YT YouTube2Text

YouTube2Text → Transcripts → BREAD-IGC virtual PhD-level course 2026 | Lecture 8 Conflict and Fragility

BREAD-IGC virtual PhD-level course 2026 | Lecture 8 Conflict and Fragility — Transcript

by International Growth Centre · 14,708 words · 2,439 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00of the bread IGC series on political
0:01economy and we have uh some new and some
0:05familiar faces here uh today for our our
0:08lecture on conflict and fragility. I'd
0:11like to introduce
0:13Andrila Dubey, professor at the
0:15University of Chicago. Thank you for
0:17joining us and Mike Callen, who we have
0:20seen um
0:22at the beginning of of this series uh
0:24and he is a professor of economics at
0:26the London School of Economics.
0:28And uh we're just really excited to have
0:30both them here. This is their specialty
0:33of specialties and we're we're really
0:35excited to just uh all be sponges and
0:39and
0:40get as much from from these 60 minutes
0:43as we can.
0:45Um
0:45if I could just uh repeat myself uh if
0:49you'll bear with me,
0:50all past lectures and slides and
0:52recordings are all on the IGC website.
0:55So, if you missed us uh
0:57a lecture or two, that's where they are
0:59and I'll I'll put that in the
1:00Um if you have questions, please put
1:04them in the Q&A box and I'll kind of
1:07cumulate those and and see which ones we
1:10we can pull out at the end for the Q&A
1:12session.
1:13And uh if anyone would like to be a live
1:16participant, please raise your hand
1:18right now and I'll randomly select 10 to
1:2015 of you and you will have the
1:22opportunity to ask your questions uh
1:24live
1:25to uh Mike and Angela.
1:28So,
1:29I think that's it from from my end. Um
1:33thank you again all for being here
1:34making the time and Andrila, I will hand
1:37over to you.
1:39>> Great. Thank you so much, Caleb. It's
1:40really a pleasure to be here with all of
1:43you today. So, I'm going to kick it off
1:45by sharing my screen.
1:55I'm going to go first and lecture for
1:57about 30 to 35 minutes. I'm going to
1:59talk about the role of economic shocks
2:01and conflict, and then I'm going to kick
2:03it over to Mike Callen,
2:06who is going to talk about the
2:07challenges of state building in
2:09conflict-affected
2:11and fragile settings.
2:13So, I want to start off by discussing
2:17with you all
2:19why we're even talking about conflict
2:22here today in this political economy of
2:24development class.
2:27For one thing,
2:29because there are more active conflicts
2:31in the world today than at any other
2:34time since UCDP started collecting this
2:38data tracking conflict events.
2:41In 1940,
2:43which began in 1946. So, there were 61
2:46active conflicts in 2024, 65 in 2025,
2:51and this is a growing problem because of
2:53the number has essentially doubled over
2:55the past decade.
2:57Now, almost all of these conflicts are
3:00internal civil wars rather than wars
3:03between states. So, interstate conflicts
3:06are very high-profile, but they're
3:07relatively
3:08few in number.
3:11Moreover, these conflicts are not
3:12distributed evenly throughout the world.
3:15They are concentrated in the world's
3:18poor countries.
3:20Um Paul Collier estimates that about 73%
3:23of what he calls the bottom billion
3:24societies,
3:26uh the the poorest countries that
3:27experienced little over 1980s and 90s,
3:30um 73% of them experienced at least one
3:33civil conflict, and the World Bank
3:35estimates that by 2030,
3:3860% of the world's extreme poor will
3:40live in fragile or conflict-affected
3:43settings.
3:44So, these statistics
3:47already begin to point to the idea that
3:50maybe poverty and low income plays some
3:53role in spurring conflict, and that is
3:57going to be the focus of the lecture
3:59here today is the role of economic
4:02shocks in civil conflict. Now,
4:05for most of this lecture, what I'm going
4:06to focus on is the commodity price
4:08shocks literature as a kind of case
4:10study in how at each stage of that
4:13literature, new data sources opened up
4:16new questions we could answer. And then
4:18at the end, I'm going to talk about um
4:21the new generation of data and what AI
4:24opens up as we look to the future.
4:28What I want to start by doing though is
4:30talking about some theoretical reasons
4:33why we should expect income to affect
4:35conflict.
4:37So, we have one set of theories that
4:39suggests higher income should less
4:42should lead to less conflict. So, for
4:44example, there's the opportunity cost
4:46theory which suggests as higher wages
4:50as wages are higher, the foregone cost
4:54of joining conflict increases and armed
4:58groups that recruit people into conflict
5:00have a harder time and find it more
5:02expensive to recruit them into conflict
5:05and that is something that predicts that
5:07higher income should less lead to less
5:09conflict.
5:10There's also the state capacity idea
5:12that richer states have more resources
5:15at their disposal to essentially
5:18suppress conflict or to buy off the
5:22opposition.
5:24But,
5:25there's another set of theories that
5:27actually suggest higher income could
5:29lead to more conflict.
5:32Higher income means there's more stuff
5:35to steal.
5:36And if we conceptualize conflict as a
5:39form of violently predating on these
5:41resources,
5:43more to steal should lead to greater
5:45conflict. And the state as a prize is
5:48essentially a subset of that theoretical
5:50idea which suggests that a richer state
5:53is then more worth capturing.
5:56There's also the possibility that more
5:58resources means more financing to make
6:01rebellions feasible.
6:02So, as you can see, the theoretical
6:06frameworks point in two opposite
6:09directions,
6:10uh and the effect is theoretically
6:12ambiguous. So, we have to turn to
6:14empirical analysis to see what we find.
6:18Now, I want to start at the beginning
6:19and what the first era of empirical work
6:23in this area found.
6:25In that first era, we had country-year
6:28level data. And that work started by
6:31running cross-national regressions where
6:34outcomes would be something like whether
6:35a country is at conflict that year or
6:38started with a conflict onset that year.
6:42And on the right-hand side of the
6:43regression, we would have something like
6:45the country's GDP per capita.
6:48And what those early studies found were,
6:50for example, that countries with higher
6:52GDP per capita had lower levels of
6:55conflict. This was the work of Collier
6:56and Hoeffler and Fearon and Laitin.
6:59Work by Ted Miguel and co-authors also
7:01found that GDP growth led to lower risk
7:05of civil war.
7:07These set of authors had
7:09different interpretations of that same
7:11findings. Fearon and Laitin, for
7:13example, favored the state capacity
7:15hypothesis, whereas the other authors
7:17favored the opportunity cost theory.
7:20But this early work, given the data
7:22available, was not able to present
7:24direct evidence on mechanisms.
7:28So, what that negative relationship that
7:30was uncovered empirically seemed to
7:32point to this side of the aisle, higher
7:35income and less conflict.
7:41So, what about this side?
7:43Does that mean that
7:46there is no evidence that these types of
7:51positive relationships running from more
7:54income to more conflict aren't there?
7:57Well, it's not necessarily because the
7:59country-level estimate is really a net
8:01effect. If some income raises wages,
8:05while other income raises the price, GDP
8:08mixes the two together, and we only see
8:10whichever effect dominates.
8:14The problem is that GDP is very
8:16aggregate. It tells us income went up,
8:19but not who it went to.
8:21And who it accrues to, labor or capital,
8:24may be exactly what determines the
8:27direction of the effect.
8:28So, this was the idea behind my paper
8:32with Juan Vargas, where we used price
8:35shocks to different types of commodities
8:37that are either labor intensive or
8:39capital intensive to see whether the two
8:43push conflict in opposite directions.
8:51So, let's walk through that in a little
8:52bit more detail.
8:54In our framework, which was formalized
8:57by Dal Bó and Dal Bó, there are two
9:00sectors. There is a legal sector,
9:04like coffee or oil, producing a
9:06commodity where workers earn a wage by
9:10working in that sector.
9:12And there is also a conflict sector.
9:17This conflict sector steals a share of
9:20production,
9:21and workers in that sector split up what
9:24is appropriated, and that gives them
9:27what they earn from working in the
9:28conflict sector, which is the return to
9:31appropriation.
9:34So, notice that here
9:37a increase in the price of the legal
9:41commodity
9:43is going to move both the wage as well
9:47as the return to appropriation,
9:51and that is going to push in two
9:54different directions in terms of the
9:56conflict outcome.
9:57So, let's walk through the top of this
9:59figure.
10:01Consider
10:03the price increase occurs,
10:05production in the legal sector ramps up.
10:07As product demand increases, labor
10:10demand increases, and this pushes wages
10:13up.
10:15As wages increase through the
10:17opportunity cost effect makes it more
10:20difficult to recruit workers into the
10:22conflict sector, there is more labor in
10:25the the legal sector,
10:27less in the conflict sector, and that
10:29predicts the conflict should fall.
10:33Now, down here in the bottom part of the
10:36figure,
10:39there's the potential opposite effect.
10:42When the price increases and production
10:44increases,
10:45what is stolen as a fraction of
10:47production also increases.
10:50This raises the return to appropriation,
10:54which predicts that there should
10:55actually be more labor in the conflict
10:57sector.
10:59And with a larger conflict sector, there
11:01should be a increase in conflict.
11:06That's what we call the rapacity effect.
11:09So, which of these two effects should
11:11dominate? The opportunity cost effect
11:14or the rapacity effect?
11:17Well, our answer was it's going to
11:20depend on the labor intensity of the
11:23commodity that is being produced. And
11:25so, we took this idea to data in
11:29Colombia, where we honed in on one
11:32labor-intensive agricultural commodity,
11:34which is coffee, and one
11:36capital-intensive commodity, which is uh
11:39natural resource, oil,
11:42to examine how these effects play out.
11:45So, on the labor-intensive side of the
11:48aisle, you're going to have a price
11:50increase
11:52having larger effects on wages than the
11:55return to appropriation. In those types
11:58of sector, the opportunity cost effect
12:00should dominate, and the prediction is
12:03that then when the price of a commodity
12:06like coffee increases, you should
12:08observe less conflict in the
12:10coffee-producing areas.
12:12Now, what about the capital-intensive
12:14side?
12:17Well,
12:18these commodities, when you get an
12:20increase in the price, that is going to
12:24move the return to appropriation more
12:26than it's going to move the wages
12:28upward. In other words, the rapacity
12:30effect will dominate, and there the
12:32prediction would be an increase in the
12:35price of a commodity like oil should
12:37lead to more conflict in the
12:40oil-producing areas.
12:43Now,
12:45notice that in order to go and take this
12:50uh idea to the data, we needed a bunch
12:54of additional things. For one thing, we
12:57needed subnational conflict data.
13:01Data that told us not just whether a
13:02country is at war, but how much conflict
13:05there is in different regions of the
13:06country.
13:07And we got that through this wonderful
13:09data source called CERAC, which tracked
13:12conflict events, attacks by
13:14paramilitaries and guerrillas in over
13:16950 municipalities in Colombia
13:20through both media-reported events, but
13:23also through a network of Catholic
13:24priests that operated throughout the
13:27country um and reported continuously on
13:30how much conflict there was in each of
13:32their locations.
13:34An analysis like this is also going to
13:36require us to have subnational data on
13:39what type of production is taking place
13:41in what municipality.
13:43And finally, it's going to require us to
13:45have exogenous variation in prices.
13:49Now, in the case of oil in Colombia,
13:51that's straightforward. Uh Colombia, you
13:53know, it's a small producer of oil in
13:55the world market, so it's essentially a
13:57price taker, and we can just use the
13:58price directly.
14:00However, Colombia is a major producer of
14:03coffee, so you can't directly use the
14:05price, but during the period of our
14:07analysis,
14:08a bunch of the other major
14:10coffee-exporting companies uh rather
14:13countries like Brazil uh in Vietnam and
14:16Indonesia were ramping up their
14:18production of exports, so we can use
14:20their exports as an instrument for the
14:23local price of coffee in Colombia. And
14:25using that approach, you can do a
14:27difference in differences analysis where
14:29you're essentially interacting the
14:31baseline production of the commodity
14:33with the exogenous price movement while
14:35controlling for municipality in your
14:37fixed effects, which is the estimation
14:39strategy that we used.
14:41Using this approach, we found that
14:44indeed coffee price shocks led to less
14:48conflict in coffee areas in terms of
14:51reduced guerrilla and paramilitary
14:54attacks.
14:56And in contrast to that,
14:58the oil price shocks led to greater
15:01effects on conflict as seen in greater
15:04paramilitary attacks in oil areas.
15:07Now, the advantage of going within
15:09country for the analysis also enabled us
15:12to make a better progress on the
15:14mechanisms behind these effects in terms
15:17of generating some empirical evidence
15:19for those.
15:20So, to examine effects on the
15:23opportunity cost effect, we were able to
15:24look at wages and hours workers worked
15:27through um
15:29labor force surveys. And then to look at
15:32the rapacity mechanism, whether there is
15:34evidence of predation, we were able to
15:36look at how much revenue was accruing to
15:38the municipalities from taxing
15:42resources like oil as well as
15:45extortionary
15:47kidnappings that were taking place of
15:49municipal officials and political
15:50leaders, which was a means of extracting
15:54resources from the municipality.
15:57And what we found in this evidence is
15:59indeed that coffee price shocks move
16:01wages and hours, which suggests that the
16:03opportunity cost mechanism is at play
16:04there, while the oil price shocks move
16:07the revenue and the political
16:09kidnappings suggesting the rapacity
16:12channel is at play with the oil price
16:14shocks.
16:19But our findings left a set of questions
16:22open.
16:24For one thing,
16:25was this rapacity effect that we were
16:27observing just evidence of predation?
16:32Or could it also reflect financing? For
16:34example, once the armed groups moved
16:36into the oil areas, they then had the
16:38resources at their disposal to be able
16:41to carry out attacks. And this is
16:42something we weren't able to examine
16:44directly in our work.
16:46It also left open the question,
16:48does this effect generalize beyond the
16:50Colombian context to other international
16:54contexts?
16:55Now, when you begin to examine that
16:57questions, there's actually two routes
17:00you can go for doing that. You can turn
17:03to cross-national data, or you can pull
17:06together a bunch of different
17:09within-country and cross-national
17:11studies in a meta-analysis. So then,
17:13let's look and see what the literature
17:16did subsequent to our work.
17:19Now, the first time
17:21that there was a cross-national analysis
17:23done of the commodity price shocks by
17:25Bazzi and Blattman, they were still
17:28measuring conflict at the aggregate
17:31country-year level, and they were still
17:33measuring commodity shocks at the
17:36aggregate country level. So, they were
17:38just looking at the whole of the
17:40commodity basket in the country.
17:43And what they found with that approach
17:44is that, you know, some evidence that
17:46the price shocks affect conflict
17:48intensity and duration, but not the
17:50start of conflict onsets, and in
17:52general, the results were mixed and
17:55weak.
17:56But it turns out that this is because
17:59the effects actually differ across
18:01regions within a country, and the
18:03country-year data were too coarse to see
18:07these.
18:09With the advent of better data, there
18:12were then a fantastic set of studies
18:17that looked at
18:19small grid cells within countries, but
18:25were able to look at this
18:26cross-nationally.
18:28So, Berman and Couttenier, for example,
18:31in their 2015 RESTAT paper,
18:34were able to work in grid cells that are
18:3750 by 50 km in size. So, to give you a
18:41sense, this is like roughly the area of
18:43the greater Chicagoland um
18:45area, and using these small grid cells,
18:48they could measure how much commodities
18:50were produced in those grid cells, and
18:52using the ACLED data source, they could
18:54actually look and see how much conflict
18:56events were taking place in those grid
18:58cells.
19:00So, once they moved to this subnational
19:02data, but across many countries, what
19:05Berman and Couttenier found was that
19:07indeed
19:08agricultural price shocks are associated
19:11with a negative effect on conflict. When
19:13the agricultural prices increased,
19:16um there was less conflict, which
19:18mirrored the effect that we observed
19:20with coffee in Colombia.
19:23They were also able to build further
19:25evidence on mechanisms. While they
19:27weren't able to look directly at
19:28opportunity cost effects, they actually
19:30presented some very interesting findings
19:32that suggested it wasn't really the
19:34state capacity mechanism at play here.
19:36For example,
19:38if you think that these price movements
19:41affect the resources available to the
19:44state to suppress conflict, firstly, you
19:46might think that, you know, the effect
19:48differs based on your proximity to
19:50capital cities where there's a strong
19:52state presence, but they didn't find
19:53larger effect near capital cities. You
19:55might think maybe these price shocks
19:57affect military spending, but they found
19:59there was no effect of these price
20:00shocks on military spending. And
20:02finally, they also didn't observe that
20:04these effects were larger where the
20:06state is better able to collect revenue.
20:14Then there was another paper by Berman,
20:15Couttenier, Dominic Rohner, and Matthias
20:18Toungi uh in their AER paper in 2017
20:21that approached this
20:24question
20:26by looking at how mineral price shocks
20:28affected conflict in Africa using again
20:31this similar approach of analyzing 50 by
20:3550 km grid cells but across the whole of
20:37sub-Saharan Africa.
20:39And using this approach, they found that
20:41increases in mineral prices led to more
20:44conflict in mining grid cells, which
20:47mirrored the oil effects that we saw in
20:50Colombia.
20:51They had a fascinating finding that
20:53these effects are larger where the mine
20:55is foreign-owned, which is consistent
20:59with the idea
21:01that firms
21:03pay armed groups to provide protection
21:05to them because when they're foreign
21:07firms, the state doesn't
21:09step in and provide this kind of
21:11protection.
21:13Now, that points in the direction of
21:15financing playing a role, but in
21:17addition to that, they also presented
21:19some very interesting and some of the
21:21first evidence that indeed financing
21:23also plays a role, which if you
21:25remember, we weren't able to look at
21:27directly in our Colombia work.
21:29So, a lot of this evidence on financing
21:32is essentially documenting that the
21:35effects of these price shocks tend to
21:37diffuse to other locations.
21:41So, for example, they find that when
21:43there's a positive price shocks to a
21:46mineral prices in ethnic homeland,
21:49that ethnic group not only fights more
21:52within that homeland, which could be
21:54consistent with both financing and
21:56rapacity and predation,
21:58but in fact, that group also fights more
22:00outside of its ethnic homeland,
22:03suggesting that it's the financing that
22:05enabled them to diffuse outward.
22:07Similarly, they also show that winning a
22:10battle in a grid cell with a mine then
22:13leads to violence elsewhere the
22:15following year, whereas winning a battle
22:18in some other type of
22:20um
22:21location doesn't produce an equivalent
22:24effect.
22:26So, through this better data, there was
22:29presentation of better mechanisms as
22:31well.
22:32Now, I mentioned there's two routes that
22:35one could take in examining the
22:37generalization of findings.
22:39The second route is to
22:42pull together a bunch of studies and
22:44conduct a meta-analysis.
22:46Now, over the time period of the last 13
22:50or so years in conflict research, in
22:52fact, there has been 46
22:54well-identified natural experiment-based
22:58studies that have looked at this
23:00question of commodity price shocks and
23:02how they affect conflict in a variety of
23:05different settings, some within country
23:07and some pulling together
23:09different studies as well. And a
23:11meta-analysis is a very nice way of
23:14seeing what the aggregate effect is.
23:17So, one such meta-analysis was conducted
23:20by Blair Christensen and Rudkin
23:23in their 2021 APSR paper.
23:27And here is what they found when they
23:29pulled together
23:31the different studies.
23:38Firstly,
23:39>> [clears throat]
23:39>> they found that indeed there is a
23:42negative
23:43effect associated with agricultural
23:47price shocks.
23:49Second, they found there is indeed a
23:51positive effect associated with price
23:55shocks to oil and gas, where they appear
23:57to lead to increases in conflict.
24:01There's a similar effect that's observed
24:04with minerals that are produced in
24:06artisanal manner,
24:07where you don't need heavy equipment and
24:10machinery to access the minerals.
24:14And in contrast to that, there is no
24:16equivalent effect with minerals
24:19where you do need actually heavy
24:22machinery to draw the minerals out of
24:24the ground.
24:25So, this contrast between the artisanal
24:28and the commercial minerals also
24:30suggests that there's some important
24:33role for dilutability of the mineral
24:36resource because these artisanal
24:39minerals are easier to essentially
24:41predate upon.
24:44And in contrast to these disaggregated
24:47effects, which is what theory tells us
24:49we should do given potentially different
24:51effects in different types of
24:52commodities, if you just pull together
24:55everything, you just get a uh a null
24:59effect.
25:03So, taking stock, what did we observe at
25:06each step of this literature?
25:10Well, what we observed is as you got
25:12better and better data in finer and
25:14finer units, this opened up new
25:16questions and gave us additional
25:18evidence on the existing ones. When we
25:20just had data at the country year level,
25:22we could just ask the question, does
25:24income matter?
25:26When we then went in and looked within a
25:28particular municipality at within
25:30country data, we then found um we could
25:33answer to the type of income shock
25:35matter. When we then went in and had
25:38many different
25:41units,
25:42but across countries, we could have
25:43asked, did the effects generalize and
25:46through which channels did the income
25:48shocks operate?
25:51So, now let's talk a little bit about
25:53what could come next. Now we have the
25:55advent of many new big data sources. We
25:57have AI tools that um have, you know,
26:01combined with computational processing,
26:04make it much easier to turn these new
26:05data sources into measures,
26:08um and let's discuss how that can change
26:10what we observe and what we can ask.
26:13So, one very exciting feature of these
26:17new data sources is that they give us
26:20the ability to measure outcomes
26:24in a many conflicts settings where
26:26statistics essentially collapse. So, for
26:28example,
26:31there's deep learning models now that
26:34can classify satellite images of
26:36buildings that are destroyed to give us
26:39a city monk panels of destruction in
26:42context like Syria
26:44or in the context of Gaza.
26:47Um
26:48sometimes satellite images and
26:50photographs rely on having clear skies.
26:53But in Gaza, the use of radar that is
26:56able to see through cloud and smoke
26:59enabled Ronaldo and co-authors to
27:02essentially measure destruction
27:05at a fine-grained level without the need
27:07for clear skies. And when they combine
27:10this with how the destruction affected
27:13night lights, they were able to measure
27:15the economic cost of conflict.
27:17Similarly, you can have incredible
27:19measures of economic activity um
27:22in the context of Sudan when the
27:24statistical system essentially collapsed
27:26because of the conflict there. Nitrogen
27:28dioxide was used as a measure of how
27:31many how much human activity was taking
27:33place because this um NO2 um measures
27:37things like
27:40engines, power plants, and Sudan also
27:42brick kilns and diesel generators that
27:45kick into place when the grid goes down.
27:49Not only do
27:51exciting new data sources like this
27:53enhance our ability to ask additional
27:56questions on
27:58uh the economics of conflict, they can
28:00help us
28:02further our study of other types of
28:04questions.
28:05So, for example, if you think about um
28:08if you contrast sort of like the vast
28:10amount of uh evidence we have generated
28:13in
28:14um questions around the economic
28:16determinants of conflict, we have much
28:18less on um
28:20the role of religion in conflict. And I
28:23think in part that's because
28:26there are real challenges to being able
28:28to measure religious adherence and
28:31religiosity.
28:33For the most part, we rely on survey
28:35measures of religion, which are really
28:37difficult to collect, especially in
28:39conflict-affected setting. They are
28:41quite coarse, and meaning they're, you
28:44know, measured at the country level,
28:46maybe once every few years. But what if
28:48you really wanted to get in and see how
28:50religious a village is, and how that's
28:52changing from month to month? Our
28:53typical measures don't really
28:55give us the ability to see that. And
28:58there's also the question of what does
28:59it mean when someone says they are
29:02um
29:03religious in a survey. You know, it
29:04could potentially reflect social
29:06desirability bias.
29:07So, what we do, and this is work with
29:10Mike Callen and Josh Blumenstock, is we
29:13use anonymized cell phone data to create
29:16a measure of religious adherence that is
29:18granular in time and space and based on
29:21observing behavior. Now, Mike, Josh, and
29:24I developed this in the context of
29:26Afghanistan, but we believe it broadly
29:28applies to many Islamic contexts.
29:32So, what's the idea behind this um
29:36measure? Well,
29:39a core tenet of Islam is that you're
29:41meant to pray five times a day. And
29:43prayer is supposed to occur during very
29:46specific prayer windows.
29:49How can we tell if people are not
29:51actually adhering to the Islamic
29:53religious norm during a prayer window?
29:56Well,
29:57you can measure the extent to which they
30:00are doing something else.
30:03What is a common activity that they
30:05could be doing?
30:07Well,
30:08they could be talking on the phone.
30:11Cuz you can't pray and talk on the phone
30:13at the same time.
30:15And that is why talking, whether on the
30:16phone or to other people, is considered
30:18to invalidate prayer in the Islamic
30:20context. So, what do we do? Um we
30:23measure how call volume changes with the
30:27start of the Maghrib prayer window,
30:31which is the afternoon prayer.
30:34Um and when we measure that change in
30:36call volume, we do find that there is a
30:39sharp dip in the call volume if you
30:42measure the call volume 30 minutes after
30:44the Maghrib prayer window opens up
30:46compared to 30 minutes before when it
30:48opens up.
30:50Now, what I want to show you is how
30:54useful data sources like this can be
30:58in capturing this type of religious
31:02behavior. So, one nice feature
31:05of the Maghrib prayer window is that it
31:09changes with sunset.
31:11To see that, what we did is we created
31:14this heat map, which plots for every
31:17minute of every day in a 2-year window,
31:21what the call volume is like throughout
31:24Afghanistan.
31:26Now, I wonder, since I told you the
31:29Maghrib prayer window
31:30starts with sunset, if you can identify
31:33the Maghrib dip in this figure.
31:37Now, what if I give you the hint that
31:38this is when sunset occurs in
31:40Afghanistan?
31:42So, then I think you can see that this
31:45light band, which reflects lower call
31:47volume, that tracks sunset, is the
31:50Maghrib prayer window. And this is nice
31:52because it allows us to separate the
31:57fall in the call volume
31:59that is associated with the prayer
32:01window from other effects like a 5:00
32:03p.m. effect because the time of sunset
32:06changes over the course of the year.
32:08But, this figure contains other very
32:10interesting signatures of religious
32:12behavior. For one thing, we notice these
32:15two lighter bands here. Why is call
32:18volume less there?
32:20Well, those are the holy months of
32:22Ramadan. So, you can see that the
32:24Maghrib dip is particularly stark during
32:26these two months, but in general, this
32:29is a time of prayer and reflection, so
32:31overall call volume is also low.
32:34Now, in the middle of the figure, you
32:36might also see that there is a light
32:39band. That is both lunch time, but also
32:42the noon prayer window. And you can also
32:45see a kind of striated pattern where
32:47every few days it looks like that middle
32:49band is higher.
32:51Those are Fridays.
32:52During Fridays there is a larger prayer
32:56that is added on to that noon prayer,
32:59which results in a longer period of time
33:02when call volume is lower.
33:06So, with this measure
33:08you can then answer all kinds of
33:10questions. And one of the questions that
33:12we ask in our paper is how does exposure
33:16to
33:18insurgent violence change religious
33:20adherence?
33:22So,
33:23I want to now
33:26look forward and
33:30see what this can open up for us.
33:34What I want to say is that
33:39there are some ground rules that we
33:40still can't break, right? Measurement is
33:43fantastic, but it is not identification.
33:47And even if we have new data sources, we
33:49still need a credible design. So, if you
33:51want to look at how, you know, changes
33:53in religious adherence affect conflict
33:56or institutional capacity, you still
33:57need some exogenous variation in
33:59something like religious adherence.
34:03And this is where the researcher comes
34:05in. It really has to be the researcher
34:07who comes up with a credible design.
34:11It also has to be the researcher who has
34:13to decide
34:15what is a question that is worth
34:17answering with this new data source and
34:20with this design.
34:23The
34:24availability of these data sources and
34:27the ability to process them through AI
34:29tools
34:31is essentially giving us reduced costs
34:33of measurement. And this cheap
34:35measurement, I think, raises the return
34:38to the value of having good credible
34:40designs and having a researcher be able
34:42to decide on good questions. What it
34:44doesn't do is produce automation or make
34:48the case for automating research.
34:51Now, what it does raise is we still now
34:53need some good validation of these
34:57machine-made measures, so we can't just
34:59argue that whatever, you know, they show
35:02is what it is, but we need a to find a
35:04way of validating that with data sources
35:06we trust. And this is in indeed a
35:08challenge for working with this type of
35:10data,
35:11but overall, I think this is just an
35:14incredibly exciting time to be doing
35:17this kind of research,
35:19particularly in context of conflict
35:22where
35:23where traditional data sources are very
35:25hard to rely on.
35:26And so,
35:28this is an exciting moment because the
35:30constraint really used to be what we
35:32could measure, but now it's simply what
35:35we think to ask.
35:37So, let me pause there and turn it over
35:40to Mike Callen.
36:03>> Okay, I wanted to just
36:05Yeah, okay, it's visible full screen.
36:07Great. It's great to see you all again.
36:10Thanks for continuing to join along with
36:12the course.
36:14A lot of what I'll say will complement
36:16what Anjuli had just spoken about,
36:18but just sort of as a point of departure
36:21and re-emphasizing something that Anjuli
36:22had started with.
36:24If we think about sort of the goal of
36:25development economics research and
36:27development practice is one of the
36:28fundamental goals is sort of ending
36:30extreme poverty in the world. I think
36:32it's sort of inevitable that you'll have
36:34to encounter ideas about conflict,
36:36fragility, and so on just because so
36:38much of sort of the remaining extreme
36:40poverty in the world is still sort of
36:42concentrated
36:44in these settings.
36:45Um so, this is sort of the most recent
36:48forecast from this year about kind of
36:50where the world's poor extreme poor will
36:52live uh based on sort of the degree of
36:54fragility or conflict experience or the
36:57degree of fragility in a country. Um and
36:59I want to kind of stop and emphasize a
37:01few key points about this. Uh first, you
37:05know, the trend if anything is
37:07accelerating and I tend to go back to
37:09this figure sort of year after year as I
37:10speak about the topic and if anything,
37:12the sort of forecasted share um of the
37:14world's extreme poor uh that's going to
37:16live in in fragile or conflict-affected
37:19countries is just growing. You can also
37:20look at other definitions, not just the
37:22OECD definition. Um they break countries
37:25down into two categories. One category
37:28are sort of what they call high
37:30fragility contexts. These would be
37:31places like Nigeria, Pakistan, and so
37:34on. Uh and then they also talk about
37:37these extremely fragile contexts. Uh
37:39these would be places like Yemen, um
37:41Sudan, Congo, and so on. Uh so, I think
37:44another sort of important feature of
37:47what's happening in terms of the global
37:48distribution of poverty is that it's
37:49actually increasingly concentrating in
37:52these extremely fragile settings, places
37:54where sort of the fundamental nature of
37:55the state is sort of unclear um and the
37:58capacity is very low and so on. Um
38:01>> [clears throat]
38:02>> What does it mean for a state to be
38:04fragile? There are many definitions that
38:07are used across sort of academic
38:08research and practitioners.
38:10One sort of common thread you'll see,
38:12something that also came up in Tim
38:14Besley's lecture, um is the fact that
38:17there's sort of a lack of predictability
38:19about what the rules of the game will
38:20be. And that's sort of the definition
38:22that the OECD uses. I'm going to This
38:25theme is going to recur um kind of
38:27throughout the lecture, but I as I I
38:28want to kind of advance this as a key
38:31definition because I think one thing
38:32that's quite consequential if we start
38:35to think about what are the implications
38:36of fragility for economic and political
38:38development uh is this fact that the
38:40future is so uncertain. Development is
38:43sort of by its nature a dynamic process
38:45that requires investments in building
38:47the state, that require investments in
38:48the private sector.
38:50Um and if you're sort of very unsure
38:53what the future rules are going to be,
38:54this sort of creates potentially
38:55fundamental headwinds um to economic
38:58development and progress. Uh so, I think
39:01it's very important if you want to think
39:03about how to end poverty in the world,
39:05um it's also sort of not very well
39:07represented in the literature, the
39:09academic research literature, um
39:11if you measure it a variety of ways. So,
39:13one thing you can do that's sort of a
39:15proxy for future research on the topic
39:18is to look at sort of the share of
39:20registries in the American Economic
39:21Association online registry of
39:24randomized control trials. Um that tells
39:26us what are researchers actively
39:28researching at the moment at least using
39:30RCTs. Um a pretty small share are in
39:32these extreme extremely fragile
39:34countries. So, something like less than
39:36a percent of all of the entries in the
39:38registry uh are there. If I look instead
39:41retrospectively at research that's
39:43already been published, if I look at
39:44sort of the top five, it's only about
39:470.2%
39:49Now, if I even just sort of break down,
39:51you know, where are countries in the
39:52very top journals in in economics and
39:54political science appearing, again, it's
39:57a very very small share um that are in
39:59these sort of fragile and low-income
40:01states. So, at least if we're sort of
40:03thinking about allocating research
40:04towards places um where we'll get
40:06insight that'll help us think about
40:08reducing poverty in the world, there's
40:09sort of a a bit of an inherent
40:11distortion here. Um
40:14Sorry, let me keep moving. Um and so, I
40:17think, you know, to the extent that
40:18you're all thinking about potentially
40:20becoming producers of research, I think
40:22there are really three kind of very
40:23important questions
40:25linked to fragility and development. Uh
40:27the first is, you know, why what is it
40:29about fragility that that that impedes
40:32the process of economic development?
40:33I've sort of gestured towards one, which
40:35is that there's a lot of uncertainty, so
40:37investment, both public and private,
40:38becomes quite challenging.
40:40Um a second sort of separate question,
40:43something actually that Jill and I are
40:44currently working on in a few projects,
40:47uh is, you know,
40:48even if we can't fix the fragility per
40:50se, what can we do? What types of
40:53interventions are there that work even
40:55in spite of the fact that a place is
40:56very challenging. Uh we're currently
40:58doing work, for example, on education in
41:00Afghanistan. Um so, what can we do to
41:02kind of help address deprivation in
41:03spite of the fact that there's so much
41:05fragility. Uh and then finally, you
41:07know, how do we understand it? Uh if we
41:09can understand the causes, for example,
41:11the literature that Andriller reviewed,
41:13um does this start to point toward
41:14pathways out of fragility? Um and that's
41:17kind of where I'll go next is to provide
41:19a little bit of a sort of conceptual
41:20perspective on on, you know, how we
41:23should think about conflict.
41:24Um so, I'm going to start by just kind
41:26of going through, you know, in
41:29particular Fearon 1995. I'll follow the
41:31progression of that paper and try to
41:33provide sort of the classical account of
41:35how we think about why why people fight.
41:38Um it's a bit different than thinking
41:39about sort of the the economic drivers
41:42and more just kind of how would we
41:43theorize about it. Uh and then I'll talk
41:45through two papers of mine that are kind
41:47of quite recent papers. Um one is about
41:50sort of trying to address an absence of
41:52state capacity in the face of conflict
41:54and the role of uncertainty and
41:55fragility and sort of impeding that
41:57process. Uh and then another paper um
42:00that's about what are the institutional
42:02consequences, so the consequences for
42:03political development that come out of
42:05conflict.
42:07Okay. Um so, turning to kind of the the
42:10big theory, I think that this is sort of
42:11foundational these these ideas are a bit
42:14foundational for for thinking about
42:15conflict. So, I wanted to make sure that
42:17I went through them. Um so, one of the
42:19most important parts of thinking about
42:21conflict is to know that it in some
42:23fundamental sense it's inefficient. So,
42:25it fighting is costly, people die,
42:28things are destroyed, um economic
42:30progress is halted, and, you know, if
42:33the parties to a conflict, whether it's
42:35two states or whether it's a rebel group
42:37and a national government, could
42:39understand where the thing would end in
42:42terms of the division of territory, the
42:44division of political power, what the
42:46political system would look like, and so
42:47on. If they had sort of this
42:48omniscience, and they could perfectly
42:51commit to agreeing to kind of following
42:54that path throughout the process of
42:55conflict, then it's fundamentally
42:57inefficient. You're just burning
42:58resources to get to a place where you
43:01could in theory get to through some form
43:03of bargaining otherwise. And so ex-post
43:06conflict is always sort of rationally
43:08inefficient because there in theory
43:09should be some way of negotiating to
43:11that outcome without having to destroy
43:14lives and property in the process.
43:17Um, so the sort of the way Fearon
43:20approaches this is to say, "Okay, well,
43:21then why would anybody do this if it's
43:23ex-post inefficient? We need an
43:25explanation for it." Um, so there's some
43:28that really aren't sort of that well
43:30suited to thinking about from a
43:31theoretical perspective, which is just
43:33sort of different versions of people
43:35like conflict. So, it could be national
43:37leaders are irrational. Um, they just
43:40will do it. They don't particularly
43:42care. They're not thinking. Or that they
43:43don't bear the costs. So, the elites
43:45whose decision it is to go to war don't
43:47actually have to deal with the war.
43:48There's actually some evidence that
43:49supports that. For example, in the US,
43:52um, you know, having a draft-age son
43:54actually reduces the support for a
43:56pro-conscription bill. So, when families
43:57actually have, you know, the title is
43:59kin in the game, they're less likely uh
44:02to want to support conflict. But he sort
44:04of in Fearon's paper, he said, "Look,
44:06what we really want is we want to have
44:08something that sort of survive, can can
44:10be internally coherent from the
44:11perspective of period of theory, is
44:13something we could actually sit down and
44:15and model as a game theory model, um,
44:17and nonetheless leads to conflict in
44:19equilibrium." And these are what he
44:21calls rationalist explanations, which
44:23are sort of the third set. Um, so I want
44:26to quickly give an overview of those
44:28just so that you've had the
44:29introduction. Um, I think they're quite
44:30useful in thinking about conflict. And
44:32and my perspective on this is is that
44:34they they always sort of exist in the
44:36presence of conflict in some
44:37combination. Um, so one thing he points
44:39to is private information. Just that if
44:42you have two sides to a conflict, they
44:44have private information about their
44:46capabilities that the other side doesn't
44:47know. Um, you also have strong
44:50incentives to misrepresent how capable
44:52you are. Uh, and so if these, you know,
44:54one side may feel for example that the
44:56other side is weaker than they are in
44:57fact and think that they could just
44:58conquer them.
45:00This can lead to certain negotiation
45:01breakdown and conflict in equilibrium.
45:05>> [clears throat]
45:05>> Another is just that there are
45:06commitment problems. So for to avoid
45:09conflict entirely, you would need to be
45:12sort of completely committed to never
45:14attacking one another even when the
45:15other side is vulnerable.
45:17If you think about disarmament or
45:18de-escalation, that process necessarily
45:21means that unless it can be perfectly
45:23synchronized. If you're reducing your
45:25ability to fight your opponents, at some
45:28point you're relatively vulnerable. This
45:30makes you vulnerable and can also cause
45:32sort of negotiation to break down.
45:35The final bit is that there may be issue
45:37indivisibility. So there may be some
45:39things where you just are completely
45:41unwilling to negotiate whether it's sort
45:43of sacred territory or a sacred issue
45:47where two sides just fundamentally
45:49cannot agree and they there's no
45:51potential for compromise or bargaining.
45:54So they just have to fight about it no
45:55matter what. So sacred religious sites
45:58are examples of these things or if one
46:00group sort of fundamentally is unwilling
46:02to be sort of subjugated by another no
46:03matter the cost of avoiding that in some
46:06sense or unwilling at least to provide
46:08some concession, then these would be
46:09sort of issue indivisibilities that
46:11could cause conflict.
46:13So I think sort of all of these are
46:15present to some extent in in sort of all
46:18of the many conflicts we see in the
46:20world and Andrea showed you a time
46:22series that if anything conflicts are
46:24just proliferating over time, which is
46:26probably also part of why there's sort
46:28of more of the world's extreme poor are
46:30living in
46:31contexts that are subject to conflict.
46:34Let me turn now to a paper with a set of
46:37co-authors where we were sort of you
46:38know, living and working amid conflict
46:40trying to do something to see if we
46:42could make the state more capable in the
46:44process. So this project was sort of a a
46:49randomized control evaluation of one of
46:51the flagship projects of the Afghan
46:53state.
46:54The goal was to sort of build the form
46:57of foundational state capacity, so the
46:59ability to identify the employees who
47:01worked for the state and then the
47:02ability to pay those employees. Uh, and
47:05when we started working on this, which
47:06was around 2015, at that point the
47:09Afghan war was already 14 years old.
47:11It'd already been sort of uh, it's
47:13received pretty substantial
47:14international investments to try to
47:16build capacity in the state. And even
47:18these sort of very basic features of
47:20what you would think of as core capacity
47:22of the state to develop and implement
47:24policy, like knowing who works for you
47:26and and knowing how to pay them
47:28reliably, um, were absent. And so the
47:30idea here was to try to see whether or
47:32not sort of new technologies, biometric
47:34registration, um, and mobile money
47:36payments could overcome these
47:38challenges. So is it, you know, was
47:40there was the was the presence of these
47:41new potentially transformative
47:43technologies a way to advance state
47:45capacity even in the face of uh, the
47:48challenges that conflict provides. Um, I
47:51think where I got to by the end of it,
47:52which has kind of now been sort of a
47:54fairly big theme in my work, was uh, it
47:57actually was incredibly challenging to
47:59move it forward. We did kind of
48:00ultimately kind of get this specific
48:02project there, but there was a lot of
48:04concern that the state would eventually
48:06evaporate, which is ultimately what
48:07happened. Um, and the whole system was
48:10collapsed when the Taliban took over.
48:11And so the fact that you're thinking
48:13about this prospectively actually makes
48:15a sort of state building initiative like
48:16this one quite challenging.
48:18Um,
48:19and so that's sort of what's lurking in
48:21the background of this paper, um, is
48:23that the uncertainty about what the
48:26future rules of the game will be, how
48:28power will be decided and allocated, um,
48:31you know, what will kind of the things
48:33that are ultimately complementary to a
48:34reform that sits at the core of the
48:36government like this one, um, is a
48:39challenge. This is something that that
48:40Professor Besley had talked a bit about
48:42in his lecture. Um, when you're trying
48:44to build the state and advance it, you
48:46need to hire people, you need to set up
48:48protocols, you need to set up rules
48:50about how you elect leaders and so on.
48:52Uh, and all of those things are sort of
48:54investments that you make now. They're
48:56costly investments that you make today
48:58in order to have a more capable state in
49:00the future. Um and so if there's the
49:02potential for conflict to fundamentally
49:04overturn these things, it creates sort
49:06of a fundamental investment problem. Um
49:09capacity is an outcome of a bunch of
49:11cumulative costly investments of this
49:12sort. Uh so the uncertainty that's
49:15inherent in fragility may weaken such
49:17incentives. Um
49:19as a sort of descriptive matter, fragile
49:22states are, you know, typically have
49:24very low capacity and also have kind of
49:26quite high uncertainty and that that was
49:28certainly the case here. Um
49:31this kind of can create the potential
49:32for sort of reinforcing dynamics where
49:34the absence of capacity makes it very
49:36hard to know what the future holds and
49:37the fact that the future is quite
49:39unclear at the same time makes it
49:41challenging to want to invest in in
49:42building capacity.
49:44Um
49:45so we tried to do that and we tried to
49:47do this in sort of a specific domain,
49:49which was education. Uh we focused on
49:51education for two reasons. I think first
49:54is just because education's quite
49:55consequential for advancing the goal of
49:57economic development, um but second
49:59because this was a very large ministry.
50:02So it employed 80% of the non-defense
50:04public sector um and there wasn't sort
50:07of systematic identification of who were
50:09the teachers and principals and
50:10administrators who worked in the system.
50:13Um and so we wanted to see if we could
50:15sort of sort this out. Um there were
50:17major issues when we started. Uh so I'll
50:21get to those in just a second, but lots
50:23of employees, for example, were were
50:25reporting that they didn't receive their
50:26salaries. Partly this can be because
50:29there issues of sort of central
50:30approvals at the ministries in Kabul. Uh
50:32partly this is because it doesn't
50:34actually the money never actually
50:35physically was getting to the teachers.
50:37>> [snorts]
50:38>> The reform had three stages. Uh the
50:40first was to biometrically register the
50:43teachers. So the goal was to just kind
50:45of first create a roster um that was
50:48internally consistent and coherent that
50:50could be used to just kind of figure out
50:51who was actually working for the public
50:53sector. Um for this purpose, the
50:55government actually uh
50:58it
50:58it procured a mobile network operator to
51:01then collect biometric data uh and their
51:04payments, their salary payments were
51:05going to be linked to having sort of a
51:08biometric means of withdrawing your
51:10payment, which meant that you would be
51:11the one who was withdrawing your
51:12payment, not somebody else. So, it was
51:13sort of meant to be an anti-corruption
51:15measure, also means of making sure that
51:17there weren't sort of workers on the
51:19payroll who were there but not actually
51:21being paid. So, a goal to just kind of
51:23get rid of ghost workers.
51:25Um
51:26the second stage was the kind of done by
51:29the Ministry of Education. The goal was
51:31to take all these data and try to
51:33reconcile it and see, you know, who is
51:34it that we would want to remove from the
51:36payroll as a result of having gone
51:38through this exercise of biometrically
51:39registering them.
51:41Um and so, they kind of match this to
51:43their existing data. They set up an
51:45internal committee in the Ministry of
51:46Education and the goal was to see
51:48whether or not they wanted to remove
51:49anybody. Uh and then the final goal was
51:52just to try to make sure that in that
51:54the payments were on time
51:56instead of delayed, that they were there
51:58in full. So, that uh people were getting
52:00the full set of money to which they were
52:01entitled.
52:03And the goal here was to do this by
52:04transferring money digitally using
52:07mobile money.
52:08Mobile money agents then would visit the
52:10school monthly to to turn those payments
52:13into
52:14to turn those payments into cash.
52:17Uh to we ran this randomized control
52:20trial at pretty large scale. So, we did
52:21it across three provinces.
52:25We created
52:26these algorithmic zones, which were
52:28meant to be evenly sized zones that
52:30covered about you know, the similar
52:32number of schools.
52:34And we divided these 345 zones into
52:36three treatment arms. One where
52:38everything turns on as quickly as
52:40possible. So, everyone gets both
52:42biometrically registered and then paid
52:44by mobile money. One where people get
52:46biometrically registered but then we
52:48wait a while to pay them with mobile
52:49money. The goal here is to try to
52:52separate these two elements of this
52:54bundle treatment that involved both
52:55registration and payment via mobile
52:57money.
52:58And then finally, we had a control
52:59group.
53:01So, I'm just going to tell you the
53:02results quickly just so that I can get
53:05through everything I'd like to get
53:06through
53:07in time.
53:09So, there were major problems when we
53:11started.
53:12Attendance among teachers was only about
53:1441% in an independent audit that we
53:16conducted. 60% of the teachers reported
53:19experiencing salary delays. The median
53:21time of those delays was 3 and 1/2
53:23months. So, there were major, major
53:24issues with both identifying and paying
53:26teachers.
53:28The you know, the RCT
53:31spanned these two provinces, including
53:33substantial areas that were controlled
53:35by the Taliban. What did we see in the
53:37end? So, first, this really did improve
53:40the state's ability to have a single and
53:42consistent registry of the employees
53:44working for it.
53:46That was also partly to the adjudication
53:48process. We ran some independent audits
53:50of this, and it seems like it was good,
53:52though it wasn't entirely perfect. It
53:55did improve the payment process, in
53:57particular in places where the mobile
53:58money infrastructure was already fairly
54:00robust, which was mainly in cities.
54:02It did have downstream benefits as well.
54:05So, it improved teacher retention. It
54:07improved student learning outcomes,
54:10which is something that we independently
54:11independently assessed, and which really
54:14was sort of the fundamental goal of
54:15improving state capacity in this domain
54:18in the first place.
54:19It also improved employees' ability to
54:23to have a form of financial access, so
54:25they could use their mobile wallets as a
54:27means of saving, and also means of
54:29performing financial transfers on their
54:31own.
54:32Um
54:33The final bit, which I'm just going to
54:35mention, links back to the point I had
54:38made at the very beginning,
54:40which is that, you know, this because
54:42this reform took place over a
54:44substantial geography in Afghanistan,
54:46where there's quite a bit of variation
54:48in terms of whether the Taliban had
54:49control, quite a bit of variation in
54:51terms of people's beliefs about where
54:53the state was headed, and and genuine
54:55uncertainty about whether the Taliban
54:57was going to win the war or whether or
54:59not the Afghan US-supported Afghan state
55:01was going to win the war. Um the
55:03strongest predictor that we could find
55:05about whether or not employees that
55:07participated in this reform in the first
55:09place, um is whether or not a community
55:11believed that the that the Afghan state
55:14would win the war.
55:15So at least this study kind of provided
55:17suggestive evidence to me
55:19that this fact that fragility creates
55:21uncertainty about what the state is
55:23going to look like. In this case it was
55:24very real. The Taliban were very viable
55:26viable threat. They ultimately won and
55:29evacuated sort of the Afghan Republican
55:31government. That makes it very hard to
55:34execute a reform like this when you know
55:36that everything could collapse.
55:39So let me now just move forward to the
55:42next paper.
55:44This [clears throat] is about
55:46the consequences of conflicts for the
55:48subsequent development of institutions.
55:51This is part of a broader program of
55:53work in Nepal. Nepal is a fascinating
55:55scenario setting to study these
55:57questions because it just is emerging
56:00from a period of substantial civil war
56:03and now is kind of becoming a federal
56:06democracy with a set of consolidating
56:08institutions. The goal of this paper is
56:10to try to understand how the process of
56:13conflict itself sort of affects the
56:16subsequent development of the democratic
56:18state. Um
56:20and so you know the kind of in the
56:22background a few important ideas here
56:25are that most rebel governments, the
56:27Taliban, the Maoists in Nepal and many
56:30other subnational governments, the rebel
56:33movements themselves set up parallel
56:35states in the areas that they control.
56:37They administer justice, they collect
56:39taxes, they deliver services
56:42and both in sort of the military and
56:44civilian wings they're cultivating
56:46leaders. The Maoists were exceptional in
56:48this regard. They even held internally
56:50they held elections. They ran schools
56:52and they ran hospitals and so on. And so
56:55it's conceivable at least that this
56:57process of running a parallel state,
56:59especially as in Nepal where there was
57:01sort of a an ideology of trying to
57:03include historically excluded groups
57:06that motivated the conflict in the first
57:07place might actually sort of transform
57:10the state. Um
57:12So, let me quickly kind of provide a bit
57:14more context and then get through the
57:16results so that we have time for some
57:18questions. Um Nepal has transformed
57:22pretty dramatically since the mid-1990s.
57:25Uh so, in 1990, it was a a full a feudal
57:30caste-based monarchy, um had a caste
57:33system that involved taking many
57:34indigenous groups that really didn't
57:37have a history of sort of a caste before
57:39the monarchy and calling them middle
57:41caste groups. So, it's basically a means
57:43of imposing sort of social control and
57:45and sort of systematically excluding
57:47from political and economic power sort
57:49of a large part of the population. Um
57:52but, this is a pretty stable regime. The
57:54monarchy lasted for 240 years. Um and
57:57then the Maoists started a civil war and
57:59made pretty clear in a set of 20-point
58:01demand that they delivered to the
58:02government that they wanted empowerment
58:04and inclusion uh for these middle caste
58:06indigenous groups called the Janjatis.
58:09Um and also that they wanted a federal
58:11democracy. Um and so, they they weren't
58:13Maoists in the sense that they supported
58:14single-party democracy. They actually
58:16really wanted multi-party democracy. Um
58:18what this figure is showing is sort of
58:21Nepal's rank in terms of how stable it
58:24is as perceived by the World Bank in the
58:26international system. Um it's quite
58:28stable at the outset of the conflict in
58:301996. At the height of the conflict and
58:32the violence was at its worst, um it was
58:34one of the least stable countries in the
58:36world. Um there was about 10 years of
58:39negotiation around the constitution
58:42after the conflict, which basically grew
58:44out of the peace agreement. Um and that
58:46was finally promulgated in 2015. And
58:48since 2015, it's made pretty remarkable
58:52progress. Um there's been sort of a
58:54rapid rebound in stability, uh though,
58:56you know, not without some some recent
58:59conflagrations. And there's been a you
59:02know, a pretty dramatic increase even
59:03relatively in GDP per capita. Also, a
59:06pretty dramatic reduction uh in $2-a-day
59:08poverty. So, I think it's a very
59:10interesting example of trying to
59:12understand how institutions develop out
59:14of the ashes of conflict that can
59:15support economic growth, support greater
59:18political inclusion, and also ultimately
59:21have a role in reducing poverty.
59:24Um so the research design in this paper
59:27is basically going to be to leverage the
59:29presence of where the Maoists controlled
59:30territory during the Civil War. And at
59:33the height of the Civil War, they
59:34controlled about half of the country. Um
59:36so the first exercise is just going to
59:38be to take a survey that measured
59:40people's engagement with the new federal
59:42democracy and their knowledge of the new
59:44federal democratic constitution right
59:46after the conflict ended. So in in 2017,
59:502 years after the new constitution, uh
59:52and then to use a spatial geographic
59:55regression discontinuity uh to try to
59:57understand if the areas that had been
1:00:00controlled by the Maoists during the war
1:00:03affected people's engagement with the
1:00:04new political regime, in particular for
1:00:06these middle-caste Janjati groups that
1:00:08had been historically excluded, um and
1:00:11that provided the background the the
1:00:12backbone for the revolution.
1:00:15Um
1:00:16So we do find pretty clear impacts. This
1:00:18is just showing you the distance to
1:00:20these boundaries I showed you on the
1:00:22prior slides. These are the high-caste
1:00:24groups, Brahmin, Chhetri, and Newar.
1:00:26These were not the subject of sort of
1:00:28the activism of the Maoists, and we
1:00:30don't really see any impacts on their
1:00:33political behavior after the new federal
1:00:35democracy emerges. But for Janjati
1:00:37groups, they have a much deeper
1:00:38knowledge of how the constitution works,
1:00:40albeit from a lower baseline, which is
1:00:42sort of consistent with them being a
1:00:43socially excluded group. They know much
1:00:45more about what the political parties
1:00:48are doing. They themselves are much more
1:00:50active politically both in social
1:00:52organizations, political organizations,
1:00:55um and party politics. Uh and they also
1:00:57have more trust in political parties.
1:01:00Uh the next thing we look at is to see
1:01:02if the way that parties internally
1:01:04operate changes as well. So parties
1:01:09in most developing countries and many
1:01:11developing countries nominate candidates
1:01:13through internal party protocols where
1:01:15they basically party elites decide who
1:01:18gets to run for office and that kind of
1:01:20plays a major role in shaping the
1:01:21political class.
1:01:23Um, and we do see that sort of these
1:01:25historically excluded groups comprise a
1:01:28much larger share of the nominating
1:01:30committees um, in political parties
1:01:32after conflict in places where the
1:01:34Maoists had control. Um, so this sort of
1:01:37suggests that they're staffing in sort
1:01:38of a key party institution that
1:01:40basically determines uh, the political
1:01:43class um, with people from this
1:01:45historically excluded group in the
1:01:47places the Maoists had control that it's
1:01:49certainly at least consistent with the
1:01:50idea that they sort of generated a uh,
1:01:53cadre of leaders that could emerge and
1:01:55and work in party politics after um,
1:01:58after the conflict. Uh, the final bit is
1:02:01to actually look at political selection
1:02:03itself. Um, this is actually kind of a
1:02:06despite the figure appearing fairly
1:02:08modest, it took a is a pretty
1:02:09challenging exercise that involved sort
1:02:11of matching millions of records in the
1:02:13census um, to people who approached the
1:02:17party to try to become candidates,
1:02:18people who succeeded in becoming
1:02:20candidates and then who became
1:02:21representatives. I'm going to spend just
1:02:23a moment on the figure so that you
1:02:24understand it. What this is telling is
1:02:27basically the ratio of the probability
1:02:30that someone in the population who is
1:02:31from a middle class group, the Janjatis,
1:02:34the ratio of their probability of
1:02:35entering a certain political category
1:02:38relative to that for higher class
1:02:40groups. So if I look at this, what this
1:02:42is saying is that Janjatis enter the
1:02:46sort of in the Maoist party, Janjatis
1:02:48enter becoming an aspirant, someone who
1:02:50approaches the party to pursue political
1:02:52office
1:02:53with 60% of the probability of a higher
1:02:56class group. Um,
1:02:58and so these underlying probabilities
1:03:00are pretty infrequent. Like it's one in
1:03:0210,000 people in the population who
1:03:05aspires to be in politics at all. Um,
1:03:07but the ratios I think are quite
1:03:09informative. Um, and so the the main
1:03:12sort of the key takeaway from this is
1:03:14that first sort of the everywhere that
1:03:17there was a people's government
1:03:19operating the ratio of Janjatis entering
1:03:23these positions to the higher caste
1:03:24groups is higher. So, that suggests that
1:03:27sort of political
1:03:28um
1:03:29that sort of political selection is kind
1:03:31of more towards equality between between
1:03:34caste groups in places that the Maoists
1:03:36had governed during the war.
1:03:37Um and it's also sort of the case that
1:03:40they look more alike um between Maoist
1:03:43and non-Maoist parties in these same
1:03:45places uh that the party had controlled,
1:03:47which sort of suggests that there's a
1:03:49strategic reaction by the incumbent
1:03:51parties to the revolutionary party in
1:03:54terms of how they conduct political
1:03:55selection. So, that's a lot. I wish I
1:03:57could develop it in more detail, um but
1:04:00there are basically three mechanisms we
1:04:01talk about. I'm not by any means
1:04:03claiming these are exhaustive
1:04:04mechanisms, but first, you know, having
1:04:06rebels changes people's political
1:04:08consciousness, at least those who are
1:04:10working like the Maoists, and then
1:04:12afterwards in the democratic settlement
1:04:15uh parties change their operation both
1:04:16in terms of who they staff as elites in
1:04:18their parties, um and then how they
1:04:20select candidates as well. So, this is
1:04:22sort of showing three ways that, you
1:04:25know, the process of revolution itself
1:04:27um can change how democratic politics
1:04:29works um in the post-war settlement with
1:04:32I think implications for sort of
1:04:34subsequent political and and potentially
1:04:36economic development. Um
1:04:39Finally, we look at sort of state
1:04:41capacity in the new governments that are
1:04:43being built under the federal
1:04:44constitution. These are local
1:04:46governments. Um they have substantial
1:04:49delegated authority. Um they have
1:04:51authority for health, they have
1:04:52authority for roads, they have authority
1:04:53for education, they raise budgets, um
1:04:56and we have quite detailed and validated
1:04:59data on how much state capacity they
1:05:01have. Um and one thing we see pretty
1:05:03consistently and clearly is that all of
1:05:05the places that where the Maoists had
1:05:07controlled, sort of using the same
1:05:09discontinuity I'd shown you before, um
1:05:11also have higher state capacity in the
1:05:14post um in the post-democratic regime.
1:05:18Um
1:05:19Okay. So, I know I've gone a long. Let
1:05:22me just wrap up and then very happy to
1:05:23take questions. So four key takeaways
1:05:27from my part of the lecture. First, I
1:05:30think to understand how the extreme and
1:05:32extreme poverty you really do need to
1:05:34look at fragile states.
1:05:37A growing share of the world's extreme
1:05:39poor live in these settings. This in
1:05:41sort of the rate at which they are they
1:05:43are increasingly in these settings is
1:05:44increasing over the time partly because
1:05:46there's more conflict. Partly because
1:05:48poverty is is disappearing relatively
1:05:51from from non-fragile states.
1:05:54And yet this is kind of quite severely
1:05:56underrepresented in development
1:05:58economics research. So certainly
1:06:00something to consider as you're thinking
1:06:02yourselves about as becoming producers
1:06:04of research.
1:06:05Second, fragility undermines the
1:06:08function of institutions, political
1:06:10institutions in all sorts of ways.
1:06:13I guess I
1:06:14a theme of this lecture is that I think
1:06:16it's quite important that it undermines
1:06:17their predictability. It's very hard to
1:06:20think about building state capacity when
1:06:21there's sort of fundamental uncertainty
1:06:23about what the rules of the game will
1:06:24be.
1:06:26>> [clears throat]
1:06:26>> Two other final points. First, you know,
1:06:29the exercise showed that it was possible
1:06:32at least for a period to use technology
1:06:35to advance state capacity in these
1:06:37states. You know, we did eventually get
1:06:39the registry and payment system to work
1:06:41and it did prove effective in a
1:06:43randomized control evaluation. But it
1:06:45was quite hard. The political
1:06:46instability created substantial
1:06:48headwinds to the reform process.
1:06:51And finally, the you know, the fact that
1:06:53conflict happens and how it happens
1:06:55really does have sort of fundamental
1:06:57effects. It's kind of a moment where
1:06:59institutions are fundamentally
1:07:00determined. And so if you think of this
1:07:02as a bit of a critical juncture that
1:07:04shapes institutions, institutions in
1:07:06turn having a major role
1:07:08in determining the process of economic
1:07:10development.
1:07:11You know, there there are clearly
1:07:12dynamics happening during the conflict
1:07:14itself that affects the subsequent
1:07:16development of the state. So with that
1:07:19I'll stop and turn to questions.
1:07:26>> Brilliant. Thank you both so much.
1:07:28There's been a lot of questions that
1:07:30have rolled in. So, I'll I'll start with
1:07:33with a few here and then
1:07:35those that are panelists now, if you'd
1:07:37like to ask a question as well, just
1:07:38raise your hand and we'll kind of go in
1:07:40order.
1:07:41Um
1:07:42there were a couple of questions early
1:07:45on about kind of the idea of
1:07:50of conflict and the frequency of
1:07:52conflict from
1:07:54more wealthy states
1:07:56into more poor states. So,
1:08:00higher income yields to
1:08:02perhaps more conflict. How would you
1:08:04connect that to colonialism where higher
1:08:06income countries stole from lower income
1:08:08countries resulting in more conflict in
1:08:11those countries?
1:08:12And a related question of why do
1:08:15developed countries often create or
1:08:17support conflict in conflict affected or
1:08:20fragile resource-rich nations? How does
1:08:23political economy explain this?
1:08:30>> Um I can come in on the first one and
1:08:33then Mike, if you have thoughts on the
1:08:34second, maybe.
1:08:36So,
1:08:39the colonialism
1:08:41point is very interesting because
1:08:44it's taking
1:08:46you know, some of the ideas about the
1:08:47internal conflict or civil wars and
1:08:50porting it over to a interstate context
1:08:54and it's making the observation that,
1:08:56you know, if
1:08:58if the theory is
1:09:00higher income leads to greater
1:09:01predation, why is it the case that the
1:09:03richer country is going in and you know,
1:09:06extracting resources from the poorer
1:09:07country? And I think that's because in a
1:09:10when you're port over to the
1:09:13interstate context, you also have to
1:09:16consider differences in the
1:09:17technological capacity of the two
1:09:20states. The richer developed countries
1:09:22having greater military capacity and
1:09:25technological capacity to go in and
1:09:28extract resources from poor countries
1:09:31is also
1:09:33you know going to be a factor for which
1:09:35way the predation happens
1:09:38and I think it also highlights
1:09:41the point that when you just think about
1:09:45GDP and whether a country is rich and
1:09:48poor
1:09:49it obscures the fact that there are
1:09:52different sectors within an economy
1:09:54there's different locations that have
1:09:55different resources and a very poor
1:09:58country might nonetheless have
1:10:01a natural resource that makes it
1:10:04valuable to extract in a particular
1:10:06location and then whether or not the
1:10:09presence of that natural resource leads
1:10:11to
1:10:13good things or bad things is of course a
1:10:15enormous literature in political economy
1:10:17termed the potential resource curse
1:10:21where some evidence suggests that the
1:10:22presence of that resource can even lead
1:10:25potentially to lower growth perhaps
1:10:27because of this extractive response.
1:10:31>> Okay, so on the second question if you
1:10:33look at the time series that Andra had
1:10:35near the beginning of her lecture that
1:10:37sort of maps out the
1:10:40the composition of different types of
1:10:41conflicts the count of conflicts using
1:10:43the ECDP data over time
1:10:46it seems to be actually increasingly the
1:10:48case that there are these
1:10:48internationalized interest state
1:10:50conflicts and that's kind of beginning
1:10:52sort of in the mid 20 2009 2010 and
1:10:55thereafter.
1:10:57So I I think the broad reason that you
1:10:59see kind of maybe there's two reasons
1:11:01these are more questions of sort of
1:11:02geopolitics and conflict per se but I
1:11:04think first just developed countries
1:11:06perceive it to be in their interest to
1:11:09get involved in these what are otherwise
1:11:11interest state conflicts because they
1:11:12perceive either there to be some benefit
1:11:14or some ability to contain an
1:11:16externality. I mean I think Afghanistan
1:11:18is sort of the extreme case of that 9/11
1:11:20certainly precipitated the United States
1:11:22involvement but you can also see
1:11:24international involvement in Yemen and
1:11:26in the Middle East for example, around
1:11:28issues of resources, but also free
1:11:30navigation and so on. I'm so I think
1:11:32it's just in partly it's increasingly
1:11:34perceived by developed countries to be
1:11:37in their interest to be involved.
1:11:39I think probably a part of why you also
1:11:41see the proliferation is actually just a
1:11:43move from a a single uni- polar moment
1:11:46to a more multi-polar moment where it's
1:11:47kind of not quite clear who is going to
1:11:50be sort of the hegemon that determines
1:11:52the or that at least is the major player
1:11:54in these conflicts and some more more
1:11:56actors are also just getting involved.
1:11:57So, I think that that's probably another
1:11:59part of it.
1:12:00That's a bit speculation and really is
1:12:02sort of in the domain of geopolitics,
1:12:04but that that certainly is increasingly
1:12:06the case and if I had to guess the
1:12:08rationale, I think it's
1:12:09you know, that that developed countries
1:12:10perceive it to be in their interest.
1:12:15>> Thank you so much.
1:12:17Um I think those both answered really
1:12:19well.
1:12:20Uh I see Kier Quirin. Sorry, I'm not
1:12:23sure how to say your name. Uh do you
1:12:24want to ask your question?
1:12:27>> Yes, thank you so much. Um so, I was
1:12:30wondering where do you see the biggest
1:12:32open questions um
1:12:34in the political economy of conflict
1:12:35today? Because is it is it still
1:12:37understanding the con- the causes of
1:12:39conflict or um increasingly
1:12:42understanding how conflict transforms
1:12:44into political institutions and uh state
1:12:47capacity also um in the long run?
1:12:54>> The reason both answer the same thing,
1:12:56but so I I mean, I think that the the
1:12:58the the kind of the motivating point of
1:13:00departure for me is just that the it's
1:13:02so it's so closely correlated with where
1:13:05we extreme see extreme poverty in the
1:13:07world and that's sort of the lens that I
1:13:08take on thinking about what questions
1:13:11motivate me um in this space of sort of
1:13:14conflict and development. And so, I I
1:13:16think the you know, to the extent that
1:13:19that institutions and states are are
1:13:21sort of very important in kind of the
1:13:23process of promoting economic progress
1:13:26and development, I think trying to
1:13:27understand how conflict interacts with
1:13:31with political institutions seems quite
1:13:33important to me, which is sort of I I I
1:13:35hope in the two examples I provided for
1:13:37my own research, you can sort of see
1:13:38that where that's where I'm oriented,
1:13:39but I'm also quite interested in just
1:13:42what are what can we do anyway? Like
1:13:45what are things that that can be done to
1:13:47sort of address deprivation in these
1:13:49settings, even if they are very very
1:13:52challenging, cuz it's just the
1:13:53deprivation is there, it's acute, and I
1:13:55don't know that we should wait for it to
1:13:56be resolved to address it. Um
1:14:01>> Yeah, and from my angle, it's definitely
1:14:04not a either/or, because what
1:14:07um
1:14:08you know, what I see is that while
1:14:11poverty could play a role in spurring
1:14:13conflict, conflict when it occurs, then
1:14:16destroys infrastructure, schooling, has
1:14:17all these adverse impacts, which can
1:14:20then further fuel poverty, and so you
1:14:22get this kind of cyclicality, where many
1:14:24countries may be trapped in a cycle of
1:14:28conflict and poverty,
1:14:30uh and breaking that cycle might
1:14:31require, you know, either breaking the
1:14:33causes or or the consequences. Um and
1:14:37and so I think both streams of work are
1:14:39really important, and I think we're not
1:14:41done with examining the determinants of
1:14:44conflict, because
1:14:46um you know, I'll go back to where I
1:14:48started, which is there are more
1:14:50conflicts every year than the preceding
1:14:53year. Um there's many um
1:14:56uh areas
1:14:58uh for example, the rise of violent
1:15:00extremism in Africa, where I think we've
1:15:02made very little progress in actually
1:15:04understanding what are some of those
1:15:06factors at play, and that is why at the
1:15:08end I also emphasized the part about
1:15:10kind of religion measures, because I
1:15:12think we haven't made as much progress
1:15:15in that area. Um so I think, you know,
1:15:18continuing to tackle and understand,
1:15:22moving beyond economic shocks, and also
1:15:24thinking about what are other factors at
1:15:25play, um could really have high value
1:15:29added for what we uh need to learn.
1:15:32>> Right, Thanks.
1:15:35>> Thanks. Okay, so going back to the chat,
1:15:37there is a question about um the Maghrib
1:15:40prayer window uh and and and in the
1:15:43study that you shared on Jala.
1:15:45Um
1:15:46this individual said that many people
1:15:49choose not to make a call
1:15:52during the Maghrib prayer window
1:15:53considering that the recipient may be
1:15:56offering prayer.
1:15:57Um time propriety in making a call.
1:16:01How do you account for the variation in
1:16:03the underlying motivation of the
1:16:05behavior that you measure in the study?
1:16:10>> Um I can start, Mike, if you want to
1:16:12add. Um so the uh so it's a great point
1:16:15because it's kind of um the reflection
1:16:18problem like
1:16:20you are choosing to take an action, but
1:16:23you're potentially responding to what
1:16:25other people are doing. So how can you
1:16:26kind of, you know, attribute um the
1:16:29behavior uh from that angle? And um I
1:16:33think one uh
1:16:35you know, way that we look at that is
1:16:38what I showed you was just the call
1:16:40volume um and, you know, making that
1:16:43phone call is a
1:16:45well, it takes two to tango uh right? Uh
1:16:48but we have other measures
1:16:51that are um
1:16:54less subject to that pairing. Um for
1:16:57example, we have measures of text where
1:17:01it's not as intrusive to text somebody
1:17:03um as it is to make a phone call. And
1:17:07most uh relevant to this, we have
1:17:10measures
1:17:11of data usage on the phone. So that is
1:17:13if you get on the phone and you're on
1:17:15the internet, and that doesn't involve a
1:17:17second person. And so we see similarly
1:17:19sized dips associated with the Maghrib
1:17:22prayer window um
1:17:24with both of those more uh private
1:17:27measures, which to us suggests that this
1:17:29is not entirely driven um by uh the
1:17:34self-reflection uh or rather the
1:17:36reflection problem, and kind of
1:17:38consideration of other people, but also
1:17:40your your own behavior.
1:17:44>> Yeah, I don't have much to add other
1:17:46than to say that it was the topic of
1:17:49pretty considerable conversation as we
1:17:50were writing the paper because it's
1:17:52actually, I think, gets to a very deep
1:17:54and interesting point about sort of the
1:17:55relative contribution of intrinsic
1:17:58versus extrinsic considerations as
1:18:00you're thinking about your religious
1:18:02practice.
1:18:03Um
1:18:03so there's a a long and detailed
1:18:05discussion of that literature in our
1:18:07perspective on what we learn about it
1:18:09from this measure in the paper, so just
1:18:11encourage you to read that.
1:18:16>> Thanks to you both. Yeah, I thought it
1:18:18was a a
1:18:19a very important question.
1:18:21Um
1:18:23The Remind, would you like to ask your
1:18:24question?
1:18:25>> Yeah. Thank you. So, I was also
1:18:28wondering how income inequality can also
1:18:31lead to conflicts like even if a country
1:18:34is very rich, if there's huge disparity
1:18:37between people, wouldn't it also lead to
1:18:40dissatisfaction which might lead some of
1:18:43the people which are facing these
1:18:45problems rise up and maybe
1:18:48lead to some kind of conflict or
1:18:49something.
1:18:54>> Um let me say that I think that is
1:18:59actually a
1:19:02interesting area where we need more
1:19:04research. Um because
1:19:08we tend to find
1:19:11the existing work that has been done,
1:19:13that tends to use survey measures,
1:19:15surprisingly finds,
1:19:17um
1:19:19uh not very strong correlations with
1:19:21income inequality and conflict, but I
1:19:23think in part that really is a
1:19:24measurement challenge because
1:19:27measuring inequality credibly requires,
1:19:30um you know, surveyed measures that are
1:19:33inclusive of people at the top of the
1:19:35income distribution and bottom of the
1:19:36income distribution, and there's many
1:19:38challenges we have in surveys and
1:19:40actually capturing people, you know,
1:19:42throughout the income distribution
1:19:43especially at the top. Um and if you're
1:19:46going to measure something like how
1:19:48inequality across countries affects
1:19:51conflict, now you're talking about
1:19:52trying to find
1:19:54commonly measured surveys of individuals
1:19:59across countries
1:20:01and um and so I think that
1:20:04leaves us
1:20:06with currently little credible data for
1:20:09answering this question and I think we
1:20:10need better data to um be able to to do
1:20:13that.
1:20:14I think part of the reason the question
1:20:16is very interesting is because in fact
1:20:19if you have um
1:20:22uh the wealthy, they could supply
1:20:25capital and resources and if you have um
1:20:29uh
1:20:29the poor, they might be uh willing to
1:20:32fight, right? And so
1:20:35potentially um
1:20:37it could be the case that you actually
1:20:41need both elements on one side in order
1:20:45to fight. So, for example, um if you
1:20:50have a ethnic group, then you might have
1:20:52the wealthy and the poor within the same
1:20:54group. So, that might lead to a
1:20:55prediction that maybe group divisions
1:20:58along ethnic lines are more likely to
1:21:00produce conflict than uh you know, group
1:21:03divisions along other lines. But, it's a
1:21:05theoretically interesting question and
1:21:07an empirically challenging one.
1:21:10>> Thank you.
1:21:12>> I mean, just quickly on that, I also
1:21:14think that it sort of links a bit to the
1:21:16literature on social movements as well
1:21:17where it's just an absence of, you know,
1:21:20the the institutional means of gaining
1:21:23voice or economic power, then you have
1:21:25need to you need sort of resort resort
1:21:27to other means of doing so. You could
1:21:29kind of think of conflict as being the
1:21:31violent competition for power. Um and so
1:21:34if they they one plausible motivation,
1:21:36certainly the one that was very present
1:21:37in the mouse conflict I was talking
1:21:39about, the leaders are very clear, the
1:21:41participants are very clear. That was
1:21:43just the absence of sort of social
1:21:44mobility for these caste groups.
1:21:50>> Thanks.
1:21:52Um
1:21:53one
1:21:54There's another question in the chat uh
1:21:56going back to the meta-analysis.
1:21:59Um
1:22:00They ask whether there are studies that
1:22:03include narcotics in their agricultural
1:22:05basis such as opium, cannabis, etc. as
1:22:09part of the meta-analysis to study the
1:22:11impact of commodity price shocks on
1:22:13conflict. And given that fluctuations in
1:22:16the price of these criminalized
1:22:17commodities can significantly affect the
1:22:19livelihoods and of of lo- of locals in
1:22:22financing of armed groups,
1:22:24they are wondering if their inclusion
1:22:26would provide important insights into
1:22:29conflict dynamics.
1:22:31>> Great question.
1:22:33Um
1:22:33There is actually work by Angrist and
1:22:37Kugler
1:22:39that looks at coca price shocks and
1:22:41conflict in Colombia. Um and and the
1:22:45reason that I think it's such an
1:22:46interesting question is because you're
1:22:47right that uh
1:22:48conflict is um
1:22:50uh rather the uh
1:22:52illicit sectors like coca are different
1:22:55because not only is there the labor
1:22:58intensity aspect to it, but
1:23:03coca as an illicit commodity is one
1:23:06where violence is actually required for
1:23:08contract enforcement. So, it may be
1:23:11labor intensive, but then there's this
1:23:13other element of violence being required
1:23:16for contract enforcement, which pushes
1:23:18against the kind of opportunity cost
1:23:21effect um of that would predict that a
1:23:24rise in the price should actually, you
1:23:27know, reduce conflict in coca areas. So,
1:23:29in fact, what Angrist and Kugler find is
1:23:32a rise in the price of coca leads to
1:23:34more conflict in
1:23:36coca areas of Colombia, which is
1:23:38consistent with the idea
1:23:40that uh uh that this
1:23:44illicit aspect of the contract
1:23:46enforcement actually dominates over the
1:23:49opportunity cost effect in that case.
1:23:51So, yes, I think we should think about
1:23:53illicit commodities um as having other
1:23:56forces at play to uh think about the
1:23:59conceptual mechanisms differently and
1:24:01different things can predominate.
1:24:07>> Thanks so much.
1:24:09Um Catherine, would you like to ask
1:24:10yours?
1:24:11>> Yeah, thank you. Uh thank you for the
1:24:14presentation. I may be off a bit, just
1:24:16want to understand what the role of um
1:24:19political consciousness awakening say
1:24:21within a younger generation, maybe the
1:24:23Gen Zs and the early Gen
1:24:25uh Gen Alpha's
1:24:26has the
1:24:27has potential in probably fueling
1:24:29already existing conflict or maybe even
1:24:31starting it due to I mean, we're seeing
1:24:35the rise in disillusionment with
1:24:37democracy. So, how how do those things
1:24:39relate to what's political economy
1:24:41around that? Just
1:24:45>> I mean, so I think that the there's kind
1:24:47of a couple elements to it. I I mean, I
1:24:49think that the the disillusionment with
1:24:51the bunch of the features of the current
1:24:52system is clearly what's at play and
1:24:54driving sort of the Gen Z and Gen Alpha
1:24:56protests. Um so, I think you can think
1:24:58of it as sort of a a social movement in
1:25:00that sense where, you know, if the
1:25:02existing institutions can't accommodate
1:25:05the people's demands, then they're going
1:25:06to take other means. Um and I think
1:25:09that's exactly what you're seeing um in
1:25:11in Gen Z itself. Uh in the case in
1:25:13Nepal, I thought it was particularly
1:25:15interesting, which I guess was sort of
1:25:16the one of the first sort of Gen Z
1:25:18conflagrations because it was all about
1:25:21sort of just taking specific features of
1:25:24the environment and making them clearer
1:25:26and more certain. So, in particular,
1:25:27they just wanted the kind of corrupt
1:25:30leaders out. Uh they didn't want to tear
1:25:32up the federal constitution and start
1:25:34from the very beginning. They just
1:25:35wanted sort of term limits on what they
1:25:37perceived to be corrupt executives. And
1:25:38actually, it was in that regard it was
1:25:40quite successful. Um so, there's sort of
1:25:42a youth line party that won 2/3 majority
1:25:44in the most recent elections. They
1:25:45happened on time. They're mutually
1:25:47respected um by everybody involved
1:25:50including the elites who lost power. Um
1:25:52so in a sense I think it's just a means
1:25:53of kind of getting a rebalancing of
1:25:55power
1:25:56that's that that lives outside of
1:25:58constitutions or institutions. [snorts]
1:26:02>> Thank you. But say in like institutions
1:26:05that I mean that weren't maybe as
1:26:06successful as Nepal and had and are now
1:26:08maybe restrategizing, what is that?
1:26:11We're seeing a lot of that happening in
1:26:12some countries at least in my country
1:26:14Nigeria, there's strategizing especially
1:26:16from this younger generation. I'm just
1:26:18trying to understand what that
1:26:19relationship might be especially and how
1:26:21contentious it might be going forward
1:26:23and what that means.
1:26:25>> I mean it's a it's a challenging thing
1:26:27to forecast but I think it really comes
1:26:28down to how fundamental is the
1:26:30reimagining of institutions that people
1:26:32will be satisfied with, right? And I
1:26:34think that like because Nepal had
1:26:35already seen a fairly fundamentally
1:26:37reimagining towards very local federal
1:26:39democracy, the Gen Z movement there was
1:26:42very clear in their discourse that it
1:26:43was just about advancing that objective.
1:26:46So it wasn't about fundamentally
1:26:47reimagining. I don't know if that's so
1:26:49in the in Nigeria but you know it that
1:26:52it certainly can lead to much deeper and
1:26:55and more substantial conflict if people
1:26:58I mean in some sense you know the same
1:26:59dynamic was present in Afghanistan as
1:27:01well actually. There was no voice in
1:27:03institution for the people who supported
1:27:05the Taliban which was a large share of
1:27:07the Pashtuns group. Large share of the
1:27:08Pashtun group could not be represented
1:27:11formally in politics and so they had no
1:27:12choice other than violent rebellion for
1:27:14many years. Um so it can lead to a
1:27:17system situation of pretty substantial
1:27:19instability if
1:27:20if people just aren't satisfied with
1:27:23with the political equilibrium that
1:27:25they're in.
1:27:27>> Thanks everyone. Um I think we have time
1:27:30for one last question. Um maybe we'll
1:27:32we'll we'll look forward towards the the
1:27:34changes in machine learning and and AI.
1:27:38Um
1:27:38I the
1:27:40Professor Dube, you mentioned a few
1:27:42methods for measuring conflict both
1:27:44spatially and otherwise. This question
1:27:47is
1:27:48in regions where granular data is scarce
1:27:51due to limited access, how does one go
1:27:54about building a novel granular data
1:27:57set? How can you use some of these tools
1:28:00that you mentioned? Um
1:28:04they say, for example, the ACLED
1:28:06provides excellent aggregated data on
1:28:08conflict events taking place.
1:28:10If I want to measure the economic impact
1:28:12of a sustained conflict in the region,
1:28:14what variables or toolkit would you
1:28:16suggest to construct a novel data set?
1:28:22>> All right. So, I think kind of broadly
1:28:24speaking
1:28:27what some of these new data sources
1:28:30point to is how
1:28:34AI is really changing the ways that we
1:28:35can learn in the sense that you have
1:28:38images that you can look at. So, a deep
1:28:41learning model can see things from um
1:28:45satellite images that might be good
1:28:47measures of um economic activity. Um
1:28:51for example, you know, the use of like
1:28:54um
1:28:55there's some uh work by uh Suri and
1:28:58co-authors
1:28:59um that suggests tin roofs are um
1:29:03uh uh
1:29:04the presence of tin roofs are indication
1:29:06of like the housing quality um and that
1:29:09could be a measure of economic activity.
1:29:11Um
1:29:12those um
1:29:14NO2 emissions that I mentioned are um
1:29:18very interesting because those are
1:29:20available almost real-time and those are
1:29:23measures of economic activity. So, if
1:29:25you want to look at the impact of
1:29:27conflict on those types of outcomes and
1:29:30um you know, night lights of course is a
1:29:32traditionally used measure uh which is
1:29:35great, but if you want to pivot away
1:29:37from night lights to some um additional
1:29:40aspects of um socio-economic outcomes um
1:29:44uh that is a great resource.
1:29:46Finally, um you know, our um
1:29:49uh co-author and colleague Josh
1:29:51Blumenstock has developed uh, measures
1:29:53for using underlying CDR data and what
1:29:56is available in phone signatures,
1:29:59also to predict
1:30:02poverty status and economic outcomes.
1:30:05So, those types of data sources could
1:30:07also be brought to bear
1:30:09in terms of being able to
1:30:12measure these
1:30:14these types of outcomes.
1:30:15But, I guess, you know, stepping back to
1:30:17the broader point about
1:30:19conflict research, these same tools can
1:30:22also be incredibly useful for generating
1:30:24measures of conflict itself where
1:30:27conflict data is not available.
1:30:29So, again turning to images, satellite
1:30:32images can show you destruction of
1:30:34buildings on the ground.
1:30:36Turning to text, you know, a lot of the
1:30:38the the conflict data that we have
1:30:40traditionally used, including the work
1:30:42that you know, we did in Colombia using
1:30:45SEDAC data or what ACLED uses or what
1:30:48UCDP uh, data uses are actually based on
1:30:51news reports. And now, language models
1:30:54can essentially do this extraction at
1:30:56scale, right? And there are even models
1:30:58that are being trained specifically on
1:31:00conflict text like the conflict bird is
1:31:03one such data source that does this and
1:31:05it does this now when
1:31:07Arabic language versions as well, so
1:31:09that you can actually get coverage
1:31:12in in
1:31:13beyond the English press to be able to
1:31:17you know, track these types of outcomes
1:31:19precisely because that type of source is
1:31:23going to be
1:31:24available where other data sources
1:31:29might not be able to give you a direct
1:31:31measure of conflict. So, now people are
1:31:33using these large language models
1:31:35combined with web search to really build
1:31:37these event data sets even going back
1:31:39historically while tracking in an
1:31:40ongoing manner what is happening. So, I
1:31:43think those are some data sources to
1:31:45look to to think about doing this type
1:31:46of work.
1:31:51>> Thank Thank you. Um, I'm afraid there
1:31:54are 80 more questions that we don't have
1:31:58time for and there there always will be,
1:31:59but thank you so much for your thoughts
1:32:03and your questions and your insights.
1:32:05Um,
1:32:07and especially thank you to our
1:32:09professors who who joined with us today.
1:32:12Um, we will be here next week for the
1:32:17the lecture on crime and organized
1:32:19crime. I know there were a few questions
1:32:20around that today and we'll go in deeper
1:32:23on that and just to note that this
1:32:26series is based on the time of the UK
1:32:30which just went through daylight savings
1:32:32time. So,
1:32:34if your country does not do daylight
1:32:35savings time, this is the new time. If
1:32:38yours does like the US does this Sunday,
1:32:41then I'll go back to the old time. So,
1:32:43just look wherever London is, that's
1:32:45what the time it would be and I
1:32:46apologize for this uh
1:32:49irregularity this week, but hopefully
1:32:51we'll have it um
1:32:52smoothie line again soon.
1:32:56Thanks everyone.
1:32:56>> Thanks very much. Yeah, thanks very much
1:32:58everybody.
1:32:59>> Thank you.

About this transcript

This page contains the full transcript of BREAD-IGC virtual PhD-level course 2026 | Lecture 8 Conflict and Fragility by International Growth Centre, generated from the public captions YouTube serves with the video. The transcript has 14,708 words across 2,439 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.