 My name is David Stifle and my co-authors are here, T.R.A. and Fali, so I'll come to them here as well. And so this is the case study using the poverty toolkit for Madagascar. And so essentially two questions that we address in this paper. The first is quite simply, what are the levels of poverty in Madagascar and how have these levels changed over time? And the second being is the do the snapshots of poverty that we get from these household surveys, and we have household surveys from 2001, 2005 and 2010, do they represent long-term trends in poverty and how much do they represent short-term shocks or some combination of the two? And for Madagascar this is a very relevant question and when we're thinking about poverty here. So the first question, how has monetary poverty changed? And so this depends in large part on the consistency of a number of things. So starting with the surveys over time, the survey instruments. And this is one thing that INSTAT has done very well. Following some work that I did with T.R.A. and back in 2000, trying to reconcile differences in survey designs in the 1900 or 1990s, there's been a big emphasis to maintain the consistency of the survey instruments and the questionnaires. And in particular the expenditure components. And so that is something that we're not so concerned about. The second is the construction of the consumption aggregate and how that is consistent over time. And here again INSTAT has been very good about that. In fact the same two guys have done the consumption aggregate for all three years. And these are guys we've worked quite closely with. So the poverty lines. So here we get into this question of specificity and consistency that Channing described earlier. And I should note that I'll skip through some of my slides because I prepared this thinking that Channing was going to talk about something else and then they changed the paper. So I'm thankful that he did so. So the original poverty lines for Madagascar, well they started with the 2001 poverty line. And each year for each of these years they've updated them for inflation and regional price variation. As is common practice. That's my sense. And so we are going to apply the poverty tool kit to here. And I should note that just as an aside that poverty lines can also be used as cost of living indices that will allow us interpersonal welfare comparisons. So this will use this not only for a threshold but also to deflate our consumption aggregates as we do poverty incidence curves. So this I'm just going to skip right through this because this is what Channing had just described. And go to our data. So the data we have here are the household survey data that are nationally representative that are collected by INSTAT. And for the three years that we have, we have growing sample sizes and this has to do with changes in the regions. Initially the strata were six provinces now they're 22 regions. For our purposes here we're sticking with the six regions for comparability purposes in calculating the poverty lines. And again the questionnaires are quite similar. So that's what we're working with. Now for our poverty estimates from using the tool kit. So this is where we the aggregate is constructed and the tool kit is applied in which we do the revealed preference tests and get these utility consistent poverty estimates. And so what we find at the national level is that poverty has risen over the course of the decade. And we can see that the significant increase occurred in the first half of the decade in terms of the head count ratio. But we see that there's actually not a whole lot of action going on among the more poverty sensitive measures of poverty. Now when we turn to urban areas we see that this is where poverty has risen considerably. Again in the first half of the decade. And here the more distribution sensitive measures also record an increase in poverty. Now in rural areas the increases are, the levels are higher certainly but the increases are not quite as large in the rural areas and in fact if we look at the distribution measures of poverty there seems to be a decline. And so naturally we turn to our poverty incidence curves to get some intuition about what's going on here. And so here you see at the national level you see that yes around the poverty line there are more poor. But as we look at the lower end of the distribution we see that the poor of the poor have higher levels of expenditures which is captured in the depth and severity of poverty. In the urban areas for the most part the time period between 2001 and 2005 we see a worsening of household consumption levels. And this persisted through the rest of the latter half of the decade. In rural areas on the other hand we see a marginal increase in poverty right around the poverty line but we do see this improvement in well-being for the poorest 50% of the population in Madagascar. So what's going on here? Well then if you look at inequality I'll put this all in context. We find that inequality over the course of the decade is decreasing. If you look at the Lorenz curves they don't cross at the national level but they do cross at the urban level especially between 2005 and 2010 and it's not surprising we get different results for using a tile and a genie coefficient. So what's driving the decrease in inequality is in the rural areas and much of inequality is explained within group which is not surprising giving the large proportion of the population in the rural areas to begin with. So some context. So these surveys were conducted in 2001, 2005 and 2010. What was going on in this decade? Well shortly after the 2001 survey was conducted there was a political crisis that disrupted urban areas in particular but the country as a whole. So just after the crisis was resolved the international community came to the aid of the elected government and provided support for the government. So there's the shock with some recovery based on support from the international community. Now right around the time of the second survey there was a rice price crisis and an appreciation of the exchange rate combined with an increase in rice prices and international rice prices meant that rice prices in Madagascar increased by some 50% and given that rice is the main food consumption item households hit generally quite hard. And so this was just around the time of the second survey so we're having repeated shocks. Now between just after the second survey, the 2005 survey we saw some growth I'll illustrate this in a minute but then just before the 2010 survey was the beginning of the most recent political crisis and one that is this is ongoing and because this was effectively a coup d'etat or at least interpreted as a coup d'etat by the international community, the international community responded with condemnation and reduction of aid to only humanitarian assistance. So all development assistance effectively was withdrawn at least initially. And so you don't have that same sort of rebound and unfortunately this is still going on. There's hope that the elections next month will take place. So how can we better understand what's going on? Well we'll use some complimentary data to understand what's going on in the time period between the shocks. We used national accounts data we used non-nationally representative data an urban labor force survey, some rural panel data that is not nationally representative but gives a sense of what's going on and non-monetary measures. So this graph here to me just gives you a great sense of what was going on between the surveys. If you look so the red dots or the head count ratios from the respective surveys you see that shortly after the first survey the political crisis occurred and there was a sharp drop in GDP, GDP per capita followed by a rise in GDP I was working on that aren't a Libran approach but followed by growth thereafter but apparently not enough to reduce poverty but you continued to have growth following up until the 2009 political crisis after which we had the household survey. So the timing of the surveys really if you just look at the surveys and the poverty over time it looks like there's a progression of increasing poverty but if you look at it in terms of the timing of the shocks the shocks seem to be pulling the economy down while there's some recovery between the survey periods. Now in terms of the sectors that is most affected by this it's the service sector and given that the service sector is a very important one in urban areas this gives you an indication of why the urban population was hit and what's interesting to see is the shift of the urban labor force following the shocks from the service sector into agriculture so the agriculture provided that social safety net in the urban sector. Now our other non-monetary measures under five stunting over the course of this of the decade we saw moderate improvements. Infant mortality rates aside from a sharp rise in 2002 following the financial excuse following the political crisis we saw otherwise we saw moderate improvements over the course of the decade. And over the course of the decade we also saw persistent rising net schooling enrollment rates low can still low in general 75% for primary but quite low for secondary on the order of 25% for lower secondary and even lower for the Lise level which is going to be a challenge for opportunities and poverty reduction in the future. So briefly these snapshots of poverty missed the underlying long-term trends and the growth and improved well-being were interrupted by these. Unfortunately the latest political crisis is in its fourth year and it seems to be like more not a short term shock but we hope that they'll move beyond that. Now I just wanted to take a couple of minutes to compare the utility consistent poverty estimates with those from the original INSTAT estimates. And so what we see here on the left are the original INSTAT estimates and then on the right we see the difference between those from the utility consistent estimates and the INSTAT estimates. And we see that the INSTAT estimates are considerably larger at least for the head count ratios and the depth of poverty though even the P2s are a bit mixed. In the urban areas we see a similar sort of pattern of increasing poverty and then a marginal increasing poverty between 2005 and 2010. The pattern is similar there. The pattern is not similar at the national level and again the levels are higher than the utility consistent measures. Considerably higher at the rural level and the pattern here is remarkably different between the latter years with a very high rise in poverty to 2010. And again higher around the poverty line but at the tails it's a bit of a mixed bag. So why the differences? The consumption aggregates are the same. So that's not the source of this. When we look at all the utility consistent poverty lines we find that they are all 13 to 46 percent lower than the original estimates. And so this could be due to differing calorie requirements because with the toolkit we allow for, okay almost there, we allow for differing calorie requirements based on demographics and fertility rates but this is not the problem. In fact the calorie requirements are higher with the utility consistent measure so that we would expect it to go the other way. So differing baskets for costing calories we'd hope to compare this to the 2001 survey but we can't find this data code and so we're in a bit of a bind for making that direct comparison. But one of the things that we can consider is that the regional 2001 poverty lines for the original estimates are updated with inflation rates calculated from the major cities each year. So you talk about the issue of rural prices. Here they're imposing inflation in the regions at the major urban areas in those regions, not the smaller urban areas. So it's just a subset of this. So this has a real issue for rural areas and further the CPI baskets that are used for inflation place greater weight on non-food items than the utility consistent basket. So I'm suggesting that the household survey is representing the consumption weights differently in specific ways. So just a brief again, the utility consistent poverty estimates suggest that urban areas are worse off while their greater percentage of rural are poor. The rural poor are actually less poor in terms of the higher sensitive measures of depth and severity. And the snapshots that we have are during a tumultuous decade and that the underlying trend of growth and welfare improvements are interrupted by these shocks. So we don't want to necessarily interpret this as a long-term trend unless Madagascar is repeatedly hit by these shocks which is not a trend that I would hope to see. Thank you. Okay, let's follow the same pattern. There are two questions there and let's take them first and then give David a possibility to respond. Thank you. So the way I'm interpreting this I'm saying it's looking, it's very hard for impossible for me to believe that you're making utility consistent comparisons between the two sets of poverty estimates. So it looks to me as you must surely be using a lower reference utility level with your so-called utility consistent poverty lines. You just don't know. So I think putting them together like that is really, really not right. I mean the official estimates may not be utility consistent internally but if they're at a higher reference utility level and that's all that's driving this I think you really should say that and make that clear. And you can't really publish it given the problems you've got. But I think then the second comment and this is a general comment but I have to go to something else so I can't... I'm not a great fan of this similar method but I'm sort of sympathetic but I think back to that basic problem of people accepting that the same utility function I'd be curious to see how all this differed with essentially the other method of doing utility consistent poverty comparisons which comes from, in the history of this we went down that route of using reveal preference decided we didn't go that direction, turned back to another route which was essentially the numerical method that we use routinely where we basically say calculate, if you make a first assumption about what the poverty line is, calculate a bundle in that line, in that neighborhood of that and then iterate till you get convergence and there's a version of that that Angus Deaton has come up with which is just an extension of it which we now use routinely for the purchasing power parity rates for the poor. And that's really become our main method of doing so-called utility consistent comparisons but in a sense we argue that they're utility consistent but without imposing reveal preference without imposing a common utility function. The common utility function here is common across commodities with absolutely no heterogeneity in any other dimension including regionally. So I'm just commenting on that difference in the literature but it'd be very interesting to do something comparative on these two approaches. There was another question please. My name is Haruna Sekabira from Uganda. I work with IFPRI. So my question just comes with your concluding remark number three. Actually that's why I have maybe trouble you could throw more light. So you were saying that shocks interrupt the estimates that perhaps you produce. So my concern is how do you or how do you generate representative poverty estimates for long term strategic planning independent of the effects of those shocks particularly for the case of Madagascar and many other countries in Africa which are characterized with several shocks each other the other time. Thank you. Quite frankly I'm not sure how you would do that. I think the important thing is to understand that these shocks do take place and disentangling those long term effects from the short term shocks so that we can interpret those changes in poverty. Otherwise we rely on the household survey to get those estimates. As for the differences in utility yes I think that we need to be clear on that. Did you want to respond to the second question? Or maybe we can take it to the end because it's a broader question. I think it's part of the broader. Okay thanks. Thanks so much.