 Thank you very much Nancy and thanks. Thank you very much to the organizers it's a great honor for me to be here today and Thank you everyone for coming to listen to what I have to say Okay, so as Nancy said I would like to talk about savings and wealth inequality and I would like to start from some very important open questions So first of all why are some people rich while some others are poor and others are in the middle? What to what extent can the government affect inequality and what instruments should they use? so In very important stepping stone to answer these questions and many others is to understand why people save The goal of this talk is to understand why people save and how this saving behavior leads to wealth inequality So let me give you a road map of this talk The first thing I would like to do is to provide some basic facts. This is not a seminar about Empirical evidence because I only have about 50 minutes So I just would like to give some basic facts to get all of us on the same page Then I would like to talk about the basic Beauty hug it a yagari more or ugly model that I will call beauty for short from now on Which is the building block of many many modders of wealth inequality And I would like to try to explain what are its main mechanisms and what it can and cannot deliver and why Then I would like to talk about richer models that I've already been developed and taken to data I Would like to talk about what I think we have learned so far from this body of literature And then I would like to conclude with what needs to be done and what I think are very important avenues for research both from the standpoint of topics, but also I will conclude with some methodological thoughts So let's get started with the basic facts as I said this will be a very short review First of all, I'm sure nobody here ignores that wealth and earnings are very unequally Distributed with a lot of poor people and a very thick tale of richer people both in earnings and wealth But wealth is much more concentrated than earnings So for the talk I will talk about US data, but these facts hold across many countries in this table You see wealth from the 1989 Soviet consumer finances data. The reason I take this period I will talk a little bit about changes over time But is that many of the models will be calibrated to this period if you look at the richest 1% About 20 30 percent of total net worth is held by various very small group of people And if you look at the richest 5 percent over half of total net worth is held by these people If you look at earnings and here report the distribution of earnings for working-age people I don't want to include people who are not working For wealth it doesn't matter a whole lot because the distribution of wealth is almost the same across the working age And the whole population So if you look at earnings and the richest 1% makes 6% which is still a concentrated distribution But much less than the distribution of wealth So one important question is going to be what kind of saving behavior or forces in the model give rise from your Concentration of earnings to the one we observe for wealth an important point that You know many people have discussed is that if anything since 1989 things have become more concentrated And so the puzzle certainly has not gone away The other important piece of evidence that I would like to mention is that Rich people you know consistently with these facts not only have high income and earnings They also have a high saving rate. Okay, so they save a larger proportion of their flows of earnings And this happens both before retirement and after retirement so I Just wanted to flash this light. There are many. This is just a partial list There are many valuable contributions about the facts and but I want to focus on the models today Okay, so let's talk about the big the building block the the beauty model of wealth inequality It's a characterized by two very important pieces one is preferences Here I'm picking the version of a basic beauty model in the life cycle But if you want to think about the infinitely lived one, you just have to let he go to infinity Okay So the person the household maximizes expected the future utility from consumption in all future periods Because I do the life cycle. There is a survival Probability there is a random variable s which is one if you're alive and zero if you're not alive and its distribution Evolves over age to be consistent with life expectancy and you derive utility from consumption The other important piece is the budget constraint so here the first important assumption is that there is one asset and it has a risk-free rate of return are and You consume these are your accrued Assets and returns and this is where the uncertainty comes from. Okay, so this is your earnings it's a stochastic process and What people do here? Even if they are identical ex ante exposed they are hit by earning shocks. They experience different earning histories And therefore they are exposed to terogeneous. Okay, so what we typically do is we close the model Computing either a steady state in which there is a constant distribution of people over state variables and people's Circle through these are states and people typically are impatient because to have a finite support people are impatient and Circle through this distribution of wealth So what I would like to do is to discuss what kind of saving behavior this kind of models deliver So here this is the saving rate by age and wealth for a median earning person And I'm going to do it by age because I'm using the life cycle version So here you have current assets and you can see that the black line is the one for a 25 year old If the 25 year old has little assets, they have positive savings But for these given earnings level as assets increase the saving rate goes to zero and actually becomes negative Okay, as people age the retirement saving motive kicks in so these saving Rates move to the right and the crossing point is higher But the bottom line is that when your health gets to be quite high your saving rate becomes a negative Okay, so the precautionary saving behavior in these kind of models is a self-insurance against earnings risk against longevity risk and then there is savings for retirement and The important idea is that there is this buffer stock of wealth Okay, so once you reach this buffer stock of wealth your savings turn negative and you stop saving and this is what is really in contrast with the Empirical evidence Especially in the US there is strong empirical evidence that people when they're very rich they keep saving at a very high rate Okay, so this is the intuition why in the beauty model and I will show you some tables later It's hard to get the very rich people because when you get rich you stop saving and you don't get truly rich okay so What are the limitations well first of all? The saving behavior is counterfactual compared to the data and second these kind of models don't generate very rich people In addition, I think there is something else These are very simple models that miss a lot of risks or other elements other saving motives And so we might miss or mischaracterize even the savings of people. We think we understand Right, not only it's clear that we are missing the rich But it's possible that we think we understand the middle income or the poor why we really don't because we mis-specified the environment And the reasons why people save is very important. So let me give you an example To try to drive this point home that it's not that we don't understand the rich It's not clear that we understand the saving behavior of other people or at least all of their main saving motives Suppose as in reality that at least in the United States Out-of-pocket medical costs and long-term care costs are a significant people risk that people face especially during retirement So what is going to happen is that the people who have low lifetimes are insured by the government through mean tested programs But upper middle income people we really face a lot of medical expense and long-term care risk So if you abstract from these risk in the models typically what we do we try to match our Average or total or some measure of assets because we want to match the resources and the economy So we are going to typically going to assume these people are very patient But then you when you evaluate government insurance What you're going to have you evaluate government insurance on people who are very patient as opposed to people who save a lot Because they face a lot of risk. So you are going to get Evaluations about these government programs that are very misguided because you are misunderstanding the savings of a lot of people So for going from the standard beauty model, I would like to talk about six important ingredients There will be others But I need to make some choices and I will try to be clear about why I think that these six ingredients are important So I would like to do You have seen these equations before and I would like to show you what are the new pieces Okay, so the first one is when we introduce bequest motives and that transmission of human capital across generations Well when you introduce bequest motives you add To the utility function some value from the assets that you leave to future generations Okay, so you are changing preferences The other thing that you do is at some point people receive this bequest when their parent dies and If you allow for transmission of human capital one very easy reduced form way to do it is to link the earnings of parents and children, okay So the second kind of story that I would like to discuss is Eterogeneous preferences, okay in this kind of story You also modify preferences, but you just assume that people have different discount factors and risk aversion The third story is going to be that it's not really preferences that are What we work on but they are rate of returns instead of everyone facing the same rate of return We have a rate of return that is idiosyncratic to the person and is stochastic The fourth story is entrepreneurship and here what we do is people can choose to run a business In which case they have a production function in which they invest some capital to run the business Or if they don't run the business they get the same earnings fluctuations. They get in the basic beauty model So this is also essentially about how you model resources here the fifth story is That maybe we are not getting the earnings risk, right? Okay, and so it's also in the budget constraint and the sixth story is that well Maybe we are getting this risk right But there are other risks for instance coming from medical expenditures and long-term care that we don't allow for in the budget constraint But they are important for a large fraction of people So let me go to the first story Why did I pick these six stories besides the fact that I wrote Four of those stories I brought on I think a model is always a very simplified version of reality and How you decide how you should make your model complex should be Grounded in empirical evidence. Okay, so for each of these stories. I will first you see these little dots here So this is the empirical evidence that underpins that story and that makes you think yes This is something important that I should actually introduce in the model So for each story, I will have the empirical evidence than the model and the results so Big question human capital so There is a lot that it was a lot of debate between Coplico fan summers and Modriani that was later settled in my opinion The idea is that when you look at capital today You can distinguish the one that has been saved by people who are alive today And the one that has been inherited by people who are already dead and in the aggregate Notice it doesn't tell us anything about distributions in the aggregate the amount of capital that is transferred across Generations we can debate if it's 50 percent or 60 or 70, but it's large Okay, so there is a lot of capital that is earned by our Parents and transferred to us the second one is more at the individual level There is a lot of empirical evidence that earnings education Social economic status of parents and children are linked, right? So this is a evidence about the human capital part while this is about physical Physical capital and why this is aggregate. This is at the individual level. Okay, so that's why I think it was worth exploring This story there is a lot of evidence that this can be important in understanding wealth and savings So let me tell you what I did in this paper What the first part that you have seen the olg with the retirement period and earnings and lifetime Uncertainties the basic model I have shown you so far what I would like to point out is that typically you have Earnings risk and you have mortality risk. So people save in these risk-free assets. There are no annuity markets As people don't typically buy them in the data in many countries. So people Happens some people live die early and they leave accidental bequests Right accidental in the sense that they would like to consume them, but they die early on and so they leave bequests So when introduce a bequest motive and I will be much more explicit about what this means You not only have accidental bequest. Those don't go away. They are still there But you also have voluntary bequests because you derive utility from transferring some resources to your children Okay, so you will have two types of bequest in this environment So I would like to remind you what the problem is you have this additional term in the utility function the intergenerational correlation of earnings and the received bequests and I would like to tell you more about what I did for the bequest motive So I picked a warm glow altruism in which you derive utility from living assets And there are two important things to say While this is a typical CRRA function this parameter It's very important if this parameter is zero the marginal utility of even as of a small bequest is very large So even poor people would try really hard to leave something to their descendants But this is a very has very come to our factual implications Because many a large fraction of people in the US if you look at people without the surviving spouse Which would be the counterpart in the model of 30% have little to no value Okay, so you need to make sure that the distribution of bequest in your model looks like something in the data The other thing we could have thought about is to have a more altruistic model You know the problem is that the altruistic model also has very strong Implications about intergenerational risk sharing so while I think the the question of how we should model bequest is important and still open I think this is a very good starting point and it matches the absurd distribution of bequests So importantly, I do not pick model parameters to match wealth inequality For instance, I pick parameters of the bequest distribution to match moments of the bequest distribution So what you get for wealth is sort of an identifying restriction is not something that you try to match by construction Okay, so let me show you The results from this model So the first line you have already seen except that I'm adding the wealth genie concentration So here you have the data from the 1989 SCF and you have the how much wealth as the richest 1% in this line What you have is I actually want to say something so This line. It's actually two different models that have the exact same implication This is the model you have seen with no bequest motive whether you give the bequest to everyone alive The accidental bequest like we usually do in many models or you give them to the children you still get these exact numbers Okay, so what is going on is that some people get a large bequest some people get a small bequest It's like winning the lottery. They're saving behavior doesn't change. Nothing changes for wealth inequality Here with the blue you see when we start adding I here I want to add this warm glow bequest that is a luxury good and you can see that the Concentration of wealth in the richest 1% doubles. What is going on here is that? This non-homotic city in the bequest motive implies that when you have either high-earning Seastories or you get a large bequest. That's when you want to leave a bequest So some of the wealth gets concentrated because it's passed on across generations And that's how some rich families arise The other thing I want to point out is that this model does a really poor job on matching the poor people right in the data about 6% has zero or negative wealth and In a life cycle model in which people start with little to no assets like in the real world These models over predict the front and you know the adding bequest that nothing good on this side of the distribution You over predict the number of poor people Okay, so when you add the human capital inheritance You have that not only some families leave more assets their children and these propagates, but they're human capital Their earnings are correlated. So there is a little more concentration So let's summarize the findings from this first Story accidental bequest and voluntary bequest and human capital So first of all, you know, sometimes you read all Bequest are actually Equalizing and so on and so on so forth What this paper tells you is that it's not as much the receipt of the bequest that matters is Whether you have a saving rate that is increasing in wealth Okay, so it's really the bequest motive rather than the passive receipt of an inheritance that changes wealth Concentration because it changes saving behavior You have also seen that adding ability Correlation of ability across generations help generate the breacher families over time But the big thing is that even with bequest and intergenerational links the wealthy are not wealthy enough and the poor are too poor Okay, so this is making progress, but it's not all of the story and you know I think when you try to calibrate these parameters to other moments You can get a Better idea why to how much this can explain. Okay, so let me go to heterogeneous preferences So again facts, why should we even think about the heterogeneous preferences? Well, I think there is a Very large body of evidence from the applied micro literature The preferences are heterogeneous You can look at different methods Euler equation versus life cycle method of simulated moments maximum likelihood You can look at US data and the PSID we have Danish registry data There is a lot of evidence that people are heterogeneous in their patients and their risk aversion So what do we have in terms of models people have studied heterogeneous preferences. I'm sure the paper 99% the people in this room know is cruz Allen Smith So what cruz Allen Smith did they took an infinitely the agent model and they found that a little bit of heterogeneity in data goes a Low way, you know in the macro tradition. You try to say it's you only need a little because we are very skeptical of heterogeneous preferences But it's still you know it makes sense, but the rich you cannot match the upper tail It's not like you know Bill Gates is twice as patient as everyone else, right? So it's there, but it's probably not the reason why the rich are so rich Even more interesting. I think as I will argue later I think using a life cycle model introduces much more discipline in what you are doing And papers by Hendricks and Gonzalo Paz pardo used life cycle models and found that first of all in life cycle model You need much more preference at a journey and what happens is that yes You can increase the rich that can be a little richer But then the poor will be poorer right because you have this gap in in preferences And they also looked at both beta and sigma at least Gonzalo did as opposed to cruz Allen Smith That mostly played with beta So what do we have on this story? So, you know on the one hand I think the evidence from the applied literature is fairly Substantive that there are these differences. So I don't think we should completely discount this explanation But I don't think that this is the reason why the rich the super rich are very rich So I think this is an important mechanism that we can use in conjunction with others But not the whole story So my third Story is about the Theruginus returns So this is a new paper that came after the models that I will tell you about And this has an exceptional Norwegian data So that you can use these are very rich data to compute returns from stocks from bonds from private equity And these authors found find that returns are very Theruginus For instance 200 basis point across the 10th and the 90th percentile of the distribution They are also Theruginus within asset classes. So whether they are bonds or risky assets And interestingly they are correlated people with higher wealth tend to have higher returns and people with the private equity Entrepreneurs also tend to have higher returns So These papers came before the evidence. So they were ahead of their time So what? Well, yeah, it's even if it's 2015. I think it was in the works well before we heard about the Norwegian data so this paper what it does essentially is to choose the The these Process for the rate of return at the individual level to match wealth inequality And the conclusion is that first of all the rate of returns alone cannot match the distribution of wealth you also need be quest motives and In my opinion, this is a very interesting explanation, but we need to move forward. Okay, so this is sort of a Perhaps an interesting Mechanism that works, but it's a bit of a black box. First of all, I think it's very important to be more serious about matching these Implied returns to data, but more importantly these rate of returns are endogenous They are endogenous to your portfolio choice. They are endogenous to entrepreneurial choice They are endogenous to a variety of behaviors that you cannot just assume that they fall from the sky especially if you want to evaluate taxation So papers that haven't endogenized to some extent and you know, they did it they didn't Mean to do it because we weren't thinking about these particular problems where paper on Entrepreneurial choices starting with the chance of quadrini There is a recent paper by Kan and Kim that thinks about this an optimal stock and bonds portfolio with the participation cost So you get an endogenous rate of return depending on if you're rich enough You get more stocks you get stocks and also you might think that perhaps there is a story about Investors sophistication and that rates of return might be related to different knowledge about different assets But the point I want to make is that I think it's important to go beyond this idea that rates of return heterogeneous and try to think where do they come from So that's why I would like to I would like to talk about this story Because I think it not only endogenize this rate of return But there is also a lot of other empirical evidence that entrepreneurship is strongly related to wealth So besides the fact that the paper is mine, of course So Let me start with the facts again Okay regarding entrepreneurs many entrepreneurs are wealthy and many wealthy people are entrepreneurs so far This is a correlation. So this is my paper with Marco Cagietti. As usual, you know, entrepreneurship is like pornography So what we try to do is we have a specific notion of what an entrepreneur is in the model And what we do is we try to go to the data and be as close as possible to that notion So when you in the survey of consumer finances when you consider an entrepreneur someone that not only is self-employed But only business and actively manages it What you find is that in the richest one percent of people over half are Entrepreneurs according to this very restrictive criterion. If you are more generous, that's 80 percent Okay, so you can see at least that many rich people are entrepreneurs according to this definition There are other I think important facts about entrepreneurs The first one is that exactly like the rich people they have high saving rates Specifically to entrepreneurship. They have a high saving rate before they enter entrepreneurship Maybe they're trying to save to get their business started and after to develop their business There is also a lot of evidence that at least part of the entrepreneurs face borrowing constraints so they need some skin in the game to be able to enter and to expand their firm and In addition, they hold very undiversified portfolios So these are all facts that the model I will show you we reconcile very nicely even though it's a very simple model So as I mentioned briefly before at every period people decide whether to be a worker or an entrepreneur So there is an occupational choice if they are an entrepreneur They are endowed with some ability theta that is persistent over time, but it's stochastic and they run the production function You can add the hiring workers. Nothing really changes What is important is that if you invest capital K there is a decreasing return And then you get back on the appreciated capital Then the other part that is very important is that there is a collateral constraint So a are your assets and K your is your working capital in the firm This simple the question is simply meant to say if you have more assets So you save more you can expand your business more Why are rates of return endogenous in this framework? Well, first of all rate of returns are the marginal product of capital for the firm So there are decreasing rates of return. So they depend in firm size and second Until entrepreneurs are constrained they will invest everything in the firm But after they will split their assets between the risk-free and the firm and so there will be that part too So this model Which I also didn't calibrate to match the wealth inequality, but we'll match other important facts of entrepreneurs Also matches wealth inequality very well Okay, and notice that the fraction of entrepreneurs that in the SCF that satisfies our definition is 7.5 percent So it's an overall small fraction of the population and you know I don't have the time to show you everything but it matches a lot of other facts that you observe in the data So let me summarize what we learn from this entrepreneurship story first of all they can generate a lot of rich people and the key mechanism is that Being able to invest in your firm is potentially very profitable So you need to save to to get this high rate of return if your firm size for at least some entrepreneurs is High the large Optimal firm size, then you will keep saving even when you're rich, right? So this is the mechanism that drives high saving rate even for richer people And the model I think Rationalizes in a very nice way all of the facts. I showed you about entrepreneurs It rationalizes a very undiversified portfolio because initially you're constrained and your return is very high The high saving rates you want to enter and expand your business and the high wealth So let's talk about the fifth story It's about do we really understand earnings dynamics or are we really with our little a one process or something like that They really representing well the risks that people face okay, so Again, I want to start from the facts. Okay, there is a lot of applied Micro-papers that have you know, I Said a few that are recent and particularly suited to make this point But there are hundreds of papers trying to understand how earnings dynamics evolve and basically one very important feature Is that they're very rich much richer than when we have a symmetric year one or even a transit or impermanent component and Importantly high earners face a lot of risk Right, you can see that if I'm a high earner and I face a lot of risk So my risk depends on my earnings level. I might save a lot. Okay, so there is an open question as to whether these is realistic method Really driving savings to a large extent or not? So let's talk about models so back in 2003 Castaneda Diaz Jimenez and victorious rule used exactly this method and the goal of their game was I want to pick why and its Stochastic properties to match cross-sectional moments of earnings and wealth So very clear exercise. Can we do it? Okay? The answer is yes We can match wealth concentration Perfectly as long as you choose the appropriate earnings process However, this is what you get right? This is what you need to do that your earnings level are normalized to one at the lowest level three ten 60 okay, so if you are a thousand at 60 like the awesome state or Tiger Woods You have a 20 percent probability of dropping of having a 99 percent or 99.9 percent depending on which of the other states you fall to and you can see here if you are super rich The only week you can smooth your marginal utility of consumption to save it on because tomorrow exogenous You lose it and it's gone. Okay So I think, you know, they said well It's really hard whether this is consistent with the data or not Because survey data at the household levels typically have a bunch of problems that we don't have the time to summarize So they don't have the higher nurse if they are a small fraction and you miss them, you know, tough luck Okay, so This is what we are doing with the Julia Fela and Gonzalo Paz Pardo. What we do is we go to data Okay, and we use a nonparametric earnings process that has dynamics over the life cycle persistent that the persistence changes over the life cycle and it's very flexible in terms of your earnings level can have very different risks Q-ness and kurtosis and we use this process to match moments from tax data. Okay, so What we find is that, you know, we think that this goes a long way to solve in this criticism that you don't really have the higher nurse In the survey data. So what we find is that one victory, okay? So remember how the poor people are too poor in this life cycle beauty models, right? So here with this kind of risk you have that the poorest 60% of people actually start looking like in the data. Okay, so there is something about understanding this earnings dynamics However, we don't find those kind of extreme drops for the higher nurse. Yes, they have more variance Yes, they have more skewness, but it's not enough to really generate bill gates or even the 1% without talking about the super rich So at least the poor people are realistically poor and When you think of another measure of how people can self-insure you can look at the variance of a consumption as people age Right, if the variance of consumption with people age spreads out a lot that there is some less capacity of self-insurance and these earnings process can match it while the previous year one has a Very large increase in inequality over the life cycle very counterfactual so My view from the earnings risk is that it is important to better think how we should move away from this year one Methodology, especially we you know we propose a very simple way to do it But this is not really why rich people save and you know this data set doesn't have the entrepreneurs This is earnings data, but if you really think it's entrepreneurship, I think an entrepreneur is not a wage shock I think you should go beyond that and you should model it explicitly, okay? So the last the last of the six stories I would like to tell you about it's a bit different It doesn't come from a standard modern which we are trying to match with inequality But I think it has a lot of convincing ideas that this is important Fact one These are our young chicks 74 year olds, okay, we are looking at people who are retired They age we observe them until very advanced ages we have a lot of them and here we are plotting their out-of-pocket so what they spend after government insurance and private insurance space and By age and we do that by permanent income, okay? You can think by education that would look very similar what is going on is that medical expenditure after age nine They really start to rise, but especially so for people with high permanent income Notice that these are also the people who live longer, right? So they can be hit by the double whammy of living a long time and being stuck in a nursing home for a number of years So they face a lot of risk, okay The other piece of evidence I would like to bring to the table to convince you that this is something important is we also have The same chicks here 74 year olds. I'm showing you two courts some started 74 some start at 83 These are they are savings by permanent income So what you see is that the high permanent income people start out return well I mean they have been retired a while, okay start out in our sample with high Assets and they really hold on to them. It's not like they start the accumulating until they are very old The poor people they never save because they rely on government programs like Medicaid and Social Security Oops while the middle income people are the one that show most this saving So what did we do with Eric and John? Well, what we did we Estimated the structural model of retirement and we took the medical expenditure from data And we took them as a shock as other people have done before to do your resources We also allowed for a consumption floor because there are government programs that do provide one So let me tell you what are the key findings from this model. So we have our 74 year old chicks, okay This is one court the dashed line is our baseline model So if you take the 74 year olds, that's how our model would predict they did save during retirement with medical expenditure and long-term care risk, okay so what you see is that you know You have this line for each permanent income level the solid line is what happens if we remove If we remove medical expenditures completely, okay So what is this graph showing you this graph is showing you that once you estimate the model like this medical expenditure Explained a large portion of savings in retirement So there is the question is you know, you're trying to think are we talking about 1% are we talking about the 5%? Are we talking about the middle income? So there is a very interesting paper by these authors that uses a data set that actually looks at upper to middle income people right, you might think well, how far up does this saving motive spread and they find that First of all they find that the same interesting dichotomy that we do say pick a A 55 year old so someone very young for us and if they have below a hundred thousand dollars What is going to happen with long-term care risk is that they know they are likely to get against the government insurance? Which is tested so they actually start this saving faster On the other hand if you have a hundred thousand or more This long-term care risk makes you save more because you're very unlikely to rely on the government consumption floor So this is introduces a very interesting inequality in savings in all the age the combination of long-term care and this kind of government insurance interestingly in terms of what I was discussing of How rich are these people really for which this matters well my answer is fairly rich So what they find in their data set again very representative of upper to Fairly rich people is that the effect in percentage terms is strongest for those at the top 20th percentile of financial wealth, which is much more concentrated than total wealth, okay So it spreads fairly high up in the income and wealth distribution So my conclusion from this is that we don't have yet a model of wealth inequality that allows for this medical expenditure But they are clearly important. They are large and they really affect the savings behavior in retirement and supposedly before And I think they it's very important to understand things. So let me Go to the conclusions and directions for future research part I first would like to say what I summarize what I think we have learned from these six stories The first thing is well, you know, I will surprise you now not everyone is middle-aged Okay, so, you know in the infinitely model everyone is middle-aged Everyone has some assets and is facing disordnance risk, but that's a really big Abstraction if you want to think about these issues In addition, you know precautionary savings is not the only reason why people save and again That is the lion's share of savings in the infinitely lead agent model So I think it's really important to model the life cycle and it's really important because we also did to do A better job of modeling retirement. So the way my friend Cristonetti puts it. It's not a happy period in which you eat cake and wait to die It's a period of really high risk. Okay, and I think Especially with increasing medical expenditures and aging we really need to worry about this even if you want to think about inequality Intergenerational links we have seen are important because the human capita I think We have seen that entrepreneurship has the potential of explaining a lot of wealth inequality and that house of earnings dynamics matter So let me go to the part where I pontificate On the beautiful papers that we should all write And Victor use rule told me I should be very inspirational. So hopefully, you know So I would like to talk about the contents and then I would like to talk a little bit about the methods as Nancy said I really think that a lot of value comes from You know the contents and building being really serious about how you go in about these models So first of all I'd like to talk about human capital Health the family rates of returns on wealth and changes in inequality over time So human capital well, we assume earnings are exogenous Maybe some papers have a little bit of labor supply, but wages are still exogenous If you're writing a theory on how inequality in wealth comes up I think it's fairly important that we start thinking how inequality in wages Translates into inequality in wealth and how these two forms of inequality come up Clearly if we think about policy, the two are going to be very interrelated. You're not going to just affect one Okay, so I think this is very important to health Well, Gary Baker told us it is a very important part of human capital when you're young health affects both your capacity to work and You know your savings through medical expenditures and the capacity to earn and when you're old Well, we I have already told showing you what happens after age 74 So I think you know first of all thinking about the whole life cycle and thinking How health evolves and how that affects labor supply and inequality in wealth and earnings is very important Going beyond that. I think we need to be more ambitious and think about what determines health and how the Determination of health interacts with wealth and inequality. It's difficult But you know, we are getting better at understanding things and we should really try to think about this I've shown you how bequests and human capital really are important Okay, so well, the family is much more than bequests and human capital I really think we need to think harder about what the family does at least more. There's some important aspects of the family for instance The family has the labor supply of both Partners with young and the wages of both partners with young and the medical expenditure of both partners when old and There is the labor supply that you can use to some extent to self-insure and there are marriage and divorce risks So I think it's very important to better understand how inequality is affected by the family Reads of returns on wealth. I think it's very important that we are better understand how they are determined Is entrepreneurship a big one? I think so, but you know, it's always good to show that your beliefs Hold have some empirical counterpart and what are the other important determinants in the paper that I mentioned about We's Opistafari and co-authors about 25% of returns. It's either a fixed effect of the person or observables So there is a significant amount of this variation. We can I try to explain Dynamics of the time super important. We have some excellent papers, but there are just two or three So, you know Understanding why Richard people are getting richer and maybe the middle class is getting squeezed. I think it's very important I have talked only about the static picture, but I think it's really important to think also about the dynamics And I would like to conclude with methods We have seen one or two story at the time I think it's about time that we try to have richer models We you know if you only have one story at the time It's first on the one hand you might be over estimating because you think it's all entrepreneurship But the other hand there might be some interesting Synergies if you have more than one in which two explanations have some kind of multiplier effect So I really think it's important to have much richer model of saving behaviors You know, this brings the question of how are we going to identify these different saving models? I think it's really important to be serious about what drives your model and what aspects of the data identify your model So this is where I think that you really digging into House of level data set is going to really help First of all, you can try to establish new facts the individual or the group level that the model has to match and Second, I think it's you know as we have seen in the superstar or the Tiger Woods shock You match cross-sectional facts you get really counterfactual Risk from the standpoint of the individual over the life cycle, right? So these are micro level data sets are incredibly important to give you a sensible Representation of the risk that people face which after all is one of the key reasons why they save