 Welcome to the 25th meeting of the Economy, Jobs and Fair Work Committee. First of all, I ask everyone present to turn electrical devices to silent or turn them off if they are likely to interfere with the sound system. For our witnesses, the sound desk will operate the speaker so no need to press any buttons when we come to your evidence. I have apologies from Andy Wightman who is on other parliamentary business and also apologies from John Mason, Dean Lockhart, who will be arriving during this session and give their apologies for their lateness. Item 1 is a decision by the committee to take items 3 and 4 in private. Are we agreed to do so? Our two witnesses for the first panel today are Dr Stuart McIntyre of the Fraser of Allander Institute and Professor David Bell of the University of Stirling. Welcome to both of you. Thank you for coming in to our economic data inquiry to give evidence, simply to remind members to keep questions short, sharp and focused and perhaps our witnesses in their answers the same. Witnesses no doubt understand that they do not have to answer every question, but if they feel they have something to contribute to the discussion, please do so and simply indicate by raising your hand if you wish to come in. I will start with a general question myself. Last week, we had a number of witnesses before the committee. In particular, a witness from the club of Rome urged the committee to consider what we are measuring and why we measure it. My first question to our two witnesses today is, are we collecting the right data? Do we understand the purpose as to why we are collecting it and the reasons for doing so and the purposes to which it can be and should be put? I will start with Professor Bell. I will make a couple of points on this. It is important to collect data for a couple of reasons. First, you should be thinking about, if I had this information, would I change some policy decision that hinges on the knowledge or lack of knowledge of this piece of data? That is a sort of utilitarian approach, a sort of cost-benefit approach. You have to bear in mind that collecting data, and I may expand on this later, is an expensive process. The second argument that I would put is that there is a wider perspective on data that a lot of democratic decision-making institutions rely on a well-informed public as well as a well-informed set of policy makers. There is a strong case for having a general basis of information that most western democracies have, and it is necessary to allow for public debate and discussion around evidence, rather than a lack of knowledge. There is an immediate approach on what information I need to make decisions about policy and then this more general approach. What kind of information do we need to be able to conduct a coherent democratic debate? Do you feel that, in Scotland, we are collecting the information and data that we need to approach matters as you have suggested that we need to? We are not too badly served, it seems to me. There are areas where information is not great, and there are also problems that the nature of information is changing. Relying on what has happened in the past and the ways of collecting data from the past are not necessarily going to last through the long term. For example, I understand that you have had a discussion of the shrinking sample size in the labour force survey. The difficulty of collecting data, either from households or businesses, is becoming more and more difficult, partly because individuals' time is more and more precious to them, partly because they see it more as an infringement on their... Well, they are constantly being bombarded with requests for information. Why should they treat a request for information from the Government, different from one from Tesco or different from one from Facebook? There are multiple pressures on people to provide data and they become more and more selective, and that is a problem in terms of encouraging people to provide data. A smaller sample size might not be much of a problem, but it could be because it might, if you want to disaggregate your data extensively, become a problem. Secondly, it might lead to biases in your data, so you really want that, if you have the labour force survey, that you are getting an accurate estimate of the unemployment rate. If people are selectively not responding to surveys, for example, young men who are often out when interviewers knock on doors are less likely to respond to surveys, so you might have a job getting a good estimate of youth unemployment, simply as a result of that process. All of these things have to be taken into account. Perhaps Dr Stuart McIntyre has a few thoughts on that. Just to pick up a couple of points that David just made there, I think that there is a distinction between high-level data, which is internationally comparable, which is frankly all a lot of people want to talk about, measure or consider. I think that there is good coverage within Scotland of all those key indicators, subject to some of the concerns about sample sizes and other things, but I think that there is a second stage to this, which is thinking about programme and policy specific data. Rather than focusing on management information about how well the programme is working in terms of consumer feedback, if you like, from people on the programme, I am thinking more about how we can capture the data that we need at the point at which we are designing policy so that we can properly evaluate it. I think that this is the area where there is a lot less attention, and economists have a toolkit of methods that let us evaluate programmes in certain cases, even when that has not been thought about. I think that capturing that data within the programme that allows it to be evaluated as something that complements the high-level stats, but at the moment there is a big focus on what is it doing to GDP. The reality is that the transmission mechanisms are so varied between that intervention and GDP that are trying to track down the contribution of a particular policy to GDP is quite difficult. Inevitably, we have to have that higher-level data, because we have to do things that are internationally comparable, but I think that there is a second stage that more attention needs to be given to, which is capturing the data that we need in designing programmes to properly evaluate them. To pick up briefly on David's point about sample sizes, this is a big issue with the level 4 survey. It is one of the reasons why it has been a move towards using more administrative data to try to get around some of the problems of poor response rates to surveys. I think that there is definitely a move in that direction. More probably could be done and should be done. To be honest, it is being done. I think that we are at the early stages of fully utilising some of the administrative data that we have. Richard Leonard has some questions that may relate back into what David Bell was talking about. You have begun to address this already, but it is really just to ask you for our benefit what you consider to be the strengths and weaknesses of the current suite of statistics on the economy that we collect for Scotland, and, secondly, from your vantage point, as people who work with this data day in, day out, are there any specific improvements that you would recommend were made? It seems to me that, for example, the trade data is pretty difficult for all kinds of reasons. There is a question of why you need the trade data, which is an open question, but, clearly, in relation to the question of democratic debate, rather than the question of, well, we are going to take a new policy direction, there probably is a case for enhancing the trade data. It seems to me, for example, that income data is an interesting issue. I was quite surprised looking at the papers for this morning. There was no mention of inequality or distributional aspects of the economy, so I take it that inequality is one of the issues that may be of considerable interest, certainly to this committee. If you are trying to compare the incomes of the rich and the poor, you have to have enough data in what I would call the tails of the distribution, the very high levels of income or the very low levels of income, that creates a necessity around the overall size of the data that you collect. Now, HMRC has got data—this is Stuart's point—about administrative data on incomes, which, well, there are issues around accessing it fully, but those kinds of data are having more access to those data, which would be of interest. Of course, we are moving into a situation in which Scotland has control over income tax, so it is not just a general interest question. It is a particular interest to the Scottish Government. I am picking up on Stuart's point, but I am making a point that perhaps has not been made to the committee. If you are looking to evaluate policy where you think that the outcome of that policy may take some time to come through, then there are great benefits in having what are called longitudinal data, where you measure businesses, households or individuals at a point in time, follow-up later, follow-up later again and so on. In relation to that, for example, I would commend the Scottish Government's growing up in Scotland survey, which started in around 2000 and followed children born in that here. If you are looking to find out what are the effectiveness of early-years intervention policies, then if you have information about children and their parents at a very early stage in their life, you know what they have been exposed to in terms of early-years interventions and then you follow them up into their 20s, then you can track back to see if those early interventions had the desired effect or did not have the desired effect. Those kinds of studies are called longitudinal studies. The UK is the best country in the world for longitudinal studies, but there are opportunities. Unfortunately, they are very expensive because you have this continuing follow-up. That is only a couple of examples that I have given you, but I think that they are quite relevant. I will pick up a few points. I agree with David with respect to the trade data. I think that one of the things with the trade data is that we have three different measures that get at trade—the index, manufacturer's exports, exports, statistics, Scotland and the HMRC data. They all tell us slightly different things, but probably the best among them is the export stat Scotland data. One of the challenges for that, if you look at the documentation that the Scottish Government put out, is response rates of firms. Firms are not required to complete that, and as a result, the sample size is not as comprehensive as it could be. There is an issue around particular types of firms that cannot be answered by larger firms or wherever. One of the things that we outlined in the submission that we made to the committee was that moving towards a situation with the Scottish Government can compel firms to complete that information. On the same basis, if ONS had that power, it would now seem infinitely sensible. It would be one way of broadening the coverage of that and improving the sample size at the same time and the quality of the data. That was just one point on that. One of the things that is quite important to recognise is that Scotland is very well served with economic data, both in terms of quality but also timeliness relative to other parts of the UK. We get GDP data for the first time this year—we had it in less than 100 days after the end of the quarter—typically 110 or 115 days. We will not find out until December this year what happened to the north-east economy in 2016. Relative to other parts of the UK, we are actually pretty well served in terms of timely quality indicators. There are things that I know that the Scottish Government is looking at that will drive improvements in some of those statistics. The Scottish Government is now getting better access to things such as the monthly business survey that it uses to produce the index of manufactured exports, which might have spill-over effects for GDP data, for instance. There is a programme of continuous improvement both on the Ons side and on the Scottish Government side to try to produce better quality, more timely data. It is worth stressing that. From an academic perspective, one of the things that I would say is that, when we are thinking about some of those headline indicators, one of the things that we need is a long enough consistent time series. If you are trying to forecast GVA or most other macroeconomic indicators, you need a long enough time series to estimate that model. If we are investing in new data series or changing existing data series, we have to consider just the compatibility of that both over time and also relative to other parts of the UK. There are still time lags. For example, productivity data and sub-regional productivity data, business research and development data, there is quite a lag in some of those things. In an era where we are facing shocks, whether it is the oil price shock or the possibility of the effects of Brexit and the new powers that the Scottish Parliament has got, it seems to me that timeliness is extremely important. I wonder whether you would comment on that. The other thing that was extremely helpful in the submission that Stuart McIntyne made was what he was saying about capital flows, inflows and outflows and the extent to which foreign direct investment data is robust or not. The Scottish Government talks a lot in its reckoning of how well the Scottish economy is doing around the E&Y attractiveness survey and so on and so forth. I wonder whether you have any views about how robust and reliable and how comprehensive a picture that that paints. On capital investment, Dr McIntyne, there is little in the way of data on investment for Scotland either in the aggregate or by the sector. It seems to me that that is a fairly important piece of information upon which to understand what is happening to innovation, investment and what the long-term prospects for the economy are going to be. I wonder whether Dr McIntyne, Professor Bell and I want to comment on those two aspects. I think that you are right. One of the obvious things that is missing from the landscapes is the better understanding of investment and, in particular, FDI. I will not comment too much on the E&Y attractiveness data, because I do not know how they put that together. However, the takeaway that everyone takes from that is that Scotland is the number one destination for FDI within the UK, which is fine based on the number of projects. If you look at the number of jobs, we are not. It is a good illustration of reading beyond the headlines, which is sometimes really important. If you look at the number of jobs for FDI projects in Scotland, it is not as much as it is elsewhere, and we do not rank as highly. However, in terms of the number of projects that we do, if you take the same project and divide it in two, you can see how that system can quite easily be gained. I will not comment much more on that, but I think that you have identified what is a very obvious gap in the landscape. The only other thing is that it is costly for businesses to supply data, and quite a lot of the delays are based around legal requirements for them to submit data. If you think, for example, even of self-employment and those people who are not under the PAYE system, they actually have around 15 months after the end of the tax year before they have to submit their last cash payment for income tax. While that persists, you will have to estimate self-employment income over that period of time. There is a question about matching up the cost in terms of lost opportunities to make decisions because data is not timely against the cost to businesses of making that data available more quickly. I think that HMRC is in the process of trying to get individuals to more or less continuously supply data on income. Obviously, with information, sorry, with IT systems now, it may be more easy to do that. It may not be necessary to continue with the 15-month delay, which may have been all right for a time when IT systems were not so sophisticated. There is just that issue about how important it is to get the data absolutely at the end of the relevant quarter. Can I just ask one very quick? Very briefly, and then a follow-up from Gil Paterson. Going back to what you said about trade flows, you did not mention input-output tables. Do you see that those have been an important measure of intra-regional trade and trade between Scotland and other parts of the world? As I said, there are three main indicators through which we get the data. What the Scottish Government statisticians then do is balance principally with the export status of Scotland data with the supply and use tables in the national accounts. The IO tables provide us with a really nice sexual coverage, which is richer in sexual detail than export status Scotland, but it is the output of the use of various different sources. If you pick up export status Scotland, it will not perfectly align with the IO data, but IO data will give you a much richer sexual picture and is constrained in balances with the rest of the economy, whereas, as we understand, the export status of Scotland is a survey of certain firms in certain sectors. Follow-up from Gil Paterson. There is a lot of talk about exports, and there are certainly questions in relation to that, but put my business hat on that looking at a competitive, particularly with Brexit, we do not have very much information on imports in a business sense, and, of course, what agencies who are charged with improving the Scottish economy have to make decisions on support. We would need that information on imports, because, in a business sense, it might be the easiest shift for a business to do to replace the imports within Scotland itself, and yet we are so light. How do we overcome that? Where do we go? How do we do it to get these numbers? I will make a couple of points. I think that the vast majority of imports into Scotland will originate in the rest of the UK and will either come from the rest of the UK or have been channelled through the rest of the UK from some foreign destination, so it might be quite difficult to track those. My second point is that a lot of international trade now works around complex supply chains, where goods cross borders, possibly multiple times. I heard last week a discussion around the car industry and some components crossing out of it and into the UK 42 times. If goods are moving back and forth across the border, in this case the Scottish border, a lot of times it becomes quite difficult to know exactly how to track those. I am just saying that there will be a set of imports that will be relatively easy to identify, but there will be a lot—for some businesses—trying to identify what goods or what components of goods are important and where there is domestic value added will be quite difficult. I guess that I would pick up a couple of things on that. One is—and this picks up on Richard's point as well—although, although not as timely as some other measures of trade, the information in the national accounts about import flows is not as detailed as we like, it is not as timely as we like, but it is there. If we think about what the alternative routes to try and gather that data might be, Brexit and requirements around rules of origin might help with the administrative data around that. However, if you look at, as David said, where most of Scotland's imports come from, it is here—the rest of the UK—and we have a huge issue about trying to estimate those flows or it is potentially the rest of the EU. The HMRC data on imports does not do a great job of capturing that either. One thing I want to note is that some colleagues, as part of—I know that Rebecca Riley was here from the National Institute last week—one of the projects under the UNS's new economic statistics centre of excellence, led by one of my colleagues—or the work plans led by her—is trying to use innovative datasets to try and understand the flow of inter-regional trade within the UK. Things like looking at financial transactions to try to help us to better understand the flow of goods and services within the UK, but it is incredibly complex to try to back that out. I agree with you—I will get personal experience—that the amount of trade that lands in England is held in warehouse to an end-shirt, and I understand that. However, the importance of that, particularly if you have the economy committee, and it must be so important to the economy to understand that question. The other thing is that we are all politicians and there is a great debate. We cannot provide the population with the answers to quite fairly straightforward seemingly questions in relation to imports and exports. He is a great fundamental. My question is whether it is not worth investing in so that we can square that circle. I agree that it is a very important part of the political debate, just the question of the cost of collecting all that information at the level of detail that would be necessary to give us accurate estimates that the public could have reliance on. Jamie Halcro Johnston is a question that may lead on from that. I was going to ask about the key gaps in coverage in Scotland, but that has been touched on quite a lot. You mentioned, Dr McIntire mentioned the north-east on the regional side. Looking at the region on a local perspective, what we can do to fill those gaps in terms of local and regional data? One of the things that I did not touch on earlier in response to Richard talking about gaps in data was that one of the fairly obvious gaps in data that the ONS, to be fair to them, has recognised themselves is around prices and regional price data. There is some work on going just now with the Scottish Government and ONS, partly funded by HMRC and partly funded by the Scottish Government, to improve the sample coverage of the living costs and food survey that produces the weights for the price indices, so that we have a more Scottish index. That is still a gap, but there is something being done in that. I meant to mention that earlier. One of the difficulties is that I read through the various submissions that the enquire, including North Ayrshire Council, had submitted something and various other people had mentioned localisation of data. There is a difficulty in some of those high-level indicators. Think about things like GNI or investment or some of those other headline indicators. The more localised you become, the easier it is to identify fluctuations with particular firms. I know that one of the things that the ONS is really concerned about is the extent to which you can identify people and activities from producing more localised data, which is one of the reasons why it sends the local data. I think that there is a trade-off here. One of the things that the ONS does, which is good in the Scottish Government, is that the ONS, when they produce their GVA data, even though it is not anywhere near as timely as the Scottish Government data, breaks it down to nuts three. Going from Scotland down to the nuts two regions and the nuts three regions, you get a lot more spatial granularity in those data, which we do not have with the Scottish data, but the Scottish data quarterly, whereas the ONS data are annual. There is more that can be done to split those out. It is, for the reasons that we identified in the submission, quite important. I will pick up three points in relation to that. First, Stuart's point on anonymisation. It is something that occupies my mind quite a lot, because I have just been doing a fairly large-scale survey on Scotland's ageing population. In the course of that, I have had to go through three different separate courses or passed three different courses on anonymisation, which is basically making sure that I never present data in a way whereby an individual or a household can be specifically identified. I will not go into the complexity of that, but it is something that the suppliers of data are very concerned about, because the last thing that they want as suppliers of data is that some firm or some individual finds that they have been identified and some information on them has been released, which they do not want to have released. In relation to prices to, again, another point of Stuart's, it is important. One particular area in relation to prices that I think is very important is housing costs. Housing costs in general do not really have good data on that. That is particularly important for differences between the generations. At the moment, the net income of pensioner families is above that of working-age families. The reason for that is that housing costs for pensioner families are much lower than those for working-age families. It is really that that makes a difference. One element of prices that I think is very important is housing costs. I would be one of them. Lastly, in relation to geographical data, we tend to think of the way that economies operate in relation to markets rather than constituencies. What is the appropriate way to think of the labour market in the north-east? Clearly Aberdeenshire and Aberdeenshire City interact very strongly, so it might not make much sense to consider them separately. Clearly there are cases and arguments why you would want to look at very detailed data and you can with, for example, DWP data on benefits. Again, it is a little bit of a caution. If you are wanting to use geographically granularised data, you have to be careful about which indicators it is that you are thinking about and whether they make sense for that granularity. I am an MSP director at Highlands and Islands. We have got Highlands and Islands Enterprise coming in next. One of the questions that I wanted to put them is how they use the information that they are able to collect now. What information is there that is collected, whether it is by different organisations or local authorities or other bodies, that perhaps is not feeding into the national, the ONS collection at the moment? Is there quite a lot of, I would not necessarily call it lost data, but data that is not part of the wider process? If you are talking about the whole spectrum of data, aside from business and economy data, what I have just mentioned, the DWP data is probably that which is most granularised, covers benefits, pensions and so on. There is data on that that is held at the local authority level. That tends to be education data, care data and some data associated with property probably. Other than that, other than exploring novel ways of addressing data using Google searches, I am not sure that there are many other existing data sets that you can interrogate. Obviously, the census, but of course it is historic. It is quite expensive. The 2011 census cost £65 million. The National Records of Scotland does try to update it in terms of births marriages and deaths in local areas, but it is never completely accurate. We will get the next shot at this in 2021. As I noted earlier, the ONS could produce more localised GVA data, although not as timely as we would like it to be in 2015. Out of the APS data of the annual population survey, we get localised information in the labour market. That is at a high level. There are other surveys that take place more regionally, which I would expect to feed into more localised decision making. My final point is that as we move towards better using administrative data, there are opportunities to tie activity more closely to localities. Subject to acknowledging that a business that is reporting in one area may have enterprises spread across, and thus we get into issues around how do we apportion their turnover or their employment or whatever across Scotland or indeed the UK. There are opportunities, but there are still challenges with using the admin data. We have been talking this morning and previous weeks about the gaps in economic data. Given that the ONS is the UK's national statistics institute, are they doing enough to fill those gaps and address those issues? I declare an interest as part of the ONS's economic statistics centre of excellence project. If you look back at Charlie Bean's review that came out in 2016, I think that pushed the ONS to up the game in a number of key areas. We are now seeing a development of—I am involved in some of this—moving ONS's regional GVA to be much more timely. In one part of it, we are working on another, but the aim is to try to improve the coverage of ONS's regional GVA data and make it more timely. Complement the Scottish Government is doing that. There are examples where the Scottish Government's HMRC, as I said earlier, is working to boost the survey coverage of the living costs and food survey so that we can start getting more of a regionalised price index for Scotland, which is important. In a sense, ONS is illustrating a willingness to listen to people. Indeed, one of the things that came out of Charlie Bean's review that he has been challenged on is to engage with users and respond to user needs. The Scottish Government has the same responsibility and does a good job on that. Part of the onus is on us as users and on the devolved administrations and the devolved parliaments to put a bit of pressure on ONS, if we think that there is something that they should be doing that they are not. However, particularly since the Bean review, there has been a real change in approach from ONS, which is quite positive. We are talking about an organisation that carries out 80 different surveys and issues 1.7 million questionnaires to more than 290,000 businesses. The data must be there to a certain extent. A quick google of ONS will find plenty of reports from ONS on England and Wales data only, but if you try to find a similar Scotland-only report, you will find it very difficult. One quick point. I think that ONS produced something like 20 per cent of national statistics, so ONS are not the dominant producer of national statistics. They do a lot. Increasingly, as the Scottish Government is getting access to the HMRC data, which has already had some access to looking at evaluations of LBTT, for instance, the work that I was saying has been going on around access to the monthly business surveys. I think that the treasure trove of ONS data is being made available. There was an issue where access protocols were not clear, and the legality of that was not clear. That has now been cleared with the digital economy act. The question now is, do we see an opening up of that treasure trove of data or not? Just a couple of points. One thing that I am critical of the ONS is in relation to its presentation of data. I find it quite difficult to go on to the ONS website and find whatever it is that I want. For example, OECD and Eurostat give you much more opportunity to interrogate quite big databases and construct your own tables without too much difficulty. They are trying to improve the ways that they visualise data. There have been great advances in data visualisation in the past three or four years. I would say that SPICE is ahead of them in terms of the reports that SPICE regularly puts out. As Stuart Scott said, the ONS is not a monopoly provider of statistics in the UK. The ESRC, which I used to be on one of the panels of, does a very good job in terms of producing other data sets. Here, since Stuart has plugged himself, I will plug myself. One of the things rather puzzlingly that was funded by UK-based departments, namely DWP and HM Treasury, was the English Longitudinal Study of Aging, which is a study that is restricted to England specifically. I have been trying for some years and have succeeded in producing an equivalent survey for Scotland, which we are going to release in December. That has really been a struggle. In general, I do not find a consistent bias towards England and Wales, although it is true that quite a lot of data are England and Wales specific. Sometimes that is due to the way that the Administrations are working, either together or differently. Given the gaps that we have, could any areas of the ONS improve and help assist? Clearly, going back to the price data, which in turn is because you have to have a basket. If you are going to have a price index, you are going to have to have a basket. Therefore, you need expenditure surveys. This is important now for VAT in particular. There is clearly a need to do more in relation to understanding how Scottish households spend their money. That is now in train with the 100 per cent boosted survey of the living cost and food in Scotland, which is paid for jointly by HMRC and the Scottish Government for precisely the reason that David identified, which is the VAT issue. Richard Marsh in his written evidence said staff—this is the ONS staff—are overwhelmingly focused on simply getting the statistics out. Relatively, little attention is devoted to how the quality and relevance of the statistics order delivery could be improved. Given those concerns, should Scotland have its own statistics authority, similar to the ONS, given that Northern Ireland has it, and Scotland's Scottish Government is already having to supplement a lot of the surveys that are carried out anyway to get reasonable information? I am not clear or told that although Northern Ireland has its Northern Ireland statistics authority, whether the outputs from that are any better than they are for Great Britain. Secondly, Scotland could have its own statistics authority, which would possibly be more focused, but I would still argue that it is important to maintain close comparability with whatever kinds of data that is being collected in other parts of the UK and, indeed, wider. I am a little concerned about what might happen in relation to Eurostat once Brexit happens, whether we will continue to collect data in the same way that it is being collected in Europe as a whole. For example, the Labour force survey in the UK is embedded as part of the European Labour force survey. I have been using the European Labour force survey over the last few days, and it asks the same questions broadly. I am not sure that I recognise Richard Marsh's characterisation of ONS. It probably was true at one point, many moons ago of ONS. One of the things that has happened more recently and has been accelerated by walking out Charlie Beans review is forcing ONS institutionally to change its culture a bit. If you look at the honesty with which ONS has, for instance, said, RPI is not a useful measure in its current form and, to be honest, frankly cannot become one, I think that they are being pretty honest about the quality of the data. I am not sure that I recognise Richard Marsh's characterisation. I guess on this thing about Nisra in Northern Ireland, I can see the kind of intuitive appeal of, well, let's set up a separate Scottish Stats authority. Nisra is not fully independent. It is an agency of the Northern Ireland Department of Finance, so it is not perhaps as independent as it might seem. To go back to David's point and build on it a bit, one thing is that we need to keep comparability, and the other is that we need to avoid duplication. If we are in the world now of ONS, HMRC and other data collection agencies UK-wide, making information much more freely available, and making the raw data, the unprocessed data, available to the devolved administrations, the need for the devolved administrations to have their own Stats authority collecting essentially the same data would not seem to be a particularly strong one to me. Before we plug on to some questions from John Mason, do you think that the Scottish Government should be involved in the production of data and statistical information, as well as being a user of it from a policy point of view, just to come in on some of the last points that Dr McIntyre was making? Well, I'm in general in favour of a statistical agency being at arm's length from Government, and in the sense that the national record for Scotland is at arm's length. The UK has the statistics authority, which oversees the ONS and the impartiality of ONS. Whatever organisational arrangements are made, my view certainly is that the organisation who is responsible for the collection and publication of data should be at arm's length from Government. I tend to agree with David. I guess that the other point, and I know that this has come up in previous meetings of the committee, and we address it in our submission, is the issue of pre-release access. It is something that, post the being review, the ONS has removed to its statistics. We have this anomaly in Scotland where we still not only have the pre-release access, but it is much more generous in terms of timescale than it was previously elsewhere. If you like that, that is an obvious area for some action to take place. Thank you, convener. My apologies to myself and the committee that I was late this morning. The other committee that I am on was meeting in Orkney last night, so there are limitations about travel, but I am quite impressed that I am here by 10 o'clock. I realise that there has been discussion already on gaps in data, and just to change that slightly to access to data and statistics, I am interested to know how you, as academics, feel you get access. Clearly, if the data is not there, you are not going to get access, but do you also have the problem that sometimes you believe the data is there and you are not allowed to see it? We did have the suggestion from one witness before that when they asked ons for information, they generally got a positive response, but when they asked HMRC for information, they did not get such a positive response. I wonder if you identified with that. I am so in access to data as something that is close to my heart, partly because, with a slightly different hat on, I access health data. There are huge qualifications around processes to go through in order to access health data. I think that it may be that HMRC is covered by different legislation from ONS data. In general, the kinds of data that I use are accessible by bonafide economists and academics from around the world through the UK data site, which is probably the best in the world for making raw data available. It is suitably anonymised to interested researchers, so that covers stuff like the family resources survey, the labour force survey. It gets more tricky when you want to access data such as the wages data. There you move up a level in terms of access, you have to go into secure access, which means that you have to effectively have your work checked in terms of whether it is revealing any information about individuals. You go to a higher level still, which I am about to go on to later this morning, because I am heading from here to the infirmary, to the bio-quarter, to access health data, and I can only do that there under the supervision of individuals whose job it is to check that I am using it properly. Do you find that satisfactory? Are you saying that those are necessary rules or that it is a little bit over the top? I would say that it is a bit over the top, but I understand that there are huge sensitivities around access to data. In England, in relation to the health data, there was an initiative that basically was disastrous, which involved the sharing of GP data without individual consent. Individuals found out about it and the whole scheme collapsed under their objections. Now it is much more difficult to access health data in England than is the case in Scotland, so there have to be safeguards. Maybe I think that they are a bit over the top, but I understand why they are there. I have much the same experience, so I am currently waiting for health data to come through of a similar nature. Inevitably, we are now developing a series of protocols that are becoming more consistent across data providers in the UK. HMRC, ONS and other providers are now working on a more consistent programme for the training of researchers to train researchers around the safe use of data, so that they are not trying to take information out through which an individual may be identified. That is all perfectly sensible and proportional. I think that academics are getting better at accessing these data and academics are getting better at acknowledging that they have to go through training to access these data. I agree with David that the UK data service site is great for researchers trying to access a range of surveys and, indeed, for researchers to put their own survey data up. However, I think that the only roadblock in the data access now, frankly, is a resource for the analysts at the other side of the desk, if you like, to get the data and inform that you can access it, because they have all the data there and they will maybe ditch off names or addresses or whatever and then put that on a safe haven for use access. I think that the roadblock for researchers now is more about the resources to those analysts. Certainly, that has been, by my experience, trying to access most recently the health data. Do you feel that things are getting worse rather than better? I think that they are getting a lot better on the software side, on the training side, on the comparability of requirements for researchers to apply for data, so that the process goes through. I think that what is getting worse is the analyst time at the other side of that to help to produce those data and inform that you can access them. I think that that has become a bit more difficult for some of those more sensitive administrative data. There is a general problem around having enough skilled people on both sides of the fence to deal with data. We are moving into an era of big data. We have to have people with the skills. Those are statistical skills by and large to deal with that. The ESRC has put quite a lot of money over the years into trying to make social scientists more quantitatively able. It seems that it has had some successes, but probably more needs to be done to get the skilled people in front of the data, because the amount of data that is becoming available is increasing exponentially. In one sense, there are gaps in data and statistics. One of the aims, presumably, is to track progress in relation to the Scottish Government's economic strategy or the national performance framework. In one sense, it is slightly flawed last year and slightly flawed this year. As long as you are comparing like with like, maybe that does not matter, or is that too simplistic a way to look at it? Part of the issue that I might have with the performance framework is that it is difficult to know whether the Government really has a handle that it can turn, which will cause one of those indicators to change for the better or, indeed, for the worse. You choose some measure of productivity, for example. What is it? How far can Government policy affect the change in productivity that may or may not have occurred? It may be something that is just a function of the economic cycle that might be unwound at some other point in time. The Government might look to take credit for it, or it might try to avoid any discussion of it if it has gone the wrong way, but it may or may not have its fault in the first place. We can tell if the productivity is improving or decreasing or staying the same, but we cannot tell why that has happened. I think that it is very difficult. If we knew exactly what the levers for productivity were, we would be turning them as hard as we could. There are lots of debates and discussions that Andy Halden at the Bank of England has been looking at recently. There is no clear set of interventions that people know of that will immediately result. I say immediately, because you may have interventions that only have an effect after five years or 10 years. I guess that that is my concern that we cannot be sure what the causal linkage is. If I add slightly to that and pick up on something that you were not here when I was talking about, one of the things that we have collectively been concerned about is the distinction between what everyone focuses on, which are the very high-level aggregate figures, whether it be GDP, whether it be productivity, whether it be employment or unemployment. Those are used as indicators of progress, which, if you look at the longer term and you smooth out some of those issues around the business cycle, they may well be. I think that the much more interesting, much more important thing is that if we do not know what might work, we might have some ideas on what might work. In each of those policy initiatives or pilots or whatever, we embed proper robust programme and policy evaluations so that we can get to the other side of it and say, well, here is the impact of this. Aggrating that up to an economy-wide impact might become a bit difficult, but at least we can say that we have tried this intervention, we have evaluated it robustly, and we have found that it has had this impact. That works, rather than we have done this initiative and GDP has gone up or down or productivity has gone up or down when the causal link between that is intangible. Jackie Baillie Stuart answered that in part, but I wonder whether I could put the same question to Professor Bell. Pre-release access to statistics still applies to ministers in Scotland, but, as we have heard, UNS and the Bank of England have ceased that practice. Should we follow them? I think that, in general, my view—I am tricky—I do not think that politicians should necessarily be in a better position than others when it comes to the release of data. Thank you very much. Can I move to public finance now? It is something that I know that you have taken an interest in. Given our new powers, do you think that we have enough information on the overall financial position of the public sector? In other words, do you see a value in whole Government accounts, assets and liabilities being set out quite clearly? I think that there is a strong case for that, and it is a case that Audit Scotland has made. Clearly, UNS, there is a debate around contingent liabilities, how do you value pensions and pension liabilities in the future. Those might change a lot with a small change in the interest rate. We have to treat them with caution, but, nevertheless, if you are looking to the medium to long-term financial position of the Government, there are some things that look to be very high probability associated with the ageing population. For example, we might need to be a bit more radical in our thinking if we are really to prepare for the changes in liabilities that we know are likely to come down the line. We make much the same point in our submission. We build on some of the concerns about the ageing build on our budget report a few weeks ago, which is just emphasising that point. If you look at the UK whole of Government accounts, all the Scottish Government bodies and non-department bodies and everything else are in there, the information is there. It is just a question of how we perhaps reparsal the information that is already there. We have heard evidence last week about the definition of inclusive growth, and Professor Bell mentioned inequality. Do we have the ability to measure inclusive growth and inequality with the data that we currently have, or any tweaks that could be made to the currently available data? For example, we have had some of our investigations around the gender pay gap, and data is out there, but it is not disaggregated into gender. That is just one example that has come up recently. Most of the inequality literature uses one of two sources for measuring inequality. One is the family resources survey, which is paid for by the DWP, and the other is understanding society, which is paid for by the ESRC, by the research councils. For granularity, I get quite a good guess at the differences by gender from those two surveys in Scotland. You would not get a guess at all of differences between different ethnicities, because although understanding society has a boost for non-white population, it will not be enough to give you a guess at that for Scotland. The other thing is that, in terms of inequalities, we know a lot about income inequalities. We know not very much about wealth inequalities. That is something that I am trying to address with my aging survey. The main source for that in England is the English Longitudinal Survey of Aging, or the Wealth and Asset Survey. In Scotland, the sample size is not that big in Scotland, and, obviously, the English Longitudinal Survey of Aging does not cover it in Scotland. Over time, we are likely to see more emphasis on wealth inequalities than has been the case recently, where most of the information research that hits the public is around poverty levels and inequality levels, as measured using income, not wealth. I agree with David Adam. I think that there is a range of data there. As always with these surveys, the difficulty once you start trying to get down to particular groups—Oxfam did something a few years ago that he wanted to focus in, particularly on less-herd groups—is quite difficult to get down that line within the existing survey data, because the number of those, almost by definition of the less-herd groups, is less likely to be sampled. If they are, they will be in the distinct minority among that sample. You run into data problems when you start drilling down a lot into different income by-characteristics of individuals. We are going to get access to administrative data from the HMRC, but none of that is really going to address looking at—we will not be able to analyse that in terms of inclusive growth around equality. Administrative data are limited compared with survey data because they are collected for administrative purposes. Therefore, they do not necessarily have all the characteristics of the individuals that you might want to do the kind of analysis by some kind of subgroup that you want. For example, the HMRC data will have gender, but it may not have ethnicity, because you do not have to fill in your ethnicity when you do your self-assessment form. The other thing is—this has been talked about for a while, and I think that it is finally happening—is taking some of those bigger administrative datasets and linking them together. You link together DWP with HMRC. Certainly, with colleagues, we have been talking about this, because if somebody disappears from the HMRC data, they may well show up in the DWP data, and if they leave the DWP data, they may well show up in the HMRC data. The linking these together gives you a better understanding of people's transition through time. However, as David said, the administrative data is collected for that administrative purpose and may well not capture the characteristics that you are interested in. For example, the benefits of linkage our survey is an hour and a half survey of people. We collect loads of information on their characteristics, so we can assign them to some subgroup. We then link that to their health data, and then we can see what kinds of people are frequently in hospital, often go to the dentist, have care packages, and so on. That is the great benefit of linkage of a survey to administrative data. You can put all the characteristics that you might want in relation to assessing inclusive growth together with the admin data, which enhances the power of the survey hugely. Presumably, if you wanted to analyse in-work poverty in Scotland, it would be a case of linking all the availability together and making an analysis of that. If you were able, for example, to link the labour force survey to DWP records, you would be able to look at somebody's benefit history, and the labour force survey is similar to our survey. It asks a lot of individual characteristics about people's health and disabilities, so you can put them all. Once you have that linkage made, it is a very powerful data resource. A couple of follow-ups from Dean Lockhart. Thank you. Just with the inclusive growth, it sounds like there is some useful data out there to benchmark it, but in terms of what we are trying to measure, is there an internationally recognised definition of what inclusive growth is? I am not familiar with it. I mean, I could think of a combination of some combined measure involving both inequality and GDP per capita growth, but I do not think that there is an accepted international measure. Insofar as you are aware, is there a definition in Scotland that the Scottish Government uses to measure changes and progress against the target of inclusive growth? I am not sure of one either. Just a similar question, but this time on innovation, another of the four eyes, is obviously a key driver of productivity innovation. Clearly, everyone is focused on something, but again, it is a similar question. Is there a recognised way of measuring innovation and progress in innovation? I think that work is being done on that. I would think of John Van Rennan, who is at MIT at X of LIC, who has done some work recently on the determinants of innovation and, indeed, on another topic that is mentioned very rarely, which is management quality. However, I would say that these are exploratory research at the moment, and may take some time before it emerges in terms of a national statistic that measures innovation. I think that the UNS has a paper out in July that was trying to scope out the possibility of distinguishing in total productivity between improvements in leo-productivity, capital productivity and multi-factor productivity. One of John Van Rennan's things is that, if we can measure leo-productivity to some extent, we can measure capital productivity, then there is another component of productivity, what is that, and how do we impact on that? Van Rennan's stuff, as David said, is looking at management practice as another form of technology, and so trying to help us to better understand. We can estimate that increase in productivity not due to improvements in labour productivity but due to capital productivity and the residual, but understanding what is within that focuses more on some of the form-level studies where we are trying to understand why some firms are more productive than others. We read about the OECD having top quartile, second quartile of innovation and productivity. How does the OECD measure a country's innovation levels? Is there some sort of benchmark that they look at? I am not familiar with the OECD's measure, but what I would say is that the OECD from experience are very good at producing methodology papers alongside that, so it is something I would be happy to look at if it would benefit the committee to do that, but I would expect that it will have made that information available. We are running short of time, but a quick question from Gil Paterson. How do you see the Digital Economy Act 2017 improving the quality and timeliness of economic statistics in the UK and what can Scotland learn from that? That goes back to what I said before. The Digital Economy Act cements in place now data access and makes it absolutely clear to HMRC that legally they can share a lot of their administrative data with other government departments and also the devolved administration. In terms of data access, I think that the Digital Economy Act moves us well ahead. There is always going to be an issue of timeliness. David touched on this earlier around the timeliness with which people have to report information to say HMRC. That imposes a necessary time lag in the data, but, through the Digital Economy Act, we are going to see much better, much fuller and much more timely access to data for both other government departments but also the devolved administrations. Can I just quote you from the Royal Statistical Society and point out that the DA will not provide devolved administrations with direct access to those data sources, but they should be able to access them through the ONS? Do you think that there should be a bit more, do you think that there is a weakness in that? A bit back to where we are at present time, rather than the devolved administration having the right to change it to shall rather than should? With your own experience, does it make any difference? I am not sure that this is a distinction without much of a difference with respect to what the RSS was mentioning. I think that this is getting at the point that HMRC will prepare the data, the data will be cleaned, the data will be assessed as being a complete data set, and then it will be put on to say the HMRC's microdata lab that will give access to government departments or the ONS's VML, and that is a data product that is there. Versus, for instance, the Scottish Government has been able to directly look up anyone's tax records, I think that that is probably the distinction, but I can't speak for the RSS, but I suspect that that is the distinction that has been made there. I thank our two witnesses and I will suspend the meeting to move to our next session.