Session 1: The first 20 years of EMU
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
| Title |
| |
| Title of Series | ||
| Number of Parts | 12 | |
| Author | ||
| Contributors | ||
| License | CC Attribution 3.0 Unported: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
| Identifiers | 10.5446/48025 (DOI) | |
| Publisher | ||
| Release Date | ||
| Language | ||
| Production Place | Sintra, Portugal |
Content Metadata
| Subject Area | ||
| Genre | ||
| Abstract |
|
4
7
8
00:00
Meeting/Interview
00:58
Meeting/InterviewLecture/Conference
02:11
Computer animationLecture/ConferenceMeeting/Interview
02:59
Computer animationLecture/ConferenceMeeting/Interview
04:06
Lecture/ConferenceMeeting/Interview
05:01
Computer animationLecture/ConferenceMeeting/Interview
07:14
Computer animationDiagramLecture/ConferenceMeeting/Interview
08:03
Computer animationDiagramLecture/ConferenceMeeting/Interview
09:18
Lecture/ConferenceMeeting/InterviewComputer animation
10:37
Computer animationLecture/ConferenceMeeting/Interview
11:25
Computer animationLecture/ConferenceMeeting/Interview
12:26
Diagram
13:20
Meeting/InterviewDiagram
14:11
Meeting/InterviewComputer animation
15:19
Lecture/ConferenceMeeting/InterviewComputer animation
16:32
Lecture/ConferenceMeeting/InterviewComputer animation
17:48
Lecture/ConferenceMeeting/InterviewComputer animation
18:40
Computer animation
19:28
Lecture/ConferenceMeeting/InterviewComputer animation
20:22
Computer animationLecture/ConferenceMeeting/Interview
21:35
DiagramLecture/ConferenceMeeting/InterviewComputer animation
22:38
Computer animation
23:31
Lecture/ConferenceComputer animationMeeting/Interview
24:25
Computer animationLecture/ConferenceMeeting/Interview
25:16
Computer animationLecture/ConferenceMeeting/Interview
26:17
Computer animation
27:04
Computer animationLecture/Conference
28:15
Lecture/ConferenceMeeting/Interview
29:10
Computer animation
30:16
Meeting/InterviewComputer animation
31:35
Lecture/ConferenceMeeting/InterviewComputer animation
32:57
Lecture/ConferenceComputer animation
34:31
Lecture/ConferenceComputer animation
35:27
Computer animation
36:57
Lecture/ConferenceComputer animation
38:03
Lecture/ConferenceComputer animation
38:52
Computer animationDiagramLecture/Conference
40:30
Computer animationDiagram
41:23
Lecture/ConferenceComputer animation
42:20
Computer animation
43:42
Computer animation
44:38
Computer animationDiagram
45:34
Computer animationMeeting/Interview
46:36
Meeting/InterviewLecture/Conference
49:47
Meeting/Interview
50:45
Meeting/InterviewLecture/Conference
51:32
Meeting/Interview
52:33
Meeting/Interview
53:27
Meeting/Interview
54:39
Meeting/Interview
55:41
Meeting/InterviewLecture/Conference
57:15
Meeting/InterviewLecture/ConferenceDiagram
59:13
Lecture/ConferenceMeeting/Interview
01:00:01
Meeting/Interview
01:01:16
Meeting/Interview
01:02:13
Lecture/Conference
Transcript: English(auto-generated)
00:01
Good morning, everybody. We're going to have this first presentation that is going to be dedicated to real convergence. In order to do that, we have a paper from Yann Imps. Yann is professor of the Paris School of Economics and the New York University in Abu Dhabi.
00:21
And I think that you have co-authored the paper with Lauren Powell's lecture in the University of Sydney. And for the discussion, we'll have Sebnem Kalenky-Otkin from the University of Maryland. Well, real convergence is a key variable in economics,
00:42
and especially in a monetary union. The launching of the euro was about nominal convergence. And well, the party line in those times is that nominal convergence could pave the way for real convergence. But perhaps we cannot overlook that the determinants
01:03
of real convergence are real factors, mainly total factor productivity. So I think that these are the kind of things that we are going to discuss now. So Yann, you have between 20 and 25 minutes to make the presentation of your paper. Thank you.
01:23
Thanks very much. It's a privilege to be here. This is a part of a project I have with my co-author Lauren Powell's. And what we're trying to do here is to take stock on 20 years of convergence. And let me be a little bit precise about what
01:41
I mean in terms of convergence. I'm going to be talking specifically about two types of convergence. One in GDP, President Draghi actually alluded to already, convergence in GDP, that is the reduction in the dispersion of GDP growth rates, is obviously of paramount importance for the conduct of monetary policy.
02:00
First, because it's one of the criterion for an optimal currency area, cycles synchronize, and that makes monetary policy easier. And the second, because of the cohesion and homogeneity that that creates within the union. Convergence in consumption, which is intimately related with financial integration
02:22
and risk sharing across member countries of the EU, and therefore is intimately related with welfare and potentially even issues of inequality. But there are many other forms of convergence. Convergence in labor market outcome, labor productivity, et cetera. In my talk, I'm going to focus on the first two. And as a matter of fact, I'm going to dedicate
02:41
the bulk of these 20, 25 minutes to discussing convergence in GDP, the first point. Time permitting, I'll say a couple of things about consumption as well. So one very, very simple, straightforward way of looking at this issue is to compute the object that you see on top of this slide, which
03:01
is a measure of absolute differences between growth rates of GDP, that's the measure to the left, and of consumption growth rates, which is measured to the right. These absolute differences can be computed for the whole cross-section of pairs of countries and are going to capture how dispersed growth rates are of GDP or of consumption.
03:21
And the simplest way to think about convergence or lack of convergence is to simply investigate whether there exists a trend, an upward trend, into these dispersion measures. Now, these are absolute differences. They are absolute values. And there's a negative sign, which means when these things go up, they go up to a zero. And so that characterizes convergence.
03:41
And so what I've done here is I've just run a simple linear trend regression to establish that both of these measures over the 20 years, the first 20 years of the EMU, have actually increased. There's a trend increase that's manifest both in GDP convergence and in consumption convergence.
04:01
The trend is a bit more marked for consumption, interestingly, than it is for GDP. That continues to be true over periods of separated roughly the first 10 years and the second 10 years. And you can see it's there in both periods. So convergence is manifest. It's salient in the data, at least as far
04:21
as that measure is concerned. And so that begs the question, why do we observe? Why have we observed this convergence? So let me talk first about convergence in GDP. And by way of motivating the question we want to ask, let me just refer to an enormously large literature that has taken interest in that object.
04:43
Now, why is that interesting? Why is it important? Why is it an important question? It is because it's key to, again, the ease with which a monetary union can function. Few people would doubt that a monetary union actually stimulates trade. So that's the first arrow I have on this slide.
05:01
There's a little bit of disagreement about how much it stipulates trade, but it's pretty clear that it does affect trade. So more trade in monetary unions. And there's also a few people in this room or elsewhere that would disagree with the fact that more trade actually creates convergence. So that's a very well-known channel going from monetary union to trade to convergence.
05:22
And then there's this feedback effect that when GDP actually converge, that actually facilitates the cohesion and the sustainability of monetary union. So one of the things that I want to really talk about today, this morning, is the second arrow, the relationship between trade and GDP convergence.
05:42
That empirically has been established on the basis of observed trade. So on the fact that activities, sectors, countries that trade more, and we observe that trade, exports plus imports relative to GDP, are more synchronized. That's an exceedingly famous paper by Jeffrey Franklin and Andrew Rose in the late 90s
06:01
that established that. But of course, deep economic integration, akin to what the EU is trying to achieve, involves more than just more trade. It also involves that non-traded sectors, particularly service sectors, would actually cater for traded sectors.
06:20
So even though business services do not trade with each other, even though lawyers or accountants do not trade across the border, or to a small extent, clearly they tend to cater for companies, for sectors that do trade over the border. So that is presumably one of the characteristics of deep economic integration, when
06:41
non-traded sectors that do not trade directly actually cater for very open activities and very open sectors. So that measure presumably correlates to a certain extent with observed trade, but it is at least conceptually different from observed trade. So what I want to do is to compute this measure of connection between sectors and then between countries,
07:02
and try and distinguish its effect on convergence from that is the well-known one created by observed trade. So how is it possible to come up with a measure like this? Let me just use a very simple diagram to illustrate what I have in mind. This is what I have in mind. On both sides of the border, you have non-traded sectors, say business services,
07:22
that actually supply inputs to manufacturing sectors, traded sectors. And the trading that we observe is between manufacturing sectors, but it is not between business services. Yet business services are actually quite connected across the border, but sort of one step removed, because they are very connected to open sectors.
07:41
So the name of the game here is to distinguish a value chain akin to what I'm showing you now, which is one that involves one border crossing, or for that matter, more than one border crossing. It can go back and forth across the border. That's also what deep integration would be about. Distinguish that value chain from one that actually does not involve crossing a border. And here is an example of that that is purely domestic.
08:02
The sort of connection measure that I'm interested in is one that would capture the black path here and filter out, in some sense, the red path here. And that's going to be something that is going to be reasonably easy to do using input-output tables that capture these vertical trade
08:21
between different sectors. Now, input-output tables actually tell us a lot of information, often quite messy and actually quite messy to expose as well, so hence the triple summations. What this complicated expression actually tells us is quite simple. It tells us that gross outputs in a given country i and a given sector r is going to serve
08:44
demand, final demand, in a variety of different ways. It can serve it directly, and that would be the first f's that you see in both of these brackets to the left. The fii r at the bottom basically says that this good i is actually consumed through final demand in the same country
09:02
and in the same sector. The sum, so the sum of the fij r, it means the same thing. It means that that good serves final demand, but across a border. So this is crossing a border, but in the form of final demand. So that's the type of trade that we used to think about, it's trade in final goods. Of course, there's also second-order linkages.
09:22
That is the second set of summations here, where at the bottom here, what you have is the possibility that that good r in country i is being sold to another sector s in the same country, and it is that other sector that actually serves final demand. You can have that in a purely domestic sense,
09:40
that's the bottom bracket, or in an international sense, that's the top bracket. Top bracket captures the various ways in which, at the second order, that good r can cross a border. It can cross a border as an intermediate good across sectors or as an intermediate good within sectors. That would be a second-order link towards the distance
10:02
from final demand. And so on and so forth. There's a third order, a fourth order, fifth order, et cetera, et cetera. So what I'm trying to do here is to disentangle, basically, to isolate the first top bracket from the bottom bracket. I want to isolate that component that really captures these intermediate inputs crossing borders.
10:21
Now, that's actually a pretty difficult thing to do, even though these data is available, because these are infinite sums. Thankfully, there's a very influential paper by Antros and Chor a couple of years ago that slightly amends the objects that I've just introduced to you by, if you see, there's a one, a two, and three, now they're weights,
10:42
that actually capture the distance to final demand. So that weigh these distances by the number of steps for each element in these brackets. Now, this object actually can be computed very easily using input-output tables, and that's the formula you see at the bottom there. The formula at the bottom there involves a, which is a simple input-output table,
11:02
a world input-output table that we have data on, and f, which is a vector of final demand. The bottom line here is that I can easily compute the sum of these two brackets by using a world input-output table and a vector of final demand. Now, by analogy to this result,
11:21
I can also compute the second bracket, which is the one that's focused on domestic linkages only, by specializing the matrices a and f onto purely domestic input-output linkages and purely domestic final demand. That's the notation that's at the very bottom of the slide. Now, having done this, I have an expression
11:40
for the left-hand side, what I've called u, and I have an expression for the purely domestic component so I can back out for any sector in any country the propensity, in some sense, the intensity with which that sector exports across the border directly and indirectly at first, second, third, fourth order, et cetera, et cetera. Now, I'm gonna call this measure
12:02
a measure of export intensity, bearing in mind it's not directly observed, it's not a measure of trade. It's a multilateral, so far multilateral measure of export intensity in the sense that it captures how closely, how export intensive, how closely linked to all of the country's partner
12:21
given sector r is in country i. So let me show you what that looks like. So what I've done here is I've done that computation for China, e-mu, EU, not e-mu, and the US. China is in black, the e-mu is in green, EU, not e-mu is in purple, and the US is in red.
12:41
So you can see export intensities have very different trends between China and the other three groups. China has an upward trend, so that makes it a very open economy in the sense of this export intensity being quite large, as in lots of sectors in China are geared towards exports.
13:00
But you can see that post 2007, the measure actually falls back down, which is presumably consistent with the rebalancing of the Chinese economy towards its domestic activity. And what's interesting is that at the beginning of the euro, China and e-mu are at similar levels, and at the end of the euro, again, China, and at the end, today, China and e-mu are at the same level.
13:23
So there's a sense in which exports intensity in the euro area is comparable with what it is in China. By the way, what I mean to say here is that the exports intensity of e-mu member countries with each other is comparable with the export intensity of China with its trade partners in general.
13:43
What's also interesting is the gap between e-mu, green, and e-mu, green, and e-mu, not e-mu, the purple one. And so there's an e-mu effect in the sense that exports intensity between e-mu members is larger than it is between EU members out of the e-mu.
14:02
With all of that, the US is actually at the bottom. It's actually a country that is largely closed according to that measure. So that's the first thing that suggests this measure is potentially interesting and says something else than just pure observed trade. The second thing that I want to do is to try and bilateralize this
14:21
so that I can say something about bilateral connection between sectors across countries. The way I'm going to do this is I'm going to specialize what I've just showed you before. Rather than using a matrix, an input-output matrix of China with the rest of the world, I'm going to take input-output matrix of France with Germany.
14:42
And that's going to tell me the export intensity between France and Germany. It's going to tell me how sectors in France, sector R in France, is actually exporting intensely to Germany. And the way I'm doing this is by introducing IJ versions of all the relevant matrices that you see on that slide.
15:00
And that gives rise to a very easy way of measuring bilateral export intensities between sectors and between countries. Now, one of the reasons why this is interesting is that this can be computed for non-traded sectors. In particular, it can be computed for services, business services. I'm actually able to say something hopefully meaningful about how connected across the border the typically non-traded sectors are.
15:26
And I'm going to call this BI for bilateral intensity. And let me show you this again by way of documenting what these measures look like. So these are pairs of sectors
15:40
between EU member countries, that's in red, between the US and its trade partner, it's in blue, and China and its trade partner, that's in black. And you can see that the overwhelmingly apparent result is that the red numbers are much, much larger
16:00
than the blue or the black numbers. EU member countries are highly integrated bilaterally in terms of sectors actually being connected with each other. You can notice, so manufacturer with manufacturer has very high numbers, which is not too surprising. These are connected traded sectors, but you can see that services, business services,
16:23
here I've sampled a few, retail also take relatively large values within EU countries. So this is an index of how closely connected non-traded sectors are across the border within the EU. You can see the numbers for business service with business service are actually not zero.
16:42
There is some connectivity between these because they in the EU tend to serve sectors that do trade with each other. Now obviously the $64 trillion question is whether that explains convergence. So the way I'm gonna do this is by running the regression that's up there,
17:01
which is on the left hand side, the measure of convergence in GDP, except that this is now measured at sector level. So I'm gonna look at how sector R in country I is converging with sector S in country J. And I'm gonna try and explain this with the bilateral intensity measure just described, which is this BI variable right to it. And I'm gonna run a horse race with trade.
17:21
Now trade here is basically the effectively traded the amounts and value that is traded between these sectors and these countries coming directly from the input-output table. So these are directly observable trade between these sectors. So it's defined in a usual way which is the amount that is being traded, that's disease normalized by value added.
17:43
Now this is a huge data set because I'm gonna look at all bilateral relations between all the sectors in this case 12 core EU countries. So what I'm gonna have is 12 countries times 50 sectors, times 50 sectors because all the bilateral relationships,
18:01
times 15 years, I only have 15 years because that's the input-output data that we do have. So that gives me a data set of 2.5 million observations whose purpose it is to try and explain the dispersion in convergence in sector value added growth rate using this bilateral intensity measure.
18:20
So let me show you the results. These are the results. So again, the idea here is that, I haven't said that, but I have a battery of fixed effects. So in particular, everything here is estimated within over time. It's an increase in bilateral intensity and I'm looking at whether that increase in bilateral intensity actually explains increase in convergence or convergence.
18:41
And so the first line is the one of interest. The first line captures whether that's significant or not and it's systematically significant and positive. And I'm adding different fixed effects, different combinations of controls really, and also trade, the directly observed trade. So a couple of things are interesting. The first one is that the trade coefficient
19:01
is actually unstable to the inclusion of fixed effects. So it's actually very hard when you use observed trade at the sector level to explain convergence in value added growth at the sector level. You actually cannot do this in any robust way. However, with that bilateral intensity measure, you actually can do that. You can see that the coefficient
19:20
is systematically significant and positive and economic significance is actually quite substantial. It's roughly three times larger for bilateral intensity than it is for the trade. So it is a relevant variable. Another way of looking at the same question is to sort of look at specific pairs of country, sorry, specific pairs of sectors.
19:42
This is still within the EMU. Specific pairs of sectors as opposed to, this is the whole cross-section of all sectors. And I'm here specializing on some pairs between the sort of three key activities in the economy. And so you can see again that what's remarkable with this bilateral intensity measure is that it really explains very well
20:01
the correlation between manufacturer and services, services and services, agriculture and services. It's actually explaining quite well the connection between services and other sectors of the economy across the border in a systematically significant way. Those are the last five columns in the table. Compare this with trade,
20:20
observed trade does a really bad job. It's a very unstable job. And that's not surprising because once again, observed trade is just probably for some of these non-traded activity, a small part of the actual linkages between these sectors across countries. Okay, so the last question is how does that aggregate up? That is, how much does the ability of this variable
20:43
to explain correlation at sector level between countries actually aggregate up to explaining aggregate cycles, which is eventually what we're interested as microeconomists. Now, absolute values really do not sum up very nicely. They're quite awful. So what I'm doing instead here is I'm using a quasi-correlation measure,
21:01
which is the object that's written here, which can be computed in every period reasonably easily. And that actually does aggregate quite nicely. It's the last double summation at the bottom of this slide that tells us that Q, I, J, T, the aggregate correlation, quasi-correlation between cycles, is a weighted average of all the sector level
21:21
without full correlation, okay? And so now the name of the game is how much of that Q, I, J, T, change in Q, I, J, T, am I going to be able to explain using the bilateral intensity measure I just described? Okay, by the way, using the quasi-correlation measure, you get the same result that convergence has happened,
21:41
that is called the distribution of these correlation at country level has shifted to the right. Countries have become more synchronized first 10 years versus last 10 years, okay? So how am I going to do this? Well, I'm going to fit a model at the sector level, which is the model that's up there, and I'm gonna get fitted values of the quasi-correlation at sector level Q, I, J, R, S
22:03
with a hat, that's my fitted value. The first one is the fitted value that's predicted by bilateral intensity. The second one is the fitted value that's predicted by trade, just to sort of run again a horse race between the standard explanatory variable for correlation that trade is. And once I have these fitted values, I'm gonna aggregate them up
22:20
using the formula I just showed you. Those are the two summations at the bottom. And then I'm gonna get a fitted, predicted aggregate correlation, as implied by bilateral intensity, as implied by trade. And I'm gonna compare this with the actual measured bilateral correlation, okay? So first thing, this is my fitted model.
22:41
So these are regressions, not unlike the ones I showed you before, where I'm trying to fit bilateral sector correlations on bilateral intensity and trade. And I'm gonna take my predicted values from this model. Here's what it gives us. I guess that's a headline result. This regression is exploring the ability
23:04
of the bilateral intensity variable to explain from a sector level aggregate correlation versus the trade variable to explain at a sector level from a sector level aggregate correlation. So the first column tells us that the model
23:21
that is only fitted on bilateral intensity does a pretty good job at explaining the increase in correlation. The R-square is 20%. That basically says that bilateral intensity explains 20% of the increase in aggregate correlation. And that's, again, built up from a model that's fitted at the sector level.
23:40
There's no macro shocks here. There's only sector shocks. Compare this with trade, just to give it sort of a yardstick of how big or small that is. And for trade, you get an R-square that's a little bit smaller. And I guess if you put the two of them, which is a tempting, normal thing to do, what you get is it's exports intensity that seems to come out on top. So at the very least, that measure is on par
24:05
with the standard measure of trade intensity that we observe in terms of explaining GDP convergence. Now, I have two minutes. And in two minutes, let me talk a little bit about convergence in consumption, which is the other thing I promised I would say something about.
24:21
Now here, this is actually very charted territory. I'm sort of revisiting in a slightly different way an enormous literature. The idea here is to run risk-sharing tests, that is tests that investigate whether the growth in consumption can be explained by financial integration. And financial integration, these regressions are well known.
24:40
There's one in the middle here. This beta that you see here is going to measure the extent to which growth in consumption can be unhinged from idiosyncratic income shocks. It's with financial integration that I can consume in a way that is not dictated by local income. And estimates of beta should be zero under perfect financial integration.
25:00
And with financial integration matters, you'd expect the interaction, which is in the bottom regression here, this beta two to be negative, more financial integration, lower coefficient. I'm going to measure this financial integration measure using the number of financial service action plan directives adopted by each country at time T. This is directly inspired from my discussant.
25:23
And ask whether beta two is negative or not. The slight wrinkle that I bring to this approach is to sort of fit it to the bilateral measures that I've discussed throughout these 20 minutes. And actually look at whether convergence in consumption can actually be explained or related
25:42
with convergence in GDP. And that's the coefficient delta that you see here. And again, same idea. You'd expect delta to be zero under a financially integrated area. And the question is the same. Do we observe delta to be closer to zero when financial integration is more prevalent? That's the delta two I have at the bottom there.
26:02
Again, how do I measure financial, bilateral financial integration between two countries? 7M has this measure which is basically looking at the number of common FSAP directives adopted by countries I and J. When you do this, you get these results. And these results actually after I found them, I discovered are pretty well known
26:20
and reasonably conventional standard. What that tells us is risk sharing is imperfect, quite imperfect. These coefficients beta on the upper panel, delta on the lower panel are really far from zero. But if you look at the interaction, what you see is that in the earlier period,
26:40
99, 2006, there's a negative coefficient which suggests that financial integration through these FSAP directive is actually achieving a larger, slightly significantly larger degree of financial integration. Whereas the opposite happens in the more recent period, the post Lehman Brothers crash from 2008. That's true across both approaches.
27:03
So let me conclude because I'm short on time. So I'm doing two things. Well, one really big thing and one really small thing in this paper. The first one is to propose a novel way of measuring convergence or synchronization in GDP growth which has to do with a measure of connection
27:22
between sectors and also between countries that actually is not based on observed trade but rather is based on input-output linkages and the proximity of two sectors across a border through their input-output linkages. What's interesting, what I find is that that does explain convergence quite substantially
27:40
at sector level. It trumps the typical explanation coming from direct trade. And in terms of how much it can explain of the aggregate change in correlation, it's explaining about 20%, which is at least on par with the standard explanation on trade. The second thing and smaller thing that I do is consumption correlation
28:00
and what seems to be happening is that the directives, the FSAP directives did have an effect on financial integration pre-global financial crisis but not post-financial crisis. And let me stop here. Thank you very much, Jens.
28:21
Now, Sebnem, it's your turn. Let me start by thanking the organizers for inviting me. It's a privilege to be here and celebrate the 20th anniversary of Euro with this distinguished crowd.
28:42
This is a topic that is dear to my heart. I spent a significant chunk of my adult life thinking about the European integration. And I think it's a great paper. So I'm going to try to highlight the key points and bring another perspective to the table.
29:04
So let me start by stating the key facts that the paper documents. So the paper shows that there's convergence in pay-wise growth rates of GDP and consumption. The explanation for this convergence is the value chain. So the idea here is there are sectors
29:21
that are not trading with each other across border but they are linked to the sectors that are trading and these sectors help convergence both at the sectoral level and at the aggregate level. And the last fact the paper shows is there's also convergence in consumption due to financial integration. I think this is an extremely comprehensive paper,
29:43
very nice contribution. It's an amazingly rich canvas of data that the paper is providing us with. And it introduced this new trade measure, the main contribution of the paper that really helps us to understand how deep the European economic integration is.
30:00
So my comments are going to be on first, discuss the measurement of GDP convergence, focusing on the role of financial integration. In this process, something Jean didn't have time to go through and then I'm going to link this to total factor productivity convergence which is another real variable we do care about when we talk about real convergence.
30:20
Then going to talk a little bit about consumption convergence and if I have time, I'm going to talk about importance of having a framework to interpret the result and to drive out policy implications. Let me start with measurement. I would like to distinguish the two concepts, GDP convergence and GDP synchronization because I believe what the paper is doing
30:42
is measuring GDP synchronization which is an amazingly important object for an optimal currency area as Jean pointed out. You can use two metrics if you would like to look at GDP convergence. The first one is the standard metric that comes from the growth literature. It's based on a neoclassical growth model known as beta convergence.
31:02
This metric is about whether or not poor countries and regions grow faster so they will catch the rich countries or rich regions. So that's the standard one. That's what we understand with real GDP convergence. There's a second measure which the papers is adopting that comes from the International Business Cycle synchronization literature
31:21
and that measure is synchronization of GDP fluctuations, the author's measure. Obviously, it's an extremely important measure in the context of optimal currency area going back to the influential work by Mandel on this, we know that desirable economic integration centers on the degree of synchronization of GDP fluctuations across borders
31:40
and we later understood that economic itself, economic integration itself affects the degree of synchronization of GDP fluctuations. So if we want to put this in a framework, let's first start with the standard measure and then go to the second synchronization measure. So the standard measure is a very simple measure
32:02
and that measure is going to give you divergence in GDP for EMU-12. You know, you may not put that much emphasis on this because at the end of the day, you are looking at 12 countries and the measure is a simple regression of growth rate of those countries on their initial GDP per capita levels.
32:20
When you do that regression, you will get a positive correlation. The beta convergence is about getting a negative correlation. If you want to say poor countries grow faster. So here, the first panel shows that this didn't happen. The literature does this conditionally, meaning conditioning on the determinants of the steady state where countries are converging to, such as you control variables as education rate,
32:41
investment rate for tilt rate. When you do that in panel B, that's the conditional beta convergence figure that shows neither a divergence or a convergence result. So no relationship between the initial income levels in 2000 and the subsequent growth in the next 15 years. Again, this is 12 observations. You may not want to do this with 12 observations.
33:01
You may want to look at regions. In fact, there's a very active literature in the US that looks at the US state's convergence. Now when you do that with regions, these are not two level regions of Europe. You do get convergence. The first figure shows you that there is convergence among regions within EMU countries that says poor regions grow faster.
33:22
This is color coded so you can see that purple, orange, these are the regions of Greece, Spain, Netherlands, Ireland, Portugal, they are kind of these top regions. They start at poorer, they grow faster. Now when you look at the second part of the sample after 2007, this breaks down again.
33:40
You get a divergent result. That's the second chart. But that might be because these are unconditional correlations between the growth rate of regions and the initial income per capita level of regions maybe a better thing is to look at the conditional correlations. And that would be this table summarizing these results where the beta coefficient is going to be the coefficient
34:00
on the initial regional GDP per capita where we are regressing the growth rate of the region on that. And you see the first column gives you divergence but the minute you control the author's favorite measure, export intensity in the second column, you get a convergence result across the European regions. And then of course this is a country level variable here.
34:21
I'm controlling, you can also do it with a country fix effect in column three for the entire sample period and column four and five before and after the crisis, you always get regional convergence. So what is this saying is if you just use the standard beta convergence measure from the neoclassical growth model, you get a no convergence result for 12 countries
34:41
in the sense that poor countries didn't grow faster. But you definitely get a astounding guess for the European regions that poorer regions did grow faster than the rich ones. And this is actually part of the whole cohesion framework. It is very important for the economic integration within the EMU. Now let's look at the second measure, which is the author's measure.
35:01
That's the synchronization of GDP fluctuations. Clearly, the GDP became more synchronized. The growth rates became more similar during these 20 years. Why is that is the case? The first thing we would like to understand is the fact that this measure of GDP convergence, which is a synchronization measure, is endogenous to trade
35:22
and financial integration. Obviously, authors realize this, and their central argument is about the endogeneity of this measure to trade. Their explanation is that the sectors that don't trade with each other, but they are linked to the trading sectors, are going to have more similar growth rates. They are going to have synchronized GDP over time,
35:41
and that's going to give you aggregate convergence at the country-payer level. Let me put this in a general framework, since this has been the box and notes of the optimal currency area literature. This is a figure reproduced from my 2001 paper in JE. What we tried to do in this figure and in that paper
36:00
is let us go and summarize the theoretical and empirical literature that's tried to understand if a currency area can be an optimal currency area exposed, even if it wasn't an optimal currency area ex-ante. So here, you will see that there were several papers that argued that if countries have similar technology, similar structure policies, more trade,
36:22
this is going to give them more GDP synchronization. Their growth rates are going to be more similar. Financial integration is going to work in the opposite, because financial integration is going to drive capital flows, induce specialization, and have trade in different goods, so that might give you less GDP synchronization. So the authors adding another box
36:41
there through trade integration, instead of just doing the intra-trade linked to the value chain, they are saying, let's also look at the part of that value chain that just don't trade with each other. And is it the case that that gives us more GDP synchronization? And the answer is yes. Let me skip this.
37:01
If we run the regression, they graciously provided their data to me, so I can do a few experiments with their data. So the first column here is their regression at the sector level. So this is, again, regressing the negative of the difference of the growth rate, so it becomes a measure of GDP synchronization on export intensity.
37:20
And you see in column one that export intensity has a very big positive coefficient together with trade intensity, which is not significant. So this is their main result, that export intensity is something that captures more than trade. I totally believe in this result, and I think it's a very important result that it shows you that the sectors with higher export intensity have similar growth rates.
37:41
Now you can add financial integration to this regression, and you are going to get a negative coefficient as shown with blue in the first column. Why is that? Because financial integration is going to make sectors and countries diverge from each other because of the capital flows and specialization patterns. But the important point here is to understand this measure is not only going to be a function of trade and financial integration,
38:02
but also going to be extremely sensitive to shocks. For example, column two is going to be a specification adopted from the IMF VO chapter in 2003, where instead of using country time fixed effects in column one, you remove them and you put a fixed effect for the crisis here, a crisis dummy.
38:20
That is the first line, the coefficient in black, bold. And it's a big positive, right? Why? Because during crisis, of course, the growth rates move together down across countries' sectors. You can enhance this specification by the interaction term of each of the variables that affect synchronization, like export intensity, trade intensity, and financial integration,
38:40
together with crisis. And you can see that the effects are going to be different in normal times and crisis times. The last column does the same exercise with aggregate data for country pairs. And you can see that, for example, financial integration makes the growth rates diverge during normal times, but then converge during crisis times. And export intensity is going to have the opposite role.
39:01
So the key here is, and the same thing you can do with the quasi-correlation measure, the key here is that all these measures capture similar things. So I'm going to show you three plots here. The first one is plotting the growth rate differential measure against the standard pairwise GDP correlation. The second one is the same growth differential
39:22
against the quasi-correlation, as Jean showed you, that they use also to link the sector to aggregate. And the last one is the quasi-correlation and GDP correlation. So these measures are going to be sensitive to shocks, policy changes, and the trade and financial integration. So it would be useful to kind of see them
39:41
in a broader framework. It will be also helpful to understand the nature of these measures when we want to think about total factor productivity convergence. Now, the paper clearly makes the case that sector level GDP become more synchronized, because the sector level growth rate differences go down over time. And that's because of the degree of export
40:01
intensity of these sectors. What does it imply for the total factor productivity convergence? On the one hand, if these sectors, although they are not trading with each other, if they are sharing a similar technology, that will bring you a total factor product of the convergence. Another story might be the fact that there might be misallocation of resources
40:21
within those sectors across firms. This is the story we pursue in my joint work with Gita Gopinath, Lucas Carabarbonis, and Karla Villegas Sanchez in 2007, where we only focused on manufacturing sector and the southern European countries, where we show that the extent of the misallocation of factors actually give you TFP divergence
40:43
within the manufacturing sector. Here, I'm going to plot you the aggregate TFP. So the first panel is going to show you four countries, Spain in black, Italy in orange, Germany in red, and France in blue. Starting in 1999, everything is normalized to one at 1999. You see a TFP divergence across these four countries.
41:02
When you look at EMU-12 and separate it as north and south, you also see a divergence to a lesser extent, getting a little bit higher after the crisis. It's important to understand how these two forces interact, because clearly, if the export intensity affecting many sectors prevails at the end,
41:22
that is going to be a force that is working against the misallocation of sector, because that's going to converge the technology and going to help to aggregate TFP convergence. Okay, so let me say a few things about consumption convergence. And at the end of the day, as a European Union, this is the most important thing from a welfare point of view, right?
41:42
So we care about consumption convergence. And as the paper correctly also puts it out, EMU countries might be having different shocks, and this was emphasized both last night's dinner speech and today's chairman's draggy speech. Countries are going to face idiosyncratic, different GDP shocks, right?
42:00
If you are in a union where countries face different GDP shocks, what matters at the end, if financial integration smooths out these shocks, so the citizens of EMU is going to have smooth convergent consumption. Authors measure this with the standard risk-sharing regressions at the aggregate level and the bilateral level. The idea behind these regression is very simple. You try to understand how much consumption
42:21
can move with GDP, okay? The less it can move, that means there is more consumption commuting. The more it can move with GDP, there is less consumption smoothing. The headline result of the paper, when they focus on consumption convergence, is that financial integration is great because it smoothed out consumption fluctuation before the crisis, but it made it worse
42:41
after the crisis. The before the crisis results are very robust, and it's very similar to what has been found in the literature so far. I would argue that the after the crisis result is weaker, why? Well, it's significant in the aggregate regressions, but not in the bilateral regressions. What might be the reason underlying this? Let's start with their measure.
43:01
Their financial integration measure is based on the FSAP directives. I mean, this is a great measure. Since I introduced it, I love this measure, obviously. But let's try to understand what the measure says and what it might imply during a crisis. This is an exogenous measure, of course. It's great. It's about two countries passing the law,
43:21
and it's extremely policy relevant. When we originally introduced this measure, we argue repeatedly, originally in the 10th anniversary of the euro at the conference at ECB, that this measure is going to help you to disentangle the effect of single currency from the financial harmonization post on financial integration. So it's clearly major relevance
43:40
for the process of further integration. But at the same time, this measure is about the laws, right? What the author's finding is that consumption can move more with GDP in pair of countries that pass the laws at the same time. Okay, but this doesn't mean these countries have contagion consumption. They might pass the law,
44:00
but it doesn't mean they are yet deep in the financial integration, integrated de facto that that law implies. In fact, the paper we wrote for the 10th year of the anniversary shows that financial integration helped consumption smoothing tremendously, but there's a different role played by cross-border liabilities and assets, assets making the job, which is the essence of the banking union
44:21
and capital markets union. And as we know, during the crisis, banks deliver, shedding the assets, so this is going to affect a lot what is going on during the crisis. Okay, so let me conclude, because I'm out of time, just let me say one thing. I think it is also important to have a framework to be able to link the sector level result
44:42
to aggregate result. The idea is this, right? There are trading sectors, not trading sectors helping the convergence. How do we link export intensity is something that doesn't vary over time as you can see here, it's flat for many countries, but we are trying to explain something time varying that is convergence over time. And to also help to have a framework
45:02
in terms of understanding right counterfactual. EMU-12 GDPs are more synchronized relative to U.S. and China for sure, but does it mean U.S. and China are subject to different temporary and permanent shocks, they have different degrees of trade and financial integration, or EMU-12 are integrated more. So a framework will help us to also compare the findings
45:21
to the right counterfactual. Let me conclude. This is a very interesting contribution that focuses on GDP and consumption convergence, two very important concepts for the economic integration in EMU, and it provides a view of integration that is much deeper than the standard trade integration because of the effect of the value chain
45:42
and the sectors that are linked to the exporter sectors through the value chain. I believe it is important to note that financial integration might cause divergence in GDP during normal times, but it is helping tremendously to consumption to converge, and this is a union at the 20th year, the banking union just getting full,
46:03
so it is going to get more and more moving forward, so this is an important thing for us to realize. And the open question on the table is that if the non-trading sectors converge due to their link to exporting sectors, this is the value chain link, why the aggregate GDP didn't converge in the terms of the standard measures.
46:21
This can be just a data issue, which will be fine, but it can also be a fact that not enough sectors yet linked to the exporting sectors. Then it will have different pulse implications, so it will be important to sort this out. Thank you.
46:41
Thank you very much, Zeban. I think that we have between 10 and 15 minutes to open the debate. Thank you very much, Richard Baldwin, the Graduate Institute of International Studies, and voxeu.org. First of all, great paper.
47:02
I think it's an innovative approach, very glad to see it. What I wanted to do is provide a little bit of a trade perspective on where we are in these kinds of measures. So two things which you didn't mention, which may help clarify why it's so important to do it the way you do it, is this difference between value-added trade and gross trade.
47:21
Because trade's crossing borders several times, the standard measures of trade have become more unreliable in actually indicating how much countries are integrated, because what you really want to know is what is the value added in the physical products crossing the border, and where that physical value added came, not just the measures.
47:42
So it's a big step forward to use these multi-product, multi-nation input-output tables to figure out where stuff actually comes from. The second trend, which you didn't mention, but I'm sure you're aware of, is called the serviceification of manufacturing, and that the true value-added of manufacturing sectors
48:03
is coming increasingly from services that are added before fabrication, or services that are added after fabrication, and so you get this, what you picked up is countries are more integrated than would seem simply according to the value-added trade. So we even have a name for that,
48:21
we call it mode five services trade, or embedded services trade. Now what I would like to suggest, and this goes to Sebum's point about the numbers, is we've actually taken another evolution in trade using the value-added tax data to figure out who's connected to which firms which are then actually exporting.
48:41
And it's only done right now in Belgium, but the National Bank of Belgium has all this data, and they can see which firms are selling to the firms that are then actually exporting. So you get kind of a friends of a friends of exporters approach to it, and what you find is that the economy, Belgium in this case, many, many more firms
49:02
are actually exporting indirectly, but with you have the value-added tax, which is the complete universe of transactions, you can get a much more defined picture. Now right now only Belgium can you do that because it's the central bank that has the VAT data and the trade data, but in principle you could connect the entire thing,
49:21
so maybe if we take this seriously, we could do a whole new project at the ECB like we did the firm level trade data, but connecting the value-added tax at the firm level to the behavior of trading, and then you can really figure out how connected different sectors are. So I would just make that as a suggestion.
49:42
Great paper, too bad I didn't think about it first, but it's really, really nice, thank you. Thank you very much, Jen. So do you want me to reply to Richard? So thanks very much. What I had in mind was those papers that do that at the firm level,
50:01
but I just didn't have the data on them, so doing it at the sector level in a global framework is what I chose, but I agree it's second best. I mean, gosh, if the data you described could be collected in new countries, that'd be fantastic, to be honest, to ask this question beyond Belgium. And you have value-added trade versus gross trade,
50:21
I think you're right. It'd be interesting to see what happens, for example, in these regressions when rather than using the trade measure which is embedding both intermediate and finals, I would actually use some procedure to filter out what's intermediate and see what that horse race would look like. I agree with that, that's a good idea.
50:42
Thanks. President Jean. Thank you. Well, first of all, it was a very, very smashing exposition, and I have to say that I see now things through a different angle. It's very innovative, if I may, and I thank you very much for that,
51:02
and also I have to say the dialogue was very, very stimulating. I have a question on chart seven of your paper. I'm not surprised by chart seven. It goes, in my opinion, very much in line with academia, and even the IMF worked on that,
51:21
so I'm not surprised to see that we are substantially more open than the United States of America, more or less at par with China, it doesn't surprise me. I think that also work has been done on the integration to the global value chain, and we are more integrated in the global value chain than the US, so it's both, I would say,
51:43
a trade and the integration of global value chain. Where I am a little bit struck is that when you say we compare the average export intensity of EU countries with the rest of the Union, so it seems that it is an intra-average integration, and you compare that with China and the US,
52:01
if I understand well, vis-a-vis the rest of the world. So, of course, the question mark is what would happen if you would divide China in various region or the US in various states? Would you have more or less the same or not? I open the question. Of course, I would like very much to understand why you consider fully legitimate to make this comparison,
52:23
which seen from in the global perspective looks a little bit asymmetric. Thank you. Thank you. I agree the comparison is somewhat diagonal. I, again, did with what I had.
52:41
I agree with you it would be very nice to have something inter-regional or inter-state in the US or provinces in China, that these data just don't exist. We don't have any regional dimension to the input-output table of the US, for example, and that's what we would need to do that. So I tried my best. In the spirit of going closer to a somewhat less diagonal comparison,
53:02
that's why I included EU, not EMU. So these are countries that are not in the EMU, but in the EU, and that includes all the trade between these countries and within these countries and between these countries and the EMU. So it goes one step in that direction, and you do find a difference which is smaller. I mean, what I'm saying is that the UK and France
53:22
are less connected on average than France and Germany, and of course, the US and South Africa are much less connected than France and Germany. So data is diagonal in that sense. It's quite obvious. What is he assuring? He doesn't contradict. We have time for two more questions.
53:46
Thank you, Jean, for a very nice paper. Just on the big picture of the message, however, I'm very much puzzled by your chart number eight in which you look at the bilateral correlation
54:02
and how your measure of convergence, the distribution has shifted to the right, indicating much more convergence in GDP growth, looking at the bilateral. Now, this is the period. You have two period. The second period where convergence has increased is 2008 and 2014, which goes against the intuition,
54:24
I think, of most people in this room that during the period of the crisis, there was a lot of heterogeneity. So I really wonder whether the cutoff year, which is 2008, is really the problem because in 2008, in the year of recession,
54:42
all countries moved together strongly because there is a big change in the mean, which is negative, but this distorts, incredibly, the picture. So are we saying really that we have performance have been more similar during 2008 and 2014 than during the normal times?
55:01
In this case, perhaps you can start. I mean, that's exactly the regression I showed. So when you do their exercise, I mean, and you account for the crisis. Crisis is a bigger effect than the export intensity in terms of increasing causation because everybody goes down. And then when you interact that crisis,
55:20
I mean, with each of the three measures, export intensity, trade intensity, then financial integration, you see that during the crisis, financial integration increases synchronization, but their variable export intensity actually diverges more, probably because trade goes down and that affects the value of change and all. Exactly, exactly, exactly, yeah.
55:41
Yeah, I just, first of all, tremendous empirical effort and really imaginative and interesting. But let me focus on the financial integration point. And there, I'm not convinced, as I think I said to you before the talk, Jean.
56:02
And I'm partly not convinced because of the measure itself. Much as I have respect for Shebnem and my colleague Elias Papayano, nevertheless, I'm not sure that, although the appeal of this measure is of course its apparent exogeneity, on the other hand, it doesn't match at all
56:23
with the data on financial integration. The ECB publishes a, every year, a piece on financial integration in Europe. And for several years, they've been publishing these graphs that are both one price-based and one quantity-based measure of financial integration.
56:42
Those measures show substantial increase in financial integration up to the crisis, and then a very precipitous drop that levels out in 2013, if I remember rightly, and then slightly starts to rise. But we're still not even at the level, at the pre-crisis level. So in a sense, the output measure, if you like,
57:03
of financial integration is telling you something very different from the input measure if you take the FSAP story as an input measure. And I'd be interested in your comments on that. I think this is a question for Sibinam more than it is for me.
57:21
Okay, can I just show you one thing? Can I have my slides back here, please, for a minute? Okay, I didn't have time to show this figure, but Richard is exactly right. So obviously, you have a de jure measure on the laws, and you have a de facto measure in terms of movement of cross-border assets and liabilities,
57:42
especially banking assets and liabilities. So let me show you this figure. This is from a forthcoming paper at IMF Economic Review, but it is the exact estimate of the authors just plotted. So they give you the aggregate estimation here. I do their same estimation with their data year by year and plot year by year.
58:00
So what do you see here? The red line is the consumption smoothing, right? And then the green line is the income smoothing, and the black line, the dotted line, is the unsummuted. So you see how unstable the consumption because of the effect of the shock and because of exactly what Richard just said, the effect of the movement in the cross-border assets and liabilities,
58:21
banking assets and liabilities. But look at, and that's the consumption smoothing is like if you look at the part before the crisis, like it's around like 60% of consumption is smoothed, and then it goes down drastically during the crisis, and the unsummuted goes up, the black line, and then consumption smoothing come back up again, then the unsummuted declines.
58:41
But the point here is look at the income smoothing and how stable it is, the green line. So that is telling you right before the crisis, EMU achieved around 20% risk sharing through the capital markets. Okay, this is ex ante smoothing. This is how much you smooth based on the cross-border asset ownership before the crisis, and it doesn't move that much with the crisis.
59:01
Part of that consumption decline there is also because of the fiscal authority, right? Everything is going to affect that measure, cross-border assets, banking assets and liabilities, fiscal authority and all that. On the other hand, the income one, how much the effect of GDP on the GNI of the country is smoothed, ex ante is very stable,
59:21
and that's around 20%, which is an amazing success for a union of 10 years when we are entering the crisis, and it goes down but not as much as the crisis. So I think this is also very important to understand because I don't think US achieved that amount of smoothing in the first 20 years of the capital market and banking union in the US. So that green number for comparison person
59:42
is 38% in the US. Thank you very much, Seth. And then I have a final question for both. What are your main conclusions and your main takeaways in terms of policy recommendations? Right, so I think what this...
01:00:00
paper says and this measure suggests is that observed trade is just potentially the tip of the iceberg, so there's much more to economic integration, deep integration, than just what is observed in what we've been using for a long time, just exports and imports between countries or between sectors. That measure is imperfectly measured.
01:00:22
It's got issues of whether it includes value added or other things. And it's also omitting quite a bit of action because it's virtually eliminating anything that is effectively non-traded. Service is the most prominent on that list. Just by way of illustration, in the data that we used in these 12 core EMU countries,
01:00:43
the correlation between services understood broadly on average between EMU countries is 0.6. It's quite a large correlation. And for most of these pairs of services, we simply do not have reliable trade data.
01:01:01
And so perhaps because we should have these data and we don't measure it, or perhaps because they effectively don't trade, but they trade indirectly. I think that measure has something to say about that. And I fully agree. I think this measure that really shows us deep integration, and it tells us the importance
01:01:23
of value chain and the importance of looking beyond what is actually traded. So I value a lot this measure. But going back to what we just talked about in Richard's comment, we can think the same thing for deep financial integration, right? So EMU has done a tremendous job in terms of integrating financial markets, but there
01:01:42
is more to do. And we can think that also in terms of how deeper we can go. Not just single currency, not just passing the laws, but how much more we can do in terms of the full-fledged banking union, full-fledged capital markets union. I mean, that's just going to add more to GDP and consumption convergence, which is
01:02:02
at the end, it's very important from a welfare point of view. Okay, thank you very much. With this very good recommendation, I think that we have reached the end of this. Thank you very much. Thank you.