 Good afternoon everyone. So the paper that I'm presenting is titled Ethnic Fragmentation, Public Good Provision and Inequality in India. And this is co-authored with Dr. Nishan Chadha. So the context of this paper is India. And we look at a remarkable feature of Indian society which is its very high levels of ethnic fragmentation. Just to give you a little bit background about India. So the Indian society has traditionally been a very hierarchical society where its major religious group, Hindus, are divided into a number of caste groups. So these caste groups are strictly endogamous and there are deep social cleavages that lie within these caste groups. The interaction between these caste groups is also very limited. So now there's a large body of work which is established that caste identity in India still has a very important influence on social, economic and political life. So with this background, we consider a caste group as a separate ethnic group and we study the impact of ethnic fragmentation based on caste lines on inequality in Indian districts from 1988 to 2012. But why do we expect ethnic fragmentation to impact inequality at all? So extensive work in the economics and political science literature has now established that ethnic fragmentation leads to a number of negative outcomes. So for example, it has been established that ethnic fragmentation leads to lower provision of public goods, collective action is weaker amongst groups in fragmented areas, there's poor quality institutions, there's loss of trading opportunities and these all negative outcomes we believe can impact the welfare of poor much more than the rich which can lead to inequitable economic outcomes in fragmented areas. But out of all these channels which are expected to drive increase in inequality in fragmented places, we are specifically interested in the channel of public goods and there are two reasons for that. One, we all know now and there's some work to support that provision of public goods is equivalent to providing virtual income to the poor who do not have to pay for these services, education, health and other public goods services in the market. So governments routinely use public goods as an instrument of redistribution. The second reason is that now revisionist work in the fragmentation literature argues that not all types of public goods are impacted by ethnic diversity. It is indeed possible to provide some specific kinds of public goods exogenous to the levels of diversity. A combination of these two reasons I'm going to appeal to to suggest later in the paper that public goods therefore could be used as instruments to mitigate some of the negative impacts of diversity. So I'll return to this point later in the presentation. So in this paper we're also interested in understanding if the increase in inequality in fragmented areas whether that's driven by between caste inequality or within caste inequality. Now if caste nearly identifies, perfectly identifies economic outcomes and there's no impact of individual characteristics on attainment of economic resources, then we would expect the impact to be driven by horizontal inequality. However, if the rich in each social group have better access to economic opportunities and the poor suffer irrespective of their caste identity we would expect the impact to be driven by within caste inequality. Now talking about data sources a little bit. So the population census of India does not release data on individual households except for the 2011 census. So therefore we turn to various rounds of a nationally representative national sample survey to construct estimates of vertical and horizontal inequality by exploiting information on consumption expenditure at the household level. We use following five rounds of NSS to construct inequality and since we are looking at the impact of ethnic fragmentation on inequality we measure ethnic fragmentation using a very commonly used measure which is given by one minus summation beta i squared where beta i is the population proportion of i-th caste group. Now we take data on different caste proportions using the 1931 population census. So we refer to, we look at historical census to construct our caste proportions because of two reasons. One, all the censuses that were conducted after 1931 they have data on only broad social groups. Now typically in India all the disadvantage castes are put together in a constitutional category known as scheduled castes. All the tribal categories, indigenous population categories are put together in a constitutional category called schedule tribes and they're typically placed outside Hindu caste system and then all the upper and middle caste are put in a category called others in this data set. So if we are to make use of these broad categories we would actually be missing a lot of heterogeneity and hierarchy that operates on ground. So therefore this is one of the reasons. The second reason is that there are some endogenity concerns with using contemporaneous caste shares, omitted variables that are difficult to capture for might impact both evolution of inequality in a given area and movement of population groups of different castes. Now since we're relying on the public goods channel we take public goods data from 1991, 2001 and 2011 census and we primarily focus on primary schools and health centers. The reason is, sorry, so the reason is that investment in education and health can be made both by the state and private households which suggests that there is a richer relationship between provision of these public goods and inequality as opposed to a public goods like roads where only state can invest. Now since our interest is to see if fragmentation drives increase in inequality it would be informative to also understand trends of overall and horizontal inequality in India. So here we look at the trend of overall inequality and what we observe is that within the sample period except for the initial fall, fall in the initial year inequality has been gradually increasing with a very sharp increase here. This was probably because this round of NSS was conducted after two successive years of droughts which probably increased inequality sharply. We observe that this trend is also followed by horizontal inequality which has been increasing as well and however, horizontal inequality is a very small proportion of overall inequality and this result is consistent with what other people have also found out about India. Now what are we empirically doing in this paper? We are basically regressing inequality both vertical and horizontal in district D on fractionalization index and we control for state specific fixed effects to basically minimize concerns that statewide differences in institutions are driving our results. We also have state specific time trends to take care of omitted variables that vary within a state linearly over time. Now we also have some district level time varying control variables like log of monthly per capita expenditure and literacy rate, urbanization rate. What do we expect? We expect the impact of fragmentation on inequality to be positive driven by a range of negative economic outcomes that I just mentioned. Since we're relying on and we're focusing on the channel of public goods, if we add public goods variables in our regression equation, we expect some of the impact of fragmentation to be driven by public goods. Therefore, the coefficient of fragmentation should ideally fall in the regression equation where we have public goods, right? So that's what we do in our regression equation two. We control for public goods and we expect the impact, the gamma one here to be less than the gamma one of equation one. Let's see what we find. So going to up tables here. So in the first six columns, we have genie as the dependent variable and then in the columns seven to 12, we have tile index as the dependent variable. I'm going to be talking about just the genie here. The results are similar for tile. So observe column one. What we see is that if we were to move from a completely homogenous district to a completely fragmented district, we would observe an increase in genie coefficient of about 0.10. So this amounts to about 70% increase in inequality as compared to the average inequality. Now this result remains robust to addition of different control variables. Now in column three and four, since our interest is to see whether public goods is driving some of the increase in inequality, we control for primary schools and health centers in columns three and four. And what we see is that about 30% increase in inequality was indeed driven by lack of provision of education and health facilities. So apart from, so in columns three and four, we linearly add public goods. So what we do in columns five and six is we interact public goods with fractionalization index. And again, we find that on average, fragmentation increases inequality. However, this increases much less if the district has public goods available. Now this suggests that if somehow public goods were to be provided in a given area that would help decrease the impact or fragmentation. Now, but current work suggests that fragmentation in fact determines the level of public goods. However, revisionist literature helps us here by suggesting that it is possible to provide some specific types of public goods independent to the levels of fragmentation. So we feel that additional research needs to be done to come up with ways to exactly provide public goods exogenous to the levels of fragmentation so that some impact of fragmentation could be mitigated. Now, we also want to understand if this impact on inequality is driven by within caste inequality or between caste inequality. What we find is that this impact is entirely driven by within caste inequality. What this might suggest is that the rich in even the disadvantaged groups have been able to overcome the disadvantage of their identity and the poor suffer irrespective of their identity. So this could be because in India, there are a range of affirmative action policies, differential treatment to disadvantaged groups, which some argue that has resulted in a class of elites even amongst the disadvantaged pointing to within caste inequality, an important dynamic of caste based inequalities that one needs to pay attention to in Indian context. Now, I'll skip this part and I'll conclude because I don't have much time left. So what our paper suggests is that inequality seems to be higher in fragmented districts, but this increase in inequality could be mitigated to some extent if public goods are provided or if we come up with ways to exogenously provide public goods in these areas. So this is a very important statement suggesting that an economic policy, a redistributive policy has the potential to mitigate negative impact of a social demographic phenomena. The second finding, important finding of the paper is that there is no impact of fragmentation on horizontal inequality. Now there's a very important caveat to this interpretation because the horizontal inequality that we have is based on broad social groups that I just mentioned. Because of these broad social groups, we might not be actually capturing the true extent of horizontal inequality. So therefore, we feel that additional research needs to be done to actually identify the groups that we have to consider in mind to capture horizontal inequality. However, at the current time, at this time, no data that is representative at the district level captures any details more than what this data set does. Yeah, thank you.