 Okay, so yes, this is on your approaches to the measurement of progress and it's actually Okay, that's not good Okay, cheers. Thanks. It's actually a summary of two papers or at least it draws on two papers I'm gonna talk primarily about the first one which is the measurement of well-being and progress which is where it was Paul and Anna and Alistair Gray I'm also gonna make some reference to work which I'm actually not directly involved with but which is part of the same project and is closely related by I'm Gordon Anderson, Leo and Anand again Okay, so I'll give a brief introduction and I'll discuss the theoretical framework that we're going to draw on for when we consider these issues And then a very large part of the presentation really is on introducing a data set. We went out and collected some data We'll discuss some new techniques that are suitable for analyzing data such as this Provides a taste of some results and then wrap up with some conclusions Okay, so well, I think it's fair to say there's not yet a consensus on how precisely human well-being should be measured But there are some guiding principles which are beginning to get some you know begin to be generally accepted I think and the first really is multi-dimensional I think you know from from a market send through to the Alcaron foster approach through to more recent work by Benjamin et al And so on and so on it's becoming you know really widely. I think it's it's fair to say it's a new paradigm really that the multi-dimensional approach is essential There's also a growing need a growing recognition. I think often needs for subjective measures Things that which reflect not only objective conditions, but also the subjective experiences that people have and I mean just as one sort of example of that You know something like affluence and technological progress From a purely objective point of view that might be seen as being unambiguously good things but whenever you if you you know whenever you bring in more Subjective things. It's possible that that the the sorts of changes that arise at Senate development process Might give rise to social isolation Depression and so on and so forth so by going out and collecting more subjective data We might learn things that we otherwise wouldn't have known Which might have implications for for policy intervention So this these are really the sort of things that we have in our mind As a framework for this this project So we discussed the development of a suite of indicators of well-being At a theoretical level our approach draws very closely on Sends contributions to the foundations of welfare economics, but we also to some extent draw on the life satisfaction literature We developed data sets for the US and the UK now This has seemed very much as being a pilot study So all of the empirical results are seen as being tentative in a way But the idea really is to illustrate a framework that could be rolled out potentially on a wider scale finest national statistical offices and so on We do illustrate how data such as these might be analyzed and with reference to two new techniques So sends capabilities approach and especially in his 1985 Formulation and commodities and capabilities has three three sort of core core equations in it Firstly the transformation of resources into what he refers to as functioning so activities and beings Secondly the production of what we might call experienced utility Which you know, although happiness and life satisfaction and so on all have slightly different connotations it could be viewed as They could be viewed with this aspect of sense approach of having something in common with the life satisfaction literature So we're going to discuss that and then finally and most distinctively of the approach of course the activities that a person is Able to engage in given their resources and personal characteristics or their capabilities So I'm just to outline our theoretical framework or you know our sort of version of sense approach if you like and Person I oh dear. Sorry Okay, I'll forget about that a person I is that endowed with a vector of resources and a vector of personal characteristics And they can use their endowments to achieve activities or function links They then have and those functions, you know, produced from their resources and personal characteristics So functions might be just you know, things people do Playing games that might be you know, or sort of the person that they are if you like and Person I the second part being that they derive experienced utility From the various activities and states that they engage in and again on some person specific characteristics and then finally they have And that you know our notation here is a little bit of a departure from sound we sound talks about a capability set here We're going to in in a sort of in a bid really to make it more empirically Operationalizable we're thinking of capability. We're going to talk of discuss vectors of capabilities Where each capability is the level of attainment that someone has not the level of attainment But perhaps the the ability to attain something in a particular dimension of well-being so opportunities freedoms and so on Our objective really is to illustrate how this theoretical framework could be applied in empirical work So in 2011 we went out and collected popular implemented population surveys and in both the US and the UK and in each country these were these were broadly broadly representative in terms of social class age gender and so on They were just pilot studies as I say we have about just over a thousand individuals in the US Just under one size of the 700 in the UK Our surveys captured all three aspects of the capabilities approach as I've just described it experienced utility with the various measures of life satisfaction and happiness in this presentation. I'm going to use life satisfaction and Various kinds of capabilities, which I'll say something about in a moment and functioning as participation Our main life satisfaction question, which we'll use just for the record Please read on a scale of zero to ten Where zero indicates the lowest rating you can give and ten the highest overall how satisfied are you with your life nowadays? So for something it's been quite widely used in the literature for capabilities Which there's there's been a much less consensus on I suppose we tried to address the opportunities and constraints that people face in a number of different domains and I mean this was we drew on quite heavily Nice nice poem and sans work in this regard and also some previous work of my co-author Paul and and and we covered your five Domains of home home and domestic and family life that the workplace the local community local environment and access to services and Each domain each of these four of these five broad domains. There were a number of different sub-domains and and Each question again was on a scale of zero to ten Ranging from strongly disagree to strongly agree and in total we captured 29 capabilities across these five broad domains Now in the interest of time, I'm not going to go through all of them I would like to give a flavor for the sorts of questions that we asked nevertheless, so so I'm going to go through two of the domains firstly home So these are the questions that people were asked again to what extent they they agreed with them I'm able to share domestic tasks within the household fairly I'm able to socialize with others in the family as I would wish I'm able to make ends meet I'm able to achieve a good work-life balance. I am able to find a home suitable for my needs I'm able to enjoy the kinds of personal relationships that I want and I have good opportunities to feel valued and loved Then in the workplace and I'm able to find work when I need to I'm able to use my talents and skills at work I'm able to work under a good manager at the moment. I'm always treated as an equal. I'm not discriminated against by people at work I have good opportunities for promotion or recognition at work. I have good opportunities to socialize at work. I Can say more about the other domains if anyone asks that I've got them all here But I'm essentially so we collected a wealth of data on a whole range of different capabilities covering a lot of different aspects of life So that's that's the data set or a quick summary of the data set We also included, you know, many of the usual sorts of variables that you expect to pick up in a household survey on age gender Geographical region it's et cetera, et cetera marital status unemployment and so on So how might we we you know use all this this sort of data? Well, of course, there's all sorts of ways that this that this kind of data could be analyzed that could be put to all sorts of uses So I just want to give really Some very brief ideas about it With so many different dimensions I suppose really there have been two broad approaches in the literature one is the sort of dashboard approach where You know each each each dimension is taken, you know, considered in its own right and considered has been, you know, intrinsically important On its own merit and then secondly, there's the the index approach and you know They've been of course advocates of both of both these approaches We're going to give a you know an example of each approach or something which is consistent with each of those approaches In a very nice recent contribution Gaston Yarnets he provided Multidimensional stochastic dominance conditions for ordinal variables such as these So when these conditions hold, you know, as usual with Stochastic dominance, we're able to make unambiguous judgments about well-being Use it, you know for a broad range of well-being or welfare functions I'm without the need to impose any specific functional form or any specific cardinal scale And and that this is this is you know, I think this is a really nice a nice paper of Gaston because and before that of course you know stochastic dominance has been used so widely in income and so on but all of the There's this issue with cardinal scales with ordinal variables and none of the Traditional statistical tests are valid and so on. So this is very nice But we've got a lot of dimensions here and we've also, you know, not got a huge sample size Even in very big samples and with just a very small number of dimensions It can be difficult to obtain statistically significant results between groups using you know with multidimensional stochastic dominance So we therefore for this paper derived univariate conditions which are Analogist to those of Yarnetsky We could drive the conditions and tests for first and second order stochastic dominance for ordinal variables and And apply those So, I mean they should look I guess, you know, pretty familiar to a lot of people The first the first condition and we're supposed that we have Big ass Categories so for example for life satisfaction s would be 11. We've got a scale from from 0 to 10 So the first orders of gastic dominance being that the the difference in the cumulative distribution functions Between the two groups for each k is less than or equal to 0 For all for all well-being functions are all utility functions If you like that satisfy a simple, you know an ordinal version of a weak monotonicity Condition and then for second order stochastic dominance and of course, you know a more restrictive version where the utility function or Well-being function satisfies a kind of a concavity condition So I'll not go through the through the test and so on but basically that's what we we're going to use so that we can compare Dimension by dimension for example for a particular capability question that we looked at The second approach which and this is the paper by Anderson et al that I'm not directly involved in but they've they've Created a multi-dimensional a new multi-dimensional index of well-being Which could also be you know tweaked as a deprivation index and the it's quite useful for data such as this I mean really the the the rationale for it for that paper and for that index is that Well very much led to the statistical problems associated with very large numbers of dimensions And even even sometimes relatively small numbers of dimensions and the constraints and the demands that that puts on the data Sometimes known as the curse of dimensionality And really it arises from two related issues Firstly the fact that intuitively similar points in k-dimensional space Become further apart as k increases and that the density surfaces become flatter So essentially this means that it becomes more and more difficult to distinguish between the two distributions Essentially massive the center of the distribution empties out if you like as dimensions increase and and the as I say the flattening of distributions makes it hard to distinguish between them so one solution to that and and something which of course has been You know without the alkydian foster approach and so on in the different multi-dimensional deprivation literature is To impose additive separability And this this this like that this solves a lot of problems in terms of the curse of dimensionality from a statistical perspective It does ever make and quite a strong normative theoretical judgment depending on how it's all it's implemented at least And at that being that there's you know generally no Complementarity between different dimensions of well-being in an additively separable approach And a lot of context we you know might be you know We might of course think that there might be quite strong complementarities between Complementarities between different dimensions, so I'm I don't have too much time to go through this detail But essentially what I'm Anderson that I'll do is they take a sort of a compromise Well, they assume that the well-being measure is weakly separable into some H number of of Different groups so what first so H being less than so there's big K numbers of you know total numbers of dimensions analyzed For some H less than K. It can be weakly separated as follows Where it each Z I? There's a vector of distinct elements such the sum of all of those Z eyes is equal to K And then for each of those z eyes in turn and each of these f by z eyes a fairly recent paper by Anderson Crawford and Lester is applied To basically provide an aggregate well-being index, you know for so that's for one of these So so you could interpret this as being that see for one sort of broad domain of well-being And this step assumes only that well-being is non decreasing and weakly crazy concave with respect to each argument So therefore Complementarities are all out there And then secondly what I the Anderson that I'll paper that I'm describing Imposes something along the lines of a well-being function as follows. So essentially It's a quadratic form which permits with a negative definite matrix and permits a situation where well-being is non decreasing in the arguments of F and W's concave well-being is concave in the arguments of F. So it's actually complementarities are large So the fact that the total number of dimensions K is reduced to a smaller number It buys a lot of advantages statistically, but yet Without going as far as imposing that out of separate building losing out on the complementarities So I'm not sure how much time do I have left Five minutes. Oh excellent. Okay, great. So I'm just going to give a test of some results You know for that we've there from these from these data sets Firstly on race and gender our small sample results And this is using the the stochastic dominance the univariate stochastic dominance techniques that I mentioned that there's evidence of Significant gender and racial disparity in the US across a broad range of indicators So whites are fine to dominate non whites at second order at least in all of the domains that we looked at And white's first order stochastic dominate in terms of environmental capabilities with a high significance and second orders stochastic dominant and In capabilities in the community domain the access to services to men and also in the more traditional household income domain and There's also evidence of significant gender disparities in the US where males first orders stochastically dominate females in most domains Not usually significantly so though that could be a sample a sample size thing But second orders stochastic dominate Significantly in both the home and household home capabilities and in household income in the UK We have a much smaller sample of non whites So we struggled to get started right may well be the reason at least why we struggled to get statistically significant results there, but we do We do nevertheless find that the whites of higher levels of well-being within the sample albeit not statistically significantly so the non whites In terms of gender disparities in the UK it's much more of a mixed picture than in the US and in our sample in fact females dominate males in more cases than not But not statistically significantly so the one statistically significant result is in household income where males and males well marginally significantly first orders stochastic dominant females We also as you know ran a number of life satisfaction regressions Now there's really not time to go into the details But what I would what I wanted to say was that in our we firstly ran baseline regressions That on life satisfaction that include all the sort of usual suspects in the literature and life satisfaction, so we have income health at these are Being married or having a partner or has been significantly and positively related to life satisfaction Unemployment being negatively related significantly negatively related very strongly significantly and we also find evidence of the You know the the widely observed you shaped relationship between life satisfaction and age which has been described as a midlife crisis and which Which which Andrew Oswald and various others have found to be true in great eps That result there's also a bit of a literature on whether or not the the the you shape is an artifact at the moment There's a small literature on that but it in any case we observe it in the data What happens then is that we add in our various indicators of capability or you know in a lot of different Specifications, but essentially what we find when we do so especially certain capabilities related to the home in the workplace We find that a lot of the usual Variables that are significant in life satisfaction regressions become insignificant So notably income being married or having a partner and also much less significance And much less size as well for the the you shape relationship with age So an interpretation of this what we think is that the development of certain capabilities might act as important transmission mechanisms via which higher income and living in stable relationships can boost life satisfaction and similarly Capability variables might be shedding some light on the specific factors They might be able to shed some line some of the specific factors associated with the you know midlife crisis that's I and also the Maybe there you know There are various you know Metric points and so on to discuss in the in all as in all life satisfaction regressions But certainly that you know there are very dramatic increases in in the R squared values when you add in even just a couple of capabilities And also they're greatly preferred by information criteria So in conclusion just to wrap up we've developed novel data from the US in the UK corresponding to the capability approach concepts There's a couple of new techniques dash both dashboards and indices home and workplace capabilities Seem to be quite important in life satisfaction regressions Finally, we just to emphasize we see this very much as not really something about these small sample empirical results But something that could be rolled out on a wider scale by national statistical offices. Thank you very much