 Good afternoon. I'm presenting a paper that looks at public investment and inclusive growth in Uganda. This is collaborative work with colleagues at the Economic Policy Research Center, notably my boss, Sarah Sawanyana, and those at Cornell University, that is Steve Younger and Elaine Hill. Basically the reason or the basis for this study was to look, was because the pace of poverty reduction in Uganda was uneven and there were calls or investigations to find out if public investment in infrastructure, that is roads, electricity, can be used to bring about some equitable growth, especially benefiting the poor. The presentation will give some background on why we focus on inclusive growth in Uganda and notably the infrastructure. The infrastructure here, it's mainly roads, health facilities, agriculture input markets and related items. Then the data we use, the decompositions that try to explain, we're basically trying to explain to what extent does infrastructure explain the special differences in poverty in Uganda. Then we also try to simulate what would happen to poverty if the infrastructure that is in the richest region is accorded to the poorest region, what would happen to poverty and then the conclusions. By way of motivation, the drivers of economic growth and poverty are a much of public concern in Uganda. We have more or less maintained positive GDP growth rates in the past 20 years. This has been coupled with poverty reduction that nearly holds over the 17-year period. Nonetheless, this is happening in an environment of worsening inequality. During the year, during 1992-93 to the period 2009-10, the genicoefficient, which is the standard measure of inequality, increased from 0.33 to 0.42. As such, this has become a matter of public concern. We're seeing very high growth rates, consider a reduction in poverty, but the level of inequality is worsening even within regions that are considered poor in Uganda. Another thing to note is that the pace of poverty reduction has reduced in the recent past. This table shows, tries to demonstrate the large variation in income poverty across space in Uganda and my point of reference here is the geographical regions. It shows the trends in income poverty mentioned from the period 1992-209-10. For the whole of Uganda as mentioned earlier, it decreased from 55 to close to 24% over this period and the trend is consistent for both urban and rural areas. When it comes to the geographical location, which is indicated by the bottom half of the table, you find that most of the progress has been recorded by Central Uganda. Nonetheless, by 2009-10, there were very wide geographical differences in poverty rates. Specifically, the incidence of poverty in the poorest region that is Northern Uganda was more than four times that of the richest region in Central Uganda. I should note that this is partly due to the history of conflict that has waged in Northern Uganda for around 20 years, although this tops around 206. Previous authors have shown that growth is necessary but not a sufficient condition for sustainable reductions. And there is also, authors point to the fact that inequality is partly the inequality that we see increasing in Uganda is partly driven by unequal access to public goods, that is roads, both Tamak, Maram roads and all weather roads, as well as access to infrastructure services. Nonetheless, also the previous literature shows that both returns and returns to education are lower in rural areas and in urban areas outside the center. So the main question here is how can public investments, if at all possible, be used to reduce such discrepancies? This is what we try to answer in this paper. In terms of methods, the general approach we follow here is to estimate regression decompositions following the Ohaka Blinda style using a national representative survey. The type of the living standards measurement survey conducted by the World Bank. And we basically try to determine the extent to which spatial differences across the four geographical regions in Uganda, the spatial differences in outcomes are due to differences in endowments or secondly returns to these endowments. For example, skilled labor infrastructure and the like. In these estimations or regressions the dependent variable is household consumption by adult equivalent and more or less a standard measure of capturing welfare in such surveys where consumption is a key welfare variables. And we estimate these regions on the four geographical regions of Uganda. Most notably we try to compare the richest region to the other three. The richest in this case as I was showing compare to the other three and in this case we are trying to see if the poorer regions had either the endowments or returns to endowments similar to what is happening in central region what could be the impact on poverty. For just as way of exposition if you have two regions say northern and central the decomposition is expressed as below is the log of consumption expenditure which is a function of a number of determinants of welfare like education attainment of the household head and the community's infrastructure. I should note from the onset that we use community infrastructure rather than individual household infrastructure to reduce issues a number of issues relating to callinearity. So the X's are the respondents so the households endowments specifically welfare producing assets that a household has or they have in their community. For example I should note and then the bitters are the returns to these endowments. When I say they have in their community when you include a variable like having electricity this refers to having electricity in the community doesn't mean that this particular household is connected to the national grid. And then we estimate the expected value of the difference in welfare of these two regions as expressed between and most especially if we rearrange those equation two we get the average difference in the log of welfare between these two regions composed of these two terms. The first term shows the return effect that is the difference in average welfare that is due to rates of return those are the bitters given any two regions for example central and northern Uganda then the second term is the endowment effect which shows the difference in the average welfare that is due to difference in values of the X's for example education attainment of the household between the two regions. In terms of the data we use the Uganda national household survey there have been other surveys in Uganda these are more or less panel surveys and are far smaller in size so this survey was conducted during May 2009 and April 2010 by the Uganda Bureau of Statistics and it covered all areas and all geographical regions of Uganda it is similar to the living standards measurement survey promoted by the World Bank in terms of sampling strategy it's based on the two stage sampling strategy in the first stage the enumeration area is the principle sampling unit and once this is selected ten households are randomly selected from each selected enumeration area a brief about the regression decompositions we make as I mentioned the dependent variable is constant throughout all these three or four decompositions we make in this case it's household expenditure per adult equivalent the explanatory variables are physical infrastructure access to roads, access to markets that is input markets, output markets health facilities phone services we also throw in some variables including access to economic activities having a factory within ten kilometers of the center of this particular community or local council one the other variables relate to the household heads characteristics and this we do it because for example education attainment is a key endowment the households age attainment as well as the special location as I mentioned the infrastructure are all at the community so we don't look at how far a household is from a given the most commonly used health facility but how far the community is from the most commonly used facility I should note there are a number of issues as I go to explaining the first decomposition between endowments and rates of return to endowments they are most the presence of infrastructure in a community or else one is highly collinear the communities that have health facilities are more likely to have electricity they are more likely to have beta roads and this as I show in our as I show in some of our results drives some of the results we find so in the same way it's very risky to interpret such individual coefficients as returns because of such collinearity so in the estimations which I'll show here we focus more on explaining what we call like the all health infrastructure all physical infrastructure or all physical infrastructure and health because of such issues of collinearity and the main interpretation for this for example for endowments is the percentage of communities that report for example having a public hospital within facilities endowments are shown by these four columns and the rates of return to endowments are shown by the first four columns if I start with endowments when it comes to infrastructure which is indicated here we find that central Uganda has access to the best infrastructure in Uganda nearly 81 of communities have access to an all season roads compared to much lower rates in the other areas of Uganda and this is more or less consistent I think with the exception of for example health infrastructure for which results we don't show here the other issue we note that again central when it comes to the other so it's mainly in the estimations we're going to compare these items of all infrastructure to the items of like education attainment when it comes to endowments relating to education again central Uganda has better access to these kinds of endowments the household heads in central Uganda have better education attainment and the differences are not as large as the ones we see with physical infrastructure nonetheless when it comes to higher education attainment that is having either having attained four years of secondary education or more then the differences between central Uganda and the rest of the country become very large partly this might reflect the fact that there are more household heads in central Uganda who have higher education attainment might be because of migration because these rates here at least are similar across the four regions when it comes here it appears those highly educated individuals move to central Uganda and partly because this is the region with the most urban areas and as I indicate later it's the region where that has better access to facilities when it comes to access to the returns to endowments again when you look at if having access to physical infrastructure improve the welfare of households in central Uganda by only 15% here compared to 4% in eastern and 11% in northern nonetheless we note that most of the infrastructure effects in central Uganda are driven by access to electricity if we take away electricity which is shown by the estimations here the rates of return in central Uganda are far less compared to the previous case yeah I think I said this so basically the returns to physical infrastructure in central Uganda I mean are highest but regional differences are influenced by electricity overall that table shows that there are no equity efficiency trade-offs for non electricity infrastructure if you provide in such kinds of infrastructure then you are unlikely to receive significant changes in the welfare distribution in Uganda then I move to the Ohaka Blinder decompositions and basically again we try to see how much of the variations in welfare is explained by endowments and household characteristics as well as that that is unexplained this particular chart again I'll start with endowments it shows that and in this case for this decomposition we are comparing the three poorer regions to the richest that is central Uganda what these results show here is that if you were to give the kind of endowments in central Uganda to the people in these other regions it will only improve welfare by between 20 and 22% I mean 20 and 25% half of this is due to household characteristics like education attainment and the other half is due to infrastructure however we should note that a lot of the variations in these places are unexplained finally we try to look at the poverty impacts of social infrastructure and as I mentioned basically we are looking at the returns to social infrastructure investments here we regress household consumption on household variables and the interpretation here can be too first it can be the change in the mean welfare associated with the particular variable or the impact of a particular variable on poverty which we take further by simulating the shift in the original welfare distribution of the regressor multiplied by its mean value and then we recalitulate the poverty headcount for the simulated distribution and compare it to the original headcounts basically the results are shown here and again I will focus on the old education and the old infrastructure we find that improving if we were to compare the two improving access to education would generate far more returns in Uganda and across the four regions compared to improving access to infrastructure we should note that although the infrastructure variable here is very high it's condition on the fact that this region has far greater access of infrastructure in Uganda as I wind up the previous chart was showing that the poverty impacts of existing education is far greater than those of physical infrastructure the differences in eastern region are especially large and for the entire country the distribution attainment contributes to around 15% for introduction in poverty whereas existing infrastructure contributes only 10 in the central region the returns to education and infrastructure are not that very greater than the other regions our results suggest that there is not a too severe trade off between equity and efficiency in public investments returns to endowments are similar in less endowed region than to those in central Uganda especially in the north but also in eastern and western Uganda as such any attempts to distribute social and physical infrastructure more equitably will not be very costly in terms of sacrifice rates of return bringing other regions physical infrastructure to the level of central Uganda will only have a modest impact on poverty as has been demonstrated so the returns to education are much bigger and this is what we propose that this should be the focus of any public interventions specifically investments in education may do better at equalizing poverty rates than investments in infrastructure thank you very much ladies and gentlemen