00:00 Uh
00:02 Everyone,
00:02 my name is Dion Filmer.
00:03 I'm the director of the World Bank's Development Research.
00:06 Welcome to this policy research talk.
00:09 You
00:11 who
00:12 in
00:13 these talks provide us an opportunity.
00:16 To present work coming out of the research department here at the bank,
00:20 with the goal of sharing the findings with colleagues inside and
00:23 outside the department along with others outside of the World Bank.
00:27 Uh,
00:27 I'd like to welcome the audience both on Webex as well as on YouTube.
00:32 So today my colleague Roberto Fatal Haef,
00:34 who's a senior economist in the macroeconomics and growth team in the department,
00:39 will discuss how distortions in firms' business environments
00:43 can help us understand productivity differences around the world.
00:47 We're grateful to have Mary Hallward-Driemeyer today as a discussant.
00:51 Mary is a senior economic advisor to the Global Director for Trade,
00:55 Investment and Competitiveness.
00:57 She was previously a senior principal specialist
00:59 in the jobs cutting solution area,
01:02 and prior to that,
01:04 she was a lead economist here in the Development Research Group.
01:08 With that,
01:08 I'll turn it over to Roberto,
01:09 who will talk for about 45 minutes,
01:11 after which we'll hear from Mary for about 10 to 15 minutes.
01:15 Uh,
01:15 we'll conclude the session with a Q&A from the audience.
01:19 If you have a question,
01:20 I'd ask that you please use the raised hand function in Webex,
01:23 uh,
01:24 that way I can call on you.
01:26 Or if you can't figure that out,
01:27 put your name in the chat and just signal to me that you have a question.
01:31 If you're following on YouTube and you'd like to ask a question,
01:34 please put the question in the chat and that'll be
01:36 conveyed to me.
01:38 So with that,
01:38 over to you,
01:39 Roberto.
01:41 Thank,
01:42 thank you,
01:42 Dion,
01:43 and welcome everyone to my talk.
01:45 It's really exciting to be able to talk about
01:47 my research in front of such a big audience.
01:50 Also,
01:51 thanks in advance to Mary.
01:52 I anticipate uh a very interesting discussion towards the end,
01:55 so looking forward to that.
01:58 Um,
01:58 the main objective of my talk is to introduce you to some of the latest research,
02:03 both on my own but also showing,
02:05 uh,
02:05 other scholars' works.
02:08 Showing how a micro to macro approach in which
02:12 the interaction between firms
02:14 and the business environments in which firms operate
02:17 can shed light on obstacles to countries' economic growth opportunities.
02:23 Now,
02:23 before I get to the core of the presentation,
02:25 let me,
02:25 uh,
02:26 remind you of what are one of the big questions or challenges
02:30 in the macrodevelopment literature.
02:32 Of course,
02:33 there are many long-lasting puzzles,
02:34 but I think that
02:35 uh most of them uh have the same common roots,
02:39 which have to do with the observation of
02:42 massive differences in living standards around the world.
02:46 So let me proxy living standards by the income per worker of,
02:50 of each country and,
02:52 and express it relative to the living standards
02:54 of an advanced economy like the United States.
02:57 I'm going to show a histogram collecting precisely these possible income levels
03:02 on the horizontal axis,
03:04 and then I'm gonna count how many countries there are
03:07 at various buckets of income along this development distribution,
03:13 this development spectrum.
03:15 And two important conclusions sort of jumped,
03:18 jumped out of the picture.
03:19 The first one,
03:20 as I alluded at the beginning,
03:21 is the observation of huge heterogeneity.
03:24 So we have countries that are roughly as advanced as the United States,
03:28 say France would be an example,
03:31 but there are countries at the very bottom of the distribution with,
03:33 with very low levels of income.
03:36 More broadly,
03:37 the,
03:37 the conclusion is the prevalence of,
03:39 of low levels of income uh in,
03:42 in,
03:42 in,
03:42 in the distribution.
03:43 Um.
03:44 A better way to,
03:46 to,
03:47 um,
03:49 to solidify this conclusion is to look at
03:51 what's the median income per worker
03:54 uh relative to the United States and that's captured by
03:58 the vertical red bar here.
04:00 So what we are seeing is that half the countries in the world
04:04 enjoy living standards,
04:05 up to a third of the living standards in the,
04:08 in the US and in the advanced economy.
04:11 So naturally confronted with this question,
04:13 one wants to,
04:14 to know more,
04:15 in particular ones want to inquire about
04:17 what are the sources of these differences in income
04:20 and hopefully this talk will provide a particular story and a particular
04:25 theory of what may be underlying these differences in income,
04:29 but uh let me take a purely accounting intermediate step
04:33 and try to decompose the
04:36 Origin of the differences in living standards
04:39 into differences in the factors of production that generate this,
04:43 the output.
04:45 So starting from the following identity,
04:48 basically the,
04:48 the output of an economy per worker is given by
04:52 a combination of its tangible factors of production,
04:55 the physical capital and human capital,
04:58 and then the efficiency with which these factors turn into output
05:02 is what is typically referred to as the total factor productivity.
05:07 With this identity,
05:08 I'm going to ask the question of
05:10 what's the relative role of factors of production versus total
05:14 factor productivity in explaining income gaps in the data.
05:19 And I will,
05:19 I will,
05:20 uh,
05:21 portray the information with the aid of the following scattered plot.
05:24 I will report the GDP per worker relative to the US on the horizontal axis,
05:30 and then I'm going to report the
05:32 relative total factor productivity of each country
05:35 again relative to the United States,
05:38 and subsequently I will report the ratio of factors of
05:42 production in each country relative to the United States.
05:47 And the purpose of the 45 degree line is to help us assess.
05:52 To what extent
05:53 either TFP gaps and factor gaps
05:56 help us account for the observed income gaps.
06:00 So the closer
06:01 the scatter plot lies
06:02 to the 45 degree line,
06:04 the more explanatory power that particular factor or input will have.
06:09 Well,
06:09 let me begin with total factor productivity.
06:11 The,
06:11 the,
06:12 the,
06:12 the,
06:12 the picture sort of speaks for itself.
06:15 We see a very strong correlation between
06:18 TFP gaps and income gaps,
06:20 and more,
06:21 more to the point,
06:23 the,
06:23 the TFP gaps line up very close to the 45 degree line.
06:28 So another way to,
06:29 to express this,
06:30 this implication is to say that
06:33 we would have gone a long way in understanding
06:36 drivers of income differences if we understand drivers of productivity.
06:43 Now,
06:43 this does not mean to undermine the importance of thinking of policies
06:48 that bridge human capital gaps around the world or that promote,
06:51 promote investment.
06:53 So the,
06:53 the
06:55 uh differences in factors of production are now overlaid to the figure with,
06:59 with the blue dots
07:01 and,
07:01 but what one must note that while there is also
07:04 a strong correlation between factor gaps and income gaps,
07:09 the magnitude of the,
07:10 of the gaps in factors of production.
07:13 It's not as big as the magnitude in income.
07:16 And er
07:18 er and it's,
07:19 you have less explanatory power than productivity in explaining the,
07:23 the,
07:23 the,
07:24 the,
07:24 the,
07:25 the heterogeneity in income.
07:27 So I,
07:28 I began the presentation with a big question,
07:30 what explains income differences,
07:32 and I sort of finished this introduction with another big question,
07:35 which is what explains productivity differences.
07:38 And the goal of the talk is,
07:39 as I said,
07:40 try to provide a particular story
07:42 for this,
07:43 for this question.
07:45 Now,
07:46 where we are right now,
07:46 we have done sort of a macro to macro loop.
07:49 I postulated an aggregate production function
07:52 and I have
07:53 uh looked at aggregate data to decompose the aggregate differences in income
07:59 and reached the conclusion that this object that we call
08:02 TFP which is kind of like a black box,
08:04 is the,
08:05 the main driver of of differences in income.
08:09 So where I'm headed is in bringing the micro to the picture.
08:13 And what I mean by bringing the micro is
08:15 by acknowledging that there's no such a thing as
08:17 a representative producer in reality,
08:20 but rather
08:21 production takes place across a large number of firms in the economy.
08:26 And the economy confronts the microeconomic challenge
08:30 of how to
08:31 allocate its productive resources
08:34 across these firms
08:36 in a way that maximizes the total size of the pipe.
08:41 And why is this progress?
08:42 Well,
08:42 this is progress because once we have an idea of what
08:46 this ideal allocation looks like,
08:49 then we can look at the particular allocation in,
08:51 in any given country
08:53 and to the extent it it differs from the optimal prescription,
08:57 then we will have identified potentially barriers
09:00 that prevented this economy from achieving its potential.
09:04 And as a result,
09:05 we will have provided you,
09:07 we will have provided a theory of GFT.
09:11 Now
09:13 Turning this intuitive argument into an actual quantifiable strategy
09:18 requires that I take a series of steps.
09:21 The first step I need to take is to uh be explicit about how I think
09:26 the output maximizing rule is achieved.
09:29 And the view I adopt in this paper and
09:32 it's the predominant view in the literature is that
09:35 the economy will operate at its potential if market forces operate freely.
09:41 So if we
09:42 let the price mechanism allocate resources combined
09:45 with the profit maximizing incentives of firms
09:48 and strong competition coming from continuous entrance of new producers,
09:53 then markets are going to take us
09:55 to the potential.
09:58 Now,
09:58 what is that uh ideal allocation going to look like?
10:01 And before I,
10:02 I,
10:02 I give you a more technical definition,
10:05 In simpler words,
10:07 what's going to happen is that
10:08 markets are going to ensure that
10:10 there's no money left on the table.
10:13 And what would be an example of money being left on the table?
10:16 Well,
10:17 imagine that under a particular
10:19 distribution of productive resources
10:22 across firms,
10:23 we end up in a situation where,
10:25 let's say producer number one.
10:28 shows a higher marginal product for its inputs
10:31 than producer #4.
10:34 Well,
10:34 that cannot be the output to maximizing allocation because
10:37 by virtue of featuring a higher marginal product,
10:42 we would be able to increase the total size of the pipe precisely by reallocating
10:47 some of the workers and the capital
10:50 from firm 4 towards firm number 1.
10:54 So,
10:54 and now I'm,
10:55 I'm,
10:55 I'm preparing a more formal definition,
10:57 at the output maximizing allocation,
11:00 the recipe for achieving it,
11:02 the economy would have to have
11:04 exploited all output increasing opportunities.
11:08 And what that means is that the marginal revenue products of each
11:12 firm in the economy would have to be equalized with each other.
11:17 Now,
11:17 in,
11:17 in the context of,
11:18 of more familiar diagrams that we are used to seeing
11:22 um
11:24 as a byproduct or another way to see
11:26 the implication of equalizing marginal revenue products across firms
11:30 is that the economy will be
11:32 producing at the frontier of its production possibility.
11:36 Well,
11:37 I know the US has many,
11:38 many problems,
11:39 but
11:39 from the point of view of this presentation,
11:41 I'm,
11:41 I'm adopting it as the efficient economy.
11:44 And so let's say the US it's at the frontier.
11:47 Now,
11:50 an important conclusion I want to draw is that
11:52 while
11:53 at the optimal allocation,
11:55 all firms will have the same marginal product.
11:58 This does not mean
12:00 that firms that we are going to be equally big.
12:04 Very much to the contrary,
12:06 because some firms,
12:07 one firm may be much more productive than another firm,
12:11 it will be able to sustain
12:13 many more workers and many more machines
12:16 until the marginal product of that firm
12:18 has been brought down
12:20 to the level of the marginal product of a less productive firm.
12:24 So underlying this efficient allocation,
12:27 there will be a firm size distribution
12:30 essentially reflecting this heterogeneity of productivity.
12:37 Now consider what would happen if this
12:39 uh ideal market-oriented world is disrupted,
12:43 and it could be disrupted by some inherent frictions in how markets work.
12:47 Think about commitment problems between creditors
12:50 and debtors in financial markets.
12:52 Or it could be government policy that no matter how well intended,
12:57 they could be distorted from an efficiency point of view.
13:00 An example of this that will uh I will build heavily on
13:04 throughout the talk is size dependent taxes.
13:07 So imagine governments that
13:09 levy taxes or enforce taxes more strongly on large firms than small firms.
13:16 Well,
13:16 this economy is one where the efficient
13:19 allocation or the output maximizing rule will be
13:22 broken down.
13:24 And
13:24 again to solidify this point,
13:26 let me put a little bit more structure.
13:28 Let's assume that,
13:30 uh,
13:30 let's assume a particular ranking of productivity
13:34 across the four firms that populate this little economy.
13:37 So let's assume that the firm one is the most productive of the four,
13:42 and then let's assume that governments enforce contract
13:45 taxes more prominently on the largest firms.
13:49 So the most productive firms will confront a higher effective tax rate
13:54 than the least productive firm.
13:56 And it's even possible that because governments sometimes want to promote.
14:01 small-scale entrepreneurship,
14:03 it's possible that even small firms can fund actual subsidies to production.
14:09 Now in this context,
14:11 the efficient allocation,
14:12 the equalization of marginal revenue products
14:15 will not be achieved,
14:16 and that's because the firm number one will
14:19 be discouraged by these higher effective tax rates.
14:23 Uh,
14:23 from acquiring as many workers and as many machines as it would otherwise have.
14:30 And on the contrary,
14:31 firm number 2 will maybe enjoy the benefits
14:34 of these promotion policies of the government and perhaps
14:38 expand beyond what they would have uh
14:41 uh uh
14:42 beyond the scale they would have achieved
14:45 in an,
14:45 in an undistorted world.
14:47 So
14:48 these idiosyncratic distortions have the potential of
14:52 misallocating resources
14:54 from high to low productivity.
14:57 And in the context of the diagrams then what
15:00 we are generating by breaking up the efficient recipe
15:04 is we are forcing the economy to operate
15:06 somewhere inside the production possibility.
15:10 And I'm not picking up on India for any other reason than
15:13 it's being the first country for which this methodology has been applied.
15:17 So that's assume that India,
15:19 then it's operating below its potential.
15:23 Now notice that uh as a byproduct of this property of misallocation
15:28 in which we take away resources from the good firms and
15:31 shift them to the less productive ones,
15:34 we are at the same time
15:35 shifting and rotating or altering the shape of the size distribution.
15:40 So in this particular er er specification of distortions I have been talking about,
15:46 we're going to have fewer of the large firms,
15:49 and we are going to have more of the small firms.
15:53 So the size distribution is gonna reflect information
15:56 about what's going on in the business environment of these firms.
16:01 And lastly,
16:01 let me consider what would entry barriers do
16:05 to break the optimal output maximizing pool.
16:10 So one can think of entry barriers as
16:13 something that inflates
16:15 the entry cost of new producers and new firms to any given economy.
16:20 So,
16:21 you know,
16:21 creating a product or opening up as an establishment is a costly process.
16:25 So there are some costs that one cannot avoid to incur,
16:28 and let's say that those are the only costs you would incur
16:32 in a,
16:32 in a,
16:32 in a
16:34 well performing market economy.
16:36 Uh this has to do with foregoing income,
16:38 it has to do with
16:39 uh preparing business plans and so on.
16:42 An entry bar is anything that artificially inflates
16:46 this cost of entry to an economy.
16:49 And the way these entry costs,
16:51 uh,
16:52 these entry barriers disrupt
16:54 the efficient allocation is not so much by
16:58 altering the allocative margin,
17:00 namely the intensive margin,
17:02 but rather altering the extensive margin.
17:05 That is by shaping or disrupting the number of producers
17:10 that will find it optimal
17:12 to circumvent these barriers and actually start
17:16 a business in,
17:18 in,
17:18 in,
17:18 in the economy.
17:19 So we'll have potentially half the number of firms
17:22 and notice that
17:23 because of the subdued competitive pressures in this environment,
17:26 firms may even choose
17:28 different levels of productivity.
17:30 So I'm denoting this potentially lower levels of productivity with set I had here.
17:36 So again back to the diagrams,
17:38 the implication of an entry barrier is that
17:41 not only the country will be operating below the frontier of its possibilities,
17:46 but the frontier itself may be shifted inwards
17:49 relative to a scenario without distortion.
17:53 Now,
17:53 at the micro level,
17:55 the implication for the firm size distribution
17:57 is going to be the opposite
17:59 than the direction in which idiosyncratic distortions
18:03 reshape the size distribution.
18:06 Because there's going to be fewer producers,
18:08 prices will adjust
18:10 to allocate.
18:12 The resources that are in the economy towards all these fewer firms,
18:17 so each of these firms will become larger.
18:19 So in equilibrium when entry barriers are prominent in a country,
18:23 then
18:24 we are going to,
18:26 we should observe more of the large firms and fewer of the small firms.
18:33 So,
18:33 uh,
18:34 at the end of this
18:35 conceptual introduction,
18:36 and I have done it in a very intuitive way,
18:38 but underlying all of these claims,
18:40 there is a paper formalizing all these propositions,
18:44 but I hope I have conveyed this idea of what I mean by the macro,
18:47 micro to macro approach.
18:50 I've shown that interacting firms
18:52 with the frictions and policies in the business environment can explain
18:56 why they don't uh produce at the frontier of its possibility.
19:00 And importantly,
19:02 I have uh emphasized
19:04 how an observable object these days in the data,
19:07 which is the frame size distribution
19:10 contains information
19:12 about the type of policies that may be affecting a given country.
19:17 So what I need to do next is to actually show you some data
19:20 corroborating the claim that there are big differences
19:24 in the,
19:25 in the size distribution across countries.
19:28 Otherwise,
19:29 if we,
19:29 if I didn't show you this,
19:31 it would be like,
19:31 this would be a very promising
19:33 theory but not very applicable in reality.
19:37 So that leads me to,
19:38 to tell you a little bit
19:40 about the,
19:40 the data I've been able to collect for this project.
19:43 So as you can imagine,
19:44 uh um
19:45 obtaining representatives from level data is one
19:48 of the scarce resources in this agenda.
19:51 So,
19:51 uh,
19:52 for this project I have collected
19:54 representative from level data for the manufacturing sectors of 21 countries.
20:00 The source of the data,
20:02 the data is a mixture of industrial surveys or manufacturing censuses
20:08 that are universal of oil firms
20:11 with 10 workers or more.
20:13 And this,
20:14 uh,
20:15 this type of data covers a few South America,
20:17 Latin American,
20:19 Sub-Saharan African and Asian countries.
20:22 And I,
20:23 I should stress at this point that
20:25 I've been able to
20:27 collect all this data only
20:29 uh thanks to the collaboration with various counterparts
20:32 in the global practices and operational units at,
20:35 uh,
20:36 at the bank.
20:36 So,
20:37 this is an example of,
20:38 of the synergies that can be created between Deery and,
20:40 and,
20:41 uh,
20:41 and,
20:41 and global practices.
20:44 And another data source I'm working with is Amadeus,
20:48 and Amadeos is a commercial database
20:50 supposedly covering all of uh European,
20:54 uh all of Europe.
20:56 Now,
20:56 uh,
20:57 it's well known that
20:58 the quality of the,
21:00 the representativeness of the data is not that good in some of the countries,
21:04 so I've selected a subset of the European countries for which
21:09 comparing the size distribution in Amadeos
21:13 with what is reported by Eurostat,
21:15 I'm reassured that the data,
21:18 the data in Amadeos is,
21:19 is represented.
21:20 So at the end of the day,
21:21 I have 21 countries.
21:23 So let me characterize.
21:26 What are
21:27 cross-country differences in,
21:28 in firm size distributions,
21:30 and I'm gonna use,
21:31 I'm,
21:31 I'm gonna answer this question looking at the average firm size in each country.
21:36 And in particular the measure of size I'll be looking at is the number of workers.
21:42 So,
21:42 for example,
21:43 the United States is the,
21:44 the red dot here,
21:45 as you can see is the richest country in my sample.
21:48 So we have the log of GDP per capita on the,
21:50 on the horizontal axis.
21:51 So the United States is the richest country and
21:54 conditional on firms with 10 workers or more,
21:57 its manufacturing sector has around 120 employees.
22:01 Well,
22:02 the main conclusion of this figure,
22:04 uh,
22:04 which is essentially
22:06 reporting a result that the literature has shown elsewhere,
22:09 but,
22:09 but I needed to persuade myself that it was true
22:12 also in my data,
22:14 is that there is a very large heterogeneity of firm sizes around the world,
22:19 and that poorer countries
22:21 produce with smaller firms than richer countries.
22:25 So to me,
22:25 this is a promising starting point.
22:27 I,
22:27 I was sort of obliged to show you that
22:30 there are differences in the size distribution because
22:32 the micro to macro approach predicts these differences to exist.
22:37 So,
22:37 we are,
22:37 we are,
22:38 we are proceeding on solid ground,
22:39 so to speak.
22:41 So what I'm going to do next is I'm going to
22:43 provide a strategy to identify the magnitude of underlying distortions.
22:50 And then once I'm equipped with estimates of those distortions,
22:53 I'm going to go back to the macro
22:55 and assess what are the productivity effects
22:58 of the distortions that I have identified.
23:03 So,
23:03 um,
23:04 what is my inference strategy going to be about,
23:07 and I need to spend a couple of slides,
23:08 uh,
23:08 uh,
23:09 talking about this and
23:11 what I,
23:11 what I will
23:12 pursue is a theory-based or a model-based approach.
23:16 And what this approach is about is
23:19 Leveraging the
23:21 insight I,
23:22 I've been saying so far that
23:24 the firm size distribution
23:27 of a country
23:28 contains information
23:30 about the distortions affecting that country.
23:33 So I'm going to propose a model of the size distribution.
23:37 And I'm going to discipline the trade-offs and the channels in the model
23:42 to replicate exactly,
23:44 exactly
23:45 the properties of the size distribution
23:47 in the United States,
23:49 which is,
23:50 as I as I said earlier,
23:51 our,
23:51 our efficiency benchmark.
23:54 So once I have a model of reality that
23:56 captures the reality of a particular and distorted country,
24:00 I'm gonna ask,
24:02 what is the combination of entry barriers,
24:05 um,
24:07 tau and allocative distortions
24:09 that must be taking place in a given country.
24:13 So that the model is able to reconcile
24:17 the size distribution of that country.
24:20 OK
24:22 So as you,
24:22 as you can imagine,
24:23 the inference is as good as the model I used to interpret reality.
24:27 So,
24:27 uh,
24:27 while it's,
24:28 it's not the objective of this talk to flesh out all the plumbing
24:32 underlying my theory,
24:34 I just want to emphasize
24:36 three key channels that the theory accounts for,
24:39 which you could argue are first order channels in any,
24:42 in any theory of the from size distribution.
24:46 So of course there will be heterogeneity of producers that's like
24:50 you know started if I,
24:51 if I didn't,
24:52 but importantly,
24:54 the heterogeneity of productivity in producers is not going to be
24:59 given to the economy,
25:00 but it's,
25:01 it's rather going to be endogenously chosen by firms.
25:05 Pretty much affected by the incentives in the environment in which they operate.
25:10 And also the model is going to have endogenous entry and exit,
25:13 which again,
25:14 uh it's,
25:14 it's like a prerequisite so as to be able to speak about the role of entry model.
25:20 So I would say that uh to a first approximation,
25:22 these are features of any theory you wanna write down to,
25:24 to interpret reality.
25:27 And one important assumption,
25:28 one needs to be transparent about
25:31 is that
25:32 the firms that will populate this model economy
25:35 will
25:35 operate in competitive factor markets.
25:38 So I'm gonna assume away any monopoly power on,
25:41 on,
25:42 on labor markets or on capital markets.
25:46 So in this context,
25:47 then let me proceed with the identification of distortions
25:51 and let me begin with idiosyncratic distortions
25:55 because I think that I have gone a long way
25:57 in explaining my identification strategy in the conceptual introduction.
26:02 And
26:03 the starting point of the strategy is basically
26:06 the rule
26:07 that breaks up
26:09 the output maximizing rule.
26:11 As you remember from before,
26:13 when there are distortions like size-dependent taxes or financial frictions,
26:18 the marginal revenue product of any given firm with productivity that
26:23 will not be equalized to the marginal product of every other firm in the economy.
26:29 So the question,
26:30 what we want to learn about is how big is this
26:33 tau set across all these firms.
26:36 And for that,
26:37 as you can see from,
26:38 from the equation,
26:39 we need information about the marginal revenue product of firms.
26:42 If I were to be able to observe this marginal products,
26:45 then I would be able to solve
26:47 for the distortion from this equation.
26:51 Now the point that Xian Klino in,
26:53 in a very famous paper published in 20099 has told us
26:58 is that indeed underrate standard assumptions,
27:03 the marginal revenue product of a firm
27:06 is nothing but
27:07 the ratio of the revenue of a firm
27:10 to the input bundle
27:12 that the firm uses in producing
27:14 its output,
27:15 its goods or,
27:16 or the services.
27:18 And,
27:18 and the key insight is that with firm level data,
27:21 both the revenues
27:22 and the inputs are observable.
27:25 So then I can go to the firm level data,
27:27 plug it into each side of this equation,
27:29 and then solve for the
27:31 er
27:32 idiosyncratic distortion that satisfies the equality.
27:37 Now,
27:37 at the end of the day,
27:38 this procedure provides us with
27:40 one distortion for each firm.
27:43 And I will choose in this presentation to summarize all that information
27:48 through the lens of one summary statistic which I argue
27:52 is the most important statistic characterizing distortions in a country,
27:57 which is
27:58 the extent to which the distortions correlate
28:02 with the productivity of the firm.
28:04 That is,
28:05 I'm interested in assessing
28:07 how much it's true in the data
28:09 that
28:10 countries
28:11 that distortions in a country
28:13 make the life harder of the productive firms
28:16 compared to the less productive firms.
28:19 And I'm gonna call this property correlatedness of distortion.
28:25 So then,
28:26 then let me put this strategy at work and show you evidence of
28:30 correlated distortions around uh around my 21 countries.
28:35 So I'm going to show you again
28:37 estimates of
28:38 the relationship between
28:41 distortions and productivity.
28:43 That's what I labeled here slope of the distortion to productivity profile,
28:48 and I'm going to project those slopes
28:51 against the income per capita of the country.
28:54 And I want to begin focusing on these blue dots,
28:57 which constitute the richest countries in my sample
29:01 and
29:02 I want to start with the richest countries because
29:05 I want you to notice that
29:08 the model does not find any significant systematic relationship
29:12 between
29:13 the size of the distortions and the productivity of the firm.
29:18 Well,
29:18 I think this is a reassurance check
29:20 both of the theoretical approach
29:23 and my choice of summary statistic
29:25 because one would expect to find that in rich countries there's no
29:29 such systematic relationship between
29:32 distortions and productivity.
29:33 We don't expect France and Finland
29:35 to tax the productive firms more heavily than the less productive.
29:39 So to me this is a reassuring property
29:41 of both the theory and the summary statistic.
29:45 However,
29:46 when we populate the figure with estimates of
29:48 the correlatedness of distortions around the world,
29:51 we do find a,
29:52 a striking pattern in which
29:55 er as you go down in,
29:57 in economic development,
29:58 then you start to find
30:00 a very prevalent correlated distortion.
30:05 Now,
30:05 the,
30:05 the,
30:06 the units don't mean much yet.
30:07 I will,
30:08 I will come to a more meaningful interpretation later,
30:10 but I,
30:10 I wanna,
30:11 I wanna take away this conclusion that
30:13 the,
30:13 the,
30:14 the,
30:14 the choice of summary statistic being
30:17 the degree of correlation between the
30:18 distortion of productivity is really informative
30:22 about the business environment in poor countries.
30:27 Now at this point you might,
30:28 if you've been sort of following,
30:30 uh,
30:30 you might wonder
30:31 why do we
30:33 think about anti-virus altogether.
30:36 It looks like
30:37 the evidence on idiosyncratic distortion is capable on its own
30:43 to tie all the knots in the in the productivity puzzle.
30:47 So why is that?
30:48 Well,
30:49 I've shown you theoretically
30:51 that correlated distortions
30:54 reduce the firm size.
30:55 Remember the diagram showing that idiosyncratic distortions rotate
30:59 the size distribution away from large firms.
31:03 Furthermore,
31:04 I've shown you empirically that poorer countries
31:07 operate with smaller firms.
31:09 And to top it all,
31:11 poorer countries have
31:12 more evidence of correlated distortions.
31:15 So everything seems to close together
31:18 and,
31:19 and pushes us to not even think about entry barriers.
31:23 What I argue then is that while that is
31:26 Reasonable from a qualitative point of view,
31:29 it's still a quantitative question the extent to which entry barriers can capture
31:35 the cross-country differences in average sizes that
31:38 I've shown you before in the data.
31:41 So that's the question I'm going to address now.
31:44 I'm going to start with my model of reality
31:47 that I calibrated to replicate exactly the American economy.
31:52 And then I'm going to use that model.
31:55 Distortions that satisfy this correlatedness with productivity
32:00 feeding the estimates of correlation from the data.
32:04 And then I'm going to compute in the context of the model
32:07 what would the average firm size be.
32:10 And I'm going to plot that model based average size on the vertical axis.
32:15 Now if the model is able to account for the data which is on the horizontal axis.
32:20 Then we should expect to see a scatter plot with
32:23 all the dots close to the 45 degree line.
32:28 What we find instead is that for some countries,
32:31 which
32:32 are the rich countries for which distortions are
32:34 not very prevalent in the first place,
32:36 uh,
32:36 the theory and the data align very well,
32:38 but that's almost by definition.
32:41 But for the majority of the countries for which we estimated prevalent,
32:44 uh,
32:45 correlated misallocation,
32:47 we find a big divergence between the prediction of a model
32:52 with only geosyncratic distortions.
32:55 and the data.
32:56 Let me illustrate this with the case of Peru.
32:59 So Peru,
33:01 Peru's manufacturing sector,
33:04 conditional on plants with 10 workers or more,
33:06 have an average size which is about 70%
33:10 the average size of the United States.
33:13 According to the theory,
33:14 if idiosyncratic distortion
33:16 had been the only distortion in Peru,
33:19 then
33:19 we would have expected Peruvian firms to be much smaller than they really are.
33:25 So,
33:26 what's missing from the idiosyncratic distortion story is
33:29 that while it goes in the right direction,
33:31 it like,
33:32 it pushes the decline of the average firm size
33:35 too far
33:36 relative to where the data tells us firm size is.
33:41 So what the model is missing then or what,
33:43 what I extract as information from this limitation is that
33:47 there is information
33:49 in
33:50 not having achieved the 45 degree line.
33:53 There there is a countervailing force missing in the model.
33:57 There is a force that is interacting
33:59 with these financial frictions and these size-dependent policies.
34:04 That is
34:04 pushing the size distribution
34:07 on the other direction.
34:09 And if you recall my,
34:10 my conceptual introduction,
34:12 uh I'm,
34:12 I'm repeating it here,
34:14 as I said,
34:14 misallocation rotates
34:16 the size distribution inwards,
34:18 so away from large firms,
34:20 but entry barriers
34:22 are exactly capable of accomplishing
34:25 what I,
34:26 what the data,
34:26 what the model is missing to match the data.
34:29 Because as you recall,
34:30 I argued that entry barriers
34:32 have the power or have the incentive of
34:35 concentrating production
34:37 into fewer and larger plants.
34:40 So then,
34:41 uh uh as you can see,
34:42 this basically,
34:44 um,
34:46 Automatically transforms into an estimation strategy of entry barriers.
34:53 What I'm essentially going to do then is
34:55 explore possible barriers of entry barriers.
35:00 Feed them into the model
35:02 and interacting them with the misallocation I've already estimated.
35:07 And I'm going to continue that exploration of the entry barrier space
35:12 until the model is capable of bringing all the blue dots back to the 45 degree line.
35:20 OK,
35:21 so
35:22 following this inference approach,
35:24 what are the properties of the entry barriers that I that I identified?
35:29 As before,
35:29 let me begin by showing my estimate,
35:32 and here the,
35:33 the vertical axis is the model-based estimate of entry barriers
35:37 plotted against the economic development of countries.
35:40 Again,
35:40 as a reassurance,
35:42 the model is finding that the most advanced economies,
35:45 Belgium is a little bit of an outlier,
35:47 but
35:47 for the most part,
35:48 the advanced economies,
35:50 the model does not find any evidence of entry barrier.
35:54 And like before,
35:55 this is both to me reassuring
35:57 not only of the theory I'm using to confront the data,
36:01 but also
36:02 about the approach to back out the distortion.
36:06 Now,
36:06 when you populate the er or when you repeat
36:08 the exercise for all the countries in the sample,
36:11 you start
36:12 identifying very big uh entry barriers in the poorest country.
36:18 Again,
36:19 units don't mean much yet.
36:20 I will come back later to a more meaningful interpretation,
36:24 but I wanted to,
36:25 to draw your attention to this strong pattern by which
36:29 as we go down in,
36:30 in economic development,
36:31 entry barriers start to show up as a prominent,
36:35 uh,
36:35 as a prominent barrier to,
36:37 to doing business.
36:40 So now,
36:41 uh,
36:41 again,
36:41 I have sort of concluded the micro part,
36:44 so I have,
36:45 I have leveraged the micro insights to learn about distortions.
36:50 Now let me close the loop and go back to macro and answer the question I,
36:55 I posted at the beginning,
36:56 which is
36:57 what,
36:58 how much can these microchannels
37:01 account
37:02 for the aggregate productivity differences in the data.
37:06 So,
37:06 uh,
37:07 I'm gonna do that.
37:08 I'm,
37:09 I'm gonna use my model
37:11 to ask a counterfactual question which is,
37:13 assume that we can remove every distortion
37:17 and liberalize every country from its
37:19 entry barriers and idiosyncratic distortions.
37:22 What are the productivity gains
37:24 that each country is going to be able to reap?
37:27 And those gains again are reported on the vertical axis,
37:31 so a number of 1.5 means a 50% gain.
37:35 And I'm again plotting that um
37:39 against incorporate worker
37:41 and as before,
37:42 distortions
37:44 don't have any bite in explaining why France and Italy
37:47 are not yet at the frontier relative to the US.
37:50 So distortions don't have any explanatory power for explaining
37:53 these tiny differences in,
37:55 in living standards
37:57 among the,
37:57 the rich countries.
38:00 Now for the rest of the countries,
38:01 we start to find large ugly productivity gains
38:05 to be ripped.
38:07 From reforms that dismantle distortions in the business environment.
38:11 And these gains range all the way from 50% in Kenya,
38:14 which you could argue is a little bit of an outlier,
38:16 but on average,
38:17 uh,
38:18 lie somewhere between 15% and 20%.
38:21 So,
38:22 uh,
38:22 I would say that,
38:23 uh,
38:24 back to the language I've used,
38:25 uh,
38:25 in the conceptual introduction,
38:27 there's a lot of money on the table
38:29 to be appropriated
38:31 by
38:32 identifying reforms
38:34 that liberalize the,
38:35 the,
38:35 the,
38:36 the economic environments of this country.
38:39 Now,
38:40 another way to portray this information is to
38:44 translate these productivity gains
38:46 in terms of what fraction
38:48 of the productivity gaps I'm supposed to understand
38:52 from the data
38:54 can be explained by this this distortion channel.
38:58 OK.
38:59 So the,
38:59 the fraction of the observed TFP gap is here on the vertical axis,
39:04 and I'm plotting it
39:05 against
39:06 the actual gap.
39:08 So uh for Ethiopia,
39:09 for example,
39:10 the United States is almost 10 times more productive than Ethiopia.
39:15 It's around 7 times more productive than Bangladesh and Ghana.
39:19 And what the model is saying is that for this.
39:24 Very poor countries,
39:26 the model can only explain up to 5%
39:30 of the observed productivity gap.
39:32 So you could argue that in this sense you could call it not a very successful.
39:37 I would still argue that 5% of a tenfold difference,
39:40 it's still
39:42 a gain worth reaping.
39:44 But
39:45 leaving aside these,
39:46 these cases,
39:47 for the average country,
39:48 the,
39:48 the fraction of the GFP gap that you could explain
39:52 from this distortion-based approach
39:56 averaged somewhere 20%.
39:58 So I,
39:58 I,
39:59 I would say that um
40:01 the question I posted at the beginning that
40:03 the microchannels are,
40:05 are,
40:05 are,
40:05 are informative about,
40:07 uh,
40:07 are a promising avenue to keep exploring.
40:11 No,
40:11 um,
40:12 Roberta,
40:13 yes,
40:13 Dion,
40:14 just to flag that you have about 5 more minutes.
40:16 OK,
40:16 yeah,
40:16 I think that that's what I need.
40:18 Great.
40:19 Um,
40:20 yeah,
40:20 so,
40:20 uh,
40:20 at this point,
40:21 I think that,
40:22 uh,
40:22 uh,
40:23 many of you might be reacting to what I've said so far by,
40:27 you know,
40:27 this is all very nice,
40:28 Roberto,
40:29 but,
40:29 uh,
40:29 what's behind all these distortions and how,
40:32 how to fix them,
40:34 and,
40:34 and,
40:34 yeah,
40:35 when we go to the doctor,
40:36 it doesn't matter how
40:37 technologically advanced is the MRI machine,
40:40 we just want to know how to,
40:41 to heal,
40:42 and I,
40:43 I will have that discussion and I hope that discussion will continue in the Q&A.
40:47 Uh,
40:48 but before I do my own discussion,
40:49 I,
40:50 I just want to pause a little bit to appreciate the progress
40:53 that has to happen to sort of corner the question at this point.
40:58 And again,
40:59 development accounting
41:01 sounds like we've been around forever,
41:03 but it's only a reasonable,
41:04 it's only a recent um
41:07 uh
41:08 exercise that researchers are being able to,
41:10 to perform.
41:11 It's been,
41:11 I think the,
41:12 the,
41:12 the first uh
41:14 uh issue of the Pen World tables which is the statistical resource
41:18 that researchers used to
41:20 Performed the development accounting was published
41:22 in the Economic Journal in 1978.
41:25 So
41:25 it's been only around for 40 years and it's still improving.
41:28 But the,
41:29 the,
41:29 the opening of the black box is also another trivial exercise,
41:33 you know,
41:33 some theorists have to figure out how to integrate
41:37 models of industrial organization into general equilibrium.
41:41 Then we needed to find uh and gather from level data,
41:45 and then to top it all,
41:46 we need to figure out how to deal with
41:47 these models quantitatively to derive the counterfactuals I,
41:51 I showed you before.
41:52 So,
41:53 again,
41:53 I'm gonna have a discussion about policies,
41:55 but let's,
41:55 let's not uh
41:57 Let's not forget about the progress that's been made in the,
41:59 in the diag diagnostic of the problem.
42:03 So let me begin offering a discussion of what potential
42:06 and what actual policies could underlie
42:09 the entry barriers and the idiosyncratic distortion.
42:13 And let me begin with entry barriers.
42:16 Uh,
42:16 one,
42:17 very promising candidate that could underlie the entry barriers
42:21 I backed out with the model is regulation.
42:24 And for regulation we have the,
42:27 we have this fantastic resource that the World Bank er
42:31 collects and,
42:31 and develops which is the,
42:33 the,
42:33 the doing business indicators
42:36 which provide a measure of the costs of
42:37 entry around the world and that's supposed to be
42:40 a regulatory-based measure of the cost of starting a business.
42:44 So one thing I can do is to say well how much
42:47 of the model-based barrier
42:49 can be explained by the regulatory component of barriers as proxy by
42:56 doing business indicator.
42:58 And again,
42:58 the 45 degree line has been my tool to assess how close they,
43:02 they match each other.
43:04 So what we find is that,
43:06 well,
43:06 of course in countries for which the model finds no barriers.
43:10 There's also no barriers,
43:12 no regulatory hurdles in the,
43:13 in the World Bank's indicators.
43:15 So that's for France,
43:15 Italy,
43:16 so it's almost by definition that
43:18 the model will do a good job
43:20 at following the,
43:21 the doing business indicators.
43:23 But even for countries like Ghana and Bangladesh
43:25 for which the model finds very large barriers,
43:28 it's striking that
43:30 they closely,
43:31 they,
43:31 they follow very closely regulated.
43:33 So regulations seem to be at the,
43:35 at the core
43:36 of barriers to entry in,
43:38 in Ghana and Bangladesh.
43:40 But we have another chunk of countries for which
43:44 the model barrier is far larger than the,
43:47 the barrier implied by the world's banks indicated.
43:53 So,
43:53 um,
43:55 This is saying that there is something else beyond
43:58 regulation and that it's for all of us to,
44:00 to figure out what,
44:01 what,
44:02 what else could it be,
44:03 you know,
44:03 predatory behavior of firms,
44:05 taxes,
44:06 whatever,
44:07 that is really making it harder to enter this country.
44:11 And to give you a sense of
44:13 how much more money on the table is being left
44:16 because of the total barriers
44:18 compared to the regulatory barriers,
44:21 I'm plotting here the differential
44:24 gain in total factor productivity
44:27 from removing.
44:29 The,
44:30 the,
44:30 the entirety of entry barriers as measured by my model-based barrier
44:35 versus having only removed the regulatory component.
44:39 And again,
44:41 uh,
44:41 of course for the countries where the regulatory component
44:44 and the model mimicked each other very closely,
44:48 then
44:48 the differential gains are very small,
44:51 but notice that for the majority of the countries,
44:53 there's an average
44:55 8% gain in productivity to be reaped
44:58 by figuring out
45:00 what else beyond regulation
45:02 is making it harder for firms to enter in Colombia,
45:06 Chile,
45:06 India,
45:07 Romania,
45:07 etc.
45:08 So,
45:09 uh,
45:09 that,
45:09 that's what I have to say,
45:11 uh,
45:11 about,
45:11 uh,
45:12 uh,
45:12 realistic,
45:13 uh,
45:14 policies that could explain what the model backs out as a,
45:17 as a barrier to entry,
45:18 and I hope we can come back to this in the Q&A.
45:20 Now,
45:21 in terms of idiosyncratic distortions,
45:23 the literature has
45:24 been around for longer,
45:26 so other scholars have explored
45:29 many other avenues that could explain
45:31 what I back out as an idiosyncratic distortion,
45:34 and this is my last slide.
45:37 Uh,
45:37 so for example,
45:38 you'd imagine that labor regulation could be an underlying cause
45:42 of labor misallocation,
45:44 so we have uh the uh
45:45 popular papers by Hoppenheim and Rogerson exploring the role of firing costs.
45:50 More recently,
45:51 we have Garicano,
45:52 Le Large,
45:53 and Van Reenen looking at
45:54 a particular size-dependent labor policy in France.
45:59 To a large extent,
46:00 the finding of this literature is that
46:02 they don't generate big misallocative effects.
46:06 Another
46:07 possible channel you could think of is size-dependent tax enforcement,
46:11 and this is,
46:12 I,
46:12 I alluded to this channel earlier because that's something I worked on with,
46:17 with my colleague Pierre Vajas and
46:19 Anders Jensen,
46:20 and we estimated from the data
46:23 the extent to which larger firms
46:25 are enforced
46:27 taxes more prominently than smaller firms,
46:30 and then we feed those empirical estimates into the model.
46:34 And we find again very small losses
46:36 coming from this single channel.
46:40 And lastly,
46:40 I think I would say the most promising of the policies that people have studied in,
46:44 in generating
46:46 big productivity losses is finance or financial frictions and
46:49 again I'm,
46:50 I'm not gonna be fair with the big literature.
46:52 I'm just
46:53 citing uh
46:55 uh whereakaboke AR 2011 and Midrick and Schul AR 2014.
47:00 These papers find that
47:02 finance could be a big obstacle to growth and to productivity,
47:06 but
47:06 more so through technological adoption channels by,
47:09 by explaining why countries cannot finance,
47:14 uh,
47:14 adopting more innovative technologies,
47:16 but from an allocative point of view,
47:18 financial frictions do not have a large effect.
47:23 So my conclusion is that most
47:25 likely misallocation or the idiosyncratic distortions,
47:28 there's no silver bullet to understand them or explain them.
47:31 They are more likely a combination of all of these sources
47:36 and most likely the relative weight
47:38 of all these channels is going to be very country specific.
47:41 In some countries finance will be more relevant than labor and so on and so forth.
47:45 But I still think it's a useful tool to count with
47:48 for uh policymakers to
47:51 help,
47:51 to help policy policymakers rank reforms and prioritize reforms.
47:55 So
47:56 one thing that will be very interesting is
47:59 if
47:59 as the World Bank
48:01 through our collaboration with,
48:03 with authorities in the governments with which we work with,
48:06 we could come up with some sort of glossary of the policies they do.
48:10 And then we could understand
48:12 what are,
48:13 of course there will be some benefits to those policies.
48:15 There's a reason why governments do them.
48:17 And then we can find a way to bring them into this theoretical framework
48:21 and assess what
48:22 they are implying from,
48:24 from for productive efficiency.
48:25 I think that's,
48:26 that's a good conclusion moving forward.
48:28 So,
48:29 let me stop here and now uh go to Mary.
48:31 Thank you everyone.
48:32 Uh you can find all the uh the replication files for,
48:36 for the paper that underlies this talk in,
48:39 in this repository here.
48:41 Uh,
48:41 of course,
48:42 you will have access to the slides.
48:44 And uh if you want to play around with the figures and the slides themselves,
48:48 they're,
48:48 they are posted in this other repository,
48:50 so,
48:50 um.
48:51 Have fun with that.
48:53 On to you,
48:53 Mary.
48:57 Thanks,
48:57 Roberto.
48:58 Um,
48:58 thank you for what I thought was a really clear talk.
49:01 So,
49:01 um,
49:02 I see there's a,
49:03 there's a,
49:03 there's a number of questions that are coming in in the chat.
49:05 What I will do after Mary's,
49:07 uh,
49:07 uh,
49:08 intervention and after giving Roberto a chance to maybe pick up on one
49:12 or two of Mary's points if he feels like he wants to,
49:15 I will call out your names and,
49:17 um,
49:18 and ask you to unmute yourself and ask your question,
49:20 and we'll do that in sort of groups of
49:23 2 or 3 questions at a time.
49:25 So Mary,
49:25 over to you.
49:27 Great.
49:27 Uh,
49:28 well,
49:28 thank you very much,
49:29 and indeed,
49:30 uh,
49:30 uh,
49:31 congratulations to Roberto.
49:32 Um there is an underlying paper which,
49:34 as he says,
49:35 goes through this formally,
49:36 but the intuition here has been very clear.
49:39 Uh,
49:39 so I think,
49:40 uh,
49:40 for those who want to delve into more formal proofs,
49:43 there,
49:43 there is that,
49:44 um,
49:45 the link that you can,
49:46 you can follow.
49:47 Uh,
49:48 so what I wanted to do was just pick up,
49:50 um,
49:50 on a few different,
49:51 uh,
49:52 elements of this.
49:53 Um,
49:54 sorry,
49:55 ah,
49:56 to be able to sort of highlight some things that I think are really useful about this,
50:01 um,
50:01 raise some of the issues where I think it would be interesting,
50:04 ah,
50:05 to delve a little bit deeper,
50:06 some of which you have a little bit in the appendix,
50:08 some may be in other,
50:10 in other work,
50:10 including potential future work.
50:12 But I think what,
50:13 what is really useful in this is you have motivated it with
50:17 incredibly important differences and we know
50:19 that trying to understand
50:21 uh what the drivers are in the very large differences across country,
50:26 in,
50:26 in productivity,
50:27 in income are critical for understanding development,
50:30 and so what you've done in this
50:32 is made some assumptions,
50:33 um,
50:34 but made a model that is sort of tractable to
50:36 try and focus on key dimensions and bring them out,
50:39 uh,
50:40 and while a lot of the data is
50:43 More cross section and static,
50:44 you have tried to bring in some of that,
50:47 ah,
50:47 sort of dynamic
50:48 dimensions,
50:49 including this emphasis on entry.
50:52 Um,
50:53 and I think,
50:53 you know,
50:53 kudos to you both in your talk and in the paper,
50:55 there is real transparency on some of those,
50:57 um,
50:58 assumptions,
50:59 and obviously,
51:00 it is a model that's trying to,
51:01 to clarify dynamics within it
51:04 that rather than necessarily assuming,
51:06 um,
51:06 that they are perfect,
51:08 uh,
51:08 summaries of what's happening,
51:10 um,
51:10 in the world.
51:12 One of the things that I think is the,
51:13 is the real contribution is
51:15 trying to sort of look at different types of
51:19 um sources of frictions and distortions
51:21 and not assuming that they're necessarily all working in the same way,
51:25 right?
51:25 So,
51:26 as you ended with the policy on the,
51:27 on the sort of,
51:28 on the
51:29 resource sort of misallocation,
51:31 there's a range of different potential policies that could be.
51:34 Ah,
51:34 but what you have done in this is sort of
51:36 differentiate the kinds of things that may be skewing firms
51:40 to be smaller
51:41 from those that might be skewing them to be larger,
51:44 and to get a sense of the relative importance
51:46 of these two different sort of forces,
51:49 ah,
51:49 in being able to understand and explain some
51:52 of the differences in performance across country,
51:55 as well as making a case for the importance of looking at them together.
51:58 Um,
51:58 that if you take just sort of an isolation of one dynamic,
52:01 you can,
52:02 um,
52:03 misunders,
52:05 you can get misguided,
52:06 uh,
52:06 sense of the,
52:07 the relative importance,
52:09 and that the,
52:09 the interaction between them,
52:11 um,
52:11 itself is something that's important.
52:14 Also appreciate that while this is sort of a model that you really
52:16 do take it to the data as you can to calibrate it,
52:19 both by using the firm level data,
52:21 ah,
52:21 as well as then taking the doing business indicators
52:24 as agreed,
52:25 one of the certainly most used,
52:27 um,
52:28 Sort of measures of trying to understand uh the,
52:31 the sort of the costs and extent of regulatory barriers across
52:34 countries.
52:35 Ah,
52:36 it does an OK but not great job in terms of what you're able to show,
52:41 um,
52:41 and again,
52:42 sort of a little bit careful on how we take some
52:44 of our empirical measures of different sources of friction or distortion
52:48 that they can be a proxy,
52:50 ah,
52:50 but a single measure is unlikely to be capturing,
52:53 ah,
52:53 the,
52:53 the range of different,
52:54 um,
52:55 kinds of
52:56 Distortions that there,
52:57 there may be,
52:58 and I,
52:58 I'm gonna come back to that particular one,
53:01 in,
53:01 in just a minute.
53:03 Ah,
53:04 the other is,
53:05 you know,
53:05 really trying to take advantage of microdata and very much like
53:08 this sort of micro to macro link back and forth,
53:11 um,
53:12 you know,
53:12 you've been able to amass a,
53:14 a,
53:14 a pretty impressive number of firm-level data sets at the same time
53:18 as my last point,
53:19 on my last slide,
53:20 there's a huge data agenda to be able to expand this and
53:23 the sort of lack of data to be able to do this for more countries is itself,
53:28 um,
53:28 I think an important message.
53:32 All right.
53:33 Ah,
53:34 in terms of some of the comments that I wanted to sort of
53:37 pick up on and some of which you may be able to elaborate,
53:39 um,
53:40 in the Q&A
53:41 is sort of interpreting the size of some of these barriers to entry in,
53:45 into the sort of misallocation
53:47 and what the implications for policy would be.
53:49 So you sort of start
53:51 with
53:51 sort of a distinction between how much the differences are accounted for by factors
53:56 versus TFP and that the,
53:57 the correlation with TFP is,
54:00 with,
54:00 with differences in income is much,
54:02 um,
54:02 much higher.
54:03 Um,
54:04 that some of the gaps in the factors,
54:07 ah,
54:08 are quite different,
54:09 but one of the huge questions is quality,
54:12 um,
54:12 and this matters both in the quality of capital,
54:15 the quality of human capital,
54:17 as well as the quality that's measured in,
54:19 in the outputs.
54:21 So,
54:21 one of the things you are also sort of saying,
54:23 if you look at the variation in firm size that indeed in lower-income countries,
54:27 firms on average are smaller compared to high-income countries,
54:31 and there's significant variation,
54:32 but it is much smaller variation than there is in this sort of value added per worker
54:36 or in,
54:37 in incomes across country.
54:38 So,
54:40 You're using this um average size to try and help explain the other,
54:43 but there's very different orders of sort of magnitude of,
54:47 of variation that's being used there.
54:49 And I would note just even in the countries that you have that the
54:52 variation in firm size amongst the higher
54:54 income countries is actually pretty considerable,
54:57 and a number of high-income countries have the same or even lower
55:00 average firm size than some in lower income countries.
55:03 So there's a range of things going on and obviously you
55:06 take this in different steps,
55:07 um,
55:08 and how much
55:09 there might be distortions pushing,
55:11 ah,
55:12 In,
55:12 in higher income countries,
55:13 some of those distortions may be smaller,
55:15 so there's sort of little need for offsetting to try and,
55:17 and rematch it to the,
55:18 the size,
55:19 uh,
55:19 distribution that you have in the data.
55:22 But I think this issue of quality is one that comes up a lot in the literature.
55:26 Ah,
55:26 so Jan De Luker,
55:27 as you mentioned in your papers,
55:28 he has a series of papers,
55:30 you know,
55:30 within the bank,
55:31 um,
55:31 the flagship on,
55:32 on productivity,
55:33 uh,
55:34 that,
55:34 uh,
55:35 Anna and Bill,
55:35 um,
55:36 so Casalito and Maloney have done,
55:38 sort of take issue with this again.
55:40 That a lot of what gets ascribed in the assumptions under the X Lina model
55:46 really also it,
55:47 it's attributing to all the differences in quality and market power and prices
55:52 to frictions in the economy and that,
55:54 that is therefore overestimating how much
55:57 these sort of policy frictions there may be as opposed to
56:00 things that are happening sort of internal to the firm.
56:03 And to the extent for your paper that that's true,
56:05 then in some sense,
56:07 it's overestimating how much you need entry barriers to push
56:10 back against that to give you the firm size distribution that you have,
56:14 and so if you were to take more
56:18 of some of these quality and market power differences into account,
56:21 perhaps some of your entry barriers similarly
56:24 might be um a little bit smaller um
56:27 than what,
56:27 what you're estimating.
56:30 Ah,
56:30 a last sort of point to me that would be really interesting,
56:32 you've got all this firm level data,
56:34 ah,
56:34 and you use the sort of average firm size,
56:36 but being able to use other moments in,
56:39 in the distribution,
56:40 how skewed they might be,
56:41 as well as some of the differences across subsectors,
56:44 um,
56:44 might be able in other work to be able
56:46 to tease out more of these different patterns and,
56:50 and understand
56:51 the,
56:51 the role that they may be playing.
56:54 So I'm just gonna pick up a little bit on,
56:55 on some of these,
56:56 um,
56:58 because you're making a,
56:59 a lot about how these different distortions are pushing
57:02 average firm size in one
57:04 direction or the other,
57:05 but it would be interesting to know a bit more on
57:07 how much it really is sort of the distribution that shifts versus
57:10 a skewing of the distribution and,
57:12 and where in that distribution,
57:14 this,
57:14 this potential skewedness is happening.
57:18 And
57:18 particularly for a lot of the dynamics
57:20 um that we're interested in and what's driving
57:22 a lot of the aggregate productivity is what happens in the upper right-hand tail,
57:26 and so understanding how much these distortions are working,
57:29 not so much on average,
57:31 but at the upper right-hand tail would be very much of interest.
57:35 So sometimes when people hear about size distribution,
57:37 they think,
57:38 oh,
57:38 missing middle,
57:39 you're not really talking about that.
57:41 Ah,
57:41 there's other sort of literature about the benefits of large firms,
57:44 um,
57:45 at the same time we know there can be difficulties if you have
57:48 two large firms,
57:50 um,
57:50 particularly ones that may not be competitive and may be
57:53 benefiting from protection of some sort or another or a lack of competition,
57:57 ah,
57:58 such that they are not in fact innovative and break down
58:01 some of the sort of correlation between size and productivity.
58:05 And so being able to disentangle a little bit
58:08 of how
58:09 the size distribution is interacting is something that
58:13 I think would be really helpful for policymakers
58:16 to understand
58:17 what's happening on that
58:18 upper tail on the right,
58:20 uh,
58:20 and the healthiness of some of the dynamics in,
58:22 in those kinds of firms that are critical
58:25 in driving,
58:25 um,
58:26 a lot of the,
58:26 the,
58:27 what you're going to end up seeing in aggregate because
58:29 they contribute,
58:30 um,
58:31 so much to that.
58:33 Uh,
58:33 and then this is outside this particular paper,
58:35 the data you have is not panel.
58:37 Um,
58:38 in the,
58:39 in the paper,
58:39 you're at,
58:40 at various points,
58:41 particularly when you're building it in the,
58:42 in the more simplified version,
58:43 you're having to take as exogenous various
58:45 probabilities of,
58:47 of some of the growth and,
58:48 and exit.
58:49 You do endogenize some of that in the,
58:51 in the annex,
58:52 but I think when we look at this in the empirics,
58:54 there's,
58:55 this is hugely important in terms of understanding what
58:57 the some of the policy advice would be.
59:00 Ah,
59:00 so I recognize that's outside the scope of your paper,
59:02 but when thinking about what the policy implications,
59:04 there is a lot of work
59:05 that is done on those dimensions
59:07 that I think is,
59:07 is really important to bring in as well.
59:10 Ah,
59:11 so turning to what we have is some of the,
59:13 the policy measures,
59:14 um,
59:15 so the doing business is one,
59:17 and,
59:17 and you bring that to the data and you sort of
59:19 look at how your estimates of entry barriers compare to it
59:23 and find some significant gaps.
59:25 I,
59:25 I want to give doing business credit,
59:27 um,
59:27 that it is really in a very comparable and as
59:30 transparent as possible way trying to measure these things,
59:33 um.
59:34 But I would say it is literally the costs of registering a business.
59:38 There can be lots of other entry costs
59:40 outside those literal registration costs that can be really critical.
59:44 So the fact that it's underestimating it
59:47 may well be true because it's not trying
59:49 to measure the full other kinds of regulatory other barriers that might be there.
59:54 Ah,
59:54 I think in my own work I've also tried to sort of just
59:56 to make a distinction between the de facto and the de jure,
59:59 so it is largely measuring the de jure,
1:00:01 what's on the books.
1:00:02 What happens in practice can be quite different,
1:00:05 uh,
1:00:05 and similar to you,
1:00:06 in the highest income countries,
1:00:08 that correlation is fairly good.
1:00:09 When you get into more middle,
1:00:11 lower,
1:00:11 middle,
1:00:12 low income countries,
1:00:13 the gap between what's on the books and what happens in practice is big,
1:00:17 but in this case,
1:00:18 Sort of in the opposite direction that it's,
1:00:19 it's actually not as bad as it is in the books.
1:00:21 Firms are able
1:00:22 on average to do it more quickly and at lower cost.
1:00:25 However,
1:00:26 and it's not shown in,
1:00:27 in this particular graph,
1:00:28 the variation within country
1:00:30 is much larger than the variation in averages across country.
1:00:34 So there's enormous heterogeneity within countries as to how they
1:00:38 experience
1:00:39 the regulatory regime,
1:00:40 and I think that dimension.
1:00:42 is also something that,
1:00:43 that matters and can help explain some of
1:00:45 these frictions and differences in performance across firms,
1:00:49 because all firms are not treated the same,
1:00:51 size can be one,
1:00:52 but it's not the only dimension um that can vary how ah
1:00:56 firms are interacting with the,
1:00:57 with the policy environment.
1:01:00 And then I just wanted to bring in some other work I'm doing with uh colleagues,
1:01:05 um,
1:01:06 ah,
1:01:07 in,
1:01:07 in FCI that are looking at the services sector.
1:01:10 So this is work with Garav Nayir and Elwyn Davies,
1:01:13 and I'm gonna be drawing in particular on some of the background
1:01:16 work that uh Elwyn Davies and Reya Saturito have been doing,
1:01:20 using similarly firm-level data,
1:01:22 um,
1:01:22 but this time on the service sector,
1:01:24 for which data is even harder to get,
1:01:26 uh,
1:01:26 than for manufacturing.
1:01:28 But just sort of to give some more stylized facts,
1:01:30 um,
1:01:31 on
1:01:31 what is an even larger section of the economy for,
1:01:34 for most developing countries,
1:01:36 and one thing that's true on,
1:01:38 on the size of firms,
1:01:39 as you show with,
1:01:40 with manufacturing,
1:01:41 the firms tend to be larger and higher income countries,
1:01:44 um,
1:01:45 but it's also,
1:01:45 so that's true
1:01:46 also comparing manufacturing to services,
1:01:49 manufacturing firms are larger than services,
1:01:51 and that the gradient at which firm size rises
1:01:54 is
1:01:54 Higher for um
1:01:56 manufacturing than service sector.
1:01:58 So there are some differences in how some of
1:02:00 these kinds of distortions in policy environment may vary
1:02:03 um
1:02:04 different sectors.
1:02:05 There are also differences within the sector as to how important uh size may be
1:02:10 as a,
1:02:10 as a measure of firm
1:02:11 performance.
1:02:13 Also that there's a lot of differences underneath
1:02:15 these broad categories of either manufacturing or services.
1:02:20 Ah,
1:02:20 there are different kinds of scale economies that can matter,
1:02:23 um,
1:02:23 for which diff again and sensitivity to different types of
1:02:27 frictions that may matter,
1:02:29 and a lot of times when
1:02:31 going empirically to the data,
1:02:32 you know,
1:02:33 going back to Rajan Zingalas and other kinds,
1:02:35 we sort of think of natural rates of,
1:02:36 of.
1:02:37 Ah,
1:02:38 needing access to finance or or the extent to which entry barriers may vary,
1:02:43 but something that in the firm level data,
1:02:45 you might be able to take again in other works sort of more advantage
1:02:49 of using some of that difference across sectors
1:02:51 to help tease out some of these differences
1:02:54 and not just purely in
1:02:56 the large aggregate.
1:02:58 One other thing that we were sort of interested in was looking over the life cycle of,
1:03:03 of firms
1:03:04 where often there's a,
1:03:05 a lot of emphasis on manufacturing as
1:03:08 um sort of prime drivers of,
1:03:10 of growth.
1:03:11 This is looking at size unemployment,
1:03:13 ah,
1:03:13 the blue bars are manufacturing,
1:03:15 and this is for
1:03:17 7 of the countries for which we have panel data,
1:03:20 um.
1:03:21 And the set of countries are sort of an eclectic mix because that's sort of where the,
1:03:24 the data in fact is available.
1:03:26 You do see in many cases manufacturing firms both being a
1:03:29 little bit larger and growing over that initial life cycle,
1:03:33 but you have differences across sector and in,
1:03:36 in
1:03:36 with services and ah within subsectors in services,
1:03:40 so there's certainly potential there.
1:03:42 And when you look at,
1:03:43 and this is just purely value-added per worker,
1:03:45 not trying to do more sophisticated,
1:03:47 um,
1:03:48 measures of productivity in this,
1:03:50 that some of the service sectors actually perform um
1:03:53 better than one might be expected,
1:03:55 given a lot of the emphasis,
1:03:57 uh,
1:03:58 a lot of,
1:03:58 a lot of the literature puts on,
1:04:00 on sort of manufacturing being a driver of both productivity,
1:04:03 um,
1:04:03 and jobs.
1:04:04 So I think there's a lot of interesting questions
1:04:06 sort of broadening the sort of coverage of firms.
1:04:09 One other last sort of point um here that's been
1:04:12 interesting that we've been playing with is this link between
1:04:14 size and productivity,
1:04:17 ah,
1:04:17 and obviously in the manufacturing literature and in a
1:04:19 lot of the assumptions you have in your work,
1:04:21 assuming very strong positive correlation between the two.
1:04:25 But that extent of positive correlation in the
1:04:27 service sector seems to be less strong,
1:04:30 that the sort of potential for scale economies in scaling up
1:04:33 to workers is quite different for a number of service sectors
1:04:37 and other measures of quality
1:04:39 of ah replication of establishments rather than a firm size,
1:04:44 but that even some more micro firms can in fact be quite productive.
1:04:47 So some of the
1:04:49 Broadening of your approach
1:04:52 outside of just the manufacturing sector to services,
1:04:54 we need to take some of this,
1:04:55 um,
1:04:56 sort of differences in the internal dynamics
1:04:58 and scale economies in some of the sectors
1:05:01 would be something also,
1:05:02 um,
1:05:02 that might be interested,
1:05:04 interesting to explore.
1:05:06 So last,
1:05:07 last slide,
1:05:08 uh,
1:05:08 is just a data agenda,
1:05:09 so I,
1:05:10 I think,
1:05:11 as always,
1:05:11 ah,
1:05:12 researchers will come when we need more and better data,
1:05:14 ah but I do think this is an area that is,
1:05:17 is true and it's been true for years,
1:05:19 ah,
1:05:19 and working in the service sector underscores it even more,
1:05:22 um,
1:05:23 but really sort of the collaboration that you've mentioned with colleagues in,
1:05:27 in the,
1:05:28 in the region.
1:05:29 And working with our,
1:05:30 our clients to really bring this data out,
1:05:33 to expand our ability to understand what is happening
1:05:36 and some of the differences and nuances across countries
1:05:38 is really critical
1:05:40 and particularly if you can do this over time,
1:05:42 that's where you really can get much more emphasis on the dynamics,
1:05:46 how firms react,
1:05:47 how they react to policy changes
1:05:49 to be able to understand it.
1:05:51 And I think the kinds of measures we need also need improvements,
1:05:54 particularly on prices and quality.
1:05:56 Uh,
1:05:57 this is to be able to improve how we can measure productivity,
1:06:00 um,
1:06:01 but in a range of different sectors that,
1:06:03 that dimension is,
1:06:03 is really critical to be able to,
1:06:05 to pin down more precisely,
1:06:07 and I'm saying in services it's missing even more.
1:06:10 Uh,
1:06:10 when we look at sort of the expansion of digital technologies,
1:06:13 I think it's raising even further,
1:06:14 uh,
1:06:15 challenges and estimation,
1:06:16 more so probably in services and manufacturing when you
1:06:19 have a lot of things that are quote free.
1:06:21 Um,
1:06:21 a lot of our estimation,
1:06:23 uh,
1:06:23 gets,
1:06:24 gets more difficult.
1:06:25 Ah,
1:06:26 I also think,
1:06:26 you know,
1:06:27 and I think that doing business is,
1:06:28 is a really important one.
1:06:29 I think the work,
1:06:30 um,
1:06:31 that's being done on the enterprise surveys too to measure some
1:06:34 of these different kinds at the firm level of distortions and barriers
1:06:37 matter,
1:06:38 um,
1:06:38 what they face in practice.
1:06:40 Not just sort of what's on the books,
1:06:42 but I think this is an area again where the
1:06:44 World Bank is very well positioned to expand our data efforts
1:06:47 to be able to help us,
1:06:48 ah,
1:06:49 do better in,
1:06:49 in informing and um
1:06:51 bringing really rigorous evidence
1:06:53 to our clients about how and why policy matters
1:06:56 and how it's impacting firms in,
1:06:58 in practice.
1:07:00 Um,
1:07:00 so thank you very much,
1:07:01 uh,
1:07:01 for this.
1:07:02 I think it's a really important agenda
1:07:04 and you're bringing some of both theoretical and some empirical,
1:07:08 um,
1:07:08 You know,
1:07:09 heavy lifting and thinking to this agenda,
1:07:11 so kudos to you.
1:07:12 Thanks.
1:07:15 Thanks,
1:07:15 Mary,
1:07:16 for those great comments,
1:07:17 uh,
1:07:17 both to the sort of the specifics of the,
1:07:19 of the presentation and the content of the papers,
1:07:21 but then also sort of this bigger picture about where some of this could go and,
1:07:26 um,
1:07:26 some of the things that it might be,
1:07:27 that might be interesting to,
1:07:28 to look at.
1:07:29 So before opening it up to general Q&A,
1:07:32 um,
1:07:33 Roberta,
1:07:33 do you have any kind of,
1:07:34 a couple of points that you want,
1:07:36 might want to respond to from Mary?
1:07:37 I mean,
1:07:38 she laid out a big agenda,
1:07:39 so I don't expect you to go through the whole thing,
1:07:41 but
1:07:41 a couple of the key important ones if you,
1:07:43 if you'd like to react to.
1:07:44 Yeah,
1:07:45 thanks,
1:07:45 thanks,
1:07:45 Dean.
1:07:46 Just,
1:07:46 just a couple of remarks.
1:07:47 I,
1:07:47 I,
1:07:47 I,
1:07:48 I was,
1:07:49 um Ryan,
1:07:50 if I could share again,
1:07:52 because I,
1:07:52 I thought about some of the extensions or,
1:07:55 or follow-up questions that,
1:07:57 that Mary asked and I have an actual answer for them,
1:07:59 so maybe I should just focus on those and,
1:08:02 and then we can move forward.
1:08:04 But,
1:08:04 uh,
1:08:04 Mary,
1:08:05 thank you so much.
1:08:05 Yeah,
1:08:05 I mean,
1:08:06 it's amazing what,
1:08:07 what,
1:08:07 what your discussion and,
1:08:08 and,
1:08:09 and the agenda you are working on as well.
1:08:11 So,
1:08:11 um,
1:08:13 I,
1:08:13 I appreciate your comments.
1:08:15 So,
1:08:15 uh,
1:08:15 again,
1:08:16 I'm gonna focus on one of the questions you asked is
1:08:19 what about looking at
1:08:20 the size distribution differences across countries beyond just the average size,
1:08:26 and that's something I've looked at and actually
1:08:29 one can think that looking beyond the average size
1:08:32 is like a measure of validation for my approach.
1:08:36 Because
1:08:37 by definition,
1:08:38 the average size is going to be matched in
1:08:40 the model thanks to the distortion I'm feeding in,
1:08:44 but what about non-targeted moments?
1:08:46 And uh I have a figure with,
1:08:48 with that
1:08:49 uh here in the end.
1:08:52 So,
1:08:53 here is what the,
1:08:54 I mean,
1:08:54 this is again evidence of validation.
1:08:56 I'm looking at what the model implies
1:08:59 for the share of firms at the top of the size distribution,
1:09:03 meaning how many firms there are with more than 250 workers,
1:09:07 and,
1:09:07 and how many,
1:09:08 and what's that share in the data.
1:09:11 And of course the goodness of fit is,
1:09:12 is not meant to be perfect because this wasn't a targeted moment,
1:09:16 but I would argue from this figure that
1:09:18 That,
1:09:19 uh,
1:09:19 when looking at,
1:09:20 at,
1:09:20 at the,
1:09:21 at the right tail of the distribution in each country,
1:09:24 the,
1:09:24 the,
1:09:24 the model is able to,
1:09:26 to populate the right tail commensurably with,
1:09:29 with,
1:09:29 with what the,
1:09:30 what the data are showing.
1:09:32 And lastly,
1:09:33 another,
1:09:33 another attribute you mentioned it would be good to have and again contrast the,
1:09:38 the model with,
1:09:38 with data is
1:09:40 what about the dynamic behavior of firms and,
1:09:42 and,
1:09:42 and you showed some life cycles of firms um
1:09:46 For services and manufacturing,
1:09:48 and that's also a piece of validation I explored,
1:09:51 um,
1:09:53 Here.
1:09:54 So again,
1:09:55 this is a non-targeted moment but because
1:09:58 the model
1:09:59 offers an endogenous
1:10:01 innovation channel,
1:10:03 firms that enter will endogenously grow over time
1:10:07 as a function of how much they are encouraged to do so in,
1:10:11 in,
1:10:11 in,
1:10:11 in the business where they live.
1:10:13 And here is to show that if you think about the US and,
1:10:16 and France,
1:10:17 for example,
1:10:17 in,
1:10:17 in the model,
1:10:18 you would expect that
1:10:19 after 40 years,
1:10:21 firms uh scale up in an order of 5-fold relative to
1:10:26 the average size of an entering cohort,
1:10:28 but in,
1:10:29 in the,
1:10:29 in the various other countries this
1:10:32 again consistent with,
1:10:33 with the evidence and
1:10:35 at additional manufacturing of life cycles.
1:10:37 Uh,
1:10:37 again,
1:10:38 the,
1:10:38 the,
1:10:38 the mix of distortions in the model can generate this flattening of life cycles.
1:10:42 In other countries like,
1:10:44 uh,
1:10:44 you know,
1:10:44 in India and Ghana being even more flat than,
1:10:47 than,
1:10:47 than India.
1:10:48 So,
1:10:48 just,
1:10:48 just to mention those are elements I've thought about,
1:10:51 but I agree with you that one can,
1:10:53 one can,
1:10:53 uh,
1:10:54 11 expand on that.
1:10:56 Uh,
1:10:56 I,
1:10:56 I,
1:10:56 I have many other reactions,
1:10:57 uh,
1:10:58 beyond,
1:10:58 but um I think that,
1:11:00 uh,
1:11:00 uh,
1:11:01 I don't want to crowd out uh the
1:11:08 Are you hearing me?
1:11:10 Yeah,
1:11:10 we can hear you.
1:11:11 Oh sorry,
1:11:11 I,
1:11:11 I thought I closed the window.
1:11:13 Yeah,
1:11:13 I don't want to crowd out the,
1:11:14 um,
1:11:15 the audience from,
1:11:16 from asking questions,
1:11:18 so I just thank Mary and,
1:11:19 and I'm happy to take questions from the audience.
1:11:22 Great.
1:11:23 So what I'm gonna do is I'm gonna read out your first names,
1:11:25 uh,
1:11:25 in the order that they appeared in the chat.
1:11:27 Uh,
1:11:27 hopefully you're still around.
1:11:28 I know Andre had to leave,
1:11:30 so I won't call on,
1:11:31 on Andre,
1:11:32 but,
1:11:32 um,
1:11:33 if you don't mind unmuting yourself and,
1:11:35 uh,
1:11:36 posing your question,
1:11:36 I'm gonna take the 1st 3 and then,
1:11:39 um,
1:11:39 we'll take it from there.
1:11:40 If you don't,
1:11:41 if you aren't able to respond relatively quickly,
1:11:43 I'm gonna move on to the next person just so we don't get,
1:11:46 get hung up and,
1:11:47 and,
1:11:47 um.
1:11:48 On that.
1:11:48 So first,
1:11:48 I'm gonna call on Hector.
1:11:50 Hector,
1:11:50 are you still there and would you like to ask your question?
1:11:54 Uh,
1:11:54 yes,
1:11:55 thank you very much,
1:11:56 um,
1:11:57 for the very interesting presentation.
1:11:58 My question is with regard to the,
1:12:01 uh,
1:12:01 use of the USA as a benchmark case,
1:12:04 um,
1:12:05 as precisely against the background that there is a
1:12:08 substantial literature by now on distortions in the US.
1:12:12 Um,
1:12:13 I think,
1:12:13 uh,
1:12:13 Thomas Philippo,
1:12:14 uh,
1:12:15 wrote a book,
1:12:16 a recent book,
1:12:16 uh,
1:12:17 about this,
1:12:17 uh,
1:12:18 where you can find,
1:12:19 uh,
1:12:19 also extensive res references
1:12:21 and just as a question whether another benchmark case might not be,
1:12:26 um,
1:12:26 with less allocative distortions in the US might be more appropriate.
1:12:33 Thanks and thanks for keeping your,
1:12:34 your question focused.
1:12:35 I'm gonna ask the others to do the same,
1:12:36 just so we can get through,
1:12:37 through everybody.
1:12:38 Um,
1:12:39 next,
1:12:39 I'm gonna call on Peter.
1:12:40 Peter,
1:12:41 I think it's Rundell,
1:12:41 if I'm
1:12:42 remembering correctly.
1:12:44 Thanks very much and thank you very much indeed
1:12:46 Roberto for a really interesting piece of work and,
1:12:48 and particularly for your
1:12:50 um recognition that
1:12:52 distortions have reasons.
1:12:54 Um,
1:12:54 they're not
1:12:55 purely a bad,
1:12:56 they're good for somebody somewhere.
1:12:58 And I'd be particularly interested in your thoughts on labor market frictions.
1:13:03 Um,
1:13:03 you alluded to the hiring and firing and other regulations,
1:13:07 but there are other,
1:13:08 uh,
1:13:09 causes for imbalance in
1:13:12 and frictions in labor markets related to
1:13:15 physical mobility and other things like that.
1:13:17 I'd be very interested to know how you feel
1:13:20 those affect your analysis,
1:13:22 particularly given that,
1:13:24 uh,
1:13:25 labor is your measure of firm size.
1:13:28 Um,
1:13:29 I am aware that it's an extremely controversial political topic.
1:13:33 But it would be very helpful to understand
1:13:35 your thoughts on it.
1:13:36 Thank you.
1:13:38 OK,
1:13:38 and last question in this,
1:13:40 uh,
1:13:40 first round,
1:13:41 uh,
1:13:41 Neva,
1:13:42 you had a question
1:13:43 for clarification.
1:13:53 Neva Southbourne,
1:13:53 are you still there?
1:14:01 OK.
1:14:02 Um,
1:14:06 Oh,
1:14:06 I see you aren't
1:14:07 around anymore.
1:14:08 OK,
1:14:08 I'm gonna go to the next person then,
1:14:10 which is Carlos.
1:14:14 Hi,
1:14:14 thank you,
1:14:14 Roberto,
1:14:15 for a great presentation.
1:14:16 Uh,
1:14:16 I was just curious about,
1:14:17 you know,
1:14:18 finding this,
1:14:19 uh,
1:14:20 effect on,
1:14:21 uh,
1:14:21 entry barriers,
1:14:23 um,
1:14:23 and being entry barriers,
1:14:25 uh,
1:14:25 a manifestation of lack of competition.
1:14:28 Um,
1:14:28 how do you reconcile the fact that this,
1:14:31 you know,
1:14:31 lack of competition,
1:14:32 uh,
1:14:33 within a theoretical model where,
1:14:35 uh,
1:14:35 competition,
1:14:36 I mean,
1:14:36 this,
1:14:37 this assumed to be,
1:14:37 uh,
1:14:38 Uh,
1:14:38 competition.
1:14:39 Uh,
1:14:40 I,
1:14:40 I was just,
1:14:40 you know,
1:14:41 trying to get my,
1:14:41 my head around that,
1:14:42 that fact and finding
1:14:44 an important effect of,
1:14:46 uh,
1:14:46 uh,
1:14:47 entry barriers and within a competitive model.
1:14:52 Great,
1:14:52 so Roberto,
1:14:53 back to you.
1:14:54 Uh,
1:14:54 Mary,
1:14:54 if you wanted to weigh in on any of these questions,
1:14:57 um,
1:14:57 just sort of wave to me physically and I'll,
1:14:59 I'll,
1:14:59 I'll,
1:15:00 I'll catch your,
1:15:00 your,
1:15:01 your,
1:15:01 I'll call on you.
1:15:03 Yeah,
1:15:03 please,
1:15:03 Mary,
1:15:03 help me out if you,
1:15:04 if you have a good answer.
1:15:06 So,
1:15:06 uh,
1:15:07 Hector,
1:15:07 uh,
1:15:08 I agree,
1:15:08 so,
1:15:08 um,
1:15:10 uh,
1:15:10 the,
1:15:10 the
1:15:12 At the root of all these strategies when you want to learn about uh frictions,
1:15:16 you have to sort of uh start from some benchmark that you think is closest to,
1:15:21 to
1:15:22 an undistorted scenario and
1:15:24 uh I acknowledge in my talk and I agree with you that the US uh especially recently,
1:15:29 uh it's far from,
1:15:30 from that.
1:15:31 So,
1:15:32 Um,
1:15:34 11 answer is that,
1:15:35 well,
1:15:35 I'm not sure there is a better alternative.
1:15:38 And second is that everything I say about other countries being,
1:15:42 uh,
1:15:42 for example,
1:15:43 when I,
1:15:43 when I mentioned uh the lack of evidence of,
1:15:46 of big distortions in the other developed countries,
1:15:49 I am sort of netting out
1:15:52 the
1:15:53 degree of distortions that there are in the US.
1:15:55 So,
1:15:56 uh,
1:15:56 when I,
1:15:56 when I define the distortions everywhere else in the world,
1:16:00 I am,
1:16:01 I'm,
1:16:01 I'm sort of differentiating them with
1:16:04 distortions that the US exhibits.
1:16:06 So this is residual distortion above and beyond
1:16:09 the distorted in the,
1:16:11 in,
1:16:11 in the,
1:16:11 in the US environment.
1:16:13 But I agree,
1:16:13 I mean,
1:16:14 if you,
1:16:14 if you have a better proposal uh that I could actually measure,
1:16:17 uh,
1:16:18 I would be,
1:16:18 I'm happy to hear.
1:16:20 Uh,
1:16:20 to Peter,
1:16:21 um,
1:16:21 labor market policies,
1:16:23 I,
1:16:23 I,
1:16:23 I discussed a couple of the best papers I've known in the literature that have,
1:16:27 um,
1:16:28 explored the role of labor market policies,
1:16:31 uh,
1:16:32 in a disciplined way,
1:16:33 you know,
1:16:33 so these are papers that were able to
1:16:36 estimate the labor policy from some feature of the data
1:16:41 in France in one case and in the other case,
1:16:43 I think it's overall in,
1:16:45 in Europe.
1:16:46 And the finding,
1:16:47 uh,
1:16:47 uh,
1:16:48 I mean,
1:16:48 I,
1:16:48 I think that the predominant view is that these are important,
1:16:51 of course,
1:16:51 but
1:16:52 relative to the magnitude of,
1:16:54 of misallocation that,
1:16:56 that I showed you exists in,
1:16:57 in many poorer countries,
1:16:59 it doesn't seem that labor market frictions on their own
1:17:03 are,
1:17:03 are,
1:17:03 are have a big bite.
1:17:05 Uh,
1:17:06 and,
1:17:06 but I should say that there's a dimension of,
1:17:08 of misallocation associated with,
1:17:10 with the labor input that they have been silent about,
1:17:13 but other scholars do have explored,
1:17:16 which is spatial misallocation.
1:17:18 And in that case,
1:17:19 there is big evidence that um countries that
1:17:22 where labor mobility within a country is hindered by,
1:17:25 by transportation barriers or whatever
1:17:28 that could leave on the table uh
1:17:30 uh uh uh more,
1:17:31 more,
1:17:32 uh more,
1:17:33 you know,
1:17:33 more opportunities to,
1:17:34 to gain.
1:17:35 So
1:17:35 I can point you to,
1:17:36 to the literature I'm aware of in this regard,
1:17:39 but uh thanks for asking that,
1:17:40 that exists and I'm silent about.
1:17:43 And lastly,
1:17:44 Carlos,
1:17:44 I'm not sure I understood,
1:17:45 so
1:17:46 my analysis,
1:17:47 there is an explicit consideration of the link between entry bars and competition.
1:17:53 So when entry barriers are high,
1:17:56 the sense in which there is less competition is that
1:17:59 wages fall and because wages fall,
1:18:02 this allows
1:18:03 the surviving firms to scale up and become too big.
1:18:06 Um,
1:18:07 so,
1:18:07 that channel of,
1:18:08 of lack of competition again is,
1:18:10 is at the core
1:18:12 of this rise in,
1:18:13 in the frame size distribution that I am,
1:18:15 uh,
1:18:15 documenting.
1:18:16 But again,
1:18:17 I'm not sure I fully captured the,
1:18:18 the,
1:18:19 the question,
1:18:19 uh,
1:18:20 what you had in mind.
1:18:27 Um,
1:18:28 well,
1:18:28 maybe this next question might allow you to,
1:18:30 to elaborate.
1:18:31 Um,
1:18:32 uh,
1:18:32 Alejandro Espinoza Wang,
1:18:34 did you,
1:18:35 you still there?
1:18:37 Uh,
1:18:37 yes,
1:18:38 you know,
1:18:38 I,
1:18:38 I just want,
1:18:38 I think Mary mentioned it too,
1:18:40 and Roberto,
1:18:41 thank you so much for the presentation.
1:18:42 I,
1:18:43 I just wanted to,
1:18:43 to,
1:18:44 to raise the issue that uh maybe some of the anti-competitive uh
1:18:49 regulations like that,
1:18:50 that are not measured by the Doing Business report,
1:18:53 maybe you can look at data from the
1:18:55 product market regulation database from the OECD to,
1:18:58 to try to,
1:18:59 uh,
1:19:00 to look for more explanations on,
1:19:02 on these frictions.
1:19:02 So I just wanted to,
1:19:03 to,
1:19:04 to flag that that maybe looking at that data could,
1:19:06 could help complement the picture that you're,
1:19:08 you're presenting.
1:19:09 Thank you.
1:19:12 Thanks.
1:19:13 And next,
1:19:13 um,
1:19:14 I'm gonna call on Sheammadas.
1:19:15 I hope I'm getting that,
1:19:16 uh,
1:19:17 pronunciation correct.
1:19:18 Um,
1:19:19 I,
1:19:19 I saw Sheammadas you had quite a few questions,
1:19:21 so maybe if you could limit it to sort of the one or two most,
1:19:24 uh,
1:19:25 important ones in your own mind.
1:19:26 Thanks.
1:19:40 It looks like Shimaas may have left.
1:19:42 OK,
1:19:42 I'm checking.
1:19:43 Yeah,
1:19:44 no longer logged in.
1:19:45 OK,
1:19:45 so let's go to Giovanni.
1:19:47 Are you still here?
1:19:52 Uh,
1:19:52 it looks like I Giovanni di Placido,
1:19:55 but you're still,
1:19:55 I see you're still here,
1:19:56 but.
1:19:58 Would you like to ask a question?
1:20:03 I
1:20:05 OK.
1:20:08 And then,
1:20:09 OK,
1:20:10 I'll move on to the next person,
1:20:11 Shahidul.
1:20:18 Shahidul,
1:20:18 are you still around?
1:20:20 Hello,
1:20:20 Shani hear me?
1:20:22 Yes,
1:20:22 we can hear you.
1:20:22 We can hear you now.
1:20:23 Yes,
1:20:23 go ahead.
1:20:24 OK,
1:20:24 uh,
1:20:24 let me show you some actually,
1:20:25 I'm a PhD student in China.
1:20:28 I'm originally from Bangladesh,
1:20:30 so I have a couple of questions.
1:20:32 Uh,
1:20:32 one is,
1:20:33 uh,
1:20:33 you know,
1:20:34 the developing countries are,
1:20:35 uh,
1:20:36 I mean,
1:20:36 they have high informal sector.
1:20:39 So entry barrier here,
1:20:40 you know,
1:20:41 uh,
1:20:42 as you explained.
1:20:44 Uh,
1:20:44 that
1:20:45 large informal sector is often associated with high entry cost
1:20:50 and
1:20:51 development.
1:20:52 So from a policy point of view,
1:20:54 your point,
1:20:54 uh,
1:20:55 your paper probably can shed some light on this,
1:20:57 particularly linking
1:20:59 these three issues,
1:21:01 uh,
1:21:01 entry cost,
1:21:03 the AP gap,
1:21:03 and size of informal sector.
1:21:06 And
1:21:07 another point is,
1:21:08 hello.
1:21:11 Yes,
1:21:11 we can hear you.
1:21:13 OK,
1:21:13 my another question is on um
1:21:16 intangible forms,
1:21:17 you know.
1:21:19 High entry cost may arise from high fixed cost of certain farms,
1:21:24 uh,
1:21:24 particularly farms,
1:21:25 those heavily apply intangible assets.
1:21:28 So their productivity performance,
1:21:30 uh,
1:21:30 is not necessarily bad,
1:21:31 what we call them superst farm or winner takes all.
1:21:35 So how do you respond is that if farm
1:21:38 Uh,
1:21:39 you know,
1:21:39 need to invest a lot,
1:21:41 uh,
1:21:42 uh,
1:21:42 particularly the platform Facebook,
1:21:45 Google,
1:21:45 and these kind of companies that,
1:21:47 uh,
1:21:49 we,
1:21:49 we,
1:21:49 we,
1:21:50 we see these days.
1:21:51 So there,
1:21:52 there is a national entry barrier,
1:21:54 but it's not necessarily bad in the sense that that.
1:21:58 Fixed cost is very high.
1:22:01 Uh,
1:22:02 so how do you respond to these two issues?
1:22:03 One is informality,
1:22:05 the other one is on intangible asset funds.
1:22:07 Thank you.
1:22:11 OK,
1:22:12 Roberto,
1:22:12 back to you.
1:22:14 Thanks,
1:22:15 Dion,
1:22:15 and thanks uh
1:22:17 everyone.
1:22:17 So,
1:22:17 uh Alejandro,
1:22:18 thank you,
1:22:18 yeah,
1:22:19 uh,
1:22:19 you give me the opportunity to talk about the PMR,
1:22:21 the,
1:22:21 the product-market regulation.
1:22:23 I definitely looked at
1:22:25 the PMR,
1:22:25 uh,
1:22:26 as a potential indicator of regulations in entry.
1:22:30 The reason I ended up,
1:22:31 uh,
1:22:32 preferring the doing business as,
1:22:33 as,
1:22:34 as,
1:22:34 um,
1:22:35 informative about the regulatory environment is
1:22:38 the PMR is good for ranking countries,
1:22:40 but the units in the PMR are difficult to interpret.
1:22:44 And because it's an indicator,
1:22:46 it's an index,
1:22:47 it's therefore,
1:22:47 it's,
1:22:48 it's hard to plug back into a model and eventually
1:22:52 translate the effects of a particular indicator value
1:22:56 in the PMR in terms of productivity.
1:22:59 Whereas the doing business has this nice attribute of expressing the,
1:23:02 the costs
1:23:03 in as a function of some observable,
1:23:05 which is the income per capita of a country,
1:23:07 for instance,
1:23:09 and then one can convert those,
1:23:10 those units into something that the model speaks about.
1:23:14 So,
1:23:14 definitely the,
1:23:15 the,
1:23:15 the,
1:23:15 the PMR is useful but uh and I can show some correlations but from the point of view of
1:23:21 telling you how much of the,
1:23:23 of the gains in productivity you could reap by
1:23:26 the PMR based
1:23:28 indicator of entry regulation,
1:23:30 I just cannot answer that question.
1:23:33 Uh,
1:23:34 so,
1:23:35 Shahidul,
1:23:35 you,
1:23:36 uh,
1:23:36 you,
1:23:36 you raised very,
1:23:37 very interesting points and thank you for uh
1:23:39 giving me the opportunity to talk about informality.
1:23:42 Um,
1:23:43 so definitely the way I think about informality is,
1:23:47 uh,
1:23:47 as
1:23:48 not so much
1:23:50 important to have
1:23:52 included it
1:23:53 in the measure of average size I used to infer the distortion.
1:23:58 But I think it's important to
1:24:00 take into account as a consequence
1:24:02 that could emerge
1:24:03 in a very uh in,
1:24:05 in a high entry barrier country.
1:24:08 So,
1:24:08 in,
1:24:08 in this presentation,
1:24:09 I,
1:24:09 my focus was on showing you an inference strategy,
1:24:13 but I'm,
1:24:13 I'm working actually on follow-up work and giving collaboration with,
1:24:17 with,
1:24:18 with,
1:24:18 with uh,
1:24:19 with the country regions
1:24:20 that
1:24:21 precisely in that direction,
1:24:23 how to extend the framework to have an endogenous
1:24:26 informal sector
1:24:28 and see how much of the informality that we,
1:24:30 we can see in,
1:24:31 in,
1:24:31 in a country can be attributable to
1:24:34 this much more meaningful measures of,
1:24:36 of,
1:24:36 of barriers to entry.
1:24:38 Uh,
1:24:38 so I,
1:24:38 I agree.
1:24:39 The,
1:24:39 the other question you made is,
1:24:40 is more complicated.
1:24:41 I mean,
1:24:41 I have an answer,
1:24:42 but it's,
1:24:42 it's longer.
1:24:43 It has to do that,
1:24:44 uh,
1:24:45 yeah,
1:24:45 there,
1:24:45 there could be like technological barriers,
1:24:47 but
1:24:48 those barriers,
1:24:49 I mean,
1:24:49 those technological entry costs are also in the US.
1:24:52 So if in the US
1:24:54 firms in that uh intensive,
1:24:57 intensive sectors is of a given size,
1:25:00 why would they have to be so much smaller or so much larger elsewhere.
1:25:04 I,
1:25:04 I,
1:25:04 I'm pretty sure I'm,
1:25:05 I'm,
1:25:06 I'm gonna leave you unsatisfied with the answer,
1:25:08 but in this amount of time that as much as I can
1:25:11 uh expand.
1:25:14 Thanks,
1:25:14 Roberta,
1:25:15 and thanks everybody for an active Q&A.
1:25:17 I,
1:25:18 I,
1:25:18 I,
1:25:18 you know,
1:25:19 this is,
1:25:19 I've really enjoyed this policy research talk.
1:25:22 Um,
1:25:22 I'm gonna start bringing this to a close.
1:25:24 Um,
1:25:25 and Mary,
1:25:26 just in case you have any last couple of comments you'd like to make before we close,
1:25:31 over to you.
1:25:34 No,
1:25:34 so I really enjoyed this.
1:25:35 I think these policy talks are great as a chance to showcase the work that DEC is,
1:25:39 is doing.
1:25:40 Um,
1:25:41 no,
1:25:41 I've just,
1:25:41 uh,
1:25:42 and Roberto,
1:25:42 happy to have a follow-up conversation on some of the nittier grittier parts,
1:25:46 that would be great.
1:25:47 Um,
1:25:48 no,
1:25:48 I've just been thinking a little bit more of this sort of this use of,
1:25:50 of the United States and,
1:25:52 and as a policy non-distorted.
1:25:54 You know,
1:25:54 this,
1:25:55 this has done a lot.
1:25:56 Um,
1:25:57 and,
1:25:57 and compared to a lot of other countries,
1:25:58 I think that assumption is not bad,
1:26:00 not to say there aren't policy distortions,
1:26:02 but
1:26:02 in a relative,
1:26:03 that's fine.
1:26:05 What's true though is the production
1:26:09 sort of
1:26:10 function that's being used in the United States will not
1:26:12 in fact be the sort of same as elsewhere.
1:26:15 And you know,
1:26:16 the United States is far more advanced.
1:26:18 A lot of the manufacturing has been offshored,
1:26:20 so,
1:26:21 and far more mechanized,
1:26:23 so your firm size as measured by a worker
1:26:25 in the United States is going to be tilted smaller
1:26:29 because of all that mechanization compared to some of the other countries
1:26:32 that are going to be,
1:26:33 they're doing,
1:26:34 while it may be the same sector,
1:26:36 they're doing very different part in that value chain within it,
1:26:39 a more labor intensive work.
1:26:41 Um,
1:26:42 anyway,
1:26:42 this.
1:26:43 It's not a first order concern,
1:26:45 but it,
1:26:45 but just to say some of that,
1:26:47 and then the data as a data point,
1:26:49 you know,
1:26:49 what counts as manufacturing,
1:26:50 what counts as services,
1:26:51 because a lot of the manufacturing in the United States,
1:26:53 in fact,
1:26:53 is
1:26:54 services in the extreme.
1:26:55 Apple
1:26:56 is a manufacturer,
1:26:57 you know,
1:26:57 you make the iPhones and,
1:26:59 and computers
1:27:00 in China,
1:27:01 the design is in the United States,
1:27:03 that gets counted as manufacturing in the United States.
1:27:05 It's a mislabel of the data.
1:27:07 So,
1:27:07 I mean,
1:27:08 there's,
1:27:08 there's
1:27:09 There's more,
1:27:10 as certification of manufacturing,
1:27:12 particularly the higher end is happening,
1:27:13 the data itself is getting messier
1:27:15 at being able to make
1:27:16 some of these um distinctions.
1:27:19 Um,
1:27:19 anyway,
1:27:20 I think there's some more interesting things that one could delve into on that,
1:27:22 but I think overall,
1:27:24 um,
1:27:24 this is great work and,
1:27:26 and really have appreciated the,
1:27:28 the discussion.
1:27:29 So thanks for including me.
1:27:32 Thanks Mary.
1:27:32 Thanks for coming.
1:27:33 Uh,
1:27:33 Roberta,
1:27:34 any last comment from you or are you,
1:27:35 are you OK?
1:27:37 No,
1:27:37 I'm very grateful.
1:27:38 I,
1:27:38 I,
1:27:38 I really enjoyed this.
1:27:39 Uh,
1:27:40 I was a bit anxious,
1:27:41 but,
1:27:41 uh,
1:27:41 at the end I think it went well.
1:27:43 Uh,
1:27:43 I,
1:27:43 I,
1:27:43 I take away many,
1:27:44 many questions and things to,
1:27:46 to think about,
1:27:46 and I.
1:27:47 I am also really looking forward to this data agenda that Mary is,
1:27:51 is putting forward,
1:27:52 uh,
1:27:53 on,
1:27:53 on,
1:27:54 and the elephant in the room is the service sector.
1:27:55 It's like 70% of the majority of the countries and
1:27:59 we are focusing on manufacturing,
1:28:00 which is usually just the tiniest of the sectors,
1:28:03 but,
1:28:03 so if you can,
1:28:04 again open that other black box,
1:28:05 that would be great.
1:28:08 Yeah,
1:28:08 as researchers,
1:28:09 we can't have a talk where we don't call for more data,
1:28:11 so.
1:28:12 Um,
1:28:13 OK,
1:28:14 just before closing finally,
1:28:15 um,
1:28:16 uh,
1:28:17 I some of you may have seen Ryan put in the chat,
1:28:19 our next policy research talk is on March 23rd and has the exciting title of Trade,
1:28:24 Robots and Industrial Development.
1:28:26 So look for an invitation to that and please sign up.
1:28:29 Um,
1:28:30 so with that,
1:28:30 thank you,
1:28:31 everybody,
1:28:31 and we look forward to seeing you,
1:28:33 uh,
1:28:34 again next time.
1:28:36 Bye-bye.
- add-style
- lp-body-content