 I should note that this is work that I've done with Tassu Woldahana, and ideally he would be here presenting this, but we invited him, but as the president of the university, he found it difficult to get away, so you have a poor substitute in me for him, but I'll do my best. This is one of the chapters in the book that Andy had pointed out. I also did one on Madagascar, and as you may have noticed, that was in the poor performing countries, and I've presented that before. I'm happy to present the success case here. This is one of the chapters in which we use the utility consistent measures for the poverty line and the standard software package that I'll talk a bit about as we go along. Just to start with the basic question, the simple question is what are the levels of poverty in Ethiopia and how have they changed? We typically measure these with household surveys, and so we have three household surveys that were available to us for this analysis, 2000, 2005, and 2011. There is a 2016 survey that was conducted, and the Central Statistical Agency has analyzed that and have written their analytical report, which has not been made public to people like me yet, and the data, we're working on getting it and analyzing it. We'll work with these three years with the understanding that there's more in the future. The question is, do snapshots of poverty from household surveys taken in these three years represent long-term trends, or are they due to short-term shocks, or is it some combination of the two? When we ask about how monetary poverty has changed using these household surveys, this measurement, as Andy had indicated earlier, really depends on the consistency of a number of things. The first one being the surveys over time. The questionnaires, are they similar? And the sampling and the like. And then once you have the data, do you construct the consumption aggregate in a similar manner? And then the poverty line is always that tricky next step before you can apply this to FGT measure, for example. And the difficulty here is maintaining both specificity and consistency. And I talk about this because this is one of the main ideas of this software package. With the specificity, you have a common consumption bundle that represents the consumption patterns of the poor in a particular region. So you'd want to use the consumption patterns to value the minimum or calorie requirements for a poverty line. And yet they need to be consistent across regions and over time, representing the same level of utility among the right at the poverty line. So the original poverty lines for Ethiopia as conducted by the CSA were calculated in 2001 based on a cost of basic needs approach, and then have been adjusted for inflation and regional price variation for comparability over time and across regions. But one thing to note is once we have a poverty line that we are comfortable with, we can think of this as a cost of living index for those around the poverty line or the poor that allow for interpersonal, intertemporal, and spatial welfare comparisons. And this lets us look at distributions of consumption. Now the approach here then is to use this please software. It's the poverty line estimation analytical software that's available on the wider website. There's a link to it when you go to the book for this project. And the idea here is to calculate food poverty lines that are for very spatial domains. So in Ethiopia we have 20 spatial domains. We have urban and rural and nine regions and then Addis and then Harari, which we have combined together. They're anchored on domain specific calorie requirements. And it's based on the least cost bundle reflecting the domain consumption patterns. There's that specificity that I was talking about. And then for consistency the software package, which is a combination of Stata code and GAMS code that you can then you apply for each your particular country, tests for revealed preference. The idea being that if households in a particular region should be choosing the least cost bundle that achieves the calorie requirements that we have a sense of for the poverty line at the prices in the region. Now if we compare this consumption pattern to other prices and we find that the consumption bundle is cheaper elsewhere, they must not have been choosing the least cost bundle. And so that violates this consistency. And so the program adjusts the consumption bundle or the consumption pattern, this consumption bundle for the poverty line, the region specific poverty line, in a way using an entropy method to minimize the distance from the original consumption bundle to the new consumption bundle that's utility consistent to come up with the food poverty lines. And then the food poverty lines are scaled up in standard ways to get region specific poverty lines. Now the data that we use are the three years of the household income consumption and expenditures survey and the welfare monitoring survey, 2000, 2005 and 2011. These are nationally representative multi-purpose surveys with sample sizes that have grown over time, which is not an issue for for comparability, but just to be aware of this. Now data issues that have us cautious in our analysis include the fact that well first of all the questionnaires are identical for all three of these years, which is encouraging for consistency reasons. But the number of food codes differs. So the categories that people are asked for in their consumption in the expenditure module increase between 2001 and 2005 and then decreased. And so we just need to be careful about this because when Mendo Pradhan years ago had did some analysis in Indonesia where he found that when you ask people with more detailed food codes they tend to report more consumption. And so we just need to be aware of this and there's not much we can do about it other than just be cognizant of this. The timing of the data collection is also something that concerned us a bit. The first two years are similar in that they were collected over two short time periods July and August and then January and February. In the 2011 however was collected over the period of one year, which now if you look at inflation rates and you notice that those the dashed lines are show the survey years. We noticed that food poverty excuse me, food inflation rates were around 35-37% over the course of the collection of the data for 2011, which gets us a little concerned in that you know we need to adjust for that and we do by adjusting by deflating using temporal price deflaters for the quarters within the survey year as a way of capturing that intra survey inflation that occurred. We also restricted our analysis for the 2011 data to just those time periods that are comparable to the earlier years and we found that there wasn't any qualitative difference and I don't report that here but just we remain a bit cautious in our interpretations and this is why it's important to consider other data as compliments to the household surveys. Now to get to the poverty estimates and I'll give it some context after this. So our initial poverty estimates for 2000 for the year 2000 were roughly 47% of Ethiopians were poor and that fell considerably by 2011 by some 23 percentage points with most of this occurring between 2005 and in fact practically all of it between 2005 and 2011 and we see that the more that the depth and severity of poverty that also show similar patterns that little change between 2005 2005 but then a big decrease between 2005 and 2011. So we see this decrease in poverty but it's kind of it's they're different experiences in different parts of the country. In urban areas we saw in fact a rather sharp decline in poverty between 2000 and 2005 and then a bit more moderated decline though still nine percentage points is quite big but this also occurred this time period you had a very high food inflation in the country and similarly the depth and severity of poverty showed similar patterns. Now rural poverty was higher than urban poverty which is not surprising. In fact while the urban areas experience greater declines in poverty early on in the decade rural areas experience more of that decline in the latter part of the decade which is why at the national level that's where we see the the big change given that roughly 85% of the population lives in rural areas and indeed you see very slight increase in poverty as measured by all of our measures between 2000 and 2005 before a decline. I hate it when people put lots of numbers up like this but I just put this up here right just to show you a number of things one is these are the these are our spatial domains for which we calculate different different poverty lines and let me just highlight the the area that that started out with the lowest level of poverty was Addis with 30 roughly 35% in 2005 and the major decline in fact all the decline in poverty in Addis as measured by our surveys was between 2000 and 2005 right and these are remarkable drops in in poverty considering that it fell by some 24 percentage points having started it at 35% the poorest region was rural afar which started with roughly 80% poor and experienced some 37 percentage point drop in in the number of people who were where the percentage of the population that was poor and unlike rural areas in general a lot of this change in fact all of this change was in the early part of the decade right so in the rural areas in general what we'll see in a minute again so what we're seeing is we're seeing very we're seeing different experiences in terms of where poverty was was declining or if it was declining in certain parts but over the course of the decade you see that there is a that all of the in all of our spatial domains poverty declined and in some some cases quite remarkably such as in rural Deirdawa 56 percentage points now as I said earlier we can use the poverty lines to to adjust the household consumption so that we can make regional and temporal comparisons and so here we we convert everything into 2011 Addis prices and plot the CDS or as as Martin Rebellion has called them the poverty incidents curves and so there are a couple things to note here so this is at the at the national level and so well we saw in the in the previous slide that there's lots of different experiences in changes in poverty if you're looking for that where the 2000 line is it's right there right underneath the 2005 line so on average nothing changed but that's on average and none of us really lives on average right and so the different regions experienced different different levels of of change but on average really nothing changed nationally then the then there was a a drop in in poverty between 2005 and 2011 and so and we see that it's almost across the board the top 95 percent of the of the population experienced this decline in poverty so regardless of the poverty line around you know where we have it right we have a close to first-order dominance but you'll notice that there's a considerable overlap at the lower end of the distribution and so we don't see the poorest of the poor are no better off and that we didn't capture with our with our other measures and this is you know kind of getting at what Martin Rebellion mentioned yesterday that while we talk about the success in Ethiopia the poorest of the poor are the poorest five percent there tend to be stuck there and we see that in urban areas so here we see a considerable decline in poverty almost across the board again between 2000 2005 and then 2011 though not as as much improvement at the at the lower end of the distribution again oops sorry trying to go through this a little quicker and then here we see the the rural areas it's basically this the story we see at the at the urban levels or at the national levels given the predominance of the of the rural population and there we go so here we see the very slight increase between 2000 and 2005 and then the the large decrease over between 2005 and 2011 okay so just very quickly inequality while we've been focusing on on poverty inequality has rose it not monotonically but over the course of the decade it rose whether we use the the genie coefficient I'm just going to use this whether we use the genie coefficient or the tile index and for urban and rural areas which again is not that surprising given that we see the that the poorest of the poor are not much better off okay just some context so since 2000 all right Ethiopia has has been hit by persistent weather shocks right periods of high inflation and they post election crisis after 2005 nonetheless over the over the course of the decade and continuing Ethiopia has has experienced what what Paul Darosh and Emily Schmidt referred to as a changing economic landscape right and this is in part due to the government's agricultural agricultural development led industrialization strategy what Joe Stiglitz referred to this morning as this emphasis on on agriculture as a as a means of growth and the idea here being that rapid agricultural growth is a means of accelerating economic transformation and reducing poverty and so what is this what is this changing landscape increased agricultural production initially due to expansion of cultivated area but increasingly evidence that it's due to higher yields this may be due to the introduction or the massive expansion of extension agents and evidence of improved modern input use whether it's maize and tough seeds or chemical fertilizer improved infrastructure it has also appeared to contribute in the the road sector there's a road sector development program which has resulted in in thousands of kilometers of roads being built that we find has integrated wholesale markets across the country and the building of the hydroelectric dams provide some the possibility of more electricity though electrification remains quite low in rural areas telecommunications with the availability of cell phones and survey area where we're very remote you go up to a hill and get a signal so that that improves communication across markets and the production safety net the productive safety nets program which was introduced in 2005 has has also appeared to provide some stability in terms of food security and and investment in in agriculture now in terms of complementary data they're in the time period so is this a snapshot we have national accounts data non non-nationally representative data and non-monetary measures I'll very quickly talk about the national accounts data and non-monetary measures to give us a glimpse here so here from the national accounts data per capita GDP as we can see so after about 2004 we see that the per capita GDP has has risen and continues to rise to give us a sense of where we're going from there I'm almost there and and if we look at a sectoral breakdown of this agriculture has been an important part of it but we see that that services as as Eric Thorbeck mentioned in his talk earlier that the services have become an important source of employment and production whereas industry has grown but not quite as fast turning to non-monetary measures we see that under age under five stunting has decreased over time so we're seeing that well we have these snapshots of poverty and improvement with the the national representative household surveys we're seeing also stunting that is falling using the DHS right and using the retrospective part questionnaire of the of the DHS we also see that infant mortality has has declined over time and persistently over time so we have these two measures of of non-monetary measures but we also find rising net schooling enrollment rates and the diets have improved in terms of food quantities and calorie intake falling food shares in terms of total consumption and a shift toward high value foods right so concluding remarks here right using this utility consistent approach to measuring the poverty lines we have snapshots of poverty but when in which we see steady but uneven progress over the decade with the headcount ratio national headcount ratio falling from 52% to to 30% in urban areas the gains were in the first half of the decade in rural areas in the second half of the decade but the discouragingly the poorest of the poor are no better off and we have Ethiopia's changing landscape that the context in which this has taken place and we see persistent improvements in in as measured by national accounts and non-monetary measures so in short these snapshots from the household surveys appear to represent long-term trends though uneven and a bit more muted when we look at the non-monetary measures thank you