 Hello everyone, welcome back to session 2. I hope we all had a nice break in between these two sessions. As Jen mentioned earlier, my name is Emma Barnes. I'm one of the coordinators from the NS side for this conference, and I'll be chairing the next session, which is the first of our research paper sessions today, covering the topic of debt and wealth. So we've got three wonderful presentations for you today. Each presentation will last around 15 minutes, with five minutes immediately after for questions. I'll let all presenters know when you have around two minutes left on your presentations. I like the last session, can all questions be put in the chat bar for me to collect and ask the presenters at the end of their presentations? So our first presentation comes from Ricky Kanaba from the University of Bath on parental home ownership and education, the implications of offspring, wealth, inequality in Britain and Great Britain. So Ricky, if you'd like to share your screen while I just do some quick introductions. So Ricky Kanaba is one of the associate professors of social policy at the University of Bath. Ricky is an applied economist whose research interests span inequality, aging and health. So over to you Ricky. Thank you very much. Thank you. So yeah, indeed, this is a joint paper with Paul Gregg, and here are interests in looking at the role of parental characteristics, in particular, home ownership and education for offspring, wealth, inequality, both across cohorts and over time, to give you kind of a precursor of the results. This is, we do find a strong growing relationship between parental characteristics and offspring housing wealth in particular. There's not really a day which goes by, where I don't see something about interest rates and housing at the moment. So without further ado, I am conscious of time, so I will flick through certain slides and slow down for others. We're very interested in how early life circumstances characteristics, which are in essence out of the control of the individual, the offspring, are related to subsequent life cycle decisions like home ownership, but also more generally living standards and thinking about how individuals use or accumulate particular types of assets. So there's lots of different strands of literature. We're talking to the first one on wealth inequalities, but also on intergenerational persistence and the role of family backgrounds and I've done some work on this recently, but also many others, possibly in this room, but also further afield. We're also interested as I said, the types of assets which individuals hold. Do you own your own home, with or without a mortgage? Do you have pension wealth, what level of wealth? So there's also a literature on wealth portfolios as well, both, again, across cohorts and over time. So I'll be brief on that. What do we find or what do we do, should I say? We document the components of wealth which drive these intergenerational wealth correlations. So I'll talk a bit more about what I mean when I say intergenerational wealth correlation in a few slides, but we're interested in precisely what types, sorry, what parental characteristics are increasingly related to individual offspring holding certain types of wealth, housing wealth, pension wealth, financial wealth. I won't have time to cover everything in the paper. I'll focus mostly on home ownership and housing wealth today because that's indeed what's driving the main story and it's probably of interest to this audience. So I have to apologise as a typo on this slide than in subsequent slides. We also do a back of the envelope calculation to try to understand how quickly this intergenerational association or link between parental characteristics, parental characteristics and offspring wealth is changing and actually our rank estimates that imply doubling in about just over 30 years. Okay, so I should say again, heads up, home ownership opportunities and housing wealth are increasingly related to parental characteristics. So maybe a surprise, maybe not a surprise. To put things into context, this is just a picture which is from a, not from a paper myself by Elvarado and his co-authors a few years ago, which just shows you a snapshot of long wealth inequality over the 20th century. This is the top 1% and essentially shows you a strong decline until about the 80s, 1980s and then sort of a reversal and there's a steady increase again. So we've seen this change and our study essentially is kind of from here onwards and we're interested in particular about breaking down the certain types of wealth, so breaking down total net wealth to understand how that's related to parental background. So we're thinking about social mobility a little bit here, but it just gives you a context of what's happened in the very long period prior to our analysis. So I want to stop on this slide for a moment because this is an important slide. What we do here is we use wealth in asset 7, I'll come onto the data set in a moment. We on the x-axis have individuals age and we take different co-awards with different individuals at certain ages and we use the panel dimension of WAS, which is really helpful because it follows individuals over time and allows us to look at their wealth holdings. And on the y-axis we've got the proportion reporting home ownership and what we do is we split this out by the parent's characteristics. So this isn't individual's own characteristics, this is their parent's characteristics. So let me give you an example. This is following individuals over six years, so between way 3 and round 6 of WAS and it's telling us for people who are 31 that started the sample at the end of 37, about 30% essentially report home ownership, a report having some housing wealth and their parents are low educated renters. Let's now flip that around or let's contrast that with people whose parents were high educated homeowners and that trend is an upward trend, it goes from about 60% to well over 80%. So I think just an important note to highlight that this whole issue about home ownership is indeed a bit more nuanced when you think about stratifying biparental background. So getting on the housing ladder is clearly very different in terms of home ownership opportunities when you cut things by essentially your parents' characteristics. So this is taking individuals in their 30s and following them. Let's try to make a note. So this group up here is obviously a very different group. These people in their 60s who experience very different home ownership and housing opportunities. These guys are born in post-world war II, sort of in the 50s, 50s, 60s or late 50s say, but it's just interesting to note that people who are from let's say relatively advantaged backgrounds report higher levels of home ownership than those who are from the least advantaged when even in their 60s. So just something to note there and if you look at the people from the most advantaged backgrounds up here, a more selective group, but nevertheless report much higher levels of home ownership than indeed the least-advantaged in this group. So it's just important I think to hold this figure in your mind given the subsequent findings that I'm going to take you through. Conscious of times I'm going to push on. I'll briefly talk about WAAS, very, very helpful for our exercise here, both cross-section and longitudinal derived measures of total net wealth and its subcomponents. So in particular we use the subcomponents in this paper and it's very handy because it also oversamples the wealthiest houses by two and a half to three times. Especially useful to carry out this exercise when we're thinking about the role of parental characteristics is the fact that there's retrospective data on individuals' parents and what are they exactly? So WAAS, when individuals turn 25 or wave two initially, asks a battery of questions regarding individuals' parents. For example, their education, economic status at the time around age 14, housing tenure, single-couple-parent household, number of siblings. So what we do is we essentially create a proxy indicator of relative wealth or relative resources or advantage in early childhood or teenage years and then we kind of try and link that to wealth of the offspring. Now if I tried to use all of this information, I would run into quite small cell sizes. So what Paul and I do is just stick to using, creating measures of parental resources based on education and housing. We think of these are relatively stable, parental education generally doesn't change much as does housing tenure. Whereas for example, you might think about economic status, people become employed, unemployed, households form and break apart. So we have to be mindful of how much we can use this retrospective data, how complex can we make these groups? So we just stick to interacting education and housing tenant, which is precisely what you saw in that figure I just showed you. As I said, these are markers of resources. So what are we trying to do? We're trying to understand intergenerational relationship between parental resources on the one hand, proxied by their parents' education and housing tenure and then offspring's actual wealth, which is measured in WAAS because, as I said, WAAS collects individual measures of wealth that construct those for us. We don't have to deal with sort of life cycle biases and wealth accumulation because we want to observe wealth at current ages. I don't want to talk too much about that given, again, time constraints. What I do want to just focus on for a moment is what we're trying to estimate. So we're interested essentially in the correlation between parental resources here as the dependent variables. I've put W pair and really it's resources, but think of it as the interaction between parental housing and education and how that's related to the wealth of the offspring. And this dependent variable here, dependent variable here could take any, it could be total net wealth, it could be housing wealth. I'm only going to focus on the total net wealth very quickly and then quickly move to the housing wealth because that's where the action is. But from a layman's point of view, all we're interested in is what's the size of this coefficient by and what's happening to that coefficient over time for different cohorts. So again, I'm going to blitz through these slides very quickly, conscious of time, but basically when we cut the data just from a cross-sectional perspective, we split there, we just basically plot, take away way three of WAS and just show what does total net wealth look like for these different groups. And remember, this isn't individuals, their own characteristics for parents. You see there's generally a pretty clear ordering here, okay? So you can clearly see that those are more advantaged backgrounds on average report high levels of wealth, both again at a given age and for across different age groups. That's this is for total wealth, this is for housing wealth, this is for pension wealth, this is for financial wealth, this is for physical wealth. So needless to say, and I point you to the paper at this point, the general trends are very similar in terms of their direction and ordering. Right, lots of table numbers on this table, I apologize in advance, but this is just to give you a heads up of the stuff that that's fine that strength of that correlation between parents and offspring in terms of their wealth. So what I'm doing is I'm lining up here, parents from least advantage to most advantage, because I've got five groups of five or six groups of parents because I've got I've interacted their education, housing, tenure. So I'm lining parents up from poorest to richest in terms of wealth. And for the offspring to the most least wealth, wealthy to most wealthy. And I just basically try and understand that relationship. So how strong is that correlation for total wealth for the youngest group, which is 2034 is their birth year. It's about point three. So that's saying, if I were to increase parents wealth by one decile person say the median 50% after the 60th. And what would happen to the offspring's wealth? Well, this relationship would generally say that the offspring would shift in their respective rank by about 3.8 rank points. Okay, so to give you a comparative view of these numbers, this is slightly lower than what you find in the US and above what you see in Scandinavia. So this is just documenting and highlighting that there's a strong relationship between parental resources on the one hand and offspring wealth. And that doesn't matter whether I look at total offspring, net wealth of the offspring, the housing wealth, pension wealth and financial wealth, their physical wealth. Whereas let's us look at all of these different components and basically tells us a similar story that it is a strong relationship. But really from a policy perspective, okay, that's that's the starting point. You might want to say, well, how is it changing over time? That's what I care about, you know, is this, you know, you've told me it's important, is it changing over time? And indeed, that's what this table is generally showing us. So if we we now exploit the longitudinal dimension of wires and follow individuals over six calendar years, roughly speaking, and what I've done is I've exactly matched up or created age groups such that the age group of 2934 at wave three by way round six, they will reach 35 to 40. Okay, so what I can do is I can basically say, well, they start this age group starts here, they end around six. And although I know that'll be this age, so I can do these sort of chain comparisons, because I know this age group will reach this age by the end of the sample period. So I can do this diagonal cross comparisons. It's okay, well, that means I can compare individuals at the same age but born six years apart, and generally find that this this number is increasing. So of course, you know, might want to say, well, how much is it changing over time? And I'll come on to that. But again, given time constraints, I want to move on. But generally speaking, what one is what am I telling you essentially, the the role of parental characteristics is becoming increasingly correlated with the offspring offspring wealth. And this is the same sort of exercise but you carried out for net housing wealth. Okay. And again, we see that this general trend upwards. Okay, so parents characteristics are becoming increasingly correlated with the wealth net wealth housing wealth of the offspring. Okay. So we're seeing this general pattern across ages when we do this sort of what we call a chain comparison across individuals at roughly the same age. Okay. Hi, Ricky, just to let you know there's two minutes left. Right. Perfect. Good. So I'm going to skip through your side because I've said all of those things. The last thing to note is, you know, how rapidly is home ownership changing across cohorts? So you're so I said, well, this is changing, but you can ask me, well, how quickly? And in this, this slide, essentially, we do that because we run this regression where we control for age and a bunch of other things. And we basically calculate over over six years, which is what for the basic we have time, a 3.6 rank point increased over that time. So that's saying that basically, what's that telling us that saying, well, parental characteristics are becoming increasingly correlated with the net housing wealth. But also, and I should have said this all the way around, I apologize, also the likelihood of whether individuals report housing wealth. So there's two things to note from this slide. Parental characteristics or resources proxied by wealth, housing, tenure or education are increasingly related to whether offspring have housing wealth. So it's just two things. It's having versus not having. And so parents matter more for having, but also condition on having, they also matter for the level of wealth they hold. Okay. So that's really, really important. And so last slide, pretty much, I showed you that figure at the start. I should say, sorry, here, back to the envelope calculation, given point one eight, multiply this by five in about sort of 30 to 30 years would imply doubling if that trend held. Now, I showed you a figure at the start, which is showing you the proportion and showing you how stratified, you know, how if you stratified by parental background, you see quite a different picture when we think of home ownership opportunities. This is exactly the same thing, but instead of proportion reporting, home ownership, this is a level of housing wealth. And you see basically a very sort of similar story here. Not only do individuals from more advantaged backgrounds report high levels of housing wealth condition on owning, they accumulate it much more quickly to the point where people who are from the most advantaged background in their well, late 30s, early 40s report an average housing wealth, which is in excess of the average reported for those who are in their 60s, but from the least advantaged background. Okay. So this whole story about housing wealth and home ownership is very much stratified by, let's say at least from our data set and our evidence by parental characteristics. And I should say all of our findings hold even after we control for individual characteristics, individuals own education, region, number of siblings, all of our main findings hold up. Okay. So this is important when we think about, you know, home ownership and living standards, major life cycle decisions, such as, you know, buying your first home. If you think about historic and recent returns to housing, you know, they're pretty strong, although I shouldn't say in the last six, six, six months or so. But, you know, indeed, owning a house does generally pay off at least in the UK. This correlation, this relationship between parents and the offspring, if current trend is set to hold, based on our estimates, it would double in about 30 years. So, you know, you don't control who your parents are, what background you're born to and their parents' characteristics. But if we think of it as a penalty, let's say from a social mobility perspective, it's mattering, it's growing more and more important for determining offspring, wealth and living standards. So just say future work, we're looking at things like sortative mating, decomposing some of these factors in much more detail. So I'll stop there and thank you very much for your time. I have Lawrence O'Brien from the Institute of Fiscal Studies to give us his presentation on the gender gap and pension saving in the UK. So, if Lawrence, if you're happy to share your screen while I introduce you. Thank you. Lawrence is a research economist at the Institute for Fiscal Studies, primarily researching people's saving decisions and economic activity in later life. He's also a PhD student in economics at the University College of London. So, when you're ready, Lawrence, onto you. Can you see the screen? Yes. Perfect, great. Thank you, Emma. So I'm presenting this work we've done at the IFS recently, together with Jonathan Cribb and Hedekai Yalainen, which is on the gender gap in pension saving. Of course, there are very good reasons to care about gender inequalities at all points of life, but the gender gap, pension gap, generally refers to the differences in incomes between men and women who are retired, although DWP recently did define it as differences in pension and wealth for those approaching retirement. And of course, should policymakers wish to use policies to change the or close the gender pension gap to understand the effectiveness of these different policies is crucial to understand what's really driving the gap. And that's what we're going to be focusing on in this presentation, and that's what we focused on in this recent report we published. And one thing that's really going to be important, and it's going to be a common thread running throughout this presentation, is that it's really important to be clear which gap we're analysing and for which group. So are we looking at gaps in pension incomes, gaps in pension participation, that's whether you're saving in a pension at all, or in pension contribution, so how much is actually entering people's pension pots. And when we're looking at which groups, are we looking at people who've already retired or people of working age, or are we being even more specific and looking at people working in specific sectors. So this presentation I'll start out by looking at people who've already retired, and any differences in their incomes and their pension incomes there, before then moving on to looking at people of working age, and show how the gaps change when we look at these different groups. So starting with the state pension income, for those who are over state pension age, this graph start with I plotted average weekly state pension income of men and women who are born in the early 1930s, and we can see that men have on average about £160 of state pension income per week, compared to about £120 for women, so there's a fairly sizeable gap here. But as we look at cohorts or generations who've reached state pension age more recently, we can see that this gap gradually closes due to both reforms to the state pension system, as well as changes in women's labour market attachment, looking at more recent generations. And in fact, when we look at those who were born in the early 1950s, who reached a pension age quite recently, we can see that the gap is almost closed. So women's average state pension income is only about 5% lower than men's, and we should think that this gap will remain small or perhaps close completely in the future due to these reforms, and again, women being more likely to work than they were in the past. So that's the state pension income. When we look at private pension incomes, however, we see a rather different story. So the gaps are both wider and have narrowed much less when looking across generations. So we can see, for example, those in the early 1930s, women have about 60% lower private pension income than men, and this is narrowed to about 40 or 45% when looking at those born in the early 1950s. So this really shows when thinking about the gap in incomes for those above retirement, for those who've reached a pension age recently, or who are going to reach a pension age in the future, we're really thinking about gaps in private pension incomes. And these gaps in private pension incomes will be driven by gaps in amounts entering people's pension pots done throughout working life. So should policymakers wish to affect these gaps in contributions, it should, of course, really be driven by any gaps we see in pension saving behavior of people of working life currently. So that's what I'm going, of working age currently. So that's what I'm going to be focusing on for the rest of the presentation. So looking across all people of working age, we can see that about two thirds of men are saving into a pension compared to about 60% of women. So there's a gap here overall. Of course, the first thing you might be thinking is that we know that fewer women are in work than men. And indeed, once we restrict our attention just to workers, we can see there's a much smaller gap here. In fact, the gap is less than one percentage point. But this isn't the whole story. Once we look within specific sectors, we can see that some gaps start to reemerge. So in particular, among people who are self-employed, the proportion of men saving into a pension is about three and a half percentage points higher, the proportion of women saving into a pension, and the gaps even larger among vector employees are about five percentage points. So how do these within sector gaps combine to leave us with quite a small gap when looking across all workers? Well, that's due to the different sectoral composition of the male and female workforces. In particular, a high proportion of male workers are self-employed than women. It's about 15% of male workers are self-employed compared to about 10% of female workers. And this is a group where pension participation rates are really low. And on the other hand, women are much more likely to work in the public sector than men. So about a third of female workers work in the public sector compared to about 17% of men. And this is where pension saving is particularly prevalent. And that's why these gaps then combine to offset, or these sectoral differences offset some of the within sector gaps to leave us with quite a small gap when looking across all workers. So that's gaps in whether people are saving into a pension or not. What about the amount going into people's pensions? So I'm just going to look at those who are participating in a pension scheme. And now what we can see is that when looking across all workers, on average, a higher amount is entering men's pensions than women. So average total, so that's employee plus employer pension contributions, are about £100 per week for men compared to about £85 per week for women. So a gap of around 14%. And this is despite the fact that when looking at all workers, we didn't see much of a gap when looking at the pension participation rates. Again, once we look within specific sectors, we can see that these gaps are larger here. This is a gap with about women having about a third lower pension contributions than men in all of these three sectors, more or less. And again, the fact that we see a smaller gap when looking across all workers done within each sector separately is due to the fact that women are more likely to work in the public sector than men where pension contributions are higher. So we have, we do see that there's a gap in pension contributions. But of course, another thing, another difference between men and women who are in work is that women tend to earn less on average than men. And these could of course lead to differences in the amount entering people's pensions. So to look at this, we now in this graph focus on, we measure pension contributions as a proportion of gross pay. And then what we see is that women actually have higher average contribution rates than men. So about 15% of women's pay is entering their pensions compared to about 12, 12.5% of the men on average when looking across all workers. Once we look within sectors, we see that there is a similar gap in favor of women when looking at those who are self-employed. So women have about 2.8% to pay higher pension contribution rate than men, although I should say that women who are self-employed and saving in a pension is quite a small group compared to a lot of the other groups in this chart. There's no real gap in the private sector among employees. Looking at public sector employees, there's actually a slight gap in favor of men by about 1.5% of pay. So overall so far, what we've seen is the differences in labor market experiences between men and women drive gender gaps in the amount saved in pensions, in particular differences in employment rates and earnings. And these differences in contributions saved in pensions will be the key determinants of the gender pension gap in pensioning incomes for future retirees. We did also see when looking within specific sectors, there are still differences in pension participation and saving rates. So the key question now is what's driving this? Is this differences in other characteristics which are really driving this? Or is there some difference in decisions between men and women who, from what we can see in the data, look really quite similar? So looking to this, what I'm going to plot here is pension participation rates for men and women who are employees by age. And I'm plotting this by age because we know that a lot of other gender inequalities open up across the life cycle. So when looking at those working in the public sector, we can see actually a fairly similar pension participation rate for pretty much all ages here. But when looking at the private sector, we see that there's a gap in favour of men at all ages. And in fact, this gap widens with age. So men have a higher pension participation rate by about two to four percentage points for those in their 20s and early 30s, widening to more like six to eight percentage points for those in their 40s and 50s. So now what we're going to try and do is see what's explaining this gap. So what I've done again, just to start with this plot, this raw gap again. So this is same as what I showed you in the previous slide. There's a gap of about two to four percentage points for those in their 20s and 30s, from rising to more like six to eight percentage points for those in their 40s and 50s. And the first thing we're going to try and do to explain this gap is restrict attention to those earning at least £10,000 per year in the yellow line here. And the reason we do that is because we know from other work, this is the threshold above which employers are obliged to automatically enroll their employees into workplace pension saving. And we know from other work that automatic enrollment has a really key impact on people's pension participation decisions. And indeed, once we do this, we can see that the gap in participation rates is essentially closed. And in fact, at younger ages, it's slightly in favour of women, but only very slightly. We can also control for other job and employee characteristics using regression analysis. So for example, there are exact amounts of earnings, their industry, their occupation, that hours worked, all these other characteristics. And once we do that, we see that again, this moves the gap slightly in favour of women, particularly at older ages. So this purple line suggests that when we're comparing men and women earning at least £10,000 per year, and with similar characteristics, if anything, women are slightly more likely than men to be participating in a workplace, in a pension, but only by about two percentage points. And indeed, really, the key thing to take away from this graph is that the gap in whether people are saving in a pension or not among private sector employees is really driven by this automatic enrollments threshold of £10,000. So that's participation rates. What about contribution rates? So again, this is for private sector employees and just those who are saving into a pension. We can see there's actually only quite a small gap for most of these ages, slightly in favour of women by about 0.5% of pay for those in their 30s and early 40s, slightly in favour of men by about 0.5% of pay for those in their 50s. Restricting to those earning at least £10,000 a year doesn't have much of an effect here. That's because we're already looking at those who are saving into their pension and then want to be further control for other job and employee characteristics. We can see that this does cause the gap to be pretty close to zero at most ages, suggesting that similar men and women who are private sector employees are contributing fairly similar amounts into their pension. So what about other sectors? So in the public sector, we see that as I showed you before, there's actually very little gap in participation rates between men and women in the public sector. Most about 90% are contributing, are participating in a pension. When it comes to contribution rates, there is a gender gap in favour of men by about 1.5% of pay. We find that this closes to approximately zero once we control for earnings and other individual and job characteristics. And the reason for this is the structure of public sector defined benefit pension schemes. These have very rigid contribution schemes. The employees have much less choice over how much to go into their pension. And the way these rigid contribution structures work is that higher earners have to contribute a higher amount into their pension. And so given that men have higher earnings than women in the public sector, this is driving the gap we see unconditionally. Among the self-employed, we do find a gender gap in participation rates by about three and a half percentage points. We find this does shrink after we control for earnings, but this does not close completely, especially at older ages. And as I already hinted at, I think among savers who are self-employed, women actually have higher average contribution rates than men. Okay, so just to summarise the findings, I showed you that while the gender gap in state pension incomes has more or less closed for those who've reached state pension age recently, there is still a large gap in private pension incomes that's closing much more slowly. When looking at those who are of working age, there is a gender gap in private pension saving. And the main determinants of this are differences in labour market experiences. So whether people are working or not, but also conditional on working how much they're earning and related to this, the hours they're working. And we know from other evidence as well as some other charts in this report that these gaps in labour market experiences particularly open up after women have children. We also saw that differences in the sectoral composition of the male and female workforces offset some of the within sector gaps in pension saving and that these gaps in pension saving within sectors are again largely driven by differences in labour market outcomes in particular differences in earnings. The fact that the gender pension gap or the gender gap in pension saving is driven by differences in labour market outcomes such as employment rates, hours worked and hourly wages suggests that the gender gap in private pension incomes for retirees will remain open for decades. And that's because we see these gaps in labour market outcomes for all generations in the labour market today, even the youngest one. And indeed this highlights that policy makers should see the gender pension gap as really a further consequence of labour market inequalities. So when we see labour market outcomes starting to diverge especially after childbirth, it's important to keep in mind this will have an effect on incomes all the way into retirement. And also if policy makers want to sort of target this gap, the root cause of this gap, it might be that labour market policies are better placed to do this than pension saving policies on their own. Finally, I think it's important to say that when thinking about how pension saving policies might affect this gap, of course you should think about how these policies affect the system as a whole. So as I showed you before, there is a gap in pension participation rates among private sector employees which is driven by the £10,000 threshold for automatic enrolment. But that doesn't mean we're necessarily saying you should scrap this threshold and automatically enrol all employees into a workplace pension. There are some good reasons for many of those earning less than £10,000 per year to not saving a pension, in particular state pension itself is worth slightly over £10,000 per year. So it's not clear that people earning less than this amount should be shifting resources into retirement. So that's all from me. Okay, on to the next presentation. So for this presentation we have Jonathan Cook, who is going to be presenting his work on who is over-indicted in Britain, the roles of health, financial literacy and risk aversion. So as you're already sharing your screen, I'll do a quick introduction. Jonathan is a professor emeritus at the University of Edinburgh Business School. He's a former founding director of the Credit Research Centre and former director of research at the University of Edinburgh Business School. He has researched into credit risk and the demand and supply of household credit since the late 1970s with several books and over-70 referred articles. Former joint editor of the Journal of Operational Research, John Isafello of the Royal Society of Edinburgh and of the Academy of Social Science. Thank you for joining us today, Jonathan, and over to you. Thank you very much for having me. Most of my work is usually on credit risk, as Emma said, but now I'm firing into aspects of demand and supply in particular, trying to identify who's over-indicted. There's quite a lot of literature on this, and I'm going to concentrate on what aspects of health might actually affect whether someone's over-indicted, whether the financially literate are over-indicted and degrees of risk aversion. The message at the end of the day will be certain types of health matters, but perhaps not in ways you might expect. We don't find any evidence for financial literacy affecting whether someone's over-indicted and risk aversion is also, we also find evidence of risk aversion. Some of these results are completely new to literature, others simply disagree with the literature, so I await comments. Okay, so usual structure. I'll say very little about the literature. I'll talk a bit about the data, the methodology, and show you the results, and then discuss what I've got. So the basic idea is to model the relationships between over-indictness and its determinants, and to look at these hypothesis I've mentioned. First thing to say is that there's no university accepted definition of over-indictness. It's multi-dimensional. The EC in 2008 came up with an operational definition, some of which in my view don't make sense. For example, over-indictness implies household cannot pay contractual commitments without reducing its standard of living. To me, that's just restating the meaning of opportunity cost. But anyway, so what I'm trying to say is there you can always argue that there's a different definition of over-indictness. So what does the literature use to actually measure it? They use these sorts of measures, debt service ratios where the household spends more than a certain percentage of income. The percentage is arbitrary. It can be done in terms of stocks, debt compared with income. Whether or not borrowing repayments takes the household below the poverty line, which is actually quite an interesting one. I haven't got time to go through that one at the moment. Delinquency, whether someone is more than a certain number of months in arrears than they're thought to be over-indicted, may or may not be true in any generic sense. Number of loans probably has got very little to do with over-indictness, but it has been used in the literature. And then a perception of the burden of debt, where you ask respondents, do they regard debt as a heavy burden or not? Do they have difficulty paying your unexpected bills and so on? Weaknesses with all of these, the last one's obviously very subjective, may not be comparable between individuals. But unfortunately, these are the measures. These are the data only really is available for these sorts of measures. Okay. So I don't want to blind everybody with science, but one point I do want to make. There is a very general, very commonly accepted way of modelling someone's consumption over time and so the demand for debt. And in a nutshell, if we assume that an individual acts as if they're going to maximise their satisfaction over their lifetime, where their income is uncertain, then we can plot, given certain parameters, what we would expect to happen to the amount of debt. Now, I'm not going to go through this equation or derive it. All I'm going to say is, if you look in this equation, this equation shows what the unimplication of this model for the amount of assets someone holds at any point in time t. And the crucial bits of these two bits here, that is that the assets that someone has does depend on whether or not they have an unexpected consumption shock. So that's actual consumption. And that's the expected actual consumption expected in the previous period. Here we've got an unexpected income shock. That's actual income. And that's the amount of income that was expected in period one. Remember, we're looking many periods into the future. So the bottom line is that if assets are negative, then someone holds debt. If they're very negative, there's a high chance they're over indebted. And this A can become very negative if they're unexpected consumption shocks or unexpected net income shocks. And this is within the lifecycle model. That's the point. Okay. What does the lifter do? The lifter looks at lots of empirical correlations. It doesn't agree. It looks at unexpected net income shocks due to poor health, loss of job, relationship breakdown, et cetera. There are other possible causes of over indebtedness, notably unexpected consumption changes due, again, poor health, car breaks down, washing machine breaks down, et cetera. Poor financial calculations may mean that someone is perhaps not on top of their repayments. It takes out more debt than they can afford. And now there is a well-known theory, given this nice name of hyperbolic discounting. Basically, you consume a lot today rather than tomorrow because you value consumption today more highly than tomorrow, jump to tomorrow. And then the difference in preferences between one period and another changes. You then appreciate that the actual consumption you did last time was too high and this induced you to take too much debt very simply. Okay. So lots of, as I said, these measures have many errors and many weaknesses. They've been used in the literature. I'm going to look at all three of these, trying somehow to get some common view of the results. But in particular, I'm going to try and correct for something that the literature, almost all of the literature except for this paper, make. And this is, if you try and regress something like whether or not someone feels over indebted, if you have missing values and the literature typically has many missing values, if you code those as not perceiving difficulty, then it's highly likely that you will not get the parameters, i.e., a model which relates to the population as a whole. In fact, it's not very clear to me what you do get. Yet that is exactly what the literature does. What you need to do if you are interested in what determines over indebtedness for the population as a whole, when you have many missing values, is to model the missingness mechanism. And that's exactly what I do. To economists, this is a standard sample selection problem. Why might debt be there? Well, it's highly likely that there's a correlation between the probability of not having debt and the probability of it being over indebted. That is, you may have been turned down for debt because banks and lenders thought you would be over indebted if you borrowed. So we're going to correct for this. So what time is passing? So I'm going to jump a bit. Poor health can increase expenditures, obviously, and reduce income. So it could lead to over indebtedness. Someone who's risk averse may adopt a very different discount rate. That is the amount to which they prefer consumption today compared with consumption tomorrow. It would increase this. And as a result, consumption may be taken earlier, facilitating large amounts of debt, or it could be the other way round. Jumping over a discount rate for the moment, looking at financial literacy, there's some literature which suggests that people who are financially literate are less likely to become over indebtedness, but it's not universal. And John Gathergood finds evidence to the contrary under certain conditions. So the data I'm going to use is various rounds of the Wealth and Assets Survey, what I think has become called the was. As you know, it's collected household and individual level. It's the interviews are in waves, and it's actually a longitudinal study. So each wave covers a two year period as I got here. It jumped from waves, wave periods, July to June, to round periods, April to March, in 2014 to 16. So we look at various waves. I'm going to use four, five, six, and round seven. And I've homogenized it all to be at rounds, in rounds. So we simply move the data in appropriate ways, such everything is at rounds, even when it's collected in waves. There are about 18,000 household interviews, 34,000 individual interviews, and the wealth are over sampled. Okay. So I'm going to use three measures that I said earlier. These are the questions in the survey. Thinking about mortgage or loans, that's an example. To what extent, keeping up with the repayments, a financial burden to you, a heavy burden, somewhat of a burden or not a burden at all. Have you missed two repayments? And thirdly, looking at total payments over net income, over an arbitrary percentage of 30, I could have used any other percentage. I picked 30. It may well be results differ if I use a different percentage, but they're measuring different things. So let's see what sort of results we get. The questions on health are two different aspects of health. How's your health in general? And do you have any long standing illness? And we're going to code these so that when I put them in the regression equations, we can look at the effects of each. Risk aversion, if you have a choice between a guaranteed payment for £1,000 and one in five chance of £10,000, what would you choose? Both have the expected value of being the same, and so the only difference between them is the degree to which you value or disvalue risk. I'll jump over the discount rate because I'm short of time. Financial literacy, if inflation is 5%, so to what extent can you compute a real interest rate? Can you read a bank statement? And can you calculate interest? That's basically what they boil down to. Okay, so as I've said, I'm going to use sample selection models because what I'm interested in is a model for the population as a whole. I'm interested in everybody. I'm not just interested in those who have debt in the sample. I want to know if we look at everybody, including those who currently don't have debt, what sorts of people would be over indebted if they have debt? And I've got the data. I've converted it into a panel. You can do this, although there's no unique ID through each round or each wave. You can actually do it by chaining, which I can talk about on another occasion. So we chained the data, we've got panel data. We're then going to use a product model to predict the probability of being over indebted. We're going to predict that we're going to have a sample selection equation, probability of having actually got debt. And when I can get a panel estimators that do this, i.e., a probate with sample selection for panel data, then I will use it. But what I found often was that, in fact, the algorithms would not converge. The maximum likelihood routine would not converge. And so in some cases, I go to a second best, but you'll see that as we go through. Just let you know, Jonathan, you've got two minutes left, thanks. Two minutes? Okay. Yes, thank you. I thought I started at 12. Anyway, okay, so looking at individual cards, mail order, HP and loans, for burden, what you can see is that those who are more risk-averse have a lower chance of being over indebted. Those who with generic bad debt have a greater chance, and those with long-standing illness have a lower chance of being over indebted. If we look at missing two consecutive payments, we get similar sorts of results. Those with bad health have a greater chance of missing two. Those with long-standing illness have a lower chance. And when we come to DSR, you can see, again, bad health gives you a greater chance and long-standing illness a lower chance. Okay. Financial literacy is only collected around seven, so no panel here. And you can see if you look at the bottom on the left-hand side, the three measures, there's nothing significant here except for payments' heavy burden for, can you compute a real interest rate? So on that basis, we don't find much evidence for financial literacy. We do, however, find the evidence, as I said earlier, and showed you in the panel data for generic bad health and long-standing illness going in opposite ways. Looking at changes in over indebtedness, if you're in a particular state, this period either not over indebted or over indebted, then the transitions are possible are to go to being not over indebted or over indebted. So if we take the difference in the indicator, that's what is indicated. So the question is, can we address this change? Or if we do, what do we get? And this is the answer. Concentrating on bad health and long-standing illness, well, you can see is a jump into bad health, sorry, into it over indebtedness, is correlated with having bad health. But jumping out of it appears not to be. On the other hand, jumping into over indebtedness is negatively correlated with long-standing illness. In terms of the discount rate, jumping into it is associated with having a high discount rate, and I was preferring consumption today rather than tomorrow, and jumping out of it is negatively correlated with it, which is consistent. Okay, so conclusions. You probably didn't spot because I didn't point it out, but there is evidence that the consample selection does have an effect on the parameter estimates, and that therefore suggests that previous literature is not as convincing as we might have thought. Poor general health increases the chances of being over indebted and becoming over indebted, that's the second bits of change. Poor long-term health is associated with a lower chance of being over indebted and becoming jumping into over indebted, and this correlation has not been considered in the literature before. High discount rate, I being more impetuous, if you like, preferring consumption today rather than tomorrow, is associated with high debt to service ratios and with them increasing, and being risk averse, people who are highly risk averse relative to others have a lower chance of becoming over indebted and a lower chance of jumping into over indebtedness, so people with high risk aversion are not taking on too much debt, it would appear. But on the other hand, we don't find financial literacy is associated with measures of over indebtedness, but it is associated with having debt, and what the earlier literature might possibly be doing is confusing the two facts. I, in this paper, have separated the two facts out. That's it. Sorry I've overrun.