Publications on the internet
UNCORRECTED TRANSCRIPT OF ORAL EVIDENCE
To be published as HC 1666-ii
House of COMMONS
TAKEN BEFORE the
Science and Technology Committee
The Census and Social Science
Wednesday 14 December 2011
Professor Tim Allen, Aleks Collingwood, Professor David Martin and Professor Philip Rees
Adrian Alsop, Jeremy Neathey, Glen Watson and Peter BenTon
Evidence heard in Public Questions 46 – 100
USE OF THE TRANSCRIPT
This is an uncorrected transcript of evidence taken in public and reported to the House. The transcript has been placed on the internet on the authority of the Committee, and copies have been made available by the Vote Office for the use of Members and others.
Any public use of, or reference to, the contents should make clear that neither witnesses nor Members have had the opportunity to correct the record. The transcript is not yet an approved formal record of these proceedings.
Members who receive this for the purpose of correcting questions addressed by them to witnesses are asked to send corrections to the Committee Assistant.
Prospective witnesses may receive this in preparation for any written or oral evidence they may in due course give to the Committee.
Taken before the Science and Technology Committee
on Wednesday 14 December 2011
Andrew Miller (Chair)
Examination of Witnesses
Witnesses: Professor Tim Allen, Local Government Association, Aleks Collingwood, Joseph Rowntree Foundation, Professor David Martin, Royal Statistical Society, and Professor Philip Rees, Royal Geographical Society, gave evidence.
Q46 Chair: Good morning, ladies and gentlemen. Thank you for coming here this morning. Just for the record, perhaps you would introduce yourselves.
Professor Allen: Good morning. I am Professor Tim Allen, representing the Local Government Association.
Aleks Collingwood: Good morning. I am Aleks Collingwood from the Joseph Rowntree Foundation.
Professor Martin: I am Professor David Martin, representing the Royal Statistical Society.
Professor Rees: I am Professor Philip Rees, representing the Royal Geographical Society.
Q47 Chair: Thank you. We have a few fairly straightforward questions. As you know, we are trying to understand more about the census process. We have been told that the British population is becoming more reticent to publish information to Government while at the same time becoming more profligate about publishing personal information here, there and everywhere. You have only to look at facebook sites to understand that. Does that suggest that the Government should step back from purchasing data from sources that they commission, and manage with the data that people freely put in the public domain? To put it slightly differently, does the census have a unique selling point that makes it special and different from other sources?
Professor Allen: I guess that trust is a much wider issue; as we know, the public’s willingness or not to offer data is part and parcel of that wider trust in institutions and in public policy and services. The bottom line for us is that it clearly behoves the public sector to gather only the data that it really needs for useful, sensible purposes for which it can account to the public. But at present there does not appear to be an adequate substitute for the basic demographic data that the census provides. That does not mean that we could not think about other solutions, but at present the census is the bedrock on which our demographic work is predicated. Of course, that leads to some very practical things that the public do care about, like how many school places you need in Leeds, London, Bradford or wherever.
Aleks Collingwood: The census is not perfect but we think that it is essential, its key role being the allocation of resources. Mainly, we need an affirmed population count. The ONS produces mid-year estimates, but by the time it comes to the next census in 10 years’ time there will be a slight discontinuity between the mid-year estimates and the overall population count. The main thing that we get from the census, which we would not get from other sources, is the ability to break things down into small areas of statistics for local area information.
Professor Martin: There are two elements to your question. It is quite telling that a major element of what seems to have been a successful publicity campaign around the 2011 census was that of explaining to the public the purposes to which the data would be put. Explaining the impacts on health care, education, planning and transportation was very valuable in justifying the data collection. My concern is that we may need alternative sources; we may have difficulty in doing another census so well, but there needs to be some framework on which the same justifications can be clearly given. Stopping the census without having an adequate strategy for how to explain and justify the replacement would be a worst possible case, because we would be relying on fragmentary evidence without being able to justify what it was being used for. That would undermine people’s motivation to ensure that they were included.
Professor Rees: Your question goes to the heart of democracy. Here in Parliament, you represent the whole people. If you moved from a legally enforceable instrument like the census or an equivalent replacement, you would be relying on voluntary information from interested parties, and that means that your democratic role would be reduced. You would be ignoring all those people who are not vocal enough or educated enough to make a proper representation of their views or characteristics.
Q48 Graham Stringer: Every 10 years, the Government spend about £500 million to collect lots of data. If the Government did not do so, what would you pay to get statistical information? Is there anything that you would pay for, if the census was not available?
Professor Rees: The census has a vital role in many of the resource allocation formulae that central Government Departments use. For instance, in allocating the NHS budget, the Department of Health has to have reliable information on the number of patients and the potential number of patients in future-the number in PCTs currently and in future, and in clinical commissioning groups, and perhaps even individual practices or individual patients. It needs that sort of flow of very accurate information, and if you don’t have it, you misallocate large chunks of Government spending.
Q49 Graham Stringer: The 2011 census had to be adjusted for a number of inner-city areas because it was flawed. Would it not have been better to have used those sources that were eventually used in Westminster, Manchester and one or two other places, rather than spending all that money nationally? Would it not be a more accurate way of getting what I agree is vital information so that we can plan our public services?
Professor Allen: That is a challenging question to answer, because you are not dealing simply with either spending or not spending. There is a cost attached to whichever way you try to access the data. For me, the question is what level of accuracy you need, and therefore need to pay for, and how best and most effectively you can reach that data. From our point of view at the moment, we are very open-minded about examining alternatives to the census, but at present local approaches still rely on the collection of and access to data, and the local approaches, which I understand you heard about last week, are still based on the public sector collection of data. The question is not whether you spend money or don’t spend money. The question is how, most wisely, you operate a statistical system that gets you to the data that you really need-the core demographic data.
Forgive me, but that is not the entire answer to your question, because trying to tease through the cost of the alternatives is quite challenging. From the LGA’s point of view, we don’t rule out the possibility of alternatives. For example, you might be able to systemise the use of things such as electoral registers and the base data that the public sector collects for other purposes, and you might, out of that, be able to create population figures that might be more up to date. However, before you went down that route, you would first obviously want to know what it might cost, but secondly you would also want to know what the weaknesses of those systems were. At present, none of those alternative systems are necessarily fit for purpose, but lots of people are beginning to think about how you might develop systems that would be fit for purpose.
Q50 Chair: Putting it simply, if electoral registration data were to be utilised, one would have to seriously invest in it to get it accurate, because it is not accurate at the moment.
Professor Allen: Yes.
Chair: That is quite an indictment of our electoral system, but that is another story.
Professor Allen: Yes.
Q51 Graham Stringer: May I turn the question round the other way? We accept that there are different ways of getting the core information that we need for planning our public services. Is there anything that could be dropped from the census? Many more questions were asked in 2011 than were asked in 1911. Could money be saved by dropping some of the extra questions, and what questions would they be?
Professor Martin: I don’t think you would save money by dropping questions within the census frame, because most of the cost and effort of conducting the census as it stands is in designing the enumeration process. The removal of one or two questions would be quite marginal on the overall effectiveness of the data collection exercise. Much more fundamental is the extent to which we are able successfully to enumerate everybody-delivering a form to every household, getting a form back from every household, and being confident that that is what we have achieved. There are elements of the census-for example, the large coverage survey, which is designed to work out how well we have covered all those people-which could be very valuable if you were evaluating another frame. In fact, if you moved to a system that used alternative data sources, some of the elements, such as the coverage survey to work out how well we are doing and what the biases are, would be ones that you would want to retain. Within the context of the census, the specifics of the questions asked are not really the ones that determine the success or failure of the whole venture.
Professor Rees: The other point is this. Roughly speaking, last time it cost £1 million a question, in terms of the add-on costs of the questionnaire. If you took off £40 million for the 40 questions, you would still be left with the £310 million needed to develop the infrastructure for administering the process. I don’t think the size of the questionnaire is moot; it has to be kept short and straightforward because everyone has to answer it. That is the real constraint on the questionnaire.
Aleks Collingwood: Essentially, I agree, but fundamentally, the population count is the most important thing about the census. One of the worries is the impact on data quality overall of all the other major surveys, as most of them are sampled from postcode sectors from the census. That would definitely affect all the other surveys. The census-the overall population count and the local area statistics-is also used, for instance, for ethnic boost samples for other huge surveys such as Understanding Society, which is one of the biggest of its kind. Their ethnic boost sample is from that. Without having this overall population count by these small areas, it would be a hugely expensive operation to work out where the minority ethnic groups are.
Q52 Stephen Metcalfe: If you were starting from scratch, if you had a blank slate, what would your individual visions be for data collection? Forget the history-we want to collect all the data that we think is important; how are we going to do that?
Professor Rees: Just to pick up on the point made by my colleague from the Joseph Rowntree Foundation, to get the population spine right is the critical thing that National Statistics have to do. I recommend that Parliament look again at something first proposed in 1753-namely, a population register. It did not pass the House of Commons then, but although you have had some experience with a national identity register, I think that a population register that had legal force and was comprehensive across the whole population would be good. It would be something that people had to do whenever they moved house, wanted a benefit, wanted to pay taxes or whatever; they would have to record their current address. That would give a really good anchor, year by year, month by month if necessary, for any additional population data that you wished to create through surveys or tapping into administrative data sources.
Q53 Chair: That would be predicated not on a 1753 version but on a modern version of data sharing across Government Departments to make any sense of it, would it not?
Professor Rees: That is the solution with a virtual population register, where you try to match individual records through various probabilistic and other techniques, so that, in the end, when you look at all the Government data sets, you are pretty sure that you have about 98% of the population. You will never get 100%, because there will still be a small gap, with people somehow slipping out of the different registers.
Professor Martin: If I had a completely clean sheet, I would probably run a census that was an unproblematic one. The failing is that the reality of delivering the census suffers, at the opposite extreme, from Professor Rees’s suggestion in that population compliance is what makes neither of them work. If I were enumerating a real population, I would probably seek to match the sources that we have in an integrated way, rather than trusting that a number from one place could possibly be matched to an address from somewhere else, and that if we got them all together we would have a reasonable estimate of a local population. I would seriously design a system that put those things together systematically, and I would put some effort into bringing about consistency between sources and the completeness of sources.
Aleks Collingwood: I would highlight how important transparency should be. Right from the start, we should be much more open with the general public about what we are doing and what the main reasons are for a census.
Q54 Stephen Metcalfe: You would stick with the census, even if you had a blank sheet of paper and could design however you wanted to collect this data.
Aleks Collingwood: Ideally, yes.
Q55 Stephen Metcalfe: As is.
Aleks Collingwood: Yes, as is. The thing about the population register is that, in those countries where it works successfully, it is a lot more accepted and supported by the general public than here.
Professor Allen: I would probably reflect the comments that have been made. Nirvana would be some form of population registration system and getting back to trust and public acceptability. I don’t think that, currently, the public would necessarily welcome the step of introducing a register. There has to be a linkage to the way in which public sector data on individuals is collected. There should be a clear public service element to it; you should be able to demonstrate to the public that by going through the registration process you are making certain bureaucratic public service processes simpler for them. In other words, your register has to be multifunctional; it is not just counting people but it is about them having the benefit of accessing services. Culturally, clearly, to shift that would be quite challenging, even if you could demonstrate quite vigorously that it had that element of public benefit attached to it, because, as you identified, clearly there is public resistance to providing data to the public sector.
Q56 Stephen Metcalfe: Whichever system we move to or stay with, are you saying that the Government have a key role in that? It is bigger than just setting standards asking people to collect data on their behalf. It has a role in that. Is that a fair summary?
Professor Rees: Absolutely.
Aleks Collingwood: Yes.
Professor Martin: One of the very important features is that there is no other organisation than the Government which can set up a system that has demonstrable national standing and independence which is not somehow funding the system because they want a particular piece of data or because of a particular regional interest. However those tensions exist, if you try to put things together without that integrated and completely national solution, people will always have possible grounds for challenging the numbers.
Q57 Stephen Metcalfe: Rather than mandating each local authority, setting a standard, and saying, "You do it", which you don’t think would work, it would have to come from the governmental level.
Professor Martin: My personal view is that the independence of the census as a nationally consistent exercise is one of the great strengths that any replacement system would have to have. It would be exceedingly difficult to mandate organisations of different sizes and shapes, with different biases inherent in their populations, to produce something that you knew was using the same methodology in every place.
Professor Rees: Another point about making the census or census equivalent local is that you would be handing over the business of producing statistics to bodies that had an incentive to produce particular numbers. They would know that, in the resource allocation formulae from central Government Departments, a large measure depends on the population. I would not recommend that kind of step at all.
Q58 Chair: Professor Allen is going to tell us that none of his members would do that.
Professor Allen: On the contrary. I am sure our members would not wish to falsify figures, but you would need, for reasons well articulated, a system that is transparent and nationally consistent. It is not impossible to do what you have suggested. For example, we have moved pretty well toward that for property location data for addressing. It is not impossible, but it is hard work.
That brings us back to the cost question. Forgetting any of the other questions, is it more cost-effective to do it once, do it properly, and do it as a national exercise, albeit with a clear engagement about why you want to do it and therefore what data you want to collect; or do you want to disaggregate it so that it becomes a data collection exercise by a whole lot of different organisations, with all the safeguards and management that that would require? I don’t have a figure in mind for the cost benefit, but you would need to through that process.
Q59 Chair: Is it the case that, whichever way you go, the starting point is very hard? Even the Post Office address file, which local authorities rely on so heavily, is full of inaccuracies. It is bound to be in a sense because it is a snapshot in time which is not likely to be accurate six months later.
Professor Allen: That is true of the Post Office’s address file. As for the way in which addresses are collected via local authorities, they have a statutory duty to create the address at the point when a new development is about to come into being. There is a system that feeds that data consistently into a national machine; it is a joint operation between the Local Government Association and Ordnance Survey. I would not claim 100% accuracy; indeed, the census went through a process of double double-checking all that; it was quite a good process because it reflected the data back to local government, saying, "Let’s improve on that." It is not 100% accurate.
Q60 Chair: Is it now moving towards an accurate property register?
Professor Allen: Yes, but it is not 100% perfect. It is not the address file; it is a much bigger database than that, in fact. I can give the reasons for that, but I am not sure that you want to go into the technicalities right now.
Q61 Graham Stringer: If we are looking at using administrative data either as an alternative or a supplement, should standards be set for its collection? If so what should those standards be?
Professor Martin: Yes, although you would have to approach it, as I have just hinted, from the mindset of designing a system that produces the required statistics, because the purposes of the various administrative sources are very different. That is our dilemma with them at the moment. For instance, the purpose of patient registration with general practitioners is extremely different from the purpose of delivering benefits to benefit claimants. The biases in those systems are not amenable to the simple fix of saying, "Record the data in this way." They are recording different data, which is their primary purpose.
It is not a simple matter of telling people to use a consistent format. That is important, but any system would have to understand the inherent differences of who is likely to be counted in those systems, as none of them will be complete. We can converge towards the concept of when someone is resident, what constitutes a household, and thinking about whether we want more consistent ethnicity recording one of those lists, but we cannot make them perform the same function when they are designed for something different. The research challenge that ONS is starting to grapple with in the Beyond 2011 programme is to do with recognising the fact that combining those lists means putting things together that are not exactly the same. You cannot completely standardise them, except perhaps at the technical level, but definitionally they are doing different things. That is where the research challenge lies.
Q62 Graham Stringer: I sort of asked a question-I tried not to-about standardisation, but I don’t really quite understand what it would mean. Perhaps you could explain to me what I meant when I asked the question.
Professor Rees: One of the problems with administrative data systems is that they capture different populations. There has to be a very clear understanding of the concept of population or property register or whatever that each system is capturing. They are not alternatives to the definition used in the census. One thing that would improve the situation would be a requirement that, when central Government Departments develop or change their administrative data systems, and there is in place a census replacement, they consult with National Statistics and the UK Statistics Authority so that they make their changes in collaboration with statistical needs. I point to the failure of the Home Office to do this when developing the electronic UK borders system, which was meant to capture people coming into the country. To my knowledge, there was no consultation with National Statistics that would have made those trip-counts-those journey counts-useful for counting international migrants. There has to be joined-up government in any census replacement that uses administrative data.
Professor Martin: May I try to answer your question more concisely, with an example? We could take two different administrative registers currently in existence in two Government Departments; both may record an individual’s address but in quite different ways. It is a substantial challenge, but a technical one, to use the standard address list that Professor Allen referred to and encourage all those organisations to record addresses in a standard way such that we do not repeatedly have to rematch them. It is a technical standard. The fact that the NHS register could include someone as resident because they are registered on a GP’s list, and therefore use a local service in a particular place, is a very different definition of residence from what we might get if we tried to trace someone from a national insurance number. The numbers are there for different purposes, and the definitional difference is one that the system has to handle. We cannot make them the same, but the technical standards have a long way to go in order to make life easier. I am trying to make the distinction on what standardisation might mean.
Aleks Collingwood: Another of the problems with administrative data is that it can be totally filled up and collected by, for instance, a member of the hospital staff, and subjective things like ethnicity may be guessed. They are not necessarily filled in by the individual, and it will have a big effect on data quality if we do not have some firm rules on how it is collected.
Q63 Graham Stringer: When Members of Parliament were concerned about the size and accuracy of the electoral register, they became convinced that the better data set was held by Experian, the people who hold data sets for the big supermarkets and analyse who shops where and what they buy. Do you think that those data sets should be made available for public use?
Professor Rees: That depends on the terms and conditions that the private sector organisations negotiate with the Government.
Q64 Graham Stringer: It is partly whether we legislate for it as well, isn’t it? We could do that sort of thing here. We could change the rules.
Professor Rees: It would be part and parcel of it for organisations to sully the information if they did not want to make it available-if they felt that it was not in their interests. Speaking as a social scientist and a geographer, the difficulty with all private sector information is that it looks very good when you talk to them, but if you say, "Show me your methodology; show me your peer-reviewed paper and report, which has been refereed and signed off," they don’t want to do that. They keep it completely secret. You have to guess how they have assembled their supposedly good statistics. On that basis, I don’t use private sector information.
Q65 Graham Stringer: Can you guess whether they are doing it to high standards? They just don’t want to tell you because there is commercial value in it. Or are they not doing it to high standards?
Professor Rees: We don’t know: they keep it secret. If you look at medical research, which is the gold standard, there are very high standards in publication and transparency. However, half of the trials funded by drug companies have to be thrown out because they are inadequate. That is a case of private organisations producing information and having to meet high transparency standards but failing. Forcing people like Experian to reveal all their marketing data would be a step too far.
The other point is that marketing companies are interested in the top end of the socio-economic spectrum; they are not interested in the bottom end. They are satisfied if they have all the people with disposable income to spend in the shops, but they are quite happy to miss out of their results all those who are short of the readies.
Professor Allen: I would add that, while I would entirely agree that these are not necessarily peer-reviewed, Tesco rely on their data for marketing and they are a very successful company. Presumably, it is very data-driven, and, presumably, they do get something right for their purposes, but the words "for their purposes" are quite important. As my colleague said, if we were to compel it to publish, the question is the degree to which the data might be different, given the challenges described earlier in relation to the use of other sources of administrative data in the public sector because Tesco’s sample will be skewed to Tesco’s particular market niche. It will focus on that, although it might want to expand to other markets and therefore use other data, but it will still be a particular cut and a particular view of the data and the information. The question mark is not whether they are very good data sets-I have no reason to doubt that they serve their particular purpose, and many of us would love to have a look at them; they would be very interesting-but they would not substitute for basic demographic data consistently and transparently collected to a standard somewhere by the public sector, or triggered and managed by the public sector.
Q66 Pamela Nash: I take you back to Graham’s previous question. Each of you highlighted problems with us using other forms of data collection from public bodies, but there are other countries that seem to be successfully doing that. The example we have been looking at is Sweden. In your opinion, is Sweden doing this successfully, or are there major problems? Do you foresee a time when the UK will not need a census or anything similar, or do you believe that the census provides something unique that can never be replaced?
Professor Rees: Sweden has some of the best social statistics in the world. I am told that at every conference by my Swedish colleagues. They start with a population register and a unique personal identification number, and citizens are used to supplying that to all the administrative registers. The task of merging the administrative registers at the individual record level is fairly straightforward. We are a long way from that. Take the names and organisations you are representing in this list. For this core, there are several mistakes in names and affiliations, so you can see how difficult it is to match lists without a unique personal identification number. For instance, it refers to my colleague Alex Benton, who in fact is Peter Benton, and so on.
Professor Martin: It would be possible to create a good system using linked lists, but that is exactly the core of your question. It will not happen, because it will not evolve to a position in which the lists are sufficient to replace the census. That has to be an active design decision. The reason it works, as you heard, is, first, that there is a common identifier; a step towards that in our world would be to use standard addressing. Some of those technical standards would help us a little, but they would not necessarily resolve the fact that we need to know whether we have the same person; also, there is a level of population acceptance, which has taken a long time to develop. If you investigate how those countries reached that point, you will see that they did not do a census and then decide halfway through not to do the next one; they developed a replacement system, in some cases over decadal time scales. We obviously have better technologies and can make changes faster, but nevertheless we cannot switch off one system and use the other one the next day. When you look at how well they work, you can see they are the product of a very long period of development and we can learn in parts from that, but it would not happen automatically.
Q67 Pamela Nash: We all have a national insurance number. I appreciate that this is not its primary objective, but each person is given one at birth and it is used by a variety of Government Departments. How far away are we from that being used as a unique identifier?
Professor Rees: When you reach 16 that is the case, I understand.
Q68 Pamela Nash: You are given your card when you are 16, but you are allocated the number much earlier.
Professor Rees: So you miss out children aged 0 to 15. The national health register is probably the most comprehensive, because it includes children, but again, it misses people out. If you put them all together, you are probably covering virtually all the population if you can match them. That is the critical research task which I know National Statistics is engaged in. There is lots of social science experience in this matching of record level. But, in the end, unless you have a unique personal identifier on each database, it is still only a probability that a person in this register is the same as the person that has almost the same characteristics in another register.
Aleks Collingwood: I am not sure how the national insurance number works for people who have migrated into the country, but that is not the main point that I wished to make. The main point is that the most important reason for the census is that it is vital for allocating services overall. The only possible alternative would be a central population register to get the overall population count, as far as I see it. In places like Sweden, as my colleague said, it is a lot more accepted and supported, but I don’t know whether that would be more acceptable here than the census; I think that it would be more beneficial to keep going ahead with the census.
Professor Allen: My response is the same as that of my colleagues. With the Swedish experience, it took 20 or 30 years to get where they are; and then there is the cultural dimension-but never say never. I think it is feasible for us to move to that system, but I would worry if we said that we don’t want another census in 10 years on the basis that we are going to achieve the alternative. I would want to see the evidence for that being achievable. There are implications not just for the practicalities of the data standards and all the technical stuff, but also for the cultural shift in how we operate through public sector departments, agencies and so on to get us to the necessary state where we could effectively substitute a different system, making it work and making it cost-effective. That is quite challenging. If you take a long enough horizon and a sensible costing, I am sure that it is possible.
Q69 Roger Williams: The structure of Government has probably changed a bit since 1753-I don’t know whether for better or worse. As the structures change, the Government may need different sets of data and different information. Should that in some way determine the style of the census and how it collects that data?
Professor Martin: There are certain core demographic characteristics for which we have not seriously changed our need over a very long time scale. The unambiguity of those in the central system will remain.
One of the difficulties with the census is that, if you do it only once a decade, you have a long time window waiting for the data and only a small opportunity to change the questions. That is, potentially, one of the great strengths of a system that uses other sources, in that you can adapt on a more measured time scale. But there is also the danger, which must be recognised by anyone making a decision which will affect the future for a long time, that, if there were to be some other change in Government policy that changed an administrative source that was not under the control of the Statistics Authority, it would be very easy to undermine your administrative population system without having a census from which to go back and check. A major change in a benefits policy that took a lot of people out of the register would clearly undermine the system that you had built. It cuts both ways, but my general feeling is that working from sources that you can change more frequently is the right move.
Professor Rees: May I add to that? The options that ONS and others are looking at beyond using the traditional census all include an element of a big social survey. Because it is usually administered by interviewers, it can last much longer and have more questions than the regular census, and you can introduce new questions. There is room for introducing a lot more questions on health of the population than we ask-we ask only a couple in the census which are very basic-and on health-related lifestyles. We do that in current national surveys, but they are all too small to get at the local and small-area detail. But if you had good methodology that linked a good population spine built on administrative data-hopefully, a population register with a large survey-you would be able to keep the information up to date to suit the current needs of society and Government.
Q70 Roger Williams: Basically, the question revolves around whether the Government should be asking specific questions or gathering large databases that can be interrogated for specific purposes.
Professor Martin: It is the latter, because we are not very good at anticipating next year’s specific questions. That is why the census remains so important. We could replace it with something else if it could answer those general questions, but designing a system that asks only specific questions is probably a short-term objective.
Professor Allen: The world changes very rapidly. For example, you are well aware that the recent census switched, although not entirely by any means, to the use of the internet and web-enabled technologies to fill in the data. It seems to me that a bedrock of core demographic data underpins some very basic human needs around health, housing, mobility and the basic functions of government, which is to understand the nature of the population and to be able to respond to its needs. Those things may change, but the consistent factor in all of that is the need for the data that underpins those decisions to make them wise decisions. In that, I would echo my colleagues. Those requirements are surprisingly stable over time.
Aleks Collingwood: I agree that the most important part of the census is the basic demographics, with age, sex and the population count being the most fundamental. Although giving a picture of the whole population in health, education and everything is important, the basic demographics are the main key for allocating resources and so on. However, there are so many very robust national surveys that we can use for other things. For instance, the English Longitudinal Study of Ageing, the Millennium Cohort Study and the Understanding Society survey, which is the biggest of its kind, all cover a wealth of information. Perhaps we should look further into linking these studies to the census.
Professor Rees: Let me give an illustration of why we need a national instrument that is comprehensive. I cite education. We have an excellent school census national pupil database, incredible longitudinal information about pupil performance and so on. Why do we need education questions in a census or the attached survey? It is because the school census does not cover all of the school population. It leaves out all the private schools. That is a decision that the body made at some point in the past. If you want to improve the school census, you would have to extend the coverage. That is the essential purpose of having a national census-like instrument, so that you can cover everyone with those kinds of questions.
Chair: I thank the four of you. It has been an extremely useful session.
Examination of Witnesses
Witnesses: Adrian Alsop, Director of Research and International Strategy, Economic and Social Research Council, Jeremy Neathey, Deputy Director of Policy, Economic and Social Research Council, Glen Watson, Census Director, Office for National Statistics, and Peter Benton, Deputy Director, Office for National Statistics, gave evidence.
Q71 Chair: Good morning, gentlemen. Thank you for attending. Would you please introduce yourselves?
Adrian Alsop: I am Adrian Alsop, director of research and international strategy at the Economic and Social Research Council.
Jeremy Neathey: Good morning. I am Jeremy Neathey. I am deputy director of policy at the ESRC.
Glen Watson: Good morning, I am Glen Watson. I am the Office for National Statistics director for the 2011 census for England and Wales.
Peter Benton: I am Peter Benton, a deputy director of the Office for National Statistics. I am currently leading the Beyond 2011 programme.
Q72 Chair: Thank you very much. We have heard that the last census cost about £500 million. In the current economic climate, is that a justifiable expense?
Glen Watson: I believe that it is. The estimate of the final cost is £480 million. In real terms, as opposed to cash terms, and taking account of inflation and population growth, that is about 35% more expensive than the 2001 census. The estimated economic value of conducting the census far exceeded that £480 million. The business case prepared for the Government and considered by the Treasury in an earlier spending review estimated that the economic value of the census to the UK was probably in excess of £1 billion. We did not try to map every single case and every single use of census data that has been made, because the number was starting to get so large; we thought that we had probably gone far enough.
Jeremy Neathey: I would essentially agree with that. As you have heard all morning, the census is probably the most effective mechanism we currently have for measuring the demographic profile of the UK’s population. Therefore, it is a critical resource for social science, and, despite the cost, the benefits to social science research are enormous. Again, as you have heard most of the morning, there isn’t a reliable alternative to the census.
I pick out a few of the issues that have been brought to the Committee’s attention. The level of geography is a key one; coverage is another. Those are key areas in which other data resources just cannot achieve the same level of coverage as the census. There is also the key question for social scientists of continuity over time and being able to measure change, which is the life blood of social science and important in all other aspects of the way in which Government work and so on. It may be expensive, but it is worth every penny. We are not averse to the idea of looking at alternative ways of doing it in future, but the cost of the last census was a worthy and worthwhile investment.
Q73 Chair: Let me put you on the spot. ESRC’s budget for the current year is about £174 million, give or take.
Jeremy Neathey: Grant in aid, yes.
Chair: We are talking about a project costing the best part of £500 million. If I was able to magic a transfer of £500 million as an adjunct to your budget for the next 10-year cycle, what would you do with it? Would you choose to run a census, or would you do other things?
Jeremy Neathey: As I say, there is no reliable alternative at the moment. There are several things that we would consider doing, but I suspect that many of them are already in train through the 2011 programme, which you might hear more about in future. In a very modest way, we are working with ONS in looking at those and ensuring that the academic community is actively engaged in the process of looking at the alternatives. You have heard some of the issues that surround trying to put an alternative in place. Those are the issues that need to be addressed head-on to see whether it is feasible to move beyond the current model. If you were to give me £500 million, I suspect that I would spend a good part of it in the same way as ONS is doing now, which is to develop an absolutely secure evidence base to take us in a direction away from the current model, with its associated costs. Essentially, we would need to build a similar evidence base to that being built by ONS if we are to transit out of the current arrangements.
Q74 Chair: If the census was stopped, you are arguing that it would be necessary to collect a lot of the data in another way-a national address register or whatever-but that other things would be necessary in a core data set. Would it be possible, adding them all together, to do it for a lesser sum, or would you end up reinventing the wheel in creating a census?
Jeremy Neathey: I suspect that, ultimately, you could do it for less. If you look at models in place outside the UK, I suspect that their investment in censuses is less, but I think that some are comparable. But the up-front investment in getting that right would still be significant, given some of the issues that have been raised about the definition of other data sources, their quality and comprehensiveness.
I shall not go over those issues again because you have heard them more eloquently articulated by others this morning, but they would still need to be addressed regardless of whether the funding was being directed through the ESRC or other sources. In any case, I suspect that the up-front investment would be significant. Over time, the cost profile would probably come down significantly once the systems were in place and could be deemed to be reliable and robust. On that journey, you would have to consider the costs and benefits of doing certain things, particularly on issues such as geography; the position of local geography would be a key factor among other issues. I suppose my short answer is that you probably could save money but it would take time. The experts here today on my right may take a different view.
Peter Benton: May I chip in at that point? The fundamental basis on which we will be evaluating the potential alternatives is by understanding the benefits that they provide for all the uses that people have articulated-the small area geography, the multivariate rich picture of the population-and setting against them the cost of the alternatives. There are very real questions about the alternative costs, how high the peak of that cost would be in setting up an alternative, and how low the cost would be in steady-state after getting over that hurdle. At this point, we do not know the answers to those questions. That is the purpose of the Beyond 2011 programme, and ONS is taking a good look at those questions.
Q75 Chair: Have you ever sat down with any of your customers in the various Government Departments and produced a cost-benefit analysis of the work you are doing?
Peter Benton: Yes. As Glen explained, in undertaking the 2011 census we looked carefully at the costs and benefits. We evaluated the benefits for the 2011 census primarily by looking at the process of resource allocation to the health service and local government, understanding the areas that might accrue if the population statistics were wrong and where the funding might move around. But there is a whole range of other uses. We did not actually value the financial benefits of all the small-area data that is used by social scientists, and one of the challenges is to understand how we might do that. We are looking to work with the social science community to put a value on some of those benefits so as to inform the decision.
Glen Watson: May I add one point? It is also important to look at the international context here. In constructing the case for the 2011 census, we looked long and hard at the cost of running censuses in other advanced western democracies. The UK census stacks up rather well compared with that of the US, which spends something like three or four times per head of population on its census, and a number of countries conduct a census every five years, including Australia, New Zealand and Canada. Consequently, over a 10-year cycle, they spend more than twice as much per head of population as we do on a census. Although the cost has grown, and although it is right that we should look at a programme of alternatives for the future, it is worth keeping the international context in mind.
Q76 Chair: Mr Watson, it would be extremely helpful to have a paper on those international comparisons, if you have any data.
Glen Watson: Yes, we can follow up with that.
Q77 Stephen Metcalfe: Between them, your two organisations are the largest publicly funded collectors of social data. What should the balance be between state-gathered data and research-directed information? If you have finite resources, which side of the fence is more valuable? Where does the balance lie?
Glen Watson: In terms of constructing a census or a replacement for the traditional field enumeration census model, I would say that we have to take account of all uses of the information, whether it is state use by central Government or local government, or whether it is by the research community, the voluntary sector or the private and commercial sectors. Our job is to weigh up the strength of the case being made for different types of question, different types of data and different types of analysis, and to come up with something that optimises the value to the country as a whole. We don’t prioritise one sector over another-state above research, or research over state-but try to come up with a balanced view that meets as many needs as possible within the budgetary constraints and the bounds of what is logistically possible and publicly acceptable.
Adrian Alsop: Indeed, they are complementary in many respects. We have a regular dialogue with ONS to ensure that that complementarity is made the most of both now and in future.
Jeremy Neathey: It is worth noting that the data currently produced by ONS and other Government Departments is extensively used by social scientists. In some respects, the boundary line is not irrelevant-I would always like more money for ESRC to collect data-but the statistics from our UK data service show that, of the top 10 data sets used by social scientists which access that service, half are funded by ESRC and half are funded by ONS. There is a sense in which the collective is the important thing in terms of access to resources. We are working with ONS and other Government Departments in a number of ways to ensure coherence. We have something called the UK data form, which brings ONS and other Government Departments together to talk about data issues and to hammer out common issues and problems; we also have a framework called the national data strategy, which tries to identify key data issues that are of interest to all organisations. We have a regular dialogue going on, which, in some respects, allows us to optimise the opportunities for social scientists by working with ONS rather than in opposition to it. We have a pretty comprehensive data infrastructure in place-it is one of the best in the world-but we could always improve on it. In answer to your original question, I think that we are in a complementary space here.
Q78 Stephen Metcalfe: The UK data strategy plays an active part in getting that balance and making sure that there are no gaps.
Jeremy Neathey: I think that it does, and I hope that ONS would agree. Certainly it is a framework that we use to drive forward our own strategy in working with ONS.
Q79 Stephen Metcalfe: You say that, potentially, we have the best system in the world for collecting and using data, but how well is data collected locally, and how does that co-ordinate with national statistics? Let me add a caveat to that. If data is collected locally and funded from the public purse, should it conform to national standards so that the data is consistent across the whole country?
Peter Benton: Local collections happen in all kinds of places for all kinds of reasons. People have their own reasons for undertaking those things, and the funding they have is for a particular purpose. It is not for ONS to say, "You must do it in this way." We can set standards and give advice on particular classifications that people might use, but the census and the information that it provides, and the information that we would need to provide in future, needs to provide the bedrock that lets others calibrate their local surveys to a national set of information in order to adjust for missing people and the bias that results. The national picture gives the bedrock that enables the local activities to function.
Jeremy Neathey: I agree that the census plays a critical role in framing how we go about the different forms of data collection that we fund. A colleague in the previous evidence session talked about one of our biggest surveys, which is the biggest in the world. It is the household panel survey called Understanding Society, which is spread over 40,000 households. But, in order to deal with issues such as ethnicity, we use the census as a frame to do the sampling, and to find out where we need to go to ensure that we get a representative spread. Going into local areas and doing analysis on where you put your boosts and how you organise the whole data collection strategy for major and expensive data collection exercises is guided by the census. In that sense, the census provides a frame, but it also provides an approach to standards on how to go about collecting data on a whole range of areas.
Glen Watson: It is not just in the academic or research community that that is true. That is also true in the private sector. I know that a lot of market research companies and other research organisations use the census results as a framework for designing and constructing their own samples and survey instruments that serve a whole host of different needs. It is true within central Government as well, so when ONS is designing a household survey-we do many such surveys-we typically use the census to provide the national framework for constructing the sample. When we try to take sample estimates from the social surveys and gross them up to something that is akin to a regional or national population, the census results are used to scale up those figures.
One more thing is related to this, which is the important part that the census plays in providing denominators for national rates, or regional or local rates. Whether it is a mortality rate or an unemployment rate, or some other rate, that is the sort of thing that we use and see in the press, and policy makers use on a daily basis. There is often a denominator sitting underneath that is derived from the census or from the population figures pulled from the census. It is quite a complex picture in terms of the spin-off uses of the census results.
Q80 Stephen Metcalfe: I have a final question. Listening to what you have said, I suspect that you will not agree, but Les Mayhew told us that the census as it is at the moment stifles social science innovation. Is that a fair comment, or do you disagree with that?
Adrian Alsop: We would disagree with that, as you might expect, because of the point about complementarity that I made. The existence of the census has not in any way prevented us from developing Understanding Society. With the birth cohort studies, we are in the process of adding social and biological data to the major data sets further to enrich their capability. While there may be other factors that bear on innovation in social science, the census does not in any way restrict our ability to innovate.
Glen Watson: My feeling is exactly the opposite. I believe that it actually encourages innovation and provides an incredibly rich treasure trove of data that other organisations can use and build on. That is true of social research instruments, but it is also true, for example, of the way that we will be exploiting the 2011 census results through the web, and the sort of web views of the data that we will be providing and the machine-to-machine transfer of small-area aggregate data that will be used by other website builders in the private, public and research sectors to power their own websites and mash together the census data with data from other sources. That is very much the way that the world wide web is going, it is the future direction of the internet, and the census will be part of it.
Q81 Graham Stringer: I return to Andrew’s initial questions about the value of the census. There was no census in 1941, and, arguably, between 1945 and 1949, there was the biggest change in the provision of public services in this country that there has ever been, with the health service, pensions and various other things, but as far as we can tell they were done pretty well. Does that not suggest that the census is less necessary than you are making out?
Glen Watson: I am not a historian. I was not around and I cannot comment in any detail on how social policy was developed in that era. I would imagine that, by the time the 1951 census was taken, it was eagerly awaited and was reporting on some incredible changes in society that had taken place over the 20-year period that possibly then sent people off in all sorts of different policy directions. It was such a different time that I find it quite hard to answer that question.
Q82 Graham Stringer: I am not surprised that you find it hard to answer the question, but it is an interesting point, is it not? We saw the planning of the health service, a change in the ownership of many industries that had previously been in the private sector, and changes in social security and pension provisions. As I said, they were done pretty well. In some ways, they were socially administered better than we do it now. The census data that would have been available at the time would have been 17 or 18 years out of date. Don’t you think we have something to learn about that before we spend £500 million on the next census? Do you think it is a field worthy of study?
Glen Watson: It is always worth looking back to see how census results have been used and how statistics generally have underpinned the development of public policy. It is also worth considering changes to how policy was developed over that period, with a greater reliance on evidence and a greater expectation of transparency. We are living in an age when, if there is a significant policy change, people expect, and have a right to expect, to see the evidence and the data on which that change of direction is based. Perhaps it was less so in the 1940s.
Q83 Graham Stringer: A central issue in the previous evidence session and this one has been administrative data compared to census data. How usable do you believe administrative data is?
Peter Benton: There are clearly a lot of potential benefits in the existing administrative data sets. As you have heard, the NHS register covers a large number of people in the population, but not all, and the electoral roll covers a large number of people, but not all. There are challenges in coverage, but there are also significant challenges with definitions.
You have heard people talk about why the information is collected; it is there to support a particular purpose. Taking the example of the NHS register, somebody might live in Winchester and travel to London during the week, and they may choose to register with a doctor in London during the week because that is the convenient place for them to visit a doctor.
Q84 Graham Stringer: That applies to everybody.
Peter Benton: Okay. If you tried to count the population from those sources without understanding some of those definitional issues, you would get a wrong result. Some of our early analysis of those sources in the Beyond 2011 programme suggests that the NHS register in London over-estimates the population by 10% or 15%. In other parts of the country it is lower and more of an accurate reflection of the rural parts of the country where people do not move around so much. If you were to use it at face value, you would make wrong decisions. The challenge is to understand those patterns of coverage and those differences in definition, and build statistical models that enable us to produce accurate outputs on the definitions that we want to use. We have some ideas about that, and we are discussing our thoughts with others, but it will take a year or two to really understand whether we can do that.
Q85 Graham Stringer: Mr Watson has just made the point that we live in a different world from that of the 1940s. People want to see the evidence, and, if the evidence is not there, they will sometimes challenge things legally. Would the public accept the use of administrative sources, even with the statistical analysis and changes that you are talking about? What are the risks involved and would the public accept it?
Peter Benton: To an extent, they are unknown at the moment, but we know that people have strong views about the census. There are some who very strongly support it, particularly genealogists, who love to have a historical record. There were others at the time of doing the census who clearly said, "We don’t think Government should be doing these sorts of things. It’s an intrusion." The range of opinions is similar on the use of administrative data sets. Some people say that it is obvious that we should use it; it is more efficient for Government; but others say that it would be an intrusion of their privacy. That debate is probably in its infancy. People are not necessarily well-informed about data sets, how they are currently used in Government and how we might use them in future. As we start talking about some of the research that we might do in the Beyond 2011 programme, opinions may start to emerge and change.
Q86 Graham Stringer: Is the core of that debate about having contemporaneous information from administrative data sets against having a national data set that is comparable over hundreds of years? Is that the basis of the argument?
Peter Benton: There are many facets to the kind of information that people want that we need to consider. There is accuracy, there is the amount of small-area detail that people can get, and there is the multivariate analysis. We have heard all those things, but the idea of frequency is really important. The census currently provides information once every 10 years. It might be possible to move to a system where we could produce good statistics every year by using those administrative data sets. The question that results is this. We may be able to do it more frequently, but can we do it as accurately, and can we get information down at the small geographic areas that people use? If we were to rely on a survey to plug some of the information gaps that we simply cannot get from administrative sources-for instance, information about carers-you might move to a survey to do that. If you had a survey, by definition, it would not give you detail down to small geography. There are some real choices and trade-offs on accuracy, frequency and geography, in particular, and that is the focus of the current consultation that we are running with users in the Beyond 2011 programme.
Q87 Graham Stringer: All the nationally collected administrative data sets require checking and correction. Which is more difficult to correct?
Peter Benton: At the moment, I would say that they are all difficult to correct. Do you mean which is more difficult in terms of people missing or being put in the wrong location?
Q88 Graham Stringer: I suppose I mean: which would take more resources and effort to check? Which would be more technically difficult?
Peter Benton: The challenge of knowing whether people are in the wrong location on a particular register is probably one of the larger ones. As I explained, it is not only about people who have two addresses. For instance, people may move house and don’t need to see a doctor for two years, and therefore they don’t tell a doctor that they have moved. So they are still recorded at a previous location. With the level of population churn that we have, that is a real issue. Trying to understand who has been correctly or erroneously included in different sources is something of a challenge, but it is not insurmountable. There are potential ways of dealing with it, but we need to do some research to see whether it is feasible, what level of accuracy it would achieve, and at what cost.
Q89 Pamela Nash: We are now in the middle of December 2011, yet we are still waiting for the first tranche of information to come from the census. Can any of you shed some light on why it takes such a long period of time to collate the information?
Glen Watson: Census day was 27 March, as everyone knows. The census field work finished in June, and we then spent a couple of months sending people out knocking on doors, reminding people of their obligations and chasing up non-responses. The capture and processing of all that information at our large processing facility in Manchester started at about the same time. The initial capture-what we call keying and coding-of some of the written-in responses on the census form has now completed, just about, very recently. We are now in the phase where we are doing further statistical processing of that data within the Office for National Statistics, and we are embarking on a process of quality assuring all the results.
It might be worth explaining that people tend to think of the census as a count, as a form that every household fills in and sends back, but that is only partly right. That was the case historically, but there are now other parts to the process. They include matching the census coverage survey records, which is a 1.5% sample of households in the country, with the returned census forms. That gives us an assessment of those households that must have been missed or did not fill in the form so that we have a basis to adjust the results and our final published results are a best estimate of 100% of the population. That is quite a complex and labour-intensive process.
We also use administrative sources from a range of places, including council tax data, data on benefits and the school census data. We use all that information, not at an individual or personal level to link it, but in terms of the statistical counts at a small-area level so that we can confront the results of that against the emerging census results. We can then ask the question whether the results are right. Does this point to any anomalies? Do we need to investigate or adjust? The combination of the adjustment for the undercount, using the census coverage survey and all this quality assurance, is quite a substantial task. For example, to help you visualise it, we have a system that can produce something like 80,000 diagnostic charts to compare the census results with all of these other sources, to check area by area that the results are plausible and that they do not need to be adjusted for any anomalies. We also have to get data from many other institutional areas of the country, for example, from military bases, prisons and boarding schools. We get data from lots of these other sources to check against the institutional part of the census form completion that was being done earlier in the year.
It is an incredibly complex process, and it will be next summer before we are in a position to get out the first set of results. The more detailed results will follow after that. In fact, the more detailed results that start coming out later next year are often the ones that social scientists eagerly await when we issue the results, for example, through micro-data samples or linkage to the longitudinal study. These are the sorts of things that social scientists are waiting for. The first results to come out next summer will be the population estimates by local authority, age and sex. They are certainly eagerly awaited by those in Government who want to use the census results to update the next round of resource allocations to local areas.
Q90 Pamela Nash: I appreciate the amount of work that must go into providing analysis of that detail, but is there any way in which the raw information that has been collected can be published at an earlier stage?
Glen Watson: I am sorry, but can I be clear that I understand the question? Are you asking whether there is any way in which the raw information can be published in terms of the census counts rather than simply taking account of the adjustments that we make?
Pamela Nash: Yes.
Glen Watson: No, we don’t do that, and we don’t intend to. That is partly because it could mislead. We are sure that there are differential response rates in different parts of the country. We set some targets this year for the response rate, and we are very pleased to say that it looks as if we have met them. For example, it seems that we have a better than 80% response rate to the census form in every single local authority, and we have achieved a national average of about 94%. Nationally, that is a bit better than was achieved in 2001; but in the local areas, where there were some problems in 2001, we have done substantially better this time. We have done a lot better in London, for example, where we had a lot of problems in 2001. If we were to publish accounts without those adjustments, it would be misleading. Some parts of the country have a 98% response rate and some parts might have an 85% response rate, and that would lead people to the wrong conclusions.
Jeremy Neathey: There is a general acceptance within the social science community about the complexity of trying to get the numbers out, and accuracy over speed is probably what the social science community would want. Of course, we would like to get hold of the data as quickly as possible, but that constraint is acknowledged. It takes time to get the data that the social scientists really want-the micro-data and some of the longitudinal data out there-but there is a recognition that delivery of the data used for resource allocation modelling and so forth probably has to come first. There is an acceptance that that is the way it has to be, even though we want to get there quicker, if possible.
Peter Benton: In the past, we used to do exactly that. After the 1991 census we published raw counts, followed by adjustments for coverage. In the end, the users said, "It is confusing. Please don’t do that next time. Just give us the best results that you can when you’re ready."
Q91 Graham Stringer: The obvious question resulting from your answer to Pamela’s question is how you know that you have returns of 85% when you are trying to do the original count.
Glen Watson: We have two mechanisms. First, as people have already said this morning, for the 2011 census we built a national address register; we brought together all the different address sources-from Royal Mail, local government, Ordnance Survey, utility companies and other places-into a single high-quality list of all the addresses. We know what proportion of returns we have had from those addresses. That is one measure.
Q92 Graham Stringer: So it is a percentage of the addresses.
Glen Watson: That is one indicator. It is a little more complicated than that, however, because it is possible that a lot of those addresses are vacant or not yet occupied. It might be possible that within an address a census return came back but two or three people were missed off it. All these things are possible, and it is the comparison between the census coverage survey and the census itself that allows us to make an estimate of the national response rate. There are methodological papers on this, and there is a very complex statistical methodology that leads us to those conclusions, but that is the basis of it.
Q93 Graham Stringer: When you are doing your check with the 1.5%, one of the things that concerned me about the 2011 census, among other things, was that, if you look at affluent areas, say Buckinghamshire, and compare them to poor inner-city areas such as in London or Manchester, the disability statistics are higher in the poorer areas. That may be true, but is it not what one would expect, and it might represent the literacy rate of people in Buckinghamshire compared to elsewhere? Do you correct or check that, as it would have a terrific impact on the funding of social services and health?
Glen Watson: We do a number of checks. The first thing is to try as hard as we can to get the response rate up in the inner city areas with traditionally hard-to-count populations; for instance, with young men and most ethnic minorities, we find historically that their response rates are lower. Our community engagement, and our publicity campaigns and marketing, was all very much focused on making sure that we got the message across to those areas, and that we did it in a nice, simple-to-understand way. For example, with our partners in local authorities, we ran something like 7,000 completion and awareness events up and down the country, trying to bring people in from the community and neighbourhoods, explaining what the census was about and how to do it. We had people who could speak the languages of the people who lived in those communities and, there and then, we could help and advise them on how to fill in the census form.
But we also do other things. For example, we do a quality survey, a small survey at the end of the census operation, where we go back out and conduct in-depth interviews with people to go through and check their understanding of the census questions and, through a conversation, to get the right answers from them; rather than it being self-completion, they have someone to advise them. It is a very small sample, but we bring those results back to head office and analyse them, and we can then compare them to the results of the census itself. That enables us to come up with estimates of how accurately and reliably the questions were answered. Part of the collection of material that we publish with the census results will be the response rates by areas and population groups, the extent to which questions were filled in or left blank, and information on how reliably we think the questions were filled in. I do not claim perfection in any of this, but we are very transparent, and we publish all of that information so that social scientists and other users can form their own opinions and draw their own conclusions about how good and reliable the data is.
Q94 Graham Stringer: Do you think that their perception about disability is correct?
Glen Watson: I know that disability is one of the barriers to being able to engage with the census process-I know that-and we have worked very hard with the Royal National Institute of Blind People, the Royal National Institute for Deaf People and various other groups to try and help people. Whether or not the returns from people with disabilities are of the same quality and reliability as others, we really need to have been through the process of publishing the results to be able to analyse that.
Q95 Graham Stringer: My final question is this. The census is a huge one-off event every 10 years. We get better at most tasks if we do them twice. Sometimes, people are not up to the task. It is my view that Mr Cook was not up to the task in 2001; he did it very badly.
I am asking the question from both ends. If you are going to have a census every 10 years, do you think there should be an attempt to recruit people who are young enough to do it twice? Secondly, if it is clearly going wrong, how do you deal with somebody who is not up to the job? It is a one-off, rather like running the Olympics. If you do it only once, you either get it right or wrong. How do you deal with that from both sides?
Glen Watson: Some countries deal with it by doing a census every five years. That gives a lot more continuity from census to census in terms of the people, the systems and the processes, and they can incrementally improve and change the way things work. That is one way of dealing with it.
I cannot envisage a situation in this country where the Government would give us the money to do a census more frequently than every 10 years. If we stick with the 10-year model, whether or not there is a traditional census or some variant in the future, or whether we have a completely different model, if it carries on every 10 years, it is important to try and get the expertise from the last census, and as much of it as you can, and incorporate it into the process for the next time. For example, I was not involved in the 2001 census, but many of my senior managers and my team were. Quite a lot of people at headquarters designing and planning and working on it had the experience of the last census. In fact, some have had experience of the last two or three censuses, and we try to retain that expertise as far as possible.
I cannot comment on Mr Cook’s experience, but my understanding is that when he came in to the Office for National Statistics most of the preparations had already been done. I came in to run this census in 2007, and you would be surprised how much of it was already decided and set out. It is a very long-term project, and the thinking and design takes place quite early on in the process.
Q96 Chair: In terms of changes in society, it was fairly obvious from an earlier conversation about mobility that the population has become considerably more mobile, and very rapidly so. Given that there is a 10-year gap between censuses, how can the design of the next census take into account changes in a society that is moving very fast? How do you go about it? How can you ensure me it is accurate?
Glen Watson: Between 2001 and 2011, we saw the population of the UK grow by more than 3 million, and two thirds of that was a direct or indirect result of migration. When doing the 2011 census, we knew that we had a very different population than we had 10 years ago.
Q97 Chair: It has also become more mobile, with people working away.
Glen Watson: Yes. The population is more diverse. There is more migration and more internal movement. There is more mobility, with more people living in different arrangements and family members living in different households, and there are more people working away from home. We addressed that in 2011 by adding some new questions and gathering new data to help us understand it; for example, you will be aware that for the first time there was a new question on second residences. That will allow us to publish population estimates that are not only on the usual residence basis-
Q98 Chair: That was a very poorly designed question, I have to say, for those of us who had to fill it in.
Glen Watson: Okay, but we tried very hard when designing these questions.
Q99 Chair: That was a good example of a question that was incredibly badly designed. If you have a second home, it should have said, "If you are living in a second home, you should be filling in both forms at the same time." It was all very vague. If you followed the questionnaire and said, "I’ll be filling in this one today, and on Thursday I’ll fill in the next one," you would have sent confusing data back to the system.
Glen Watson: We tried various alternatives and tested them on thousands of people. They tend to work very well in most cases, but there will always be examples and there will always be people for whom it did not work awfully well.
Q100 Chair: Perhaps I can’t read. I found it incredibly difficult to follow the logic of that question.
Glen Watson: Okay. I am not sure what to say about that. The reason that we added that question, and the reason why it is important, is that we will be able to publish statistics not only on the usual residency basis but also on some other population bases. We will be able to publish estimates of working week populations. We will be able to publish estimates of students, whether they are in their term-time address or at home. For the first time, using the census, we will be able to publish estimates of short-term migrants who are in the country for only a few months.
That is part of the richness that we will be publishing, and social scientists will be able to use these different cuts of the data. The other reason why we add questions like that is because, if we do not, people will find excuses not to fill in the census. Unless we include something about visitors or second homes, people who are away from home on census night will find a reason not to fill it in at all, and it is quite important that we do not miss people.
Chair: Gentlemen, thank you very much indeed.