Science and Technology Committee - Minutes of EvidenceHC 322

Back to Report

Oral Evidence

Taken before the Science and Technology Committee

on Wednesday 18 January 2012

Members present:

Andrew Miller (Chair)

Stephen Metcalfe

David Morris

Stephen Mosley

Pamela Nash

Graham Stringer

________________

Examination of Witnesses

Witnesses: Richard Bartholomew, Joint Head of the Government Social Research Service, Department for Education, and Jenny Dibden, Joint Head of the Government Social Research Service, Department for Work and Pensions, gave evidence.

Q101 Chair: I welcome the two of you here. Thank you for coming. Would you kindly introduce yourselves?

Richard Bartholomew: I am Richard Bartholomew, joint head with Jenny Dibden of the Government Social Research Service representing social science across Government and externally.

Jenny Dibden: I am Jenny Dibden, joint head of the Government Social Research Service.

Q102 Chair: To help put this in context, perhaps each of you would explain your role as half of the Chief Social Scientist for Government.

Richard Bartholomew: We completely share the role. Jenny and I have worked together very closely over the years. We work together very well. It depends on the availability of particular things. We will tend to major on one issue and then the other will pick it up, but we often share responsibility. It is a completely shared role which gives us more input in terms of time and so forth. We are as one; so address us both, please.

Q103 Chair: We are interested in the use of social science by Government. Who is ultimately responsible for setting the parameters for what is required by Government, and whether the information they receive is adequate?

Jenny Dibden: We are from the Government Social Research Service, but there are five analytical professions in Government. We work very closely with the other analytical professions. There is a Heads of Analysis group to ensure strategic coordination, chaired by Sir Nicholas Macpherson at the Treasury, on which sit all the heads of profession. In terms of coordination and deciding the strategic direction for analytical work in Government, we work through the Heads of Analysis group. Sir John Beddington is also on it. Within Departments, all the various heads of professions, departmental directors of analysis and chief scientific advisers also work together. That collective working and focus on issues within but also across Departments means we have a serious amount of input in deciding what the agenda is and securing it.

Q104 Chair: There is a fuzzy group that is the customer; there is not a single customer.

Jenny Dibden: Within a Department, individual Ministers have responsibility for particular areas and the analytical community within it will support them. Depending on the issue, we would support them too. Across Government, it depends on what the issue is. If it was particularly pertinent to science-

Q105 Chair: Are the issues driven by Ministers or by you?

Jenny Dibden: It depends. We have an agenda. For example, one of the things you have been talking about in this inquiry is census data, but there are also other sorts of data. We have an agenda on data which we have been very keen to push for some time, but it fits very well with Government priorities-

Q106 Chair: But who has established that?

Richard Bartholomew: Ministers will have specific needs and policy officials will advise on them. Our role through the directors of analysis and chief scientific advisers is to take an overview and see where we need to push and look at cross-cutting issues. We have set up a number of groups to look at different cross-cutting issues, because we need to see the whole picture of what Government and the public need in terms of information, working also with the National Statistician. There is that overview as well as the specific needs of each Department expressed by Ministers and their senior policy advisers.

Q107 Chair: That includes the public need for data and, for example, the desire of universities to have access to it?

Richard Bartholomew: We are very active on access to data, and we have an active policy of putting our data into the archives. There is a public use of the data that individual Departments collect that is also an important use, and national statistics is very much about getting data to the public. The new approach to transparency is all about getting data out to local users so that they can use it. That is also part of our role.

Jenny Dibden: It is also important to remember that all the analytical work on which Government draw for reasons of policy strategy and delivery is not done just within the analytical community in Government; it is also done externally. It is either contracted out or we draw on the vast amount of work done in academia, often funded by the research councils. The sum total of input on the analytical side is both inside and outside Government, and it is important we make sure that the external community is able to do that work.

Richard Bartholomew: If I may give you a specific example of the new major birth cohort study funded through the Economic and Social Research Council and Medical Research Council, recently I have been asked to be part of its governing board. It is clearly a study of huge benefit to the academic community in the future and ultimately the public, but one can also see very specific interests for departments, particularly my own, the Department for Education, in terms of looking at the development of children. That is a joint enterprise. We bring these things together by working alongside the academic community and the research councils to identify the areas of future data we need to pursue commonly.

Q108 Chair: From the outside, the structure of social science in Government seems to be confusing, to say the least. I take it from your response that there is not a single Minister who has coordinating responsibility. Is this because your unit has effectively been the most successful and broken down all the silos in Government or would it be better run in a different way?

Richard Bartholomew: David Willetts is the Minister for Science, including social science, so in terms of science funding he has broad public responsibility for that.

Q109 Chair: You do not report to him.

Richard Bartholomew: No, but we have a very strong network that reports through Sir Nick Macpherson to the Cabinet Secretary and the relevant Minister in each case. I do not think it would be that helpful. There are specific issues for each area of policy relating to social science, so it is quite difficult to say that that is all social science. That would probably marginalise the role of social science. We need to be very integrated into each department’s needs if we are to have an influence in making sure the evidence is available to make informed policy decisions.

Q110 Chair: It seems a very sensible approach. Could you apply that to other aspects of science in Government? Do you have a diffused group like you?

Jenny Dibden: You regularly see Professor Sir John Beddington in his role as Chief Scientific Adviser, but he too works through a network. He has a network of chief scientific advisers in Departments, much like we have heads of profession. They work alongside the departmental directors of analysis and all the analytical heads of profession in Departments too. The way in which John is able to do his job and deal with the issues of the day is often through the individual chief scientific advisers but also because they meet as a network, much like the departmental directors of analysis.

Q111 David Morris: In your opinion, how central and important are the census data to Government Departments in their planning?

Jenny Dibden: We have spoken with a whole range of Departments about it. They have also been feeding into the Beyond 2011 consultation that ONS has been doing, much as we have. The view from the users we have consulted is that currently there is no obvious replacement for the census, but they are confident that ONS will do the work required to assess all the alternatives and look at the potential trade-offs we might have to make in terms of quality, speed of data and so on. At the moment it remains very important to Government for a number of reasons such as our ability to forecast, assess potential policy changes and those sorts of things.

Q112 David Morris: What do you think are the major advantages and disadvantages to the Government of collecting these data via the census? Could there be better ways or methods of collecting similar data? What would your opinions be on that?

Richard Bartholomew: The main advantage and the absolutely unique thing about the census, which would be very hard to replace, is the quality and comprehensiveness of local data. That is crucial for many Government decisions, particularly by local authorities but also in other ways-for example, in education in terms of pupil projections. It is that unique aspect that is very hard to replicate through other approaches such as surveys. It is crucial to have that 10-yearly benchmark to give local data against which you can calibrate all the other large surveys. Much of the rest of research, as well as spending decisions and so forth, is dependent on having that regular common approach done on the same day, the same basis and with the same definitions. There may be alternatives to that, to which we are certainly open, but that is the unique feature that even the largest surveys cannot easily provide. There are methods like imputation of data for local areas, but that can become unreliable. If you do not have that single benchmark, although you need other things as well between the 10-yearly intervals, you cannot calibrate back to that single objective point. The difficulty is that you will get increasing bias in all the surveys and will not know its extent unless you have that benchmark.

Q113 Pamela Nash: We know that, at best, at the moment census data are two years old, and therefore two years out of date, before they are in a usable form. If more up-to-date data become available to the Government, do you feel it is flexible enough to respond to that and change policy in time to make a significant difference?

Richard Bartholomew: Absolutely. A lot of the role of members of Government Social Research, statisticians and others is to collect other sorts of data in between the 10yearly intervals. You cannot look at these in isolation; it is a combination. Many of the policy evaluations we do look at the more immediate impact of policies, changing course if they are not working properly, and exploring new options. It is not simply that we can rely on the census for these things; we need a combination of these different surveys. But, to make sure the surveys themselves use accurate samples, they need to relate back to a common basis that is provided by the census. It takes time to produce accurate statistics. The first results of the most recent census are coming out in July, but, again, you can use other data to supplement and compensate for that 18-month or two-year gap. That lag is a very important issue for local authorities, but you have to look at what the alternative would be to that. If there is one, we are open to it, but even large surveys-which may be an alternative-do take time to do.

Q114 Pamela Nash: Do the Government have access to the raw data before it is published?

Jenny Dibden: No, they do not; it is released to everybody at the same time.

Q115 Pamela Nash: So there is no pre-fixed data. Do you think that, apart from the timeliness and the amount of time it takes to publish, any other aspect could be improved that would allow the census data to be of more benefit to you and your colleagues?

Jenny Dibden: One of the benefits of the ONS work on Beyond 2011 is that it is taking a requirements approach. It will look at the requirement for data within Government in our case, and externally, and see whether the census as well as the other methods for delivering that data can be improved in any way. The programme of work more generally has value because it allows us to explore all data sources and see whether we can make better use of them.

Q116 Pamela Nash: Is there anything you would like to add?

Richard Bartholomew: It is a very good source at the moment. It is not suitable for everything. A choice may need to be made looking at the different options between the core purpose, which it has had for 210 years, of the head count-the numbers of people-and the other questions, some of which could be and are collected through other survey methods. If the need is to reduce expenditure-I know it is a costly exercise, although you have to look at the overall value for money of what it provides and who uses it-you might look at reducing the amount of data you collect from every household and supplementing a basic collection with a sample-based collection of more detailed information, which I think is the model used in the United States.

Q117 Stephen Mosley: I want to ask about the mechanics of how a
Department uses the data. If, say, the Department of Health wants some data from the census, how does it get them in a usable format? Does it approach ONS and get the raw data? Does it say to ONS what data it needs and they are supplied, or does it just get the data ONS publishes, which are the overall figures but not specific details about individuals? What is the information, and how does a Department get it?

Jenny Dibden: The individual records from the census are not released for 100 years; so nobody has access to those at all. The ONS works on the principle that data are made available to all. My working assumption is that Government have no extra access to census data. It is very important for us that the external community has equal access, because a lot of the work on which Departments depend in terms of analysis and use of the census is done by external academics.

Richard Bartholomew: We would never see the individual identity of the people who respond. Equally, ONS has a very good approach in safeguarding individual data so that there is no indirect identification. One of the two big innovations in the last 10 to 20 years that have enabled more analysis by both Government and academics of individual level data is the release of samples of anonymised records where there are individual-level data but they are disguised so that it is impossible to work out the identities even indirectly. It is quite a complex process, which statisticians could explain to you, for doing that.

The other one, which I think has tremendous benefits, is the linking of the 1% longitudinal sample of the people who take part in each census. It is the same individuals tracked through time, again under lots of safeguards. That is available both to us and academics under very secure conditions. You cannot take the data away; you have to go to an approved centre, which is ONS or now other approved centres, I believe, where you can analyse anonymised individual records. That is a tremendous resource for looking at causal impacts over time, such as the effects of unemployment on health and various other conditions, because it is linked with vital statistics on death registrations and so forth, and certain registered health conditions. That is a really positive development. We look for further work on that because it gives you that causality, and some interesting results are now coming through. We can link four or five censuses. It is more advanced in Scotland.

Jenny Dibden: An example of that in the Department for Work and Pensions is that the longitudinal study allows us to calculate mortality rates by social class, which is very important in forecasting state pension.

Q118 Stephen Mosley: As to those particular data, the Department for Work and Pensions has its methodology as to how it interprets them and comes out with results. Do you share that methodology with other Departments-for example, the Department of Health?

Jenny Dibden: It is very important in scientific work, including social science, that methods are shared. I do not know specifically what we do in terms of estimating state pension using the longitudinal work, but certainly we would routinely get external experts to look at our methodologies both before we actually use them and then when they are published and used. It is very important that we expose our work. Just as academics expose their work, we would expect to expose our work. If you think about potential policy changes, Government now routinely publishes on the internet equality and impact assessments, and you can see the way in which analysts have worked through the potential policy changes.

Q119 Stephen Mosley: You are talking about mortality rates and so on. DWP is doing the work and the Department of Health effectively does the same work. They can share it, but there does not seem to be a formal process to allow them to share those data as yet. Does that mean they are both doing it separately?

Jenny Dibden: The community of statisticians and economists and the rest of analysts in Government attempt to join up, and colleagues work across Departments. One of the benefits of this very forensic examination of whether we should have the census in the future and other sources is that all Departments-everybody-are being exposed to the real detail of uses of the census. Certainly, within Government there are groups looking at whether we are maximising the use made of a whole range of data. We are also part of external groups-for example, those run by the Economic and Social Research Council-looking at the social science data requirement for the country that affects both external and internal users.

Richard Bartholomew: A key part of the role of the four analytical professions within Government is to network and share methods and approaches so that there is not duplication. I do not think you would find much duplication-certainly not in methodology-and people separately developing it. If we can see it happening, it is our role as heads of profession to make sure those people start working together; so I do not think that is a major concern for us.

Q120 Stephen Metcalfe: Departments across Government have seen their budgets squeezed over the last 18 to 20 months. Has that had an effect on the individual Departments’ investment in research and development in this area in the way they gather, analyse and use data?

Jenny Dibden: As Heads of Analysis, we are always concerned about the capability and capacity of the analyst community in Government. Clearly, with the spending review settlement it was very important that we began to understand the implications of that for analytical capability and capacity both within Government and externally. Heads of Analysis undertook an exercise, which Sir John Beddington has spoken to in the House of Lords Science and Technology Committee, that looked at how Departments were dealing with the spending review settlement specifically in relation to analysts and what they spent on the research budget. The conclusion that we reached on the early returns was that analysis was not being disproportionately affected, which was important. There were indications that in some Departments spending would be preserved, or potentially increased, and in other Departments there would be reductions. We did that work at the very beginning of the spending review period, and it is quite difficult for Departments to anticipate exactly how they will spend over the coming years of the spending review. We have gone out again to see how that is panning out, what the current estimates of spend are and whether there are any issues.

On the back of that, the Heads of Analysis group also wrote to permanent secretaries saying, "We have done this exercise. We will continue to monitor, but it is very important that within your own Departments you have an understanding of your analytical capability." Another aspect is that, when Departments were proposing to eliminate or reduce the scope of particular data sources, we asked them to make sure there were not dependencies in other Departments on those data sources.

Richard Bartholomew: My experience is that it is some of the evaluations of the initiatives of the previous Government where the need has changed. My Department has brought to an end some evaluations, but we have also started other new research-for example, in the area of adoption breakdown, which is a priority for the Government. Therefore, there are opportunities. There has been some overall reduction, but it is not disproportionate compared with other reductions in expenditure. There are certainly definite areas of growth in demand for research.

Q121 Stephen Metcalfe: Presumably, there has been some reduction. Therefore, the Departments have to ensure that the research they undertake is of most value.

Richard Bartholomew: Yes.

Q122 Stephen Metcalfe: Is consideration given to the longevity of that research-for how long it will be useful?

Richard Bartholomew: Yes.

Q123 Stephen Metcalfe: What is the life expectancy of a piece of work, and what should someone be aiming for?

Richard Bartholomew: It is difficult not to give a very woolly answer because it depends on what it is and how precise it is. If you are monitoring a policy, you probably need much more regular sweeps of data to see its unfurling impact. If you are looking at a broader change in social attitudes, such as the British Social Attitude Survey, with an underlying change, clearly there is less justification, because from year to year things will change less rapidly. That is a debate we always have with our colleagues and ministers about the frequency we need. It is difficult not to give a woolly answer, but I suppose the spending review horizons are around three years; so there tends to be a regular survey at least every three years, and sometimes there is a good case for doing them annually. It is a question of affordability and the likely change you will get in that period. Some you would need to do much more regularly.

Q124 David Morris: Health and education are the most obvious parts of Government undergoing significant structural change. Do you feel these changes mean that the Government need for social data gathering is changing?

Richard Bartholomew: As a result of the structural changes?

David Morris: Yes.

Richard Bartholomew: My view, working in education, is perhaps different from DWP where the Department for Education is responsible but delivers through a number of different bodies such as schools and local authorities. It is not fundamentally a different picture in terms of the structural change. There is more emphasis now on making sure that the data we provide and collect is the right data for the local commissioners and providers of services, not just for Government. I think Government, certainly in the statistical and research areas, has a key role in making sure there is a common set of data available using common definitions-this is related to making the data more transparent and usable to commissioners of services and to the public-to ensure they are consistent data. In terms of statistics on social care for children and in our educational statistics it is very important that data are there for the people to judge schools against a common benchmark of standards. All the evidence suggests that is the way to get improvement and consistency. There is autonomy in the way schools are run, but there is a common process of accountability using common standards, both nationally and, as far as you can, internationally, hence the emphasis on the new PISA study that has been running for a number of years. Those international comparisons have become increasingly important in education.

Q125 David Morris: Given the level of funding dependent on formulae, where do you think that the greatest effort is needed in firming up social data collection?

Richard Bartholomew: In a sense it comes back to the reliability of those local estimates of numbers, the point about timeliness and, if we continue with the 10-yearly census, finding ways of updating that. If we replace it with a different system, we should still get the regularity, because I know that for local authorities it is a key issue, particularly in large urban areas where there is a lot of change and very mobile populations. We need to be fleet of foot to address that and make sure we use other sorts of data where we know there is rapid turnover. There are possibilities in relation to the school census, which is quite sensitive in terms of the intake of five-year-olds, for example, to local change. By using that in combination with other data-and other departments will have data that are equally sensitive to very quick changes-that is where one would seek more development.

Q126 David Morris: It has been said that administrative data would be a more valuable resource if the data collected were cleaner and more easily collated. What would be the best first step to standardise administrative data collection?

Jenny Dibden: The Beyond 2011 programme will look at the question of standardisation. One of the difficulties here-or it is certainly something you need to consider in some depth-is why we have administrative data. Essentially, it is to administer things like benefits. If we move to a process of standardisation, we have to ensure that the purpose of the system remains. We still need to be able to pay benefits; that is the primary driver of that system. There are parameters within which we need to consider standardisation. I am sure there is scope for some standardisation, and we will need to make sure that we get the incentives right to do that and to ensure that there is a good early warning system so that the administrative system does not get changed and the first we know about it is when the statistics are impacted. There is a whole basket of issues being worked through, but I do not think we can get away from the fact that these are administrative systems for particular purposes and we need to see how we can make more use of them for statistical purposes. But it is quite difficult to see how you can turn them into statistical systems when their primary purpose is administrative.

Richard Bartholomew: To add to that, a specific thing that I think is important is the identity of individuals. There has been some discussion in other evidence about address registers. Clearly, there is a big problem about identifying that it is the same individual, if you are going to use this for census purposes, to avoid double counting. We know that sometimes people are not in administrative databases at all or their information is very out of date, and some may appear twice. We spend a lot of time trying to iron out those problems. That matching process to make sure you count each individual only once is a key challenge. It needs to be looked at very hard before we rely entirely on those administrative sources for population counts.

Q127 Chair: As an aside, Mr Bartholomew, clearly this Committee is interested in why young people end up studying science and engineering. Has the Department ever commissioned any longitudinal work in this area to try to track the exposure that young people get to good science education and the outcomes?

Richard Bartholomew: We have a longitudinal study of young people that has been running for some years. It has followed one cohort, and they were 13 to 14 years old when we first interviewed them and their parents. The purpose of it has been to track their decision making in some detail, because it is interviews about their intentions and ambitions to look at their subject choice before they finalise their GCSE decisions. We have then tracked them through to 19. Those data are now available. At least in the early stages of their career decisions we can see their educational choices, what influences they had and how that varies by social class and other background factors like parental influence.

Q128 Chair: But, interestingly, the Association of Science Education would argue that that is starting far too late; you need to start at school or pre-school.

Richard Bartholomew: That is right. It is probably a little more difficult with primary age children, but some research has been done in that area. We have certainly tracked them from 14. I agree that may be an area we need to look at. Maths, in particular, is the key to a lot of these things in terms of interest in science. It would be interesting to try to pick that up in younger age groups.

Q129 Chair: The reason I ask is that it is not just obvious to this Committee, but to a very wide part of the population that that would be a very interesting study to do. However, is it the case that it is not done because there is not a sufficiently strong drive from a Minister to get that work done because the Minister will not be around when the outcomes are published?

Richard Bartholomew: No, I do not think so; I think that underestimates Ministers’ interest in future developments. That is why they are interested in longitudinal data.

Chair: I have known a lot of Ministers in my 20 years here.

Richard Bartholomew: Even if that was the case, I do not think it is true. It is our role as the people responsible looking to the longer term. It is a key role of Government Social Research to look to the longer term and what Ministers and their successors will need several years down the road, and to lay the basis for that. You have to argue quite hard for that investment, but the announcement by David Willetts on the decision on the birth cohort study shows there is a real understanding of the value of that longitudinal data.

Q130 Pamela Nash: If a decision is made to discontinue the census and replace it, what do you envisage the Government will have to do to allow more data sharing between different Departments?

Jenny Dibden: The ONS work will identify the things that need to be done to facilitate data sharing. One of the important aspects of that work is the testing process through which ONS will go to identify how best to make use of alternative data. One of the things to which we need to pay close attention is the overlap between any new methods and the census.

As to issues that it needs to look at, there are legal issues to do with access to data, linking data and also matching data. There are also technical issues about definitions. Do you know whether an individual in one source is the same individual in another source? There is also a public aspect to it. All of the proposed alternatives that ONS will need to look at must be assessed against the criteria of whether the public will find them acceptable. One of the reasons we have a census in this country is that we do not have a population register. There is a perception that the public in this country would not be keen on a population register, although they exist in other countries. There is definitely a public aspect to being able to match data.

Q131 Pamela Nash: Is that perception because of the extended level of data sharing of personal information?

Jenny Dibden: Population registers in other countries have been running for many years. I guess there is a question about how long in this country we have not been comfortable with the concept of a population register. If we set one up, it would take many years to get to the point where it was robust. The public seem to have some difficulty with the concept of data matching and sharing, much as, on the face of it, they had with the census.

Q132 Chair: Who says that?

Jenny Dibden: Quite a lot of work has been done on trust in science, data and statisticians. To take the example of statisticians, if you look at surveys and work done on whether people trust statisticians, yes, they do. Do they trust statistics? No. There is a whole range of reasons for that.

Q133 Chair: If you asked a simple question of the public that was posed in the interests of children, "In child support cases, should the CSA have the right to access the DWP and HMRC records of both parents and so on?", you would get an overwhelming yes for that.

Jenny Dibden: Yes, but I think that is about making the case. It is like the census. The census had to work really hard this time to make the case for why people should fill in their forms. If you are a member of the public, you may well see the link between your census form and the allocation of resources for health or education, but you may not. ONS did a very good job of explaining to all of us just how important it was to fill in the census form and that that data were subsequently used. That is the case for the whole range of data sources. It depends on the issue. You have cited a specific issue, but generally on a source like the census there is a whole range of issues.

Richard Bartholomew: I think there is a case for that, and we would welcome a public debate about it. As analysts we can see the benefits of linking up those data, but clearly there are public concerns. Whether they are shared by everybody, or the majority, we do not know, and it depends on what you are actually talking about. As an example, the ContactPoint database for children that my Department established under the previous Government, which has now been discontinued, was one in which each child would have a record but with very minimal data. There were huge safeguards created around that to reassure people about the uses of those data. It was not going to be possible to use them for many research purposes because of those safeguards, but clearly there was a very strong view in some quarters that that was an infringement of civil liberties and people were concerned. Yet it was set up for the very laudable reason of trying to track children at risk to see whether they had had contact with social services and other agencies. Only about 5% of children or something of that order would have that kind of record; 95% would not. But there were huge concerns about that process and that contributed to the decision to discontinue it, as well as the overall cost of the exercise. There is public concern and you need a debate about it, because there are trade-offs here in terms of the cost of the census versus better use of administrative data. We can see the analytical value, but the public need to be convinced and reassured that they will be used properly.

Jenny Dibden: One implication if they are not reassured is that they will stop responding to surveys. By definition, surveys are voluntary; we do not make people fill them in, so we need them to buy in.

Q134 Pamela Nash: Over recent years how have the Government changed their way of using the data collected from the census and from other sources that are used between every 10 years, and has the collection changed as well?

Richard Bartholomew: From my perspective, we are making much more use of administrative databases. Certainly, in education there has been a huge transformation. We have had the schools census for about a decade. It took a number of years to develop and it has evolved over time. That provides data on about 93% of children. It does not include the independent sector so you have a gap there, but it includes basic information about children and their educational performance. It does not include a lot of information about their social background. It is a census and it gives us tremendous analytical power. We do not need to do some of the more ad hoc surveys we might have had to do 10 years ago to look at more basic characteristics. We can use those administrative databases in combination with surveys to look at attitudes-obviously administrative databases do not include information on attitudes-to pinpoint, perhaps, primary children and their interest in science. That is a huge transformation. It has not happened overnight. There has been a lot of very hard work done by statisticians in setting it up, but it is a tremendous and internationally recognised data source. That is also true in other Departments. One of the major changes is greater use of statistics and greater power of statistics, but there are still opportunities clearly in terms of data linkage.

Jenny Dibden: Within my own Department, two decades ago we would have been able to look at the data in a single system. We now link for analytical purposes data from a whole range of systems. We also link data with HMRC, which now allows us to look at, for example, not just whether somebody has moved off benefit but whether they have gone into work.

Q135 Pamela Nash: I am very interested in how the information collected is shared with the Scottish Government and the Welsh and Northern Ireland Assemblies. How has that progressed since the devolution of power? How much power do they have in collecting the data in the first place? Are they still centrally collected? Personally, I had to complete two forms and I noticed they were different.

Jenny Dibden: The census we have been talking about today mainly is that which covers England and Wales. There are separate censuses in Scotland and Northern Ireland. Clearly, it is in everyone’s interest for there to be coordination of what is in the census returns so that we are able to do comparative analysis across countries. As part of Beyond 2011, ONS will be looking not only at the census in England and Wales but also working with colleagues in Scotland who are looking at the future of the census in Scotland. There is a lot of join-up between the different censuses at a professional, statistical level.

Q136 Pamela Nash: The data collection is coordinated, but are the data collected analysed together?

Richard Bartholomew: You need to discuss that with the statisticians at the ONS, but they are combined for UK statistics. Some of the devolved Administrations may want slightly different questions. There are Welsh language questions and other issues like that. That also gives freedom. There are benefits in terms of slight differences, but from the UK perspective, those statistics are combined. Certainly, the professional statistics network, the Government Social Research network and the economists are cross-UK, and that is a benefit. We link up quite regularly with our colleagues in the other three countries to make sure we are joined up.

Q137 Chair: There is a seamless link between you and your colleagues?

Richard Bartholomew: There are different needs and they are devolved Administrations, so there are different emphases, but we make sure there is common ground on data collection.

Q138 Chair: You work on a collegiate basis.

Richard Bartholomew: Yes, absolutely.

Jenny Dibden: We have a head of profession for social research in Scotland, Wales and Northern Ireland, and we go through a self-assessment process for each head of profession within his or her own Department. I do one for the Department for Work and Pensions. That is peer-reviewed, as it happens, by Scotland, so that we can look at issues in my Department and the strength of analytical capability in social research in Scotland and learn from each other.

Richard Bartholomew: My own staff at a more junior level have annual meetings with their opposite numbers in the other Administrations to look at the data collected on children-children’s care and so forth.

Q139 Pamela Nash: I suppose that is what I am trying to get to. There are different levels of power devolved to the different Administrations. Health is completely devolved to Scotland, whereas you cover the entire United Kingdom.

Jenny Dibden: We cover it; that is right.

Richard Bartholomew: Education is just England, so there are different challenges there.

Jenny Dibden: Going back to administrative data, DWP, for example, has information on the whole population.

Q140 Graham Stringer: You mentioned a couple of examples of longitudinal studies in the evidence you have given. The written evidence we have received says how valuable those studies are. Can you explain to the Committee whether the information in those studies and data sets has affected policy?

Jenny Dibden: The example I gave was to do with state pensions. The longitudinal survey allows us to look at mortality rates by social class, and that is used in forecasting state pensions. Clearly, that is available for policymaking and informs it. There have been changes in policy on state pensions which will have been informed by that. We have a whole modelling system for looking at pensions, both state and private.

Richard Bartholomew: I do not know whether you are talking of administrative data or longitudinal studies like the birth cohorts, but one of the big messages from those in recent years, certainly from the 1958 and 1970 cohorts, is the social mobility question. That is very much on the political agenda of all parties, is it not, in terms of the apparent reduction in social mobility between the 1958 and 1970 cohorts? That is a really strong influence.

Q141 Graham Stringer: But that has exposed an issue without leading to policy change so far. Are there any areas where you can say, "This is the information we have from these longitudinal studies, and this is the policy change we have enacted"?

Richard Bartholomew: I think that is one, in terms of Governments addressing it. For example, the pupil premium in my Department is certainly influenced by the concern about social mobility and improving the chances of children from disadvantaged backgrounds through a specific set of interventions. Clearly, you need a number of interventions with different age groups for the purposes of social mobility, but it is crucial in terms of the arguments as to why one should invest both in early years and the pupil premium.

Another one, which has justified the increased investment in early years, is the EPPE study-the Effective Provision of Preschool Education study-which was set up about 12 or 13 years ago to look at three-year-olds and their experience of early years and the different qualities of early years. That had a seminal influence subsequently in all spending reviews in making the case for improving quality in early years’ provision. Through that study we can relate the quality of the provision to the outcomes of the children at five and subsequent stages. We are now trying to put forward proposals to repeat that study. Obviously, 12 or 13 years is a long time and we need new evidence on that.

Q142 Graham Stringer: When there are policy changes are you consulted in terms of what impact they will have on your data sets? Generally, is consideration given to the coherence of your data sets when there are changes in policy?

Jenny Dibden: By "policy", you mean changes in the way it impacts on a computer system that has a bi-product of producing statistics. It is that sort of change.

Graham Stringer: Yes.

Jenny Dibden: Yes.

Graham Stringer: I had not thought of precisely that.

Jenny Dibden: All major change projects, even small ones, in Government are managed and go through a process of impacting. Within a specific Department we would be among a range of stakeholders that would have sight of that change and would be able to input as part of the impacting process the implications of that change. If a change was to be made to one of DWP’s administrative systems as a result of a policy change, we would have sight of that and be able to say that that will impact on our ability to produce statistics or forecast. That does not mean to say the change will not happen, but it allows us to know what that change is so that we can make provision for explaining any discontinuity in the data series.

Q143 Graham Stringer: That is interesting. Can I go back to David Morris’s series of questions where he was asking about the input of data sets into public expenditure? Is there any difference in the accuracy of your data sets on a regional basis?

Jenny Dibden: One of the strengths of the census is the ability to look at very small geographical units for the variables that the census collects. One of the reasons why at the moment the census is so valuable to Government is that it allows you to go down to those very small geographical areas.

Q144 Graham Stringer: I understand that. I am pursuing a slightly different point. It is not really about the focus but the accuracy of it. It is fairly well reported that fewer forms are returned from poorer inner city areas-sometimes from ethnic minority communities-and therefore there is less accuracy in those areas. Do you take account of that? How do you deal with that? When I talk to my health bodies, they say that our population is massively underestimated, and they always have done, not going back just to the last census but two or three censuses.

Jenny Dibden: That is right. ONS does some very important work alongside the census to look at the quality of it and also its coverage so that it can understand where there has been under-coverage. You are absolutely right. Looking back at the response rates for the 2001 census, although the overall response rate was 94%, for the main metropolitan areas it was 92%, and it went down to 78% in Inner London. That was a huge cause for concern, and certainly in some cases local authorities challenged the basis on which they were allocated resources. ONS has put in a huge amount of work this time to try to boost response rates, particularly in London, but also in other metropolitan areas, to push them back up to a certain level, but if you are using those data you need to take account of that.

Q145 Graham Stringer: Just forget the 2001 census, which had particular problems. If you were to go back to the 1991 census where there would still be lower rates, when you are doing the statistics and advising a spending Department on population, do you say, "These are the actual statistics we have got, but you really need to add on 3% or 4% to get to the real number", or do you just leave the statistics as they are?

Richard Bartholomew: That is what the ONS do in terms of adjustment to make overall population estimates from the census data they have collected to allow for that under-counting and poor response rates you will get from some particular areas or groups, like young men, who we know do not respond well. Therefore, an adjustment to the figures is made before they are used to make sure we are doing everything to compensate for that potential under-representation.

Chair: Thank you very much for your attendance this morning; it has been very helpful.

Prepared 20th September 2012