Official Statistics: 2011 Census Questions - Public Administration Committee Contents

Examination of Witnesses (Questinos 1-128)


19 NOVEMBER 2009

  Q1 Chairman: Let me welcome our witnesses this morning for a session looking at the preparations for the census, in particular looking today at the suggested questions in the 2011 census. We have had hearings previously on general preparations for the census and have not really discussed the questions, but, because the draft census order has now reached the House of Commons and there will eventually possibly be debate and decision on it, it seemed useful to try to explore some of these issues at this point. We are therefore delighted to have Professor Philip Rees with us from the University of Leeds, Keith Dugmore from the Demographics User Group, and David Darton from the Equality and Human Rights Commission. Thank you very much for coming to help us. You have given us some papers but would you like briefly to say something by way of introduction, each of you, as to what your perspective on the questions issue is at this point? Mr Darton, would you like to start?

Mr Darton: Yes, thank you. As I said in the written submission, our major concern is the omission of a question on sexual orientation, or, more precisely, sexual identity. I can rehearse the reasons for this if you are interested.

  Q2  Chairman: In a nutshell.

  Mr Darton: It is primarily because they are a significant group in the population in terms of their overall size, and in particular we are concerned that there is a lot of self-segregation that goes on in this group in terms of self-segregating themselves into particular occupations and particular localities that feel safe, and that that may be very bad for the productivity of the country and for their own life choices, and you can only pick up that sort of information and data in the census at that local level, given the size of it. The fact that sexual identity questions are being asked in the future or in other surveys does not compensate for the fact that they are being eliminated from the census, and, of course, the census is the main source for the planning that local authorities and service providers do. Finally, we think that there would be an accurate enough picture of sexual orientation distribution around the country from the census in the sense that it would accurately reflect the sexual identity of people in the sense of being able to identify themselves in their household, which is a particular aspect of identity which is valuable in its own right. We are not over concerned about the fact that everybody chooses to identify differently in different spheres and so therefore it is a very particular measure but a very valuable particular measure.

  Q3  Chairman: People are worried already about the size of this census, rightly or wrongly. Could I just ask you, if this is your proposal for inclusion do you have a proposal for exclusion to match it?

  Mr Darton: We have a letter from the ONS that says that the reason they are excluding it at the moment is not because of a shortage of space or a cost issue. I think it is right, as I said in my submission, that in the end there is a political judgment to be made in terms of the priorities of different questions. Having said that, I think there are some—and I am hesitant because this is not necessarily our centre of expertise and I think you need expert advice—that on the surface look to us as though they might be candidates. We are not sure, for example, that the accuracy and salience of a question which asks people about their intention to stay in the country. I think there are two questions related to that. We have not seen evidence that suggests that that gives a very reliable estimate, so they may be candidates. We are not entirely certain that the bank of questions on immigration is the best way of getting at immigration questions, so that may be one area. There is another issue which is slightly separate, which is whether it is really required by the Welsh Language Act to leave a space in the England questionnaire so, whilst it would be inadequate only to do things in England, there is some space in the English part which I think could be utilised better and we are certainly not convinced that the Welsh Language Act prevents that happening, which I think has been suggested. I think, although this is a different part of the questionnaire because it is the household survey, that we might question the number of rooms question on the questionnaire because we are not sure that cognitive testing suggests that it is very accurate anyway and it is available from other sources. I think there are some areas and we are cognisant of the fact that this has to be good value for money and nothing is cost free.

  Q4  Chairman: Thank you very much for that. Mr Dugmore?

  Mr Dugmore: I am representing the Demographics User Group, which is on behalf of large commercial companies who use the census. I think it is probably sometimes overlooked that people naturally concentrate on central government's and local government's use and public service planning, but we as a group have been going for more than ten years and the census is an invaluable source of information for decision-making by large retailers and financial service companies and so on. The way we see it is that there are quite a lot of questions in the census which are of broad interest to almost all groups, and one goes back to a long history here of what has been asked in 2001 and 1991 and 1981 and so on, and then at the margins there are questions which are of perhaps great interest to some user groups but not to others. In our case, we are very interested in the broad demographics and social composition status and so on of the population. When it comes down to, for example, carers, I would have to be straightforward and say that that is not an immediate matter of interest to us but we recognise—and I also in the past have chaired the Statistics Users Forum, which is a much broader umbrella of census users—that there are a lot of people in local authorities or social services or whatever who are very interested in the carers' questions. I think it would be unusual for any user group to support all questions. From our own viewpoint we would not be immediately looking towards national identity or citizenship or numbers of bedrooms as questions that are of great significance to us, but obviously there will be other users who will lobby for that and ONS does not have an easy job of reconciling these conflicting demands. The one question that we would like to have seen in is income, which is of interest, obviously, to the commercial world but I think also to decision makers across all public services, and it was obviously a finely balanced judgment for ONS to make. We naturally pointed to the fact that income has been asked in several other countries for many years and has been achieved. We also look to the fact that ethnicity was a very sensitive question back in 1981, where I remember the Haringey census as an enumerator, where there was a lot of public concern about ethnicity but the ground was set and people made the case and it was successfully asked in 1991 and by 2001 it had become embedded. It would have been good if it had been done with income, which indeed is being done in Scotland, so that is our one particular extra if we were looking for one.

  Q5  Chairman: How do you feel about sex?

  Mr Dugmore: It is okay.

  Q6  Chairman: That is the right answer! No, no. How do you feel about the argument about sexual orientation from the commercial world?

  Mr Dugmore: I think it is fair to say that it would be well down our list of priorities. There might be some particular target marketing in seeking to define population segments. I certainly would not rule it out that some of the members of the group might say, "That is quite an interesting one; yes, we are launching a new product", or whatever, "and we are wanting to target particular members of the community", but it is not an obvious one for us.

  Q7  Chairman: People write about the gay pound though, do they not?

  Mr Dugmore: They do.

  Q8  Chairman: That is a commercial—

  Mr Dugmore: Yes. I think it is of potential interest but not right up there in the obvious ones.

  Q9  Chairman: Thank you very much. Professor Rees?

  Professor Rees: Thank you very much for inviting me to speak with you today. I am an academic, so I have no statutory right to demand anything in the census but I use everything that is in there and exploit it to understand what is going on with our population and our society. My submission put the case again for the income question. When I was co-ordinator of the ESRC census programme, which you heard about in June from my successor on this, David Martin from Southampton University, in 2001 we launched a strong case for including income, so I have repeated that. The essential case is that if you want a variable that distinguishes between people who are poor and people who are rich, ask them directly about the money they receive. A lot of academic work and national statistics work has shown that proxies constructed from other variables measuring deprivation in the census just do not discriminate sufficiently between people, between areas, do not identify those pockets of policy that need public action. Obviously, having suggested an income question, which I think the design of has been tested out thoroughly, it is not intrusive. The income question can be administered personally within the household and the husband can answer it A, B, C, D, and the wife can answer it A, B, C, D, and they do not know what their respective incomes are, so it is well designed for confidentiality even within the household. What would I give up if space had to be made and you were persuaded that income was a question to include? I think the new questions on migration are extremely useful; I will be using them, but I think they are not quite fit for the purpose that they are intended for. They are intended to fill the large gap in our international migration statistics but that gap is not just for the year before the census, 2010-11; it is for the whole of the decade. We need a robust continuing system of measuring migration better, so I would have taken the money from the budget and used it to persuade the Home Office to demonstrate that the e-border system, into which enormous amounts of money are being invested, can be used to generate statistics. They have not yet demonstrated that and I have been pressing them in other evidence I have given to the UK Statistics Authority on that in the migration report that came out in July. I think I would drop the rather awkward question about intention to stay. The people who I have asked about this feel it is awkward. It does raise spectres in their mind of are they going to be deported in the near future; is this a way of collecting information, which the census does not collect. I have absolute faith in the integrity and confidentiality of it and of the data that is collected, but people's perception is that this question could lead to unfortunate consequences. The other reason I would drop the intention question is that it sets a precedent for asking people about views of the future. The census is designed to ask what is the situation now or what has been happening over a year or five years, back to your birth date, if you like, so it is information that is relatively factual, whereas intention is pure opinion. I know there is a reason to try and connect it to the international passenger survey but I do not think that is the right solution to the gap in migration statistics. In terms of evidence on the value of the income question, I will just refer you to my PhD thesis, 1979, using a splendid set of income data from the US census of 1960, and demonstrating this point, that it is a much better discriminator of people in poverty, or, in the case of commercial organisations, people who are rich in income terms than any of the other variables. It is better than occupation, it is better than education, it is better than housing value. Income is a critical variable for analysing our social and economic fabric.

  Q10  Chairman: Just on that, is there an argument though that says because we have not had an income question in the past that is a good reason for not having it now because we do not have the virtue of what you are describing, which is the time series that will give us the information that you have talked about?

  Professor Rees: No, I think that is an argument that would mean that there was no innovation at all, so we would not have an ethnic identity question, we would not have had lots of questions if we had not asked them for the first time.

  Q11  Chairman: Mr Darton knows about the future because he is the Director of Foresight. On the sexual orientation issue, so that we can move on to other ones, it did occur to me, thinking about that, that one of the big reasons for wanting census data is for public policy reasons: you want to do things with it. What I am not clear about is what public policy reasons make us want to know about people's sexual identity. We do not provide different services, do we, for people with different sexual orientations?

  Mr Darton: I think we do potentially provide differently tailored services in areas like health and so on. I think though that one of the main reasons for knowing about it is to be able to assess the size of the population in local areas because if, for example, a local area is completely devoid of anybody who expresses a minority sexual identity the question has to be asked, is there something about the culture, the nature of that area, things that local authorities and local partnerships do have some control over, which is in a sense excluding quite a big proportion of the population and may therefore need changing in order for that area to be the most productive to get the best employment force it needs and so on. I think it would be wrong to assume that the census on its own gives you a direct policy response. Clearly, it gives you evidence that there is something that needs looking at. It gives you evidence that there is something that you need to target your policy development resources or your research resources on, and, with a forecasting hat on, and this is well known, we are going to be in a period of pretty tight public expenditure constraint over the next ten years. We have recognised in Parliament that there is a duty for public authorities to try and ensure equality, and it includes sexual orientation, and that steps must be taken where it is shown to end bullying, et cetera. We need to know the numbers of people so that that money is spent effectively. People see the census as a cost but actually these requirements are not being made for the sake of political correctness or anything else. It is because it affects real people's lives and we need to have basic—

  Q12  Chairman: We are not asking people whether they are short or tall or fat or thin or any of these kinds of things, are we, all of which would be fascinating to know?

  Mr Darton: No, but I think there is a difference which has been recognised. Sexual orientation is one of the protected grounds in legislation. We have recognised that with ethnicity, disability and so on there are some characteristics where there is the potential for disadvantage or discrimination. That is a matter of public concern and that has been recognised, so I think sexual orientation is different in that sense, and if we are serious about that legislation we need the data that allows us to act effectively on it and to distribute resources accordingly.

  Q13  Chairman: You mentioned health issues, and I can see that of course there may be—but we do not have separate housing, separate transport system, separate all kinds of provision for people with different sexual proclivities, do we? Your point about if you find out there are not many in an area, suggests a kind of cultural repression. It may indicate there are not any.

  Mr Darton: Yes, I think it may. It suggests areas for further investigation, but I think the same point could be made that we do not have separate services necessarily for people of different religions, yet we ask about religion. We do not necessarily in all areas have separate provision for groups that are measured in the census. What this is doing is directing you to ask the question: is it important enough that you need to start thinking about different services?

  Q14  Mr Walker: Mr Darton, why can you not just leave gay people alone, just let them get on with their lives? We all here do surgeries. If someone was to walk into my surgery and I said, "By the way, before we start, are you gay?", I would expect to be told to bugger off—sorry, to get lost, or go and jump off a cliff. Why can you not just relax about this? We are all pretty relaxed about it. Why are you not relaxed about it?

  Mr Darton: Because that is not what is being fed back to us by our gay and lesbian stakeholders. They are concerned that they have the ability to identify themselves when they think it is appropriate. The majority of them do think it is appropriate to identify themselves within the census. I also think that if you say this is a private issue and not a public issue that is completely wrong. There is a huge amount of evidence on the devastating impact on people's lives that comes from harassment, bullying, targeted violence. We have recognised all of that. There are also, or have been, considerable examples, but we do not know the quantification, for example, of discrimination in the workplace. That actually affects the productivity of the country and it affects the economy. I think it is frankly ludicrous to suggest that this issue is purely a private one and I do not think most gay and lesbian people would agree with you.

  Q15  Mr Walker: I think you spend a lot of time with pressure groups who clearly want to promote their interests, but I personally find your approach and the approach of people like you deeply patronising, and I know, like you, a number of gay people. They do not all share your view. In fact, they are fed up at being defined by their sexuality and you seem to want to perpetuate this problem. You are determined to make it a problem and keep it going as a problem.

  Mr Darton: To be frank, I think it is patronising to say what you have just said because people define themselves in different ways in different situations. Nobody wants to be identified by a single characteristic all the time and on every occasion, but people do want the right to identify themselves with certain characteristics when they feel that that would be helpful to them. If I were a gay man I would not want to necessarily identify myself as such at this particular point or when I am socialising or something, but in a health surgery I may well want to. There are different occasions when people want to identify different aspects of their identity, and it seems to me that if we are serious about the legislation that we already have in this area, if we are serious about having legislation in this area because we recognise that this is an identity issue which causes real distress and problems in people's lives sometimes, but we are not willing to measure it at all, unlike any of the other characteristics, then, frankly, I think that is patronising.

  Q16  Mr Walker: Finally, Chairman, out of solidarity with gay people, if we get this in the census I shall define myself as gay then, because, quite frankly, I think it is a ridiculous question and I am more than happy to define myself as homosexual. I am not but it is such a silly question and I might as well—

  Mr Darton: Okay, that may be your view but I have to say that we have done some research recently on how people would identify themselves in different survey situations. I have to say that because we do not have decent population estimates this has to be taken as large-scale qualitative work, but we interviewed 3,000 people who had declared themselves to have a minority sexual orientation and then asked them how they would define themselves in a situation in the household where other people knew what they would say, and the vast majority of them would identify themselves in the same way, but there is a minority who, for one reason or another, choose in those circumstances to identify themselves differently, and one can make models to create estimates as a result of that information. I do not want to say that that data is robust in numerical terms, but it is clear. If you had 3,000 people that you had spoken to as a result of, say, having 300 focus groups around the country and they were all broadly saying the same thing and they had a very diverse range of the population, you would take some notice of that as qualitative evidence, and, frankly, that qualitative evidence does not suggest what you are suggesting. It suggests that most people would take it seriously and that most people would answer in a way that is appropriate.

  Q17  Chairman: Just finally on this, if it turned out that large numbers of people did not want to declare their sexual identity on a census form, and you said yourself just now that you would not want to proclaim it, filling in an official census form is quite a thing. If people do not want to do that there would be no point asking the question, would there, and we know on the income question that the answer against the income question is that it shows—or at the least the argument is—that it has a depressing effect on returns?

  Mr Darton: I think this is partly a matter of judgment. We are respectful of the fact that some people still see this as a personal issue and one that they do not want to declare. We are suggesting that it is a voluntary question in the same way as the religion question. There is no suggestion that anybody is going to be forced against their will to declare their sexual identity, so that is one point. The second point is, frankly, there is no direct evidence that suggests this would have a dampening effect on the overall completion rate of the census. When we asked the ONS about this they did send some information but it was mainly saying that they were worried about it because there was some sign in relation to controversial questions like the income question, for example, that there could be some dampening effect. There is not actually any direct evidence that this one would have a dampening effect, I do not think, and it is also partly a matter of how well the census is promoted and what is said about it at the time, and as for why this would be any greater an issue than, say, was the furore around having the ethnicity question introduced for the first time two decades ago, I am not sure. It feels as though the standards being applied in this case are somewhat different.

  Chairman: Right—let us move from sex to religion.

  Q18  Kelvin Hopkins: The British Humanist Association, and I think your own organisation, have raised this with the ONS. I think there is a fair point, I hope you agree, that the way the question is phrased at the moment, "What is your religion?", tends to depress the numbers of those who are not necessarily religious and so you get a false reading. Some of us have lobbied for a change to that question and I am wondering what your view is.

  Mr Darton: I think this is extremely difficult and it is the one I have to be frank about and say we are less certain about what the correct solution is compared to some of the other points that we have been making. I think there are two related issues. One is making the question non-leading, not assuming that you are in a religion, and we and others have suggested a different stem to the question and we believe that it could be made less leading in that sense. The other aspect is the issue about whether, as the legislation does, you attempt to do anything on the religion question which covers beliefs other than religious beliefs. That, I think, is much harder. I agree with the ONS that in the testing they have been able to do at the moment the term "belief" added to the question confuses people and it is not clear what people are responding to when they respond to that. Our view is that we should change the question to make it less leading in the stem, so it should be more neutral than, "What is your religion?"; it should be more along the lines of, "Do you have a religion?", or some such thing. It does need further testing and we are sympathetic to the problems and issues that the ONS have in doing this, but we do think there should be some more work done and that it is possible to have a less leading one. The other area of complexity, and this is similar to the sexual orientation one we were talking about earlier, is that there are different concepts, obviously, of belonging to, being affiliated to, having beliefs and so on associated with religion. The ONS has come to the view that a broad affiliation question is a better basis for their modelling estimates of the other things than having a question which asks directly about belonging. I think that is a matter of judgment. I do not necessarily disagree with it. I do not think there is overwhelming evidence for it, but as long as one is clear what aspect of religion one is measuring in the question and it is non-leading then those are the two criteria and we would be happy with an affiliation question which took out the bias in the stem.

  Q19  Kelvin Hopkins: I must say I have come up with my own formulation as a compromise, in a sense, because I know it opens up all sorts of qualifications if you have a complicated question, but if the question was, "If you have a religion, what is it?", you would get a much better response, if you just put that clause at the front of the question.

  Mr Darton: I think there are a number of options like that. The problem is that they have not all been tested. In terms of the best out of a number of options one has to make a judgment on the basis of what is tested. I think there is just about still time to do some more testing, I am not certain, but I certainly think with the current question there is enough evidence to suggest that it is leading and it does give considerably higher estimates for particularly loose affiliations with Christianity than other questions would give.

  Q20  Kelvin Hopkins: That touches on another problem where people have a heritage which they are often very proud of—a Jewish heritage, a Sikh heritage; but may not be religious. Expanding or building cultural heritage into the ethnicity question, "What is your ethnicity?", I believe would overcome that problem. Have you given any thought to that?

  Mr Darton: We have given some thought to it. Again, I think there is no easy solution. When you talk to the sorts of groups you have just mentioned there seems to be a division. Some of them are happy to see it encapsulated in the ethnicity questions; others prefer it to stand alone in the religious question. There has been a concern, I know, that Sikhs in particular would be less counted, as it were, if it were taken out of the religion question. I do not think there are easy answers in this area but I do think that in the absence of a perfect answer it could be made better by making the stem of the question less leading.

  Q21  Kelvin Hopkins: I just think that if the question was ethnicity or cultural heritage, just broadening that slightly, would overcome the problem, certainly in my constituency where I have a very wide range of people from a whole range of different ethnic, religious and cultural backgrounds.

  Mr Darton: It seems to me that you may be right or you may not be and one would have to test how people out there responded to the term and how they interpreted the term. In the absence of that test I think for now leaving the ethnic question broadly comparable with the last one is probably the right decision. That would be my view on balance but I do accept that it needs further exploration for next time.

  Q22  Paul Rowen: I would like to ask a question about migration and ask Professor Rees to start with. Why do you think the migration questions were included?

  Professor Rees: There is a great concern that we do not have accurate international migration statistics so we do not know very reliably how many people are coming into the country. We do not know even less reliably how many people are leaving the country and we do not know where they are going. The Office for National Statistics uses a very small part of the international passenger survey to make their national estimates. It is of the order of 4,000 or 5,000 respondents coming through airports or seaports, so our knowledge is very lacking. This was all reviewed by the UK Statistics Authority earlier in the year. It was reviewed last year by the Treasury Select Committee. My interpretation, which the Director may confirm or deny, is that adding additional questions into the census was seen as satisfying this demand. The argument I would make—and as an academic I will use all the answers but only for one year—is that we need a more fundamental think about how to generate accurate figures for the total numbers involved across our borders and settling and staying for different lengths of time.

  Q23  Paul Rowen: If I could take you up on that, how accurate is it that you are going to be? People come into this country for a number of reasons. There are asylum seekers. There are failed asylum seekers, people who are trying to avoid the system. There are people on visit visas. There are people on spouse visas. There are people on work permits. How does the question, "How long do you intend to stay in the UK?", provide any sense of or reliable information about why people are coming in? If I am on a spouse visa I might want to say, "I am going to be here for ever".

  Professor Rees: I think what you are arguing for is even more migration questions in the census.

  Q24  Paul Rowen: No, I am supporting you. I actually think that it is quite ridiculous, or would you not agree, that a ten-year census is trying to measure something that is a snapshot because actually it is a dynamic that is moving and changing all the time? You cannot do something that is reliable in one census, because of the range of issues that that one question poses, for it to have any meaningful statistical value.

  Professor Rees: Yes, I would agree with you. The kind of investigation intended behind those questions could well have been done with a targeted survey of migrants.

  Q25  Paul Rowen: Is not the real reason that the Home Office at the fag end of the last Tory Government abolished the system of counting people in and counting people out? The UK Borders Agency has been singularly unable to manage its caseload to have any reliable figures as to where people are in the system, and even the new e-border system does not count everybody because if you are a child you are not going in and out of the country; you are not properly recorded. Is it not a political attempt to say, "We are doing something about this", whereas in reality it is a sticking plaster to cover a gaping hole?

  Professor Rees: That would be my general view, but, just to defend the Office's proposals, behind the question is an attempt to link up and understand what the current instrument, the international passenger survey, which does ask an intention question, is providing and test its reliability. I think that is the thinking behind it. I do not think it is the right thinking.

  Q26  Paul Rowen: I get failed asylum seekers coming into my surgery. They have lost their right to remain in the country and many of them are sleeping on people's couches, or whatever. Do you honestly think that when the survey is done those people are going to put their hands up and say, "Yes, I stop at that flat and I am an illegal immigrant. Please deport me"?

  Professor Rees: No, of course not.

  Q27  Paul Rowen: So is it not a meaningless, useless question?

  Professor Rees: Yes. The intention question, I think, will get lots of incorrect answers.

  Q28  Paul Rowen: Putting the other side of the thing, one of the reasons, I suspect, this has been included is that people like local authorities complained bitterly after the last census that the government grants that were targeted to them were based on census data that were widely inaccurate. If you take a borough like Haringey, for example, where the numbers of people moving in and out of the borough are huge in any one year, is this an attempt to provide some reliable data? Do you think that is the reason?

  Mr Dugmore: There has obviously been real pressure to try to improve migration statistics in any way that can be done, and I think that in looking around for almost anything that would help the census was thought to be one thing that might help a bit. I have real doubts about the intention to stay question, as to whether people would answer it. The very people you are interested in are less likely to answer it. I would support Phil's comments about the use of administrative data, not just the e-borders data but in particular national insurance number data. I had an inquiry only yesterday from one of the big retailers who are interested in the local Polish populations around the country, a predictable question, and one naturally turns to national insurance numbers as an indicator.

  Q29  Paul Rowen: But if you are a non-EU and you have not got the right to remain, or if you are a student, you will not be given a national insurance number.

  Mr Dugmore: That is quite true. It is not perfect, but in the commercial world people are often looking for good pointers and insights rather than the seeking of perfect numbers. One would naturally look towards administrative sources because they are continuously refreshed month by month, quarter by quarter, or whatever, and typically they are available for small areas, and when you look at the numbers you can think, "That is plausible, 8,000 of a particular category in Oxford.", or whatever it is, "Yes, this is giving us a sense of some reality", and so I think that is where the real progress will be made on migration.

  Professor Rees: If I could comment on that, you made a comment that some local authorities were unhappy with the 2001 census results, the population counts for their areas. I do not think that was the fault of the census. It was that their expectations were based on the rolled forward population estimates from the previous census, and those are: add births, take away deaths, add immigrants, take away emigrants. The problem was that the migration counts were poorly estimated. Westminster was allocated far more immigrants than had actually arrived because when you arrive at Heathrow airport you are asked where you are going, so you say central London, and you get allocated to Westminster but you may be going somewhere quite different within London because your knowledge of the geography of the place is probably non-existent, so Westminster's beef last time with national statistics was a result of being misled by the population estimates. That is why getting the migration statistics right year by year through the decade is so vital, so that they are reasonably close, local authority by local authority in 2011, to what people are expecting to happen in the census.

  Q30  Paul Rowen: So is this question statistically reliable or is it just a political gesture?

  Professor Rees: You will have to ask that of the Director.

  Q31  Paul Rowen: You are a statistician and, given the points I have made about the various people who are here for various reasons and whatnot, is that going to produce reliable statistics?

  Professor Rees: I do not think the research has been done on it, the in-depth research on the reliability of the question.

  Q32  Chairman: As I understand it, though, Professor Rees, your essential argument is that instead of bothering about, as it were, trying to refine the migration questions in the census it would be much better simply to put some money and effort into making sure that all the data sources that would be available to the Home Office were put together properly so that we get some reliable migration figures from that collection of source data.

  Professor Rees: That is my basic argument, yes.

  Q33  Chairman: Is that just idiosyncratic to you or is that an argument that is widely shared?

  Mr Dugmore: I would say it is widely shared and, thinking back to the Treasury Sub-Committee a year or so ago on counting the population, that was a theme that came out pretty strongly, that administrative sources were ahead on tracking the population.

  Chairman: We will pursue that a little later.

  Q34  David Heyes: It is interesting, is it not, that, in terms of the arguments that are deployed against the inclusion of those migration questions in the census, very similar arguments are used in the opposite direction in relation to income? You argue strongly that income should be included, Professor, and yourself, Mr Dugmore, and yet all the same criticisms could be made that you have just been making about including migration statistics, that people will be coy about giving truthful answers, the deterrent effect. In fact, the argument is that they are of less use for not including income. Help me to understand that.

  Professor Rees: The income question has been used in national government surveys for a long time. It is a well tested question in the labour force survey and its successors and in the forthcoming integrated household survey a household income question is used, and also in the expenditure on food survey, which was going into the integrated household survey. In that sense it has a much better pedigree than the set of migration questions. It has also been used successfully, as I said, pointing to my PhD thesis, in other countries on a regular basis. What is the alternative? The alternative is for Parliament to ask Her Majesty's Revenue & Customs to generate from their administrative database the equivalent neighbourhood statistics on the distribution of income, again using very broad and non-disclosive classes. That would be an excellent source of information year by year.

  Mr Dugmore: If I may just add to that, in the case of migration often one is looking at ebbs and flows and changes of fairly short order if you think of migration from eastern Europe and so on, and a lot change in a few months, or certainly a year or two. With an income question, yes, I think over a long period of time there may be some gradual shifts, but a lot of the measure would be fairly fundamental and would last for several years. When one looks at social classifications across the country it is quite striking that, whilst there might be a few areas that change dramatically, many areas do not, and I fancy that if income were to be asked this time and ten years hence the broad distributions across the country might well be broadly similar.

  Q35  David Heyes: I think Professor Rees was saying that you would get better quality data if HMRC could be obliged to provide it than you could ever hope to get out of the census.

  Professor Rees: Yes, we could.

  Q36  David Heyes: You would still argue for it being included in the census. How do other countries get round this? Scotland presumably are satisfied that they can deal with the kinds of problems that have been listed. There are comparable countries that have dealt with this. What is the difference? Why do they do it and we cannot?

  Professor Rees: In the US case they have asked for income for decades, so everyone is very used to supplying that information at the time of the census. In Britain it is routinely asked in household surveys, in commercial surveys. When you fill in your warranty card for your latest electronic household apparatus you are often asked for your income in more detail than in the proposed census question, so people by now are fairly comfortable about reporting their income, a very blurred and non-disclosive form of their income. Just to pick up your point earlier about the impact of an income question on response, that certainly was not the case in the census test in Scotland. They decided there was not an impact. In the case of England and Wales there was a statistically significant difference. It was between—and here I am doing it from memory so it would need correction—51 and 52.5 %, so 51 % in those test areas which had an income question responded to and 52.5 % in those areas which did not. It turned out, because of the size of the England and Wales census test, to be statistically significant but in terms of importance I do not think it was sufficient to eject the question.

  Mr Darton: Although there are alternative sources of income data, they could be better, such as the suggestion made for the Inland Revenue. The advantage of having it on the census is, of course, that it helps you understand and take a lot more value from some of the other questions on the census. It is clear that the experience of various categories of people that you might measure in the census, say, somebody with a disability in their service needs, is likely to be quite different, depending on their level of income, just as one example. I think the case for putting the income question in, and we raise it as putting it in the census, is that there is a clear and well-established relationship between socioeconomic conditions and other things that we are trying to assess with the census and it is therefore important to have it in the same survey.

  Q37  David Heyes: Just one last question for Professor Rees, if I may. You would argue that the deterrent effect is fairly negligible, as evidenced by the testing that has taken place. Is there any evidence, perhaps from overseas, of the numerical difference between the response on a test, which people know is a test and therefore does not have legal power behind it, and the actual census, which is accompanied by the kind of publicity campaign that says you must do this, it is a legal obligation? Does the power of compulsion make a difference to the response rate?

  Professor Rees: It does in general in the census survey. The response rate is around 50 to 60 % in the tests and it is going to be 92 to 98 % in the census. I do not know of any evidence from other countries to answer directly your question but the Director of National Statistics or the Director of the Census may be able to answer that question.

  Q38  Chairman: In your opening remarks, Professor Rees, though, on income you said that a reason for having it in the census was that the other sources where it was used and tested had defects attached to them. Now, though, you have been arguing that in fact their pedigree is shown by the fact that they have been so successful in the labour survey and in the household survey and therefore it would be easy to do it in the census. One of these things has to be true, but I do not think they can both be true, can they?

  Professor Rees: Asking an income question in the surveys does not deliver you the detailed geographical information and tables that you need, so that is the key reason for having an income question in the census. When I referred to the other instruments what I meant was the other variables which are used essentially as proxies for income, so unemployment, for example, or very low skilled employment and activity rates in the labour force, which can be measured with the existing questions in the census. ONS and others have then attempted to construct a synthetic income estimate for small areas, so there is a ward income value. The problem with that is that it does not contain any more information about the distribution of income than the variables that have been used on the right-hand side of the equation to predict this income, and income distributions are much more varied than that. The illustration I use with my students is this. Assume you have knowledge of the distribution of lawyers across your city, and there are lawyers living in most of the wards, and then you apply the average income that you know from your survey of lawyers. You would be overestimating the lawyers who lived in poor districts, because there are poor lawyers (perhaps), and underestimating them in rich areas. In particular an income question will enable you to pick out the clusters of poor people in rich areas, but if you use area averages in your indices of deprivation which feed through to the funding distribution formulae in probably a dozen government functions you are going to miss out those pockets of poverty in rich areas.

  Q39  Chairman: An income question would simply be the one that asked what your income was each week?

  Professor Rees: Yes, that is right.

  Q40  Chairman: It would not ask what the source of the income was.

  Professor Rees: No. You would be asked to include benefits as well as earnings.

  Q41  Mr Prentice: But surely the simple thing to do would be just to ask HMRC to publish figures for small areas? Obviously, they would not have information on people who do not pay tax, for example, so that is a problem, but why do we not just ask the Inland Revenue to publish the information they have for smaller areas than they do at the moment?

  Professor Rees: That would be an excellent suggestion.

  Q42  Mr Prentice: It would be dead easy, would it not?

  Professor Rees: I do not know.

  Q43  Mr Prentice: Yes, it would.

  Mr Dugmore: Could I add to that in terms of yes, I think it would be a great thing to head for; we need to get some good information on income in small areas, but it would stand in isolation and, of course, the great advantage of the census is the possibility of relating income to other variables as well, so that would be the attraction of asking it still within the census, unless the HMRC records were to be merged into a census database, which is quite a long way ahead, I think.

  Q44  Mr Prentice: And that would hold true even if the information were published at a neighbourhood level, a very small area level? You could make the correlation, surely?

  Mr Dugmore: Yes, I think it would give material to play with, but it is rather like some of the other statistics we have from administrative sources at the moment on unemployment and claimants and so on. They tend to be single slices and you can look at small areas and think, "Yes, they are there and not somewhere else", but you cannot then analyse them in terms of housing tenure or ethnic origin or some of the other questions you are asking.

  Q45  Mr Prentice: The last census was three pages long and the next census is going to be four pages. I read somewhere in my briefing material that when the methodology was tested there was no decline in response just with the addition of the extra page. How far could we go in asking people to fill in the census? What about a five-page census next time?

  Professor Rees: There is a certain cost to the public purse of each page. The Director of the Census can give you an estimate of that. It is a serious matter of public expenditure. The key innovation in the administration of the 2011 census is the ability of the householder to fill it in online via the internet, and that makes it very cheap to add further pages should this Committee or Parliament wish.

  Q46  Mr Prentice: It is the deterrent effect of having the extra page, that is what I am after. It would have no effect on the response rate, that is what you are saying?

  Professor Rees: No, it would do.

  Q47  Mr Prentice: What about a fifth page then?

  Professor Rees: As you expand the number of questions there must be a deterrent effect on response. Not necessarily on surveying the form, which is a legal requirement, but in completing all of the questions. That is where the internet version is very useful because it is very easy to go through, you are not disturbed by having to read questions you do not have to read, there is a routing through it. I was privileged that National Statistics asked a set of people, including academics, after the census rehearsal this October to test out the internet questionnaire. I think it is the best internet questionnaire and easiest to complete I have ever done.

  Q48  Mr Prentice: I understand all that, but when the testing was done in Newham and wherever else, the fourth page was not seen as a deterrent but you are telling us there may be a problem with accuracy. I think you did tell us that people would not necessarily fill in the forms as accurately as they might have done if the census was shorter. Is that not what you told us?

  Professor Rees: I think I am straying outside my area of expertise here.

  Q49  Mr Prentice: Okay.

  Professor Rees: It is simply that if you were to look at the total body of information provided by 60 million people over the 40 questions, or slightly less than that, in the 2001 census and asked how many of the data items—60 million times 35, or whatever it was—actually depended on completed answers from forms, you knock out five per cent because of non-response in the census and those that had to be invented, but then there is the whole edit and imputation process which the census very skilfully implements to fill in all the forms that have not been completed. My estimate, and I think it has been confirmed by others, is of that huge data matrix only 60 % of it actually depends on people who have put some mark on the piece of paper.[1]

  Q50 Mr Prentice: Amazing.

  Professor Rees: Yes, it is amazing. That has always been the case. It is the dirty secret of the census! That is why we need a very sophisticated statistical methodology which is built into the edit and imputation operation, it is built into what is called the one number census, add that five per cent missing. There is incredible expertise that is gathered in National Statistics to produce a really good reliable product because Joe Public does not manage to get to the end of the form quite often.

  Mr Prentice: We know the dirty secret now!

  Q51  Kelvin Hopkins: Is not the important thing to have meaningful time series and if they make the same mistakes in every census you get the meaningful time series? If you suddenly clean it up and make people answer more questions, in a sense the time series breaks down.

  Professor Rees: What you need is more reliable ways of filling in those missing answers and the best statistical methodology is used for that. There is a phrase in the statistics profession that is "borrowing from strength", so if you have a household and only half the record is filled in you look for another household nearby that matches your completed variables and then you borrow the response from somebody who has returned it. That is the basic operation that goes on.

  Q52  Chairman: Another dirty secret!

  Mr Dugmore: No further dirty secrets here! I would just like to pick up the point about the number of pages, and I am sure you are right that logically it must begin to fade and I think those of us who have lobbied for questions have lurking behind the feeling that you cannot just keep on adding more and more questions. I suppose the pleasant surprise was when the test was done for 2007 and the difference between three and four pages in statistical terms did not show any difference. I am sure had it been five or six decay would set in. We do have evidence there is not such a difference.

  Q53  Mr Prentice: Can I just ask Professor Rees, because you are the population expert, in future immigration into the United Kingdom British people marrying overseas spouses will be a large component, and we have read the speculation in the papers about the UK population going beyond 70 million, is it possible to forecast meaningfully future trends in intercontinental marriage, or is that just impossible?

  Professor Rees: I do not know anyone who has done it. I speak as a husband of an immigrant, and long may that flow continue. I am currently engaged in a project which is attempting to forecast for the UK and its local areas the ethnic composition of the population and we will have to make estimates of the flows of immigrants and emigrants to and from the different origin countries that people come from.

  Q54  Mr Prentice: It is whether it is a first generation thing, or second or third generation thing, people going back—we are talking about the Asian subcontinent here—for their spouses. That would have a huge impact on future immigration into the United Kingdom, would it not?

  Professor Rees: I do not know. If you look at the estimates of regional origin of immigrants over time, for those from South Asia, although the immigration for participation in the labour force has gone down substantially and is controlled by various regulations and laws, the total immigration has been very flat, the family reunification immigration has continued, and will continue. One of my PhD students born in the UK of Pakistani origin found his wife in Pakistan.

  Q55  Mr Prentice: That is the point; it is the non-controlled part of immigration. The controlled part is the points system and all that kind of stuff. My question is, given it is such a big percentage of total immigration into the United Kingdom, whether it is possible to forecast and you are saying it is not really.

  Professor Rees: I would have a go. Obviously it depends entirely on the legislative framework for that migration. I would anticipate a continuing flow of migrants from South Asia marrying spouses in this country, yes.

  Chairman: Before we go into the wider issues, all of which are fascinating, we will have to thank you for the session this morning. Thank you for expressing your general views and your reservations about certain questions and on the whole your endorsement of the questions in general, but raising the particular issues you wanted to talk to us about. Thank you very much for that. We shall pursue those now with those responsible for it. Thank you very much indeed.

  Q56 Chairman: Let us move seamlessly into our second half. We are delighted to welcome Jil Matheson, who is the new National Statistician. We met your predecessor on a number of occasions and we are delighted to welcome to you. Glen Watson, who we have met before, is the Census Director at ONS. Thank you very much indeed for coming. I am sure we are going to ask you about the issues that we were talking about in the previous session, but can I ask you a general question to start with. How open are you at this stage to making amendments in the kinds of areas that we have just been talking about?

1   Note from witness: The correct figure is 90 %. Of the 10 % of data that was imputed in the 2001 Census, 6 % was missed because people did not return a questionnaire at all, and 4 % because some questions were left unanswered by those that did return a questionnaire. Back

previous page contents next page

House of Commons home page Parliament home page House of Lords home page search page enquiries index

© Parliamentary copyright 2010
Prepared 18 January 2010