Oral evidence: The Future of the Census, HC 1090Tuesday 25 February 2014
Oral evidence, 25 February 2014.
Witnesses
Keith Dugmore, Director of Demographics User Group, Demographic Decisions Ltd, Sarah Henry, Head of Intelligence and Performance, Manchester City Council, Juliet Whitworth, Research and Information Manager, LGA, Piers Elias, Demography and Modelling Officer, Tees Valley Unlimited, and Professor Les Mayhew, Professor of Statistics, City University, gave evidence.
Chair
Could I ask each of you in our second panel to identify yourselves for the record, please?
Sarah Henry
I am Sarah Henry. I am Head of Intelligence and Performance in Manchester City Council.
Juliet Whitworth
I am Juliet Whitworth. I am the Research and Information Manager at the Local Government Association.
Piers Elias
My name is Piers Elias. I work for Tees Valley Unlimited. I am a local government representative on the Census Advisory Group, and I also co-chair the Population Liaison Group that meets the ONS and central Government to discuss methodology and ensure local authority needs are catered for.
Keith Dugmore
My name is Keith Dugmore. I organise the Demographics User Group, which represents some large, commercial end‑users of census data.
Chair
Thank you. I should add that we are also waiting for Professor Les Mayhew, the Professor of Statistics at the Cass Business School. Unfortunately, he is launching his book today. I will give that a plug for him. This is unfortunate for us, but he will be turning up and I will fire some questions at him when he arrives. Before we go on, I have noticed you all listening to the previous session. Are you burning to say anything in reaction to what you have already heard?
Keith Dugmore
I have one quick point on what Mr Flynn raised about the merits of alternative methods. I am completely with him on the possibility of using administrative data to a greater extent, but when it comes down to samples and sample survey, those of us who are concerned about statistics in very small areas see this as a great danger. It could render a census useless if we could not get statistics for very small areas.
Piers Elias
I have two points. First, the 2011 census was better than the 2001 census. We are getting better data for two reasons. One is that the address register was properly checked, updated and cleaned by local government and by the ONS. Secondly, the response rates from the hardest-to-count authorities were much higher than in 2001. For those two reasons, it is wrong to say that 2011 was worse than previous censuses.
Chair
I can see assent from other members of the panel. You do not need to say anything if you do not need to.
Sarah Henry
Yes, I agree.
Kelvin Hopkins
You have heard the previous session and I asked questions about the value of the census. It seems to me from earlier comments that for every pound spent on the census, the private sector alone benefits by six times that amount every year. Does that not suggest there is a value?
Sarah Henry
I agree that there is huge value in the census. It is very difficult to put an exact pound on that value, but looking back at decisions that are made at a local authority level, I think that the census sits implicitly or explicitly behind every decision that we make, with a focus on operational decisions as well as big strategic decisions. All our ward councillors are very aware of the place that they represent, the people whom they represent and the intelligence that they get to inform them. That comes from the census—not only the census, but it comes from the census.
Kelvin Hopkins
We have heard that the census has improved over the last decade. That is one good thing.
Sarah Henry
Certainly from our perspective.
Kelvin Hopkins
Is there not an argument that the alternative, administrative methods of collecting data should be an addition to the census rather than instead of the census?
Sarah Henry
Certainly now, they should be in addition to. I agree with the previous panel that we are nowhere near replacing the census with administrative data yet.
Piers Elias
I would just add that the ONS use a lot of administrative data in their population estimates. They do actually link individual data between the GP patient register and the higher education data on students to try to improve their methodology on migration from year-to-year, so it is already happening. We would encourage the ONS to continue to do that. What we do not get is the characteristics that are sitting on those admin files and we would be very interested to hear of the ONS doing more work on digging those out.
Kelvin Hopkins
My first two questions were leading questions, but now, an open-ended question to each of you: what do you personally think is the value of the census?
Keith Dugmore
You mentioned commercial views and I will just put our perspective on that. I have worked in local authorities before, but in the commercial world a lot of companies use census data to try to understand local markets. They use this for locating stores, in particular. Just look at the extent to which companies like Sainsbury’s have been opening stores over recent years, or Whitbread opening Costa Coffee. All these things involve a lot of investment and having reliable small area statistics is vital for those sorts of decisions. There is also the question of what they put in the stores and how they vary the offering according to local markets. A third area, which has been mentioned already, is organising market research sample surveys. Again, the major companies there rely heavily on census data to target particular groups. When one thinks of the opening of supermarkets, for example, a large one might be an investment of £25 million to £30 million; smaller ones might be £5 million. Even a few companies being involved in that soon add up to big numbers. When you think that we know that many thousands of companies use census data, the numbers become potentially very big indeed.
Juliet Whitworth
I would add that local authorities are major users of particularly the small-area data. This is the case below local authority level at ward level, and even below that. High-level information is useful, but you can miss the granularity. You can think that you have got an area that is full of owner-occupiers, for example, and miss pockets of private‑rented housing or social housing. This information and this granularity are really essential for local authorities to be able to target their resources, and, as Sarah said, to be able to make not only strategic decisions but also day-to-day decisions. I would argue that it allows delivery of central government policy as well, such as with things like the localism agenda, doing neighbourhood development plans. They often do not fit within normal administrative boundaries, and because of the level of detail that is available from the census, you can build up from really small units to understand those neighbourhoods where work is going on. It is really crucial. Local government has perhaps struggled to place a numerical value on that small-area data. That is not because it does not have a value, but more because it is quite difficult to measure that targeting of resources and that sort of thing.
Sarah Henry
If we did not have the census and we had to carry out endless surveys, that would not only be very time-consuming, it would be very resource intensive for us. We need to take that into consideration. It is helpful to distinguish between the count that we get from the census and the rest of the population characteristics that we get from it. Both are extremely important, especially for democratic accountability at very local levels. Ward members need to understand their population and the people they represent; they need to understand households and place. That is really important for planning for growth and reform.
Piers Elias
The other thing the census gives us that would be very hard to replace is flow data. For example, travel‑to‑work data is very important for local authorities. It helps with transport planning; it helps businesses decide whether the catchment area of the place they are going to invest in has a big enough pool of resources or skills. We would miss that without the census.
Kelvin Hopkins
On the point of granularity, I represent Luton, which has a large Asian population. In the past, people have talked about Asians in general terms. In fact, in Luton, we have overwhelmingly Kashmiri and Bangladeshi, whereas in Leicester they are Gujuratis from East Africa, professional-background Indians. These are very different populations generalised as Asian and the granularity is really important.
Sarah Henry
Manchester has the same issue.
Paul Flynn
How many times do you think the manager of my local Sainsbury’s would consult census figures for his decisions?
Keith Dugmore
The manager of a local Sainsbury’s?
Paul Flynn
Yes, my local Sainsbury’s.
Keith Dugmore
I do not think that he would refer to them directly, but he would be talking to the site location team at Sainsbury’s headquarters, who would inform him what the local market was and how it differed from others.
Paul Flynn
The local Sainsbury’s came to my constituency 35 years ago and they relocated to a bigger site. The decisions they take are based on what the customers are buying. What they will order for next week depends on what they have sold in the previous week. Are you serious about saying that the census is something that they look to guide them in their daily decisions?
Keith Dugmore
Yes. There are two elements to that. One is changing the estate of stores. Companies such as Sainsbury’s, which is a good example, have grown their number of stores from roughly 500 to about 1,000 in the last decade. They have needed to decide where to put the new ones. They do not have data.
Paul Flynn
How could the census help? I am genuinely puzzled by what you are saying. They will do it on the basis of population they know, on the profitability of the shops, and on the customer and the customer’s taste. That is not in the census.
Keith Dugmore
They are concerned in getting information about the numbers of people and types of people in local areas and also questions of access; they are interested in transport and so on. They will then look at the market potential in different parts of urban areas and decide that Balham is better than Tooting or whatever. When they are locating new stores, those are the sorts of decisions that are made. I mentioned Costa Coffee, and they are very interested in where people are working. It is not so much where they live but where they are going to be at lunch time. Boots is another example of people popping out to buy things at lunch time. The information about where people work that we get from the census is absolutely vital for locating stores.
Paul Flynn
The big supermarkets do their own market research. They have their own experience of what is selling in their shops, what people are coming in and it is based on that. It is not a decision taken by a census; that was 10 years ago.
Keith Dugmore
I can say definitely that that is not the case. They certainly have their own market research—
Paul Flynn
If Sainsbury’s in Manchester are making a decision, should they base it on what the census says or what the local authority says of the population of Manchester? There is a lot of difference between the two.
Sarah Henry
I do not think it is either/or. When you make those kinds of business decisions, you take everything into account, including the census and including your own market intelligence. You also use various products that are on the market that are based on the census, such as the predictive modelling that Acorn and MOSAIC provide. It is very difficult to disentangle the census from all the other bits of intelligence that we use for various decisions, and that local businesses use for their decisions. The census is the bedrock. Even if they do not think they are referring to it immediately, they might still be referring to it.
Paul Flynn
We know there are major imperfections in the census: 5% of the population do not take part in it. Many of these are migrants who might have language problems or who do not trust authority; many of these might be elderly people who feel they cannot cope with it. There are all kinds of other reasons. We already know there are those people trying to manipulate it. We already have a lot of garbage going into census forms. Is it not inevitable that garbage will come out of the other end?
Sarah Henry
No, because there are statistical methodologies that we use to improve them. That kind of bias applies to all the other methodologies that are being proposed. Surveys are worse than the census in that respect, because there is self-selection in whether you decide to participate in the survey or not.
Paul Flynn
This is another, rather elaborate way of describing guess-work that comes up as an element of that figure. Have any of you examined what is happened in Finland? They have gone on to an administrative system there that we hear is cheaper, timelier, more efficient and more scientific, by the sound of it. Why are we lagging behind in the past?
Sarah Henry
As we heard from the previous panel, it is easier to carry out a census there.
Paul Flynn
It is always easier to be inert and carry on with the system that you have been using for a long time and that people are reluctant to change. Is that what it is? Does Finland have a system that is more scientific, cheaper and more reliable? Yes or no?
Juliet Whitworth
I do not know a lot about Finland’s system in detail, but the other issue is that they did not arrive there overnight. It took some time. We are not saying that we should not move to administrative records. That is very likely to be what we aspire to do. What we are saying is that we want to make sure that when we make that move, the administrative method is as good as it can be and gives us what we need, and that we do not just junk something before we have got an administrative system that really works.
Paul Flynn
Could I ask Professor Mayhew, who has caught his breath now—?
Chair
Can I just ask Mr Elias to make a comment?
Piers Elias
You were saying that the ONS guessed the last 5% of the people who do not fill in their forms. First, you have got the 94% who have filled in their forms. The other 5% are built up by comparing with administrative data sources. Some of them are actually genuine people whom they can find and they put them back into the census. It is not all guesswork. A lot of it is done by maths and matching with administrative data.
Paul Flynn
It is elevated guesswork; I will give you that.
Chair
That is what administrative data is.
Paul Flynn
Yes, indeed. It is based on a large, reliable sample. Can I ask Professor Mayhew and say that I raised these points in 2005 as a member of the Committee about the upcoming census? I thought it was a biblical solution rather than one that matched modern science.
Chair
Welcome, Professor Mayhew. Could just identify yourself for the record before you answer.
Professor Mayhew
I am Professor Les Mayhew from Cass Business School, City University, London, Faculty of Actuarial Science and Insurance.
Paul Flynn
Is there a better way of collecting the information we require from the census, rather than having this 10‑year attempt to interview everybody?
Professor Mayhew
I am an advocate of an administrative‑based system. I have a small company that has been estimating population at the local authority level, up to the size of cities as large as Birmingham, for about 30 years.
Paul Flynn
How does the accuracy of your figures compare with the census?
Professor Mayhew
Very well. It depends where you are. If you are in London, we certainly found 4% to 5% more people when we compared our figures with the then ONS mid‑year estimates. We were commissioned by the Olympic boroughs to do this because they were concerned that the census would not adequately count their populations.
Paul Flynn
What was the comparable cost?
Professor Mayhew
The comparable cost for six boroughs was about £200,000.
Paul Flynn
We know the last census was £482 million and there is a lot of information on that. Would you agree that the last census will probably be the last census? Would it not be an improvement to have a continuing census going on with a tiny sample, compared to the 100% sample that we seek now? It would give us information that would be timely, of good value and more reliable.
Professor Mayhew
You have to look at this in layers. There are the data available in administrative systems. There is absolutely no doubt that central Government data systems could provide a good population count and even a household count if you did it properly. Whether you would get all the attributes you needed that you get in the census, I doubt. However, you would also get other attributes that you do not currently get, such as whether a person is on income support or not. I also think that there are data that you will continue to need to collect via surveys. These are softer figures, like how many hours a week you spend caring for an elderly person and that sort of thing. Surveys will remain the appropriate way for doing that. The other area that I do not think has been mentioned is the enormous amount of data assets that are held at a local authority level that they use for a multitude of things: locating health centres, assessing child care, or looking at troubled families. There is a whole multitude of applications there for which the census does not really give you any adequate information at the level of granularity you need. I think there is a model that says, “Yes, the centre does the big population count, but it then tags people to local areas”. Then local areas can add or append their own attributes to that data for the purposes they need at that point in time. At the moment they cannot do that, because the granularity is not sufficient; the data are out of date; geographical units are wrong; and so on and so forth. You have got the beginnings of a model there which involves local and national collaboration, overlaid with a survey to get the softer information that is more difficult to collect from administrative systems.
Paul Flynn
Before you arrived, Mr Dugmore was saying that one of the advantages of the 10-year census is that it did have granularity, and there were small areas in which it showed the results that perhaps would not be apparent. Would it not be better to have targeted investigations into those particular areas to provide a more accurate picture, and not one that came up by accident in the statistics?
Professor Mayhew
I certainly think that the proposed administrative approach needs to be road-tested in some big areas. Different models need to be tested that involve combinations of local and central input. However, I do think that if the centre does not provide local areas with the statistics they need for their local usage, they will create their own statistical systems that suit their needs. In fact, I know many local authorities are already doing that.
Sarah Henry
Yes. I agree with that, but if we lose the census, we will lose some of the benchmarking. There is a long way to go in order to agree how we carry out any administrative option. Personally, I do not think there is enough time between now and 2021, for sure. The legislation needs to be thought out carefully; it needs to be enforceable; different Government Departments would have to have duties around population statistics set out in law, which they do not have at the moment. The other point is that in order to do the kind of work that Professor Mayhew is talking about, data need to be matched. In order for it to be matched, it needs to be identifiable. The experiments that have been done to date by the ONS did not include that exercise.
Paul Flynn
These arguments were the ones presented in 2005, very much from people who are comfortable with the present system and the status quo, and do not want to change or are afraid to change. We have a system that is far less than perfect. Nobody believes that the fourth biggest religion in Britain in based on Star Wars, and there are other wildly inaccurate figures for Westminster and Manchester. We cannot talk about this as though it is the gold standard of censuses. We seem to hear that better systems are coming out at a fraction of the value from Finland and other places. Is that not true?
Sarah Henry
Nobody is arguing that the census is perfect, and in 2005 Manchester City Council would for sure have had a lot more to argue against it. In 2011, though, to the ONS’s credit, they learnt a lot of lessons from 2001 and they worked with us in partnership in a way that we think can be built on to improve the next censuses even further.
Paul Flynn
Could you give me an example of something that came out of the census we would not know in any other way?
Sarah Henry
Yes. For example, they used us in order to generate the address base, which they did not do in 2001. In 2011 all local authorities contributed to that, and the address base used for the census was far improved. Come 2021, we believe that we could help in ways that we did not in the past. For example, we could go round to people’s houses all the time. We can help collect forms or, hopefully in 2021, we will be going round with mobile devices and looking at how people fill in their forms online. We can make the census better and cheaper that way, if we build on the partnership.
Paul Flynn
Why do we need a census to tell us how many people are on income support?
Sarah Henry
I do not think we ask that question—
Paul Flynn
No, it was something Professor Mayhew mentioned.
Sarah Henry
—But about unemployment and the kind of skills people have. That kind of information is not captured at the moment by DWP.
Piers Elias
I will just to come back to Professor Mayhew’s point about local authorities using administrative data sources that they hold. We do use a lot of council tax data, Land and Property Gazetteer data, and, for example, the electoral register, which has recorded nationality since 2006. We do most of that as a proxy for international migration to the area. There are the two issues with it though. One is that it does not cover the whole population; it only covers the 17-plus population. Secondly, according to Audit Commission figures, 10% of us are not on electoral registers. These admin data sources are potentially very useful. If we can tighten up the legislation that means people have to re‑sign up when they move or something like that, that would be really useful. On the point about the address register, there was no legislation after the 2011 census to allow it to be continued to be used because of intellectual property rights arguments between Ordnance Survey, Intelligent Addressing, and the Local Government Association. If you could get rid of that and put some legislation that allows us to keep that cleaned address base to start with, we have an address register.
Professor Mayhew
I just wanted to say that whilst I agree with Piers, the data that are available locally include the school pupil census; it includes other data about older people, social care, and so on and so forth. Our methodology combines all these datasets together alongside the GP register. We have a system for sifting through and confirming every individual in the population by the address that they live in so that we can confirm that person on the basis of several pieces of evidence. Secondly, we can assign them with a household, which we can also find—is it a lone parent household? Is it an older person living alone? Is it a three‑generational household? All those things become possible if you combine these datasets in the right way and it all can be made fairly automatic. The challenge is to actually clear the bureaucratic undergrowth to enable these things to happen automatically, instead of having all these disputes between different data owners, sometimes under the guise of data protection and things like that. All these things could be done in a data safe haven. People can be accredited to use this kind of data, and the outputs that come from these processes could be fully de‑identified and anonymised as well. That is something that we do very regularly, so we know it is possible.
Keith Dugmore
Can I pick up two points? For the local authorities themselves, there is probably a lot of interest and a lot of strength in using their local data for the census. Commercial companies, though, are particularly interested in getting a consistent national view, so I think it is very important that some of that data is produced from the centre, rather than having local variations. Dare I say it, national view actually extends to the United Kingdom as a whole. There is an interest in not only getting England and Wales, but adding Scotland and Northern Ireland so that one can look across markets across the entire UK. On the second point about the National Address Gazetteer and the fact that it was created for 2011 and has now become a commercial product through Ordnance Survey, this is one example of where it is important that Government puts its foot on the accelerator to try to improve access to the address file, and some of the Government Departments’ administrative files. Otherwise, there will be a handwringing in the next five years with people saying, “Well, we would like to do it by 2021, but it is all very difficult”, and another decade passes. The ONS has made a lot of progress in their research for Beyond 2011, looking at the Department for Work and Pensions, the National Health Service and so on, but it needs to be propelled more quickly in the next couple of years. Then we might see where we are as we near the date of the next census.
Sarah Henry
I agree that the matching exercises are extremely useful. In Manchester, in preparation for 2011, we actually did a similar exercise using our own administrative data. Admittedly we did not have access to everything that the ONS would have access to, but we did that. We ended up with a range that the actual census figure fell in, but at the higher end, so there is something that needs to be considered around people who are not listed on any of the registers. For example, we found out that there were 20,000 homes in the city centre of Manchester where nobody was registered with a GP. The service was not nearly as universal as everybody thought it was. When you come to urban areas where there is a lot of change in the population, it is quite difficult to enumerate. You need a foundation that is consistent nationally, but we think that for big urban areas—Manchester and Westminster are two good examples—we need to think very carefully about how we use administrative datasets.
Greg Mulholland
If I could just ask this to you specifically, Sarah Henry. You have clarified to some extent the concerns that Manchester had. It is very much one of the councils where there were problems—45,000 people were missed in the 2001 census and Manchester City Council raised concerns. Westminster City Council still has concerns. We heard about those concerns last year. Are you now saying that Manchester City Council is entirely happy with the reliability and the accuracy of the census, even though Westminster still has exactly the same concerns?
Sarah Henry
We were much happier in 2011 than we were in 2001, and we worked very hard with the ONS to achieve that. I am aware that Westminster was less happy. Obviously, their population seems to be harder to enumerate, but I still think that if we build on the partnership, we can make the census better. That is my point. I am definitely not saying that we have got to the point where we have made it perfect for everybody, but it is not about throwing the baby out with the bathwater.
Professor Mayhew
It is very easy to select Manchester and Westminster as outliers. They have been known for a long time. Let me give you an example in the other direction. I have worked on and off in Newham for about seven or eight years. When we last went there, the mid-year estimates said there were 243,000 people living in Newham. When we did our administrative work in 2011, in time to coincide with the census, we identified over 300,000 people. When the census finally published its results, they also found 308,000 people in Newham, so for several years there has been this discrepancy of over 60,000 people in Newham that has affected schools planning, service planning and a whole host of other things. They were particularly keen to get my view, to produce a different perspective, a different evidential approach, and come to that figure. In fact, we found that in several locations in London, none as big as Newham. Newham was a very special case, with lots of transient in-migration and other things, but we found quite substantial discrepancies in a host of other London boroughs.
Chair
So it would be inaccurate to say that it was the census alone that identified the miscounting of the population in Newham.
Professor Mayhew
They started from a baseline in 2001 that may have been a false baseline. Other people can comment on that. Then they used their methodology, adding births, subtracting deaths and adding net migration, year-on-year, up to 2011. The figure they got to was 243,000, but the actual count of population in Newham was over 300,000.
Chair
That suggests that depending upon administrative data is what led them to the wrong estimate of their population.
Professor Mayhew
Potentially.
Paul Flynn
What was the difference with yours? How did you achieve your figure of 300,000?
Professor Mayhew
We combined all the datasets that I have mentioned. We had our rules for deciding whether that person was genuine and lived at that address. We came to that figure and, interestingly, the 2011 census came to an almost identical figure, but they had been working on a completely different assumption in Newham for several years previously because of the false baseline in 2001.
Chair
Even if you got your estimate correct, it suggests that the census was the only qualitative assessment that could be produced that would verify your figure or not.
Professor Mayhew
That is true.
Chair
How can we then consider getting rid of the census?
Professor Mayhew
There is one thing we do that I have not mentioned. We have used random checks on households to see who is actually living there and compared that with our own figures on a sample basis. I agree though. I am not saying that the 2011 census would be an independent check or not. In fact, we were the independent check on the census. By the way, we produced our figures within six months, compared with something like 18 months.
Chair
My understanding is that your position is that the census should be scrapped?
Professor Mayhew
My position is that there are much better ways of doing it, and we would save a hell of a lot of money if we did.
Chair
On the other hand, you have verified that, if we had not had the census, we would not know the true population of Newham because we would not have been able to cross‑check your estimates. There is no other benchmark your estimates could be cross‑checked with.
Professor Mayhew
Well, suppose you did my exercise every year or every quarter, which would be feasible within the envelope of money that you currently spend on the census.
Chair
Sorry, can you just answer the question, please?
Professor Mayhew
You would have ongoing checks and verifications on who was inhabiting particular households in your patch, and you could analyse that data and use it as partial confirmation of what was going on in your area. I am convinced that this is a cheaper and more accurate alternative to the census.
Paul Flynn
What percentage of the households did you contact, and how many do you think you would need to contact to have a continuous picture?
Professor Mayhew
In the one place that we did this, which I think was the London Borough of Brent, we contacted something like 1,000 people who were on what was then called a “citizen’s jury list”. We are a small organisation.
Paul Flynn
That would be a tiny percentage of the total population.
Professor Mayhew
Gallup and other organisations use small samples. You choose your confidence limits and select your sample accordingly.
Chair
We have the delight, on this panel, of having people with different views and we would like to hear them.
Piers Elias
On the GP patient register nationally, at the 2011 census, it was found to be about 4% overinflated. It ranged from plus 15% to minus 15% overall for local authorities. If you go into the age groups and look at the 20 to 30-year-olds, you find huge disparities of up to 50% differences in counts for people on GP registers compared with the census. That is because students are not very good at re-registering when they leave university. It is because of the migration. People might come to the country, not register where they first live, and then move. When an international migrant comes into the country, they get something called a “Flag 4”. Newham actually raised that idea of using Flag 4 status as a way of counting international migration. It is used in the current method of allocating international migrants at a local authority level. We have to have a benchmark to check the admin sources. Without that, we can end up going in the wrong direction, and that has issues of fairness for funding for local authorities; some are getting too much, and some are getting too little. The ones who get too little tend to shout the loudest.
Sarah Henry
In the interest of balance, the census is not perfect; we have determined that, but the administrative datasets are not perfect either. That is why we need to carry on working on both methodologies. It is still worth having the validation every 10 years, in order to make sure that error is not creeping in if there is a significant shift. It also gives us added value of being able to cross‑reference and cross‑tabulate lots of data that are collected at one single point. You cannot do that in a survey, otherwise the surveys will be too long and nobody will participate. The statistical confidence that you cite assumes that you have randomly sampled and that everybody has responded. That does not happen. There is self‑selection in surveys. Neither solution is perfect: keep that balance.
Paul Flynn
How do you explain that a survey that attempted to interview 100% of the population came up with a figure that was wildly inaccurate, and a survey that interviewed a tiny fraction of 1%, conducted by Mr Mayhew, produced a figure that was accurate?
Sarah Henry
I have not looked at the detail, but I am assuming you included lots of other variables in your estimate, not just one. You also include far fewer questions in that survey than are in the census.
Professor Mayhew
It depends whether you are just trying to verify the population count, or whether you are trying to collect 200 other variables from each individual. There is a substantial practical difference between those positions.
Paul Flynn
For the enlightenment of the nation, Professor Mayhew, would you tell us what the title of your book is, which you have just published today, I understand? I think we need to read it.
Professor Mayhew
That was on paying for long‑term care.
Paul Flynn
Oh, was it?
Professor Mayhew
It was carried out with Demos. I can give you a copy now, if you want. I have just rushed from the launch of that to here. I am talking about my other passion, which is population statistics and better use of them.
Chair
We are very grateful you got here. Juliet Whitworth, do you have a comment?
Juliet Whitworth
It has been picked up since, really. Sarah made the point that, although the census is not perfect, and the 10‑year census is not perfect, the administrative method is not proven to be perfect yet either. We do not disagree in the sense that we are saying “never” to administrative data. That seems a really sensible way to go. Local government would really like more frequent data, if that was possible, and administrative data is a good way of getting that. What we are saying is that we should not rush into it before we know it is good enough. Running the two in parallel in 2021 would allow us to assess that.
Keith Dugmore
I agree wholeheartedly with what Juliet has just said. One thing about administrative data we may be overlooking is that it can offer us new topics that we do not get at the moment. It is not just a matter of replacing questions in the census. There has been a long interest in information about income that has never been asked in the census. When we look at HMRC as a data source, we think, “Wouldn’t it be good to be able to harvest some data there, and produce aggregates and anonymous statistics for small areas?” There are some distinct pluses with admin data, as well as just substitutions.
Chair
Again, I am drawn to a question I asked the other panel. Is it really helpful of the ONS to distil this into a binary question between two options, either administrative data or a census, when in fact you all seem to be saying that you need both?
Keith Dugmore
I agree with the point you are making. When I read the consultation document, my feeling was that people are bound to say, “Do I vote for A or B?”, and it risks not thinking about a third option. In particular, the second option, which was described as “administrative data plus surveys”, really is primarily surveys, plus a little bit of administrative data. It would not give us anything much at a small‑area level. Some of us in the commercial world would urge the ONS to continue the good work and do it more quickly, and to try to get more access to administrative files swiftly in the next year or two, which probably involves legislation on data sharing.
Chair
Is the binary question helpful, Mr Mayhew?
Professor Mayhew
What we might be looking at is a hybrid, where the main system is administrative‑based, and you produce frequent—it is open to question what the frequency should be—population counts down to the local level. The local level can append and attach its own data. Some of that is centralised. Centrally produced data will have attributes on it such as those that have been suggested. Then you might carry out surveys, a bit like the one that I have done, to verify that accuracy, but you would do it on a smaller scale, and you would work in more difficult areas where counting population was slightly more problematic.
Chair
This would involve not a decennial census, but perhaps a much more regular but much more limited census.
Professor Mayhew
Yes. You would just rotate it and do it on a directed basis to areas that you were concerned with.
Chair
I would like to have the other members of the panel on this question. How helpful is it to have a binary question in this consultation?
Sarah Henry
It is not helpful, and that is why we did not answer the question in that way, but responded that we would prefer a hybrid.
Piers Elias
The ideas for a third way have only just started coming out, really, towards the end of the consultation. There were some ideas from the British Academy that Professor Sir Ian Diamond has suggested, where you have an online census and then you use those records for individuals to update in the future. That is either as their own updatable piece of information they update when a certain event happens in their life, like they move home, or every three years or whatever.
Juliet Whitworth
We also did not respond in a binary way, and responded to say “both” in the first instance.
Chair
Very interesting.
Paul Flynn
Can you give us some idea of how much business relies on the census, now out of date? It will be 2023 before we get information on the next one. Given the day‑to‑day reports that are put out on a weekly, monthly, annual basis by the ONS anyway, do they not rely most on the regular reports that come out and are much more valuable?
Keith Dugmore
The quick answer is they do not. The particular value of the census is the fact that information is for small areas, and that is what they need to make their decisions about local store catchments and so on. The ONS’s publications on national trends or regional trends may be useful in other parts of their businesses, but when it comes down to investment in local stores, that is of no use, really. It is the fine geography of the census that is valuable to them. If one went back two or three years, before the 2011 census came out, the 2001 census was still being used for store location decisions, albeit 10 years old, because there was nothing better.
Paul Flynn
There was in Newham.
Piers Elias
A lot of characteristics do not change very quickly. For example, the pool of skills you have in your area is going to be very similar from one year to the next. If you have a survey that is trying to gather data over five years, for example, which is what we would need to get small‑area data, you have 10%—or 14% in London—of people moving every year, so it is going to be a very confusing statistic. At least a census is a snapshot; it is a point in time, and we can understand that much more easily.
Chair
Supposing we went for option two, as it is described in the consultation paper, what would we lose that cannot be replicated in the way they described?
Piers Elias
We lose flow data, so that is travel to work and migration statistics. We lose a lot of information for small parishes. In Suffolk, it is almost a complete wipe‑out in terms of characteristic information. You get population counts, but nothing else. Just about every single parish in Suffolk is below the threshold needed. That is the same across several rural authorities
Chair
Do all members of the panel agree with that?
Professor Mayhew
Not completely, because I think you gain and what you gain could be argued to be more than what you would lose.
Chair
You think we can do without all that data that Mr Elias described.
Professor Mayhew
I am considering what is relevant for local authorities in their day‑to‑day work. What gets them excited is how many people will turn up at school on the first day of term, how many people will need social care in the future, how many post offices there are in the area and whether these are accessible to vulnerable groups.
Chair
Your argument is, essentially, that the data you lose, which Mr Elias has described, is actually not data that you need.
Professor Mayhew
There are uses for it. I am just saying that the local authority users have quite different perspectives and needs. That is why I argue that they should be enabled to append their own attribute data to national administrative data produced at the centre. That would overcome that problem. There is not one spanner that fits all nuts, basically.
Chair
You envisage replacing the national census, to a degree, with a whole lot of local authority censuses.
Professor Mayhew
No. The vision is that the centre would produce the core data, counting the population and certain key attributes that are available from administrative systems. Local authorities would be able to append their own data to that data, as and when they need it. Then the other part of it, the softer data—the issues about religion and so on and so forth—could be picked up in surveys. You would lose something at a local‑area level, and I agree with Piers on that, but you have to consider, in the great scheme of things, what the priority is and which is the most important data needed by users. Local authorities have been a forgotten element in this whole debate for many years. They do not voice their views as loudly as they might. Often, we hear the academic perspective or other perspectives instead, which are valid, but do not fulfil this need.
Chair
Mr Dugmore, it could be argued that the users you represent, if they really want this data, should collect it for themselves, and pay the cost of collecting it for themselves.
Keith Dugmore
I have two things on that. Firstly, I must pick up the point that the ONS’s second option in the consultation involves some admin data, but would only provide statistics by age and sex at small‑area level, and it does not provide anything else. It is important that is recognised when people are looking at the consultation. It is a very limited admin option.
Chair
I see Professor Mayhew nodding at that.
Professor Mayhew
Yes, I think they ought to consider adding a lot more attributes to that data. I completely agree with that. What they are proposing is very limited. They cannot even construct households from the way they are doing it because the data is anonymised.
Chair
Is option two an Aunt Sally option?
Professor Mayhew
Possibly—it is an unambitious option.
Keith Dugmore
It is a very cautious option. It really is.
Chair
You were going add something else, Mr Dugmore.
Keith Dugmore
On the point as to whether anybody else could do it better or companies do it themselves, the great attraction of having something centrally organised through Government is that you get universal, even coverage across the country, and it has many benefits for non‑commercial users as well—the great British public. The Government would either do that as it always has done as a traditional census or mobilise its increasingly large administrative data resources to come up, again, with something that provides good coverage across all areas of Britain. Nobody else can do it.
Sarah Henry
Arguably, they have paid for the census to be done in the most efficient way, through the taxes that they pay. The Government is spending their taxes. It would be extremely inefficient, across the whole of the industry, for people to produce their own censuses.
Chair
That is an extremely powerful point. I am almost running out of questions here, because we have dealt with this so comprehensively.
Paul Flynn
Has any other member of the panel apart from Professor Mayhew any comment to make about why it is thought to be essential to spend the ratepayers’ money on Professor Mayhew’s investigations? Were they wasting their money, or is there a really serious defect in the census material?
Piers Elias
The reason is usually to do with local authority funding. They tend to be local authorities who are undercounted.
Paul Flynn
That is very tight, and I am sure they do not have money to waste on this, but they took the serious decision because the census material on which they could take major decisions was not reliable. In Newham, they would have been taking the decisions that were hugely inaccurate, based on an inaccurate thing. It does expose the census as being far from perfect. There is this idea that any census can be perfect, which it is not going to be, but here were have a clear example of a major defect in the census.
Piers Elias
It is not a major defect in the census.
Paul Flynn
243,000 population compared with 300,000.
Piers Elias
The ONS use the GP patient register that Professor Mayhew uses to count the population in the mid‑year estimates, so it is obviously a problem with the stock of the GP patient register at the start of 2001. That is not an issue to do with the population being wrong in the census.
Paul Flynn
The result was so granularly accurate that it missed out 50,000 people in Newham alone. How can you defend that?
Piers Elias
It could have been that the base from 2001 was too low.
Sarah Henry
The other thing that this Committee might have discussed before this is relevant as well. It could be that that formula that you use in order to progress the population year‑on‑year to get your estimate—predominantly inward migration and outward migration—is wrong. There are quite a lot of factors. In Manchester, partly because of our experience, on top of the census, we have our own forecasting model that we use, but again, it uses, as its foundation, the census. It just adds more in, takes more out in a different way from the way the ONS did.
Chair
This raises a question. This question touches on the whole quality of what Government has been doing for a number of years around this whole issue. The Newham surprise should not have been a surprise. The Treasury Committee remarked on how unreliable the International Passenger Survey data was for forecasting local populations. We knew that already. How effectively is the ONS handling this whole issue?
Sarah Henry
For urban areas, it could be more effective.
Chair
What should we be recommending in our report that the ONS should be doing that they are not doing?
Piers Elias
We need a bigger survey than the International Passenger Survey to catch people who are actually going to stay here every year.
Chair
We know that on that particular issue. I am talking generally about conducting this consultation, preparing for more administrative data. What are the top three things that the ONS should be getting on with that they are not getting on with?
Keith Dugmore
Better access to administrative data really is the key here. You have mentioned international migration. I understand there is the e-Borders system that, in principle, tracks people going in and out of the country, but I do not think it is used for statistical purposes
Chair
Since we produced our report on migration statistics, the Government are pressing very hard on e-Borders information, but that is something that one Department is doing in one particular area. What should the ONS be doing in this general area of understanding our population and their habits?
Sarah Henry
They should make sure that they do not disaggregate the census from the work that is done in between censuses. It is important to look at the whole package. The other thing they should be doing is exercises that match data based on identifiable information. The fact that they are trialling anonymised data, in my opinion, is flawed and unnecessary, because the census itself is not anonymised. That is very much identifiable, so stick with that.
Piers Elias
The ONS should be lobbying you to introduce some sort of compulsory re‑registration when you move for local government services, or offer some incentive to make more people sign up to their GP, because the GP patient register is crucial to ONS estimates between the censuses, or to make sure the electoral register is 100%. 10% of the country is not on these registers.
Professor Mayhew
They need to get their skates on. They ought to carry out a lot more local studies in the vein that we have been testing in the past. They should then take that learning and feed it into the system that is going forward from 2020. These open‑data issues are incredibly important. One of the things that happened with the passing of the Health and Social Care Act in April 2013 was almost like data coup d’état. All data that were previously available at primary care trust level, with their abolition, migrated into the Health & Social Care Information Centre. No personally‑identifiable or address‑identifiable data is allowed to come out of that system. This is really the left hand not knowing what the right hand is doing, because, in the meantime, we have a public health function moving to local authorities, the setting up of health and well‑being boards, which rely on an accurate and up‑to‑date population to conduct any number of strategic and local issues. All that data, now transferred to the centre, is denied to them, and all that data is data they previously had access to. They are the very people who have the capacity to link that data and use it. It is a totally bizarre situation, and I do not know why it has not had more publicity than it has had, because that is driving a wedge. It does not apply across the whole United Kingdom. In Scotland, where I also do work and I am helping them, that is not the case with these data. It really is a case of the left hand not knowing what the right hand is doing.
Chair
Are there any other comments?
Keith Dugmore
Just to pick up further on administrative data and something Sarah said, it is important that the census is not seen as a single object. The whole process of producing neighbourhood statistics and more open data is really important. I would encourage the ONS to start producing aggregate statistics from more administrative files and publishing them, albeit saying that they are experimental statistics or whatever. Culturally, there is often a feeling that we cannot publish it until we know it is perfect. If data were put into the wider domain, I think that users would be able to say, “Well, this looks pretty good” or “There are some difficulties here.” To begin to churn out statistics that may be useful to people, while they cannot be perfect, would help us get towards better data for the next census.
Piers Elias
Starting with the 2011 census, the ONS published alongside the census estimates plausibility numbers from different administrative data sources, and they do intend to start publishing them in the mid-year estimates. I am not sure they are going to do it for the 2012‑13 set, but we are hoping they will. That will give local authorities more confidence in seeing that, for example, their five‑to‑nine year olds are within the figures that the pupil censuses are saying. We would encourage them to do that.
Professor Mayhew
On that point, if you use administrative data, you can get the exact age of individuals, whether they are 5.5 years old or whatever, so you are not forced into artificial age bandwidths for the purposes of developing a new service. Frequently, with the census, you are forced into certain age groupings, which may or may not be appropriate for the service you are trying to calibrate and provide.
Chair
This is just something that is clarifying in my mind. There are three kinds of data. There is census data, which attempts to collect from 100% of the population. There is administrative data, which attempts to collect from 100% of the population. Then there is survey data, which is the sampling so championed by my colleague, Mr Flynn. The administrative data, alongside the surveys, could replicate a great deal of what is in the census. The question is how much, and whether we have the capability to do that. Have I summarised it correctly?
Sarah Henry
In terms of the counts, that is right. There is still a question about people who are not registered on any of the administrative datasets for good reason, or counted on administrative datasets and do not match the criteria we would need for censuses. For example, people living abroad, and that kind of thing, are still registered with GPs locally, or can be.
Chair
Those tend to be the people that a census misses anyway.
Sarah Henry
It depends what the weather was like when the census was conducted. Our big issue is with the proposals around the surveys now, and the information that we would get at a level that matters to our local councillors—ward levels and even lower than that. At the moment, we think that the administrative data solution needs developing further, but it can be good for the count, and maybe the general demographics. In terms of what the survey would add, we think it is a long way from being a good solution. We need to think about how to boost it, how to change it and how maybe certain areas, like urban areas, could add to it, but it still needs a lot of development in the way it is presented in the consultation.
Chair
My last question, in summary, is how many marks out of 10 would you give the ONS for the conduct and supervision of population data?
Sarah Henry
Six.
Professor Mayhew
Is this off the record?
Chair
I guess that is less than 6.
Professor Mayhew
Well, I think they do some things well and other things not so well.
Chair
Average score?
Professor Mayhew
Five or 6—that is an average.
Kelvin Hopkins
This is just a resource problem, though. If we got the immigration statistics right, we would solve a lot of problems.
Juliet Whitworth
I have to do an average of my local authority colleagues, so I would say 6.5.
Keith Dugmore
I would go for a mean of 7.
Piers Elias
Seven.
Chair
Thank you very much indeed. It has been a very informative session. I hope you feel you have communicated what you needed to tell us. Do send in any further comments if they occur to you, and we can publish those as evidence or just informally to help us draft our report. Thank you very much indeed. The Future of the Census HC 1090 37