Public Audit Committee

Meeting date: Thursday, February 3, 2022

Official Report 538KB pdf

---

“Administration of Scottish income tax 2020/21”

Agenda item 2, which is the principal item on this morning’s agenda, is an evidence-taking session on the report “Administration of Scottish income tax 2020/21”. I am pleased to welcome once again to the committee room the Auditor General for Scotland, Stephen Boyle, who is joined online by Mark Taylor, audit director, Audit Scotland.

I am also particularly pleased to welcome to the Scottish Parliament’s Public Audit Committee Gareth Davies, Comptroller and Auditor General, National Audit Office, who is joined by Darren Stewart, audit director, National Audit Office. I think that this is the first time that Mr Davies has given evidence to the committee. Unfortunately, Mr Davies, we are able to take your evidence only in online form this morning, but I hope that in the not-too-distant future we will be able to welcome you to the Scottish Parliament to meet the Public Audit Committee in person.

I invite Stephen Boyle to give a short opening statement. I will then ask the NAO’s Comptroller and Auditor General to make some opening remarks, but over to you, Auditor General.

Good morning, everybody. Scottish income tax remains a key part of the package of new financial powers that was implemented as a result of the Scotland Acts 2012 and 2016. The purpose of today’s evidence session is to look at its administration, and the reports before the committee relate to 2020-21, the fourth year in which the full amount of non-savings and non-dividend tax collected by Her Majesty’s Revenue and Customs is payable to the Scottish Government.

It is also the third year in which HMRC has published Scottish income tax outturns in its accounts. Those outturn figures relate to 2019-20, and the difference between actual United Kingdom and Scottish tax outturns and the amounts forecast at the time is then adjusted for the 2022-23 budgets in what is known as a budget reconciliation. The reconciliation for the 2019-20 outturns will result in a budget reduction of £34 million in 2022-23. HMRC’s annual accounts also include an estimate of Scottish income tax for 2020-21, but that does not yet affect the Scottish budget.

HMRC collects and administers Scottish income tax as part of the overall UK income tax system. The NAO audits HMRC’s accounts, and the Comptroller and Auditor General is responsible for reporting to the Scottish Parliament on HMRC’s administration. I report to this committee as part of an additional assurance process on the NAO’s work in line with a recommendation dating back to 2014, made by a predecessor public audit committee. I also explain some of the impacts that the reconciliation has on 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, and I am also satisfied that the findings and conclusions in the Comptroller and Auditor General’s report are reasonably based. The C and AG has concluded that the outturn on Scottish income tax was fairly stated, and that provides the Scottish Parliament with valuable assurance with regard to this aspect of the Scottish budget.

As ever, we look forward to answering the committee’s questions but, as you suggested, I will briefly hand over to Gareth Davies.

Thank you. First, I will briefly explain my role in this process. As the Auditor General said, I am required to report to the Scottish Parliament on the outturn figures—in this case, for 2019-20—and HMRC’s estimates of the revenue from Scottish income tax for 2021. We have also examined HMRC’s administration of the system as it applies to Scottish income tax and the costs recharged by HMRC to the Scottish Government under the service level agreement. As the Auditor General said, the methodologies with regard to the outturn estimate have remained broadly similar with the previous year and I have concluded that both are reasonable.

This is the seventh annual report that we have produced, and HMRC’s administration of Scottish income tax has now reached what is essentially the implementation of business as usual. The rules and processes are well embedded and my report paints a picture of the consolidation of progress made in previous years. Clearly, HMRC’s focus must now be on refining its processes to maintain an accurate and complete record of the Scottish taxpayer population and on continuing to monitor the risk of non-compliance that might or might not arise as a result of divergence between UK and Scottish tax rates.

I am sure that the committee will be interested in hearing about the impact of Covid-19 on the figures in the report and in subsequent years. In summary, it is fair to say that it has had less of an impact on the collection of income tax in the UK as a whole and in Scotland than might have been expected at the start of the pandemic. In fact, the outturn is slightly higher than the estimate for 2019-20 and represents an increase on the previous year. The estimate for 2020-21 also represents a further increase. In the context of a pandemic, that might be slightly surprising, but it is also very welcome as far as tax collection is concerned.

That position is consistent with what we have seen in our audit of HMRC for the UK as a whole, with Covid support schemes stabilising employment. Now that those schemes have ended, HMRC will be paying a lot of attention to the impact on tax payment rates, the level of taxpayer debt to HMRC and the impact of all that on the amount that is finally collected.

Throughout our audit, my team and I have worked closely with the Auditor General for Scotland and Audit Scotland colleagues, and I am very grateful for their collaboration on this work.

I look forward to answering the committee’s questions.

I thank Mr Davies and the Auditor General very much indeed for their opening statements. As you will expect, we have a significant number of questions.

I want to begin where you left off, Mr Davies. It seems a little bit counterintuitive that at a time of huge collapse in the economy, gross domestic product, gross value added and all the other measures of economic performance, the estimates suggest an increase in the tax take when the pandemic was at its height. We keep coming across references to estimates, samples and assumptions, but can we rely on the accuracy of those estimates?

You are right to be cautious with regard to the full impact of the pandemic, not so much on the assessment of tax liabilities—after all, there is more clarity on the amount of tax owed by taxpayers through self-assessment and the pay-as-you-earn scheme—but on the collection of tax due. We need to bear it in mind that the outturn that we are reporting on here—which is, if you like, the most accurate figure in this set of figures—is for 2019-20, and most of that year was pre-pandemic. The first full year of the pandemic’s impact would therefore be 2020-21.

As we have seen, the estimate is robust and shows a healthy level of tax collection, but I would sound a note of caution that the risk of high levels of non-collection and non-compliance has not yet been completely eliminated. As I am sure that we will go on to discuss, HMRC has made some reasonable estimates on the impact of those factors and they have been taken into account in coming up with this particular estimate.

However, there is a higher level of risk than normal because, until we have seen exactly how much tax is collected rather than assessed as due, we will not know the full picture. I am sure that we will get into some more detail later on how those estimates have been constructed.

On your broader point about the impact of the pandemic on the economy, the total tax take for the UK has reduced, but it is taxes other than income tax that have suffered the brunt of that impact, particularly VAT. A £15 billion reduction is anticipated in VAT, which is the biggest single fall in the tax take for the UK. There are also reductions in hydrocarbon duties—another significant amount—and some other business taxes. However, income tax on individuals has held up. I think that that is because of the Covid support schemes that were put in place; the furlough scheme and the self-employed income scheme heavily protected levels of income through the worst phases of the pandemic. That has resulted in higher levels of income tax than might have been expected.

Thank you. Both PAYE receipts and self-assessment receipts have grown, according to the figures that are presented in the report. Is that correct?

That is correct, yes.

There was, if I remember rightly, a postponement of the deadline for self-assessment tax returns. Has that had any impact on collection rates?

It has certainly had an impact on the amount owed by taxpayers to HMRC. At this point, I am talking about the UK-wide amount, because we have not done a separate analysis of tax collection as regards Scottish taxpayers. However, at this stage we have no reason to think that the picture is different in the various parts of the UK.

As part of our audit of HMRC, we reported separately on the large growth in tax debt owed by individuals to HMRC through the course of the pandemic. As you say, people were allowed to defer payments. Not surprisingly, that led to a significant increase in the total amount owed, peaking in the late summer of 2020. Since then, it has been falling, as people realised that they could actually pay the taxes that were due—again, partly because of the economic support schemes in place at the time.

However, the amount still outstanding relating to 2019-20 and 2020-21 is higher than it was pre-pandemic, so HMRC still has a major job to do in collecting the outstanding amounts. Of course, in general, the longer taxes remain due, the harder they are to collect, because people have to keep up with their current tax bills, let alone pay arrears from previous years. Your equivalent committee in Westminster held a specific hearing on the collection of outstanding tax debt by HMRC and it is reporting on that shortly.

Thanks, Mr Davies. We will look forward to that.

One of the things that is concerning this committee a little bit is the extent to which we are still reliant on estimates, samples and assumptions rather than hard data. The system of a Scottish income tax arrangement has been in place now for three years; why is it that we are still so reliant on estimates, samples and assumptions rather than being able to rely on three years’ worth of Scottish income tax outturn information? I will ask Mr Davies to answer that first and I will come to Stephen Boyle second.

Some element of estimation is inevitable in a system of this kind; the challenge is to make that as small as possible. An element of estimation is always likely to be included just because of timing, essentially. The bulk of the figures in the outturn for 2019-20 are certain and are known, so the vast majority of it is based on accurate data. However, the final tax take depends on the effectiveness of recovery action, tackling non-compliance and pursuing difficult cases, sometimes through the tribunals and so on. That all takes a long time and is not complete by the date that the estimates have to be reached for the purposes of reporting to the Scottish Parliament.

It is in those areas where recovery action is outstanding that you will have to estimate the likelihood of success and the amount that will be collected, but that will always be the prime area where things are difficult to resolve in the time available before these figures have to be reached. I assure the committee that the vast majority of the numbers in the report are based on actual tax returns and money received and collected.

09:15  

I believe that you raised this issue in your commentary on the NAO report, Mr Boyle, so perhaps you can give us your views on it.

The headline for us is that we are satisfied with the NAO’s approach and, through our work, we are content with the judgments made by our NAO colleagues.

On the circumstances surrounding the estimates, I would perhaps make a couple of points, one of which Gareth Davies has touched on already. First, there is undoubtedly volatility by virtue of the pandemic. Although the estimates are reasonable, we do not yet know with sufficient accuracy what further challenges to the estimate of the tax take are coming down the line or the effect of individual behaviour. As has been touched on in aspects of the report, there are still risks to the tax take in future as a result of a couple of factors. For example, there is the divergence that we are already seeing in the different tax regimes in Scotland and elsewhere in the UK and which might yet lead to changes in behaviour. I am sure that the committee will want to come back to that issue.

It might be helpful by way of reassurance to compare the estimate here with that of the Scottish Fiscal Commission; the fact that they are relatively close to each other should provide an element of additional validation. Therefore, although there is uncertainty, HMRC’s own estimates, those of the Fiscal Commission and the assurance activity carried out by our NAO colleagues give us as much confidence as is available that there is a reasonable basis for the estimate.

I have a final question before I open discussion up to the rest of the committee. This arrangement is quite new and evolving, and we are to some extent learning as we go along, but does HMRC have any plans to change its approach in future, given the extent to which it relies on estimates for the reports that we at the Scottish Parliament get on income tax take in Scotland?

From our work with HMRC on this, it is clear that it keeps the entire calculation under review, with a focus on areas where it has to make adjustments and estimates. Although most of those adjustments and estimates are small compared with the overall figure, they require judgment to be applied and, as a result, HMRC reviews its approach.

It is fair to say that we have seen a mindset of continuous improvement. HMRC is very keen to ensure that the estimates use the best available information, it challenges whether previous approaches have been accurate enough and it will suggest changes that it thinks will improve quality. We have seen that approach in recent years and we think that it is healthy.

Of course, HMRC does not do this on its own. There is governance under which it and the Scottish Government test whether the approach is sound and sign off any proposed methodology changes. The agreement of both parties is required before any such changes are approved and implemented, which I think is a good control with regard to the methodology’s development.

As for your main question, however, we have seen evidence that HMRC is keen to improve the process over time. It also takes quite a lot of pride in the accuracy of its estimates, so it is keen to understand anything that departs from its assumptions and to propose changes as a result.

Thank you very much indeed. I call Colin Beattie, who has a series of questions.

Mr Davies, I have been involved with public audit committees since the Scottish rate of income tax was introduced several years ago. I want to begin with a very simple question, and then I will focus more on the actual report.

One of the things that jumps out of the report is the 70 per cent increase in the number of missing Scottish postcodes. It still represents a small proportion of the taxpayer population, but if it includes large numbers of high-net-worth individuals, it could have a significant effect on the tax collected. What is behind that increase? What is driving the error?

I will bring in my colleague Darren Stewart to answer that, as he oversaw the detailed work on the issue and might be better placed to give you further details.

In conducting our work, we posed that question to HMRC, but I do not think that it has clarity with regard to exactly what is driving the increase in missing postcodes. What I would point out—and this is reflected in the report—is that it does case work on those postcodes. Where they relate to Scottish taxpaying individuals—in other words, those who are in employment, have pension arrangements and so on—it updates its records to ensure that the correct amount of tax is being collected. It is a trend that we have observed and played back to HMRC, but there has been an assurance that, for postcodes where there is an impact on tax collection, some remediating action has been taken.

Do we have any idea at all of the impact on the actual revenues that are being collected?

From our work with HMRC, we have a relative degree of comfort that the missing postcodes are not driving significant understatement—which I guess is what we are interested in here—of the revenue that is being collected, but we have suggested that some analysis be carried out on identifying the root causes as part of the continuous improvement that Gareth Davies referred to earlier.

So you are satisfied that HMRC is taking steps to deal with the issue.

I think so, based on the work that we have undertaken.

You think so. Do we know whether the problem will recur?

The issue will be a key focus of our forthcoming report for 2021-22, and it is an area that we have pointed out to HMRC and which we will be investigating with it in preparing that report.

Surely it indicates that there is a flaw in the process.

The key thing is to identify the root cause and take steps to address it. We will be taking a keen interest in the issue for our 2021-22 report.

Given that the problem has been happening every year since the system was instituted, it must endemic to it.

It is difficult to say without having undertaken that analysis, but it is clearly something that we are keenly interested in and which we will be taking forward with HMRC.

But if it happens every year, there is clearly a flaw in the system.

In the report, we say that it represents, as you have said, a relatively small number of individuals compared with the overall 2.5 million Scottish taxpayer population. Given that, as you have pointed out, the system is still maturing, we would expect HMRC to be looking at such marginal gains to see what continuous improvement can be made.

From our work on similar systems elsewhere in Government, we can see that it is at heart a data quality issue. Such large data collection systems require continual effort to maintain data quality. You might think that if you have a correct postcode you can be assured of its staying correct, but that is not always the case with these systems. For example, records might be updated with other changes and inadvertently errors are made in the postcode field.

The important question is whether HMRC has sufficient checks in place to identify where such errors are arising and a robust system for correcting them. Data quality issues will always arise in such systems, so the issue is how good the system is at detecting and correcting them quickly, and that is what we will focus on in our work.

The convener has already highlighted the multiplicity of references to estimates in the report. That is not terribly encouraging when we are trying to plan expenditure against the income that we are likely to get from taxation.

When the Scottish rate of income tax system came in, the first year or two were a bit of a shambles, to be honest, but we kind of expected that. However, it disappoints me that I do not see improvements or the elimination of recurrent problems since then. As the system refines itself and HMRC identifies the weak points and variables in the system, I would have thought that HMRC would work to eliminate those anomalies. I do not see that happening—I see the same stuff again and again. You might think that it is still relatively early years, but I would have hoped that some of those anomalies would have been eliminated by now.

My overall view is that the system for establishing the outturns is now robust and is giving you reliable data. We have tried to explain in the report why the data will never be 100 per cent predictable or accurate. In answer to the convener’s earlier questions, I was explaining why there will always be a need for estimates in that process and, by definition, estimates will have a margin of error built into them. The question is whether the system is giving a meaningful outturn figure that can be relied on for budgeting purposes, as you said. In the past three years, we have seen a reduction in the variance of the outturn from the original estimate. We point out in the report that, in comparison with equivalent outturns, the level of accuracy has been reasonably high in recent years. The 2018-19 outturn—the year before the one that we are looking at—was 0.9 per cent lower than the estimated amount. The final outturn for 2019-20 is 1.1 per cent higher than the equivalent estimate. Therefore, over those two years, there is a variance of around 1 per cent in different directions. Given the inevitability of some level of estimation, I do not think that that is a surprising or unreasonable level of variation. Clearly, the challenge is to make that variation as small as possible, but I do not think that the system is giving figures that cannot be relied on. I think that the system is as good as the current approach can make it, bearing in mind the need for annual improvement to spot trends and deal with them.

Does that 1 per cent equate to approximately the same as the UK figure?

We do not have that calculation here, because it is not used for the same purpose in the UK budget, but I will ask Darren Stewart whether he has better data on that for the UK as a whole.

[Inaudible.]—publish the relative quality.

Yes—because the same system is being used by the same organisation, I would not expect to see any significant difference in the quality of the estimate for tax take for the UK as a whole, but Darren might want to add to that.

As Gareth Davies said, there is not an equivalent model or estimate in the context of UK tax revenues, but there are other estimates that are fundamental to preparation of the UK-wide trust statement, which records all the tax revenues for the UK as a whole. Among other things, HMRC is required to estimate the amount of revenue receivable at the end of the year. That is where activity has been undertaken but tax returns have not yet been submitted. I think that the figure of 1 per cent is certainly not inconsistent with that, but we would have to go away and look at that on an estimate-by-estimate basis. However, the figure of 1 per cent that we are talking about here is certainly not an outlier in that respect.

I will ask for a bit of clarification on one or two specific areas of the report.

As I said, anybody reading the report would have a concern at the level of estimations right the way through it. I know that you consider that the system that is in place is robust, but that is dependent on having a system to produce those estimates. It is very difficult to get a grip of the facts, so let me ask one or two questions.

In paragraph 6 on page 4 of the report, you talk about HMRC producing

“a provisional estimate of Scottish income tax revenue for that year.”

Is that in line with what you do for the UK?

09:30  

I come back to our previous answer, which is that we do not have an equivalent process for the UK in which we audit how close HMRC’s calculation of income tax comes to prior estimates, because we are auditing the entire tax take for the purposes of the HMRC accounts. Our work in that respect focuses on the accuracy of the figures being disclosed in each year’s annual accounts. We pay a lot of attention to the same issues, such as accurate identification of the income tax that is due, HMRC’s performance in collecting that tax and the level of debt that has been built up by taxpayers, but, as we have tried to explain, there is no process in the UK-wide accounts for estimating and comparing outturns that is equivalent to the process that we are discussing today. The same systems are being used consistently for UK tax, Scottish tax and now Welsh income tax.

With regard to paragraph 12 on page 8 of the report, I find it astonishing that taxpayers do not have to bother with advising a change of address. I realise that this is not within your powers, but how can the system possibly be robust if people can just change address willy-nilly and vanish?.

It is a feature of the UK-wide tax system. It would be up to Parliament, if it so wished, to legislate to make what you have referred to a requirement on taxpayers, but it has clearly chosen not to do so. Most taxpayers would probably regard it as in their interests to ensure that the information that goes to and comes from HMRC was accurate. However, as you have said, HMRC has to work around that. As a result, it has the now well-established address cleansing process that we describe in the report and which we have covered in previous years. The database of addresses is a key part of that process, and, as we have discussed, it throws up data quality issues every year.

Would making taxpayers responsible for updating their addresses with HMRC help? It might, but people who are determined to evade the system will still attempt to do so. However, it is not our role as auditors to recommend policy changes; we are here to observe how the system operates in practice and where the outstanding issues appear to lie.

In paragraph 15, you say:

“HMRC ... estimates Scotland’s share of net losses was £800 million ... based on a proportion of the UK figure, rather than ... Scotland-specific data”.

How exactly does it calculate that share of the net losses? Is it simply a percentage based on the volume of taxpayers that we have or is there some other esoteric formula?

I will bring in Darren Stewart to answer that question.

What you have described is broadly in line with our understanding of how the system works. As we have drawn out in the report—and as was signposted in written evidence that HMRC provided to a predecessor committee—the key thing to note is that, at present, there is no bottom-up assessment of compliance yield or tax at risk for Scotland. An assessment of tax at risk is made at the overall UK level through the strategic picture of risk and those elements that are relevant to Scotland—that is, the non-savings and non-dividend aspect of the income tax regime. The amount that is attributable to Scotland is, broadly speaking, calculated based on Scotland’s share of overall non-savings and non-dividend income compared with the UK as a whole.

But in your report you say that the dividend and savings elements are bound up with the income tax figure for the UK, which means that we are taking a proportion of that as part of our loss. That does not seem right.

It is probably worth clarifying that this is essentially income foregone; it is not tax collected, with an amount attributed through the budget as an outturn relevant to Scotland. In other words, this is not an adjustment made by HMRC to Scottish income tax take in calculating the outturn. It is, as I have said, income foregone.

How do you account for that?

From an overall UK and HMRC perspective, it is not accounted for in the trust statement and accounts. The statement includes income that has been assessed, the liabilities that have been established and the amounts that have been recovered and reported by HMRC.

So it is a notional figure. What is its purpose?

It is like the tax gap assessment that HMRC conducts at the overall UK level and which, based on 2019-20 data, was last assessed at around £35 billion. This is HMRC’s estimate of the impact of abuse and avoidance of the system on amounts that it could, in theory, have collected but which it has not been able to, because of non-compliance, evasion and things of that nature.

I could go on, convener, but I am conscious that other members have questions. Perhaps I can come in later, if there is time.

Absolutely. On that note, we will press on with Sharon Dowey, who has a number of questions.

Good morning. On compliance of Scottish income tax payers, one issue that has arisen during the scrutiny of previous NAO income tax reports is whether any evidence of behavioural effects has emerged as a result of Scottish income tax rates and bands diverging from those in the rest of the UK. In 2020-21, Scottish taxpayers paid more income tax than taxpayers south of the border on earnings of over £27,000, with a difference in excess of £1,500 on those earning over £50,000 per annum. The NAO report states:

“HMRC continues to assess as ‘low’ the risk of non-compliance as a result of divergence between Scottish income tax and the rest of the UK”

and

“HMRC has not identified any significant or widespread instances of taxpayers changing their address to obtain a tax advantage.”

What is the definition of “significant” or “widespread” in this instance?

Essentially it is the concept of materiality that we as auditors use—in other words, anything other than the odd case here or there. Darren Stewart will correct me on this if I am wrong, but I am not sure that we are aware of any cases of proven abuse through declaring a false address to avoid paying Scottish income tax. We can come back with a further update on that, but there has been no significant impact from any such behaviour on the tax take as set out in the outturn for Scotland.

In paragraph 2.36 of our report, we set out the three activities that HMRC undertakes to check and mitigate that risk and to pick up any examples that might occur, and we think that such an approach is reasonable, given the scale of the risk. However, we and the Auditor General for Scotland recommend that this area be kept under particularly close attention. Given that the divergence in tax rates is quite new, it might take some time for evasion behaviour to emerge in the system, and HMRC needs to continue to pay the issue a lot of attention.

So it is not causing concern just now, but you are keeping an eye on it.

My other question concerns HMRC. It has limited performance data available about its compliance activities in Scotland. Unlike its income tax system, which flags residents as Scottish, HMRC’s compliance system cannot readily identify people living in Scotland. Therefore, it cannot easily track and monitor compliance activity in Scotland, which affects its ability to collect performance data about the extent of Scottish non-compliance.

Why is there limited performance data and compliance activity in Scotland, and what, if anything, is being done to address that?

I will bring in Darren Stewart in a moment, as he will be able to add more detail, but the general point is that the data that identifies Scottish taxpayers for the purposes of the self-assessment and PAYE systems—the S prefix on tax codes—is not available for the compliance systems to use in a routine fashion, because of the design of those systems. There would have to be significant updating and replacement of compliance systems to enable the use, in an automatic way, of the data that identifies the country of the taxpayer. That is the central reason why HMRC is not able to do that at the moment. If that became a priority, that would be a good reason for making the necessary investment to bring those systems up to date.

I do not have much more to add to that. As Gareth Davies has outlined, the key point is that the use of the S prefix in the self-assessment and PAYE systems allows us a good opportunity to identify the Scottish taxpaying population, but the separate compliance systems that HMRC operates do not have an equivalent mechanism for identifying compliance activity undertaken in respect of Scottish taxpayers.

The written evidence that HMRC submitted to one of your predecessor committees came to the conclusion that, although it was possible to make the system work in that way, it would be incredibly resource intensive, and it was not something that it was minded to do at that time. The task was feasible and technically possible, but it was not practicable. However, as Gareth Davies says, that might change if there were a substantive need. I know that HMRC engages with the Scottish Government around the balance of compliance work and the activity that it undertakes, and, clearly, that work would come with a cost benefit consideration attached to it.

We are going to come on to questions around the cost of administering Scottish income tax shortly, and Willie Coffey will ask about the identification of Scottish S-code taxpayers. Before that, however, Craig Hoy will ask a series of questions.

Good morning. The report examines the impact of Covid-19 on HMRC compliance activities. It says:

“COVID-19 has continued to have an impact on HMRC’s compliance and debt management activities. Across the UK, there were 29% fewer civil compliance cases opened and 26% fewer cases closed in 2020-21 than in 2019-20”.

Can you give us a snapshot of why that is? Is it because internal processes in HMRC have been impacted by Covid—for example, due to people working from home, which means that there is less capacity—or is it because the outside world has become more complex because of the pandemic?

The answer to that has changed over the pandemic. Initially, a big operation was undertaken in order to equip HMRC staff to work from home. More importantly, there was a diversion of some HMRC staff resource to the furlough scheme set-up and delivery. If you remember how quickly that was stood up and delivered, you will understand that it required the diversion of people who were previously working on things such as tax compliance. That was understandable and was probably the correct prioritisation of resources in that situation.

As things developed and the numbers joining the furlough scheme reduced, people could be taken off that work. At that point, some of the compliance activity was restarted and the home working arrangements were bedding in.

Since then, staff have been able to return. In fact, in recent months, they have not just been working at full capacity in relation to compliance activity but have added to that, because of the need to catch up on the activity that had to be postponed during those phases of the pandemic. The resources of HMRC is, largely, the answer.

On tax collection, taxpayer behaviour changed, because taxpayers were given more time to pay. As we have already discussed, tax debts built up in many cases.

09:45  

People who struggled to pay tax during that period are now paying their present tax and have therefore almost forgotten about that period, which may impede the recovery of that tax. Do you have a concern that, in capacity terms, HMRC will be so busy in its forward-looking work that it may end up not fully delving into that period retrospectively?

HMRC recently gave detailed evidence on that to the Public Accounts Committee at Westminster, based on the report of ours that I mentioned in which we looked at the build-up of tax debt during the pandemic. I am sure that we can make that evidence available to the committee if you would find that helpful. In essence, it described a recovery plan for steadily bringing down the level of tax debt. HMRC has targets for doing that and can show how it deploys its resources to achieve those targets and how it is making steady progress.

As you would expect, the questioning of the committee was challenging in relation to the speed at which tax debt could be recovered, bearing in mind that the longer a debt stays outstanding, the harder it is to collect. That is work in progress. I think that HMRC would say that it is hitting its milestones for bringing the total debt down, but there is a long way to go.

Would you expect there to be an upscaling of those activities on compliance and debt management as we come out of the pandemic?

There has been already. The issue is about getting that balance between staying on top of current collections and tackling the backlog. At some point, there will be a question about whether, in essence, an amount has to be written off. However, we are not yet close to that point.

We talked about a separate analysis of Scottish compliance and how the effort and cost would perhaps be too great to do that. Do you have any insight as to whether there are plans to make available an analysis of Scottish income tax debt? Is such work being undertaken, and would the Scottish Government, HMRC or yourselves benefit from that?

Darren Stewart will correct me if I am wrong, but I do not think that we are aware of there being any work under way on that. That goes back to the prioritisation of HMRC’s resources and, as I understand it, HMRC is not currently prioritising that kind of analysis. Is that your understanding, Darren?

That is in line with my understanding. It is certainly a topic that has been discussed with our colleagues over at HMRC. There are benefits to disaggregating debt compliance and other data based on sectors and regions. However, although that would be important and valuable management information, I am not aware of any plans to do that for Scotland at the present time.

Mr Davies, in your opening remarks—or perhaps just after—you said that there was no reason to predict that the levels of income tax debt attributable to Scotland would be any different from those in the rest of the UK. How could you come to that conclusion if no substantive analysis has been conducted?

It was not an assertion that that would be an accurate measure; it was more about saying that nobody has put to us any reasons why there would be behaviour in Scotland that was different from that in the rest of the UK.

I will ask one quick question of Mr Boyle. In your report, you note that further analysis of taxpayer behaviour

“along with the relative success of compliance activity in Scotland and the Scotland-specific tax gap”

would help the Scottish Government to

“assess whether any Scottish income tax compliance risks are emerging”.

Do you have any concept of, and will you elaborate on, what those risks might be and their potential scale?

What we set out in our report is along the lines of the conversation that we have had this morning. As the tax policy choices of the respective Governments diverge, it might inform some of the behavioural choices that taxpayers make as they identify themselves through their employer or self-identify as Scottish taxpayers. That is captured in the NAO’s report, which illustrates, by bands, the percentage differentials of different incomes that taxpayers might have.

If the difference in the tax choices of the respective Parliaments increases, there is further potential risk of non-compliance and non-identification, and, by extension and in due course, there is a potential threat to the Scottish budget.

As we set out in our report, it is important that that is kept under review. There is an opportunity for non-compliance and it is one of the risks that HMRC will need to manage, along with its interaction with the Scottish Government.

The risk might be someone buying a bolthole in Berwick-upon-Tweed and registering themselves there while working in Edinburgh, for example.

That is the classic example. However, through our own work, which supports the NAO’s conclusions, we know that it is not currently a material risk, although the potential for it remains, and it might grow if the two Parliaments continue to make different choices about tax bands and how they are applied.

Can you just confirm that you have no evidence of flight of income tax payers from Scotland to England, Auditor General?

That might be more of a question for the C and AG, convener, but, in our own work, we have not identified a specific pattern of behaviour. Rather, it is a growing risk.

If I may, I will refer back to Mr Beattie’s question about the increasing number of missing postcodes. Although we agree that that is not necessarily a significant risk, there is a point about equity that we touch on in our report. Scotland-based taxpayers ought to be identified and be paying into and contributing to the Scottish budget. We also note that there is perhaps the opportunity for HMRC to target a low number of missing Scottish postcodes. Those are part of our findings on this work.

Thank you. I now turn to Willie Coffey, who is joining us virtually. Willie has a number of questions that he wants to put. Over to you, Willie.

We know that 54 per cent of Scottish taxpayers pay less income tax than they would if they lived in other parts of the UK. Nevertheless, identifying Scottish taxpayers is an issue for us.

We note that, in 2020, HMRC identified more than 30,000 records of taxpayers who were not registered as Scottish with S codes. In March 2021, that figure rose to 39,000. Why is that happening? A simple calculation might show that that represents £1 billion in earnings being incorrectly taxed. Give us a little flavour of why you think that those numbers are going up so significantly, and what is being done to reduce them?

I think that you are referring to the administration of the tax codes with the S prefix. In June 2019, there were 67,500 cases; by April 2020, the figure was down to 31,000. HMRC identified 31,400 taxpayer records—about 1.3 per cent—where employers were not operating a tax code with an S prefix. The most recent data that we have shows that, by March 2021, the figure was around 39,000 cases. It came down from the higher level of 2019 but, as you say, it is still a worryingly significant number that needs to be addressed.

I go back to the comments that I made when we talked about not tax codes but postcodes, because the same general point applies. This is an issue of data quality in a large system, and so it requires permanent vigilance. It requires the quality of the data to be tested to ensure that, where there should be an S prefix, there is one; identifying any exceptions; trying to understand why they are occurring—and, if that shows that there is a weakness in the system, correcting that weakness. That is a continuous process for the good management of any large data system such as this.

Our approach will be, first, to monitor performance to see whether the number is going up or down; secondly, to find out from HMRC whether it is identifying any systemic reasons for the level of error in the system; and, thirdly, to find out what HMRC is doing about those issues if it has identified them. That is our role in the continuous vigilance process—our role is to ask what is being done to ensure that the data quality is as high as possible. We recognise that there will always be an error rate; the challenge is to make that error rate as small as possible.

Is there any estimate of the loss of revenue as a result of the issue? We read in our briefing notes that some employers might repeatedly be failing to apply the S code correctly, but we do not have any information on that. Do you have any? Are there employers who are repeatedly not applying the code correctly? Fundamentally, it is against the law not to apply the code correctly. What are we doing to encourage, let us say, employers to apply the code correctly and legally?

I do not think that we have a specific estimate on that element. The issue forms part of the overall figure for leakage from the system, which my colleague Mr Stewart described, but we do not have a specific estimate for that particular element.

However, you are right that there is an issue with action being taken in relation to persistent offenders, or employers that are persistently found to not be applying the code accurately or to not be applying it at all. We have not been able to get a clear answer on the approach that is being taken to that small number of employers. However, we will follow that up with HMRC in our work on next year’s figures.

You will probably recall that, in the early days of the system, even some members of the Scottish Parliament were not coded as Scottish taxpayers—I think that that applied to 45 out of 129 MSPs, which is a huge number to get wrong. Can you give us an assurance that that problem has now been corrected completely and that the 129 of us and our 59 Scottish MP colleagues are being correctly coded as Scottish taxpayers?

We believe that that embarrassing position has been corrected. Obviously, we have paid attention to the issue, having highlighted it in a report several years ago. Some focused action was taken. In fact, HMRC applied a manual check to exactly the group of people you mention to avoid the issue recurring. Our finding so far is that the exercise has been successful and HMRC has maintained that level of accuracy since then. However, because of the significance of the issue at the time, it remains on HMRC’s radar.

I am glad to say that I was not one of those 45 MSPs; I was correctly identified initially.

As we have you in front of us, Mr Davies, will you say something about the audit function that might apply to the UK Government’s shared prosperity fund when it comes in? As you probably know, the European Union structural funds were completely within the scope and under the gaze of Audit Scotland and the Scottish Parliament, but we do not have any details on the follow-up arrangements that will apply for Scotland with respect to the levelling up fund or shared prosperity fund—whatever it is called.

Do you have any information from an audit perspective on where the audit function will rest? We think that the value to Scotland and the Scottish Parliament was more than £1 billion over seven years, but we have no idea whether the amount will be the same with the new arrangements. However, I am asking you specifically about the audit function rather than the politics. Can you share any information with the committee on that?

I cannot yet share any detail about how that will operate. Obviously, we will discuss that with our colleagues in Audit Scotland.

The Auditor General and I work closely together on all the matters where we have functions that either abut or overlap, so I am confident that, between us, we will be able to explain clearly the audit arrangements that are proposed for each of those funds. We are just not clear about the arrangements yet because the Government in Westminster has not got to that level of detail.

10:00  

I agree that having clear accountability to the relevant Parliament is really important, but rest assured that the two audit offices work closely together to input our views on what is practical and feasible from an audit perspective. Clearly, the policy questions are not for us, but the practical administration of an effective audit regime is, so we will be inputting our thoughts on that as we get to that point.

Do you have any indication on when we might know about the arrangements? The committee has been asking that question over recent weeks. Is there any indication as to when we will get some clarity?

Both the funds are moving through the legislative process and answers will be required to that question about the audit arrangements. However, I cannot give you a useful date at the moment, I am afraid.

Okay. Many thanks for those responses.

We have a little bit of time in hand before we wind up, and Colin Beattie wants to come back in with a couple more questions on the cost of administering the Scottish income tax system.

I refer back to our previous discussion. There are four bullet points in paragraph 1.23 on page 17 of the report that I find very telling in terms of the robustness of the figures around Scottish income tax. To me, they clearly indicate that there is a real problem in calculating the figures.

We talked about the notional £800 million figure. Bullet point 3 says that the figures that are being used

“do not exclude tax from savings and dividend income”.

You cannot possibly get something accurate out of those figures, because that is not an area within the Scottish Government’s tax authority.

Bullet point 2 refers to the “differing proportions” of types of taxpayers north and south of the border. The figures are completely distorted, because London, for example, is massively overrepresented in terms of the top income tax payers. We do not have that situation in Scotland. If we are using a methodology to calculate the figures that does not account for that incredible difference, how can the figures possibly be accurate?

That is a methodology question at heart, and it is a question that will need to continue to be addressed through the governance arrangements between HMRC and the Scottish Government. You are making perfectly valid points about the limitations that those factors bring to the process. They are all agreed elements of the methodology at the moment, so it is not as though what is being done is not known or visible to both parties—the Scottish Government and HMRC. The committee might like to take evidence from the Scottish Government and HMRC on the question.

Are those factors under consideration in discussions about future methodology improvements? What would be required to address the issues? Again, as we have said on a number of points, it would be for HMRC, and not me, to answer that but I am sure that HMRC would say that it is a question of cost benefit analysis: what would be the extent of the improvement if we could disaggregate data in that way, and what would be the cost of being able to do that?

However, you are right—the issue should be continually scrutinised because the process needs to improve over time. If there is a cost-effective way of eliminating some of those essentially rough elements of the estimation process, that should be considered under the governance arrangements that are in place between the two parties.

I turn to one or two specific issues. Paragraph 1.2 on page 10 says:

“HMRC calculates the final outturn figure from several components”,

but it does not actually mention what those components are. I do not know whether explaining them will require a lengthy response. If it requires just a short one, that will be fine, but if not, you can perhaps drop us a note.

Is the response to that question straightforward, Darren, or should we write to the committee with a bit more detail?

We could follow it up in writing, but in the first instance, I would point to figure 3 on page 12 of the report, which signposts the different elements that make up the calculation, including extractions from self-assessment and PAYE systems and other estimates and adjustments that are made to arrive at the outturn.

We would be happy to write to you, expanding on the figure.

That would be excellent.

Paragraph 1.5 on page 11 says:

“In some areas of the calculation, data are not available in sufficient detail to identify income tax liabilities, reliefs or other adjustments relating to individual taxpayers.”

Again, that will require making a huge estimate.

In financial terms, it is not significant compared with the overall estimate. It relates to the small number of people on whom HMRC essentially has no access to data, because of various sensitivities, but it is not a material financial element of the estimate.

But it comes back to the theme of estimates and assumptions that runs through all of this. For example, paragraph 1.11 on page 13 says:

“HMRC deducts an estimate of the Scottish share of tax reliefs given against PAYE liabilities.”

How does it reach that figure? What is it based on? Is it based on some of the flawed data that you have highlighted in paragraph 1.23?

Again, HMRC will be able to set out the detail of that, but I think that it falls into the same category as other issues that we have been discussing. In other words, to improve the situation would require having more detailed information than HMRC currently holds in its systems. It is a trade-off between the cost of collecting that more detailed information and the benefit of having a more accurate estimate.

I also note paragraph 1.16 on page 14, which says:

“HMRC calculated both deductions by estimating the Scottish share of each tax relief claimed across the UK using historical data.”

I do not know what those “historical data” are.

It means data from previous years rather than from the year that we are discussing. However, on this and the other points that have been raised, the committee might like to take evidence directly from HMRC, as these are clearly operational issues for it. Our role as auditor is to point out where estimates have been made and the limitations on the accuracy of the data, and our overall conclusion is that this is a robust—

I accept the point, but I am trying to highlight issues with regard to uncertainties around the figures that are being produced.

Flicking through your report, which was an excellent bedtime read, I note that in paragraph 2.13 on page 21 you say:

“HMRC confirmed the residency status for 92% of Scottish and Welsh pension scheme members ... and provisionally estimates that 94% have been confirmed following the 2020-21 tax year.”

That percentage does not seem high—in fact, it seems quite low. You would think that pension scheme members would be easy to pick up.

I do not think that that is out of line with the UK-wide figures. Of course, tracing pensioners is a big exercise for pension schemes, because, apart from anything else, they need to know that pensioners are still alive in order to keep paying their pensions. This is an area where I think there could be further developments in data matching between different parts of Government with an interest in accurately locating pensioners.

Again, this is another example of a situation in which improvements might be available as data becomes easier to match, but the question is, how much resource would that take, and what would the financial benefit of doing that be?

It is known that, in Scotland, a higher proportion of people give to charity than is the case in the rest of the UK. On that point, I note that paragraph 2.14 on page 22 says:

“HMRC does not have systems in place to establish the residency of taxpayers donating to charity”.

All those anomalies add up. You might say that each one has a small impact but, when you look at them in aggregate, it starts to become a concern.

The agenda for discussions between the Scottish Government and HMRC about the governance arrangements and how the methodology could be improved over time is a long one, and I think that you are pointing out perfectly valid items for that agenda.

I could go on, but I am conscious of time.

Thank you.

I would like to capture the essence of what we have been saying today. I say to Gareth Davies and Darren Stewart that our approach today has not been pointed at you; we have taken that approach because we are trying to understand where there are gaps and where improvements could be made.

We reflect that this is year 4 or 5 of the distinctive Scottish income tax system, and yet, even at this stage, we do not have Scotland-specific tax gap data—that is, data about the amount that should be paid versus what is actually paid. Further, we do not have Scotland-specific data on compliance and non-compliance or on income tax debt, all of which we have been probing this morning. You are right to say that the reasons relate to the service-level agreement that is in place between the Scottish Government and HMRC, and we need to consider whether it would be useful for us to explore that avenue further.

In previous years—and, to a limited extent, today as well—we have been told that the issue is all about HMRC’s limited resources. If the service-level agreement were adjusted, would it be possible to get that data disaggregated to a Scottish level, so that it could be used to inform the decisions that this Parliament needs to make around income tax rates as well as to give us an understanding of the revenues that would be generated by the rates that are set?

I hope that you do not think that I am dodging the question when I say that I think that only HMRC can give you a useful answer, because it understands what would be involved in improving the various elements of the data that we have been discussing. It could explain the resource implications of various approaches, but I could not give you a useful answer to your question.

As an auditor, I am keen on accurate figures, so I would say that the more that some of the estimates can be reduced in variability, the better the database for arriving at figures, which is better from an audit point of view. However, clearly, you need HMRC to describe what would be required in order to deliver the improvements that you have been discussing.

Just to be clear, the report says that the £0.7 million administration fee is

“accurate and fair in the context of the agreement between HMRC and the Scottish Government.”

Could you tell us what you mean by “fair”?

Our starting point for that is the service-level agreement that has been negotiated between the Scottish Government and HMRC. We have examined the information that HMRC has given us for the costs of implementing that service-level agreement. We can see how it has attributed staff costs, information technology costs and so on to carry out the work that is required under the SLA. We think that that calculation is supported by the data that we have seen and that it is a fair estimate of the time and the other costs that are involved in operating the agreement. Clearly, if you had a more highly specified agreement that got into some of the areas that we have been talking about today, it is likely that HMRC would require more costs to be covered. However, on the basis of the agreement that is currently set out, we think that the cost is fairly stated.

On that note, we will conclude the session. I thank Stephen Boyle, the Auditor General, who joined us in person at our meeting today, and Mark Taylor, who was on hand to answer any questions that we had for him—in the event, I do not think that he was called on, but I thank him anyway.

I give particular thanks to Gareth Davies and Darren Stewart from the NAO. As I said at the beginning, we hope to see them in person at some point in the not-too-distant future, because that would help us in examining the important role that they play in keeping account of how HMRC is functioning in this area, which is of particular concern and importance to us.

10:15 Meeting continued in private until 11:00.