24
Examination of witnesses
Professor John Friedman, Jen Woolford, Dr Mark Brewin and Dr Emma Gordon.
Q158 The Chair: Good morning and welcome to an evidence session of the Social Mobility Policy Select Committee of the House of Lords. We are fortunate today to have a panel of four people who are going to give us help on data. I am Eliza Manningham-Buller, the Chair of the committee. I am not going to introduce everybody, because I want to move on to questions quite quickly.
We have Professor Friedman here from Brown University, who we are grateful to for getting out of bed so early to help us in our understanding of this subject. The first question is to him. It is a fairly general question to ask what policy insights the data that you are collecting gives you. In other words, at the end of this investigation by the Select Committee, we are going to recommend various suggestions to Government. What policy insights can you help us move towards?
Professor John Friedman: Thank you again for the invitation to the committee. The data that we have seen in the United States and, most recently, in the UK suggests a really critical role for social capital and, in particular, the role of economic connectedness, which is the extent to which children from lower socioeconomic status backgrounds have connections to broader society and to other children and families from higher socioeconomic status backgrounds.
Economic connectedness seems to be the single most powerful factor in explaining upward mobility, which is the ability of children to have opportunities as adults that outstrip what you might have expected, given fewer resources in their family when they were growing up.
This suggests both policies that directly try to increase these types of connections, and ways in which policies that, themselves, are not about connections can be made more effective.
I can give two examples of that. First, the role of social connections and of a community in an individual child’s upward mobility suggests targeting policies not just at individuals, but at communities. The intuitive reason is that, if you help not only that child, but all the children around them, you get a double bang for your buck, in that you are helping them and helping the environment and the connectedness around them, and both of those will help that child succeed.
Targeting communities, especially where there has been a historical lack of opportunity, is also just a very good thing to do. We see that that opportunity is highly dependent on the quite narrow neighbourhoods—within 500 metres or a kilometre—in which you grow up.
Secondly, this research suggests that, in the context of schools, universities and further education, we should think about conveying not only the technical skills—or the human capital, as economists like to talk about it—but also the social capital part of it. That is thinking of schools and universities as communities themselves, where we need to foster connections, but also thinking about some of the soft skills or, as some talk about, the navigation skills that students need to translate those technical skills into employment or to have the expectations and the aspirations for which learning those skills are a part of it.
For instance, helping students to navigate the choice of course in higher education, or the transition from university or workforce training into the workforce, is a way in which the most successful programmes that we have seen, mostly in the United States, focus not only on the skills training, but also on this social capital and connectedness.
The Chair: Can I ask the other panel members, possibly starting with Dr Brewin, whether there is a potential here to get more information out of the tax data of parents and children to inform some of the thinking that Professor Friedman has described?
Dr Mark Brewin: Thank you very much for the invitation to this committee. Indeed, HMRC is already supporting work led by DfE on providing income and earnings information, working in partnership with our colleagues in DWP, to support a product called the longitudinal educational outcomes dataset. As I said, this is a DfE product, but HMRC is very much supportive of it.
We are already, and have been for a while now, providing data from HMRC that can then be linked and matched to individuals and a record, essentially, over time that allows them to understand the earnings of that individual as they progress into employment and, indeed, self-employment. That can then help unlock potential around the opportunities mission that is key to the Government at this point in time. We are already doing some of that work, and have been for some time.
Part of your question also alludes to whether we can link between the records of parents and children, as it has been described by Professor Friedman. That is, again, another piece of work that is being led primarily by DfE, but HMRC is certainly supportive and providing input and data through our ability to share data under a certain legal framework through the pupil and parent data matching product. That is really a product that DfE is leading on. It would be for DfE to describe how far it is in its development, but certainly the way in which HMRC can disclose information in secure and legal ways enables us to link that data with DWP benefits data to match back to pupils and their progress over time.
LEO currently includes data up to around 40 year-olds, so they have the educational outcomes linked through to their earnings information or, indeed, their benefits information up to around the age of 40 years.
The Chair: I apologise. I should have asked you to say what your role at HMRC is.
Dr Mark Brewin: I apologise. I am a deputy director inside the chief data office of HMRC, and I have a remit that covers data sharing. One of the key parts of my role is ensuring that HMRC, where we are able to, shares data that we hold for our tax and administration functions through legal gateways to support wider benefit across government. HMRC recognises that the information we hold on individuals around earnings and self-employment is a key asset for government, so we are keen to make use of that to inform wider policy-making across Government.
The Chair: Professor Friedman’s job is described on the screen, but I should just ask Ms Woolford and Dr Gordon in turn to say what you both do.
Jen Woolford: Thank you for inviting me to be part of this session today. I am director of public policy analysis at the Office for National Statistics. Within that role, my teams have strong links with both central and local government to understand their information needs, to provide responsive analysis, and to support them in developing analytical capability. We also publish a range of policy-relevant statistics on household income, wealth and crime. My background is in census and demography, so a spread of the data that we hold.
Dr Emma Gordon: Thank you very much for inviting me. I lead the Administrative Data Research UK programme at the Economic and Social Research Council. It is our job to work with the major government data owners such as HMRC, DWP and the Department for Education to open up access to linked administrative datasets, which are population-wide, for research use in secure, trusted research environments. The one for ADR England that we use is the Secure Research Service that is managed by the ONS.
The Chair: Does either of you want to add anything to the answers that we have already heard from Professor Friedman and Dr Brewin?
Dr Emma Gordon: We work really closely with DfE, HMRC, DWP and HESA to open up research access to the longitudinal education outcomes dataset. Through that and the work that has been done on it by academics, we have that feedback loop going into those government departments to improve it over time. We have had one iteration of that improvement so far, and there are opportunities for further iterative improvements as we go forward. We are also opening up access to longitudinal education outcomes data for Scotland, Wales and Northern Ireland over the next few years.
Q159 Lord Ravensdale: I have a brief follow-up question for Professor Friedman. You have a really impressive, granular dataset there, and you suggested some policy interventions that result from that. I just wondered whether you had had the opportunity to evaluate the effectiveness of some of those policy interventions with the long-term view that that data would allow you to get.
Professor John Friedman: The answer is yes and no. Some of these policies are newer, because they are being informed by these more recent results, and so we have not yet been able to track these children or adults who have been benefiting from the programme to see the outcomes. There are some programmes that have been in operation already, and part of this set of policy recommendations comes from data that evaluates the effectiveness of those programmes historically.
To give you one example, I mentioned education and workforce training programmes. One such programme in the United States is called Year Up. One of the distinguishing features of this programme is, again, that it is focused on young adults and giving them not just the skills to, for instance, serve as quality control adjusters in banks or as administrative professionals within healthcare organisations, but also an understanding of some of what comes through economic connectedness for those who have had it, for example how you greet other individuals when you walk into a professional workplace, how you put a résumé together to highlight what you have done, or the importance of showing up on time, in that, if you have to be at the desk at 9 am, you really have to be there at 9 am.
These are things that, from one perspective, are really quite straightforward, but, if as a young adult you have grown up in an environment without exposure to people who are working in these types of environments, it is very easy to see how you just would never have learned that. If not addressed, that is going to put you at a serious disadvantage when then trying, even with the technical skills, to move into the workplace. This particular programme generates enormous earnings gains for the participants. This has been validated through randomised control trials as well as a broader analysis conducted by Opportunity Insights.
Q160 Lord Watts: I was very interested to see the map, because it indicates that, even within prosperous areas, you have outlying areas of deprivation. That is a common factor with the UK. There is a group of people who think that you can have trickle-down economic policies, in the sense that, if you centre your investment in the city region, for example, that will filter out to the outlying areas. That does not seem to have happened in New York, according to your map, or in many of our areas in the UK. Is the trickle-down theory something that has any credibility?
Professor John Friedman: Let me not pass judgment on a grand economic theory, but there are certainly many cases that we see in the data where very low levels of mobility for children co‑exist with a very economically vibrant city centre.
I would draw your attention to the example of Charlotte, North Carolina, in the United States. This is a city that is one of the economic engines of the American south-east. It is the headquarters of many large corporations, as well as a wide range of other important businesses. This was a city that, in an initial set of rankings that Opportunity Insights put out, ended up as 50th among the 50 largest cities in the US in terms of economic mobility for children. Again, this is despite being one of the fastest-growing cities in the US.
Why was this the case? Why was there this wide gap between the economic growth of the city—and these are really good new jobs that were being created—and the economic mobility of children? It is exactly about the economic connectedness. When the city was growing, children who were growing up in lower socioeconomic status communities and families were not coming to the labour force with the human or social capital to get those jobs that were being created. Instead, the banks, healthcare organisations and other companies were just hiring people who were moving into Charlotte. There can be enormous divides even in very prosperous cities, and pockets where opportunity is just not present.
The Chair: Thank you. Can I just say this to the panel? We have something like nine more questions that we want to get through. The rest of you may have further answers to add to Professor Friedman’s, but it would be very much appreciated, if possible, if we could keep questions and answers crisp. Otherwise, we will never get to the end of our session. Thank you all very much for that. Unless Dr Gordon or Ms Woolford had anything to add to what has been said, could we move on to the second question?
Dr Emma Gordon: Just very briefly, using this LEO data that we have for England—the linked educational outcomes, benefits and income data—we have had some natural policy evaluation opportunities with things such as Sure Start. The IFS has published a couple of reports recently on that. DfE is working on a flagging method whereby, if a child has been in an education intervention, that will be flagged on the record within the LEO database, so you can follow them up over time to see not just the impact of the education outcomes but also later income outcomes.
Q161 Lord Harlech: That leads nicely into the second question. To the whole panel, but to Ms Woolford first, can you outline some of the main datasets currently and typically used to analyse socioeconomic background, disadvantage and social mobility in the UK? What are the advantages and disadvantages of the datasets used? How could these be improved or other ones looked at?
Jen Woolford: From an ONS perspective, there are two key datasets that we use. There is the Labour Force Survey and Annual Population Survey. Then there is our longitudinal study. We also have the opinion survey, which collects some opinions on and perceptions of social mobility, but, to get into the detail, those are the two main datasets that we use.
The first of those—the Labour Force Survey, which is supplemented by the Annual Population Survey to get down to smaller areas—collects information on an individual’s occupation and the occupation of their parent if they are still living at home; if they are not living at home, it has a question about the parent’s occupation at age 14 and where they were living. You have that information to give you something about social mobility, and then you have all the questions on the LFS around their social demographic characteristics, employment and so forth. That is one of our key datasets that we use.
The key challenges that we have with the LFS at the moment are the well-documented quality issues and the low response rates. We have taken a lot of steps recently to put more interviews on to the survey, to get back to face-to-face interviewing, and to increase the incentives. We are seeing that coming through with improved response rates and better outputs from the LFS.
We also have a programme of work developing the transformed Labour Force Survey to get that on to a more sustainable footing. We are soon publishing our plans for improving our economic statistics and our surveys, with an increased priority on that for ONS, so moving more people on to doing that work. That should be published shortly, and you will have more information.
I do not know whether you have heard about the longitudinal study already, so, if I go into too much detail, let me know. That takes about a 1% sample of the population, from the 1971 census onwards, and links them census by census. It also links data on births to cohort mothers, cancer registrations up until about 2015, and mortality. You can follow people for the last—I was born in 1971, so I should know this—50-odd years to see what their circumstances were growing up and what their educational attainment is compared to their parents. That is a really rich and wealthy source of data.
Generally, that is provided in the Secure Research Service, which I am sure we will say more about. It is available then to ONS analysts, but also academics, other government analysts and private citizens to do research under very secure circumstances to make sure that the data are protected. It is only a 1% sample, which is big for survey data.
Lord Harlech: Yes, that is still a lot of numbers, is it not?
Jen Woolford: It is. It is half a million people, but, if you really want to get down into local areas, it would not be able to do the sorts of maps that Professor Friedman was showing, for example. That is a challenge if you really want to get into specific subgroups.
The other thing with that longitudinal study is that it is built on censuses of England and Wales. There are sister studies in Northern Ireland and Scotland, which are slightly different, but we try to bring the data together as far as we can to allow UK analysis.
Dr Emma Gordon: Jen has given a really comprehensive account of the benefits of survey data, where you can go really deep into the data, but it is not population-wide. The administrative data that we make available for research use is the opposite way round. It is very broad, but you are entirely limited by the bits of information that were picked off the Government systems and services at the time that the data was collected. LEO is an administrative dataset, not a survey dataset, so it has complete coverage of the population that is covered by those systems and services.
We have an indicator only for parental income at the moment, which is very incomplete. It is a free school meals indicator, so it is binary and you have no sense of scale. That would be the biggest improvement that you could bring to LEO, so that you have much more granular detail about parental income, and then you can follow children up through education and into their own earnings.
Lord Harlech: We need to tie that one data point with something else to build up a picture and then track it through time.
Dr Emma Gordon: Yes.
Dr Mark Brewin: I do not have much to add. The key point, certainly in administrative data source terms, is the one that Emma brought out. Ultimately, linking through from the child to the parent is going to be really important for the future use of LEO. You will be hearing from other departments in the future, but that comes back down to a product called the pupil parent matched dataset, and how that bears fruit over the coming years.
Professor John Friedman: This is a challenge that we faced in the United States as well. One of the efforts that we have made is to go back to some of the earlier census data. All these data are stored on magnetic tape files, which are totally inaccessible to research, but, through about a five-year collaboration with the Census Bureau here, we are scanning those records and using optical character recognition and optical mark recognition. Then there is a linking process to digitise and, ultimately, add those data, going all the way back through the 1950 decennial census in the US, to this longitudinal data history.
That is going to be transformational in terms of giving us a chance to understand not only the backgrounds of people who are older adults today, but some of these monumental social changes that happened in the 1960s, 1970s and 1980s. I do not know what the capacity is, or where old data are lying around in the UK, but efforts such as that, while expensive relative to research, are very cheap in the grand scheme of things and can really add tremendous value.
Q162 Lord Hampton: A lot of these questions have been answered. Professor Friedman has answered my next question pretty fully, so can I ask a separate question to Professor Friedman very quickly? You talk about tax records and Facebook data. Rather innocently, what do you mean by Facebook data?
Professor John Friedman: We use data at the social media company, Facebook, to see who is friends with whom. That, of course, is a crude measure on the face of it. Then we validate it in various different ways. To be clear, the data from Facebook, as part of Meta, are not linked with tax records. Those are analysed entirely within the secure data systems at Meta. There is an enormous amount that you can learn about social connections from those data. When used in creative ways, you can link that, as in the UK paper I showed you, to get some measures of social mobility as well.
Lord Hampton: Thank you very much. That was really interesting.
To the whole panel, you have all talked to an extent about the types of insights into social mobility that are gained through being able to share and link. What are the problems? Dr Brewin, you talked a couple of times about certain legal frameworks and gateways. If you could wave a magic wand, what laws would you change to be able to get these datasets linked? What kinds of insights could you get from them?
Dr Emma Gordon: The Digital Economy Act has been completely transformative for how we have been able to collaborate with UK government departments to open up access to population level administrative datasets.
The one caveat is that health bodies are excluded from using the Digital Economy Act to open up access to health data. As you can imagine, a person’s health status and the conditions that they have can very much impact their education, their earnings, and their subsequent social class as well. Being able to have that contextual information of their health data alongside their education and income data would be incredible.
To give you an example, we have opened up a linked dataset that is education outcomes over 25 years with their health data for England. It took us four years to not create the dataset but to figure out a legal way to open it up for wider research use. It is completely unsustainable for me as a programme director to do that every time we want to link in health data. Removing that exclusion from the Digital Economy Act would be transformative to get that information. It does not have to be every bit of information, but just small indicators of health status would hugely improve that situation.
Q163 The Chair: Can I just ask a supplementary, Dr Gordon? I thought that Health Data Research UK was precisely set up to do this. Have I misunderstood something?
Dr Emma Gordon: My understanding is that Health Data Research UK is set up to scale up access to health data. When I talk about health data, it is all of the hospital episode statistics data. It is GP records. It is genomics data. It is largely focused on opening up access to health data. What I am talking about is linking health data to that government administrative data.
The Chair: But there is an organisation and, in principle, if the law allowed, there could be that exchange.
Dr Emma Gordon: Yes. The Health and Social Care Act facilitates research access to health data, but I would argue that that does not have all the safeguards in it that the Digital Economy Act does in terms of how researchers have to access the data. Under the Digital Economy Act, there has to be a Digital Economy Act-accredited, trusted research environment such as the ONS Secure Research Service. The research has to be in the public good, and you can take only the statistical outputs of analyses out of that environment, not the individual data. That is why it has allowed us to open up so much data. It is a balance between benefit and risk, and all the risks are mitigated within that system.
The Chair: Lord Hampton, does that answer your question?
Lord Hampton: If you could send us something about that, that sounds like an avenue we could go through.
The Chair: For our witnesses online, if we do not ask you specifically but you have particular points that you want to make, just put your hand up and we will come to you.
Jen Woolford: From my perspective, as well as the legal barriers, there are some real cultural and technical barriers as well. All of us who own data are really passionate about keeping it secure and safe. The idea of sharing it has taken a while to become more embedded, but we are getting there across government.
Technically, it can be quite expensive and difficult to extract the data from the systems and provide them in a way that we can use. Once we have those datasets, you have all the challenges of linking them, and linking them as accurately as you can, so that you can do this kind of analysis over time or across different data sources. The data do not necessarily collect what you want from a statistical perspective either, so the issue is then how you get from what is in the administrative systems to what you are interested in. Income is a really good example of that, because people’s income comes from so many sources that creating an income variable from admin data is quite hard and we have to make compromises.
If there was one single thing that would make our lives a lot easier, it would be to have a unique identifying number that goes across all of those datasets. That would really help the accuracy of our linkage and speed up that process as well. We are trying to create that in ONS by having an index that links together things such as your tax number, your NHS number and your passport number, just to make that easier to do. A single identifier would really help us to do that in a more effective way.
Lord Hampton: Good news seems to be going into the Children’s Wellbeing and Schools Bill, does it not? The unique identifying number seems to be gaining a lot of traction.
Dr Mark Brewin: Just to add to the descriptions from Dr Gordon, when the data is presented to the researchers, we just need to remember that it has also been de-identified. In that sense, it has been protected. That then links back to some of the other pieces of legislation, certainly when we are sharing personal data, that we all need to be cognisant of, such as UK GDPR and all of those principles. They are not a barrier. They are a way that enables departments and, indeed, researchers to carry out their research, and to do it in a safe and secure way.
Data protection law, in that sense, will have specific bits that say, “This must happen”, but, at the same time, it should not be seen as a barrier to enabling all of the work that is currently going on under the Digital Economy Act. It is a way in which we can facilitate and make sure that the rights of the individuals are protected. It is not, in itself, a barrier. That goes back to how the data is being protected when it is being presented to the academics or the other researchers who are analysing the data, which is really important in making sure that we retain the trust of the UK population when we are doing all of this research.
Q164 Lord Ravensdale: Ms Woolford, you have already mentioned the Secure Research Service in the context of some of the data, but I wonder whether you could perhaps expand on that and, for the purposes of our evidence, outline what the Secure Research Service aims to do and the value of its current output for understanding social mobility. Following on from that, how will that be developed by the Integrated Data Service, and how will that improve data on social mobility in the future?
Jen Woolford: The Secure Research Service is a platform provided by ONS that allows accredited researchers secure access to de-identified sensitive data for the purpose of conducting statistical research. We apply five safes to make sure that everything that we are doing is really safe and secure, and that we are protecting individuals’ data as well. We make sure that it is an ethical and safe project, that the researcher is somebody we trust, that we have safe data in there that cannot be easily identifiable for individuals, and that there are safe outputs. We have put lots of safeguards in place to make sure that the data is secure, that the analysis being done is ethical and legal, and that it is being done for the public good.
The SRS holds most ONS survey data, as well as the longitudinal study that I mentioned earlier, so it provides access to all of that data to analysts. At the moment, it has about 400 datasets on it. We have nearly 2,000 users, 800 live projects, and 150-plus organisations, so it is a really broad use of it. About 55% are academic users, 25% government, and 20% private and third sector, so a broad range of users. Last year, we had about 250 articles published on a really wide variety of things, but mainly on education and economy.
Through our partnership with ADR UK, we have several linked data assets in there that can provide information on social mobility, but Emma is probably better placed to talk a bit more about those linked datasets than I am.
Dr Emma Gordon: Since 2018, the ADR UK programme has been funding the Secure Research Service to expand and improve it. It was always there, largely to provide a service to external researchers to gain access to the ONS social survey data. We have funded it so that we can hold these large linked population-level administrative datasets from across government as well. We have also worked with them to find ways of smoothing the research journey as much as possible, while making sure that the integrity of those five safes that Jen mentioned remains.
We have opened up access to things such as LEO and the ECHILD dataset that I mentioned, linking education outcomes to health data. We have also opened up access to all of the courts and prisons data, working collaboratively with the Ministry of Justice through the Data First programme that we fund. In our next investment phase, we have ambitions to take that work forward a step further.
The work that we fund with the Ministry of Justice is called Data First, because we wanted it to figure out how to link and clean its data first, and then link in other bits of information such as benefit and income records, so that you can see what happens when somebody comes out of prison, and before they go into prison, to test where the bottlenecks are and which government services need to improve.
We also have a commitment to open up access to more linked data to support econometric analysis, and this more health and administrative data moving forward. There are lots of opportunities to improve data to support social mobility analyses, working with all the government departments that I mentioned.
Lord Ravensdale: Thank you. I just wanted to clarify something that came up in the previous question about the unique identifier. In the IDS dataset, you already have an anonymised identifier. Can you just explain, in terms of the unique identifier, why you need that in the context of the anonymised identifier that you already use? What would the key change there be?
Jen Woolford: We have developed an anonymised identifier, and that is really helping make our linkage higher quality and more efficient to do, but there are still problems when people come in. We still have challenges linking data, and those challenges do not apply equally to everybody. For people with non-European names, for example, which can often be misspelt in datasets, those links can be hard to make, so you can end up with broken links and it looks like you have more people than you actually have. For people moving into the country, it can be more challenging to establish those links for the first time. We are improving the quality all the time in what we do.
One problem that we have is when people move around, or in and out of the country, and how quickly that gets updated in the administrative systems. If we have a way of identifying more quickly that somebody is a migrant and has left the country, that improves our understanding of the population.
The Chair: We are not even halfway through, and so I hope that our witnesses can be a bit flexible at the end. You possibly have to go off and teach, Professor Friedman, but we will hope to hang on to you slightly beyond our cut-off time.
Professor John Friedman: It is summer here. I can hang out as long as you would like.
The Chair: And the Brits can just stay here.
Q165 The Lord Bishop of Lincoln: Dr Brewin, you alluded earlier to the pupil parent matched dataset. I would be interested to know what the current status of that is. I ask that in the context of the fact that most administrative data collected by government relates to individual adults. The Social Mobility Commission states, “information is not routinely collected in these administrative data sets about a person’s wider household structure. But this is needed to understand the different factors that may contribute to social mobility outcomes”.
Could such a household dataset be created that contributes to Professor Friedman’s narrow geography, making it even narrower in identification? What, if any, are the issues that would need to be addressed in creating such a dataset? I am thinking of how that links with longitudinal survey data, such as the Millennium Cohort Study.
Dr Brewin, since you raised the parent pupil set in the first place, perhaps you would go first.
Dr Mark Brewin: The parent pupil database is a DfE product, so DfE is leading on it. HMRC, as I say, can provide and disclose information under various legal gateways to enable and support that product. The wider question is how you construct and understand households from an administrative data source perspective. If we think back, HMRC’s primary role and function is the tax and customs authority, and UK tax is, essentially, at an individual level. There are elements where HMRC collects information for our functions, such as child benefit and tax-free childcare, but those two elements do not always create, if you like, a position that says, “We now understand what the household looks like”.
In the example of child benefit, it is usually the claimant and the child only, not the partner. Even then, we do not have the full number of parents claiming who are eligible to claim for child benefit, so that is a partial match. Similarly, tax-free childcare, which HMRC also administers, covers only a very small percentage of households in the sense of those that are actively claiming for tax-free childcare, and so, again, it is a very partial picture.
The challenge more widely is to understand how you can create what can be quite a dynamic situation for households. Define what a household is and what it means to many people, and you may get slightly different versions of it. Then what is the administrative data that can then create a patchwork of information that enables wider government to understand what constitutes that household and what its components are?
It sounds simple to define and describe in these terms, but, when you are looking to use administrative data sources so that you have the full population, it can be quite complex, particularly when you think of concepts of households or houses of multiple occupancy and whether there is a relationship between those. It is about establishing a relationship between those and then maintaining that information.
Certainly from an HMRC perspective, there is wider work that we are supportive of, with government, to understand households much better, but it is not something that HMRC, with its own functions, can actively create in its own right. It is not part of our tax administration functions, in simplistic terms.
There are a number of challenges around what a household is and then demonstrating, through administrative sources, the links between people over time and recognising that there is flux and change in that household.
The Lord Bishop of Lincoln: Thank you. This is a particular challenge for the ONS.
Jen Woolford: It is, indeed. We have the advantage of our surveys, many of which are collected on a household level, and the census, which is such a rich information source but comes only every 10 years. We have been doing some work over the last few years on our future of population and migration statistics programme, which has looked at how much of this information we can construct from linked administrative data. One of the challenges that we have is that we have addresses for people, but not information about households and families, which, as Dr Brewin has already said, is very difficult to define.
Around the time of the 2021 census, we looked at how closely we could match between people at an occupied address and households as collected in the census. Overall, we found that there was a really high correlation there, but that the households where you do not have that correlation are the ones that you are most interested in from a policy perspective. Those are the ones that we really need to understand.
We are doing some work to think about whether we can use what we know from the census and from surveys to look at that relationship between what you collect and what you get from your administrative data, and to do some modelling to think about whether we can get to better estimates of households and numbers of households.
As Dr Brewin said, one of the big challenges is defining it in the first place. Every census, we go through a big consultation about how we are going to define a household this time. It used to be sharing an evening meal, but that is not a particularly helpful definition of many families and households now, so we will go through that again in preparation for 2031, should the Government recommend that we do one. We may well find that what we have to do is construct something according to the data available to us, rather than according to what, ideally and in a perfect world, we would want to have.
Dr Emma Gordon: The ESRC has, for many decades, funded a whole suite of longitudinal population studies. As a by-product of the work done during Covid, we recently funded a platform called the UK Longitudinal Linkage Collaboration, which brings a lot of these social survey datasets, where it is consented data. During Covid, we were linking a lot of the electronic health record information to understand the impact of Covid.
Since Covid, we have been working with HMRC, DWP and DfE to find out if we can feed the cleaned administrative data that we have created for research use within ONS through to the same platform, where there is consent to do so. We hope that that will be coming through in the next couple of years, enabling us to get the benefits of both.
The Lord Bishop of Lincoln: Is the Millennium Cohort Study one of the things that you have been involved in?
Dr Emma Gordon: I believe that the people in the Millennium Cohort Study are around age 60 now. In terms of understanding social mobility now, I do not know whether that is such an important one, but things such as Understanding Society would definitely be really important, because that follows people consecutively.
The Lord Bishop of Lincoln: Professor Friedman, you talked about narrow geography earlier. We are talking about even narrower geography, if we are talking about households. What observations do you have to offer about how you would assemble such a dataset?
Professor John Friedman: The household is very important. The local environment is very important. As this committee has been talking about, the details depend on what comes easily through administrative records, and then where you fill in the gaps. In the United States, we have different challenges, because we have a household taxation system. We tend to get households of a certain type very easily, but then, if there are two unmarried parents living together, that is harder and we have to figure that out separately.
I do not think that there is any magical answer here. It is just important to make sure that the perfect not be the enemy of the good. These are datasets of millions of people. There are going to be errors in the data, and the question is not, “Do you have a perfect dataset?” but, “Have you captured things sufficiently to align with the patterns that you want?” Just as important, as one of the witnesses said, is to do so in a way that gives you good results across a broad range of society, not results that are feasible for only one segment.
Q166 Lord Young of Cookham: I am happy to forgo my later question if I can just press Jennet on this one follow-up. We are looking at social mobility and specifically at NEETs.
The Chair: Could you define “NEETs” for Professor Friedman? He may not know what it means.
Lord Young of Cookham: NEETs are young people not in education, employment or training. We have heard that a good indicator is whether the child has been in care, whether they have a poor attendance record at school, or whether they have been in contact with the judicial system. What might also be relevant is whether the child has a learning difficulty, in which case the DWP would have that information.
It would be really useful if, when a child leaves school, the careers service or Jobcentre Plus knew that this particular individual was going to have more difficulty finding employment or training. Is there any way that that data could be put together so that whoever is trying to help that individual has it? Is the unique single identifier that is being introduced under the Children’s Wellbeing and Schools Bill of any relevance to this?
Jen Woolford: If the data exists out there on systems, it is technically feasible to bring it together and to use it, and a unique identifier would always help that. In terms of what the ONS can do, we do this work for statistical purposes, not operational purposes, so it is always on an anonymised basis. It is outputs of statistics rather than about a named individual, so it would have to happen under different legislation if you were going to bring the data sources together and then say, “This is what we know about Jennet Woolford”. That is an operational use and, therefore, does not come within the remit of ONS or our legal permissions, as it stands.
Q167 Lord Evans of Rainow: I was going to ask a question to Dr Brewin regarding something that he said, which is relevant to my question. You talked about HMRC and household income. HMRC takes money away from households in taxation, and then gives it to the Department for Work and Pensions, which then distributes it to other households in benefits. We have heard about gaps in social mobility data, for example that occupational data, which is not routinely collected by government as part of its administrative function, could be collected by HMRC from employers through the PAYE and national insurance system.
With that in mind, HMRC has this data, and it has an interest in collecting as much tax as possible, but, in any dealings with the Treasury, it is also reluctant to give the money out without it going to government departments, so that it is not wasted. With that in mind, are there other gaps that might be usefully filled by HMRC to help produce a dataset that would enable decision-makers and Ministers to better target that welfare spending for DWP and the Department for Education?
Dr Brewin: If we start with the fact that HMRC collects information to support its primary tax and customs function, everything that we collect from individuals—and UK tax is at an individual level—has to support a tax administration function, in simplistic terms.
If we are looking to collect more information—and this has been part of a recent public consultation, but with particular emphasis on a number of additional attributes to be collected through employers under the PAYE system or, indeed, through self-assessment processes—we have to be clear that it supports a tax admin function, unless new primary legislation is created and creates a new function for HMRC to undertake that.
When we hold and are using that information, we are able to then disclose and share with other parts of government within certain legal frameworks. What you are describing sounds much more like a public service delivery kind of subject area rather than necessarily any statistical and additional work from a research perspective.
Under some recent work that we did through the public consultation, we were asking whether we should collect more information around occupation, for example, and whether that might help with our tax administration function, but also whether it would be beneficial to more general and wider government use.
When we are collecting more information, certainly through an administrative system, we need to think about wider benefits to government and, as in this case, wider benefits for research on social mobility, but also the additional burden and costs that that might place on employers, and the self-employed in the particular case of this public consultation.
After lots of consultation and then listening to the feedback through that public consultation, which took place in 2022 and then reported in 2023, there were only a certain number of additional data items that government decided would be beneficial for the tax system and to support wider government aims. On balance and taking into account all those aspects, occupation was not one that was taken forward and brought into the data collection from an HMRC perspective.
That specific case really just highlights the more general concept of how it supports, certainly from an HMRC perspective, our tax administration function, but also of weighing up and understanding the benefits as well as the burdens that that might place on employers, in this case, or, indeed, the self-employed. More generally, what is the utility of that information? What does it look like? How good a quality is it that would then enable other and wider benefit and use?
The Social Mobility Commission did ask questions around why we are not collecting occupation data more routinely, but, as I said, that was something that Government considered and that HMRC specifically responded on in 2023. The decision was that, on the balance of those two elements, we were not going to collect occupation data.
Lord Evans of Rainow: So we have a gap that HMRC could collect but is not collecting. Can we ask Dr Emma and Jennet to comment on that, please?
Dr Emma Gordon: Occupation is tricky, because you also get, apparently, within occupation, differences in income. Occupation itself is not a super clear marker of social mobility. It comes down to social surveys, where it is consented data and you can ask these questions, get really detailed information about family background through a slice of the population, and follow them up over time.
With administrative data, it is much more about piecing together the bits of information you can, so benefits and income data, and the potential to increase that through looking at parental income and, through tax information, at wealth. There are a lot of pieces of the jigsaw that you could add there, but it is a different measure from occupation.
Jen Woolford: The main source that we would have in the ONS is that longitudinal study, which collects detailed information on occupation and industry only every 10 years. If you are following somebody through who was born in the 1960s and, therefore, appearing in the 1971 census, you have a lot of information there about the parental occupation, industry and demographics, and then you can follow that person through year after year.
Lord Evans of Rainow: What are the datasets that would best help Ministers and, therefore, government to identify and, therefore, target limited resources to those people who are potentially going to be NEETs? What would be the key datasets?
Dr Emma Gordon: Dr Brewin mentioned the pupil parent matched database that the Department for Education is putting together. The department is actively considering whether that could be linked through, in some form, to the longitudinal education outcomes dataset. The limitations are that, while it is really rich information, it is only for families claiming universal credit. It is deep, rich information on probably the most at-risk part of the population, but not everybody.
Lord Evans of Rainow: Could you recommend to the committee in writing what would be the perfect dataset for us to recommend in our report on social mobility?
Dr Emma Gordon: Yes.
The Chair: I am sorry to put you to more work. Professor Friedman, is there anything that you want to add on this subject? Otherwise, I will press the committee on to the next question.
Professor John Friedman: Occupation is really hard, as the witnesses have said, but it is oftentimes easier to get the firm that someone works for, which, combined with income level, is often something that is really close, not only to occupation per se, but to what you are looking for from a broader social perspective.
Lord Evans of Rainow: Do we still have occupations on birth certificates?
Q168 Baroness Garden of Frognal: Lord Young has, effectively, asked my question, which was about young people not in education, employment or training. We were recently in Blackpool, where they had done an immense lot of work to try to help those young people at risk, and collected a great deal of data. How far can or does central government work with local authorities to try to collect data on young people?
Jen Woolford: There are a couple of aspects to that. One is that, if a number of authorities are going to collect similar data, it would be more straightforward if central government developed and tested a question, which was available and well tested for each of those areas to use. Not only would it be more efficient, but it would mean that the data that each of those areas is collecting are consistent, so you could do those comparisons. That is quite a central role, so methods are developed only once.
Baroness Garden of Frognal: Is that happening? Is central government doing this?
Jen Woolford: What we have in my area is a team called ONS Local. They have boots on the ground in all the regions and countries of the UK, and have those links with local authorities. They are talking to local authorities all the time about what their information needs are, seeing whether we can answer those needs centrally for them and, if not, supporting them in getting the data themselves. If Blackpool is doing something and they then see that the West Midlands are doing it as well, they will put those two areas together and say, “Do that together”.
We can also help point people to central data that might help. We also help doing the analysis if necessary. It is quite a small team, so it cannot do everything everywhere, but we have that infrastructure in place to try to introduce greater support across different areas and to help them to build on what each other are doing.
Dr Emma Gordon: We have funded some data collection trying to get data from local authorities directly to create a nationally representative administrative dataset. I have to say that it did not work for us. As Jen said, a much more efficient way to get that data is for education authorities, for example, to feed that data through to the Department for Education, as the central government department responsible for that particular system or service. Then we can make it available for research use.
The Chair: Lord Young, did I hear you say that you would skip your question?
Lord Young of Cookham: In the interests of making progress, yes.
The Chair: It was a question about the challenges. You have covered some of it, Dr Gordon. Unless there is anything you want to add, perhaps we could move on, because we are out of time now.
Dr Emma Gordon: Lord Evans mentioned birth certificate data and whether you could take occupation data from there. There are two issues with that. First, when women have just given birth, their income is likely to dip, because many do not work when they have a very young child. There is incomplete information about the father, who might not be on the birth certificate. It is one source of information, but is not complete, and it is a snapshot at one particular time in a life.
Lord Evans of Rainow: It is a start.
Dr Emma Gordon: Yes.
The Chair: I have been reminded that this session goes on to 11.45 am, so I apologise. We usually stop at 11.30 am. Lord Young, would you like to ask your question, in case there is anything more that anyone wants to add?
Q169 Lord Young of Cookham: It is a two-part question. From the Government’s point of view, is the research that is being carried out presented in such a way that you can use it? From the point of view of those doing the research, are the Government making the fullest use of all your hard work?
Dr Emma Gordon: That is a very good question. As a programme, where we have made the biggest gains is where there are mutual benefits for the government department, as the data owner, to work with us in order to open up access to research data. We make sure that we close the loop and that that evidence gets fed back into government departments. Civil servants working in government departments always have a huge list of competing priorities, and it is really difficult for them to do this really deep, rich analysis.
It is what academics build their careers on, so it is our job to make sure that we close that feedback loop. We have a fantastic website. We have lots of policy impact case studies on there, and we run events bringing policymakers and government analysts together with the researchers to create those feedback loops, so that it becomes more normal for them to look outwards for that evidence. Ministers are always going to have to make decisions about policies, whether there is that evidence there or not, but, if we can make it as easy as possible for them to find those evidence syntheses, then we are doing our job.
Jen Woolford: I agree with everything that Emma said, but we are very constrained in what we can do within the ONS. We have to make sure that we are spending our resources and expertise on the most impactful things, and particularly on things that only we can do. We can make the data available more broadly for people who can link it to other sources and do more detailed analysis.
The other thing that we are doing is trying to make the data more accessible, so that people can play with our data more easily just through our website. We recently launched “explore local statistics”, which is an online tool that allows people to go in through the lens of geography and say, “What do I know about my area?” There is a whole range of survey and administrative datasets behind that, which people can use in one place to understand their local area.
The Chair: Thank you. Let us now move on to Lady Ramsey’s question. Just for the benefit of the witnesses, she was not bored of listening to you. She had a parliamentary question that she had to go and ask during the early part of this session. That is why she disappeared briefly.
Q170 Baroness Ramsey of Wall Heath: I apologise for that. It may mean that you have answered some of this question and I have missed it, in which case please do not feel that you need to repeat it. Several witnesses during the course of our inquiries have recommended that employers record and report on their socioeconomic pay gap in the same way that they report on the gender pay gap. I am interested in what you think about that. If this were done centrally, how could it be collected and managed?
Dr Emma Gordon: I worry a bit about that, because there are quite a few questions that you would have to ask in order to get to that information. The standard question is, “What was the wage of your highest-paid parent when you were 14?” I cannot answer that. I would love to give that information, but I have no idea. There is recall bias. The very people you probably want to help most might be less willing to give that information, even if they did know it.
Even if every employer in the country was asking those questions, it would have to be voluntary. You are probably going to get white, middle-class people answering it much more readily than people at the bottom of the income scale. Would it solve the policy question? I am not sure.
The Chair: Thank you. This brings us rather miraculously to our final question. After that, I will be asking the panel to say anything that we have failed to ask you, but the last question is from Lord Watts.
Q171 Lord Watts: We are interested to know whether you can make some recommendations that we could use for how the system could be improved, but it seems to me that we have lots of information. What I do not know is whether that is co-ordinated. For example, would it be best to have one department leading on this to identify what needs to be collected, bring all the parties together and identify the gaps, and then use the academics to fill the gaps in?
Dr Emma Gordon: You have just described how our programme is going to work going forward. Until now, we have worked with DWP, HMRC, DfE, DHSC/NHS England and MoJ in—this is probably not the right word—a very transactional way. “Will you work with us to create this dataset? Thank you. Let’s make it available. Here are some results back”.
Moving forward, those five departmental groupings are going to be core partners in ADR England, so that we can have those discussions collectively, at a strategic level, and start thinking about where the most important data gaps to fill in the LEO or ECHILD dataset are. We would not be bombarding one department, usually NHS England, with, “We need access to health data”. It would be part of the conversation and we can agree priority linkages and where to fill those gaps.
It is an iterative process. It will not happen overnight, but we have really strong relationships with these government data-only departments, as well as collaboration with the ONS and our partners in the devolved Governments. I should add that we work right across the UK. We have a really good system and way of working, with those mutual benefits aligned.
Lord Watts: Do we have a lead department and a lead Minister?
Dr Emma Gordon: We do not have a lead department, because by definition it is complex and cross-cutting. I do not think that it could ever be one department’s individual priority to fix this, because it is the deep knowledge of the data and the systems that collect it that you need in order to make sense of it. Jen might say something about the Integrated Data Service. ONS tried to do this, but, if the levers are not there for the benefits to work both ways, it is really difficult. Now, having a Minister would be useful.
Lord Watts: I was not suggesting that anyone was excluded from the process, but that leadership is important in any organisation, and we do not seem to have a lead Minister or department. It is left to the departments to decide what they bring to the party.
Dr Emma Gordon: I would argue that they are all very committed to working with us.
Lord Watts: It was not a criticism.
Jen Woolford: The Statistics and Registration Service Act and the Digital Economy Act open up those gateways with departments for the ONS. That is part of why we have been developing this index and the linkage technology as we onboard datasets to allow us to use them more quickly and easily.
A big thing is that, as the data starts to flow more freely and those pipelines are established, we will be able to bring data in with far less effort, which then frees us up to go and look for more data sources or to think about how we might want to augment others. At the moment, the Integrated Data Service, as Emma said, is a platform for government users to come in and access linked data and to do that work.
Q172 Lord Evans of Rainow: I have two quick points and a question. A Minister in the Cabinet Office might be helpful, who would then have an overarching remit. My question is further to that from Lord Watts. How would artificial intelligence help with this question? Is it an opportunity? Is it a threat? Will it have any impact?
Dr Emma Gordon: That is a very good question. I mentioned that the Digital Economy Act sets a framework for how data will be accessed by researchers, where it is only the results of statistical outputs, and only de-identified data, that researchers can get access to. All these conditions around how a trusted research environment works mean that doing research using this data and using AI methods is very difficult, if not impossible.
There is also the scale of these datasets. It is a terabyte of data if you are just holding it in a form that facilitates statistical analysis. As I understand it, if you were to transform that data so that it could be analysed using AI methods, you would need an infinitely greater amount of scalable compute power.
I can currently fund the ONS to provide a free service for researchers to access to do statistical analyses on this data, and there is an incredibly rich amount of research coming out. I could not do that if it was a platform that had that scale of compute, because it would be so much more expensive to run.
Q173 The Chair: Thank you all very much indeed. In our final moments, are there any recommendations from panel members on data that you would like the committee to consider when it writes its final report, which we are about to do? You are our last set of witnesses apart from Ministers, so we are very grateful for your insights.
Dr Emma Gordon: For me, it would be the removal of the exclusion of health bodies from being able to use the Digital Economy Act to facilitate research access to health data.
The only thing that it would have been really nice to say something about, which there was some reference to in the NEET conversation, is care-experienced children. One of the huge benefits of administrative data is that we now understand a lot more about the outcomes of care-experienced children versus other children. That is a particular group who are very difficult to get through survey datasets, because, almost by definition, they have quite chaotic lives. They change address a lot. The people looking after them change a lot.
Through administrative data, we can follow them. Through the Ministry of Justice Data First programme, working with the Department for Education, where we have linked education outcomes, and the police national computer, you can track the experience of care-experienced children versus their peers. That is the only other thing that I would like to mention.
The Chair: On the reference to crime statistics, the Ministry of Justice suggests that it is a pretty depressing trajectory, from what you are saying.
Dr Emma Gordon: It is incredibly depressing.
Jen Woolford: It might be a bit ambitious, but something resembling a population register would really help us. Building from having a unique number, if we could be assured that people were telling us quickly where they were moving house or leaving the country, that would really help us from a statistical perspective. There are potentially lots of ethical, political and public opinion barriers to that.
Dr Mark Brewin: I would focus less on data, but more on leadership within departments, and functions and teams with expertise in departments, to support and facilitate data sharing. Several studies and reports have been produced over the last three to five years that have looked at how to improve data sharing more generally, not specifically for social mobility. Key to many of those is the right level of resource and prioritisation across departments to support data sharing with the expertise.
Some of that then comes down to prioritisation from government more generally, because these are fairly finite resources with a particular skillset. Ultimately, it will be a decision for government and Ministers about how they prioritise it, but that leadership within a department to facilitate and support the kind of work that Emma is describing is really important.
The Chair: Thank you. We started with you, Professor Friedman. You certainly helped shape this session, and we are very grateful. We are going to let you have the final word. I cannot guarantee that, because some of the committee may want to, but I think you are going to have the final word.
Professor John Friedman: I really appreciate this invitation. I agree with everything that the other panellists said. There was one thing that I wanted to add. What is very important is not only the formal legal restrictions or permissions for what can or cannot happen, but the way in which the administrative and bureaucratic process of implementing those happens.
In the United States and, it sounds like, in the UK, there are a lot of situations where data exist, and there is no law that specifically says that you cannot do something. Maybe you can interpret something to say that you can; maybe you can interpret something to say that you cannot. There is just a fundamental conservatism built into this where, for someone in a deputy Minister’s position, it is just a lot easier to say no. Why take the risk that something is going to go horribly wrong, even if there is a lot possibly to be gained? Having a great study is very helpful for society, but maybe not personally for the bureaucrat who is making the decision.
It is about not just being really clear on what can and cannot happen legally, but setting up an environment where it is more like, “Let’s see if we can do this. Let’s err on the side of sharing data in a way that fits within guide walls”, as opposed to having a system where you have to have 19 people say yes and, if one person says no, and people’s opinions differ, the whole thing can fall apart.
That broader attitude, in addition to what the laws are formally and what data are stored in what places, can have an enormous impact on the practical ability of researchers and of government departments to make progress on these things.
The Chair: Can I repeat the committee’s thanks very much for this very helpful session? I hesitate to put extra work on you, but, if you get back to your offices and you think of something that we have failed to ask you or that you failed to say, which you think is germane and important, please would you let the committee know? Otherwise, that is the end of this session. Thank you.
