Culture, Tourism, Europe and External Affairs Committee

Meeting date: Thursday, September 17, 2020

Official Report 637KB pdf

---

Scotland’s Census

Our next agenda item is an evidence session on Scotland’s census. I welcome to the meeting Paul Lowe, chief executive, National Records of Scotland, and registrar general for Scotland; Pete Whitehouse, director of statistical services, National Records of Scotland; Anne Slater, director of operations, National Records of Scotland; and Jamie MacQueen, who is a lawyer with the Scottish Government. I remind everyone to give broadcasting staff a few seconds to operate your microphones before beginning to ask your question or to provide an answer. I would be grateful if questions and answers could be kept as succinct as possible. Before we move to questions, I invite Paul Lowe to make a brief opening statement of two to three minutes.

Thank you very much, convener. Good morning to you and to the committee members.

Before I get to the substantial part of my introduction, I will take this opportunity to record my thanks to colleagues in NRS for their very significant contributions working with a range of partners during this terrible pandemic. Their commitment to progressing essential work in the Covid response, providing a range of important services and progressing the census has been truly substantial and impressive.

Since colleagues were last in front of the committee, it is fair to say—somewhat mildly—that the world has significantly changed. In March 2020, my organisation was working to deliver the census for March 2021. At that point, our programme was on track but there remained significant work that needed to be done. The census order was in place and we were working hard with the committee and others to progress the regulations. We were learning the lessons from a very successful census rehearsal and moving into the next phase, in which we were about to undertake further significant development work, testing and the onboarding of very substantial field and programme resources.

Our delivery confidence was based on an appreciation of the skills, capacity and commitment of NRS staff and our partners and where we were in delivery of our plans. However, contingency was not limitless and there remained a challenging but achievable amount of work to do. Success was conditional on there being no major and sustained disruption to our work.

The immediate and continuing impact of Covid has, unfortunately, delivered a level of disruption that most of us have not experienced in our lifetimes. It has led NRS to recommend that the best way of securing the long-term value and benefits of the census is to move the census date to March 2022. As the organisation tasked to deliver the census, we have not taken this matter lightly. We have reached that conclusion following detailed impact assessment work, which I have shared details of in our submission to the committee and a summary of which we have provided. The work included an independent assessment of the status of the programme, which by May was reporting that the programme had shifted down to a red status.

10:30  

As the committee is aware, the Scottish ministers received NRS’s recommendations and subsequently informed the Scottish Parliament and the committee of the intention to move the census to March 2022. This is the timescale to which we are now operating. Currently, NRS is in the phase of an intense replanning exercise and we are working with our contractors and partners, with the Northern Ireland Statistics and Research Agency and with the Office for National Statistics to support the delivery of their respective censuses, and with data users to meet their needs.

The impact of Covid has been and continues to be tragic and substantial. My responsibility with the census is to ensure that we gather and provide a census that submits vital and accurate data in a safe, secure and efficient manner. We get only one opportunity to ask the people of Scotland these questions in each cycle and it is important that we do so in a way that allows that to be done effectively and with high quality, not only to meet short-term needs but for many years to come.

In our assessment, the decision to move the census to March 2022 provides the best opportunity to put in place a census of the right quality to deliver for our users. I thank you for this opportunity, and I welcome any questions from the committee.

Thank you, Mr Lowe, and thank you also for your submission to the committee, which is very helpful. You say in your submission that the census delay, as you have said now, is due to the pandemic. However, you also say in the submission that, prior to the pandemic, the census programme board had reported that the delivery confidence assessment had moved from amber-red to amber. You said in your comments that it had moved to red, but maybe you misspoke. Is it correct to say that it moved to amber?

To clarify, we have had a number of independent assessments carried out on the programme. An assessment was undertaken at the end of February or in early March, before the Covid pandemic. That put the programme at amber status, which means that delivery of the programme is feasible. For a programme that is a year or more out to delivery, an amber assessment is relatively common in such reviews. A follow-up review was undertaken towards the end of May, several months on and some time into the pandemic; it suggested that the impacts of the pandemic were more significant and recommended that the programme be put at red status in light of the Covid challenges.

I see. An amber status does not suggest to me that you are on track. In traffic lights, as I recall from my highway code, amber means wait; it does not mean go.

To clarify, there is a set of very specific definitions that is linked to the project review methodology for gateway reviews. I am a gateway reviewer and I review other projects. This is based on what is called a delivery confidence assessment. An amber assessment means that delivery of the programme is seen to be feasible but there are challenges. It is based on assessment of what needs to happen over the period ahead. It would be common for a major programme a year or more before delivery that was in a reasonable position to have an amber or at best an amber-green assessment. If a programme were rated amber-red or red, that would be a more concerning position. We have had a range of independent assurance reviews undertaken in the census programme, and we have a programme of them going ahead. All of those pointed out that significant progress had been made in the programme in the period just before the pandemic and that, at that time, delivery in March 2021 was deemed to be feasible.

I am looking at an Office for Statistics Regulation report from October 2019 that looked at how all the census authorities across the UK were progressing—“Assessment of Compliance with the Code of Practice for Statistics: 2021 Censuses in the UK—Preliminary findings”. That report suggested that NRS had experienced difficulties. For example, in paragraph 4.5 it said that NRS had told the regulator that it was facing challenges that included

“procurement issues, concerns over effective decision making ... and contingency planning arrangements.”

It went on to say that you were putting a new governance structure in place to deal with those things. It acknowledged that you were dealing with those difficulties but it said:

“Nevertheless, there remains a delivery risk for census outputs in Scotland and we welcome the ongoing dialogue with NRS as it continues to manage these risks.”

It strikes me that you had quite significant difficulties already, and the fact that you are now in red, despite having delayed in order to catch up, suggests to me that there are serious problems within NRS.

When I came into the organisation, there were some challenges with the delivery of the census programme, and those were picked up, as you say, in elements of the OSR review in October 2019. I will hand over to Pete Whitehouse in a few minutes, but I will follow up with assessment activity from OSR in the past few weeks, which is much more positive. That red assessment was made in May, before the decision was taken to change the census date, so the delivery confidence assessment of red is based on the status of the programme if there was an expectation to deliver in March 2021, based on Covid. That same report stated that, if decisions were taken to look at the timeline of the census programme—and that was under consideration at the time—the delivery confidence assessment of the programme would be amber or better. The red assessment is based on the challenges of Covid, not just those at that point in the programme but the future likely challenges of Covid for delivery in March 2021; it is not an assessment of the health of the programme if it was being delivered to March 2022.

It might be helpful if Pete Whitehouse said a few words about the OSR review process and the most recent follow-up activity on the report that the convener references.

Good morning, everybody. The OSR is an organisation that works with the statistical census agencies—the ONS, NISRA, the Welsh Government and us—to help us reflect on the progress that we are making but also to ask us questions about how we are meeting our users’ needs. Its focus is very much to ask, “How are you ensuring the quality of the data that you are going to produce? How are you explaining to the public, to those from whom you are going to take data and to those to whom you are looking to provide data where you are in that process? How are you going to deliver the public value that is required?” The conversations largely happen within the organisations. Colleagues from the OSR will come in and speak to colleagues in NRS, as they do with the ONS, NISRA and others. People have an informal and open conversation, because it is a collaborative exercise. It is an opportunity for statisticians in the main to talk to each other about the challenges that they see.

Talking about those challenges is not about saying that we cannot do this. It is much more about saying that we are aware of the challenges ahead of us and of the work that we need to do. What advice can the OSR give us? What advice can we learn from our colleagues across the UK to be transparent and make sure that the information is out there for the public? It is largely within that context. The OSR sets out a series of assessments, which we go through and we publish on our website and make available for others to read and to come back to us on.

We are continually working with the OSR. We have had conversations very recently about the work that we are doing to make sure that our methodology is in the public domain, that our rationale for the work that we are planning for the next year into March 2022 is clear and understood and that all that information is there. I would characterise the OSR as very much focusing on whether we are doing what we need to do to let people know where we are with our programme delivery and how we are learning from each other to ensure that we do is to high quality.

Thanks very much for that. That is all very well, and I understand that it is a collaborative process but, nevertheless, the report said that there was a delivery risk, and Paul Lowe has acknowledged that there were some problems, as were outlined in that October report, and that he had been tackling them.

I appreciate that the decision to cancel the census next year and move it back a year was a difficult one. However, from the paper that we have received, it does not look as though everybody else has taken that decision—that is, the rest of the UK will proceed with the census. You gave examples from Australia, Canada and the US where, although it is challenging to deliver, there are still attempts to go ahead with the census. You have concerns about the response rate, which is moving from a predicted 95 per cent down to a 60 per cent or 80 per cent response rate under the various scenarios that you looked at for possible delivery. I would like to understand why those estimates are so low. Given that the ONS and the rest of the UK are going to proceed with the census, I assume that they do not feel that they will have the same drop in response rate. Why is the response rate in Scotland predicted to be so low?

What is happening internationally with censuses is that the Republic of Ireland has announced this week that it is delaying its 2021 census to April 2022, and we understand that India is also contemplating doing that. The impacts of Covid on the censuses in Australia and New Zealand are still under contemplation but they have announced that the public rehearsal in Australia and major public testing in New Zealand are being delayed. In North America, the US census was live when the pandemic hit—their reference date was 1 April. The implications of Covid for the US census are still to be evaluated, but they have more than doubled their field force collection. They will have field force agents out in the US until the end of October—for twice the length of their anticipated period—and concerns have been expressed about the response rates that they have seen.

The position of NISRA and the ONS is that they are continuing to work towards March 2021. I cannot speak in detail about their internal decision making around why they are proceeding on that route. That is, obviously, a matter for them and for the respective Governments that are overseeing the census in those parts of the United Kingdom. There are differences in the design—

I am sorry to interrupt, but you are giving quite a complicated explanation for why the ONS is proceeding, which is to do with its ability to mitigate and the different ways in which it collects data. I find that quite difficult to follow. Is it because of the size of the organisation or the set-up? I am trying to work out why we are different in Scotland.

There are differences in the design of the censuses in different parts of the UK, and certainly in Scotland. It might be helpful to bring in Pete Whitehouse, our chief statistician for NRS, to answer some of the more technical questions.

There is a concern about the impact of Covid on response rates across the whole of the UK. The ONS, independently of the census, has been working for some years to develop what is called administrative data, which is data sources that exist in other parts of the Government and public sector. If there are low response rates from the census, it hopes to access and use that administrative data to see whether it can fill gaps. The ONS is a hugely capable organisation and it has the ability to progress solutions like that in a way that many parts of the world cannot, but those are experimental and they also involve data that is not available or legally accessible in Scotland, so we do not have access to that.

Part of the equation is that there is a recognition that response rates are a concern and an issue for everybody, but the ONS has access to solutions that we do not have in Scotland to compensate for any gap that might arise. If we proceeded and had low response rates, we would not have a way of accessing data from other places to fill gaps in the same way. That is the risk that we would face in Scotland. Pete Whitehouse might want to add to that.

10:45  

We start in a place where there is concern across all census organisations about how to conduct a census in the field in these times. The ONS is a significant organisation that runs social surveys and economic surveys, does a lot work on statistical methodology, has a lot of links into academia and has a lot of IT knowledge and scope. We and the ONS were both ready in February and March 2020 to run our respective census designs and to put our censuses in place. As Paul Lowe said, we still had a lot of work to do but our assessments were, as were the ONS’s, that these things were manageable. Our census design was suitable and appropriate for delivering what we were intending to do.

Covid came in and hit our ability to do that. Our considered view was that, if we went in March 2021, we would be likely to get a response rate of somewhere between 60 and 80 per cent. The impact of that would be that we would be likely to miss lots of vulnerable individuals—

I am sorry to interrupt again—and, convener, I will take this as my second question—but I still do not understand why the response rate would go from 95 per cent down to 60 or 80 per cent. Is it because there would not be a field force? You have told us in previous meetings that we are moving to an online method because that is how everybody works now and that would improve response rates from the previous census, in 2011, which used a paper-based method. Why would there be such a big drop?

The ambition for a successful census is somewhere around the 90 per cent response rate, whether that is online or on paper. In 2011, we had a 94 per cent response rate, so that gives us what we are aiming for. Through our work in looking at options, given the challenges that we were facing with this Covid pandemic, putting in place a census in March 2021 would have meant running it without a field force. The field force is the people who are out in the streets, helping people to fill in the census and knocking on doors. That helps with uptake but it also helps people who might need support with the paper forms. If we do not have that in place, one of the options is to take an online-only approach. However, that would mean that we would lose the digitally excluded people or people who do not feel able to do that. If we took a completely paper-based approach, there would be issues with what that would do to the response rate. Some people might get the form and decide not to respond, so we need a field force to help with that. If we use both of those approaches—online and paper—we will still miss a good number of people.

One of the points of the census, and why it is so important, is that, as Paul Lowe said, it allows us, once in a cycle, to get information at a very detailed local level about all our people in Scotland. It is not like a survey whereby we select 6,000 households or whatever. The census is about trying to get the really detailed information that we require at a local level, so our caution has to be around how we put a census in place that allows us to get as close as possible to a good mid-90s response. Our conclusion, through our work, was that, if we had a census in March 2021 but did not have a field force, we would miss 20 or 30 per cent of our population. It is not that we would miss the average; we would miss the very people for whom the census is the most important tool in gathering that information. That takes us to our decision. We said that a census in March 2021 would not deliver the necessary quality.

As Paul Lowe said, the ONS has different options in front of it partly because of the work that it has been doing for a number of years to gather information from different bits of the public sector, local authorities and other big GB and UK organisations. That was not part of its census design but, because it was doing that work as part of one of its major strands, it was able to use that to give itself some confidence that, if response rates in England and Wales hit 80 per cent, it could understand who was missing and could use that evidence to improve its statistical estimation.

That is the difference. We do not have that information, so we would be left with a response rate that was low and biased and we would not be able to have confidence that we could produce the outputs that our users need. The ONS’s view is that it will work very hard—and it will—to make sure that response rates are as high possible but that, if they turn out to be low or problematic, it has other data and statistical methods that it will apply. What it has also talked about is that there will need to be some consideration of the precision of those estimates and the timescale that it takes to deliver them.

I will stop you there because we have to bring in other members of the committee. I realise that this is a technical subject, but I ask for answers to be kept as succinct as possible.

I will try to keep this brief. What is the estimated cost of postponing the census?

We have decided to reschedule the census based on achieving best value and value for money. We are currently doing detailed replanning of the census, which will include the additional cost of delivering the census. I am happy to write to the committee setting out the outcome in more detail once we have concluded that exercise, which will be soon.

It is not a comparison between the original situation and the new census date. It is likely that any census organisation would, in attempting to deliver a census in 2021, incur Covid-related costs in delivering the census according to plan. The question is not as binary as being about the original plan versus rescheduling. We will be happy to provide that information to the committee.

Thank you. That would be helpful.

I also want to touch on the public policy implications of census data being gathered in Scotland in a different year from its being gathered in the rest of the UK. What will be the data implications and wider public policy implications of the two systems running out of sync?

I will say a little bit, then Pete Whitehouse can say a bit more from a statistical perspective. All the bodies that take censuses in the UK are working very closely together on this. Our collective view is that the issues are surmountable; that is certainly the view of the national statistician. We agree that it is not an ideal situation, but there are existing methodologies and approaches that we can use. We produce lots of data based on the census in years when the census is not run; there are ways of bridging data gaps using existing census data and other surveys and methodological approaches.

One of the key things is how we deal with population estimates. The census is one of the key tools that we use every 10 years or so to allow us to add births, subtract deaths and add migration. That approach produces our population estimates, which go into helping with allocation formulas. We also get information on rates from lots other statistics—most recently, Covid and mortality rates, for example.

We are very conscious of the need to ensure that the information is provided. We have experience—this happened after 2011—of censuses delivering their population estimates at different times. At that time, the nations came together to agree that we could produce a UK population estimate based on census data and rolled forward census data from a previous period. People run their allocation formulas and look to readjust them when the new census data comes on stream.

We are very conscious that users such as equality groups and particular interest groups will be looking for specific data. We will have conversations and engagement with them and we will ensure that we make it very clear on our website how we will help them to get information as we run through 2022, 2023, 2024 and onwards. We are very clear that that is an important component, so we will be putting relevant work in place. The aforementioned Office for Statistics Regulation will help us to ensure that we speak to and engage fully with all UK users of census data, as well as those in Scotland.

A number of groups that are already in place across the Governments’ statistical services—the ONS, the Welsh Government, NISRA and ourselves—will be looking at the matter. That is an issue for my directorate in NRS, because we produce the census, to a degree, but are also a customer of the census because we are responsible for population estimates for Scotland. The issue is very close to us and will help us to support data users across Scotland and the UK.

On public policy and finance policy, what statistics will be used, with the rest of the UK, for the purposes of the block grant adjustment?

After the 2011 census there was a nine-month gap between Scotland census population estimates and those of the rest of the UK. We have population estimates. In 2011 England, Wales and Northern Ireland had their 2011 census data. We were using our 2001 data, which we would be rolling forward. Each year we add births, take off deaths and add migration information, which gives population estimates. We use the census to calibrate that and to see whether we need to make adjustments. That is the process that will happen. We will have those conversations. We managed to do it successfully in 2012 and we will do so again.

There is work going on across the UK on our migration statistics and how we use administrative data from different sources, such as national health service data, to inform that. The quality of the estimates should improve.

Good morning. First, I want to pick up on another aspect of the delay to the census, which is that it will reduce the time between this census and the following one from 10 years to nine years. What planning has been done to squeeze 10 years’ work into nine years? What are the implications of that?

The census programme is approximately six years in delivery. It does not take the full 10 years, so there is scope within delivery to compensate for the delay. The dates for when future censuses are set are matters for the Parliament and the Scottish Government. NRS will produce recommendations after the census in 2022, but it will be for ministers and Parliament to fix the date of the next census. NRS will progress with delivery against that.

I suppose that that opens up a question: if the census takes you only six years, what do you do for the other four?

NRS undertakes a broad range of work outside the census. I am happy to go into that in more detail, but we gear up our staffing and resources in step with our delivery of the census programme. A number of people come into the organisation to deliver the programme, then leave at its conclusion so that we are not sitting with a high level of staff for the full 10 years.

My final question will be short. We will be a year out of step, and you have discussed some of the issues around that. What public policy areas will be affected by the delay? Will it affect health service planning, infrastructure decisions and so on? In other words, it is not just an issue for NRS; it is an issue for all public bodies that depend on the census.

11:00  

We know who our data users are and we engage with them very closely. To come back to an earlier answer, I repeat that we produce, outwith the census, a range of statistics and information that are used by stakeholders. Each year we produce mid-year population estimates, for example, which are used by a range of stakeholders. As Pete Whitehouse explained, because we are also responsible for the registration system, we have very good figures on births and deaths in Scotland.

The other factor through which population changes is migration. There is an existing UK model for migration statistics, but, outwith census consideration, the ONS, NISRA and NRS are working to produce a more robust migration model. We produce lots of statistics and information that helps them, and we will work closely with them to make sure that the issue is mitigated.

I would like to go back a step to before Covid—if any of us can remember what life was like then. On delivery of the census in March 2021, answers to the convener’s line of questioning seemed to indicate that there was a problem. Did you consider additional resourcing or capacity to deliver in March 2021? Was that an option? Did you flag that up with the Scottish ministers?

I will clarify the initial point. There were historical challenges in the census programme. When I started at Christmas, in December 2018, a number of assurance reports indicated that there were challenges in the programme. However, since then a number of mitigations have been put in place to improve the programme. I reconstituted the programme board, we brought in additional resources and changed governance of the programme. A number of assurance reviews came out late in 2019 and into 2020 that indicated that substantial improvements in the programme had been achieved. The independent assurance reports that I received in February and March indicated that, before the Covid pandemic, delivery of the programme was perfectly feasible for March 2021.

As part of the options appraisal exercise, we looked hard at the option of still delivering the census in March 2021. First, we looked at our original plan and considered whether we could deliver the census with all the planned shape and colour according to that timescale and whether there were mitigations.

Additional resourcing was an option that we looked at, but I am afraid that there are some situations in which additional people cannot buy you out of trouble. For example, it is really important to robustly test the census. A lot of our testing happens in the year in the run-up to the census going live. It is important not only that all the IT solutions and so on work as expected, but that the data in them is protected from cyberattack and other things. The last Australian census, for example, was a digital census, and it was closed for the first three days because of a cyberattack. The issues that we were facing included truncation of the testing window, which started to feel very risky in the light of the census being digital. We looked at such things as part of our consideration of whether we could hold to the March 2021 date.

That is helpful.

How confident are you that you will not be back at the committee some time next year saying that you cannot make the new date?

I am very confident of that. We are delivering what we planned to do in March with contingency time. The scenario that presents a real risk is our being unable to find a solution to Covid. That is less likely now, based on our knowing more about the virus, about how we manage the issues and about how we will manage the census even if Covid is still around. I do not have concern about that; the assurance reviews that I am having done and the independent advice that I get all indicate that the programme is in a good place for delivery in March 2022.

I asked the question because, as we established at the outset of this session, NRS was already at amber before Covid. I think that you said in response to an earlier question that the detailed replanning exercise is not yet complete. If so, how can you conclude with reasonable confidence that you will meet the next deadline?

I will briefly clarify the point about amber status. Amber status means that a programme’s delivery is feasible. A programme that is a year or more out from delivery is unlikely to have green status because there is a considerable amount of work still to deliver. At the point when assessment was undertaken, amber status was a reasonable assessment. For context, I point out that the ONS programme was also at amber at that time. That status is not an unusual or negative critique of the health of the Scottish census programme at the point before Covid arose; in fact, it is quite the opposite. It is important to clarify that.

The question about replanning is legitimate. We had a plan before Covid for how we would deliver the programme. We knew all the components, their relationships, when we needed them and how we were going to test them. The replanning work is about how we move that plan to a later delivery date, rather than do it according to a shorter delivery timescale. Before Covid we had confidence that we had a plan for delivery in March 2021. It is not the case that there were unknowns in that plan; it is about how we match that plan with an elongated period for delivery. The risk in that is less than it would be in a scenario in which we try to do it according to a more truncated timescale.

Thank you for that answer. Time will tell. It would be useful if NRS could keep the committee updated on its progress on that. Thank you.

We are very happy to do that.

Thank you. That would be very helpful.

You say in your report that you need at least a 90 per cent response rate for a census to be valid. The rate was 94 per cent in 2011, which means that there were still 330,000 people in Scotland who did not respond. For the census in 2022, is there any focus on areas from which there was a particularly low return. Were there specific areas where the rate fell below 90 per cent? What steps are being taken to ensure that we get a higher return next time, given that now we are talking about an 18-month preparation time?

That is a very important question. In a second, I will hand over to Pete Whitehouse, who can talk a bit about how we adjust statistically for areas where we do not have responses. The issue that Kenneth Gibson has articulated is partly at the heart of why we recommended delaying the census. Our trying to deliver it in March 2021 risked there being a higher number of people not responding.

Our work for the census includes community engagement work. We know from previous censuses the populations that are more difficult to get results out of; they vary among countries and over time, but we know them. We built into the census a very active community engagement programme in which we would engage with and go out to speak to different groups in order to drive up awareness and participation in the census. That is important activity before the census goes live, but because of the restrictions that have been caused by the pandemic, we have not been able to do that. That was partly why we recommended delaying the census.

Active community engagement and communications and an education programme are important in targeting those groups, because they are the people about whom we need to know most. There are things that we can do to compensate where there are relatively small numbers. Pete, do you want to say something about that?

We have a number of approaches. Some are based on understanding of how we use our field force—our support to people as the survey takes place. We know the areas that have traditionally had poorer response rates, and we know where we need to place our additional support in order to help people in those areas to respond. We have thresholds; we expect response rates at local authority levels and so on that give us sufficient coverage in a general sense, but we have also to ensure that we do not miss key parts of the community. We have those kinds of data; we learn from colleagues across the UK and from our own experience.

We do other things that are about using other data. It is not well known that after the census we run another survey called a census coverage survey, which helps us to understand whether we have missed particular areas or bits of areas, and to understand from a statistical perspective how we can adjust census data.

We also use other data that we can get hold of to quality assure the estimates that we get from our census. It is, very much, a statistical process that engages with people through the media and communications with their communities in order to help them to fill in the forms. That is aligned with our understanding of where people need more support and with a management information tool that tells us how the census is actually running so that we can reallocate field force support.

A number of statistical approaches help us to use other data, to look at where we are missing information and to provide better-quality estimates, so that we produce good-quality output based on all that.

From a stats perspective, it is key that we continue to learn, including from our colleagues, and that we provide what we learn to the UK and internationally, because we all have to deal with the issue.

Thank you very much for those two comprehensive answers, which are very helpful. I am wondering about the level of variance. Both witnesses have said that there are certain communities in which, traditionally, there has been a lower return from censuses. If we are talking about 90 per cent being the accepted level at which a census is valid, are there many such areas, and what is the variance? Is it 80 to 99 per cent? Is it 92 to 96 per cent? What is the level of variance in Scotland?

Even with all the community outreach and so on, when do you consider the return level to be such that the information from an area is not valid? You can make statistical projections and so on, but, given that we are trying to base a number of public services on accurate information, when does the information that you receive go below the level—it may be 90 per cent, as you say in your report—at which you can consider it to be robust?

I do not have all those numbers here, but I am very happy to provide a paper to the committee on our thresholds. We have expectations of response levels at, say, local authority level, which we have set out in our quality assurance document—they are published—which show the response we anticipate and need to have.

11:15  

When we talk about areas or individual communities in which response rates are more challenging, it is not that we accept that; it is more that we are saying that we know we need to focus more on those areas through our community work and media, to encourage and engage with people to make sure that the census returns are made.

We explain to our users the quality of the estimates that we produce—we do that with the Office for Statistics Regulation and other UK census organisations. When we have gathered the data, we assure it to the levels we do and, when we understand that there is missing information, we produce estimates that say it is within plus or minus 0.5 per cent, for example. There will be a presentation of the accuracy of the information—that is what we will do, as the ONS, NISRA and all other bodies do.

I am very happy to provide a paper that talks about how we are ensuring that we get the coverage and the completion rates that we need and what we do to make sure that all the data users are clear about the precision and accuracy of the data.

I have one more question.

It will have to be a very succinct question and answer, please.

Sorry, convener.

When you talk about individual communities, what scale of community are you talking about? Is it 500 houses or 1,000 houses? Is it a particular town? What do you define as a community in trying to put together the statistical analysis?

I am thinking of communities very broadly as groups in the population across the whole of Scotland who have some shared characteristics that mean that they are unable to respond very well. There may also be some local areas where there are aspects of, say, digital exclusion or other things. We have to think about all of this in the round. I am happy to provide clearer detail in a paper if that would be helpful.

It would be. Thanks very much. I appreciate that.

Our last questioner will be Oliver Mundell.

Has any additional resource been requested from the Scottish Government and the Scottish ministers?

We are in regular contact and engagement with the Scottish Government about the resources that are needed to deliver the programme. As I said earlier, in the presentation, it is likely that there will be some additional costs and resource requirements in running the census for an additional year. We are concluding our work on that through our replanning exercise, and we are happy to write to the committee to set that out in more detail.

So, no specific request has been made so far?

Sorry, can you clarify, in respect of—

For additional funds. You have not asked for any money already.

Not yet, no. That will come after we have concluded the replanning exercise. That will then feed into a resource budget request for the census for the revised date.

Thanks. I have one other substantive question. Were the Scottish ministers relaxed about the proposal to delay or did they push back at all?

Obviously, I cannot comment on the detail of ministerial thinking. What I can say is that we were asked for a range of additional information, as well as being asked questions on our recommendation. The conversations went on over a number of weeks. Some of the questions that the committee has asked today are the sorts of questions that ministers have asked of us about the firmness of our assessment of the need to shift the date of the census, about whether any other mitigations could be taken, and about exploration of the options that are set out in the paper that we have shared. There were a range of issues.

I do not think that anyone—certainly not me—has taken the prospect of shifting the date of the census lightly. I did not join NRS for this outcome. It has arrived as a result of Covid, unfortunately.

We have a little bit of time in hand. I will ask a couple of very brief supplementary questions, and I ask you to keep your answers brief. Mr Lowe, when Annabelle Ewing asked you for an assurance that the census would be delivered in 2022, you said that the only factor that might militate against that was our not having solved Covid by then. However, a lot of international experts are saying that we may not have a vaccine by then. I take it that you are not planning around a vaccine or elimination of Covid, are you? We all hope that it will be eliminated, but I presume that we have to plan for a continuation, local lockdowns and everything else that we are seeing. Are you planning for all of that?

I am sorry if I articulated the sentiment not very elegantly. There is a lot more understood about this virus and a lot more understood about how we would operate a census if Covid still remains—and our assumption is that that is a distinct possibility. We will plan to use mitigations if Covid is still endemic in the population in March 2022. There is hope that it might not be around then, but, if it is around, we will take steps to manage that.

That is great. Thanks very much.

I also want to raise a slightly different issue, which relates to some of the committee’s earlier work. You may have noticed on the committee’s website some correspondence, in the past couple of months, between the committee and Professors Lindsay Paterson and Susan McVie, who are very distinguished statisticians at the University of Edinburgh. Professors Paterson and McVie have been doing some work with ONS on the guidance for the sex question in the ONS census that is going ahead next year. In the course of that work, they wrote to you about this, and you wrote back to them, stating that

“no further revisions of the Scottish guidance could take place because the CTEEA committee had approved the guidance.”

You said:

“As set out in Ms Hyslop’s letter to the CTEEA … the agreement by the Scottish Parliament to the Census Order … confirms that the sex question and associated guidance has been agreed.”

As I clarified in a letter to Professors Paterson and McVie, that is not actually the case. The committee recommended, in its report dated 4 March, that the Scottish Parliament approve the draft census order, but the committee has no locus with regard to the guidance. In fact, the cabinet secretary clarified this to the committee. In her oral evidence on 30 January 2020, she said:

“The guidance is completely separate from the legal processes that we are considering.”

She continued:

“The committee’s legal responsibilities relate to the order, which sets out the subjects to be included, and then the regulations, which will set out the questions. ... I know that the committee has become heavily involved in the guidance issue, but its legal responsibility relates to the order.”—[Official Report, Culture, Tourism, Europe and External Affairs Committee, 30 January 2020; c 21, 35.]

I have written back to Professors Paterson and McVie, pointing that out.

Since you are here, would you like to take the opportunity to clarify that what you told them was not the case and that the committee does not have the ability to approve or disapprove the guidance?

Thank you, convener. I apologise for that error in my original letter. I can confirm that I have written to the professors in the past couple of days, to confirm and clarify that position.

Thank you very much for that. Claire Baker has a supplementary question.

On the delay to the census, have you had an opportunity to speak to organisations that have an interest in the census—for example, the veterans community, which ran the count them in campaign? I understand that the collection of that information could be important to them in terms of the planning of services and that the delay will have an impact on some of that work. Were you able to have discussions with some of the stakeholders?

Yes, absolutely. We have started some of those conversations already. I do not know whether Pete Whitehouse wants to say anything in more detail, but I know that we have been engaging with a number of veterans organisations on the issue.

When the decision was made, we wrote to a number of the stakeholders that we know have very specific interests in the census, as well as putting general information out through the website. The veterans are one of the groups that we are keen to engage with. The Office for Statistics Regulation regularly talks to us about how we are engaging with the users who were expecting data to come through in March 2022 that is now going to come through in March 2023. It is absolutely part of our work as we go forward. We will be talking with them, and we are very happy to engage with all our stakeholders if they have any questions about what the delay means for how they use the data as it comes through at a Scottish level and at a UK level.

Thank you very much. No members are indicating that they want to ask supplementary questions, so we have completed our evidence session a few minutes ahead of time, which I am really pleased about, given that we were running a little bit late and it is quite a technical topic. I thank Mr Lowe, Mr Whitehouse, Ms Slater and Mr MacQueen for their evidence today. The committee will consider the evidence that it has heard in private shortly.

That concludes the public part of the meeting. I will allow a couple of minutes to enable members to have a comfort break before we resume in private session.

11:27 Meeting continued in private until 12:07.