 Okay, thank you for your interesting presentation. I have a question about correlation you find between satisfaction with government and subjective well-being. Of course people can feel positive about their life and feel positive about government, so these are just positive people. So I think it would be useful if you could use a governance index from another source. And so maybe also use time dimension because you have some time dimension in the afrobarometer data. And secondly, when I saw your outliers of subjective well-being, so the case of Uganda and Kenya, Sierra Leone or Liberia was the other one. It made me think that actually what seems to be most important or extremely important for subjective well-being are shocks, so covariant shocks or idiosyncratic shocks, and of course, you do not have information on shocks in the afrobarometer survey, so I was wondering whether the things you find, so the correlation between subjective well-being and income, maybe partly driven by positive shocks and negative shocks, so I think it would be worthwhile thinking about that. Thank you. Do you want me to respond? Okay, great. Yeah, great. Yes, I just want to find out regards to the the data you used, I was still trying to appreciate how you computed the consumption wealth index. You tried explaining it, but I was trying to get that as if you could throw some highlight on that, but does your data permit you to do some desegregation beyond, so looking beyond the community attributes? I'm asking because depending who is responding to the surveys that you're using, the community attributes might vary slightly. For example, if there are persons living in the community with a disability and whether or not these people are captured in the data, and if it allows for desegregation, you're able to see what the variation would be. Thanks, Erin. This is interesting, but I didn't first a clarification. Your interaction effect is between the Gini index within the reference group and the own consumption wealth or the community consumption wealth? The own. The own. All right, then I Okay, that makes it more interesting even, but I think then there's why is that? I mean, okay, that just needs explanation, but interesting. The second point, I think you're missing an important interaction effect that you need an interaction effect also between the own income, the own consumption wealth index, and the community consumption wealth index. The point is that if you go back to the paper with the auction, we're actually not arguing that there is no relative deprivation effect in Malawi. We're arguing that it's not relevant for poor people in Malawi. In fact, we show there is a relative deprivation effect amongst better off people, particularly in urban areas. Malawi is a very high inequality country, and you can see a huge differences in the importance of relative deprivation, depending on your own level of living. And I think that's plausible when we tell a story about why that would be the case, but you're missing that interaction effect. I actually don't think it's going to show anything, but you might as well put it in. If you mind like a sip, I don't write these down, we'll go away. So thanks. These are all great questions, and this is indeed a work in progress, so this is very helpful. So first of all, adding a governance index is a great idea, I think, and that would be really helpful. And I am quite interested in this idea of shocks, and especially thinking about work about set point theory and hedonic treadmills. And so in some ways, one would, you know, if set point theory is true, something like the extremely high subjective wellbeing when season in Liberia wouldn't make sense, right? And so maybe there's a question here about, you know, a lot of the set point theory and hedonic treadmills sort of work is coming out of not nearly as catastrophic a set of circumstances as civil war and post-election violence and so on. So I think you're quite right, maybe there are things like these sorts of shocks that really transform subjective wellbeing for a longer duration than maybe has been found in studies from the West. So that's really exciting, and I need to figure out how one can do that. Secondly, your question on data, this all can be disaggregated. There are, the Afro-Aphrobrometer data are at the household level, and the reason initially why I aggregated up is in order to get reference groups, but then also I was quite interested in demographic and health survey, health information in which I needed to kind of create communities. It turns out, or create community averages, it turns out those don't matter as much, but your point is taken that there is a lot of variation across communities and that's, I mean, across individuals in a rural area. And so to use a rural aggregate measure loses some of that, and I agree this kind of comes back to an issue with the Afro-Aphrobrometer data where there's not a lot within each particular country, there's not a lot of variation across different regions. And so I'd have maybe eight people in one, you know, in one region and maybe 12 people somewhere else. And so it was quite hard to figure out how do you estimate it, that kind of, that degree of fineness. And so I think I have to kind of think about other, other data sets that would allow me to kind of get at that disaggregated point, but your point's well taken. In terms of marching, the rebellion's points, I mean absolutely, I didn't mean to sort of gloss over the findings of Lakshin and rebellion at all. I do think that there is a really interesting story there. And I think to me what's quite interesting is for the poor respondents, finding neighbors and family mattering quite a bit is a really interesting mutual insurance kind of social protection piece that's really valuable. And I will certainly try to add, I will try estimating community wealth index interacted with the own wealth index. I think that could be quite fruitful. And so what I find thinking about the marginal effects, I don't think I have the slide with me, but the marginal effects when I'm varying the consumption wealth index over the genie, I don't find any statistically significant findings. It's only sort of this direction where I have specific genies for each community that I'm finding a small change in the consumption wealth index matters a lot. And I'm not quite sure why that might be. I think it might be that an increase in a genie is maybe a little bit harder for respondents to sort of maybe notice. I'm not sure. It's an interesting puzzle, though, which is why I'm finding it one way and not the other. Thank you, Michael. This is interesting work. I think you need to kind of go back to the subjective poverty lines now. In a sense, it's like it's not closed. I mean, if you start with your motivation of the concerns that the South African government had about the objective poverty measures, then to go back to those, you now have to construct the subjective poverty line based on your aggressions, which will then be a function of, in other words, fix the subjective welfare, back out the poverty line in money income space, which is now a function of all the other Xs, which include your social wage. Now you've got a poverty line which properly reflects the social wage as the government claims, and then redo the poverty measures with that metric. And that's exactly the same as redoing the poverty measures with a subjective welfare, predicted subjective welfare. And I encourage you to use either predicted subjective welfare, which is exactly the same as using either the poverty lines based on income but subjective poverty lines, given that subjective welfare is monotonic in income. So I encourage you to do that rather than use what you call subjective poverty based on the answers to the question. Because I think the answers to the question are so contaminated by personality effects, heterogeneity in all kinds of ways. But I really believe there is a signal in the noise of subjective data, and you've extracted that signal, I think, very well. The other, just to briefly mention, I've been doing now for a few years the project on trying to establish whether there are significant biases in the types of aggressions you're using by using vignettes. What we do is we go to households, we ask them their subjective welfare, we've done it in three countries now, we ask them their subjective welfare, and then we describe people in some detail with their living conditions, how they live, how much they eat and so on, and we ask the same respondents to tell us what the subjective welfare is of those people. So that way we can recalibrate the scales and see if the aggressions are significantly biased. The news is very encouraging in all three countries. There is a bias, but it's quite small. So you really can learn from these kinds of data. Actually, I think this is probably just an amplification of Martin's point just to say that you know there's this vigorous debate in South Africa about what the objective poverty line actually should be. And I wondered whether in the earlier part of your paper you'd looked at what it would look like with several different objective poverty lines because they're very far apart and of course they reflect different opinions about what it means to be poor. Feeding on from that a little bit then, it does strike me as an interesting thing that you're asking one respondent in the household to talk about the subjective poverty of that household but making a lot of your discussion is about the intra-household process and money metrics and intra-household. I don't know what the evidence says about whether it matters who you talk to in the household and whether it's fair to actually think that you've solved the issue with a single respondent from the household subjectively. Yeah, I think on the issue of subjective poverty lines, my co-author and I have talked about this to some extent. Some of the additional things we've considered are which measures of subjective well-being. We have a sort of battery of those types of questions. So one of the first two questions that we've discussed amongst ourselves are what you include on the right-hand side of the equation and what differences you might get based on what you include on the right-hand side of the equation with various combinations of the subjective types of questions. We also have the question that you've used yourself and others have used in Albania, the sort of subjective ladder. So I mean there are sort of different options in terms of calibrating a subjective poverty line. We looked at this as sort of a first attempt to see if we could draw something out of what might be important before sort of a significant next step. But that's something we've been sort of discussing at length. Second question, sorry, I didn't bring a pen to the podium. Poverty lines. Yeah, yeah. We didn't use different poverty lines. SA came out with three. So one possibility would be some sort of robustness based on different poverty lines. They are quite different. I think what you would expect is as the objective poverty line gets lower, you'd have greater overlap between subjective and objective, but you might then find that the protectors of social poverty change in their sort of order of importance. That's something we're looking at and we're not quite sure if it's going to be more of a robustness check or if it's something that could be worth exploring further. Sorry, Murray, I forgot yours as well. It was just about asking a single respondent of their subjective well-being and thinking that that represents the household, especially as you go on to talk a lot about intra-hustle things and equivalent scales and stuff. Yeah, I mean the work that we found interesting there was Bookwalter, who sort of looks at South African data and suggests that the person responding for the household generally tends to evaluate the well-being of the household based on household characteristics rather than personal characteristics. The difference with that work is they had to assume that the head of the household was the respondent, because we got more nuanced data and were able to identify exactly who was responding, but whether that has an implication or how much further we could take it with this particular data. And obviously Dory, my co-author and others have found that when you control for individual level factors from NIDS, where you actually get individual information on each individual in the household, it would make a difference. And I would say a similar response to whether the person responding is a main income earner or her or his position in the household would probably make a difference. So we weren't controlling for the individual contributions of that person to the household, but we were able to control for her or his personal characteristics, including health education. So I think we maybe got at some of that, but could do better probably. Thanks very much. Yeah, quick question, but maybe I'm missing something really basic, so forgive me. But it seems to me that an interesting null hypothesis that you could test in a more rigorous way perhaps is whether employment leads to happiness. That would seem to me probably a more, well, a useful thing to look at, right? Because then you could use your, you know, being in the treatment as the instrument for being employed. I don't know whether you've considered that. They would just seem pretty obvious to me as an addition to what you're doing. Sure, sure. Trying to use happiness in this way so directly for policy to derive policy restrictions is tremendously important, but I think there's one problem with it in the European happiness literature. It's well established that changes are much more important impact on happiness than levels. You win the lottery, you're happy, then you get used to it and you're miserable. You break your leg, you become crippled, you're miserable, you get used to it, you become happy again. And if a temporary intervention like this, you can imagine people become employed, they get used to it, they're happy for a little while, they get used to being employed, the happiness goes back to the mean again. And whereas the people who were employed because of the voucher and then once the scheme ends or whatever intervention it ends, they go back to where they were before, they're miserable. And so this will pull down the happiness impact of any temporary intervention. And I think this is an important, something that you have to examine in studies like this or else you're going to be under-reporting the happiness impact of interventions. There seems to be another interpretation, I'm not sure about this, but let's walk through another story which suggests that your results are not at all surprising. Suppose that unemployment is picking up a latent personality trait. Just bear with me, there's some latent personality effect which is constant and we know that personality influences the answers to happiness questions. It's hugely important in the psychological literature. So what would happen then, I mean if it is, and there's a rationale for that as well, that it's a latent personality trait, the rationale would be that elementary economics, that holding constant income, being unemployed just means you have more leisure, so your happiness should surely rise. Holding constant income, if you're not holding constant income, of course there's no different story altogether, if you're controlling for income, you should expect that. So that all hangs together, says okay, unemployment effect is just a personality, latent personality trait, the economics and the psychology point in the same direction. What would happen then if you exogiously switch people into employment through your randomized control trial, nothing, you would have no effect at all. So try, I mean think it through, but if you did that, you see if it's a personality trait, nothing to do with the, it's not an effect of unemployment per se, it's an effect of a correlate of unemployment which is in your editor, which is the personality trait, then the exogenous switching will have no effect, because the personality trait is what's driving the aspect of subjective welfare. Can you say something about that? Okay, well great, thank you very much. Very nice comments. Yeah, yes, exactly, but I mean like when we started off, that wasn't what we were interested in, that's what I'm saying, it is the first round that we're actually presenting these results, and we were hoping that we would find a bit more on the active labor market policy itself, because that was our interest, that was what we were supposed to evaluate, but you're absolutely wrong, I mean one can take this absolutely further from there, so yeah, I agree. So in order, yes, happiness as a policy, but rather sort of, and that I think is sort of the idea also of the paper that I sort of don't like to say, but if you're the co-author, because they also say, I mean like how long this effect might last is not clear. And maybe one way for us to also look at it would be to see actually people that sort of transitioned into a job more recently towards the interview. Do they actually report differences? So can we use potential to the time dimension from transitioning into the interview as a measure for that? I think that might be a way forward to sort of try and unpack that, but it's a good point, thank you very much. Yeah, the exogenous, yeah, but in order to do that, clearly we need the panel dimension, I would assume, or not necessarily. I know what I'm like, okay, but that sort of, would that go back to basically your argument as well? I mean, so if, for example, a personality trait, so if it doesn't change, right, so it basically means also you basically just have a potential change, so your response is just to change, and then you come back to the level that you would have anyway. I think it's a different point, but it's very hard to get data on that, but I think there is, I know there's some Dutch work where they have these wonderful panels of the whole Dutch population, which enables them to look at the question of is to compare employment experiences with subjective well-being, and they came up with a finding a few years ago, I don't know whether they still would say this, that it's not unemployment that makes people miserable, it's miserable people who become unemployed, which is interesting. But that's my point, so I wouldn't need the panel dimension. No, no, but I hear what you... What I'm saying is not something that you can go and test, it's a theoretical interpretation of the argument. Now, if I put an empiricist hat on, how would I test it? Well, I would need either to control for personality attributes as data, or I'd need vignettes. Because the personality attributes are saying that the subjective welfare scales are contaminated by something in the error term, we have to clean that out. So I'd need to recalibrate the scales using vignettes, or I'd need personality data. So you don't have either, so that's not the point. In a sense, you could... It's just to say, when you think about the results, what do you learn from them, to clarify that this is other interpretation, which you didn't point to, and I can't see why it wouldn't be valid, just thinking it through intuitively. I'll take that question just now. We'll just... quickly. So, yes, we do have some information on the jobs, and the differences aren't significant. So the kind of jobs that you would get through the treatment are not necessarily different to the jobs that people got in the control either. In general, what type of jobs do they get? Do they focus on personal possibilities? Yeah, so now, absolutely. I mean, again, and it comes back to the kind of sample that we've got there. It's predominantly less educated. So the kind of jobs that you would get would be contract and would be more menial. So they're not very demanding or most probably stimulating jobs, absolutely. And the one measure that we do have is actually sort of a self-reported measure if they're considered actually a job that we started playing around with as well. But that's across both groups. So this is just to follow Martin's point about another, yet another interpretation would be one of kind of a William Julius Wilson culture of poverty-style interpretation where you have this astronomically high unemployment rate among a particular subset of a community or of a country and it's quite possible that as you kind of hint at at the end, maybe employment is not something that they see as important to what makes them happy, you know, and if no one else around them or very few people around them are able to get decent jobs, maybe that's not something that they're anticipating as being important for them. Okay, we have the gloves here. Thank you very much.