Public Audit and Post-legislative Scrutiny Committee

Meeting date: Thursday, February 4, 2021

Official Report 498KB pdf

---

“Administration of Scottish Income Tax 2019-20”

Agenda item 2 is on the report “Administration of Scottish income tax 2019-20”. I welcome Stephen Boyle, who is the Auditor General for Scotland, and Mark Taylor, who is audit director for Audit Scotland. From the National Audit Office I welcome Gareth Davies, who is the Comptroller and Auditor General, and Darren Stewart, who is the director of financial audit.

I understand that the Comptroller and Auditor General and the Auditor General for Scotland will make brief opening statements.

Good morning. This is my first such meeting with the committee. Thank you for the invitation to give evidence. My first meeting would have been last year, but we had to deal in correspondence at that time because of the outbreak of the pandemic. I am pleased to be here.

As the committee is aware, I am required to produce an annual report on the administration of Scottish income tax, as part of my audit of Her Majesty’s Revenue and Customs and, specifically, to give assurance on the calculation of Scottish income tax in three key areas. I will briefly remind the committee of the three areas that our report “Administration of Scottish income tax 2019-20” covers.

First, the report covers the outturn figure that has been calculated for 2018-19 and HMRC’s estimate for the following year, 2019-20. Secondly, we look at the rules and procedures that are in place to administer the tax system and how they affect calculation of the Scottish income tax level. Finally, we look at the costs that HMRC recharges the Scottish Government under the service level agreement. Our report sets out our findings in those areas.

The methodologies for the outturn and the estimate have remained broadly the same as they were the previous year. On the basis of my audit work, I have concluded that both figures are fairly stated. As I have set out, it is clear that the impact of Covid is a significant factor in everything that we are looking at. The impact on the 2018-19 outturn is, not surprisingly, pretty limited, given that the pandemic began in spring 2020. However, it is a relevant matter to consider, so I have explained how the estimate for 2019-20 has been adjusted to a small extent to reflect the impact of Covid-19. Clearly, the impact will be much more important in relation to the 2019-20 outturn and the 2020-21 estimate. I am sure that the committee will want to return to that.

My team and I have worked closely with the Auditor General for Scotland, as we have done in previous years. I am grateful for his team’s co-operation with mine and for the further work that it has reported to the committee.

Good morning.

Scottish income tax remains a key part of the package of new financial powers that have been implemented as a result of the Scotland Act 2012 and the Scotland Act 2016.

The purpose of today’s meeting is to look at the auditing of Scottish income tax. The report that is before the committee relates to 2019-20, which is the third year in which the full amount of non-savings, non-dividend tax that HMRC collects is payable to the Scottish Government, and the second year that HMRC has published Scottish income tax outturns in its accounts.

The outturn figures relate to 2018-19, and the difference between actual United Kingdom and Scottish tax outturns. The amounts that were forecast at the time are adjusted through the 2021-22 budget. That is known as budget reconciliation. Reconciliation for the 2018-19 Scottish income tax outturns results in a reduction to the 2021-22 budget of £309 million. In 2021-22, the Scottish Government plans to borrow £319 million to cover forecast errors arising from income tax and social security reconciliations. HMRC’s annual accounts also include an estimate of Scottish income tax for 2019-20, but that does not affect the Scottish budget.

I, too, welcome the improvements to the methodology that is used to make the estimates. Given the on-going pressures that Covid-19 will put on Scottish budgets, it is important that the methodology be kept under review, as more data becomes available.

HMRC collects and administers Scottish income tax as part of the UK’s overall income tax system. As the Comptroller and Auditor General noted, the NAO audits HMRC’s accounts, and the C and AG is responsible for reporting to the Scottish Parliament on HMRC’s administration of Scottish income tax. I report to the committee to provide additional assurance on the NAO’s audit work, in line with a recommendation that the Public Audit Committee made in 2014. I also explain what the findings mean for the Scottish budget.

In summary, my report says that I am satisfied that the NAO’s audit approach was reasonable and covered the key audit risks. I am also satisfied that the findings and conclusions in the C and AG’s report are reasonably based. The C and AG has concluded that the outturn for Scottish income tax was fairly stated. That provides the Scottish Parliament with valuable assurance on that aspect of the Scottish budget.

I am delighted that the C and AG and his colleague Darren Stewart have been able to join us. As the C and AG noted, we work closely with our colleagues in the NAO. My colleague Mark Taylor leads on that work for Audit Scotland. As ever, we will do our best to answer the committee’s questions.

I thank the Auditor General and the Comptroller and Auditor General for their statements.

Colin Beattie will open the questioning for the committee.

I want to ask about the impact of Covid-19. It is estimated that there will be a reduction of £7 million for 2018-19 and a reduction of £25 million for 2019-20, but the estimate for 2020-21 has not yet been determined. At paragraph 9, the Comptroller and Auditor General’s report says:

“The impact of COVID-19 on future outturns for 2019-20 and 2020-21 is yet to be determined.”

However, in the previous paragraph, it says that HMRC is making an allowance for £25 million for 2019-20. If the impact has not yet been determined, where did it get that figure from?

I will answer initially, and then I will invite my colleague Darren Stewart to add to what I say, based on his detailed audit work.

We are talking about the stage at which the estimates have been developed. The very small amount for 2018-19 relates, in essence, to collectability of the outstanding debt, because the tax that is due is from well before the pandemic started. The pandemic took hold in February and March of 2020, so it affected the very end of the tax year. There will be a slightly bigger impact on 2019-20, but most of the tax that is due had been properly calculated and quite a lot of it had been collected by then.

The £25 million is an initial estimate of the impact of the pandemic on 2019-20. Much more will be known by the time that becomes the outturn figure for 2019-20. In next year’s report, we will audit calculation of that outturn figure and will be able to give much more detail on how the estimate for the Covid-19 impact has been updated.

I ask Darren Stewart to add to that.

Gareth Davies has captured the key points. I draw the committee’s attention to his point that, given the timing of the pandemic, the 2018-19 and 2019-20 adjustments reflect the possible issues relating to recoverability of income tax that was due at that time.

I will give you the overall UK context. On collectability of tax revenues across the board, including from income tax, in its 2019-20 accounts HMRC made a similar adjustment to reflect the fact that the pandemic would have the potential impact of financial hardship on businesses and individuals, which might result in tax debts not being recovered. That is what the adjustments do.

It seems to be strange and anomalous that, having said that there are two particular years that cannot be determined, you have made an allowance against one of the years.

In paragraph 12, reference is made to the need for “behavioural analysis”. Will that be for the accounts of the Scottish Government or those of HMRC?

In the report, we are talking about HMRC understanding taxpayers’ behaviour. It is key that we keep a close eye on whether incentives develop for people to manipulate their address, for example, in order to avoid differential tax rates across the countries. HMRC spends a lot of time trying to understand taxpayers’ behaviour and how new risks to tax collection arise, and it uses that information to target its compliance activity. We are saying that HMRC will need to maintain its understanding of taxpayers’ behaviour, particularly if it spots changes in that behaviour, because it might need to refocus its compliance activity. That explains that reference in the report.

So, that is UK-wide analysis, rather than something that is specific to identifying behavioural changes in Scotland.

It is a UK-wide programme of work. Most compliance activity is based on a set of UK assumptions, but we note the importance of HMRC maintaining a view of whether a Scotland-specific risk to collection of revenue emerges, and of it adjusting its work accordingly. At the moment it is not doing that. In the report, we have set out some of the Scotland-specific activity that the HMRC does, but that is not in that behaviour territory, at the moment.

In the first bullet point of paragraph 1.7, reference is made to

“sensitive records, which are not accessed by HMRC to calculate Scottish income tax”.

What are those “sensitive records”?

I ask Darren Stewart to comment on that.

In the context of the estimate, the numbers are relatively small. We can write to the committee to provide absolute clarification on that, but I think that that is a reference to the people who complete the modelling and estimates not requiring access to substantive evidence that supports those taxpayer records.

It would be useful for the committee to understand that, because we do not know the dimensions of the “sensitive records” or the impact on the calculations that are being made.

The report also talks about the proportion of the Scottish share of tax, and the adjusted Scottish share is 6.9 per cent. Obviously, there is a huge disparity between the number of top-level taxpayers in Scotland and the number south of the border; there is a particular distortion around London and the City of London, which affects everything. Given that you estimate that Scotland’s share of the total income tax take of the UK is 6.9 per cent, I would like to understand a little better how HMRC model work allows for the fact that—from memory—there are something like 300,000 top-rate taxpayers in England and Wales and only 13,000 in Scotland. It is a huge disparity. How does HMRC work that through and bring it down to that 6.9 per cent figure?

09:45  

The core process is to identify Scottish taxpayers by address. That is how the outturn and estimate are built up. Provided that address data is accurate, it will capture accurately taxpayers and the rates that they should be paying. The quality and accuracy of that database help to drive the figures from which comes the proportion of higher-rate taxpayers.

As far as the estimate is concerned, percentage calculations are applied to the UK-wide figures. I invite Darren Stewart to comment on the estimate side of the question.

You have touched on one of the matters that we draw out in the report, which relates to the limitations of the estimation methodology. With regard to the basis on which the estimate is prepared, the Auditor General for Scotland’s opening remarks were helpful. We should remind ourselves that the estimate that HMRC produces is purely for financial reporting purposes. It is the Scottish budget, informed by the work of the Scottish Fiscal Commission, that has a monetary impact on outturn.

However, given that the basis for the estimate is the survey of personal incomes, I think that it is right to say that the number of higher-rate taxpayers in the rest of the UK relative to the number in Scotland has a skewing impact. That is one of the limitations of the approach. In the written evidence that we provided off the back of our 2018-19 report, we pointed to some analysis that the Office for Budget Responsibility had done on the impact of those assumptions, and it drew a similar conclusion.

In the past, the committee has rightly made a point about using outturn data, as it becomes available, to refine the estimate over time. That is one of the ways in which HMRC seeks to address the skewing effect. We might come on to this in a moment, but that process includes a calibration adjustment that takes historical experience and applies it to the current-year forecast. In the round, that takes account of the skewing impact.

Mind you, the calibration is not working too well.

In paragraphs 1.22 and 1.23 of the report, reference is made to the fact that the application of the calibration adjustment to the 2019-20 provisional estimate resulted in a reduction of £300-odd million, which is a 2.3 per cent differential. In the context of the amounts of money that we are talking about, that can be a lot of money.

What worries me is that, according to bullet point 2 in paragraph 1.23, it is not understood why the situation is arising. Does the calibration have a fundamental flaw in it, because it has failed to be validated twice? Given that it is not fully understood why the 2018-19 overestimate came about, how confident can we be as we move forward?

I have two or three points to make. It is absolutely right to point out that, in the context of the Scottish budget, the numbers in question are significant. The calibration adjustment is quite a simple process of applying historical experience to the current-year estimate. That involves the inherent assumption that the level of overestimation in the past will be present in the same way in the future. In that respect, it is relatively crude.

As far as the direction of travel is concerned, although the numbers are very significant, the level of error—the difference between HMRC’s estimate and the outturn that is produced—is reducing over time, but there is a question to be asked about whether the gap is being narrowed sufficiently. In the context of our audits, the 0.9 per cent difference between the 2018-19 estimate and the outturn is within the reasonable tolerance that we would expect. That is purely from an audit perspective, but the monetary sums are still incredibly significant.

If the sequence of differences continues in the same direction that it has been going in for the past two years, that will suggest an underlying problem that is not just the normal variation that one would expect from estimates. We would expect HMRC to pay attention to that and assess whether, as your question suggested, what is happening is systematic rather than just random deviances from estimates. It is really important that we keep an eye on that.

I want to expand on Mr Beattie’s point about the impact of the tax gap and the volatility that we are currently experiencing. I draw the committee’s attention to paragraph 38 of our additional assurance report, which calls on the Scottish Government to think about compliance risks that have not manifested or been identified, but which may crystallise as the tax systems in Scotland and the rest of the UK diverge. The Scottish Government should consider, in discussion with HMRC, what further steps it might want to take. The Government should note in particular the recommendation from the UK Public Accounts Committee on a Scotland-specific tax gap analysis, which might help understanding of some of those compliance risks in taxpayer behaviour.

Does Colin Beattie have any more questions?

I have one more question, convener. Do you want me to ask it now?

Yes, please.

Paragraph 1.24 of the NAO report, which is all about sampling and so on, states:

“The data used for PAYE include all income types and do not exclude tax from savings and dividend income.”

How safe is that? We are doing a calculation that is pretty important for estimating Scottish income tax revenue, which does not include savings or dividend income. The predominant value of that will be south of the border. What is the methodology for making the data valid and relevant to the Scottish rate of income tax?

Darren, would you like to come in on that?

It is not for me to advocate for HMRC, but that touches on one of the inherent challenges of producing such a complex estimate when there are multiple data sources, high volumes, complex estimation techniques and uncertainties. I think that HMRC’s choice is in how far to disaggregate that data in applying it in the methodology. Because all types of income are included, including those that are not directly relevant to Scotland, or the calculation of Scottish income tax, that would have a beneficial impact in broad terms in the context of the estimate.

The only other thing that I would say in that respect is that the calculation of the outturn, which is ultimately what is compared against the Scottish budget, leads to the reconciliation differences that we described earlier. When those reconciliations are done, the calculation is based on the actual position, which is why the outturn is calculated a year in arrears. It is a pretty high level of precision: I think that 90 per cent is the number that we calculated in the report.

I will start with a question for Gareth Davies, on the accuracy of records of Scottish taxpayers’ addresses. Paragraph 11 of your report says:

“Matching records could not be identified for 21.2% of Scottish taxpayers (1.1 million records)”.

Does that mean—I hope that it does not—that more than a fifth of tax codes could be wrong? I assume that it does not mean that.

No, it does not. That is the first figure that was arrived at from a straightforward matching of the address and postcode information that is held by HMRC with information from other sources. An initial long list of queries needed to be addressed. Later in our report, we set out in more detail how HMRC worked through that long list to boil it down to areas in which there appeared to be genuinely inaccurate or missing information. If you follow that down, you come to a much smaller number of pieces of genuinely incorrect or missing postcode information. That figure should not be taken to mean that a very large percentage of records is not correct.

Okay. So what is the number now?

I might struggle to find the relevant table, but I think that 350,000 addresses required further work. The number of records that were clearly in error or missing reduced from 61,000 for the previous year to 31,000 for this year, so the number had roughly halved. That still means that quite a few were not accurate, but there will always be a level of turbulence in the data with people changing addresses because they have moved house and other address changes of that kind. However, that figure gives at least a sense of the scale of the number that remained to be corrected.

So you think that the number is around 31,000.

Yes—for cases in which it was not possible to match information and there was evidence that the address was out of date or incorrect. Clearly, that is 31,000 too many but, given that we have a fluid population, we will always have some changes in the data.

Our report says that HMRC has not identified what an acceptable number of cases would be to allow the system to be regarded as fully compliant. Clearly, it could never be zero. We just pointed out that it has not yet set such a number.

Given that there is no legal requirement for someone to tell HMRC if they move, how easy is it for people to fiddle the system?

It is getting harder because of the availability of accurate data-matching techniques. Because of enhancements in big data analytics, it is much more of a challenge for someone to keep their accurate address from HMRC now than it would have been a few years ago. However, there will always be such a challenge. The reason for our emphasising the point in our report relates to what I said earlier. If the difference between the tax bands and the rest of the system in the two countries continues to diverge, there might be more risk in that area, so having a completely accurate database then becomes a higher priority.

I suppose that that goes back to Colin Beattie’s earlier question about taxpayers’ behaviour. If it will pay someone to lie about their address and it is easy to do that, they might be tempted to do so.

That is right. HMRC was already treating that area of compliance activity as a high priority, but the pandemic has meant that it has had to direct resource for a lot of that—not just the part that affects the Scottish income tax calculation, but all such activity—to other areas, such as the furlough scheme. Therefore, the pandemic might have an impact on that aspect of HMRC’s work for a year or two. However, as things stabilise, it will be important to bring that back into focus, to allow us to understand whether there is evidence of changing behaviour and, if so, what implications that might have for the level of compliance activity that will be required.

Initially, it was reported that 45 MSPs had not been given the Scottish tax code for 2019-20. That seems extraordinary. Do you have any idea how that came about?

10:00  

I do not. That was in last year’s report, was it not? We followed that up this year and HMRC has, not surprisingly, strengthened its checks on whether Scottish parliamentarians have the correct tax code. It now has a specific manual process to ensure that that is the case, and that operated successfully in the year that we audited. We noted in the report that it has since automated that process, because having a separate manual process carries cost. We will check next year whether the automated version is delivering the same level of accuracy.

Do you know what the figure is now?

I think that it was zero at the end of the year that we checked. Is that correct, Darren?

That was a specific area of focus for this year’s report. Based on the work that we did and the inquiries that we made of HMRC and our Audit Scotland colleagues, there were not any further instances. That is what we would expect, given the level of scrutiny that was applied in the previous year and the previous report.

Did you ever look at MPs who represent Scotland for a similar figure?

Yes. They are included. When we say “Scottish parliamentarians”, that includes MSPs, MPs with Scottish seats in Westminster and, in the relevant year, MEPs.

I have one more question; I do not know who will know the answer to it. One of the tools that HMRC was looking at was postcode scans of taxpayer records. Can you tell us anything about how that works?

Does Darren Stewart want to say what we know about that process?

There are a number of different aspects of that. HMRC applies a multilayered approach to ensuring the completeness and accuracy of the Scottish taxpayer base, given that that is the fundamental risk to accurately determining the level of revenue. We have talked about the third party data verification exercise. The postcode scan aspect of that is essentially about interrogating its own records and comparing them against different sources of information across the different systems that it maintains—for example, the pay-as-you-earn and self-assessment systems. There is a control that sits a level above that. It looks annually at the Office for National Statistics record of new postcodes to make sure that its baseline list is complete and that it is accurately benchmarking the taxpayer records that it has against the ONS database.

I hope that that is helpful in giving you a little context.

Neil Bibby, please. Do we have Neil Bibby?

I will come back to Neil in a minute.

We have been speaking about errors, adjustments and areas of focus. A tiny difference might be a nuisance to manage, but at the end of the day it will catch up. Which areas have the potential to cause the biggest absolute gain or loss to the Scottish finances?

Would the Auditor General like to pick that up?

Like members, I am sure, we have been closely following the events of last week regarding the Scottish budget, particularly the Scottish Fiscal Commission’s assessment of the estimates and the likely inescapable impact that the Covid-19 pandemic has had on projections for Scottish public finances. What is clear from that report is the level of volatility that remains in the system and the challenges of making reliable estimates. We have seen volatility in the past in the estimates of tax take and the subsequent reconciliations, but that set of risks seems ever more present.

That is borne out when we look at some of the numbers. The SFC’s forecast for 2021-22 is now a reduction of £634 million in income tax, and it signals that the conditions have been met for a Scotland-specific shock. However, even allowing for that context, it is anticipated that the situation will bring additional budget. Given the way that Barnett consequentials and the block grant adjustment flow through, it is anticipated that they will result in a positive reconciliation at some point down the line due to the divergence between the Scottish Fiscal Commission and Office for Budget Responsibility assumptions. That demonstrates the sheer volatility and complexity of the system.

I ask Mark Taylor whether he wishes to add anything about what we are seeing and, in particular, what it means for the Scottish budget.

I thank Mr Bowman for his question. On a technical point, I note that normally in an accounting system, timing differences will correct over time. What is interesting about how the system that we are discussing works is that, once the outturn is struck, that is it, and the year is over and done with. Some of the things that are based on estimates in the outturn are wholly based on those estimates and, when the actual figures come in, there is no self-correction for them in the way that we would expect in a normal accounting system.

The best way to get a handle on that is to look at figure 2, which comes before paragraph 1.5 in the Comptroller and Auditor General’s report. It shows the estimated amounts, but no information is available on how far out those estimates will ultimately be. It may be up to six years after the outturn was struck before the actual figures are known, but the figures are fixed and we will not get a self-correction. The Scottish budget is fixed at that point. The estimates might be over or under, but that will not resolve itself. That speaks to Mr Bowman’s question about how the timing differences work. The estimates are baked into the outturn, and that is where the risk is.

I think that it is fair to say—colleagues might want to say a little more about this—that there is likely to be more estimation because of Covid. More time is available to taxpayers to complete their tax returns and greater estimation is likely. There is also greater risk in how HMRC might put the figures together in the future. That might be where the risk to the Scottish budget will come through. It is important to note that that can go either way, but we will never get a true sense of the extent of that given the way that the outturn process and system work.

Thank you for that. Are you saying that the estimates are in effect actuals in the calculation, in the sense that they are fixed?

In the calculation of the outturn, yes. To be absolutely clear, the figures in the calculation of the provisional estimate for 2019-20 do not affect the Scottish budget at all. However, in the calculation of the outturn, the figures that are estimated in figure 2 of the Comptroller and Auditor General’s report are fixed at that point and are based on those estimates.

Do we have an opinion on those estimates?

The overall assessment that the Comptroller and Auditor General makes about the figures for the outturn includes an assessment of the reasonableness of the estimates. I am sure that colleagues would be happy to speak to that.

I ask the Comptroller and Auditor General whether he is comfortable with those.

That was a good summary of the report’s conclusion on that point. We checked the methodology that has been—[Inaudible.]

I am sorry, but your microphone went off there. Could you say that again?

Yes. We checked the estimated part of the outturn in the way that has just been described and our view is that it has been calculated reasonably and in line with the agreed methodology.

The point that I would add to the previous comments is that there needs to be a live discussion between the Scottish Parliament and HMRC because there needs to be more accuracy on some of the issues, or the ability to go back and, for example, check whether the estimates that were used in the outturn calculation were subsequently adjusted. Those would be additional processes in HMRC, so there would be a resource implication. There needs to be a continuing dialogue between the Scottish Parliament and HMRC on what level of accuracy is required and what resource is necessary to achieve that.

Thank you—that is helpful.

I now go to Neil Bibby.

One of the compliance risks that have been identified is the hidden economy, where income is not declared. I understand that that would be difficult to quantify. Given that, how can we be confident that there is not a Scotland-specific risk? What, if any, estimates are there of the amount of uncollected tax in the hidden economy?

[Inaudible.]—through the concept of the tax gap, which HMRC uses calculate a UK-wide tax gap. We regularly review the methodology that HMRC uses to make that UK calculation and regularly report on that to the UK Parliament.

Our view is that HMRC has a strong approach to the issue, and it is one on which it continually seeks challenge from academics and others with an interest in the reasons why tax that is legally due is not collected from everyone. The hidden economy is one of the big sources of that gap.

The UK-wide tax gap has reduced slightly. In answer to your question about whether there are any estimates, the figure is around 5 per cent. In other words, if we could identify all the taxes due that should have been collected from the hidden economy and that people have avoided paying in other ways, there would be about a 5 per cent increase in the tax take.

On whether a Scotland-specific tax gap is calculated, we note in the report that it is not. HMRC’s methodology for calculating the UK tax gap goes not allow it to disaggregate that to the four nations. The Public Accounts Committee in the UK Parliament recommended that HMRC considers whether it could move in that direction and focus specifically on Scotland. HMRC’s response to the recommendation was that its resources and methodology currently do not allow it to do that, and it does not have any immediate plans to try to do that.

I am sure that there will continue to be a dialogue between the UK Parliament and HMRC on the issue. Clearly, the committee might have its own views on that, too. However, at the moment, HMRC is saying that it does not plan to calculate a Scottish tax gap, because its methodology does not permit it to do so.

That answer is helpful—thank you very much.

The cash-only economy is a significant part of the hidden economy. Is there any estimate of the impact of Covid-19 on that?

Not specifically, and I certainly cannot help with a Scottish estimate on that. However, the committee might be interested in some UK-wide work that we have done on cash in the economy, which was part of our audit of the Bank of England.

The clear trend through the pandemic has been of reduced use of cash. Lots of retailers were insisting on card transactions, rather than on, for example, cash in the first phase of the pandemic. There was also lower economic activity as a whole. The cash authorities have been responding to that and trying to ensure that people who still rely on cash are not disadvantaged by the general decline in cash use, particularly if it is becoming difficult to use cash for essential purchases, which is a big concern.

I know that that is not quite the focus of your question, but there is certainly a focus on the impact of the pandemic on the use of cash, including the impact on the hidden economy and any lost tax as a result.

Do members have any further questions? Is Alex Neil with us? Are you there, Alex? No, he is not.

As there are no further questions, I thank all the witnesses very much for their evidence and for joining us in this virtual session.

10:14 Meeting continued in private until 10:40.