Finance and Constitution Committee

Meeting date: Wednesday, March 13, 2019

Official Report 603KB pdf

---

Scottish VAT Assignment

Good morning, and welcome to the sixth meeting in 2019 of the Finance and Constitution Committee. As usual, I ask everyone to ensure that their mobile phones do not interfere with proceedings.

Agenda item 1 is a round-table evidence-taking session on Scottish VAT assignment. I warmly welcome to the meeting Mark Taylor, assistant director, Audit Scotland; Charlotte Barbour, director of taxation, Institute of Chartered Accountants of Scotland; John Cullinane, tax policy director, Chartered Institute of Taxation; Professor Graeme Roy, director, Fraser of Allander institute; John Ireland, chief executive, Scottish Fiscal Commission; and Dr Paul Mathews, senior analyst, Office for Budget Responsibility. I am grateful to you all for coming along to help us with an area that might, on the face of it, seem technical but which, when you start digging, becomes quite complicated. I am sure that you will be able to give us a significant amount of help in this morning’s discussion.

I intend the discussion to be as free flowing as we can make it, but it has been structured around three themes. If you want to contribute, please catch either my eye or Jim Johnston’s eye, and we will make sure that you can get in and say your piece. We need everyone’s contribution to be as successful as possible.

The three areas for discussion are: first, how VAT can be assigned to Scotland effectively; secondly, the robustness and transparency of VAT assignment methodology; and, thirdly, VAT forecasting and risks to the Scottish economy. It is inevitable that our discussion will flow across those areas. If I feel that we have exhausted an area, I will move on to the next.

I invite Willie Coffey to start the discussion on our first theme of how we assign VAT effectively to Scotland.

Good morning, everyone. As we know, the first 10p of the standard rate and the first 2.5p of the reduced rate are to be assigned to Scotland using a fairly simple estimation of the total VAT take across the United Kingdom. I want to ask a few questions to start off our discussion. First, could VAT be assigned to Scotland in a more effective or, indeed, more accurate way? Secondly, how much of an influence could it have on policy development by the Scottish Government? Finally, is it fair to hold the Scottish Government to account for any fluctuations in VAT receipts when we have no levers to influence any of this?

Right—that is your starter for 10. Who would like to kick off by answering that very simple question? [Laughter.] Does anybody want to respond? I see that Graeme Roy wants to come in. Thank you very much, Graeme. I was going to break Mr Coffey’s question down a bit, but go for it.

There are a couple of things to address in that question. First, we know that one of the big challenges with assigning VAT is how to come up with a robust estimate. In the papers that have been published so far, we see what you rightly call a relatively simple way of doing that, which is based on our share of consumption across the UK. However, even with that approach, the potential for variation in the estimates is still significant. If you look at the numbers in “Government Expenditure and Revenue Scotland”, you will see that they are plus or minus a couple of hundred million pounds simply because of statistical uncertainty. There is a big question about the ability to get a pinpoint accurate estimate of VAT revenues in Scotland.

I guess that that comes back, first of all, to the complexity of the VAT system, which is much more complex than the income tax system. Secondly, it is much more complex to estimate VAT revenues. Because you have to worry about what people are spending from day to day across a wide variety of products, you cannot have the same level of certainty or the same accuracy of estimate as you can have with income tax. The question for Parliament, therefore, is about the level of risk that it might want to take on, given the level of uncertainty that exists purely from a statistical point of view—never mind as a result of volatility in the economy.

On your second question, which was about the ability to be accountable for or to grow VAT revenues, the principle behind the assignment of VAT is the correlation with the strength of the economy. In other words, if the economy does well, you benefit; if it does worse, you are accountable for that. That raises questions about the levers that the Scottish Government has, first, to control the economy and, secondly, to control changes in VAT revenues. Are there any levers that we can use to encourage people to spend more on certain vatable products? Do we want to do that, or do we want to encourage people to spend more and save less? There are a host of different issues around your ability to use your levers to move VAT revenues, but that, again, links back to the complexity of estimating these things.

From the outset, ICAS has said that we need to be careful about the consequences that flow from this as to whether it would lend itself to the Scottish Government looking at the kind of industries that generate VAT. Because the food industry, for example, tends to be zero rated, it might not go there. However, I am not sure that the correlation between individual policies and this particular VAT assignment model is that strong. I would have thought that the issue is much broader than that and that it is more about the general strength of the economy. That assumes, of course, that the economy equals spending, and it might not.

There are two aspects to this, the first of which is the fluctuation and volatility that are due to policy intentions. If there were more VAT receipts in Scotland due to the strength of the Scottish economy—let us ignore, for a moment, the problem with estimating these things—that would be something that was intended; in other words, that would be what you would get if you had a fully devolved tax. Whether something is fair is one question; whether you want it to happen is more a political decision.

One issue that I have seen nothing about—although Charlotte Barbour has just alluded to it—is the distributional side of things. One issue in the discussions about the Scottish rate of income tax was the progressivity of the system. Generally speaking, the belief is that, if incomes are spread more evenly, there will be more consumption; on the other hand, some of that consumption might be of food, which is zero rated for VAT. I have not seen an awful lot of analysis of how those policies might fit together. If policies are adopted that boost the Scottish economy and therefore spending in a general sense, you will expect to get more VAT receipts. However, as I have said, I have not seen much analysis of that broken down either distributionally or otherwise.

Of course, the other side of that is the unintended volatility that arises as a result of our having to use lots of estimates. I tend to take the rather simplistic view that it does not matter so much if the estimates are inaccurate on day 1 as long as they constantly remain inaccurate in the same way. After all, you will just keep giving up the same amount of block grant. Nothing will change; so, in one sense, what does it matter?

To be honest, this is a statistical issue—it is not really an area for tax specialists. I did not see much in the official papers about the volatility of the estimates, but some of the comments about them seem to suggest that they might be quite volatile indeed, and I get the impression that it is an area of risk for the Scottish Parliament.

09:45  

As the committee well knows—it has spoken about this on a number of occasions—ultimately, what matters is the interaction between the UK tax take and the Scottish tax take, however it is assessed. There is something about how those UK policies play out in Scotland and how Scottish Governments and Parliaments can match, or diverge from, the policy perspective at the UK level. It is about the interaction of those two things.

Clearly, there is no ability to vary tax rates or the tax model, but the Government can do other things around encouraging economic growth. At a step removed from that is consumption and how it flows through to the VAT model.

I reiterate a point about the estimation process, which I know we will come back to. It is essentially a new part of how the fiscal framework operates. There have been small degrees of it elsewhere, but, in essence, the whole adjustment is based ultimately on an estimate, which is fundamentally different from what the position was before. That raises questions, which the committee will want to get into, about people’s trust in the estimation process and the ability of all parties to rely on something that they know is an estimate.

I emphasise that, if something is consistently wrong, that is less of an issue. The challenge is in knowing whether it is consistently wrong. It is about the extent to which the volatility sits on both sides of the equation—the block grant adjustment and the tax take in Scotland—and how much of that is driven by the statistical estimation process, to pick up on Professor Roy’s point. The issue is the extent to which those two volatilities are correlated and whether they cancel each other out or potentially build on one another. Because the tax calculation basically works out what the UK position is and what the Scottish position is and takes away the Scottish position from the UK position, there is a greater danger of such amplification in this process than there is elsewhere.

Given that the SFC will have to forecast on the back of estimates—if I have understood how this will work in the future—how concerned should we be that we will never have any real outturn figures, from a Scottish perspective, on VAT?

There are two dimensions to that question. The first issue, which is not for the commission, is the political acceptability of that and people’s trust in the assignment data and model. We can talk about the technical aspects of that later.

From our perspective, it is more of a grey continuum. The difficulty that the assignment process for forecasting presents is that you can imagine that there is underlying VAT revenue for Scotland—that is the conceptual thing that you are trying to work with—but, because you have a statistical model for its estimation, you introduce noise. That noise is hopefully, but not necessarily, random—it moves around. The job of the commission will be not just to estimate the underlying trend but to forecast the random error in the assignment model, which is very hard. The simplistic answer is that, if you have random noise and something moving randomly, the best estimate will take the average. However, on a year-by-year basis, guessing or forecasting the noise that is in the model will be difficult.

The conversation is already identifying problems with the estimative model. You talked about basing an alternative approach on outturn data. From a practical point of view, what would be involved in that? Clearly, it would mean creating separate Scottish VAT points for businesses, which would be a major burden on businesses. Has anybody done any thinking about what the impact of that would be on businesses and what the cost would be of taking that alternative approach?

Our membership has always been adamantly against that. You see the problems that there are in disentangling with regard to Brexit—pulling Britain out of a European market. We would have similar problems here if we had to set up a Scottish VAT network. The whole point of VAT is that it is designed for a single big market, because it goes in and out with production, adding value as it goes along. In terms of broad principles, members in business and in practice would be very much against the imposition of all that administrative work. Even then, I am not convinced that you would truly nail down what was Scottish VAT. In the immediate term, businesses are under pressure on other issues—whether Brexit or making tax digital for VAT—and I do not think that they would at all welcome an approach that was based on outturn data.

I do not know whether that is of any help.

That is the response that we would expect in the present circumstances. I was not suggesting that we should use that approach; I was asking whether, given that we cannot do that, we should be concerned about whether the system that will be used is robust enough.

I was toying with that very problem as I came down the road this morning. We would not want to account specifically for Scottish VAT. We see the problems with the assignment model, but I think that, ultimately, that model is better than trying to identify true outturn figures.

I have just one point to make about that. It would be understandable to think that an outturn-based approach might be very burdensome but that it might produce a more accurate result. However, in one respect, I do not know that it would give an accurate result, because it would be necessary to design rules that said what Scottish VAT was. The place where supply is deemed to take place for VAT purposes has been redesignated from time to time, even within the European Union. When the Scottish income tax came in, an approach to defining Scottish residency was adopted that was different from the approach to defining UK residency, for a mixture of practical and other reasons. Therefore, I do not think that we would end all discussion and that people would say, “Ah—there’s the simple answer: that’s a Scottish receipt.” It would be necessary to consider and design a whole set of rules to define what a Scottish receipt was and then make businesses follow it. An outturn-based approach would be burdensome and would not end all the argument about whether we had got it right.

That leads into the area that Patrick Harvie is interested in. It might not be the right time to go there; if it is not, please forgive me.

Is there ever a right time?

I am interested in why we are going down the road of assigning VAT. As I said to colleagues before the meeting started, I was on the Smith commission and, although I did not argue against the proposal to assign VAT, I was never a wild enthusiast for the idea that simply assigning a proportion of VAT would achieve any objectives. I understand the desire that we had to have a big number for the total proportion of the Scottish budget that came from taxes that were in some way under the control of the Scottish ministers or the Scottish Parliament, but, in the absence of the ability to use policy levers to change the rates or bands or to redesign VAT—for example, to incentivise or disincentivise different kinds of consumption to achieve social or sustainability objectives—and in the absence of a great deal of control over things such as how much of the money that people spend goes through shops on high streets in Scotland as opposed to Amazon accounts, what will we achieve through the assignation process? If it has ended up being more complicated and less convincing than we thought it might have been, should we step back and think again at this point? There will be a review of the fiscal framework. When that takes place, should we review the assignation of VAT as well? In the light of what happens in the next few weeks in UK politics, for example, should we decide whether a different approach is desirable or achievable?

Let us look at what the Smith commission was trying to do. Mr Harvie was on it and I was not, so I will not try to second-guess that.

I apologise again. [Laughter.]

As I understand it, the Smith commission was trying to build in accountability for the performance of the Scottish economy having an impact on the budget. That is fine in principle, but, when we look at the practicalities of the methodology for using VAT to do that, we find that there is a great deal of uncertainty within the statistical measures. There is a big question mark over whether, even if the Scottish economy was changing relative to the economy of the rest of the UK, that would be reflected in the VAT revenues that were received in Scotland. Given the uncertainty about the calculations, it could well be the case that that does not happen. There is a fundamental question about whether the link on accountability is achievable with the assignation of VAT, given the complexities of that process. The principle of assigning VAT is solid, but there is an issue with the practicalities of doing it, given the costs and the risks involved.

A second question that the committee might want to reflect on is about the transparency of the system. We have already seen the challenges around transparency even when people understand how income tax works, with what happens when the statistical estimates around income tax change. Those estimates are based on a much bigger and more robust sample than the estimates for VAT, so there is a question about transparency around the system for that.

The final point about the Smith commission concerns the ability to use Scottish Government or Parliament levers to influence the budget. That would not happen with VAT. I am talking not about the ability to change VAT rates but about being able to trace how changes in Government policy have an impact on VAT numbers. There will be question marks around that, as well.

Does it not also rest on a deeper assumption that, at an overall country scale, more consumption of VAT-chargeable goods and products is, by definition, a good thing? The Government will be accountable for maximising that, however it does that.

Yes and no. Yes in the sense that, if Scotland’s economy is growing more quickly than that of the UK, so you are not changing the pattern of levels of consumption or encouraging people to consume more but the economy is just growing more quickly, that should show up in Scotland’s relative performance on VAT receipts. However, you are right that there could be a situation in which, for the same level of growth, if you encourage people to spend more—including on certain types of product—and to save less, that would boost Scotland’s relative performance on VAT. The assumption in your comment would be right in that context; you would be encouraging people to spend more, and there is a question about whether that is right or wrong.

So, there are some perverse incentives in terms of whether there is economic value in getting people to save less and spend more in those circumstances.

Yes, in that context.

I want to pick up on Patrick Harvie’s reference to the review of the fiscal framework that is due in the next couple of years. We have discussed previously whether consideration should be given to greater flexibility around the Scotland reserve, given the volatility of income tax receipts. We have had many discussions about forecasts.

If I am correct, Professor Roy said that there is a confidence interval around VAT of a couple of hundred million pounds. If we look at some of the other taxes that we will be discussing, next year’s forecast for landfill tax is about £104 million and the cumulative forecast for land and buildings transaction tax is about £646 million. A significant element of public spending will be up in the air depending on which way VAT receipts are forecast to be, if we are talking about £200 million either way. For example, £180 million has been committed to the attainment challenge, with £120 million for high schools. If we take on VAT assignment, will we have to look again at the fiscal framework and the provisions around the Scotland reserve, to give us greater flexibility, or is there enough flexibility in the current arrangements to manage that volatility?

As you said, how all those things interact creates the challenge that is the extent of the management job. If all the bets go against you, it could be a sizeable number; if things cancel one another out, it is less of an issue; and if all the bets go for you there is a real challenge as to how to spend that windfall in a way that does not commit you to spending that you will not necessarily have the funding for in future years. That challenge is inherent in the system.

On reviewing the fiscal framework, at this stage we do not have any sense of the actual numbers and the reconciliations that will take place for the big taxes—certainly not for income tax or VAT assignment. We talk about that in conceptual terms and we have seen what has happened in the devolved taxes, but as social security, income tax and VAT assignment play through we will begin to get a much greater handle on the size of that volatility. It is important that we have a degree of track record and more of a sense of the numbers, to support the review of the fiscal framework.

A challenge for us all, at the moment, is to get a sense of how much we are talking about. We have some revised forecasts that give us some sense of that, but until we have more of a track record that will be hard. In relation to VAT assignment, we have a conceptual model but no detail behind it yet. So far, we have no sense of how it will play out historically and how much the level of volatility will contribute. It is a big challenge.

Given the forecasts—with the caveat that they are forecast reconciliations, I think—of reductions of £145 million and £472 million, would it be prudent to delay the implementation of VAT assignment until we have seen the outturn for this and the previous tax year with regard to income tax?

10:00  

One problem is that we will not have any outturns.

We will not have any outturn for VAT, but to get a sense of how the outturn matches up to the forecast and what volatility exists with regard to income tax—

We will begin to get a picture of that in July—was the first figure £185 million?

It was £145 million.

We will find out in July how real that figure was.

On the point about reconciliations and volatility, there are two moving parts: the SFC’s forecast could move in totally different ways from how we choose to forecast, because the SFC is, quite rightly, completely independent from the OBR; and we and the SFC will go at different times, which can be particularly tricky.

Our UK VAT forecast is quite stable and ticks along nicely; the previous year plus a little bit of growth is a fairly good predictor of what will happen next year at the UK level. That is mostly because we have not had any major policy changes. Who knows whether policy changes will come up in future? You might have quite a risk if a UK budget had a policy event; the next forecast would require the Fiscal Commission to react and feed through what that would mean for the Scottish assignment. Those differences could feed through at different points in time, so the reconciliation that would ultimately happen when the outturn occurred could be done on two different bases, with the OBR looking at one set of policy positions and the Fiscal Commission looking at another.

I appreciate that you are looking at this from a technical OBR perspective. Were the UK to leave the European Union without a deal, it has been suggested that the impact would be on a par with what happened in 2008. If the policy response were to reduce VAT to 13 per cent, would the pulling of such a policy lever make it almost impossible for this year to be an effective transition year for VAT, or would it make no difference?

It is very difficult to know what is going to happen—

I give that as just one potential example.

Yes, in relation to consumption. My point is that we have had a really stable period of VAT policy, compared with such things as income tax, stamp duty and LBTT, for which the policy designs have been chopped and changed a lot. As you said, if there were a desire to boost consumption in the short term and VAT was used as a policy lever, things could move quite dramatically.

That has been very helpful in kicking off the discussion, but we need to get to the nitty-gritty, so our next topic is the robustness and transparency of VAT assignment methodology. I invite James Kelly to kick that discussion off.

We drifted into elements of the issue in the previous discussion. Let me put it in context: what is trying to be achieved here is to look at the UK VAT calculation and to calculate the Scottish share and assign an element of that Scottish share to the Scottish budget. Given the budget’s importance to the Scottish Parliament’s expenditure decisions, it is important that the assignment is accurate.

The Governments have jointly produced a paper that sets out a methodology to achieve that aim. The methodology will have to be robust, so that it will produce accurate calculations, and transparent, so that people who view the calculations and, ultimately, the assignment to the Scottish budget, will have confidence in it.

In this section of the discussion, we will look specifically at the methodology that has been outlined by both Governments in their paper. How does it serve the purposes of being robust in delivering an accurate VAT assignment and giving transparency to inspire confidence in the calculation?

Does Graeme Roy want to kick off again?

There are a few things in that question. It would be great to see a lot more detail from both Governments about the methodology. We are talking about a 16-page report with no numbers and no estimates of sensitivity about how those numbers are changing or moving around—for £5 billion of revenues. That is exactly James Kelly’s point: a lot more information is needed about the robustness of the methodology. It is all very well to talk about a methodology, but what does that mean in practice, in terms of the estimates?

For example, the way the methodology works for about 70 per cent of VAT receipts involves looking at consumer expenditure, and the methodology relies on one survey: the living costs and food survey. To give you a sense of the assumptions, a maximum of 500 households were surveyed in Scotland for the GERS report over the past few years. The UK and Scottish Governments have boosted those numbers as part of the new methodology, but we are probably talking about 1,000 households maximum in the living costs and food survey.

To put that in context, 2 per cent of taxpayers were surveyed for the survey of personal incomes. That is about 50,000 respondents, versus a few hundred in the living costs and food survey. The robustness of that estimate is always going to be lower than it would be for other taxes. That creates problems, as Tom Arthur mentioned. The spread of confidence and errors is quite large—a couple of hundred million pounds on either side.

To take another example, simply as a result of the Government boosting the numbers of the sample in the living costs and food survey, the most recent GERS numbers in 2016-17 took £300 million out of the Scottish VAT estimates. That had nothing to do with a difference in performance; it was just because the Governments increased the sample size in the survey that there was a revision down of £300 million in the Scottish VAT share and a revision up of £300 million in the UK VAT share. That illustrates the potential sensitivity of the methodology to revisions in the analysis that goes into it. Much more transparency about the potential for that to be an embedded feature of VAT assignment would be helpful.

I want to pick up on Graeme Roy’s points on surveys, in particular the living costs and food survey, and how robust and reliable that information is. Given that the transition period is only one year, and 2019-20 will potentially coincide with other major events around Brexit, is using survey information based on only one year the best approach? Why one year? Maybe one year would normally be ok, but 2019-20 might be an unusual year. Why not three years? Is there any way for the methodology to reflect that or to use a rolling survey?

I do not know whether it is the same with the new methodology, but the way that GERS works at the moment is to use a three-year average for the survey. That is not for any structural reason other than simply that the quality of the survey is relatively weak, in the context of the numbers. Therefore, a three-year average has to be used, to give some degree of confidence. You are right to raise the issue.

However, that opens up some interesting questions. If the whole purpose of the process is to improve accountability on a year-by-year basis, the use of an average over a number of years means that that link between accountability, policy changes and actual outcomes is broken, because the actual numbers are being smoothed over a period of time.

In the next iteration of material that comes from the Government, it would be really helpful to see just how sensitive the estimates are to whether one-year, two-year or three-year spreads are used, and what happens when different numbers of people are surveyed. If an extra 100 or 200 people are added into the estimate, does that radically change the results? The results from last year suggest that it does; the addition of more households potentially will lead to fluctuations in the numbers that come out. Much more information about just how sensitive that methodology is to the data that goes into it would be helpful.

Perhaps we need to pursue with both Governments the size of surveys and the timeframes.

The fundamental point that Graeme Roy makes is really important. There is a real need for much greater transparency on the details of the assignment model. We have a slight advantage, because we had to produce a forecast in December, so we have seen some preliminary versions of the assignment model and the data. In fact, in our December report, we in effect published the 2016 outcome for the assignment model in its very preliminary form, which had not been agreed by the joint exchequer committee.

My take on the sampling issue is slightly different from Graeme Roy’s. Basically, that just underlines the difficulty of two people working in the same field getting their hands on the information, and the fact that we desperately need it to be published. In our “Statement of Data Needs September 2018”, we asked clearly for the two Governments to publish the assignment model and the underlying data through an official statistics publication, on which all users should be consulted. That has not yet happened, and we still do not have published data, although we have access to some data.

To go back to the technical issue about the survey, my understanding is that the preliminary assignment model uses just one year rather than three years and that the sample size has been boosted—it has been doubled. For the next iteration that is to be published, the number of households responding will go up to 720. In the current version, which uses 2016 data, there are 360 households, which is 0.03 per cent of households in Scotland.

The issue with the 2016 data was not so much that the sample sizes increased; it was that there was sample variation. Because the sample size was relatively small and there was an estimated switch between zero rate and exempt goods and reduced rate and the standard rate, that caused a drop in VAT receipts.

All that underlines the volatility issue—we think that that was a rogue drawing of the sample—and the need for much more transparency and the publication of the underlying details of the assignment model.

Did you say that the boosted number was 720?

There were 720 respondents. With such surveys, you get about a 40 per cent response rate.

So we are basing our outcomes on the VAT estimates of 720 respondents.

No, not quite. You are using the information from those 720 households to estimate a split between exempt, zero-rated, reduced and standard-rated VAT. Once you have that weighting, you can then use the consumption data to estimate VAT. That is a detail, but it is an important detail.

It is a very small number, though.

It is a small number, but you should not necessarily be thrown by the small sample size.

I will try not to be.

It goes back to the need for transparency—we need to know, and have published information on, the confidence intervals around that sample size.

I think that we are all thrown by that small sample. Why should we not be thrown by it? Why should we feel reassured that 720 respondents is a big enough number?

You should feel reassured because the statistical properties of samples do not vary in size with the sample size; they vary in size with the inverse of the square root of the sample size. [Laughter.]

Which means?

Crudely, it means that, as you increase the sample size, you get diminishing returns in the scale of the increased accuracy—

Is that clear, Mr Tomkins?

Let us move on.

What was that you said about transparency, Mr Ireland? [Laughter.]

As you would expect me to say, transparency is really important. There are two elements to that. One, which we have talked about, is about publishing the workings and the detail, so that people can understand it. It is already apparent with the methodology that there are lots of details—we have talked about one of them, but there are many more, which are built on Office for National Statistics estimates and Scottish Government economic estimates, which themselves are built on lots of details.

I absolutely subscribe to the view that transparency on all the workings is important, but there should also be transparency on the interpretation and assessment of things such as the extent of risk. We need a professional assessment of that from those who make the assessments. We have some of that detail for the things that the methodology is built on and that are already published by the ONS and the Scottish Government; the issue is how that is aggregated up so that transparency includes a sense of the things that we have talked about today.

10:15  

The paper refers to unregistered traders, on which there could be variance between Scotland and the UK. What is the impact of the cash-only economy and is that impact higher or lower in Scotland than in the rest of the UK? Could that have an impact on the VAT figures?

Willie, do you want to come in?

Businesses are VAT registered in Scotland. Is it not a safer methodology just to pick out the locational data that is coming in, in much the same way as we did with income tax, which we all know was a great success in estimating the numbers of Scottish taxpayers?

Charlotte Barbour might give you a reason why that would not work.

It is more difficult than just looking at individual traders. There are large businesses based here whose business is done everywhere, which raises the question of what is Scottish, and lots of consumers buy stuff online, which raises the question of where they are buying it from.

What are we trying to measure? The whole concept of VAT is such that it sits around businesses. Any business that we consider will have input tax going into it and output tax coming out, and we marry one with the other. It is like a production line where there are half a dozen processes with some going in and some coming out. Also, a lot of business goes on across the UK, as we have discussed with the committee before. It is therefore difficult to pin down what is Scottish VAT.

As we discussed earlier, and as John Cullinane said, it is just a difficult issue, and it is difficult to pin down the model that Mr Coffey suggests as well. Both models have their difficulties.

I want to bring the discussion back to an area that I have concerns about. Fundamentally, if we put rubbish into the system, we will get rubbish out. From what I understand, in broad terms, the VAT assignment methodology works by applying VAT rates to the estimates of expenditure from Scotland through something called the total theoretical liability model. However, I have not seen anywhere, in any of the descriptions from the Scottish or UK Governments, how successful that model is. Maybe Paul Mathews can help us. If that is the model for calculating the amount of VAT, how is the output from it compared with the actual outturn? Does it work? There is no evidence in any of the papers that we have on how robust the model is.

I am happy to comment on that. It works. For the purpose for which we use it, which is forecasting, it is okay, but—

Wait a minute. You say that it is okay, but what does that mean?

None of our forecasts is 100 per cent accurate, but our VAT forecast at the UK level has a lower level of errors than our average tax forecasts, so our UK VAT forecast is doing okay.

Is that the forecast compared with the outturn?

Yes. The key thing is—

What is the variation in there? What is the difference that leads you to say that it is okay?

I will answer a slightly different question. We know the amount of money that has come into HM Revenue and Customs—the receipts. That is a definite, known outcome. The VAT theoretical liability is estimated or built up from a lot of ONS estimates of consumption, and then various judgments are made about what is standard rated and what is zero rated—that is similar to the assignment methodology—and we get the total theoretical liability.

That number is a lot higher than the actual VAT receipts, and the difference between the two is known as the tax gap, or the VAT gap—it is a residual. There are lots of things going on in there that we do not know about. It could be that genuine aspects of the tax system are just not captured in the theoretical liability model—for example, businesses that are under the threshold—but it could also be that things such as avoidance and evasion are taking place.

The VAT gap is quite large, at about 8 per cent. That is not such a problem for the forecast because, as we discussed earlier, if the error is constant across the forecast, which is what we assume with the VAT gap, other than some measures and policies that the UK Government has introduced, it is doing generally an okay job in holding the forecast fairly flat. However, it is a large difference.

I guess that the real question for us, because we are talking about a tax that is based on relative differences between the two jurisdictions, is what will happen to the Scottish tax gap. How do we know how it will perform and at what level will it be compared to that of the rest of the UK?

We will never know, because we will never have Scottish receipts. The difference is between theoretical liabilities and receipts, which is known at the UK level, but we will have just theoretical liabilities in Scotland.

I can perhaps help to tie things down by giving some numbers. We have been looking at the OBR’s forecasting record on the difference, because it gives us a sense of how difficult it will be to forecast VAT in Scotland. If we go from the OBR’s March 2012 forecast onwards, the OBR’s two-years-ahead forecasting error was 2.4 per cent, on average. Paul Matthews is right that that is lower than a similar sort of thing across all the OBR’s fiscal forecasts, but it is higher than the income tax forecast error, which is 1.7 per cent. We can therefore get a sense of the relative magnitudes there.

On the VAT gap point, my understanding is similar to that of Paul Matthews. Certainly, when we are doing our forecast model for the assignment model, we use the UK data on the tax gap to move from the Scottish theoretical tax liabilities down to our forecast of Scottish receipts.

I will say a bit about the difficulty of working out even conceptually where VAT belongs. Last night, I paid a hotel bill that would have had VAT on it. That hotel consumption took place in Scotland, but I am not a Scottish resident. If I had come up here for ordinary, run-of-the-mill business reasons and that VAT was quite properly reclaimed, it would depend on the outputs of that business and what took place where. If we just took the receipts from the hotel chain and if it had hotels in England, it would have an extra job of work to break down what took place in Scotland. Even then, though, that would not get to “the right answer”.

We have forecasts that possibly stand to be corrected by outturns, but we also have estimates for things where it is not totally clear what the conceptually right and accurate answer is. That is the difficulty.

On the tax gap issue and the methodology that will be used for Scotland—frankly, I just do not know about that—will it just be assumed that the tax gap is the same in Scotland?

Yes, because it will be based on the theoretical liabilities in Scotland and those in the UK. In effect, it ignores the tax gap. There are many reasons why the tax gap could change in lots of odd and strange ways, but I do not think that that will be a particular problem for the assignment methodology.

Do members have any other questions on this area before I ask about a different one?

Before we move on from methodology, I note that the paper that has been produced by the Governments says:

“The Scottish VAT assignment model is similar to the VAT revenue sharing arrangements the Canadian Government uses with some Canadian provinces.”

Does any of our guests know what Canadian provinces that refers to, how that arrangement is working out and whether the affected Canadian provinces are happy with it?

How robust is the approach in Canada? If it affects what we do here, we need to know.

Angela Constance is looking at me.

I am casting the question widely.

To give a bit of context, we are publishing a new forecast this afternoon, so we have been busy. I read the Governments’ paper on the train up here and made a note to look into that exact point on the Canadian provinces, because I found it interesting. Clearly, it is useful to look at the international context, and we have tried to do that with income tax in order to understand behaviour changes—we have looked at different states in the US and different parts of Switzerland. Looking at international comparators could potentially be a good step forward for this work.

Does anyone here know anything about the international comparators?

I can make a couple of points about the Canadian example. We have to be careful about making direct comparisons between models that are contextually quite different. In Canada, there is a common goods and services sales tax, which goes across all of Canada but in many cases that is on top of a domestic sales tax, which is levied in the province. That means that there is outturn data, which is helpful as a benchmark that can be compared against what is happening at the federal level.

There are two other points in relation to the Canadian example. First, the quality of Canadian data is really high and it produces many economic accounts at province level. We do not have that in a Scottish context. Effectively, in Canada, national accounts are produced at a state or province level, which provides much more accurate information that can be drawn on and compared against.

The second point relates to one of the principles that the Smith commission was heading towards, which is that in Canada the system is designed to be a mechanism of helpful equalisation and to increase the share of Canadian tax revenues that flow through the provinces. It is less about giving provinces the ability to grow their economy and more about apportioning national revenues across the provinces. The Scottish context is quite different and is about injecting accountability into the Scottish budget, rather than the other parts of the UK. Canada has quite a different system.

Would it be good to look at the data that is produced elsewhere, how it is configured and how the evidence is found? I appreciate the point about context and the fact that the politics and systems are different. However, knowledge about the range of information and the terms on which it is gathered might be helpful.

Very much so. The whole fiscal framework leads us to the question whether the UK system for public finances and statistics is robust enough to cope with devolution and the transfer of significant public finance powers to devolved Governments. Places such as Canada have much more information and accessible data. The question is whether the pace of change in statistics and data that are available in the UK can cope with what is happening in the context of the wider fiscal framework.

Is it appropriate, or safe, for us to use the Canadian model here?

My understanding of what the Treasury is talking about in the document is that the general principle in Canada is an assumption that the purpose is to apportion national tax based on consumption in the provinces. That would seem to be the sensible way to approach VAT in the Scottish context. The question is whether we have sufficient robust data and information that would allow us to do that in a way that would minimise risk and maximise confidence.

I just do not believe the numbers, and I want to make sure that I am not wrong. I am often wrong, so let us just test this. One thing that will be applied to the system is the VAT that is taken from domestic tourism. Paragraph 4.36 of the paper “Scottish VAT Assignment: Summary of Assignment Model” says:

“According to 2014 tourism figures English and Welsh visitors spend approximately £1.7bn in Scotland and Scottish visitors spend about £1.2bn in England and Wales.”

It cannot be right that a population of 50 million people spends £1.7 billion, while Scotland, with a population of 5 million, spends £1.2 billion. Is my sense that that figure needs to be challenged correct? If no one has any thoughts, I will challenge the figure. Intrinsically, it just cannot be right.

I will not comment specifically on that figure, but the document mentions many other data sources, beyond the living costs and food survey. Maybe that figure is right and maybe it is wrong but, if it is constant, it should not be a problem. However, if it chops and changes and the 2014 tourist figures are replaced by a new survey, using a different methodology and suddenly those estimates change dramatically, that will feed through the assignment model and the results will be quite different.

There are other areas where the committee might have some concern. An example is where the methodology uses the annual survey on hours and earnings. That is an okay data set, but we are looking into a much better one, which is real-time information from HMRC. If the data sources are moving, the error is potentially moving too, and thus the assignment.

10:30  

I agree with what Paul Mathews has said. That highlights the point that we need much more transparency in the assignment model and the data sources that it uses, and an understanding of the variability in the methodology of those data sources. Everyone’s interests would be well served by the two Governments publishing something sooner rather than later.

Our final theme is issues around VAT forecasting risks to the Scottish budget. We have already gone into that, but Murdo Fraser will take it further.

We have touched on the issue to an extent, but I will start by asking about VAT forecasts. I will start with John Ireland, as he will have to do that happy job in due course. My two starter questions overlap. First, how does forecasting VAT compare in difficulty to forecasting income tax receipts, for example, which is something else that the Fiscal Commission is doing? Secondly, what is the short-term risk of forecasting error and how might that be managed in the context of the Scottish budget?

Forecasting VAT is not more difficult than forecasting income tax, but it is different. Mechanically, it is quite straightforward, because we take the components of the economic forecast, particularly consumption, put them into a relatively simple model and out comes a result. The income tax modelling is much harder, because there is much more data and a relatively large sample size, and we have the individual micro data. In terms of technical difficulty, forecasting VAT is relatively straightforward.

However, the question was probably about how difficult that is in conceptual terms and how difficult it is to get an accurate forecast. The experience of the OBR is interesting in that respect because, even with a richer data set and outturn data, its forecast error on VAT is larger than its forecast error on income tax.

As I have said before, the real conceptual problem for us is not so much estimating VAT assignment levels; it is that the surveys that are used in the assignment model are volatile and we have to predict that volatility, which is really hard. The example that we have to hand is the December forecast that we published. We think that there is a rogue drawing from the living costs and food survey for 2016, which suppressed consumption. That caused the forecast to be lower. How do we make that judgment? The reconciliations are based around that rogue drawing, so how can we guess what that rogue drawing would be in advance? That is tremendously hard.

What impact will that have on the Scottish Government budget? As the joint exchequer committee has not yet made a decision, we do not know what the arrangements will be. Let us assume that it is like income tax. If we think about the data that goes into the assignment model, the UK blue book is particularly important because of the outturns at the rest of the UK level. The lags there suggest a similar lag to that involved with income tax, of 18 months to two years. If we think about income tax and the reconciliation issues with it, we will have a good idea of the likely mechanics of the issue.

On the numbers, we took the OBR average forecast error of 2.4 per cent. If we apply that to the magnitude of Scotland, we are talking about an average error of, say, between £100 million and £200 million—the low hundreds. That gives you a sense of the potential issues. As everyone keeps saying, the reconciliation depends on the combination of forecast errors for Scottish VAT and VAT for the rest of the UK, so it is not just about that number. We are talking about something that is not as large as income tax in magnitude, but it is pretty chunky.

When we say “forecast error”, it is not necessarily a bad thing—it does not mean that we got something completely wrong. It can just be a result of policies that we were not aware of at the time of the forecast. If there is a policy change, our forecast for the year ahead will obviously be different, because the policy has changed.

VAT has been relatively stable as a policy area compared with income tax, which has been changing all the time. As we know from income tax, trying to understand policy costings is pretty hard, because it is hard to understand what the behaviours will be. That issue could add another slice of variance. There is the economy forecast error plus the methodology error in relation to the survey, and then there is the policy costing error on top of that, so we could have quite wide confidence intervals.

To go back to Angela Constance’s earlier point, would it be advisable to spread this over three years instead of just taking an encapsulation of what is going on in one year, to make sure that we are reducing some of that risk to the Scottish budget? Would that be the right thing to do, or not?

The division of labour between the commission and the Governments is clear. It is our job to do the forecasting and it is the Governments’ job to design the assignment model.

However, statistically, if you pool sample sizes by averaging over years, you increase reliability.

You would increase reliability—so that would reduce the risk.

It would be helpful to see what happens to the numbers when you start to do that. Does it reduce the risk? What happens when you boost the sample size? Do the numbers move around a lot? You can then get an idea of whether the Scotland reserve mechanisms and the borrowing powers are sufficient to cope with the level of risk that exists within that. In the next iteration of the papers that the Governments produce, it would be helpful to see just how robust the numbers are.

This goes back to a more fundamental question, on which the committee might want to reflect. If you have to rely on three-year moving averages for the VAT revenues and you have to worry about smoothing them over time, how does that link back to the point about accountability, which is why the policy is being assigned in the first place?

The idea is that you should be held accountable for movements in the budget. That happens with income tax. If your outturn is lower, you have to pay back the forecast. However, if you are smoothing out VAT revenues over three, four or five years, all you are really doing is replacing Barnett with something similar; it is just smoothed out over a number of years. You break that link between accountability and the policy implications.

On the sample and how it plays out, I would quite like to look into—or potentially the Governments could look into—the outliers in the sample and how much they matter. We have 720 households in the sample. If we removed the most extreme-spending one, how would that change the share? How would it feed through? What would the end result be for Scottish VAT?

We are talking about £5 billion here. Would the change be millions or tens of millions of pounds because of individual households that happened to be at one extreme of the sample? In future years, if the sample changes and more of these outlier households come in, how much effect will that have?

I will respond to the second part of Murdo Fraser’s question about how manageable all this is. There are two aspects to that. First—this is a question that Audit Scotland has been asking for a while—what is the Government’s policy on reserves and borrowing, and how does it set up that policy such that it copes with whatever uncertainty lies around the corner? That is an important matter.

Secondly, when adjustments are made, how do spending plans cope with that inevitable volatility? To what extent does everything get baselined, meaning that you would be storing up problems for the future? To what extent are you able to modify your spending plans by using a short-term response in order to cope with volatility? Key to that—this is where the real challenge lies—is knowing what permanent and recurring effects are due to economic performance, and what is just, to use John Ireland’s term, “noise” in the system. It is a real challenge for Government and Parliament to get a sense of those two components. The issue is how financial planning and management get ready for the inevitable volatility that is around the corner.

You have talked about transparency. David Eiser commented on that in his paper to the committee, and we should reflect on his words. He said, of the paper, “Scottish VAT assignment—Summary of VAT assignment model”:

“The Paper states that the model calculates the VAT incurred in each of these sectors at UK level by using ‘confidential HMRC data’ and ‘intermediate consumption figures’ but gives no further information on this, and then says that the ‘Scottish share of VAT in the UK exempt sector is calculated using labour market data’.”

It is all complicated. In effect, a lot of the calculation is based on confidential HMRC data, to which we do not get access. Is that appropriate, given that we are designing a new model for Scotland?

I recognise that the Scottish Parliament is likely to have an appetite for a degree of independent assurance over some of those calculations. Questions arise about how auditable that information is. Given all the challenges that we have talked about today, to what extent could you audit that information to give you that degree of assurance? In the context of the recent agreement between the two Governments about how audit and accountability work, through the audit and accountability framework, to what extent are Audit Scotland and the National Audit Office able to sit down, design and agree an approach that gives the assurance that you are looking for and is workable, given the complexities that we have talked about?

Our starting point is that we recognise that as a legitimate question that we need to explore, but we also recognise that there are challenges in giving you the assurance that you seek, which is what I infer from your question. We are trying to explore and understand that, and having more information about how the model works and where the numbers come from is a necessary part of doing that.

I mention on behalf of HMRC that it is governed by a statute that makes it a criminal offence to reveal anything to do with taxpayers’ affairs or any information that could be reduced to that. Therefore, quite understandably, it has an institutional aversion to risking doing that. I totally see that that does not make for transparency or accountability, or for ease of explanation to the public, particularly if adverse changes cost them more taxes, but it is not just a question of HMRC behaviour—to solve that problem at the root, one would need to look back to that statute.

That is a good balance to have put in.

We regularly bump up against the taxpayer confidentiality issue when we are trying to understand why forecasts have changed. HMRC is very robust in defending taxpayer confidentiality. There could be a tricky issue here. If outturn changed quite dramatically, and the answer to the change was that something had happened in some confidential data that you could not be told about, which meant that you had less—or more—money, that might be a hard message to explain or sell.

I agree.

We have had access to some of the data, to do our forecasts. I think that, once the Governments have published the details of the assignment model, you will get a very clear sense of where and how important the confidential data is. I would delay answering that question until you see the details of the assignment model.

Will we get more than just the summary that we have been provided with already?

I am going to a workshop at the end of March, at which the Governments will discuss the assignment model in more detail. I understand that they are hoping to publish something in spring.

Will the workshop involve only the two Governments and the relevant agencies?

No.

I have been invited, too.

It is an open workshop.

10:45  

I will finish off the session with a question on risk. What can the Scottish Government do to increase the growth of Scottish VAT relative to VAT revenue for the rest of UK? What can we do to increase the amount of VAT in Scotland? That brings us back to the very beginning of the session and the question about what the point is of doing anything if we do not have any levers.

That takes us back to the earlier conversation with Patrick Harvie. The whole purpose of the Smith commission’s idea is that you can better link the tax base in Scotland—and therefore the Scottish budget—to the performance of the Scottish economy. All other things remaining equal, if the Scottish economy can grow more quickly than the rest of the UK’s economy, our VAT base will grow more quickly than that of the rest of the UK, so Scotland benefits from that and accountability in Scotland is built in. Of course, if the VAT base grows less quickly, we take on the risks associated with that.

There is a debate to be had about how you grow your economy more quickly, and all the issues that go with that. There is also the subtle point that, to achieve a certain level of growth, you could try to get people to spend more money on vatable goods. That could be done by increasing the proportion of income of people who spend a lot of money on vatable goods. It is a bit like the incentive to have more higher rate taxpayers, because they pay more income tax; in this case, the incentive would be to encourage faster growth, with people spending more money—if the mechanism worked.

That takes us into interesting questions about whether you would want to take that approach, and whether it is consistent with ideas of inclusive growth and ideas about shifting money away from consumption into savings and vice versa. It is quite hard to see what unique levers the Scottish Government could use to boost VAT, other than—

Generally growing the economy.

Trying to grow the economy more quickly.

Thank you very much for your contributions. That was a very useful beginning to our discussions—I get the sense that this is only the beginning. I hope that it will inform our sessions with Government. You have greatly helped us to understand some of the challenges that exist. It is obvious that a lot of work has to be done.

I suspend the meeting to allow a change of witnesses.

10:47 Meeting suspended.  

10:57 On resuming—