Family Finance Surveys User Conference 2026 - Session 1

Hello. Hello everyone. Uh it's great to be here and I'm glad to welcome you all to the DWP latest developments and updates from the surveys team within the department work and pensions. My name is James. I'm the chair of this session. I've been a member of the family resources team for a few months now. Uh and I've been a government statistician for a few years. Uh let's see here. There we go. Wonderful. Um so first up we're going to have a little introduction from Joanna who is the head of surveys team. She's been a government statistician for 35 years and has been head of surveys branch for 15 years. So over to you Joanna. You can press it myself. Okay. Which one? Just click on that. >> This one. Yeah. Okay. Uh thanks very much for that introduction James and welcome to the update from DWP. Um I'm just going to do a few uh brief reminders and setting the scene uh and then my colleagues will be uh providing more detailed updates across their own publications. So the presentation uh yes so it's just a reminder that today we'll be focusing on analytical developments. We're not going to be talking about policy issues. Uh also a reminder that our statistics are developed under the code of practice for statistics and that means that updates on developments on DWP statistics are published in the DWP statistical work program and that's updated every month. So we're obviously sharing some news today. Uh but otherwise do keep on looking at the statistical work program for further updates. Uh and I say it's good to see so many people here today and joining us online. Uh obviously this is an important part of our user engagement uh to be able to update you and uh talk to you. Uh but DWP also makes use of stats usernet as part of our wider usment user engagement as well. Uh you can see some background to stats user network on the slide. Do follow the link to sign up and see set up a profile if you are interested. Uh so just a brief update uh then in terms of one of the projects that we talked about last year um before we focus on the updates from my own team uh and just want to highlight um yeah uh this recent update in the children in low-income families publication. So based on uh so that means that results on the new afterousing cost measure have now been published. Uh again you can see the detail of those new measures if you follow the follow the links. Uh any queries on this though please do send a message to the team inbox that's on the slide. There's no one from this team here at the event today. And also given the research interests of attendees at the event I do just want to highlight the publication of the DWP areas of research interest that was published last month. Uh do take a look at the link again for more information. The ARRI invites the research community to partner with DWP. Uh so there might be opportunities there for you. Uh again there is no one from the specific team here today but please do contact Victoria Whitaker for more information if you're interested. I know I'm going backwards in the slides. Just bear with me a minute. Why am I still going backwards? I got it. Right. There we are. >> Sorry about that. >> All right. So, yeah, there we go. There's the email address for Victoria Whitaker, uh, for the DWP areas of research. Uh, I'll move on to the next one for you. There we go. That's it. Thank you, Joanna. Um, oh yes, now we go on to the family resources transformation for you, >> which is still me. Yes, [laughter] >> here we go. Okay, but yeah, so before I start talking a bit more detail about the developments, um, before we turn to the specific project updates uh, particularly relating to the FRS, I'm pleased to tell you that last night a number of team members attended the Royal Statistical Society Excellence in Statistics Awards. uh the integration of administrative data into the FRS and related outputs uh won the 2026 Campion Award for Excellence in Official Statistics. Um the Thank you very much. Thank you. [applause] Thank you very much. Um certainly yes, we were very pleased. Um and uh yeah the the message from the panel uh in considering the awards it said that um the overall project was highlighted as a strong example of ambitious efforts to improve a set of well-used outputs using a well- constructed methodology. The work is likely to be highly useful to a wide range of users and helps plug long-standing gaps in a major survey. Uh Don Burke, the F FRS transformation lead, accepted the award on our behalf last night. Uh and as I said, we're all really pleased about this recognition for a project uh and we were reflecting on the journey that has brought us here. Uh you'll be hearing then a bit more about uh the output in the coming uh session. Uh and obviously as users, I do hope you value the improvement in quality that we've been able to deliver this year. So John's joining us online today. Uh and so I'm just going to briefly uh give a very broad um reminder of the high level approach for the integration of the administrative data into the FRS. I'm not going to be going into the detail of the methodology today. Yeah. Because we've talked about the detail of the methodology at previous events. Some of you, if you've been here the last couple of years, you'll remember that uh we've published the first integrated data sets and publications just in March this year. This built on the research that we've talked about previously at this event and the technical paper that was published in March 2024. Um at this stage, the integration of administrative data resolves a substantial proportion of the long-standing undercount of benefit receipt in the survey, but it does not resolve it completely. A new technical paper and tables were published in March and that outlines the approach implemented this year and demonstrates the effect on individual benefits in more detail. As I say, I'm not going to go into the detail of the approach today. I'll highlight the broad steps drawing on the content of the technical paper. Other speakers will highlight the effect on their specific outputs. If you do have questions on the detailed approach after reading the U published technical paper, do contact John or the team inbox for more information. So the very broad steps um that we take in terms of the new approach um the first step is the production of the lookup file. So we can link survey records to administrative records held by DWP and since 2020 we can find NOS for more than 95% of respondents on average on a UK basis. There are of course differences in the match rate across countries and regions. Then we aim to identify the benefits that respondents are in receipt of at the time of the interview and the amounts they are paid on an ongoing basis. We also aim to ensure the benefit amounts recorded reflect the rates that apply in the relevant financial year. We use a number of different data sets across the different benefits. We use the central payment system to identify the actual amount received and the amount of any deductions or recoveries. And in the vast majority of cases, the total weekly benefit amount identified uh by the matching code in the specific administrative data set for the benefit does equal the central payment system net amount plus deductions or recoveries. But in some cases, there are discrepancies that have had to be investigated. As I say, the details of this approach across different benefits is outlined in the technical paper. For the support proportion of UNL cases, we take the self-reported benefit information from the survey and there are other benefits where there's no administrative data available and in that case we also use the survey responses as we have done previously. Additional credibility checks were then used uh before the benefits tables were compiled and the new benefit variables values were then included in the relevant FRS tables. We had outlined research previously on adding additional benefit controls to our grossing approach in our March 2024 technical paper. But we have not been able to implement a revised grossing rate grossing regime at this stage. Therefore, using the current approach, administrative data linking resolves around 50% of the benefit under count at UK level but does not resolve it completely. There is some variation in this reduction again across country and regional level. The remaining undercount is due to an under representation of benefit recipients in the FRS sample as reported in the technical paper in March 2024. The technical paper published this March included detailed tables illustrating the effects on each benefit. And the use of link data affects both the reporting of benefit receipt and then also improves the reporting of benefit amounts received which are included in the FRS and related outputs with a focus on implementing our initial research into data sets to be published as accredited official statistics. Other research strands were paused, but now research work is continuing to look at the integration of other administrative data s data sources, including data from HMRC. Future plans for implementation will be published in the DWP statistical work program and relevant release strategies. I hope that sets the scene for everyone for the developments that underpin the latest FRSbased release from the team. And now I'm going to hand over to Freya who will tell you more about the FRS itself. Thank you. >> Thank you, Joah. Uh here comes Treya step. [applause] Wonderful. Thank you. >> Perfect. Yeah. Perfect. So, thank you very much, Joanna. I'm here today to talk to you about three separate publications that we produce and surveys branch. Going to start off with the family resources survey as trail just then. So for those of you that don't know, we are an annual publication and we interview UK private households and compile lots of lovely statistics off the back of that which we'll go into. So we're obviously compliant with the code of practice for statistics. There's three pillars in that and I'm just going to outline how we are indeed aligned. So under trustworthiness, we have our release strategy which continues to give transparency on our future plans and is available on the below.gov.uk website link. In terms of quality, we very much remain a large-scale survey uh with a big pool of respondents getting nearly 30,000 adults in last year's survey. And we obviously have a keen focus on accuracy trying to have face-to-face interviews where possible uh referencing documents and also the admin benefit linking as just trailed by Joanna. In terms of value, we try to make sure that our survey and survey outputs are as useful as possible to the DWP policy arena. So we have lots of information on income, not just work and earnings but pensions, state benefit receipt and other sources such as investments. We also have lots of information on characteristics. So demographics, tenure, disability, carers and child maintenance and child care. There's also plenty of other facets available such as material deprivation, pension participation, and food bank usage. So our most recent publication was in March 2026. Uh that's available in three separate locations. So for all gov.uk UK uh for some slightly more detailed analytical breakdowns on stat explore and the micro data is available on the UKDS or ONS secure research service. So what's new for 2425 our most recent publication obviously the integration of uh benefit admin data all the wonderful things that Joanna just trailed there. Um so information from major state benefits and tax credits is now based on admin data sources rather than survey responses where the uh linking is possible. This process is at core of the periodic review of the FRS that the office for statistical regulation are currently running. So alongside a brand new 2425 data set, we also produced a back series back to 2021 2022 and this is to make use of our new availability to do the admin linking on the benefits data. And there were some other alterations which you can also see in our background information and methodology report. So we just talked about what happened in March 2026 looking ahead through 2026 and indeed through to 2027. Uh we're continuing to look at our uh response monitoring in the field, looking at our sample sizes, compositions, mode, and also just making sure there's uh enough capacity within the field. Our next publication will be in March 2027. So this will be the 2025 2026 set of FRS results and it will also include a further three admin linked back series years. So that will take us from 201819 to 2020 2021 with that updated admin approach. This is as far back as we can go based on uh admin linking rates but that will produce a fall back series with the admin link data to 201819. There'll also be a grossing rebasing. Currently grossing uh population estimates are rooted in census 2011. We now have census 2021 and census 2022 for Scotland population estimates available and so we'll produce new grossing factors based on this updated um population back um data set. And for that we're going to produce a back series back to 2012 2013 up to 2024 2025. And this is very much to allow for comparability over time and try to be uh you know as useful as possible. There's also going to be changes to council tax ban. So the way that uh council tax has been improved with an automated reading of CT band from the BOA. Uh going to also pursue a similar approach for Scotland via ONS and Scotland uh assessments association. There are some other things to note. Uh the 25 26 data set will have an updated number of food parcels from food banks variable. The 2425 data set will have uh new variables well updated variables for growth CT annual and CT amped as an error was found with the uprating of this these variables. So they'll be reissued for 2425. The impact of this is minimal but they will be included in the March 2027 publication and for 2021 the data set will have a fix to the main educational variable. So this was the back then DV highqual is now educ. There's also going to be additional questions on whether a child maintenance arrangement is working. So that's me there for the F FRS update. Now going to step through two other statistical outputs uh that we as a team produce. So that's pensioners incomes and take up uh both links there and also the relevant team inboxes today. So going to start with pensioners income. This is a annual accredited official statistic and it reports on pensioners income and it is rooted in FRS data. The statistics themselves examine how much income pensioners get each week and where they get that income from. Uh it looks at how the incomes have changed over time and variations in income by different pensioner types. The estimates are normally based on a sample of around 7,000 pensioner units. uh in the most recent year which is financial year 2024 2025 there were 6,300 pension units in the sample the data for pensioners income is available not just go it's also available on statics explore and the UK data service uh because pensioners income uh statistics are based on the FRS data they also have the benefit of being rooted in admin linking from 2021 2022 which we can see here in this graph you can see the break in the time series where the admin linking is brought in again that's admin linking for benefit receipt Um, and you can see here that slightly increases the pensioners income. Uh, and that is to be expected because benefits tend to be under reportported in the survey. Uh, and so the admin linking helps reduce that somewhat. You can see in the last four years, pensioners income has been relatively stable. In the financial year ending 2025, pensioners had an average median income of £455. Uh, the average median income for a pensioner couple was £650 per week and single pensioners had an average income of £332 per week. So that's pensioners income. Now I'm going to talk about the takeup statistics that we have. So this is again an annual publication. Uh this is normally published in October whereas the last two we talked about normally published in March. Uh and take up basically refers to the receipt of benefits someone is entitled to either by case load so percentage of eligible people and expenditure percentage of money claimed. This is again based on FRS data. So the data is matched to administrative records to produce estimates for main income related benefits. This is currently pension age only and this covers pension credit and housing benefit for pensioners. So looking at the most recent results which is 2425 uh six and 10 of those entitled to pension credit claimed the benefit and 71% of the total amount of pension credit that could have been claimed was claimed for housing benefit for pensioners. Nine out of 10 of those entitled to it did claim the benefit and 87% of the total amount of housing benefit for pensioners that could have been claimed was claimed. A little bit more about income uh take up statistics. So the department previously published estimates of working age take up of income related legacy benefits. The introduction of UC however meant that methodology had to change. We're now at a point where we successfully worked through some data issues and compound plan pan plan to publish the first estimates obviously in compliance with the code of practice for statistics. So those are intended to be published uh in October 2026 and this will be take up of universal credit or equivalent statistics and this will also be alongside the pensioner take up statistics shown in the previous slide. Um this composite measure will assess the take up of universal credit or equivalent income related legacy benefit. They'll be labeled as official statistics in development as we continue to refine the methodology and indeed the changing benefit landscape. This initial release will include financial year owning 2024 and financial ending year ending 2025 estimates and it will reinstate the reporting on working age population which was last done in the financial year ending 2016. Uh all of this information of course can be found on the homepage and the department for working pension statistical work program also has more details on this development. I believe I'm going to pass over to Angela now for households below average income. Thank you. [applause] Mine is you can just click the button to move forward. Okay. Hello. Um so yeah, I'm Angela. I'm going to talk to you about households below average income um publication which again as Freya said um as with the FRS and the pensioner incomes is an annual publication and our latest release published in March. Um it produces estimates of number and percentage of in individuals uh by measures of various measures of poverty including those that you can see in the in the boxes. Um but there there's lots of information on and HBI is commonly known as the poverty statistics. Um very similar again to FRS and PI. Um latest release relates to 2425 survey year. Um and you can see on there kind of where you can find um extensive set of tables and data available on stat explore on the UK data service um and our team mailbox um if there's any queries after after this um just briefly in the orange boxes new for the 2425 publication again as far as mentioned um HBI is based on the FRS and so our big development for the latest release is the integration of the survey and admin link data um and back updates to the back series again as as Freya said back to 2122 and then two specific things just for HBI the last two orange boxes um we also updated the absolute low-inccome reference year to 2425 so that's previously um 201011 um and absolute low income is one of the poverty measures if you're familiar with that um and we also introduced the new deep material poverty measure for children into the HBI publication. So that was previously first published um the year before um as an ad hoc statistic but now will be reported on going forward annually um within HBI. So the next few slides I've just got um sorry my um just got a few slides showing the latest results or some of the latest results from the 2425 release. So this slide um is all about median weekly household income. So the slide on the left shows the income distribution before housing costs um for the total population and you can you can see the kind of distribution um of household income on there and skewed skewed to the right and the median household income of £719 per week before housing costs. Then the house on the chart on the right um also shows medium weekly household income but over time um and again with the pensioner income series as Freya showed you can see the break there for where we've linked um the admin data from 2122 um and again it that shows that that has had an impact of increase in incomes um uh and again you can see that 719 um weekly income before housing costs and 600 123 um income be uh after housing costs. And then uh this chart just focuses on a couple of the poverty measures and and these are the two um headline measures used for the recently p recently published child poverty strategy. So there are as I said there are me many different measures of poverty within our report and within our tables. Um but on the left hand side you can see a chart showing um percentage of children in relative low income or relative poverty. Um after housing costs and before housing cost but after having housing costs that's one of the um government's preferred measures um that's referred to in the child poverty strategy. Um and you can see there again the kind of breaken series for 2122 um due to the introduction of admin data. Um and that's actually lowered um poverty rates very slightly. Um you can see after that and then on the right hand side the new measure that I just mentioned that was introduced um as the second headline measure for the child poverty strategy the percentage of children in deep material poverty. So again, we've only got two years for that. Um, and you can see that that on the right hand side, that bar chart, um, has reduced by one percentage point from 14% to 30% um, of children in deep material poverty. And the way that that's measured is um, looking at um, the material deprivation questions and 13 essential items out of that. And if children are lacking four or four or more of those, then they're classed as being deep material poverty. Okay. And so, and then this final chart um just shows the impact of admin linking for benefits on the income distribution. Um and I've just realized actually that heading isn't quite right. It's not really headline results for 24 25 because this chart is 23 24 um as the title below that says. So 23 24 is the last year that we've or the most recent year that we've got both um non-lin published data and admin linked data so we can compare them. Um so for 2425 onwards um we've only got the admin linked. So this chart shows uh the change between those two years. So the impact of linking on that single year. Um and as you can see there you can see again that um linking increases reported income levels because it's more common to under reportport rather than over report benefit receipt in the family resources survey which fraud mentioned. Um and Lincoln has a greater effect on the lower part of the income distribution which you can see there by the sort of the 10th percentile on the sort of steep um part of the chart there. Um and that's because a greater percentage greater proportion of income for people at that bottom end of the distribution comes from benefits. So again, it's what we'd expect. And then finally, I'm not going to go through there's quite a lot of detail on here, but um a lot of it refers to again what what Freya said and what Joanna mentioned. Um so on the left hand side um just to note the things we did in our latest release in March, the big one that the admin linking um but also uh the deep material poverty measure for children and the absolute low income reference year um has uh are the changes that you'll see in that release. And then on the right hand side the developments for 2526 again very similar to to what Frey has mentioned already um will FRS developments will feed through to HBI next March. Um but again as Joanna said do keep an eye on our um release strategy for other development updates that will be coming for for next March. I think I'm handing over to Helen. Sorry Helen gone too far. >> [applause] [cheering] >> Click on the button there. >> That one. >> Okay. Thank you. >> Thank you. Okay. So, um just a just a change in um present present. Well, am I a presenter here? I'm not I'm not Abby. Sorry. [laughter] Hopefully, it's going to get better from this point onwards. Um but my name's Helen Smith. I also work in the income dynamics team at DWP. Um so um I'm going to talk a little bit about income dynamics which um unlike a lot of the other uh publications um discussed by my colleagues at DWP is based on understanding society data rather than the FRS. Um so understanding society um is managed by the University of Essics. Um and it's a longitudinal panel survey. Um and so our statistics are longitudinal and as as such they differ um from HBAI um and the cross-sectional measures um included there. So the key statistics we report on um are rates of persistent low income. So that's um individuals who are in relative low income for any three out of four consecutive survey waves. uh we also look at short-term movements into and out of low income and the events associated with these and that's across two survey waves and we also look at longerterm income mobility. So that is across the full income distribution and um that's over a longer time frame. Um so our most recent publication included data from wave 15 of understanding society covering the calendar years 2023 and 2024. Um and our analysis starts right from the back uh the start of understanding society. So the recent publication covered the period 2010 through to 2024 inclusive. Um so continuing on a theme of sort of developments in in data sources here. Um the main development for for this year's release was the inclusion of understanding society's general population sample boost uh which was launched in 2022. uh and we brought this into our longitudinal sample used to produce statistics on movements into and out of low income. Um that's our entry and exit analysis and the events associated with these movements. Um so um so what is the um I'm going to call it the GPS 2 because that's what the University of Essics call it and it's quicker. Um and why was it introduced into USOK? Um so it was introduced to improve um the survey's representativeness. So specifically to address the effects of attrition and um immigration over time and to improve the coverage of minority samples. Um it was introduced into the uh understanding society in wave 14 which was 2022 and 23 and it resulted in an additional uh 5,761 household interviews um and um there has been quite a bit of analysis conducted by the University of Essex um and it concluded that ultimately the the boost achieved its aim of improving the surveys representativeness. Um so members of the GPS2 uh have been included in our cross-sectional analysis from wave 14 um the the wave in which it was introduced um and then this year we brought them into um our longitudinal analysis carried out out over the most recent two years uh waves 14 to 15 because we had two waves worth of data on that boost. So the table probably sums it up a little bit better. Um, so we've included them in our cross-sectional income measurements from 22 to 23 onwards and then this year in our two-wave longitudinal analysis, but we haven't we can't include them yet in any uh longerterm measurements because we don't have data on them for sufficient numbers of waves. Um so just a little bit in terms of like our key uh headlines and perhaps um some of the effects um that we've observed which may be linked to the boost. Um so these are our rates headline rates of persistent low income across the two most recent four-wave periods um for children, working age adults and pensioners. And then on the right hand side for all individuals and the the lighter shaded bars are the previous four-wave period and the darker ones are the current the most recent four-wave period. Um and what we observed was some small decreases across most categories um that you can see there. Um generally persistent low-inccome trends um are stable. They haven't changed much since uh 2010. Um but we in a way we weren't that surprised that we saw some decreases because across the most recent period because we have seen lower slightly lower median incomes since we introduced the boost. Um and um those median income values are used to measure who's in poverty in any single wave. Um and so seeing lower levels of uh lower levels of incomes overall, but we have an otherwise unchanged sample. Um we weren't that surprised to see um slightly small slightly lower rates of persistent low income in the most recent wave four-wave period. Um so in terms of movements into and out of relative low income over the two most recent survey waves. Um so the headlines here we've got entries on the left hand side and then exits on the right hand side. Um and yeah overall children are more likely to enter low income than working age adults or pensioners. And then in terms of exits from income, low-income working age adults are more likely to exit from low income compared to children or pensioners. Um and in terms of uh bringing in the boost sample and kind of some of the changes that we saw over the most recent period, uh we did see slightly higher rates of uh entry into low income. Um and we saw lower rates of exit from low income. And again, this was something that we weren't entirely surprised about uh given what we know about the uh booth sample members being uh representing populations on slightly lower incomes. So um so yeah, that made sense to us. Um I think that's that's my last slide actually. Um but um in terms of looking forward, we we we will be bringing in the boost um for uh persistent low-inccome estimates, but it won't be next year. It will be the year after when we've got four waves worth of data. So that's me. Um sorry I I did wing it a little bit. Um I don't know. >> Thanks Helen. Thank you for filling in at short notice Helen. Uh next up is Neil maybe. There we go. For the wider developments from the universal credit low income statistics. You can just press the >> or the arrows as well probably. >> Okay. There you go. >> Yeah. Thank you James. I'm Neil for short. You can call me also Neil. Um we've this is the last um section um before questions and answers. So there's only two slides on uh UC low-inccome stats, but I will take a little bit of time uh to unpack them. If you were here last year, um we gave a taste of last year. So this is an update. Progress not as quick as we would have liked. When is it? has been um some steady progress and I can be a little bit more definitive um about what we're going to publish. Um so um universal credit um covers I think over 7 million families these days and over 12 million individuals in those families, adults and children. Therefore um a big chunk of the population. The admin data for those families is pretty decent in terms of income. Obviously, income from UC itself and from earnings, which is part of the calculation, but we can also imputee uh income from child and disability benefits as well to end up with a pretty comprehensive um family income picture for families on universal credit. We also have housing cost data for renters because housing support for renters is included within universal credit and we've been working at imputing mortgage interest payments um for people paying a mortgage. Um so we can also do housing costs um for those folks. Just to be clear, housing costs for poverty stats purposes are just the interest payments for those paying a mortgage. They're not the repayment um part. So it's mortgage interest payments. Um therefore we can produce something like the income distribution um that you can see on the slide. It's similar to an HBI income distribution in terms of like pounds per week um that Angela presented to a similar format. I should say it's towards the lower end because it's families on MU. This is illustrative as it says on the watermark. Um, we've put a poverty line of about£320 per week on there for illustration purposes. I think that's 2425 absolute poverty line based on the 1011 median. But you, you know, you put the poverty line um wherever you like, but to illustrate, it then tells us who's below the low-inccome line, who's in poverty, and who's um not in poverty. Um and also as it says on the title um we can split up between children and adults um quite happily. So we can do poverty figures from UC admin data uh only uh and therefore because it's admin data only. Um we have no sample size issues millions and millions of families like I said therefore you can do local level figures and you can also do much more timely than is possible for um HBI and children in low-income families round about within six months um of the reference period. Um is this the end of survey data? Um I'm certainly not going to claim that at a surveys family finance surveys user conference. Absolutely not. Um it only covers those on universal credit by definition. You need surveys to cover the whole population to find people who are in poverty. Um not on benefits. You also need surveys data to actually put people together into households. Um I've deliberately talked about families. A lot of you folks will know that families and households are different. Um technically they're often the same, but you can have multiple families in the same household. So this is a family level um poverty measure. And then the other thing I should say and where we do still have to use survey data for these statistics is we need to know what the poverty line is and that comes from HBI. So we're going to say who's in poverty for families on UC with reference to the UK poverty line from HBI. We just simply then have to up rate that from the latest HBI year 2425 at the moment to early 2026 or whenever we're publishing um figures from. So last bullet point uh on the bottom there, official statistics in development, just because it gives us a bit of scope to um fine-tune some things methodologically. And we're going to publish quarterly alongside um the DWP's use, the official statistics and indeed the quarterly benefit statistics in general in those months. Can't be explicit about when, but it's definitely months now. We're not talking years in the in the coming months as it says there. Okay, second and last slide on this before we get to questions for the whole presentation. Um, this is the approximate format that we'll be publishing. We'll publish in stat explore in due course, but there's a fair bit of leading time to do that. So, it won't be stat explore to start with. It will be fixed tables similar to the fixed tables that children in low-income families um publish. So as you can see there you can do geographies um down the side that happens to be local authorities. We'll do parliamentary constituencies and regions as well. Um again it's admin data. So you'll be able to get to ward level and super output area level and all those good things but only when we go to stat explore. We're not doing that as fixed tables just because of the volume um of data. Anyway, geographies um down the side and then some metrics um across the top. If you look at the orange uh columns, they're going to be poverty numbers and poverty rates for families on um universal credit and will split by children and adults. Um that's helpful, but we know already even from the published HBI data that poverty rates for individuals and families on UC doesn't vary that much geographically. So for HBI, we'd pull three years of data together to do region figures, but we'd be looking 40 45% um ballpark and it wouldn't vary that much across the UK. That's because you've already subseted to families on universally on universal credit, low-inccome um families. And in some senses, there's some areas with high housing costs that are quite affluent. affluent in terms of deprivation measures and poverty rates altogether, but they actually score slightly higher poverty rates just for those on universal credit because they tend to have what's called housing shortfalls, high housing costs. The local housing allowance rates within UC are more likely to give a gap and then AHC after housing cost UC measures actually brings the poverty rate up a little bit. So the orange figures h are useful but for child poverty which is the blue columns we also want to give an idea of how many children in UC are in poverty not just as a proportion of children on UC which is going to be 40 to 45% in most cases but as a proportion of the child population as a whole which gives a lot more um differentiation um across geographies as it says there we're doing that for under 16s that because that's because we don't have local level um population denominators for children. For 16 to 19 year olds, you've got to divide them up between children and working age adults and we impute that for HBI at region level. Um but it's lots of work and we can't do that reliably at local level. So a bit like syllif you get um proportions of children in poverty for they're there for under 16s. Um similarly um for here other reason for doing that for children is that most children in poverty are on UC. I mean it's difficult to be definitive about that. There are children in poverty who are not on UC. Um whether that's because families aren't entitled in the first place in the benefit system, no recourse to public funds or again uh owner occupiers that their income might be high enough to be just outside just above the UC threshold for their family type. But then if they're owner occupiers paying a big mortgage, their housing costs might bring them under and take up issues and so on and so forth. So the the dark blue there, number of under 16s in poverty on UC is not all children on UC, but it's it's a it's it's a majority. It's the vast majority. There is a pitfall. You're probably thinking that's not quite a complete numerator, but it is a complete denominator. Um all children under 16. So we think it's helpful, but it takes a bit of unpacking. Hence one of the reasons for official statistics and development. Um, I think that's all apart from two quick things. It nicely leads us to the last bullet point, which is to say our background information methodology document will explain similarities and differences with children in low-income families and HBI um and what to use um for what. Uh and then also very lastly just to point to the child poverty strategy. So again, one of the particular interests in child poverty is because of the child poverty strategy that was just released for the UK last December this morning. The child poverty strategy baseline and evaluation monitoring framework I think it's called no uh monitoring and evaluation framework I don't know something like that. Um it's the it's the publication of how we are going to measure the country is going to measure um progress against the child poverty strategy. So it talks about the key metrics that are going to be used. Um head the two headline ones are from um HBI relative low income after housing costs and children in deep material poverty. But then there's some underlying uh indicators um as well. So um that's published this morning. Tim's report by DWP about disability benefits all is also published um today. So um yes, today is a today's a big day. Um that is all. I shall put the last um slide up which is contact details for uh all of the products and hand back to James. >> So thank you very much everyone from the DWP. We're going to finish this session with a research presentation from Glenn Bramley. So um Glenn Bramley is professor of urban studies at the institute for housing uh social policy housing inequalities research and his research career spans 55 years and more recently um he's been focusing on housing homelessness and severe poverty and is the quantitive lead for the JRF supported destitution in the UK studies. I'll pass over. >> What what is the advance? >> You can click through there. >> Okay. Thank you very much for that introduction. Um, and I've got slightly more slides than time. So I will have to caner over some of them a bit quickly. Um, but uh I was asked to put acknowledge the data sources used which are primarily various uh family resources surveys and households below average income. Also, some of this work was done in the UK data service secure lab and um uh the the I'm also referring quite a lot to the destitution the UK survey that we do with Roundtree. Um the fifth wave of that is just in the process of being an analyzed and prepared for publication in the autumn. So some of my references will be informed by the work on that but will the actual some of the tables will be from previous rounds of that work. Um so uh I think background to this is that um there's rising concerns about not just poverty but I think deep poverty and I think we would argue that the trends in deep poverty or severe poverty may be somewhat different from those that are uh apparent from the main publications on on public poverty that we've been talking about. Um the and indeed uh there wasn't much talk about destitution until we started doing this work around about 2014, but it has now come back into the uh common currency. Uh traditionally perhaps mainly associated with either the third world or 19th century poor law kind of debates. Um but there's clear evidence from the studies we've done with Roundtree of quite a dramatic rise in the scale of destitution at least since 2017. uh where we have fairly consistent measures. Um and it needed really special surveys to show this for various reasons that will um emerge as we go forward. Um and really uh this paper is perhaps about ways in which can insights from this work be brought to bear on some of the mainstream uh monitoring and mapping of of poverty and I have a particular interest in mapping down to local level in this as well. Um, one particular publication I would refer to you to is a an article in the journal of poverty and social justice by Braramley and Fitzpatrick in 2023 which both discusses the conceptual basis as destitution and also reports on a a methodology where we composite together data from destitution survey and a mainstream survey in that case UK hls. Um now then yeah the definition of destitution we use with round tree is has three elements to it officially two elements but really three elements. Uh firstly there's a material deprivation definition. So we have six uh um material deprivations you know one relating to food clothing toiletries heating lighting and shelter. The shelter one is sleeping rough. Okay. So, they're the fairly extreme end uh uh and the most essential essentials if you like. And so, if you're lacking two or more of those, um you're destitute. Um secondly, your income's so low that you can't afford those essentials. And we have three bases that we feed into an analysis of what we think those income thresholds should be. and um that that with those get updated each each year each each wave of the survey roughly every three years now. But there is a third and implicit criterion for destitution that's embedded in the survey we do which is that people have really exhausted any assets and informal support that they would normally be able to access from family etc. uh and that is proxied by the act of having used a crisis service and that's we do the survey of people who are using crisis services could be advice services like cababs could be food banks could be hot food services uh uh free free you know lunch things or whatever um uh soup runs homelessness services of all kinds and um uh services oriented to migrants um so uh That's really the implicit third criterion. Um, if we go forward, I think that the mainstream surveys uh that we've been mainly talking about here uh up till now uh all face challenges in measuring the severe end of of poverty and destitution. Falling response rates are a cause for concern. Um in earlier in my career I I thought we generally expected response rates from official household surveys in the 60% sort of 60 to 70% range. Was a bit shocked to discover that the latest FRS response rate is 31%. Okay. Without wishing to blow my trumpet too loudly, destitution in the UK, the most recent survey we got 59.5% of the people using crisis services in the week that we were surveying them in their local area. So um obviously we were throwing resources at those particular areas but I think the challenge of low response rates to the the credibility of these services surveys is is is is significant. Um then who are we missing is the key key question perhaps. I think you've got different segments like small single households living in flats behind entry phones who are hardly ever home uh etc. that that's uh mobile private renters. These are groups that I think are typically under well under represented migrants perhaps overcrowded in informal tenencies. Um there's certainly been a big problem with uncompensated attrition in the UK hls. I was pleased to note that they've finally addressing resources to that problem from one of the presentations earlier this morning. um the force switch to telephone and online methodologies um and uh uh obviously COVID was a big problem and we've never fully recovered from that I think um and there is really no coverage of quite high-risk non-household population and people who are temporarily resident where they're staying so-called sofa surfers and so on. Um so uh we think that our destitution survey is a way of of addressing these particular groups and getting some evidence about them. So the sort of research questions this paper and there's a longer paper which I can send people if they want um uh that that goes through this um uh six research questions. Is it possible to identify and measure destitution within mainstream poverty related surveys? So can we could we actually say how many people in the F FRS sample are actually destitute and I've tried to do that and there are some problems about it. Um it's very good on income not so can't exactly match our material deprivations and the the third criterion of using crisis services but uh we can approximate it but also it's difficult to do it consistently across years because of the changes in in various aspects in FRS. So I've only been able to do it for single years whereas really I'd prefer to pull you know three or four years which I've done with some other measures like high food insecurity and so on. Um uh so what does if we can do it what does it tell us about the recent level pattern and trend in destitution uh and and other more measures of of more severe poverty? Can we develop predictive models using F FRS with local contextual variables added to the the predictors for measuring uh uh these things and and developing indicators at a lower level which um and what do these models tell us about what are the key predictors that are that are driving things um and can we then have synthetic models say that we run on census and other data that's available at the local level and possibly the new one that we heard about shortly about the universal credit. I mean, universal credit is always a good predictor in that my the models I've developed so far, but we could build on the work that we've just heard about. Um, and [clears throat] what about the people who are in severe poverty who are not usually resident private households or in those chronic non-responding groups? How can we capture them? So, um, destitution in the FRS, uh, the income uh, and savings criterion are very straightforward. In fact, your data in FRS on this is better than our data in in DUKS. So, uh that's good. The material deprivations can be approximately matched with some creativity, but also some inconsistency between years. I've got a table that illustrates the detail of that, but it's a bit too nerdy to perhaps dwell on in this presentation. Um and then the third implicit criterion, uh we cannot exactly match, you know, have people use crisis service. We we can see that they've used a food bank but not the other ones like advice services and so on. We've used various proxies like receiving or loan a grant loan or grant from former social fund local welfare fund uh local uh um household support funds those sort of sources re uh uh having to repay loans from working age benefits use of food bank in last year. So it's sort of a a bit of a mi mishmash but trying to match that third criterion. Um another issue is the time period of assessment. Um in for the DUKS we we we both we report both uh a point in time measure uh people who've used services in a week and uh have have reported deprivations in the last month. But we also can by asking a lot of questions about services that have been used in in the past year and imputing massively for the missing values there. Uh we can get an annual figure as well. Roundtree like to report the annual figures mainly which are a period prevalence measure which are about you know three or four five times the weekly incidents. If we're comparing with something like FRS it's somewhere between the two but it's closer to the point in time. Uh so you have to bear that in mind if we look at you know is it 1% or 4% or 5% depends what you're which measure you're using. Um those are this is you won't be able to read this fortunately probably. So uh this is the um uh trying to match our six material deprivation criteria. They're not even in the right order here, but that the we we've had to sort of it's not exactly the same for the three groups because material deparation questions are asked differently for the three demographic groups. Um but we've tried our best to do that, but it's also had to change from year to year. Um this is some results from the latest FRS on um uh an attempt to replicate destitution which is the dark blue uh uh a measure of which I like of severe combined poverty which is below 40% of AHC income and either um uh well mainly uh being above the high threshold score on the on the on the material deprivations. The way this is presented has changed slightly from pre used to be 25 points on the old scale but it's it's equivalent on the new scale and or reporting you know immediate financial debt type issues and that one I think has has what I would describe as the most sensible regional distribution here. Um I I'm a little bit concerned about some of the regional picture that's revealed from uh on on both my destitution measure and on the the low 40 AHC income measure um that I don't quite believe that Yorkshire and Humber is that much worse than the Northwest or that the East Midlands is worse than the West Midlands for example or that the Southwest is as low as that suggests. So I think there are some problems with single-year estimates from FRS and uh at this regional level. I would group the regions frankly you know to Midlands, South London, London is clearly the highest on all of these um yes uh moving on again. Yes. So um there are both whether we look at descriptive things in terms of demographic um issues or or indeed when we come on to modeling you can see clearly that um single working age and loan parent families have much higher risk um three to four times uh and similar for destitution and quite a lot higher for severe combined poverty very elevated risk for three plus children. This is before the recent changes in benefits. um no worker households, social renters pretty high, private renters high but not as high. Various ethnic minorities um black mixed and multiple uh are a group who are quite deprived I think on a lot of these measures as well. Um and the other group which includes often recent migrants. Um and then there are protective factors being of retirement age, the number of workers in the household, higher uh NSX etc. And so those kind of variables also crop up when we do the the the multivariant models to predict. Um there are two distinct modeling approaches I've sort of exemplified. I'm not talking so much this time about the one that's in that article in the journal of poverty and social justice which is this composite where you put the two data sets together for a common set of about 40 or 50 variables and try to predict across that. Um I'm going to talk uh I think it is an interesting approach but u another approach is to to do uh to model on the on the um on the F FRS uh for the household population and then use another source which uh uh or various other sources to capture the non-household populations. We've we've managed to get introduced into the industry of deprivation this year a measure called core homelessness which we've developed with crisis over the last six or seven years and that is available at local authority level and we can use the destitution survey to say how many people who are in the various in the five categories of core homelessness are uh destitute and or in our three risk groups migrants complex need and um uh other UK. So that's the approach that I've actually used in the latest destitution work that I'm doing at the moment uh for getting local level um estimates. Um so uh I I should have jumped. Yeah. So comp but I I won't say much about this composite approach. I think it's quite an interesting approach as well. Um but uh there are two articles now. There's also one on homelessness in housing studies just coming out as well using the same approach. Um and I am now doing composite modeling using F FRS as the main survey but I haven't published anything uh from that work yet. Um so the the these notes summarize some of the main findings and again it's similar to what you get with those descriptive statistics that I perfor strongest variables income savings financial difficulties parental help uh receiving universal credit mental health problems etc. So all of those do emerge from that kind of modeling and um so more recently um as I say I've been modeling in FRS with attached local data. This is in the UK secure lab. Um and this this again there's this slide lists the strongest predicting variables for uh severe poverty or or destitution, social and private renting census. there's a census deprivation score, you know, 1 2 3 4 at a household level. That's a strong predictor. Um, and um, this is, you know, working with census, we don't have income data in the census. Um, black ethnicity, rental affordability ratios, and so on. And, uh, so there's a list of variables on that slide that that are significant predictors. And the point of this is then to find equivalent variables from the census and other published data. Certainly that's available at local authority level potentially down to MSOA level um to have run a synthetic model that we've calibrated um to predict for the in this case for the household population um and so uh I've I've I've continued with that that work. Um uh there are some issues. I mean, one of the things I've found this time is that the the predictive model seems to display a wider variance than some of the direct measures that we use like universal credit and um sanctions and we we have a you know a number of things that we observe from stat expat and other sources. Um, and I think it there are a couple of factors that underly this. One is that if we're using logistic regression, you have to allow for the fact that everything interacts. There's automatic interaction and as you as you go up the the steeper slope of the logistic curve, then effects are stronger and secondly you also get you have negative variables in the prediction. So that so you can get a wider variance. So I think you need to have a balance between this kind of predictive model and um uh direct measures of pro known problem groups if you like um in in developing an overall uh predictive indicator set. Um now I want to just say a bit more about the non-household and under reppresented populations. Um and uh that can include people who are in institutions um uh hard to contact groups um as I've mentioned but people who may be shy of official sounding contacts that might include migrants with uncertain status perhaps particularly in the current political climate um people in the gray and black economy um victims of domestic violence and abuse and so on. Um now in our DUKS survey we estimate that 37% of weekly users of point in time users of crisis services uh were excluded would have been excluded from mainstream surveys. If we go to the annual period prevalence it's 25%. So it's not a trivial issue especially when you're talking about severe poverty. Um, and I've talked about a couple of different approaches to to uh modeling this just I probably should I'm running out of time I think probably um but um this is uh from the last round of distribution study again looking at the regional distribution of um uh uh uh this time using a mixture of direct indicators comparing with a predictive model across the different regions. Um and at that time in that time London seemed to be less dominant in the picture. Um perhaps characteristic of that period when we were in the midst of the cost of living crisis that arose from the which affected more of the group we called UK other households just regular households who suddenly have in crisis. um whereas I think now we're going back to the pattern as it was more before that uh where there's more emphasis on groups like migrants and people with complex needs for example but there's also that we can do analyses by different typologies this is the the 2011based typology of local authorities so you can see that the high scores [clears throat] are in London cosmopolitan suburbia business and education centers um uh to some extent manufacturing traits. Those are the sort of high ones. The low ones were the rural, prosperous England, coastal and so on as you would expect, but it gives you another slight different perspective. Um so I'm coming to my last two concluding slides, which I probably don't need to go over in great detail because I think I've made my my points pretty clear. Um, so I think we I I think I've sort of semi achieved my six objectives, but in some cases with with some limitations like as I've tried to explain in relation to FRS data on some of the criteria that we're using. Um it is worth saying that the destitution in the UK studies indicate that on the preferred way that we present it which is annual period prevalence uh destitution rose strongly from 27 through 2019 even more strongly to 2022. um massive increases in numbers and I will suggest that what we will publish later this year will be a leveling off but not a big fall in those numbers and in fact an increase in the migrant migrant subgroup and to some extent the complex needed a slight fall off in the UK um other category um but if the the I think the most important finding that we will report um will be that The point in time measures are still way up on 2022 and that's telling us that people are in a more chronic situation destitution and they're visiting crisis services frequently um more frequently um and they're stuck and and this is echoes some of the findings of trussle trust from their uh statistics for example um so I think um we are making some progress in in ability to generate useful predictive models models and I think some of the work I've heard about this morning, particularly the last presentation, could be fed into some of this modeling of local local estimates. Um, sorry, what have I done there? I' I've seem to have missed my last slide. Oh, there we are. Yeah. So, I think this repeats points I've already made. I don't really need to say this again. So, thank you very much for your attention. [applause]