YT YouTube2Text

YouTube2Text → Transcripts → BREAD-IGC virtual PhD-level course 2026 | Lecture 5 Politicians and Political Agency

BREAD-IGC virtual PhD-level course 2026 | Lecture 5 Politicians and Political Agency — Transcript

by International Growth Centre · 13,587 words · 2,048 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:04Welcome everyone. Uh we're back with
0:06another lecture in the series of bread
0:10IGC course on political economy for
0:132026. This is the fifth lecture in the
0:16series and as you can see we have um
0:19Fred and Claudio back with us who are
0:23conveners of the course and and shared
0:25um a lot of what we heard in our kind of
0:28initial introductory lecture but this
0:31lecture specifically will be on
0:33politicians and political agency. So um
0:36just maybe a little bit of background uh
0:39professor Fenan is a professor of
0:42economics and business business
0:43administration at UC Berkeley and
0:46Professor Ferras teaches economics at
0:49University of British Columbia and uh PU
0:53Rio. We're really excited to have both
0:55of them here. Um, as in the past, if
0:59you're interested in being a live
1:01participant, so you can have your video
1:03on and ask questions kind of face to
1:06face in the Q&A session, please just
1:08raise your hand and I'll randomly select
1:1110 to 15 people. Uh, and to address some
1:15of the common questions up front, we do
1:19um require that there are 10 out of 12
1:22attendances to receive the certificate.
1:26your attendance is taken automatically.
1:28So you please don't send into the Q&A
1:30box, hi, I'm here, my name is. We It's
1:34all taken care of. Um, if you're
1:37interested in reading lists for next
1:39week, they're on the website. If you're
1:41interested in this lecture or past
1:43lectures for slides or the recordings of
1:45the lectures itself, those are on the
1:47website as well. And I'll put that link
1:49in the chat.
1:51I think that's everything. So, uh, thank
1:53you for being here. I hope everyone
1:56benefits and we can all learn together.
1:58And I'll pass over to Fred.
2:01>> Great. Thank you, Caleb. Hi, everyone.
2:03Um, welcome. I hope you're enjoying the
2:07course so far. Let me go ahead and share
2:10my screen here.
2:13Um, yeah. So, uh, we're going with, you
2:17know, so Claudia and I are going to be
2:18talking about politicians, political
2:20agency, and political accountability.
2:24Um so by
2:27uh hopefully you can see my screen. Um
2:31so by by way of motivation what I would
2:34like to show you is this graph here.
2:37Um so what we see is that during the
2:40past 50 years much of the developing
2:42world has transition to democracy.
2:46And even though we've had some recent
2:48backsliding, um over half the world's
2:52nations are classified as either liberal
2:54or electoral democracies, uh which are
2:57characterized by sort of these
2:59reasonably free and fair elections and
3:02high levels of participation and you
3:04know competitive political party systems
3:06and so forth.
3:08And yet despite these democratic
3:11advances, what we see from the falling
3:15uh figures
3:17is that many democratic nations,
3:21particularly those that are less
3:23developed, the ones that I have kind of
3:25circled here,
3:27um
3:29struggle to deliver, you know, equitable
3:32access to resources or opportunities or
3:34basic, you know, basic essential public
3:36goods and services. And we can we can
3:38see that clearly when we in the upper
3:40graph where we plot the human
3:42development index and so forth. And we
3:45also see that in the um we also see that
3:49these lower income countries
3:52um are also characterized by having much
3:54more political corruption or
3:56clientalism. And even if we condition on
4:00you know GDP per capita we see a lot of
4:03variation in in these in in um in these
4:08what we would call political pathologies
4:10and so forth. Okay. Um so you know one
4:15question that we're trying to understand
4:17is is why do these democracies sort of
4:20malfunction?
4:22uh especially when you know one of the
4:24promise of of democracies is to be able
4:27to offer you know public goods and and
4:31uh and and and well-being. And what
4:34we're going to argue during the course
4:36of this lecture and we're going to also
4:38give you you know the tool you know the
4:40reasons for why we make this argument is
4:43you know is that democracies malfunction
4:46because of what we would call weak
4:47accountability uh mechanisms that fail
4:51to constrain the government's abuse of
4:53political power. And our argument is
4:56going to center around three sort of
4:59distinct yet related channels of
5:00accountability which have to do with uh
5:03what we would call vertical
5:04accountability
5:06um horizontal accountability and
5:09diagonal accountability. So the the
5:12notion of of vertical accountability
5:15um is the ability to to punish sort of
5:19uh hold politicians accountable uh at at
5:23the polls. uh whereas horizontal
5:26accountability has to do with the
5:28accountability that we often that we see
5:30with for instance the judiciary system
5:33or the legal system and then uh the and
5:37then um uh diagonal accountability has
5:40to do with you know civil society and uh
5:44media and those types of organization
5:46and their ability to hold government or
5:49politicians to account. And so those are
5:51the three different uh dimensions of of
5:54accountability. And the way we're going
5:56to introduce these different is it uh
5:59these different notions of
6:00accountability is in the context of a
6:03political agency framework. And so I
6:05know that that Patrick Francois went
6:08over a class of models uh a couple
6:11lectures ago. Um and political agency
6:14models which uh were featured which you
6:17know he introduced through I think
6:19through uh the Farahan model or or at
6:21least mentioned it um is is a is one of
6:25the core classes of models that we see
6:28uh in political economy these days. And
6:31the notion of a political agency model
6:34is that you have two kind of you have
6:37two actors. You have uh what we'd call
6:39the we'd have the politician which we
6:41which would play the role as the of the
6:43agent and then we have the voter who
6:46would play the the role of the of the
6:49principal. And uh and the reason why
6:52this this this model is interesting is
6:55that there's an incentive problem
6:57between these two actors. uh the
7:00politician has preferences that uh
7:04differ from that of the voters. For
7:06instance, maybe the politician wants to
7:09steal money from office whereas the
7:11voters wants the politicians to provide
7:14public goods instead instead. And so at
7:17the core of this model is this incentive
7:19problem that exists between politicians
7:21and voters.
7:23On top of that, politicians enjoy
7:26anformational advantage. Okay. And this
7:29be and this may be you know uh due to
7:32what we would call moral hazard or or
7:35monitoring in the sense that uh voters
7:38don't observe the actions of the
7:40politicians and as a result uh
7:43politicians will take advantage of this
7:46uh of this uh lack of information to act
7:49opportunistically.
7:51Um or another informationational
7:53advantage um that's you know not
7:55necessarily mutually exclusive either is
7:58one that we would call adverse election
8:00or or selection um which is that uh
8:04voters don't necessarily observe the the
8:07politician's type. And what do we mean
8:09by type? Well, that could be their level
8:12of honesty for instance or it could be
8:14their level of competence. Um and so you
8:18know and typically what voters want or
8:20what they need to do is they need to
8:22select the most competent policy maker
8:24or those whose motivations are most
8:27likely in tune with with what they want
8:29or what the public interest is. And the
8:32inability to observe this type of person
8:36also complicates the matters for voters
8:38in in in their choice to reelect an
8:42incumbent or not. And so how does how do
8:46voters solve this this uh these agency
8:50problems? Well, the the core mechanism
8:53in which they do that is through
8:55elections. Okay. And this is very
8:58different from a from a a class of of
9:03principal agent models that you see in
9:05economics more generally. Okay. So
9:07there's you know there's a long history
9:08of principal agent mo models typically
9:11when when thinking about you know firms
9:13managers and employees and this you know
9:16we've you know political economy took th
9:19those models and adapt them to uh to
9:22politics and the key difference between
9:25the standard principal agents models
9:27that principles here in in the context
9:30of of political agency have only a very
9:33crude incentive structure which is
9:35reelection. Whereas in the standard
9:38principal agent models, they can you
9:39know they'll have different contracts at
9:41their disposals. They can incentivize
9:43workers in different ways. Whereas here
9:46uh reelection is the only sort of
9:48incentive structure that voters can can
9:51impose on on on
9:54um on politicians uh with a few
9:57exceptions that I'm going to get into.
9:59And so the role of these elections is to
10:02achieve accountability. And as we
10:05mentioned in you know even in the very
10:07first lecture we hinted at this there
10:09are two different forms of
10:10accountability there's what we would
10:12call djury accountabilities which these
10:14are the formal rules that affect
10:17accountability a classic example would
10:19be like term limits so term limit you
10:21know if if you're in your final term
10:24term then you can't be reelected right
10:27and this is important because a lot of
10:30studies as Claudio is going to show have
10:32used sort of these djour jury forms of
10:35accountability, these formal rules, uh,
10:38in order to help identify the behavior
10:40of politicians and so forth. Uh, but
10:43there's also something else that we
10:44would call de facto account, de facto
10:46accountability. And the and these are,
10:48you know, what we would call real
10:50accountability in the sense that there
10:52are instances in which politicians can
10:55still displease voters. You know, they
10:58can still be corrupt. they can still
11:00engage in clientalism and all that stuff
11:03without the threat of of of real
11:06sanction. And this could and you know
11:08this could be because they do a lot of
11:09other stuff as well. And so that's what
11:12we would call de facto accountability.
11:15And so uh the key feature of any
11:17political agency model is is the
11:20importance of reelection incentives
11:24in shaping the behavior of uh incumbent
11:27politician. And as a result, as you can
11:29tell, this is going to require kind of a
11:31dynamic structure, right? There's going
11:33to be multiple periods uh in in these
11:36classes of models. Um before so before I
11:39introduce the model, I want to also
11:42describe a couple other sort of key
11:44modeling issues. And you know this is
11:47these are very important because
11:48depending on what you assume in terms of
11:50your in terms of these issues it may
11:53affect sort of our empirical strategy
11:56once we try and test these these models
11:58more formally. Um and so the first
12:01question is what is the the nature of
12:04uncertainty? Okay, what is the source?
12:05Is it that you can't perfectly observe
12:09the politician's act actions? Or maybe
12:12it's because you can't observe the
12:15politician's competency levels or
12:17motives.
12:18Or maybe there's even voter uncertainty
12:21about what the best policy or the exact
12:23policy that was implemented. So maybe
12:25voters don't even observe the policy
12:27that was implemented or, you know, or or
12:30the or their or their benefits from this
12:33policy because it takes too long. and
12:35and what you're going to see in the
12:36model that we're going to present is
12:38that there's going to be this
12:39uncertainty along all three of these
12:41dimensions.
12:42Um the other key issue is is what do
12:46politicians know that voters don't know?
12:48Because that provides what we would call
12:50a multitasking
12:52uh issues where you know maybe you know
12:56why do we why do you wonder why
12:58politicians always like to to um to
13:02build bridges? Well, it's because it's
13:04very observable. It's something that
13:06voters observe. What what when you know
13:09at the same time politicians are also in
13:11charge of, you know, making sure the
13:13water is clean, which is something
13:14that's less observable. So maybe you
13:17know because voters don't observe they
13:20observe the bridges more than the than
13:22you know the quality of the water
13:24politicians are going to divert their
13:26effort towards uh uh building bridges as
13:30opposed to making sure that the the
13:32water quality is is clean. And then the
13:35other component of the the other
13:37modeling issue that that is worth
13:39emphasizing is the fact that you know at
13:41the end of the day voters are going to
13:43vote you know during an election and
13:45they're going to vote retrospectively
13:47right so they're going to observe what
13:50what are what do you know either the
13:52actions or the utility of of what
13:54they've received um prior to making
13:57casting their vote. So they're not
13:59necessar necessarily forecasting out you
14:02know what they expect how they expect to
14:04be in the future and so forth. Okay.
14:09Okay. So, let me present uh this
14:11political agency model that's due to uh
14:15that's a a modification of of of what
14:17Bezley wrote in his book. And I know
14:19that that Tim uh gave a wonderful
14:22lecture last week, I believe. Um or
14:25yeah, I believe so. And um and so this
14:28is this is largely attributed to to to
14:31the wonderful work that he's done. And
14:33so, as I said, we're going to have two
14:36agent two actors in this model. There's
14:39going to be the the politician and then
14:41the voters. Um, as I mentioned before,
14:44these models are dynamic. So, we're
14:45going to need two periods.
14:48And then in each period, the politician
14:50is going to take one action. And, you
14:53know, we're going to denote that by EFT.
14:55And it's going to be an action that's
14:57either going to be a zero or a one. So,
14:58the pol so the politician is just going
15:00to choose zero one. We can think of this
15:02as like the politicians either going to
15:04provide good or you know work to provide
15:07goods or not. Um and then uh the payoffs
15:12of the voters and the politicians are
15:15going to depend on the state of the
15:17world. Okay. And then the state of the
15:19world is also only going to take on you
15:22know two values either a zero or a one.
15:26And and that's relevant for the
15:27following reason. uh first the
15:30politician only observes ST um before
15:35they need to act and so you know right
15:38away we're we're introducing some uh
15:41some uh asymmetry and information here.
15:44So only the politician is going to
15:45observe the state of the world and then
15:48the payoff of the voter is going to
15:50deter be determined on whether or not
15:54the politician sets his action equal to
15:57the state of the world.
15:59So if the if the politician sets the
16:02action so either 01 to the state of the
16:06world 0 or one if they're equal then the
16:09voters are going to receive a payoff uh
16:11called delta. If he decide if the
16:14politician decides not to equate his
16:17action to the state of the world then
16:18the voters will receive a payoff of
16:20zero. Okay.
16:26On top of that, there's going to be two
16:28types of politicians. There's going to
16:30be a politician that we're going to call
16:32a non-corrupt politician denoted by NC,
16:36and then a corrupt politician denoted by
16:39C. And we're going to let pi be the
16:43proportion of non-corrupt politicians in
16:46the population. Okay? Um, and that's
16:50going to matter because if a if uh if
16:54voters decide not to reelect the
16:57incumbent, then they're going to draw a
17:00population, a challenger, they're going
17:02to draw a challenger from from the
17:04population at random. And so the voters
17:07will then know that with probability pi,
17:11that politician is going to be
17:12non-corrupt. So keep that in mind for
17:14when we talk about how voters vote.
17:18Um all politicians are going to receive
17:20an ego rent e for holding office. And
17:23this is just another motive that
17:25politicians may have for holding office.
17:28And uh and then non-corrupt polit and
17:31the distinction between noncorrupt
17:34politicians and corrupt politicians is
17:36that non-corrupt politicians their
17:39incentives are aligned with the voters.
17:41And what do we mean by that? Well,
17:44they're going to receive a payoff equal
17:46to delta if they set their actions equal
17:49to the state of the world and zero and
17:52then and nothing otherwise. Okay, so
17:55they're so they're clearly aligned with
17:57the voters. The corrupt politicians on
18:01the other hand have the opportunity to
18:04receive some rents are from setting
18:08policy not equal to the state of the
18:10world. Okay, where r is going to be you
18:14know randomly distributed from some
18:17distribution with you know a support
18:20with of with a mean with a sorry with a
18:24mean meu and some support over zero r.
18:27So that's just the definition of that
18:28distribution.
18:30On top of that there's going to be some
18:34probability of getting caught gamma if
18:37they choose to steal. And so their
18:40expected cost of of stealing is going to
18:44be is going to be you know gamma gamma
18:47r. And so notice that what gamma is do
18:51the role that gamma is playing in this
18:53model is that it's going to be our our
18:56lever into uh promote into uh horizontal
19:01accountability. So this would be we can
19:03think of this as kind of a reduced form
19:05way of introducing horizontal
19:08accountability into into the model.
19:13Okay. So the timing of the game is as
19:16follows. So at the beginning of period
19:19one, a politician is elected at random.
19:24there's going to be some state of the
19:26world ST that is revealed only to the
19:29incumbent
19:31and as well as their type uh whether
19:33they're corrupt or non-corrupt if if
19:35newly elected. Okay, so this is so
19:37they're going to so the what are the
19:39sources of of um information asymmetry
19:44or uncertainty is the state of the world
19:46and the politicians type. So so the
19:49voters don't observe either one of those
19:51things.
19:53If the politician is corrupt, he's going
19:56to observe the rent that he could steal.
19:59So remember that's that's a draw from
20:01this G distribution.
20:04And then given that either, you know, if
20:07it's a corrupt or non-corrupt, the
20:09incumbent's going to choose
20:12his his policy decision uh EFT
20:16voter. This is going to then translate
20:18into some utilities uh for voters.
20:22But voters are only going to observe
20:24delta with some probability kai before
20:27heading to the poll. So we can the
20:29motivation for this might be we just
20:31don't know what the effect of that
20:33policy is going to be until much later.
20:35So we don't know whether we're going to
20:36benefit or not.
20:39After the elections, the corrupt
20:41politicians are going to draw a new
20:43rent. We'll call that R2.
20:46the same set of decisions and payoffs
20:49are going to be realized in the second
20:51period and then importantly after the
20:54second period the world ends. So the
20:56game is over. Okay? So there's no third
20:59period and fourth period. So you know at
21:02that point there's no more reelection
21:04incentives. And so that's that's you
21:06know that's how we're going to kind of
21:08introduce re-election incentives in the
21:10in the model in the sense that the first
21:12period the incumbent will have this
21:14possibility of getting reelected into
21:16the second period but in the second
21:18period because the game ends there's no
21:20reelection incentives. So this is going
21:22to provide us with the variation that we
21:24need to think about how reelection
21:26incentives affect politician behavior.
21:32Okay. So here is the basic timeline of
21:35the game uh just shown graphically uh
21:38you know I hopefully hopefully all this
21:41is clear but um and so let me go ahead
21:43and and and skip this slide.
21:46So so what's what's the solution to this
21:50game? Well we're going to solve it. Uh
21:53the solution concept will be something
21:54called perfect basian dash equilibrium.
21:58Um and basically what we do there is we
22:01start uh you know in the final period
22:03and then work our ways uh backward.
22:07Okay. So what are what what is the
22:09behavior of non-corrupt politicians?
22:12Well remember non corrupt politicians
22:16have their incentives completely aligned
22:19with voters. So what are they going to
22:21do? They're going to set EFT equal to S
22:24of T in both periods. Right? So they
22:27have every incentive to provide the
22:30public good because they're just going
22:31to that's you know that's you know
22:34that's that's what maximizes their
22:36welfare.
22:38The noncorrupt politician well what
22:40would a non-corrupt politician do in
22:42period two? Well, in period two, he's
22:46the politician's going to set their
22:48policy not equal to the state of the
22:50world because in that case, he'll be
22:53able to extract rents and he's better
22:56off if he's able to do that.
22:59Where the decision gets a bit more
23:01interesting is in the first period. So,
23:05in the first period, what's the
23:07trade-off that a corrupt politician can
23:09do? Well, in the first period, he could
23:13set EFT equal to SFT. Now, remember, by
23:17setting policy equal to state of the
23:19world, he's not going to get anything.
23:20He's not be going to be able to um to
23:24extract any run in that case. And so,
23:26his utility will be zero in that. But
23:30what he might be able to do is convince
23:33voters
23:35to be sort of quote unquote a good type
23:38politician, right? So voters are trying
23:41to determine whether their incumbents
23:43good, corrupt or non-corrupt. And so if
23:46you know if the politician sets policy
23:49equal the state of the world then he
23:51might be able to pull this is why we
23:53call it pooling equilibrium in which he
23:57pretends to be a good type and then he
24:00may get elected into the second period
24:03in which case once he's in the second
24:05period as we just discussed he can then
24:07steal some rents right so on the other
24:11hand maybe he can on the other hand he
24:15may decide just to extract rents in the
24:18first period
24:20and then and in that case there's a good
24:23chance that that voters will observe
24:26that he's that they didn't get the
24:28public goods or there's some probability
24:29that and then if they observe that they
24:32didn't get the pro the the public goods
24:35then in that case for sure they know
24:37that the politician is corrupt and so
24:40then they won't reelect him. Okay. And
24:42so and so there's this tradeoff that
24:45politicians face and and in particular
24:48there's this particular threshold where
24:51we're where we're going to see whether
24:53or not there's a pooling equilibrium and
24:55that's what's given and that's the
24:57expression that's given below in one of
24:59these sub bullet points. Right? So if
25:02the rent remember there's this random
25:04draw of of rents that the politician can
25:07can achieve and if it's below this RC
25:11and what's RC there? RC is this
25:14discounted return. So the benefits of
25:16holding office in the second period. And
25:19so if if the rents that they that they
25:24receive in the first period is less than
25:27R RFC, then they're going to they're
25:29they're going to refrain from from rent
25:32extraction. They're not going to engage
25:33in corruption. However, if it's larger
25:36than RFC, then they're going to they're
25:39going to set policy not equal to the
25:41state of the world so that they can
25:43extract rents. So hopefully the logic of
25:45that is is very clear. Um we can see
25:49this uh graphically uh uh in the
25:53following way. So each corrupt
25:54politician draws this rent R from G.
25:58If R is less than RF RFC, then stealing
26:02is not worth losing office and he'll
26:05refrain from from holding office. Um and
26:09so he'll look honest in the process. But
26:12if RFC is greater or sorry if R is
26:16greater than RFC then he's going to go
26:18ahead and steal. Okay. And so you can
26:20see the proportion of Paul. So what this
26:23graph on the left shows is the
26:24proportion of of corrupt types based on
26:28the draw that are going to refrain for
26:32corrupt from from stealing um versus
26:35those that are going to uh extract rent
26:38in the f in the first period. And what's
26:41important is um
26:44uh there's these there's this uh the
26:47this gamma and kai which is kai is the
26:51probability that you observe if you
26:53recall it's the probability that you
26:55observe the your utility
26:58um if you the higher you move those
27:01things meaning the higher you the the
27:04higher that kai is the more likely
27:06you're obser you observe your utility
27:09the more you the more the this this
27:12decision point will shift to the to the
27:14right and the less the less you're going
27:16to you're going to steal or the more
27:18politicians you're going to get we're
27:20going to refrain and the same logic
27:22applies for gamma which is this
27:24probability of being punished okay and
27:27I'm going to be more explicit about
27:28these comparative statics in a second
27:30okay so what are what are the voter
27:32strategy here so the voter strategy is
27:35to reelect a non uh politician since the
27:39game ends in period 2. So what does that
27:42mean? Well, if they don't observe delta,
27:44then their beliefs are going to remain
27:46the same as their prior. What was their
27:49prior? It was pi because you know the
27:51pool of noncrupt politicians was pi. As
27:53I mentioned, if they do observe delta,
27:57well then voters are going to update
27:59their belief according to bae rule that
28:02the incumbent is non is non-corrupt. And
28:05if you work out the bay rule, basically
28:08it gives you the following equation in
28:10the middle here that you know we don't
28:12need to go over in in much detail, but
28:15basically what this tells us is what's
28:17the probability that the politician is
28:19non-corrupt conditional on on on delta.
28:24Okay. And that's what that's what this
28:26expression gives. Notice that there's
28:28this kai here which is you've observed
28:30delta and then you have this ex this
28:33probability here and then and this is
28:36your updated probability that the person
28:38is is is is
28:41um is corrupt or uh sorry non-corrupt
28:45conditional on on observing delta and
28:48then this is your the the your one minus
28:52kai is the probability of not observing
28:54delta and then this is the your your
28:56belief of of being noncorrupt
28:58conditional on not observing delta. And
29:01so given given this expression what it's
29:04what one can show is that this
29:06expression is greater than um than uh
29:11than pi and so voters have a dominant
29:14strategy to reelect the incumbent if
29:16they observe delta. Okay.
29:20Um so so given that uh the you know
29:25accountability is a key is key to
29:27disciplining politicians and increasing
29:30uh voter welfare. And as I mentioned to
29:33you there in this model there's three
29:36levers for accountability. There's
29:39vertical accountability which appears
29:41through this G function this G RFC
29:45component. And we can think of these we
29:48can think of vertical accountability as
29:49like having free and fair elections that
29:52incentivize politicians to behave due to
29:55the reelection pressures. And so this G
29:58of RC is what's capturing this
30:00reelection uh incentives. And then as I
30:03mentioned to you before there's also
30:05this horizontal accountability component
30:07which is capturing by this parameter
30:09gamma. and and one can interpret this as
30:12a re as a reduced form of way of
30:14reflecting you know the checks from the
30:17courts or other oversight institutions
30:20like audit institutions and so forth and
30:22so a higher gamma raises the corruption
30:26threshold RFC as I showed you
30:28graphically making rent extraction less
30:30likely
30:32and then the third component is this
30:35diagonal accountability uh kai and we
30:38could we could argue that this is going
30:40be determined or a function of the role
30:42of media and civil society in societ in
30:45in in in the location. And so if you
30:48have, you know, independent watchd dogs
30:51or independent media, then the
30:54presumably this diagonal accountability
30:56is going to increase and so a higher Kai
31:00is going to strengthen uh deterrence and
31:03then uh reduce the likelihood of
31:05corruption.
31:07Um so in the next couple slides I I go
31:11through and I explain how we we derived
31:14this welfare
31:16uh function that I that I present in the
31:19other other slide. But for the sake of
31:21time because I want to I want to give
31:24Clauddio time to to go over uh a lot of
31:27the the empirical research that's that's
31:30been you know that's been growing over
31:32the last 5 to 10 years. Let me go ahead
31:35and and skip uh this this part here. Um
31:40which which you know you can go over
31:42these slides uh at your own leisure and
31:45so forth. And and let me just uh finish
31:48off on on the empirical implications of
31:52this model and why this model has been
31:55such a workhorse for thinking about uh
31:58incumbent behavior and so forth. And so
32:02uh what we and so as a result a lot of
32:04these you know uh a lot of these uh
32:07empirical implications from this model
32:09have been borne out in the literature as
32:11as as Claudia is going to mention but
32:14basically what we see is that you know
32:15when there's more media more information
32:18and this can be in in the context of
32:20better radios newspapers internet we see
32:23that corruption goes down if in places
32:27where you know courts function well
32:29where there's audit agencies. There's,
32:31you know, you know, independent
32:33prosecutors, better oversight in
32:35general. We also see in those places
32:38corruption is much lower. There's also
32:41interesting interactions between uh
32:44between uh diagonal accountability and
32:46horizontal accountability uh where the
32:49effects of media becomes even larger
32:52when oversight is stronger. Um and then
32:57you know and then also we and and
33:00importantly we also see that uh based on
33:03term limits that mayor that politicians
33:06who are term limited are going to engage
33:09in more corruption than politicians that
33:12are not termlimited. Um and this can and
33:15and we can show that in a variety of
33:16way. Now there's a lot of empirical
33:18challenges associated with these
33:20predictions namely in trying to find
33:23exogenous variation
33:25because things like you know whether we
33:28have good judiciary system or whether
33:30the media is independent those are all
33:32endogenous. Um and then the other
33:35challenge is that corruption is rarely
33:37observed.
33:39So we need to we need to find a credible
33:42measure of corruption which is something
33:44that the literature has has struggled
33:46with for a long time. Um and then
33:48there's you know and then we both in
33:50this model which I didn't quite discuss
33:52but you know if you get interested in
33:54this model and go in more detail what
33:56you'll realize is that there's two
33:58mechanisms at play. There's a discipline
34:00mechanism in the sense that what
34:01re-election incentives are doing is it's
34:03it's refrain it's it's providing the
34:06incentives for politicians to refrain
34:08from corruption. But there's all this
34:10but there's also a selection mechanism
34:13role that elections play which is what
34:15elections do is they help voters
34:20u uh select the best politician. And so
34:24there's this mix of the, you know, mix
34:26of of good and bad type politicians that
34:28can also affect aggregate behavior as we
34:32would see it. And so with this, let me
34:36go ahead and and pass it uh pass it to
34:39to Clauddio to discuss some of the
34:42empirical evidence uh associated with
34:45this model.
34:59Okay, great. Uh, thanks Fred.
35:04Maximize this.
35:06Is this
35:08Okay, what what I'm going to do is uh
35:12provide you. We don't have a lot of
35:13time. We only have like half an hour. So
35:16what I'm going to try to do is convey
35:19some of the empirical evidence that we
35:21have uh for some of uh this topics that
35:24we have been discussing. If you are more
35:27interested, one of the readings that uh
35:29we told you that you could read is this
35:31chapter that Fred and I wrote for for
35:34the handbook of political economy that's
35:35forthcoming. So you can uh read a lot of
35:38uh the evidence that we know on the
35:40different types of accountability. What
35:42I'm going to do because we don't have a
35:44lot of time is actually focus on uh the
35:47electoral accountability, the vertical
35:49type of accountability and then some of
35:52these other discussions about horizontal
35:55judicial systems and media and all that
35:58you can find that in that in our in our
36:00chapter. So mostly uh I'm going to focus
36:03on two different types of questions
36:05which the one way to think about it they
36:08are kind of like the two sides of of of
36:11the accountability that we've been
36:13discussing about one is the question on
36:16whether elections discipline
36:18politicians. Is it the case that if
36:20politicians going to face elections the
36:22re-election perspective
36:25does that as the model will predict will
36:28discipline politician to behave better
36:31get reelected and then being able to
36:33enjoy those rents either monetary rents
36:36or ego rents in the second period that
36:39they're in office. I mean of course the
36:42model is simplifying a reality that's a
36:44little bit different because politicians
36:45maybe can be reelected for many periods
36:47and all that. The model is very simple
36:50but the overall implication is that
36:52elections will discipline politicians.
36:54So that will be kind of the first uh
36:56part that I'm going to discuss. And then
36:59the second part that I'm going to
37:00discuss that's also needed in order for
37:02this to happen is that the only reason
37:05why politicians will be disciplined by
37:08elections is if they believe that voters
37:11know something and once voters know
37:14something they're going to hold
37:15politicians accountable. So the
37:17prediction that comes out of that is
37:20that information should affect the way
37:23that voters vote. Voters will use that
37:26information to say reward good
37:29performance or punish that performance
37:32because without that then politicians
37:34will not anticipate that if they don't
37:36behave well they will get reelected. So
37:38I'm going to discuss both types of
37:40evidence. It's going to be focused on
37:42the readings that we asked uh you you to
37:45read, but I'm going to bring in a couple
37:47of other pieces of evidence as well as I
37:50as I present this this evidence.
37:54So, as Fred was mentioning, there are
37:56many challenges in answering this uh
37:59research questions. Uh
38:03one is that re-election incentives and
38:04voter information not randomly assigned,
38:07right? So, we cannot just purely
38:09compare. Oh, let me compare politicians
38:11that face re-election incentives with
38:13those that do not if the re-election
38:17incentives are coming from their choice
38:19of running for election again or not
38:22because that's an indogenous choice that
38:23they make. Uh so that's going to be a
38:26challenge. It's also the case that it's
38:29going to be hard to measure uh things
38:31that we care about. So if you want to
38:33for example measure rents then we need a
38:35measure of corruption and people going
38:37to try to hide corruption and stuff. So
38:40how do we actually get data to measure
38:43those things? And then we have we need
38:46some type of exogenous variation and and
38:48and most of the literature have used
38:52either variation in changes in policies
38:54that generate change in term limits.
38:57They all of a sudden some politicians
38:59can get reelected, others cannot.
39:02Sometimes people have used closed
39:03elections, timing and and and things
39:07like that. And on the information, a lot
39:09of the evidence they're going to
39:10provide, they're going to come from
39:11either new policies that somehow provide
39:15information in a random or quasi random
39:18way or for example differential media
39:21markets across localities where some
39:24places have more media than others or I
39:26mean there's a more recent evidence also
39:28using field experiments to understand
39:30the role of of information.
39:35So the the first question that we're
39:39going to ask here is well if voters use
39:43the information to enforce
39:45accountability is it the case that
39:47politicians that face re-election
39:48incentives behave in a different way
39:52and I'm going to answer this question
39:55describing some research that friend and
39:57I did many years ago where we look at
40:00the setting of mayors in Brazil.
40:04And this is an interesting setting
40:06because we have the two things that I
40:09was describing that we need in order to
40:11study this one is variation that some
40:14politicians can get reelected and others
40:17cannot. So that's kind of like the first
40:19piece that we need. And then the second
40:22piece that we need if we want to study
40:24for example rents is some measure of
40:28corruption that in our case is going to
40:31come from federal audits that were done
40:34by the federal government in local
40:36governments. And what we're going to do
40:39is that we're going to exploit the fact
40:41that
40:43up until 1996 in Brazil like in many
40:46countries local politicians were not
40:49allowed to get reelected. They had four
40:51years in office and then they had to
40:53leave. There were no possibility of
40:55re-election and there was a
40:57constitutional change that allowed
41:00mayors and other politicians as well but
41:03our study we're focusing on mayors to
41:06get reelected. Now because this was
41:09allowed in 1997
41:11then what happened is in 2000 some of
41:14those people got reelected while others
41:17did not get reelected. So that means
41:20that in the period that we're going to
41:22study here, which is this period between
41:262000 and 2004,
41:29this period here, we're going to have
41:32mayors that can get reelected and mayors
41:35that cannot get reelected. I don't know
41:39how to do an ad here.
41:42So, we're going to have two types of
41:43mayors. We're going to observe their
41:46corruption level and then we're going to
41:49compare if we compare mayors that cannot
41:52get reelected with the ones that can
41:54controlling for the fact that maybe some
41:57characteristics of these municipalities
41:59are different. How do they fair? How do
42:01they compare one to the other?
42:05So, as I was saying in the beginning,
42:07this is data from a controllers's office
42:11that randomly selects through a public
42:13lottery municipalities to get audited.
42:16And what we're going to do is read the
42:18audit reports and come up with a measure
42:20of corruption in terms of the amount of
42:23money that was diverted over invoice
42:25procurements that had like corrupt
42:27practices etc. And with that, we're
42:29going to have a measure of the share of
42:33resources in that municipality during
42:36that window of time that had corrupt
42:39practices. Or another way to think about
42:41is like what share of money got diverted
42:44for corruption in those places. And
42:47we're going to have a those measures
42:49were a little bit less than 500 places
42:52across 11 lotteryies that this was done.
42:56And we're going to run the following
42:57regression. We're going to regress a
42:59corruption level against an indicator on
43:03whether in that municipality mayors can
43:06and cannot get reelected. We're going to
43:09control for a rich set of municipal and
43:11mayor characteristics.
43:13And we're also going to try to address
43:16the fact that once you compare first and
43:18second term mayors, the beta coefficient
43:21that we have in this regression might
43:24not only represent differential
43:26re-election incentives, but might also
43:29represent, for example, different
43:30selection or different experience.
43:34So, what we're going to find is that
43:38mayors that are in a first term and can
43:40get reelected are significantly less
43:44corrupt than mayors that are in a second
43:46term and cannot get reelected. So
43:49re-election incentives
43:51implies that these mayors are afraid
43:54that voters are going to find out that
43:56they're stealing money and are going to
43:58refrain from red seeking from corruption
44:02in a way that they look 27% less corrupt
44:07than the mayors that cannot get
44:09reelected.
44:14Another way to look at these results is
44:16to say look let's use a different design
44:18where we not only compare for example
44:20all the mayors in a first and second
44:23term but let's focus on those that are a
44:27little bit more similar in terms of the
44:30municipalities where they are.
44:32If we do a regression discontinuous
44:34design where we compare localities where
44:36a mayor got reelected by a small margin
44:40and a municipality where the mayor lost
44:43by a small margin, then what we have is
44:45a quasi experiment where by a little bit
44:49of a margin of victory, we have a mayor
44:52that's a second term visav mayor that is
44:54a first term. And when we do this
44:56comparison, we find actually results are
44:59not that different. sometimes even a
45:01little bit larger than the original
45:03results that I showed you before. So
45:05what you see in this figure is to the
45:08right places where you have someone that
45:10got reelected now is in the final term
45:13are more corrupt that places where you
45:16have a new mayor that now face
45:18re-election incentives than those people
45:21they are actually less corrupt.
45:26Now, if re-election incentives really
45:28matter, we should not observe that only
45:31for corruption. We also should observe
45:33that mayors that face re-election
45:35incentives are trying to put more effort
45:39in order to try reelect it. And what we
45:42do to test that is we look at grants
45:45that in the case of Brazil but that's
45:47true in many other countries where
45:49mayors local politicians can try to
45:52bring resources to their locality or
45:54their municipality
45:56by suggesting infrastructure projects to
45:59build roads to build schools to bring
46:01health clinics to bring tractors to
46:04bring machines etc to their localities.
46:07If we believe that re-election
46:09incentives really matter, then what we
46:11should find is that these mayors, they
46:14should be trying to attract more
46:16resources.
46:18So what I'm displaying here in these
46:20figures are the coefficients of first
46:24term versus second term mayors,
46:27whether they received a grant, the
46:30amount of grant they receive, and the
46:32share of the grant that they receive
46:34based on what they apply for. So the
46:37positive coefficient means that mayors
46:41that have a re-election incentives get
46:43more stuff. Not only because they ask
46:46for more stuff in terms of the discrete
46:49variable, but even the amount of stuff
46:51that they ask for and that they get is a
46:54function of the effort that they make.
46:56But we also find that this effect tends
46:59to be larger the closest that you get to
47:02an election. So the election year his
47:04was 2004. And you can see how if you
47:08have re-election incentives, you try to
47:09bring more stuff to your municipality
47:11and you make a larger effort the closer
47:14you get to the election day.
47:18Okay. So, so now let me switch to a
47:23complimentary type of evidence to
47:25convince you that politicians really
47:28care about the fact that voters know
47:31something about them and react to them.
47:34So this is a very different setting and
47:35a different very different research
47:37design. So this is Bergian and Qualters
47:40looking at Delhi municipal counselors
47:43studying a newspaper report card on
47:46their performance before an election. So
47:49what these researchers did is they put
47:51together a report card with a
47:54description of what this politician is,
47:58their biography, what they did in terms
48:01of how they spent a pot of money that
48:04they can distribute in terms of topics.
48:06This is a setting where a lot of the
48:09individuals are poor. So these are like
48:11in really poor neighborhoods in in
48:13Delhi. So part of what this variation is
48:17is going to capture is whether
48:18politicians are really delivering these
48:20projects for the poorest of the poor or
48:23not.
48:26And what they find is they're going to
48:30go and they're going to pro tell
48:33politicians, a subset of politicians
48:35that this information is going to be
48:37shown to voters, but a randomly set of
48:40politicians are not going to have their
48:44performance disclosed to voters. So
48:47that's the first experiment that they
48:49do. And then the second type of
48:51experiment to complement this one where
48:54is where they vary how public is the
48:56information that is going to be provided
48:59about the politicians.
49:02And what they find which is the more uh
49:05I I'll say the more meaningful uh result
49:08here in this experiment is that if you
49:11look after you tell politicians that the
49:16way they're spending the money is going
49:17to become a public information what you
49:20observe is in these settings that a lot
49:23of people are really poor and legally
49:25Islams is the prop poor spending
49:29significantly increases. So politicians
49:32respond to the fact that information is
49:34going to become available that people
49:36are going to know how they spend the
49:37money and are really afraid of
49:39re-election incentives and therefore
49:41they want to make the best thing that
49:43they can do in order to try to get
49:46reelected.
49:52So let me provide a third uh example on
49:55this before I switch into the role of
49:58information. I'm going to be but I also
50:01want to provide an example that's about
50:03the subsaharan Africa context. So this
50:06is work by uh Roger and Smani where is a
50:10little bit different. They look at local
50:12water systems in three countries in
50:15Nigeria, in Sierra Leon, and in
50:17Tanzania.
50:19And they're not going to be looking at
50:21electro incentives, but they're going to
50:23be comparing
50:25today, many years after the decision
50:28about water infrastructure was built,
50:32the quality of the infrastructure, if it
50:35was built before the election or after
50:37the election. So the idea here is that
50:40if incumbents before an election have
50:43incentives to show to the voters that
50:45they're doing things right, the
50:48infrastructure that you build before an
50:49election will be better, better quality,
50:53more long living compared to the
50:55infrastructure that you build after the
50:57election.
50:59And they have water points all around
51:03built all around Nigeria, Sierra, Leon,
51:05and Tanzania.
51:06And then what they're going to do is for
51:08the different countries they're going to
51:09look at house of representatives, local
51:11councils and national assembly depending
51:13on the type of politician in in each one
51:16of the countries. And what they're going
51:18to find is that if you look at the
51:20functionality of the water system today
51:24and you compare the ones that were built
51:26after the election compared to the ones
51:29built before the election, the ones
51:31built after the election, they do way
51:33worse in the long run. consistent with
51:35the idea that the election provides
51:38incentives for politicians to build
51:41better infrastructure compared to if
51:44you're building the infrastructure after
51:45the election already has passed.
51:51So, so this first part of the lecture is
51:54is is trying to provide you evidence and
51:56testing some of the the parts of the
51:59model that Fred was talking about. This
52:01idea that if politicians face
52:04re-election incentives, they're going to
52:06make more effort and they're going to
52:08behave in a better way. Now, this only
52:11makes sense, of course, if they think
52:14that if voters learn about something,
52:17voters are going to make them
52:18accountable at the polls. So, what I'm
52:21going to do next is provide you evidence
52:23in different settings of voter account
52:27making politicians accountable at the
52:29polls starting with corruption, but then
52:31I'm also going to provide a couple of
52:33other examples. So the first evidence
52:36comes also from the work that friend and
52:38I did also many years ago that uses the
52:42same variation that I was mentioning
52:44before. But in this paper here we using
52:48the fact that the places that were
52:50decided in order to be audited in these
52:53localities were chosen at random. So
52:55there was a public lottery. Some places
52:58were chosen to be audited before the
53:01election that occur in October 2004.
53:05other places were chosen to be audited
53:08after the election that occur right here
53:10in October 2004. So what we have the
53:14facto is a natural experiment where for
53:18all these voters here they're going to
53:20learn about the audit results and
53:23whether their incumbent politicians are
53:25corrupt or not. While for all these
53:28other voters here, they eventually going
53:32to learn what happened when the audits
53:34come to town, but they only going to
53:36learn about it after the election
53:38already has passed. So you can think of
53:40this as a treatment group that's going
53:42to get the information and a control
53:44group that also is going to get the
53:45information, but only after the election
53:49has already passed.
53:53So what we find in this paper is that
53:58politicians that are revealed to be
54:01corrupt are significantly punished in
54:04the polls. So this is the number of
54:06corrupt violations are found and these
54:09are the reelection rates. You can see
54:11that in blue this is the control group.
54:14So in the control group the average
54:16re-election rate is something close to
54:1843%.
54:20While the treatment group when the
54:23information that the politician is
54:25corrupt comes in, you see that the
54:28re-election rate goes down to 30
54:30eventually 20%.
54:33So that's the punishment that voters are
54:36doing when information about politicians
54:39come out that they are corrupt. On the
54:41other hand, if there is information that
54:45the politician is clean, so there's no
54:47corruption found in that municipality,
54:50then this good news is actually rewarded
54:52by voters and the reelection rate
54:55increases from something around 43%
54:59for something around 54 or 55%. So
55:04almost like 10 percentage point. is a
55:06massive increase in the re-election rate
55:09when voters find out that their
55:12politician comes out clean when
55:14governments do this type of audits in
55:17the municipalities.
55:20And also related to something that Fred
55:22was was discussing about the media, what
55:25this paper shows is that this effects
55:27that I just shown you are magnified in
55:31places where you had local media to
55:34provide that information and diffuse
55:36that information. Now, in the case of
55:38Brazil, this is we're looking like more
55:40than 20 years ago where internet
55:41penetration was really really low.
55:44Social media didn't really exist.
55:46National TV don't really cover the small
55:48local municipalities. So, a lot of the
55:51information flows, they go through
55:52radio. So, if you compare places with
55:55and without radio, in those places, you
55:58really get this magnified effects of
56:01this new information and how politicians
56:04are being punished for corruption and
56:06rewarded for uh good performance.
56:11Now, is it the case that bad information
56:14always hurt the incumbents? So, I'm
56:17going to complement what I just showed
56:19you with a couple of uh other pieces of
56:23of of research.
56:25But one thing that's really important
56:27that some people have shown is that
56:29sometimes bad news is not great news
56:33because it disengages people from the
56:35electoral process. So Shan and Qualters
56:39have shown that in the case of Mexico
56:41for example when you provide information
56:43that politicians are corrupt. If there
56:46is a perception then that all
56:48politicians are corrupt this can
56:51disengage people from the electoral
56:53process and instead of people showing up
56:55to punish corrupt politicians they
56:57actually don't even show up to vote. So
56:59in countries where for example voting is
57:02not mandatory people choose to turn out
57:05to vote or not information on the
57:07quality of politicians can actually
57:10reduce turnout.
57:13The finding that information also is
57:17going to make voters punish corrupt
57:20politicians really depends and this is a
57:22point that at show in this paper from
57:26Mexico. It really depends on what's the
57:29baseline
57:31on perceptions that people already have.
57:33So what they do is that they do a
57:35similar experiment here except that they
57:38deliver leaflets doortodoor in a
57:41randomized way where they report the
57:44share of the mayor's federal
57:46infrastructure that is misspent. So they
57:49go to neighborhoods and they randomize
57:51that information and then they look at
57:54how people voted and what they find is
57:58actually that the leaflets were not
58:00worse than expected in the sense that
58:04you're providing information but in a
58:06setting where everybody believes that
58:09politicians are extremely corrupt. It
58:12can be that the information that you're
58:14providing is actually better than what
58:15people thought. Oh, I thought people
58:17were stealing a million dollars per
58:20month. Now people come in and tell you,
58:22oh, they actually only stealing a
58:23hundred thousand. What happens? Well, in
58:26the case of Mexico, they find that
58:28actually some politicians actually on
58:29average get rewarded. So, they get
58:32rewarded for being less corrupt than I
58:35initially expected. So, this is
58:37interesting because in this type of like
58:40experiments, I try to provide
58:41information. We need to be careful how
58:43to frame it, how to think about the
58:45types of information that that we are
58:48providing.
58:50The final piece of evidence that I want
58:52to give you is something about not only
58:56misbehavior of poor corruption, but how
59:00is it the case that voters think and
59:02reward the quality of the performance?
59:06And here I'm going to give you an
59:08example of some work that I have been
59:09doing with qualters looking at the
59:11quality of education in Brazil. So the
59:14setting here is a setting where in many
59:17poor and middle inome countries is
59:19really hard for voters to have a
59:22benchmark to compare the quality of
59:25public service delivery.
59:27Brazil back until uh 2005
59:31did not have a comparable metric for
59:34voters to know if their public schools
59:36were doing well or were not doing well.
59:40So what happens since then is that a
59:42national test standardized was created
59:46where all schools were obliged to take
59:48the test and then the information of the
59:52quality of the education for primary
59:54school was then delivered online made it
59:58public all the newspapers made pieces
1:00:00about it etc. Now, the way that this
1:00:03policy was implemented right before an
1:00:06election in 2008
1:00:08kept out of the system rural schools and
1:00:12schools that had less than 20 students
1:00:14to take this test. Meaning that we have
1:00:18a natural experiment when we have a
1:00:20comparison group that did not have the
1:00:23results of school quality that we can
1:00:25compare to voters that receive the
1:00:29information on school quality.
1:00:32So what we find is that in the top of
1:00:35the distribution, if voters knew about
1:00:39their schools in their municipalities
1:00:42doing well compared to like average
1:00:44schools, they rewarded in approximately
1:00:48two percentage points the incumbent
1:00:50mayor responsible for education.
1:00:53suggesting that if voters learn
1:00:57something where they can infer good
1:01:00quality, they can reward also
1:01:03politicians in the positive way and not
1:01:06necessarily only in the negative way
1:01:08respect to uh corruption.
1:01:11But then in this paper what we do is
1:01:13that we ask the following question which
1:01:15is do politicians then respond to this
1:01:19once there is information about the
1:01:21quality of schools then what we do is
1:01:24that we track infrastructure investment
1:01:27in schools and in this figure here what
1:01:29I'm doing is that I'm comparing schools
1:01:32in pink that were in the bottom on the
1:01:35quality distribution with schools in the
1:01:38blue that are in the top of the quality
1:01:40distribution.
1:01:41and I'm plotting the effects of
1:01:43providing the information after 2007 on
1:01:47an infrastructure index on school items
1:01:50such as equipment, facilities, services,
1:01:53etc. And what we find is that policy
1:01:56makers seem to be making more
1:01:58investments in infrastructure in those
1:02:01schools where information came out that
1:02:04those schools are not doing well in this
1:02:06comparison uh test scores.
1:02:09So to finish up uh key takeaways is that
1:02:14we have
1:02:17evidence
1:02:18that voters respond to cred public
1:02:21information which is something that is
1:02:23implicit in this political agency
1:02:25models, right? Like the idea behind this
1:02:27political agency model is that there's
1:02:29some asymmetric information between
1:02:32politicians and voters. But once voters
1:02:35get some information, they use going to
1:02:37use that information either to infer the
1:02:40type of politician that's trying to get
1:02:42reelected or the effort that the
1:02:45politician made and that's going to be
1:02:47used in order to reelect or not
1:02:49politicians.
1:02:51Media sometimes amplifies that
1:02:53information. So free media in many
1:02:55countries is really important in order
1:02:58to alleviate problems of asymmetric
1:03:01information and we also have evidence
1:03:04that politicians are going to respond to
1:03:06that. If they can get reelected they're
1:03:09going to reduce corruption. We have
1:03:10evidence of that. They might provide
1:03:12more visible infrastructure. They might
1:03:14invest in performance uh that maybe poor
1:03:18people are going to care about. Right?
1:03:19So they're going to care about many of
1:03:21those things. Now, of course, with this,
1:03:25I don't want to say that accountability
1:03:28is always going to work because there's
1:03:30going to be many, many settings, and
1:03:31this is how I'm going to finish the
1:03:33lecture where accountability is not
1:03:36going to work. And this is a little bit
1:03:37outside of the model that Fred
1:03:39presented. But if we try to think in
1:03:42practice why accountability does not
1:03:44work, it's really important to have some
1:03:47of these problems in mind. So, first we
1:03:50have old problems. And in the chapter
1:03:52that I mentioned in the beginning, we
1:03:54discussed many of these problems. Vote
1:03:56buying politicians that want to offset
1:03:59bad news just distributing cash or
1:04:02distributing food or distributing things
1:04:04that politicians think voters care
1:04:06about. Sometimes there is manipulation
1:04:09at the polls. There's fraud in
1:04:10elections. And
1:04:13worse than fraud, sometimes there's
1:04:15political violence when when you use
1:04:18intimidation, guns, groups, coercion in
1:04:23in order to either oblide some voters to
1:04:26vote or dismiss and block some voters to
1:04:29showing up uh to the polls. So these are
1:04:32problems that have been around for
1:04:34decades in many low and middle inome
1:04:35countries that we all know about and
1:04:38that can be problems when we think about
1:04:40political accountability. Compounded
1:04:43into those problems, I want to emphasize
1:04:45two more things that are becoming more
1:04:48and more important uh these days. One is
1:04:51polarization and of course in many poor
1:04:53and and developing countries
1:04:55polarization is an issue especially for
1:04:57example when v people vote along ethnic
1:05:00lines. So they can be important issues
1:05:03of polarization along ethnic lines for
1:05:05example. But nowadays even polarization
1:05:08in terms of ideology and increases in
1:05:10polarization can make politicians less
1:05:12accountable when someone would rather
1:05:15have their own corrupt politician rather
1:05:18than for example voting on someone from
1:05:20the other tribe.
1:05:22The second thing is the rise of social
1:05:26media and the spread of fake news and
1:05:29the at the same time decline for example
1:05:32of professional newspapers etc. that
1:05:36makes information very abundant but way
1:05:39less credible. And because we're we're
1:05:42saying that information is key in order
1:05:43to allow voters to learn about the type
1:05:46and the quality and the effort of
1:05:48politicians, then widespread social
1:05:50media and fake news can really be a
1:05:52problem and a threat to the way that we
1:05:55think about uh accountability.
1:05:58So, I'm going to stop there. I think I'm
1:06:00already uh out of time. I guess we have
1:06:03some minutes for questions. So, KB, I'm
1:06:06gonna uh
1:06:08get it back to you.
1:06:10>> Thanks so much, Claudio and Fred. Um I
1:06:13can see that some questions were
1:06:14answered throughout in the chat, but now
1:06:16is obviously the main time where where
1:06:18we can ask. Um if there are participants
1:06:21that would like to maybe come and ask uh
1:06:24as a live panelist, raise your hand and
1:06:26I can see if I can promote you.
1:06:28Otherwise, we'll just kind of toggle
1:06:29back and forth between who we have here
1:06:31and the ones in the chat.
1:06:34So Masoud maybe we'll start with you.
1:06:40>> Um salutations. Thank you professor. Um
1:06:44my question is uh regarding the the
1:06:46modeling of the first lecture in the
1:06:48first lecture and then the results in
1:06:51the latter lecture. In the first
1:06:53lecture, um, corruption and and the the
1:06:57best action of each politician is
1:07:00defined as if the politician acts
1:07:02according to the world in accordance
1:07:04with the world, the world will duly
1:07:07reward him. But the problem with this
1:07:10model is that uh I do not understand
1:07:13that when when when did we concluded
1:07:16that if the politician and always if a
1:07:18politician acts in accordance with the
1:07:21world the world will provide voters with
1:07:25any kind of rewards or benefits it in in
1:07:28many cases for example what comes to my
1:07:31mind in uh when the Carter was the
1:07:35president of United States that wasn't
1:07:37the case I believe Carter believed act
1:07:39in his best interest of his people but
1:07:42the world did not actually reward him or
1:07:44his people. So when the result came in
1:07:48when the empirical when the empirical
1:07:50result came in there were more about the
1:07:52allocation of results the the perfect
1:07:55allocation of results than what the
1:07:58model actually was about. The model was
1:08:00about getting the rewards but the
1:08:02empirical results was about changing the
1:08:05actions of the politician regarding how
1:08:09they anticipate the behavior of the
1:08:12voters not the reward or or the how the
1:08:16world are going to are is going to
1:08:20reward the their own their actions. uh
1:08:23in reality they were anticipating how
1:08:26the voters will react to their actions
1:08:31not how the world will react whether to
1:08:33the voters or whether to the
1:08:38>> so I can start and then Claudia you can
1:08:40jump in. I I went over the model uh
1:08:42fairly fast. So the so let me just
1:08:46restate the main purpose of the model.
1:08:48The main purpose of the model is to
1:08:50highlight the incentives that
1:08:53politicians face when delivering public
1:08:56goods or deciding to extract rent.
1:09:00And what the model makes very clear
1:09:01based on its structure is that when
1:09:05politicians have reelection incentives,
1:09:08they're going to behave differently
1:09:11than politicians who don't have
1:09:13re-election incentives. So that's why
1:09:15all the subsequent papers are testing
1:09:18this behavior of politicians, right?
1:09:21They're trying to understand, oh, when
1:09:23faced with this incentive,
1:09:26are politicians going to behave
1:09:28differently than not? And so that's
1:09:30that's the link between the model and
1:09:34and the empirical uh results that then
1:09:37follow.
1:09:39>> I see. But my point is whether if the
1:09:42incentive is
1:09:46uh how we anticipate the water's
1:09:48reaction
1:09:51or how we how we perceive the actual
1:09:55consequences of the politician's actions
1:09:59because there are two different things.
1:10:02One politician
1:10:04may act in accordance with what he
1:10:06thinks the water realtors will perceive
1:10:08that that this action will reward them
1:10:10but that action will not actually reward
1:10:12the voters. Okay. How you modeled your
1:10:16um question, how you model how how you
1:10:18model that is there was a connection
1:10:21between how the action will uh the world
1:10:24how how will reward the action but the
1:10:26empirical data will more likely uh to
1:10:30the spectrum that they show that how uh
1:10:32voters will treat the anticipation and
1:10:34the perception of voters how how they
1:10:37try to manipulate that and there's a
1:10:39difference and I I like to know your
1:10:41opinion on that. Sure there's uh yeah of
1:10:45course there's there's a difference the
1:10:47the the point of the model as and I
1:10:50think uh Patrick Francois highlighted
1:10:52this in his lecture is that you know
1:10:54there's certain empirical tests one
1:10:57would be to test the assumption of the
1:10:59model which is something that Clauddio
1:11:01presented in the second half the other
1:11:03is to test the implications of the model
1:11:06now all models are going to be
1:11:07reductionist in some sense um but you
1:11:11know and so and So you know it's not
1:11:14going to capture exactly every aspect
1:11:17that and in fact it's up to you as a
1:11:19researcher to decide what is the aspect
1:11:22that you're trying to explain. Are you
1:11:24trying to explain differences in the
1:11:27behavior of politicians or are you
1:11:29trying to explain the phen you know a
1:11:32phenomenon in which politicians may act
1:11:35appropriately and yet still voters may
1:11:38still not uh be rewarded in some sense.
1:11:41So those are, you know, it depends for
1:11:44that latter question. Maybe you wouldn't
1:11:46use a political agency model in that
1:11:48case. But what's important for me, I
1:11:50want to make sure that you understand is
1:11:51that in these models, voters don't
1:11:54observe the actions of politicians.
1:11:56That's the that's the entire point of
1:11:58this model because if they did observe
1:12:00the pol the actions of politicians, then
1:12:02there wouldn't be this asymmetry in
1:12:05information and and so they can and they
1:12:07can punish the politician accordingly.
1:12:09So it's precisely because they don't
1:12:11observe the action and they only observe
1:12:13the state of the world or the benefits
1:12:16uh you know either all the time or you
1:12:19know some of the times that you get that
1:12:22you get uh that it delivers the
1:12:25predictions that that that we see.
1:12:28Thanks so much. Um we have a few
1:12:31questions in in the chat around kind of
1:12:33the accountability failures in
1:12:35developing countries and I guess the
1:12:38underlying concern or or or core
1:12:40question is that the model assumes um
1:12:44that elections will discipline the
1:12:45politicians but sometimes that doesn't
1:12:47happen and so there is a one question
1:12:50from more blessing who asks why do
1:12:54accountability mechanisms fail in
1:12:56developing countries even when
1:12:57democratic institutions exist Is it an
1:13:00issue of democracy itself or more the
1:13:03inability of the institution to
1:13:05constrain power?
1:13:09Uh I can I can start and then so this
1:13:13was something that that Claudia alluded
1:13:15to in his last slide which if you you
1:13:17know if you wanted to read our h
1:13:19handbook chapter we go into quite you
1:13:21know quite some depth uh because the
1:13:23question that that we try to address in
1:13:25that is precisely why are these
1:13:29democracies malf what we would call
1:13:31malfunctioning now you're exactly right
1:13:34like the you know the model that that I
1:13:37presented today that we also present in
1:13:38the handbook chapter takes you know this
1:13:41very strong assumption that we do have
1:13:44free and fair elections right and so if
1:13:47you re if you you know relax that
1:13:49assumption then it you know the then the
1:13:53world gets much more complicated because
1:13:55and this is exactly what we see in a lot
1:13:58of uh lower inome countries which is you
1:14:01know the as Clauddio mentioned the use
1:14:02of political violence the use of uh
1:14:06election rigging
1:14:07the use of vote buying, clientalism, all
1:14:10of these things, you know, work to
1:14:13weaken these accountability mechanisms.
1:14:17Um, you know, even, you know, why is it
1:14:20that why is it that uh politicians will
1:14:23buy media markets or
1:14:27imprison uh or you know, you know,
1:14:30imprison uh uh journalists and not let
1:14:33them into their press conferences and so
1:14:35forth. Again, those are those are
1:14:37actions that politicians take to weaken
1:14:40what we would then call diagonal
1:14:42accountability. And so, you know, there,
1:14:44you know, there are in the model that we
1:14:46present, our actors are very our
1:14:48politicians are very limited in their
1:14:50actions. All they can do is provide
1:14:52public goods or steal and a much you
1:14:55know, as we know from from experience
1:14:58and from viewing the world, politicians
1:15:01can do much more. They can engage in
1:15:02buying. they can, you know, threaten
1:15:04their their their enemies. They can do a
1:15:07variety of different things that we see
1:15:09in the news almost constantly. And so
1:15:11that's, you know, it's it's those types
1:15:14of additional actions that we feel
1:15:16weaken this accountability mechanisms
1:15:18that we, you know, that we wrote in
1:15:20theory. Um, and so
1:15:26>> great.
1:15:29Well, uh, let's go then go back to a few
1:15:32of our our live attendees.
1:15:35Emagos, do you want to ask your question
1:15:38and then we'll move to Hillary. And if I
1:15:40could just ask that we keep the question
1:15:42quite brief so we can hear as many as
1:15:44possible.
1:15:54>> Okay. Thank you, professors Clauddio and
1:15:57Fred. Um I have a couple of questions
1:15:59actually and all of them are centered
1:16:01around the relationship between these
1:16:04incentives we are talking about and then
1:16:07accountability
1:16:08especially when politics itself is
1:16:11framed or conceptualized as as as a
1:16:14game. So I want you to understand if uh
1:16:18institutions that are supposed to
1:16:19enforce accountability are
1:16:22themselves controlled by actors that are
1:16:25participating in the political game.
1:16:27then who actually creates the incentive
1:16:30to be accountable and then two I would
1:16:33love to know if um we can actually find
1:16:36some ground on which we can rationalize
1:16:38asking politicians to behave against the
1:16:42um uh the very incentives that are
1:16:46generated by the political system when
1:16:48in fact or should I say perhaps now um
1:16:52it is the duty of institutional design
1:16:54to [clears throat] like change the
1:16:56payoff of structure to change the payoff
1:16:59structure for um for the game itself. So
1:17:02I don't know if you understand the
1:17:03question. So they're actually two in
1:17:05one. The first is the relationship
1:17:07between accountability and then
1:17:09incentive because I want to know um in
1:17:13developing economies we know that there
1:17:15are certain institutions that actually
1:17:17exist but those institutions are
1:17:19controlled by actors that are already
1:17:20participating in the political game. So
1:17:22in that instance who actually provides
1:17:25the incentive for these politicians to
1:17:27be accountable to the people. And then
1:17:29the second one as I said I would love to
1:17:31know if um we can rationalize asking
1:17:34politicians to behave against the
1:17:35incentives that are generated by the
1:17:37political system itself when I think it
1:17:40is the duty of institutional design to
1:17:42change the payoff structure for the
1:17:43game. Thank you Sam.
1:17:46>> Okay I I can start answering that. So,
1:17:49so I think those are two really good
1:17:52questions. Uh, on the first one,
1:17:58part of what a future lecture is going
1:18:00to bring is going to be a discussion
1:18:01about political selection that we have
1:18:04not discussed here like like who are
1:18:06politicians, where do they come from, uh
1:18:09what do they represent, how do political
1:18:11parties select them, like that part
1:18:14we're not discussing here. We're
1:18:15starting here saying look these are the
1:18:17politicians. How do we make them behave
1:18:20better? But then there's a question
1:18:22about selection, right? Like who are
1:18:24these people? How do political parties
1:18:26bring people into the political system,
1:18:29etc. And how people self- select? Who
1:18:31are the types of people that want to be
1:18:33a politicians in in a country? Do they
1:18:35want to do they have civic virtue or do
1:18:38they just want to to get rich for
1:18:40example? Right? Uh so in the future
1:18:43lecture, I think there's going to be
1:18:45some discussion of that.
1:18:48uh that is also related to your second
1:18:50question because the types of people
1:18:52that run for office are a response of
1:18:54the institutional design, right? So, so
1:18:57countries have institutions that will
1:19:00map what do you need in order to get
1:19:03elected? Is it a proportional
1:19:04representation or a majoritarian system
1:19:06for example? What's the role of
1:19:08political parties? How do political
1:19:10parties get funding? all those rules,
1:19:12they're written rules that will affect
1:19:15the power that parties have and the
1:19:18self- selection to who runs for office.
1:19:20So, so I think the question that you're
1:19:22asking about can we expect people to
1:19:25behave in different way if that's the
1:19:27incentive that the rules create, I think
1:19:29the answer is no. I think people are
1:19:30rational and they're responding exactly
1:19:32to the incentives that are created by
1:19:34the political system. And if we want to
1:19:38transition to reduce corruption, improve
1:19:42public service delivery, we need to
1:19:44think of what are the rules of the game
1:19:46that we create the right incentives for
1:19:48people to behave in the right way. Now
1:19:51that might not be enough, right? So, so
1:19:53maybe it's necessary, but it's not a
1:19:55sufficient condition because a lot
1:19:57[snorts] of it does come from culture. A
1:20:01lot of it does come from informal rules
1:20:05of the game, right? And and that's the
1:20:07distinction that Fred made in the
1:20:08beginning between the rules that are
1:20:10written down visav officially what's
1:20:12going on. You can have a very detailed
1:20:14limit on campaign financing rule in a
1:20:17country but 80% of the politicians get
1:20:20their money under the table in an
1:20:21illegal market and therefore is useless
1:20:24to have that ruled. Right? So I think
1:20:26those are really really important
1:20:28questions that I think we're just
1:20:30beginning to understand the answer in
1:20:32terms of empirical evidence.
1:20:35>> Thank you so much sir.
1:20:39>> Thank you Mr. Garass. Hillary would you
1:20:41like to ask your question?
1:20:44>> Yes. Yes. Uh my question is very clear
1:20:46prof I'm Hillary from Kenya Nairobi. uh
1:20:50we have seen uh post uh January 6th uh
1:20:54of the United States and now the
1:20:56low-inccome countries or middle inome
1:20:58countries like in Kenya now applying the
1:21:01use of informal groups and militias to
1:21:03try to tame their uh the public or
1:21:07supporter or anti-uporters uh for them
1:21:10is this the new emerging area for
1:21:12political uh experiency and how is that
1:21:15now working with the model that you
1:21:17presented earlier? Thank you.
1:21:26Am I clear?
1:21:29>> Yeah, I got that. Um, sorry, Fred.
1:21:34>> Go ahead, Fred, if you want.
1:21:37>> Sorry, he broke up on my end. Can you
1:21:38repeat the question?
1:21:42Yeah, my question is just about the use
1:21:45of informal groups, militias to
1:21:48basically bring political order against
1:21:51the use of formal government channels
1:21:53like police because of the
1:21:55institutionalization of certain
1:21:56accountability oversight mechanisms that
1:21:58would hold police to account. And we see
1:22:01this system is uh mirrors what happened
1:22:04in the on the 6th of January in in the
1:22:08United States and many other countries
1:22:10in in Africa. Now do you find the state
1:22:12is using militia to defend their
1:22:14position?
1:22:15That would be my question.
1:22:19>> Yeah. So that that I mean so that's been
1:22:22going on for quite some time, right? So
1:22:24even before January 6, it's you know
1:22:27been going on in Latin America for a
1:22:29long time. It it speaks to, you know,
1:22:32this issue of of, you know, the monopoly
1:22:35of violence that the state has um and
1:22:38whether, you know, to what extent do
1:22:40they do they do they control uh
1:22:44control the the the you know, allow for
1:22:47it. And so um you know how that you know
1:22:51how that you know speaks to
1:22:53accountability I I would put that more
1:22:55in the in the category that we've
1:22:57already discussed with uh which is you
1:23:00know the use of political violence.
1:23:02Political violence doesn't necessarily
1:23:04have to be before an election. It can
1:23:05also be after the election. Um and we've
1:23:08seen and we've seen plenty of examples
1:23:10of that and we highlight several
1:23:12examples in in the in the chapter. So um
1:23:16maybe I missed some other aspect of the
1:23:18question but uh this is you know this is
1:23:21this is very much a way in which the
1:23:24government uh uh weakens accountability
1:23:27in a in a very uh serious way.
1:23:35>> Yes. Okay. I'm okay.
1:23:40>> Great. Um well there are a couple
1:23:42questions in the chat about I think
1:23:45there there's a huge number of
1:23:47participants from subsaharan Africa and
1:23:49many are are saying some sort of
1:23:52variation of the question of uh how how
1:23:57do elections in subsaharan Africa
1:23:59actually redistribute bargaining power
1:24:01between citizens and policies when
1:24:03there's so much inequality and and do
1:24:07you have any suggestions on what
1:24:09accountability mechanisms might be most
1:24:11appropriate for for things like um when
1:24:16a political party has already started
1:24:18saying that they've won the election
1:24:19before they actually do and they all and
1:24:23they do um and pushing any, you know,
1:24:26the narrative that any outcome that's
1:24:29not the outcome they want is fraud. I
1:24:32also want to say that this happens
1:24:34certainly outside of this case was
1:24:37Nigeria or subs Africa. We're seeing it
1:24:39all over the place. So maybe just a
1:24:41chance to respond to that.
1:24:52I mean, I have a couple of thoughts, but
1:24:54if if uh I mean, look, there I mean this
1:24:59is this isn't based on any research
1:25:02evidence that we've have. Well, with the
1:25:03exception of one one example that I'll
1:25:05give, but like um you know, it's not a
1:25:08perfect solution, but uh independent
1:25:11election observers is is is critical for
1:25:15trying to ensure that there it's a free
1:25:17and fair election. We have plenty of
1:25:19evidence that that that can be an
1:25:21effective policy uh in in in making, you
1:25:26know, and in increasing the integrity of
1:25:28of an election. Um the the other issue
1:25:33is is which has been pointed out uh
1:25:37uh in the literature is that in a lot of
1:25:39countries in subsaharan Africa voting uh
1:25:42along ethnic lines is so powerful that
1:25:46you know that's also perhaps a root of
1:25:49the problem. Although there is, you
1:25:50know, interesting work by by Katherine
1:25:53Casey who's going to be a speaker uh uh
1:25:56you know, in a in a couple weeks who has
1:25:59shown that like that voters are willing
1:26:02to cross party lines when they're given,
1:26:05you know, better information about the
1:26:07candidates. And so she's done a lot of
1:26:09work trying to you know institutionalize
1:26:14debates within within you know within
1:26:17Sierra Leone and other countries uh
1:26:20precisely
1:26:21by in order to provide voters with the
1:26:24adequate information that they need to
1:26:26make certain uh decisions about about
1:26:28who to elect and hope and then and maybe
1:26:31that will break up some of the the the
1:26:34voting along along ethnic lines which
1:26:36which really dampens accountability,
1:26:38right? So, um it's it's easy to show
1:26:42theoretically and in fact there's a
1:26:44annual review piece that I that I wrote
1:26:46with with um with Ernesto Delo that if
1:26:51if you have strong party affiliations,
1:26:53whether it's through ethnicity, whether
1:26:54it's through religion, where you're
1:26:56willing to, you know, you basically are
1:26:58just going to vote for the person for
1:27:00the group member that you belong, you're
1:27:02going to lower the average quality of
1:27:04politicians in that case. And so, um,
1:27:07and so that's something that that will
1:27:09probably be discussed when when we
1:27:12discuss political selection in in a
1:27:13couple weeks.
1:27:23Shall we move on to the last question
1:27:24then?
1:27:26Great. Okay. Um, Rohan, do you want to
1:27:30ask your question?
1:27:34>> Hi. Um thank you so much for the
1:27:37lecture. Uh I have a completely
1:27:39non-technical question. Um so something
1:27:42that I found helpful in your papers is
1:27:44your discussion of kind of how
1:27:46corruption plays out in practice. So for
1:27:48example over invoicing or um fraudulent
1:27:51procurement. Um, and my question is,
1:27:54what kinds of resources and experiences
1:27:58do you use personally and do you
1:28:00recommend to kind of inform your
1:28:03practical understanding of how
1:28:04corruption plays out? Um, for context, I
1:28:07was in a political economy seminar um,
1:28:09where this question was asked and the
1:28:11presenter said, you know, go watch the
1:28:13mechanism um, or mechanism on Netflix.
1:28:16So, I'm I'm curious about your thoughts
1:28:18and recommendations on that. Thanks.
1:28:23Uh
1:28:25I mean so what what I'm going to answer
1:28:28I think is just a general uh answer not
1:28:31necessarily only to our work but but in
1:28:34general right I think the knowledge of
1:28:36what's going on comes from a mix between
1:28:40reading local news and what journalists
1:28:44talk about
1:28:46uh and also talking to people uh more on
1:28:50the ground both auditor s and and people
1:28:53in governments and things like that,
1:28:55right? So, so I think that's something
1:28:56that in general uh development
1:29:00economists need to do for like any type
1:29:01of mechanism they trying to understand
1:29:03on any paper uh to be honest, not not
1:29:06ours. I mean, in our case, Fred and I
1:29:08spent uh many many afternoons reading uh
1:29:12audit reports and then reading
1:29:13newspapers and then in Brazil talking to
1:29:17people. So, so I think it's is really a
1:29:19combination of all these different
1:29:21things that I think allows you to
1:29:24understand what's going on. In our case,
1:29:26the audit reports were really detailed.
1:29:28So, really helped us in order to
1:29:29understand what exactly were the
1:29:31technologies that politicians and
1:29:34bureaucrats were using to divert the
1:29:37resources, right? But in other settings,
1:29:40you might not have exactly that. You
1:29:42might need to do say focus groups with
1:29:44people and interview say policy makers
1:29:47or interview individuals and then uh try
1:29:50to understand that.
1:29:52>> Thanks a lot.
1:30:01>> I was on mute. Thanks so much everyone.
1:30:04I think we're at time now so we will
1:30:06have to to call it there. But big thank
1:30:10you to Claudio and Fred for coming back,
1:30:12for giving us another round. Um, and for
1:30:16all of you for coming and attending. I
1:30:17just want to flag that next week is our
1:30:20halfway mark. Uh, and a really exciting
1:30:23lecture by Adnan Khan and Asen Kuaja on
1:30:26bureaucracy and implementation. So,
1:30:28please make sure you you join for that.
1:30:30It should be a great one. And uh, we'll
1:30:33see you next week.
1:30:35>> Yes. Thank you very much everyone for
1:30:37for being here.
1:30:39>> Thank you very much. Have a good day.
1:30:42Thanks everyone.

About this transcript

This page contains the full transcript of BREAD-IGC virtual PhD-level course 2026 | Lecture 5 Politicians and Political Agency by International Growth Centre, generated from the public captions YouTube serves with the video. The transcript has 13,587 words across 2,048 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.