Public Audit and Post-legislative Scrutiny Committee

Meeting date: Thursday, May 9, 2019

Official Report 525KB pdf

---

Administration of Scottish Income Tax 2017-18

Item 3 is on the administration of the Scottish rate of income tax 2017-18. I welcome our witnesses: Jim Harra, deputy chief executive and second permanent secretary, Her Majesty’s Revenue and Customs; and Jackie McGeehan, deputy director for income tax policy, HMRC. Thank you very much for coming.

I invite Jim Harra to make a brief opening statement.

Thank you, convener. I would like to make two short points. First, I reaffirm HMRC’s commitment to administering Scottish income tax effectively, which means providing a good service to the Scottish Government and to Scottish taxpayers so that they can have confidence that they are paying the right tax at the right time. As part of that, we can always improve; we have joint oversight and governance arrangements with the Scottish Government, including the Scottish income tax board, which I have asked to oversee the steps that we take to gain better assurance of our work.

Secondly, I want to clarify our inability to share the strategic picture of compliance risk that we have produced for Scottish income tax. It is part of our overall strategic picture of risk for tax compliance, which we do not publish for operational reasons. However, we will include an assessment of the compliance risks and a summary of our compliance activity in our annual report in September, and I will be happy to answer any questions on our compliance plans.

Thank you very much. I ask Colin Beattie to open the questioning.

I have had a long-term concern about the accuracy of the figures for the Scottish rate of income tax in being able to capture the correct number of Scottish taxpayers. Do you agree that the base figures from 2016-17 are the key to the figures going forward?

Yes. Identification of Scottish taxpayers is our prime operational concern in making sure that we get the administration of Scottish income tax right. That is not a one-off process but an on-going one. There is no perfect benchmark against which we can measure ourselves—there is no perfect index of Scottish residents.

We have undertaken considerable work to identify Scottish taxpayers and to gain a level of confidence in how accurate our index is. Our assessment is that we have 98 to 99 per cent accuracy. That does not mean that the other 1 to 2 per cent is wrong; it means that we have no means of corroborating that. We have taken extensive steps and I think that the index is as good as it can be at this point, but we are continuously improving it.

At a previous evidence session, the National Audit Office said that the box to tick to identify a person as a Scottish taxpayer was on the self-assessment forms for the 2016-17 tax year but that

“whether it was or was not ticked, no action was taken on it at that time”.—[Official Report, Public Audit and Post-legislative Scrutiny Committee, 28 February 2019; c 10.]

Does that create a concern about the baseline figures that we are working from?

No, I do not think so. First, our prime way of identifying Scottish taxpayers is through the address data that we hold. All our efforts go into validating that address data and keeping it up to date. In the case of people who file self-assessment returns, they are asked to confirm their address; those who file their returns online are also asked to tick a box to say whether they are a Scottish taxpayer.

Over the past few years, we have been learning how we can use the responses in that box to best effect. For example, in this year’s tax returns, when someone did not tick the box to signify that they were a Scottish taxpayer, we initially calculated the tax on the basis of what they told us in the self-assessment return—even when we held address data that suggested that they were a Scottish taxpayer. The intention was to pick those cases up afterwards to understand the reason for the discrepancy. However, we quickly learned that there was a large number of errors, in that taxpayers should have ticked the box but did not. Therefore, we quickly changed our business processes so that we do not initially calculate income tax on the basis of what is put on the self-assessment return, but based on what our data tells us is the residence of the taxpayer, and we inquire into those discrepancies afterwards.

On the question of how we can make best use of taxpayer representation on the SA return, it has been a learning process for two years, and it might well continue to be.

But if the 2016-17 figures are the baseline from which we are working and no follow-up work was done at the time to validate those baseline figures, are we not working in the dark?

I realise that HMRC went back to some of those figures and made alterations in subsequent years—I think that 30,000 taxpayers were picked up. However, nothing was done with the original baseline 2016-17 figures. Are they now irrelevant as a baseline?

We did a lot of work to validate the Scottish taxpayer base in 2016-17, including work to validate our address data against other databases. You are right that we did not use the SA return entries as part of that validation. Our experience—particularly with the 2017-18 returns—is that they are of limited use as a validation, as they throw up a large number of discrepancies. We have found that the address data that we hold is much more accurate than what people put on their self-assessment returns.

But if the 2016-17 outturn figures are being used by the Scottish Fiscal Commission and the Office for Budget Responsibility for their forecasts, are they working from incorrect figures?

No. Our view is that they are working from figures between 98 and 99 per cent of which we are confident are correct in terms of the identification of Scottish taxpayers. For the remaining 1 to 2 per cent, we have no corroboration; some will be correct and some will be incorrect.

But the National Audit Office has said that no action was taken by HMRC at that time in relation to whether a person elected to be a Scottish taxpayer; it has said that there was no follow-up work on that.

That is correct in relation to the entries on the self-assessment tax returns. However, it does not mean that we did not take any action to validate the Scottish taxpayer index—we did. For example, we validated the addresses that we hold against third-party data sets and followed up on discrepancies and mismatches between them.

In our experience, the entries on the self-assessment tax returns are not the best indicator of whether the people concerned are Scottish taxpayers, because the error rate in them is higher than we would like it to be. Therefore, in order to make good use of that data, we must continue to improve the accuracy of what taxpayers put on those returns. In the meantime, we have taken the decision not to rely on what people put on those returns.

So you are saying that the 2016-17 outturn was corrected by HMRC on the basis of addresses.

Yes. In order to create the 2016-17 population of Scottish taxpayers, we put a flag on our systems for everyone for whom we held a Scottish address. In addition, when we knew that people had more than one address, we took steps to ensure that the one that we held was the correct residential address. We tried to match the information in our data set with that in other address data sets, such as the electoral register and the data sets of credit reference agencies. Where there were mismatches, we looked to see whether we could corroborate which address was right; more often than not, it was the HMRC address that was most up to date. When we could not find any corroboration, we contacted taxpayers to explore the discrepancy with them. As a result, we are confident that the accuracy level in the Scottish taxpayer base is between 98 and 99 per cent.

In previous evidence sessions, we had information that showed that there were concerns about how companies that have cross-border operations identify their resident employees. That was a considerable problem, at least in the initial stages. Has it now been eliminated?

It is HMRC that identifies whether someone is a Scottish taxpayer, not the employer. We notify the employer that we regard that person as a Scottish taxpayer and that, therefore, they should deduct Scottish income tax.

We have found that although the vast majority of employers do that correctly, in some payrolls they make mistakes, because they have not got their software or the administration of their payroll right. When we get their first return of the tax year, we can run a scan against it to identify whether there are cases for which they have not deducted Scottish income tax when we expected them to. In such cases, we reissue the code and work with the employer to make sure that they get that right.

If an employer gets that wrong, it does not affect the amount of tax that is assigned to the Scottish Government, because that is calculated by reference to HMRC’s systems. However, it could result in the wrong amount of tax being deducted from a taxpayer during the year, which would mean that, when we did a reconciliation at the end of the year, we would find that they were due a repayment or had an unexpected tax bill. Therefore, it is important that we work with employers to get that right from the outset, for the sake of taxpayers. However, there are fail-safes to make sure that we pick the issue up. In particular, we carry out a scan in-year and a reconciliation after the end of the year.

You said that you believe that your accuracy is 98 per cent or thereabouts. How does that compare with the rest of the United Kingdom? I realise that it is a slightly different scenario, but you also identify UK taxpayers, so you must have a percentage accuracy for that, too.

As I said, there is no benchmark against which we can measure ourselves. All that we can do is do as much operational work as we can to ensure that we get the identification right, and then carry out corroboration exercises with other data sets. We do not do that in the case of English and Northern Irish income tax payers, because there is no need, but we do it for Scottish taxpayers and we now also do it for Welsh taxpayers. We would expect to achieve that 98 to 99 per cent accuracy level for both those populations. Our task is to maintain that and to improve on it over time.

You say that there is nothing to benchmark against, but you are obviously benchmarking when you say that you are achieving about 98 per cent accuracy. What is that based on?

What I mean is that there is no perfect benchmark. We do benchmark ourselves; we corroborate the information that we hold in our data systems with information that is held in third-party data sets. Where there is a mismatch, or an inability to match, we look into that to see whether it tells us anything about the accuracy of our data set and, if necessary, we correct our data.

That work has shown us that the address data that HMRC holds, which is the key thing that triggers us to put a Scottish taxpayer identification flag on someone, is generally more up to date than the third-party data sets that we corroborate against. From our work, we believe that we can corroborate and be confident in 98 to 99 per cent of our identifications. As I say, there is no reason to believe that the balance of them are all wrong—some will be, but some will be right. We will continue to drive that level up, but we have taken all reasonable efforts; there is no perfect benchmark against which we can check ourselves.

To follow up on Colin Beattie’s question, if the data set is based on the addresses that HMRC has rather than what is put on the self-assessment forms, does that mean that there are also mistakes whereby English taxpayers are wrongly identified as Scottish taxpayers? If so, how many cases are there of that?

Where there is a discrepancy between what the taxpayer puts on their tax return and what our database holds, we will inquire into that and resolve it. The question is what we do in the meantime when it comes to calculating that person’s tax. Our initial decision in this year’s self-assessment run was to calculate the tax based on what people said on their return. However, we quickly learned that the level of error in returns was such that it was better to calculate the tax based on what we knew from our own database and to follow up on the discrepancy afterwards.

Obviously, if somebody notifies us of a change of address on their self-assessment return, that will represent an update to our database. However, when a Scottish address is provided but they do not tick the box to say that they are a Scottish taxpayer, that is a discrepancy that we have to inquire into.

So there are discrepancies both ways, which adds to the workload. It is easy for us to think about discrepancies one way leading to a high workload, but if there are discrepancies both ways, that will lead to an even bigger workload. How many discrepancies or queries were there?

10:15  

We had to correct 30,000 when we stopped the business rule. In the meantime, we had calculated tax for about 30,000 taxpayers based on their tax returns, but in hindsight we thought that it would be better to assess it based on the information that we held. We are carrying out checks into those and subsequent cases.

You are right that the error rate is too high, which means that it will give lots of false positives that we have to inquire into. One of our tasks is to improve the accuracy of what taxpayers are putting on their returns.

Just to clarify, were there 30,000 checks and clarifications or were there fewer?

There will be more than that. There were 30,000 cases in which we had to make a correction, because we concluded that it was better to initially assess tax based on our data set. However, there will be other cases after that in which our calculation is based on the address that we hold, but we will have to find out—

How many?

I do not have a figure for that at this stage.

Are the checks and clarifications made by letter or phone call?

They are made by letter and by phone, although it is generally done by correspondence.

That sounds like a lot of work, a lot of resource and a lot of money.

Yes. A lot of it will be unnecessary work because it will be a simple error that someone has made. That is why we have to look at the design of the tax return and the guidance that we have given to see whether we can push up the accuracy rate of the information that taxpayers give us.

How many members of staff has it taken and how much has it cost HMRC to make those clarifications?

I do not have figures for that.

Are you able to put those figures together and provide them for the committee?

We can give the committee whatever we have on that.

There is a particular issue with manual entries—MSPs are a good example of that, with about a third of us being classified wrongly. How many manual entries were there from Scotland? How many people were incorrectly identified?

Our general business rule is to set a Scottish taxpayer flag by reference to the address. That process, which is automated, applies to the vast bulk of Scottish taxpayers. However, there is a different legal test for parliamentarians: you are automatically a Scottish taxpayer if you are a parliamentarian, regardless of where you live. Therefore, we have to switch off the automated business rule and apply a manual process instead. We have a special tax office that looks after parliamentarians and carries out that process. It is just the parliamentarians who are subject to that process.

Does it apply to sensitive individuals in the round or does it apply purely to parliamentarians?

It is purely for parliamentarians.

It applies to MEPs and MPs with Scottish constituencies and to MSPs.

How many MPs or MEPs were wrongly classified?

We identified 45 MSPs whom we had sent an incorrect 2019-20 code, which did not identify them as a Scottish taxpayer. That was because of a clerical error in the tax office that looks after the affairs of Scottish parliamentarians. We identified that before the start of the tax year, stopped the problem and corrected the cases that we had got wrong. We alerted the MSPs whose codes were wrong, apologised for that and issued them with a revised code. We have also taken some steps with the office to ensure that the process runs more smoothly.

So you are saying that there was not an issue with any MPs or MEPs.

We have double-checked all Scottish parliamentarians and, once the self-assessment process is complete, they will all have been taxed on the correct basis. In the meantime, 45 MSPs got the wrong code for 2019-20. As I said, that was corrected before the start of the tax year, so no one was impacted.

However, we have to improve our assurance of that process, because it must be repeated each year. We are also looking at whether we can automate it, although I cannot promise that we will be able to do that.

Do you understand why Scottish parliamentarians, who set Scottish rates of income tax, are extremely concerned by the fact that a third of them were wrongly identified and how that reflects the challenges that might exist in respect of the wider population?

I can understand that concern and I share it. As you might imagine, I was not pleased to learn that the mistake had been made. I am satisfied that there is no read-across to the rest of the Scottish population, because there is a special clerical process that applies to parliamentarians and that is what went wrong.

For the main bulk of the Scottish population, the process is automated and based on the address set. We have done extensive work to gain assurance about the accuracy of that set. However, I share your concern that what happened should not have happened.

I have a final question. How is work going in relation to public consciousness of the Scottish rate of income tax and your communication with employers and employees on how they fill their forms out differently?

In administering Scottish income tax to individuals, our main message is that they need to tell us when they change their address. They need to keep their address information up to date with us, and the vast majority do that.

Until recently, for a lot of people, there has not been any difference in the tax that they pay, so they do not need to do anything. Although we want them to be aware of the fact that there is a Scottish income tax, we do not want them to worry or do anything different. Therefore, our messages to employers and employees have been about address changes. We will continue to work with Scottish Government colleagues on how else they want to raise awareness and how we can assist with that, but, for HMRC administration, it is really about the addresses.

I will pick up on the issue that Anas Sarwar raised about MPs and MSPs. We are not asking these questions because we are concerned about our tax affairs; we are concerned because, compared with the bulk of the Scottish taxpayers, MSPs, MPs and MEPs are a very small group of about 200 people. When it comes to confidence in the wider system, it worries me that a significant clerical error happened with such a small group and when there is a special process in place. I am sure that you will be delighted to hear that I was one of the people who got an incorrect tax code. When an error such as that is made for 45 people out of a small group of 200 people, how can the Scottish public have confidence? There would not have been address issues for many of those people because we have all been Scottish parliamentarians for many years, so we will not have lived elsewhere for a significant number of years.

I understand your concern. I will explain a bit about the process that we use in applying the special rules for parliamentarians and how that went wrong.

A special rule—one that is not based on residency—determines the tax status of parliamentarians, so we have to switch off the normal business process that would allocate their Scottish taxpayer status, depending on where they live, because there are some Scottish parliamentarians and Welsh parliamentarians whose main residential address is not in Scotland or Wales. We have to disapply that process, which means that our computer systems’ default is to treat them as non-Scottish and non-Welsh taxpayers.

But that is only for MPs. You misidentified 45 MSPs who are all resident in Scotland.

Because an MSP’s taxpayer status is determined not by their residence but by a different legal test, in our computer systems, we have disapplied the normal rule, which runs on their address, that applies to other Scottish taxpayers. Therefore, as far as the computer systems are concerned, the default is that people are treated as not Scottish and not Welsh taxpayers.

We rely on a clerical process in the office that deals with parliamentarians’ tax affairs putting the flag back on the returns. Staff in that office did not realise that, having set the flag for the previous year, they would have to go back and repeat that, because the computer system would remove the flag again. There was a misunderstanding. Having put the flag on an MSP’s record once, they believed that it would be retained on the record. In fact, they had to repeat the exercise before we repeated the coding run. Therefore, before we picked up our mistake, they missed that process for 45 MSPs, so we had to go back and correct those codes.

I acknowledge that it is a clunky process, but we had to put it in place to make sure that the special rule for parliamentarians works. If we did it the other way round, there would still be some errors. That is the process that we have adopted. Unless we can automate it in some way, we will have to gain better assurance of that clerical process, so that we do not make the same or another mistake. Jackie McGeehan and the Scottish income tax board have been looking at how we can do that.

I understand.

It is worth saying that we are looking at what we can do to automate all or some of the process so that we do not need to rely on individuals going back and making the changes manually every year. As has been said, a very small number of taxpayers are affected, but we will look at how best we can automate the system.

It sounds as though your checking processes, as well as the technology, need to be tightened up. I repeat that our concern is about how we can have confidence in how the wider population is dealt with, given that mistakes have been made with such a small sample.

I entirely understand and share that concern. I repeat that an entirely different automated process is used for the general population of Scottish taxpayers who are not parliamentarians, so there is no direct read across.

Have there been any miscoding errors in relation to Scottish MPs and MEPs? You said that 45 MSPs have been affected, but have any Scottish MPs or MEPs been affected?

I am not aware of any who have been affected. I am aware that 45 MSPs had a coding error for 2019-20 that we had to correct.

I am not aware of the numbers. One or two people might have been affected, but I do not know the numbers.

I should say that I was one of the 45. The clerical error applied only to the MSPs, not the Scottish MPs.

That is as far as I am aware. We checked every parliamentarian’s 2019-20 tax year and earlier tax years to ensure that they were paying the correct tax and that the correct flag was on their records, so that their tax would be attributed to the Scottish Government, not the Westminster Government. We were satisfied that that would be achieved once we had completed all those checks and had made the corrections, between the pay-as-you-earn system and the self-assessment system, to the 45 codes.

Did the problem apply to any of their Scottish lordships? There must be hundreds of them.

We looked at all parliamentarians and my understanding is that no peers were affected.

Peers were not affected.

I will pick up on the error that resulted from 30,000 people failing to tick a box. Jim Harra said that, previously, the information and knowledge in the system that was used to identify where a person was from was based on their address. However, on the new form, people were invited to tick a box saying whether they were Scottish. Did the information that was taken from the box-ticking part of the process take priority over the information on addresses that you already had? Was that the error?

In essence, yes. Our orthodox approach is to start to calculate a person’s tax bill on the basis of their self-assessment. If we have any doubts about their self-assessment, we make inquiries afterwards. After we put the tick-box question on the self-assessment return, our default position was that whether a person ticked the box was the basis on which they were self-assessing their liability. We decided that that was how we should calculate someone’s tax bill and that we would make inquiries afterwards if we had any concerns.

However, as I said, we quickly learned that the error rate from people failing to tick the box was so high that it made more sense for us to base the initial calculation not on what people put on their tax return but on what we believed, on the basis of our data, to be the correct position. Therefore, we flipped over the business process. If our information on a person’s address suggests that they are a Scottish taxpayer, we will make our assessment on the basis that they are a Scottish taxpayer, even if they have not ticked the box to say that they are. We will explore the discrepancy with the person afterwards. If it turns out that they were right not to tick the box—our information on their address might be out of date, for example—we will fix the situation by repaying any excess tax. Following our experience of the initial 30,000 cases, we concluded that taking that approach would give a more accurate outturn than would be achieved by relying on what people put on their returns.

You made the box-ticking part of the process the new principal determinant of where a person lived instead of relying on the information that you probably already had about them.

That was the initial business process, which we have since disapplied.

You no longer take that approach. You already had information suggesting that the people were probably Scottish, but you chose to make the determining factor whether they ticked a box.

That was the initial position. I am not saying that we would not return to that approach if we were able to get the number of people who tick the box to a sufficiently accurate level, but we will not do so until the positive affirmation of residency status is correct in a sufficient number of cases.

In hindsight, was that not a daft thing to have done?

It was based on an assumption about the accuracy of people’s work information, which proved to be unfounded. By the time that we had picked that up, 30,000 cases had been dealt with in that way, and we decided that we would go back and adjust them.

10:30  

You mentioned that you are 98 to 99 per cent confident that your identification process is now accurate. Is that assessment based on your statistical approach? Is the estimate that you have declared a confidence interval, to use a statistical term?

Our statistical unit has calculated that figure. We tried to match the addresses that we held with those from other data sets. We achieved matches in around 80 per cent of cases, and the addresses were corroborated in 96.5 per cent of those cases. We then worked on the 20 per cent of cases in which there was no match.

For example, if we had an address for someone in Dundee but could not match that with any other data set, we looked up other factors such as their employer. If their employer was in Dundee, it was clear that the address that we held for the taxpayer was correct, because they were employed in the local area, so that counted as corroboration. However, if their employer was nationally based, we could make no assumption about the taxpayer’s address, so we did not treat it as having been corroborated. For cases in which we still could not achieve corroboration, we wrote to the taxpayer and asked them whether the address data that we held for them was correct.

That is how we arrived at the figure of 98 to 99 per cent of cases being corroborated and 1 to 2 per cent being uncorroborated. In the uncorroborated cases, some addresses will be right and others wrong.

Are you saying that, if we were to apply statistical principles of corroboration and accuracy, a standard 95 per cent confidence interval might be applied to data sets such as yours?

I am not a statistician, so, if I get this wrong, I will have to write to you to clarify the position. I think that the figure is basically 100 per cent. We have counted everyone whom we regard as corroborated as having been corroborated and everyone whom we do not so regard as having not been corroborated. I am not aware of there having been any confidence interval involved at all, because there had been a 100 per cent scan of our database. However, I will be happy to write to you if I have got that wrong.

Thank you.

Anas Sarwar has a supplementary question on that point.

You told Mr Coffey that you wrote to taxpayers for clarification. What if you had been writing to the wrong address? If the address that you had used was based on the data that you had but that data was wrong, you would not have got clarification, because you would have been writing to the wrong address.

We found that, in the interval between running the scan and issuing the letters, taxpayers sent us a significant number of address updates unprompted. However, in a small number of cases, they did not do so. You are right: there will always be cases in which we cannot get a response or cannot find the taxpayer, so we will have no way of corroborating—or not—the address data that we hold. We will never get to a point of perfection.

In such a case, would you go by the address that the taxpayer had put on their form and the fact that they had ticked the box or by the information that you held?

We would treat such a case as uncorroborated and so falling within the 1 to 2 per cent of cases that I mentioned. Nevertheless, we would proceed with administering the taxpayer’s tax affairs on the basis of the data that we held, which would include the Scottish address.

I want to move on to a slightly different topic. HMRC’s assessment of compliance risk in relation to Scottish income tax is based on what is called a strategic picture of risk, which I understand is updated in September each year. I believe that, in this instance, HMRC did not provide its assessment to the Scottish Government until December and that the reason given was that pressure on resources meant that it could not be provided earlier. Why were resources not available for a process that seems to be critical, and what is the impact of the late provision of that information?

Each year, we aim to complete the strategic picture of risk process by around July, which gives us time to get it to the Scottish Government by September. You are correct in saying that, last year, we did not get it to the Government until December. We have taken steps to ensure that it will be on time this year.

It is part of a larger, cross-UK exercise that we carry out to complete a strategic picture of risk for the country as a whole. A subset of that data is the Scottish strategic picture of risk, which focuses specifically on compliance with Scottish income tax. A general pressure of work meant that that exercise was not completed as fast as we would have liked. It feeds into our compliance plan, which we publish every year.

Our current assessment of the Scottish income tax in the first couple of years—2016-17 and 2017-18—is that the differentials between Scottish income tax and income tax in the rest of the UK were not sufficient to create any material additional compliance risk. However, we will complete that exercise for later years, because the differential is now getting more significant. In the year in which we made the compliance plan, the difference in the threshold for the higher rate meant that there was a maximum differential of about £400, which we concluded was not sufficiently large to create a risk that we would need to take specific action to tackle.

There are three levels of risk. First, there is the risk that someone will not declare their Scottish residency or will try to disguise it. The second risk is that someone might try to convert income that should be subject to the Scottish income tax into something else that is not subject to the Scottish income tax, such as a capital gain. The third risk is that someone might generally underdeclare their income, which is a general risk across the tax system. Depending on the level of differential, those risks could differ between Scotland, Wales and the rest of the UK. However, when we were making the compliance plan, we did not believe that we had reached that threshold.

You are saying that the previous late provision of the strategic picture of risk has not had a significant impact but that any such late provision could have an impact in the future. If that is correct, will you be providing one in September 2019?

Yes. I am not saying that, at this stage, the differential is sufficient to trigger additional compliance risks that would mean that we would want to carry out specific additional compliance activity. We have not yet reached that view with the Scottish Government. However, you are right in saying that, as the differential is now wider, it is more important that we act timeously. You have my commitment—as does the Scottish Government—that we will be doing that on time this year.

You will be doing that in September.

We will be doing it in September.

I have a final question, Mr Harra. You talked about the three tiers. To what extent does the strategic picture of risk identify the risk of Scottish taxpayers misrepresenting their residency? Can you give us a bit more information on that?

We have undertaken with the Scottish Government to assess that risk annually, and we will base our proposals for any compliance activities on that assessment. We have some baseline data for 2016-17 on the Scottish population and movements of people in and out of Scotland, which enables us to monitor trends as well as to risk assess individual cases.

At the moment, to the extent that there is a risk, it is largely among the highest-net-worth taxpayers. Our compliance strategy for dealing with high-net-worth individuals is to make individual risk assessments, with a compliance relationship manager appointed to each of them, whose job is to monitor that taxpayer on an individual basis.

At this stage, we monitor that risk both by looking at trends in the population as a whole and by monitoring specifically what high-net-worth individuals are doing or saying to us. We have a special devolved income tax compliance officer for high-net-worth individuals, to make sure that Welsh and Scottish income tax risks are taken into account in the regular risk assessment of those people.

I want to follow up on that, Mr Harra, because, as you said, the differential between Scotland and the rest of the UK is now bigger than it was at the start. From your trend analysis, particularly of high-net-worth individuals, is there any evidence that as the differential increases there is any increase in the relocation of such people from Scotland to other parts of the UK? I am talking not about people deliberately dodging or evading tax but about people taking a legitimate decision, from their point of view, to relocate.

I will give you the figures that we have. First, our definition of a high-net-worth individual is someone with wealth of £150 million or more.

No parliamentarians come into that category—that I know of.

I will take your word for that.

In 2016-17, we identified 608 high-net-worth individuals whom we regarded as having Scottish taxpayer status. In 2017-18, we identified 609, so there was one more than there was in the previous year. However, there was movement within that. Our records show that 16 of the 608 people we identified as having high net worth in 2016-17 said in the following year that they no longer lived in Scotland; they had moved either within the UK or abroad. However, there was a corresponding change the other way, so that we had one more high-net-worth individual in 2017-18, net, than we had in the previous year.

Overall, the differential is not having a net impact on the number of those people.

Our records show that, between 2016-17 and 2017-18, 16 people changed their residence. I have no way of knowing what their motivations were for doing that. However, according to our records, the overall number of high-net-worth individuals stayed almost constant between those two years.

Right.

Your definition of a high-net-worth individual is someone who is worth £150 million—is that right?

Yes. That is our main test. We have income levels as well but, broadly speaking, that is it.

What about below that? For example, the claims has been made that the differential is leading to an exodus of consultants and people from the health service who would be paid, depending on their status, somewhere between £90,000 and £150,000, or perhaps slightly more. Have you done any trend analysis on people who are high earners but do not have anything like the scale of net worth—£150 million—that you have defined?

No, HMRC has not done that. Our interest is perhaps slightly different from the Scottish Government’s. Our interest is in whether anyone is behaving non-compliantly in relation to the Scottish income tax. A decision to move residence is not a non-compliant action—it is something that people are entitled to do—so we do not have a direct interest in that. The policy makers in the Scottish Government might have an interest in it; they have access to data and we would help them in any way that we could, if they wanted to monitor that kind of thing.

Broadly speaking, we divide taxpayers into high-net-worth and affluent categories and a third group that I will call the mass market.

What does the trend analysis show among the affluent?

I do not have that information.

Could you send to the committee?

I can see whether we have it. Obviously, in the case of high-net-worth individuals, each case is looked at individually, which is how we can get that data. I will see what we have in relation to the affluent group.

That is very helpful.

You said earlier that peers, who are parliamentarians, are exempt from your special scheme for elected members, which covers MSPs, MPs and MEPs. Why are peers exempt from that scheme?

It was a decision that the Government made when the Scotland Act 2016 was finalised. My assumption—I am afraid that I am guessing, because I was not involved in it—is that peers do not represent a constituency. They are not representing a geographical area, so the link with Scotland is through where they live; if they live in Scotland they will be Scottish taxpayers.

I think that there is a specific definition of Scottish lords. There certainly used to be. There have been a number of examples of people who, at the same time as being MSPs, have sat in the House of Lords. Would people in those circumstances be allowed to register as non-Scottish?

They would be Scottish taxpayers.

So their elected status would take precedence.

Yes.

10:45  

Do you know how many of the 800-odd lords we have, who do such a wonderful job for themselves, are registered to pay tax in Scotland?

I am afraid that I do not.

Is that information that you would be able to give us? Obviously, I am not looking for individual names.

We can take that away and have a look at it. I cannot immediately think of any reason why we could not disclose that.

It would be very helpful to know how many lords pay tax in Scotland as opposed to the rest of the UK.

My third and final question is on the treatment of allowances and reliefs. Recently, I saw a calculation that revealed that the pension tax relief alone is worth about £47 billion a year, which is very heavily skewed towards the higher end of the earnings scale. Obviously, we do not have control over that. Does where people pay the pension contributions that they get relief on impact on the revenue that is assigned to Scotland?

Yes, it does.

In what way?

It is the net tax that we collect from Scottish taxpayers that is assigned to the Scottish Government. To the extent that someone’s tax bill is reduced by a tax relief such as the relief for pension contributions, that has an impact on the tax receipts of the Scottish Government.

I want to zero in on pension tax relief, because the Chancellor of the Exchequer appears to have his eye on that. Would you be able to tell us—not now—how much of the £47 billion of pension relief relates to Scottish taxpayers?

We can certainly take that away and ask our analysts whether we can produce that information.

That would be very helpful. Thank you very much indeed.

We have spoken about how taxpayers are allocated to Scotland and to the rest of the UK. How does someone get into the system? When is a person’s account with HMRC opened? People arrive in this country by being born here or by flying in and so on. When do you start to take an interest?

It will vary. Someone might have a record with us at any age. We do not automatically set up a record for taxpayers at birth, but someone might have a record with us because they claim tax credits or child benefit, they have a taxable source of income or they have claimed a repayment. Our estimate of the number of Scottish taxpayers—the 2.5 million figure—relates to the number of people who pay income tax. There will be a larger number of people on our database who are potential Scottish taxpayers, in that our data tells us that they are resident in Scotland, but their earnings might be below the personal allowance or they might have income that is not non-savings, non-dividend income; 2.5 million is our estimate of the number of active taxpayers.

When someone triggers your interest, is that the point at which you make a decision about whether they are Scottish or non-Scottish? How does that work?

There is a flag on our system for everyone who we believe meets the criteria for being a Scottish taxpayer. For example, someone who lives in Arbroath who is in the early years of self-employment and who is earning only £8,000 a year while they set up their business will not pay tax because they are below the personal allowance, but there will be a flag on the system to say that they are a Scottish taxpayer. That means that if, at any point, their income goes above the level at which they have to start paying income tax, that will automatically be identified as Scottish income tax.

When an employer takes on a new employee, they are now required to tell HMRC that person’s address. At that point, if it is their first job, we will know that they have a Scottish address.

But someone in Scotland who had no employment and no income would still be a Scottish taxpayer.

They would be a Scottish taxpayer if they started to have taxable income. Someone who is not currently a taxpayer might or might not be on our database already.

You said that the 2.5 million are people who pay tax, so the others on your system could be in employment but could also have allowances, whether in relation to pension or whatever. I just want to make it clear that pension payments are perfectly legal; there is no issue about that. It is good to save for one’s pension.

I would agree with that.

How many people are there on your system over and above the 2.5 million?

From memory, I think that there are about 3.75 million people on our database.

In total?

That is the total number of people on our database who have a flag against them to say that our records indicate that they are resident in Scotland. Of those, 2.5 million are active taxpayers.

Do you apply the same rules across the UK when it comes to pursuing those who evade tax, or is there is a difference between Scotland and the rest of the UK in that regard?

At the moment, there is no difference between Scotland and the rest of the UK in the compliance activity that we undertake; we do that on a UK-wide basis.

As I mentioned earlier, in the future, we might have a specific Scottish income tax compliance plan if we believe that there are specific risks in relation to Scottish income tax that we need to manage but, at the moment, the Scottish Government gets attributed to it a proportion of the yield that we bring in from our compliance activity across the UK.

Do you have any indication of whether evasion is higher or lower in Scotland?

We have no evidence to suggest that it is higher or lower in Scotland. There are some regional differences in the make-up of taxpayers that affect the inherent compliance risk. For example, employees tend to have a very low compliance risk, because they do not have the same opportunities to evade that, say, self-employed people have, and there are regional variations in how that population is made up. We are not aware of any national tendencies to evade or not to evade.

Where is Scotland when it comes to the balance between employees and self-employed people?

We tend to go on who is on self-assessment and who is on pay as you earn. Roughly speaking, the balance is about the same—about 16 to 17 per cent of taxes are paid through self-assessment and the balance is paid through pay as you earn.

The balance is roughly the same as in the rest of the UK.

Yes.

As other members have no further questions, I will ask a small techie question. Will the 2018-19 revenue estimate also take account of the 2017-18 Scottish income tax outturn data, which will be the most up-to-date information at that point?

Yes, it will. When we made estimates in previous years, we had no outturn data, so they were made purely on the basis of the survey of personal incomes, but this year, for the first time, we will be able to make our estimate based on that and the outturn for the previous year, as will the Scottish Fiscal Commission and the Office for Budget Responsibility.

Thank you very much indeed.

I close the public part of the meeting.

10:52 Meeting continued in private until 11:35.