Public Administration and Public Affairs
Oral evidence: Civil Service People Survey, HC 575
Thursday 20 October 2022
Ordered by the House of Commons to be published on 20 October 2022.
Members present: Mr David Jones (in the Chair); Ronnie Cowan; John McDonnell; Tom Randall; Lloyd Russell-Moyle.
Questions 1 - 37
Witnesses
I: Sarah Henry, Director for Methodology and Quality, Office for National Statistics; and Peter Lynn, Director of the Institute for Social and Economic Research, University of Essex.
Written evidence from witnesses:
– Office for National Statistics
Examination of witnesses
Witnesses: Sarah Henry Director and Peter Lynn.
Q1 Chair: Good afternoon. Welcome to this session of the Public Administration and Constitutional Affairs Committee. Today the Committee will be holding its first evidence session in our inquiry into the Civil Service People Survey. This introductory session will focus on survey design and methodology, looking at best practice for ensuring the accuracy and validity of the results in large-scale surveys such as this, as well as ensuring the anonymity and confidentiality of respondents. We are joined this afternoon by Sarah Henry, who is the Director of Methodology and Quality at the Office for National Statistics and Peter Lynn, Director, Institute for Social and Economic Research at the University of Essex. Good afternoon to you both. Thank you very much for coming here today. Perhaps you would introduce yourselves formally for the record.
Sarah Henry: I am the Director of Methodology and Quality at the Office for National Statistics.
Peter Lynn: I am Professor of Survey Methodology and Director of the Institute for Social and Economic Research at the University of Essex based in Colchester.
Q2 Chair: We understand that the Civil Service People Survey is conducted as a census-style survey. What are the strengths and weaknesses of using this particular approach for a large-scale staff satisfaction survey rather than a sample-based survey?
Sarah Henry: The main advantage is that you can let everybody have an equal opportunity to participate. Anybody who wants to participate can simply log in or get a paper survey, whatever the methodology is, and share their views.
You can also make it quite cost-efficient because you can access everybody in the same way. It is open to anybody to respond. If you are going to do a sample, you need to plan how you are going to do that to make sure that the sample is representative. In both cases, you need to put some effort in to maximise the response rate. You are either encouraging everybody to respond or you are encouraging the people who you have sampled to respond. Particularly with a sample, it is very important that you maximise the response rate.
However, if you are doing a sample, you can argue that you can target your sample in a way that directly addresses your area of concern to make sure that you have the right representation and some will argue that it is easier to manage that sample.
Those are the broad benefits and disadvantages.
Q3 Chair: Do you think that the census-style survey is appropriate for this exercise?
Sarah Henry: Yes. For a staff survey where you want to give everybody an opportunity to participate it is a very valid method.
Peter Lynn: I agree that the main advantage of the census approach is avoiding the need to design and select a sample, which is an aspect of survey design that could potentially go wrong, and could be open to criticism in terms of whether the sample is fully representative. You avoid that issue.
On the other hand, the sample-based approach has the advantage, first reducing the overall burden on people taking part—fewer people have to spend time taking part—and potentially reducing the costs—fewer data to handle, and so on.
In most other aspects the two approaches are broadly similar. They involve the same tasks and have the same risks, concerns and possible outputs.
One of the main deciding factors in whether to go for one approach or another in a survey of this kind is simply to do with the sample size that you would obtain. You need the sample size to be big enough to allow you to produce precise estimates. If the only way for it to be big enough is to carry out a census or something close to a census, you might as well carry out a census.
Q4 Chair: I understand that the people survey is about 130 questions long. What would you say, Mr Lynn, is a suitable length for a questionnaire such as this to achieve an optimal level of engagement from the target audience?
Peter Lynn: What we know broadly from survey methodology and the survey methodology field is that the length of the questionnaire is not something that participation is hugely sensitive to. It is far more important that the survey asks questions that are perceived to be sensible, relevant and pertinent to the stated aims of the survey. If the questionnaire should be 10 minutes longer or 10 minutes shorter does not tend to make a huge difference to how likely people are to participate.
Sarah Henry: I agree. The real question is the motivation of the respondent to participate. Some will argue that a shorter questionnaire would attract more people to participate. Sometimes you see that people start completing a questionnaire and abandon it at a certain point and then you need to deal with a decision about what you are going to do, whether you are going to take the responses given or ignore the whole questionnaire. You need to take that into consideration but I agree that the motivation for participation is the key factor and the length of the questionnaire and the time it takes is a secondary question; they are two slightly different questions.
Q5 Chair: I understand that this particular questionnaire takes on average about 20 minutes to complete, which seems quite a long time. Do you think that is an acceptable length of time for surveys such as this?
Sarah Henry: I am not sure if it might take a little bit longer than 20 minutes. It depends as a respondent on how much attention you pay to each question and how much thought you give to each of the responses and whether you are filling in the open-ended questions and so on. You will get people who complete it in 10 minutes; others will take a lot longer. I think 20 minutes is an acceptable time for members of staff to give feedback to their employers. If you look at it from that perspective, the length of time is acceptable. If you are taking the question as a stand-alone for all surveys—how long would you expect to spend on a survey—there is some evidence that 20 minutes is about the maximum span for a respondent.
Q6 Chair: Do you have anything to add to that, Mr Lynn?
Peter Lynn: Simply that this question relates also to the motivation for taking part. If a survey is seen to be important and serious and has useful benefits, people are quite willing to spend time, to spend longer than that if necessary, taking part. There are many surveys with an average length of considerably longer than 20 minutes and people take part.
The opposite is also true. If the survey lasts three minutes, the respondents can perceive it to be not sufficiently serious and to be not worth taking part in because it is not asking about the things it should be asking.
Q7 Ronnie Cowan: If you are not creating a representative sample, how do you know that those people who volunteer, that cohort, are representative of their organisation? It could be just the people who want to keep their bosses happy and say nice things; it could just be the people who want to have a moan and say unkind things.
Sarah Henry: That is true. The question is what you mean by representativeness. The only way you can draw a sample that you think is representative at the moment is based on the demographics. The civil service does have demographic information about its employees but the question is a little bit different—how do you know if the sample that you should be drawing should look at representation on a completely different set of criteria? For example, to follow on directly from your question, how do you make sure that you have a sample that is equally representative of people who are happy and people who are unhappy? The decision was made in the civil service to make very clear that the responses are from people who responded to the survey rather than pretend they represent the entire civil service. There is an important nuance there. That is a good way of addressing the difficulty here because the question is how would you weight it and how would you sample something representative?
Ronnie Cowan: That was my next question.
Q8 John McDonnell: Has it always been a census approach?
Sarah Henry: It is not my survey but as far as I know from my experience with it.
Q9 John McDonnell: On the qualitative side, you have the free write-in element but you do not publish that?
Sarah Henry: As far as I know, the free write-ins are not published but are shared without being linked back to the respondent. It is very important to recognise that it is just a list of sentences, which is shared, as far as I know, with the relevant Government Department. You might want to confirm that with the survey owners.
Q10 John McDonnell: Are any other qualitative measures used at all?
Peter Lynn: Not that I am aware of. I do not think either of us is deeply familiar with this survey. Neither of us is involved in running it.
Sarah Henry: There is an element of qualitative work prior to the survey being agreed. In the engagement exercise, they seek quite a lot of feedback—and it tends to be qualitative—from different Government Departments to help design the survey in the first place, particularly the questions in the survey.
Q11 John McDonnell: I would be interested in knowing a bit more about the qualitative work that goes on in the preparation of the survey.
Can I come on to some questions about the survey delivery? There are three questions. Participation in the people survey is optional but managers within departments are responsible for encouraging employees in their area to complete the survey. This is a difficult one. Is it appropriate for managers to be given that role of potentially pressurising participants?
Peter Lynn: It is a difficult one. I think usual practice on most surveys is to try to ensure that every sample member receives the same information about the survey, the same encouragement, a kind of standardised invitation and reminders and so on. This clearly departs from that in some respects, but not in all respects. One could argue that that might result in the nature of the response or the type of people who respond, differing, between departments or depending on who their manager is and what kind of messages are sent out.
John McDonnell: Would you care to comment on that, Sarah?
Sarah Henry: Just to say that I understood the word “encouragement” slightly differently and maybe a bit more positively. You get an e-mail telling you about the people survey and in it you are encouraged to participate; it is not as coercive as that might sound.
Q12 John McDonnell: What is the best practice for encouraging participation? What would you envisage as best practice?
Sarah Henry: Certainly making everybody aware, as Peter Lynn said, in the same way, and inviting people in a positive way to participate in the survey. That is good practice. Anything that is threatening in any way, if I can use that word, would be very bad practice and would influence the response in a way that is probably biased and therefore would make the results less useful.
John McDonnell: The surgeon with the knife before they operate.
Peter Lynn: Best practice can also involve ensuring that you communicate different types of messages, different reasons for taking part, because different reasons will appeal to different potential respondents. Some people will respond to an altruistic motivation, and others will respond to a more egotistical motivation, so best practice usually involves mixing those different kinds of reasons, possibly in a combination of an initial approach and then a reminder letter of some kind and another reminder with slightly different messages at each stage.
Q13 John McDonnell: Just to give people more assurance, is there any engagement, for example, with the trade unions or the other professional bodies representing the staff to encourage them?
Peter Lynn: I don’t know whether there is but that type of engagement is typically seen as good practice at the survey design stage to make sure that all the necessary concerns or issues are recognised when the survey and the invitations are being designed.
Q14 John McDonnell: Do we know if that happens?
Sarah Henry: I thought that it did but it is worth confirming when the survey ends.
Q15 John McDonnell: That leads to my other questions. Do you think it would be appropriate for the staff survey to be made compulsory rather than voluntary?
Sarah Henry: No. Even if you did manage to get everybody to respond, the quality of the responses would not be worth the effort, for sure.
John McDonnell: The write-in element might be interesting.
Sarah Henry: It would be.
Peter Lynn: Agreed.
John McDonnell: Agreed. Yes. Thanks.
Q16 Ronnie Cowan: I am of the opinion—experts are not—that if people are guaranteed their anonymity, their responses maybe be more honest. You do not want to be critical of your line manager if they are going to see the responses and know who you are. It is common sense that you would last a very short time in your job, which is uncommon in some places; but more common in this place.
If participating organisations could review their people survey results by team—and teams can have fewer than 10 respondents in them—is there not a danger that people can be identified by their answers?
Peter Lynn: You are absolutely right that both the requirement to keep responses anonymous and the perception of that process by the respondents are important to engage people to respond. Indeed, there may be a risk of responses being identifiable if responses are released based on very small numbers of people to people who know who those people are, such as managers. My understanding is that even with this survey and surveys like it, that risk is still pretty minimal because they do not release the open text answers where somebody might have written in somebody’s name or some particular event; it is just the data, just the quantified data, that is released at a team level.
Q17 Ronnie Cowan: Taking that one stage further, what would you consider to be the minimal acceptable sample size to share with an employer or a manager?
Peter Lynn: I am not sure that one can generalise because it does depend, crucially, on the nature of the data that is being released. If you are releasing quite fine-grained data where you have the response to a particular question broken down by age and gender and years in the job and there is only one person in that team who falls into one particular category, that becomes identifiable. But if you are not providing that kind of breakdown, it is not necessarily identifiable.
Sarah Henry: I agree with that. If you are releasing data that cannot easily be cross-referenced to other information, that also helps with anonymity.
Q18 John McDonnell: When you are talking about cross-referencing, what actions can you reasonably take to make sure that your responses are reliable? If responses are going to be exposed to the line manager, how can you nail that down? Is cross-referencing part of that? I have seen questions where I think, “I have answered that question already” but it takes someone very clever to cross-reference further down,
Sarah Henry: If you are releasing aggregate results and there is no access to individual-level records, first of all you can do more disclosure control on those aggregate results if you think there is a risk—you can swap records and there are other methods you can use to guarantee more anonymity—but generally, summary results are hard to cross reference on an individual basis to other information, particularly if you are asking very general questions. If you start asking very detailed questions that contribute to identifying people, it becomes easier to re-identify people. But the people survey questions do not really fall into that category. The demographics are quite usual types of questions.
Q19 Tom Randall: The 2021 people survey had an overall response rate of 62% but there was quite a variation within that rate. The Scottish Prison Service had a response rate of 21% and the rate went as high as 97% for the National Infrastructure Commission.
Do you consider there to be a minimum acceptable response rate for a census-style survey for these sorts of results to be considered sufficiently reliable?
Peter Lynn: There isn’t really. It is a continuum. More is better. Less is not necessarily worse, though it can be. The concern here is not the response rate. The concern is whether the respondents are similar to the non-respondents and therefore representative of the totality. That is what we want, the responding sample to be representative. If response was something that happened at random, it would not matter whether the response rate was 20% or 80%. The reason we think it matters is because these things do not happen at random. People choose not to take part for particular reasons, which may be associated with the response they give.
Whether having a range of response rates across different Departments in this case, or any other kind of sub-group, matters depends mainly on what you are going to do with the data. If somebody is going to use it to compare Departments and the nature of the response is rather different in one Department from another, you might worry about big differences in response rates. But if that is not what it is being used for and it is only being used within Departments, you can take each case on its merit and you are not assuming comparability between Departments.
Q20 Tom Randall: If there is a low response rate, are there actions that can be taken to reduce any risk of bias in that data from a particular organisation?
Peter Lynn: Yes. Broadly there are two types of things that survey researchers do. One is to try to maximise the response rate. You may see that a response rate appears to be low in a particular sub-group or department and take actions to encourage more response. The other is that at the end of the day—maybe this is what you are asking about—when you have your 25% response rate, or whatever, do you need to make some kind of statistical adjustment in the way you use the data to make estimates about the total population in which you are interested. It is generally best practice on surveys to consider that question head-on and look at the available evidence about the nature of the non-response and assess, evaluate, whether some type of statistical adjustment is warranted.
Q21 Tom Randall: Ms Henry, what further actions would you typically expect to be taken to quality-assure short data in a large-scale survey such as this one?
Sarah Henry: Responding to that question and also carrying on from Peter Lynn’s response, in order to understand whether you have an acceptable response rate, it is good practice to consider whether you understand why you have the response rate that you have. For example, you can look at the trends of the response rates to surveys in your organisation rather than comparing them with other organisations where maybe access to the survey is easier or there are different motivations for participating and so on. You can look at all sorts of things like that to validate. “Am I happy with that 70%-something response?” or, “Do I usually get 30%-something?” “Do my members of staff have the opportunity to participate in other surveys in which I get similar response rates?” These kinds of broader questions that help you gain confidence in the response you have do help you to decide how much weight you put on the information you are gleaning from the survey in order to inform the decision making that follows. If you have a lot of trust in the responses you have, you can take that and say, “I am now going to develop my action plan to address specific things that I trust people were telling me and really meant to tell me through this survey”.
If it is an anomaly and you are very surprised that you have had a very low response or even if you are very surprised to have a very high response rate, it is worth looking further to see whether the underlying question of whether you trust the information that is now presented can be answered. That is the more important question than the number, so to speak.
Q22 Chair: How do you look further? What further exercise would you carry out?
Sarah Henry: I cannot speak for all Government Departments but most Government Departments have more engagement opportunities than just the people survey. The people survey is just one pillar in a wide-ranging strategy to get employees to participate so I would look at those. I would also, in a qualitative way, keep a finger on the pulse to see what is happening as the survey data collection exercise is happening, what are people saying on the various staff forums, what are people saying in various meetings. At the end of the day, it is a relatively easy sample to access, although it is large because you know who everybody is, you work with them every day, and you can get a vibe of whether people are participating positively or not.
Q23 Chair: Potentially further interviews or maybe even further questionnaires?
Sarah Henry: You could do that if there was something you wanted to drill into further. If you wanted to have another independent data collection of the same type of questions you could run a parallel survey or a follow-on survey, yes, for sure, but the more independent bits of information that you get to complete your picture, the better you will know if you trust the information presented to you or not.
Q24 John McDonnell: Continuing on that point, Peter Lynn and what you said earlier about the numbers participating in individual Departments, you said that the quantum is not the key issue but rather whether it is representative and that provides information on that Department and that it is more difficult when you try to compare Departments but that is what we use this for. We do use it to see how it is going in that particular Department, what morale is like, what responses there are and how it compares with other Departments. It is very difficult to do that because they are meeting different challenges but you can map the passage of management and individuals, even, can I say, Secretaries of State, in different Departments in that way so it is quite critically important that you do get good participation across the piece, isn’t it?
Peter Lynn: Yes, and this is generally true of most surveys. This is a common issue. Surveys are comparative. They are used because people want to compare sub-groups of some kind or compare over time or compare some particular thing, so it is important that the nature of the non-response is known to be broadly similar across whatever groups are being compared and that is why best practice would normally be to explore the data as far as you can to see if that is the case. In the case of this survey, it is clearly a survey where one has a lot of statistical information about the whole of the initial sample. You know what grade they are, age, gender, all these kinds of pieces of information, so you can compare the profile of respondents across whichever Departments or whatever other sub-group you are interested in to see whether the nature of any response tendencies is similar across Departments and if not, one can adjust for that statistically.
Q25 John McDonnell: On the Chair’s point, it would be useful to know what other mechanisms are being used to reinforce the census approach in terms of qualitative work, further investigations and so on. We can get the bald statistics about what is happening but take the question of morale. You can see whether or not there is a high level of morale or if people are deeply miserable in a particular Department but it would be useful to know what is behind that from other forms of survey, what the causes are and what the consequences are.
Sarah Henry: I agree. Essentially what you would do with the survey alone is take the responses to that question and cross-tabulate them with other questions. You will glean something out of that but you will not get information about the data that you have not collected. It is definitely worth investigating further, particularly if you have a response to a question that you are not happy with. That is what the action plans will cover, I am sure.
Q26 John McDonnell: You can usually track the passage of Dominic Cummings from Department to Department.
Anyway, if we can go now to the publication and distribution of results questions, the first is to what extent should the survey participants be informed about who will be able to access the data and the individual Departments when taking part?
Sarah Henry: The more transparent you are in these kinds of surveys, the better.
Peter Lynn: It is essential that respondents or potential respondents are informed as to exactly how the data are to be used.
Q27 John McDonnell: To Sarah Henry, to follow up on that, is it common practice to share the results of the survey with users prior to publication, for example, showing results scores on a dashboard as the responses come in?
Sarah Henry: Do you mean live?
John McDonnell: Yes.
Sarah Henry: I have not seen it happen on this particular survey. I think there is probably no harm—no, sorry; that was inaccurate. The thing that we do get live commentary on so to speak is the number of responses we are getting.
John McDonnell: Right. So you will get that?
Sarah Henry: We get that information but not—
John McDonnell: But not necessarily the results?
Sarah Henry: That is correct. I cannot off the top of my head see much value in providing ongoing information about the breakdown of the responses to individual questions but it certainly is helpful to see how the survey take-up is going because then you can direct your encouragement.
Q28 John McDonnell: How long does the survey take; not the individual forms but the whole passage of events on it?
Sarah Henry: It depends on whose viewpoint.
John McDonnell: From the surveys going out to individuals and then the results?
Sarah Henry: From the respondents’ perspective, I think they get a few weeks. You often get extensions. If the question is do people have enough time to respond, probably yes.
John McDonnell: Also if something critical comes from the information as it is ongoing.
Sarah Henry: As far as I know, the results are not analysed as the data is being collected. I think it only gets analysed automatically in one go after the data has been collected.
John McDonnell: I see.
Sarah Henry: Again, that is one thing that is worth checking with the team that owns the survey.
Q29 John McDonnell: The final question from me is about the issue of sensitive data such as personal demographic details, for example experiences of bullying and harassment. How should access to that sort of sensitive data be managed?
Peter Lynn: Providing that the data are appropriately anonymised, the sensitivity of the question should be irrelevant. If it is still not possible to identify who has given what answer if the data are appropriately anonymised, that is as equally true for sensitive issues as non-sensitive issues.
John McDonnell: Again that relates to the scale, the point that was made by Ronnie Cowan, the numbers of respondents.
Peter Lynn: Correct.
Sarah Henry: There is a difference if that information is being provided as a response to the bank of questions around, for example, bullying and harassment or if it is information being disclosed in the open-ended questions. Government Departments have safeguarding policies so if information that is disclosed through the survey highlights a concern that needs addressing, the safeguarding policy, I am sure, will kick in.
John McDonnell: It is a key issue because we have experienced the exercise in Parliament itself and it has been difficult to get accurate information because people feel that the release of sensitive information is so difficult for them.
Q30 Chair: Currently the results of the people survey are published on gov.uk in an aggregated format. To what extent is it appropriate for data relating to employee experience to be made publicly available?
Sarah Henry: I guess the question is who is going to use it and for what purposes. If people are using that information for research purposes, for shaping the employment experience within the civil service and in this example in particular, it needs to be made public. Certainly from the UK Statistics Authority’s perspective, transparency around the results of surveys from data that we have collected is a fundamental criterion to instil trust in Government statistics and these are Government statistics. Unless there is a very good reason not to publish something, it is good practice to publish.
Q31 Chair: Does it not amount in any sense to a disincentive to participate?
Sarah Henry: From a respondent’s perspective, there might be a handful of people who would find that—
Chair: They would not complete it anyway.
Sarah Henry: —but it should not be because they cannot be identified.
Q32 Chair: Mr Lynn, do you agree with that?
Peter Lynn: Yes. It does raise the important question of expectations. For any survey, managing expectations and particularly on behalf of the survey respondents is very important. It is important that upfront the respondents understand how the data are going to be used and what they are going to be used for and by whom, so they need to know that the data in a purely aggregated anonymised form is going to be made public and it would be normal practice to convey that when you are introducing the survey at the start. I think that is crucial.
Q33 Chair: What is considered to be an appropriate format for publishing data sets online? The results are published as a spreadsheet rather than as a report with accompanying commentary. Do you consider this sufficient to enable people to interpret the data accurately?
Peter Lynn: Practice does vary between surveys and between organisations. It can be perfectly appropriate to provide data in this kind of spreadsheet form. There are advantages in perhaps additionally publishing some kind of commentary that explains the data and summarises key findings so that people looking at the data would have some idea of what they might be looking for, and what kinds of patterns might be of interest.
Sarah Henry: I agree with that. I think the data should be made available in a machine readable format that is easy to read. Excel might not be the best solution but making that summary record for machine learning is good practice.
Q34 Chair: Considering previously raised concerns about participation rates and the variation between Departments, do you consider that the people survey data are reliable?
Sarah Henry: Broadly yes but I will go back to my previous answer on this to emphasise that if you do not trust some of the data for some of the responses—and that can be in a Department, it can be in a sub-set of a Department, in a directorate or a division—there is a decision that needs to be made about what you do with that data? Do you take that data out? Or do you leave that data in to be included in the wash of summary statistics? The main point here is the trust that you have in the responses. If you decide that you do not trust them, that is when you need to go to that next level of conversation, “What do I do with it?”.
Q35 Chair: So to that extent, as published you would consider them to be reliable?
Sarah Henry: I do not have information about any data that would not be reliable but if it was my job to look at that in detail, I would go over the information and if there was something that looked abnormal—for example an outlier in responses from particular Department compared to the history, compared to what they normally do—if something looks unusual, you need to look at it again. If it looks similar to what it looked like in the past as you are eyeballing it, it will probably pass.
Peter Lynn: My response would be that to have a good understanding of whether survey results are reliable, one needs to look closely at the survey process and be reassured that from start to end, the survey is carried out following best practice, in good ways, no mistakes have been made and so on, rather than just looking at the data the comes out at the other end, which could have come from anywhere.
From my limited understanding of this survey, it seems to me that most aspects of it are carried out according to good practice. There probably are some things that could be improved and they probably could be made more reliable but these are probably marginal things.
Q36 John McDonnell: Just on that, do you survey the stakeholders in the process itself, those people who have participated and those people who have used the data? Do you survey them about their confidence in the process? It sounds like a survey of a survey, I know, but there has to be some understanding of how valuable this is or whether people have confidence in it at the end of the day.
Sarah Henry: I would say yes, the surveying of people—using that word in English rather than methodologically—can be qualitative and certainly in my experience—I am both a respondent and receiver of the information—we scrutinise quite a bit whether we have confidence or not in the results we get and whether we can explain them. Sometimes you might get a result that might look like an anomaly but you can explain it—people were working very hard during Covid; you can understand that they were very stretched and they said so in the survey, which would have been different from previous years, for example.
Q37 John McDonnell: Does that mean, for example, you will survey the Permanent Secretaries of Departments or managers, about if they have confidence in the survey itself, or others? Will you survey the trade unions, for example? Who surveys them? Who engages with them to see whether they have confidence in the process?
Sarah Henry: Essentially, because this survey is repeated every year, it is cyclical, so people participate prior to the survey going out to inform how it is going to be tweaked, what is going to be different, what can we learn from previous experience before we go into the next survey. It follows that that means they will also be commenting next time round on the experience. We are continuously engaging with them and improving our process.
Chair: Thank you both very much for coming here today. It has been immensely helpful. If there is anything that you feel you could add to the evidence you have given today, please could you write to us? We would be very pleased to hear from you but that is the end of today’s session.