Economy, Jobs and Fair Work Committee

Meeting date: Tuesday, November 7, 2017

Official Report 404KB pdf

---

Economic Data

Agenda item 3 is our economic data inquiry, and I welcome to the meeting our witnesses: Sir Charles Bean, member of the budget responsibility committee, Office for Budget Responsibility; Ed Humpherson, director general for regulation, UK Statistics Authority; and Jonathan Athow, deputy national statistician and director general for economic statistics, Office for National Statistics.

I will start with a fairly general question. What do you think of the current provision of economic statistics in Scotland? Are there any gaps in coverage in those statistics and do you have any specific recommendations for improving them? I should say that not all of you have to answer each and every question but if you want to come in, please indicate as much by simply raising your hand. Who would like to kick off?

I am going to throw the ball to Jonathan Athow, and I have quite a good reason for doing so. My review, which came out well over a year ago, flagged up regional statistics in general as one of the areas where more work was needed; however, it also recognised the problems in that respect. Obviously, if you want lots of detailed regional statistics, you will need lots of information, and the finer the degree of disaggregation you want, the more information you will have to collect.

I found that the only way of unlocking that was access to administrative data, which has, of course, been facilitated by the Digital Economy Act 2017. Given that the ONS has been moving forward with plugging some of the gaps that I flagged up in my review, it is probably better for Jonathan Athow to talk about the work that it is doing and its plans with regard to the regional statistical estate and—obviously—apropos Scotland.

We are on quite a journey here. The way in which devolution is changing in Scotland has changed user needs, and user needs are where we start from in thinking about our statistics. Following Sir Charles Bean’s review, there is now an added impetus to making certain that additional geographical detail is available. Obviously that applies to Scotland, but it also applies to Wales, Northern Ireland and the English regions. It is an area where we are changing.

Recently, we have started to do a few things differently. We have published public finance statistics for all the countries and regions of the United Kingdom, building on the Scottish Government’s work in “Government Expenditure and Revenue Scotland”. Things can be done, and things are moving. Next month, for example, we plan to publish what is called a balanced measure of gross value added. Gross value added is akin to gross domestic product. Previously, we have published those statistics for Scotland, but there have been two different measures using the income and production approaches, and we are now going to produce a single measure to give users in Scotland a much clearer picture of economic activity in Scotland.

There are lots of plans in place to improve statistics, but, as I have said, they come from user need—in other words, from what the Scottish Government wants and what other users in Scotland are looking for. We are thinking quite widely about this. On Thursday, we will publish a study on how practical it would be to publish a different estimate of inflation for the countries and regions of the UK. We have always thought that to be quite difficult, because of the way in which prices are collected to inform inflation statistics, but providing that information, which has been a gap in the past, is exactly what we are looking at and thinking about.

The answer is that we want to be as responsive as we can to user needs. As Sir Charles Bean has said, though, there are real challenges. In some cases, the way in which survey samples are collected for statistical purposes makes it very difficult to get down to very fine geographical levels. People often want statistics for not just Scotland but particular geographies in Scotland. For example, city deals are now emerging in Scotland, and people want to understand what is going on in their local area. That is always going to be very difficult with the data collected from surveys, and one of the things that we are now trying to do is unlock the potential of administrative data. One of our first data sets is VAT data from Her Majesty’s Revenue and Customs to allow us to understand economic activity in a particular area and give us a better understanding of what is going on. We are only starting to explore that, but that data underpins the balanced GVA statistics that I talked about earlier.

It is an evolving area. There are some gaps, but we are trying to fill them. As I said at the beginning of my remarks, we are guided by user need, and if this committee is able to highlight particular areas or gaps that it has heard about from users, that will be very useful in informing our work programme and our conversations with the Scottish Government about how we work together to make certain that there are good economic statistics for Scotland.

11:00  

I speak from the perspective of the person who heads up the Office for Statistics Regulation. What we do is oversee statistics that are produced in any part of the United Kingdom, certainly those produced by Jonathan Athow and his teams in the ONS but also those produced by the Scottish Government, Whitehall departments, the Welsh Government and the Northern Ireland Administration. That gives us good capacity to look across the provision of statistics in the round and to form an overview.

From that perspective, I would say that the Scottish economic statistics produced by the Scottish Government are in a good place in two senses. First, I think that there is a very clear framework for developing the statistics and their production, starting with the national accounts, with some good sectoral information on different sectors of the Scottish economy as well as some good labour market information. When you have that kind of clear framework, you can see much more clearly where the gaps are; they are harder to see if you have a fairly random collection of data releases. With a clear framework, you can see where something is stronger and where it is weaker.

Clearly there are some weaknesses, which brings me on to my second reason why I think the Scottish Government is in a relatively good place. There is an appetite for improving things and addressing the gaps, and there is a track record of having done so. You have heard a lot from witnesses in previous evidence sessions about the improvements that have been made over time to Scottish economic statistics. However, there are significant gaps. For example, Jonathan Athow has mentioned prices. Until recently, those data have not been available for Scotland. There are also questions about trade information on exports and, particularly, imports, and there are, of course, questions about the timeliness of regional economic statistics.

What really stood out for me, though, was the evidence that you received from the Scottish Fiscal Commission on how more timely and more comprehensive labour market information on income and earnings would help it to forecast economic matters and income tax receipts. That strikes me as a very salient issue that it would be very sensible for the Scottish Government to address in partnership with the ONS.

In short, we think that the base is good, but there is clearly an appetite for improvement, and some things really need to improve.

You have touched on a lot of issues that we are going to explore a bit further.

The Digital Economy Act 2017 has been mentioned, and Mr Athow particularly mentioned that we are moving towards more use of administrative data rather than surveys. What is the United Kingdom picture at the moment? Will the 2017 act fix everything or are there other things that we need to do? How does the UK compare to other countries? We hear a lot of good stuff, but it tends to be about small countries such as Estonia, Denmark and the Netherlands. Maybe it is just not possible to do those kinds of things at UK level. I am interested in where you all see the UK going on the issue.

It is not just small countries that have made effective use of administrative data. Canada, which is regarded as having one of the best national statistical institutions, relies heavily on administrative data. Basically, whenever they want to run a new survey, they have to make the case that they cannot get the information from existing data sources. It is notable that the Canadian statistical authority is a much more innovative and agile institution as a result. It is regarded as one of the best employers in Canada and one of the best places for economists to work. You should not think of administrative data as the preserve of small countries. There is a lot of potential there.

Equally, administrative data will not solve everything. Of course, the key thing about administrative data, and private sector big data, is that the information is not generally collected with statistical use in mind. It is collected as a by-product of collecting taxes or, in the private sector, of supermarkets pricing their products and registering what is being sold and so forth. It is then a question of statisticians levering off that. You need to use judgment and expertise in the way that information is exploited, and you often need to use surveys to fill the gaps. In my review, I took the view that, although unlocking administrative data is certainly helpful and an important step forward, it is not a silver bullet that will solve all the UK’s or Scotland’s statistical problems.

I should say that the 2017 act ended up going further than I thought was possible. In my review, I focused on facilitating ONS access to administrative data that is held elsewhere in Government, most obviously by the tax authorities, although not solely them, because lots of other departments have similar information that may be of use. I did not wade too far into the question of access to private sector data but, in fact, the act as implemented gives the ONS rather more scope than I anticipated for using private sector data. However, most importantly—because this is the first step—it removes the obstacles to effective use of administrative data across Whitehall ministries and so forth. Protocols need to be put in place, which I think the ONS is consulting on, but I am personally pretty happy with how the 2017 act came out.

Jonathan Athow might want to expand on that and give some details of how he anticipates its operating, but we are now in a much better place.

We are maybe not comparable with Canada, but are we comparable with Germany, the United States and Japan?

Potentially, we could be comparable with Canada. That is what we should aspire to; indeed we should aspire to not just being comparable to Canada but to overtaking it.

Underpinning John Mason’s question is the issue of whether we are keeping up with leading practice. We were not in the first group of countries to use administrative data, but we now have the 2017 act and we have been able to learn from how other countries did it. For example, other countries focused primarily on Government-held data, but we thought that there is advantage in also having powers over private sector data. Members may have seen it reported in the newspapers this morning that we have been looking at mobile phone data to understand commuting patterns and where people travel to work from. Mobile phone data could also be a rich source to help us to understand what is going on in the economy, and we know that private sector data is very important.

We have also been able to learn from the techniques that were used by the countries that went first. From December this year, we will use VAT data to put together our national accounts and our measures of GDP. We have worked closely with the Dutch statistical office, which has done that for a number of years, to learn how it uses that data.

Sir Charles makes the relevant point that such data is not primarily for statistical purposes, so time needs to be invested in understanding it. Different methods or statistical techniques might need to be used to understand it. The data might fit some parts of the economy very well. We have found some sectors of the economy where the VAT data describes what is going on very well and other areas where it does not. That is sometimes about understanding the way that it is collected. With VAT data, some companies will send in one return covering a large number of businesses, which sometimes makes it difficult to apportion how much economic activity is happening in a particular company or a particular local organisational unit.

Specifically on that point, such companies also would not split the data between Scotland, Wales and the north of England.

Exactly. You have to understand the nature of how that data is submitted and decide whether to apportion it or whether you still want to use a survey for that. You may want to survey some of the very complex companies, because that is better than the administrative data.

The Scottish Government largely does not have access to VAT returns and, even if it did, the data would possibly be at UK level. Is there something that we can take forward on that? Should we look to the ONS to get a bit more detail, or should we do something ourselves?

We want to use that data to understand industry and geographical breakdown. For some companies, that will work well. If all a company’s activity is based in a particular area, the VAT data will work well for that. It is about understanding where that approach works and where it does not. We will want to talk to the Scottish Government to understand its data needs, and we can then work through how we meet those needs. We need to consider whether that is done with administrative data, survey data or a blend of the two, and how we go about doing that.

As we go forward, we want to start making the underlying data available. One revolution in statistics in the past 10 or 20 years is that more of the underlying data is now made available for anybody to access, whether that is an academic or Government researcher. ONS data is mainly survey data at the moment, but it is available to researchers to interrogate. As we get these rich data sources, we want to make certain that they are also available for people to do their analysis, because we cannot do every particular cut of the data that people want. We want to make the data available for people to do their own analysis, but we also want to work closely with the Scottish Government and other users in Scotland to understand their needs.

We have received quite a lot of evidence, both oral and written, about the poor response rate to key surveys that are issued by the Scottish Government. For example, the response rate to the global connections survey, which looks at export data, is pretty low. We are examining what can be done about that. Is the solution to give the Scottish Government the kind of powers that the ONS and the Northern Ireland Statistics and Research Agency have to compel businesses to respond to surveys?

Under the Statistics of Trade Act 1947, we have the power to compel businesses to complete business surveys. Given that, response rates from businesses are in general quite strong. With some of our surveys, we have what we call a particular Scottish boost to the sample, and we can of course use our powers to make certain that that boost that the Scottish Government wants is collected effectively. I do not see a particular issue with the business surveys.

The issue for businesses is one of burden—they would rather be getting on with running their business than filling in forms for us. That is where the use of administrative data could be very helpful. We can reduce the burdens in that way.

With individuals, response rates are a real challenge. Fewer and fewer people now pick up their landline telephones, and we all have experience of somebody trying to talk to us about payment protection insurance or to sell us something. We find a large degree of scepticism when we try to collect data over the phone or even sometimes face to face. With individuals, it is a challenging environment and our response rates for our surveys are falling. That is a concern, because falling response rates could lead to biases in the surveys if we are not careful in understanding what is going on. That is a challenge.

11:15  

In particular, we are finding that fewer and fewer people are willing to participate in our labour force survey, which underpins pretty much all of our employment statistics. We are trying to get under the surface of that and consider whether we can shorten the forms or the questions. People might sometimes be put off by the fact that they have to be on the phone for 30 minutes or so. We are considering whether we could use behavioural insights or nudge techniques and whether there are ways to provide incentives for people to respond.

We are trying a mix of things, but it is a challenging environment, and many countries round the world are finding that people are less willing to participate. Perhaps we can use administrative data to take out some of the questions. We ask people what their income is but, for employees, that information should be available through the pay-as-you-earn system. We might be able to use administrative data to reduce the burdens and prop up response rates. There is a real challenge around individuals—as I said, we are less concerned about businesses and more concerned about burdens on individuals.

I will not give a view on the general power to compel, but Richard Leonard is right that the response rate for the global connections survey is relatively low, with of the order of 1,700 responses received out of 5,500 surveys sent out. The question is whether you would want to compel if you had the power. The question is about balancing the cost, which would be the burden on respondents, with the benefit, which would be how much more information you would get from a more complete survey. There are statistical approaches to considering what is going on with non-respondents, which is called the non-response bias. It might be that having 1,500 responses gives a sufficiently reliable picture because there is no difference between the respondents and the non-respondents. However, it could be that the non-respondents have a very different structure or different ownership and that, therefore, quite a lot of information is missing. The way to think about it is to consider what you would get from a greater response rate and whether that merits the costs of compulsion were you to have the power to compel.

I have one point that might be worth injecting into the discussion, which connects a bit with Jonathan Athow’s comment about nudge techniques. One approach is to give the person who is being surveyed an incentive to complete the survey or provide the information so that he or she gets something in return. That might be a financial incentive or it might be about what the survey is used for or information that is provided back to them that helps them in their business. Very often, surveys are a burden on the person filling in the form, as it takes time but, if they get something in return, they may be more likely to participate.

Maybe I am being a bit too naive and altruistic, but it seems to me that knowing what trade patterns are, where there are gaps and where there could be boosts in order to inform public policy could benefit individual enterprises. That seems to me to be an area worth exploring. The other thing that crossed my mind on the less altruistic side of my brain was to ask what remedies you have if people do not co-operate.

I do not know the exact powers, but we certainly have the power to fine businesses if they do not comply, although obviously we do not want to do that.

On your other point, I very much agree but there is a challenge. Sometimes a statistical office will say that something is very important from the point of view of a survey, but we need to tell the wider story about why the surveys are important. That is also true with the individual surveys. We explain to people how our understanding their wealth assets and pension savings informs Government pensions policy. Our best field researchers are the ones who make those connections and explain why. I could not agree more on that. The data feed into important decisions that affect businesses and individuals in Scotland. In some ways, people are helping to providing a public benefit through the statistics.

We have heard from witnesses that statistics on non-market activities and wellbeing are not as well covered. Is that just because there is a lack of demand for them? The Organisation for Economic Co-operation and Development talked about other countries using time-use surveys. Is the ONS considering anything similar?

We certainly see that there is a wider need to go—with the phrase that we use—beyond GDP. Obviously, GDP is very useful. It informs a lot of forecasting around the economy, tax collection and revenue generation more widely. It is very important as a concept.

We have used time-use surveys to look at non-market production. We published the results on that last year, and they show some interesting trends: non-market production has grown slightly quicker than GDP in the recent past. That information is out there. We would like to think about how we can make our time-use surveys more sophisticated, but they are quite burdensome. You can see that it quickly becomes quite burdensome when we ask individuals, “How did you spend your time today?” However, doing that gives us insight into what is happening with patterns of childcare and care for the elderly, how much volunteering goes on and how much those things are worth. There is quite a lot in that space.

We have published our wellbeing statistics; the next batch came out this morning. We are trying to make certain that we are providing a wider picture on wellbeing. An area where the UK leads is environmental accounts: understanding systematically the benefits that accrue from biodiversity and different sorts of environmental amenity. We are trying to pursue those areas because they are very important to people’s quality of life and people want to know what is happening with, for example, our forests or peatland or things like that. Those are areas where we are trying to do more.

GDP statistics are in some ways decades ahead, because they are so well developed and well resourced, but that means that those other interesting areas can catch up quite quickly. We try to make certain that we cover them. We obviously have a limited amount of resource and have to balance that quite carefully, but the issues that you raise are prominent in our thinking.

It is important to distinguish two separate strands in your question. One is non-market activity and the other is the broader question of wellbeing and those other indicators. On the first of those, one issue that I discussed in my review was the fact that shifts across the production boundary might be taking place. Conventionally, statisticians have said that they will measure what takes place in the market economy and not the stuff that takes place in the non-market economy. A lot of economic activity takes place outside the market economy: work that is done at home, charitable activities and so forth.

There are some exceptions. Imputed rent of owner-occupiers is a classic example of something that we measure in the national accounts that is not within the conventional production boundary. The reason for that is that we do not want to distort the conventional market measures if there is a shift between owner-occupier and rented accommodation. Statisticians would think that there is a case for measuring activities that are sufficiently big and might change in importance.

At the current juncture, it is quite plausible to say that there are quite a lot of activities that previously would have been done by intermediaries in the market economy. I always give the example of travel agents. In the old days, when you wanted to book a holiday you would spend an hour talking to the travel agent about where to go, then he would get on the phone and book the flights and hotel for you, and you would go and buy a guide book. These days, you do the research online, you book directly with the carrier, you might stay at Airbnb rather than a conventional hotel, and you use Google Maps to get around. A lot of that activity has been taken out of the conventional market economy, so potentially the market economy statistics, such as GDP, are being distorted.

The solution is not necessarily to expand the measure to include all those other things, because they may be difficult to measure continuously. However, at a minimum we would want to know how big the distortions might be, which is where time-use surveys might come in. I suspect that we will need to do more of them going forward, but there may be clever ways of getting information on time use. Obviously, a lot of data on internet usage is collected by the digital companies.

On the broader issue of wellbeing, in which there has been a lot of interest recently, I should say that from my perspective—talking just as an academic economist—I am sceptical of single measures of wellbeing, such as gross national happiness and things like that. I certainly do not think that GDP is a sufficient statistic for welfare or anything like that, and economists will always tell you that, but I do not think the solution is to get a load of indicators together, throw them in the pot, put some arbitrary weighting system in and say, “Here is my better measure.” GDP works as a concept because it can use prices to weight things together and prices are a measure of marginal utility. As soon as we start incorporating other things, it is not clear what the weighting should be. A more fruitful way forward is something like a scorecard approach, in which you decide what indicators you need to look at. Instead of trying to add up the 20 or so indicators that you might want to look at, you look at the detail and discuss which ones you, as a policy maker, think are the most important.

I very much agree with what Sir Charles said about the distinction between GDP as, in effect, an internationally comparable and consistent measure, and something that might be more of a dashboard or a series of indicators that responds to different policy imperatives and policy drivers. One of the interesting things that the committee has been hearing from your witnesses has been the fault line regarding what needs to be done to replicate at the Scotland level the best possible range of economic statistics that are available at the UK level. That fault line lies behind a lot of the gaps that I mentioned in my introduction, so that replication is important. It would be very useful and it is very much needed.

There is a fault line between that replication and thinking about what things we actually care about in the context of Scotland as policy makers and people who scrutinise Government performance. In that context you are in a space not so much of replication, but of innovation and thinking about what the things are that really matter. You have talked a lot about the four Is. They are a particular emphasis in Scotland’s economic strategy and developing a way of getting a good handle on them is something that could be quite innovative.

Jonathan Athow talked about personal wellbeing measures, statistics on which were published this morning by ONS. They are relatively underappreciated as a resource in this space. They are often spoken about slightly dismissively as being a happiness measure. There is now a growing time series of personal wellbeing over time. It is quite possible to drill down to quite local levels, because the data set is large enough, and it is quite possible to correlate the personal wellbeing responses that people give with other aspects of their life. Wellbeing is quite a rich policy area and it has tended to be slightly marginalised. I would encourage some creative thought on that.

11:30  

I am curious to know whether the survey data shows that we are happier or not.

I have a supplementary question. Obviously, you are starting to collect a wealth of detail. What is the size of the Scottish sample? Would you advise us to collect those statistics ourselves or to boost the sample?

I do not know off the top of my head, but I know that it is a representative enough sample for Scotland. I can come back to you with the exact number.

We tend to do it by tagging some of these questions on to existing surveys. Often there is a small number of questions. We will have gone out to talk to somebody about whether they are working, their job, their pensions or something like that, and then we will tag on the questions at the end. It is often quite easy for us to add a couple of questions to the end of a survey. It does not add much to the respondent’s burden, but it gives us a much richer source of data.

Then, of course, we get the whole-person perspective, because we can relate how an answer on a personal wellbeing question relates to what a person has said in other parts of the survey about their employment status, their pension status, their health or whatever it might be. That is why it is quite powerful.

Indeed. For example, one of the bits of analysis that has been done has been looking at how happiness varies according to people’s sexual identity and issues related to that. Such analysis gives us a very rich data set that goes much beyond economic statistics to help us understand wider social phenomena.

Can I go back to what you said about VAT receipts? Clearly, the Scottish fiscal framework agreement between the UK Government and the Scottish Government has a notional figure for VAT receipts. From what you said, it sounds as though the calculation of VAT receipts allocated to Scotland may not be reliable.

The key thing is to understand that there is a distinction: VAT is collected from companies, but the incidence is borne by individuals. I do not know exactly how the methodology works, but if you are interested in where the receipts are generated from, you are more interested in where people are spending, rather than which company collected the VAT. You can develop a methodology to look at that; it is just that, again, you will be answering a different question. It goes back to understanding how the data is collected, and the VAT collection data probably does not help you understand where the burden of that VAT is paid from.

What methodology is used to calculate the Treasury’s calculation of VAT that is allocated to Scotland?

I do not know off the top of my head.

Could you come back to us on that?

Yes, I can, but I am certain that the Scottish Government would have been involved in that work as well.

The problem is that the Scottish Government does not have any of its own statistics. It relies entirely on the UK—HMRC and the like—to provide the statistics. The other reason why the issue is important is that there is political controversy over VAT in relation to police and fire services. It would be interesting to know whether the methodology can state explicitly that VAT receipts from the police and fire services are included in the notional figure for VAT. Clearly, if they are already included, the situation is neutral, but if they are not included, there is a deficit. It would be useful if you were able to answer that question as well.

You referred to GERS, and a lot of very good work has been done north and south of the border on GERS down the years—I think that the original version was published when Roy Jenkins was the Chancellor of the Exchequer, way back in the late 1960s. In addition to VAT, there are big numbers in there for corporation tax and capital gains tax, for example. How reliable are those figures?

Often when you get into apportionment you need to make some assumptions. In relation to corporation tax, for example, we know from other sources how the workforce is distributed across Scotland and other parts of the UK. Sometimes employment is used as a proxy for where the economic activity takes place. Sometimes it is about using proxies rather than actually understanding where the tax arises. If a company has a single office and operates only in Scotland, it is quite easy to understand where it is. The situation with supermarkets is more complex, and we would look for things such as employment. Again, that may not be a perfect approach, but the way in which the tax collection system works often does not allow you to use tax collection to identify where the economic activity arose. You have to use proxies to do some apportionment.

Our briefing says that in Canada most of the statistics are gathered from the bottom up rather than the top down. Is that why Canada is able to produce more accurate figures at provincial and regional level?

I do not know enough of the details about how Canada works, but I know that they apportion their GST—their VAT system—and that they have to use a whole series of proxies for that.

Is that the goods and services tax?

Yes. In theory, you could start to ask businesses to apportion their activity between different geographical areas, but that could be quite burdensome. How does a business allocate overheads? That is always going to be a challenge. Even if you start to ask businesses for more granular data, that may not be any more reliable than using other data to apportion income against VAT.

What is likely to be the margin of error in corporation tax estimates and receipts?

I would not know off the top of my head, but I can give you a note to explain those sources.

That would be very helpful. Thank you.

My question is specifically for the ONS. The levels of devolution across the UK are somewhat variable in terms of statistics and so on. The ONS focuses in particular on developing statistics for the UK as a whole, but it also develops some regional statistics. To what extent are the unique or specific needs of the Scottish Government, or Scotland as a whole, factored into the ONS’s primary objectives?

I go back to my opening remarks. Everything that we do has to be user focused. What do people need the statistics for in the decisions that they need to make? In that sense, we are very alive to the needs of Scotland, but we are also alive to the needs of Wales, Northern Ireland and the English regions, given devolution within England. That is a real focus for us. We have a track record of working with the Scottish Government to innovate, do things differently, boost sample sizes in Scotland and provide more detail where that is needed.

From our point of view, there is a real willingness to work together. We have regular meetings of what is called the inter-Administration committee, where we sit down and compare priorities. We include not just economic statistics but wider statistics on population and so on. We also have an economic statistics group that involves the ONS, Scotland, Wales and Northern Ireland and meets every quarter to look at common issues and needs. We do a lot to try to make certain that we incorporate all the different needs, whether those are UK-wide or Scottish needs.

Sometimes, some things are easier to deliver than others. Some things require extra resource or new data collections. We try as far as possible to meet all user needs, and I hope that we have a good reputation with the Scottish Government for working constructively with it.

When we assess statistics that have been produced by the ONS—so they are, of course, at the UK level—we assess them against the code of practice to determine whether they meet the high standards required of national statistics. One of the things that we look at is the extent to which statistics serve user needs beyond the UK level. We recently published an assessment of regional gross value added, and we looked a lot at the extent to which the questions that those statistics addressed spoke to the interests of policymakers not only in Scotland but in the English regions and in Wales and Northern Ireland. Sometimes my job is to prod or nudge—with various degrees of politeness or firmness—the ONS into making change. We make interventions in those areas.

My next question moves on to look at that process for developing objectives. Mr Athow touched on that with reference to the inter-Administration committee, and Mr Humpherson said there is possibly an element of encouraging the ONS to develop those processes.

If the Scottish Government said at one of those meetings that it wanted to go in a different direction and wanted data on X, what would the turnaround timeframe be for developing that new statistic?

I will be slightly unhelpful here and say that it would depend on what X is. Inflation statistics have been a longstanding issue, with people wanting to know more. We are lucky in the sense that we have now brought in a wider range of academics to work with us, so we can now get them to think about how realistic that is and what we need to do. In some cases, we can respond very quickly. If the data is already sitting there and just needs to be cut in a different way, that can often be done quite straightforwardly. If it is something that requires a whole new data collection exercise, I think that we would just sit down with the Scottish Government and work out the best way of doing that. That might involve a combination of, for example, Scotland commissioning its own surveys and the ONS taking one of our existing surveys and modifying it slightly in Scotland. All the options would be open and we would think about the best way of meeting those needs.

Sometimes there are very difficult conceptual issues to deal with. Understanding exports to or imports from Scotland is conceptually challenging. The area is of interest to not just Scotland but Wales, and it is quite challenging. Sometimes it is not just the data collection but the conceptual issue that is challenging. For example, what does “an export from Scotland” mean? Measuring those things can be quite challenging. However, we would be very open minded and there are lots of different ways of answering those questions.

Can I just inject a suggestion? At present, key users—the Bank of England and the Treasury, for example—have an annual process of interaction with the ONS in which we say what has worked well and what has not worked well, and in which we signal priorities for future development—basically, what we would like to see. When I was at the bank, we often said that we would like to see completion of the flow of funds, for instance.

I guess that there is a question about the extent to which the Scottish Government is consulted formally in its role as a user of statistics and whether it might usefully be brought into that process. It is simply a mechanism whereby, on a regular basis, the Scottish Government could flag up areas that it thought ought to be developed.

Section 20 of the Statistics and Registration Service Act 2007 provides that

“The Board may itself produce and publish statistics relating to any matter relating to the UK or any part of it.”

Can we take it from that that the UKSA has unfettered powers to produce any statistics for any part of the UK on any matter?

The collection and publication of population statistics, for example, rests with the National Records of Scotland. Specific elements, such as the census, are devolved. We are the census provider in England and Wales, and the Northern Ireland Statistics and Research Agency is the provider in Northern Ireland. The answer to your question is, broadly, yes, but there are some areas where there are longstanding constitutional differences and we do not go into those areas.

11:45  

Section 23 of the 2007 act says:

“The Board may not without the consent of the Scottish Ministers produce and publish Scottish devolved statistics.”

I presume that the board has never tried to publish devolved statistics, to which the Scottish Government has said no.

No, indeed. Such statistics would be, for example, the population statistics that I was just talking about.

Yes. It seems that there are no statutory constraints to getting better statistics that are relevant particularly to users in Scotland. We heard last week from the Scottish Fiscal Commission, whose need is perhaps of a different order of importance, because it is involved in forecasting and, on the basis of those forecasts, we set budgets that affect the amount of tax that people pay and the amount of economic activity that we anticipate will be generated in the economy.

The current arrangements between the UK Statistics Authority, the ONS and the Scottish Government are via protocols, memoranda of understandings and joint meetings that take place on a regular basis. Given that, as I understand it, those arrangements work pretty well, is there a need to formalise them along the lines that Sir Charles Bean has just talked about, making them part of a much more formalised process whereby the needs of, for example, the Scottish Fiscal Commission can be flagged up pretty early so that, once it takes on its powers, everything is in place to allow it to do the job to the best of its abilities?

There could well be. Through the UK Statistics Authority board, we are accountable to the Scottish Parliament, the Welsh Assembly, the Northern Ireland Assembly and the UK Parliament, and we have been thinking about how we make that real. We lay our annual report before the Scottish Parliament, but could we do something more systematic that would talk about what we had done to meet Scottish user needs and use that to elicit further feedback? I am open minded about how we might strengthen that relationship to make certain that all needs are articulated early.

We do go out and engage all users. We have something called an economic statistics and analysis strategy that sets out our priorities for the year. We consult on that and ask whether it addresses people’s needs. However, if there is more that we could do that you think would be relevant to Scotland—particularly to the Scottish Parliament—we would be open minded about that.

There is something in what Sir Charles Bean has suggested, and the Scottish Fiscal Commission is a good exemplar of it. That body has some important new responsibilities and has identified a specific need or a lack of completeness in the data that is available for a fundamentally important thing. It has spoken about the time lag for its access to income and earnings data, and it is thinking about whether there are more timely ways in which it could get insights into income and earnings in Scotland. That resonated with me, because the Office for Statistics Regulation has, for a couple of years, been prompting the ONS on the question of more comprehensive income and earnings statistics as well as more timely ones.

The general point is that the Scottish Fiscal Commission, an important user with new responsibilities, is exactly the sort of player who could feed in to some kind of forum and say, “These are our needs. What are the best ways to meet them?”

On the practicalities of doing this, we have some figures for the money that the Scottish Government spends on boosting the various surveys to get better results, which are modest sums of money in the bigger scheme of things. I take it that those are ad hoc arrangements. The Government says that it wants to boost a survey and talk to the ONS; you say that that would not be very easy because budgets are tight this year. The Government says that it is happy to chip in something to help and you then have a negotiation. Is that how it works, or is there some formal protocol to what the Scottish Government will and will not pay for—what it expects you to do?

We set a requirement to meet the particular needs of a certain sample size. If Scotland wanted to increase the sample size, we would expect that to be funded on a cost recovery basis unless there was something very trivial. If it was a very simple thing to do, we would not ask for cost recovery on that.

More generally, we expect cost recovery when the approach in Scotland differs from the UK approach. That is how it has worked in the past. We ask for cost recovery on any additional costs from boosting the samples, and that seems to work well. It gives Scotland the statistics that it wants, and it does not mean that we have to trade off anything else within the ONS. It is then for Scotland to think about whether it is worth the additional cost to boost the samples.

You have covered quite a lot of what I was going to ask about around inclusive growth and regionality, so I will ask you about something else. Forgive me if it sounds as though I expect you to have watched every second of our inquiry so far, and tell me if I am off the mark.

Our inquiry has been going on for a number of weeks. Has anything that you have heard in the evidence, in preparing to come here today, struck you as something that you did not already appreciate or know, which has given you pause for thought or suggested how your existing bodies can better serve Scotland’s interests? That may involve not what you just talked about with Andy Wightman—our paying for more—but changing slightly how you operate when conducting the existing surveys.

The Scottish Fiscal Commission is rapidly developing an interesting and challenging new function that needs to be carried out. That stood out for me as a particular issue. We are also at a very interesting stage in the sense that there are huge numbers of new requests coming through. I see people asking for lots of different elements, and I am surprised by the breadth of the requests that are being made as devolution widens and deepens. I think that those needs will become a bit more sharply focused as new frameworks come in, as new measurement frameworks are developed and as the Scottish Government gets to grips with its new powers. The breadth of what you have heard about covers virtually everything in my responsibility in the ONS and goes a lot wider. It has been a real tour of all the issues on people’s minds. The next stage is to decide which of those requests are the most prominent in terms of crunchy statistical needs. I am interested in seeing how we narrow down that very wide field into some priorities that we can then start to work through.

I am not sure whether this is going to answer your question but I will tell you my reaction to the previous evidence sessions.

One thing slightly surprised me and one thing quite excited me. A couple of witnesses talked about the need for a stronger ecosystem of research and analysis outside the Government in Scotland. That rather surprised me, because I see quite a strong range of organisations in academia. There are some good think tanks. The Fraser of Allander institute is a superb institute that does great work. I was quite surprised by some of your witnesses, and I was not sure what their benchmark was. I think that there is a really good basis.

The thing that excited me builds on that. A few witnesses talked not so much about the statistics and data that are produced by the Government as about the statistics and data that are produced by a range of other actors: enterprise bodies, research bodies and so on. They asked whether there might be some way of linking those up into a consistent framework, which would make the data more usable. That is really exciting, and it maps very well on to something that we are keen to advocate—that the code of practice that we have for official statistics could be adopted much more widely, including by providers of information and statistics that are not necessarily Government bodies.

Some things in that code are specifically designed for the Government context—things such as pre-release access and publication at 9.30. You might ask why publication at 9.30 matters. It matters because the producer of the information—the official body, which is the Government—wants to demonstrate a commitment to its users that it will publish at a particular time, come what may and regardless of political drivers. That commitment to being trustworthy could be satisfied in a number of ways, but the principle of trustworthiness is something that a range of other bodies could easily adopt, describing how they comply with that.

Similarly, we talk about the quality of statistics—how that is explained and how you make quality transparent to users—and, indeed, about the value of statistics and how statistics produced by any organisation answer key questions. We think those high-level pillars of trustworthiness, quality and value are not the unique preserve of the Government.

You have had quite an exciting debate with the wider actors in your ecosystem in which you have been thinking about whether there are ways of adopting universal principles that could help users to marry up different sources in a confident way. That is very innovative, and I would be happy to talk about it more. It is an exciting development.

My question is on pre-release access to statistics—a matter that I believe the convener has raised with the First Minister previously.

You have changed your approach to ministers getting access to statistics in advance, but the Scottish Government unfortunately seems to be stuck doing something entirely different. Would you advise it to change in the same way as you have done?

Yes.

Yes.

I always like to end on a positive note.

I say that as someone who has been on the other side of the fence, as a key user when I was on the monetary policy committee. At the end of the day, having very tight rules around pre-release access prevents abuse. Sometimes, that abuse is due to accidents rather than deliberate—somebody knows something in advance and they say something that reveals market-sensitive information that should not have been out there, and that brings the process into disrepute. Sometimes, however, people may deliberately exploit pre-release access for personal gain. There have been examples of that in the past.

Ministers like having early access because they can be briefed about what the numbers mean, which means that their initial reaction is informed rather than ill-informed; therefore, there is an argument for it. Nevertheless, I think that the balance of arguments supports pretty strict limitations on pre-release access.

12:00  

My first answer to your question is yes. A more structured answer is that the act under which we operate in the Office for Statistics Regulation prevents us from imposing any changes to pre-release access arrangements. The act specifically says that our code of practice, the core enforcement tool, cannot deal with the granting of pre-release access, so it is the choice of the producer of the statistics whether they grant pre-release access. The ONS recently chose to remove it and, shortly afterwards, the Bank of England followed suit. The Scottish Government has chosen another way.

I cannot say “Yes” and then enforce that decision—the act prevents me from doing so—but I can make some advocacy comments. The things that I would advocate in this space are twofold. First, why do we care so much about this? It is because, at the heart of what statistics are about, they are a public asset. They are there for public consumption as information that enables the public to understand the nature of the world, the nature of policy and the nature of the decisions that are being made. Secondly, that vision of a public asset is underpinned by statistics being equally available to all and not partially available to some audiences but not others, and by their being available as soon as they are ready. Pre-release access seems to run against those two principles.

Pre-release access makes the job of the Scottish Government, as a producer of statistics, a little bit harder. It could, no doubt, be argued that it is beneficial because it allows ministers to have access to the data and to have the content of the statistics explained to them. Nevertheless, you have to set against that the issue of whether there are perceived or actual breaches, as Charles Bean identified. There is a perception that one set of actors—ministers—gets a privileged access that others do not get. Therefore, for the Scottish Government to establish its trustworthiness—to use that term again—it needs to work that much harder to demonstrate the integrity of its production process, and I think that that makes its work harder.

In saying that, I am not making any judgments or accusations about the people or what they do. They are genuinely highly professional statisticians who do an excellent job. I just think that pre-release access makes their work harder.

Should the Scottish Government be both a producer and user of statistics, or do you think that the set-up is fine and it is just the pre-release issue that you might have a difficulty with?

I will give you my first thoughts on that. Good statistics can emerge from a whole range of institutional arrangements. If we look across the UK, we see that Northern Ireland has NISRA, which is a separate agency of the Department of Finance. NISRA operates a hub-and-spoke model; in other words, there is a central NISRA that has about 200 people, but a lot of its statisticians are also embedded in the various departments in the Northern Ireland Administration. That very powerful model ensures that individual departments, in producing their own statistics, also have a powerful professional voice.

In England, you have ministerially led departments that are very effective producers of high-quality statistics. If I could single out one example, the work by the Department for Work and Pensions on households’ below-average income is an absolutely superb statistical output that emerges from a ministerially led department.

In Scotland, you have what you might call a three-peaks model. You have a strong statistical hub in the Scottish Government, but there is also the National Records of Scotland and Information Services Division Scotland, which produces the national health service statistics.

I have no presumption as to whether any one of those models is superior to the others; what really matters is very clear adherence to the code of practice. That code enables statisticians to operate independently across all of those institutional settings, with clear development of professional standards and a focus on quality. I am—if I can put it this way—agnostic on the question of where production takes place but, to go back to my earlier point, I think that the issue of pre-release access creates a little bit more of an uphill struggle with regard to perception.

The issue goes beyond pre-release access because if production takes place within a department that uses the information, tensions might arise, so you must have very clear protocols and make it clear that the producers of the statistics in a department are independent and can produce those statistics without interference. The whole structure for statistics, with the ONS, the UKSA and so forth, goes back to 2007 and has its roots in the perception that some statistics—crime statistics, if I remember rightly—

It was unemployment, too.

Yes. There was a perception that some statistics were basically being massaged or fiddled by the ministers in charge of the departments. That might not have been true but, nevertheless, that was the perception, and the key objective of the Statistics and Registration Service Act 2007 was to establish clarity of independence and production, which needs very clear rules of operation in the department or organisation involved.

Obviously, one of the questions that I investigated in my review was whether that process was working. I found that it was, by and large, but it was clear that on some occasions producers were being put under inappropriate pressure by their minister or senior civil servants. You must have a robust framework in place; it is possible to have this process, but things need to be made clear.

But what happens if the process is abused by a senior civil servant, as Sir Nicholas Macpherson did during the referendum? Nothing happened to him.

If there is an abuse of process, it is up to—

Why was he not done under the code?

You will have to ask John Pullinger. I do not know whether Jonathan Athow wants to—

Mr Neil, I am not sure whether it is within your remit to decide who should be done under the code.

But it is within Mr Humpherson’s remit.

Then we will let Mr Humpherson answer, if he so chooses.

I am not going to comment on that case, as it was some time ago. What I would say is that, in Scotland and in other parts of the UK, we repeatedly make public statements not only about how statistics are produced by statisticians but about how they are disseminated and used by Government departments. We have a long stream of correspondence in which we have made some quite firm public statements to remind all those involved with the production of Government information—not simply the statisticians—that they have responsibilities to the public. That is what we will continue to do.

As a Highlands and Islands MSP, I represent a huge and diverse area with differences in the speed of local economies. My question, therefore, is whether administrative data could play an increasing role in providing more useful localised data for the regions of Scotland—and even more local than that—and whether it could be used to get an idea of regional productivity.

The answer, in short, is yes. We certainly need more administrative data. An area that we want to explore next is PAYE data, which will allow us to understand in a much more granular way people’s earnings from employment in a particular area. We could—and would certainly want to—get down to quite a fine level with that, although we would have to be very careful that the data did not become disclosive.

As I have said, VAT data allows you to get down to finer levels, but there are some challenges with that. In a previous role, I was asked by an MP at Westminster how much VAT was being paid in one particular postcode, which I believe was a Scottish island. However, only one company was trading there and because it had had a VAT repayment, it looked like no VAT was being paid on the island. That is the function of how the data is collected, and then you need to understand, given that data, how much of what is going on locally can really be inferred.

Such an approach is not a magic bullet, but it allows you to get a lot more granular data. We would never be able to get that level of detail with surveys; indeed, with VAT data, we have 30 or 40 times as much data as we collect through our business surveys. The level of detail that we have is astounding, although there are difficulties with how it is interpreted.

The data would give much greater understanding of local performance. With our balanced gross value added measure, we are looking at not just Scotland but the next layer of detail down. Obviously, with the small number of people involved in the Highlands and Islands, you might not get much granularity or resolution but, in theory, we can extend our work through the use of administrative data sources.

One of the central themes of the Scottish Government’s four Is economic policy is inclusive growth. As we have asked in other evidence sessions, is there a recognised economic definition or statistical measurement of inclusive growth, or does it fall into the category of wellbeing, which you have already mentioned and which is an amalgamation of different measurements that might mean different things to different people?

I think that, if there were a scorecard, it would fall into that category. One of the criticisms of GDP is that it is an average measure and that what is important to people’s wellbeing is the distribution of income and wealth. We try to provide more detail on those things, but they are quite difficult to measure as you need a good and very detailed understanding of households. There are some definitions and shortcuts that we use, but you have to be very careful. We sometimes use shortcuts in relation to below-average income in households, with net poverty or low income measured on the basis of certain percentages of the household median. Some people would say that those figures are arbitrary, so you have to be very careful that you do not just choose one; you have to look across the whole suite.

Similarly, the main measure of inequality is something called the Gini coefficient, but that measures only one aspect, and there are lots of other different aspects of inequality that might be salient. Moreover, what do we mean when we talk about inequality? We often focus on inequality in income when inequality in wealth might be more relevant, and there is also the issue of social mobility. There are many aspects to understanding these issues, ranging from the technical aspect of not choosing one particular measure that might have perverse results through to an understanding that inclusiveness and, indeed, social welfare are multifaceted things in themselves.

The key is to define the question carefully at the beginning; we should not start with the statistic and then ask, “Now, what question can we throw at it?” We need to think about and try to define precisely the question that we want to answer and then look for the statistic—or set of statistics—that most closely corresponds to it. I think that Jonathan Athow is right: this is territory for which you probably need a portfolio of indicators.

I thank our witnesses very much for coming today. We now move into private session.

12:12 Meeting continued in private until 12:51.