YT YouTube2Text

YouTube2Text → Transcripts → Lecture 2: Development accounting: methods and findings

Lecture 2: Development accounting: methods and findings — Transcript

by CEPR & VideoVox Economics · 14,243 words · 2,550 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:05Hello. I see people are streaming in
0:07still. I wanted to welcome you welcome
0:09you all to the
0:11second class
0:13of STEG's virtual course on key concepts
0:15in macro development.
0:17Um this is course is offered um
0:20course as you know free
0:22um from STEG's uh CEPR program
0:25which is supported by the United uh
0:27Kingdom's a foreign Commonwealth and
0:30Development Office. Uh we're thankful uh
0:32for them for their support.
0:35Um
0:37Uh last week uh Richard Rogerson gave
0:40the introductory lecture.
0:41If you uh missed that
0:43um the both the lecture slides and um
0:46the talk are available on our website uh
0:50via YouTube.
0:51Um today we're very thrilled to have
0:53Julieta Caunedo
0:54who's going to talk about development
0:57and income accounting.
0:59Um
1:00Julieta has uh said for those who are um
1:03a panelists uh Julieta has said that you
1:06can you know interrupt her with
1:08questions. That's fine just as you would
1:10in an ordinary class.
1:12Um
1:13She may you know depending on how many
1:15we have she may not be able to answer
1:16all of them she might to continue with
1:18the the lecture uh just as a normal
1:20class would be. Um but otherwise if
1:22you're not answering asking a question
1:24please mute.
1:25Um for those who are joining as
1:27participants uh you can type questions
1:30into the Q&A you can vote them up. Uh
1:33Doug Gollin uh and I um
1:36will be answering those in the Q&A.
1:39Uh there will also be a
1:42um a TA session for this lecture.
1:46Um and I actually don't know the timing
1:49on that. Julieta do you know the timing
1:50for that or Mandy Yes so it's going to
1:53be on Monday the same time.
1:56Um
1:56and I'll explain in a bit what is that
1:59session going to be about. Okay.
2:02Um
2:03and um
2:05uh Julieta, we'll leave a little bit of
2:06time at the end uh for answering
2:09questions uh as well.
2:11And I think that's it. Um
2:14well, let me thank I should I know, I
2:16should
2:17uh let me do a couple of the things. One
2:19is um
2:21um
2:22lectures notes for this lecture
2:25uh are also available uh online and this
2:28will be available uh on YouTube as well
2:31uh following the lecture. Um let me also
2:34thank uh Mandy Chan and Kirsty
2:37um McNeill at CEPR for helping organize
2:40this as well. Okay, Julieta, go ahead.
2:45All right. Thanks uh
2:46uh
2:47Daron and and Richard and Joe for for
2:50asking me to to this lecture. I'm
2:52thrilled to see that there's so much
2:54interest in the field and I'm hoping
2:57that um STACK will become a home for
3:00many students that are thinking about
3:02these issues um
3:05and I would like to have some sort of
3:07references of people that are working in
3:08in similar issues and thinking about
3:10this. Um so, welcome to hopefully your
3:13intellectual home.
3:15Um what I want to do today is to
3:18uh
3:19talk about an important issue, which is
3:22development accounting. So, let me just
3:25share my screen. So, this is your second
3:27lecture uh for the virtual macro
3:30development course. This is going to be
3:31a course on on income accounting.
3:34Um a few organizational issues is that
3:38uh Raul Morales, who's a fifth-year
3:40student at Cornell, he's going to be uh
3:43doing a TA section on Monday.
3:46Uh you will see that a lot of what I'm
3:49going to do today is going to be
3:52measurement
3:53and it will require data analysis.
3:56Um so I generated small repository that
4:01you can access and you will find the
4:02link through the slide back and also
4:06through the Stack Web page where you can
4:09see there is minimal code that will
4:11allow you to do a lot of the work that
4:13I'm going to describe today and
4:15hopefully play around with the data. So
4:17what Raul is going to do on Monday is
4:19work work with you to make sure that you
4:23questions about it if you have questions
4:24about it
4:26um
4:26he will be able to answer those for you.
4:28Okay?
4:29So um
4:31without further ado let me just
4:33uh dig in into the issue of the today's
4:35lecture.
4:36So
4:38the first thing to to
4:40that you should be asking yourself is
4:42you know, what is income accounting?
4:44So as as Richard mentioned at the
4:46beginning of of his introductory class
4:49last week
4:51he said, you know, the main question for
4:52macro development
4:54is the the you know, the observation
4:56that
4:57there are poor and rich countries in the
4:59world and those differences are
5:01persistent.
5:02And we want to understand why.
5:04Okay? But to be able to understand why
5:08we need to be able to
5:10you know, to understand the sources of
5:12those income disparities. So income
5:13accounting
5:14is going to
5:16have a particular approach to answering
5:18that question.
5:20So you know, we want to know if the
5:22reason why poor and richer countries are
5:24different is it because richer countries
5:27use more labor
5:28or because they have better educated
5:31workers
5:33or because they have more physical
5:34capital
5:36or the capital that they use is
5:37different altogether.
5:39Or even more important is it because
5:41they have similar you know, factors
5:45workers and physical capital
5:48but then somehow they managed to do
5:50better use of those inputs.
5:53Okay?
5:55So what income accounting refers to is
5:58the approach to answering those
5:59questions.
6:01So what we are going to do in income
6:02accounting is to measure
6:04to the best possible each of these
6:06ingredients,
6:08and then we're going to measure also the
6:10relative contributions
6:12of each of these ingredients.
6:14So for example, it might be that and
6:18this is not going to be true in the data
6:19just so you know, but it might be that
6:21there are large differences in physical
6:23capital, but physical capital contribute
6:26very little to differences in income
6:29across countries. So then those
6:31differences are not going to be very
6:33important in explaining
6:35uh why Malawi is poor and the US is
6:38rich.
6:39Okay?
6:40Now, spoiler alert alert here. I'm going
6:43to start from the end. I'm going to tell
6:45you what we know so far or what seems to
6:47be the consensus in the literature.
6:50The current consensus is that
6:51disparities in things that we can
6:53directly measure like capital and labor
6:57account for at most 50% of the
6:59differences in income per capita.
7:01So there is still a lot
7:04that
7:05uh
7:06you know, that we can't directly
7:08measure. We have some ideas of what is
7:10generating that additional 50%.
7:13Um
7:15but from a pure accounting sense,
7:18capital and labor only account for about
7:20half of the difference that we see.
7:23Now, there is an issue with interpreting
7:26this finding
7:28in the sense that
7:31the fact that, you know, how we use
7:33those inputs is accounting for 50% of
7:37the differences, you know, at the very
7:39least,
7:41um may feed back into accumulation,
7:43right? So,
7:45um,
7:47it is important to say that the capital
7:50and labor, or the capital and labor that
7:52we see in an economy, is not exogenous
7:54to that economy. How much you invest in
7:57Argentina depends on the environment of
8:00the economy in Argentina.
8:02Okay?
8:03And the links to the world, etc., etc.
8:07Now, um, as you will see in the rest of
8:10this course, of the macro development
8:12course, we are going to there's going to
8:14be, you know, very important scholars
8:16explaining,
8:18um, you know, the role
8:20of this better use of inputs, and how we
8:23think about that, particularly the role
8:25of misallocation and the role of
8:27frictions. But that's going to come
8:29later in the in the in the course.
8:31Today I will have very little to say
8:33about this.
8:34Okay?
8:35There is another issue, which is the
8:37issue of heterogeneity.
8:39Um, you know, how much capital and labor
8:42account for
8:44in the aggregate may not be the same
8:47across sectors or across firms.
8:50Okay? So, I'll talk a little bit about
8:51that today. And then there is a final
8:54question, which is
8:56are we doing the best we can in measures
8:59measure inputs,
9:01um,
9:02properly?
9:03And the biggest challenge in doing this,
9:06when you think of data across countries,
9:08is to do it in an harmonized way.
9:11Okay? So, I'll get to these issues
9:14as we move through
9:16through the course.
9:17So, let me just tell you what's going to
9:19be, uh, the plan for today.
9:22I'm going to give you the very basics of
9:23this. I'm going to spend quite a bit of
9:25time on this,
9:26uh, um,
9:28because it it is important. The first
9:30thing that I'm going to do is I'm going
9:32to give you a framework to think about
9:33the data.
9:35And in general, I think when we are
9:37going to go and get our hands dirty with
9:40data, it's useful to have a framework
9:42because that's
9:43an organizing
9:45um
9:45path
9:47that you can use
9:48to put order on how you look at the
9:50data.
9:51Okay?
9:52We're not just going to run a million
9:54random regressions and come up with
9:56something.
9:57We want structure. That's what the
9:59framework is going to do.
10:01And then I'm going to get very detailed
10:03on the data we have available,
10:06what we can measure, what we know
10:09um that are important margins
10:12that we should explore further. Um
10:15I think what it what is important about
10:17this is that, you know, I think the
10:20any of you that have a an econ major
10:23have had some interaction with with
10:25accounting, typically national accounts.
10:28Okay, that would be the first or second
10:30class in a standard intro to macro or
10:32intermediate macro course.
10:35And I don't know about you, but I
10:36remember the first time I see this, most
10:37likely I was rolling my eyes thinking
10:39this is not very
10:41you know,
10:42this is not fun.
10:44This is what are we doing this?
10:46It turns out that it's actually quite
10:47important.
10:49Okay? And I don't know, perhaps I've
10:51been thinking too much about this in the
10:53last couple of weeks,
10:54uh but I think it's critical to our
10:57understanding of the development
10:59process.
11:00Okay? So, we need to get dirty with the
11:02data somehow.
11:04Uh then I'm going to discuss what the
11:05main findings of the literature are and
11:08different approach approaches to think
11:10about um
11:12um
11:13the accounting per se.
11:15And then I'm going to just
11:17uh
11:19bring up some issues that I think are
11:21important, but mostly as a road map to
11:23think about open questions that I think
11:26are going to be important if if you guys
11:28are guys are interested in in field and
11:30you shouldn't take this as a as a
11:33uh
11:35you know, as a roadmap of what people
11:37think is interesting. This is mostly
11:38what Julieta think it's interesting.
11:41Okay?
11:42Um
11:43So, I'll I'll get to the details as I as
11:45I go along into the selected issues.
11:48Now,
11:49before I get into uh the framework, are
11:52there questions?
11:56Okay.
11:57So, let me get on.
12:00Okay. So, the framework that we're going
12:03to use to think about income accounting
12:06is the one the neoclassical framework.
12:10Okay? So, we're going to think of a a
12:12production technology that is
12:14neoclassical, where output is produced
12:17with physical capital K,
12:21H or human capital, so a measure of
12:24quality-adjusted
12:25labor force. So, basically, we want to
12:28somehow measure that guys that finished
12:31college
12:32um are different than people that do not
12:36um go to school altogether.
12:39Alpha is going to be the capital share,
12:42and this technology is going to be
12:44neoclassical because we're going to
12:46assume constant returns to scale. Okay?
12:48That's why the the share on labor is
12:50going to be 1 - alpha. And we're also
12:52going to assume Hicks-neutral
12:54technology.
12:55That's going to be A, and that's going
12:57to be our measure of total factor
12:58productivity.
13:00So, so far, I've been talking about this
13:03idea of efficiency with which we use
13:06inputs.
13:07Through the lens of the model, that's
13:09going to be this total factor
13:12productivity, this A.
13:14Okay?
13:15Now, if I give you this,
13:18one temptation
13:20is to say, "Ah, this is nice. This is
13:23log-linear."
13:24Right? So, So, can take logs on this. If
13:27I truly believe that the production
13:29technologies are there,
13:31um have these particular features that
13:33are that are Cobb-Douglas, then I can
13:35take logs,
13:37and then I should get a linear
13:38regression a a linear regression of
13:40sorts.
13:42Right? So, I have output,
13:43I have capital,
13:45I have measures of labor input,
13:48and then this A that I cannot really
13:50measure, this, you know, concept of
13:53efficiency with which you use factors,
13:56is going to be an error.
13:58Okay?
13:59Now, there there are
14:01many reasons why this is a bad idea.
14:04Okay?
14:05So, first,
14:08for these these um these estimates to
14:11give you unbiased estimators of what the
14:13what these uh
14:15uh you know, factor shares are and the
14:17contribution of these of these each of
14:19these elements, you need a orthogonality
14:21condition between TFP and inputs.
14:25And that's very unlikely. This is a
14:26story This is what I was saying before,
14:28right? If
14:30uh it is unlikely that if you live in an
14:33economy where inputs are not very well
14:36used,
14:37I'm going to invest a lot in that
14:38economy. Say, I'm going to accumulate
14:40capital in there.
14:42Okay? So, the orthogonality between the
14:44errors and the and the uh regressors
14:47here is likely to be violated.
14:50And more important, there's a huge
14:51literature
14:52that um in in in industrial organization
14:56that speaks about the challenges of
14:58estimating a production function,
15:00let alone a production function in the
15:02aggregate.
15:04So, the income accounting um
15:06approach is going to take these
15:09production neoclassical production
15:10technologies seriously,
15:12and it's going to
15:13try to measure to the best possible
15:16each of these ingredients. Okay?
15:20And taking this structural restriction,
15:23which is the shape of the production
15:24technology, we're going to try to
15:26measure what the role of each of these
15:28components is. Okay? So, as I said,
15:31remember
15:33um
15:36what things are observable.
15:41So, we observe
15:43we observe output,
15:45GDP,
15:47we observe physical capital, and we
15:49observe somehow human capital. We'll
15:52talk about whether this is really
15:53observable or not and what are we
15:55missing.
15:56But, the idea is that we need some sort
15:59of way to figure out what this object
16:01is, which is a residual.
16:04Okay?
16:06Okay.
16:09Now,
16:11again, another temptation is to say,
16:14"Okay,
16:15I can measure K, I can measure H, I can
16:18measure Y. Let me compute and I I know
16:20what the alpha is.
16:22And I'll I'll I'll It's okay if you
16:24don't know right now uh what the alpha
16:26is. I'll I'll tell you how we measure
16:27this.
16:28Um
16:31and then you can say, "Okay, let me
16:32account for the the role of capital in
16:35this in this uh
16:39in this equation."
16:40Okay?
16:41Now, the problem is that there are
16:44important um feedback effects between
16:47TFP and investment.
16:49Okay? And that we know from the insight
16:52of the one-sector growth model.
16:55Okay?
16:56Again, this is something that um
16:59we can get to if you have questions
17:02about the details of the one-sector
17:03growth model, I'll be happy to take them
17:05either at the end of the class or or
17:07online, and even Raul can can go through
17:09some of these details uh on Monday.
17:11Okay?
17:13Now, there is a an important takeaway
17:15from the one-sector growth model, which
17:17is that differences in TFP are going to
17:19induce differences in capital.
17:22So, if I just say, "Okay, how much of
17:23the variation in output is explained by
17:25capital?" And I don't take into account
17:28the fact that some of the variation that
17:30I can't that I see in capital is coming
17:33from TFP,
17:36then I'm going to I'm just not going to
17:37get the right answer, right?
17:40But, what is important from the one
17:42sector growth model is that we know that
17:44the capital output ratio is actually
17:47independent of TFP in steady state.
17:50What is a steady state? Steady state is
17:52an allocation
17:54where output
17:56uh
17:56capital output and capital are constant.
18:01Okay?
18:02Now,
18:03the Euler equation of a standard growth
18:05sector one sector growth model, to view
18:09um
18:10the Euler equation is basically the
18:11equation that tells me what the optimal
18:14accumulation of capital is along the
18:18equilibrium path of the economy. Okay?
18:21It's saying that if I invest
18:23one unit in capital today, I should get
18:26the expected return
18:28on capital tomorrow. Okay? Now, that's
18:31the equation that you have in the
18:32bottom.
18:33And it just says
18:35um
18:37that
18:38the this Sorry, that the capital output
18:40ratio or the inverse of the capital
18:43output ratio, which is the object that
18:44you have here,
18:45okay? Is going to be a function in
18:48steady state. So, this is the the Euler
18:50equation
18:51in steady state.
18:54The capital output ratio is going to be
18:56a function of parameters. So, it's going
18:58to be constant and independent of TFP.
19:02What are those parameters? Well, the
19:03parameters are the depreciation rate for
19:07capital.
19:10Okay?
19:11So, you know, if I had a car, I need to
19:13go to the garage
19:15every year for maintenance, etc. All
19:18that because, you know, using the car is
19:21going to make it break more often.
19:23That's the depreciation rate. Okay? In
19:25the context of the neoclassical model,
19:27this is just a This is just a a number.
19:30Okay? And then it is come factor.
19:33Which is going to in steady state in a
19:36one sector growth model is going to give
19:38you a measure of the impatience of the
19:40agents. Okay? How much do I value
19:42consuming today versus consuming
19:44tomorrow?
19:46Okay? And that's why the return to
19:47capital, which is
19:49um the good that allows me to transfer
19:52consumption today to consumption
19:53tomorrow, is going to depend on how
19:56impatient I am.
19:57Okay? But the point is that in a
19:59standard neoclassical world
20:02in in the world of the one sector growth
20:03model, this object in steady state is a
20:06constant.
20:09Okay?
20:11So, here the one sector growth model is
20:14giving us a lot of structure on how to
20:16look at the data and how to do the
20:18accounting. It's saying, "Look, look,
20:20look. Yes, you know
20:23that
20:25K is a function
20:28um of A.
20:30But honestly, if you just look at
20:32capital output ratios, then you're going
20:33to kill that relationship.
20:35So, let's do that. Okay?
20:38So, how do we do that?
20:40Well, we're going to rewrite
20:42neoclassical production function like
20:44this.
20:47So,
20:50the neoclassical production function was
20:53this object here.
20:56Oh, sorry. I use L. I should use H.
21:03Okay?
21:05So, divide by Y over Y to the alpha.
21:10And I can collect that down here.
21:15Okay?
21:16And now
21:18I can solve
21:20for Y.
21:22So, if you solve for Y
21:27you're going to end up with this.
21:37Okay?
21:38So, I basically have
21:41the same
21:44the same production technology
21:48that you had before just rewritten in
21:50terms of the capital output ratio.
21:54Okay? And that's if I further divide by
21:58the head count of the number of workers
22:00in the economy, that's L.
22:02If I divide by the number of workers, I
22:04end up with a equation that I'm going to
22:07be using for accounting throughout.
22:11What this equation says is that the
22:13amount of income per worker in an
22:14economy
22:16is going to be a function of
22:18you know, the efficiency of usage of
22:20factors, some modified function of that.
22:24The capital output ratio.
22:26Okay?
22:28And
22:30my measure of human capital.
22:33Okay?
22:36You can write this in logs if that helps
22:39the accounting, and I'll show you that
22:41given depending on what measure of the
22:42accounting you're going to use this this
22:44might be useful.
22:45Um
22:47but for now this is the equation that
22:49we're going to be using throughout. Our
22:51accounting is going to decompose output
22:54per worker
22:56into capital output ratios
22:59modified by a factor share of capital
23:03and a measure of human capital.
23:05Whatever is not accounted by the capital
23:08labor capital output ratio and human
23:11capital is going to be a crew to
23:15efficiency of usage of factors, TFP.
23:19Okay?
23:20Are there questions about this?
23:27Good. So, we can move on.
23:30Okay. So, let me remind you
23:32uh that the the question that we are
23:34asked after is how much of the variation
23:37in income per worker is accounted for
23:39variation in capital output ratios,
23:42human capital, and the residual.
23:45Okay. So, I basically narrow down the
23:47question the the income accounting
23:48question.
23:50With that, I'm ready to go and tell you
23:53how we're going to measure these
23:55uh elements.
23:59Okay.
24:01The good news of living in 2021
24:04and not in the
24:06second half of the of the
24:0920th century is that a lot of work has
24:11been done
24:12uh to help us measure these objects.
24:15Okay?
24:16The good news for you guys that are
24:18going to be writing the next generation
24:20of of
24:21you know, research on this topic is that
24:23not everything is done.
24:25So, there's still questions.
24:27Okay. So, um as Richard mentioned last
24:31week, uh the the go-to measure
24:34uh for macro development is the the the
24:37Penn World Tables.
24:38Okay?
24:39This was a data set that was created in
24:42the '60s by uh to measure uh differences
24:45in living standards.
24:48Okay?
24:50Now, there is a question
24:53of, you know, whether do want to measure
24:56living standards or we want to measure
24:58productive capacity?
25:00And since
25:02basically the productive capacity of an
25:04economy
25:05and since version 8.0 of the Penn World
25:08Table, which is right now run by the uh
25:12uh
25:13by
25:14the University of Groningen and the
25:15Groningen Centre for Growth um there
25:18have been substantial changes that are
25:20going to allow you to to measure
25:23productive capacity using the Penn World
25:25Tables. Okay?
25:27Now my
25:29two cents here before we even get into
25:31the data is that in general you should
25:33be very careful uh on how you set the
25:35data and record the vintages that you're
25:37using because there this data is in
25:40constant um
25:42um change. People are you know the the
25:46the um
25:47uh the group that is putting together
25:49this data is working really hard to make
25:52as many improvements to the data as
25:53possible. So that means that the changes
25:55changing. The data change.
25:57Now what I did for this class is I
25:59generated a repository,
26:01a GitHub repository. For those of you
26:04that are not familiar with with Git,
26:07um you can pretty much just link on the
26:10on the slides. It's going to take you to
26:13a webpage where you can download all the
26:15material that is relevant for these
26:17things. You don't need to work in that
26:18environment.
26:19Um you will need to have access to a
26:23software uh like Stata to be able to run
26:25the code that you have there.
26:27But if you don't want to, the the code
26:30that even or if you don't want to or you
26:31don't have access to to that software
26:33right now,
26:34um the code is annotated in a way that
26:37is going to be kind of a class on its
26:39own on what are the steps that you need
26:41to be able to measure each of these
26:43inputs if you were to do income
26:45accounting. Okay?
26:47And and Raul will review uh that with
26:50you on Monday.
26:53Okay, so without uh
26:57more details about this, although I'm
26:58happy to talk about them if the audience
27:00has questions, let me just talk about
27:02how we measure output and work. So, I'm
27:04going to start with the output part of
27:06this, then I'm going to talk about
27:07employment and labor.
27:09Okay? Um
27:11Again, the good thing is that coming
27:13after Richard, Richard talked a lot
27:14about these things. So, uh you will see
27:16some repetition. But, it's good to see
27:18it twice. Hopefully, the second time it
27:20is clearer.
27:21Um
27:22nominal output uh
27:25is going to get um typically from the
27:26national accounts, okay?
27:29But, what we're interested is in a
27:30measure of of real outputs or a
27:32unit-free measure of of uh how much of
27:36an economy produces, okay?
27:39And then the question is, okay, then
27:40that means if I have a nominal number, I
27:43I and I want to turn it real, what do I
27:45do? I divide by the the the price level.
27:48Okay?
27:50Now, what is this price level?
27:53This was a concern uh
27:56um it draws some some discussion among
27:58economists and and of the two candidates
28:01to do something like this
28:03are, okay, maybe I want to denominate
28:06say the GDP
28:09in Brazil
28:11in dollars
28:13and now it is directly comparable to
28:17the US GDP in dollars.
28:20Well, the one way to do that is to just
28:23use the exchange rate of the real to um
28:26to the dollar
28:28and transform everything in dollar
28:30values.
28:31Okay? Now, the problem is that the
28:33exchange rates are good measures of
28:36um
28:37of tradeable goods.
28:39Okay? And something that economies um
28:43uh
28:44in particular Bela Balassa and Samuelson
28:47figured out early on is that well, there
28:50are departures from what is called
28:52purchasing power parity. Okay, so let me
28:54let me get a little bit into into this.
28:58So, the the PPP prices,
29:01and I'll tell you what they are in a
29:02second, are built from the international
29:04comparison program. Okay?
29:06Um which basically surveys goods that in
29:11theory should be of comparable quality
29:12across countries. Okay, and they do a
29:14lot of effort to compare similar goods.
29:17But more important, they really want to
29:20have the same basket of goods priced in
29:24different countries.
29:26Okay?
29:27What does this basket include? It
29:29include consumption and investment good
29:32prices.
29:33It also includes tradeable and
29:35non-tradeable goods. So, a haircut,
29:37okay?
29:38And an apple.
29:40But it doesn't necessarily include
29:43imports and exports. If you go to the
29:44documentation of Penn World Tables,
29:46you're going to see that after version
29:489.1, they have done a lot of effort in
29:52trying to have proper deflators for
29:55exports and imports. Okay?
29:57The reason why when I do this income
29:59accounting exercise, I'm going to be
30:01thinking of closed economies is that it
30:05turns out that that component of export
30:06and import from a pure accounting
30:08exercise, it's not
30:11is not important in explaining the large
30:13disparities in income across countries.
30:15Okay? And there is a nice paper um
30:18written by the uh Penn World Tables um
30:21you know, group discussing the
30:23discussing this issue. Okay?
30:26Now, in general, the issue of comparing
30:28prices on poor and rich economies
30:30is a complicated ma- matter. So, what is
30:33the issue here? The issue is that the
30:35bundle of goods that people consume in
30:37India is
30:39systematically different from the bundle
30:41of goods that people consume
30:44in um
30:46in Peru.
30:47Okay?
30:48Staple goods are different, right?
30:52You know, the diets of people change
30:54dramatically across countries that are
30:56even in in the same income
30:59uh bracket. Okay?
31:01So,
31:03you know, if you want to if you're
31:05interested in the issue of comparing
31:07prices across country, I think the the
31:08paper I did on Al Hester is a is a good
31:10reference
31:12to think about this.
31:13Now, how does the
31:16the PPP price is
31:18how is it constructed?
31:20Just so we have an idea of what is that
31:21we are actually doing
31:24when we deflate using
31:26um PPP prices instead of exchange rate.
31:30So, the the may the basic idea is that
31:33if we use the exchange rate, we're only
31:35pricing goods that are tradeable, okay?
31:38Whereas, if you use the PPP prices,
31:40you're really comparing the same bundle
31:42of goods across countries. Why is that
31:44important? I'll tell you in the next
31:46slide.
31:47Okay?
31:48Now, how do you construct these bundles
31:50these bundle of these prices, sorry?
31:53What I want you to get from these uh
31:55from from the procedure
31:58is that you're basically going to have
32:02an aggregator, so the sum across
32:04countries,
32:05of a price of a consumption good.
32:07Suppose that all that there is in the
32:08economy are apples. Okay?
32:11So, you're going to you're going to
32:12price apples
32:14uh in different countries, that's PJC.
32:19Sorry, let me just mark this.
32:22You're going to price
32:23prices in different countries, that's
32:24PJC, okay?
32:26And then,
32:28um
32:28you're going to weight this Sorry,
32:31you're going to turn these prices into a
32:33common
32:35unit, okay? Using an exchange rate,
32:38that's going to be the PPP exchange
32:40rate. I'll get back to this in a second.
32:42But then you're going to weight these
32:43prices, the price of an apple in
32:45Bangladesh versus the price of an apple
32:48in Brazil, uh using
32:52the domestic consumption in country J of
32:56that apple relative to the world
32:59consumption of apples.
33:02Now,
33:04a direct implication of this is that
33:08because the weights have um the size of
33:11the economies here, prices from rich
33:13countries are going to have more weight
33:15than prices from poor countries.
33:17Okay?
33:19That's first takeaway.
33:21Second takeaway
33:23is the following.
33:28This is what is called the Penn effect
33:30or the Balassa-Samuelson effect.
33:32So, this is what I did. I took
33:36I took
33:40You know,
33:41output in current in current dollars
33:44from the as reported in Penn World Table
33:47in a country, in this case I picked
33:49Argentina, my country of origin, China,
33:53and Finland, a rich economy. Okay?
33:56And then I did two things. I could the I
33:59could turn this into common units using
34:02the exchange rate
34:04or using the PPP deflator.
34:10Okay?
34:11Now, what you see is that in relatively
34:14poor countries, like Argentina and
34:16China, GDP looks bigger
34:20when you deflate with PPP than when you
34:22deflate with exchange rates.
34:25Why is that? Fir- Sorry, first of all,
34:27this difference seems small because I'm
34:29using a log um a log um these are logs
34:33of output per worker. I should have
34:35changed the scale. I'm sorry for that.
34:37I'll I'll do it. But basically, the
34:39difference in these in terms of, um,
34:42PPP dollars is about $5,000.
34:46Okay? So, for a country with an income
34:47per capita of about 20,000, which is
34:49Argentina, this is a big deal.
34:52Okay?
34:53Now, why is that when I deflate with
34:55PPP, I get, uh, that poor countries have
34:59seem to have higher value added?
35:01Well, the Balassa-Samuelson effect says
35:03the following. If I have an economy
35:06where the production you are,
35:08uh, relatively less productive in
35:10tradables, okay?
35:12What that's going to generate is that
35:14the overall price level of the economy
35:15is going to be cheaper. Not only of
35:17tradables, also of non-tradables.
35:20Okay? This is the idea that if
35:24you know, if you want somebody to help
35:26you in your house,
35:27in the US or in Europe,
35:29in terms of the income of the average
35:31guy out there in the street, it's going
35:34to be pretty expensive.
35:36Okay? It's going to be a good chunk of
35:37your income.
35:38Whereas, if you are in Turkey, in places
35:41in South America,
35:43you know, getting somebody to come and
35:45help in your house is going to be
35:47relatively cheaper.
35:49You know, somebody helping in your house
35:51is highly non-tradable.
35:54So, how is this possible? Well, that's
35:55the Balassa-Samuelson effect.
35:57It's the idea that even the prices of
35:59non-tradables are going to be cheaper in
36:01poor countries because the price of
36:04because they're relatively less
36:05efficient in producing tradables.
36:08So, what that's going to generate is
36:09that if you look at income disparities
36:12deflated using exchange rate, you're
36:15going to overestimate
36:17the differences in income across
36:19countries.
36:21Okay? So, the idea is that poor
36:24countries are going to look even poorer
36:26because you're not adjusting for the
36:27fact that the non-tradable goods
36:29actually
36:30are actually cheaper
36:32in those countries.
36:34Okay, are there questions about this?
36:41Okay, so if I want to measure income
36:43income per worker, now I need to go to
36:45the denominator. I need a measure of
36:47employment.
36:49Well,
36:50then I need to
36:52uh take a call. I could just use
36:54headcounts
36:55or I can use hours work.
36:58Until Penn World Table
37:00um
37:019.1,
37:03there were actually no harmonized
37:05measures of average uh working hours
37:09available
37:10uh within the Penn World Tables.
37:12Even now in version 10, you're going to
37:15see that there is available a measure of
37:17hours work,
37:19but the measure that is available there
37:21is not going to available is not going
37:22to be available for all the countries in
37:24the sample and for the whole time
37:26period. Okay?
37:27So, a lot of the you know, the two
37:30handbook chapters that were uh assigned
37:32as as as reading material for this
37:35course still use measures of workers
37:39that are just headcounts, employment.
37:41Okay?
37:43Now, why is this important? Well,
37:46there's a paper by Biggs, Fulton, and
37:47Lagakos that shows that our works
37:50decrease with development.
37:52So, people in rich countries work less
37:54hours than people in poor countries.
37:57Why is this important? Because if in my
38:00accounting,
38:01I compute, remember the object here is
38:05output per per worker.
38:08If I divide by a systematically
38:12um um
38:14lower level of employment in poor
38:17countries,
38:18then I'm going to then I'm I'm to tend
38:20to increase the amount of output per
38:22work.
38:23But what happens is that the amount of
38:24hours worked is higher in in poor
38:27countries.
38:29Okay? So, what you have
38:31um
38:33in the what you have in this in this in
38:35the exercise that you have in the
38:36repository is you have the alternative
38:39to use either employment that maximizes
38:42the sample or to use employment times
38:45the average hours worked, which is going
38:46to give you a total number of hours
38:49uh um
38:50full-time equivalent workers uh in an
38:53economy. Okay? So, you can play around
38:55with the with the with the
38:58uh data that way.
39:00Okay.
39:01Sorry. I wanted to ask something related
39:03with these working hours. Yes.
39:05I was thinking when I was reading this
39:06paper in the recommendation that um
39:09is there any way to control for the
39:12actual hours worked? Because I can't
39:14imagine that we have very long shifts
39:16like they have the example of Egypt with
39:1857 hours. I highly doubt that people,
39:22you know, like committing a big effort
39:24working for 60 hours a week. I mean, I
39:27don't know. For me it's kind of odd.
39:29And coming I mean I I'm I'm Spanish and
39:31here we have longer working shifts than
39:33for example France, Germany, stuff like
39:35that. But maybe we we relax a little bit
39:38during the working hours, right? Because
39:40it's a longer shift. Yeah, so you're
39:42Okay, if I get correctly
39:44uh your your question, what you're
39:46trying to or you're asking is
39:50are all hours uh
39:52do all hours entail equal effort? And if
39:55what I'm really after is a measure of
39:56employment, I agree with you. I don't
39:58think we have
40:00uh even for developed countries, I don't
40:02think we have good measures of of this.
40:05Okay.
40:06Um now, the fact that there are 57 hours
40:09in Egypt, I don't think it's very
40:10surprising in the sense that
40:13uh if you ever walk through Cairo, you
40:15will see that people barely sleep. So,
40:18uh
40:18it's not so surprising that the hours
40:20are are longer there. But I agree with
40:22you, there is an issue of of uh
40:24productivity as signs to this. And in
40:26part that's going to be coming in in the
40:29amount of output that you produce per
40:30hour work.
40:31Right?
40:32Yes. So, that would be the direct
40:34measure of of that efficiency.
40:37Mhm. Okay?
40:40Okay.
40:41So,
40:42um
40:47So,
40:49uh
40:52In measuring the the the next thing that
40:54we need to measure is the capital output
40:56ratio.
40:57Okay?
40:58Now,
40:59the current
41:00So, think about what is capital in an
41:02economy. Capital in an economy are
41:04production lines, the trucks, the cars,
41:08um
41:09but also, you know, the the land that
41:12you have available, the buildings.
41:14Now,
41:16Pingle table considers the following
41:18categories. So, they're going to
41:19consider machinery, computer and
41:21communication equipment, transportation,
41:23structures,
41:24and other assets, in particular, IPP,
41:27which is intellectual property products,
41:30and software.
41:32The reason why I'm highlighting
41:33computers and communication equipment
41:35and IPP and software is because these
41:37are equipments that particularly since
41:39the 2000 have had rapid changes
41:42in
41:43uh
41:44um
41:45in that option of of this type of
41:48investment and the prices.
41:50Um
41:52and are potentially important in the
41:55economy, and at least in developed
41:57economies, in the economies that are
41:58coming up and are becoming available to
42:00us. If you go to the data set that uh
42:03Richard also mentioned last week, uh the
42:05total economy data set, there you're
42:07going to see that you have different
42:09time series time series for capital, for
42:12what is called
42:13uh communication equipment, ICT
42:15equipment, and ICT equipment, sorry,
42:19and non-communication equipment. Okay?
42:22So, if you're interested in these
42:24particular capital goods, there is
42:26different disaggregations that are
42:27available in sister data sets to the
42:31Penn World Table.
42:32Now, there is a question here of what is
42:35productive capacity versus
42:37productive capital versus natural
42:39resources.
42:40So, for example,
42:41countries
42:43in in the Middle East that mostly live
42:45out of oil rents are going to look
42:49extremely rich because they produce a
42:51lot with a very little input.
42:54But, it's not clear that the standard
42:56neoclassical framework is what really,
42:59you know, a good theory to think about
43:01the output produced or the output the
43:03the richness of the workers producing in
43:06in those economies.
43:07So, Caselli and Feyrer wrote a paper
43:09where they're taking these ideas
43:11seriously,
43:13trying to clean up what is productive
43:16capital versus natural resources.
43:20And
43:21a crude adjustment to output
43:26that you can do is to just eliminate
43:29from output the rents that come from
43:30natural resources.
43:32The World Bank has put together a data
43:34set that is now available through the
43:36World Development Indicators that is
43:38going to allow you to know how much of
43:40your GDP are rents associated to these
43:43natural resources.
43:45So, what you will see is what I did in
43:46the accounting that I'm going to present
43:48is I eliminate I took away from the
43:51value of GDP
43:53the
43:54the rents from natural resources. Cuz if
43:56you don't do that, you're going to have
43:58countries like, you know, Qatar and
44:01Benin being super rich. They're
44:03basically living out of their
44:06their oil rents. Okay?
44:09Uh, so how do you properly measure
44:11capital?
44:13Well, the standard way to measuring
44:15capital is the permanent inventory
44:16method.
44:18Uh, which means that you're going to use
44:21uh, the following law of motion.
44:23So, suppose you have the stock of some
44:25equipment I, in this case trucks,
44:28and you want to know how much stock
44:29you're going to have in each period of
44:30time. That's K.
44:32There was a typo in the notes, so I
44:33added a plus one on the
44:36on the
44:38uh,
44:40on the left-hand side here. I'll update
44:42it after the class, so you have the link
44:44corrected. Okay?
44:46So, how much capital am I going to have
44:48tomorrow? Well, I'm going to have
44:50whatever investment I have today, plus
44:52the value of the undepreciated stock.
44:55Where delta here is the depreciation
44:57rate for the stock.
44:59Okay?
45:00Now, this is the stock of trucks. Then I
45:03need to aggregate.
45:04Uh,
45:05to aggregate it, I'm going to use some
45:07weights to aggregate trucks to computers
45:09to structures to buildings. Okay?
45:12Now,
45:13uh, a couple of issues. Let me first
45:15focus on on the question on top. Okay?
45:19Now,
45:21what the what the framework says is that
45:23delta should be physical depreciation.
45:26Now, the measures out there for physical
45:28depreciation typically include economic
45:30depreciation, which is this idea that,
45:33you know, you have your phone,
45:36and your phone makes calls,
45:39sometimes, you know, it allows you to
45:41text,
45:42and if you have a, you know, a
45:44smartphone, then you are even able to
45:46connect to the internet.
45:47Fine. The services that you have are
45:49what they are.
45:51But, you know, 2 years after you hold
45:53that phone,
45:54that phone is worth half the price.
45:58It may be exactly the same phone, still
46:00making calls, still connecting to the
46:02internet.
46:04Why is that? Well, that's economic
46:05obsolescence.
46:06The reason why the price of that capital
46:08good falls is because there are better
46:10goods available in the economy.
46:12Okay?
46:14Now, the second issue that is important
46:16to understand in measuring in in
46:19measuring capital stocks is that you
46:20need some sort of
46:23initial stock of capital to be able to
46:26generate this iteration, right? You see
46:27investment in each period, but you need
46:29to have some initial capital to use
46:31here.
46:33Okay, so there are two options. One
46:34option is to take uh seriously a
46:37one-sector growth model here, call it
46:40the solar model, but a one-sector growth
46:41model,
46:43and say, "Okay, capital in the initial
46:45period is going to be uh investment
46:48adjusted by the depreciation rate of
46:51that capital,
46:52which
46:55For some reason, this is not working.
46:57Okay, the depreciation rate of that
46:58capital,
47:00and the growth rate of the economy.
47:03All right? That's going to be my
47:04initial.
47:05The second approach that um uh people,
47:09in particular after Penn World Table 8,
47:11have been using is to uh
47:14calibrate
47:16so that the measure of the stocks that
47:18you generate are consistent with the
47:21share of capital to to output, which
47:24happens to be, at least Finstra in 2015,
47:27this is a AR paper,
47:29uh added used that this ratio is stable
47:32across countries and and time.
47:34This is a different version of this is a
47:36uh
47:37um
47:38uh the Kaldor facts. Okay?
47:41Now, the problem is when you go
47:42empirically and look at the data, it
47:45doesn't seem that this is so constant.
47:48So, the newer version of Penn World
47:50Table, and here you have the link in the
47:53in the in the
47:56in the slide deck, so you can see the
47:59increase in the the change, sorry, in
48:02this in this ratio.
48:05Um so, basically, what they're doing is
48:06they're trying to match
48:08countries, given their income level, to
48:11that curve that they estimate using the
48:13data from all these countries through
48:15time. Okay?
48:17Uh sorry, that seems a little bit big,
48:19but that was the best I could do, given
48:21that I didn't want to spend more than um
48:23you know, 5 minutes talking about how
48:25you construct this this initial capital
48:28stocks. If you have questions about
48:30them, I'll be happy to answer at the end
48:32or offline.
48:37Okay.
48:40Okay. So, the second issue is that now
48:43we figure out how to measure the stock
48:44of trucks. Now, we need to somehow add
48:48trucks
48:49to computers.
48:51How do we do that? Well, we're going to
48:52have some weighted averages of these
48:54stocks.
48:55Okay.
48:57There are various ways to do that. One
48:59way is to use
49:02the expenses in the sorry, the stock of
49:04capital value that whatever prices for
49:06the capital you have, divided the
49:08overall value of the capital stock in
49:11that economy at the point in time.
49:14The other way to do this is instead of
49:15using the prices of the capital, is to
49:18use the rental prices, the rental rates.
49:22Why does this matter? Well, because the
49:24first approach is going to tend to
49:26overstate the value of long-lived
49:28assets.
49:30So, the idea is that if I have a truck
49:32that can be around for 50 years and
49:36whose price is fairly stable through
49:38time, and it's giving me very little
49:40service in each period, then I'm going
49:42to tend to overestimate the value of the
49:44truck and underestimate the value of the
49:47phone.
49:48Right?
49:52Using these rental rates,
49:54um,
49:54is perhaps a desirable
49:57better way to measure capital services
49:59associated to these long-lived assets.
50:03And the way they're going to infer these
50:05rental rates is using the concept of the
50:08user cost of capital.
50:10So, how do we do this? We want to
50:11measure what is the user cost What is
50:14the user cost of capital?
50:17Well, one way to do that
50:20is to
50:23do the following.
50:27So, suppose I can
50:30you know, I can buy a truck or I can put
50:33the money in the bank.
50:34Okay?
50:35So, suppose that the price of the truck
50:38is PIT.
50:40PIT. If I put it in the bank, it's going
50:43to give me a return RT.
50:46Or
50:48I can buy the truck, rent it,
50:53Okay?
50:56And then I'm going to end up with the
50:59undepreciated value of the truck.
51:02Okay? Suppose it was There was some wear
51:05and tear.
51:06And I can sell it back in the market.
51:09At whatever price it's available for
51:12trucks tomorrow.
51:13I think we all understand that equation,
51:15right?
51:16So, what this is saying is that if it is
51:18true that you are indifferent when you
51:20buy this truck, you're indifferent
51:21between putting the money in the bank
51:23and getting whatever return or whatever
51:24return you get when you purchase the
51:26truck the truck, then the user cost of
51:28capital can be inferred as a residual.
51:32Okay? So, basically, the equation that
51:34you have here for the user cost of
51:36capital is just solving through for
51:40R, which is the user cost.
51:42Now, it turns out that after you do all
51:44this work,
51:45and again, you can play with this
51:46through the code that you have available
51:47in the repository. You're going to see
51:50that the amount of capital services
51:53are not so different.
51:56Okay?
51:57And this is just an empirical issue.
52:01I'll be happy to talk more about
52:02adjustment to the investment series, but
52:05I just don't have time to go through all
52:06this. It's already, you know, um hitting
52:09the limits. Now, the next thing, so we
52:11have two more things that we need to
52:13measure. We measure output, we measure
52:16the capital output ratio, we need to
52:17measure the factor share for capital,
52:20and we need to measure human capital.
52:21Where are these factor shares estimated?
52:24Well, under the assumption of constant
52:25returns, you can literally uh infer the
52:27capital share as a residual between the
52:30labor share uh of income,
52:33basically how much all the workers in
52:35this economy are paid for as a fraction
52:38of GDP, um
52:40um
52:42Sorry, as a residual of that that uh
52:44that labor income over overall. Now, the
52:47time series suggest that this number
52:50that the capital share is about a third.
52:53I mean, this is kind of uh
52:55um
52:56color facts.
52:57Now, the cross-country data, when you go
52:59and do this with the with the national
53:02accounts data,
53:04shows that there are massive
53:05differences. Okay?
53:07What I'm going to do for the exercise
53:09that you have in the code, and what also
53:10Hall and Jones and Jones in his chapter
53:13does is to adjust the assume that alpha
53:15is 1/3.
53:17Now,
53:18uh Daggolyn,
53:20uh or Steen Daggolyn, uh who should be
53:23somewhere in the audience,
53:24um had an influ- influential paper where
53:27he said, "Well, the issue is that you
53:29really need to consider
53:31self-employment."
53:32And in national accounts, typically
53:34self-employment is treated as capital
53:35income.
53:36So, once you adjust for self-employment,
53:38which happens to be disproportionately
53:40more prevalent in poor countries, then
53:43you get an in- a labor share that varies
53:47between
53:48a third and point eight. So, the capital
53:50share is going to go sorry, two thirds
53:52and point eight. And the capital share
53:55is going to go between one third and
53:56point two. That's much smaller
53:58variation.
54:00Penn World Table has their own
54:02estimates, the new one, the Penn World
54:04Table 10 has the new estimates of the
54:06factor shares, but again, there is a lot
54:07of variation. So, that's why we didn't
54:09use it. It's in the code. You have the
54:11code written, you can play around with
54:12it.
54:15Okay, finally,
54:17human capital.
54:19I'm going to give you a brief overview
54:20of how we measure human capital. You're
54:22going to have the full supplemental
54:24lecture that Turchin is going to teach
54:26to get into the nitty-gritty of of how
54:29we measure human capital, how we do it
54:30properly, and what is at the frontier.
54:33Now, the idea of measuring
54:36human capital is basically this idea
54:39that I think Juan had before, that you
54:41know, workers or an hour of a worker is
54:46not the same.
54:48Neither through the day nor, you know,
54:51of workers of different skills. So, for
54:53example, a college educated worker is
54:57going to have different input than an
55:00educated worker. Okay? Now, the
55:02differences between those two workers
55:04are as what what we're going to call
55:06efficiency units.
55:08So, the idea that if my efficiency unit
55:10is one, it's basically that's a
55:12normalizing factor. Think that an
55:14educated worker
55:17produces half of what an educated worker
55:20produces, then the efficiency units of
55:22the uneducated worker are going to be
55:24half
55:25of the other one. And it's just a way to
55:28add
55:30factors of production, in this case
55:31human capital, in a way that take into
55:33account this heterogeneity that we know
55:35exists in the data.
55:37Now, the typical way to map this
55:39efficiency or just think about how a
55:40worker is different is to say, well,
55:44workers differs in the level of
55:46schooling
55:47and also in their experience in the in
55:49the job, right? There are people that
55:50have been doing the same job for 30
55:51years. It cannot be that they produce
55:53the same thing that the guy that just
55:55started the job. Okay?
55:58Now, in in a very nice paper, Bills and
56:00Klenow argue that there are also
56:01differences
56:03in
56:04kind of the quality, even if you have
56:06the same schooling and the same
56:08experience in Bangladesh and in France,
56:12it turns out that the quality of the
56:14input that you're putting in is
56:15different and that's going to generate
56:16differences in return. Okay? So, what
56:19I'm going to do today, I'm going to
56:20assume that the quality of that that
56:24schooling is the same across is
56:26identical across countries.
56:28And then we're going to parameterize
56:29that function fee, which is going to
56:30give me the returns to to to schooling
56:34and and experience as two terms. One
56:37that is just a function of schooling,
56:38the other one that G that depends on my
56:41experience.
56:42Now, what is the the the story of this
56:45specification, it goes back to the long
56:48labor literature that thinks about
56:50earnings in terms of
56:52in Syrian regression. Okay? The idea
56:54that the earnings that you earn are a
56:57function are a function of your
56:58schooling in this case. So, in the
57:00Mincer specification,
57:02uh an additional
57:04uh year of schooling, so as
57:07uh higher is going to give me theta more
57:11in terms of income.
57:14Um
57:16And then the idea that experience
57:19also matters, it has a return, and it
57:21actually the return to experience is uh
57:24convex. Okay?
57:26Now,
57:28in this cross-country analysis, what
57:30we're what we're going to do is we're
57:31going to follow the Penn World Table uh
57:34for their own measures of human capital.
57:37So, in particular, what they do is
57:39they're going to use the educational
57:40attainment data from Barro and Lee, and
57:42also they are merging now uh data from
57:45Cohen, Soto, and Lecker that what that
57:48what it does is
57:50uh it controls
57:52it has better measures of what
57:54attainment means in a harmonized way
57:55across countries, okay? And then the the
57:58returns are only going to have returns
58:00to schooling to education, and this is
58:02going to come from uh
58:04Sakellaropoulo's '94, okay? Now, the
58:08human capital index that they're going
58:09to generate is going to have this uh
58:13piecewise linear
58:14uh shape that is going to say that the
58:17first four years of schooling
58:20is going to uh increase each year is
58:23going to increase your income by uh 13%
58:26points.
58:27Then, the next four to eight years is
58:30going to increase your income by 10% and
58:33uh sorry, this is a typo. This should be
58:37larger than eight.
58:43So, if X is larger than eight, sorry, my
58:46handwriting is not best here. Let me
58:48just do this.
58:51So, if S is larger than eight, then
58:54you're going to get an additional income
58:56uh a marginal income of uh 7%. Okay?
59:01Now, as you see
59:03the Penn World Table right now is using
59:05a measure of human capital that doesn't
59:07depend on experience.
59:09Now, what How important is experience?
59:12Well, there is a nice paper by Lagakos,
59:13Moll, Porzio, Tian, and and Toth Ralston
59:17that argue that the experience profile
59:19is steeper for workers with higher
59:21schooling.
59:22Okay? But, then not only that, which is
59:23important for for income per capita, is
59:25that the experience wage profiles are
59:27twice as deep in rich in rich than in
59:30poor countries.
59:32So, the idea is that the you know, the
59:34return for for your
59:37for staying one more year in the job is
59:39much higher in rich countries than in
59:41poor countries.
59:42So, what that's going to do is that it's
59:43going to change the shape of your human
59:46capital input
59:48into that production technology. And I
59:50really encourage you to to read the
59:51paper through, get to the nitty-gritty,
59:53and I guess Todd will will uh revisit
59:56some of these these issues next week.
59:58Julieta, I have another question. Uh I
1:00:00was When I was checking this part as
1:00:02well, I was thinking like
1:00:04um I mean, yes, they they they were
1:00:06saying that in um underdeveloped
1:00:08countries they have more experience on
1:00:09average, right? But then I I I was
1:00:11wondering if I mean, and now you're
1:00:12mentioning that you actually have more
1:00:13return to experience in the in Western
1:00:15countries, in more developed countries.
1:00:17So, I was wondering like this may be a
1:00:20problem related with um job switching
1:00:22and probably sector switching. Maybe in
1:00:25underdeveloped countries they don't
1:00:26really switch jobs that much because I I
1:00:28I don't know if you have any insight of
1:00:29this because I don't know if they switch
1:00:31jobs more, sectors more. I don't I don't
1:00:33have any idea.
1:00:34Um
1:00:36the evidence seems to be that there is
1:00:38more switching, but the evidence is very
1:00:41new. The paper to look into this is also
1:00:43a paper that uh Schulman is writing with
1:00:46uh Kevin Donovan.
1:00:48Um if you just go through the webpage,
1:00:50you're going to see, and the data is um
1:00:52the data is available. There is another
1:00:54paper
1:00:55to look into this
1:00:57uh by uh uh
1:00:59Markus Poschke, but I'm not sure if they
1:01:02have the occupational turnover.
1:01:05So, so poor countries switch more?
1:01:09That uh I think that's the I think
1:01:11that's the the takeaway. The problem is
1:01:13in the paper that really thinking about
1:01:16uh unemployment employment flows.
1:01:18They're not thinking about occupation
1:01:20occupation flows. I was thinking a bit
1:01:22of maybe some sort of friction, right?
1:01:24Like yeah, so
1:01:26The data I know these guys are thinking
1:01:28about these issues. I don't think we
1:01:30have a definitive answer. Not that I
1:01:31don't not that I know. Maybe Joe or Doug
1:01:35we have a better answer about that, but
1:01:37I don't think we know right now.
1:01:39In part because it's very difficult to
1:01:40measure.
1:01:42So think about it. The real challenge in
1:01:43doing all these things is that you need
1:01:45to measure this stuff in a harmonized
1:01:47way across the world.
1:01:49And that's the the the harmonization.
1:01:50It's not because conceptually we think
1:01:51it's difficult. It's just because it's
1:01:53difficult to measure this harmonized
1:01:55way.
1:01:57Okay, so with that let me just get to
1:02:01the results.
1:02:02Jingong, do you have a question? Yeah,
1:02:04can I question about the price
1:02:05adjustment? So earlier we talked about
1:02:07using PPP to adjust the different prices
1:02:09in different countries. And
1:02:12for instance, if we select the US as a
1:02:13base country, we we may want to convert
1:02:15all the other
1:02:18GDP into the US dollars. And this is
1:02:21this is a important for doing a
1:02:23cross-sectional comparison. But what if
1:02:24we want to do a time series comparison?
1:02:27If we for instance want to compare the
1:02:29the GDP of China over time and after
1:02:31converting
1:02:33the same Y into USD, then
1:02:35which inflation which deflator should we
1:02:38use? The US one?
1:02:38are different
1:02:40I guess the answer depends on what is
1:02:42that you're
1:02:44you're really interested in. Even when
1:02:46you do growth accounting, which is kind
1:02:48of the
1:02:50the same exercise that I'm running here
1:02:53across countries, but for the country
1:02:54through time.
1:02:56You want to have a PPP measure of these
1:02:58things exactly because of the issues
1:03:00that I was discussing before. The
1:03:01exchange rates are just a measure of the
1:03:05prices of tradeable goods.
1:03:07Right? And we know that there are
1:03:08systematic different In a world where
1:03:10the world
1:03:12Sorry, and we know that there are
1:03:13systematic differences in the prices of
1:03:15non-tradeable across countries and
1:03:17through time.
1:03:19So, something that the newest version of
1:03:21the Penn World Table does, so it used to
1:03:23be that they would use a benchmark year
1:03:26from the ICP
1:03:27and they would use that to deflate the
1:03:29whole thing.
1:03:30The newest versions of the Penn World
1:03:31Table, what they're doing is they're
1:03:33using different benchmark years to
1:03:35deflate different time periods.
1:03:38Which is exactly what you want because
1:03:39the bundles of goods are changing
1:03:42through time and across countries. So,
1:03:44they make
1:03:45they make sure to to generate the
1:03:48deflator that takes the same bundle
1:03:50through time across countries.
1:03:52Okay? So, the answer, if you're going to
1:03:54do a a growth accounting exercise, is to
1:03:56use PPP.
1:03:58And the reasoning is what I was what I
1:04:00was saying is that
1:04:02we live in a world where uh there are
1:04:05deviations for prices in power parity
1:04:07and the reason for that is the
1:04:09Balassa-Samuelson effect. The idea that
1:04:11if you're less productive in the
1:04:12tradable good, your overall price level
1:04:15also of the non-tradables is going to be
1:04:16lower.
1:04:17Okay? So, there are systematic
1:04:19differences in the relative prices of
1:04:21haircuts to um shoes.
1:04:25Okay?
1:04:27Is there any more question Are there any
1:04:28more questions?
1:04:33Okay.
1:04:34So, after all this work
1:04:37uh let's just try to see what we can
1:04:40find. So, how much variation in income
1:04:42per worker is accounted for variation in
1:04:45factors of production?
1:04:47And how much is just deficiency in using
1:04:49those factors or the of the
1:04:51productivity?
1:04:53Well, what I did here is I literally
1:04:55replicated the table that is in Jones
1:04:572016, the handbook the handbook chapter,
1:05:00with the same countries, with the only
1:05:02difference that I used the data in 2017
1:05:04instead of the data in 2010.
1:05:07Also, my natural adjustment uh my
1:05:10adjustment for natural resources is not
1:05:12what exactly what he did.
1:05:14Um
1:05:16um
1:05:17but
1:05:18you know, in terms of the ranking of
1:05:19countries, what you're going to see is
1:05:21that there was some reshuffling
1:05:23here on the top,
1:05:25but the ranking among poor countries
1:05:26pretty much what he had there.
1:05:29Now, how do you interpret this thing?
1:05:31I'm going to pick Argentina because
1:05:33that's the the country I I come from and
1:05:35it's rather arbitrary. I could have
1:05:37picked any any other uh one, but also
1:05:40because the numbers are easy.
1:05:42Okay? So,
1:05:44um
1:05:44what we did here is we normalized the
1:05:46United States to one in each of the
1:05:48factors. Okay?
1:05:50So, what are the income differences
1:05:53between Argentina and the US?
1:05:55Well, one way to think about it is
1:05:58to do one over 0.4. That's give you
1:06:01approximately 2.5.
1:06:04Okay?
1:06:05And now let's try to think of how much
1:06:07of this comes
1:06:09from measured inputs, which are the
1:06:11first two terms, and how much of that
1:06:13comes from the residual. This is what is
1:06:15produced in this uh last column here.
1:06:19How do we measure that? Well, we know
1:06:21that
1:06:23what comes from
1:06:25TFP or the C is one over 0.5. That gives
1:06:29you a ratio of two.
1:06:31Okay?
1:06:33So, what is that 0.6 coming from?
1:06:35Well, we know that capital there is
1:06:38basically no differences and there is
1:06:40about the
1:06:41um
1:06:43uh
1:06:440.2 differences in human capital.
1:06:46Okay? So, if you do
1:06:491 + 2 +
1:06:51uh
1:06:521.2,
1:06:54you're going to get
1:06:56um
1:06:58Yeah, you're going to get about 0.6.
1:07:02Okay? Which is the number that you have
1:07:04in the last column.
1:07:06And you do the same for each country.
1:07:08And then what I did is I produced the
1:07:10averages for the whole sample, not just
1:07:12the ones that are presented in this
1:07:13table, and then uh I computed the same
1:07:17exercise. There you get to about 70%.
1:07:21Now, why is that?
1:07:23Well,
1:07:25what seems to happen is that if I comp
1:07:28if I plot the share due to TFP,
1:07:32okay? The share of income disparities
1:07:34due to TFP, the last column that I
1:07:36showed you before, on the Y axis,
1:07:38against income per capita
1:07:41income per worker, sorry, in the log
1:07:42axis, then what you're going to see is
1:07:45that
1:07:47there is a strong negative relationship.
1:07:49Okay?
1:07:50So, what this is saying is that
1:07:53poor countries, sorry, that
1:07:56TFP
1:07:57explain more most of the difference
1:08:01between those poor countries and the US
1:08:03than it does in rich countries.
1:08:06All right? That's takeaway number one.
1:08:10Takeaway number two, which you typically
1:08:12don't see in income accounting
1:08:13exercises, but I thought it was I was
1:08:14just curious and I thought it was
1:08:16important to see, is that these effects
1:08:19are
1:08:20I would say stable. Yes, there is a bit
1:08:23of a of a trend in the contribution of
1:08:25TFP.
1:08:26I would say it's it's stable. It's
1:08:27around 60%.
1:08:29Okay? The contribution of of TFP.
1:08:32There is a little bit of a trend and
1:08:33that's a question too of measurement,
1:08:35etc., etc. So, hopefully, you know,
1:08:37we'll figure we'll figure that out soon.
1:08:40Now, another approach to thinking about
1:08:43the contribution, instead of doing that
1:08:45exercise that I that I did before, is to
1:08:48think about a variance decomposition
1:08:50inspired measure. This is what Caselli
1:08:52in his handbook paper does.
1:08:54Okay?
1:08:55Um
1:08:58So, you can take a factor only component
1:09:01of output per worker.
1:09:03Okay?
1:09:04That's basically everything that we can
1:09:05measure. Capital output ratios, factor
1:09:08shares, human capital.
1:09:10And then you can have uh then you can
1:09:12write output per worker as
1:09:15this measure, which is what we don't
1:09:17know,
1:09:19you know, TFP, and then
1:09:22what we can what we actually see.
1:09:27If you do a log variance decomposition
1:09:29of this, you're going to see that the
1:09:31variance of log output is going to be
1:09:34explained by the variance of
1:09:36um
1:09:37the residual term
1:09:39plus the variance of the measure the
1:09:42things that you can measure
1:09:44and the covariance between those two.
1:09:48Now, what is the income accounting
1:09:50exercise in this environment
1:09:52uh
1:09:53translated to this environment?
1:09:55Remember, what we're trying to answer is
1:09:56how much of the income variation is
1:09:58accounted for variation in factors of
1:09:59production.
1:10:01Well,
1:10:02this is
1:10:04uh equivalent to asking what's the
1:10:06variation in output when
1:10:10say the productivity of all the
1:10:11countries is the same.
1:10:12But if the productivity of all the
1:10:14countries is the same, then the
1:10:15covariance between log Z and log uh and
1:10:18what you can measure is going to be zero
1:10:21because Z is flat, okay?
1:10:23Now, the key measure that Caselli
1:10:25proposed is to use a measure of success,
1:10:28basically how much we can explain from
1:10:30measure stuff, as uh the variance in the
1:10:34log of
1:10:35uh output measure versus the variance
1:10:38that we actually see in the data.
1:10:41Alternatively, you can follow the Cleno
1:10:43Rodriguez Claray uh approach, which is
1:10:46to say, "No, no, well, this covariance
1:10:48term is important. I'm going to assign
1:10:49it equally between TFP and um measured
1:10:54inputs."
1:10:57Now, if you do that,
1:10:59uh the measure of success that you get
1:11:01is the following. So, in our data in
1:11:042017, you get that the variance of
1:11:06output per worker is about 0.65.
1:11:10The variance of output per worker that
1:11:11you can measure is uh
1:11:14a third
1:11:16of that uh sorry, 18%. So, then what you
1:11:19can actually measure of the variance is
1:11:21about the third.
1:11:22Everything else, this is consistent with
1:11:25the number that we got from from the
1:11:26other measure.
1:11:27About 70% on average is stuff that we
1:11:30cannot properly measure. Of course, I'm
1:11:33using a very uh you know, the way I
1:11:35measure capital and labor is as a
1:11:37standard as it can be. The papers that I
1:11:39mentioned through the whole data section
1:11:42um
1:11:43increase the the the share accounted for
1:11:47things that you can actually measure,
1:11:49okay?
1:11:50So, um this is this is the plain vanilla
1:11:53accounting that you can that you can
1:11:55cover.
1:11:56Now, one more question and and with this
1:11:58I likely close. I leave the selected
1:11:59issues for for the discussion. I'll
1:12:01bring them up uh
1:12:03I'll discuss them if if we want because
1:12:05I really want to hear from you if you if
1:12:07you have uh questions and thoughts.
1:12:09So, let me finish with with the
1:12:11following.
1:12:12Um
1:12:14so far we have been thinking of an
1:12:15economy as a homogeneous unit, okay,
1:12:18where the production technology is what
1:12:20it is.
1:12:21But, we know that there are sectorial
1:12:24disparities, right? The production in
1:12:26agriculture is not the same as as in
1:12:28non-agriculture.
1:12:30Potentially, the production of tradables
1:12:31and non-tradables is different.
1:12:34Um
1:12:36so, I'm going to follow here the the
1:12:38exercise that Caselli did
1:12:40and I'll tell you why why this is
1:12:42important.
1:12:43Uh it's not arbitrary that we're picking
1:12:45agriculture and non-agriculture.
1:12:48So,
1:12:49uh
1:12:50you can think of total output in an
1:12:51economy as
1:12:54the
1:12:55um
1:12:57an employment weighted um average
1:13:00between the output produced by
1:13:02agriculture and the output produced
1:13:04everywhere else.
1:13:05Why is this important? Well, because
1:13:07pretty much all the workers in poor
1:13:09countries are employed by the
1:13:12agriculture sector.
1:13:13Okay?
1:13:15And pretty much none of the
1:13:18workers in the rich countries are
1:13:20employed in the agriculture sector.
1:13:23Now, the challenge for which we can't
1:13:25just have
1:13:26you know, disaggregations for many many
1:13:28sectors is that to be able to construct
1:13:31this these measures of output at the
1:13:33sector level, you need PPP deflators by
1:13:35sector. And those are actually very hard
1:13:37to construct.
1:13:39For the case of agriculture, uh
1:13:41Restuccia, Yang, and Zhu
1:13:43uh constructed measures using producer
1:13:46prices that came from
1:13:48uh Food and Agriculture Organization,
1:13:50FAO. Okay?
1:13:52There is a an early paper by Baldi and
1:13:54Solow on the manufacturing and service
1:13:57uh service PPP using firm-level data um
1:14:01that I actually wasn't aware of uh until
1:14:04recently, and I think it's you know,
1:14:06it's nice to to take a look in in terms
1:14:09of uh giving us ideas on how to measure
1:14:11these these PPPs for broad sectors.
1:14:14Now, going back to agriculture.
1:14:16So, how
1:14:19So, one way to think about this is to
1:14:21say, "Okay, how important is
1:14:23agriculture?" We know that countries
1:14:26rich and poor countries are very
1:14:28different in their employment shares.
1:14:30And they're also uh potentially
1:14:32different in how much This is a typo,
1:14:34sorry.
1:14:38This is A bar, and this is A bar because
1:14:40this is the share outside agriculture,
1:14:43and this is the output outside
1:14:44agriculture.
1:14:46All right, so how important this
1:14:48So, the exercise that Caselli does is
1:14:50the following. He says, "Look,
1:14:53real output per worker the uh the log
1:14:55variance for the period that he's using
1:14:57is 1.1.
1:15:00Suppose that let me just rewrite the the
1:15:04the equation here. So, output is
1:15:08the price in agriculture, the output in
1:15:11agriculture, and the labor share in
1:15:12agriculture.
1:15:16Uh you have it here. Sorry, I have it
1:15:17down there.
1:15:19Okay.
1:15:20So, what happened if
1:15:23I put the output in agriculture that the
1:15:26US has, and I change nothing else?
1:15:29All that I do is I give the poor
1:15:30countries
1:15:32uh sorry, I give all the countries the
1:15:35output that the US has.
1:15:37Then the log variance in output per
1:15:38worker pretty much disappears.
1:15:42Right?
1:15:45Alternatively, suppose I don't want to
1:15:48give them the technology of the US in
1:15:49agriculture, I give them the technology
1:15:51of the US in non-agriculture. This
1:15:54subject.
1:15:55What do I get? Well, it drops by about a
1:15:58half.
1:15:59So, yes, it's important.
1:16:01But obviously it's not as important as
1:16:03difference in agriculture. What happened
1:16:05if I give them the employment shares of
1:16:06the US?
1:16:08Again, it decreases, you know,
1:16:12uh quite a bit, close to 70%.
1:16:16So, these two components, the labor
1:16:19share in agriculture and the um the
1:16:22output in agriculture, seems to be key
1:16:24drivers of income disparities across
1:16:27countries.
1:16:28Okay?
1:16:29And that's why there has been a lot of
1:16:32um research on these issues.
1:16:35Now, because I I spoke too much, maybe I
1:16:38I I was awfully um
1:16:41um
1:16:44optimistic on how much I
1:16:46could discuss.
1:16:48Let me just uh leave you with the
1:16:50following thought.
1:16:52So,
1:16:54you know, trade trade economists tend to
1:16:56think of an economy between, you know,
1:16:57tradeable and non-tradeable goods.
1:17:00Growth economists think about
1:17:02consumption goods and investment good
1:17:04capital.
1:17:05And then the macro development
1:17:06economists think about, you know,
1:17:07agriculture, manufacturing,
1:17:10and services.
1:17:12But the the big idea that I want to
1:17:14leave you with is that
1:17:16relative prices in all these sectors,
1:17:18and hopefully you saw that from the
1:17:19measurement issues that we have
1:17:21discussed in, right? We went from
1:17:22consumption to investment to tradeable
1:17:24and to non-tradeable to sectors.
1:17:26There is a lot of information about
1:17:29productivity that comes in relative
1:17:31prices.
1:17:33Okay? So, while this the measurement of
1:17:35relative prices seems like an a geeky
1:17:37issue, it's actually quite important
1:17:39because it has a lot of information
1:17:40about productivity difference, which are
1:17:42first order to understand income
1:17:44income disparity.
1:17:46Okay? So, with that,
1:17:49I'm going to just let you go, and I'm
1:17:52going to open the floor uh in the last
1:17:5510 minutes uh for questions.
1:18:00Raul, do we have any any questions from
1:18:02the chat?
1:18:05Hello, Julieta.
1:18:07Hello.
1:18:08Hello, yeah. Um thank you very much for
1:18:10the class.
1:18:12Okay. Uh I would just like to raise a
1:18:15point. Okay. If I may. Okay.
1:18:18Mhm.
1:18:18Yeah. Uh
1:18:20According to findings, yeah, we
1:18:24we we notice that development accounting
1:18:27highlights human capital as an important
1:18:30factor of economic growth.
1:18:33And when we look into the literature,
1:18:36social capital is not a point which has
1:18:40been widely discussed.
1:18:42And however,
1:18:45uh we can see that there are some
1:18:48economic benefits of social capital.
1:18:51For example,
1:18:51you Can you be a little bit more
1:18:53explicit about what I have an idea what
1:18:55social capital means, but I just want to
1:18:57make sure that we we are on the same
1:18:59page. What do you mean by social
1:19:00capital? Yeah, uh maybe some aspects
1:19:03like
1:19:05uh confidence,
1:19:06integrity, lectures
1:19:09Mhm. among people.
1:19:11And yeah, what I'd like to raise is that
1:19:15uh we have many uh some economic
1:19:18benefits of social capital like people
1:19:21may find job uh through social networks,
1:19:25for example.
1:19:26And trust encourages more efficient use
1:19:29of of credit. And also
1:19:32um social networks uh
1:19:35are important for regional groupings of
1:19:38innovative activity sectors
1:19:41because they they can enable
1:19:43dissemination and sharing of knowledge,
1:19:47which are really important for economic
1:19:50growth, too.
1:19:51So, my question is uh
1:19:54what's your opinion about that? How can
1:19:56we take into account social capital? So,
1:20:00uh personally, I think that social
1:20:02capital may reinforce human capital and
1:20:05then uh contribute to to uh economic
1:20:08growth.
1:20:09Okay. Thank you. Thanks. That's a very
1:20:11good question, Samson. So, uh
1:20:13the answer comes into in two parts. The
1:20:16first one is that a lot of the features
1:20:19that we think about uh in terms of
1:20:21social capital are actually or say
1:20:23you know,
1:20:24um also the the way you manage uh your
1:20:27workers, etc., etc., those are all going
1:20:30to be
1:20:31uh in the accounting framework that I
1:20:34presented are going to be sucking to
1:20:36this productivity term.
1:20:38How efficiently do you use the same
1:20:41input of production?
1:20:43The The answer is in terms of that
1:20:45particular channel,
1:20:47I don't know what's the in the in an
1:20:49accounting sense, I don't know what the
1:20:50I can't don't know the to pinpoint what
1:20:52the the the actual magnitude of that
1:20:55effect is. But what I would encourage
1:20:57you to think about is the fact that uh
1:21:00you know, the
1:21:02differences in in what you will need is
1:21:06that the differences there are
1:21:08systematic differences in social capital
1:21:11across
1:21:12uh
1:21:13across countries and across income.
1:21:16And that I'm you know, I'm not a
1:21:18specialist in the field of social
1:21:20capital, but I'm not I'm not sure if
1:21:23that is that is true and if anything uh
1:21:26if it will go in the right direction,
1:21:28right? Is social capital more valuable?
1:21:32Uh likely social capital is more
1:21:33valuable in in poor economies than it is
1:21:35in rich economies, but I'm really in the
1:21:37world of hypothesizing
1:21:39at this point because I don't know what
1:21:40the actual answer to that is.
1:21:42So
1:21:43as I said, it's stuck in the Z effect.
1:21:47Uh since the real question is whether
1:21:49are systematic differences in income uh
1:21:53in terms of social capital and I am not
1:21:55aware of uh um
1:21:58uh research that documents this thing.
1:22:00There is a lot of research in terms of
1:22:02management practices, but that's a
1:22:04slightly different than what you're
1:22:05asking.
1:22:06Okay? Yeah, because uh yeah yeah,
1:22:10because uh
1:22:11I can see that
1:22:13uh when we talk about economic growth,
1:22:16uh we we we may not see the the direct
1:22:19effect, but when it comes to welfare,
1:22:23you know, I really think that
1:22:25uh social capital should not be put
1:22:27aside because it's not only a matter of
1:22:30what do you know, but it also involve
1:22:34who do you really know. Yeah, yeah,
1:22:36yeah.
1:22:37Okay.
1:22:38Thank you. Uh, so I would rather do you
1:22:41figure out your your your speaker
1:22:44situation? Okay, yes. Can you hear me
1:22:46now?
1:22:46Mhm. Okay. Yeah, my question is actually
1:22:49with respect to the concept of
1:22:51informality.
1:22:52Uh, do you think the TFP accounts for
1:22:55this?
1:22:56Yeah, so this is this is a a nice uh,
1:22:59question.
1:23:00Some of that is accounted there. The
1:23:02real issue is whether we're measuring
1:23:05employment properly.
1:23:07Right?
1:23:07Okay. Cuz the issue of informality is
1:23:09that you have a large uh, and this is
1:23:11what what that guy in paper is
1:23:13is in part about. Uh,
1:23:15he uses to measure properly uh, input
1:23:18shares. But,
1:23:20you know, self inform informality and
1:23:22self-employment, at least in developing
1:23:24countries, go uh, very close together.
1:23:27Okay? And then the question is the
1:23:28measures of employment that are there
1:23:30are measures of formal employment.
1:23:33Uh, but
1:23:34there is room to to think about uh,
1:23:37informality better. I know the people
1:23:39table
1:23:40um, has been doing an effort to put
1:23:43informal workers to measure informal
1:23:45workers uh, in a better way. But do you
1:23:47really share that view that was shared
1:23:49in the paper with respect to
1:23:52um, comparing self-employment and
1:23:54informality? I I think there there is a
1:23:56kind of difference between the two.
1:23:58No, there is a difference between the
1:23:59two, but what I'm saying is that a lot
1:24:01of the informal workers happen to be
1:24:04there are many of them that happen to be
1:24:06self-employed.
1:24:07That's all. That's on one side, but we
1:24:10we see characteristically that in
1:24:11developing economies
1:24:13uh, a huge chunk of transactions are
1:24:16done in the informal sector that are not
1:24:18accounted for.
1:24:19And and it's kind of deliberate or also
1:24:23because
1:24:23Oh, I see. I see. There are There are
1:24:25two issues. Okay, so there are two
1:24:26issues. It's not only whether you count
1:24:28employment or not, there is also the
1:24:29issue that GDP whether GDP properly
1:24:32measures the employment.
1:24:33Absolutely. Absolutely. Absolutely.
1:24:35There is actually an ice paper by
1:24:37Paulina Restrepo Echevarria that takes
1:24:39this idea very seriously seriously and
1:24:41tries to adjust measures of GDP by the
1:24:44size of the informal economy.
1:24:46And I think my understanding is that
1:24:48national So, to the extent that the
1:24:50informal economy is reflected in the
1:24:52national accounts, this is going to show
1:24:54up in
1:24:56in the bank world tables. The bank world
1:24:58tables are not going to produce new
1:25:00adjustment to the national accounts.
1:25:02One is my understanding is that in the
1:25:05developing world there is
1:25:07strong efforts to measure the informal
1:25:10sector in the value of of output. Some
1:25:13of them you know, the the measure of
1:25:15that success is yet to be seen, but I
1:25:17totally agree with you that
1:25:20you know, properly accounting for the
1:25:21size of the informal economy is actually
1:25:23quite important in understanding
1:25:25differences in income. Could you please
1:25:26reference the paper? Because for me that
1:25:28is very important because
1:25:29Yes, so let me just take your name. I'll
1:25:32I'll follow up with you with a link to
1:25:34the to the
1:25:36to the paper. Okay.
1:25:40Are there any more questions? Yeah, can
1:25:42I ask one more question, please?
1:25:44Yes. are also asking about this paper.
1:25:47Who was that?
1:25:49Some people in the chat are also asking
1:25:51about the
1:25:52the this paper. Okay, yeah. The the the
1:25:55question about informality, maybe you
1:25:57share it on the platform for
1:25:59What I can do is I can share it because
1:26:02I need to look for the link. So, what I
1:26:03can do is I can
1:26:06I can put it in my webpage and I can ask
1:26:08Mandy to post it in the in the
1:26:12in the website. I can also put it in the
1:26:14repository where you have the rest of
1:26:16the materials for the class. I think
1:26:17that was really nice. Yeah, thank you.
1:26:19All right.
1:26:20Okay, thank you. Andres, do you have a
1:26:22question? I believe I just posted a link
1:26:25to that paper in the chat. Ah, okay.
1:26:27Awesome. Thank you. I think it's the
1:26:28right one. Thanks, Doug.
1:26:31Yeah, thank you, Julieta. Um so,
1:26:35I've been thinking,
1:26:37is there some way you can disentangle
1:26:40like
1:26:41allocative uh efficiency with uh
1:26:44technological efficiency?
1:26:47Um may- maybe by assuming in sectors or
1:26:49something like that. I get the
1:26:50impression that
1:26:51um Yeah, so if you notice, um I think
1:26:55this is related to things that
1:26:57you know, there is a whole course
1:26:59related to this to this issue. And part
1:27:01of what the the the lectures that are
1:27:03coming up are going to do is they're
1:27:04going to put you at the frontier of what
1:27:06we know in terms of allocative
1:27:07efficiency versus difference differences
1:27:10in technology. So, a lot of those issues
1:27:12are going to be addressed in the sec- in
1:27:13the module on misallocation.
1:27:16Okay? I think Richard and Pete Klenow
1:27:18are going to present like what our know-
1:27:21the state of knowledge is
1:27:23uh in terms of how much of this is
1:27:26uh you know, differences in technology
1:27:29and how much of this is proper
1:27:31misallocation. I think your question is
1:27:33a good question in terms of something
1:27:34that I wanted to discuss, but I didn't
1:27:36have time to, which is this idea that um
1:27:40you know, different countries may be
1:27:41choosing different technologies given
1:27:44the endowments and the frictions that
1:27:45they face. Okay? And this is the idea in
1:27:48the paper by uh Caselli and Coleman, uh
1:27:52the world technology frontier.
1:27:54Okay? There they're taking seriously
1:27:57skilled and unskilled workers
1:27:59and they're thinking about factor bias
1:28:01technical change. So, basically, the
1:28:03idea that technical change affects
1:28:04skilled workers differently than
1:28:05unskilled workers.
1:28:07But also this idea that and I think what
1:28:09is uh so impressive about about that the
1:28:11findings that paper is that it seems
1:28:14just from a an accounting sense that
1:28:18poor countries use unskilled workers
1:28:21better, quote unquote, productivity
1:28:23terms than rich countries. So, that
1:28:25suggests that they're using different
1:28:27technologies.
1:28:28So, it's not that they're in a vacuum.
1:28:29It's not that, "Oh, well, we're not
1:28:31adopting the technology
1:28:33um
1:28:34and we're just worse, right? They're
1:28:36they're basically adapting the use of
1:28:38factors to the technologies that they
1:28:40can adopt. And they're adopting
1:28:42technologies that are complementary
1:28:44potentially to uh the factors that they
1:28:47have available. The question is to what
1:28:48extent they're doing that um in a
1:28:51constrained efficient way and uh to what
1:28:54extent this is, you know, misallocation
1:28:56of of of factors of of production. But
1:28:58that there's going to be a whole module
1:29:00about that. So, hopefully a lot of those
1:29:02questions will will come back again
1:29:04then. Sue, do you have a question?
1:29:07Uh hello. I have a very question.
1:29:09Hello.
1:29:10Uh so, I think in the accounting like
1:29:14equation that we have talked about like
1:29:17like assumes that there's So, so
1:29:20basically capital and labor are like
1:29:23separable, so there are no like
1:29:25like like uh so, any complementary
1:29:28relationship between them?
1:29:29Yes. Yeah, so I Mhm.
1:29:32Yeah, so what like if there is like the
1:29:35like difference in the usage of like
1:29:38factors, then
1:29:40uh what if whether it might affect the
1:29:43our income accounting or not. That That
1:29:45was my question.
1:29:46Thank you, Sue. So, this is another
1:29:48thing that I I didn't have a chance to
1:29:50discuss uh and it's part of my selected
1:29:53issues.
1:29:54There is a
1:29:55uh so, there is there is an issue on how
1:29:58we think about
1:30:00you know, measuring inputs and
1:30:02aggregating the efficiency of say, for
1:30:04example, workers or even capital, right?
1:30:06Capital of different vintages that are
1:30:08different. The way kind of the
1:30:10efficiency units or the standard
1:30:12approach to doing these things is to add
1:30:14them up linearly. And if anything, you
1:30:17know, compensate for the fact that say a
1:30:19high school worker is going to produce
1:30:21slightly more than a
1:30:24an uneducated worker. And the way we
1:30:27infer those differences in production is
1:30:29by looking at wage differentials.
1:30:31Okay?
1:30:32Now, there is a nice exchange between
1:30:36Ben Jones is all referenced in the
1:30:38slides. Ben Jones,
1:30:40Caselli and Ciccone.
1:30:42And then Ben Jones again.
1:30:44They literally get engaged into this
1:30:46discussion of, you know, what happened
1:30:48in a world where skill and
1:30:51skill and unskill workers are
1:30:53complementary, for example.
1:30:55Right? And Ben Jones says, well, if you
1:30:57do that, then the role of human capital
1:30:59is going to be much larger.
1:31:02But then, you know, Caselli and Ciccone
1:31:04said, well, wait, wait. You need to be
1:31:06careful because then one needs to really
1:31:09think that whatever differences in wages
1:31:11that we see between skill and unskill
1:31:14workers are not
1:31:16at all connected to differences in
1:31:18technology.
1:31:20Because if they are, part of that is
1:31:22really differences in, you know, in TFP.
1:31:24And then at that point, the role of
1:31:26human capital can be as little as you
1:31:28want and as big as you want.
1:31:30And then there is a nice response from
1:31:32from Ben Jones to to this discussion
1:31:34that I think is worth for for the
1:31:36students here to to read. Kind of
1:31:39saying, okay, yes, we can consider on on
1:31:42on this front. But let's think really
1:31:44about how to do accounting in a world
1:31:46where there are all these
1:31:47complementarities in factor of
1:31:48production.
1:31:49So, I think everything is there to be to
1:31:51be written. The one thing that I want to
1:31:53mention is that once you go into the
1:31:55world where all these inputs are not
1:31:58perfect substitutes in production,
1:32:00the simple accounting
1:32:03uh tools that I have brought up
1:32:05cannot really be used really off the
1:32:08shelf. And I think a
1:32:11even more important question is, what's
1:32:12the elasticity of substitution between
1:32:14these factors?
1:32:15Okay, even if you're going to go that
1:32:17way.
1:32:18I you know, there is an active
1:32:19discussion in developed economies, in
1:32:21particular in the US, as of what those
1:32:22elasticities are.
1:32:25Um so, I think you know, there there are
1:32:27a lot of open questions.
1:32:29Uh unfortunately, I don't know the
1:32:31answer, but I think yes, this is a
1:32:33feature of the data that is potentially
1:32:34important. It's also important uh in
1:32:37terms of what Andres was asking of, you
1:32:39know, the choice of technology given a
1:32:41constraint set.
1:32:43So, you know, I'll I'll look forward to
1:32:45the papers that have will come up on
1:32:46this.
1:32:48I'm sorry, but I have to to uh cut off
1:32:50the conversation. Again, uh there will
1:32:53be the um TA session for Monday.
1:32:55Uh we want to thank uh Julieta. Um it
1:32:58was a very big crowd. Uh people are
1:33:00dropping off now, of course, but um
1:33:03uh one thing for students to know is
1:33:05that basically everybody who's teaching
1:33:07here is doing this out of the generosity
1:33:09of their
1:33:10their hearts. Uh I also want to thank uh
1:33:12Mandy and Kiersten one more time. Uh
1:33:14next week, we will have on a Thursday,
1:33:16Todd Schoellman will be discussing more
1:33:18about human capital. Uh so, some any
1:33:20questions about human capital, you can
1:33:22bring those to next week. And uh that'll
1:33:24be Thursday, and then on Friday, uh same
1:33:27time, will be uh Bertolt Herrendorf
1:33:30talking about structural transformation.
1:33:33Thank you all.
1:33:34Thank you.
1:33:35Thank you. Julieta, fantastic.
1:33:39Thanks, Doug. Thanks, Joel.

About this transcript

This page contains the full transcript of Lecture 2: Development accounting: methods and findings by CEPR & VideoVox Economics, generated from the public captions YouTube serves with the video. The transcript has 14,243 words across 2,550 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

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

Free YouTube transcript tool

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