|
Philip Verwimp:
[Welocme to] this second session this morning on Poverty, Conflict, and Agriculture. Actually I have two hats on in this session. The first is I am the president of the Dutch/Flemish Chapter of Economists for Peace and Security. In that capacity I would like to thank Jamie and Thea for inviting me to this conference and for having a panel that we contribute to this conference. I think it’s very important that different chapters keep having close relationship with each other and in strengthening the ties, and I’m very grateful for that.
Secondly, I’m also the co-founder of the Households in Conflict Network, which is a group of scientists doing micro-level research on causes and consequences of conflict. We are interested in how households and individuals and rebel groups during, before, and after conflict, what are the health, the poverty consequences of conflict, what are the contributing factors; and we try to analyze this in an empirical way with micro-data from large-scale samples or from qualitative samples. We go often in the field. Most of our members are development economists who often go in the field. And this session we’d like to present three of our research results in as many papers.
The first presentation will be by Tom Bundervoet, who is a Ph.D. student at University of Brussels. And the second presentation will be Eleonora Nillesen, who is a Ph.D. student at the German Institute for Economic Research in Berlin. And I will take care of the third presentation. Thank you.
“Livestock, Crop Choice, and Civil War: Evidence from Burundi”
Tom Bundervoet:
Thank you, Philip. So I will talk about savings and activity choices during civil war, with an empirical reference to Burundi. As Philip said, it’s a micro-level study of the consequences of war at the household level. So the paper’s concerned with the effects of war on the welfare of households effected by the war.
Obviously there are many different channels to reach civil war, and the insecurity, the risk related to civil war reduces household welfare. I will focus on one possible channel, and that is the impact of war on activity choices, so the choices households make to generate their income basically.
The hypothesis is that war and the heightened insecurity or risk related to war drives households into the low-risk activities, low-risk subsistence strategies, like growing of low-risk food crops, abandoning the markets, the market-oriented activities, because these are more risky than the auto-subsistence strategies, and so on.
However, to test this in a direct fashion we would need panel data. We would need a set of households before the war, and then re-interview those same set of households some years into the war to see how their activities have changed, all other things being equal. However, this panel data we are going to collect between July and December this year, so we don’t have it yet. So we have a cross-section, a cross-section from Burundian households of 1998. So I will work with a cross-section.
Since there is only a cross-section, I have used an indirect approach. I have tested the standard economic risk arguments, the standard economic risk associations between savings of households and their activity choices. This risk argument is that households with more savings, the wealthier households, will tend to engage in higher-profit, higher-risk activities, since if these activities turn out wrong, they can use savings to smooth consumption basically; while the poor households with less assets, with less savings, have to engage in the low-risk activities since they do not consist of enough, they don’t dispose of enough [sic] means to smooth consumption if things go wrong in the absence of credit markets, as is the case in [total?] Africa. These standard economic risk associations we tested for many developing countries, and generally, is validated by the data, meaning the higher the savings of the household, the lower their allocation to save activities, the higher the allocation to risky, more risky, more profitable activities.
However, the previous studies assume that assets, a household’s assets, are certain, so completely risk-free, completely sure. However, during war, assets of households, which in rural settings is household livestock, are very very risky, because it’s easily pillaged, it’s easily stolen, both by government forces and by rebels. I’ll present evidence on this later. So if we adapt a micro-economic theoretical model of the association between savings and activity choices to assume risk in the assets, then the theoretical prediction is of households, even the richer households, will not engage in the profitable activities because assets, their livestock, can easily be stolen during a conflict. So everyone, more or less regardless of savings, would engage in the safe, low-risk, low-paying activities. So welfare implications obviously would be, theoretically, a decrease in household income, since all households engage in low-profit, low-risk activities, and an increase in poverty levels. That would be the prediction of a theoretical model.
So the empirical setting is Burundi. There’s been a civil war between 1993 and 2006, a very bloody civil war with over 300,000 deaths. The macro-economic impact of the war was very negative. The income levels per capita fell from $210 before the crisis, which already is very low, to about $100 in the year 2000, making it the poorest country in the world at that time. And the people below a nationally defined poverty line rose to about 70 percent, which is huge.
Now I have to change some slides.
The assets, the livestock holdings during civil war, were very very risky in Burundi. These are aggregate figures of the evolution of live animals, so the savings between 1990 and 1998. For all types except for poultry, we can see very very large changes, negative changes: minus 20 percent for cattle, minus 28 percent for goats, minus 45 for sheep, and so on. According to the U.N., this is mainly due to deaths and pillaging during the crisis. So the assets in Burundi were very risky during the civil war, so there is a motive for testing whether the standard economic association still holds.
I’m using a 1998 cross-section, a representative household survey carried out by the World Bank with data on 2,908 rural households in Burundi. Then to identify the effects of the civil war, we explored the fact that not all regions were equally exposed, equally effected by the war. In particular, there were three provinces in the country which have been very much effected by the war. (I think I have to make this smaller. Okay.)
Those three northwestern provinces of [Sibitoke, Ubanza, and …?] surrounding the capital have been very effected by the war, at least between 1993, the beginning of the war, and 1998, the survey. Of course, you don’t have to take my word for that. There is a lot of evidence: reports by Human Rights Watch, “Provinces of Persistent Insecurity,” the three war provinces. These provinces were the bases of the two main rebel groups, the FNL and the FDD rebel groups. They’re also called zones of persistent combat, because they were always fighting in these provinces. There’s also some quantitative evidence from a demographic survey by the United Nations Population Fund, where each of the respondents was asked whether or not his parents were killed during the war. These are the results. So within the three war provinces 12.4 percent of all respondents had at least one parent killed in the period 1994-1998. In the rest of the country this was only 3.9. Then in the years immediately preceding the survey, this was 6 percent in the war provinces, and 0.8 percent for the rest of the rural provinces. And this shows that the choice of these provinces as war provinces, war regions, is a very plausible one.
Then I already showed that assets were risky during the period in Burundi, and especially in the war regions. This gives the evolution of livestock savings expressed in tropical livestock units. Tropical livestock units is a standardized measure of livestock. Before the war in 1993, average livestock holdings per household were 3.93 in the war regions, and 2.14 in other regions; so actually the war regions were the richest ones before the war. Five years into the war, average per household in the war regions was 0.68; in other regions 0.62. The change, the difference, is more than double in the war regions than in the other regions, which again shows that in these regions, savings were particularly risky, because they were stolen basically every day.
So savings, household savings, are proxied by the value of household livestock holdings in 1998. Then to proxy low-risk, low-return activities in most of Africa, and especially in Burundi, unskilled agricultural wage employment, so a farmer which has small land holdings or even no land holdings at all who goes working in the fields of another farmer. It’s a very low-paying job, low-risk obviously. Also the cultivation of low-risk food crops, such as cassava or sweet potatoes, is also a very safe, very low-return activity.
To proxy the labor allocations, or the allocation of household labor to the safe activities, we used the income share of these activities, so the share in total household income of the safe activities. It’s of course a quantitative study. Then the specification—So the allocation, the income share of the safe activity is the PEIJ, which are the coefficients of savings and the coefficients of savings [interacted?] to the civil war region. So actually the beta-1 is the impact of savings on safe activities; so we expect to this to be negative. The beta-2 is the impact of risky assets--so assets in civil war regions are risky assets--on safe activities. According to the model, this should be positive and significant.
So the results--I’ll very quickly show the results—are presented here. The first analysis is the safe activity of unskilled or farm agricultural work. So as you can see, the savings expressed at per capita, have a negative effect on allocation of safe activities-- that’s a standard result; we expected that—but in a civil war the coefficients turn positive and significant. This means that the relation between savings and activity choices differs significantly between the war provinces and the non-war provinces. Basically, this says that within the war regions, assets or savings do not play any role at all in determining safety choices. Why? Because assets are very risky, so obviously everyone, even the wealthier ones, would on average allocate more to safe, low-paying activities with, of course, negative welfare [implications?] in the longer run.
The second analysis is the same one but with additional controls. Same results, as you can see.
The third one pools all safe activities together. So we have one measure, food crops and [off-] farm work for safe activities, and again we see that the results are significant and do support the hypothesis that when in war, during wartime, when savings are risky, households allocate more to safe activities relative to the [?] than if assets were not risky, as is often assumed in wartime.
These are some checks for the robustness of the results, with the allocation of cassava, a low-risk, low-return food crop, and allocation to maize, a higher-risk, a high-return crop. We see that households with higher savings allocate less to cassava, as expected, and more to maize, also as expected; but within the civil war regions, this becomes positive for the low-risk, low-paying alternative, and negative for the high-risk, high-paying alternative, which supports the results of the previous ones actually.
[So the overall conclusion is that households in the war regions do not reduce allocation to a low-risk, low-paying activity when their savings, when their asset levels increase?], a precautionary motive because their savings are very risky. They can be stolen, pillaged at any moment, or killed.
The consequences [are] massive increase in poverty levels in the three war provinces over the period 1993-1998. They were the richest provinces before the war, with a poverty head count of only—in Africa—only 22.4 percent in 1993, and 52.5 percent in 1998. So more than 30 percent increase, which is obviously very very large. And this can probably at least partly be explained by retreat into subsistence, low-risk, low-paying activities during conflict, during wartime. So thank you very much.
PV:
--on the poverty trap in a conflict situation—households having to return to low-yielding, un-risky activities which bring them very low income.
I would like to now give the floor to Eleonara Nillesen for her presentation on “War, Peace, and Coffee: Farm Level Evidence from Rwanda.”
“War, Peace, and Coffee: Farm Level Evidence from Rwanda.”
Eleonara Nillesen:
Thank you, Philip. I’m going to present some of the preliminary work that I did, together with the Philip, on the coffee sector in Rwanda, taking a micro-economic perspective. As I said, this is a joint work with Philip, affiliated with University of Brussels, and, as he said before, he’s a co-founder of the Households in Conflict Network. I myself am a first-year Ph.D. student at the German Institute of Economic Research in Berlin.
So the objective of our research was to provide a micro-economic assessment of what happened in Rwanda after the 1994 genocide at the household level, and then with a specific reference to the coffee sector. The rationale is that what we observed was that there was a sharp decline in coffee cultivation after the genocide, and we kind of expected that because a lot of men were actually killed during the genocide. Men were the main force who actually did cash crops, coffee. A lot of men were killed, so we would expect a sharp decline in coffee cultivation after the war. So that was the rationale to do this study. What we used were we used two separate national representative surveys. One was collected in 1992, and the other one was collected in 2001.
Let me first give you a brief introduction to the policy of coffee in Rwanda before the genocide. Under the regime of Habir Rumana, they used the so-called carrot-and-stick approach. On the one hand it was very motivating for farmers to actually cultivate coffee, because fertilizer was made available, new plants were very heavily subsidized. But on the other hand, once farmers actually started to grow coffee, it was nearly impossible for them to stop, because they were closely monitored to not uproot the coffee trees or to abandon their coffee trees. That was the stick approach.
Coffee exports during this regime contributed to about 60 to 80 percent of the state revenues, with this range depending on the annual output and the market prices. Basically what the state did was they bought the coffee for a fixed price from the farmers, so the farmers were always sure that they would get this price. And then if the international price was high enough, the state would actually make some profit, because they already bought the coffee from the farmers at the fixed price and could sell it again at a much higher price. In that sense, the state budget could increase severely. Initially the additional revenues that the state would get would go into this fund, but that didn’t quite work out, so the coffee sector was very closely linked to core state functions, because the fund didn’t work properly. That was about the policy.
Let’s have a look at how farmers worked, the livelihood strategies: By the end of the ‘80s, nearly all Rwandan farmers actually lived in rural areas, and most of them were farmers. Main food staples included beans, bananas, sorghum, sweet potatoes, and cassava. Cash crops were coffee in competition with bananas. And, as I said before, man’s labor was mostly diverted to cash crop production; women’s labor to food crop production. This was a relatively successful system to avoid famine most of the time, combining food crop production, cash crop production, in combination with small livestock production. Some also had additional income from off-farm activities and beer brewing.
That was all before the genocide. Now we wanted to link this up with what happened during the genocide. First we take a regional perspective. What we used was we had province-level data on coffee farms in 1992 and 2002. Then we used the conflict shock index that was developed by Philip and a colleague of Philip’s, Patricia Justina, to assess the impact of conflict shock on coffee. We listed four shocks that they had identified: shocks of civil war, the genocide, mass migration, and counter-insurgency. And what’s important here is that this really only province-level data we have on these shocks, so we couldn’t really distinguish here what the effect of a conflict shock would be at the individual household level, because we didn’t have that data. But what we did do was use this province-level data, link it with province-level data on coffee farms. What we found was that there is indeed a positive, statistically significant correlation between conflict shocks and [?] of coffee cultivation.
That was the regional perspective. Now we go really to the micro-level, and we use the household survey data, first of all in 1992—that’s the first column of figures—and then the 2002 data set, and in the third column we showed the differences. I’m just going to highlight a couple if interesting results, because we’re only going to look at the third column. I think we’re only interested in the differences here.
What we see is that, for example, we see a very high increase in the number of female heads of household; it increased by about a quarter, so that’s really big. This is also important: This is only for the non-coffee farmers. What we also see for the non-coffee farmers is that real income for [?] equivalence actually increased by nearly 12 percent; but, at the same time, we see that there’s a small increase in people that were getting poorer. There’s even a larger increase in people that were getting extremely poor. So despite the fact that the real income actually increased, there were lots of people getting poorer. This is an indication that inequality actually increased.
Looking at the coffee farmers, we see a similar pattern, but the magnitude is slightly different, because here we see that female heads of household increased by nearly 47 percent, which is three times as much as it did in the non-coffee farmer sub-sample. Here we also see that real income per adult equivalent increased slightly, but the number of people that were getting poorer is even a lot higher than with the non-coffee farmers. This is 11 percent. And people living in extreme poverty even increased up to 17 percent, again indicating a very large increase in inequality.
So then finally some coffee-specific indicators: What’s interesting here is that what we see is that the number of trees actually decreased sharply over the years; but coffee production actually increased, and then, as a result, a logical result, the kilogram per coffee tree also increased very sharply. Income from coffee—apparently coffee still is not a major source of income for farmers, because in ’90-’92, it only constituted 9 percent of farmers’ income, and then in 2002, it was even going down to 7 percent.
Basically we did the same analysis when we divided our sample up into land quintiles, using three quintiles, so we had the very land-poor farmer, a middle group, and the land-rich group. Basically we saw a similar pattern. These are the main results. IN 1992, there was a clear relationship between land endowment and coffee cultivation; so people with more land were more likely to grow coffee. But in 2002, that relationship completely disappeared in our sample. The second result was that the number of trees increased by land-owning [?], so people with less land were less likely to grow coffee than people in the highland group, the land-rich group; but it decreased over time, as we saw in our previous result. The number of trees sharply declined between the two data sets.
And then our final results: The kilogram per coffee tree increased over time, so this indicates an increased efficiency; but that was less for the land-rich farmers than it was for the land-poor farmers.
Just very quickly on coffee practices: What we see is that the number of farmers that stopped during the genocide or just after the genocide is pretty big. It’s about 26 percent, and then just after the genocide another nearly 30 percent. We also see a sharp decline in coffee farmers that invested in trees in the last six years. That went down from 41 to nearly 12 percent only.
Then we developed an index of quality of coffee cultivation. This index consisted of whether or not they used fertilizer, they used mulching, these kind of things. We also see here that farmers that were still doing all these kind of things in 1992, this figure severely dropped from 48 to about 14 percent.
Then a final note on the prices that they received: We expressed the prices that they received. So here, in 1992, that was 100; in 2002, that was 200. The other list of categories are expressed as percentages of those prices. So what we observe here is that farmers in 1992 were much more flexible and much more positive actually, because here prices, if they would go down to about 80 percent of what they received in the previous year, they would abandon their coffee trees. In 2002, they were much more careful. Prices could go down to 60 percent, and then they would abandon their coffee trees. Same with the uprooting, but then slightly different numbers. And here farmers would increase the number of trees if prices would go up to about 150 percent of what they received last year. Then in 2002, they had to double before they would actually increase the number of trees.
I will very quickly say something about the econometric analysis that we did. Basically we wanted to establish a causal link between the war and coffee cultivation. So what we did was an econometric analysis; but then we had a problem because we didn’t know quite which way the relationship would work. We focused here on property rights and investment in coffee; but then we had this dual relationship: Do property rights enhance investment, or does investment actually increase security of getting land?
Very quickly, what we did, we estimated a [?] model, so probability of cultivating coffee defined by whether or not they had access to the land, the size of the land they had, and the number of socio-economic characteristics. Just to finish, some results: Focusing on the second column—I won’t go into the first column because that’s a result that wouldn’t quite be, [couldn’t] be trusted, because of problems in the econometric specifications; but we need to present this as well to make this comparison. What we see here is that, indeed, ownership status or land rights are a positive predictor of cultivating coffee. That’s a very large coefficient and statistically significant at the one percent level. I would like to leave it at that, since we’re running out of time.
PV:
All right, thanks, Eleanora.
“Civil War and Child Health in Burundi”
by Philip Verwimp
So my presentation will be another empirical application of the household level effects of civil war, and I’m focusing on the health effects for small children. Why is this an important issue? Because there’s a literature out there that says that negative health shocks early in life, say before age five, have persistent effects over the lifetime of the child. So when you’re negatively effected as a young child, you carry this on for your entire life. And so we want to see if the civil war in Burundi has a negative effect on child health, and then we know that it has welfare consequences for the entire [?] of these children. And for this audience I don’t have to emphasize that Africa was effected by civil war throughout the ‘80s and ‘90s, with a lot of research cross-country, or micro-research, on the cause of this war, but there’s very little micro-level research on its consequences.
Especially people focus on mortality, and of course it’s very important; but war also has a lot of indirect effects, namely, it’s not because you’re not killed by the war that you’re not negatively effected. There are a lot of people killed years after the war stops because, for example, during the war the health infrastructure was destroyed and they couldn’t get care, or they couldn’t buy malaria tablets, and so they die from malaria. So all kinds of indirect effects from war occur.
And as I said, there’s a link between shocks early in life and long-run negative effects of adult welfare. And more and more in the economic literature, we find papers showing this long-term effect of shocks early in life. And so our research question is what’s the impact of armed conflict on health status of young children?
Some of the literature out there is Alderman, who as you know has a paper on Zimbabwe where they show that healthier children complete more grades, and they use panel data so they exclude family income and all kinds of other things that effect the completion of grades. They really link this to child health. And child health is measured as anthropometric status, so the child reaches a certain height for its age. That’s what we use as a health indicator.
We find the same effects in Pakistan. There’s a paper on long-term effects of the famine in China on adult health and work capacity. Children who were effected by that famine when they were very young reached less height as an adult and have therefore also less physical strength; and in a rural economy people learn less when they don’t have physical strength, because they can’t do much work as a strong adult.
There’s a negative effect even when children are in utero and their parents are effected negatively, for example, by the influenza, which we talked about this morning. They have lower birth weights, and when they have lower birth weights, it affects them throughout their lives. So we expect long-term effects for Burundian children when their civil war deteriorates their child health.
Now, what method do we use? It’s very important to do this econometrically correct, and that’s why we spend most of our attention to do this econometrically correct. So what we will do is we use the variation that the civil war had in Burundi over the regions, so it did not occur at the entire territory at the same time. Some provinces were affected in ’95, some provinces in ’96, some provinces in ’97. So what we can use is the variation over space, as well as the variation over birth [cohort?]. So what you want to see is if the child living in that province at a time it was effected by the civil war. Then we say, okay, this child is exposed to the civil war. If you were not yet born at the time of the war, then you’re not effected; or if you’re living in another province that was not effected by the civil war, then you’re not exposed. So we compared the health status of exposed children with non-exposed children to pin down exactly what the effect of the civil war [was] on child health. So we control for selection effects, estimate problems and birth [cohort fixed?] effects. We examine Burundian children born between ’93 and ’98; it’s in fact the same data set that Tom uses in his earlier presentation. And we combine this data set with secondary literature from the historical and political science literature on where did the conflict occur and at which time; because we want to link the age [cohort?], the birth [cohort?], with the incidence of civil war at his or her residence.
Research summary: The health status of women and children is very poor. Civil war negatively impacts child health. […?] children exposed to civil war are between 0.44 and one standard deviation lower, depending on specification, and we argue that this is the causal impact of the war, and it’s not due to a spurious relation.
One of the policy implications of our work is that early interventions may yield a big impact, and the impacts are big exactly because children carry over this negative effect throughout their lives. So if you intervene early in life, then it yields positive effects throughout life.
A short overview of civil war in Burundi: It’s basically a series of coup d’etats followed by or preempted by insurgencies. And in fact the Burundian conflict is often described as a struggle for state resources, where the person or the group [that] is in power uses state resources to provide for his or her kin. It’s basically a zero-sum game, and Burundian coffee is often defined as a zero-sum game. Either you are in power as a ethnic group and you provide for your kind, or you’re not in power, and then you’re at the losing side of the game.
The conflict in ’93 started in the north to the middle of the country in ’93, and then spread out gradually over time and over the provinces. So in our econometric analysis, we use this fact that there’s spatial variation in the conflict, and not all good [cohorts?] were effected at the same time. Here you see which provinces are effected at which time; for example, the first province, [...?], they were among the first ones effected. […?] provinces are later effected. So this a schedule on which province was effected by the civil war at which time.
We used a survey that was collected by the National Bureau of Statistics, together with the World Bank. It has about 3,900 rural households, and on these rural households we’re able to use 1,169 children under five with complete data. We quote the civil war [?] as one if the child was exposed to the civil war, meaning it was living in the province at the time the civil war occurred in that province. And so explored province and cohort of birth variation.
This is just a descriptive table saying that only 40 percent of children are adequately nourished in Burundi, meaning that 60 percent of Burundian children are undernourished, which is one of the highest percentages in the world, especially when we know that about 33 percent, or one-third, is severely malnourished. So that’s a standard delineation lower than -3, if it’s severe undernourishment. These children are not exposed to the civil war, and these are exposed to the civil war. So the exposed ones have a much more detrimental health status.
Before doing the final analysis, we have to counter two arguments: The first is that what we observe is not an effect of age. Because from the health literature we know that as children get older, they accumulate the negative effects of having not enough food; so children of four and five years old have a worse health status than a child that’s only one year old, because it’s a cumulative process. That’s one thing. And we know also that if children are older, they have more chance of being effected by the civil war, of course, because they’re already four or five years old, compared to a child that’s only one year old. She or he has only one year of time to be effected by the war. So we first have to exclude the fact that maybe it’s just an effect of age that we observe.
[Now speaking from the exhibit and away from the mike, so less audible on the recording] So we would do a […?] table […?] children not exposed to the civil war […?] and children exposed to the civil war […?], and then we can compare horizontally and vertically. We know that older children have a worse health status. As I said, that’s a normal process. In peaceful countries, we always find that. […?] we see a significant effect of the civil war. So older as well as younger children have worse health status; and importantly, the [….?] result is not significant, meaning that comparing old and young, and civil war to non-civil war, we don’t find any effect. So the effect is really attributed to the fact that they were exposed to the civil war, and it’s not just an age effect.
Here you see a polynomial analysis, where the bottom line is the children effected by the civil war, and the upper line is children not effected by the civil war; so you can also see here that it’s not an age effect. It occurs for all ages and all […?].
The second thing that we have to exclude is the fact that maybe it’s due to poverty. Imagine that the civil war hit the poor provinces first and what we observe is worse child health just because of poverty. […?] compare children not exposed to the civil war compared to children exposed to the civil war in poor versus non-poor provinces, again with the […?] data. And it seems that poor versus non-poor has no significant effect; but inside the poor provinces the children exposed to the civil war have worse health status, and it’s the same thing for the non-poor provinces. So this table also shows that in fact the observed negative effects of child health have nothing to do with poverty. They all have to do with the exposure to the civil war. And the correct way to do that is in a multi-variant framework, which I will come back to later.
First, the numbers of these 1,169 children: A little more than half living in poor provinces and non-poor provinces were exposed to the civil war. So it’s not the case that only the poor or only the non-poor provinces were exposed to the civil war. Both were exposed to the civil war.
What we also observe is that in fact it was first the richer provinces early on in the civil war that were effected by the civil war, not the poor provinces. And we observe a particular negative effect of the civil war in the rich provinces. So that’s another indication for us that it has nothing to do with the poverty status of the province. Here, for example, we have some poverty indicators, we have the percentage of households with any education—this is before the war. Thirty percent for early effected in the war, 36 for late effected, and 36 for no civil war. And the same for […?] so this is the percentage of people living under the poverty line. And we see in fact that the provinces not effected by the civil war are poorer, because they have a higher percentage of poor people. Again, for [?] analysis, […?] effected in the civil war, meaning in the ’94-’95 period, […?] worse health status compared to children in provinces later effected in the war and children in living in provinces not effected by civil war.
So to do this econometrically correct, what we do is we have alpha-j for province […?], we have delta-[?] for birth [?], and then what we do is we interact the conflict province at the time when the war was occurring there with the fact if the child was living there in that province. That’s beta-1. So we’re interested in beta-1. We also include another [?] measure, because there’s a child development literature out there saying that children are only vulnerable to negative shocks—[end Side A] –the mother and so on, and we interact […?]. Why is that important? Because you can say that maybe the situation in the provinces changes over time, so we want to capture the change over time. Maybe one province got very rich over time, so we have to control for that. And we do that in two ways. We interact […?] with the poverty status of the province, so poor versus non-poor; but we also interact all provinces with all birth [?]. That’s a very long list of controls to capture if there are province trend effects. So we want to control for province trend effects exactly because we’re interested in the beta-1.
And so these are our analyses. And the first regression, we don’t control for trends; we only control for province […?] effects and birth […?] effects. The coefficient of interest is 0.443; so children effected or exposed to the civil war in rural Burundi have a worse health status of 0.443 standard deviation compared to international standards. When we then include household characteristics, like age of the mother, or medical status of the mother, nothing really changes in the observed effects, and our most important specification in the fourth one, because there we include this trend effect; namely, we multiply the poverty status of the province with the […?] effects. So this is econometrically the most correct way to do it. And so we observe an effect of 0.6 standard deviations lower. So we can now most [be?] sure that this is really the causal effect of the war on child health.
That’s the Households in Conflict Network, which I co-founded. We have numbers of papers there that you might be interested in as an audience. Thank you.
Okay, I’m also moderating this session, so I have to put on my other hat again. People who have questions, please come forward and tell us your questions.
Q:
Actually I found the papers really interesting because it gets at the heart of what Jamie Galbraith saying last night [about] getting at the costs of war in a real kind of a way. It was interesting to see the real unambiguous conclusion is that certainly in Africa, or certainly in Burundi, war effects children, probably more dramatically than others.
But let me ask a question about the first two papers and make sure I understand them, and see if I can reconcile the results.
So if I understand the second paper, Eleanora’s paper, it looked like what happened in Rwanda is that there were some winners and losers, and that the dispersion of earnings rose because of the war, if I understand the table right. And if I look at the first paper—and this is where I’m not sure I understand, but I think I understand—is that if indeed there is unrest, then what would happen is that the farmers are going to lower their risk and get involved in the more safe enterprise. Well, if you get involved in the more safe enterprise, that then kind of lowers rates of return and should be a narrowing in the dispersion of earnings. So what I don’t understand is the difference between the two papers. It looks like in one paper unrest widens dispersion of earnings; and in another, it looks like it has tendencies toward narrowing the dispersion. And so the question is, do you see differences in your results, and how do you reconcile them?
EN:
I’ll respond first, I guess. Yes, I think that what you’re referring to is the table that I showed on the real income per adult equivalent, and it shows that over the years it increased, but then there is an indicator, when you look at the poverty line and extreme poverty line, that there is a larger proportion that’s actually getting poorer, or extremely poor. Indeed, that suggested that there are winners and losers.
I think that one of the basic things is that my paper is really on Rwanda, which may differ from what has happened in Burundi. So that might be a very simple explanation of the difference. I don’t know if that is indeed the explanation, because we haven’t compared that.
Apart from that, right now I wouldn’t be able to come up with any distinct conclusion on how that could happen.
TB:
It’s a very interesting remark you made. All I can say is for the case of Burundi, if we compare the last pre-war data with the 1998 data, we indeed observe a kind of leveling effect. The dispersion in welfare, or the difference in poverty levels before the war was more widespread than the difference in poverty levels five years into the war; because the poor have gotten a little bit poorer, the richer have gotten a lot poorer. So it’s a leveling effect, and it’s a very good remark. I can’t really [?] right now.
PV:
I also want to respond to that. I’ve seen also in my work on Rwanda that the rich, the people who are rich before the war are heavily struck by the civil war and the genocide, so they lose their wealth. So that’s one thing; that’s leveling off. But on the other hand, of course, you had a huge influx of migrants who bring livestock, maybe human capital, and so they become the new entrepreneurs and the newly richer people. And if you then do a post-conflict survey, then you have to include those people in your analysis. And those explain also explain why after the conflict you have more dispersion again, because the new migrants, in this case the Tutsi coming from Uganda, have more wealth than the people who experience the conflict per se. And in Burundi you didn’t have an influx of richer people. And so I think that would also help explain the difference.
Yes, Alan, please.
Alan:
In the Rwanda data, didn’t it end in 2004, or something like that?
PV:
2002.
Alan:
So that’s eight years post-conflict; whereas the Burundi data ends when?
TB:
1998.
Alan:
So the war is still going on. That might be the answer, that the end in one case is while the conflict is still going; while the other on, it’s eight years after. So in one case you may be getting the war effect; and in the other, you may be getting the peace effect.
Q:
This is not directly related to the data, but I was wondering if any of you could comment on what were the demographic outcomes within the two countries in terms of Tutsis and Hutu and people coming in from Uganda, and so forth, how it all panned out in displacements and different economic levels. I don’t know if you can say anything about that. It’s a rather general question.
PV:
Well, it’s huge, of course. You had ’94 some two million people leaving the country, mostly Hutu, and at the same time you had a large influx of Tutsi coming from Uganda. They settled in the east of the country and in the capital. After two, three years, lots of the Hutu who migrated to Congo in refugee camps came back. So that’s the Rwandan situation. A large exodus of people, of course many people killed during the genocide, and a large influx of Tutsi from Uganda.
In Burundi you had a completely other situation. You didn’t have a large influx at one point in time. You had a lot of people living in Tanzanian refugee camps over the course of 10, 15 years, and who came down, let’s say, in piecemeal. But internally you have people living in camps. So that’s a different situation. I don’t know in which way your question was actually targeted.
Q:
My question is mostly about Tom’s paper. It seems like the people who are effected by conflict, that they might have just been making rational choices to engage in these low-risk activities; so I was wondering if you could maybe conflict on the outcomes. I don’t know if you have the kind of qualitative data on the outcomes of those activities. Could they have been making better choices, or were they really just sort of responding rationally to the fact that their assets were risky, and so it made more sense to do low-risk activities?
TB:
To answer this we have to wait for the panel data, because we could just observe the direct effect of war on choices without the detour of the assets, let’s say; but we don’t know, or we can’t prove, how the situation was before the war. We can’t say yet there has been a war, and the war provinces have more safe activities, because you don’t know the situation before the war, or if there’d not been a war, would the situation anyway be the same, or not? So for that we use panel data to be sure. So I can answer this next year, but not right now.
Ann Herpel:
Hi, my name’s Ann Herpel, and I’m not an economist, but I’m a coffee drinker. I was trying to understand your figures, because you said at the beginning—and this might have been pre-conflict—but you said 60 to 80 percent of state revenues were derived from coffee. Was I correct? Okay, but then at the very end your last slide seemed to indicate that, prior to conflict, only 9 percent of household income was derived from coffee, and it had fallen to 7 percent after the conflict. So I was wondering what makes up the difference for household incomes, and where else do people derive their income, and also how is income related to the size of the plot they own? Because you just made the comment that richer Rwandan coffee producers lost more than poor ones, and so when the conflict had ended, how did the distribution of the land fall out again. Was there kind of a concentration, where people were able to come in and buy larger pieces of coffee-producing land because of the result of the conflict that people displaced, or that it was even more disbursed and people were on smaller plots, many more people were on smaller plots, which didn’t produce much income then, because they were on smaller plots.
EN:
What I think, but I’m not an expert on Rwanda yet, but I think what happened after the conflict was that a lot of people lost their right to the land and also they abandoned their coffee trees; so when they finally got back after the conflict, if they ever did, there was not much possibility to actually buy coffee cultivating land and start cultivating coffee again. So that would explain, in part, why there was a slight drop in income from coffee; because it does require three years, actually, to plant coffee and to get revenue from coffee. So after the conflict, people were not very willing to make these real big investments anymore.
The issue on land is that, if I’m correct, after the conflict, the government actually—You want to comment on that? Okay. So after the conflict, the government had this policy that they were going to redistribute land, so there was absolutely no real choice of people to actually go and buy land. But the government did some redistribution policy to give— I don’t know if they actually gave back land to people who previously held rights to the land, but there was this whole resettlement and redistribution policy going on. People were not very able to actually go and buy new land.
PV:
So if you think of farm income, you have to distinguish subsistence income. So people grow crops for their own consumption, and when we confuse income as economists, we also include that. So income from subsistence crops counts for about half of farm income. Then you add some 10 percent of cash crop sales, like coffee; then people have a wage income from working on someone else’s farm or working off-farm sector; and then people also have transfers, coming from household members who live in the city; and people also have some income from some livestock. So all these income sources, then, if you add them, you come to about 100 percent. But most of […?] for the average farmer is from subsistence income.
And when I say that rich people are more struck by genocide and civil war, it’s not only rich coffee farmers; it’s richer people in general, meaning people, for example, who have a small company, or who make something that could robbed, or who sell radios, for example, or who have a lot of cattle that can be stolen and traded. So those are the richer people, not necessarily the coffee farmers.
Wim was first, I’m sorry, Jurgen.
Wim Naude:
[…?] in the selection of crops due to the effects of the war. [……?] if the war leads to […?] diversification of crops [……?].
TB:
If I understand correctly, it’s comparing the ex-ante poverty level with the ex-post poverty level.
WN:
You’re saying, ex-post, poverty has increased, but you’re also saying that the risk has decreased.
TB:
The risk as a result of the conflict? The risk in assets?
WN:
Yes, yes. […?]
TB:
[…?] activity choice.
WN:
[……...?]
Jurgen Brauer:
First of all, thank you very much for the three papers and for the Households in Conflict Network effort. I think that’s really great work. I have sort of a similar question, I gather, in this case with respect to Rwanda. A number of years ago I came across a very fascinating paper on bean production in Rwanda. Apparently in Rwanda, they describe themselves as the bean capital of the world. There are apparently hundreds of varieties of beans. Rwandans really like that. That of course is a food crop. I gather that’s what the women grow, rather than the men growing the coffee. So it seems to me that in your analysis, when you focus on the coffee, you perhaps overemphasize what’s going on, because you’re not really looking at the totality of the household. I was wondering--you pointed to that to some degree--if you can give us some flavor of what really is happening in the household as a joint unit, not just what is happening to one part of the household, that is, the men, but also what is happening in the household, that is the women, perhaps even the children, in the production process. I wonder if that might not influence your results.
EN:
I think that’s an interesting point that you make. This is preliminary work, and in our discussions, we [?] that indeed we would look at the more broader picture. One of the reasons that we wanted to narrow our focus on coffee was because in our data set we had a variable that explains whether or not people migrated because of the conflict. We wanted to establish that this [?] between genocide and investment, in this case, investment in coffee. And what we did actually, because we didn’t have a very specific indicator of shock, conflict shocks, at the household level, we proxied this by using the migration status of people. So people that migrated were, according to our framework, had experienced a larger shock in terms of violence than people who did not migrate. We related that really to investment, because what we thought was that actually migration, and people who wanted to migrate, as I think I said before, are people that were displaced. They didn’t have any access to their land anymore. They lost their right to their land, so there was a need correlation between losing the land of their rights [sic] and being a migrant, being in a refugee camp. So in an econometric sense, that was a economic feature that we used for this specific analysis, and then focusing in on coffee. So that was basically our reason.
PV:
So it was not presented, this analysis [of] migration and land rights because of time element, but we were interested in coffee because it’s an investment over time, and if you migrate, you may lose your land rights. So were interested in this relationship between property rights and investment. And coffee is an investment, whereas beans are not an investment. But your point is well taken, and if I may add something to beans, beans are a bit of, let’s say, a mystery in Rwanda. Why is it a mystery? Because it’s the largest consumption crop for people, so people like beans a lot. Why? Because it gives protein. But it doesn’t give high yields to land. And in a land-scarce country, a farmer needs high yields to land. So the Rwandan farmer has the problem that he wants, or she wants, to consume a lot of beans, of protein, but she or he can’t grow them because he or she has to grow more strongly yielding crops on a small plot of land. So what it does is to grow more high-yielding crops and to exchange them on the market for beans. So it’s not true that all Rwandan farmers cultivate the beans they eat. They trade it for other crops, high-yielding crops. Half of the beans that are consumed in Rwanda are imported from neighboring countries exactly because Rwanda is land-scarce, and it would be irrational for farmers to grow beans, because they’re very land-intensive and low-yielding.
Q:
First of all, my compliments on all three papers, and informing our understanding of the economy of Rwanda and Burundi. The president of Rwanda, Paul Kagame, has been touring in the United States several times seeking investment in Rwanda. Now in the light of your analyses and the fact that many American well-disposed investors would like to help Rwanda recover from its genocide, what would be your advice to American NGOs and profit-oriented investors wanting to respond in both an economically and socially production way to this solicitation of investors by the Rwandan government?
PV:
The plight of Rwandan citizens is like enormous, yes, so I would definitely wish that Americans would invest there, that’s for sure. I wish all the best for its population, that’s for sure.
There is some good examples of it. Rwandans used to get a fixed price for their coffee, as you saw in Eleanora’s paper, which meant that they were not really interested in the quality of the coffee, because they got a fixed price no matter what the quality was. And it was often, then, that they produced low-quality coffee. Now it’s a liberalized system under President Kagame, and so now the farmers get a price according to the quality of their coffee beans. And they’ve been working with USAID and with American coffee companies to produce high-quality coffee that can be marketed on the American market. Now there are some successful results with high-quality coffee in the south of Rwanda. It’s called the [Mabarra Project?], [after] the name of that commune. And so the farmers actually now learn that if they have high-quality produced, they get a better price, and the government doesn’t take the premium. Themselves, they get the high price; where, under the fixed price regime, the government would get the profits if the international price was high. But now they get the profits. Of course they have to make more effort, they have to produce high-quality coffee. So that’s a good example of working with American coffee companies for high-quality coffee.
Now even in peacetimes, it’s very difficult to do something profitable in Rwanda, because it’s landlocked, it’s land-scarce, it’s 1,500 kilometers away from the nearest port; so it has a competitive disadvantage vis-à-vis all its neighbors who have ports; so it’s very difficult to do anything productive. The Rwandan government has heavily invested in information technology and computer science, more like the Indian model, where, for example, they could do transactions, computer-based transactions in countries that are independent of location and have trained people for that. That’s one thing.
So the difficulty for Rwanda is that who will benefit from the development? And as you know, there are many people in prison, Hutu, perpetrators of genocide, and most of the Tutsi offspring are in the university system. Over the longer term you create an imbalance there, where you educate the Tutsi, and the Hutu are in prison. So that’s a big difficulty for long-term development. So I would suggest that any potential investors should also look at the equity of its investment, who’s benefiting from its investment. Is it really the population at large, or is it a specific group attached to the elite.
Q [Mike Intriligator?]:
You probably know Paul Collier at Oxford, who’s done a lot of work on conflicts in Africa. [?] has worked in the previous session. He gave a paper at the Copenhagen Consensus; I was his discussant at that conference. It came out as a book, Bjorn Lomborg was the editor: Global Crisis, Global Solutions. The gist of his paper this: that most internal wars and civil conflicts are in Africa, which of course is absolutely true; and he went on to say that the cause of these conflicts or civil wars is poverty, and the way to overcome that poverty is through foreign aid. So the way to reduce conflicts in Africa is providing more foreign aid.
I totally disagree with every part of that analysis, because there are a lot of poor countries that are not involved in conflict at all. Foreign aid, I think, actually in some sense stimulates conflict, because when you support a regime, the people on the outs are jealous, they want to come in and take those funds for themselves. Most of the [?] regimes have corrupt leaders who stash their money in offshore bank accounts […?] Hutu, and Charles Taylor, and others in Africa. And people on the outside don’t want to take over to help the country; they want to get the money themselves in their own private bank accounts; they want to stash it there. So foreign aid actually acts as a stimulus to more conflict.
And that’s the gist of my analysis. I was curious about your reaction to Collier’s proposals on how to prevent civil wars in Africa.
EN:
I guess I’m the one to respond to that because, in addition to this paper, I do have another paper on foreign aid and the risk of civil conflict in sub-Saharan Africa. What we did was a cross-country analysis.
The story of [?] and the argument you give is I think only one side of the story. Indeed, there is theoretical and also, to some extent, empirical evidence that once you increase foreign aid flows, the price of actually capturing the state is getting bigger, and it provides an incentive for people to go and overthrow that state.
On the other hand, the government can also become so powerful with this foreign aid, [it’s done?] a sort of perverse effect […?]; but nonetheless it can become so strong that it actually provides a business […?] to go and fight. That’s the kind of perverse effect of it, but which reduces risk of conflict.
There can also be, of course, a benevolent effect of aid that it actually is being put to good use, which reduces risk of conflict.
In our analysis, what we find is that foreign aid flows do reduce conflict risk by about 6 percent; increase in aid flows, 10 percent […?] and a 6 percent decrease. So that would kind of counteract against that argument that it would stimulate conflict.
I guess it’s a very difficult issue, because we find this result; others found either nothing or the opposite result. My main conclusion would be that this is only part of the story, and it’s really a very macro approach that they use. What you really need to do is take it down to the micro level and look what the chains of transmission, channels of aid are or can be, and what they really are doing, and what provides an incentive to actually […?] or not, and that they’d have an influence on that.
Q:
[……?] Now what happens to people who don’t have an option of engaging in any kind of activity?
TB:
Well, as long as people had some labor, you can engage in an activity. It will be unskilled labor on other people’s farms, so you can always engage in some activity.
Q:
[…...?]
TB:
If you don’t exercise any activity, your total income is zero theoretically speaking, so the […?] would be skipped for the analysis, and we would [?] control for selection effects, of course.
Q:
I’ll tell you why I think that issue is important, because it’s quite possible that those who have no option of [?] income earning will enhance conflict in the area. I could think of several studies. One comes to mind, an [?] study by Eric Hoffman on primitive rebels takes exactly that line, that indeed there is a seasonal pattern to armed [revolution?] during the period of surplus labor. If you take that kind of approach to […?] conflict, then you see that it’s significantly effected by whether or not people have the choice of engaging in any kind of activities. [……?]
TB:
[….?] very poor households […?] could easily join the rebels [….?]
Q:
[…?] I think it’s very important that you should allow for that in your analysis.
PV:
All right. Thanks very much, everybody. [end tape]
|