 Hello everyone and welcome back to the conference. So this is going to be session 3B, Long-Term Trends and Structural Dynamics. So we have three speakers today and they're going to be talking about long-term trends in part-time work in the UK and disability employment gaps in the UK and the role of education and they're measuring the impact of Brexit on migration to Wales. OK, so our first talk today is by Rachel Scarf from the University of Edinburgh. Rachel is an early career researcher and she researches labour economics focusing on non-standard contracts such as zero-hour contracts and part-time work. She's particularly interested in the interaction between firms, technologies and workers' preferences and how this affects labour markets. She also has a side interest in studying labour markets in sports to see what they can tell us about labour markets in general. OK, so over to you Rachel. Great, thank you very much. Yes, as was just said, this is a paper about long-term trends in part-time work in the UK and it's kind of part of a general research interest of mine which is different patterns of work and how those have been changing over time. So there are two parts of this paper that I'm writing at the moment. The first part is to document some facts about trends in part-time work and I've kind of split them into two categories. There's kind of fact one, which is that there's been this long-term increase in the quantity of part-time work in the UK and I'll talk a bit more about that, what that means in a minute. And then fact two is that at the same time there's been an increase in the relative price of part-time work. So part-time workers now earn more per hour, obviously less absolutely because they work fewer hours but they are more per hour relative to full-time workers than they did previously. And then the second part of this paper, which I may have time to talk a little bit about today, is to build a kind of structural model which incorporates part-time work. So what I want to do is to kind of, these are two facts that aren't necessarily, generally we think that if the quantity of something goes up, the price might go down and that's not what we see in the data here. And so the aim is to build a structural model that can perhaps explain workers and firms' preferences for part versus full-time work and then perhaps explain whether these changes that we see are something to do with changes in workers' preferences on the supply side or whether they're changes from firms kind of on the demand side. And the reason that this is kind of important in many ways, obviously, I think that, but particularly I think it's got a kind of theoretical importance because a lot of the models that we have as economists of the labour market assume that earnings are linear in hours. So if you work another hour, the extra amount you get doesn't change if you're already working 10 hours a week or if you're already working 40. And as I'll show you, that's just not what we see in the data and therefore maybe our models are not capturing something important. And it has a policy importance to this question and this topic, I think, as the inequality in hours, if more people are working part-time, that could feed through into inequality in earnings. So there are policy reasons why we might be interested in these trends too. So for the empirical part of the paper, I used the quarterly labour force survey and that's kind of ideal for this purposes. We've got this long term data, so it covers the whole business cycle, so if you want to look at trends, that's quite important. One thing that has the other data set I work with is the ASH, the annual survey of hours in earnings. And that doesn't have the kind of individual characteristics, so things like education and particularly important, obviously for part-time and full-time work is kind of family situation, which we see in the LFS. And it also captures, the ASH sort of fails to capture the very low wage end of the labour market, so people who aren't earning enough, for instance, to be paying income tax are less well captured in the ASH. And those people are generally people who work part-time. The people that I kind of focus on here, so I define someone as being part-time if they work 30 or fewer hours a week. And everything I'm about to show you is kind of, is not sensitive to exactly where that cut-off is. So I've tried everything using 35 hours a week as the cut-off, everything using 25, and I find the same trends. I look at people aged between 16 and 64, so that's a whole of someone's working life. Again, if I narrow that to 25 to 55, again, I find the same trends. And I just look at people working between 5 and 70 hours a week, so not people doing very, very few or sort of implausibly, who report that they do implausibly long numbers of hours a week. So the first fact that I said I would talk about was this increase in the quantity of part-time work. So what we see is these three charts show for men, for women, and for everyone what we call a part-time share. So that's the percentage of people who are working with a part-time job. And you can see very, very different trends for men than for women. For men, that part-time share has increased from about just over 6% in 1994 to over 12% today. For women, the trend has generally been downwards with a little uptick during the financial crisis. So that's the first key fact, is that there's been this increase in the part-time share. And as well as looking at part-time versus full-time dichotomy, I also looked at what kind of hours those part-time workers were doing. So this is the proportion of part-time workers doing very, very few hours on the left, so less than 10, and doing between 21 to 30, so more hours on the right. And you see that fewer and fewer people are working very, very low hours and more people are working two or three days a week. And that trend has been carrying on to increase even after the financial crisis. And a consequence of those two facts is that in the economy as a whole, more of the work that is being done in our economy is being done by part-time workers. So this is what I call the part-time share of all hours. So all of the hours of work that are being done, and then I work out what percentage of them are being done by part-time workers. And you can see on the right that it's been increasing, sort of jumped in the financial crisis. It's not clear if this is kind of that fall since then is kind of going back to kind of a trend up or if it is going to be a fall. And crucially you can see that that trend is driven almost entirely by increases amongst men. So this is really a story of men working, more men working part-time, and those men working conditional on being part-time, working for longer hours. So that was kind of the first fact that I found was of interest, was this increase in the quantity of part-time work. And then the second fact that I found was interesting was a change in the price, so a change in the wages here of part-time workers. So all that I do here is I regress hourly wages. So hourly wages on whether or not someone is working part-time. And so these three lines plot the coefficient beta. So it's basically the difference in average hourly wages between part- and full-time workers for men, which is a black line, for women, which is a blue line, and then kind of an average for everyone, which is the red line. And that's what we call the part-time pay penalty. And you can see that it's big, particularly amongst men. And that's because part-time men tend to be working in very, very low-wage jobs. And then at the other end of the spectrum, very, very high-wage jobs, which tend to be full-time, are mostly done by men. And so there's a bigger part-time pay penalty on average for men. But part of that is just due to the kind of the different types of job and occupations in industries where you find full and part-time workers. So what I then do is I include a set of controls where I control for kind of work characteristics, so age, education, experience, all these kind of things, and firm characteristics, and then job and occupation. And I call that the adjusted part-time pay penalty. And you can see that it's much smaller and it's very similar for men and women, but that even for very similar people working in very similar-type jobs, full-time workers are kind of catching up, their pay is catching up, hourly pay, sorry I should say, is catching up to full-time workers. And that's the kind of, the second thing that I think is quite interesting to think a little bit more about. So then the kind of question kind of sort of comes is where does that part-time pay penalty come from? So where is this kind of jump in hourly wages? So what I do here is I look at sort of weekly earnings. So it's earnings rather than hourly wages now. And I regress that on a dummy for kind of bins of hours. So for everyone working five to 10 hours I put them together and 10 to 15 and so on and so forth up to 65 to 70. So that allows me to kind of plot out an hour's earnings profile. And you see that for people doing kind of fewer than about 30 hours it's kind of linear. So that sort of means that their hourly wages constant. If you earn, if you work at an hour more you earn however much more. But then between about 30 and about 35 there's suddenly a big jump. So at the point which we kind of tend to traditionally think of as the full-time number of hours suddenly people's earnings increase very, very quickly. There's this kind of discontinuity. And that's the kind of... And what is interesting about that is that this is... So on the left we have kind of in the 19th... So I did this by sort of the periods where we have the different SOC codes so that I could kind of control for occupation. Kind of consistently. So this is the kind of the 90s on the left and the 2010s on the right. And that shape of that hour's earnings profile is very similar. That hasn't changed very much. So though the part-time pay penalty has changed that discontinuity is still there. And what I want to sort of where I'm moving now is to think about what it is that creates that discontinuity. So we know that part-time work is very concentrated in some industries, hospitality, service. And there's evidence to show that it's associated with lower occupational skill requirements. Jobs with lower skill requirements are more likely to be part-time. And I want to think about if it's related to the different types of tasks that part-time workers do at work. So maybe it's that jobs with more sort of coordination and analysing longer tasks are more likely to be full-time. And maybe those with shorter, more defined tasks are more likely to be part-time. And what I've started to do, and I'd be really interested to hear from anyone who's thought about this before, is to link the occupational data in the LFS to this O-Net, which comes in the States, database of occupational task requirements. And then sort of beginning to investigate which types of tasks can predict whether someone works part-time and which types of tasks are associated with a higher part-time pay penalty. Yes, and then I think I've got a few minutes left. So I'm just going to briefly talk about the kind of structural model of part-time and full-time work that I'm building. And as I said, the aim of the model is to explain both workers and firms' preferences for full-time versus part-time work and to incorporate that shape of the wage distribution that we see, the earnings distribution that we see. And then the aim is to use that model to try and think about whether the changes that we see are driven by changes in worker preferences, so maybe people want more part-time work than they did, or whether they're driven by changes in firm technology. So maybe the kinds of work that we have now are just more kind of amenable to being done part-time and firms are demanding more part-time workers. So just to give you a very quick, and I'm not going to kind of show you the model in any detail, but to give you a very quick idea of how it works, the idea is to simplify things and to say that firms produce using a combination of output from two occupations. And there's a kind of, I call it, I don't like that term, but I call it simple occupation, where all the tasks are very well-defined and they all have the same length. And then there's a kind of more complex occupation where tasks are less well-defined, they have a sort of length that can vary and is uncertain. On the firm side, it's a kind of near-classical model, firms decide how many workers to employ and how many hours, and workers decide which occupation and how many hours. And then market clearing gives you some equilibrium earnings in both the occupations. And what I find is that people who, in the model, people who want to work part-time because they've got high disutility of labour, they sort into those simple occupations and they work for fewer hours in those simple occupations. And then people with a high disutility of labour, so with a low disutility of labour who want to work more hours, they sort into the more complex occupations and they work more hours in those and they earn a higher hourly wage and that's where that difference, that part-time pay penalty comes from. And then the idea is to calibrate the model and to look at how the changes in parameters can drive changes in the relative wages and the relative quantities of part-time and full-time workers. So just to kind of plot that out, perhaps to make it a little clearer. So this is in the equilibrium of the model, this is the earnings as a function of hours. So as you work more hours on the horizontal axis, your earnings are on the vertical axis. And for the simple occupations, that's that straight line, those earnings are linear. But for the complex occupations, those earnings are convex. So as you work more hours, your hourly wage increases. And then the green lines here are some utility and difference curves for some workers. So U1 is the utility and difference curve for a worker with a very high disutility of labour. And you see that for them it makes sense they earn more if they work in the simple occupation and they do that for a few hours. But then U3 is the utility and difference curve for someone with a very low disutility of labour. And you see that they can earn more if they work in the complex occupation and they choose a higher number of hours and their hourly wage, which is higher. And so this gap between the two is the part-time pay penalty that we observe. Yes, so just to sum up. So the conclusion so far of this project is that there has been this long-term increase in both the quantity and the relative price of part-time work. Those trends are much more, that's being driven mostly by men. And that suggests that maybe there are differences in preferences for men and women that might be important. And then what I've done so far is to create a model that can replicate that part-time pay penalty and kind of suggest how people might sort into full-time and part-time work. I've talked a little bit about my plans to extend the empirical work and I also plan to extend the model. So at the moment everybody in my model has the same productivity, but in fact obviously that's not the case. And so extending it in that would allow people to work for a long hours in low-pay occupations, which we do see in the data. It's not the case that everyone in low-wage occupations works for a few hours. And then the other extension is to incorporate gender differences in preferences. So maybe then we can start to explain why some of these trends are different for men and for women. But yeah, but that was all that I wanted to talk about today. So thank you very much for listening and I'll be interested in any comments or any questions that you have. So thank you. Thank you Rachel. That was really interesting. I learnt a lot. So we do have a question from Andrew. He says, thank you for your interesting presentation. And he has two questions. So how are you going to convert the U.S. omet data into UK SOC codes? Yes, that turns out to, yes, an excellent question. So what I do at the moment is I map them to ISCO codes and then I map those ISCO codes to SOC codes. I'm aware that that's not ideal. I don't know if that's something that you've worked on at all. I'm definitely looking for a better way to do that, for sure. I think there was a second question. Have you looked at the part-time workers have a second job? So I should have put that on the kind of extensions. The percentage of people with a second job is fairly stable. But one thing I want to look at is whether the kind of the total hours for those people has been increasing or decreasing and whether it's the case that people are increasingly moving to having one main part-time job and then a smaller part-time job on the side. Obviously that then feeds into all sorts of things about something that is very prominent at the moment is the idea of the gig economy and whether people have a main job and then a gig job on the side. So that's definitely something I'm looking into at the moment. Fantastic. Well, I think it's time to move on. But thank you very much for your talk, Rachel. Thank you. OK. So our next speaker then is Mark The Ryan. Let me have a look. So yes, Mark is going to be talking about decomposing the disability employment gaps in the UK, the role of education. So Mark is a reader in economics at the University of Sheffield and his research centres on health wellbeing work and is based on the statistical analysis of large-scale datasets. And he's currently leading an awful foundation project on unpacking the disability employment gap. OK, then. So when you're ready, over to you, Mark. OK, thanks. I'm probably looking at the wrong screen because they say it's on my right-hand screen, which is not where the camera is. But anyway, we'll carry on. So yeah, this is joints work with Andy, who just asked the question there. Jenny and Christina, my colleagues here at Sheffield. And Andy actually did all the analysis in the Secure Lab. So this is part of a large enough field project which is about unpacking the disability employment gap. And what we're doing here in this particular analysis is focusing on the role of education and asking the question potentially how could changing education affect the disability employment gap. So what is the disability employment gap? It's the difference between the employment rates of disabled people and non-disabled people. So if you take 2019 before the pandemic in the UK, among working-age non-disabled people, 81% of them were employed, whereas the employment rate was only 53% among disabled people. So the difference between those is this 28 percentage points disability employment gap. And what we aim to do in this, and this is quite a hot policy issue the government's concerned about it. They in the past had a target to halve the disability employment gap. Disability rights advocates and charities are also interested in it because they want to get rid of the structural barriers to employment of disabled people. So what we're aiming to do in this particular analysis is to try and disentangle this gap into the bit that is because people are different. So the characteristics of disabled people are different from potentially different from the characteristics of non-disabled people in addition to the difference in their health status. And secondly, the part that is because people are treated differently or behave differently in the labour market. And so we're calling those differences structural differences or structural barriers. And we are focusing on education because this is a so-called modifiable characteristic that also is itself the target of policy. So disability is a protected characteristic under the Equality Act of 2010. So it's illegal to discriminate in employment matters. But nevertheless you can see that on that diagram there the disability employment gap is much bigger than the equivalent gaps between men and women or between white people and people from ethnic minorities. So why is this? Could it be that despite this legislation, this discrimination in the labour market, that would be an example of a structural barrier? Could it be that having a disability inherently makes you less productive and employable than not having a disability? Could it be that disabled people tend to have lower qualifications and skills than non-disabled people? So that's something that we will be able to address in this research looking at education. Or could it be that disabled people are employable, just as employable as non-disabled people, but it's just that there's a lack of workplace accommodations and other adjustments in the wider society which prevents them from taking up jobs. So that would be another structural barrier. And then finally, could it be that disabled people are less likely to participate in the labour market? Choose not to participate in the labour market. So they're less likely to be in work than non-disabled people. I don't think anybody's, and nobody's suggesting that work is appropriate for all disabled people. So there may be an element of a decision or choice in this gap. So we'd never expect that gap to be down to zero. This shows how that gap varies over different educational qualifications. So you can see here there's a very sharp educational gradient going from, if we take people with no qualifications, then the disability employment gap is 40 percentage points and it goes down to 14 percentage points for people with a degree. So those green figures of the 2019 figures, if you just take the ends of that, you can see that actually the gradient got even steeper between 2014 and 2019. So clearly education is playing some kind of role here because it's this very steep educational gradient. If you take those figures and you weight them or multiply them by the percentage of people in each category, then if you add those segments together, you'll get the full disability employment gap. So that's what we've done here. There are now 11 categories. These are the 11 academic and vocational qualification categories that we're going to be using in the analysis. But for example, if you take that top left segment, the 25%, that's a function of the 40 percentage point gap for people with no qualifications that I just showed you and the number of people who are in that category. So one way to think of this is that if we eliminated the gap for people with no qualifications, then that would shrink the overall disability employment gap by 25%. So this is one way that can kind of inform how we might want to target policy on different groups. And I'll come back to this later on when we talk about the structural gap. So this is the overall gap at the moment. The data that we're using are from the APS in 2019 and we're not using the full working age population because we only want to look at people who've finished their full-time education. So we're limiting it to people who are 25 or older. And that means that our disability gap will be a bit larger than the 28 percentage points that I had in the earlier slide because the older you go, the bigger the gap. So it'll turn out to be 33 percentage points. We're defining disability using the two questions in the LFS APS that enable you to construct the same definition as in the Equality Act. So that's whether you have a long-term health problem or illness and whether that reduces your ability to carry out day-to-day activities. We're including self-employed people here. So it's whether someone is either an employee or self-employed. And we have our 11 academic vocational qualifications which I just showed you in that donut. And then we're including a bunch of other demographic characteristics that are listed there at the bottom of the slides. Now, I can't, I don't have time to go through this in any sort of detail, but basically what we're doing here is no AXACA-lined decomposition. So we're basically estimating employment equations that show how much qualifications which are the cues and the other characteristics, the Xs, how much they affect people's probability of being in employment and the betas there are those effects or the returns to those characteristics. We then take the averages after estimating the equation and take the difference of those averages which gives us the DEG. And then by rearranging these algebraic terms, you can apportion parts of the gap to differences in characteristics, so education and other characteristics or the differences in these returns, the betas. So the impact of the different characteristics and that's what we're calling the structural component. Now there's lots of different ways you can actually do this because ultimately it's just kind of, it's almost like an accounting exercise or a statistical decomposition. And one contribution of our paper is we're trying to clarify how you actually interpret these things. So what I'm going to do for, I haven't got time to discuss that now, but what I'm going to do for the rest of the presentation is whiz through some tables and just select particular numbers and ask you to take it on trust that those are the right numbers to select and the most meaningful figures to give you. So this is the first table and the key numbers here are the differences in the means, looking at the differences in qualifications between disabled and non-disabled people. And really the big differences are at the two ends here. So what you can see is that for, if you take degrees, 39% of non-disabled people have a degree and it's only 24% of disabled people. At the other end, if you take no qualifications, 6% of non-disabled people don't have a qualification and that's 17% among disabled people. The other qualifications in the middle are much more equally distributed. So it's really at these two ends where we see this big difference. So the question for us is, is this big gap in degree level qualifications and no qualifications, is that driving the disability employment gap? So this is our first decomposition table and we'll give us some information on that. So what you see there at the top is that there's overall this 33 percentage point gap and if we look at the proportion, the part of that is due to differences in education, that's just four percentage points, which is 12% of that gap. And that's driven, as you would expect from that previous table by these differences in degree and no qualifications, more or less equal contributions there. Another way to think of this is in terms of a sort of counterfactual or a kind of thought experiment, policy thought experiment where what we try and do is increase disabled people's qualifications to the level of non-disabled people, leaving the structural barriers in place. And if we did that, this is telling us that it would reduce the gap by four percentage points or 12%. If we do another thought experiment and forget about increasing education but think about trying to dismantle some of the structural barriers, then what this column here tells us is that we could potentially reduce the gap by 28 percentage points. So that's 85% of the gap, although that does include some of the preferences to work that I talked about earlier, which I'll come on to later. So obviously that's a huge amount. What we can do is look at how that 28 percentage points is distributed across the different qualifications. So the left-hand column here is essentially doing a similar thing to what we did in the donut earlier on, where we look at how that structural gap now is distributed across qualifications. So what the number there at the top for degrees says is that if we took people with degrees and we were able to eliminate the structural differences or the structural barriers, then that would take 4.6 points off the total gap, the total structural gap of 28 percentage points, so that's 16% of that gap. If we did the same thing for people with no qualifications, so if we took the 17% of disabled people with no qualifications and removed the structural barriers, that would get rid of 6.5 points or percentage points of the gap, so 23%. So again, that's telling us something about how we might want to target different policies. Now that structural gap I've just been talking about there, as I said, includes preferences. And to get a handle on that and how much they're actually coming into influencing this, we've tried to narrow it down by defining a so-called involuntary disability employment gap. Now we're not too sure about the name here. We've in the latest iteration of the paper calling it a want to work gap. Because the way this is defined is we now do the same analysis, the same calculations. But for people who are not working, we only include people who are ILO unemployed, so by definition they're looking for work. People who are not ILO employed, they don't satisfy the full criteria, so they're inactive, but they're still seeking work. And people who are inactive are not seeking work but say they would like work. And these are taken from this variable in LFSAPS INEC AC05, which is a big list with this multitude of different economic activity categories. And so what we can safely say is that in this sample, calculating the gap in this way, choice shouldn't be coming into it because all of these people say they would like to work. And you can indeed see that the gap is much smaller. It's now half what it was before overall. But if we look into the decomposition, we still find that 92% of that is structural. So despite the fact that choice isn't playing a role here, we're still finding that 92% is structural and in fact that's higher than it was before. And looking at education, if we tried to equalise education, we'd closed the gap by 8%. So again, in fact that's smaller than it was before. So this is some evidence here that it isn't just choice. That's affecting whether disabled people get into work or not. I'll skip that slide for reasons of time and just go on to the conclusions. I guess we haven't really, at this stage, got a list of policies. We're not quite at that stage yet. But I think this is providing some useful pointers for policy. If we just look to education and didn't try and reduce structural barriers, we could make a dent in the gap, but it wouldn't be a huge dent. We're talking about something like 12%. It would be bigger for women, which I haven't had time to show you, but essentially that overall conclusion doesn't change that much. So what we're seeing here is the importance of these so-called structural barriers that make up about 85% of the overall gap. And even when we go to the so-called involuntary disability employment gap, then we still find it that that's driven by these structural preferences. We can look at how the structural barriers differ over qualifications, and what that shows is that if we address them for people with no qualifications, then that would probably have the greatest single impact on the disability employment gap. So I think I'll finish there, and that's the link to our main project there, if anybody's interested in finding out more. Okay, great. Well, thanks very much then, Mark. And I will move on now to our last speaker of this session. So, Noel Gin, who's also known as Dan, comes from the university. He's coming from the University of Cardiff University, and he's a PhD in economics candidate at Cardiff Business School. And his PhD research topic is measuring impacts of Brexit on migration and regional economic growth in Wales. So he holds an MSD in economics from the University of Illinois and is an active member of the Regional Studies Association. Okay, so over to you, Dan. Thank you so much. Hello, everyone. I'm Daniel. I'm currently a PhD candidate at Cardiff University, and it's been an honour for me to have this opportunity, and actually today's top part of my presentation is measuring the impact of Brexit on migration in the case of Wales from 2016, referring to 2022. So I will divide my today's presentation into the five parts. The first part is introduction. So in this section, I'll introduce the history of Brexit as a political and economic agenda, debates regarding Brexit and its migration issues. Very briefly, and most importantly, reasons why I chose Wales as the very case of my research project for the PhD thesis. I'll start this section with a quote from former president of the European Commission by Jean-Claude Juncker, who said that, of course, Brexit means that something is wrong in Europe, but something is also something I was wrong in Britain. Then I'll talk about the history of Brexit from the past to the present very briefly. So as we can see, that in this section, I'll talk about the history of Brexit on this figure, which shows the data of before the 2016 referendum. I would say that it's just like a type race, because you can see that the approval rates of remaining EU and leaving EU were quite close. Basically, the proposal of Brexit could date back to a 1975 referendum of leaving EC, the European Commission, the predecessor of EU. And besides, Brexit began to grow since 1990 and reached an extremely high level since 2004, a great enlargement, which included six new countries in EU. And then, the 2016 referendum just decided the Brexit to become a reality. And finally, we have experienced a transition period in time. Why migration matters to Brexit? So basically, as we can see in this figure, Common Go, migration to and from UK during this period, we can see that a number of EU migrants to UK increase very rapidly, and EU migrants started to occupy more than 50% of total migrants to UK in 2014, roughly right here, right after last two EU expansions in 2007 and 2013. So some economists claim that the migration issue ran. And some economists claim that this trend, this historical trend could explain why the migration issue ran top three with high weights of new EU Brexit arguments many posts before the referendum. So how did Brexit reshape the migration policy here in UK? And so, according to the final agreement between UK and EU regarding to Brexit, all migrants to UK are now divided into three categories as shown in the table. So basically, all EU migrants now have to apply for point-based workers visa after Brexit. It is very important to notice that before Brexit, they didn't have to apply for workers visa to get a migration status. And so for only non-EU migrants earning no less than £25,600 per end can apply for the point-based workers visa. Basically, many economists also claim that migrants from EU countries to UK are estimated to be more impacted by Brexit-related restrictions on migration than those from non-EU migrants. And that's by the least so we have experienced that 2020 COVID-19 pandemic and more importantly 2022 Russian invasion of Ukraine that could bring unexpected impacts on migration to UK and Wales as well. So then I will explain why I chose Wales as a case of this, of my research project. I will give major two reasons. First reason is a strong economic and trade relationship between Wales and EU. So as we can see both in both charts that the strong trade relationship between EU and Wales exceeded almost all other economies. So Port is in turn to claim that Wales has one of the strongest trade and economic ties with EU region among all UK subregions such as Northern Ireland and Scotland. And the second part is just like just like what many economists have claimed including Hoover in 1969 claimed that migration decisions are highly motivated by economic ties between moving and move out countries. So as we can see in this chart showing the number of migrants to Wales from 2010 to 2020 we can see that the Wales and EU share a close relationship regarding migration and Wales has become according to the data from Stetswels Wales has become the second most popular destination for EU migrants to UK since 2005. Finally, a country literature on Brexit's impacts on migration to Wales is quite inadequate. I have to admit then so I hope to and contribute to this topic with more empirical results. Finally my research questions are these three. First one is how did Brexit impact migration patterns to Wales after this referendum and can these impacts being differentiated with respect to other external shocks such as the pandemic and the war in Britain. Second, to what extent will Brexit impact and reshape future migration patterns to Wales? And then finally what public policies, especially migration related to administrative policies that Welsh Government implement to assure a health so-called healthy and balanced migration governance proposed by the Council. And so then I'll go through the two major models and methodologies that were used The first one is to review previous data and to check whether Brexit had impacts on migration to Wales. So basically I'll use FEBD based gravity model. So it's based on a station form of a classical gravity model of migration just like the general form of regional model is shown below. So we can see that MIJ is migration from region I to region J. So PIPJ are population sizes for both regions and DIJ assistance. So SIJ means pole factors and SSI means pole factors consistent with the definitions of classical gravity model. So then take the log linear form and look at this equation and then we're going to FEBD methods. So basically Palmer and Choker developed this method fix effect factor decomposition method to allow estimations of time variables in my model is just like which is very important such as in distance between moving and move out regions based on FEM setup. So according to my previous calibration of this model, so this FEBD method is highly consistent with the kinds of my model especially for some variables such as distance and location which makes it a half-wise linear. Finally the empirical model is just like a force. So here P means population, D is distance, U is means unemployment rate I here is move out regions J as well. So G is GDP per capita in both regions and EU this is very important variable which checks whether the country of the move out country belong to EU or namely had the free movement rights to UK at time 2 minus 1 and here COVID this variable is also very important. It checks whether COVID pandemic had some indiginated bias to the process of regression of this program. So as I have mentioned MIJT will be divided into three groups and regression analysis repeated four or three times in accordance with top three industries receiving most migrant labour in Wales to check whether there are some heterogeneity issues in different industries and then to forecast the future trends of migration to what I use in Niger model co-developed by the National Institute of Economic, Social Research and College Business School and it also uses the spatial DSG approach so all the variables used in Niger model are identical to those used in FEDB-based gravity model except one extra dummy variable which is Ukraine IT minus 1 to check whether the move out country is continuous to Ukraine because numerous papers have estimated that Russian elevation of Ukraine could last four months causing possible impacts on regression decisions. So basically we're going to data. I use these four major data sources LFS data migration and survey data from Wales, from Wales, the virtual COVID-19 database, World Bank database and finally in this section I will introduce three major parts of results so descriptive statistics and empirical results of those models. First is descriptive statistics so note that all descriptive statistics about have not been localised but localised values were applied in modeling to make it easier for reaction analysis and to eliminate any statistical bias. So as we can see this it shows the result of FEDB-based gravity model and first generally more people in the move out country contribute to more migration from this country to Wales and you can see the co-efficient of population mobile PI in three regression analysis processes were all positive and then distance discourage migration because the coefficient were all negative and then higher employment rate in the move out country contribute to more migration to Wales and then migration to Wales tended to be from regions with lower GDP per capita to the region with higher GDP in Wales and then note that this second column is to test whether the EU free movement had impact in the process of Brexit on migration and to control for other variables right here. So basically results show that EU free movement attracted migration as shown in this table and for the second column shows the the second column tests whether COVID-19 pandemic imposed any endogenity bias or had impact on migration decisions to UK and without show that COVID-19 search cost reasons of not. There are some discussions so basically I also conducted further analysis on whether Brexit led to changes in the number of EU on EU migrants in three categories and top three industries namely manufacturing industry education of Wales as mentioned. I repeated the question procedure four six times in total and results show that number of EU migrants earning less than 30,000 pound per anime and manufacturing industry decreased most significantly due to Brexit related migration. Also continue to test whether the 2016 referendum cost perceptual concerns of publishing EU and UK free movement also inserted another dummy robot just like EU IJ to test whether the mobile country belong to EU at time T. So we will show that the referendum contribute to a rapidly increasing the number of EU migrants in Wales and such cause of effect is statistically significant. Also to test the robustness of whether COVID-19 brought the endogenity issue out instead of inserted three IV but no statistically significant coefficients for these three IV were found showing that COVID-19 did not cause endogenity to changes in migration. Next part I use the nitrogen model to forecast the future trend of migration to Wales showing you start from 2021 to 2026. It shows that EU migrants are expected to increase rapidly in 2022 and remain at a high level of increase until 2024. Also for non-EU migrants are expected to increase steadily from 2021 to 2025 and express slightly increasing. Yes. Finally I also use the ArcGIS and GeoNames tools to give a big picture of geographic divergence of migration to Wales in the future in a given period from 2020 to 2026. So here we can see that in major cities and economic centers in Wales such as Cardiff right here, Newport right here and Swansea all of the cities are expected to receive more EU and non-EU migrants in the future during this period. There are also some discussions previous literature also noted that the industrial heterogeneity could lead to regression in the focus bias but for my research project the conversion of Welsh-based Niger model has not included such approach or data to forecast how migrants would change migration decisions to Wales. Basically if I would like to conduct such research that requires difference in different method and especially more importantly the firm or individual level data in the near future and individual level data so basically Niger model will include this data in the near future. Okay, so I'll conclude my presentation in these four points. First one is Bresid. According to my empirical analysis and empirical analysis, Bresid had causal effects on the reduction of EU migrants to Wales in major industries especially after the referendum. Second, Bresid did not impact non-EU migrants to Wales in major industries possibly due to the plans of losing immigration policies such as the extension of PSW visa in 2018 and the start of BNO visa application from Hong Kong starting in term time. The second one is EU migration is expected to be encouraged by 2022 Ukrainian war and according to many economists and researchers it could be this war could last for at least one year due to the current situation. So I expect that the impacts of Ukrainian war on migration to UK could also last for at least one year due to the preparation. Finally my suggestion is just like the Welsh Government maybe have to cooperate with UK Home Office to develop a balance and merit-based immigration policy that meets the requirements of economic growth and industrial transformation and more importantly the different the dynamic the dynamic structures of migration to Wales. Thank you so much and that's all of my today's presentation. Please refer to your questions and