecting and preventing fraud" and therefore it is carried out for a "statutory function of the local authority ... relating to crime prevention". Unless the council is using the full electoral register for such a statutory purpose, then, it had been argued, the exercise made unlawful use of the electoral register, as it fell foul of a prohibition under Regulation 107 of the Representation of the People Act asserting that the council should not make any use of the register or of any information derived from it.

The Audit Commission denies, however, that it receives copies of any 'claims' made by the taxpayers in question.

It has also clarified that in its opinion, the comparison between the full electoral register and council tax data sets provides 'no evidence of fraud'.

The Data Specifications sent to local councils by the Audit Commission require councils to 'code' all Section 11(1) discount recipients as either "single" or "disregard". It is these codes which are used by the software to produce lists of "hits" - and not actual claims made by taxpayers.

The Commission asserts that in the two "single person discount" exercises, there is "no ambiguity as to the reason for their entitlement". It has also asserted that the matches relate only to those whose entitlement was determined on the basis that there were no other adults living at the property".

The law states, however, that entitlement is determined by the following criteria: it arises on any day when there is either only one resident who does not fall to be disregarded or there is more than one resident but only one falls to be disregarded. Council Tax demand notices are issued at the start of the tax year in question, with discounts deducted on the basis of two assumptions, both of which relate to entitlement to a particular amount and rate of discount and not to any particular basis within Section 11(1).

The Audit Commission also denies that it has ever thought that one could ascertain entitlement to a discount using the electoral register:

the Commission does not consider that it is possible to identify whether an individual is entitled to a discount under section 11 through individual matches between the electoral register and council tax data sets. We reiterate that it is only following an investigation of matches that entitlement can be determined. It has therefore not instructed counsel on this basis.

The Audit Commission has also clarified that the output does not indicate that there has been a failure to provide information required by law, as any additional voter whose sole or main residence is at the dwelling in question may fall to be disregarded.

However, other guidance issued by the Audit Commission includes the following assertion: "One of the most common reasons for the discount being wrongly paid is when one of the members of the household turns 18 and the applicant (SIC) does not inform the council." This guidance is provided to participants via a secure, log-in protected web site run by the Commission.

In 2009 the Audit Commission refused to comply with a Freedom of Information Request for a redacted version of this guidance showing only those parts of it which provided guidance on how to interpret the output arising from the comparison of the two data bases.

In the course of responding to F of I requests for its guidance on how to interpret, the Audit Commission denied providing guidance on how to investigate as this would be "outside its remit". It also refused to provide the guidance in question on the grounds that it included guidance on how to investigate.

This self-contradiction was commented upon critically by the Audit Commission's Independent Complaints Reviewer, who said that it gave "an unintended impression of duplicity". The Audit Commission's refusal to provide a copy of its guidance on how to interpret the output of the computer processing was upheld by the Information Commissioner. Following the launch of a Tribunal Appeal, however, the Audit Commission produced, in 2011, various copies of the Audit Guide in question, ranging from 2007 onwards.

This information is now, therefore, in the public domain. It shows that the Commission does in fact provide guidance on how to carry out investigations as well as guidance on how to interpret the output. These investigations involve passing on and obtaining a variety of sorts of personal information and meta data, including, it would appear, the fact that one has been identified as a "hit" using NFI computing algorithms.

The Audit Commission's legal briefing (reference above) also clarifies the administrative steps which must be taken by councils when administering the discount, and the information which must be provided to the taxpayer from whose tax the appropriate amount has been deducted.

"A demand notice is a written notice that an authority must serve for each financial year on every person liable to pay council tax under regulation 18, in accordance with regulations 19-21. Under regulation 20, the demand notice shall require the making of payments of the authority’s estimate of the chargeable amount made, as respects part or whole of the relevant year, on various assumptions, including in regulation 20(3)(e) that if, by virtue of regulation 15(1), the chargeable amount is assumed not to be subject to a discount on the day the notice is issued, that it will not be subject to a discount as regards any day after the issue of the notice; and in 20(3)(f) that if, by virtue of regulation 15(2), the chargeable amount is assumed to be subject to a discount on the day the notice is issued, that it will continue to be subject to the same rate of discount as regards every day after the issue of the notice."

"Regulation 3(1) of these Regulations (“the 2003 Regulations”) provides that a council tax demand notice shall contain the matters specified in Schedule 1. Under paragraphs 16 and 17 of Schedule 1, these matters shall include the days (if any) when the amount payable under the notice was calculated by reference to section 11 of the 1992 Act; the reasons for the discount and its amount; a statement that if at any time before the end of the following year the person to whom the notice is issued has reason to believe that the amount of council tax payable is not subject to any discount or is subject to a lesser discount, he must notify the authority of his belief within a period of 21 days beginning on the day on which he first had that belief; and a statement that if the person fails without reasonable excuse to comply with that notification requirement, the authority may impose on him the penalty which is specified in paragraph 1(2) of Schedule 3 to the 1992 Act."

Nothing in this briefing, however, acknowledges the different definitions of residence in the two frameworks, as noted by lawyer Paul Russell at the time when the controversy over NFI demands for the full electoral register was at its height. Indeed, the briefing does not deal with CT law on sole or main residence or with electoral register concepts of residence.

It does not, for example, mention the fact that some people, including students, may lawfully register to vote at more than one address, but that only one of these, usually the university town address, may be their "sole or main residence" for CT purposes.

It does make it clear, however, that in law, the discount is not deducted "on the assumption" that the situations indicated by the codes which the NFI expects councils to attach to personal data will apply at the time of upload, which takes place much later in the tax year, after the electoral register canvass in the autumn following the issuing of the demand notices in spring.

The Code of Data Matching Practice (cited above) makes no provision for any right of appeal against an "assumption" that one has acted fraudulently, or indeed, for the individual to be informed of that assumption. It may, however, be widely shared with public and private sector organisations. Appeals against incorrect adjusted demand notices have to be made through a civil appeals process in which the assumed "criminal" is the appellant and has a burden of evidence.

Valuation Tribunal Appeal Number 2004M29604/257C is an example of a successful appeal against a determination that no discount applied to a dwelling. Hull Billing Authority had determined that the residence of X was also the sole or main residence of her daughter, Y. X's discount had been highlighted by the council using products marketed by Experian Ltd., of Nottingham. This information included information from the electoral register. (In late 2011, the Guardian Newspaper Consumer Rights journalist Miles Brignall reported that a growing list of local authorities are handing over lists of residents granted a Section 11 discount so that Experian could "grade each according to the potential risk of fraud.")

In the Tribunal case in question, Hull had obtained from Experian Ltd personal information enabling it to identify Y by relying on the exemptions provided under Section 29(1) of the DP Act. The Tribunal Hearing Report does not make it clear, however, which section of that provision was relied upon, whether this was for the collection or assessment of tax or in connection with the prevention and detection of crime. Experian's web site clarifies that this is done "where fraud is suspected." On that basis, X was suspected of fraud, but the matter was dealt with by a civil procedure and there was no requirement for the council to present a prima facie case that fraud had taken place. This is cheaper for local councils, and, as the matter is dealt with on a "balance of probabilities" basis, rather than a "beyond reasonable doubt" basis, protection for the taxpayer/suspect is relatively limited.

In respect of the civil case in question, X was supported by a community legal advice centre. Reference was made to R(Williams) v Horsham District Council [2004]. The outcome was that the decision of the council to regard the property as the sole or main residence of Y was deemed "perverse" and on that basis Hull was ordered to re-instate X's discount within two weeks of the order.

Experian is not the only credit reference agency to have established a business stream in connection with highlighting people as statistical fraud risks. Capita works with Equifax to provide a very similar service. They also undertake to "Notify the council of households that are making fraudulent SPD claims and require notification of ceasing SPD benefit." They take payment for this on a no-win, no-fee basis. Among those using this service are Lambeth and Westminster Councils.

The Role of the Information Commissioner

The Code of Data Matching Practice refers to the Information Commissioner. However, that Commissioner has made it clear that he will not consider any complaints about the NFI if it involves making a judgement about any of the legal frameworks pertaining to the data being matched, on the basis that he is a "creature of statute" and such considerations would in effect be ultra vires.

Checks and Balances on the National Fraud Initiative

("awaiting content")