Finance and Constitution Committee

Meeting date: Wednesday, October 28, 2020

Official Report 526KB pdf

---

Pre-Budget Scrutiny 2021-22

Our second agenda item is an evidence session with the Scottish Fiscal Commission as part of our pre-budget scrutiny. The session will focus on the commission’s analysis of the HM Revenue and Customs 2018-19 Scottish income tax data outturn figures, its “Forecast Evaluation Report” and its evaluation of income tax forecasts for 2018-19.

I welcome to the meeting Professor Francis Breedon, who is a commissioner; Professor David Ulph, who is also a commissioner; Claire Murdoch, who is head of social security and public funding; and David Stone, who is head of economy and tax.

I ask Professor Breedon to begin with any opening remarks that the SFC might wish to make.

I will be fielding the questions, convener, and David Ulph will give our opening statement, so I will hand over to him.

Good morning, and thank you for asking the commission to give evidence.

At the start of September, we published our main “Forecast Evaluation Report”, which covered our devolved taxes and social security spending forecasts for 2019-20. It did not include an assessment of our income tax forecast because, as a result of Covid, HMRC delayed the publication of its income tax outturn data beyond the usual publication date in July.

On 23 September, HMRC published the income tax outturn data for 2018-19. We wrote to the convener on that date with an initial assessment, and on 5 October we published our income tax evaluation. I should add that both our evaluation reports cover the pre-Covid-19 period.

I will briefly cover a few points from our income tax evaluation. Income tax funding in the 2018-19 Scottish budget was set using our forecast of Scottish income tax revenue, and our forecast of the block grant adjustment, which in turn was based on the Office for Budget Responsibility’s forecast of income tax revenue in the rest of the United Kingdom.

With the publication of outturn data for 2018-19, we now have final values for 2018-19 Scottish income tax revenues and the BGA. The differences between our budget-setting forecast and outturn and, similarly, the differences between the budget setting forecast and the final BGAs will be corrected through a negative reconciliation of £309 million to the Scottish budget for 2021-22.

The simplest way to understand where that figure of £309 million came from is to look at the forecast and outturn growth rates. As members are aware, what really matters for income tax funding and the Scottish budget is the growth rate of income tax revenues in Scotland since the initial deduction year—2016-17—compared to the growth rate of income tax revenues in the rest of the United Kingdom over the same period.

Our forecast in February 2018 predicted that Scottish income tax revenue would grow by 8.6 per cent between 2016-17 and 2018-19. Outturn data shows that tax revenues actually grew by 7.8 per cent, so we overestimated growth in Scottish income tax revenues. The income tax BGA was forecast to grow by 4.8 per cent, with outturn data showing that it actually grew by 6.7 per cent. The growth in the BGA was underestimated when the budget was set. Our moderate overestimate of growth in Scottish income tax revenues and the moderate underestimate of growth in the BGA reinforce each other, resulting in the negative reconciliation.

Before putting some numbers on that, I need to say a bit more about the errors in forecasting Scottish income tax and the associated BGA. When we made our February 2018 forecast, we did not have accurate outturn data on Scottish income tax revenues in 2016-17—that data became available only in July 2018. We estimate that, of our headline forecast error of £621 million, around £538 million arose because of the lack of historical outturn data. We calculate that, had we had accurate historical outturn data to work from, our forecast error would have been closer to £83 million, or less than 1 per cent.

That data issue had no direct effect on the Scottish budget reconciliations, since the forecast of the BGA was also affected by the error arising from the data issue, and the two effects cancelled out. Therefore, we believe that, of the £309 million reconciliation, our forecast error contributed £83 million and the error in the forecast of the BGA contributed £207 million.

Following a request from the committee last year, alongside our income tax forecast evaluation report, we recently published a paper looking at the effects of the distribution of income on the growth of income tax revenues, and we would be happy to discuss that with the committee today.

With that, I hand back over to you, convener, for the question-and-answer session. My fellow commissioner Francis Breedon will co-ordinate our answers.

Thank you. I will begin with some of the numbers that you have gone over, particularly the reconciliation of £309 million. I want to get some stuff on the record so that people understand why we have gone from a forecast negative reconciliation in 2018-19 of £555 million to a negative reconciliation of £309 million.

We have been over some of this territory before. We previously asked you whether the negative reconciliations were likely to significantly change when the outturn figures for 2018-19 were published. The SFC responded by saying:

“we are now close enough to having the final figure to say that we would be surprised if it was very different.”

We were told:

“this year’s £204 million figure settled down into that area some time before we had the final outturn figures.”

The SFC added:

“It would certainly be wise for the Scottish Government to assume that next year’s reconciliation figure will be in the area of £550 million rather than to hope that it will turn out to be very different, which is unlikely.”—[Official Report, Finance and Constitution Committee, 12 February 2020; c 20.]

As we know—as Professor Ulph has described already—the reconciliation figure of £309 million is just over half that £555 million forecast figure. You told us in February that you would be surprised if the figure were very different. It would be helpful if you could explain just what was going on in relation to the change between those figures. I see that—initially, at least—Professor Breedon is going to field that question.

I will give you the mechanics of what happened before moving on to the other issues.

The outturn for 2017-18 was very weak. When we were undertaking our forecasts, we projected that that weakness would continue. However, in fact, the 2018-19 data show that there has been a significant bounce back in tax receipts. We expected another relatively weak outturn but we got a significant bounce back instead, and that is why the reconciliation has changed—we did not initially predict that that bounce back would take place.

If we gave the committee the wrong impression, we apologise for that. We said a number of times that we were presenting a forecast, and that, therefore, it would not be entirely accurate. The point that we were trying to make was that, even though it was a forecast, it was something that should be considered in budget setting—that is, the budget for the year should be developed in light of the consideration of the fact that there might be a reconciliation. The forecast consideration is an important piece that should be considered because, although the reconciliation fell on this occasion, there may be another occasion on which the reconciliation will be seen to have become bigger when we get the final outturn data. I think that, in making that point, we incorrectly gave the committee the impression that we were firm about the data. I think that, on other occasions, we explained that it was a forecast. I apologise if we gave the wrong impression.

I was not looking for an apology; I was only after an explanation, and I think that you gave a pretty good one.

Figure 2 in your supplementary evaluation report deals with your forecast for income tax in 2018-19. It puzzles me a wee bit and it would be helpful to get some clarity on it. My main issue is to do with why the forecast error falls from 0.9 per cent in December 2018 to 0.6 per cent in May 2019, but then rises again to 1.6 per cent in February 2020. For the record, it would be useful if you could explain how the forecast error more than doubled in the period after the end of the tax year, given that much of the outturn data, especially that relating to the pay-as-you-earn system, would have been available.

The reason is fundamentally the same as the one that I just discussed: the rebound in the outturn data was much greater than we expected.

You are right to say that, in general, we expect the forecast—[Inaudible.]—and, sometimes, a bit of data can throw you off, even quite late in the day. I think that that is what has happened there. There is no guarantee that the forecast errors will always shrink as you get closer to the end of the process.

I am not sure whether you mentioned the RTI monthly data. David Stone might want to add something on that point.

We have been watching the RTI data closely since it first started being published. In 2017-18, we saw that the RTI data appeared to be a very poor predictor of outturn—there was quite a large difference between what the RTI data suggested was happening to income tax and what was actually in the outturn data. We were reserving judgment on that data and trying to understand more about what was driving the difference. I think that we have made that point you before; we also made it in our statement of data needs.

With regard to the 2018-19 data, we had RTI data published alongside that as well, and the RTI data appears now to be a better predictor—in relation to that year, at least. We now need to understand whether that is a one-off or whether RTI is a useful predictor for us in estimating changes in income tax outturn data. We are still talking to HMRC about that and trying to understand more about it. We would certainly like to use RTI data more in our forecasts, but we are still trying to understand more about that relatively new source of data and its relationship to what really matters, which is the outturn estimate of income tax.

10:15  

Thank you. After I ask this final question, I will go to Murdo Fraser.

Just for the record, because we all get caught up in the language that we understand, can you explain what RTI means, so that people watching the meeting at home can appreciate what it is?

I will hand over to David Stone again to answer that, because he is the expert on this.

RTI stands for real-time information, which is one of HMRC’s systems for collecting and processing income tax data.

Okay. Thanks, I just wanted to get that on the record so that people who may be watching the meeting know what is going on.

Good morning. I, too, have a couple of questions on your supplementary forecast evaluation report on income tax, just to get a little bit more clarity and background.

Figure 1.5, which is on page 9 of the report, looks at the amount of money that is projected to be raised from the introduction of the five-band income tax system, which came in in 2018-19. I have looked at the way in which you project income from that system, and there has been quite a substantial reduction in the forecasts from February 2020 compared with those from February 2018. The gap between the forecasts grows to a difference of £48 million in 2022-23 in relation to the originally forecast sum of £267 million, which is roughly 20 per cent of the original forecast and is quite a substantial drop in the figure that you originally forecast two years ago. Can you give us some background on that and explain why you have revised down your projections so substantially in relation to that new income tax banding system?

I apologise—my phone went off just then so I missed part of the question. I will hand over to David Ulph, who I hope heard the whole question.

There are a number of different factors. One is that, part of the data issue that we have been talking about means that we have a better understanding of what the data outcomes have been. Also, the growth of income tax revenue has not been quite as strong as we initially thought that it would be, which contributes to that effect.

We also started to get more understanding of some of the impacts of distributional factors on the link between the growth of income and the growth of tax revenue, and perhaps we had not fully taken account of the fact that having a tax system that is a lot more progressive would, in and of itself, start to contribute to the growth of tax revenue in Scotland. Our understanding was perhaps not as full as it is now of what those effects were likely to be. However, David Stone might have a better understanding of some of the detail of the changes in forecast that Murdo Fraser referred to.

I am happy to talk about that for a bit. We are constantly updating and revising our forecasting models and, as a natural part of that process, the policy costings will move around a little bit. With this one in particular, we are comparing the policy costing forecast from February 2020, which, as we have seen elsewhere, is when we revised down the overall level of income tax revenues in Scotland, following the publication of the 2017-18 data, which was lower than we expected. By bringing down the overall level of income tax revenues in our forecasts, we also revised down the policy costing a bit. That is probably not the only factor; there will be some other changes within the model as well as we revise things such as the number of taxpayers in different bands.

I would not say that there was any one big factor behind the revision—apart from the outturn data, perhaps. It is certainly not a new view on behaviour change, for example, which works in fundamentally the same way in this updated costing as it did in the previous costing.

Thank you. I have a follow-up question that links nicely to the number of taxpayers in Scotland, which you referred to and which is referenced in figure 1.6. There was a quite substantial error, as you acknowledge, in the overall number of taxpayers, which is 33,000 fewer than the number that you previously predicted. That seems quite a large number. Can you give a bit of the background to that and explain why you think that the number of taxpayers has fallen rather than increased, which is what you originally thought would happen?

The key thing about the number of taxpayers relates to those who cross over from not paying tax at all into the tax system. Clearly, changes to the tax system have changed the number of people who do not pay tax at all. That is where the changes have mainly been. Given that, a fall in the number of taxpayers is entirely what we would expect. I hope that that answers your question.

It does to an extent, but it accounts for only half the total, does it not?

Generally, the number of those who cross into—[Inaudible.]—tax—[Inaudible.]—barriers is pretty hard to pin down precisely, because small changes in people’s circumstances can take them into or out of paying tax. I am not sure how precisely we ever estimated the number of taxpayers, particularly in that lower band. From our point of view, that group does not contribute much to the tax receipts. Knowing whether someone pays a small amount of tax or no tax at all does not figure enormously in our forecast.

David Stone might want to add to that.

Sure. I pulled out a figure the other day that shows that 268,000 people in Scotland were in the starter rate band in 2018-19, which is 10 per cent of all taxpayers. Those people are within £2,000 of the personal allowance, so even quite small changes in the personal allowance can reclassify them from being a taxpayer to not being a taxpayer. Our forecast of the total number of taxpayers is sensitive to that. Similarly, around the higher-rate thresholds, a lot of people are within that range, and even small differences in our estimate of what a taxpayer’s income is and how that changes relative to the higher-rate thresholds can reclassify taxpayers from one band to another.

It is also important to recognise that, because of the way in which we construct our forecast—we give a detailed forecast of individual taxpayers’ incomes—whether they are just above or just below the higher-rate threshold does not have a massive effect on income tax revenues and their income tax liabilities. However, it does mean that we can get some volatility in the way in which we classify them between one band and another.

Thank you.

Professor Breedon, I have some questions on social security, but I appreciate that they might be for Ms Murdoch to answer. It is still early days for the Fiscal Commission when it comes to social security forecasting, but I would be grateful if you could plainly explain the implications of the Fiscal Commission getting forecasts wrong; how you would rate your performance thus far; and whether you think that it is advisable for contingency to be built into budgets to take account of potential forecasting errors.

You made the point that it is early days. We are very much in the learning phase on social security and how things work. When a new benefit or policy is introduced, the first year is difficult to forecast, because there are lots of things going on. As things bed down, forecasting becomes easier.

Similarly, when new benefits come in, because we have learned from the ones that have already been introduced, we know a lot about what take-up we might expect. We are hoping that our performance will improve, but we will always have the same problem when a new policy is introduced, which is when the biggest errors are likely to occur.

On the issue of contingency, you make a very fair point. The budget process is difficult, and demand-led benefits are obviously—[Inaudible.]—is there and has to be managed, so I think that it is sensible to have some buffers in the system to allow for those errors.

Professor Ulph has said that he would like to contribute. Have I got that right, David?

You have indeed; thank you.

To respond to the first part of Ms Constance’s question, we took the learning process very seriously. We made some quite big mistakes in our early forecasts on the best start grants, and we learned from that that we must understand in a lot of detail exactly how much resource the Scottish Government is putting in to promote and advertise new benefits. We listen carefully to people from the Scottish Government and Social Security Scotland about their plans for all the new benefits, and we think extremely carefully about how to incorporate all that information into our forecasts. For the Scottish child payment, for example, we increased our forecast of the take-up rate to 80 per cent, which is very high. We did that partly because of some of the effects of Covid.

Following on from that, we know that we are on the brink of economic upheaval, so it is a reasonable assumption that there will be an increase in the number of people who are eligible for low-income benefits. It would be good to understand how all that is factored in.

I am aware that at times there is a reliance on Department for Work and Pensions figures and that some of the comparable DWP benefits have lower uptake. The Scottish Government has a legal duty to prioritise and promote uptake. The DWP does mid-term alignment, whereas I do not think that the Scottish Government does. How is cognisance taken of all those factors when the Fiscal Commission develops forecasting models for new benefits?

We are already thinking about the impact that Covid will have on various benefits. We will need to look at that carefully when we do our next forecasts. You are right. As David Ulph said, one thing that we have learned is the importance of thinking about not only the policy but the communication plan for advertising a benefit or whatever to the public. Understanding how much money has been spent on a comms plan is a factor that we now integrate much more into our forecasting process. We discuss what the comms plan is for a particular policy, how quickly we expect a benefit to be taken up and what level of take-up we expect.

On the more technical question on mid-term alignment—[Inaudible.]

You broke up a bit there, but I think that you were trying to say that Claire Murdoch would answer that question. Claire, do you want come in at this stage?

10:30  

Yes. I will come back on a couple of those points.

Angela Constance mentioned the Covid-19 effect. In September, we published an updated forecast on the Scottish child payment. That was our first forecast since February, and we were able to factor in the effects of Covid-19. We have assumed that, because there is a higher eligible population, more families will receive universal credit and will therefore be eligible for the Scottish child payment. We think that, in the next financial year of 2021-2022, Covid-19 will lead to £15 million of additional spending on the Scottish child payment. We will also consider the effect of Covid-19 when we produce our updated forecasts for the best start grant, the funeral support payment and other low-income benefits.

It is worth pointing out that although the effect on those benefits is significant for people, the effect in cash terms on the Scottish budget is not that large. For example, disability benefits account for a much larger proportion of spending. We think that there might be some Covid effects there, too. The DWP is changing some of its processes. There has been a reduction in the number of claims for some disability benefits, but we think that some of the changes that the DWP has made will increase spending. We will factor all that into our forecasts.

With regard to the question about using DWP information, Angela Constance is absolutely right. When we made our forecast for some of the benefits, we had to look at what was happening in the DWP’s systems before those benefits were devolved. The best start grant replaced the sure start maternity grant. We looked at take-up of the sure start maternity grant and assumed that take-up in Scotland would be higher. As David Ulph said, our assumption was not high enough, so we subsequently increased the take-up rate. We look at the systems and we think about what the Government is doing. We are learning from that process, as David Ulph said.

We now get more regular information. We are getting information on spending as we go from Social Security Scotland and from the DWP, and we will consider how we can factor that into the forecasts that we will publish in December.

I have covered a lot of ground, but I think that David Ulph has more to add.

You did cover a lot of ground; maybe David could provide a shorter answer.

I have a short answer that goes back to Angela Constance’s point about comparison with what is happening in the rest of the UK. We know that there has been a big increase in claims for universal credit—claims have doubled. People will ask whether that will lead to a doubling of the uptake of some of the benefits in Scotland.

We looked at that carefully. As far as we can see, a lot of the additional claimants for universal credit are people who do not have children, who would not become claimants for some of the Scottish benefits. Therefore, we do not think that it would be appropriate to increase the number of potential claimants for some of the Scottish benefits by anything like the amount by which the number of universal credit claimants is going up. We have looked at the issue carefully.

The information about who is likely to take up universal credit is interesting. The only question that has not been answered is the one that I asked about what the implications are when social security forecasts are wrong. What happens when you get it wrong?

I am afraid that that has to be managed in the budget. It is a difficult process. It will be a particularly difficult process this year; we appreciate that.

Indeed.

Angela Constance has covered a lot of the territory that I wanted to ask about. On a pragmatic level, an issue that you faced was a delay in launch dates being confirmed and communicated to you. That is relatively easy to fix. What practical arrangements do you have in place with the Scottish Government to ensure that you receive the earliest possible notification of such matters? Obviously, early notification leads to more accurate forecasts.

That is a good point. As has been said several times, we are all learning about the process. We are developing our relationship with the Government when it comes to talking about introduction, timing and—as I have mentioned—communication plans, and things have improved a lot. We have a good working relationship there.

Even in the best of times, launch dates can be a little bit variable because of events that are beyond anybody’s control. We certainly hope that we have a process that means that there will be better communication about that, but even in the best of worlds, those dates can change for reasons that neither side can necessarily predict. That will be an on-going issue. New benefits are difficult even without that uncertainty about launch dates and we appreciate that that will continue to be a potential issue even with the strong relationship that we have.

Absolutely. If we can drive out those forecast errors, it will have less impact on the budget, which is the significant consideration for us.

Before I move on to another issue, my final question on social security is about the assumptions that you made for the fair start Scotland programme. Were they based simply on data that was given to you by the Scottish Government or did you look at programmes elsewhere? There was quite a difference between the outturn figure and the forecast figure that you had in place.

You are right; we took data from a variety of sources. We had very detailed discussions with the Scottish Government and with other sources, and we made our own judgments on the basis of all that information. Given what we experienced, we would probably make a better judgment the second time around, but that is part of the process of learning about issues such as take-up, which is very hard in the first year of a programme. It was particularly hard in that case, as that was pretty much the first programme that we looked at. We tried to get all the information from all the sources that we thought would be helpful. Now that we have had the experience of seeing a Scottish benefit being introduced, we will have a much better information base on which to make those judgments.

I go back to the territory that was explored by the convener, which was to do with Scottish income tax. As it currently stands, the cumulative impact of forecast and known reconciliations for the past three years from 2017-18 is about £700 million. That will have a negative recurring impact on the budget going forward. I listened carefully to the response about the potential application of real-time information, which it is clear will make a difference to modelling. Will that enable you to revisit the forecast for 2019-20 or is it applicable only to future forecasting?

It is applicable to future forecasting.

So we stuck with your estimate for 2019-20 as it stands.

Yes.

One of the reasons for negative reconciliations is weaker performance in the Scottish economy relative to that of the UK. Your report helpfully explains why growth in Scotland differed from your forecast, but I am interested to know whether you would make any analysis of why growth was lower in Scotland than it was in the UK as a whole.

That is quite a big question. There are two big factors, one of which is to do with the fact that population growth and demographics are different in Scotland. That is one of the contributors. The other one is to do with the oil and gas sector. Investment in that sector has had an impact on other sectors around it. Lower oil prices have led to lower investment, which has had an impact. Those are the two bigger factors that I would mention in relation to the difference in growth performance.

I refer members to my entry in the register of members’ interests in the context of business rates.

My question is on non-domestic rates, the error on which—£24 million—was one of the smaller errors in the forecast. That is a relative error of 1 per cent, which in forecasting terms would normally seem pretty good. However, unlike other taxes, which are based on aspects such as the number of taxpayers or turnover, NDR is a tax that is based on floor space or property space, which is one of the most tangible metrics going. Professor Breedon, could you provide an explanation of the NDR error? Is it down to a small number of high-value properties being demolished?

As I recollect, the issue is often the appeals process—the timing of appeals is difficult—but I will hand over to David Stone, who might be able to add to that.

I would be happy to. Francis Breedon is right. As a tax base, NDR is very stable, because it depends only on the amount of business floor space, which makes it easier to forecast. However, we cannot forecast how any individual appeal will turn out; all that we can do is to look at the history of the appeals process and try to judge what effect appeals might have on NDR revenues in the future. Therefore, although we had an aggregate forecast error of £24 million, our error as regards the total effect of appeals on NDR revenues was £29 million. That more than accounts for all of the error in our forecast.

As on previous occasions, you have focused on the appeals element, but there are business properties that are being destroyed. Surveyors in Aberdeen have compiled a list of properties that are about to be demolished, which they have put into three categories. The first of those consists of properties that it has been formally declared will be destroyed, which have a total rateable value of around £11 million. That will result in a £6 million annual loss in rates. Surveyors in Aberdeen are aware of further business properties with a rateable value of £10 million that are about to be destroyed. Beyond that, the potential exists for another group of properties with a rateable value of £10 million to be destroyed.

How well sighted are you with regard to those three levels, which surveyors on the ground see, when it comes to predicting destruction of floor space and, therefore, of the tax base?

It is fair to say that we have not looked at that in detail. As David Stone explained, our error is—[Inaudible.]—it was coming from, and we focused our effort on that. You have given us a useful insight, and we will certainly consider whether we should look into that data in more detail to help us to improve the forecast.

David Ulph and David Stone both want to come in. I will leave it to you to choose who should do so.

I reckon that David Ulph should go first.

I want to reinforce Francis Breedon’s point—we do not spend a lot of time looking at the destruction of rateable value. Alexander Burnett has given us a useful insight that we should focus on in future.

We have thought about the appeals a great deal, because we know that they are a big source of error, but there is far less that we can do about that because, inevitably, how appeals turn out is quite a random process. We know something about the appeals that are in the system, but it is very hard to predict when and how they will be settled. Alexander Burnett has raised a useful point.

Well done, Alexander. Is there anything else that you would like to pick up on?

No. That is fine—thank you.

Good morning. I would like to ask about the future forecast modelling that the SFC will use for the current financial year and beyond. Given the fundamental impact of the pandemic on the tax system and policy measures such as the job retention scheme, how will the SFC change its methodology and modelling to forecast tax revenues for this year and beyond?

Given the significant uncertainties that are involved, is there an increased risk of forecast error? That would be understandable. Given the uncertainties, is there a concern that forecasts for the current year might be subject to increased uncertainty?

10:45  

Yes, that is a fair point. It is a difficult period for forecasting and, inevitably, forecast errors will be very large. The forecasts that we made for what this year would look like will be very wrong, because we did not factor in the pandemic. Even after the pandemic, the speed of recovery will also be hard to forecast.

There is a point that members know about, but which is worth reinforcing: for the framework, as well as for our forecast, the errors that are made about the rest of the UK in the BGA forecast matter. On this occasion, we and the OBR will have made very similar errors. Going forward, the important thing is how Scotland performs relative to the rest of the UK and how our forecast performs. Although very big errors are going on, we have had a few occasions when the error from the BGA and our errors have been in different directions. On this occasion, we are pretty sure that they will be in the same direction.

A couple of members have supplementary questions. I will let Dean Lockhart conclude before I bring in Patrick Harvie.

Thank you, convener. I have a brief supplementary question, which might be difficult for the SFC to answer now. Is operational data available for the time being, such as uptake of the job retention scheme, that would indicate whether the impact of the pandemic on the Scottish income tax system might be less or more severe than elsewhere in the UK? Or is it too early for the SFC to have an indication of that?

From the data that we have so far, our judgment is that the economic impact has been relatively similar but, as you pointed out, it is still early. We are still monitoring that carefully. We have significantly increased our use of real-time data to spot differential impacts, because that will be important for the framework. At the moment, our judgment is that we are not yet discerning important differences, so the impact is slightly reduced.

We are certainly making as much use as we can of the high frequency data and the new data that is coming out, including the data on the job retention scheme. We will use that as much as possible in our forecast in order to create a baseline.

However, the critical point is that we are trying to understand the effect of the pandemic on the economy and basing our forecasts on that, but, of course, the pathway of the pandemic from this point forwards is highly uncertain. We cannot predict how the pandemic will turn out over the next few years, and the outlook for the economy is very much tied to the outlook for the pandemic. Therefore, coming back to the start of that discussion, forecast errors are likely to be larger.

Thank you. That is a very helpful response.

I recognise that we have moved on to looking ahead rather than looking back, which was the main part of the session, but, as some of the witnesses will remember, I have previously asked questions about how behavioural effects are worked into future forecasts and projections, as well as assumptions about behavioural effects. Given that we are living through a year unlike anything that any of us can remember, surely it becomes not just difficult but impossible to work out in advance any kind of reliable forecasts about people’s behavioural changes, in relation to work or other aspects of the decisions that they make, which will impact on the budget, such as social security take-up.

You said that the forecast error is likely to be larger, but that is just a consideration of the direct economic impacts of the pandemic. Can you convince me that there is any remotely meaningful way to understand behavioural effects in a completely unprecedented year that is affecting everyone’s sense of security, mental health and willingness to make changes in their lives?

That is a fair point about behavioural effects. The pandemic is resulting in behaviours that are very unusual and hard to forecast. All aspects of forecasting, including behavioural impacts, will be very hard to manage in the next few forecasts. We are trying to make the best forecasts that we can, so we will try to allow for behavioural impacts when we think that we can. Equally, we will continue to look at the evidence from other countries and historical data from Scotland to see whether our assumptions are borne out. If the historical data bears out those assumptions, we can be more confident about using them in the future.

However, it is a fair point that a lot of forecasting is based on historical experience and we are going through something that has rarely been experienced in history, so we do not have much to go on with a lot of these—[Inaudible.]

We lost your sound a wee bit at the end there, Professor.

I will not pursue it any further. We can move on.

I have a supplementary question. When someone says to me that large forecast errors are expected—for understandable reasons—it begins to start alarm bells ringing about the potential level of spending available to the Scottish Government. If we have forecast errors that are larger than normal and the OBR forecast error is in the opposite direction to the forecast errors of the SFC, we are potentially looking at a serious situation for Scottish spending. How can the committee reflect on that and ensure that we are in a situation to dampen anything that goes off the scale and creates a significant difficulty? I appreciate that that is a big question, but it needs to be asked.

Although we are anticipating big errors, as I said, the likelihood is that our errors will be of the same style as those of the OBR. That is why, as I said before, trying to work out whether the impact in Scotland is different to the rest of the UK is very important—in some sense, it is as important as the overall impact. There will be situations where the errors compound each other, but this is more likely to be a situation where the errors offset each other somewhat. It will be difficult for all parties, even bearing in mind that concern. If we get a differential impact in Scotland, that will be quite difficult to manage.

I want to reinforce Francis Breedon’s point. What matters for the Scottish budget is the difference in performance between the Scottish and the UK economies. So far, at least in aggregate terms, they have performed in similar ways. There are some details that could matter at some point. However, we have been looking at this quite carefully and we do not see any significant difference at the aggregate level so far.

This year, the two errors have compounded one another, but that might not be repeated in future years. That is not a forecast—that is just a view for you maybe to reflect on.

All the signs at the moment are that the two economies are going in the same direction. The OBR is facing exactly the same challenge that we face in forecasting. If it is making errors in one direction, we are quite likely to be making errors in the same direction. I reiterate Francis Breedon’s point that it is extremely hard to forecast what could happen over the next year or so because of the way that the pandemic will track.

There seems to be a lot of hope in that—[Inaudible.]—which I understand.

I want to return to some of Murdo Fraser’s questions, on which I will ask a couple of supplementaries. He talked about the five-band income tax system in figure 1.5 in the “Supplementary Forecast Evaluation Report: Income Tax”. The comparison was between February 2018 and February 2020. Will you clarify whether the figure for February 2020 is the actual figure? It was quite a pessimistic forecast, so is the £48 million difference for 2022-23 the final figure?

I ask David Stone to answer that.

We will never have a final figure for a policy costing; we can only ever estimate the effect on the economy. The figures provided for February 2020 are our latest estimates.

We will never have a final figure because we are trying to compare how much we know has been raised in tax revenues with how much would have been raised under the previous system. We can never know that for certain; we can only estimate it. Now that we have final outturn data, we could come up with a better estimate for the 2018-19 policies using that. We could look into that for our next report. Those figures might change again a bit, but I would not expect them to change radically.

Okay; thanks. I should have said “latest” rather than “final”, but that is helpful.

Murdo Fraser also asked a question about the number of taxpayers. Is there any evidence that, for example, taxpayer numbers have fallen because people have incorporated to avoid income tax?

No—I do not think that we have evidence for that. As I said, most of the changes have been as a result of people coming into the starter band. Unless one of my colleagues would like to contradict me or add anything, I do not think that incorporation is an important factor.

I am not sure whether David Stone wants to contradict you or complement what you have said.

We get a fair amount of detailed modelling and numbers from HMRC on its estimates of incorporations in Scotland, which is factored into our forecasts. At this point, I cannot say whether those numbers have changed much, but I do not think that the scale of or variation in incorporations is enough to explain the numbers. The primary factor explaining them is small differences between our forecasts of the effect of the personal allowance on the very lowest earners and what happened in practice. That is the major factor behind it.

That is helpful. We have already talked about real-time information. Will you clarify whether that applies only to PAYE income tax payers, or whether it includes people under self-assessment? How accurate are we when it comes to self-assessment?

That is another one for David Stone, I suspect.

RTI relates only to PAYE information. It is important to understand that people who file with self-assessment may have some PAYE returns as well. We might get some information about the SA community from PAYE, but not all their tax comes through PAYE. Therefore, RTI is only a partial record of the income tax that is paid in Scotland. We do not really get a final figure for the self-assessment population until we get the outturn data. There is no timely way of monitoring self-assessment returns until we get that final outturn figure.

Even within the PAYE information that we get, as I mentioned, we saw a difference between what RTI suggested was happening with PAYE income tax revenues and what the outturn data said about it when those were published. Therefore, even within PAYE, there is a bit of a gap between the two. We are still trying to resolve that, so that we can make better use of the data.

11:00  

We have already talked a bit about the relationship between the SFC and OBR forecasts. Do you base your forecasts on the same information? Your analysis is that £83 million of the forecast error was attributable to your organisation and that £207 million was attributable to the BGA forecast error. Does the OBR agree those figures?

We have a good relationship on data with the OBR; we share information and a lot of the same approaches, which is helpful because, as we have discussed, it is the difference between our two forecasts that matters.

We have tried to give you illustrative breakdowns of where the errors come from, which mathematically fall out. We could break down the errors in different ways to show different information, because there are a lot of moving parts, such as the— [Inaudible.]—data and the growth rates in the two forecasts. How the data is broken down is up to judgment, but the way in which it is broken down is not subject to dispute.

The final area that I want to ask about is the statement of data needs that you produced. I understand that the Organisation for Economic Co-operation and Development visited last year and said that system is one of the most complex that it knows of in the whole world.

I am interested in the previous improvement requests that you have made, which are listed in annex A of the statement of data needs. For example, the table shows that the request on “More timely and detailed data on Scottish wages and earnings” is “High” priority but that there has been “No progress”. Will you comment on that?

In general, the flavour of the statement of data needs is that a lot of progress has been made on a lot of the areas that we have asked about in the past. The earnings data is an outstanding issue. I do not think that that is anyone’s fault; it is just a difficult issue to resolve. However, for us to make our projections, it is an important bit of data, so we still want progress on that.

On your other comment—[Inaudible.]—wants to raise on the data needs, I would say that, this time around, it is incumbent on us and the Scottish Government to try to explain as simply as we can what is going on in the Scottish budget, because it is always complex, and it is becoming more complex as more moving parts are added. As I said, we wanted to highlight in the data needs document that we will all have to work hard to explain what is going on. I hope that we will manage to play our part in that.

As well as giving us some insight into what is happening with income tax revenues, the RTI data tells us something about employment and earnings in Scotland. RTI has a lot of potential to tell us more about earnings in a timely way. HMRC and the Office for National Statistics have been working hard on developing that as a source of information, but it will take some time. It is quite new, and we have been talking to them a fair bit about what they can publish and when, but the advent of Covid had an effect on some of the publication plans. We have hopes and expectations for RTI, but it will take some time to come through.

My final question is, again, on the statement of data needs. It suggests that, for the review of the fiscal framework in 2022, there should be consideration that the Scottish Fiscal Commission

“should be given a statutory right of access to information held by UK Government departments”.

Does that mean that you are not getting information at the moment? Is a statutory right to access information the way that we should be going?

As we have said, the relationship with all those departments is really good, so it is not a criticism. A statutory right would make getting data an easier process. We are not critical of the relationships that we have; we are simply saying that it would make our job a lot easier if we were to have that statutory right. That is why that comment is in the statement.

You have outlined some of the challenges in making predictions for 2018-19. What lessons do you draw from that about making predictions for 2021-22?

As that is a difficult question and I have answered quite a few, I will hand this one straight over to David Ulph.

One of the lessons that we learned in 2018-19 is that we probably overadjusted our forecast too much on the basis of the outturn data for previous years, so we have probably learned to place a bit less weight on just one year’s data. We have systematic discussions all the time about whether we should consider data over a longer period and about how much weight we should give to the most up-to-date data. That is an on-going discussion in relationship to all our forecasts. We certainly learned a lesson from that exercise.

Different types of risks bedevil the forecasting exercise from here on in that the ability to learn much from past episodes will be limited.

If you cannot see me, it is because a cloud is going past the window.

You have alluded to some of the difficulties that there might be in making forecasts. There are two large known unknowns: Brexit and Covid. If you look back at the 2018-19 forecast, can anything at all be done to try to make your prediction exercise flexible or deft enough to cope with those two unusual situations?

Not a great deal can be done. Obviously, we are waiting for details on Brexit, as is everybody else. We are forecasting on broad-brush assumptions of an orderly exit, but we hope to get more detailed information in the next few weeks.

With regard to the pandemic, we are, once again, having to draw information not so much from recent history, but more from the consideration of other countries and the more detailed research that people have done about the past and so on. This is a difficult forecasting period for that reason.

I do not see any other requests to speak in the chat box, so I warmly thank Francis Breedon, the two Davids and Claire Murdoch for their evidence today.

Meeting closed at 11:08.