Transcript

Transcript of an IMF Economic Forum: Contractionary Currency Crashes in Developing Countries --The Mundell-Fleming Lecture, IMF Annual Research Conference

November 5, 2004

MR. RAJAN: Good afternoon, everybody. Welcome to the Fifth Mundell-Fleming Lecture. Today we have Jeffrey Frankel, the James Harpel Professor of Capital Formation and Growth at the Kennedy School of Government, with us. Professor Frankel has a very distinguished record both in academia and in policymaking. He directs the program in international finance and macroeconomics at the NBER, where he is also on the Business Cycle Dating Committee. This is the one that declares recessions on and off.

He was nominated by President Clinton to be a member of the Council of Economic Advisers, and his responsibilities included international economics, macroeconomics, and the environment.

Before going to Harvard in 1999, he was professor of economics at the University of California at Berkeley, and he has had appointments in a number of public policy institutions like the Federal Reserve, the Institute for International Economics, and the International Monetary Fund. He has visited us a number of times before, and his research interests include international finance, monetary policy, regional blocs, and international environmental issues. I'm sure everybody in this room has read some piece of writing that Jeff has done at some point.

And, with that, can I hand it over to you?

MR. FRANKEL: Thank you for that introduction, Raghu, and thank you all for coming on a Friday afternoon.

Next month is the tenth anniversary of the Mexican peso crisis, so I thought I would take currency crashes as my topic. It's a great honor to be giving the Fifth Mundell-Fleming Lecture. It's an honor because of Mundell and Fleming and those who have given the lecture before and to participate in the Jacques Polak Annual Research Conference.

I'm going to talk about a number of different aspects of devaluations in developing countries. It's rather striking when you think back over the last 20, 30 years of research how much our interests have shifted away from industrialized countries towards developing countries. If you look at the working papers in international finance and macro in the NBER, a heavy majority now focus on emerging markets and other developing countries. Certainly if you look at the caseload of the IMF, there has been a huge shift in emphasis over the last two decades in that direction.

To some extent, we are applying to developing countries, models and other tools that originally were developed for industrialized countries like the Mundell-Fleming model itself, which 20 years ago you would not have thought of applying to developing countries because they weren't sufficiently integrated into the world financial markets to really talk about high capital mobility. But now I think that is one of the tools we apply, and there are many others as well, and then there's also the question of what is different about developing countries and emerging markets that requires that we make some modifications in our macro models to describe what's going on there, particularly after the big financial integration of the last 20 years, and the crises, of course, if you start from the Mexican peso crisis in December of 1994 and go up through the last big ones, Turkey and Argentina in 2001, the last decade has been quite a turbulent time.

There's actually three different papers I'm going to talk about today, and they correspond to three points. I'm going to start off by talking about political costs of currency crashes and devaluations; and then talk about contractionary effects, why have devaluations often been contractionary in recent episodes; and then talk about an aspect of what countries can do to reduce vulnerability to these crises. I hope to describe to you some recent research, but also try to fill in the connections between these different topics with a bit of literature review and common sense, and hopefully it will all hang together well enough.

I'm going to take as my starting point not Mundell and not Fleming but, rather, a paper by Richard Cooper, which is now 33 years old, called "Currency Devaluation in Developing Countries." It was a Princeton essay, has been widely cited over the years, and one of the reasons it's cited widely is the statistic that leaders lose their job 30 percent of the time within the year following a devaluation. This is a impressive statistic and one of the signs that people must consider it impressive is that it is still cited so widely 33 years later. He had a control group, which was 14 percent, so it works out to an approximate doubling of the rate at which a leader loses his job or her job in the year following a devaluation.

Well, I propose to update this calculation. First, I did a citation count. Since 1972, the Cooper paper has been cited 84 times, which is, nowhere near Mundell and Fleming, but is still pretty respectable, particularly since, as I say, there's no downward trend; people still cite it. I also thought that maybe if I updated this that people would cite me that often.

His criterion for what constitutes devaluation was 10 percent. Now, if we did that today, many countries all the time would qualify, so I think we've raised the bar on what we consider an interesting devaluation. So in the calculations that I'm going to present to you, the threshold is 25 percent and, in addition, certainly in the high-inflation '80s, there are some countries that would qualify for that every year, even every month. So it has to be not just 25 percent, but it has to be an acceleration relative to the rate of depreciation over the preceding year. And if you've had a currency crash within the last three years, I might window that out. It has to be a new one.

Before I give you the numbers, you can just look at the bar charts. I'm going to emphasize three calculations: the rate at which the leader—prime minister, president, whatever, I say premier here—loses his or her job over the subsequent six months, over the subsequent 12 months, and then the finance ministers and central bank governors. In each case, it is a substantial increase.

When I do it over 12 months, it's a substantial increase, but it's not quite statistically significant. Over 12 months, it goes from 20 percent to 29 percent. It's a 45-percent increase in the rate at which they lose their jobs, not quite statistically significant at conventional levels.

If I narrow it down to the six months after the devaluation, then this is very highly significant statistically; 23 percent of the time the leader loses office within six months of the devaluation, and in normal periods it's only 12 percent. So that's almost a doubling. And this number is significant at not just the 1-percent level but even the tenth-of-a-percent level, very—or at least, I guess, at the 0.5-percent level, very highly significant.

We did it also for finance ministers and governors. It turned out to be a little harder to get the data than I expected, but the way we did it is finance ministers and central bank governors are Governors of the IMF and come to the Annual Meetings. And so we used that as our measure. And here the rate of job loss is much greater. Turnover is greater in normal times, but also the effect is even more significant statistically. In normal times, a 35.8-percent probability of losing your job; within 12 months of a devaluation, 58.3 percent.

So why are the political costs so high? Why do leaders so often lose their jobs? The most obvious interpretation is there's high economic costs, the countries have recessions in particular, and I'm going to spend a lot of time, really most of my talk on that. But before we go directly to that assumption, let's think about other possible explanations that maybe the politics is important or maybe the political costs exist even independently of the economic costs.

So one possibility is that elections cause devaluations rather than the other way around. And you can think of plenty of reasons why this might be, and some of it came up in a session this morning about postponing devaluations until after an election.

It is striking in how many of the highly visible currency crashes of the last ten years how often the run-up, the beginning of the capital outflows has occurred during a presidential election year, but the final or worst crash and ultimate devaluation and the IMF program and all that have been after the election, after the new person comes in.

In the old days, we would have said, well, that's because of the political business cycle that governments follow very expansionary policies in elections years in order to get re-elected, and then they pay the price later on in terms of inflation or currency crash.

That does not seem to be what's going on as often anymore, I think maybe in the '70s, but less so now. And I don't know if Allan Drazen is still here, but he has an interesting paper that says it's really in countries that have recently made the transition to democracy, that's where you can fool the voters and gun the economy in the election year and then have the price in terms of devaluation and inflation come after the election. But the voters who have been through that a few times catch on, and so that doesn't happen as much anymore.

You could think of other reasons why devaluation would come after an election. It's politically costly to do, and there are lots of reasons to want to postpone it. There's some papers by Ernesto Stein and Jorge Streb looking at that. And for a new government—in a sense, it's almost optimal. If it's politically costly, the best time to do it is when the new government comes in and can blame it on the predecessor and get it out of the way right at the beginning. There was a paper of Sebastian Edwards some time ago that showed that a disproportionately high percentage of devaluations occur during a new administration.

So, anyway, that's all an interesting subject, but that can't be what's going on in this data because we've looked at the timing very carefully, and we're talking about leaders who lose office after the day of a devaluation, and if the devaluation comes a day after the change in leadership, we're not counting that. So that's not what's going on here.

The second possibility is that the devaluations are often part of a more comprehensive package of macroeconomic stabilization or structural reform, often under the auspices of the IMF. These programs tend not to win a lot of popularity points, and there are cases of slashing offood subsidies that leads to riots in the street and the government is overthrown. I mention in the paper the example of Nimeiry in Sudan who I think lost office in this way in 1985 or so. And maybe it's not the devaluation per se that's causing the problem. I cite a number of papers on the history of political instability and IMF programs.

Well, this one is easy enough for us to test. We controlled for whether or not the country had an IMF program, and I'm counting any IMF program here, plus or minus three months of the devaluation. The list of the IMF programs that I'm counting is an appendix to the paper. If anybody knows anything that should modify the way we use those, let me know.

Well, what we find is when you condition on whether there's an IMF program, it does not affect the basic results. I was actually surprised to see that just the effect of an IMF program itself does not actually raise the probability that the leader loses his or her job, which may be that it's just not as salient as we think, or it may be that in some cases, the old view that the IMF can perform as a scapegoat to do reforms that maybe the leader knows they have to do anyway, but if you have the IMF there, you can blame it on them, maybe that sometimes actually does work after all.

In any case, the IMF program itself didn't seem to make a difference on average, and so if you condition on the IMF program, then, the results for devaluation are the same. Among countries that did have an IMF program, the leader lost office 21 percent of the time—either way, with or without the IMF program. After a devaluation, they lost office 21 percent of the time, pretty much the same as in the full sample. And, again, that's almost double the rate in normal times.

So the second one was maybe the explanation is that it's a proxy for IMF programs or other painful reforms. That one doesn't seem to be the explanation.

The third one is that the government is perceived as having broken a promise. Often they have broken a promise. Sometimes they're clever enough to avoid going on TV or newspaper in the months before the devaluation and talking about the exchange rate, but often they can't get out of it, or they may really believe they're going to be able to maintain the peg, and they do make statements, commitments, promises of one sort or another. And possibly the voters are punishing them for abandoning this promise, even if it works out well is not damaging, and is not associated with a recession.

Well, to do that one, we took a sub-sample of countries, and I have a small team of research assistants speaking a number of different languages who went over to the Widener Library at Harvard and looked through the newspapers for the month preceding each of the devaluations that we could do and looked for signs of either the leader or the finance minister or the central bank governor saying we weren't going to devalue. And there's an appendix that describes those cases. Someone mentioned—was it Roberto Chang? Someone mentioned Lopez Portillo saying he was going to defend the peso like a dog. And we couldn't find that one, actually. I mean, we count that as a commitment because during the month before there were plenty of statements by him and by his ministers that they weren't going to devalue, so that's pretty clear. We could not actually find the quote about the dog, and so I don't know, maybe somebody knows the answer. There's two possibilities. One is he said it more than a month before the devaluation. That's quite possible. But the other thing is it's actually amazing how often things that you and I and everyone knows that a president said, they never actually said. Bill Clinton never said, "It's the economy, stupid," for example. I've got a whole list of these.

Anyway, we classified the cases of devaluations according to whether there was a promise beforehand not to devalue, and the sample size is much smaller here. What we find is in cases where they promised not to devalue, the executives lost jobs half the time within six months; whereas, when they did not make a promise, they lost the job only a fifth of the time within six months. And within 12 months, it rises to two-thirds of the time that they lose their job if they devalued and made the promise; whereas, if they devalued and hadn't said anything about it beforehand, it's one-third of the time.

Now, this is if anybody said something, if any of the three were quoted in the newspaper. You might ask, well, is the head of state more likely to lose the job, or the prime minister, I should say, if it was him specifically or her specifically who made the public statement? And the answer is we only have two of those in our sample, and in both cases they lost their jobs, so that's 100 percent. But the fact that that happens so rarely tells you something, that usually they send out the finance minister or the central bank governor to be the one to say, no, no, we'll never devalue.

So we're narrowing down the list here, and where I conclude is this explanation, the government is perceived as having broken a promise, well that does seem to be an effect. There is this broken promise effect. But even if you look at the ones that devalue without having made that promise, they still lose office statistically significantly more often than in normal times. So that can't be the complete explanation.

Even in those cases where no assurances had been given over the preceding month, the rate of job loss at 20 percent is still greater than in the cases where there was not a devaluation. Still, almost twice as great.

Actually, let me say it's easy to talk about breaking a promise or whatever, and someone asked this morning why do they make these statements given that so often they have to reverse themselves. My own personal favored view is not that they're being duplicitous, not that they expect to devalue, but that they really mean it and they're determined. In some cases, they may be underestimating the forces they're up against, how big the speculative attack could be, and not adequately thinking what's going to happen when they run out of reserves and they won't have a choice. But my preferred interpretation is that they know exactly what they're doing and they know there's a risk that they'll be forced to devalue, anyway. But they're really determined to do their best to hold the line, and this is a way of signaling. It's a device for buying some time, for buying some credibility, because you're saying I care about this so much that I'm willing to risk my job. We don't have that, in our economic models, the cost of losing a job, not in most models. But I think that that is sort of what's going on.

And so after the fact, if it's the minister who said it, they are likely to have to lose their job, but it doesn't necessarily mean that they were being short-sighted or duplicitous when they made the statement.

Anyway, although the broken promise effect is there, even controlling for that, there's still a statistically significant effect that they lose jobs. So I conclude from that that some of the reason for these political costs must be that it's economically costly, and that's what I will devote the remainder of my talk to.

Many, not all—Brazil in '99 we heard was one exception, but in many, certainly in the most prominent cases, the currency crash was followed by a pretty sharp recession. Why is that? After all, in the more standard models, devaluation is supposed to be expansionary. It makes you more competitive. You export more. There's the line about the British Chancellor of the Exchequer "singing in the bath." Now, it turns out that one he did say, except people often say that he said he was singing in the shower. And it wasn't the shower, it was the bath. But he did say that one.

So devaluation often works the way it should, and the British case and the Italian case also that year, '92, were examples of this, that it stimulates competitiveness and you export more and your growth rate goes up. But so often in emerging market countries that's not the case.

I'm switching here from developing countries to emerging markets. I did the breakdown of the job loss by income level, and it turns out to be mainly a phenomenon of the middle-income countries, not the rich countries at all, and not very much the poor countries. It's there. It's just not as strong. But it is very strong among middle-income countries so that's part of my justification for switching over now and talking more specifically emerging markets, middle-income countries.

Why did output fall in so many of these cases after the Mexico peso crisis in '94-95 and East Asia in '97-98 and the other cases that we know about? One view is—and I have Sachs and Stiglitz up here as the most visible proponents of this view—that the problem was excessive expenditure reduction through excessive fiscal contraction or excessive monetary contraction, raising interest rates too high. Stiglitz has a good argument here. We like to think in our models, let's say the Mundell-Fleming model, that if you raise interest rates, you attract a higher capital inflow. But if it raises the probability of default because the domestic banks and corporations can't meet these higher interest payments, then that has two effects. One is it may lead to defaults and domestic contraction; it's greater than it otherwise would be. But also it may not even succeed in attracting the additional capital inflow because the foreign investors obviously care not just about the face value of the interest rate but they also care about the probability of being repaid. So maybe when you raise interest rates, it doesn't attract further capital inflow, and this is pretty well grounded, I think, in things Stiglitz had done earlier about credit rationing and imperfect information. I think that's a perfectly valid argument.

Is that really what is going on here? I don't think it exactly is.

What about overoptimism on Fund targets, another theme that came up in a paper this morning. If you're overly optimistic about your estimated growth rates and that leads you to be overly optimistic about your forecast for the budget deficit, you may unwittingly be imposing a big fiscal contraction on a country without realizing it. And the Fund, for better or worse, did admit to sort of having erred in that direction in the early stages of the East Asia crisis.

Well, to my mind, this can't be the main thing that's going on because they basically all went into recession, and the question is what's the alternative. Our standard textbook theory says that, yes, you've got expenditure-reducing policies like raising interest rates or fiscal contraction. But you also have expenditure-switching policies, namely, devaluation. And if you have the right combination of the two, you should be able to meet your external balance constraint, the balance of payments constraint, and yet avoid a recession—in other words, continue to satisfy internal balance as well.

So here is a graph that I mean to represent the standard textbook model, and some of them I'm relating to taking from a little thing that Paul Krugman wrote called "Swansong," which is a play on the words "Swan diagram." But here it was in the context of Brazil in January '99. The vertical axis is the exchange rate. The price of foreign exchange is on the vertical axis. The interest rate is on the horizontal axis, so expenditure switching is to move up and expenditure reduction is to move to the right. So here's the exchange rate and here's the interest rate. And, of course, either a devaluation or a high interest rate can help you with external balance. So external balance could be defined as a trade balance, and devaluation improves the trade balance and a higher interest rate, expenditure reduction reduces spending, and some of that spending would have fallen on imports or other tradable goods. So they both improve the trade balance, and for a given trade balance, if you've got more of one, you can have less of another. There's a tradeoff between the interest rate and the exchange rate or a tradeoff between expenditure reduction down here and expenditure switching up here. That's why the external balance line slopes down.

The internal balance line, say output equal to potential output, slopes up because an increase in the interest rate is contractionary, but if you also devalue, that's expansionary because it stimulates foreign demand for goods, which makes up for the loss of domestic demand for goods. If you raise the interest rate, you lose domestic demand for goods. If you raise the exchange rate, you raise the foreign demand for goods. And the two offset each other, so you have the same level of output.

So that's the conventional story. Let's say you're East Asia, sitting here at this point in the diagram in 1997, and suddenly you have an adverse shock, or you're Mexico in 1994. So this point which previously was consistent with external balance, now suddenly you find yourself on the deficit side of where you want to be. The external balance line has shifted out. And for this purpose, it doesn't matter why it shifted out. It doesn't matter whether the government did something wrong, the government, or whether in the case of Mexico, it was the uprising in Chiapas or the assassination of Colosio, or was it excessively expansionary policies? Or maybe it's external affairs, the U.S. raised interest rates or change in world market conditions for your export commodities. It doesn't matter. For some reason, you find yourself now in the deficit side of this. We're still here, and we find we're now to the left of the new external balance line.

Well, the standard textbook story would say that you want to do some combination of both expenditure-switching and expenditure-reducing policy so you can get back on the external balance line and yet you can also stay on the internal balance line, and you can move up here.

Then how would we interpret the critique? I don't want to overinterpret them, put words in their mouth, but let's put Sachs and Stiglitz here to make it concrete. You could interpret them as saying the problem was that there was not enough devaluation, not enough expenditure switching. In the extreme, a fixed exchange rate, initially at least, we stay on the horizontal line. So we moved out here. Yes, we satisfied the new worsened external balance constraint, but only at the cost of a big recession because we're now well below this line. And so the right answer should have been to have less expenditure contraction and more devaluation.

Well, if that's the argument, I just don't think it can work because these countries had huge devaluations, Indonesia, you know, more than 50 percent, and the other countries as well, very large devaluations. So I don't think it quite works.

Perhaps what they're really saying is that we don't accept this new external finance constraint. We want to do something to shift this line back in, like convince international investors to put more money in or convince the IMF and the World Bank to put more money in or restore confidence or something and shift this line back down. So that's quite possible.

Basically, even admitting the Stiglitz point that a high interest rate may not attract a lot of increased capital doesn't change the basic logic, doesn't change the fact that the internal balance line slopes up. It maybe says you have to move your instruments even more. Maybe the line is flatter. But it doesn't change the basic logic.

To my mind, what you need to change the basic logic is a devaluation that's contractionary, that that is what changes the answer. If devaluation is contractionary instead of expansionary, then both lines shift down. You know, devaluing by moving up here doesn't improve demand for your goods. It has the opposite effect. And the external balance line slopes down. So then maybe we can understand what happened in the East Asia crisis as initially we were here, external balance line slope moved out, it's very difficult to say where these two lines cross. Or maybe they don't cross at all. Very difficult to say what combination of exchange rate and interest rate would satisfy the new external balance constraint without losing internal balance.

So if devaluations are contractionary, internal balance line slopes the other way, the new intersection is hard to find, and perhaps these countries at least if they wait until late in the game when they've run out of reserves, that at that point maybe no combination of expenditure reduction and expenditure switching will get you to where you want to go. They're going to have to have a recession.

My next topic is why is a devaluation contractionary. That article of Cooper's that I mentioned of 33 years ago listed six different possible contractionary effects of devaluation, and then there's more. I have a textbook with Caves and Jones where we have ten listed, which I hope is comprehensive. Some are more important than others, and I've listed four or five here.

The first four require rapid passthrough. They're different. Each one has to do with some passthrough of exchange rate change to some prices or some wages and not to others. The first one is imported inputs. Price of oil goes up. Oil is an input to production. That's an adverse supply shock. The second one is consumer inputs go up if there's real wage rigidity, particularly countries like some South American countries that had indexed wages in the '80s, then a devaluation raises the CPI, raises nominal wages. But if some non-traded goods haven't gone up, then that's an adverse supply shock in the non-traded goods sector. So you can get it that way.

The third one is that if traded goods prices generally go up, not just imports but exports as well, that that affects the distribution of income. And, of course, we could just stop there. Maybe even if it's not contractionary for GDP in the aggregate, maybe certain interest groups that lose out the most from a devaluation are the ones that are overthrowing the president, or if it's a non-democracy—if it's a president, voting him out of office, or an autocracy, overthrowing him. And you can think of cases—take an African coffee- or cocoa- exporting country where the price goes up and those people actually benefit, but the urban workers or elites lose from a devaluation. And if they have more political power, that's what happens.

So you can imagine that, but I can think of an awful lot of examples. I think more examples that happen to go the other way, where the sectors that are traded goods have more political power than non-traded goods. So I can't see getting really a generalization out of that, that that's why devaluation has these big political costs. So let's go back to talking about ways in which it can reduce all of GDP, not just some sectors.

Well, this one I have in mind, Carlos Diaz-Alejandro many years ago had an argument—he was thinking of Argentina—that the devaluation redistributes income towards the rural landowners because they're exporting the wheat and the beef, and it helps them but it hurts the urban workers; and that if urban workers have a higher marginal propensity to consume and the rural landowers are wealthy and have a low marginal propensity to consume, then out of that kind of redistribution you could get a decline in overall consumption, and so that could be contractionary.

And the fourth one is the real balance effect, that to the extent the general price level rises, that's a reduction in the real money supply, and that could be contractionary.

So all of these require some kind of passthrough. I'm going to spend a lot of time talking about the balance sheet effect. Someone else this morning said you can't talk about any of these without talking about the balance sheet effect, and I think that's right. It's the one that has gotten all the attention recently, and I think properly so. But before I turn to the balance sheet effect, let me talk about passthrough, and here I'm plugging some work I've been doing with Shang-Jin Wei and David Parsley on passthrough for developing countries. So it's a slight bit of detour, but I hope it will be worth it.

The bottom line of our paper is that the passthrough coefficient fell fairly dramatically for developing countries in the 1990s. It has long been a generalization that passthrough is slow or incomplete for industrialized countries, for large industrialized countries, and especially for the United States. But the usual view is for small developing countries, passthrough is very high. It's basically the starting point of the small open economy model, is that passthrough is very high.

Well, after the large devaluations in Asia and other emerging market countries over the last ten years, in each case most of us were afraid there would be a big passthrough, certainly to import prices and quite likely more generally to the general CPI, and that we'd be back reigniting the inflation of the '70s and '80s.

Well, it didn't happen, and this is one of the few pleasant surprises of these crises. The output situation was worse than was expected in most cases, but the inflation situation was better, that in most cases there was not the increase in prices that was expected.

This has been pretty widely remarked upon, but we're not aware of many empirical studies of that. Now, I need to make a few qualifications about the literature, because passthrough literature is huge. There have been some papers on a decline in the passthrough coefficient among industrialized countries. So John Taylor had a paper on that four or five years ago, and there have been a number of others since, extending it—his was mainly on the U.S., but extending it to other industrialized countries. And in many cases saying the reason is that the whole monetary environment has changed. In the '70s and '80s we had high inflation rates. In the long term, people were accustomed to that environment, and every time the exchange rate changed, they were pretty quick to pass it through; whereas, now we're in a low-inflation environment.

There's a paper by Campa and Goldberg, a recent paper that also finds that it declined a lot for 30 industrialized countries. They attribute it to composition effects, that it's not primarily because of the low inflation, falling inflation, but, rather, that we're talking about different sectors, different goods.

There are some papers that have examined this empirically and said that it looks like passthrough was much lower in these devaluation crises of the late '90s. But they virtually all work with CPIs. And I think it is very important to distinguish what kind of passthrough we're talking about. There's almost two different literatures. The passthrough literature that is motivated by industrialized countries is talking about passthrough specifically to prices of imports. But when you talk about passthrough in Latin America or in the context of these crises, usually people go straight to the CPI, which in part is a matter of data availability, but at least I think one should be clear about it, and they are quite different things.

What we did in this paper was get data for very narrowly defined products. We found eight products where you could identify the exact brand name and get data on prices in 70-something countries or 100-something cities, and they are products like, very narrowly defined, a Time magazine, a bottle of cognac with a brand name, a couple of them are actually alcoholic beverages, a carton of Marlboro cigarettes, a roll of Kodak film, very specific examples with prices in lots of countries. And part of the reason for doing this is to emphasize a difference between passthrough to import prices versus passthrough to the CPI, and many of the papers in the existing literature talk about composition effects. I mentioned Campa and Goldberg, and Michael Knetter before them for industrialized countries, but now for developing countries there's the hypothesis of Burstein, Eichenbaum, and Rebelo, which is that what looks like low passthrough is actually because of a change in composition, that even when you think you're looking at pretty narrowly defined goods, the composition changes, that people were importing luxury brands from abroad, and then after the devaluation they switched to a lower-quality bar of soap, for example. It looks like a bar of soap is a bar of soap, but, no, it's lower quality and that is showing up as the price not going up, but it's really because there was a shift in the product, a shift in composition. So this is one of our motivations for looking at very narrowly defined goods, so we can rule that out.

We also look at prices of local competitors, so the local brand of beer is the competitor for the alcoholic beverage, the local newspaper is the competitor for the Time magazine and so on. And we also look at the CPI to be comparable with the rest of the literature. But we mainly concentrate on this one, imported goods prices, and this is just an average passthrough for the sample period. I'll show you some regressions in a minute.

But, first, it's reassuring that what you would expect intuitively holds. The greatest passthrough is for the import goods at the dock. That's 80 percent. You get less passthrough for the identical goods at the retail level because there's costs of distribution and marketing and retail. That one's about 50 percent or just over 40 percent. And it's a little bit lower for the local competitors, and it's lower still for the CPI. So this is reassuring. It's what you'd expect. It also reminds us you want to be very explicit which kind of passthrough you're talking about, that one or that one, because they're quite different.

We do find that historically passthrough has been much higher for less developed countries than for rich countries. This is on average. Here are the regressions, and I'm violating my own rule here, which is you're not allowed to use a font below 20 points because nobody can see it unless you're in the front row. But I'm putting this slide up—and the fuller table is in the paper. I'm putting this slide up first to show you that we really ran some regressions. It's not just graphs. But also I want to give you the feel for this triangle, even though there's a lot of columns missing in this version of it.

What we first did was estimate passthrough just on average, unconditionally. This is for rich countries, 0.13, and then this is the increment for developing countries, and passthrough of the change in the exchange rate and passthrough of any changes in the price of the good in question in the foreign currency.

And then I put in the time trend, we put in the time trend, and then we conditioned on income per capita, and then we added more and more variables. These are variables that are said to be important in the literature on passthrough. We have tariff rates and we have distance. Well, why is that important? You can't have failure of the law of one price. You can't have incomplete passthrough in the first place if you don't have some kind of barrier to arbitrage. So passthrough models, broadly speaking, can be divided into two categories: those where the barrier is in international trade, tariffs or non-tariff barriers, or transportation costs that prevent perfect arbitrage between the foreign country and the dock; and then there are those models that say the main barrier is between the dock and the retail level, and that's the costs of distribution and retail. And, you know, probably they're both important, but we have tariffs and distance in here to get at the first factor. We have wages to get at the cost of retail. And we have size, we have long-term inflation rate, we have long-term exchange rate variability, we have a dummy variable for the U.S.

We can consider lots of different cases. There's too many in this all, so I'm just going to summarize some of the ones that are most salient for today's purposes.

Here's the time trend. When time was zero—at the beginning of the sample period, which is 1990, that's when this data set starts, the passthrough coefficient was 0.3 for rich countries and it was 0.3 plus 0.5, or 0.8, for developing countries. So the conventional wisdom is right that for developing countries it's closer to 1 and for rich countries it's only 0.3. But there are strong downward trends of 0.25 per year for rich countries and 0.5, the sum of the two, for developing countries. So others like Taylor have noticed this one, but here we have, I think, the first comprehensive estimates showing that the trend is twice as great for emerging market countries.

And then one by one we add in other variables and try to see can we explain what determines passthrough. Is it that small countries have more passthrough or poor countries have more passthrough and maybe what's changed in the '90s is poor countries are getting a little bit richer or small countries are getting a little bit bigger or maybe tariffs are changing or transport costs are changing or real wages are changing. I like that one because that could be a Balassa-Samuelson effect. Rich countries have higher wages; therefore, they have higher costs of distribution; and, therefore, they have lower passthrough. That could explain the difference.

So if you think of a poor country selling that carton of Marlboro cigarettes, it's probably going to be a retailer, a vendor out on the street that has no overhead, literally, and it's just the cost of his or her labor, which is pretty low; whereas, if the country gets richer, the vendor moves indoor and gets a roof over their head and gets higher wages and some overhead they have to pay, and then you could get more of a wedge and you get lower passthrough.

Anyway, the short story is we find that many of these determinants are statistically significant, if not for predicting passthrough to the retail import prices, then for some of the other I mentioned, like the CPI or the competitors. But they don't actually do very much to explain the downward trend. The real wage one might explain it, and inflation, lower inflation might explain it a little bit. It explains part of a downward trend. But even after conditioning for these things, we're still left with a pretty big downward trend.

In this estimate, you actually can explain it for the industrialized countries. It's the poor countries that it's still mostly unexplained for.

So here's just kind of a summary. At the start of the sample, passthrough for these import prices was 0.8. On average in the sample it's 0.5 — [tape ends].

— what you get if you stop halfway through in terms of the trend, if you stop halfway through.

So hopefully that tells us some things about passthrough, but let's now get back to pick up the overall thread of my lecture, which was why are devaluations contractionary. And I think that those arguments that depend on high passthrough, either to imported inputs or to consumer prices and, therefore, wages or the CPI, they're less valid than they used to be, and that's part of the reason perhaps why we're paying more attention to the balance sheet effect.

I guess the other reason we're paying more attention to the balance sheet effect is that debts are bigger than they used to be, and it is important.

So the balance sheet effect, of course, is that a bank or a corporation that undertakes debt in dollars or some other foreign currency and doesn't hedge but has its domestic revenues often in domestic currency, in Indonesian rupiahs or Korean won or Mexican pesos, it may be in perfectly good shape even when the economy starts to get into trouble. It may be capable of servicing its debt until you devalue by 50 percent and suddenly your debt service costs have doubled relative to domestic currency, relative to your earnings and your revenues, and so then you're in trouble. At a minimum, you maybe have to lay off some workers and cut back, or maybe you go bankrupt, and that is the balance sheet effect. And it is indeed everywhere. I list 15 papers that are more on a theoretical side. There's many more. I'm sure there must be another 15 people in this room who have written a paper who are not on this list that have a balance sheet effect in it.

And then on empirical evidence, I'm citing two papers here, one by Cavallo, Kisselev, Perri, and Roubini, and one by Guidotti, Sturzenneger, and Villar which show contractionary effects.

The empirical work is pretty difficult because, despite all the nice phrases we have about original sin and currency mismatch and the balance sheet effect, and despite all the very nice theoretical models we have, actually measuring these things is pretty tough. Often you have to use imperfect proxies to figure out exactly what the currency denomination of this debt is.

I'm not giving you any original research on this here, but instead I'm giving you a graph that I stole from this paper that I just mentioned, Cavallo, Kisselev, Perri, and Roubini, and this is a different Cavallo than the one I'm going to be mentioning in a minute. On the horizontal axis is the magnitude of the foreign-denominated debt times the magnitude of the depreciation that took place. And on the vertical axis is the loss of output. And that is, I think, a pretty striking relationship.

Now, the paper runs regressions and does all kinds of other things. I don't mean that that's all here. You want to do things more carefully. But I think this is a pretty striking graph. Down here this is Indonesia in '97, Bulgaria in '96, big devaluations, big dollar liabilities, big adverse balance sheet effect, big fall in output.

What determines the balance sheet effect? Well, it's a combination of a devaluation with a balance sheet that was already vulnerable, in other words, currency mismatch. So balance sheet effect is the trouble you get into when you devalue and you have currency mismatch. Currency mismatch is what happens when your local revenues are in local currency and your debt service is in foreign currency, and original sin is a possible source of currency mismatch, to try to be precise about that terminology.

Ricardo Hausmann came up with the phrase "original sin," and Eichengreen and Hausmann's paper was the one that popularized it and made it widespread. I think it properly refers to a structural inability to borrow in local currencies. So I would use it to say any time a country borrows a foreign currency. I think Hausmann means that to be something that a country is born with, that's in its fate, it can't borrow in domestic currency no matter how hard it tries because foreigners just don't want to lend to them in foreign currency. And if you try to hedge, that doesn't help because someone has to take the other side of the forward contract, and you still need ultimately to have a foreigner somewhere who's willing to take an open position in domestic currency.

So what are the origins of this? Well, historical fate is possible, and Ricardo was deliberate that he wanted to conjure up the fall in the Garden of Eden, that maybe it's the fault of one of our ancestors who was irresponsible, you know, back in the '30s, but we still have to live with this 60-70 years later. So it could be historical fate. It could be unjustified discrimination on the part of international investors that are too lazy to try to figure out what's actually going on and differentiate among different emerging market countries.

Some people argue that the culprit is adjustable pegs. That's come up earlier in this conference, that if you have a fixed exchange rate, or at least the illusion of a fixed exchange rate, banks and firms will not think they're taking a risk, they're not aware of the risk, and so they will expose themselves to dollar liabilities. And it's been offered as an argument in favor of floating rates, that the exchange rate is variable, that your debtors will be forced to confront this source of uncertainty, and, therefore, they won't incur the dollar liabilities in the first place or they will hedge them.

Now, I used to slightly make fun of this argument. It comes from Barry Eichengreen and some others. Alan Blinder actually will sign on to my caricature version. My caricature version is to say, well, your local borrowers are so myopic, so unaware, they underestimate this exchange risk so much that you want to create a lot of it, you want to maximize exchange rate volatility gratuitously in order to discourage them from the currency mismatch. But now my colleague Andres Velasco has a model in which it's completely rational. That even with people having rational perceptions of exchange rate volatility, the equilibrium in which a country commits to a floating rate and you get more exchange rate volatility and the private financial markets adjust to that and denominate in foreign currencies to the appropriate extent is a model that maximizes welfare. Anyway, I think the empirical evidence is—I read it as pretty weak so far that the exchange rate regime really will determine this.

Instead of those, having mentioned those, I am going to mention two other possible suggested influences on the balance sheet effect. So without denying any of the other things, maybe international financial markets do discriminate against poor countries. Maybe you can never live down what happened 60 years ago. Maybe original sin is right. But let's think of some respects in which countries maybe do have control over their own fate. Is there some way that they can minimize a currency mismatch and, therefore, minimize the balance sheet effect if the currency crash comes.

I'm going to talk about something in the short run, and then I'm going to talk about a longer-run structural aspect, and that's going to be the last of my topics. We'd like to leave at least a little time for questions.

First, what I meant by the short run and what I call the procrastination phase. In the literature on sudden stops or the literature on speculative attacks, you would get the impression—and in theoretical models it's usually the case—that it all happens in one instant, that foreign investors lose confidence and pull out, and that happens at the same time. You know, one moment everybody is happy to lend to you, and the next moment they pull out and you have the crash.

What seems to happen in reality is that that process takes some time, between six months and a year, if you measure it in terms of the loss of reserves. Here's a graph for Mexico. So in Mexico, January or February of 1994 is where the turnaround was. Up until then, money was flowing in, reserves were high, everything was great. I mentioned the uprising in Chiapas, the assassination of Colosio, the Fed increase in interest rates. Who knows why. It was just an embarrassment. In East Asia, we don't have enough proximate causes to explain exactly what changed in the summer of 1997. Nothing much really changed. In the case of Mexico in '94, too many things changed in early '94, so we don't know which of them bears responsibility. But for some reason or another, for one of those reasons, capital started flowing out again, and then there was a lag of almost a year before the crash actually happened.

Here's the crash by our criterion of the 25-percent fall of the exchange rate. Here's the date of the IMF program in January or February. And so this is the period I'm talking about. And I'm still happy to call it sudden stop. You know, sudden stop originally refers to jamming on the brakes in an automobile. I have a theorem that all important lessons about emerging market crises can be phrased in terms of automobile accidents—everything. You know, moral hazard, most hazards, dagger in the steering wheel, and there's whether you should have ambulances and speed bumps, and it's all—everything. Optimal sequence that had to do with whether you paved the roads, paved an off ramp from the highway going right into a primitive village. No, you don't want to do that, you want to develop the village first and teach people to walk on the sidewalks and put in some traffic lights before you dump the freeway traffic interchange through the village. Well, that's the optimal sequence of liberalization.

But that's true of the sudden stop, too. I remember my driver training class, and about half of it was on the subject of what happens during the three seconds between the time that you see the hazard in the road and the time that your car stops. You know, how quickly do you jam on the brakes? I still don't remember whether you steer into the skid or out of the skid. But what you do during that period of time is very important. And so I think that's true here as well, and I'm going to call this the procrastination phase. And Pedro Aspe's name came up before. Pedro Aspe was my classmate in graduate school at MIT in the '70s. I think he was a fantastic Finance Minister, and nothing I am going to say should mean a lack of respect for his abilities. But there are four ways to procrastinate that are in most common use, and it's sometimes called gambling for resurrection. Mexico is a good example.

I should say in the paper I recently did with Shang-Jin Wei for the World Bank, we've looked across a very large number of countries, and we found that this phase typically lasts six to 13 months. Again, it starts when reserves peak and begin to come down, and it ends when there's a crash defined by the 25-percent rule. And so it typically lasts between half a year and a year. And some of what I'm about to say for Mexico is also true for many other countries, but Mexico is a pretty clear example. And since we are at the tenth anniversary of the Mexican peso crisis, I am going to focus on that.

First, officials announced that they won't devalue. We saw that. The officials, the political leaders and the ministers, announced they won't devalue. I suggested that maybe they're not being short-sighted or stupid or duplicitous. They're buying some credibility with this. They're saying, "I am so determined to defend the peg that I'm willing to, in effect, put my job at risk for it."

A second way you can procrastinate is to run down reserves. That's a pretty obvious one. And Mexico did a lot of that. But it is interesting, about the middle six months of the year reserves pretty much stabilized.

The third way you can do it is to shift the composition of debt, change the maturity structure from long term to short term. And one of the papers that was presented here yesterday said there's a reason countries borrow short term. It's because it's cheaper for them to do that. They have to pay higher interest rates for long term. And that there is a shift in maturity in the course of the crisis.

And the fourth is a shift in the composition from local peso-denominated, in this case, to dollar-denominated.

So the point is the currency composition and the maturity composition gets worse during this period of sudden stop, aside from whether you were born with it, you went into it a year before the crisis, it gets worse during the course of the crisis.

Here it is for the currency of composition. Mexico's debt was primarily in pesos a few years before the crisis. For the foreign investors, it was the cetes that were the investment vehicle, and starting in 1993, but accelerating dramatically in 1994, the composition of Mexican debt shifted away from cetes towards tesobonos, the famous dollar-linked debt, until at—by the end of 1994, this measure—it depends what the denominator is. This measure is looking at tesobonos relative to tesobonos plus cetes. And by that measure we're up to 70 percent here at the date that the final crash came. So that's a very clear shift, and the reason is obvious. Foreign investors are nervous. They're reluctant. So you are accommodating them by shifting the currency composition to reassure them, and it is a sort of commitment device. It does succeed in buying time.

This is the other one. This is the shift in maturity, two different measures of maturity. One of them is the average maturity of the debt, and the other one is the ratio of long-term debt to total debt. And by both measures, the maturity of Mexican debt leading up to the crisis, here's November '94, here's the crisis right here, and over the preceding two years the maturity had been shifting from long-term to short-term.

What all four of these ways of gambling for resurrection have in common, all four of these ways of procrastinating, what they have in common is two things: First, they work. I think they do buy some credibility. They do buy some time. And in some cases, they may end up working. If something comes to your rescue like a fall in world interest rates, which I think is probably what came to Brazil's rescue at the end of 1998, and maybe is why the Brazilian devaluation in January '99 or February '99 did not cause a big recession the way many people had feared. So something might come to your rescue. But if it doesn't and you have the currency devaluation anyway, what all four of these have in common is you're in for more trouble than you would have been otherwise. You've run down your reserves, so you have trouble defending a peg which actually may be reasonable, and then that's what leads to overshooting when the currency goes even farther, which is what happened to Mexico. The first devaluation didn't work. And it's bad for confidence, of course, not to have the reserves. While your officials, they've just lost credibility, maybe are forced to resign, so you're in worse shape there. And my main point is if now your debt is heavily shifted towards short-term and floating rate debt, and the currency composition is heavily shifted towards the dollar, then you're in trouble because it doesn't matter whether you raise interest rates and don't devalue or devalue and don't raise interest rates or do both, they're all contractionary. You're very vulnerable to any of them, either the increase in exchange rate or the increase in interest rate. And that's the sense in which I think at that point there's really no alternative. You're pretty much going to have a recession no matter what.

Lesson. After inflows cease, adjust to the new external balance early rather than procrastinating while there's still time to exploit the conventional framework of expenditure-switching versus expenditure-reducing policies and you have some hope of maintaining internal balance. The alternative of procrastinating, it may turn out that you're a genius if something comes to your rescue, but it runs the risk that you're in a situation where it doesn't matter what the combination of devaluation and increase in interest rates is. You're going to have a contraction no matter what.

Well, I have one more thing to say, one more actually bit of research, another paper to summarize about something that countries perhaps can do in the long run to reduce vulnerability to sudden stops and contractionary currency crashes. And what I'm going to describe is a paper with Eduardo Cavallo, and it attempts to be a contribution to the literature on is globalization good or bad for these emerging market crises. More precisely, is openness to trade good or bad? And there's plenty of arguments and some empirical papers on both sides of this.

So let's define openness as openness to trade, most often in this literature measured simply as a ratio of trade to GDP, although you might want to do something else. And there are some people who say that openness raises vulnerability to sudden stops, so this is kind of the anti-globalization view—not that I'd put Milesi-Ferretti and Razin in that camp, but people who talk about vulnerability to real shocks, I think in 1997, with those crises, I don't know if it's gotten enough attention that the export markets of those countries had turned down pretty sharply over the preceding year, and that might have been part of the precipitating factor. Obviously, these were capital account crises, but possibly they are precipitated by adverse trade shocks. Or financial shocks, if the crisis has the effect that you lose all credit, including trade credit, and then you can't export, you can't trade anymore, that's a big cost if you're an open economy. It's not that big a cost if your ratio of trade to GDP was low to begin with. And maybe being open to trade always goes hand in hand with being open to financial flows, that trade happens via multinationals, you can't really do trade without letting the multinationals in, but you can't have the multinationals in if you don't let them move money back and forth.

These are various possible arguments that have been given as to why openness maybe increases vulnerability to sudden stops and to currency crashes. And there is some empirical support for that in these papers by Milesi-Ferretti and Razin, which look at current account reversals.

Then there's a lot of people who argue the other way around, who say that openness to trade actually reduces vulnerability to sudden stops and to currency crashes. Sachs had a paper in 1985 which asked the question: Why in the international debt crisis of the 1980s have the East Asian countries come through it pretty well while the Latin American countries did very badly? The ratio of debt to GDP was roughly similar. If they compare Korea to Mexico or Argentina, no big difference there. Why did the Asian countries do better? Jeff said it was because they have higher ratios of trade to GDP. And so, therefore, the necessary adjustment cost is less. If you've been receiving $10 billion a year in inflows and you suddenly lose that, so you have to go from a trade deficit of $10 billion to a trade deficit of zero, well, that is a very tough adjustment to make if trade is a small share of GDP. But if trade is a large share of GDP, it's no big deal. A relatively small expenditure reduction or a relatively small devaluation will do it for you. And that argument has been extended by others recently.

Another argument is that trade is a hostage that reassures creditors. So this goes back to the classic paper by Eaton and Gersovitz, but Andy Rose has some recent empirical evidence that's very interesting looking at bilateral trade, and he finds support for the idea that countries are very reluctant to default on debt with partners with whom they trade a lot because they're afraid they'll lose the trade credit and they'll lose the trade. And the partners know that, and so they have more confidence in—a major creditor country has more confidence in a debtor country if it trades a lot with them because it's a form of taking hostages. They know, it's obvious that the country has a lot to lose from a crisis, so they worry less that the country will default.

And here is the argument I made about the marginal tendency to import that's kind of a Keynesian term, but you could think of the share of spending that falls on tradable goods. You get the same answer, that if this is low, if openness is low, then it will take a really big adjustment to generate a given amount of export revenue that you'd need to service your debt. And so, therefore, open countries are better off. The adjustment that they have to make won't be as extreme, and the people know that ex ante.

A couple of empirical papers. I mentioned—well, here are two empirical papers. Edwards actually, I think, was the Mundell-Fleming lecturer last year, my predecessor by one year had some of that. And Calvo, Izquierdo, and Mejia. Many of these papers are motivated by the experience of Argentina. Argentina is kind of the classic case. The ratio of trade to GDP has historically been low, partly because, I would say, being a fan of the gravity model, it's partly because they're located far away from the rest of the world. But it's also partly because of obviously a history, a past history of protectionist trade policies.

So here are a number of papers that are motivated by Argentina that say that the interaction of having a low ratio of trade to GDP and a high dollar debt, that's toxic, that's what really dooms you.

I've listed former Treasury Secretary Paul O'Neill here because he said during the early phases of the travails in Argentina something like, well, of course, nobody wants to lend to them; they don't export anything. I should look—in line with my comment before, I should get the exact quote before I go around saying that. But I remember it very clearly.

So here is some empirical evidence. The vertical axis is simply openness measured as the ratio of trade to GDP. And the first bar is episodes where there was neither a sudden stop nor a currency crash. The second bar is episodes where there was a currency crash but not necessarily a sudden stop. And the third one is episodes where there was both, and just to present these crude facts, you can see that the open countries seem to be in better shape, and the countries with lower ratios of trade to GDP seem to be more prone.

Here's the same thing the other way around, this is per currency crashes. These are the open countries. They have fewer currency crashes. The countries with low ratios of trade to GDP have more currency crashes.

Well, this paper with Eduardo Cavallo looks at trade to GDP as a determining factor of two things: we do sudden stops and we do currency crashes. And the contribution is supposed to be that we take into account endogeneity of trade.

I'm running a little late on time, so I'll skip through all the possible reasons why trade might be endogenous—it's not hard to think of them, and I think you could imagine it—and go to the solution, which is to use the gravity instrument for trade. So to take Argentina as a case, the fact that it is located far from the rest of the world—or actually in my calculations, Fiji is the remotest country in the world, and so you would naturally expect Fiji to have a low ratio of trade to GDP. But you also want to take into account size and various other factors and whether it's in a trade union, a customs union with another country, and all the factors in the gravity model. And then you add them up across bilateral trading partners to get a gravity prediction of the ratio of trade to GDP, and that's the instrumental variable.

And we find that the effect is even stronger when we use instrumental variables. What is the effect? Well, this one first is a probit model of sudden stops, and the ordinary probit model shows that trade has a highly significant, statistically, effect, and it's negative, meaning more open countries are less prone to sudden stops, and the effect is even stronger when we use the gravity instrument.

I mentioned in addition to looking at sudden stops, we also looked at currency crashes. I thought of maybe having a slide, since these terms are now thrown around so much, having a slide—and maybe I can still put it in the paper, the IMF Staff Papers, if people would want me to, defining, you know, as a public service, using definitions that seem to me are in current use for current account reversals, sudden stops, speculative attack, currency crisis, and currency crash. Those are five different things. They tend to be related, tend to be correlated, but they're different things.

Without going into detail on the sudden stop, I used the definition of Calvo, and on the currency crash I used the definition that Shang-Jin Wei and I used in this other paper. And the same thing, trade openness is highly significant statistically or significant enough statistically, and stronger when we use the instrumental variable.

Two more slides. Conclusion about this. An increase in trade openness of 10 percentage points decreases the likelihood of a sudden stop by approximately 32 percent. How much is 10 percentage points? That's the difference between Argentina and Australia. If Argentina became as open as Australia, it reduces the probability of sudden stops 32 percent. What about currency crashes? Also statistically significant. And there's also some evidence that openness reduces the output costs associated with the currency crash.

All right. Last slide, takeaways. "Takeaway," this is sort of a Harvard Business School term, sorry about that. Three takeaways from this lecture, something you should get out of having sat here for an hour and a half or however long it was.

The first one, updating the Cooper statistic of 33 years ago. Leaders are twice as likely to lose office after a devaluation and, similarly, for finance ministers and central bank governors, especially if they promised not to, but even if they didn't.

Second, passthrough coefficient has declined sharply over the last 12 years. That could be important for lots of things, but in this context the point is that can't be where the contractionary effects of devaluation are coming from; rather, they're coming from the currency mismatch and the real balance effect.

What can countries do to avoid the contractionary effects of devaluation? Well, first, they can avoid weakening of their balance sheets during the procrastination phase by adjusting promptly rather than waiting, rather than gambling for resurrection. And, second, in the long term they can seek structurally to become more open economies.

Thank you.

[Applause.]

MR. RAJAN: Thank you very much. I was signaling to you because I wanted to leave some time for questions, and we have about eight minutes. So let's take questions from the floor, maybe take three and then you can answer them.

QUESTION: I wonder if I could ask you for a bit of clarification of the title. It refers to contractionary currency crises and—

MR. FRANKEL: A clarification of what?

QUESTION: Of the title of the paper where you refer to contractionary currency crises. I take it you do mean demand contractionary. The reason I ask you is because there is a famous model of crisis, which you referred to, by Aghion, and they had low passthrough and they had balance sheet effects, and they had contractions, but it's only on the supply side. It comes one period later. And the reason is it's a small open economy, and the assumption is that any fall in domestic demand can be substituted by exporting. I take it you do not mean that, you do mean—

MR. FRANKEL: No, by contraction, I mean fall in output. It can come on the supply side. So, for example, the classic argument that import—the price of oil goes up, that's on the supply side. I mean just fall in output. It doesn't have to be demands.

MR. RAJAN: Okay, back there?

QUESTION: [inaudible, off microphone] trigger, and similarly Korea, Malaysia, Indonesia. All of them had [inaudible]. And the other country, of course, is South Africa which always [inaudible].

MR. FRANKEL: Am I supposed to gather them together?

MR. RAJAN: Let's gather some. I think the last one was an aberration. Right there?

QUESTION: Thank you. I just have two questions. One is that as far as the contractionary effect is concerned, do you think there is any difference between the small devaluation and the larger devaluation, for example, small devaluation less than 10 percent, larger devaluation more than 30 percent? That is one question.

The second is: Is there any—if I understand clearly, there is a definite relationship between balance sheet effect and sudden stop, which means that there's a strong relationship between the contractionary effect of devaluation and sudden stop. Is that true? If the devaluation is not followed by a sudden stop, what will be?

Thank you. I'm from IIE, visiting fellow. Thank you.

MR. RAJAN: One more question.

QUESTION: [inaudible, off microphone].

MR. FRANKEL: Okay. Let's see. Why don't I take those in order? And from the lights, I can't see people, but I think that was Arvind who was asking the first question. The question was why am I so skeptical that the exchange rate regime determines the currency mismatch. You know, I don't mean to completely dismiss that, but partly it's because exchange rate regimes are just sort of a messier thing in reality than we often discuss. The countries you named, I wouldn't describe them as having a long history of fixed rates. The one that comes closest is Thailand, which had been fixed to the dollar for about a year or two. Before that, it had a basket peg with some variability, and the others all had some variability.

You know, it's become very well known in recent years that if you try to estimate de facto exchange rate regimes, what are countries actually doing based on looking at fluctuations in reserves or fluctuations in exchange rates, you get very different answers than what is called de jure the official IMF classification. It's less often recognized that these different competing classifications schemes, their correlation with each other is just as low, the different de facto regimes, is just as low as their correlation with the IMF de jure.

Why is that? The truth is that exchange rate regimes are messy. You can pick the 20 or so countries that have very rigid pegs, fully dollarized, EMU, the currency boards, that's fairly clear. Fairly clear. You know, the currency board people always tell you it wasn't really a currency board, but—

[Laughter.]

MR. FRANKEL: You know what I mean. But—Argentina. But that's fairly clear. And on the floaters, you know, maybe we could agree who are the free floaters. But almost everybody is in between, and when you try to categorize in between, it's pretty tough. So those countries, Thailand, Korea, Mexico, they were on intermediate regimes. And when you try to push them into the—say it was basically a peg that was the problem, it's not that clear. These are countries that had experience with fluctuations in the exchange rate. The idea that someone from Mexico or Thailand had never seen the exchange rate move before, move a long way before, so it never occurred to them they had to hedge, I just don't—I don't actually buy that.

The second question, I think the gentleman visiting from IIE about the connection between sudden stops and contractionary effects, the correlation, in our results they go together. They seem to be—for example, openness reduces vulnerability to both. You can think of theoretical reasons why they might go together, and my co-author Eduardo Cavallo has a theoretical paper in which they go together. But knowing that in the event of a currency devaluation things are going to be really bad undermines the confidence of investors and makes it more likely to happen. But there's just as good theories that go the other way. One is Mike Dooley, that the knowledge that a currency devaluation is going to be devastating for the country and it's going to cause a recession, that's precisely what reassures foreign creditors that you're not going to devalue needlessly, and so that boosts confidence. So, a priori, you could get either relationship, but at least with respect to this exact question, we get that sudden stops go the same direction as currency crashes.

And Yusuke on trade openness versus financial openness, yes, I mean, the literature on does financial openness raise or lower vulnerability to currency crises, that's a very large literature. I think it's hard to get clear results. Shang-Jin and I sort of tried to in this paper, and we didn't really. There's plenty of good papers that come down on both sides. This paper is not about that. This paper is supposed to be about trade openness, which is more of a finite literature and more of a finite subject.

MR. RAJAN: Well, thank you very much. That was an excellent presentation.

MR. FRANKEL: Thank you.

[Applause.]