YT YouTube2Text

YouTube2Text → Transcripts → Lecture 1: Course Overview by David Atkin and Dave Donaldson

Lecture 1: Course Overview by David Atkin and Dave Donaldson — Transcript

by International Growth Centre · 16,002 words · 2,630 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:51Uh hello everyone. We're going to wait a
0:52few more seconds until we start.
2:01Okay, so um thank you very much for all
2:04joining us today. This is the first
2:05lecture in the BRAC IGC virtual PhD
2:08course.
2:10Uh Dave Donaldson and I are delighted to
2:12have the opportunity to organize this
2:14course
2:16for all of you.
2:17Uh and as hopefully you will know it's a
2:19course on firms and development and each
2:21week there's going to be
2:23different kind of guest lecturers on
2:25topics that are particularly
2:27close to their research expertise.
2:30Today we're going to be giving a kind of
2:32overview and a kind of framework for
2:34thinking about the subsequent seven
2:36lectures.
2:37So a few words on logistics.
2:40So each week about 20 of you have been
2:42chosen at random to participate as
2:45panelists. You'll be able to kind of
2:47raise your hands and potentially ask
2:49questions
2:50as in a normal Zoom call.
2:53The rest of you are
2:55kind of more
2:56traditional attendees of an event and
3:00feel free to ask questions in the Q&A
3:02and in the messages
3:03and either Dave or I will kind of select
3:06some of those questions and ask them out
3:08loud. We'll have time for a few
3:10questions today. As it's a kind of an
3:12overview there probably won't be
3:14that much time for questions but in
3:16later lectures there'll be kind of more
3:18back and forth we expect.
3:22So let me dive straight in and
3:25talk to you a little bit about the
3:27course
3:29what we'll be covering
3:31over the next seven lectures. So
3:33first I'll very briefly
3:35provide an introduction and then I'll
3:36present five facts and fictions about
3:39firms in developing countries and then
3:41I'll hand over to my colleague Dave who
3:44be presenting a simple framework which
3:46we think will
3:47help us kind of organize the content in
3:50this next seven lectures as well as
3:52think about kind of interactions between
3:55different topics.
3:58Okay. So
4:00you know, what's this course about?
4:01Well, you know, I think the big
4:03motivational question is the same as in
4:04many development courses which is why
4:06are some countries rich and some
4:08countries poor. And here we'll be taking
4:10a particular kind of lens on that
4:13which is
4:15you know, focusing on firms. Why firms?
4:17Well, you know, firms large and small
4:19provide employment and income for most
4:22households in the world as well as
4:25producing the goods and services that
4:27take up a kind of lion's share of
4:28household budgets. So you know, surely
4:31firms writ large play a huge role
4:34in
4:35the development process.
4:38You know, a separate but closely related
4:39question is are firm related policies
4:43potentially
4:45tools we
4:47should be using in order to reduce
4:49poverty, boost growth,
4:51increase development. That's a slightly
4:54more nuanced question but you know, the
4:56large role that firms kind of
4:58play in the economy I suggest there are
5:00certainly possibilities
5:03for firm related policies
5:05to play a role.
5:06And if we look at kind of
5:09simple associations in the data it's
5:11certainly true that the richer countries
5:14in the world here I have GDP per capita
5:17in logs on the x-axis and I have the
5:20share of employment in firms from World
5:22Bank Enterprise Surveys on the y-axis
5:25and we certainly see a strong positive
5:27relationship. Countries where
5:30a larger share of firms
5:33workers are employed in firms of size
5:35five or more are wealthier and that's
5:37true also in changes. And so that
5:41at least
5:42without making too much of causality,
5:44this is obviously clearly an
5:45association,
5:47is definitely
5:48supportive of the fact that potentially
5:52the growth of firms is a kind of a key
5:53part of the development
5:55process.
5:56So you know, can policy makers
6:00intervene to generate firm growth?
6:03That's a question we'll be trying to
6:05think about in this course.
6:07If they can, should they be doing so?
6:10And if so, exactly how?
6:13And you know, here myself included many
6:17economists and policy makers
6:19you know, have seen the kind of rapid
6:20growth in East and Southeast Asia
6:25plus the kind of a huge expansion in in
6:27market activity, private sector
6:31employment as well as
6:33market activity more generally and have
6:36kind of linked that
6:37to fairly extensive firm related
6:40policies
6:41in those countries
6:43and
6:44seen that as some evidence that indeed
6:47there is
6:48a kind of a place potentially for firm
6:50related policies and understanding
6:52which ones and how to do them
6:54seems of first order importance.
6:57And so this mini course is going to you
6:59know, take the evidence from kind of
7:02vibrant and growing literature on firms
7:04and development to try and kind of
7:05sketch out the issues, what we know
7:08about them so far and kind of where the
7:10open questions
7:12remain.
7:14Okay, so with that let me dive
7:17kind of straight in. And the first thing
7:19I'm going to do is present kind of five
7:21facts and fictions about firms in
7:23developing countries. Why do I say facts
7:25and fictions? Well, you know, all of
7:27these are kind of things you'll you'll
7:29hear a lot about firms in developing
7:31countries. Some of them we have very
7:33good evidence for. I'll kind of think of
7:34those as facts. Some of them the
7:36evidence is more limited or ambiguous or
7:40more work still needs to be done and
7:42those are some of them may be closer to
7:44fictions. So let me
7:47kind of
7:48dive in and present those five facts.
7:52Okay, so the first fact is one I'm sure
7:55all of you are aware of and that is that
7:58aggregate productivity is much lower in
8:00developing countries than it is in
8:01developed countries. What do I mean by
8:04aggregate productivity? So here's kind
8:06of one version of this fact. We have
8:09GDP per capita on a log scale on the
8:12x-axis and on the y-axis we have total
8:15factor productivity. What is this? This
8:17is a residual when you regress
8:20log output on the capital an economy has
8:23plus the labor it has where you adjust
8:25the labor for the kind of level of
8:27schooling of those workers. And so here
8:29we're thinking about kind of what can't
8:32be explained by
8:34differences in inputs used in different
8:38countries.
8:39And it turns out that this this object,
8:42this aggregate total factor
8:43productivity, explains more than half of
8:46cross-country differences in GDP per
8:48capita. And this this is kind of the
8:51fact that lies behind this kind of claim
8:53that if we are able to increase
8:55aggregate productivity, we'll be able to
8:57go a long way to closing the gap between
9:00the world's poorest countries and the
9:01world's
9:02richest countries.
9:04So you know, this is one of those
9:06facts where you know, obviously
9:09you can debate
9:11exactly how to
9:12calculate aggregate total factor
9:14productivity but I think few people
9:16would disagree
9:18that there's huge differences
9:20in such an object between rich and poor
9:22countries.
9:24Now
9:25just because we can agree on that fact
9:27doesn't necessarily mean we can agree on
9:29why that's true. And broadly speaking
9:32there's two different reasons why total
9:35factor productivity on aggregate might
9:37be
9:38much lower
9:39in developing countries.
9:41The first is that you know, firms
9:44themselves might be less productive. So
9:46think about comparing
9:48two firms which are in some sense
9:50equivalent,
9:51you know, the one in the developing
9:52country might be much less productive,
9:55much less able to turn certain sets of
9:57inputs into outputs
9:59than the
10:01equivalent firm in a developed country.
10:04So that's kind of one
10:06route that could lead to lower aggregate
10:08productivity. The second is
10:11related to the kind of allocation of
10:13resources across firms
10:15and it could be the case that in
10:16developing countries resources are
10:20allocated much less
10:22optimally
10:24or much less efficiently such that
10:26um
10:27the
10:29firms that aren't able
10:31to produce a lot of output with a set of
10:33inputs
10:35have a much larger share of the economy
10:38than firms that are. So
10:40inputs are kind of skewed towards
10:42you know less
10:43productive firms would for example also
10:46lower aggregate productivity even if the
10:50kind of typical firm in the two
10:52countries was equally productive. Then
10:54we kind of think about that as
10:55misallocation which has been kind of a
10:57key object of study in the recent
11:00literature.
11:02Okay. So that's our
11:04kind of first
11:06fact. We're now going to explore those
11:08two channels whether
11:10you know firms are less productive or
11:12whether
11:14least productive firms firms account for
11:16larger shares of GDP. So first are firms
11:19less productive in developing countries?
11:21So this is actually a surprisingly
11:23difficult question to answer because
11:25measurement
11:26is obviously challenging.
11:28You know one approach is you know focus
11:31on countries where we have you know
11:32maybe rich data on
11:35firms including their input use and then
11:37calculate like we did across countries
11:39in the world this residual from
11:41regressing kind of the output of an
11:43individual firm in terms of quantities
11:46on the inputs they're using. And what's
11:48left we're going to call productivity or
11:50TFPQ
11:52and we can kind of compare the
11:54distributions of that across countries
11:55and
11:56Shane and Cleary have famously done that
11:58in in some recent work and there it
12:00certainly does indeed seem that poorer
12:02countries
12:04you know
12:05looking at any point in the distribution
12:09um
12:10firms are less productive in
12:13less developed countries.
12:16Another approach is to try and
12:18circumvent some of those measurement
12:20issues which kind of mean it's kind of
12:22hard to know exactly how much less
12:24productive they are
12:26and instead focus on a kind of single
12:29industry where we're able to better
12:30understand the technology and then maybe
12:33we're able to do a much better job
12:35controlling for the inputs our firms are
12:37using even going as far as controlling
12:39for the machines firms are using. You
12:41know one approach
12:44one paper that kind of follows this
12:45approach is is the work of Gregory Clark
12:47who goes back to 1910 for this
12:50comparison and looks at the textile
12:52industry and what's kind of interesting
12:53there is
12:55firms across many countries in the world
12:56were all using the same machines
13:00essentially the same set of machines
13:01being made in the UK at the time
13:04to produce textiles and then we can
13:06compare
13:09you know how productive workers are on
13:11the same machinery.
13:12And here we see stark differences for
13:15example in the number of looms that were
13:17operated per worker comparing New
13:19England
13:21Canada at the time versus say India or
13:24China there was a factor of four
13:27factor of six difference
13:29between
13:31the number of looms being operated per
13:34worker. So kind of one measure that of
13:36productivity that's kind of much more
13:38direct. And in kind of further work
13:41Clark has
13:42you know updated some of these over time
13:44so we can get some more recent
13:45comparisons.
13:47The one particularly interesting measure
13:49is doffs per hour so a doffer for those
13:52not very familiar with the textile
13:54industry is the person who takes off the
13:58spindle which has been
13:59kind of filled with yarn as part of the
14:01spinning process
14:03and basically takes it off puts it in
14:05baskets and then they go
14:08off to sell those. And that doffer their
14:12task is basically removing things
14:14and taking them away and in India they
14:17were doing say in 1944 to 49 when we
14:21have measures for all three we're able
14:23to doff 124 of these an hour while in
14:26the US a single doffer was taking 770
14:29spindles off. So you see really quite
14:31dramatic differences
14:33in in this case worker productivity on
14:36the same machines across different
14:38countries. So that's
14:39kind of another way where at least where
14:41we look we see potentially large
14:43differences in in productivity between
14:45developed and developing countries.
14:49Another approach is to think about kind
14:51of the technologies the ideas
14:53the management techniques that firms are
14:55using.
14:56There we're going to have to typically
14:57rely on surveys asking firms what
14:59they're using what they're doing.
15:01Recently there's been some progress made
15:04here by the World Bank who has this nice
15:06new set of data on technology use across
15:09a variety of different developing and
15:11and more developed countries and here
15:13what they're going to do is they're
15:13going to get experts to try and
15:16write down you know how sophisticated
15:18different technologies are in particular
15:20industries and then find out what
15:22technologies firms are using and then
15:23create a technology index and you know
15:25this is early days with this type of
15:27work but already there's interesting
15:29patterns that emerge such as the fact
15:31that
15:32the kind of gap between
15:34the kind of more developed countries
15:36like Korea who are at the frontier
15:39and the less developed ones such as
15:41Burkina Faso which are far behind the
15:42frontier that gap is is kind of very
15:44large for the most productive firms
15:47they're the ones at the frontier in
15:48Korea
15:50but the kind of less productive firms
15:52actually the differences in in
15:54technology use is is not so extreme.
15:58Now an important caveat here is you know
16:01the fact that different firms use
16:03technologies doesn't necessarily mean
16:04they're doing something suboptimal
16:07obviously factor prices differ
16:10etc. And it may be optimal to use
16:12different technologies across different
16:13countries and
16:15making welfare statements is going to be
16:16difficult and this is something Dave
16:18will discuss more precisely when he
16:21turns to the framework in a few minutes.
16:24So another approach is you know rather
16:27than something as ambiguous as
16:29technologies where you know different
16:31firms may want to use different
16:33technologies depending on factor prices
16:34for example. Uh
16:37you know
16:38another area of productivity that people
16:40have investigated a lot of detail is
16:42related to management techniques
16:44management technologies and
16:47particularly influential set of work
16:49pioneered by Bloom and Van Reenen had
16:52the ingenious idea of focusing just on a
16:55subset of management technologies which
16:57could be vertically ranked. And what do
16:59I mean by vertically ranked? I mean
17:01policies which
17:03potentially everyone could agree or
17:05techniques everyone could agree
17:08some are better than others
17:11maybe it's not more costly
17:13to do a
17:15better
17:17uh
17:18strategy or a better practice than
17:21another. And so an example of this might
17:23be having a performance based pay system
17:26rather than a flat pay system or
17:28you know organizing your inventory in
17:31kind of a way that means you can easily
17:34find what you're looking for and stuff
17:36doesn't get lost due to
17:41damage. And so once we've kind of
17:44created these types of practices which
17:46can be vertically ranked and you know
17:48it's going to be a little debatable
17:49how vertical some of these are but if we
17:52agree on that then we can then compare
17:53across countries how good their
17:55techniques are and that will allow us to
17:58potentially make some progress here.
18:00And when you do that you see stark
18:02differences in management practices
18:04across countries that line up pretty
18:06strongly
18:07with
18:08GDP per capita.
18:11Okay. So this is all about
18:16you know technologies
18:18management techniques potentially
18:20differing across firms and some of this
18:22will be related to the fact that
18:25firms in developing countries
18:27may be more productive than in developed
18:30countries are more productive than those
18:31in developing countries if we're
18:32comparing you know two equivalent firms
18:35say
18:36doing similar things
18:38across the two.
18:40You know a second question we may be
18:42interested in is
18:45you know how
18:46resources are allocated across those two
18:48types of
18:49across less and more productive firms
18:51and that may also generate differences
18:53in aggregate productivity as we talked
18:55about a few minutes ago.
18:58And this leads to our third fact. And
19:01the kind of first part of it is that we
19:04when we look we see potentially very
19:06large distortions in input and output
19:10markets
19:11both in developing countries and in
19:13developed countries but maybe more so in
19:14developing countries. So what do I mean
19:16by distortions here?
19:19Think about say input markets we're
19:21talking about poorly functioning capital
19:23markets onerous labor regulations
19:26poor institutions that make it difficult
19:28to write contracts that will be
19:29enforced. All of these types of things
19:31will distort firms input use
19:35compared to a kind of a
19:36undistorted economy.
19:38There also may be distortions on output
19:40markets things such as markups or taxes.
19:44Um so you know do we know that
19:48these distortions are larger in
19:49developing countries than developed
19:50countries? Certainly there are many
19:52literatures
19:54that uh uh
19:55kind of assume that is true and provide
19:58various pieces of evidence in support of
20:00that. In a recent handbook chapter, Dave
20:03and I have tried a
20:05extremely heroic in the sense that we
20:07have to make assumptions that are
20:08likely not to be true, but allow us
20:12to make progress and we hope are not too
20:13far
20:16from the truth. Uh
20:18and
20:20what we're going to do is we're going to
20:22take World Bank Enterprise surveys, use
20:24kind of quantitative responses in those
20:26surveys
20:27to try and calculate wedges across
20:31both inputs
20:32and outputs and various different types
20:34of wedges. And by wedges I mean anything
20:37that drives a gap between the price a
20:39buyer pays and the cost
20:42to the seller.
20:43And we can kind of think about kind of
20:46wedges created by regulations or the
20:48need to pay bribes to get things done.
20:50Uh wedges due to crimes
20:53such as kind of theft of of of output or
20:56inputs. Uh
20:58wedges or distortions due to markups,
21:00domestic taxes as well as input
21:02distortions whether they come from
21:04tariffs on input imports or or capital
21:07market distortions raising interest
21:09rates or or or or rationing credit or
21:12labor market distortions through
21:14regulation etc.
21:16And when you go and
21:18try and do that it does
21:20jump out these distortions do appear to
21:22be larger particularly the input
21:24distortions
21:25regulation crime
21:28um
21:29uh output distortions as well appear to
21:32be larger in developing countries than
21:34in developed countries, but you know,
21:36it's very difficult to measure these
21:37objects
21:39and so they should be taken with a pinch
21:40of salt. We can also do this across
21:42different types of firms as well.
21:45If you pick a particular distortion and
21:47a particular country, you can often go
21:48further. Uh
21:50for example, some nice recent work has
21:52looked at contracting frictions
21:55which arise when it's
21:57in poor institutional environments it's
21:59different to difficult to have contracts
22:02in force. Uh and what they do is they
22:05look at across Indian states which have
22:07different qualities of institutions and
22:09in particularly
22:11have different kind of backlogs in their
22:13court systems on how long it's going to
22:15take for a case to actually come to
22:17trial and they show that in states with
22:20more congested courts then firms are
22:22using different bundles in order to make
22:24products to make the same products and
22:27in particular the bundles of
22:28intermediates they're using tilted
22:30towards standardized intermediates which
22:32you can buy on markets and don't require
22:34relationship specific contracts for
22:36example
22:38versus more kind of complex inputs. So
22:40here's an example of input use being
22:42distorted as a result of a particular
22:47contracting friction that we can
22:48identify.
22:51Uh we can also look at various output
22:53market distortions. Here as I mentioned
22:56the evidence that these are more severe
22:58in developing countries is
23:00much more nuanced. In fact, I would say
23:02we
23:03have no answer to that question yet and
23:06so this would fall into one of the kind
23:07of fictions despite the fact you often
23:09hear that
23:11developing countries have
23:13you know, very
23:14uh
23:18uncompetitive markets either because of
23:20crony capitalism
23:22or because of
23:24huge rents going to intermediaries and
23:27such like. Uh when we actually try and
23:29measure markups for example following
23:31the method pioneered by Bob Hall and
23:33more recently Jan De Loecker
23:36the cost minimization approach then we
23:38see
23:39huge dispersion in markups across
23:41countries and when we're attempting to
23:43do a cross country comparison if you
23:44look at the African data here, you know,
23:46you have two neighbors who have both
23:49some of the lowest markups in the world
23:51and some of the highest markups in the
23:52world and it's hard to take these data
23:55too seriously.
23:57Uh you maybe can make more progress
23:59again by focusing on particular
24:00industries
24:02where the measurement challenges are
24:04slightly ameliorated.
24:08Uh and here's some recent work's been
24:09looking at cement. Cement's obviously an
24:12important input in the construction
24:14industry which
24:16if
24:17there's kind of big distortions due to
24:18markups will obviously affect
24:19intertemporal investment decisions. Uh
24:22and
24:23here we we see that cement prices as
24:26well as other important inputs
24:29much higher in in the poorest countries
24:31than they are in in wealthier countries.
24:33And this is you know, at first blush
24:34surprising because we might think that
24:37in fact
24:39uh
24:39input prices should be lower in the
24:41poorest countries since
24:43you know, labor's an important
24:45input in in cement manufacturing these
24:47products
24:48and labor's particularly cheap in the
24:49poorest countries,
24:51but in fact as you see the prices are
24:53higher in the poorer countries
24:55for these goods.
24:57So you know, obviously the first thing
24:59we might want to know is is this markups
25:02a distortion or is this just higher
25:04cost? Maybe the other ingredients in
25:06cement much costlier in developing
25:08countries.
25:10And if it is markups for policy reasons
25:13we might want to know if is this due to
25:15kind of uncompetitive contract conduct
25:18the erection of barriers to entry by
25:21you know, politically connected firms
25:22for example, that's something you might
25:24be able to
25:25do something about or is it simply the
25:27fact that these markets are quite small
25:30and so particularly if there's kind of
25:32high fixed cost industries like making
25:34cement, there may not be room for many
25:35firms and that's naturally going to
25:37raise markups. And some recent work has
25:39shown
25:40that maybe that latter explanation is
25:42true although there are also high costs
25:44in these industries, but again this is
25:46very early days as very very little we
25:48know about the level of competition in
25:50developing countries, the level of
25:52markups and
25:53this is kind of one of these areas where
25:56we may need much more work.
25:59Okay, so up to now we've talked about
26:01the fact that the level of distortions
26:02might be much larger in developing
26:04rather than developed countries. But
26:07in order to kind of dramatically bring
26:09down aggregate productivity, often what
26:12matters more is not the level of
26:13distortions, but how those are
26:16distributed across different firms. And
26:18so for example, you know, a tax is a
26:20distortion and if every firm faces the
26:23same tax, that's not going to
26:25misallocate resources across different
26:27firms. It's only if different firms
26:30face different tax rates that we may get
26:33this effect we talked about where
26:36less productive firms are allocated
26:38larger kind of shares of the market or
26:41allocated more resources which leads
26:43them to have larger shares of the market
26:44which brings down aggregate
26:45productivity.
26:47So what do we know about whether or not
26:49distortions vary across the size
26:51distribution or vary across other
26:53dimensions. Here I'll mainly be focusing
26:55on the size distribution.
26:57So
26:58you know, a classic work is the work of
27:00Jim Tybout. So this is a
27:02uh
27:03as far as I know the only really
27:06comprehensive survey on firms and
27:08development although it's somewhat dated
27:10now.
27:11And in it he documents something called
27:13the missing middle which is you know,
27:15something people have talked about
27:16previously. And here the idea is that
27:18there's kind of many many small firms in
27:20developing countries. There's also a
27:22fair number of big firms. What's really
27:24unusual is the lack of firms in the
27:26middle of the distribution. And what
27:28might be behind that? Well, one
27:30kind of
27:31two sets of explanations that would
27:33generate that is both the presence of
27:35high regulatory barriers and a lot of
27:37distortions
27:39in the formal sector and that means that
27:42these medium and large firms are kind of
27:44subject to you know, high taxes
27:46um
27:48a lot of regulatory burdens etc.
27:51Um
27:52and so that's naturally going to mean
27:54there's going to be a lot of small
27:55firms, but at the same time big firms
27:58are kind of big enough to get the
27:59attention of politicians to benefit from
28:01cheap credit
28:03subsidies of other types or through
28:05generating political connections and
28:07then and handouts etc. And so in that
28:10case you'll see kind of a lot of small
28:12firms, but also
28:13some big firms doing well and it's those
28:15firms in the middle
28:17who are
28:19really
28:20suffering from these distortions and so
28:23we'll find this missing middle. And why
28:25do we care? Well, of course if we're
28:27trying to design policies to help firms,
28:30which firms we decide to help is going
28:31to depend critically on where these
28:33distortions
28:35are largest and
28:37this evidence would suggest we'd want to
28:39target firms in the middle of the
28:40distribution. But let me return to these
28:42size dependent distortions in the
28:43missing middle in a few minutes
28:45before thinking of this is something of
28:48a fact as you'll see it will be
28:51disputed.
28:53A second large literature
28:55closely related to size dependent
28:57distortions is asking this question why
29:00are there so many small informal firms
29:02in developing countries?
29:04And here LaPorta and Shleifer offer kind
29:06of three views that have been very
29:08influential. The first view they call
29:10the romantic view is that there are lots
29:13of small firms because
29:15uh you know, essentially it's all these
29:17regulations and difficult difficulty
29:19uh borrowing
29:21credit frictions etc. Uh and uh
29:25those essentially mean that high quality
29:27entrepreneurs
29:28are unable to grow their firms and
29:31instead remain small. And this is an
29:32idea closely associated with Hernando de
29:35Soto.
29:36And in that in that world, you know, the
29:38most distorted firms are these very
29:40small firms who find it say particularly
29:42hard to borrow
29:43and you're going to want to go and help
29:45those firms
29:47in order to
29:48boost aggregate productivity.
29:51A second
29:53kind of view is a parasite view. And in
29:56that view,
29:57these small firms are actually very
29:59unproductive. They're not great
30:00entrepreneurs. In fact, that they're not
30:02very good entrepreneurs. But the reason
30:04they're able to survive is because
30:06of the taxes and the regulations that
30:08bigger firms face, and they're able to
30:11avoid those by staying small under the
30:13radar
30:14or informal and avoiding
30:17those
30:18costs. And in that world, it's the
30:20opposite. It's kind of the bigger firms
30:22who face the most distortions, and you'd
30:24want to help those rather than the small
30:27firms. In fact, uh the small firms
30:28should be driven out if you kind of
30:31reduce the distortions on the bigger
30:33firms. So, you see very, very different
30:34policy prescriptions from these two
30:37theories. And then there's a third, a
30:38dual economy theory, where there's kind
30:40of two sectors, a modern and a
30:42traditional sector. And this traditional
30:43sector's more like a buffer. It offers a
30:45kind of insurance, a source of
30:47subsistence
30:48and no growth.
30:51Okay, so you know, what's the evidence
30:53for these different stories and for
30:54these size-dependent distortions? Well,
30:56surprisingly, a little evidence. But one
30:58recent paper that's instructive here is
31:01the work of Shen O'Connell. They do a
31:03couple of things, but what they're going
31:05to do is is like Shen O'Connell is focus
31:07on three countries where we have
31:09very good data that kind of spans the
31:11entire firm size distribution. And
31:14they're going to focus on India,
31:15Indonesia, and Mexico.
31:17And what they're going to show first is
31:19that in those countries at least, there
31:20is no missing middle. There's kind of a
31:23non, you know,
31:26non-bimodal distribution. And
31:30their their argument is that actually
31:32those kind of facts and type outcome
31:34from basically
31:36kind of poor data coverage, where you're
31:37often missing certain types of firms
31:40when you're combining various different
31:41data sets. That's kind of the first
31:43takeaway, which is maybe we shouldn't be
31:45helping these in the middle or focusing
31:48on them because there doesn't appear to
31:49be this missing middle.
31:52The second takeaway is, you know, what
31:54do we think about that romantic view? In
31:56that romantic view, it's these kind of
31:58small firms who are most constrained. As
32:01Dave will discuss more precisely, what
32:04we're looking for here to know if the
32:05small firms are most constrained is
32:07whether they have higher value marginal
32:09products than bigger firms.
32:12And it's very hard thing to kind of
32:13measure the value marginal product, but
32:16Shen O'Connell looks for the average
32:18product of both capital and labor in
32:22with different kind of firm size across
32:25these three countries.
32:26And at least looking at average
32:28products, we find the opposite, which is
32:30that the bigger firms have higher
32:31average product of capital, say value
32:33added per worker.
32:35It's the bigger firms who on this
32:37dimension of fear appear more productive
32:40on average. Now, of course, they may not
32:43be more productive on the margin, and
32:45that's what really matters here for
32:46knowing whether or not we we should be
32:48allocating more resources to them.
32:50And so again, this is one of those
32:53kind of facts which is very much
32:56uncertain and ambiguous until we have
32:58kind of some better measures of value
32:59marginal products as opposed to average
33:01products.
33:03They also shed some light on the
33:04parasite view. And that parasite view
33:06says we should be helping the kind of
33:07most productive firms because it's these
33:10kind of least productive firms
33:12that are kind of surviving only because
33:14they're avoiding regulation.
33:16And if that's true, then what we'd
33:17expect to see is kind of large
33:20a bunching of firms below regulatory
33:22thresholds if indeed many firms are
33:27staying alive because they're avoiding
33:29regulation. And at least when we look at
33:31labor regulations that bind at 100 firms
33:33in India or or VAT thresholds at various
33:36points in Indonesia or tax regimes in
33:38Mexico, we don't see this bunching,
33:41which would at least be supportive of a
33:43parasite view.
33:45So, kind of the takeaway here is, you
33:47know, I think there's still a lot of
33:49ambiguity here on which types of firms
33:51are most distorted and where policy
33:53should be kind of focusing. And we'll
33:56kind of talk more about that later in
33:58the course.
33:59So, my final fact before I hand over to
34:01Dave is related to industrial policy.
34:05And in particular,
34:08related to kind of policies which are
34:10there to target externalities. You may
34:12want to help particular sectors because
34:14we think those sectors
34:16generate a lot of learning that spreads
34:18to other firms or to other sectors.
34:21Or you might want to target particular
34:22firms, which again have large
34:24spillovers. Or more generally,
34:26industrial policy could also be targeted
34:28at just the correcting those distortions
34:30we talked about earlier.
34:33So,
34:34when thinking about industrial policy,
34:36certainly there was a
34:38consensus that built up amongst
34:40economists in the '80s
34:42and the '90s
34:43that we really shouldn't be encouraging
34:46it
34:48as economists. And in fact, we shouldn't
34:50be doing it all. And this often,
34:53you know, this was part of what became
34:54termed as the Washington Consensus. And
34:57the logic here was twofold.
34:59First,
35:00what is there's an informational
35:01constraint. It's really hard to
35:03understand which firms you want to you
35:05want to help. It's hard to know which
35:07firms generate the most the most
35:09externalities, even which industries
35:11generate the most externalities.
35:13And it's hard to work out which firms
35:16are most distorted. As we talked about,
35:17we still don't know whether big firms or
35:19smaller firms are more distorted, let
35:21alone particular industries or sectors.
35:24And so that's kind of the first, an
35:25informational constraint.
35:27The second
35:29is that in practice, many of these
35:30policies seemed like they ended up being
35:32captured
35:35by politicians in order to reward firms
35:38who who maybe helped them in other ways.
35:40And because of this capture,
35:43essentially we should tie our hands and
35:45not do these policies at all because
35:47they just end up being corrupted.
35:50So, this was certainly the conventional
35:52wisdom amongst economists. There's been
35:53some pushback over the years, maybe most
35:55notably from Danny Rodrik. And you know,
35:58his argument is that, you know, these
36:00types of issues that it's, you know,
36:02information is hard
36:04and they're subject to capture,
36:06you know,
36:07plague many other policy areas, maybe to
36:10a different extent. But say for example,
36:12education policy, health policy. And in
36:14those cases, we don't hold up our hands
36:16and say we don't know what to do.
36:18We should do nothing. Instead, we kind
36:20of try and kind of muddle through the
36:22best we can because we think they're too
36:23important to ignore. And that would
36:26certainly be his view and, you know, one
36:27I I think I subscribe to as well.
36:30The kind of second point
36:32that he'd make is that, you know, a lot
36:33of the reason that we kind of shied away
36:35from industrial policy was this idea
36:37that previous industrial policies had
36:39been abject failures, particularly those
36:41in India or Brazil with things like
36:43import substitution strategies.
36:45And you know, his response to that is
36:47it's so difficult to infer causality in
36:49these cases, particularly since they're
36:51often, you know, helping, you know,
36:53firms that are most distorted, we may
36:55expect to grow the slowest absent any
36:57policies.
36:59And so,
37:00you know, it's really hard to know
37:01whether or not the the policies that
37:02were put in place historically did
37:04actually
37:05lead to benefits or not. And again, this
37:07is something we're starting to revisit
37:09and and and Reka's going to talk a lot
37:10more about in later lectures.
37:13You know, one thing I think it's less
37:14debatable is that everyone is doing
37:16industrial policy.
37:18Countries didn't stop doing it just
37:19because economists
37:21told them not to.
37:23And, you know, in that light, we may
37:26want to think about, you know, telling
37:28them how to do it better
37:30if they're going to do it anyway. And,
37:33you know, here there's kind of new
37:34evidence which is on this
37:38picture here
37:40from Reka and co-authors trying to
37:42document
37:43the types of policies different
37:45countries are actually doing over time.
37:48Okay, so those are my kind of five facts
37:51and fictions.
37:52Obviously, some things we're more sure
37:54of than others.
37:56And now,
37:57we'll kind of turn to Dave, who'll give
37:58a kind of overview
38:00of
38:01the
38:03kind of course through setting up
38:06theoretical framework, which will kind
38:07of guide the next seven lectures and
38:09allow you to start piecing together some
38:11of these facts as well and putting them
38:13in various boxes.
38:15Okay, so let me undo my share and then
38:17I'll hand over to Dave.
38:25Thanks.
38:27Hope you can all hear me.
38:30And see my screen.
38:36So, um
38:38yeah, my
38:40my goal here is to in the remaining time
38:43to uh
38:45talk about, you know, um
38:48like a
38:49an organizing framework, I guess you
38:51could call it, sort of the the the bare
38:52bones of an economic model
38:55that allows us to think about sort of
38:57how to label different concepts that you
39:01know, that appear in the study of firms
39:03in the context of developing countries
39:04and have already appeared in David's
39:06discussion and will kind of continue to
39:08appear in in the coming lectures. And
39:10so, kind of consider this a preamble,
39:12lots of
39:13lots of details to be filled in by the
39:17the lectures that follow,
39:19but something that hopefully kind of,
39:20you know, wets your appetite at at the
39:22very least.
39:23Um
39:25So, you know, we
39:27we're going to sort of, you know, go
39:29fairly
39:30abstract and and fairly general at this
39:32point. I mean, you know, my the way I
39:34think about that is that the generality
39:36is useful because we kind of want to
39:38want to understand the limits of
39:39arguments, you know,
39:42the assumptions, the minimal assumptions
39:43that are needed to make arguments.
39:45Hopefully, that's the value of
39:46generality. Abstract, you know, I find
39:48it useful because it's sort of it's a
39:50sort of means it's a fairly flexible,
39:53you know, way of thinking about the
39:54world and and that can sometimes be a
39:57curse. It's maybe too flexible, but it
39:58can also sometimes be,
40:00you know, lead yield powerful insights.
40:03And so, for my my own thinking, I find
40:05that that a useful feature.
40:07Anyway, so we're going to imagine that
40:09there's
40:10there are these firms
40:12indexed by the letter I and they each
40:15kind of produce a single good.
40:17You know, we can we can embellish that
40:19and talk about multi-product firms and
40:21things, but today we won't. But these
40:24goods, you know, they they combine to
40:26produce GDP or development, you know,
40:30in some narrow sense of the word
40:31development. And so, one way of thinking
40:34about that is that they enter some sort
40:35of aggregate national utility function.
40:37So, there's sort of a vector of goods
40:40whose quantities are consumed, this
40:41vector big Q here, bold Q. And
40:45obviously, focus on national utility
40:46ignores distributional considerations,
40:48you know, people might have different
40:49preferences, different incomes, that
40:51might be bad.
40:52But but, you know, the the main focus of
40:56um
40:56of a lot of this literature so far has
40:59not been on such distributional
41:00concerns. It's been on sort of aggregate
41:02productivity, you know, sort of the the
41:03question David started with, which is
41:05sort of why are nations
41:07richer than others. And that's
41:09inherently a national aggregate
41:11question.
41:12Okay, so then sort of how are these Qs
41:14produced? Well, we're going to sort of
41:15think very abstract here about them
41:18being produced in the following way. So,
41:21so, you know, this is a this is some
41:23function that's going to map um
41:26inputs,
41:28which are these kind of Xs.
41:30That's those bold Xi is the inputs that
41:33this firm is using in this function, you
41:37know, some some arbitrary vector X. And
41:41this thing A that I'll come back to in
41:43just a second. And we want to sort of
41:44think about this this this this function
41:47F is actually kind of being derived from
41:49some other function that I'll call sort
41:50of F tilde here, which um
41:53is actually the firm is maximizing over.
41:55So, this is not your usual
41:56profit-maximizing decision that that we
41:58can talk about in a bit, but this is
42:00almost like a sort of a trivial
42:02precondition for maximizing for for
42:04profit maximization, which would be that
42:06this firm sort of thinks to itself,
42:08well,
42:09given that I'm going to put in X inputs,
42:12like let's just be concrete about that
42:14for a second, imagine like one worker
42:17and that's it, no machines, just one
42:19worker. What is the best way of
42:22organizing production conditional on
42:24using only one worker to produce my
42:26output? So, that's what this function
42:28here says.
42:29So, we're going to maximize over like
42:31ways of producing, quote unquote, like
42:32or techniques as I'll call them here,
42:34these these Ts.
42:36Now, of course, there's kind of probably
42:37some constraints on the techniques that
42:39are available and that's what this says
42:41here. You're maximizing over the Ts
42:42subject to the the Ts being in some set
42:45that I'll call Ai.
42:47Okay, so
42:48you can see this is this is a fairly
42:50general way of thinking about things.
42:51So, you know, there's there's there's
42:52there's a this firm right now knows only
42:55about the set Ai. It doesn't know, you
42:58know, how else to produce conditional on
43:01using one worker. If if it decided to
43:03use one worker and one machine, then of
43:05course, different techniques would
43:06probably be useful for for the right way
43:11to maximize output given using one
43:13worker and one machine. And all and all
43:15of that all of those discussions are
43:16kind of embodied in this.
43:18So, anyway, we're going to ignore the Ts
43:20for most of what follows, you know,
43:22you know, today. However, you know, in
43:24subsequent lectures, things that might
43:25look like Ts, you know, kind of will
43:27turn up and they're important because
43:29they're they basically give rise to what
43:31this function, the F function, is. You
43:33know, this is where
43:34firms figure out the right way to sort
43:36of turn inputs into outputs, the best
43:38way to turn given inputs into outputs.
43:41Anyway, so so from what follows, we're
43:43mostly going to focus on sort of this
43:44function, which is where, you know,
43:45there's inputs that go in. How many
43:47inputs go in? We're going to talk about,
43:49but for now, just imagine there's an
43:50arbitrary vector of inputs going in. And
43:52then Ai is this thing that indexes
43:54basically like how big is the set, how
43:56good is the set of techniques that this
43:58firm knows about in terms of making its
44:00output.
44:02Okay, so a crucial thing in today's
44:04discussion is going to be this at the
44:07bottom here. So,
44:09you know, we're imagining that these
44:11inputs, these Xs, are
44:14in fixed supply at the national level.
44:16So, that's what this bottom constraint
44:18here says. It says that when you sum
44:19those little Xs
44:21across for a given type of input M, so
44:24by type of input, I mean like labor, for
44:27example, or, you know,
44:29rubber
44:31rubber gaskets, the given type of
44:33material input. For a given type of
44:35input, you know, when you sum those
44:37input usages across the firms I, they
44:40have a limit, which is this big XM. You
44:42know, so for the type of input M, but
44:44there's a limit on nationwide how much
44:46of the thing we have. So, this is going
44:47to be kind of crucial. We'll talk later
44:49about, you know, relaxing it, like where
44:51does big X come from? But discussions of
44:54aggregate productivity, typically like
44:56the ones David's already shown you,
44:57typically hold big X constant. They sort
44:59of ask, given the workers we have as a
45:02nation, how much in sense develop or GDP
45:05or U can we produce? So, they sort of
45:07they do already condition on aggregate
45:10sort of input availability.
45:12Okay, so, you know, just to sort of say
45:14a few more more notes about that
45:17previous discussion. So, you know,
45:18techniques didn't
45:20there was no price of using techniques,
45:23right? So, they were just ways of
45:24producing. Right? So, and remember,
45:27think about this, we're conditioning on
45:28the Xs. So, so often when you think
45:30about ways of producing, you're thinking
45:31about should I use two workers or one
45:34worker and a machine? That's not
45:36techniques in this language. Techniques
45:38is like conditional on using one worker
45:40and one machine, how best should I do
45:41it? Okay.
45:43Now, of course, as I already mentioned,
45:44there's the the the the the
45:46there's this sort of set of constraints
45:48A on the techniques the techniques of
45:50the firm knows how to use. I'll come
45:52back to that kind of constraint in a
45:54second, but
45:55Okay, so then when we think about a
45:57technology, you know, it's it's really a
45:58recipe for mapping
46:01inputs into output, right? So, so that's
46:04that's going to be our definition of a
46:05technology here. And we're indexing
46:08those technologies both by by F at the
46:11Ai, you know, so so
46:13so, you know, obviously, at this level
46:15of generality, the F is not restricted,
46:18you know, it can be different across the
46:19firms because the Ai, remember, is
46:21indexing a set. So, things since that
46:22can be different, you know, everything
46:23can be different. It can offer kind of
46:27flexible returns to scale. It can offer
46:29non-homotheticities, which is to say at
46:31like low scale, you might
46:33in some sense produce more using a
46:35labor-intensive mix and at high scale,
46:37you might use produce more using a
46:38capital-intensive mix. All those sorts
46:40of things are totally possible. But an
46:43crucial thing to note about A is that A
46:45kind of by construction is going to be
46:46an index of like how good the function
46:48is in the sense of like holding input
46:50constant inputs constant, if A goes up,
46:52so does Q kind of necessarily. Okay, so
46:56now just to reiterate, you know, this is
46:58nothing to do with
47:00how many Xs will go into this firm, you
47:02know, which Xs will go into this firm,
47:04will the firm be big or small, will it
47:05be capital intensive, et cetera. All
47:07those are great questions, we're going
47:08to come back to them, but but for now,
47:10just conceptually, those are not the
47:12technology. The technology is the menu
47:14of options available.
47:16Okay, so, you know, just finally to to
47:19reiterate, you know, the distinction
47:20between techniques and inputs is
47:22crucial.
47:23Inputs are constrained at the national
47:25level. You know, they they remember I
47:27told you about this crucial constraint
47:29aggregate constraint here.
47:31So, if one firm's going to use
47:33a bit more of an input, some other firm
47:35has to use less of the input. There's no
47:37way of getting around that aggregate
47:38constraint. However, techniques are sort
47:40of free in the sense that, you know,
47:41they're again, they're constrained by
47:43the Ai set, but otherwise, they're free.
47:45So,
47:46so we would sort of hope that,
47:48you know, that they're being used when
47:50they're kind of appropriate, when
47:51they're worthwhile. The good techniques
47:53are being used up to the constraint.
47:56Okay, so so then that begs the question,
47:59well, what does A capture? Right? So,
48:01remember, we're we're imagining that A
48:02is not something the firm can do
48:04anything about. It was a constraint in
48:06the firm's problem. Hiring more labor is
48:08not going to raise your A in the way we
48:10wrote it down.
48:11Hiring more labor does let you produce
48:14more
48:15holding A constant, but it doesn't sort
48:17of A would be like even with holding
48:19labor constant, I can produce more. So,
48:22think about A as like things that are
48:23sort of, you know, they of course have
48:25to obey the laws of physics, the laws of
48:26thermodynamics are inside A. There's
48:28only sort of, you know,
48:30there's only so much output we can get
48:32for a given amount of input.
48:34But also in a, you know, obviously in
48:36some sense, they're also going to have
48:37to reflect limits of scientific
48:39knowledge. Maybe sort of nobody knows
48:42how to, you know, squeeze more um
48:45uh
48:47you know, tons of wheat out of an acre
48:49of land than
48:50than than a certain thing that's
48:52embodied in A.
48:53Finally, however, it also appealed to
48:56two other very important notions that
48:58might make Ai differ across firms and
49:00this is what David has already alluded
49:02to. Um
49:03It would be sort of just we could start
49:05with ignorance. So, you know, remember,
49:07Ai indexes the set of techniques that
49:09the firm knows about. So,
49:11so you could imagine that, you know, for
49:13some reason that is interesting and we
49:15should understand and and try to,
49:18contribute to, but for some reason, firm
49:21I just doesn't know about a wider, you
49:24know, the wider set of techniques that
49:26some other firm, not I, kind of does
49:29know about,
49:30even if they were producing the same
49:31thing.
49:32So, that's an interesting hypothesis
49:34that sort of ignorance or knowledge
49:36limitations, variations across firms,
49:38across countries is is an important
49:41feature here. And and second, I'd also
49:43point to kind of in some sense mistakes
49:45or you might call irrationality, you
49:47know,
49:48if you wanted to embrace the possibility
49:50that that the firms are not set of
49:52maximizing over techniques, that's okay.
49:55Just give them a shrunken A. You know,
49:57just basically sort of define A as the
49:58ones that they do maximize over for the
50:01for the purpose of this discussion. So,
50:03you know, think about that as kind of
50:04like we know about some
50:06but for whatever reason we're
50:07deliberately not using them even though
50:08they would literally let us produce more
50:11with less or more with the same inputs,
50:13you know, nothing to do with approximate
50:15profit maximization, just to do with in
50:17some sense being on your own production
50:19possibilities frontier as a firm.
50:21We usually assume they are. This is just
50:23allowing for the possibility that
50:24they're not and indexing A by the
50:26possibility that they're not.
50:28Um
50:29Okay, so those are those are A's. Now,
50:31let's go to the X's in this function.
50:33Remember the X's are super different
50:36because um are
50:39they're in constrained national supply.
50:42They're they're they're they're kind of
50:43costly. They're scarce resources. And
50:46so, when you think about what's an X and
50:48what's an A, I find this very helpful
50:50because you know, you know, again,
50:52remember the crucial thing about X's is
50:55they're in fixed national supply, which
50:56of course means that if one firm is
50:58going to get more of an X
51:00of a given type of X then another firm
51:01has to get less of that type of X.
51:03There's no way we can break that
51:04constraint that's embodied here. So, you
51:07know, that's of course how economists
51:08have traditionally thought about
51:09capital, labor, human capital, maybe
51:12even types of human capital we could
51:14call managers, management services, that
51:16kind of thing. It's often how we think
51:18about material inputs, you know,
51:19intermediates that are not necessarily
51:21material, but like, you know, legal
51:23services that firms hire are are
51:25intermediates. Um but, you know, I I
51:28encourage you also to think about other
51:30things the firms use as inputs that are
51:33in some sense, you know, r-
51:35you when they use when they use them as
51:36inputs, they're using up society's
51:38scarce resources. Um
51:40like R&D.
51:42You know, that's using sort of a
51:43particular type of time
51:45to to try to produce an input that the
51:48firm will use.
51:49Um search, you know, we often think
51:52about that as as maybe separate, but
51:53it's not. It's, you know, the firm is
51:55using its time to sort of search for, I
51:57don't know, customers or suppliers or
51:59workers.
52:00There's marketing and advertising
52:01expenses, which is the firm using its
52:03resources to
52:05um to try to
52:06tell the world about its products or
52:08um
52:09that sort of thing. Transport costs
52:11would very much belong here. And then
52:13finally, I'd make a comment that, you
52:14know, nothing here, even though there's
52:16no time subscripts on anything, nothing
52:17here is inherently static. We just would
52:20need to let the X's kind of index
52:23you know, inputs at different points in
52:25time. And then that allows one to think
52:27about phenomena like, you know,
52:28adjustment costs, hiring, firing costs,
52:31learning by doing inside the firm, where
52:33we know that if we sort of put in a lot
52:35of input in the past, we probably are
52:38you know, can produce more today. That
52:40would all be inside this kind of Q as a
52:41function of X function.
52:44Um all right, so so so with that all
52:46said, we can now ask them how this
52:48relates to development. So,
52:50um so one way to think about development
52:52is to ask, well,
52:54what's the best we could do
52:56as a society as this sort of
52:58hypothetical nation in this model.
53:00Um and that's of course a well-studied
53:02economic problem that goes back to um
53:05uh
53:06Hotelling,
53:08uh Samuelson wrote wrote seminal pieces
53:10on this. Uh you know, the the planned
53:12economies of the um
53:14of the
53:16of the last century used these results.
53:18Mathematicians developed these results
53:20to help those planned economies try to
53:23achieve the answer to this question,
53:24which is what's the best that we could
53:26do
53:27in this environment, right? So, so of
53:29course that's like maximizing the
53:31national utility, where the choice the
53:34arguments here the Q vector, how much of
53:37apples do we want and how much of
53:39oranges do we want, etc. Those are all
53:40inside this Q vector.
53:42Subject to the limits of technology,
53:44which this planner that's solving this
53:46problem can't do anything about. You
53:48know, they can't change the F, they
53:50can't change the A. Um so they they
53:52cannot change the technology. They have
53:54to obey this law of technological kind
53:57of life, that whatever we produce has to
53:59be no bigger than F at the inputs we put
54:02in
54:03for every firm,
54:04every good. And then finally, also that
54:07other law of life, which is that the
54:08inputs are on aggregate supply
54:10constrained in aggregate supply, this
54:12one here. And of course, we don't want
54:15it doesn't make sense that inputs go
54:16negative. So, that's what this
54:17constraint is doing.
54:19Um
54:20Okay, so of course embodied in this
54:22maximization problem is the question
54:24basically, you know, that David was
54:26alluding to, which is like, should that
54:28firm over there be bigger? Should that
54:30firm over there be more capital
54:31intensive? You know, I I was in the Q&A
54:34a lot of questions like that were coming
54:35up and and those answers to this
54:38question are, you know, embodied in this
54:40admittedly abstract, but sort of, you
54:42know, general framework. Anyway, so as
54:45you probably have seen before, some of
54:48you, you know, the the solution to this
54:50is not hard to write down.
54:52The first order conditions look as
54:54follows.
54:56The you know, basically for
54:58the basic it's it's crucial distinction
55:00is between sort of things that are
55:03happening, let's call them. So, sort of
55:05inputs of type M that are actually going
55:08into firm I in positive amounts,
55:11strictly positive amounts. So, let's
55:13call that sort of the active firms and
55:15inputs kind of set. Um so for
55:19I's and M's that are in the active set,
55:21then the following should be true, which
55:23is that sort of the
55:25marginal utility that's created from a
55:28little bit more of the output of that
55:30good I
55:31times the marginal productivity of
55:34putting a little bit more of input M
55:35into firm I. So, you know, in order to
55:37get more of that output QI, that that
55:40that sort of
55:42you know, pair of products there, dUDQ
55:45and dFdX
55:46should be equal to a constant, which
55:48I'll just call lambda star.
55:51Where the constant, notice it does not
55:53when I say constant, it does not have an
55:55I on it.
55:56So, within any M, like sort of for
55:58example, within the type of input M I'm
56:01calling labor,
56:03the what this of course embodies is is
56:05sort of that every firm has this every
56:08active firm that's using some labor has
56:10the same dUDQ times dFd labor. All
56:14right, that's what that says. So, I'll
56:16put more economic intuition on that in a
56:17second, but but for now that's what the
56:19math says. And then, you know, what
56:21about those firms that are not those
56:23sort of firms and input types that are
56:24not in the produced or active set,
56:27that's the ones down here. Well, the the
56:29statement here is just that those are
56:31characterized by the fact that
56:34you know, in some sense the dUDQ times
56:36dFdX of those firms,
56:38you know, at that amount of zero,
56:40basically, has to be less than that
56:42aforementioned constant. So, it's kind
56:44of like saying, you know, for the firms
56:46that are producing using a given type of
56:48input, equalize their utility
56:50productivity. For the firms that are not
56:52using that thing, make sure their
56:54utility productivity, if they were to
56:56produce, is strictly less than that
56:59value that that is equal across the ones
57:01that are producing.
57:03All right, so so, you know, that is the
57:07the best allocation. Um
57:10the you know, what good is the best if
57:13if the real world isn't going to achieve
57:15the best.
57:16But, you know, I find it helpful to
57:17think about comparing the best to what
57:20actually happens in order to think about
57:22what went wrong.
57:24So, so this is going to let us kind of
57:26think about the word word kind of um
57:29misallocation that I'll come back to in
57:30a second, but
57:32but let's just first talk about the best
57:34allocation. Uh sorry, the the actual
57:36allocation and how it compares to the
57:37best one.
57:38Um okay, so to talk about the actual
57:40allocation, this is just going to be
57:42like a label. So, let's just call the
57:45actual allocation whatever it is.
57:47I don't know where this came from.
57:49You know, we'll get there, but for now,
57:51whatever it is, it has a Q, which I'll
57:54call Q bar. Bar for being like the one
57:56that actually happens.
57:58And then all the firms in the economy
58:00have, including the inactive ones,
58:03which, you know, for which this would be
58:04zero, have an X bar for all the
58:07different types of M. So, that's what
58:09this notation means, kind of, you know,
58:10all the the vector of all the firms in
58:12the world in our economy, I mean.
58:14Um and let's imagine that the actual
58:16allocation obeys this aggregate input
58:19constraint. You know, it doesn't
58:20obviously it couldn't break it because
58:22that's physically impossible. It could
58:24be this could be a less than sign. You
58:26could imagine the actual economy leaves
58:28some inputs unused, just leaves them
58:31idle. That's a totally valid
58:33possibility, I suppose.
58:35Um but not something we're going to sort
58:37of entertain here. Um
58:40just for the purposes of some of these
58:41discussions, it's not so central. But,
58:44but for now, imagine that it holds with
58:46equality even in the actual allocation.
58:48So, so this economy might be using the
58:49wrong mix of which firms get the inputs.
58:52That could be totally screwed up. That's
58:54what we'll call misallocation in a
58:55second, but it's not so screwed up that
58:57like some people are just sitting idle.
58:59That, you know, in other words, this
59:01holds.
59:02Anyway, so the second thing, just to
59:03make this easy to talk about, is we're
59:05going to assume that sort of again, this
59:07is really just
59:08don't think assumption, think sort of
59:10like almost like a
59:11a convention, you know, in some sense.
59:13Like for for thinking about issues
59:15conceptually, think about this as a
59:16convention. But,
59:18imagine that households in a sense are
59:21not distorting their consumption
59:23decisions. So, that's to say that, you
59:25know, through whatever means, markets,
59:28whatever, you know, something or other,
59:30households are acting as if they're
59:32paying a price P bar
59:34and they're choosing those cues that
59:36they are buying that enter that utility
59:38function such that the the P bar is
59:41proportional to the marginal
59:42productivity. So of course, you know,
59:44our usual model of like frictionless
59:47optimizing
59:49untaxed
59:51um
59:52consumers without any buying power of
59:54their own does display this. You know,
59:56price is proportional to marginal
59:57utility at the point the consumer
59:59chooses when it maximizes.
1:00:01Um so I'm just sort of in some sense
1:00:03pursuing a normalization or a convention
1:00:06where that is true even though of course
1:00:07that might not happen. The the reason we
1:00:09do this is sort of I guess
1:00:11I think people often think about that as
1:00:13almost sort of like consumption side
1:00:15distortions.
1:00:17Um
1:00:18whereas most of the firms literature has
1:00:20been more about what you might loosely
1:00:21call firm side distortions. And so
1:00:23that's a bit of an arbitrary
1:00:25distinction. Uh you know, a government
1:00:27could impose a tax and it it it's not
1:00:30always well posed to ask well, who's
1:00:32distorted the consumer or the firm?
1:00:34There it some sense the problem is that
1:00:36they're not
1:00:37the consumer and the firm aren't facing
1:00:38the same price and
1:00:39you know. But anyway, but we're going to
1:00:42just you know, basically follow this
1:00:43convention and in a sense kind of write
1:00:45this like that. So so the the the power
1:00:50of that I think for the purpose of
1:00:51thinking about firm firm efficiency or
1:00:53production efficiency is the is the
1:00:54following.
1:00:56It's just a definition, but let's let's
1:00:58unpack it. It it if it's defined well,
1:01:00so this goes by the name the value
1:01:02marginal product of input type um
1:01:06M, the Mth input. So if this was labor,
1:01:08you might call it value marginal product
1:01:10of labor or VMPL. But whatever it is,
1:01:12let's just call it VMPX because the M's
1:01:15don't have to be labor of course. Um and
1:01:18it's just defined as follows. You take
1:01:19that price P bar that the households are
1:01:21paying
1:01:22and you multiply it by dF dX of the type
1:01:25M at the actual point of production.
1:01:29The X bar.
1:01:30So um so obviously for any allocation
1:01:33and any set of prices, I can stick that
1:01:36allocation and those prices into this
1:01:38function, take the derivative at that
1:01:39point, and calculate the left hand side,
1:01:42and I'm just defining that as the value
1:01:44marginal product. Why are we doing this?
1:01:46I'll I'll get to that in a second. But
1:01:47for now it's just a definition.
1:01:49Um so you know, a couple comments
1:01:51however on this definition just to bear
1:01:53in mind when you think about this stuff
1:01:55in the literature and in in your own
1:01:58thinking.
1:01:59Um
1:02:00Three three sort of comments. So so the
1:02:02value part, the V, is really important.
1:02:05Um firms probably produce different
1:02:07things, different goods. That's product
1:02:10differentiation. They probably produce
1:02:12and by the way, when you think about
1:02:13product differentiation, of course we'll
1:02:15bear in mind that that includes things
1:02:17that might be a little hard to see like
1:02:19where is the firm physically located and
1:02:22you know, is it open 24 hours or
1:02:24something? You know, so we have to
1:02:26remember to think broadly about product
1:02:28differentiation. But anyway, um whatever
1:02:31it is, the cool thing about the V in
1:02:33VMPX is that the P bar, remember, is
1:02:36capturing the marginal utility of that
1:02:39good at the allocation. So or at least
1:02:41proportionally. So so it's
1:02:44so it's going to sort of perfectly be
1:02:45almost like a sufficient statistic for
1:02:48the social value that the firm is
1:02:51creating on the margin. That's dU dQ.
1:02:53That's what I'm calling dU dQ at Q bar.
1:02:56Um so that's the kind of really cool
1:02:58role played by P bar. Other sort of
1:03:00things it captures are sort of possible
1:03:02quality differentiation. Um you know, if
1:03:04this firm is making a highly quality
1:03:06adjusted version of its good, um then
1:03:09that would all be embodied in P bar cuz
1:03:11quality presumably shifts the utility
1:03:13function up. Um
1:03:15it also crucially captures diminishing
1:03:16marginal utility. You might imagine that
1:03:18that there's some firm that's kind of
1:03:20so
1:03:22big and productive in your economy that
1:03:25it's kind of
1:03:26you might think, you know, they should
1:03:28just take over the universe. They're
1:03:29they're so good at what they do that no
1:03:31other firm making anything remotely
1:03:32similar should exist. But of course, you
1:03:35have to sort of think on the margin.
1:03:36It's like you know, the the the the
1:03:38consumer is like that good less and less
1:03:40presumably the more they consume of it
1:03:42and the price again is a perfect
1:03:44sufficient statistic for that marginal
1:03:45utility. Um finally is just units. I
1:03:48mean, we we when we look at firms in
1:03:50data, we never know the units of what
1:03:51they're producing. They're some
1:03:52arbitrary thing like number of widgets
1:03:54or tons of you know, tons of
1:03:57rolls or something, doffs as David had.
1:04:00So anyway, the the the beauty of prices
1:04:02is that they convert the physical units
1:04:04of whatever good is being produced that
1:04:06this thing is measured in into utility
1:04:08relevant units, which are the relevant
1:04:09units.
1:04:10Anyway, second point, marginal. So I've
1:04:12already alluded to this, but obviously
1:04:14the productivity of a firm on the margin
1:04:17in terms of the the
1:04:18the marginal productivity of an input
1:04:20inside a firm could be very different
1:04:22depending on which margin we're talking
1:04:23about. So this is the physical
1:04:25manifestation of that. The the other one
1:04:28was more in the utility, the eyes of the
1:04:29consumer's manifestation of that. That I
1:04:31like less and less that firm the more it
1:04:33produces. That's in here. Um the firm
1:04:37itself might be less and less productive
1:04:39the more it physically puts in, the more
1:04:41it puts in. So that would that would
1:04:43mean declining marginal utility of an
1:04:45input or it's a marginal productivity of
1:04:46an input. Anyway, the point is that it's
1:04:49the margin that matters for thinking
1:04:50about, you know, everything here in
1:04:53today's lecture and in much of the
1:04:54course that follows. So anyway, the M is
1:04:56kind of very important in this VMPX
1:04:59definition.
1:05:01Just highlight that of course, you know,
1:05:03that's obviously different than the
1:05:04average product of input X, which we
1:05:06might define as just the total revenue
1:05:08produced. The average value product
1:05:10might just be revenue per unit of the
1:05:12input like revenue per worker or
1:05:14something.
1:05:16There are lots of reasons to expect
1:05:17positive correlations across firms maybe
1:05:20or even across the world within a given
1:05:22good between VMPX and VAPX.
1:05:27But I can also think of lots of reasons
1:05:29to expect a negative correlation or a
1:05:30zero correlation. So so for those
1:05:32reasons kind of I always sort of try to
1:05:34remind myself VMPX is what I wish I
1:05:37could measure and other things like VAPX
1:05:40are what, you know, unfortunately might
1:05:42be the best thing we can measure. And so
1:05:43we have to sort of ask how correlated is
1:05:46the thing we can measure with the thing
1:05:47we wish we could measure
1:05:49and be wary of, you know, settings where
1:05:51that correlation might not be so
1:05:52obvious.
1:05:53Um okay, final comment is that you know,
1:05:56that this thing is defined as it is,
1:05:57which is price times dF dX. That's
1:06:00obviously different from
1:06:03the what I've written down here, the
1:06:05derivative with respect to X of open
1:06:07bracket price times F.
1:06:09Uh
1:06:11the the distinction
1:06:13um
1:06:13uh is that you know, sometimes you can
1:06:15think about a firm as their revenue is
1:06:19kind of price times quantity or in other
1:06:20words, price times F. And so the firm
1:06:23part of the firm's problem is asking
1:06:25questions related to this, which is if I
1:06:27put a bit more X into my firm, well, how
1:06:29will it affect my revenue? And that has
1:06:31that works through two channels. One is
1:06:33how much will I physically produce? The
1:06:34other is how will I optimally change my
1:06:36price as I physically produce more. But
1:06:39all of those discussions are not
1:06:41germane. They're not part of this
1:06:43definition for for starters. And the
1:06:47usefulness of this definition, as we'll
1:06:49see in a second, is to do with this
1:06:52definition, not this one. And so you
1:06:55know, so so thinking through how will
1:06:57the firm change its prices useful for
1:06:59lots of questions, but for misallocation
1:07:02it's not a first order usefulness as
1:07:03I'll sort of argue in a second.
1:07:06You know, there are settings like
1:07:07famously studied settings, monopolistic
1:07:09competition with CES preferences is one
1:07:11that comes to mind where it just so
1:07:13happens that sort of across firms um
1:07:16their marginal sorry, if they're all
1:07:17facing the same CES preferences, then
1:07:19their marginal revenue product of an
1:07:21input would be proportional to their
1:07:23value marginal product of that same
1:07:25input. But that's kind of a coincidence
1:07:27that happens and and I don't think it's
1:07:28very generic or very general. So that's
1:07:31why I'm mentioning this kind of caveat.
1:07:35Um okay, so
1:07:37you know, the
1:07:38Sorry, so the um
1:07:42What happened here?
1:07:44Yeah, okay. So now we can talk about
1:07:45misallocation. So remember the best
1:07:47allocation was one where as I wrote it,
1:07:51dU dQ times dF dX is equal to a
1:07:54constant. But remember, all we've done
1:07:56here is sort of pointed out that it's
1:07:58often reasonable to follow the
1:07:59convention that prices reflect dU dQ.
1:08:02And so in other words, we can write what
1:08:04the planner would do as set VMPX equal
1:08:08to the con to that constant for all
1:08:10active firms and inputs. Um
1:08:13So so in other words, the planner would
1:08:15just sort of make sure that everywhere
1:08:17in the economy every time the right
1:08:19firms, the active set of firms are being
1:08:21used, within that set the VMPX for a
1:08:24given type of input like labor would be
1:08:26constant across all those active firms.
1:08:29So the corollary kind of is obvious that
1:08:32you know, since that's the best, every
1:08:33other allocation has to display
1:08:35something different. And one thing that
1:08:37would be different is this. That that is
1:08:39there exists at least one I and and or M
1:08:43such that
1:08:44you know, the value marginal product in
1:08:46that I is not equal to the
1:08:48aforementioned constant, you know.
1:08:50And so we often sometimes capture that
1:08:53by another definition that sort of some
1:08:54people might call a wedge,
1:08:56which would be the sort of the ratio of
1:08:58the value marginal product of an input
1:09:00divided by that constant. And we would
1:09:02observe that sort of the the planner
1:09:04would want that thing to be one for
1:09:06every firm, you know, equal to a
1:09:08constant and the constant would be one
1:09:10for every firm. But the misallocated
1:09:12actual economy presumptively would not
1:09:15have this. It would have wedges that are
1:09:17not equal to one. It would have wedge
1:09:19dispersion. In other words, it would
1:09:20have some firms that are lower than one
1:09:21and some that are higher than one. And
1:09:25And you know, it's wedge dispersion that
1:09:27would be indicative of misallocation.
1:09:29Um okay, so you know, of course why
1:09:31might misallocation happen? We haven't
1:09:33talked about that. We just said it said
1:09:34suppose there was an allocation
1:09:36uh it would be misallocated if and only
1:09:38if it displayed this that that is
1:09:41variance or dispersion in the value
1:09:44marginal product of a given input across
1:09:46uses in the economy.
1:09:48Um why does my my misallocation happen?
1:09:51Well, a lot of this course is going to
1:09:52be about that, but
1:09:53we already know some answers as as
1:09:55economists who sort of studied things
1:09:56like the first welfare theorem. Um you
1:09:58know, economists call these things
1:09:59market failures. We sort of know that
1:10:01market allocations achieve the first
1:10:03best kind of you know, if market
1:10:06failures don't exist and probably always
1:10:09almost always don't achieve the first
1:10:11best if market failures do exist. So uh
1:10:14so you know, examples of market failures
1:10:16that um that exists you know, and are
1:10:19and are important and and
1:10:21some of the discussions you're going to
1:10:23see in coming lectures aimed to sort of
1:10:24try to quantify some of the costs of
1:10:26these these these market failures and in
1:10:28terms of the misallocated cost um would
1:10:31be taxes on production, subsidies on
1:10:33production, regulations that that are
1:10:35imposed on production. Those things of
1:10:36course might have some higher objective
1:10:38like raising government revenue which
1:10:40maybe is good even though it might uh
1:10:42for some reason the best way to raise
1:10:43government revenue might have to distort
1:10:45production, but it would distort
1:10:47production. So for instance in aggregate
1:10:48productivity being low, we would be
1:10:50interested in asking if there are sort
1:10:52of taxes that distort production. Um and
1:10:55hence lead to misallocation. Um things
1:10:57like corruption uh probably would. That
1:10:59would be uh a case where firms are asked
1:11:02to pay bribes let's say uh or do other
1:11:04things that are directly unproductive,
1:11:06but um
1:11:08but indifferent extents on the margin,
1:11:10right? So uh to the extent that's true,
1:11:12that would act like a wedge uh that is
1:11:14dispersed across firms. Uh expropriation
1:11:17risk would be another example. You know,
1:11:19imagine that one of your exes like
1:11:20investing that you're a firm that you're
1:11:22thinking well, if I invest, I'll just
1:11:24get held up by the government
1:11:25expropriation let's call that. Um and
1:11:28that would mean that that that that I
1:11:30probably wouldn't invest which would
1:11:31mean that sort of the value marginal
1:11:32product of me investing would be high.
1:11:35You know, I if I were to do it, I might
1:11:36keep doing it until the value marginal
1:11:38product is of investing is the same as
1:11:40every other firm, but given that I know
1:11:42I'm going to get held up, I'll hold back
1:11:44on that investment and hence the VMPX
1:11:46might be higher in my firm than in some
1:11:48other firm that's not facing that
1:11:50expropriation risk.
1:11:51Um okay, we can go on. So we can think
1:11:53about asymmetric information. You know,
1:11:55that's as many manifestations of things
1:11:57like um
1:11:58uh you know, hold up between buyers and
1:12:00sellers,
1:12:01um credit constraints. We often think of
1:12:04a micro foundation that is asymmetric
1:12:05information that the lender can't verify
1:12:08that the borrower firm is kind of using
1:12:12the loan in in the way that the firm
1:12:14said it would and to the extent that
1:12:16that's not visible or verifiable or
1:12:17monitorable, uh that would um that could
1:12:21lead to you know, uh inefficient
1:12:23allocation of credit which which we
1:12:25sometimes call a sort of
1:12:27sometimes manifests itself as a credit
1:12:29constraint, a a contract that a bank
1:12:31might write with a firm that says you
1:12:33can borrow at an interest rate R, but
1:12:35only up to some limit. You know, that
1:12:37that that would be a constraint if the
1:12:39it would be a binding constraint if the
1:12:40firm wanted uh could you know, to
1:12:44borrow more than that amount.
1:12:46Um incomplete contracts appear in other
1:12:48environments. Uh you know, the that um
1:12:51that you know, are familiar. Uh
1:12:54you know, at a higher level we can think
1:12:55about market power. You know, of course
1:12:57market power you could also
1:12:59micro found as a deriving from for
1:13:01example, asymmetric information between
1:13:04firms and their customers. Uh the
1:13:06inability to perfectly price
1:13:07discriminate is a form of asymmetric
1:13:09information and and market power sort of
1:13:11happens when firms
1:13:13kind of
1:13:15you know, can set prices, but can't
1:13:16fully uh price discriminate. Um of
1:13:19course you know, I described that sort
1:13:20of in their in the output market
1:13:23sense of like a standard oligopoly, but
1:13:25of course you know, we also know that
1:13:26firms when they buy probably might have
1:13:28some market power too.
1:13:29Uh when they buy labor or materials for
1:13:31example.
1:13:32Um
1:13:33I
1:13:35Pure externalities, I'll come back to
1:13:36that, but uh that could be embraced
1:13:39here. There's a little subtlety there
1:13:41that I'll come back to, but for now it's
1:13:43certainly a market failure.
1:13:44Um
1:13:45irrational firms often don't get talked
1:13:47about as a market failure, uh but um but
1:13:50I think it's important uh element of
1:13:52discussion in in um
1:13:55uh
1:13:56uh in the in the type of literature
1:13:59we're talking about of the the next
1:14:01lectures. Um you know, that I would call
1:14:03as opposed to the the type of irrational
1:14:06rationality I mentioned earlier, here I
1:14:07would kind of really just point to
1:14:09failure to profit maximize. So so there
1:14:11is an X that is profit maximizing. The
1:14:14first welfare theorem assumes that firms
1:14:16choose that, you know, that they're
1:14:18price taking and profit maximizing. Um
1:14:20if they're not profit maximizing, then
1:14:22they're not we're not then we can't
1:14:24apply the welfare theorem. It acts like
1:14:25a market failure. Um and uh one example
1:14:29of that would be a you know, not just a
1:14:30firm that's kind of stupid in the sense
1:14:32if you if you forgive the word, a form
1:14:34of irrational, but of course you could
1:14:35imagine the firms are a collection of
1:14:37agents the of actors. There's no need
1:14:40that that there's no reason to expect
1:14:42they would necessarily do what's sort of
1:14:44best for profits and we might call that
1:14:46some agency problems. Anyway, you know,
1:14:49the bottom line of course as you've seen
1:14:51probably in your study of
1:14:52microeconomics, macroeconomics of market
1:14:54failures is that effectively at the end
1:14:56of the day, what what's going wrong is
1:14:59that a given firm, let's call it I, is
1:15:03paying acting as if it pays on the
1:15:06margin more or less for a given input
1:15:09than some other firm does. And that's
1:15:11what this basically says. Now, if you if
1:15:14you're struggling to think about that,
1:15:15you know, think about market power. It's
1:15:16like um
1:15:18you know, we don't normally phrase it
1:15:19this way, but it acts this way which is
1:15:22to say that like that firm who's
1:15:24thinking of hiring let's say some more
1:15:26labor and hence would expand, they know
1:15:28that one cost of expansion is that they
1:15:31would um slide down their own demand
1:15:33curve and hence have to lower prices and
1:15:36that would harm harm the profits they
1:15:38were making on infra-marginal units. And
1:15:40so in some sense the cost of hiring
1:15:42labor is not just the wage of labor,
1:15:44it's also the markup uh that they were
1:15:46getting on on existing units at the
1:15:49start of that thought experiment. So
1:15:50that's a case where if one firm has a
1:15:52high markup and another firm has a low
1:15:53markup, even though they pay the same
1:15:55price for labor, they're acting as if
1:15:57they're paying a different price on the
1:15:58margin for labor and that's why it it it
1:16:01too comes down to this bottom line.
1:16:03Um
1:16:04Okay, so I you know, the reason this I
1:16:08think is useful uh you know, this this
1:16:10discussion is it sort of lets us think
1:16:12about the difference between sort of
1:16:15misallocation as a source of low
1:16:16aggregate productivity and bad
1:16:18technology if it kind of quote unquote
1:16:20as a source of low aggregate
1:16:22productivity. Um there's no easy answers
1:16:25to the question of like which of these
1:16:27is bigger.
1:16:28I think to some people, you know, it's a
1:16:31could be thought of as almost a little
1:16:32bit tautological because some people say
1:16:35well, everything you can think of as an
1:16:37A, I could just call misallocation and
1:16:39other people say well, everything you
1:16:41think of misallocation, I could just
1:16:42think of as a bad technology. Um you
1:16:45know, a lot of that is
1:16:48a lot of some of that is definitional,
1:16:49some of that is semantics, some of that
1:16:50is philosophy, but uh so you'll kind of
1:16:52have to make up your own mind and stick
1:16:54to your own conventions and then be kind
1:16:55of internally consistent. But but here's
1:16:57how I sort of think about some of those
1:16:59debates, uh some of those discussions.
1:17:01Um
1:17:02So the key thing about misallocation as
1:17:04I've kind of set this up to in order to
1:17:06define this as I previously did is that
1:17:09you know, we're talking about inputs
1:17:10that are in fixed aggregate national
1:17:12supply. So um so the problem that's
1:17:15going on is that that input is being
1:17:18used in the wrong way within my nation
1:17:20across the uses.
1:17:22So um so in some sense, you know, as I
1:17:24already stressed, the key thing to
1:17:26remember there is that like if you're
1:17:28thinking about improving the world by
1:17:30fixing misallocation, you must be
1:17:32thinking about fixing the world with a
1:17:34reallocation. You must be thinking about
1:17:36basically taking some firm J and
1:17:39removing labor let's say from it in
1:17:42order to free up labor to put it into
1:17:44some other firm I. That's you know, once
1:17:46we've got an aggregate input constraint,
1:17:47this is just mathematical necessity. So
1:17:51in some sense if you're thinking I you
1:17:52know, I'm thinking about something
1:17:53that's wrong with the world that in
1:17:54principle policy could make better and
1:17:56you're thinking of labeling that thing
1:17:57misallocation, you're inherently
1:17:58thinking that there's an input that's in
1:18:00fixed supply and that you could move
1:18:02around. And by corollary, you'd have to
1:18:04make some firm have less of it and
1:18:06another firm have more. Um so in some
1:18:08sense you're you're sort of thinking
1:18:09about something that's going to harm
1:18:10quote unquote one firm and help quote
1:18:12unquote another firm at least to the
1:18:14extent that like
1:18:15you know, shrinking the amount they have
1:18:17is harming a firm.
1:18:18Um
1:18:19Okay, so another less obvious uh point
1:18:22about misallocation is that it is always
1:18:24generating a rent or what sometimes
1:18:26people might call revenues, but um but
1:18:29it's in other words, if you're thinking
1:18:30about misallocation that in some sense
1:18:31just makes resources from society
1:18:34disappear into thin air, nobody gets
1:18:36them, nobody gets any money from them,
1:18:38then it's not misallocation. It's bad
1:18:41technology. So the reason there are
1:18:43rents is that sort of as soon as VMPX if
1:18:45somebody's paying W for the input M WM
1:18:48for the input M, but their sort of value
1:18:50marginal product of using that thing,
1:18:52again, this is just a definition, so it
1:18:54has to follow is VMPX. If VMPX divided
1:18:57by minus one is not zero, then in some
1:19:00sense they're making money off of hiring
1:19:02that input and that's what we could call
1:19:04or often define as a rent. And if
1:19:06they're selling Q of those inputs
1:19:09Q outputs at the time, then Q times that
1:19:12marginal rent would be the total rent
1:19:14enjoyed. So of course in the case of
1:19:15markups, we're used to calling these
1:19:17things pure profits and hence kind of
1:19:18thinking about them as rents. Um in the
1:19:21case of taxes, there are where where a
1:19:23tax on labor would drive a wedge between
1:19:26the the W that the that the laborer
1:19:29earns and the MPX that the firm gets out
1:19:31of that amount of labor
1:19:33um that the firm acts like it's paying.
1:19:35Um, and uh those rents go to the
1:19:38government, the the person who collects
1:19:39the tax revenue. Uh but either way,
1:19:41there has to be a money trail. So, if
1:19:42you're sort of thinking about things
1:19:44that are
1:19:45corruption due to misallocation, ask
1:19:46yourself the question, is there a money
1:19:48trail?
1:19:49Um, anyway, so by contrast, uh you know,
1:19:52bad technology, quote unquote, as a
1:19:54source of aggre- low aggregate
1:19:55productivity is in this sort of
1:19:58framework something about the A's.
1:20:00Remember, it's about the set of tech the
1:20:01constraint on the set of techniques the
1:20:03firm I knows how to use.
1:20:05Um, so in that sense, you know, that the
1:20:07the A's could be something that could be
1:20:11improved like for free, almost, you
1:20:13know, without possibly an aggregate
1:20:16constraint. So, in principle, you know,
1:20:18I could teach one person how to fish and
1:20:20teach someone else how to fish. And and
1:20:22that would increase person number one A
1:20:24one's AI and also person number two's.
1:20:26Um, so, you know, these these A's are
1:20:29not only exogenous to the firm by
1:20:31definition, the way we set it up, a
1:20:32constraint the firm can't do anything
1:20:34about, but um but they're also some
1:20:37things that do change exogenously in the
1:20:39world around us. Like, you know, a
1:20:40hurricane happens and that's bad. And
1:20:43the way you would see that in this model
1:20:44is that would reduce some firms' AIs.
1:20:47Um, also, some exogenous gifts of good
1:20:50technology do arise. We sometimes think
1:20:53about these as externalities in that
1:20:54context, but, you know, for example, if
1:20:56you have foreign firms in your economy
1:20:58and you think that they're
1:20:59uh
1:21:00in some sense teaching, whether, you
1:21:02know, whether by deliberately or not,
1:21:04teaching domestic firms, uh or maybe
1:21:06there's spillovers through worker
1:21:08mobility or something. These things
1:21:09often get called FDI spillovers. Those
1:21:11are kind of from the purpose of the
1:21:12firms in this country, those are
1:21:13exogenous, you know, gifts of of better
1:21:16A's. Um, they came for free, you could
1:21:19say. Uh obviously, there might there
1:21:21might not be free to attract FDI, but
1:21:22for the purpose of this thought
1:21:23experiment, this transfer is free.
1:21:26Um, uh you know, finally, however, you
1:21:28know, you can think about some policies
1:21:30that governments might pursue that would
1:21:32the whole point might be to raise A's.
1:21:34This could be like teaching firms
1:21:36something to again raise that sort of
1:21:38ignorance constraint on the techniques
1:21:39they know about. Um, we think of
1:21:42education that governments do sometimes
1:21:44in the in that sort of light. Um, but
1:21:46another example would be building
1:21:47infrastructure. So, if if infrastructure
1:21:49is the type of things that tends to be
1:21:51the firms don't produce, the governments
1:21:52produce, then you can sort of think
1:21:54about it in a reduced form way as,
1:21:56you know, when I build good roads in my
1:21:58country, I'm just helping the firms to
1:22:00have higher A's.
1:22:02They can produce more Q with the same X.
1:22:06You know, they can they devote less
1:22:08resources to transporting because the
1:22:10roads are better. And and of course,
1:22:13those A's might not come for free. You
1:22:14know, it costs the money the government
1:22:16money to to build the roads and improve
1:22:18them. But
1:22:19but
1:22:21but from this framework, they think
1:22:22about them for now as A's. Finally,
1:22:24another good example would be removing
1:22:25red tape. We don't think that that might
1:22:27just be making the organ of government
1:22:29the organ of customs procedures just
1:22:31more efficient.
1:22:33That just lets firms produce more Q for
1:22:35fewer work for this for the same number
1:22:37of workers. So, that that would be way
1:22:39I'd think about red tape. Um, anyway, so
1:22:42the bottom line of thinking about this
1:22:43distinction for me is that sort of, you
1:22:45know, improving A is usually going to
1:22:47sort of push out it by definition it
1:22:49pushes out one's firm production
1:22:50possibility frontier. It probably it
1:22:53sort of it it therefore also pushes out
1:22:55the the country's production possibility
1:22:57frontier. And in and and that might be
1:23:00sort of free and so it doesn't come at a
1:23:02cost. You know, by contrast, um changing
1:23:06the X's
1:23:07uh is um is not free. Remember, it's not
1:23:10about moving out the PPF, it's about
1:23:11sort of maybe choosing a different point
1:23:13along it. Oh, sorry, sorry, sort of
1:23:15getting the the economy to a point on
1:23:16its own PPF would be reallocating the
1:23:18little X's. Um, now, you know, it it's
1:23:22it's very unclear to me if I take a
1:23:24given firm and ask, should I make that
1:23:25firm bigger, you know, raise its XAI, or
1:23:28not? It's very unclear to me whether the
1:23:30answer is yes or no. Remember, when I do
1:23:32it, I'm going to help that firm get
1:23:33bigger, I'm going to make some other
1:23:34firm smaller. Is that a good
1:23:36reallocation? It's very not much not
1:23:38clear from to me. So, when I think about
1:23:39misallocation, I sort of think the
1:23:41answer is not at all obvious. When I
1:23:42think about improving A, the answer is
1:23:44almost always that yes, that will be
1:23:46good. Um, there are some caveats to
1:23:48that, but it it almost I think in
1:23:50realistic scenarios, improving the
1:23:52nation's PPF is likely to be good.
1:23:54Um,
1:23:55okay, so
1:23:56uh
1:23:57you know, that the the the final thing I
1:23:59wanted to sorry say about this is that
1:24:02um you know, there's a couple caveats
1:24:04about everything I just said that should
1:24:05be borne in mind. And I would also
1:24:08highlight that I think these are areas
1:24:09where it's not just that I've omitted
1:24:11them and it's these are areas where I
1:24:13think gaps in kind of research knowledge
1:24:16might be might be particularly high in
1:24:18my opinion, but uh maybe you know,
1:24:20that's just my opinion. But anyway, I've
1:24:21taken the big X's as as exogenous. Uh
1:24:24but they might be, of course,
1:24:26endogenous. There are many reasons to
1:24:28think that in fact that they go in kind
1:24:29of a good direction. That like um that
1:24:33countries that are more have higher
1:24:34aggregate productivity also have higher
1:24:36big X's.
1:24:37Um, you know, maybe the incentives to
1:24:39invest time in your own education are
1:24:41higher in an aggregate higher aggregate
1:24:43productivity country. So, to the extent
1:24:45that big X is something like human
1:24:46capital, that makes sense.
1:24:48Um, second, input-output linkages. So,
1:24:50everything I said here about sort of the
1:24:51variance of wedges being sufficient for
1:24:53misallocation is true if um
1:24:57the inputs are in fixed aggregate supply
1:25:01and probably not true if inputs are not
1:25:03in aggregate fixed supply. So, an
1:25:04example of inputs that are not in fixed
1:25:07aggregate supply are non-factor inputs.
1:25:09In other words, intermediates that are
1:25:11produced by other firms. Um, so then you
1:25:14get something called double
1:25:14marginalization where basically it's not
1:25:17whether the wedges are dispersed that
1:25:18matters, it's what what I if you read
1:25:21this very nice paper, you'll see that
1:25:24sort of in some sense what matters is
1:25:26whether there's dispersion in sort of
1:25:28the amount of double marginalization
1:25:31across all paths from raw factors of
1:25:34production to final consumption. If that
1:25:36thing is not dispersed, then you still
1:25:38don't have misallocation. But um but if
1:25:40that that sort of path-wise notion of
1:25:42dispersion of of sort of aggregate
1:25:44wedges is dispersed, then you do have
1:25:46misallocation.
1:25:48Um, third, you know, I I I skipped over
1:25:50pure externalities. You can always think
1:25:52about those without loss here as just a
1:25:53firm that's a multi-product firm. It
1:25:55makes both its good and the externality
1:25:58good, like whether it's pollution or
1:25:59knowledge spillovers or whatever you
1:26:00want. Um, you could also, of course,
1:26:02think about that in this framework as
1:26:04like the firm I firm not I is affecting
1:26:07AI's productivity AI directly.
1:26:09Um, finally, you know, I've glossed over
1:26:11extensive margin concerns. You know,
1:26:13nothing I said was incorrect, but but
1:26:16it's it bears when the world has
1:26:18increasing returns to scale
1:26:20technologies, then this statement, value
1:26:23marginal product of X is constant among
1:26:24active firm pairs, is still true. It
1:26:27just needs to be used with some caution
1:26:28because sort of something that economies
1:26:31markets could very much get wrong is not
1:26:33the MPX among active participants, but
1:26:35who are the active participants. If
1:26:38you're sort of thinking of like that
1:26:39your economy is missing some superstar
1:26:41firms that just haven't even been born
1:26:42yet, you're thinking in this light. Um,
1:26:46okay, so I with that said, you know, I
1:26:48want to sort of foreshadow
1:26:50very briefly sort of some themes that
1:26:52are going to come in in the the next few
1:26:54weeks, starting tomorrow.
1:26:56The next lecture is tomorrow at the same
1:26:58time. Um,
1:27:00and so, you know, Eric Verhoogen will be
1:27:03talking tomorrow and this will be
1:27:05followed next week by Jin Kai sort of on
1:27:07the sort of themes that I that I kind of
1:27:09relate. These characterizations are not
1:27:11perfect. Themes will overlap, of course,
1:27:13but
1:27:14you know, really sort of getting into
1:27:15the weeds of like, you know, what are
1:27:17these good techniques? What are these
1:27:19AIs that are limits on good techniques?
1:27:21What are some policies that appear to
1:27:22sort of help firms to sort of have
1:27:25access to better techniques and hence
1:27:26raise their AIs. Um,
1:27:29also, you know, we glossed over sort of
1:27:31what's happening inside this very
1:27:33interesting and important max problem.
1:27:35You know, I I sort of wrote it down very
1:27:37glibly, but as economists, a lot of what
1:27:39we do should be sort of just very quite
1:27:41positive understanding of how that
1:27:43works. What are the techniques
1:27:44available? Um,
1:27:46okay, so then and you can think about
1:27:48understanding management and training in
1:27:50that light, I believe. Um, uh Pete
1:27:53Klenow is going to lecture give lecture
1:27:55four and sort of really get into a lot
1:27:57of the details of the misallocation
1:27:59thread I've already opened, uh thinking
1:28:01about how do we measure this stuff? How
1:28:02do we quantify this stuff? What you
1:28:04know, which wedges look big, which
1:28:06wedges look small? Is this is this sort
1:28:08of misallocation hypothesis uh
1:28:10quantitatively um
1:28:12uh
1:28:13important, which it it very much seems
1:28:14to be.
1:28:15Um, lectures that come later will then
1:28:17dig into sort of particular input
1:28:19markets and output markets. So, sort of
1:28:21capital labor distortions, uh
1:28:23contracting distortions in input
1:28:25markets.
1:28:26Um, McKenzie Woodruff and Machiavello
1:28:28will talk about those theme themes.
1:28:30Think about those as sort of measuring,
1:28:31diagnosing, understanding policies that
1:28:33fix wedges in input markets.
1:28:36Uh Lauren Bergquist will talk about in
1:28:38lecture seven about um competition in
1:28:40product markets. And um and think about
1:28:43that as kind of wedges that are due to
1:28:44imperfect competition in the in the
1:28:46output market.
1:28:48Finally, in our last lecture, Rebecca
1:28:50Usha will, as David already mentioned,
1:28:52will will talk about industrial policy.
1:28:54So, this is, you know, one space of
1:28:56policy that that that many people think
1:28:59is is interesting because it it part of
1:29:02its effectiveness might hinge on a claim
1:29:05or the belief that it
1:29:06it might improve misallocation and or
1:29:09just sort of improve the spread of good
1:29:11technology. Uh you know, a lot of debate
1:29:13about whether that's true, as David
1:29:14already alluded to, but
1:29:16but Rebecca will kind of lead you
1:29:18through some of that discussion and what
1:29:20we know and what we don't know, etc.
1:29:22about those ideas.
1:29:23Um, I want to just flag before we end by
1:29:26sort of, you know, want to flag that
1:29:28this is an incredibly active and
1:29:30interesting research area. We haven't
1:29:32been able to put together, and I'm sure
1:29:34your patience would eventually wear
1:29:35thin, like an infinitely long course
1:29:37that would cover everything. We tried to
1:29:39focus on some important highlights and
1:29:41we're very grateful that these lectures
1:29:42have volunteered their time to to
1:29:43contribute, but you know, I just want to
1:29:45sort of leave this last slide up and and
1:29:47in order to highlight there are lots of
1:29:48things that aren't going to get as much
1:29:50attention as they might need and that
1:29:52are great active research areas that
1:29:54where where a lot more work is needed.
1:29:56Um, development firm space interacts
1:29:58with other fields like international
1:30:00economics, IO, public finance,
1:30:02environmental, labor, urban,
1:30:04the fluctuation side of macro as well as
1:30:06the aggregate side of macro that I've
1:30:07already alluded to, household economics
1:30:09cuz a lot of firms are household firms,
1:30:11uh, behavioral economics, political
1:30:12economy. So, rich intersections between
1:30:15all of these areas and um, firms in
1:30:17developing countries that aren't going
1:30:19to get as much attention in the in the
1:30:20coming weeks as as maybe, uh,
1:30:22you know, we wish they could. Um, and
1:30:24then just I I'd again reiterate that one
1:30:27could have very legitimate objectives
1:30:29other than aggregate productivity such
1:30:31as distributional concerns, um, for
1:30:33example, self-employment offering more
1:30:35than just profits to the people who are
1:30:37choose to be self-employed, um,
1:30:39especially in developing countries
1:30:40perhaps. And um, and also PF, you know,
1:30:44public finance, governments raising
1:30:45revenue to fund public goods and using
1:30:47firms to achieve that goal might distort
1:30:49aggregate productivity, but might
1:30:51nevertheless be good because that's the
1:30:53best way to fund public goods that are
1:30:54important for other goals. Um, and so,
1:30:57you know, we we're, you know,
1:31:00I think a lot of what I've talked about
1:31:02today has been fairly glib on
1:31:04non-aggregate productivity,
1:31:06uh, character considerations, but uh, I
1:31:08didn't want to sort of belittle those
1:31:10considerations. They're they're
1:31:11extremely important. Um,
1:31:12and and finally, you know, there there's
1:31:14a classic problem that sort of, you
1:31:16know, a lot of very important and
1:31:18interesting firms are informal and data
1:31:21limitations and other limitations have
1:31:24often sort of hindered our ability to
1:31:26understand that side of the economy
1:31:27fully and and so this sort of informal
1:31:30versus formal distinction is an area
1:31:32that, you know, another area that we
1:31:34wish could be covered even more than it
1:31:36than it will be covered.
1:31:37Uh, okay. So, with that I'll stop.
1:31:39Uh, I
1:31:40want to thank everyone for being here
1:31:42and again thank the lectures that are
1:31:44going to follow in the in the coming
1:31:46days starting tomorrow 10:00 a.m.
1:31:48Eastern time with Eric Verhoogen talking
1:31:51about firm upgrading.
1:31:53Uh, and thank you all for for being
1:31:55here.
1:31:56Uh, we'll stop now.
1:31:58Thank you very much.

About this transcript

This page contains the full transcript of Lecture 1: Course Overview by David Atkin and Dave Donaldson by International Growth Centre, generated from the public captions YouTube serves with the video. The transcript has 16,002 words across 2,630 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.