YT YouTube2Text

YouTube2Text → Transcripts → BREAD-IGC virtual PhD-level course 2026 | Lecture 6 Bureaucracy and Implementation

BREAD-IGC virtual PhD-level course 2026 | Lecture 6 Bureaucracy and Implementation — Transcript

by International Growth Centre · 15,979 words · 2,347 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Thank you.
0:02>> Okay. Well, welcome everyone. Uh this is
0:04week six in our political economy
0:08lecture series put on by Brad IGC. Today
0:11is bureaucracy and implementation and we
0:14want to give a special welcome to Adnan
0:16Khan and Austin Quad uh for being our
0:19guest lectures today.
0:21Um, as many of you probably know,
0:24Professor Adnan Khan was the chief
0:26economist to FCDO uh in the UK from 2021
0:31to 2025 and previous to his current role
0:34as professor of public policy at LSE. He
0:37actually served as the research and
0:38policy director at the IGC. Um, and he
0:42taught public policy at the Harvard
0:43Kennedy School.
0:45Aim Kuaja is the co-founder of the
0:48center for economic research in
0:50Pakistan. Uh a lead academic for IGC
0:53Pakistan and the director of the Harvard
0:55center for international development.
0:57He's the current currently professor of
0:59international violence and development
1:01at Harvard Kennedy School. So um very
1:04very warm welcome to you both. Thank you
1:07so much for making time to be here. Um
1:10I'll just say two notes of housekeeping.
1:13If anyone would like to be a live
1:16panelist to interact um with Anon and
1:19Awesome, please raise your hands and
1:21I'll randomly select 10 to 15. As you
1:23know, you can ask Q&As's live uh at the
1:26end. If you do do this, please have your
1:28camera on and your mic muted. And the
1:32second thing is I'll put a link in the
1:33chat, but all the past lectures, all the
1:35past slides record and recordings are on
1:38the IGC website. If you have if you
1:40missed any, I know we got a lot of
1:42common questions there.
1:44Without further ado, I will turn it over
1:46to Adnan and um wish you all the best.
1:51Looking forward to
1:54Thank you, Carb. Hello everyone. Uh
1:56really great to to have this session
1:58with you and thank you for joining. Uh
2:01as a former bureaucrat, it gladdens my
2:03heart to see so many people interested
2:05in bureaucracy. Uh so I and aim will uh
2:09tag team and we'll switch between
2:11ourselves but let me start uh despite my
2:16enthusiasm or our enthusiasm for
2:17bureaucracy. Historically there have
2:19been different views on bureaucracy.
2:21While Bever thought this was the most
2:23efficient and rational way of organizing
2:25societies. Others disagreed. As Juan
2:29Meis put it nobody doubts that
2:30bureaucracy is thoroughly bad and that
2:33it shouldn't exist in a perfect world.
2:35However, like in today's session, we are
2:37we have I think uh in the research on
2:39bureaucracy, we have moved past these
2:42ideological perspectives and the
2:45question that uh the literature and
2:48especially in this session we are trying
2:49to address is how can we make
2:51bureaucracy more effective and that's
2:54the question we'll tackle uh in today's
2:56session and the readings that we have uh
3:00also deal with some aspects of those.
3:02Let me start with the one that uh which
3:04is the
3:06uh one of mine like a bureaucracy and
3:09development which is a review paper on
3:12the state of the literature on
3:13bureaucracy
3:14uh both on personnel economics system p
3:17perspective but also where the gaps in
3:19the literature are. We start with a
3:22question of uh does bureaucracy matter
3:25and uh we gathered some data and others
3:28and as you can see in this graph what we
3:30establish is that increases in
3:33bureaucracy scores bureaucracy
3:35bureaucratic effectiveness scores are
3:38correlated associated with higher levels
3:40of economic outcomes in this case GDP
3:44per capita and yes this is a correlation
3:48uh but correlations are also meaningful
3:50in this sense uh they do reveal uh the
3:53fact that bureaucracy does matter. Our
3:55standard framework for thinking about
3:57how bureaucracy matters is the principal
4:00agent framework
4:02uh which is the problem of one uh the
4:06principle one side trying to motivate
4:09the other side the agent uh it is the
4:12cornerstone for thinking about incentive
4:15problems and in this session we're
4:17talking about different layers of this
4:19citizens to politicians is one layer of
4:21this principal agent problem and then
4:23politicians to bureaucroates need
4:25another layer.
4:27However, like I think we have moved
4:29beyond this simplistic
4:31uh or like this view of using a one
4:36layer principal agent framework. And
4:39with this let me hand over to my
4:41colleague Aim who will talk about u
4:44moving beyond the principal agent
4:45framework and how we currently think
4:47about this and also then share a few
4:50slides a few studies. I'll come back
4:53then at the afterm and then we'll have a
4:56question answer. Over to you Aim.
4:59>> Great. Thanks. Um
5:03>> you guys give me one second. I'm just
5:05going to share my screen as well.
5:44D, have you given me share access?
5:50>> Yeah, I have stopped sharing. So, should
5:52be okay for you to share.
5:56>> It doesn't seem to be working. Give me
5:59one second, guys.
6:13>> Yeah, now we can see your screen. So,
6:15I'll be good.
6:18>> Uh, you can Sorry, guys. We're just
6:34Okay, they should work now.
6:36>> Yeah, we can.
6:38>> Perfect. Um so I'm going to just um move
6:42on from where Adnan was. Um and you know
6:45one thing as Adan said with the
6:47principal agent framework and what this
6:49is one of the ideas we're trying to sort
6:50of share over here which is the simple
6:53principal agent framework actually as
6:55much as it's been used in economics in
6:57fact is not a good description of what's
6:59happening. Um, and one of the points
7:01we're trying to make is when you do sort
7:03of studies like these, um, you want to
7:05have a much richer sense of what the
7:07framework is. And so what we're going to
7:09show you now is is a is a richer
7:12framework u which gets at some of these
7:15um uh ideas uh uh better.
7:28So the way to think about these ideas
7:30now is here's a sort of wider framework
7:33and in this framework think of what's
7:34happened instead of having a typical
7:36principle and an agent we realize that
7:38this space is kind of much more
7:40complicated right and anytime you think
7:43about an intervention you have to look
7:45at the systemic response or the system
7:47level response to that. So in the
7:49framework that Adan had showed before
7:51what we were really thinking of is look
7:53there is a there is a bureaucrat a
7:56senior bureaucrat called that a gov
7:58department or a senior policy maker
8:00there's a tax collector uh but then
8:02those tax collectors are potentially
8:04interfacing with taxpayers they're
8:06interfacing with politicians they're
8:08interfacing with service providers and
8:10so it's quite critical to think about
8:12all of these players and how they
8:13interact with each other and in reality
8:15what's beginning to happen in all of
8:17these things is there is a sense in
8:20which all of these players are linked
8:22with in relationships to each other.
8:24Right? So just to give you an example, a
8:26tax collector is going to a taxpayer and
8:29trying to raise taxes. So that arrow
8:31between the tax collector and taxpayer
8:33shows that relationship. Uh a taxpayer
8:36may have a relationship with a
8:37politician and you can see the downward
8:39arrow there because tax ultimately is a
8:41political subject. Similarly, a tax
8:44collector may be is embedded in a tax
8:46department and so tax collector is
8:49interfacing with mid-tier and senior
8:51tier bureaucrats. All of this by the way
8:53is under the context of service
8:55provider. So remember what's happening
8:56is you pay taxes so that you get
8:59services delivered. This is the kind of
9:01social compact with the state. And so
9:03you can think of all of this. Now this
9:05is still not the full complication. I'm
9:07just showing you as a way of saying this
9:10space is rich. So whenever you do
9:12research studies you want to start with
9:13a more complicated space primarily
9:16because your understanding of what
9:17happens how the uh policy change you're
9:20doing will really vary on what the based
9:23on what the relationship between these
9:25different actors is and what Alana are
9:26going to do is uh now demonstrate four
9:29different studies these are papers that
9:31we have shared with you u we're not
9:34going to go through the technical
9:35details of the paper or we're happy to
9:36do that in Q&A but we want to represent
9:38what it means for a study to to explore
9:42different spaces in this kind of um
9:45diagram if you will. Um so uh the next
9:48figure shows uh all of these X's
9:50represent what we would say as a
9:52friction. A friction is something which
9:54is blocking that exchange from
9:55happening. So for instance, if a tax
9:58collector isn't incentivized properly by
10:00the government department or the senior
10:01policy maker, a tax collector may not
10:04collect tax uh as they should, right? So
10:06that's a friction um between the tax
10:10collector and the government department.
10:11So the Red Cross is saying something
10:12isn't working over here. Now I've
10:14obviously shown this picture that
10:16everything is a friction and actually
10:17that may be a reality in the settings we
10:19live in. But what we're going to do is
10:21we're going to start one by one doing
10:23studies which we try and take one of
10:26these X's out and make it like a check
10:28mark. Try and take that friction away.
10:30Um so in this context we're going to ask
10:33two really big questions right and we're
10:35going to show you studies the first two
10:37studies are going to answer question one
10:39uh the second two studies which are none
10:41we'll talk about are going to answer
10:42question two right so the first question
10:44is how do you incentivize a bureaucrat
10:46in our case we're going to talk about a
10:48tax collector how do you incentivize a
10:50tax collector to do the job they're
10:52supposed to do which is raise taxes
10:54without creating too much disruption I
10:56want to say that carefully it's not just
10:57about raising taxes it's raising taxes
10:59in a way that is acceptable to a citizen
11:02otherwise all sorts of um um um protests
11:06will happen right um and the second
11:09question is and this is a deeper
11:10question which is it's not just once you
11:12realize that tax collectors um are
11:15working in a certain area there's a
11:17question of how do you allocate a tax
11:19collector to a certain area what makes
11:21sense right because you have this choice
11:23you can put people in different places u
11:25you can also change the relationship
11:27between the senior bureaucrat and uh
11:29where individuals are. So Dan will talk
11:30about both of those questions. So let me
11:32go one by one. So these are different
11:34studies that we're talking about. So the
11:36first study is introducing performance
11:39pay. Now performance pay is what you
11:41would think the private sector pays
11:43which is based on your performance um I
11:47give you uh a financial reward right? So
11:49that's a standard performance pay
11:51contract. And here the challenge is
11:53between three different players. So the
11:56challenge is the senior policy maker the
11:59the the little cross you see on the top
12:02right hand corner the senior policy
12:04maker is trying to create an incentive
12:08for the tax collector in this case the
12:10principal in the standard framework of
12:12NAN gives the principal in this case is
12:14the senior policy maker they're trying
12:16to achieve something the agent who
12:18they're working with is the tax
12:19collector okay so that's the standard
12:21principal agent framework but remember
12:23this framework if you incentivize as a
12:26tax collector and this is where a a
12:28partial analysis would miss this whole
12:29thing. It's critical how the tax
12:32collector interface with the citizen
12:34then and there's a there's a missing
12:36there's a broken link potentially there
12:37as well which is the taxpayer is saying
12:40why should I pay taxes to you the tax
12:42collector is trying to develop a
12:44strategy to try and raise taxes right so
12:47so think of this as a game between the
12:49principal and the agent government and
12:51tax collector and then you could imagine
12:54the tax collector becomes the principal
12:56now and the taxpayer is the agent which
13:00the tax collector is deciding a strategy
13:02on how to collect taxes which leads to a
13:06taxpayers's um behavioral change right
13:08and that behavior could be I will pay
13:10your taxes I refuse to pay your taxes or
13:13I go back and complain to the local
13:15politician that I'm having to pay too
13:16many taxes that's the decisions that you
13:18get to make okay so this is the setting
13:21now what did we do so this is a study in
13:23the study what we did was we essentially
13:26decided to give performance incentives
13:29two tax collector. So, we're intervening
13:31in that middle agent that you saw on the
13:33picture. And what we're going to do is
13:35we're going to randomly allocate. This
13:37is a randomized control trial. I assume
13:39you guys are familiar with it. If you're
13:40not, then you know there's some great
13:42stuff by JPEL on what an RCT is. You
13:44should definitely uh look at that. But
13:46it's a simple way to do causal
13:48estimation, right? And the idea in an
13:49RCT is we randomly select a group which
13:52is a treatment group, give it the
13:54program or policy. there's a control
13:55group which is a comparison group and we
13:57compare these two groups over time to
13:58see the impact of a policy. So we
14:00conduct an RCT and in the RCT we take
14:04half of all of the tax areas. So tax
14:06unit is a neighborhood um and we put
14:09them into one of four conditions. Either
14:12we make them a control so nothing is
14:14happening business as usual or we
14:16introduce three different styles of
14:18performance pay. So one is pure money.
14:22If you raise more money as taxes, we
14:24will give you a percentage roughly 20 to
14:2640% of the money you raised. Think of
14:28this as a pure revenue scheme. Okay?
14:31Then there's a revenue plus scheme where
14:33you're worried again, you're worried
14:34that if you have people incentivized to
14:36collect money, they may go overboard,
14:38right? They may be harsh to a tax to a
14:41taxpayer. Uh and so you want to make
14:43sure that you don't upset the taxpayer
14:46too much. And there's also something
14:48else, the assessment accuracy. So, if
14:51you don't know your own tax liability,
14:53what a tax collector could do is overt
14:55tax you, right? Because they're trying
14:57to raise money. You don't know that you
14:59owe um uh $10,000 or make it easier,
15:03$1,000
15:04uh or in our case, a lak rupees. Um you
15:08don't know that you owe uh uh uh that.
15:11And the tax credit come in and says to
15:13you, actually, you don't own a lak
15:15rupees. You should be giving us um uh
15:17two lakh rupees. um a lakh is 100,000,
15:20two lakh is 200,000. And so it looks
15:22like you're making a um a a a fair
15:25payment, but in fact you're being overt
15:27taxed. So what we did in this in the
15:28revenue plus scheme is we went object
15:31the actual value of the proh property
15:34tax and what you were being told you
15:36should be paying and if there's any
15:37inaccuracy there, you're under taxing or
15:39overt taxing, you'd be penalized for it.
15:42And the last one, this flexible bonus is
15:44what you imagine Wall Street might do or
15:46many companies do, which is there's no
15:48formula per se, but the boss decides you
15:51did a good job and the boss gives you a
15:53bonus. Right? So you go from very
15:55mechanical, money- based only to very
15:58subjective. And we wanted to see what
15:59would happen uh in these three things.
16:01And our outcomes was how much money you
16:03collected and whether the citizen was
16:05deeply dissatisfied because you
16:07collected money. And remember this is
16:08very important for a politician. What
16:11they're looking for is what is the cost
16:14in terms of votes lost for money raised.
16:17That's the game they're playing, right?
16:18Because they would raise one more taxes,
16:20but they can't have a protest happening
16:22or a revolution happening. In fact, if
16:24you want to study the history of taxes,
16:26one of the first set of people to be to
16:28be to be um targeted in the French
16:31Revolution was tax collectors. Actually,
16:33they were so annoying for citizens that
16:35they went after them. So, you can
16:36imagine that this could be a issue. In
16:39fact, there's a real interesting history
16:41in the US. The the earlier tax
16:43collectors who were incentivized for
16:44money were called tax ferits. A ferit is
16:47an animal if you don't know and it kind
16:48of goes everywhere and it collects
16:50stuff. And so these guys were considered
16:52to be feds, right? So, so we look at
16:54these three schemes and we're going to
16:56see what happens. And it turns out
16:57incentives work. When you pay,
16:59unsurprisingly, when you give tax
17:01collectors money, they actually raise
17:03taxes by about 10 percentage points. In
17:05fact, they raise so many taxes that the
17:07additional tax money is more than enough
17:09to cover the entire cost of the scheme.
17:11So, so that's sort of the good news. Um,
17:14however, and there's no change in
17:16taxpayer satisfaction or assessment
17:18accuracy. So, we don't lose anything in
17:20collecting that money, right? Um, but
17:23here's what happens and this is where a
17:24systems view. So, right now we were
17:26treating the tax collector as one
17:28person, but there may be different types
17:30of tax collectors. So imagine that first
17:32graph I gave and you say let me split
17:34the tax collectors into two different
17:36types the taxpayers sorry and so what
17:39happens is these tax collectors are
17:41super smart. They don't go around and
17:43treat every taxpayer the same way right
17:46so they're like look I need to raise
17:47more money who can I get more money from
17:51so here's what they do for most
17:53taxpayers they actually don't raise more
17:55taxes but they increase the bribes they
17:58receive and we'll come back to why
18:00that's happening. This is very
18:01interesting. A citizen isn't paying more
18:03taxes, but they end up paying more
18:05bribes. Obviously, the government
18:07doesn't get bribes, but the tax,
18:08remember, the tax collector can earn
18:09money in two ways. I get paid more by
18:11the government or I get bribed more,
18:13right? You can have a whole ethical
18:15discussion on that, but these guys are
18:17just maximizing the money they can
18:18raise. And so, that's what they're
18:19doing. But for some taxpayers,
18:23they increase taxes massively and the
18:26bribes fall for those taxpayers. So what
18:30you're doing is some guys you're
18:32actually bringing properly into the tax
18:34fold and some guys you're just getting
18:36more bribes from. It turns out the guys
18:38you get from the the fewer smaller set
18:41of people you have pay more is enough to
18:43raise that money. Now imagine the the
18:45scale of this. You're only getting this
18:4710% from just a few people. So think
18:50about how much you must have been under
18:52taxing before for that to happen. And
18:55there's an interesting question to say
18:56who do they target? These guys are not
18:57stupid. So if you are given this task
19:00and you want to get money from a few
19:01people, guess who will you will go for?
19:04You go for people who are rich, they
19:05have money to give, but they're not
19:07influential.
19:09If you go for someone who's rich and
19:10influential, guess what's going to
19:12happen to you? You can imagine there's
19:14going to be a call to the head of the
19:15department by some local influential guy
19:17saying, "What the hell is Adnan doing?
19:19He's coming and collecting taxes. Get
19:21rid of this guy." You guys have all been
19:23there. You all know exactly. So this is
19:25exact where they're smart. They only
19:27target those people who can give them
19:28money without giving them grief. Right?
19:31So again, understand that when you give
19:34a simple revenue incentive, it plays out
19:36in a much more complex way which you had
19:39missed if you had just done a very
19:40simple principal agent framework. And
19:42that's an essence of what they're doing.
19:43Now you can ask what in the world is
19:45going on. So theory is so the other
19:48thing we want to talk about is how
19:50critical theory is. Theory is not just
19:52math which you do to impress someone.
19:55theory is telling you a story of what's
19:57going on. And if you know that story,
19:59you can actually change your policy to
20:02better take advantage of what's going
20:03on. And what's going on is the real
20:05story.
20:07So in a normal principal agent
20:08framework, we say the tax collector is a
20:10lazy guy. They're not doing their work.
20:12It's like teachers, they don't teach as
20:13much. We should give them some money.
20:15They'll teach more. That is totally not
20:17what's going on over here. It is not the
20:19standard incentive theory where you get
20:21some guy to work. These tax collectors
20:23are working really hard
20:25because if you want to raise bribe, you
20:27need to know exactly what the person is
20:29due. It's a lot of work to raise a
20:31bribe. Again, I don't want to say which
20:33one of you have tried that. I assume
20:35most of you have experienced the other
20:36side of it. It's not easy for a
20:39policeman to stop you all the time to
20:41raise a bribe. There's a negotiation
20:43what happens. And so the real story here
20:45is a bargaining story. What's really
20:48happening is a tax collector knows
20:49everything. They know how much money
20:51someone is due. They have that
20:52information. They just don't share it
20:54with their boss. Okay? So, when you give
20:57them money, think of a standard
20:59bargaining story. Two things happen. If
21:01I get money by giving you by having you
21:04pay proper taxes, you've increased my
21:06bargaining power as a tax collector. If
21:09you don't pay me the bribe, I'll I'll
21:11just tax you and I'll make money off the
21:14bonus.
21:16And I, and this is the beauty of theory,
21:18and you can write this model down. write
21:19a basic model and in the paper you'll
21:21see this of Nash bargaining.
21:23Imagine two people trying to bargain
21:25about something. If you make someone
21:27stronger,
21:29they will end up getting a higher share
21:31of that deal. By having a tax collector
21:35be able to So when you talk about
21:37bargaining, you talk about the outside
21:39option. What happens if the negotiation
21:41breaks down? What can you walk away
21:44with? And the standards theory of Nash
21:46bargaining is you bargain netting out
21:49your outside options, your walk away
21:52points. And again, if you don't do
21:53bargain, you know, this is very
21:54standard. Like u for instance, for those
21:56of you ever bought something from a from
21:58a place where you can bargain a shop,
22:01you're buying a handiccraft, right?
22:03There's some guy from the US who comes
22:05into u Nigeria, goes to a nice store and
22:08says, "I'm going to buy this beautiful
22:10uh mask." Right? The worst thing you can
22:13do as a bargainer is make it sound like
22:16you want the mask, right? Because you've
22:20weakened your outside option. You've
22:21said, "I really want this, and if I
22:23don't buy this, I can't get anything
22:25like this." The best thing you can do in
22:26bargaining is to go for two seconds, and
22:29you all know this. Look at the mask and
22:31say, "Ah, it's not that great. I saw a
22:33really good mask in the store next
22:35door." What you're doing is you're
22:36strengthening your bargaining position
22:38so you can get a price, better price.
22:41That's exactly what these tax collectors
22:42are doing. They're saying, "Hey, if you
22:44don't pay me a bribe, I don't care
22:46whether I tax you fully because I'm
22:48going to make more money by bringing
22:49more money to the government." So, we
22:51strengthen the tax collector. And then
22:53two things will happen. Either the
22:54bargaining will stay and you get more
22:56bribes, you get a better price, or the
23:00bargaining breaks down and you say,
23:01"Screw it. I'm just going to make you
23:03pay taxes." And that's exactly what's
23:05happen. And this is beautiful now
23:07because now that you know this, you can
23:10decide well you can introduce a better
23:12program which would actually check for
23:14bribes. You want to make the bribing
23:16equilibrium less and less likely because
23:19that is what the government wants. They
23:21want the money to go to their pocket,
23:22not to the pocket of the tax collector.
23:24Right? So this is study one. Now again
23:26think of what happened. We gave a much
23:28more nuanced view of what was going on.
23:30Okay? I'm going to speed through the
23:32others um so that Adan has time as well.
23:34Okay? I want to be respectful that we
23:36have some time for Q&A. Don, we have 45
23:38minutes and then 45 minutes for Q&A. Is
23:40that correct?
23:42>> It's 30 minutes at the end for the Q&A.
23:45>> Oh, perfect. So, we have an hour. So, I
23:46have more time, but honestly, I'd love
23:48to hear from people. So, we'll try and
23:49are do best to finish even sooner if we
23:51can. Um, a lot of you have probably
23:53ideas over here. So, we'd love to learn
23:55from you all as well. Okay. So, now
23:57study two. So, study one used money.
24:01Okay. Now money is costly, right? If you
24:05have a tight budget constraint, why
24:06should you pay money? So then you start
24:08thinking, what else does a tax collector
24:11value? They value money, of course.
24:16Now, here's the beauty. The first
24:17scheme, if you have a very corrupt
24:21system, the first scheme won't work
24:24because no matter how much money you
24:25paid, the tax collector can get more
24:28bribes. So they'll always choose bribes.
24:31Okay. In fact, when we came up with the
24:33scheme, it was very interesting. Adnan
24:34and I with our uh co-author Ben were
24:36designing this scheme. And the first
24:38question we wanted in the first one is
24:39to say how much revenue, how much profit
24:42should we give these guys? And what I
24:44was trying to figure out from tax
24:45collectors is how much average bribe do
24:47you get? I want to pay more than the
24:50average bribe. Now, obviously, if you go
24:51to a room and say what's the average
24:53bribe you get? They're not going to say
24:54that, right? Because that's a problem.
24:58So, instead, what we did was it was
24:59fairly fascinating. One of the texts
25:01which was brilliant. He knew what I was
25:02asking. He knew exactly what I was
25:04asking. So he said, "Asim you're a big
25:06professor. You're an economist. I have
25:08four kids.
25:10Let's make a household budget for me."
25:13So we sat there time and we made a
25:15standard household budget. And then he
25:16said, "This is my pay." And it was
25:18significantly below their household
25:19budget. And then he just smiled and
25:20didn't say anything. So he is telling
25:23me, "You better pay me enough that I can
25:25cover my household budget because the
25:27bribe that I take covers that." That's
25:28what he's telling me. Right? It's a
25:31beautiful way of helping each other. And
25:34I didn't judge him, right? I didn't say
25:35uh because sometimes in economists, as
25:37economists, you know, people do value in
25:39ethical judgments. That's a different
25:41discussion. My view is sometimes you put
25:44people in a position where they have no
25:45choice. That's what he's telling me.
25:47He's like, you think bribe is an
25:48unethical unethical thing. Is it ethical
25:50for me not to be afford able to afford
25:53housing and and paying for my kids? Is
25:56that ethical? So, he's raising a
25:58different philosophy to you. And I'm not
25:59going to judge one or the other. I'm
26:00just going to figure out, hey, if I can
26:02pay this guy incentives, he's better off
26:04or she's better off. U actually most of
26:06the tax collectors tend to be men in
26:08this context. Um and the government is
26:10better off. So can I make a situation
26:12where both people get to do the right
26:14thing and both people are better off.
26:16Okay, but that costs money. Now imagine
26:18you're in a world with this heavy
26:20corruption.
26:22Okay, so then you're like, what do I do?
26:24Money's I can't just I can't pay enough.
26:27But here's a think theory. Here's a
26:29really interesting thing. Corruption
26:31occurs in tax collection in a very
26:33specific way. This happens in education.
26:35You want to be posted in a lucrative
26:38place because if you're in a lucrative
26:40place, right, then you can charge more
26:44money, you can get more bribes. No one
26:46wants to be unfortunately in the poor
26:48area. First of all, you you feel bad.
26:50And I've had conversations with we think
26:52tax collectors don't have morale. They
26:54do. So when they go to a poor area where
26:56there's nothing provided the roads are
26:58crap, the sewage is there. The trash is
27:00there,
27:02by what moral right can you say, pay me
27:04taxes and you'll get stuff. The citizen
27:06will say, "What the hell am I getting,
27:08right? Why should I pay you tax?" And
27:10the tax collector kind of agrees with
27:12that, right? You have to, it's a
27:14conversation you're having. Imagine put
27:15yourself in a position of tax collector.
27:16You walk up and I'm going to make Adnan
27:18again the taxpayer. Adnan is now in a
27:20poor area. And I ask Adnan, "Hey man,
27:23pay taxes." And Adan says, and this
27:26happens, by the way. Adnan takes out his
27:28slipper and starts running after you to
27:30hit you to say, "Are you insane?"
27:33Okay,
27:35now imagine I'm seeing someone on the
27:36screen. I'm seeing Guru on the screen.
27:38Guru, I'm going to reach out to you.
27:40Guru turns out to be a super rich uh
27:43taxpayer.
27:45And I go to Guru. And Guru has beautiful
27:47roads paved, the trash gets collected.
27:49Guru is seeing services being provided.
27:51So I can go to him and Guru is rich. So
27:54the tax he owes me, it's a fancy guru,
27:56you live in a very fancy house, you four
27:58stores, four stories, you have a
28:00beautiful neighborhood next to you.
28:02There's a little park next to you which
28:03the government is maintaining. I go to
28:05you say, "You need to pay me a lot of
28:06money." And Guru says, "Yeah, sure.
28:10Okay. Now you would love to be in Guru's
28:12neighborhood. You don't want to be in
28:14Adab's neighborhood. And the more
28:17corruption you have, the more you like
28:18guru.
28:21Now, this is smart. What if we use that
28:23as a way to incentivize you? Think about
28:26what I just said. I'm going to
28:27incentivize you by the promise of future
28:30bribes.
28:33If you do well, you get to move from
28:35Adnan's neighborhood to Guru's
28:37neighborhood. Now, this scheme is really
28:40interesting.
28:42This works extremely well in a high
28:45corruption area.
28:47The first scheme worked well in a low
28:49corruption area. So now you as a policy
28:52maker, as a academic researcher coming
28:54in, the first thing you should do is am
28:56I in a high corruption area or a low
28:57corruption area.
28:59If you're in high corruption area, you
29:02will know that transfers is super
29:04salient. And by the way, I didn't put a
29:06link over here. There's a real link
29:07between tax collectors. I could have put
29:09that friction in. Actually, I should
29:10have between tax collector, guess who?
29:13I'll have you think for a second and
29:15I'll give you the answer. There's a real
29:16important link when you talk about
29:18transfers and postings between an actor
29:20and a tax collector. And that link, by
29:23the way, is the politician.
29:26The politician would and you do this by
29:29if you ever sit in a government office,
29:31you'd be surprised how many local
29:33politicians show up. Sometimes I joke
29:35with my uh bureaucratic friends to say I
29:37don't know when you do your actual work
29:38because you're just entertaining
29:39politicians. And invariably the request
29:43is Kibri, I'm going to reach out to you
29:45now. I see you on the screen. The
29:47request is Gibri is really unhappy where
29:49he is. Can you transfer Gibbre to the
29:53following place?
29:55And by the way, that gets votes because
29:58Gibbury comes from a community where
30:00there's 500 people who vote with him.
30:03And as long as you produ move Gibber to
30:05where he wants those 500 votes come to
30:07you.
30:09So that's the scheme. So what do we do?
30:11We essentially create groups of 10
30:13people 10 tax collectors who are in
30:16neighboring communities. We don't want
30:18people to travel too much. And we say
30:20whichever one of you did the best this
30:23year in tax collection gets to choose
30:26which one of these 10 neighborhoods they
30:28want to go to next year.
30:31Now this was not easy because we had to
30:33break the political favoritism here
30:35because now you are taking the transfers
30:37and making it objective. Right now it's
30:40political, it's subjective, it's bribe
30:42based. There's a whole deal making going
30:44on over there and you're changing that
30:45deal making. So we do this right. Don't
30:48ask how you can ask Adan afterwards how
30:50we did this and that will tell you. Um
30:53and we basically
30:57implement this policy. Okay. So year one
31:00performance will affect year two
31:02postings.
31:04And turns out this works. It works
31:08almost as well as giving money.
31:12So think about this. This cost the
31:14government nothing financially.
31:17No budgetary implications. And you get
31:19more money. But there is a cost. Let's
31:22not be naive. Nothing comes for free.
31:24There is a political cost. You as the
31:27government are giving away one of the
31:29most precious things you do to dispense
31:32favors, which is posting people not
31:34based on their performance, but based on
31:37how influential they are politically or
31:40how much they're willing to bribe you.
31:42By the way, seats get sold,
31:45right? Tax areas get sold
31:49unsurprisingly because it has value,
31:52right? Always remember people are super
31:55smart.
31:56They're going to do what is best in
31:58their interest to do. Okay. Now, there's
32:02interesting theory here as well. The
32:04theory in the first paper was collusion.
32:07There's deeper theory here. And I want
32:09again you guys to think about it. When
32:10you think of these systems pictures,
32:12don't just look at the results of what
32:14you did. That's just a basic thing.
32:17Think about what the world is doing.
32:19What what are you which world are you
32:21in? Right?
32:25So in this world, think about the tax
32:27collector's problem. Imagine you're a
32:29tax collector. You're in a group of 10
32:33and you're told this is the program
32:35you're going to go under. We have to ask
32:37ourselves, just like we had to ask
32:39ourselves in the surf study, how much
32:41money should you pay? We have to ask
32:43ourselves, will the scheme ever work?
32:45Now think intuitively, where do you have
32:47the highest incentives to your effort to
32:51get the scheme to work? Well, if each
32:54one of you cares about a different
32:56neighborhood, it's not about money.
32:59Adnan likes the neighborhood one because
33:02he lives there. Guru likes neighborhood
33:04two because his aunt lives there. Kibra
33:08likes neighborhood three because
33:10neighborhood three is really nice to
33:12him. They're nice people. They smile
33:14when he shows up. Each one of the three
33:17of them have different preferences for
33:19different places. If you're in that
33:22world,
33:23then the incentive is weak because no
33:26one cares about there's no popular
33:28place. You have competition when
33:30everyone wants the same thing.
33:33If people want different things, there's
33:35no competition,
33:37right? For those of you are applying for
33:39civil service jobs, it's insanely
33:42competitive.
33:44It's harder to get into har it's easier
33:46to get into Harvard relative to doing
33:48well in the IS exam in India for
33:50instance it's factually true by the way
33:53IIIT exam is more competitive than MIT
33:56okay u because it's a very desirable
33:59position you can become an IT guy you
34:02get to work in Bengaluri Bengaluru you
34:05get money from some nice IT firm sitting
34:08in the US that's a great life everyone
34:10wants it
34:13the second thing. You could have a world
34:15where everyone wants one thing, but some
34:18people are so far behind that there's no
34:20point. They're like, there's no way I
34:23can win this thing, so why even try? So,
34:26you have to be in a world where people
34:28want the same thing and a lot of people
34:31are very close to winning,
34:34right? Again, think of it an exam.
34:37If you're someone who's taking this
34:39public entrance exam and you know
34:42there's no way you're going to get the
34:43marks you need, why even bother? You
34:46only do it because your mom and dad want
34:48you to do it. So, you're like,
34:48"Whatever,
34:50I'll do it." But you're not going to
34:51work hard cuz you're not going to win.
34:54But if you almost miss it, you took it
34:57last year. The cutff point was 562.5
35:03and you get 550.
35:05You're like, "Man, I got to study more.
35:06or I can do this. So these two So our
35:09prediction from the model is this scheme
35:12will not work as well for everyone.
35:15It'll only work in places where there's
35:18a popular choice and people are very
35:21close to each other and we test for that
35:23in the data and we find exactly that
35:26right we find exactly that the scheme
35:29works better in places which have these
35:32features to it right uh so this is what
35:35I told you this is the effect now why am
35:37I telling you this I'm telling you this
35:38because this you could have figured out
35:41before even implementing the scheme this
35:44is why theory is Beautiful. You don't
35:46need to wait for a year or two years.
35:48That's a lot of money and effort. You
35:50can use theory to say where will this
35:52scheme be likely to work. And if you
35:54live in a place where people have
35:56idiosyncratic preferences,
35:58they choose they prefer different things
36:00and people are very far from each other.
36:03This is not going to work. Just don't do
36:05it. And these two factors you could have
36:07asked. You didn't need to collect any
36:09real data. You can look at performance
36:11from before historically. Where does
36:12money get collected more from? And you
36:14can also look at people's preferences.
36:17In the case we were in, turns out
36:19everyone did want everyone wants Guru's
36:21neighborhood. So everyone agrees on that
36:24and people's performance is very
36:26similar, right? Guru is just as good as
36:29Adan is just as good as Gib. You put
36:31them in any place, they'll work just as
36:32hard. So the incentives work.
36:36So let me summarize. I'm going to hand
36:38over to Adan now. What if I told you if
36:41you remember nothing else except these
36:44two things it's been a success for us.
36:47One think of the space as a system
36:50space.
36:53Don't oversimplify it. The world is
36:55complicated. Give it respect to treat it
36:57as complicated. Just because people are
36:59poor or uneducated or whatever, don't
37:02think they don't know what they're
37:03doing. Your assumption should be people
37:05are smart and are optimizing
37:08and they're all connected in some way.
37:09You do something over here in one part
37:11of the policy space, other things will
37:14move. And if you don't know how they're
37:15moving, you won't know the effect of
37:17this. Okay,
37:19that's point one, system thinking,
37:21interconnectedness thinking. Point
37:23number two, use theory.
37:28Theory is your friend. If you're scared
37:30of math, don't worry. Just tell stories.
37:32It's fine.
37:34We only use math for theory because it's
37:36a nice way to express theory. But math
37:38is storytelling.
37:41So if you come to writing a story,
37:42instead of this collusion fancy math
37:44bargaining model, you simply needed to
37:46say people are in this bargaining thing
37:50and the person who wins the bargaining
37:52is the person who can walk away. That's
37:54very intuitive. That's all you needed.
37:57The second thing you don't I can in the
37:59paper we do this elaborate math. You
38:01look at it and you're scared. I looked
38:02at it after a year and I was scared. I
38:04don't know what we were doing. But the
38:06intuition is very simple.
38:09Merit based transfers will work where
38:11people value positions and are able to
38:14win. Very intuitive. You could have
38:16written that in three lines in English.
38:19Okay, that's it. Systems and use theory.
38:22I'll hand over to Adnan now.
38:27>> Thank you. So let me yeah let me start
38:31sharing my slides.
38:35Okay. So
38:38let me go down.
38:42Okay. So uh so far we have talked about
38:46like how do we incentivize civil
38:47servants u and as talked about like we
38:52need to think about the system
38:53perspective. How does the state operate?
38:55How does bureaucracy operate? So just
38:57thinking of one link is not enough
38:59because one thing affects other things.
39:02Um
39:03and
39:06sorry I'm getting too many uh
39:09distracting things. Okay. So let me now
39:13like turn to the question of how do we
39:15allocate uh bureaucrats or bureaucratic
39:18effort in in in fact like when we are
39:21designing an organization who should we
39:24incentivize
39:25and what kind of linkages should we be
39:27considering and if we are to giving
39:30authority and control within
39:32organization should we give be giving it
39:34to the frontline agent the supervisor or
39:37like if giving it to one means like also
39:40affects the other how Does it affect the
39:42provision of incentive to the
39:44organization? So this project uh is on
39:48public procurement that we did in uh
39:50again in Pakistan. So um first of all
39:54public procurement is a fascinating
39:56area. So like it's a textbook moral
39:58hazard problem and agent buys goods or
40:00something she wouldn't use with money
40:02she doesn't own. So there are multiple
40:04moral hazard problems. It's also an area
40:07where there's a huge disconnect between
40:09what the procurement manuals say and
40:12what is happening on the ground. All the
40:14procurement manual at least the 30 or 40
40:16that I have seen around the world they
40:17all talk about value for money. What
40:20happens on the ground is um the system
40:23only monitors whether the rules are
40:25followed or not but there are three
40:27courts in the system or not. And now
40:30having three courts in the system is
40:31very different from an outcome
40:33compliance. So a formal rule compliance
40:36substitutes for outcome compliance.
40:38Partly it is because it's very hard to
40:40measure what is an efficient procurement
40:42or not. So in this project one of the
40:44thing that we did was u we work uh we we
40:49basically used statistics we collected
40:51all the attributes of goods for instance
40:54like if you're talking about paper like
40:56it whether it's A4 A6 A3 gage per square
40:59meter like is it 80 g or others also
41:03like it's quantity and other things not
41:05too many attributes for a for say a
41:07paper for laptops and other goods
41:10generic goods this technique works very
41:12with generic code doesn't work for very
41:14idiosyncratic goods like a bridge is
41:16very different.
41:18Um and then once we have all the
41:20relevant attributes and the prices one
41:22can use regression one can use
41:24statistics to get quality adjusted unit
41:26price for the same good. Once we have
41:29that we can solve this problem of like
41:32how do we compare the performance of
41:34different public uh procurement officers
41:37and different public bodies. So again
41:39like economics is also about like data
41:42measurement and using theory to to get a
41:45sense of what data to collect and how do
41:47you compare very simple
41:50call it statistic 101. Uh but then it
41:54allowed us to compare the performance of
41:56different offices and officers over a
41:58period of time and also across each
42:00other. First thing we discover there's a
42:03huge variation in prices paid by
42:06different public of of bodies for
42:08exactly the same good. A4 paper was our
42:11favorite like but we did it for all
42:13other goods. Uh very different prices
42:15even controlling for all of the things
42:17that we can think of like quality gamage
42:20like attributes and other things. So
42:22there is some inefficiency in the
42:23system. Some people are systematically
42:25doing something differently than others.
42:27Then we focused on where is the source
42:30of inefficiency. Should we incentivize
42:32the frontline uh procurement officers or
42:35is it some other inefficiency in the
42:37system? So it turned out like the
42:39frontline procurement officer can't make
42:41the payment before a supervisor and
42:44auditor in this case and the office of
42:47the accountant general. In other systems
42:49it's named differently has to approve
42:51the payment before the payment goes out.
42:54So the actual authority is held by by a
42:56different agent. Um
42:59so this gave us the idea of how do we
43:02test like just not just the provision of
43:03incentive but also allocation of
43:06authority. Again a field experiment
43:09where we provided uh kind one uh uh
43:13business as usual and to some we
43:16provided incentives and to some we
43:19provided autonomy that they could make a
43:22payment without getting the auditor's
43:24prior approval. They were still subject
43:26to exposed auditing and everything
43:30and some that had this combined
43:35and our standard model will tell that
43:36incentive should work very well. That's
43:38incent like what we got from the earlier
43:40papers also. Uh we
43:44collect all of this spending data. We
43:46have all of these like very micro very
43:48detailed data. What we find is autonomy
43:52treatment works very well. It reduces
43:54prices by up to 9%. It doesn't change
43:57the quality, quantity and composition
43:59remains unchanged. And the combined
44:02treatment is as good as the autonomy
44:03treatment alone. It doesn't do anything
44:05beyond the autonomy treatment. And the
44:08incentive treatment is mostly
44:09ineffective in most of the zone. It does
44:11work for a subset of things. Um and I'll
44:14tell you what is going on there. So what
44:16is going on there? Incentives do not
44:17have an effect and autonomy has a big
44:19effect. Uh makes for a big saving. uh
44:24just to explain like again economic
44:25theory in this paper also we have a
44:27model but you can already tell the
44:29intuition of the model in simple words
44:31without necessarily having to go through
44:33the all of the maths which is um
44:37incentives do work when people who are
44:40incentivized have the control um
44:44when people who are incentivized
44:47uh also have the authority to make
44:49decisions Um
44:56okay and um but it doesn't work when the
45:01actual authority is held by supervisor
45:05then it's a question of between the
45:07supervisor and the frontline agent who
45:10is more efficient and honest. If it is
45:13the frontline agent then giving more
45:15autonomy to a frontline agent who is
45:17more honest, more efficient improves
45:20outcome. That's what we discover in most
45:22of our cases. However, giving authority
45:24to like to to a frontline agent who is
45:27more dishonest or inefficient than the
45:30than the supervisor doesn't improve
45:32outcomes. So that's exactly what we
45:34discovered. Autonomy treatment is
45:36effective because giving autonomy to the
45:38procurement officer means less power to
45:40the editor and the agency problem is not
45:43just at the frontline agent. It's also
45:45like at the level within the
45:46organization at the level of the
45:48supervisor and shifting authority lowers
45:51prices. If the auditor is relatively
45:53more inefficient and corrupt. So
45:55shifting authority from a more
45:57inefficient and corrupt auditor to a
45:59frontline agent lowers prices improves
46:02outcomes. This is exactly what we're
46:04discovering and a huge like if you look
46:07into all the procurement manuals most of
46:09the procurement manual that I saw more
46:11than 40 it completely like flips on its
46:15head most of them are based on the
46:16principle the frontline agent is the
46:18dishonest one is the inefficient one
46:21let's resign more and more rules to
46:23constrain the behavior of the frontline
46:25agent and at least in this case I'm not
46:27saying this case is relevant for all
46:29cases what we discover is giving more
46:32autonomy works because it it turns out
46:34that the frontline agent is um is
46:39uh is less dishonest and more efficient
46:42than the frontal than the than the
46:44supervisor. Let me go to the the last
46:47paper very quickly because I want to
46:49raise uh have more time for you to come
46:51in. And this paper is about like uh not
46:54just about like tax collector and u the
46:58senior uh policy maker link but also
47:01like about like the link between tax
47:04collector and taxpayer. In fact, this is
47:06about like how do you make teams of tax
47:08collectors and who how do you decide
47:11which team should be talking to or
47:13dealing with taxpayers. This is a paper
47:15by Jonathan Bel and and colleagues on
47:18Kananga state in DRC in Congo. Uh one of
47:22the like poorest state very low tax to
47:24revenue. Um what they do here is they
47:28introduce property taxes and what they
47:30do in this paper is um
47:34uh property tax collectors go in pairs
47:37and the question that they're asking is
47:39who should we be pairing with whom? Is
47:41it Guru and Kra? Is it um Aram and um
47:45and Sanjana or who should be paid with
47:47whom? Uh does it make difference to have
47:51like people who are like of certain
47:53types and does it make difference who is
47:57dealing with what type of households?
48:00So the two level of assignments are one
48:02is the team assignment of tax collectors
48:05and the other is uh which households are
48:08assigned to which tax collectors and as
48:11you'll see it makes a big difference. So
48:13the optimal assignment in this case can
48:15involve positive assertive matching like
48:17which means that like if you pair high
48:19type collectors together
48:22the counterpart is low types with low
48:24types that makes a big difference. Um
48:28and if you send the high type types to
48:30household with a high propensity to pay
48:32in other in other word places where
48:34there is potential the marginal return
48:36to effort is high where they can
48:38actually make a difference that makes a
48:40big big difference. U these are the
48:42numbers where like teammate matching
48:45alone increases compliance by more than
48:4720% whereas matching them to to the
48:51right team increases by 14%. So huge big
48:55impact.
48:56Um and now my last light on this one
49:00before the conclusion is like like with
49:03all papers I think one message that we
49:05are both Asim and I are telling you is
49:08when you're looking into papers are you
49:10thinking about your context don't just
49:12look into what works and what doesn't
49:14work which is important the most
49:16important question is what is the
49:18economic insight what is the mechanism
49:20why is something working or not working
49:22what is the theory behind it and what
49:24does this paper or this
49:26Or if you designing some policy or
49:28others, what does that policy tell you
49:30and what are the results tell us about
49:32the underlying mechanism in this case?
49:34Like turns out that the high high payers
49:37matching high ability tax collector with
49:39other high ability tax collectors. It
49:42had an outcome. It had an impact not
49:44because of skills. They had more more
49:46skills.
49:48Uh basically they were using the same
49:50mechanism than others. It was because
49:52they were making more effort, exerting
49:54more effort. they were making more
49:56visits, spending more days. Uh so high
49:59high pairing uh in this particular case
50:02uh increased effort and effort increased
50:05compliance because uh they were going to
50:09the households more. They knew when to
50:10find the owners. They also knew when the
50:14owners will have cash that they will
50:15make actually the payment
50:18and which households to go and where
50:20when to go there. In other words, there
50:23were complimentarities in in effort.
50:25High ability collectors work harder when
50:27match with other high ability collectors
50:29and they're increasing more effort also
50:32helps them um connect with households
50:35collect more with household with have
50:37higher uh propensity to pay. Uh my last
50:40slide I think these are the things that
50:42both Asim and I have said uh I'll very
50:45quickly uh share this and then open to
50:47very interested in hearing from you. Um
50:50first thing like I think uh what we are
50:53both saying is don't do not just think
50:56of one link think in systems. So low
51:00state capacity means there are some
51:02friction between actors and fixing one
51:05link alone is not sufficient. You hear a
51:09lot you go to policy circle these
51:11frontline agents are this person
51:12particular agent is either inefficient
51:15or corrupt. We need to do something with
51:17that. that may work. One lesson that I
51:20we hope that you learned from today's
51:22session is think system terms. One thing
51:25changing one thing can also change
51:27something else. Fixing one link shift
51:30something else also. So if you raise
51:32performance pay can increase revenues
51:35but it can also increase bribes as we
51:37showed and incentives can work if
51:40they're designed for that particular
51:41setting. Performance pay can raise its
51:43revenues as we asim showed in the first
51:45paper. But it doesn't work when the like
51:48when the frontline agent is not doesn't
51:51have the the space to be uh to to make a
51:55difference. When the bottleneck is
51:57coming from the guy at the frontline
51:59agent doesn't have the authority the
52:00space to make a difference the
52:03constraint is coming from the supervisor
52:04or the auditor. Allocation can be a
52:07powerful tool as AIM showed in the uh we
52:11don't spend anything and just making
52:13people like incentivized through
52:14transfers and posting can make a huge
52:16difference. Uh in fact it had two3 of
52:19the same impact as a performance pay. Um
52:22you think of like return to investment
52:24being massive because you're not
52:26spending anything and you're getting a
52:27big return. uh and as he showed in the
52:31last example in DRC just better matching
52:34of collectors could raise revenues by
52:3638%. Which is which is a big big deal.
52:39Uh who holds authority matters. So if
52:41you have agents multiple chain of like
52:44agents and principles then it matters
52:46who is more misaligned and usually our
52:49systems when you hear these discussions
52:51they will put all the blame on the
52:53frontline agent as I showed in our
52:56example at least in that case the
52:57frontline agent turned out to be more
53:00honest and more efficient than competent
53:02than the than the the supervisors.
53:04Designing more rules and constraining
53:06their behavior will backfire in this
53:09case. uh theory can be very powerful
53:13when will incentives matter more what
53:15allocation uh how do we design
53:17allocation because it helps us
53:19understand when something works or
53:20doesn't work and last thing which is
53:23hugely important implementation is not
53:25just a management problems we need to
53:26think as an economist in terms of uh the
53:29economic incentives the economics of it
53:31but also the political economy of it so
53:34what is the friction that is is not
53:36allowing implementation to happen um So
53:39it's another word implementation. It is
53:41a design problem. So use theory to to
53:44design. Use data and testing to refine
53:47it. And that's the way to think of
53:49implementation. It's not just a question
53:51of like say consultants coming up and
53:53saying you this guy needs to do this or
53:55that. That will not work in a system
53:57where it's uh there are friction, there
54:00are constraints and there are incentives
54:02that are affecting this. Let me stop
54:04here uh and stop sharing my slide. so
54:09that we can then
54:11uh turn to you
54:14Karab over to you for running this.
54:16>> Thank you Adnan and thank you awesome.
54:19Uh we'll kind of transition now to the
54:21Q&A portion. So for those that have
54:24questions that are panelists, please
54:26just raise your hand. We'll go one by
54:28one and we'll switch out between um the
54:31the questions submitted through the Q&A.
54:33Guru, why don't we start with you? And
54:35just a reminder if we could
54:37>> tell one thing I I
54:39volunteer one thing which I may regret
54:41uh which is if people want to type their
54:43questions in and we can't get enough
54:45questions in I think Adan and I and I'm
54:48I'm nominating I'm voluntelling Adnan
54:50for this which is we will try and
54:51respond to your questions. Uh we could
54:54try to do it in a shared document gall
54:56if that's possible so others can see our
54:57responses as well. I'm trying to
54:59maximize learning. Um, so you know,
55:01happy to answer verbal questions, but
55:03also happy to answer typed in questions.
55:04And Adan and I are both going to be
55:06looking at the chat as well. So if one
55:07of us answers, the other could be typing
55:09in a response.
55:10>> Yeah. And if you guys questions that you
55:12want to pull out, please flag.
55:14>> Uh, I answered the one or two that were
55:16there. So I don't think there's anything
55:18new that I see right now.
55:20>> Okay.
55:20>> N you want to move to my room?
55:23>> Go ahead.
55:26>> Sorry. What did you say? Let's open up
55:28questions. Sorry.
55:29>> Yeah. Adnan, do you want to move to my
55:30room?
55:32>> Oh, okay. Okay, I can come.
55:34>> Sorry. Adnan specially flew from London
55:36to be in Boston just for you guys so we
55:38could be together.
55:42>> And just to note, awesome that uh the
55:45chat is mostly disabled. It's just the
55:47panelists. They can use that. The Q&A
55:48box is where everything's being uh
55:51>> see. Sorry about that. I will try to
55:53open that.
55:55Um, okay. Maybe I'll start with one
55:59question uh in the cluster and then we
56:01can go to Guru. Um, there were a couple
56:05questions earlier on about kind of
56:07incentives versus distortion and Solomon
56:10asked asked one question that kind of I
56:12think sums it up. He says, "How can
56:15governments design performancebased
56:17incentives for tax collectors to improve
56:19revenue collection without encouraging
56:22excessive taxation or harassment of
56:24taxpayers or political backlash?
56:31>> I'm looking at the answers. Why don't
56:32you respond?"
56:34>> Yeah. So, happy to answer the question.
56:36I hope you can hear me. So as you say in
56:39the paper like uh like I think it's what
56:41we find is um tax collectors in
56:44developing countries are
56:45underincentivized. So there is a case
56:48for incentivizing them paying them for
56:50better performance. Um what works in our
56:54case was like tying uh performance to
56:57pay to performance. Just raising like
57:00their salaries alone is not sufficient.
57:02It's the the connection that matters.
57:04But as the questioner asked about the
57:07downside consequences, bribes, what we
57:10also recommended to the government was
57:12yes incentivizing and introducing
57:15performance pay but also introducing
57:18grievance redressal mechanism, audit
57:20mechanism in order to to dilute uh this
57:24this tendency or this like of of
57:27basically this result and eventually as
57:31it happened in our case also we we
57:33recommend ed to the government to also
57:35like uh upgrade their systems so that to
57:38remove as much of the human interface as
57:41possible. So the systems were digitized
57:44and there's in that system also there's
57:46less opportunity there's less
57:48possibilities of bribes um but uh the
57:52element of like human interface will
57:54never disappear. So it's important to
57:56also not just be designing the stick
57:59mechanis the carrot mechanism but also
58:01the stick mechanisms.
58:04>> Thank you
58:07Guru. Do you want to take yours?
58:09>> Yeah. Yeah. Thank you. Uh hi professor
58:11it was uh great presentation. My
58:13question is on the area one of the
58:15municipality that we work in India. So
58:18the scenario is like uh uh with respect
58:21to the property tax like uh there were
58:26there were like you know huge increase
58:27in property tax and then like the
58:30municipal council which is like you know
58:32uh there are politicians. So there is a
58:34lot of bargaining happens between the
58:36politicians and bureaucrats. So in that
58:40way what happens is that based on our
58:43data we could be able to see that
58:44there's a lot of excessive control has
58:47been holdled by the bureaucrats only. So
58:49the negotiations the interactions that
58:51happened in the municipal council which
58:53are not come out. So I'm just wondering
58:55you know how to uh study this and also I
58:59wanted to know like you know how this uh
59:02increase in tax affects the marginalized
59:04communities in Indian context. So I just
59:06want to get your inputs. Thank you.
59:09>> Let me answer the second one first and
59:12then if wants to. Um I think you're
59:14exactly right. Look um there's a deeper
59:16question that someone asked in the chat
59:17and I answered which is look are we
59:18assuming that services will be provided
59:20in lie of taxes. In a lot of countries
59:23that's not the case right you you don't
59:25pay taxes because you don't get anything
59:26in return. So we have a separate project
59:27going on where we're trying to if you
59:29look at the link I've shown a link
59:31between service providers as well right
59:33at the bottom of that uh image. We're
59:35actually trying to improve that link so
59:37that you can directly tell the local
59:41department through the politician what
59:44you want in your neighborhood. I want a
59:46street light. I want a pothole repair. I
59:48want um better uh security. Um you know
59:52um and if you can tie a tax payment to
59:55this political response, then it allows
59:58even marginalized communities to have a
1:00:00voice, right? So I think that's really
1:00:02important. It's not that you're paying
1:00:03taxes. people are willing to pay taxes
1:00:05if they get something in return. I think
1:00:07that's the kind of challenge over here.
1:00:08The other thing with marginalized
1:00:09communities, if you do this well, there
1:00:11is um reallocation. So poorer
1:00:14communities should get more than the tax
1:00:16they pay. They should be in the sense
1:00:18that a rich community gets to pay more
1:00:20taxes. You should take some of that
1:00:22money and allocate it to a poorer
1:00:24community as well just because of sort
1:00:25of progressive uh uh um uh behavior. In
1:00:28terms of politicians, absolutely. This
1:00:31is a highly political thing. And so in
1:00:34that third project which we did which is
1:00:35still in in in the process of being
1:00:37written up, we had a whole bunch of
1:00:39stuff where we tried to get local
1:00:40politicians excited by this, right? If
1:00:43we can say, hey, you can get more votes
1:00:45by delivering better services, then it's
1:00:48a win-win,
1:00:49>> right? Um and so we're trying to do
1:00:51that.
1:00:54>> Yeah. Thank you. So I just wanted to add
1:00:56that you know the incentive thing also
1:00:58they have implemented in that
1:00:59municipality that you know best
1:01:01performing municipality using that as a
1:01:03factor to collect more tax. Yeah thank
1:01:04you.
1:01:05>> Yeah great thank you.
1:01:10>> Thanks Kuru. Um VJ asked a question
1:01:14about I mean a lot of the the bribery or
1:01:17or incentives kind of I think comes
1:01:20around maybe more physical money. Um,
1:01:23have you guys do you have a sense of how
1:01:25things might shift with the digitization
1:01:28of payment and that will that help or
1:01:32hurt bribery, help or hurt corruption
1:01:34streams since it's more centralized?
1:01:40>> I'll take it. So digitization can help
1:01:43in some ways in our system also like the
1:01:46property tax system that we mentioned
1:01:47like uh the digital in the original
1:01:50system all the records were manual that
1:01:52increases the possibilities of obviously
1:01:54like uh [snorts] negative behavior. Uh
1:01:57right now it's all digitized that in
1:01:59reduces the possibilities. However, the
1:02:01human element is still there like
1:02:03properties are still assessed by by tax
1:02:05collectors and others and that's in our
1:02:07case was the biggest margin of uh the
1:02:10side bargains. Um overall as the
1:02:14literature shows like digitization can
1:02:16help reduce corruption by removing some
1:02:19human interface but it doesn't
1:02:21completely like bypass the incentive
1:02:22problems uh
1:02:25that are related to
1:02:28>> that's fine.
1:02:29>> Okay. Wonderful. Thank you. Um,
1:02:32Sanjanna, do you want to ask your
1:02:34question?
1:02:38>> Yeah. Uh, thank you, Pepe. This was
1:02:40great. Um, I had a question about the
1:02:44collusion nash bargaining model. So,
1:02:46it's it's great to know that one can
1:02:48think beyond the principal agent
1:02:50framework. But within that model, you
1:02:54also mentioned that when you're thinking
1:02:55about systems, you have you sort of have
1:02:58like this long chain of principles and
1:03:01agents and they kind of become one
1:03:03becomes the principal and the another
1:03:05becomes the agent. in in u in in this
1:03:08taxpayer setting if the bureaucrat or
1:03:12like the audit officer someone at the
1:03:14supervisory level is having multiple
1:03:17principles and those principles
1:03:20essentially dynamically change for you
1:03:21when you work every day. um like today
1:03:25maybe in this po in one posting you know
1:03:28I feel that the citizen is is like my
1:03:32key principle when when it comes to my
1:03:34work performance and in another posting
1:03:36I realize that the supervisor above
1:03:39because I have a strong social network
1:03:41uh with that person in the next posting
1:03:44how does that um just complicate our our
1:03:48systems thinking in terms of like having
1:03:50multiple principles when we're working
1:03:52as as a supervisor.
1:03:55>> Great. That's a great question, SA. Um,
1:03:57actually, honestly, the principal agent
1:03:59framework is less and less useful in my
1:04:01mind precisely for the reasons you said.
1:04:04Um, in fact, what I was trying to do is
1:04:06even though when you first learn ecom
1:04:08theory, you put it in a very simple
1:04:09principal agent framework just to get
1:04:11the math done. The reality is exactly
1:04:13what you said. They're multi-principal
1:04:15issues. There's peer effects. There's So
1:04:18if you think of the image I drew, I
1:04:20actually didn't need to use the
1:04:21principal agent framework or language at
1:04:23all. I just did it so you were familiar
1:04:25with what I was saying. Ultimately, each
1:04:27of these uh individuals are in a game
1:04:29with each other, right? And you have to
1:04:32decide how that game works. So for
1:04:33instance, if you take multiple
1:04:34principles, you strategically will
1:04:36decide, do I weigh one principle more
1:04:38than the other? Right? You have multiple
1:04:40bosses. Which one do you respond to? So
1:04:42that's the first question. And you will
1:04:44decide that based on who can benefit you
1:04:46more or who can cost you more, right?
1:04:48And then you may decide actually it's
1:04:50more sophisticated than that. There are
1:04:52certain things where the principal
1:04:54cares, principal A cares more about X
1:04:55and principal B cares more about Y. And
1:04:58you may change, you may mix your
1:05:00strategy. Sometimes you're really
1:05:01focusing on X. When you do, you go and
1:05:02sell yourself to principal A. Say, look,
1:05:04a great job I did in X. And then you try
1:05:06and do Y. This just makes your life
1:05:08complicated. But honestly, this is life.
1:05:09Like I assume for those of your PhD
1:05:11students, if you have multiple
1:05:12adviserss, you know how to handle this,
1:05:14right? They have different advice.
1:05:15You're like, who the hell did I listen
1:05:16to? Uh you can also have mechanisms to
1:05:19to deal with this where you can put the
1:05:20adviserss in a same room and say, "Okay,
1:05:22this is great. Can we have a meeting
1:05:24together?" So then advisor A and advisor
1:05:26B disagree with each other and you're
1:05:27just sitting there letting them fight it
1:05:28out and saying, "Guys, I'm happy to do
1:05:30whatever you say, but why don't you
1:05:31first agree what we need to do?" So you
1:05:33can imagine, if you have the power to do
1:05:34that, you can imagine doing the same
1:05:36with multiple bosses. It's not easy
1:05:38because they don't like each other. We
1:05:40advisers like each other mostly. So
1:05:42we'll be happy to come in the same room.
1:05:44Uh so I think that's the way to do it.
1:05:45But what I was trying to say is to me
1:05:47this is beautiful because it means I
1:05:49love rich spaces. If all the world was a
1:05:52principal agent framework, you're done
1:05:53after grad. What's the point? Right? Um
1:05:57and this idea that wherever you are and
1:05:59I I want you guys to realize that you
1:06:01should theorize yourself. Don't just
1:06:03read a paper and say this is the model,
1:06:05right? because we never get it right.
1:06:07Your context will be different. So you
1:06:09need to think about your context and be
1:06:10comfortable modeling. Like the worst
1:06:12paper is when you copy someone else from
1:06:14Finland or whatever. Sorry, I hope
1:06:15there's no Finns over here and say, "Let
1:06:17me just apply it here." I want you guys
1:06:19to recognize that you will become
1:06:20experts and that I honestly mean this.
1:06:23I'm not being facitious. We will learn
1:06:24from you. Sure, Adan and I've been
1:06:26teaching for 30 years, but every time I
1:06:27talk to one of you, you give me some
1:06:28insight which I hadn't thought of
1:06:30before. That's the value you bring to
1:06:32the table. Be confident in doing that.
1:06:37Yeah, thank you.
1:06:38>> Thank you.
1:06:39>> Um,
1:06:40>> just messaged me and said that she has
1:06:42to hop off, so I'm going to give her a
1:06:44chance to ask her question. Um, sorry,
1:06:46Felix. I'll get to you after.
1:06:49>> Thank you so much. Thank you so much.
1:06:50Uh, thank you professors for your very
1:06:52informative um, lecture today. I'mma
1:06:55van. I'm doing my PhD economics from Kar
1:06:57Business School. uh my PhD uh thesis
1:07:00basically studies state capacity
1:07:02institutions and economic growth and I'm
1:07:05using uh cross-country panel data uh one
1:07:08of the challenges I'm facing is that
1:07:09there is very limited uh within
1:07:11variation because of the a slowmoving
1:07:13nature of both institutional and state
1:07:15capacity variables so uh I'm having
1:07:18issue of causal identification I know uh
1:07:21and I understand that in today's lecture
1:07:24the methods were RCT maruro and
1:07:26experimental but giving your vast
1:07:27experience
1:07:28Um I would like to uh take your advice.
1:07:32How would you recommend identifying
1:07:34credible causal mechanisms in macro
1:07:36level research where randomization is
1:07:38not feasible?
1:07:41>> Yeah, thank you. Let me start. Uh I
1:07:43think one of the way of doing that is uh
1:07:45what my colleague at LSC uh Tim Bestlay
1:07:49and Torston person. So they have a whole
1:07:51agenda. I think he gave the earlier
1:07:53lecture on state capacity where uh they
1:07:56both have a theoretical framework but
1:07:58they're also looking into macro data of
1:08:00how uh state capacity measures have
1:08:02changed over time in between in within
1:08:05and across countries and how they are
1:08:08correlated with uh uh changes in
1:08:11economic outcomes.
1:08:13Many of those papers have been
1:08:15published. Uh but take a look at those
1:08:17that literature. Um yes u how much if is
1:08:22causal is an open question. Uh but to me
1:08:25that's the best like um
1:08:28uh the best literature on this subject
1:08:31out there on the macro question that you
1:08:33you talked about.
1:08:34>> Okay. Thank you so much.
1:08:38>> By the way, I apologize. Many of you
1:08:39said I messed up the slides. I was
1:08:41trying to do a job where we could both
1:08:42be together and clearly um that didn't
1:08:44work as well. So, I'm really sorry about
1:08:46that. We'll send our slides so you guys
1:08:48can look at them, but I apologize for
1:08:49that.
1:08:53>> Yeah. All all good. Um Felix, do you
1:08:56want to ask your question?
1:08:58>> Uh thank you so much. Uh thank you,
1:09:01Professor. I think uh it was good for to
1:09:05go on because my question also was about
1:09:07the the multiple principle because uh my
1:09:11concern is that some context that this
1:09:15principle agent operates or as we are
1:09:18being told we think it in the terms of
1:09:20system but we have a context where we
1:09:23have external principles like in the
1:09:26case of uh IMF World Bank detecting the
1:09:30kind of taxation that need to happen. So
1:09:32in this manner you realize that
1:09:34politicians and both the tax collectors
1:09:38are acting on behalf of the external
1:09:41principle instead of domestic principle.
1:09:43And this happened in Kenya I think when
1:09:45the genesis protested against the some
1:09:48of the what they were thinking of
1:09:49physical policy taxation policies that
1:09:51were coming from the MF instead of even
1:09:54the government or the parliamentarians.
1:09:56they thought that the bills that were
1:09:58being the taxation bills that were being
1:09:59passed by the parliamentarians were not
1:10:02actually emanating from from from Kenya
1:10:05rather than from the IMF. So my concern
1:10:08is that how how is the external um
1:10:11environment now affects uh the kind of
1:10:14you know this relationship uh uh of of
1:10:17of agent principles in in even in a
1:10:20systematic manner and the second concern
1:10:23I was thinking of is where the
1:10:25politicians and the tax collectors
1:10:27merger to to in into an extent that we
1:10:30can't differentiate them uh what's their
1:10:33role in in the matter to an extent that
1:10:36the tax h their concern is not
1:10:40actually to the interest of the people
1:10:42or or the taxpayers but the merger to
1:10:46fulfill personal interest. So again what
1:10:50is this merger between the politicians
1:10:52and um that collector effect because in
1:10:55some situation the environment is more
1:10:57personalized than institutionalized.
1:10:59Thank you.
1:11:01>> Great excellent questions Felix. Let me
1:11:03quickly answer them. So look the first
1:11:04one um I grew up in Nigeria and Pakistan
1:11:07so I'm very familiar with both of the
1:11:09things you said I have deep love for
1:11:10both places um you're exactly right
1:11:13there is a sense let me put it this way
1:11:16um when we engage with outside actors
1:11:18I've seen two types of engagement both
1:11:20are bad one is
1:11:24please tell us what to do you bring
1:11:25money and mana from heaven we will do
1:11:27whatever you want us to do right uh the
1:11:30other is the exact opposite you're the
1:11:32foreign devil get out of here. I don't
1:11:34want you. Um you're you're Kenyan so we
1:11:37can do a bunch of history where Kenya
1:11:39itself has played both things. Uh um I'm
1:11:42happy to go offline on that. Um both are
1:11:44wrong because you want to engage
1:11:47intellectually and this is why this is
1:11:49why look we the IGC we bred we as
1:11:52academics are really committed to
1:11:53developing talent smart people in all
1:11:56sorts of parts of the world because you
1:11:58need to produce a local body of
1:12:00knowledge. It's not that the IMF in
1:12:01particular is trying to be, you know,
1:12:04bullish about it. It's just that there
1:12:06is no one equivalent to them who can
1:12:08speak their language and challenge them
1:12:10in many of our countries. And I think
1:12:12I'm hoping all of you guys will
1:12:15eventually reach that point. Some of you
1:12:16will become academics, some of you will
1:12:18be in think tanks, but you will be able
1:12:19to give that independent voice and then
1:12:21the conversation changes. You don't need
1:12:23to be in either extreme because you have
1:12:25power. You have control. The power is
1:12:27knowledge. And I think the the the the
1:12:29most unfortunate thing we've done in
1:12:31places we all come from is that
1:12:32knowledge isn't as invested as it should
1:12:35be, right? And my view is that's exactly
1:12:38why we run a course like this. It should
1:12:39be, right? You should be empowered. And
1:12:41I think that conversation will change
1:12:42and it's just like it's bargaining. I
1:12:44have power, you have power. We can
1:12:46decide where we end up. And it won't be
1:12:47necessarily my thing. It may be your
1:12:49thing, but we'll come into some sort of
1:12:50agreement. I think that's exactly what
1:12:51should happen. Um on your second
1:12:54question with collusion between the
1:12:56politician and the tax collector, I
1:12:57think you're right. Look, there's a
1:12:58deeper question about democratic control
1:13:00and accountability over here. Um and
1:13:03that's uh political science. There's a
1:13:05lot of stuff to say why did this happen?
1:13:06Like I said, we have a third paper which
1:13:08once it gets done we'll share with you
1:13:09guys which is trying to get at these
1:13:11these kind of connections. I would say
1:13:13look um uh Jonathan Wagel, Augusta
1:13:16Bjano, there's a whole team Gabe Durk
1:13:18there's a whole team of phenomenal
1:13:19researchers working on DRC. The one
1:13:21paper we gave you DRC just Google any
1:13:23one of those guys there's a brilliant
1:13:25paper they have on local chiefs and tax
1:13:27collectors it's this lovely paper to say
1:13:30if you combine local chiefs with tax
1:13:32collectors do you get better performance
1:13:34or not and there you'll see the paper
1:13:36there cases you do you don't but you can
1:13:37think of the local chief as a co-op as a
1:13:39as an actor who may be representing
1:13:41depending on how strong the local
1:13:43systems are the community and then the
1:13:45question is because they live in the
1:13:46community so if they start ripping off
1:13:48people in taxes someone's going to knock
1:13:50on their and say what the hell or they
1:13:52won't enjoy them to have gi together or
1:13:54whatever you know you are you're
1:13:56breaking meals together so so I think
1:13:58that's the way to kind of get this there
1:14:00is a beauty of mixing the informal with
1:14:03the formal and I think that's the way we
1:14:05should think about it one is not
1:14:06necessarily worse than the other and in
1:14:08a lot of places we are told that the
1:14:09informal is not right and the formal is
1:14:11the best thing or explicit I don't think
1:14:13that's true in a lot of our societies
1:14:15right and but again I think we need to
1:14:16do the research to show it not just
1:14:19express the sentiment And that's what
1:14:20we're equipping you all with.
1:14:26>> Thank you. Thank you.
1:14:27>> Thanks.
1:14:28>> So, if we look down, we're just trying
1:14:29to type answers as we're speaking. We're
1:14:30not not paying attention to you guys and
1:14:32we're taking turns to do this because
1:14:34one of us answers and the other types.
1:14:37>> Yeah. Yeah. We're we're maximizing
1:14:39output here. Um Anon, I guess then a
1:14:42question for you. Um,
1:14:44compared with incentive reforms, when
1:14:47there are large-scale bureaucratic reg
1:14:49organizations, it seems particularly
1:14:52difficult to evaluate sometimes because
1:14:54they're endogenous and and change many
1:14:57margins simultaneously.
1:14:59What research designs do you find most
1:15:01convincing for identifying the causal
1:15:03effect of uh of a reorg?
1:15:09>> Yeah, tough question. So like for
1:15:11research purposes we want to separate
1:15:13the intervention like design if say if
1:15:16it is an RCT or a field experiment you
1:15:19want to design the arms in such a way
1:15:21that only one thing changes at a time so
1:15:23that you can separately uh identify the
1:15:26impact of certain intervention. Sometime
1:15:28interventions are bundled in the sense
1:15:30like the intervention only works if
1:15:32there are other conditions that also
1:15:34operate in which case it'll be hard to
1:15:36separate the the the impact of like one
1:15:39small component of that one. In real
1:15:42life like many interventions and many
1:15:44reforms that happen are bundled
1:15:45intervention. A lot of things happen at
1:15:47the same time. Uh that may be a good way
1:15:50of like undertaking real world reforms.
1:15:53But in order to understand the mechanism
1:15:55of what is it that created an impact it
1:15:58is also important to test those to
1:16:01evaluate those and for those it is
1:16:03important to also have a design
1:16:06mechanism that separately like
1:16:08identifies the impact of different
1:16:09things. Sometimes harder to design do it
1:16:12in practice uh because of the real world
1:16:15pressures of like rolling out a reform
1:16:17on on a big scale. uh but often it is
1:16:20possible to have like either benefit
1:16:23from a like phase roll out or some other
1:16:25way of like designing an intervention so
1:16:28that we also learn at the same time.
1:16:33>> Brilliant.
1:16:36Um if there are any panelists that have
1:16:39a question you know just raise your
1:16:41hand.
1:16:42Yasir you have one. [clears throat]
1:16:46>> Uh thank you. Thank you very much. So I
1:16:48have a question about like the from the
1:16:51panel about the insensitive and the
1:16:54bureaucracy. So I'll uh I'll like as a
1:16:58case study we'll take like two cases one
1:17:02let's say Pakistan and one a United
1:17:05States of America. So here if you see
1:17:08the bureaucracy in Pakistan where the
1:17:11civil service is a kind of like a legacy
1:17:14system from the British imperiali uh
1:17:18civil service system which has been
1:17:20transferred to the country uh and since
1:17:24the since then it's been going as the
1:17:26way it seems the bureaucracy over there
1:17:29is much more insensitive wise as
1:17:32compared to like any modernday developed
1:17:34countries the British itself self and
1:17:36the United States of America or for
1:17:39instance any other country still we
1:17:41don't see the the productivity of the
1:17:45bureaucracy in terms of services and
1:17:47delivery
1:17:50in every aspect not not up to the par as
1:17:54we see in like other countries. So apart
1:17:56from like insensitivizing
1:17:58the bureaucracy, what other factors do
1:18:01you think are like needed to be
1:18:04considered or make it like more
1:18:06productive and
1:18:08beneficial for the community in general
1:18:10and the system in total?
1:18:13>> Let me thanks let me take that question.
1:18:15Look, part of this is saying why are
1:18:17people poor and the answer is because
1:18:19they're poor, right? Um there's an issue
1:18:21in many parts of our world where there
1:18:23just is a massive resource constraint,
1:18:24right? And that resource constraint is a
1:18:26capacity constraint. It's just people
1:18:27are struggling. There's so some of it is
1:18:29look as the as the the rising tide lifts
1:18:32all boats, right? So um uh and again the
1:18:35answer is a much deeper question. That's
1:18:37a much longer kind of growth
1:18:38productivity kind of question beyond the
1:18:40scope of what we're doing. But I I think
1:18:42we should just take a deep breath before
1:18:43we judgely why things are not working to
1:18:45say man if if the US was to deal with
1:18:47these economic conditions how well would
1:18:49it fare? We already seeing now I'm going
1:18:52to get myself kicked out of this
1:18:53country. You're already seeing how
1:18:54fragile the US political system is given
1:18:57a few people are totally changing how we
1:18:59think of immigrants, how we think of
1:19:00policy. So it's not that the US isn't
1:19:02fragile. I think the answer is all
1:19:04systems are fragile and when stressed um
1:19:07they struggle and so the question is not
1:19:09can we design a system which reduces
1:19:11stress that's very hard. Can we design a
1:19:13can we try and reduce stress to begin
1:19:15with right and that's a different kind
1:19:17of question right? Uh and so that's the
1:19:18way I would answer it uh uh quickly but
1:19:21again I'm happy to have a longer chat uh
1:19:22on this calip can we prioritize um women
1:19:25and people from different places of the
1:19:27world as well because I um no offense to
1:19:30the people I'm hearing right now but I
1:19:31love getting diversity so if they are in
1:19:33particular women I think we've only had
1:19:35one question so far from moment or maybe
1:19:37two um so sorry I no offense guys but
1:19:40I'm just trying to
1:19:42>> Nope that's absolutely fine I've just
1:19:45been pulling them out as they've come in
1:19:47Um there is a question from Clauddio
1:19:51Pimementel who I'm going to guess is
1:19:54from slightly different background. Um
1:19:58and their question is does the variation
1:20:01in public procurement effectiveness stem
1:20:04more from individual buyer
1:20:06characteristics and skills like
1:20:08bureaucratic fixed effects or from the
1:20:10design and rules of the organization
1:20:12itself? um more like organizational
1:20:15practices and and which element is more
1:20:18feasible to transform through public
1:20:20policy.
1:20:23So I would say both uh our paper
1:20:27basically like focused on one aspect of
1:20:30the generic goods where the supplier
1:20:31effects were less important. The princip
1:20:34the agency problems were lying within
1:20:36the government partly because of lack of
1:20:38measurement of what a good procurement
1:20:40is. uh so our paper focus on that one.
1:20:43There's another paper by Michael Best
1:20:45and others for public procurement in
1:20:47Russia. U I think the the the name of
1:20:51the paper gives it away which is
1:20:52individuals and organizations as sources
1:20:55of effectiveness. If we postallocate
1:20:57good bureaucrates and good procurement
1:20:59officers to good procurement
1:21:01organizations they can have huge impact.
1:21:04Um in other like but in other word both
1:21:07are sources of like effectiveness. uh in
1:21:10both of these this type of literature we
1:21:12are abstracting away from what is
1:21:13happening on the supp on the supplier
1:21:16side of the market in certain uh goods
1:21:19like that is the most important market
1:21:21when there is either collusion or
1:21:23something else going on there's less
1:21:25competition and others going on uh so
1:21:27depending on the type of good and the
1:21:29type of like market uh it's either
1:21:32agency effects within government our
1:21:34paper showed even for a highly
1:21:36competitive good like generic goods
1:21:38think of paper think of laptops. These
1:21:40are openly comparative available in the
1:21:42comparative markets. There the relevant
1:21:45margin is agency problems within
1:21:46government and inefficient rules. On
1:21:49other cases if you're thinking of like
1:21:51say construction projects or other
1:21:53projects uh there the relevant margin is
1:21:56what is happening on the either the
1:21:59supplier side are the collusion among
1:22:01suppliers sometime their cartels or the
1:22:04collusion between a supplier and and the
1:22:07procurement officials.
1:22:13>> Thank you.
1:22:14Um, there's a couple more questions kind
1:22:17of out around equilibrium responses. So,
1:22:20I'll lump them together.
1:22:22When bureaucrats respond to incentives,
1:22:25citizens also change their behavior, how
1:22:28important are these equilibrium effects
1:22:30for interpreting the estimated treatment
1:22:32effects?
1:22:34And then a a related question is how
1:22:37should we think about the long run
1:22:38effects of bureaucratic incentive
1:22:40reforms when agents can learn to game
1:22:42the system or change their behavior?
1:22:44Could an intervention generate large
1:22:47shortrun effects that dissipate as
1:22:48bureaucrats adapt? And how can we
1:22:50empirically distinguish genuine
1:22:52institutional change from a temporary
1:22:54response to incentives?
1:22:56>> Great questions. Uh the citizen response
1:22:58is is spot on. Look, a lot of this is
1:23:00and this is the costlier part, but I
1:23:02think we just need to accept it, which
1:23:03is when you do these interventions, you
1:23:05have to measure the response of many
1:23:06actors, including citizens. So, when we
1:23:08did our projects, we had household
1:23:10surveys running in addition to tax
1:23:12collector surveys, in addition to
1:23:13administrative data. Um, the data
1:23:15requirements feel heavy. But what's
1:23:17happening now, which is beautiful, is
1:23:18you have all sorts of sources of data
1:23:20now, right? You have you have admin
1:23:22data, you have social media data, you
1:23:24have you have countries where people
1:23:25protest um and tweets go viral. Uh,
1:23:28right? And so you can use a lot of this
1:23:29data to figure out citizen response. Um
1:23:32but but you kind of need to do uh uh all
1:23:35of this in terms of the the dynamics of
1:23:38this. Uh I think that's spot on. I I
1:23:40answered a question around this which is
1:23:42look you can tell stories where things
1:23:44fade out or you can tell stories where
1:23:46things stay the same or they grow. And
1:23:49it kind of really depends if you can
1:23:50work out these long-term uh
1:23:52relationships, these long-term uh
1:23:55interactions. Um let me give you uh
1:23:58examples of each. So imagine if the tax
1:24:00collectors are raising taxes um by
1:24:03basically squeezing the few people who
1:24:05could pay taxes. At some stage you're
1:24:06just going to run out of space, right?
1:24:09How much more can you get because the
1:24:10people who pay have already paid? So
1:24:11that's going to decay the effect even if
1:24:13the incentives are there. Now the other
1:24:15story is the opposite which is imagine
1:24:17the big game over here is not there's
1:24:19there's more and more spaces um which
1:24:22are untaxed areas. So a lot of places
1:24:24have a tax rating area especially urban.
1:24:26There's lots of places on the periphery
1:24:27of those areas which are effectively
1:24:29urban but are non- taxpayers. Now if you
1:24:32give incentives the biggest way you
1:24:33raise taxes by the way is increasing the
1:24:35tax base actually not just getting high
1:24:37repayment. Now imagine if you give
1:24:39people incentives now they want to work
1:24:41harder and go to an area which wasn't a
1:24:43taxed area. Now your scheme actually go
1:24:46increases over time right? Um and so
1:24:49that's kind of the idea. Now again no
1:24:51scheme will work permanently. You have
1:24:53to keep that's why, you know, I think of
1:24:54policy as a process, not a product. I
1:24:56think we make a mistake and we say,
1:24:57"Hey, we did X and it worked really well
1:24:59and it'll keep working." That's just not
1:25:01true. The world is complicated. So when
1:25:02you do X, someone will respond with it's
1:25:04like a game of chess, right? Think of
1:25:06policy like a game of chess. If you're a
1:25:07good chess player, you think 14 moves
1:25:09ahead. So you kind of need to do that in
1:25:11policy as well. And you learn along the
1:25:13way. So you don't want to ever fix
1:25:15yourself. Which is why when I think of
1:25:16evaluation, often we think of evaluation
1:25:18as saying something was good or bad.
1:25:19That's a mistake. Think of evaluation as
1:25:22calibration or reccalibration. You're
1:25:24building a machine which is running. You
1:25:26do a test. Something doesn't work. You
1:25:27figure out why it's not working. You fix
1:25:29it. You keep going. Right? That's kind
1:25:31of the way to think about this.
1:25:33Gabriella has her hand up. Can I have
1:25:35call her?
1:25:36>> Hello.
1:25:38>> Thank you so much for the lecture. Um so
1:25:41I worked in the projects at DRC uh with
1:25:44professor WGLE and one of the one of the
1:25:47questions that I had was about
1:25:49innovation because it's very hard when
1:25:52you have hierarchy and like hierarchical
1:25:54status between for example politicians
1:25:57and the tax collectors and how like
1:26:00innovation can play around in these kind
1:26:02of situations. Have you worked in um
1:26:04sort of projects where uh innovation is
1:26:08not it's like um pushed forward? Um
1:26:11because sometimes tax collectors have
1:26:12great ideas but they are not the ones
1:26:14who can implement them.
1:26:17>> Yeah, it's a brilliant question. Look um
1:26:19I'm trying to do a project. It's a
1:26:22really deep question to say how does
1:26:23learning happen in organizations
1:26:25particularly bureaucratic organizations
1:26:27and it's a real challenge. There are so
1:26:29many forces to prevent learning and by
1:26:31the same token there's so many forces to
1:26:34prevent innovation and it's not even
1:26:36like the lower guy can't innovate. Even
1:26:38the top guy can't innovate. It's just
1:26:40bizarre because a top guy you know the
1:26:41the standard story they say is in any
1:26:43system where there's um high cost of
1:26:46failure. No one wants to stick stick
1:26:48their head up because the only thing
1:26:49that'll happen is it'll get chopped off
1:26:51if you don't succeed. Right? I'm just
1:26:53think of little gophers coming out of
1:26:54the ground and saying no way should we
1:26:56look around what's happening. Um, look,
1:26:58the system really has to celebrate
1:27:00failure. I don't know how else to say
1:27:01it, right? And that's a real deeper
1:27:04think of entrepreneurship in general,
1:27:05not even the public sector. Take the
1:27:06private sector in our part. Gabriella,
1:27:08where are you from?
1:27:09>> DRC or I'm from Dominican Republic, but
1:27:12I work.
1:27:14>> Yes.
1:27:14>> It took from one DR to a DRC.
1:27:16>> Yeah, exactly.
1:27:18>> So, um, you know, even in the private
1:27:21sector, entrepreneurship is very limited
1:27:23because in the US is a fascinating
1:27:25culture. I have professors of mine,
1:27:27friends of mine who are at HBS who are
1:27:28teaching at HBS because they failed in
1:27:31three companies. And then they come and
1:27:33teach how failure, how you deal with
1:27:34failure. And I'm like, what the hell?
1:27:36You failed three times. In our part of
1:27:38society, no one would even talk to you.
1:27:40You'd be embarrassed. Your parents would
1:27:41never introduce you to anyone, right?
1:27:43Good luck getting married. Um, and in in
1:27:47[snorts] this society, what I love about
1:27:49the US is there many things I don't
1:27:51like. Uh, but what I love about the US
1:27:53is there's an embracing of risk-taking.
1:27:56If you do, you know, let me say this to
1:27:58you guys. If you haven't said no, if you
1:28:00haven't had people say no to you in your
1:28:02life, you're doing something wrong
1:28:04because it means you aren't asking
1:28:05enough. It's a very different way of
1:28:08thinking about it. And I think that's
1:28:09the kind of but the question you rais is
1:28:11really deep. I mean, I'm doing a study
1:28:12right now. We're introducing an
1:28:13innovation in education and we're trying
1:28:15to see if it spreads or persists or
1:28:17diffuses. Exactly this question. Um
1:28:19there's a whole field called
1:28:20implementation science which is trying
1:28:21to figure out how scaleup happens of
1:28:23ideas. Uh there's a really nice paper in
1:28:26a totally different field by um looking
1:28:29at soccer ball industry in Pakistan and
1:28:31innovation spreading in the soccer ball
1:28:33industry. It's by Eric Verhugan and
1:28:35co-authors. So if you like that there's
1:28:36some cool work there as well. But if
1:28:38you're interested in this great study
1:28:40question to study.
1:28:41>> Yes. It's part of one of the questions
1:28:43that I want I would like to answer in a
1:28:45PhD. Thanks so much.
1:28:49Thanks to everyone. Uh I I apologize
1:28:52that there are many many many more
1:28:53questions that we won't be able to to
1:28:55answer now, but um hopefully next week
1:28:58there will be another chance for more
1:28:59people to to ask new ones. U professors,
1:29:03thank you for your time and thank you
1:29:05for for giving uh so much insight into
1:29:08this process.
1:29:11>> Cal, can you share I hope um we're happy
1:29:13to answer these. Can you share the Q&A
1:29:16and there questions we haven't answered?
1:29:18Uh, I hope there's a way to get a
1:29:19transcript of that and if there is, uh,
1:29:21we'll send our responses to you and
1:29:23maybe there's a way you could spread
1:29:24that out to the to the to the community.
1:29:26>> Yeah, I just copy and pasted everything
1:29:28into a Word doc and I'll I'll trim it up
1:29:30and send it over to you guys.
1:29:32>> Awesome. Really appreciate it. And guys,
1:29:33just the last word from us is keep at
1:29:35it, right? I I'm completely biased
1:29:38towards more PhDs mostly because I feel
1:29:40like we need to generate knowledge in
1:29:42our parts of the world and that gives us
1:29:43power. It's it's never resources. It's
1:29:47never natural resources. It's never a
1:29:48strategic positioning. It's just people
1:29:50and you represent a certain segment of
1:29:52people who've committed one way the
1:29:54other perhaps by mistake or perhaps by
1:29:56conscious choice. You're here now make
1:29:58the most of it.
1:30:00>> Great to hear from you and we expect
1:30:02great things from you in future. No
1:30:04pressure.
1:30:06>> Thanks. so much.
1:30:07>> Thank you.
1:30:08>> See you next week, everyone.

About this transcript

This page contains the full transcript of BREAD-IGC virtual PhD-level course 2026 | Lecture 6 Bureaucracy and Implementation by International Growth Centre, generated from the public captions YouTube serves with the video. The transcript has 15,979 words across 2,347 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.