 Ac mae'r ystafell yn dweud yn ystafell, mae'n gweithio ar y cyfnodd yn cyfrifiadau, ac mae'r methodologi yn y cyfnodd yn galw i gael allan gwell. Mae'r cwestiwn, y sample, y imputio, y cwylwgr ymddangos. A os ydych chi'n cael ei fod yn gweithio ar y dyfod y pandemig, ac mae'n gweithio'r ysgol o'ch gweithio ar yr ysgol, to the pandemic and what's changed since. So what is the labour force survey? It's a survey of employment circumstances of the UK population. Before the pandemic was probably the largest, now it's probably more one of the largest continuous household survey in the UK considering that we've also got the COVID infection study going on at the moment. And it's got over 1000 variables, including questions from the questionnaire, so questionnaire variables, as well as derived variables for analysis and the topics it covers is a general household characteristics, job and industries and occupation, employment patterns, hours worked, earnings and benefits, health, wellbeing and sickness and education and training. The overall sample design covers people living in private households, as well as nurses living in NHS accommodation, we also cover students, but not in halls of residence, they're basically covered at their parents address, the sampling frame we use is the postcode address file, as well as an NHS accommodation sampling frame that was specifically produced for this survey. We also use the telephone directory to match telephone numbers to addresses that are based north of the Caledonian Canal. I'll explain later why that is. And other communal establishments are however excluded from the sample. The sample is a random sample stratified by postcode and is representative of the UK population. With a few exceptions and that's basically because we only include the NHS accommodation as a communal establishment. So the LFS sample covers around 75,000 issued addresses and around 40,000 households take part every quarter and this equals around 90,000 individuals. So when we look at the overall population in the UK each respondent represents about 800 people. Now obviously that differs based on their characteristics. So there's certain sample members like people living in 75 plus households that have larger weights than others because we don't interview them at every wave. And also an important thing to mention is that the LFS is north of sufficient size to allow analysis below regional level. And this is where the annual population survey comes in, which I'm going to mention in a minute. Also an important thing to notice that the labour force survey is a longitudinal survey where we follow addresses over the course of five waves in three monthly intervals. So we replenish 20% of the sample every quarter. And this depicts the wave structure here on the slide. So in any given quarter let's take the first one January to March 2020. You have a cohort of respondents that is in wave one, one in wave two, wave three, etc. So a cohort from each wave that makes up the LFS dataset on a quarterly basis. The annual population survey sample serves the purpose of conducting analysis at local authority level. So the boost cases basically ensure that we have a set number of cases at local level. And we cover around 200,000 households or we contact around 200,000 households and around 140 of those take part and that equates to around 280 individuals. So what is the connection between the two surveys? So these are not two standalone surveys, they're actually connected. And in the way that part of the labour force survey is used in the annual population survey. So the annual population survey is basically made up of part of the labour force survey as well as the labour force survey boost. And this here depicts the wave structure of the boost. This means that respondents of the boost are contacted in four waves in 12 monthly intervals. So we replenish 25% of the sample every quarter. So in any given year, so let's take the first one here in 2017, you have a cohort from each wave similar to the LFS. Looking at the APS, so as I said that combines the LFS specifically wave one and wave five cohorts in any given year plus all waves of the boost cases. And that together basically gives you a bigger sample for the annual population survey to conduct analysis at lower level. Moving on to imputation, there's two imputation methods we apply on the on both surveys. The rolled forward imputation is basically used as a successful where a successful interview happened at one wave. But then there was a non-response at the next wave and you can identify this with the variable I outcome and where this equals six. You know that there was a rolled forward imputation and that's used on the LFS person and household microdata. Roll forward imputation however only occurs once. So where someone has been a non-respondent in two consecutive waves, we don't roll the data forward into another wave. Dona imputation is basically used where there's been a non-response at wave one or there was a second consecutive wave of non-response. So this is where then a case with similar characteristics is used for donor information. Non-responders can be identified as an outcome equals three and you can find those on the LFS and APS household microdata. So this kind of imputation happens on the LFS and APS. Then over to the data collection. So pre pandemic, our approach was that we did face to face interviews at wave one and then followed up with telephone interviews at wave two. The exception was cases north of the Caledonian canal. They were already interviewed over the phone from wave one onwards for efficiency reason really because it was easier to contact them over the phone. We've got approximately 600 field interviews and 200 telephone interviews to do the field work on the two surveys and we allow proxy responses on this survey and around a third of the data is collected that way. The field process looks like it covers basically five stages. The sample is drawn on a quarterly basis about three months in advance. Sampled addresses then get divided into 13 weekly stints. Stints is basically sort of interview areas and they are allocated to interviews across Great Britain and Nisra conducts the field work in Northern Ireland. The first advanced letter and an unconditional incentive of £10 then gets sent out about 10 days before the start of the field period. And the second advanced letter sent out to interviews by interviews locally about five days before. And then start of the data collection starts, which is normally a week with an additional one to follow up on any interviews. Quick thing to mention, but I think you're going to cover this in one of the practice sessions later in terms of how do I identify. Which questions are asked when certain variables are wave specific others are quarter specific. So in the user guide that contains all the variables or the information about the variables on the data. You have acronym acronyms in the top right corner. So for the first one here you can see in the green circle there's an acronym W1F that basically tells you that the question is only asked at wave one or at first contact. In the other green circle you can see a reference to AJ. That basically means that the question is only asked in the quarter April to June. So the second quarter of the year. So this is just to know sort of what sample size you would have for the question you want to use in your analysis. Also very quickly to mention that obviously we want to ensure that we have good quality data. So therefore we have implemented various things to ensure that people understand the question. So we obviously use show cards for some of the questions during the pandemics. Obviously with telephone interviewing not really the case. So the questions are being read out more, but for face to face interviews that's normally being used. And we have also implemented soft checks and hard checks in our questionnaires to make sure where there's any oddities in terms of conflicts in answers being given that interviews can check those and correct as they go along. Now quite a few things to mention in terms of what's happened since the start of the pandemic. I've got here a timeline that shows you. Sorry I just clicked through. Shows you the changes that happened from the start of 2020. So beginning of March 2020, the public awareness started affecting survey participation. And as of 17th of March, face to face data collection was suspended. So we had to switch to telephone interviewing in Wave 1. Lockdown also then began on the 23rd, which meant that interviews couldn't go out anymore. So we rolled out telephone interviewing to all field interviewers by the end of March. And the problem was also obviously because we are sampling from the postcode address file that we didn't have telephone numbers for all of our addresses. So therefore we extended the telematching process that we normally just do for addresses north of the Caledonian Canal to the entire sample. And in addition to that, towards the end of April, we also changed our advanced material and set up an online portal to enable participants or householders to provide us with their phone contact details via an online portal. And that as well as the telematching provided us with phone contact details for about half the sample. Obviously that then minimized our achieved sample quite considerably. For that reason, we doubled our Wave 1 sample size to make up for that. And we then trialled a field strategy called Noctonedge late in the year on some other surveys first before rolling out in April 2021 on the LFS, which was called Noctonedge, where interviews basically were able to go out again, knocking on doors, not to do face-to-face interviews, but to follow up on respondents where we didn't have phone contact details to obtain phone numbers. Here, this slide basically shows you what happens with our response rates when the pandemic started. So you can see in March 2020 it hit rock bottom really from mid 50s to around mid 20s. And it stayed around that level until we introduced Noctonedge in April 2021, where we saw a 10 percentage point increase. That level then for a while at around 37%. Our response rate saw a bit of a decrease end of last year, but that was not down to Noctonedge losing its effectiveness. That was more down to resources being stretched as well as Omicron affecting interviews going out to do Noctonedge again. This here shows you the achieved sample size, basically also dropping obviously with the start of the pandemic. And when we doubled our sample size with July 2020, we were actually then achieving a sample that was above pre-pandemic level. So we tweaked this later in a year again. So we adjust around the level that we had before the pandemic. What happened though with the start of the pandemic due to the mode switch? We could see that there was certain types of respondents that we didn't see in the sample anymore. So it was a certain level of response bias introduced. So the turquoise bars on the left basically show you the proportion or the distribution of tenure for owner occupies, owners with mortgage and renters before the pandemic. Then the green bar shows you what happened when we switched to telephone mode. So you can see that we then had more owner occupiers and fewer renters in the sample. And then when we introduced Noctonedge, which is the purple bar, you can see that the distribution was going in the right direction, basically bringing more of the harder to reach people back into the sample. And the right dark green bar shows you the proportion of cases that were subjected to Noctonedge. So they're pretty much where we were before the pandemic. So the use of a Noctonedge strategy in the field was a good thing for us as it improved the quality of the sample again. Obviously tenure is something where we normally don't see any movements or very rarely see movements in the data. So because of this response bias, we introduced a tenure adjustment in the calibration of our weights. There is some information on our website. So I've got a link on the slides later on where you can find a bit more information on that. So here is just a few slides that basically show you how our unemployment rate, unemployment rate and the economics inactivity rate changed by this tenure adjustment. So you can see here that all three rates saw a movement there after this was applied. We obviously also with individual characteristics saw some differences in the sample. So we got more older and fewer younger respondents in the sample. And again with Noctonedge this was improved. Similarly, when looking at nationality, we saw that we got more UK and fewer non UK born in the sample. And that was quite an important characteristic that we also needed to address in our weights. And due to the lack of population estimates covering the year 2020, our methodologist had to look at alternative ways of adjusting population movements in our weights. So we wanted to estimate year-on-year population growth in each rolling quarter in 2020 onwards. And the real-time information tax data from the HMRC was the best available data really to help us estimate these movements. Although obviously that's just based on employees, it was the best available data at the time. The assumptions were that the change in population growth rates of the non-UK sub-population is the same direction as the change in the RTI employee growth rate. And that the magnitude of change in the population growth rate does not exceed that of change in the RTI employee growth rate. And here on this slide, you can see what the labour force survey data was before the RTI adjustments in the middle columns. And you can see that the UK born saw quite big increases, whereas the non-UK born vice versa saw very big decreases. So when then the RTI based method was applied, we saw much more reasonable changes in the population over the course of 2020. Again, there's some links in the slides where you can follow up and read up more on this method. And also if you happen to attend the user conference tomorrow, there'll be a presentation about that where a bit more detail is provided. We also updated our questionnaire with the course of the pandemic where we touched upon a few topics where follow-up questions were added to help measure when, for example, people have been on sick leave or they were away from work, worked fewer hours, more hours than usual, etc. That basically these follow-up questions were identifying whether that was down to COVID or the pandemic to basically see whether there is a movement in the time series, whether that was down to the pandemic or potentially to other reasons. And yeah, so that was basically all the changes due to the pandemic. And looking ahead, what's ahead of us, we finished a reweighting exercise in 2021. There will be another one over the course of this year where we basically update our weight adjustment in comparison with the latest RTI data. And here's the link to the methods paper where you can, when you have access to the slides, you can follow up on that. The office is currently also working on a roadmap back to in-house and face-to-face interviewing. We did a small-scale trial, end of last year, and the largest scale trial is planned for spring 2020. That won't be on the LFS though, so we want to see first sort of how this impacts on response rates before we look at when and how to deploy that again on the LFS. There's also ongoing development alongside the LFS on the labour market survey. The labour market survey together with a combination of admin data will eventually replace the labour force survey. This is an online first mix mode data collection and the latest results from the 2019 mix mode test that was published on the website, which you've got on the next slide, a few links to follow up for more information if you like. And again, if you attend to conference tomorrow, then there's a bit more information on where we are with that provided as well. And that is the end of my presentation, so I can take some questions now if we have time before we go to the next slide, to this or to the next session. OK, so I'm Simon Woods, but I work as the introduction said earlier, I work with Martina on the labour force survey and the annual population survey. So this session will basically give a very brief kind of overview to the different types of LFS and APS data sets that are available through the UK data service and the SRS. And essentially as well is kind of what to consider when when looking to do analysis using the LFS and an APS. Essentially what the main purpose of the data sets, some limitations on the data sets as well and some quite fairly basic dos and dos that you should look to follow. And it's probably worth pointing out at this stage that most of the information that I give here will be or is available via the labour force survey user guide 10, which is available on the website. So if you don't read any of the data, any of the user guide, sorry that are that are available, I suggest you read user guide 10 is quite an informative and short user guides. So can we skip through to the LFS person? Martina, please. OK. Yeah, was it the one you wanted? Yeah, that's great. So essentially the labour, the LFS person data sets are probably the most widely used data set that we produce both from an internal and external point of view. And it's really the main source for the labour market overview statistical bulletin that is published each month. And its main kind of analysis is or main use for analysis is really to produce person level statistics such as employment and employment and an economic activity in activity levels, etc. Broken down by personal characteristics such as age, sex, ethnicity, etc. These are produced on a kind of a rolling quarterly basis. So these help publisher, as I said, the monthly labour market release. Externally, we always used to just publish them on a calendar quarter basis. But since the pandemic, due to the importance of the labour force survey, we've now started to publish them every month on a temporary basis until we kind of deemed that. I wouldn't say the pandemic is over, but essentially reporting on the pandemic is kind of slowed down a bit. There are essentially kind of two main weights on the LFS person data sets. And that's the LFS person weight and the LFS earnings weight. And apart from if you're analysing earnings or carrying out earnings analysis, you should always use the LFS person weight when doing any types of analysis. Just a couple of things to be aware of with the LFS person data sets or even the APS person data sets. Person data sets should not be used for any family or best practices, not to use them for any family or household type analysis. And this is really because the person level data sets do not contain everyone that is sampled in the household. They will only contain the people that respond to the survey in that wave. Again, as Martina mentioned, the LFS data sets should only be really used for analysis at the UK and regional level and certainly not below the regional level. And they're also not to be used for any personal wellbeing or sexual orientation analysis. We've had this in the past where people have picked up, I think it was in when they were first introduced, they were available on LFS person data sets. So again, please don't use them for any personal wellbeing or sexual orientation analysis. And this is because they're not asked in every wave of the LFS as Martina was pointing out in her slides earlier. It's just finally on the LFS earnings questions. They are also not asked on every wave of the LFS. So they're only asked in waves one and five of the LFS and that's why there is a specific earnings way to use with those variables. Next slide please Martina. It's already on the longitudinal one, sorry. Yep, that's okay. So the LFS longitudinal data sets, there are two versions of these. There are the two quarter longitudinal data sets and the five quarter longitudinal data sets. But essentially the two quarter longitudinal data sets are the primary source for the labour market's published flow estimates. And really these kind of are helpful to analyse those kind of moving in and out of things like employment, unemployment and inactivity from one quarter to the next or over. If you're looking at the five quarter data sets over those five quarters to see how people's economic activity behaves if you like. And as the kind of title suggests these only contain individuals that are responded for two or five consecutive periods. These, unlike the LFS person data sets are only produced on a calendar quarter basis, both internally and externally. And just a couple of quick things to be aware with these data sets, especially with the five quarter longitudinal data set is that the sample sizes can be quite small. So around the two quarter longitudinal data set is only around 25,000 individuals now and the five quarter is only around 4,000 individuals. So you do need to be careful that your sample size is enough for your specific analysis that you're interested in. And lastly with these data sets they don't also include every single variable that's available. If you looked at the LFS user guides they won't contain all the LFS variables that are available on the LFS person data sets. And this is because they're mainly focused around the labour market requirements for these flows estimates. So there's only around 500 variables I say only. There's around 500 variables on these data sets compared to well over a thousand on the LFS person data sets for example. If you can skip to the APS person data sets slide Martina please. So essentially as Martina pointed out earlier in her slides that the LFS and APS are quite intrinsically linked. The APS is made up partly from the waves one and five of the LFS and then the boost for the APS. But essentially the APS person data sets provide a much larger sample size especially if you want to carry out analysis below the regional level. So down to local authority potentially depending on which variables you're looking to use. But the APS person data sets are really the primary source for labour markets regional level estimates. But they're also used certainly over the last five years or so in more wider ONS publications. Excuse me. Like the personal wellbeing publications, the sexual orientation publications and the smoking prevalence publications. So over the last so many years it's become much more than just what it was originally designed for an employment focus survey. Is much more now like a population focus survey. And this is really due to the larger sample size that it holds, but also the wide range of other variables that it also covers along with these kind of new variables that we introduced. And again, like the LFS person data sets, the main purpose is to produce person level statistics broken down by personal characteristics. So again, you should never use the APS person data sets for any family or household analysis because essentially you might not have everyone in the family or that household on these data sets. As I mentioned earlier, you only have responding individuals. There are four data sets published APS person data sets published every year on a rolling annual period. So you'll have things like January to December. You'll have then they'll move on a quarter and then you'll have the April to March the following year, etc. So there are four data sets published every year for people to use. I suppose the biggest challenge sometimes with these data sets is that there are lots of weights. On the APS person data sets. So please be very careful, especially when you're doing analysis on various things just to make sure you are picking up the correct weights. For the large majority of analysis, most people should just use the APS person weight. But there's things like, as I mentioned earlier, the sexual orientation weights, the non-proxy weights, which is to be used for the well-being and the quality of work variables. And there's also a fairly new APS earnings weight on there on January to December data sets as well, which has been introduced over the last 18 months or so. Again, the main point of these data sets is because of a larger sample size, you can do analysis below the regional level down to a local authority level. But just to be aware that not all variables that are available on LFS person data sets or LFS data sets as a whole will be available on APS data sets. And like Martina Pondidog just now, and this is largely because they might not be asked in all waves of the LFS, but they're also not asked every quarter on the LFS as well. So just to be aware of that, and that's where you need to kind of reference the user guys to try and make sure you're picking up the correct data sets to use. Next slide please Martina. Just quickly, this is a fairly new kind of product over the last couple of years, the APS two year longitudinal. So essentially it has the same kind of methodology and purpose as the LFS longitudinal data sets. It's still really classed as a little experimental, so it was originally kind of brought in to replace the LFS five quarter longitudinal data sets. Because it's such a small sample size these days. And again, this data set contains individuals that have responded for two consecutive years in the APS for January to December periods. But just to make you aware that this isn't yet available by the UKDS, but it is available via the SRS. Next slide please Martina. Again, this is a, I wouldn't say a fairly new product is being available now for a couple of years. But essentially this is quite heavily used, now widely used across within O&S and across government. The APS pool data sets or the three year pool data sets. Again, it's a person level data sets. So you shouldn't be looking at carrying out any family or household analysis using this data sets. But essentially its biggest benefit is that you have a much larger sample size compared to the single year APS. If you ever get access to it, don't think there's going to be three times the size of the single year APS. But it will have a round because of the kind of duplication and the kind of wave design and the rotational design of the LFS and APS. There will be lots of duplicate cases over those three years that we kind of combine together. But you will get a sample size of just under around 500,000. I think it was 470,000 something like that in the last one we published for 2018 to 2020. And the key thing with these data sets is really they should be used just as a point in time estimate. They shouldn't be used really for any time series analysis. And this is because you've essentially when you've got one pool data set compared to the next essentially two years data from those two data sets are the same consecutive data sets. So essentially all you're doing is kind of comparing an APS period three years apart. So it's essentially, as I said, just a kind of time point in time estimate analysis to use the APS pool data sets. Next slide, please, Martina. So the moving on to the household data sets now. So essentially the LFS and the household, sorry, the LFS and the APS household data sets. I really used his main purposes. They are used for the labour markets, workless households, statistical bulletin that's published. The LFS data sets are used for the UK level workless households and the APS is used for the regional workless households statistical bulletin. And again, both these data sets, both the LFS household and the APS household, their main purpose is to try and produce person level statistics, such as employment, as I mentioned, but broken down by characteristics of a family and the household in which people live. So that's kind of a key difference when you're using the LFS household data sets. Essentially, you need to be looking at the different types of families and households in which people live as part of your analysis. Sometimes people get confused when you mention household data sets. They just think it just contains one kind of case per household on the data sets. That isn't true. So these data sets are still at an individual case basis and they will also have the same people in LFS household data sets as the kind of person data sets. But these household data sets will also include the non-responding individuals that I said were missing from the person level data sets. And these non-responding individuals will be imputed for via the donor method that Martina mentioned earlier. Essentially, you can do household and family level analysis because you now have everyone in the family and household on these data sets. So if you ever want to do analysis, kind of a mixture between personal and household level analysis, you should always, if you can use the household level data sets. The next slide, please, Martina. So essentially, I've just said everything I wanted to say about the household data sets. But again, the APS household data set is only published once a year. So we only published them for the January to December period, unlike the APS person data sets where you have four year. And again, these are more for kind of below regional level that you would use the APS household data sets. Next slide, please, Martina. So Martina mentioned the kind of obviously why we create weights on the LFS and APS. Obviously the LFS and APS is just a sample. So we need to somehow kind of estimate that for the people that we don't interview. So the methods for the weight in for pretty much all the data sets is broadly similar. And that historically you're always used a range of characteristic variables in the weight in calibrations, things like age, sex and a range of geographies. But of course, as Martina said, that since the pandemic, we've also added in things like a 10 year constraint and a country of birth constraint. And we've also introduced a kind of a non response adjustment as well to try and tackle the issues that we've come across that Martina was mentioning around the data collection as well. And the key thing I suppose to take away from this from this session, especially with the weights is that please make sure that anytime you've kind of looking at data sets to make sure you're using the correct weight. And obviously this is because the different weights will filter the target population that you're trying to create analysis for. So, you know, so things like sexual orientation, wellbeing and income or earnings weights. If you use those incorrectly, you will have obviously incorrect target population totals. And the reason for that is because they're not asked to everyone, you know, things like sex ID, the wellbeing variables, earnings variables that they're not asked to everyone in the kind of household if you like. So, for example, the wellbeing questions, you can't answer those by proxy, you have to actually just ask the respondent themselves. Next slide, Martina, please. So, Martina mentioned things about the variables earlier. Again, I won't go over it too much, but essentially there are well over a thousand variables available on most data sets, LFS and APS on person and household data sets. Again, these are a range of questionnaire variables, variables that we derive ourselves from what we've already collected to try and reduce the questionnaire length. They also include a range of geography variables to help people kind of do different types of geography regions, etc. Just to be aware that not all, as I mentioned earlier, not all variables will be on every single data set. So, please reference the user guides to kind of give you an indication. We have so many questions, not only coming from externals, but even internals where people have asked for certain questions and finally say why they're missing. And if we looked at the user guides, we could see straight away that they're not asked in the LFS, certain quarters or certain waves, etc. So, please reference the user guides just to make sure that the questions that you're interested in are in these various data sets. Just to be aware that there are what's called missing values on data sets for each variable. So, a minus nine on a data set will mean that that question, individuals weren't rooted to that question, so it hasn't been asked to that individual. And then a minus eight means that they were asked, but they either replied they don't know or refused to answer that specific question. The UKDS or the end user license will include a kind of a cut down and reduce set of variables compared to the more extended version, which is available through the SRS. And the UKDS doesn't have really any actual personal identifiers. They're kind of more randomized. But essentially all this is trying to do is from the UKDS version is to try and protect responders from any identification. So that's the critical thing with the end user license, obviously. Just to add that on the final point on variables is that, as Martina said, that we introduced a whole range of COVID variables in the questionnaire back last April. So initially we did withhold those from any data sets published on the UKDS and SRS, but these have now since been published in recent months, so they are available. Next slide please, Martina. It's already in data pooling. So as a general rule of thumb, because we publish the APS pool data set, et cetera, we don't kind of advise anyone trying to pool data sets together to kind of get a larger sample size. And because this is really because we've come across people getting into real trouble because of the complexity of the survey design and not understanding that. And this is because obviously because of the wave structure, you get lots of duplicate cases from one quarter to the next or one year to the next. But essentially as well, the critical thing is you won't have an appropriate weight to use if you're trying to pool data sets together yourself. And that's all I want to say on that slide, Martina. Thank you. So this is, you know, there is a flowchart in the back of user guide 10, which I kind of alluded to earlier, which is probably a very helpful user guide if you're a new starter to the LFS and APS. And as I said, there's a flowchart at the back which kind of directs you through this kind of slide. So it helps you, it kind of asks you questions to try and pinpoint which data set and which weight you should be using for the different types of analysis that are available. So I won't talk through this slide, but essentially you can kind of feed down and ask yourself this question, all these questions and then the flowchart will should help you. Next slide please, Martina. And again, I think the main thing to take away from is that all this information, there's loads of information out there on our user guide web page. I know that there are lots of endless user guides for people to kind of read, so it can be quite daunting, but essentially all the different user guides that are available are listed on the screen. Some are extremely long like a user guide one, which goes into extreme depth about the LFS methodology. Then there are kind of more useful and probably useful user guides, things like user guide two, three and four, which contains the information that's asked in the questionnaire and the flowcharts for things like derived variables, how we kind of create those derived variables. But there's a whole range of different types of user guides there depending on which analysis you'd like to do. But essentially the critical things is, you know, use the correct data set, use the correct weight and obviously make sure that the variables you use in what you think they are and that they available for all the periods you're expecting them to be available for. I think that that's it from me. Thank you.