Finance and Constitution Committee

Meeting date: Wednesday, September 11, 2019

Official Report 634KB pdf

---

Scottish Fiscal Commission (Forecast Evaluation Report)

Under our second item of business, we will take evidence from the Scottish Fiscal Commission on its forecast evaluation report. I welcome Dame Susan Rice, chair; Professor Alasdair Smith and Professor Francis Breedon, commissioners; and John Ireland, chief executive.

I invite Dame Susan Rice to make an opening statement.

Good morning. Yet again, I thank you for the chance to give evidence on our work.

Last Wednesday, we published three reports, including our annual forecast evaluation report and statement of data needs, to which I will refer. I am also conscious that the convener wrote to me in June asking the commission to provide an evaluation of the Scottish Government’s forecast of 2017-18 income tax receipts. There is a detailed explanation in our report of why that forecast was £941 million, or 8.6 per cent, higher than the outturn that was published by Her Majesty’s Revenue and Customs in July. I will come back to that in a minute; I will also mention the implications for next year’s budget.

Before I do that, I want to highlight something that you have all been supportive of in the past and in relation to which I feel that we have made some progress. Our evaluation report gives some examples that show the importance of good-quality data sources and information. One source is income tax itself—Her Majesty’s Revenue and Customs outturn data were not available until last year. Now that we are able to make use of that new data source, our forecasts have improved.

We face similar challenges in forecasting the new benefits. We underestimated spending on the new pregnancy and baby grant by £2.5 million, or 59 per cent, because of an unexpectedly large number of claims in the first few days after launch. The early data on claims for the grant, combined with a better appreciation of the Scottish Government’s approach to launching new benefits, led the Fiscal Commission to increase its spending forecasts in May.

Although forecasting new and reformed benefits will always be difficult, as the benefits become established and we have new data sources, we would expect our forecast errors to reduce over time. During the next few years we should expect some volatility around and fairly large forecast errors for social security, as more benefits are devolved and the Scottish Government introduces reforms to them.

I am pleased that our second statement of data needs documents that we have made good progress since last year, including signing a memorandum of understanding with the Department for Work and Pensions. It also sets out our request for additional data and information for the coming year—we keep raising the bar.

I will now return to the income tax forecast for 2017-18, which was made by the Scottish Government in February 2017—this addresses the question that the committee posed to us. At the time—this was before the commission went into statute—the Fiscal Commission’s role was to scrutinise the Scottish Government’s forecasts. We did that for the 2017-18 forecast and judged it to be reasonable at the time. We know now that the forecast overestimated revenue by £941 million, or 8.6 per cent. That is a significant error and 2.8 percentage points higher than the Office for Budget Responsibility’s average two-year-ahead forecast error, which is a benchmark that we use.

We estimate that £820 million of that total £941 million error was due to what we have labelled the 2016-17 baseline error, which was the result of having to use an imperfect survey data source that was three years old to estimate 2016-17 income tax revenues as the starting point for, or baseline of, the 2017-18 forecast. Taking that amount away leaves £121 million of the £941 million still to be explained. Our report shows that roughly £90 million of the error was due to the Government’s economic forecasts—total earnings growth was overestimated by 0.2 percentage points. We describe that as quite a small forecast error. The remaining £30 million is explained by a combination of modelling issues, HMRC incorporations estimates and changes in the costs of United Kingdom policies.

We all know that the income tax forecast matters because it feeds into the £204 million negative reconciliation in next year’s Scottish budget. Our report takes some time to explain the reconciliation that arises from differences in both our forecast and that of the OBR. We hope that we have made a start in the report in helping a wider understanding of the complexities of how the fiscal framework operates. The committee has copies of our new graphics, which try to pick up on those issues. We are happy to discuss them in due course.

I know that the OBR has written to the committee with an explanation of its UK income tax forecast error. I will not stray too far into its territory. The key point in our explanation is that the £204 million reconciliation—I emphasise that it is a reconciliation between forecasts—has little to do with the £820 million 2016-17 baseline error that I have just described. That is because the £820 million error equally affects the amount of additional funding that Scotland receives through income tax and the amount of funding that is taken away through the block grant adjustment, so the net effect is zero. The reconciliation arises from the combination of the error in forecasting growth in Scottish income tax and the error in forecasting growth in the block grant adjustment, using OBR forecasts. We think that the block grant adjustment was underestimated by around £83 million and growth in the Scottish income tax revenues was overestimated by £121 million. Together that gives us the £204 million negative reconciliation.

I have just gone through many numbers, and that is a lot to absorb aurally. Members will find those figures in our report. I expect that we will go through them again as we answer some of your questions. I thank you for your attention and patience.

Thank you, Dame Susan. Yes—there are a lot of numbers. I will begin with a question that is not on numbers, so as not to throw any more into the picture. I have been asking myself—and others might be doing the same—about the extent to which the forecasting process will improve once we have more outturn data for Scottish income tax and whether that is likely to reduce the size of reconciliations in future. If that were to be the scenario, within what timescale should we expect it to happen?

With any devolved tax or benefit, over the years, as we get more outturn data, we will have a sounder base on which to prepare future forecasts. The answer to the first part of your question is that forecasting will continue to improve. The first outturn data, which we got a little over a year ago, made a huge difference to the quality of our forecast. The timing will vary for each of the different taxes. Does Alasdair Smith want to speak to that?

As Susan Rice has already indicated, once we deal with the baseline error issue, the two forecast errors that go into the 2017-18 reconciliation are not very large, at roughly £100 million each for the block grant adjustment and our tax forecast. As time passes, we hope that everyone’s forecasting performance will improve, but given that those are relatively modest forecasting errors it would be a mistake to suppose that the reconciliation issue will somehow go away.

The 2017-18 reconciliation of just over £200 million is a significant number arising from not very large forecast errors and, at the moment, it looks as if the 2018 reconciliation will be a lot bigger. That is a feature of the fiscal framework within which we are operating. It would not be sensible to suppose that the issue will go away with the passage of time. We have to recognise that it is a feature of the system.

It is helpful to keep in mind that when we use the term “reconciliation” we are talking about reconciling differences between forecasts. The BGA is influenced partly by the OBR’s forecast of what is happening at the UK level and partly by our forecast, so it is a reconciliation of forecast differences. Outturn data is what actually happened, and that is when we talk about forecast error, because we can compare what happened with our forecast. They are two somewhat different processes.

I think that most of us understand the baseline issue. “Reconciliation” is probably the wrong word; it was a different type of adjustment that is now out of the system. However, Alasdair Smith has just told us that, over the coming years, while it may not be on the scale of the baseline issue, there is the potential for a reconciliation of a couple of hundred million a year—maybe a bit more or a bit less—which might be negative or positive. That might not be a big number statistically, but it could still be a big number in terms of what any Government can do with public expenditure.

That is correct.

So you are telling us that, because of the fiscal framework, we are in for a bumpy ride on the issue for a long time.

That is a fair assessment. It is something that has to be factored into the budget process. As Susan Rice said, two forecast errors go into the process—ours and the OBR’s. The OBR has a long track record of forecasting and yet it still makes errors, so we have to assume that that is an on-going situation.

I might put a caveat on that and say that it is not simply because of the fiscal framework. The fiscal framework is complex—that is absolutely fair; we are all still feeling our way through it—but other factors also influence the forecasting process: the quality of data that I have been talking about. That is why we keep asking for better and more timely data.

Also, it is important to keep in mind when we talk about income tax reconciliations that we will have reconciliations across taxes and our social security benefits over time as well. It is complicated.

09:45  

I understand that it is complicated—it is bumpy. What can the committee do to help to make things less turbulent for both sides of the equation—the United Kingdom Government and the Scottish Government?

One of the things that are driving the reconciliation errors is the issue of estimates of total earnings. In a sense, total earnings from the economic forecasts feed into our income tax forecasts, and quite small errors in those earnings forecasts are driving quite a lot of the reconciliation errors.

One of the things that we asked for in our annual statement of data needs was for some work to be done on Scottish earnings data. As you will remember from previous committee meetings, we have a number of indicators for the average earnings data for Scotland but we do not have anything that allows us to see clearly what the position on average earnings is.

One of the things that could help is pressure from the committee on the Office for National Statistics and the Scottish Government to start thinking about and working on improving the quality of earnings data in Scotland.

Good morning. I have a couple of questions on income tax reconciliations more generally but, before I come to them, I want to follow up on the convener’s first questions about the scale of the reconciliations.

Professor Smith, you said that the first three years were outliers, in a sense, and that, all being well, things should settle down thereafter to a much smaller range of variation. One of the things that the committee has been considering is whether the Scottish Government’s borrowing powers in the current fiscal framework are sufficient to deal with the peaks and troughs that might arise from forecast error. What are your views on that, and does your report lead us to any particular conclusions about the levels of borrowing powers that are currently available to the Scottish Government?

It is not really for the commission to advise on any redesign of the fiscal framework. You said that I had implied that, somehow, the reconciliations issue is going to settle down after two or three rocky years. I am sorry if I gave that impression; that is not what I meant to say.

The 2017 reconciliation that we are looking at in this year’s forecast evaluation report arises from two relatively modest forecasting errors. It looks as though there is going to be a bigger reconciliation next year. It would be ambitious to say that, in a few years’ time, we will see smaller forecasting errors than we have seen this year. We are talking about relatively modest errors, and it is more productive to ask—as you went on to do—what that says about the framework.

The numbers that we are talking about are relatively small—by the standards of reasonable forecasting errors—compared with the amount of income tax that is raised in Scotland. However, they are big sums of money relative to the Scottish Government’s borrowing powers. That is where they bite—the Scottish Government’s borrowing powers are sufficiently limited that coping with ups and downs—hopefully, there will be ups in some years, and not always downs—of several hundred million pounds involves quite a lot of variation relative to the borrowing powers. However, as I said, it is not for the commission to advise on whether the Scottish Government’s borrowing powers should be changed.

I would like to add a footnote to that, which is that, as the social security benefits come over to Scotland and Scotland is responsible for making that expenditure—and there is an unknown quantity, because we do not know how many individuals will qualify at any one point in time for any particular benefit—the borrowing powers may be used to smooth some of that. I am not commenting on the policy, but the borrowing powers will become more important and perhaps more actively required.

I would like to ask more generally about the income tax receipts as against forecasts. You indicate in your report—and we have heard this from you in previous evidence—that part of the explanation for the gap that has arisen is that, relative to the rest of the UK, there has been slower growth in Scottish income tax, which is a consequence of slower overall wages growth. To what extent is that down to faster-growing employment rates in the rest of the UK compared to Scotland and/or faster wage growth in the rest of the UK, per capita?

Part of it is down to the demographics in Scotland. Compared with the rest of the UK, we have an increasing demographic of people who are coming to an age at which they cease to be actively employed in so-called day jobs. My colleague Francis Breedon might want to answer the question more specifically.

I am not sure about that—it is quite a difficult issue to unravel. An important issue is that one of the upsides in the UK has been the amount that is raised from the higher income brackets, and that has been less of an element in Scotland. With regard to your question, I am afraid that I do not have the numbers. Perhaps John Ireland can say something.

There are some relevant numbers in the full report, which I do not think that you have in front of you. I point you to table 2.5 on page 23 of the full report, which looks at forecast and outturn growth in income tax economic determinants in Scotland and the UK. You are asking about employment. The table shows that the Scottish Government’s forecast for employment was 0.3 per cent growth, and the outturn data figure was 1.5 per cent. Therefore, the forecast, which went into income tax calculations, underestimated employment growth. UK employment growth over the same period was at 1.0 per cent, and the OBR’s forecast was 0.1 per cent. What happened there was that the OBR also underestimated employment growth.

I could read out the rest of the table, but the figures all go in a similar way. The important thing is the growth rates of total earnings. In Scotland, the outturn was 2.4 per cent and, in the UK, it was 4 per cent.

So, the important issue is earnings growth.

The real outturn to look at is total earnings growth.

Okay. To go back to the point that Professor Breedon made a second ago, the difference that has arisen has come about because, relatively speaking, there are more higher earners in the UK average than there are in Scotland, and it is among higher earners that the larger wages growth has been.

That is certainly where the income tax revenue increase in the UK came from.

Thank you.

I want to continue on the theme of wage growth, and then I will let John Mason in, because I know that he is interested in the income tax area, too.

In our report on the 2017-18 budget, our adviser highlighted that, at that stage, there was lower wage growth in Scotland than there was in the rest of the UK, and that that was a risk to the budget. Considering where we are, that appears to be a pretty astute analysis. To what extent did the SFC examine that risk when you were assessing the forecasts in 2016? I know that they were Scottish Government forecasts, but you were looking at their detail.

I think that I am the only one who can answer that. It is hard for me to say to what extent we did that. At that stage, all of us were learning. We examined previous years’ growth and considered what we thought was happening in Scotland. Everything has moved on since that point in time. The whole of the UK—only for a couple of years—was coming out of the impact of the financial crisis. Since that time, we have been asking for better and more timely earnings data for Scotland—that is really important to us. We considered the issue that you raised, but we had to do so in a way that involved a backwards look, because the process was just starting.

The issue of data is important. Data on Scottish income and earnings is not as good as the UK-wide data, as John Ireland and Susan Rice have said. However, we must also recognise that lots of people are puzzled about what has happened in labour markets over the past 10 years, not only in Scotland and the UK but across the developed world. None of us has a good understanding of the way in which labour market behaviour has changed in that period, so the issue will be a continuing source of uncertainty.

I have a question on the data. It is three years since I was previously on the committee so things might have changed a little bit, although I see that Dame Susan Rice is still here.

When I was on the committee previously, the problem was that the data was just not there, and with the best will in the world, nobody could give it. The main problem was that the 2016-17 data just was not there, but we now have it so we understand better. Is that correct?

How solid is the data now? Are we expecting improvements going forward? I had a letter from HMRC telling me that it assumed that I do not live in Scotland. How much do we trust the HMRC data? Is there also room for improvement in that?

I will give a quick answer then turn to John Ireland to put some flesh on it.

The 2016-17 baseline data was based on a report that used three-year-old data in a time of economic change and turmoil. However, it was the only data that we could use. John Mason is right: whether it was right or wrong, it was not what we would have wanted at that point in time. We needed to do something.

In an ideal world, all our data collection and sources would be absolutely where they should be and then we would devolve taxes. Clearly, that is not the real world, so we devolved and the data has improved along the way. It is certainly better now than it has been. Increasing years of better data also amplified that positive effect. John Ireland might want to give you some more details.

The data that was not there back when the forecasts were made was the outturn data. We now have two years of outturn data, which is helpful because it allows us to compare the survey of personal incomes, which is the sample survey of administrative data, with the actual outturn. We do have outturn data now, and that is a helpful thing.

We have been working with HMRC and the Scottish Government to improve the survey of personal incomes for Scotland. It is certainly true that, since the forecasts that we are talking about were made, the survey of personal incomes has also improved, particularly in the treatment of earnings at the very top of the range. They used to be composite records for the UK as a whole; now they are for Scotland. There have also been improvements in the underlying survey of personal incomes.

The other element of data that we have now that we did not have back then is something called real-time information, which comes from people’s pay-as-you-earn submissions. That is now becoming available for Scotland.

We are concerned about its relationship with the underlying outturn data. We do not think that enough work has been done to match the story that the real-time information is providing with the outturn data. In our data statement, we have asked HMRC to do more work there.

If I understand you correctly, there has been quite a big improvement—

Yes, there has been a lot of improvement.

—but you are looking for further improvement.

Yes.

It is worth adding that it suggests that the track record of the data will get better and we will understand the data better when we see a longer time series of outturns. At the moment, we have a range of indicators and we are still trying to work out which ones are the most useful. The more history we have, the more we can judge. That element will also improve.

In the past, I got the impression from HMRC that it saw itself as a UK body and it was not part of its job to provide Scottish, Welsh or any other subset data, and that we might have to pay for that if we wanted it. Has that attitude changed?

The Fiscal Commission has worked hard with teams and individuals in HMRC and the relationship is growing. This is about relationships and two different teams working together. HMRC cannot do everything for us when we want it because it has a bigger brief, but it certainly has been helpful and has listened to us.

During the past two years, the income tax team at HMRC has done a lot of work and has worked closely with us and the Government on these issues.

Personal relationships are good. The HMRC team came up to Scotland and spent almost a week talking to us and the Scottish Government, which was very good. We have regular liaison meetings by phone and down in London. Those relationships are good and HMRC is taking our data needs around income tax very seriously.

It is true that the Scottish Government pays HMRC money for the administration of the collection of Scottish income tax. My understanding is that some of that money is now finding its way to the analysts in HMRC, which is helping with the process. Those relationships are positive and HMRC has been responsive on income tax.

10:00  

Are you getting all the information that you need from the Scottish Government as well?

Yes. The primary supplier is HMRC, but the Scottish Government has certainly been helpful on income tax.

I remind members of my entry in the register of members’ interests in relation to property.

My question is on land and buildings transaction tax, in particular the residential LBTT forecast error shown in figure 4.4 of the report. What further analysis or breakdown can you provide on the decrease within the tax brackets shown and can you provide any further analysis or breakdown by region of the drop in transactions?

We cannot provide a breakdown by region, but we can say that LBTT is a difficult tax to forecast because it is related to the number of transactions that occur. We need to know, first, how many transactions occur; secondly, what is happening to average house prices; and, thirdly, what is the distribution of those transactions.

We struggled with that third element this time round. On average, about 4,000 transactions generate about half the revenue from this tax, so you can imagine that if we get just a few of those wrong, we can really make an error. The key element of our forecast error came from the forecast of those higher value transactions, which attract a higher tax rate and are higher value and are therefore quite important for the total revenue from the tax.

Can you provide a further breakdown of the decrease by bracket?

Not directly by bracket, no.

You comment in the report that there has been a drop in the top two brackets so I presume that there is some analysis available.

It is fair to say that it was not a drop. The value of transactions in the higher value group had been trending upwards so we forecast a continued increase in the number of transactions in that higher bracket, but in fact it was pretty much flat. The error came from us expecting further increases in the share from that high-value group.

Anecdotal evidence from estate agents suggests that, in some parts of the country, sales in the top two brackets have dropped to near zero. In figure 4.4, you show the drop in share in the top tax brackets and the drop in total transactions separately, but many individual transactions are part of a chain of transactions, so when you lose one, there is a multiple effect. Would it be fair to say that it is the same problem or would you do any analysis on the relationship between that drop in top tax brackets and the total transactions?

Over the summer, we have been rebuilding the model that we use. We are trying to allow for a more variable distribution of transactions across the different values. We were slightly imposing a shape on the number of high-value transactions relative to the number of low-value transactions and now, we are allowing that to be a bit more data driven. Indirectly, I think that that will help to deal with the point that you raise because it is a way of allowing more flexibility to come through on those issues.

Okay. So you are not planning to collect any of that transaction data by region in Scotland.

Not at this stage, no.

Your main report highlights that total transactions made up £18 million of the £42 million residential LBTT forecast error and £22 million of the error came from the share in the top tax brackets—that is £40 million out of the total £42 million error. Were you able to identify the reasons for that shortfall? Was it a result of a slowdown in the economy, policy change or uncertainty over Brexit, for example?

We do not really have anything detailed to say on that. Fundamentally, from our point of view, it is an error. As I said at the beginning, land and buildings transaction tax is a difficult tax to forecast in any situation because the number and scale of transactions vary so much year on year. We just see that as part of the general difficulty in forecasting such types of tax revenue.

To be fair, you could almost ask that question the other way round. What caused the steeper growth in those top two tax bands in the previous couple of years, and have we returned more to a norm? I do not know, but you can ask the question either way.

Yes. On the non-residential side, you say in your report:

“The non-residential forecast was close to the outturn figure, although we think this was the result of a cancelling out between more reliefs and higher price and transaction growth.”

Will you say a wee bit more about the compensating errors in there and what the difference would have been if those things had not happened?

The key element that we saw as being where the errors are coming from is in the area of appeals, which we think will come through later on.

That is on non-domestic rates, as opposed to non-residential LBTT.

Sorry—I thought that Gordon MacDonald mentioned non-domestic rates.

I will throw in a comment on non-residential LBTT. Table 4.10 shows that our forecast for that was £190 million and the outturn was £193 million. The most important thing to say about that is that the forecast was very close. You will see that, underneath the £193 million, there were some ups and downs, but there will always be some ups and downs. Let us just say that it was a very good forecast.

I appreciate that and I know that it is difficult to predict the future—I am not suggesting otherwise. However, the forecast for residential LBTT was higher than the outturn and the forecast for non-residential LBTT was below the outturn. Were different methodologies used there? Will there be any changes in how the forecasts are calculated?

They are very different methodologies. On the residential side, we have a relatively complex model because the data allows us to do that, and the banding helps there. On the non-residential side, it is a relatively simple model.

I have found the relevant bit in the report. On the non-residential side, both price growth and transaction growth were above our forecasts, so we underforecast them. However, it is a much simpler model for non-residential LBTT, partly as it is so volatile and lumpy because it is dominated by a very small number of commercial transactions.

Angela Constance has some questions on social security forecasting.

Good morning. It seems to me that how well plugged in you are to social security policy is central to how accurate your forecasts are. Some of the benefits are a wee way off, but they are coming on stream one by one. The Government has made great play of the people who use or receive benefits being crucial to informing policy. As well as the stakeholders, we have experience panels and the Scottish Commission on Social Security, which is scrutinising the regulations. How do all the factors in the development of policy inform your forecasts. Do you, for example, scrutinise social security regulations?

I want to use words correctly here. As you know, the commission does not get involved in development of policy: we would not go to the Government to say that its regulations have flaws.

However, to do our forecasts, which we understand are very important, we need to understand what population might be eligible for a particular benefit, and what proportion of that population might take it up, because not everyone who is eligible will do that.

When a new benefit comes online, we need to understand from the Government how it might reform or change eligibility, timetabling or whatever. We do not have much past data to use for the new benefits, so we need to understand the basic elements, such as the point in the year when the new benefit might be launched. Those are some of the elements that we consider.

Are you looking for us to—

Let me be clear: I am not suggesting for a minute that it is your job to comment on social security policy. Rather, I am suggesting that you need to understand what social security policy is, what the proposals are and how they will affect behaviour and take-up rates.

Let us look at the best start grant. The Government has a legal responsibility, as part of the Social Security (Scotland) Act 2018—we passed it aeons ago—to promote uptake of all benefits, not just the devolved ones. There are some very distinct policy differences in relation to the best start grant around, for example, second and subsequent children, which has a bearing on the financial outlay and, therefore, the forecast. I am keen to know whether you can demonstrate that you were, from the outset of your forecasts, tuned in to the policy. Did things such as the commitment around second and subsequent children figure in your work?

Absolutely. We discussed that at length, and judged that someone who had had the grant for their first baby would be much more likely to take it for their second child because they would know how the benefit worked and it would be easier for them to access. The answer to your question is yes.

Let me back away from the process of forecasting. As you know, Social Security Scotland is a new body. The commission knows what it is like to be a new body—it is a lot of work. We have engaged with Social Security Scotland and have tried to learn what it is about. That understanding is part of what you are asking about. Social Security Scotland is also learning as it goes and as it launches benefits. It is an iterative process.

We have had some private meetings with the Social Security Committee and will speak to it in a meeting of a similar format to this in a couple of weeks. We have tried to engage in various ways.

I will pursue the example of the best start grant, because it is a good one. We have had a lot of detailed discussions with the DWP and the Scottish Government on each of the social security benefits. We are always aware that understanding the details and detailed changes in the benefits is very important if we are to make reasonable forecasts. In the case of the best start grant, eligibility for children beyond the second child is clearly an important factor.

The striking thing that we learned from the best start grant is that the details of how the policy is launched by the Government can have a very big effect—it had a dramatic effect in that case. The Government seems to have launched the grant with a great deal of vigour, and parents who were entitled under the previous UK benefit quickly learned that it was better not to apply for the previous benefit, despite being eligible for it, because they also came into the window of the best start grant, which was more valuable to them. Take-up of the best start grant by parents of children who had been born before the grant was launched was one factor that led to the forecasting error.

The social security team in the Scottish Government has learned, along with the commission, that there are strong sensitivities to detailed issues in respect of how quickly potential beneficiaries learn about a benefit and how easy it is for them to compare the benefits of that one with the benefits of another.

I do not want my next question to sound obtuse—

But—

No buts. When my son hands in his maths homework, the teacher always tells him that he needs to show the workings that he did in his head, and not just give the answer. If the committee was to ask commissioners for your workings on forecasts for benefits, could you show us the work that was done at the time?

We have put a great deal on our website and have tried to be very transparent about what goes into our models and the bases.

10:15  

The forecast reports that we published contained a lot of information about the judgments—especially those about take-up—and how they were informed by issues such as the ones that you have mentioned. A lot of the workings are shown there.

The other part of the workings is our models, which are based very much on the rules for the policy and tend to use Excel spreadsheets or similar formats. We share them with the Scottish Government so that it can look at our workings very closely. Part of the forecast process is meetings with policy advisers and analysts from the Government, at which they go over our homework and mark it, probably with more vigour than a maths teacher would—at least, that is what it feels like at times. [Laughter.] Our workings are pretty transparent.

We have not yet shared our models outside the Government—partly because no one has asked us for them and partly because some of them contain information from the DWP that is not in the public domain.

There is one other element. In all our forecasts—especially in social security forecasts—there is a very strong element of judgment. There will, for all the individual forecasts, have been the kind of modelling and interaction with policy teams that John Ireland described, and there will have been meetings between commissioners and Government officials.

However, the question comes down, in the end, to whether we think that take-up of a new benefit will be a lot higher than take-up of the old one because it is more generous, or a lot lower because it will take time for beneficiaries to learn about it. It is not a matter of having maths workings, but of having heard all the evidence, some of which will have been quite qualitative and subjective. We can then say that take-up rates will probably fall within a certain range, which is a very important element in all social security forecasts.

I want to ask about your appreciation of what you and your partners have learned and what everybody needs to do differently. There was no fiscal consequence of the forecast error in carers allowance and I understand that you have—as the committee and other parliamentarians urged—had to wait for two Governments to come to agreements about work, timescales and so on that you could not have known about at the time of your forecasts. What have you learned from that process? Bigger and more complex benefits are coming down the track—for example, around 350,000 people will receive some form of disability assistance.

On the specific issue of carers allowance, most of the forecast error was due to our forecasting the cost of that benefit for an entire year. We had a fairly vague sense that it might be launched in the summer, but we did not know when the carers allowance and supplement would run for the whole year. Again, that was a matter of judgment. It was our first go and our judgment was that we should do something. When we looked at our forecast again in relation to the part year after carers allowance had been launched, our forecast error was really quite small.

We have also talked to Social Security Scotland. We have learned that the more timely the information we have about its plans to launch a new benefit, and the more details we have of that benefit, the better. We have also learned that the timing of the launch can make a big difference to the outcome.

To pick up on Ms Constance’s question about £3 billion-worth of disability benefits being devolved, I agree that that is a very big issue that loops back to our earlier discussion on reconciliations.

We will be looking at £3.5 billion of devolved social security expenditure, the forecasting of which will—with the best will in the world—be subject to errors of the kind that we have already discussed in a much more modest context. Those forecasts will depend on the details of eligibility for new benefits, which are different from the previous benefits, and on how the benefits are launched. We will make forecasts about spending of the order of £3.5 billion that is subject to quite a lot of uncertainty—to add to the uncertainties that we were talking about earlier in relation to income tax forecasting. It will be a very significant issue.

Taking all that on board and looking to the future, does the commission have any thoughts about what would help or hinder the budget-setting process, particularly when you are trying to manage demand-led spend? Has any consideration been given to forecasts that are based on ranges of spending rather than on one fixed amount, and that take into account the fact that behavioural changes can be difficult to predict or understand?

We already do a lot of those tentative analyses. The trouble is that it is difficult to translate them to use in the budget. We can give a range of start dates and take-up rates, but somewhere in the process a decision must be made about which to use. That is where the issue arises.

Our approach is to cost and use stated Government policies. We do not do a lot of trial runs.

We have seen uncertainties in measuring gross domestic product, and there are concerns about the strength of the link between GDP and tax revenues. You said this morning that we need better data on earnings, for example. However, GDP is an indicator of how well the economy is performing, so how interested should we be in GDP forecasts?

We have to be interested in GDP forecasts because they are part of the backbone of our overall economic forecasts, which have a direct link to income tax forecasting, so forecasting GDP is important.

Ultimately, our interest is in tax revenue, and GDP is a step along the way there.

There is one proviso, which is that the Government’s borrowing requirements depend upon a test of relative GDP growth between Scotland and the UK. Neil Bibby’s question is perhaps aiming at some of the Government’s recent work to look at alternative measures of GDP that capture different aspects of wellbeing.

My question was not specifically related to that, but I know about that on-going debate. I am not saying that we should not be interested in GDP. The question was not about whether we should be interested, but about the extent to which we should be interested, given the factors that have been raised this morning.

One reason to be interested in GDP is in respect of looking at the longer run. That is an important part of how our GDP forecast is constructed, even if the headline numbers are what our GDP growth rate is for the year ahead. If you look at how the forecasts are done, you see that we pay a lot of attention to long-run trends of productivity. The short-run GDP forecast reflects, to some extent, our view of where the Scottish economy is going in the long run, which is one of the most important questions to ask about it: what are its long-term prospects for raising what have, in recent years, been disappointing levels of productivity growth and, therefore, long-run growth.

There are no further questions for the Scottish Fiscal Commission, so I thank you for your evidence this morning.

10:24 Meeting suspended.  

10:33 On resuming—