 I am talking about human capital investments and the impacts of that, though. So this is where I hope this paper will make the contribution to this panel. So there's been a long-running debate with differences in developmental outcomes at the subnational level, like explained by variations in human capital over institutions. So recent work around the world at the subnational level shows, for example, that usually incomes are very highly correlated with human capital, work by Giannioli, it all shows that, and also by Cimuglu and Dell. But then this work also usually has very large residuals that are not explained. And so, for example, Cimuglu and Dell argue that this could be due to the quality of local institutions. Obviously, this is a point that somebody like that and Cimuglu would make. They don't really have support for that, empirical support for that, but they show this in a theoretical model. Now these studies have a problem that they cannot really deal with and the genetic of institutions and human capital at the subnational level. These countries control for country contacts or country fix effects, but they cannot really control for the fact that institutions and education and also incomes may be endogenous at the subnational level. So this is where I try to make a contribution by going back to the origins. This is why I use the historical approach. I try to trace current variations in incomes by looking at historic variation in missionary schooling in Madagascar and in colonial institutions using a similar indicator as the one used by Archimoglu, Johnson Robinson, and their famous paper. So I will look at the long-term consequences of settler institutions. And I hope that this subnational focus also helps to resolve some of the deadlocks that you see in this literature on colonial institutions and colonial human capital. There has been a large literature on that, mostly represented by Archimoglu, Johnson Robinson, and Glaser et al. This literature is usually at the cross-national level, so there you really have huge endogeneity issues and also problems with unobserved country contacts. So by going subnational, I hope to resolve this deadlock as well. So the main results, just working through them very quickly, I find very robust long-term impacts of colonial settlement institutions. So this is really support for the institutionalist argument to my own surprise I find very weak impacts of past human capital investments. I suggest that two stories are driving these results. One is kind of a reversal of fortune story as also developed by AJR in the 2002 paper whereby initially disadvantaged areas as measured by initial population densities then developed after the colonial period because they received superior property rights institutions. I explain the weak regional impact of human capital investments of missionary school investments by spillovers between districts. Basically when you invest in schooling you always have the possibility that people who are educated then move on to districts that are more developed. And this I think explains some of the results that I find at the spatial level. Okay, so what is my identification strategy? First I try to exploit historical variation between missionaries and settlers in time. So in Madagascar missionaries arrived in 1820. That's when they started to build the first schools. And then they had school networks in place when the French colonizers arrived. By 1896 this is when French colonial rule started. The school networks of most missionary societies in Madagascar was already well in place. I present some numbers in the paper. I'm not going into this now. Just to show that this is not an exceptional situation. David Livingston, a famous missionary, of course, he traveled Africa from 1850 to 1873. So well before the scramble of Africa when colonial rule was established more or less across the entire territory of Africa. I also explored variation in space. So these are the maps of the distribution of missionaries and of settlers. Missionaries tended to settle or tended to build the missionary stations in their schools in the central highlands because they suffered very high rates of mortality. So they preferred the more temperate highland regions. So these are the darker regions here. Whereas settlers were mostly interested in cash crop production. So they went into the coastal areas where they could produce spices, vanilla, rice, sugar cane, these kind of things. So I exploit this special variation at the subnational level and the correlation between these two variables is almost zero. Now you're going to argue that, okay, if they are in such different regions then everything is probably driven by geographic effects. Then I'm just measuring these geographic effects. So I also try to deal with this by using some instruments. This is my estimation strategy. The dependent variable is the log of mean district household consumption from an official poverty assessment. I have that for all of the 110 districts of Madagascar. And then I basically run a horse race where I try to predict these household expenditures at the district level using my information on missionaries and on settlers. The IV for missionaries are dummies for the stages of expansion of the pre-colonial marina empire. I won't go into that instrument because it's really just a robustness check. I have to admit it's not a very strong instrument. But this is what I use for identification here. The IV for settlers are somewhat stronger and relates to the literature. The population densities in 1936, so before settlement really took off. This is also the instrument that has been used by AJR in the reversal of Hodgens paper. They use it at the cross-national level. I use it here at the sub-national level. So it's a replication of the identification strategy if you want. So what's measured by my variables? Missionaries. That's the number of churches per thousand inhabitants per district in 1904. Now historically churches were linked to missionary schools. Often the missionary schools actually doubled as churches on Sundays. So they built a school to attract followers basically. And then on Sundays they had services in these schools. In my paper I also combine information on Protestants and Catholics mostly because these two groups were usually in the same regions. There was a lot of competition as in many other parts of Africa between Protestants and Catholics. Usually the Catholics who came later went to areas where the Protestants were. This makes it very difficult to identify separate effects statistically for these two groups. So that's why I combine them. I tried separate regressions which did really produce different results. Settlers, that's the population proportion of French non-military personnel. Now Madagascar followed the extractive model of colonial exploitation but they also encouraged settlement. They settled through small-scale settlers mostly from Réunion which was a nearby French colony and is still a French département d'outremer. And this settlement variable mostly measures quality and local property rights. That's my argument about this variable. Because where the French went to settle or to build agricultural businesses they introduced formalized land rights, formalized land titles. And they left informal, traditional, customary land titling systems in place in areas where they didn't settle. So this introduces a lot of variation in the quality of local property rights institutions. I should also add that this is contextual historical evidence which I find supported in two studies that I read that the settlers who came to Madagascar were usually not very well trained. So that's been this argument that the settlers maybe didn't bring institutions but also their own human capital which then makes it difficult to identify the effects of institutions in human capital. This doesn't seem to be the case in Madagascar. These settlers were not very well trained. This is just a historical quote to illustrate that. Now this is a summary of my results. Again I find no effect of the missionary schools. This was to my own surprise. And I find quite robust effects of colonial settlement. Now you're going to argue that missionaries arrived in Madagascar much, much earlier and settlers arrived. This measures settlers in 1950. So there's a very large time difference here. I also tried to predict district wage levels in 1938 using my missionary verbal. Same result. So this doesn't seem to be explained by the time difference. These are just the regression tables. So this is the result for my missionary verbal just to show you again that also when I use geographic and historical controls I find really no significant impact of missionary schools on current district incomes. This is the effect for the settler variable, so for the colonial institutions effect. Here I find quite robust effects as you would expect from the scatterplot. This is just a simple barbed-rate correlation and this is where I add controls for geography and some historical controls. I think this step here is particularly interesting. Here I also add historical wage levels. So wage levels at the time when this verbal was recorded, my settler verbal was recorded. So this is where I try to pass out the institutional impact of settlement from the economic impact. So this is where I try to pass out basically the effect of persistence in incomes since the colonial period. So there's still a fairly robust effect here. And then these are my IV estimates. So the result also holds up in the IV estimates using historical population densities as an instrument. Now what explains these very different results? This is just where I'm trying to build some theory about the processes. I think the settler effect is mostly explained by this reversal of fortunes story that could be explained by this reversal of fortunes story that AJR introduced. Here are my first stage results for settlers just to illustrate again that settlers went to areas that were historically disadvantaged. They had these areas had low population densities so they didn't support very large populations, these areas. This is manifested here in the negative first stage result. What I also find is that former settlement regions today still have stronger land titling systems. The share of agricultural plots that are titled in settlement areas is higher than in other regions. So this suggests that really there's some sort of institutional persistence going on here that might explain this. So persistence in the quality of local property rights. I also find that the past presence of settlers does not predict the cultivation of cash crops today. You could argue that the settlers went to some regions to cultivate cash crops. That's why they were mostly in tropical coastal regions. And this could just explain persistence and incomes. It doesn't seem to be the case. So today these areas are not much more likely to produce cash crops. I also find that in these areas there's more manufacturing activity going on. I control for population sizes and for urban status. It's not just that these are urban areas but also in these rural areas there seem to be more manufacturing activity. So it's really a story that is consistent with this standard idea that better property rights encourage economic activity and this could then explain the better economic outcomes. Now this is also when trying to come up with an explanation why the effect of missionaries is so weak. And I argue that this could be due to human capital spillovers. The only result or the only outcome where I find that missionaries has a strong robust effect is in school supply. School supply mostly in the private sector. Many of these schools are operated by churches. So there's a pretty robust correlation here. And also secondly schools. It's less strong in the public education sector because of massive school investments after the colonial period. But still we see that overall former missionary areas still have better school supply today. But we also see that they don't really have higher adult education rates. These adult education rates are higher in former settlement regions. So this suggests that there is probably some human capital spillover going on that people who are trained in these regions then migrate to the economically more dynamic former settlement regions. This could of course contribute and perhaps explain the lower economic impact of missionary schools that I find at the geographic level today. So this already brings me to my conclusions. Am I going on time? Okay. So just to summarize again, this was a test using historical data of the relative impacts of institutions and of human capital investments and I find strong support for the institutional argument building on this reversal of fortunes theory and weak or no support for the human capital argument. However, in this study as in many other studies that look at impacts of colonial interventions perhaps there's a specification error here because I'm looking at regional impacts and when you look at human capital impacts you also have to take into account these billovers that are just documented. So this is perhaps also a lesson for the wider debate about the consequences of interventions in the colonial period because these studies usually only look at regional impacts. Obviously this also has implications for policy. I mean when you build schools to develop a region you have to take into account that perhaps people are going to migrate away from that region so that you're not going to see economic impacts in that region but that this will spill over into other areas. Okay. Thank you very much.