Public Audit Committee [Draft]

Meeting date: Thursday, February 22, 2024

Official Report 682KB pdf

---

Administration of Scottish Income Tax 2022-23

The fourth item on our agenda is consideration by the committee of reports prepared by the National Audit Office and the Auditor General for Scotland on the administration of Scottish income tax for the tax year 2022-23.

I welcome our witnesses: Stephen Boyle, the Auditor General for Scotland, who is joined by Mark Taylor, who is an audit director at Audit Scotland. I am very pleased to welcome from the National Audit Office Gareth Davies, the Comptroller and Auditor General, and Rebecca Mavin, who is a senior analyst.

We have quite a number of questions to put to you. I will ask Mr Davies to give us a statement but, before that, I invite the Auditor General to address the committee.

Many thanks, convener, and good morning, committee.

Scottish income tax remains a key part of the package of financial powers implemented as a result of the Scotland Act 2012 and the Scotland Act 2016. The purpose of today’s evidence session is to look at the administration of Scottish income tax up to 2022-23. It was the sixth year in which the full amount of non-savings and non-dividend income tax collected in Scotland by His Majesty’s Revenue and Customs was payable to the Scottish Government. It was the fifth year for which HMRC published Scottish income tax outcomes in its accounts. Outturn figures are for the year 2021-22.

The differences between the actual United Kingdom and Scottish tax outturns and the amounts that were previously forecast will now be adjusted through the 2024-25 Scottish budget. That is known as a budget reconciliation. There will be a negative reconciliation of £390 million to the 2024-25 budget. It is the largest budget adjustment to date, which is attributed to the unexpected extension of the furlough scheme and relatively weaker economic performance in Scotland compared with the rest of the UK following our emergence from the pandemic.

HMRC’s annual accounts include an estimate of Scottish income tax for the 2022-23 year. HMRC collects and administers Scottish income tax as part of the UK’s overall 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 of Scottish income tax. I report to the committee to provide additional assurance on the NAO’s audit work, in line with the recommendation of the predecessor Public Audit Committee in 2014. I also explain what the findings mean for the Scottish budget.

I am satisfied that the NAO’s broad audit approach was reasonable and covered the key audit risks. I am also satisfied that the findings and conclusions of the C and AG’s report are reasonably based.

The C and AG has concluded that the outturn of Scottish income tax was fairly stated, which provides the Scottish Parliament with valuable assurance of that aspect of the Scottish budget.

Thank you, convener, and good morning to the committee. As in previous years, the content of my report follows the requirements that are set out in the legislation and covers three things: first, the outturn for 2021-22 and HMRC’s estimate of Scottish income tax revenue for 2022-23; secondly, the rules and procedures that are in place to administer the system; and, thirdly, the costs that are recharged by HMRC to the Scottish Government under the service level agreement between the two.

On the outturn and estimate, the methodologies remain broadly the same as last year’s, and I have concluded that both are reasonable.

External factors such as high interest rates and inflation, wage growth and high and volatile energy prices have created additional uncertainty in HMRC’s estimate of tax revenues for 2022-23. Nonetheless, I consider that the approach that HMRC has adopted in reaching the estimate is reasonable.

HMRC’s compliance activity—its activity to detect underpayment of tax in the UK as a whole—began to recover in 2022-23 after being stopped for a period during the pandemic. However, the activity level remains lower than before the pandemic. Across the UK, the number of compliance cases that HMRC opened in 2022-23 was higher than in the year before but still 17 per cent lower than in 2019-20, which was the last financial year before the pandemic started.

The total tax debt—tax that is due from taxpayers but not yet paid—was £43.9 billion in March 2023 for the UK as a whole, which was £4.7 billion higher than the year before. It remains at a level that is roughly three times the average annual figure from before the pandemic. Therefore, tax debt is a significantly bigger problem for the UK as a whole than it was before the pandemic.

HMRC continues to assess the risk of non-compliance as a result of divergence between the Scottish income tax system and the rest of the UK in terms of the marginal rates that are charged. We continue to assess the risk as low, but there is a risk that that position might not be sustainable in the longer term as further divergence of tax policy occurs.

My team and I work closely with the Auditor General for Scotland and colleagues at Audit Scotland throughout our audit, and I am grateful for their co-operation and contribution to the work. I look forward to answering your questions.

Thank you very much, indeed. The committee will be particularly interested in looking at the methodological approaches and the implications of greater divergence between the income tax systems of Scotland and the rest of the UK. Perhaps I could begin by putting an important question to you.

On the one hand, you say that HMRC estimates are reasonable, but, in paragraph 1.21 of the NAO report, you identify areas of methodology in which there is a degree of what we could call woolliness. You say:

“HMRC does not fully understand the causes of the over-estimate in 2021-22 and any socioeconomic factors contributing to the over-estimate may be different in 2022-23.”

You say that there is a big reliance on sample data in the revenue estimates and that pay as you earn and self-assessment amounts that are apportioned to Scotland do

“not reflect the differing proportions of each type of taxpayer between Scotland and the rest of the UK.”

I wrote to you last year about that, and I might return to it.

You say that PAYE assessments include areas that are not subject to Scottish income tax variation, such as dividend payments, savings interest and so on, and you note that assumptions are the basis of HMRC’s estimates of PAYE liabilities. Those are all holes in the methodology that you have identified. How does that provide you with the comfort to be able to say that what we have before us is reasonable?

As we say in the report, these are not new issues for this year; they have been features of the system since it was designed. Although there have been marginal improvements, and new challenges arise each year, we have made these points in previous similar reports. It is not a deterioration; it is a continuation. However, as I said in my opening comments, the environment for 2022-23 was unusually volatile. In that year, we dealt with energy price spikes and all sorts of unusual economic activity. The impact of those uncertainties could be higher in 2022-23 than in previous years, even though the list is the same as in previous years. That is an important point.

Our conclusion that the methodology is reasonable is not the same as saying that we are giving assurance on the number itself. That is an important distinction in audit terms.

Clearly, this is an estimate for a particular purpose. It is not an accounting statement with all the technical requirements to be materially accurate. One thing that we can be sure of is that the estimate will have an error in it somewhere, one way or the other. The question is whether it is useful enough for the purpose that it was designed for. Our conclusion is that you can take the same level of assurance on it as on estimates in previous years while being aware that volatility in the economy in 2022-23 means that the estimation error is likely to be higher.

We have made the point in previous years that it is important for parliamentarians to understand what is going on out there, because, in the end, that guides policy decisions and decisions around the Scottish budget and where income tax is set. That is why we are especially interested. I would be interested to hear from you about what you would do differently if you were in the shoes of HMRC.

The challenge is that most of the improvements that you could attempt to make to the estimate would come with a cost. It would involve the collection of more granular data, the implementation of new systems and more staff activity to determine higher levels of testing of things such as the accuracy of the tax base. Although it would be possible to construct quite a detailed action plan for making improvements in those areas, they would all come with a cost.

That is why, each year, my answers to those questions are that it is a matter for the Scottish Government and HMRC, because only the Scottish Government can quantify the value of such improvements compared with alternative uses for the money. That has to be a policy decision for the Government rather than for an auditor. It is possible to set out where technical improvements might be made, but they would all come with additional costs.

I will ask the Auditor General for Scotland about this. When the Scottish Government was in front of the committee last year, we heard that the service level agreement between the Scottish Government and HMRC was being renegotiated. We had been asking the Scottish Government to look at how the renegotiation could include discussion about releasing more of the data that we think should be available, some of which is fairly basic information about the tax gap and the extent to which fiscal drag is having an impact on tax returns. We were led to believe that that was within the scope of those renegotiations.

However, the current service level agreement looks broadly similar to the previous one. Looking at a fairly useful indicator, which is Mr Davies’s point, the fee paid by the Scottish Government to HMRC—which, let us bear in mind, is to support collecting data on £13 billion to £14 billion in tax—is £600,000. Maybe I am looking at this from the wrong end of the telescope, but if it is a question of paying a bit more to get better-quality data, I think that the view of this committee would be that it would be money well spent.

At the heart of the judgment of both the NAO and Audit Scotland is the increase in compliance risk. As we project into future years, it is already building on circumstances that the committee has seen in previous reports plus what the recent developments in the divergence of income tax rates between Scotland and the rest of the UK might mean for taxpayer behaviour. There is all the more reason, therefore, for the Scottish Government to make a judgment about what additional comfort and assurance it wants from the estimates, given how central they are to its spending ambitions as set out in the budget.

09:15  

The pandemic undoubtedly had an impact on the volatility of the estimates. As I said in my opening remarks, £390 million is the highest budget reconciliation to date. However, convener, you make an important point that, relative to £13.9 billion, the £600,000 service level fee does not feel entirely commensurate with what is at stake.

There is a limit to what more I can say about that. In the past 12 months, we have seen progress being made on the availability of some data sets and I contend that that is a step in the right direction, but there is more to do in terms of transparency and how the Scottish Government and HMRC keep under close review the extent to which they can rely upon these estimates in future as they see the future data outturns.

Mark Taylor might wish to say more about the data sets and what comes next.

The provisional outturn, the estimate itself, sits alongside the information that the Scottish Fiscal Commission provides in its regularly updated forecasts. That information is prepared differently, looking more at the granular data and real-time information that is available from the tax system. We have always been clear that the estimate is based on a share of the overall UK position. As more data about what happens in Scotland becomes available, there will be opportunities for a different approach. As others have said, alongside the work that the Fiscal Commission does, there could be a cost-benefit judgment about whether the investment in and different approach to this estimate is good value. I think that we would all recognise the reality that the estimate that is prepared here is put to one side and that Government and Parliament focus on the SFC’s forecasts and updates. We think a bit more value could be brought out of this estimate, given its HMRC genesis.

The sense I had from the evidence that we took last year from HMRC and the Scottish Government was that maybe the data that we are looking for is not available and that, no matter what the price, it could not be extracted from the system and disaggregated in the way that we think it ought to be. Do you have a view on that? I will ask Mr Davies first and then come to Mr Boyle.

Certainly, some existing data is not being separately analysed, but it could be.

Could you give us an example of that?

I mentioned the tax debt information in my introduction. Clearly, it is possible to attribute that tax debt to individual taxpayers and to know which tax code applies to those taxpayers. That is one example. It is currently done on the apportionment basis that was just described but, with more resource devoted to it, it could be analysed more accurately. Some systems issues make that less simple than I made it sound, but those issues are not insoluble.

I do not have much to add to that, convener. This is a live issue. To give the Scottish Government some credit, we have seen some progress in the past 12 months. It is a cost-benefit decision but one that should be informed by the increasing risk in this environment.

It is also fair to acknowledge the Scottish Government’s increasing transparency around some of the governance, for instance through the publication of minutes, which had been a long-standing issue with the Scottish Income Tax Board. It is for that board and the Scottish Government more widely to take a view about where to go next with their ask of HMRC by way of compliance.

If I may, I want to draw your attention to paragraph 1.13 of our report, which is titled “Estimated further liabilities”, because it is an example of where improvements in systems have significantly reduced the inaccuracy of the estimation. It is useful to have a concrete example of what we are talking about. This is a fairly obscure bit of the outturn calculation, but there are PAYE liabilities that are not straightforward to calculate. They are often determined just once a year when employers have agreements with HMRC on particular employee expenses and so on. As that paragraph says, there used to be an estimate uncertainty of £10 million a year in this part of the figure, but that has now been reduced to almost nothing because of an improvement in the system. At first, you might have assumed that figure of £2,731 was a typo but it is actually the remaining uncertainty. It is a very good example of something that is not material to the overall estimate but of where an improvement in the system has allowed an area of uncertainty to be essentially eliminated.

Thank you. That is helpful. I now turn to Jamie Greene, at his first meeting of this committee.

I have been thrown in at the deep end. Thank you, convener, and good morning to our panel.

Before I move on to my line of questioning, I might just conclude the previous one. We have had a bit of discussion about the variance in outturns versus forecasts. That flags a concern to me with the 2022-23 estimation from HMRC of £14.9 billion and what variance we might see in that outturn. Surely there can only be one data set at the core of all this. No matter who you speak to—the Fiscal Commission, analysts, policy makers or HMRC—I presume that they are not using different data sources, so a robust set of data that can help to produce the forecasts must sit at their core. Given the huge variance in outturns this year, can we be confident that the data set and the analysis of it are sufficiently robust for policy makers and analysts to produce the forecasts that affect decisions about tax rates, for example?

It is the same data sets that are being analysed here but, crucially, they are at different stages of maturity. It might seem like old news to be talking about the outturn for 2021-22, given that we are in February 2024, but the tax timetable means that self-assessment tax returns have to be in by the January after the end of the tax year and they then have to be processed and calculated. It is only in late 2023 that the accurate picture for 2021-22 can be calculated.

As we say in the report, you can be confident about the outturn figure because it is based on tax actually calculated for real taxpayers. It is based on their tax due. Therefore, 99.3 per cent of the figure is actual rather than estimated. It is only at that point that you can be confident in the accuracy of the number. The estimate for the following year is made before HMRC has received most taxpayers’ tax returns, and certainly before HMRC has tested whether the returns are accurate and calculated the tax due.

The difference in the maturity of those figures is at the heart of the variance. As we discussed in our first conversation, how much more accurate can that second estimate—in this case, the estimate for 2022-23—be? At the moment, it is largely based on an apportionment of estimated liabilities for the whole of the UK. While HMRC goes through checks and procedures to make sure that as much of the estimation uncertainty as possible is eliminated, we are still left with a fairly significant amount. Furthermore, that position is even more uncertain when the economy is as volatile as it was in that year. How much more cost and effort do you want to put into improving the accuracy of that more recent estimate, knowing that it will be another year before you get the actual accurate figure?

I will perhaps come on to that volatility. Your opening comments—repeated in the opening paragraphs of your report—gave quite a broad-brush overview in using phrases such as “continuing economic uncertainty”, “tax policy divergence”, “weaker economic performance”, and so on. Those are quite generic terms. What specific work has Audit Scotland done on the root causes of why that figure of £390 million was so vast?

I will start and then I will bring in Mark Taylor. We are not challenging the reasonableness of the methodologies. For all the reasons that Gareth Davies sets out in his paper, and based on our own consideration of the Fiscal Commission’s report, there are different sources and time differences and, layered on top of that, economic volatility, because some of the forecasts and estimates had to be made as we were in the midst of and emerging from the pandemic. At paragraphs 64 and 65, and in exhibit 2 in our report, we set out some of the circumstances that were specific to the scale of the budget reconciliation that was to be deducted from the 2024-25 Scottish budget of £390 million.

Factors that I mentioned in my opening remarks are relevant here. New circumstances came about after the Fiscal Commission’s forecasts were made. The furlough scheme was extended. I also mentioned the extent to which economic performance in different parts of the UK rebounded after the pandemic.

As the C and AG sets out in his report, tax revenue in 2021-22 increased significantly compared with what it was the previous year, as we would have expected. I know that the committee is aware that the requirements of the fiscal framework as they relate to the spending power of the Scottish budget are based on the relative performance of the Scottish economy compared with other parts of the UK. While both rebounded significantly, it is important to note that Scottish income tax rates rose by 14.9 per cent compared with 15.4 per cent in other parts of the UK. That was an element of success, but it impacted on the overall scale of spending provision through the Scottish budget.

Those are the high-level factors. Mark Taylor might want to elaborate on where our own work has taken us and where we might go next in our activity on economic performance.

Specifically and in comparison, is it possible to quantify that underperformance in numerical terms?

The starting point from our perspective is that we are not the economists in the room. The Scottish Fiscal Commission is the economist in the room and a lot of the detailed work that it does in analysing its own forecasts and what has happened through time is publicly reported. It is at the heart of the commission’s work, and it engages with Parliament’s Finance and Public Administration Committee on that.

We draw a lot of our understanding of what is happening in the economy from the commission’s work and we use it to understand what it means for public finances and the risks for public finances. As the C and AG pointed out in his introductory remarks and earlier answers, at the heart of what we do, high uncertainty and volatility are baked into the system, and that is, in turn, a consequence of the system that was set up and agreed.

Underpinning everything else, budgets need to be set four years in advance of when the tax is collected. That is the heart of the system and our work over a number of years past has been, and continues to be, focused on how government in Scotland responds to and manages the risk. It will continue to be the focus of our work.

As the C and AG says, post the pandemic and in a high-inflation environment, the Fiscal Commission has been clear that the levels of uncertainty and risk have been higher. The SFC has been quite positive in its outlook for the tax position going forward, but it also makes the point that risk is mostly on the downside and it thinks that that will be the position. Given that the factors that sit behind that position are earnings and levels of inflation and how they impact the various systems if those expectations do not materialise, there is a bit of a downside risk going forward. We continue to stay alert to that in our engagement with and audit of Scottish Government and how it responds.

09:30  

I suspect that one of the biggest areas of interest, and perhaps concern, on the part not just of the committee but of the Government relates to the effect of tax divergence—Scotland having differential rates compared with the rest of the UK.

You state that, broadly, historically speaking, there has been little divergence, but that it has increased for this tax year and that it might increase in future tax years. The known unknown is what effect that will have on people’s behaviour and how it will affect the tax base and the resulting tax take.

What work has Audit Scotland or the NAO done on understanding how any such behavioural impact will affect the final outturn numbers? What are the risks? How high or low are they?

The section of our report from paragraph 2.16 onwards covers the issue that you have raised. HMRC is steadily building its work in this area to understand any emerging evidence of taxpayer behaviour on the question of whether the divergence that already exists is driving different behaviour. For example, are there higher levels of non-compliance as a result of people wrongly declaring an address on the other side of the border? The conclusion so far is that there is very limited, if any, evidence of that.

However, the point that we are making is that, in the year that we are looking at here—2022-23—the level of divergence was fairly low. Obviously, it has been increasing steadily, in 2023-24 and in 2024-25. Therefore, as an expression of audit risk, in our view, that area is one that requires increasing attention from HMRC.

In response, HMRC is building a couple of databases, one of which involves a longitudinal survey of taxpayer behaviour over time. It is building up several years of data on cross-border movements—it is recording all cross-border tax movements in both directions and looking at any trends in that data over several years to see whether particular changes in tax policy lead to any changes in behaviour. I think that it is sensible to study that data over time, and it is good that HMRC has been collecting it for a while in anticipation of such effects. It is too early for that data to shed any light on the question at this point, but I think that having it in place will be very important for the future. That is one issue to watch and to seek evidence on from HMRC.

There is also HMRC’s regular cleansing of its data on the system and its regular matching of data across different systems, although that is becoming increasingly difficult because of stronger data protection rules. The amount of data matching that HMRC can do is reducing over time, but it is still an important component. HMRC also looks routinely for missing or inaccurate postcodes.

In addition to the standard controls that you would expect in such a system, there is a process of standing back to look at long-term trends in data. Together, those two things are designed to shed more light on the issue that you raised.

We support the judgment of the C and AG and the NAO about the increase in risk to compliance and forecasting from the divergence of tax policy, and the need for the Scottish Government to have a clear view on what activities it wants HMRC to undertake so that it can be satisfied that it understands the scale of the risk and how it is being managed on its behalf. That goes back to the earlier discussion with the convener about the extent of the activity that is being undertaken on compliance so that the Scottish Government can have as reasonable forecasting as possible to support future Scottish budgets.

I draw the committee’s attention to paragraph 53 of our report, which sets out a range of possible behavioural responses by taxpayers to the divergence of tax rates. The committee will be familiar with some of those—we have explored some of them already.

I also highlight paragraph 56 of our report, which refers to the data set that the C and AG referred to, which is tracking how taxpayers in Scotland are responding to the divergence of tax rates over time. That does not matter only to HMRC and the Scottish Government. There is an important point here about transparency and the need for forecasters, auditors, policy makers and scrutiny bodies such as this committee and Parliament to have a clearer picture of taxpayer behaviour when they set budgets. They will be able to do that if such information is more readily available to support income tax decisions in Parliament than it has been to date. There are signs that progress is being made in the publication of such data, but there is still more work to do and greater scope for transparency in future.

We know what some of the risks are, as they are well documented and are talked about often. However, quantifying those risks will be difficult until we see what people start doing with their money. Let us take, for example, the scenario of a person who earns £50,000. Such a person would pay 20 per cent more tax in Scotland than they would do elsewhere in the UK, which is a substantial difference. The number of people who are affected in that way must be easy to find out, but it might be more difficult to forecast what they might do. I guess that that is part of the problem. One presumes that, if there is a huge shift in behaviour, that will affect future budgets and how much money is available to the Government. Is that fair to say? Is that a risk in how we forecast budgets?

Those are the factors that the Scottish Government needs to take into account to satisfy itself. It depends on the extent to which the Scottish Government has good-quality, robust data to base policy decisions on, and data that is as widely publicly shared as possible. The C and AG and Audit Scotland are satisfied that the risk that taxpayers’ behaviour will change as a consequence of divergence is increasing. The Scottish Government needs to be on top of that risk as it relates to its forecasting, given how central the forecasts will be to Parliament’s budget setting in future.

To conclude that part of our questioning, are you saying that there is no figure or estimate for tax evasion or avoidance in Scotland?

HMRC produces a UK-wide tax gap estimate, which includes the things that we have mentioned. Essentially, it reflects the difference between the amount that would be brought in if everyone paid exactly what they should under the law as it is set out and what is actually brought in. That estimate is of the order of £30 billion to £40 billion for the UK as a whole. The only estimate that is available for Scotland is an apportionment of HMRC’s total figure. We are talking about one of those bits of data that are not separately calculated for Scotland but are apportioned from the UK figure.

Thank you. That is helpful.

I will move things along by bringing in Graham Simpson, who joins us remotely.

I want to follow up on a couple of things.

Mr Davies has just mentioned the UK tax gap, which I think he previously said was £43.9 billion. We do not yet have a breakdown of where that debt falls in Scotland, England and Wales, but would it be possible to get such a breakdown? Would it be desirable for us to get that?

There are so many large numbers in the report. I think that you are referring to the tax debt figure, are you not—

I am.

—rather than the tax gap. These terms all sound very similar, but the tax debt is essentially the amount that is owed by taxpayers to HMRC for the UK as a whole. It is tax that has been certified as being due but which has not yet been paid. The figure for Scotland is another example of an apportionment of the total. At the moment, HMRC does not build up the figure for Scotland from individual taxpayers in Scotland. As I said in my opening remarks, the figure as a whole is roughly three times what it was before the pandemic, so it is a much bigger problem for all the countries in the UK receiving the cash that is due.

I make it clear that the figure for Scotland is an apportionment of the total rather than an accurate build-up.

Surely we know who owes what.

Yes. As I said earlier, that is an example of where greater investment in resources at HMRC might be required. If the Scottish Government decided that it wanted that figure to be built up accurately for Scotland, it would have to agree on how to achieve that with HMRC.

When you think about HMRC, you might imagine there being one big system that calculates the tax due from each taxpayer, monitors the amount that is paid and looks at any debt level and what is being done about it. However, that is not how HMRC is set up. In practice, there is a separate system for every bit of the tax system. There is a PAYE system and a self-assessment system, which involve two completely separate information technology systems. There is also a tax debt management system. That explains why it is not as simple as reading off all the data that you might want to know about taxpayers from a single database. It is much more complex than that.

Okay—so, if we wanted to know what the tax debt in Scotland is, the Scottish Government would have to make a request to HMRC.

Yes. There would have to be a variation to the agreed approach. At the moment, the basis for all the figures that are set out in our report is covered by an agreement between the Scottish Government and HMRC. Our audit assessment that HMRC is applying the rules reasonably is in the context of that agreement. That agreement would have to be varied—that is, improved in the way that you have described—if that system were to change.

I will now ask what might be described as a daft-laddie question. Perhaps somebody can help me out. If I lived in Carlisle but I worked for a company in Dumfries and travelled to Dumfries every day to work, which tax rate would I pay? Would I pay the English tax rate or the Scottish tax rate?

It is our understanding that where you live forms the basis of your tax affairs, so, in that scenario, you would pay UK or England-based tax rates.

Is that definite? In other words, my colleagues who lived and worked in Dumfries would pay a different rate of tax.

That is correct. It depends on your postcode. If you live in a Scottish postcode area, you will have an S code attached to your tax number, which will dictate that you will pay Scottish rates of tax, as set by the Scottish Parliament, rather than English or Welsh tax rates.

I think that Mark Taylor wants to come in.

Yes—I want to offer a slight clarification. Mr Simpson would pay the Scottish tax rate irrespective of where he lived, because there are certain rules for the rates that MSPs and MPs pay.

A slightly less flippant point is that there are very specific rules that dictate where people pay tax. The system is largely based on residency, but there are specific exceptions, one of which relates to MSPs.

So, do not think that you can just jump on the M74 and pay less tax, Mr Simpson.

I want to assure Mr Taylor that I was not trying to wriggle out of paying tax. [Laughter.] It was merely an example. I was thinking of somebody else—someone who was not an MSP.

That takes me on to another point. From all the papers that we have here, it is quite clear that HMRC does not know where a lot of people live. Why is that?

The basic fact is that it is not a legal requirement to tell HMRC where you live. That would be a measure that the UK Government, as the host of HMRC, but also the Scottish Government could legislate for—they could legislate to make it a legal requirement to inform HMRC of any change of address. Currently, that is not a legal requirement. That is one reason why HMRC has to invest so much time and effort in investigating the accuracy of taxpayers’ postcodes. That is a simple explanation for why the process is not straightforward at the moment.

Do you think that it would be a good idea to change the law so that if, for example, I worked in Dumfries and I moved to Carlisle, I would have to tell HMRC that I had moved to Carlisle, or vice versa?

09:45  

Thankfully, auditors are not legislators. When it comes to what might appear to be a natural audit recommendation for an orderly system, there are lots of other factors to take into account, such as privacy and so on, which are for politicians, not for auditors.

However, in order to be able to apply the right tax rates now that there are different rates of tax, surely it is important for HMRC to know where people live.

Absolutely.

I am sorry to hark on about this, but let us say that I am on PAYE, as a lot of people are. I know that it is not a legal requirement to tell HMRC, but whose responsibility is it to tell HMRC where somebody lives? Is it the employer’s or the employee’s?

If you are on PAYE, it is the employer’s responsibility. HMRC does a lot of work with employers to ensure that they code accurately. It is the employer that ensures that the tax code for each employee is accurate. It is the responsibility of the employer to get right the S prefix that is used for Scottish taxpayers. Through the PAYE system, HMRC has regular communication with employers about that and it carries out regular checks.

Do you think that the level of accuracy of the information in question is greater for people who are on PAYE than it is for those who are not?

There is an extra level of checking involved, because the employer is involved as the intermediary between the taxpayer and HMRC. Obviously, self-assessment, by definition, involves a direct tax return from the taxpayer to HMRC. Some people in that position might have tax advisers, while others might not. It is certainly the case that, through the employer, there is an extra step and an extra control in place for PAYE.

Okay. I will leave it there for now, convener.

I will move things swiftly along and invite Willie Coffey to ask some questions.

Good morning. I want to pick up on the issue of employers that habitually do not apply the Scottish tax code for their employees. In the NAO paper, we read that there were more than 37,000 such cases. We ask this question every year: why is there a recurring issue with employers failing to apply the correct tax code for their employees in Scotland?

The issue continues because employers’ diligence in relation to that responsibility varies. That is the straightforward answer. HMRC has in place good arrangements for checking and identifying such cases. When they are investigated in more detail, HMRC finds that there is usually not a significant issue in relation to the level of tax involved. I think that that will provide some reassurance to the committee.

As auditors, our experience of such systems tells us that you can never relax and assume that you have created a secure and reliable system. You always have to put energy into checking compliance. For example, new employers might not be used to the system and might need to be introduced to it to understand how it works, and changes in ownership might mean that a previously reliable employer becomes unreliable. Our bitter experience as auditors suggests that you have to keep putting energy into checking compliance and communicating what is required. You cannot assume that, once you have done that for a few years, everything will be understood and implemented, because that simply will not be the case. Constant vigilance is needed.

When we spoke with HMRC, we were told that the number of employers that persistently get tax codes incorrect is very low, so its activity with employers to rectify such mistakes is effective. It is not as though the same employers make the same mistakes year on year.

I was going to ask whether the same employers are involved. For the fifth year, there seems to be the same number of cases—about 30,000. A member of the public would ask why that is the case. Is it the same employers making the mistakes? How many employers are we talking about? Is it a small number, or is it in the hundreds? Can you give the committee any further information to help us to understand the issue?

We do not have data on the number of employers.

It is welcome that the Scottish Government is taking an interest in the matter. As colleagues in the NAO have reported, the Scottish Government has requested data from HMRC on the number of staff employed by each employer that is using the incorrect tax code. You are right about the overall number of employers, but I accept what Rebecca Mavin said about the patterns that sit below this. The Scottish Government has set out the correct next step so that it can be absolutely clear on the scale of the issue. Are the employers large or small? Is there any discernible pattern that influences the overall compliance risk to the collection of Scottish income tax?

At a previous committee meeting, we asked for more information about the number of employers that habitually make such mistakes—we might even have asked for their names. Is it possible for the committee to get sight of that information?

The committee would have to make a request to HMRC. We certainly do not have that information.

Can you give the public some reassurance on the matter? Is the balance different each year, or are we dealing with the same offenders every year?

We cannot be precise, for the reasons that we have said, but the evidence that we have seen from HMRC suggests that the issue is, in essence, churn. Once HMRC spots a problem with a particular employer, it secures corrective action by that employer, but then another problem crops up somewhere else. The number of cases that are solved and the number of new problems balance out each year, which is why the number appears to be stable, but the same companies are not involved every year.

That is our impression, but the extra information that the Scottish Government has asked for will be helpful in getting into the topic.

Is the money recoverable in relation to companies that should have applied the Scottish code? Is the tax ultimately recoverable when they correct the code for previous years?

Yes. When errors are identified and the employer is responsible for the incorrect coding, the employer is responsible for making that good.

I have a very quick question, which relates to this thing about third-party data checks to get assurance. The report says:

“1.4 million records (27.0% of records analysed) could not be corroborated by comparison with the third-party data.”

That seems to me to be quite a high number. Do you have any concerns about that, Mr Davies?

That is not massively different from the proportions in previous years. The most recent exercise was only just available when we produced the report, so we did not have all the detailed analysis from HMRC. For our next report, we will follow up on that and give you the final analysis, because it will be available by then.

On the face of it, it seems worrying; you cannot get third-party confirmation for a large number of records. However, it is an area in which, for example, the data protection rules that I described earlier kick in. It is not straightforward to make comparisons between different data sources, particularly if the data that you are trying to compare it with was not collected for that purpose. Quite rightly, data protection rules are tough on that. If you did not state the purpose for which the data was to be used, it cannot be used for that. That is one reason for the high number.

Secondly, the inaccuracy could be on either side of the third-party comparison. The fact that the residency status does not match for a particular taxpayer does not mean that the tax code is wrong; it might mean that the other data set is out of date and inaccurate. It is important to be clear. That does not necessarily mean that there is an error in the tax figure. Rebecca Mavin might be able to add some more detail if that would be useful.

I can add some limited detail. Since the publication of the report, we have had some preliminary analysis from HMRC on this year’s results. Of the 27 per cent of records that were unmatched, there were various reasons why that was the case, as Gareth Davies has said. HMRC was able to match 83.5 per cent of the unmatched records to Scottish taxpayers who did not have a tax liability, either because they were abroad or were under 18, or because it was a duplicate record. Of the remainder, HMRC told us that almost all of those were temporary reference numbers, such as a temporary national insurance number that was given to someone while they awaited a permanent one. HMRC is confident that it has now matched more than 99 per cent of all records using other data sources. We will have to interrogate that in more detail in next year’s report. The initial number of unmatched records does not necessarily correspond to an unmatched record; it is just that, initially, it does not match with the third-party data.

That is helpful. It might be that, in due course, we invite HMRC in and get into more of the detail on that. Perhaps the Auditor General would like to say a word on the issue.

You will see from the reports that third-party data exercises are carried out every two years at present. In our report, we ask the Scottish Government to consider the appropriateness of the frequency of the exercises, given the volatility that we have discussed already this morning, and to consider whether undertaking them more regularly might provide more certainty to the Government in its information-resultant forecasting decisions that it makes around the budget. We think that it is important for the Government to be satisfied about how regularly the exercises are undertaken.

You mentioned the Scottish income tax board earlier. One of the things that resulted from our work last year was that it now produces minutes. They are not the most illuminating or comprehensive minutes that I have ever read—they are, basically, an outline of the agenda and a couple of bullet points beneath. Nonetheless, one set of those minutes describes the fact that the board is planning to be more strategic. Do you get any sense that that is the case?

I am probably not in a position to comment with any degree of authority about whether that is the case. We welcome the increased transparency and the activity of the income tax board. We have not done any specific audit work on whether it is adopting a strategic or operational position, but we can certainly consider that as part of a future programme.

Some of the areas that we have discussed in the last hour would presumably fall within the remit of a Scottish income tax board that was seeking to be more strategic in its approach.

There is no question about that. I guess that I am resisting putting a label on whether it is or is not being strategic; instead, we are focusing on the specifics that we have in front of us that relate to its satisfaction with compliance, third-party data checks and so forth. All of those will be decisions that the Scottish income tax board and the Scottish Government more widely should have under consideration. Whether they choose to identify those as strategic or operational is a matter for them. We will keep that under review.

Okay. We might not be able to rely on the minutes to help us to get to the bottom of that. Anyway, we go to Colin Beattie.

Thank you, and good morning.

I am sorry to bang on about stuff that has been partially discussed already, but I have a big concern about the uncertainties of the tax revenue and the way that it is calculated.

In 2021, I took down a list of all the anomalies in the NAO report for 2020-21, in which there were estimates, guesstimates, apportionments and goodness knows what else. I did the same for 2021-22, and it is almost a carbon copy. We have your report for 2022-23 and, although I have not gone to the trouble of taking out line by line, it is pretty much the same again. As you said, Mr Davies, there have only been some marginal improvements. Is HMRC in breach of the service level agreement that it has in place with the Scottish Government?

10:00  

No, we do not think so. We think that it is reasonably applying the requirements of the service level agreement. As I said before, that is a long way from saying that the estimate that it produces is accurate to any particular percentage level. Given what is required in that agreement, however, the approach that HMRC is taking is reasonable.

Our audit provides assurance that the processes in HMRC systems to deliver the estimate under that agreement are working as intended. Our audit gives you that assurance. It is important to note, as I have said, that that is not the same as a normal audit of a set of financial statements, in which there is a much higher level of assurance on accuracy. This audit says, in essence, that HMRC has applied the processes that it said that it would apply in its agreement with the Scottish Government.

I am not a legal person. However, if you look at the SLA, you will see that, in clause 23 on page 5, the first bullet point says:

“HMRC will identify the Scottish taxpayer population and collect from it the correct rates of SIT to ensure the Scottish Government receives the correct amount of income tax revenue each year”.

Patently, it does not. Continuing on, there are another five bullet points about, for example, identifying and maintaining

“an accurate and robust record of the SIT taxpayer population”.

You cannot say that HMRC is doing that, or is it?

We can say that the outturn figure is fairly stated, albeit two years after the end of the year. The figure that you get for 2021-22 in this report, for example, is not full of estimates; it is an accurate figure that is based on actual tax due from Scottish taxpayers. It is subject to the same level of uncertainty as the whole UK tax take on levels of non-compliance and so on, which is the tax gap that I referred to earlier on. That level of uncertainty is no different for Scotland that it is for the rest of the UK.

With respect, you have already said that that is an apportionment of the UK. There are no specific figures for Scotland.

No, not for the tax gap. The point that I was making is that HMRC will be able to say that, for 2021-22, it has accurately identified the amount that is due from Scottish taxpayers to the Scottish Government. What is much more full of uncertainty is the estimate for 2022-23. This time next year, obviously, we will report to you the accurate figure for 2022-23.

As I said, I am not a lawyer, but, as I go down the list, the clauses seem to be very simple and straightforward and, from your report, it is quite clear that the HMRC is not complying with the SLA. It is okay to say that, in a couple of years’ time, we will have a real figure, but that does not say that it is timely or relevant information or that it is received by the Scottish Government in such a way and such a speed to enable it to discharge its duties in respect of rate setting, forecasting for SIT and all those other things. It is a fairly solid SLA, but it is not being complied with.

We think that you can take reasonable assurance that it is being complied with. However, it does not generate an accurate estimate for the financial year that has just ended—you have to wait another year to get that, and that is the point. It is not right to say that HMRC is not delivering an accurate record of tax due and paid to the Scottish Government—it is. However, it takes two cycles for that to be accurate. At the moment, the estimate is necessarily full of estimates and will, therefore, be out by an amount in either direction.

It is important that we are clear on what we are talking about. If the SLA said, “Deliver an estimate for the most recent financial year that is accurate to within 0.1 per cent”, you would be absolutely right that HMRC is not delivering that at the moment and would need much more accurate and expensive bases of estimation to deliver that. What it is delivering is an accurate list of tax due and collected from Scottish taxpayers. You can take the same level of assurance on the accuracy of that as the UK as a whole can take for HMRC’s performance.

Again, if we look at this literally, the SLA says that

“HMRC will identify the Scottish taxpayer population”,

and it also says that it

“will identify and maintain an accurate and robust record of the SIT taxpayer population”.

You cannot say that it is doing that.

We think that it is taking a reasonable approach to that. Of course, a reasonable approach does not mean 100 per cent accuracy with no errors. That would not be achievable in any conceivable tax system. The question is: what is reasonable? As we said earlier, given the state of the legislation—there being no legal requirement to declare one’s change of address to HMRC at the time that it happens—the approach that is being taken at the moment is in accordance with the agreement with the Scottish Government. Therefore, it is reasonable. There is plenty of scope for improving the degree of assurance that you can place on the various controls in the system. We have discussed lots of those this morning. However, that is not the same as saying that HMRC is not complying with its SLA.

There does not seem to be a lot of wriggle room here. The SLA says that HMRC is

“to provide the Scottish Government with sufficient relevant and timely information and data for assurance purposes and to budget effectively”.

Can you categorically say that it does?

We have talked about lots of areas in which it takes two years to get accurate information. The basis for how the estimate is compiled versus the outturn for the year before is a shared understanding between the Scottish Government and HMRC, and how that will be approached is agreed in some detail each year. I think that it is transparent. It is not as if the Scottish Government thinks that it has agreed one thing and HMRC is delivering a completely different service. This is the service that it has been commissioned to deliver, with all the uncertainties in the most recent year that we have described.

Where is the agreement on all the uncertainties? Is it an agreement that says that apportionment will be done in this way and estimates will be done in that way?

Broadly, yes. Each year, there is a discussion about the approach that will be taken to build up the estimate. That is a transparent process, and it is understood. Obviously, there is a report from us every year confirming that, as well. It is understood on the basis that that estimate is produced. As we have described, it would be a policy decision, in essence, for the Scottish Government to ask for a more granular approach to some of the estimates in order to get a more accurate figure. As we have said, there is a strong case to be made for that, but it would come with a cost.

Do you measure performance against the SLA?

Not in the sense of routine monitoring of every line of it. We have three clear responsibilities, as we set out at the start of the report. One of them is whether HMRC has correctly applied the rules and procedures that apply to Scottish income tax. We test that in our audit work, but that is not the same as monitoring every line of the SLA.

Auditor General, do you look at performance against the service level agreement and compliance with the clauses that have been agreed?

It may be helpful if I provide some clarity on what our role is here, Mr Beattie. It goes back 10 years now. Under the Scotland Act 1998, Audit Scotland does not audit HMRC. That is clearly a role for the C and AG and his colleagues in the National Audit Office. As I alluded to in my opening remarks, my role is one that I inherited from my predecessors and it was taken on at the request of predecessor committees. As Auditor General for Scotland, I provide an additional assurance report to the Scottish Parliament through this committee.

My role is to consider the NAO’s approach to identifying the key risks to the successful administration of Scottish income tax, to look at some of the National Audit Office’s audit files on a sample basis and the audit evidence that it has accumulated in arriving at the judgments in its report, and to take a wider view on the quality arrangements in the National Audit Office. As I mentioned, I am satisfied that all those factors have been appropriately discharged by the National Audit Office.

What I do not do—and clearly do not have the powers to do—is to have any direct engagement with HMRC on its compliance with the service level agreement that you asked about.

Given that it is an agreement between the Scottish Government and HMRC, would you not have a role from the point of view of looking at what the Scottish Government is getting?

I probably have to repeat my previous answer in a different way, Mr Beattie. I do not have a role. It is a matter for Gareth Davies and his colleagues to audit HMRC in relation to the engagement that it has with the Scottish Government through the service level agreement. However, my role, which I discharge through my reporting, includes considering and pointing out options where the Scottish Government can take further steps to satisfy itself about compliance, accuracy and the quality of the data.

As a layman, I would still say that the service level agreement is being breached.

I will move on to something different, although we touched on it when we talked about Scottish income tax behaviour. Auditor General, is there enough published data and research on taxpayer behaviour to assist the forecasting by the SFC, policy development by the Scottish Government and scrutiny by the Scottish Parliament?

There is more to do. There is scope to publish more data, building on some of the activity that the Scottish Government agreed with HMRC this year on the data sets. Echoing a point that Gareth Davies made earlier, I note that we state in our report that there is increasing risk in this field given the divergence of tax rates between Scotland and the rest of the UK and how that might influence taxpayer behaviour in Scotland. I think that both the NAO’s report and our paper, given the judgments that we make, say that this is an area of risk. The publication of more data and the creation of more transparency would inform the decisions that the Scottish Parliament takes, ultimately, on the Scottish budget. There has been some progress, but there is more to do. The Scottish Government needs to determine how close an eye it wants to keep on the impacts of behaviours.

Also at play here is the time lag in the availability of data. As Gareth Davies described, we operate a couple of years behind in relation to the cycle. Some of that is through necessity but, as the convener mentioned, it is important for the Scottish income tax board to take a view on how robust the data is, on the frequency of checks and on how transparent the Scottish Government wants to be for its own purposes, but also for parliamentarians more widely.

As you say, it will be a couple of years before we really understand the full impact of what is happening now.

That is correct.

Okay. I will move on. Do Audit Scotland and the C and AG believe that the strategic picture of risk could be more bespoke for Scotland? We keep coming across places where data is not available for Scotland as such, and there is also apportionment from UK-wide figures. To be honest, the UK has a different make-up of taxpayer. It is all distorted by London. That goes for any place outside London because it dominates, and any apportionment will be somewhat distorted. Can we do this better? Can we get better figures for Scotland?

The strategic assessment of risk is the HMRC process for looking at, essentially, the risks to collecting all the tax that is due in line with tax law. It looks at non-compliance, tax evasion, economic trends and so on. As you say, that is an important bit of analysis.

I think that you are right. As the UK diverges both economically and in tax policy, that analysis needs to become more granular and to take the increasing divergence into account if it is to remain a useful guide to the allocation of HMRC resources to maximise the tax yield, minimise evasion and meet all the other objectives.

The matter needs to be kept under careful review. There is a need to improve the accuracy in relation to the nations in the UK and potentially, as you say, regionally within countries in the UK as well. How much that should be done is a policy question for ministers rather than a question for the NAO, but we will continue to report on the matter and, as part of our role in the UK, to hold the UK Government to account for how useful and accurate its strategic assessment of tax risk proves to be.

Continuing on risk, I have a question on cross-border migration. A trend analysis is referred to in paragraph 2.33 of the C and AG’s report. Has that report on cross-border migration trends been published? I see that HMRC expected the analysis to be complete by January 2024. It has perhaps been under a bit of pressure, but do we know where that analysis is?

That refers to cross-referencing with information from the Land Registry about cross-border moves. The analysis had not been completed when we asked recently post-publication. We are still awaiting that data.

Is there a revised date for it?

We were not provided with one. We were just told that it was not ready yet. The analysis has been done in previous years, but we do not have any information on it for this year.

10:15  

Do we know whether it is being done this year?

That would be a question for HMRC, I am afraid. We just inquired about the delivery of it. We do not know about that yet.

My next question leads on from that. Mr Davies, will you update the committee on the jointly-funded work by the UK, Scottish and Welsh Governments on cross-border migration trends over a period, which is mentioned in paragraph 2.34 of the report, and the HMRC work on retirement and migration, which is mentioned in paragraph 2.35? Where are we on those pieces of work?

I think that that is the work that I described earlier. The work to build up a longitudinal assessment of taxpayer behaviour across borders is a welcome development because it is the only way in which we will be able to form a robust view on the impact of divergent tax policy on behaviour. It represents a good investment.

It will be some time before we see any impacts of the increasing divergence, because it will happen in the current year and next year. The ability to see such impacts showing up in the data is at least two or three years away. However, if we did not start collecting the data now, we would not have the baseline that we will need to measure any changes. The work is a good development.

You will monitor that going forward.

Yes.

Okay—thank you.

We have gone over our time allocation. I think that some members of the committee still have questions, but we will follow them up in writing if you are willing to answer questions on that basis and if that is acceptable to the committee members who still have questions to put to you, including Graham Simpson.

Thank you for the evidence that you have given us this morning. It has been illuminating. We know that, at points, you have been challenged by the committee, and we thank you for your honesty in your responses to that. We have other work that we may need to do. In previous years, we have, in turn, invited the Scottish Government and HMRC to give evidence. The committee will consider how useful that would be on this occasion.

I thank Gareth Davies and Rebecca Mavin for their attendance and their evidence. Auditor General and Mark Taylor, I thank you as well. With that, I will draw the public part of this morning’s proceedings to a close.

10:18 Meeting continued in private until 11:34.