.png)
.png)
KEY POINTS
- Tax is not lost through tax avoidance schemes.
- Most tax losses come from the black economy.
- Will results from HMRC's random enquiries give correct figures of tax losses?
- The difference between fraud and innocent mistake.
- An objective body should collate data on tax losses.
My eye was caught by an intriguing title: 'Methodologies for measuring direct tax losses', in the list of supplementary documents published at the time of the Pre-Budget Report.
The Chancellor had not referred to the paper in his speech and I had not noticed any other references to it in the Pre-Budget material. The document explained that it:
'… discusses the methodological issues associated with estimating direct tax losses, sets out HMRC's approach to further development of direct tax loss measures and describes the more robust results available.'
Direct tax 'losses'
Making a lightning deduction, I decided that the paper was concerned with direct tax losses. But what were they? The paper contained a glossary of technical terms which helpfully defined such terms as 'econometrics' and 'evasion', but there was no mention of the subject of the paper itself, 'direct tax losses'. Nor is there a definition to be found within the paper.
Most of the paper is concerned with methods of measuring the amount of tax due under the law which, for one reason or another, is not collected. At paragraph 2.16, however, it lists five 'significant types of losses' that cannot be covered by random enquiries. One of these is 'the use of avoidance schemes/devices to reduce liability'.
So perhaps 'direct tax losses' bears some relationship to a concept to which HMRC speakers have been referring for some time as the 'tax gap'. In his Hardman lecture on 18 November 2004, the Permanent Secretary to the Treasury, Gus O'Donnell, defined the tax gap as:
'… the difference between the tax actually paid and what Parliament intends to be paid.'
A conceptual confusion
This is to confuse the legal fiction that Parliament, which consists of 646 MPs, the Lords spiritual and temporal and the Queen, have a common intention when they pass an Act with an actual state of affairs.
Even if we were to accept that that legal fiction does correspond to the actual state of affairs, it is absurd to think that Parliament has in mind a particular sum of money which it intends to be raised by the entirety of its tax legislation.
What is more, though one might, at a stretch, describe the difference between the amount which Parliament intended to be raised by its legislation and the amount which it actually does raise as a 'gap', it is conceptual nonsense to refer to it as a 'loss'. A thing which does not exist cannot be lost. It cannot be said too often that the term 'tax avoidance' contains a conceptual confusion.
'Tax avoidance' involves undertaking transactions which attract a lower tax liability than other transactions which one might have undertaken, but did not. The result is that a liability, which might come into existence, never does so. Tax is not lost; the liability to pay the tax simply does not arise. This is not simply a matter of semantics. How does one measure something which does not exist?
In its consideration of 'losses' from tax avoidance, the paper starts by saying:
'Tax losses arising from the use of avoidance schemes/devices cannot be measured solely through the results of enquiries.'
How can they be measured from enquiries at all? The purpose of an enquiry is to determine whether the tax due under the law has been properly accounted for and paid. If an adjustment is made, it is because the taxpayer has been found not (or at least has conceded that he has not) to have paid the tax due from him under the law.
That tax is not tax which he has avoided. It is tax which he has either attempted to evade or has underpaid under a mistake. So the process of the enquiry can give no information about tax foregone by reason of tax avoidance, unless one defines that tax loss as being the difference between the tax which the officer or his superiors considers ought to have been due under the law, and the tax which was actually due.
That would be a hopelessly subjective figure giving no useful information at all except, perhaps, as to the political prejudices of HMRC officials.
The paper goes on to say that the: '… avoidance disclosure regime [under FA 2004 Part VII] … is starting to provide a greatly enhanced view of the use of marketed avoidance schemes and the potentially associated losses.
'A full publication schedule of quarterly information at six monthly publication dates has been in place for nearly 18 months.'
I have demonstrated in earlier articles in Taxation (see Related lLinks above) that HMRC have encouraged widespread non-compliance with the disclosure regime through the publication of misleading guidance.
Even if that were not the case, the disclosure regime could hardly provide information about tax lost through tax avoidance, as it is not concerned with tax avoidance but with the obtaining of a tax advantage. When the disclosure regime was introduced in 2004, the professional bodies suggested that the regime should be based upon a test of tax avoidance, but HMRC rejected this suggestion.
Furthermore, the regime does not require disclosure of quantitative information about the amount of tax avoided by transactions taking place under the disclosed strategies. It is true that through the provisions relating to disclosures by taxpayers, HMRC should be able to determine what actual transactions have taken place implementing the disclosed strategies.
However, that information will only reveal the tax which actually became due under the law and not the tax which would have become due had the taxpayer adopted some other course of action.
So from the very weakness of the sources of information to which the paper refers, it is clear that HMRC have no reliable method of quantifying what it incorrectly describes as tax losses from tax avoidance.
Measuring real tax losses
Fortunately, the paper is mainly concerned with quantifying real tax losses. That is, the difference between the tax actually due under the law and the amount collected, and here it does have some interesting things to say.
It explains that there are two possible approaches to measuring revenue loss. The first is the top down method which involves: 'Constructing an estimate of the total tax liability from sources independent of data collected for tax purposes, which is then compared with actual receipts.'
It explains that this is not possible for direct taxes because: 'There are no independent sources of data describing the taxable income for the various taxpayer populations. Moreover the structures of the taxes are more complex and a range of tax rates may apply to different parts of taxpayers' income.'
So the paper concludes that the only available method is 'the bottom up method'. What is meant by this is that tax losses are estimated from the information arising from HMRC's compliance activities. The problem with this approach is that it misses those persons, assets and activities which are unknown to HMRC. Almost by definition the greatest losses will come from this area of the unknown.
Even within the sphere of its activities, the information which is available to HMRC is mainly not of a type which can produce statistically significant results. The paper explains that 'most compliance work is risk based'. A sample based on such work would be skewed towards taxpayers whose returns are likely to reveal under-declarations.
Random enquiries
HMRC does have random enquiry programmes for income tax self assessment, corporation tax self assessment and employer's PAYE and National Insurance and, as the paper reminds us, 'HMRC are also developing random enquiry programmes for other taxes, specifically inheritance tax and stamp duty land tax'.
The paper, however, notes that: 'The random enquiry programmes will not identify all incorrect returns or the full scale of losses, especially where independent information from third parties is not available to verify the data supplied by the taxpayer. This means that tax loss estimates produced through random enquiries underestimate the full extent of the losses.'
This is puzzling. The burden of proof in an enquiry rests on the taxpayer. Where information is lacking, that is far more likely to result in an over assessment than in an under assessment. We shall return to this very significant matter later in the article.
In fact, the random enquiry process has characteristics unnoticed in the paper which undermine the statistical validity of results based on it.
To provide statistically valid results, an enquiry would have to provide a high degree of probability that the extent of the taxpayer's under or over declaration had been accurately identified. An amount which is immaterial in relation to Government revenues, or even to the income of the taxpayer concerned, might be statistically significant when extrapolated across the taxpayer population.
Even in random enquiries, the officer conducting the enquiry takes account of the costs of conducting an enquiry and the value of the time expended on it in deciding how far to pursue his enquiries.
He should also avoid imposing burdens on the taxpayer which are disproportionate to the tax at stake, particularly when dealing with innocent errors. Ordinary enquiries will not therefore provide the degree of certainty that an Inspector has identified the correct liability that statistical reliability requires.
If HMRC are to develop their use of random enquiries for statistical purposes, the temptation will be to ignore issues of proportionality in order to obtain statistically valid results. To do so will create another significant shift in the relationship between HMRC and the taxpayer. Random enquiries were first introduced with self assessment in relation to the fiscal year 1996-97.
For the first time, individual taxpayers were put to the expense of proving the correctness of their returns where there was no reason to think that their return was anything other than correct. That marked a fundamental change in the relationship between the revenue authorities and the taxpayer.
There is all the difference in the world between HMRC saying to the body of taxpayers that they wish to impose significant costs on particular taxpayers chosen at random, because to do so is necessary to deter taxpayers from under-declaring their income, and saying that they wish to do so simply in order to provide themselves with a set of statistics.
Unmeasured losses
The paper lists what it regards as the five 'most significant [categories of] tax losses' which cannot be measured from the results of random enquiries as follows:
- non-payment;
- large business;
- the use of avoidance schemes\devices to reduce liability;
- the informal economy; and
- individuals who are not issued a return.
Non-payment
The paper does not say what it means by 'non-payment', although it seems to refer to situations where HMRC have quantified a liability but have failed to collect it. It seems odd to include this as a separate category. The liability will either arise under the self assessment process subject to random enquiry or it will arise under one of the headings of other types of losses, subject to the methodologies appropriate to that particular sort of loss.
Large business
The inclusion of 'large business' as a category of tax loss is also rather puzzling.
The paper explains that: '… a full enquiry can involve a complete examination of all books and records. For the largest businesses this is not a practical approach because of the huge resource requirement it would place on both HMRC and the business.
'However HMRC carries out regular risk assessments for all large businesses to identify the key areas where tax treatment is in doubt.'
As we have seen, risk assessment produces statistically unreliable results because it skews the sample. As all companies and unincorporated businesses are subject to enquiry under self assessment, it is difficult to understand why the results of random enquiries into large businesses could not be used.
It surely is not necessary to examine every item of the books and records of a large company to produce statistically significant results. Auditors have been using statistical sampling systems to produce estimates of the errors in company accounting records at high degrees of statistical probability for decades. The paper does not explain why similar approaches could not be applied by HMRC.
The paper goes on to comment that: 'Other countries have used variations in effective rates as an indicator of tax performance. These methods are more difficult to centrally apply in the UK context because there is no group basis for company taxation (unlike many other leading OECD countries) and our accounting standards make it more difficult to identify the UK element of quoted companies' commercial profits.'
Again this is very puzzling. The difficulty of estimating tax losses from accounting profits is more fundamental than differences in group accounting methods; it is based on the fact that successive governments have refused to base corporation tax liabilities on accounting profits, because they regard such profits as being too easily manipulated.
Whether that is a reasonable view or not, it explains why the rate of tax on accounting profits disclosed in the accounts of companies does not correspond to corporation tax rates.
HMRC have confidence
So it is pretty clear from the paper that HMRC do not have statistically significant information, other than that which is available from the random enquiry programme, and that the random enquiry programme itself has a significant risk of statistical bias. In spite of that, the paper concludes that:
'HMRC does have confidence in the figures that come from random enquiry programmes and analysis of non-payment [sic].'
Are the results persuasive?
The report goes on to describe the results of random enquiries for individuals, small and medium-sized employers and small and medium-sized companies, providing a summary of statistics and results and explaining the nature of those statistics.
Some of the results revealed are quite extraordinary.
The proportion of returns where liabilities are understated is remarkably high. In relation to the self employed, it is estimated that about half of the returns submitted understated the true tax liability. For other individuals, the proportion of returns which understated liabilities are still high at 16% to 18%. In relation to small and medium-sized companies, the percentages are 37% to 43%.
This is a truly staggering level of non-compliance, if it is true. It would indicate either a remarkably high level of mistake or dishonesty, or both. Should we draw that conclusion from the paper? I do not believe we should.
The paper itself seems to show a peculiarly skewed view of the enquiry process. For example, among all the statistical difficulties which it very sensibly points out, the one thing which it should be relatively easy to identify is the proportion of the tax loss determined under the random enquiry process which is caused by fraud or negligence on the one hand, and the proportion which is caused by innocent mistake on the other.
That is because penalties are imposed for loss of tax due to fraud or mistake, and Inspectors conducting enquiries are highly motivated to impose such penalties where they can. There is surely a very significant difference between understatements relating to fraud and negligence on the one hand and to innocent mistake on the other. Yet the paper does not distinguish between the two.
If we take errors arising from innocent mistake, they should as often result in overpayments as in underpayments. Yet the possibility of there being a statistically significant degree of overpayment balancing the loss from innocent underpayment is not discussed in the paper.
Although we have a tax system which everybody admits to be vastly and overly complicated, the possibility that large numbers of taxpayers pay more than is due under the law through their mistakes or those of Revenue officials is not considered.
Just as significant is the possibility that the enquiry process results in significant adjustments which taxpayers know or suspect are not due under the law, because the burden of proving that that is the case is disproportionate to the amount of tax at stake.
Tax Inspectors conducting enquiries are motivated to obtain adjustments favourable to HMRC. Every accountant dealing with tax returns with whom I have discussed the matter has said that he has known taxpayers to pay tax not legally due, simply in order to bring an enquiry to an end, because the professional fees and time involved in establishing the correct position in law is greater than the amount of tax at stake.
Any paper which purports to deal in an objective manner with 'tax losses' should at least take account of the possibility of this countervailing 'tax gain'.
Independent approach?
The failure to deal with these possibilities surely raises a question as to whether the paper is truly objective and therefore whether HMRC are the correct body to deal with estimating tax losses.
Taxpayers generally are likely to have more confidence in the objectivity of the data obtained and the conclusions drawn from it, if an independent body is responsible for overseeing fiscal statistical information. Estimates of tax losses are important. They may well result in significant policy changes in relation to tax administration and HMRC's powers.
These matters are too important for the information on which the decisions are based to be left entirely under the control of HMRC.
Simon McKie is the chairman of McKie & Co (Advisory Services) LLP, telephone: 01373 830 956, e-mail: simon@mckieandco.com
