 Hello, everyone. I'm Shushmita Baulia. I'm a postdoctoral researcher in University of Turku in Finland. And this is a joint work with Kunal Sen of UNIWIDER, who's probably presenting in a parallel session at the moment. And Mila Nussela is also in WIDER and Labore, another research institute in Finland. So, let's start. This is about income diversification in the long term. And we are looking at intergenerational outcomes in the Kagera region in Tanzania. So, livelihood diversification, it's a common phenomenon in rural agricultural settings in developing countries. A lot of factors that have been identified for this is they can be used as risk reduction strategies, responses to household shocks, or as an asset accumulation strategy. And especially in the Kagera region in the academic literature, already it has been found out that geographical mobility is one of the important drivers of how individuals choose to move out of poverty by choosing diversification strategies closer to urban areas. And in the literature, diversification components have been classified in different ways. Sometimes it is classified across sectors, for example, that is agricultural versus non-agricultural, sometimes by location on and off farm, and sometimes by function, self-employment versus employment by some other entity. Our point of interest is income diversification. We are going to cut across almost all these, and we are looking at the major income sources that the households choose to diversify in, and that could be within agricultural, outside agriculture, and it could be employment by someone else or self-employment. Before moving on, some stylized facts about diversification in Africa. Diversification outside agriculture in Africa has been less varied than in other continents. The reason is lack of industrialization and structural transformation came to Africa quite slowly. And then also in Africa, evidence supports that in Africa, we have diversification out of accumulation or born out of opportunities and access than in case of Asia, where it is born out of survival strategies out of desperation. But with all these, now the question is can these diversification strategies be associated with long-run welfare? And being within the theme of the conference, we also try to say something whether it reduces inequality in the long term or not. So there's only short-run evidence so far on positive association between income diversification and welfare. The closest literature on medium to long-run diversification which we find is in Ethiopia, Burkina Faso, and in Tanzania where the time period of study ranges from five to 10 years. The main takeaway from all these studies and all of them agree upon is that relatively wealthier small-holder farmers, they have the sufficient assets to make, to successfully diversify to non-agricultural sectors. And the poorest households are often facing entry barriers. And there is no empirical examination as of yet on the intergenerational outcomes. So that's what we are going to look at. So what we do, we are studying the long-run intergenerational implications of household income diversification in the 1990s. And we are looking at the outcomes of the young adults in 2010. And we examine outcomes such as consumption, their educational attainment, professional diversification in their current households. And we are using propensity score analysis. And we are using both a dichotomous setting for diversification. And we are also looking at continuous diversification settings to understand the heterogeneity effect. And we are using the two-decade long data, longitudinal data, from the Kagera region in Tanzania. We also explore patterns of income diversification and how the income groups have evolved in their income among the dynasty households from early 90s in comparison to mid-2000s. And the last line, which is in shadow, I'm not going to talk about it today in the interest of time. But we also explore diversification within agriculture. We also explore diversification within crops in the farming households. And we see if those are enough or do we really need a firm diversification. So preview of results, we do confirm that income diversification can be used as a successful tool to increase consumption of the future generation. And it does put the future generation on a sustainable livelihood path and with higher human capital. And again, I'm not going to go there today, but here's the result. We show that off-farm diversification is crucial for this as agricultural diversification alone is not enough to bring this development. And we also document that both the push and pull mechanisms, that is the accumulation and the survival strategies, are at play in diversification. And we are also able to document that the poorest households are pushed to the diversification trajectory. And after a decade, they have a higher relative income position in the income ladder. And this is the Kagera region for you. So it's at a political and economic crossroads and adjacent to Uganda, Rwanda, and Burundi. And the population size in the early 90s was 1.55 million. And it is a predominantly agricultural society. And also it was hit hard by the HIV AIDS epidemic. So it's important to see how their diversification strategies have helped them over the generation to survive. And then data and summary statistics. So we are using the Kagera Health and Development Survey. So the baseline panel have four waves. They were bi-annually collected between 1991 to 1994. And we are interested in the income diversification information from these households, which are also the main households of the panel. And then the first follow-up comes in 2004, where we could track the individuals or the split of households. And we are going to look at how their income distribution has changed in this time point compared to this. And the last follow-up that is Wave 6 is in 2010. And here we can track the young adults who were children here. And then we look at their consumption and other welfare outcomes when they're adults. Yeah, so it's a pretty neat sample given the long-term it was tracked. It has very low attrition. So we are starting with a little over 900 baseline households. And they were interviewed more than 90 percent were interviewed four times in the baseline panel from 1991 to 94. And we have 93 to 92 percent re-interview rate in the follow-up panels. And then we have the split of households or at least one individual from the household they were tracked during the follow-ups from the main households. So we consider them from the start of the baseline to the end of follow-up. We consider them under the umbrella of a domestic household or a dynasty. Right. And it's a pretty rich data set. And we can look into various socioeconomic outcomes from there. And here is the... So here we have the six income sources that we use to build our diversification information. So we are looking at self-employment in agriculture, employment in both agriculture and non-agricultural sectors where you are employed by another entity. And then you could have self-employment in non-agricultural sector, some business, and then rent transfer and remittances and finally non-labor income, which could be inheritance or pensions or savings interests and such. And from here we can see that 31 percent of these households in the 90s had one income source and about 91 percent were self-employed in agriculture. Most of the households had two income sources and the primary being self-employment in agriculture and the secondary being business or employment by someone else. And now here we are looking at... So we have the income percentiles according to the income distribution of the baseline. And here we are looking at the shares of different income sources that the households in these income brackets have. So across the board we can see that this blue line, which is self-employment in agriculture, that is the primary source of income for most of the households in all income brackets and the prevalence of other shares, it kind of increases in the higher income brackets. And this is the diversification index that we build. So we have the six income sources and we use a transformed hair-fingled Simpson index of household income diversification. It looks like this creature here and here SI is basically the share of income source in the household's income basket and it is zero when there is only one income source and then it is one minus one by N when all the N income sources contribute equally to the total household income. So it considers all the shares and also the evenness of the shares. All right, so now we would like to look at who actually diversified in the different income brackets and how much and where are they after a decade, that is in the first follow-up of 2004. I try to go through it point by point. It may be difficult to explain. So here what we are doing is we are looking at the evolution of the mean income and diversification of these different income groups, which are according to the baseline income distribution. And we would like to see what is their mean income per capita in the wave five and we also look at their corresponding diversification figures. So let's consider this bottom most income percentile of the baseline wave. So their real income per capita is somewhere here and then we try to see where the mean income is according to the wave five distribution. So these squares are basically the upper cutoffs of the different income brackets of wave five. So what we see is that the bottom most percentile their mean income in the baseline wave was somewhere here and now they have gone up. So what is this yellow dot? It is basically the upper income cutoff of the 26 to 58th percentile of wave five. So we see that the relative position has somewhat gone up. Similarly, for the second income group, they were somewhere here and then we see that they also go up somewhere here and this is below this green square, which is the cutoff of the 51 to 75th percentile of wave five. So they have also their mean income has also gone up the ladder and these guys are also following a similar pattern. But what about the rest of them as we go up the income brackets? So these guys, they were somewhere here and then we see that the mean income increases a bit, but they are still under the green square, which is still the same percentile. So their relative position hasn't changed much in wave five. And also the next income bracket is in the same income group as they were in the baseline. And the top most income bracket, they somehow go down a bit. This income bracket, I wouldn't trust much because they have quite a variation in the income and it could be a bit volatile to trust. And then what about their diversification? So we can see that everybody is diversifying in all income brackets. So probably the poorer ones are diversifying out of the push strategies and the upper income brackets are diversifying out of accumulation or pool strategies. But anyway, everybody is diversifying. We see that it goes down a little in the wave five. There could be one reason for this because wave five, it was collected over one season and it doesn't cover all the season. So we could be having some seasonal effect here and not looking at the entire picture. But anyway, the main takeaway from this picture is that income shuffling upwards was more pronounced in the lower income brackets and we see diversifying happening across all the income groups. So both push and pull mechanisms are at play. But yeah, in the wave five, everybody is diversifying a little less. Now the question is that, okay, we see that these poorer groups, their mean income has increased and they have gone up the ladder. So who are the poorer guys in wave five? So this is a bit of speculation exercise. But still, so what I'm doing here is that we are trying to track the mean income and diversification of the income groups from wave five to the baseline. So now we have the bottom most percentile of wave five according to wave five's distribution. And when we look that, we see that they were somewhere high up in the baseline wave. So when we look at their corresponding diversification, we see a lot of quite a difference. It was, it is not as, as less, I mean the difference in the earlier graph was much less. So one, we could understand, try to understand it this way that this may be a sample of people who have this who have this random shock in the wave five. And we see their diversification is also quite low and they end up having a really low mean income during this period. This is a bit of speculation. We need to probe this further. But yeah, this could be understood this way. And similarly, we see this in the second income group as well. So they were also somewhere high up. But this is this small group of people somehow ended up in the, in the bottom percentile of fifth wave and their diversification is also quite off in the two, two waves. Okay. So half of my slides are still left. So quickly going through now the intergenerational outcome analysis. So now we are interested in looking at a cohort of surviving descendants whom we, who, who could be traced in 2010. They are young adults now, and they were children in the baseline. So they were less than 15 years old in the original household in the baseline. And we have, sorry, we have 1313 individuals like this belonging to 549 dynasty households. So little descriptive stats about them. We have 50% of the sample as male and the households they belong to in the 1990s had a real income of mean, real annual income per adult equivalent of 545 USD. The median income percentile was 26 to 50th in, in the nineties in the household that they belong to and so forth. And now we are looking, we are defining our diversification variable in two ways. First we use the simple dichotomous dummy variable. It is equal to one if the main household of the dynasty had more than one income source out of those six income sources in the baseline. And then we are also using the Herfindl Simpson index that I showed you earlier. And we are looking at, yes, we are using the basic benchmark of OLS, and we are using propensity score analysis to avoid the selection bias with observational data and such. And we pick one of the rigorous methods of propensity score. And then in the continuous case, we are using a dose response function of generalized propensity score and the results. So for the dichotomous case where it is a dummy variable for diversification, we see that we look at outcomes such as food consumption in the current household of the young adult and the non-food consumption in the current household, their educational attainment, the educational status of their spouse, the migration patterns, and also professional diversification. We have the number of different professions in the current household of the individual that we are tracking. So we see, for example, with the OLS, we see that food consumption due to diversification in the baseline household. In the current household, the young adult, they have food consumption increased by 9%. And then non-food consumption is even more 21%. Education in years, it increases by a few months. And also, probability to migrate for school is also more in the diversifying households. And with the propensity score analysis, also we see the results don't change meaningfully. They are similar, more or less. And so this is just the summary that we see more or less positive outcomes in all the outcome variables we look at. Just to mention that we see a hint of associative matching in marriage because the spouse's education is kind of correlated with the education level of the individual as well. And then in the continuous setting as well, we see positive and statistically significant outcomes in almost all the variables. And this is the final, these dose response functions. So on the left, for each outcome variable, on the left, we have, shall I just stop? So I mean, just to wrap up, it's that we find that the dose model captures heterogeneity. And it confirms that with increasing amount of diversification, the outcomes like the estimates were higher for these outcomes. And just to wrap up, so the effects are within codes because we don't claim anything causal here. We are mostly saying we find association in positive welfare outcomes and income diversification. And yes, and we also can say that for this on and off farm diversification, both are important. And finally, we could also, we see some association that both push and pull motivations are a plane income diversification and the poorest households are going up in the income ladder more. Thank you.