Global current account imbalances and exchange rate adjustments.

Obstfeld, Maurice ; Rogoff, Kenneth S.

THIS IS THE third in a series of papers we have written over the past five years about the growing U.S. current account deficit and the potentially sharp exchange rate movements any future adjustment toward current account balance might imply. (1) The problem has hardly gone away in those five years. Indeed, the U.S. current account deficit today is running at around 6 percent of GDP, an all-time record. Incredibly, the U.S. deficit now soaks up about 75 percent of the combined current account surpluses of Germany, Japan, China, and all the world's other surplus countries. (2) To balance its current account simply through higher exports, the United States would have to increase export revenue by a staggering 58 percent over 2004 levels. And, as we argue in this paper, the speed at which the U.S. current account ultimately returns toward balance, the triggers that drive that adjustment, and the way in which the burden of adjustment is allocated across Europe and Asia all have enormous implications for global exchange rates. Each scenario for returning to balance poses, in turn, its own risks to financial markets and to general economic stability.

Our assessment is that the risks of collateral damage--beyond the risks to exchange rate stability--have grown substantially over the five years since our first research paper on the topic, partly because the U.S. current account deficit itself has grown, but mainly because of a mix of other factors. These include, not least, the stunningly low U.S. personal saving rate (which, driven by unsustainable rates of housing appreciation and record low interest rates, fell to 1 percent of disposable personal income in 2004). But additional major risks are posed by the sharp deterioration in the U.S. federal government's fiscal trajectory since 2000, rising energy prices, and the fact that the United States has become increasingly dependent on Asian central banks and politically unstable oil producers to finance its deficits. To these vulnerabilities must be added Europe's conspicuously inflexible economy, Japan's continuing dependence on export-driven growth, the susceptibility of emerging markets to any kind of global financial volatility, and the fact that, increasingly, the counterparties in international asset transactions are insurance companies, hedge funds, and other relatively unregulated nonbank financial entities. Perhaps above all, geopolitical risks and the threat of international terror have risen markedly since September 2001, confronting the United States with open-ended long-term costs for financing wars and homeland security.

True, if some shock (such as a rise in foreign demand for U.S. exports) were to close up these global imbalances quickly without exposing any concomitant weaknesses, the damage might well be contained to exchange rates and to the collapse of a few large banks and financial firms--along with, perhaps, mild recession in Europe and Japan. But, given the broader risks, it seems prudent to try to find policies that will gradually reduce global imbalances now rather than later. Such policies would include finding ways to reverse the decline in U.S. saving, particularly by developing a more credible strategy to eliminate the structural federal budget deficit and to tackle the country's actuarially insolvent old-age pension and medical benefit programs. More rapid productivity growth in the rest of the world would be particularly helpful in achieving a benign adjustment, but only, as the model we develop in this paper illustrates, if that growth is concentrated in nontraded (domestically produced and consumed) goods rather than the export sector, where such productivity growth could actually widen the U.S. trade deficit.

It is also essential that Asia, which now accounts for more than one-third of global output on a purchasing power parity basis, take responsibility for bearing its share of the burden of adjustment. Otherwise, if demand shifts caused the U.S. current account deficit to close even by half (from 6 percent to 3 percent of GDP), while Asian currencies remain fixed against the dollar, we find that European currencies would have to depreciate by roughly 29 percent. Not only would Europe potentially suffer a severe decline in export demand in that scenario; it would also incur huge losses on its net foreign asset position: Europe would lose about $1 trillion if the U.S. current account deficit were halved, and twice that sum if it went to zero.

We do not regard our perspective as particularly alarmist. Nouriel Roubini and Brad Setser make the case that the situation is far grimmer than we suggest, with global interest rates set to skyrocket as the dollar loses its status as the premier reserve currency? Olivier Blanchard, Francesco Giavazzi, and Filipa Sa present an elegant and thoughtful analysis suggesting that prospective dollar exchange rate changes are even larger than those implied by our model. (4) William Cline argues that an unsustainable U.S. fiscal policy has substantially elevated the risk of an adverse scenario. (5) In our view, any sober policymaker or financial market analyst ought to regard the U.S. current account deficit as a sword of Damocles hanging over the global economy.

Others, however, hold more Panglossian views. One leading benevolent interpretation, variously called the "Bretton Woods II" model or the "Deutsche Bank" view, focuses on China; that view is forcefully exposited in this volume by Michael Dooley and Peter Garber. This theory explains the large U.S. current account deficit as a consequence of the central problem now facing the Chinese authorities: how to maintain rapid economic growth so as to soak up surplus labor from the countryside. For China, a dollar peg (or near peg) helps preserve the international competitiveness of exports while attracting foreign direct investment and avoiding stress on the country's fragile banking system. Is this argument plausible? Set aside the fact that China maintained its peg even through the Asian financial crisis of 1997-98 and as the dollar soared at the end of the 1990s (presumably making Chinese exports much less competitive), or that China risks a classic exchange rate crisis if its fortunes ever turn, say, because of political upheaval in the transition to a more democratic system. The real weakness in the Bretton Woods II theory is that the Chinese economy is still less than half the size of Japan's, and less than three-quarters the size of Germany's, at market exchange rates. So, while running surpluses of similar size to China's relative to their GDP, Germany and Japan actually account for a much larger share of global surpluses in absolute terms. (After all, Germany, not China, is the world's leading exporter.) And surplus labor is hardly the problem in these aging countries.

U.S. Federal Reserve Chairman Alan Greenspan, in a 2003 speech at the Cato Institute and in many subsequent speeches, offers an intriguing argument. (6) He agrees that the United States is unlikely to be able to continue borrowing such massive amounts relative to its income indefinitely, and he recognizes that the U.S. current account deficit will therefore narrow substantially someday. Greenspan argues, however, that increasing global financial integration is both what allows the United States to run such large deficits and the saving factor that will greatly cushion the process of unwinding those deficits.

We completely agree that increasing global financial integration can explain larger current account deficits, particularly to the extent that greater trade integration helps underpin financial integration, as in our original analysis. (7) Indeed, this was a major point of our first approaches to this problem. A narrowing of the U.S. current account deficit must ultimately be the result, however, of more balanced trade, because the trade account is overwhelmingly the main component of the current account. And, as seemingly open as the U.S. economy is to financial flows, international product markets remain quite imperfectly integrated.

Thus any correction to the trade balance is likely to entail a very large change in the real effective dollar exchange rate: our baseline figure, which assumes a moderate speed of adjustment and that the world's major regions all return to current account balance, is 33 percent. A much smaller dollar devaluation is possible only if the adjustment is stretched over a very long period (say, a decade), in which case labor and capital mobility across sectors and economies can significantly reduce the need for relative price changes. On the other hand, should adjustment take place very abruptly (say, because of a sudden collapse in U.S. housing prices leading to an increase in saving, or a dramatic reallocation of global central bank reserves toward the euro), the potential fall in the dollar is much larger than our baseline estimate of 33 percent, primarily because sticky nominal prices and incomplete exchange rate pass-through hamper adjustment.

True, in a recent Federal Reserve study, Hilary Croke, Steven Kamin, and Sylvain Leduc argue that sustained current account imbalances in industrial countries have typically terminated in a relatively benign fashion. (8) But their threshold for a current account "reversal"--the country must have run a deficit of at least 2 percent of GDP for three years, and must have improved its current account balance by at least 2 percent of GDP and a third of the total deficit--is a very low bar compared with where the United States stands today. (Croke, Kamin, and Leduc are forced to choose a low threshold, of course, because current account deficits of the size, relative to GDP, of the recent U.S. deficits, although far from unprecedented, are not the norm.) Most important, the United States accounts for over 75 percent of global deficits today, as we have noted, and so any comparison based on the experience of small countries, even small industrial countries, is of limited value.

In addition to Chairman Greenspan, a number of academic researchers have emphasized how some important changes in the global financial system, particularly over the past ten years, have changed the nature of international financial adjustment. Philip Lane and Gian Maria Milesi-Ferretti, in a series of papers, have documented the explosion of gross asset flows. (9)

They and Cedric Tille, as well as Pierre-Olivier Gourinchas and Helene Rey, have shown that asset revaluation effects from dollar depreciation can have a significant impact on U.S. net financial obligations to foreigners. (10) Gourinchas and Rey point out, in fact, that the historical extent of such revaluations suggests that the United States might need to adjust its trade balance by only two-thirds of the amount that would be needed to fully repay its net external debt; even this, however, would still imply very large dollar movements. We agree that the size and composition of gross asset positions are increasingly important, and our model simulations in this paper explicitly take account of the revaluation channel. We find, however, that valuation effects mute the requisite exchange rate changes only modestly.

The growing financial globalization that these authors and Chairman Greenspan emphasize is, moreover, a two-edged sword. Enhanced global financial integration may well facilitate gradual current account and exchange rate adjustment, but it might also promote the development of large, unbalanced financial positions that leave the world economy vulnerable to financial meltdown in the face of sharp exchange rate swings. The net foreign asset revaluation channel might help modestly, but a rise in U.S. interest rates could well wipe out the benefits. Because the United States borrows heavily in the form of low-risk bonds, while lending heavily in the form of equities and high-risk bonds, it is especially sensitive to even a modest rise in the interest rates it pays on its foreign debt. Indeed, we show that, in terms of exchange rate adjustments, the adverse effect of a 1.25-percentage-point rise in the interest rate that the United States pays on its short-term foreign debt is similar in magnitude to the benefits gained via the valuation channel, even with a 20 percent dollar depreciation. More generally, although increased global financial integration and leverage can indeed help countries diversify risk, they also expose the system to other vulnerabilities--such as counterparty risk--on a much larger scale than ever before. All in all, although we believe that growing financial globalization is largely a positive development, it does not justify excessive confidence in a benign adjustment process.

This paper begins by trying to put the recent U.S. experience with current account imbalances in historical perspective. We hope this first section will provide a useful reference, although some readers will already be familiar with the essential elements. One historical observation that is important for our later analysis is that the United States (so far) has had the remarkable ability to consistently pay a lower rate of interest on its liabilities than it earns on its assets. Some component of this differential in returns has been due to luck, another to huge central bank holdings of U.S. Treasury bills, another perhaps to the unique and central role of the dollar in international finance. Still another, which we have already emphasized, is the fact that Americans hold a much larger share of their foreign assets in equities and high-risk (equity-like) bonds than foreigners hold of U.S. assets (and thus benefit more from the equity premium). An open question is whether this advantage can continue in the face of large and persistent U.S. deficits.

We then provide a nontechnical summary of our core three-region (Asia, Europe, and the United States) model. Readers interested in the technical details of our model can read the theoretical section that follows, and the most adventurous can venture into appendix A, where we fully lay out the structure. Our model simulations calibrate the requisite dollar decline against European and Asian currencies under various scenarios. Most of our analysis focuses on real exchange rates, but, by assuming that the regions' central banks target GDP or consumption deflators (or sometimes, in the case of Asia, exchange rates against the dollar), we are able to extract nominal exchange rate predictions (relative to the initial position) as well.

As noted earlier, our baseline simulation, in which Asia's, Europe's, and the United States' current accounts all go to zero, implies that the dollar needs to depreciate in real effective terms by 33 percent (and in nominal terms by a similar amount). Because the trade balance responds to an exchange rate change only with a lag, this exercise slightly overstates the necessary depreciation relative to today's exchange rates. However, our calibration assumes flexible prices and does not allow for possible exchange rate overshooting, which could significantly amplify the effect. A halving of the U.S. deficit, with counterpart surplus reductions shared by Asia and Europe in the same proportions as in the first simulation (arguably a more likely scenario over the short term) of complete current account adjustment, would lead to a depreciation of the real effective dollar of 17 percent. In our base case the real value of Asian currencies would need to rise by 35 percent and that of European currencies by 29 percent against the dollar.

If, however, Asia sticks to its dollar exchange rate peg as the U.S. current account deficit narrows, the real effective value of the European currencies would have to rise by almost 60 percent. Indeed, to maintain its dollar peg in the face of global demand shifts that fully restore U.S. current account balance, Asia would actually have to better than double its already massive current account surplus. Even halving these numbers (corresponding, for example, to the case in which the U.S. current account deficit falls only by half), one can still appreciate the enormous protectionist pressures that are likely to emerge if Asia tries to stick to its dollar peg in the face of a significant pullback in the United States' voracious borrowing.

It is perhaps surprising that, despite Asia's current account surplus being several times that of Europe (which we define broadly here to include the euro zone and the other largest non-Asian, non-U.S. economies), the required rise in the Asian currencies relative to the European currencies is not even larger in the global rebalancing scenario. As we shall see, a couple of factors drive this result: one is that Asia's economies are relatively more open than Europe's to the rest of the world, so that a given exchange rate change has a bigger impact on trade; the other is that a large, unanticipated dollar depreciation inflicts brutal damage on Asia's net foreign asset position, a factor we explicitly incorporate in our calibrations.

The analysis highlights two important but widely misunderstood points about the mechanism of U.S. current account deficit reduction. First, real dollar depreciation is not a substitute for policies that raise U.S. saving, such as reductions in the federal fiscal deficit. Instead, depreciation and saving increases are complements: exchange rate changes are needed to balance goods markets after a change in global consumption patterns, whereas dollar depreciation that is not accompanied by U.S. expenditure reduction will lead to inflationary pressures that, over time, will offset the initial gains in U.S. competitiveness. The second, and related, point is that it makes little sense to ask how much dollar depreciation is needed to reduce the current account deficit by 1 percent of GDP. Exchange rates and current account balances are jointly determined endogenous variables. As the simulations in this paper illustrate, there are numerous different scenarios in which the U.S. external deficit might be erased, all with different implications for the dollar's foreign exchange value.

Although our model is considerably richer than those previously advanced in the literature (including our own earlier studies), it remains subject to a wide range of qualifications and interpretations; we try to emphasize the most important ones. Nevertheless, we view the simulations as quite useful. The paper's final section highlights the main conclusions that we draw from the technical analysis.

The U.S. Current Account and Foreign Wealth Position, 1970-2005 and Beyond

The main analytical contribution of the paper is its modeling and numerical calibration of exchange rate and net foreign asset valuation adjustments under alternative scenarios for reducing the U.S. current account deficit. Our framework is intended as a tool for assessing risks and evaluating policy options. At some level, however, the exercise must entail an assessment of how unstable the current trajectory of external payments imbalances really is, along with the likelihood of adjustment taking place in the next few years. In order to think about this overarching issue, it is helpful to understand the history of the problem.

Perspectives on the U.S. Deficit

Figure 1 traces the U.S. current account balance as a percent of GDP from 1970 to the present. After fluctuating between +1 and -1 percent of GDP during the 1970s, the current account began to go into deep deficit during the mid-1980s, reaching 3.4 percent of GDP in 1987. After recovering temporarily at the end of the 1980s and actually attaining a slight surplus in 1991 (propped up by a large, one-time transfer from foreign governments to help pay for the Gulf War), the U.S. current account balance began a slow, steady deterioration throughout the 1990s, which continues today. As already noted, U.S. international borrowing in 2004 accounted for about 75 percent of the excess of national saving over investment of all the world's current account surplus countries.

[FIGURE 1 OMITTED]

What are the proximate causes of this profound deterioration in the U.S. external balance? That, of course, is the $666 billion (and rising) question. Since, in principle, the current account balances of all countries should add up to zero, the U.S. current account deficit--equal to the excess of U.S. investment over national saving--has to be viewed as the net result of the collective investment and saving decisions of the entire world. German demographics, OPEC oil revenue investment decisions, depressed investment in Asia--all these factors and many others impinge on global interest rates and exchange rates and, in turn, on U.S. investment and saving. We do not believe there is any simple answer.

Nevertheless, U.S. fiscal policy clearly has played a dominant role in some episodes. The current account balance equals, by definition, the sum of government saving less investment plus private saving less investment. Because the Ricardian equivalence of public debt and taxes does not seem to hold in practice, the big Reagan tax cuts of the 1980s almost certainly played a role in the U.S. current account deficits of that era. Similarly, the Bush II tax cuts of the 2000s have likely played a role over the past few years, preventing the current account deficit from shrinking despite the post-2000 collapse in U.S. investment. Currency over- and undervaluations also loomed large in both episodes, usually operating with a lag of one to two years. For example, the peak of the U.S. current account deficit in 1987 lagged by two years the peak of the real trade-weighted dollar exchange rate (figure 2). The weak dollar of the mid-1990s was matched by a pause in the U.S. current account's decline, and the dollar peak in early 2002 was followed again, with some lag, by a sharp worsening in the external balance. Admittedly, both correlations with the current account deficit--of fiscal deficits and dollar appreciation--are fairly loose. As figure 3 illustrates, U.S. fiscal deficits have expanded massively in recent years compared with those of the rest of the world. But, as the figure also illustrates, Japan has run even larger fiscal deficits relative to its GDP than the United States, yet at the same time it has consistently run the world's largest current account surplus in absolute terms.

[FIGURES 2-3 OMITTED]

Indeed, during the 1990s the major proximate drivers of the U.S. current account balance were a declining rate of private saving and rising rate of investment. The U.S. personal saving rate, which had been stable at around 10 percent of disposable personal income until 1985, has steadily declined since, reaching a mere 1 percent in 2004. The declining private saving rate has apparently been driven first by the stock price boom of the 1990s and then by the still-ongoing housing price boom. (11) Were the U.S. personal saving rate simply to rise to 5 percent of disposable personal income, or halfway toward its level of two decades ago, more than half of today's current account deficit could be eliminated.

During the late 1990s U.S. investment was robust, as shown in figure 4, so that the United States' high external borrowing really was, in principle, financing future growth. Today, however, the picture has changed. As figure 4 also shows, the main proximate driver of recent U.S. current account deficits has been low private and government saving rather than high investment. So much for the prominent view of former Treasury secretary Paul O'Neill, who argued that the U.S. external deficit was driven mainly by foreigners' desire to invest in productive U.S. assets. The more sophisticated analysis of Jaume Ventura is also inconsistent with declining U.S. investment. (12)

[FIGURE 4 OMITTED]

Another important factor contributing to the U.S. current account deficit since the late 1990s has been the persistently low level of investment in Asia since the region's 1997-98 financial crisis. Indeed, today, sluggish investment demand outside the United States, particularly in Europe and Japan but also in many emerging markets, is a major factor holding global interest rates down. Low global interest rates, in turn, are a major driver in home price appreciation, which, particularly in the United States with its deep, liquid home-equity loan markets, contributes to high consumption.

International Assets, Liabilities, and Returns

Naturally, this sustained string of current account deficits has led to a deterioration in the United States' net foreign asset position, as illustrated in figure 5. In 1982 the United States held net foreign assets equal to just over 7 percent of GDP, whereas now the country has a net foreign debt amounting to about 25 percent of GDP. Accompanying this growth in net debt has been a stunning increase in gross international asset and liability positions, as figure 5 also shows. From 29.5 percent and 22.3 percent of GDP in 1982, U.S. gross foreign assets and liabilities, respectively, had risen to 71.5 percent and 95.6 percent of GDP by the end of 2003. This process of increasing international leverage--borrowing abroad in order to invest abroad--characterizes other industrial country portfolios and is in fact much further advanced for some smaller countries such as the Netherlands and primary financial hubs such as the United Kingdom; see table 1 for some illustrative comparative data. (13)

[FIGURE 5 OMITTED]

The implications of the reduction in U.S. net foreign wealth would be darker but for the fact that the United States has long enjoyed much better investment performance on its foreign assets than have foreign residents on their U.S. assets. This rate-of-return advantage, coupled with the expansion in foreign leverage documented in figure 5, has so far allowed the United States to maintain a generally positive balance of net international investment income even as its net international investment position has become increasingly negative. Figure 6 shows two measures of U.S. net international investment income. (14) The first, net foreign investment income (income receipts on U.S. assets owned abroad less income payments on foreign-owned assets in the United States), is taken from the U.S. balance of payments accounts and comprises transactions data only, that is, actual income earned on assets. Interestingly, this balance has not yet entered negative territory, although it could do so soon. Over 1983-2003 the income return on U.S.-owned assets exceeded that on U.S. liabilities by 1.2 percentage points a year on average.

[FIGURE 6 OMITTED]

A more comprehensive investment income measure adds the capital gains on foreign assets and liabilities, reflecting price changes that could be due to either asset price movements (such as stock price changes) or exchange rate changes. The Bureau of Economic Analysis (BEA) incorporates estimates of these gains into its updates of the U.S. international investment position, although they do not appear in the international transactions or national income accounts. As one would expect, figure 6 shows this net income measure to be much more volatile than that based on investment income alone. Although it is negative in some years, cumulatively this balance is even more favorable for the United States than the smoother transactions measure. On average over 1983-2003, the total return on the United States' foreign investment, inclusive of capital gains, exceeded that on U.S. liabilities to foreigners by a remarkable 3.1 percentage points a year. (15)

To understand better the implications of the U.S. rate-of-return advantage, let [r.sup.W] be the rate of return on foreign assets, [r.sup.U] the rate of return on liabilities, F the stock of net foreign assets, and L gross liabilities. Then the net total return on the international portfolio is [r.sup.W]F + ([r.sup.W] - [r.sup.U])L. This expression shows that, even when F < 0 as it is for the United States, total investment inflows can still easily be positive when [r.sup.W] > [r.sup.U] and the stock of gross liabilities is sufficiently large. The expression also reveals, however, that the leveraging mechanism generating the U.S. surplus on investment returns also heightens the risk associated with a possible reversal. An unresolved but critical question is whether the United States' favorable position in international markets will be sustained in the face of a large and growing external debt. Should the United States at some point be forced to pay a higher rate on its liabilities, the negative income effect will be proportional to the extent of leverage, L.

Part of the historical U.S. international investment advantage is a matter of chance and circumstance. Japanese investors famously bought trophy properties like Pebble Beach golf club, Rockefeller Center, and Columbia Pictures at premium prices, only to see those investments sour. Europeans poured money into the U.S. stock market only at the end of the 1990s, just as the technology bubble was about to burst. However, a deeper reason why the United States' net debt position has accumulated only relatively slowly is that Americans hold a considerably larger fraction of their foreign assets in equities (both portfolio equity and foreign direct investment) than do foreigners of their U.S. assets. At the end of 2003, Americans held almost $7.9 trillion in foreign assets, of which 60 percent was in equities, either foreign stocks or foreign direct investment (here measured at market value). Foreigners, by contrast, held only 38 percent of their $10.5 trillion in U.S. assets in the form of equity. Given that equity has, over long periods, consistently paid a significant premium over bonds, it is not surprising that U.S. residents have remained net recipients of investment returns even though the United States apparently crossed the line to being a net debtor in the late 1980s.

A major reason why foreigners hold relatively more U.S. bonds than Americans hold foreign bonds is that the dollar remains the world's main reserve and vehicle currency. Indeed, of the $3.8 trillion in international reserves held by central banks worldwide, a very large share is in dollars, and much of it is in short-term instruments. (16) Figure 7 illustrates the burgeoning reserves of Asia, now in excess of $2 trillion. According to the BEA, over 45 percent of the $700 billion stock of dollar currency is held abroad, and this is probably an underestimate. (17) (Note that, when one speaks of the United States enjoying rents or seigniorage from issuing a reserve currency, the main effects may come from foreigners' relative willingness to hold cash or liquid short-term Treasury debt, rather than from any substantial inherent U.S. interest rate advantage.) In any event, our empirical analysis will take account of the systematically lower return on U.S. liabilities than on assets elsewhere, and will ask what might happen should that advantage suddenly disappear in the process of current account reversal. (18)

At present, as we have noted, the net U.S. foreign debt equals about 25 percent of GDP. This ratio already roughly equals the previous peak of 26 percent, reached in 1894. A simple calculation shows that if U.S. nominal GDP grows at 6 percent a year and the current account deficit remains at 6 percent of nominal GDP, the ratio of U.S. net foreign debt to GDP will asymptotically approach 100 percent. Few countries have ever reached anywhere near that level of indebtedness without having a crisis of some sort. (19)

If large, sudden exchange rate movements are possible, the greater depth of today's international financial markets becomes a potential source of systemic stress. As we have documented, the volume of international asset trading is now vast. Although many participants believe themselves to be hedged against exchange rate and interest rate risks, the wide range of lightly regulated or unregulated nonbank counterparties now operating in the markets raises a real risk of cascading financial collapse. In a world where a country's current account may adjust abruptly, bringing with it large changes in international relative prices, a persistently large U.S. deficit constitutes an overhanging systemic threat.

A sober assessment of present global imbalances suggests the need for a quantitative analysis of how a U.S. current account adjustment would affect exchange rates. We take this up next.

Summary of the Analytical Framework

Here we summarize the main features and mechanisms in our analysis. After reading this section, readers who are primarily interested in our exchange rate predictions can skip the following section, which presents the details of the model, and proceed directly to the discussion of our numerical findings.

We work within a three-region model of a world economy consisting of the United States, Europe, and Asia. These regions are linked by trade and by a matrix of international asset and liability positions. Each region produces a distinctive export good, which its residents consume along with imports from the other two regions. In addition, each region produces nontraded goods, which its residents alone consume.

A key but realistic assumption is that each country's residents have a substantial relative preference for the traded good that is produced at home and exported; that is, consumption of traded goods is intensive in the home export, creating a home bias in traded goods consumption. This feature builds in a "transfer effect" on the terms of trade, which provides one of the key mechanisms through which changes in the international pattern of current account balances change real and nominal exchange rates. A reduction in the U.S. current account deficit, if driven by a fall in U.S. spending and a matching rise in U.S. saving, represents a shift in world demand toward foreign traded goods, which depresses the price of U.S. exports relative to that of imports from both Asia and Europe. (The international terms of trade of the United States deteriorate.) Because the U.S.-produced export good has a larger weight in the U.S. consumer price index (CPI) than that of foreign imports, whereas foreign export goods similarly have larger weights in their home countries' CPIs, the result is both a real and a nominal depreciation of the dollar.

This terms-of-trade effect of current account adjustment has been prominent in the literature, but it is potentially less important quantitatively than is a second real exchange rate effect captured in our model. That effect is the impact of current account adjustment on the prices of nontraded goods. The CPI can be viewed as made up of individual sub-CPIs for traded and nontraded goods, with the latter empirically having about three times the weight of the former in the overall CPI, given the importance of nontraded service inputs into the delivery even of traded products to consumers. The real exchange rate between two currencies is the ratio of the issuing countries' overall CPIs, both expressed in a common currency. Thus a fall in a country's prices for nontraded goods, relative to the same-currency price of nontraded goods abroad, will depress its relative price level just as a terms-of-trade setback does, causing both a real and a nominal depreciation of its currency. Because nontraded goods are so important a component of the CPI, ignoring effects involving their prices would omit much of the effect of current account adjustment on exchange rates. Hence this additional mechanism, absent from much of the policy discussion, is critical to include.

When the U.S. external deficit falls as a result of a cut in domestic consumption, part of the reduction in demand falls on traded goods (exports as well as imports), but much of it falls on U.S. nontraded goods. The consequent fall in the nontraded goods' prices reinforces the effect of weaker terms of trade in causing the dollar to depreciate against the currencies of Europe and Asia. As noted, in our calibration this second effect receives more than twice the weight that terms-of-trade effects receive in explaining exchange rate movements.

We consider several scenarios for U.S. current account adjustment, involving different degrees of burden sharing by Europe and Asia and the resulting effect on those regions' bilateral and effective exchange rates. For example, if Europe's deficit rises to offset a fall in America's deficit, while Asia's surplus remains constant, the dollar will depreciate more against Europe's currencies, and less against Asia's, than if Asia and Europe shared in the burden of accommodating the U.S. return to external balance. In terms of its trade-weighted effective exchange rate, the dollar depreciates more under the second of these two scenarios. Because Asia trades more with the United States than Europe does, bilateral depreciation against Asia's currencies plays the more important role in determining the effective depreciation.

We also consider the effect of dollar exchange rate changes in revaluing gross foreign asset positions, thus redistributing the burden of international indebtedness, as well as the possibility that the adjustment process, especially if disorderly, could entail higher interest payments abroad on U.S. short-term foreign obligations. Finally, key parameters in our model govern the substitutability in consumption among various traded goods and between traded and nontraded goods. In general, the lower these substitution elasticities, the greater the relative price changes caused by current account adjustment and the greater, therefore, the resulting terms-of-trade and exchange rate responses. Because the values of these elasticities are quite uncertain and can differ between the short and the long run, we quantitatively examine their role in generating our numerical estimates.

The Model

The three-country endowment model we develop here extends our earlier small-country and two-country frameworks. (20) We label the three countries (or regions), whose sizes can be flexibly calibrated, U (for the United States), E (for Europe), and A (for Asia). The model distinguishes both between home- and foreign-produced traded goods and between traded and nontraded goods (with the latter margin, largely ignored in many discussions of the U.S. current account deficit, turning out to be the more important of the two quantitatively in our simulations). Our focus here will be on articulating the new insights that can be gained by going from two countries to three, particularly in understanding different scenarios of real exchange rate adjustment across regions as the current account deficit of the United States falls to a sustainable level.

Four features of our model are of particular interest. First, by assuming that endowments are given exogenously for the various types of outputs, we implicitly assume that capital and labor are not mobile between sectors in the short run. To the extent that global imbalances close only slowly over long periods (which experience suggests is not the most likely case), factor mobility across sectors will mute any real exchange rate effects. (21) Second, we do not allow for changes in the mix of traded goods produced or for the endogenous determination of the range of nontraded goods, two factors that would operate over the longer run and could also mute the effects on real exchange rates of current account movements. Third, our main analysis assumes that nominal prices are completely flexible. That assumption--in contrast to our assumption on factor mobility--almost surely leads us to understate the likely real exchange rate effects of a current account reversal. As we discuss later, with nominal rigidities and imperfect pass-through from exchange rates to prices, the exchange rate will need to move more, and perhaps much more, than in our base case in order to maintain employment stability. Fourth, we do not explicitly model the intertemporal allocation of consumption, but rather focus on the intratemporal price consequences of alternative patterns of production-consumption imbalances.

The Core Model

Although notationally intricate, our core three-region model is conceptually quite simple. We assume that consumers in each of the three regions allocate their spending between traded and nontraded goods. Within the category of traded goods, they choose among goods produced in each of the three regions. The equilibrium terms of trade and the relative price of traded and nontraded goods (and thus both bilateral and effective real exchange rates) are determined endogenously. Given assumptions about central bank policy (depending, for example, on whether the central bank aims to stabilize the CPI deflator, the GDP deflator, or a bilateral exchange rate), the model can also generate nominal exchange rates.

We begin by defining [C.sup.i.sub.j] [equivalent to] country i consumption of good (or good category)j. The comprehensive country i consumption index depends on U.S., European, and Asian traded goods consumption (T), as well as consumption of domestic nontraded goods (N). It is written in the following nested form:

(1) [C.sup.i] = [[[[gamma].sup.1/[theta]][([C.sup.i.sub.T]).sup.[theta]-1/ [theta]] + [(1 - [gamma]).sup.1/[theta]][([C.sup.i.sub.N]).sup.[theta]-1/ [theta]].sup.[theta]/[theta]-1] i = U, E, A,

with

(2) [C.sup.U.sub.T] = [[[alpha].sup.1/[eta]][([C.sup.U.sub.U]).sup.[eta]-1/ [eta]] + [([beta] - [alpha]).sup.1/[eta]][(C.sup.U.sub.E]).sup.[eta]-1/[eta]] + [(1-[beta]).sup.1/[eta]][([C.sup.U.sub.A]).sup.[eta]-1/[eta]].sup.[eta]/ [eta]-1]

[C.sup.E.sub.T] = [[[alpha].sup.1/[eta]][([C.sup.E.sub.E]).sup.[eta]-1/ [eta]] + [([beta] - [alpha]).sup.1/[eta]][(C.sup.E.sub.U]).sup.[eta]-1/[eta]] + [(1 -[beta]).sup.1/[eta]][([C.sup.E.sub.A]).sup.[eta]-1/[eta]].sup.[eta]/ [eta]-1]

[C.sup.A.sub.T] = [[[delta].sup.1/[eta]]([C.sup.A.sub.A]).sub.[eta]-1/ [eta]] + [(1 - [delta]/2].sup.1/[eta]][([C.sup.A.sub.E]).sup.[eta]-1/[eta]] + [(1- [delta]/2).sup.1/[eta]] ([C.sup.A.sub.U]).sup.[eta]-1/[eta].

We do not assume identical preferences in the three countries. On the contrary, we wish to allow, both in defining real exchange rates and in assessing the effects of shocks, for a realistic home bias in traded goods consumption, such that each country has a substantial relative preference for the traded good that it produces and exports abroad. (22) Home consumption bias gives rise to a "transfer effect," whereby an increase in relative national expenditure improves a country's terms of trade, that is, raises the price of its exports relative to that of its imports.

In the equations above, the United States and Europe are "mirror symmetric" in their preferences for each other's goods, but each attaches the same weight to Asian goods. Asia weights U.S. and European imports equally but may differ in openness from the United States and Europe. Specifically, we assume that 1 > [beta] > [alpha] > 1/2. We also assume that [delta] > 1/2. For example, if [beta] = 0.8 and [alpha] = 0.7, then the U.S. traded goods consumption basket has a weight of 0.7 on U.S. exports, 0.1 on European exports, and 0.2 on Asian exports. (A very similar--and for many exercises isomorphic-model arises if one assumes that all countries have identical preferences, but that international trading costs are higher than domestic trading costs.) (23)

The values of the two parameters [theta] and [eta] are critical in our analysis. Parameter [theta] is the (constant) elasticity of substitution between traded and nontraded goods. Parameter [eta] is the (constant) elasticity of substitution between domestically produced traded goods and imports from either foreign region. The two parameters are important because they underlie the magnitudes of price responses to quantity adjustments. Lower substitution elasticities imply that sharper price changes are needed to accommodate a given change in quantities consumed.

Price Indexes and Real Exchange Rates

Using standard methods, we derive exact consumption-based price indexes. (24) Define [P.sup.i.sub.j] [equivalent to] the country i exact price index for consumption category j. The corresponding overall CPIs, in dollars, are

(3) [P.sup.i.sub.c] = [[[gamma]([P.sup.i.sub.T]).sup.1-[theta]] + (1 - [gamma])[([P.sup.i.sub.N]).sup.1- [theta]].sup.1/1-[theta]], i = U, E, A,

where subscript C denotes the comprehensive consumption basket. (Our main analysis is in terms of real prices and exchange rates, so all prices can be expressed in terms of the common numeraire.) In equation 3,

(4) [P.sup.U.sub.T] = [[[alpha][P.sup.1-[eta].sub.U] + ([beta] - [alpha]) [P.sup.1-[eta].sub.E] + (1 - [beta]) [P.sup.1- [eta].sub.A]].sup.1/ 1-[eta]]

[P.sup.E.sub.T] = [[[alpha][P.sup.1-[eta].sub.E] + ([beta] - [alpha]) [P.sup.1-[eta].sub.U] + (1 - [beta]) [P.sup.1- [eta].sub.A]].sup.1/ 1-[eta]]

[P.sup.A.sub.T] = [[[alpha][P.sup.1-[eta].sub.A] + (1 - [delta]/2) [P.sup.1-[eta].sub.U] + (1 - [delta]/2) [P.sup.1- [eta].sub.E]].sup.1/ 1-[eta]].

Here [P.sub.i], i = U, E, A, is just the price of the differentiated traded good produced by country i.

We assume the law of one price for traded goods, so that the price of any given country's traded good is the same in all regions. (In practice, of course, the law of one price holds mainly in the breach, partly because of the difficulties in separating out the truly tradable component of "traded" goods.) Because of the home export consumption bias we have assumed, the price indexes for traded goods [P.sup.i.sub.T] can differ across countries even when the law of one price holds, reflecting the asymmetric consumption weightings. As a result, changes in the terms of trade, through their differential effects on countries' price levels for traded goods, affect real exchange rates.

There are three bilateral terms of trade, three bilateral real exchange rates, and three real effective exchange rates. The terms of trade are

(5) [[tau].sub.U,E] = [P.sub.E]/[P.sub.U], [[tau].sub.U,A] = [P.sub.A]/ [P.sub.U], [[tau].sub.E,A] = [P.sub.A]/[P.sub.E] = [[tau].sub.U,A]/ [[tau].sub.U,E].

Here, for example, a rise in [[tau].sub.U,E] is a rise in the price of European traded goods in terms of U.S. traded goods, that is, a deterioration in the U.S. terms of trade. Bilateral real exchange rates are

(6) [q.sub.U,E] = [P.sup.E.sub.C]/[P.sup.U.sub.C], [q.sub.U,A] = [P.sup.A.sub.C]/[P.sup.U.sub.C], [q.sub.E,A] = [P.sup.A.sub.C]/[P.sup.E.sub.C] = [q.sub.U,A] /[q.sub.U,E].

A rise in [q.sub.U,E], for example, is a rise in the price of the European consumption basket in terms of the U.S. consumption basket, that is, a real depreciation of the dollar.

As we have noted, asymmetric preferences over traded goods allow the terms of trade to affect traded goods price indexes. The United States' price index places a comparatively high weight on U.S. exports, whereas Europe's does the same for its own exports. Thus the U.S. traded goods price index falls relative to Europe's when Europe's bilateral terms of trade against the United States improve. Denoting a percent change with a caret, we can logarithmically approximate the evolution of the relative European-to-American traded goods price ratio as

(7) [P.sup.E.sub.T] - [P.sup.U.sub.T] = (2[alpha] - [beta]) [[tau].sub.U,E].

(Exact formulas for relative price indexes, which we use to generate the numerical results reported below, are given in appendix A.) This expression equates the difference between European and U.S. price inflation in traded goods to the European consumption weight on its own exports, [alpha], less the U.S. consumption weight on imports from Europe, [beta] - [alpha], all multiplied by the percentage increase in Europe' s terms of trade against the United States. Observe that the terms of trade against Asia do not enter this expression. Given the bilateral Europe-U.S. terms of trade, changes in the terms of trade against Asia enter the European and U.S. traded goods price indexes symmetrically (that is, with identical consumption weights of 1 - 1]) and therefore drop out in computing their log-difference change.

Similarly, the evolution of the Asian price level for traded goods relative to that of the United States also reflects terms-of-trade movements. But because, under our assumptions, Asia trades more extensively with Europe than the United States does, the prices of European exports have a relatively bigger impact on Asia's average import prices. This is shown by the following logarithmic approximation:

(8) [P.sup.A.sub.T] - [P.sup.U.sub.T] = [[delta] - (1 - [beta])] [[tau].sub.U,A] + [(1 - [delta]/2) - ([beta] - [alpha])][tau.sub.U,E]

The weights on the terms-of-trade changes here simply reflect relative consumption weights, as before. Now, however, given the bilateral Asia-U.S. terms of trade, an improvement in Europe's terms of trade vis-a-vis the United States raises Asia's price index for traded goods relative to that in the United States when, as we assume in our simulations, the Asian consumption weight on European imports, (1 - [delta])/2, exceeds the weight attached by U.S. consumers, [beta] - [alpha]. Such third-country asymmetries cannot be captured, of course, in a two-country framework.

Bilateral real exchange rate movements follow immediately from the expressions above. For Europe and the United States, for example, the log change in the bilateral real exchange rate is simply the consumption weight on traded goods times the log change in relative traded goods price indexes, plus the consumption weight on nontraded goods times the log change in relative nontraded goods price indexes:

(9) [q.sub.U,E] = [gamma](2[alpha] - [beta])[[tau].sub.U,E] + (1 - [gamma])([P.sup.E.sub.N] - [P.sup.U.sub.N]).

Analogously, between the United States and Asia we have

(10) [q.sub.U,A] = [gamma][[delta] - (1-[beta])][[tau].sub.U,A] + [gamma][(1-[delta]/2)-([beta]-[alpha])][[tau].sub.U,E] +(1 - [gamma]) ([P.sup.A.sub.N] - [P.sup.U.sub.N]).

We emphasize one key aspect of these expressions. The weight on nontraded goods is likely to be quite large because of the large component of nontradable services included in the consumer prices of goods generally classified as entirely tradable. In our simulations we therefore take the weight on nontraded goods above, 1 - [gamma], to be 0.75. An implication is that, although the terms of trade certainly are an empirically important factor in real exchange rate determination given home consumption bias, relative prices for nontraded goods potentially play an even larger quantitative role.

Solution Methodology

The methodology we use to calculate the effects of current account shifts on relative prices is essentially the same as that in our earlier papers, extended to a three-region setting. (25) Given fixed output endowments, an assumed initial pattern of current account imbalances, an assumed initial pattern of international indebtedness, and a global interest rate, relative prices are determined by the equality of supply and demand in all goods markets. Changes in the international pattern of external imbalances, whether due to consumption shifts or other changes (including changes in productivity), shift the supply and demand curves in the various markets, resulting in a new set of equilibrium prices. These are the price changes we report below, under a variety of current account adjustment scenarios. (The global sums of external imbalances and of net international asset positions are both constrained to be zero.)

There are six market-clearing conditions, covering the three regional nontraded goods markets and the three global markets for traded goods (although one of these is redundant by Walras' Law). The five independent equilibrium conditions allow solutions for

--the U.S. terms of trade against Europe, [[tau].sub.U,E]

--the U.S. terms of trade against Asia, [[tau].sub.U,A]

--the price of nontraded goods in terms of traded goods in the United States, [P.sup.U.sub.N]/[P.sup.U.sub.T]

--the price of nontraded goods in terms of traded goods in Europe, [P.sup.E.sub.N]/[P.sup.E.sub.T]

--the price of nontraded goods in terms of traded goods in Asia, [P.sup.A.sub.N]/[P.sup.A.sub.T]. One can then calculate the three bilateral real exchange rates, for which these five relative prices are the critical inputs. Because of the asymmetric preferences over traded goods, there is, as we have noted, a transfer effect in the model (wealth transfers feed into the terms of trade and through that channel into real exchange rates), although it is more complex than would be the case with only two countries in the world. Finally, we will also want to define and analyze real effective (loosely speaking, trade-weighted) exchange rates:

(11) [q.sup.U] = [([P.sup.E.sub.C]).sup.[[beta]-[alpha]/1-[alpha]] [([P.sup.A.sub.C].sup.1-[beta]/1-[alpha]/[P.sup.U.sub.C]

[q.sup.E] = [([P.sup.U.sub.C]).sup.[[beta]-[alpha]/1-[alpha]] [([P.sup.A.sub.C].sup.1-[beta]/1-[alpha]]/[P.sup.E.sub.C]

[q.sup.A] = ([P.sup.U.sub.C]).sup.1/2][([P.sup.E.sub.C]).sup.1/2]/ [P.sup.A.sub.C].

Three extensions to the analysis add to its relevance and realism. (26) First, we ask how real exchange rate changes translate into nominal exchange rate changes; this depends on central bank policy. In general, this turns out not to he a critical issue empirically; the other two extensions are potentially far more important. One of these is to take into account how exchange rate changes affect the net foreign asset positions of the different regions, because of currency mismatches between gross assets and liabilities. (27) This valuation effect is significant, but its impact on aggregate demand is of secondary importance compared with the primary demand shifts emphasized in our preceding analysis. Finally, our third extension takes into account the effect of a rise in relative U.S. interest rates (due, say, to concern about government deficits or erosion of the dollar's reserve currency status). This effect, which works to worsen rather than ease the adjustment problem, is also significant, although again it is less important (at least over the range of interest rates we consider) than the primary effects of a rebalancing of global demand.

Model Predictions

With these critical behavioral parameters in hand, we are now ready to explore the model' s quantitative predictions for global exchange rates and the terms of trade under various scenarios for rebalancing the U.S. current account. We first need to think about parametrizing the model.

Choosing Parameters

As we have already observed, the critical parameters in the model are [theta], the elasticity of substitution in consumption between traded and nontraded goods, and [eta], the elasticity of substitution in consumption among the traded goods produced by the three regions. The lower are these elasticities, the greater the exchange rate and price adjustments needed to accommodate any interregional shifts in aggregate demand. Most of our simulations will be based on a value of [theta] = 1, which is high relative to some estimates suggested in the literature. (28) We will also report results, however, for an even higher elasticity of [theta] = 2.

Our baseline choice of [eta] = 2 as a representative aggregate trade elasticity is a compromise between two sets of evidence. Estimates based on trade flows within disaggregated product categories cover a wide range but typically include many values much higher than [eta] = 2. (29) On the other hand, conventionally estimated aggregate trade equations, as well as calibrations of dynamic general equilibrium models, tend to indicate much smaller values for [eta], typically 1 or even lower.

A number of mechanisms have been suggested to explain this discrepancy, some echoing Guy Orcutt's classic skepticism about the low elasticities seemingly implied by macro-level estimators. (30) Aggregation bias lowers estimated macroelasticities because the price movements of low-elasticity goods tend to dominate overall movements in import and export price indexes. (31) Another issue is that macroeconomic estimates of business-cycle frequency correlations tend to confound permanent and temporary price movements, in contrast to micro-level cross-sectional or panel studies centered on trade liberalization episodes. (32) In taking [eta] = 2, we try, in a crude way, to address these biases while also recognizing the empirically inspired rules of thumb that inform policymakers' forecasts. We also include an illustrative simulation of the case [eta] = 100 (in which all traded goods are essentially perfect substitutes). That simulation shuts down the terms-of-trade effects and thereby shows how large a role is being played by substitution between traded and nontraded goods, the channel we have emphasized elsewhere. (33)

We set both [alpha] and [delta] equal to 0.7; these are the consumption weights that Americans and Europeans, on the one hand, and Asians, on the other, attach to their own domestic products within their traded goods consumption baskets. That choice is plausible based on our discussion in an earlier paper. (34) We set [beta] = 0.8, implying that Europe and the United States alike place weights of [beta] = 0.1 on each other's traded goods, and twice that weight (0.2) on Asian goods. Asia, by assumption, distributes its demand evenly across the other two regions (placing a weight of 0.15 on the exports of each). So, in our model, Europe and the United States both trade more with Asia than with each other. We assume that all three regions produce the same number of units of tradable goods output.

Appendix A discusses in detail our assumptions regarding gross liabilities and assets for each region, as well as the currencies of denomination of these stocks. The point we stress here is that, to a first approximation, the United States is a net debtor (to the tune of 25 percent of its GDP, or 100 percent of its exportable GDP), and greater Europe has approximately a zero net international position. Our model' s third region, Asia, therefore is left as a net international creditor in an amount equal to 100 percent of U.S. tradable GDP. U.S. gross foreign liabilities are almost all in dollars, but U.S. gross foreign assets are only about 40 percent in dollars. We assume that greater Asia's gross liabilities are equally divided among U.S., European, and Asian currencies (because Japan borrows in yen), whereas Asian gross foreign assets are 80 percent in dollars and 20 percent in European currencies. For Europe we assume that gross foreign assets are 32 percent in dollars, 11 percent in Asian currencies, and 57 percent in European currencies. In our model, 80 percent of European gross liabilities are denominated in European currencies, and the balance in dollars. These numbers are very rough approximations, based in some cases on fragmentary or impressionistic data, but portfolio shares can shift sharply over time, and so there is little point in trying too hard to refine the estimates. As we shall see, these shares do imply large potential international redistributions of wealth due to exchange rate changes, but those redistributions themselves have only a secondary impact on the exchange rate implications of current account adjustment.

For nominal interest rates we take a baseline value of 3.75 percent a year for U.S. liabilities but 5 percent a year for all other countries' liabilities. This assumption captures the "exorbitant privilege" the United States has long enjoyed of borrowing from the world more cheaply than it lends. (35)

Turning to current accounts, we place the U.S. external deficit at 20 percent of U.S. tradable GDP. (36) This is consistent with a U.S. current account deficit of 5 percent of total GDP, a reasonable baseline if part of the 2004 deficit is due to temporarily high oil prices. Because we find our simulation results to be approximately linear within the parameter space we are considering, it is easy to adjust the prediction to the case in which the 2004 deficit of 6 percent of GDP persists. In any event, what matters most for our calibration is how much the current account balance adjusts (for example, from 6 to 3 percent of GDP). We assume an initial position with Europe's current account surplus at 5 percent of U.S. tradable GDP and Asia's at 15 percent. (37)

A final benchmark to establish is our initial reference value for measuring subsequent exchange rate adjustments. This issue was less critical in our earlier two papers, because trade-weighted effective exchange rates move more slowly than the bilateral exchange rates that we consider below. In our basic model prices are flexible and economic responses to them are immediate. In practice, however, there are considerable lags: Michael Mussa, for example, posits the rule of thumb that the U.S. trade balance responds with a two-year lag to dollar exchange rate changes. (38) In that case, if today's current account balances reflect averages of exchange rates over the past two years, it would be more accurate to think of our simulations as giving exchange rate changes relative to two-year average reference rates rather than current rates. Table 2 presents some resulting reference exchange rates. (The Chinese and Malaysian currencies have been pegged over the past two years, and so their current and average rates are the same.)

Simulations

With the model and our parameter assumptions in hand, we are ready to consider alternative simulations. Underlying much of our analysis is the assumption that demand shocks (such as a rise in U.S. saving) are driving the redistribution of global imbalances. This seems by far the most realistic assumption, given the magnitude of the external financing gaps.

Tables 3 through 6 lay out the results of three scenarios under which the U.S. current account balance might improve by 20 percent of tradable GDP or, equivalently, 5 percent of total GDP. (All simulations include the effect of exchange rate changes in revaluing the regions' foreign assets and liabilities.) In the "global rebalancing" scenario (the first column in each table), all regions' current account balances go to zero (with trade balances adjusting as needed to service interest flows on the endogenously determined stocks of net foreign assets). Looking first at bilateral real exchange rates, in table 3, we see that Asia's exchange rate with the United States rises by 35.2 percent, and Europe's rises by 28.6 percent (we define the real exchange rate such that these changes indicate real depreciations of the dollar). Europe sees an improvement in its terms of trade against the United States (a rise in the price of Europe's exports relative to its U.S. imports) of 14.0 percent, and Asia sees an improvement of 14.5 percent.

What are the implications for nominal exchange rates? To answer this question we must specify monetary policies. We consider two possibilities: that central banks stabilize the domestic CPI, and that they stabilize the domestic GDP deflator. Table 4 reports the results. Under CPI targeting, the monetary authorities hold overall price levels constant, so that the only source of real exchange rate change is nominal exchange rate change. As a result, nominal and real exchange rate changes are equal, as can be seen by comparing table 4 with table 3. (39) Because none of the three regions is extremely open to trade, movements in CPIs and in GDP deflators are fairly close, and, as a result, nominal exchange rate changes when the GDP deflator is stabilized differ very little from those under CPI stabilization.

The appreciation of Europe's currencies against the dollar is smaller than that of Asia's under the first scenario, because Asia starts out in our simulation with a much larger external surplus than Europe does, and so it has more adjusting to do. But the Asian currencies' appreciation against the dollar is mitigated somewhat by the fact that Asia trades more with the United States than Europe does. (40) We see in table 5 that Europe's real effective currency appreciation--represented, as is traditional for such multilateral indexes, by a positive number--is much smaller than Asia's: only 5.1 percent versus 20.9 percent. Again, this reflects the greater weight of the dollar in Asia's trade-weighted real exchange rate than in Europe's. Notice that, as in table 4, nominal (under GDP deflator targeting) and real effective exchange rate changes are again quite close numerically.

Another factor underlying the equilibrium exchange rate responses is that dollar depreciation implies a much bigger reduction in Asia's net foreign asset position than in Europe's. (Table 6 shows the impacts under GDP deflator targeting.) Asia has 80 percent of its assets, but only 34 percent of its liabilities, in dollars. Thus, under the global rebalancing scenario, dollar depreciation raises Asia's gross liabilities relative to its gross assets, pushing its net foreign assets down (as a fraction of U.S. tradable GDP) by 60 percent. Europe, by contrast, has only 32 percent of its assets and 20 percent of its liabilities in dollars. The fact that Asia loses so much on the asset side implies that its trade surplus shrinks by less than its current account surplus does. Because trade surpluses are what drive the constellation of real exchange rates, the real appreciation of the Asian currencies is mitigated. In sum, thanks to Asia's greater openness and to the fact that Asia suffers particularly large capital losses on foreign assets when the dollar falls, Asian exchange rates do not need to change quite as much as a model-free, back-of-the-envelope calculation might suggest.

The tables cover two other possible scenarios. The second column in tables 3 through 6 analyzes a "Bretton Woods II" scenario, in which Asia clings to its dollar peg. (41) We calibrate this case by setting the U.S. current account balance to zero and endogenously varying Asia's and Europe's current account balances in a way that both maintains Asia's bilateral nominal exchange rate with the United States (assuming GDP deflator targeting) and absorbs the fall in U.S. borrowing. (Of course, nonmonetary policy instruments such as fiscal policy would have to be used to attain just the fight constellation of current account balances.) In this case the bilateral real exchange rates of the European currencies against the dollar must rise spectacularly, by 58.5 percent, and they would rise against the Asian currencies by 59 percent. This result also is approximately linear in the change in the U.S. current account balance. Thus, under the Bretton Woods II scenario, eliminating only half the U.S. current account deficit would raise the real value of the European currencies against the dollar by as much as would occur in a global rebalancing scenario that eliminates the U.S. current account deficit entirely.

For Asia to maintain its nominal exchange rate peg in the face of a balanced U.S. current account, it must drive its own current account balance significantly further into surplus, from 15 percent to 31 percent of U.S. tradable GDP. And Europe would have to move from a surplus equal to 5 percent of U.S. tradable GDP to a 31 percent deficit! (See the footnotes to table 3.) When Asia pegs its currencies to a falling dollar, its own traded goods become more competitive and its imports more expensive relative to domestic nontraded goods. Both factors shift world demand away from Europe, which, by assumption, is passively absorbing the blow, and toward Asia. These calibrations make patently clear why sustaining Asia' s dollar peg is likely to be politically unpalatable for many of its trading partners if the U.S. current account deficit ever shrinks. Asia would be extremely vulnerable to a protectionist backlash.

As table 6 shows, the sharp appreciation of Europe's currencies in the Bretton Woods II scenario also decimates its external asset position, which declines from balance to -70 percent of the value of U.S. tradable production. Asia suffers somewhat, and the U.S. net asset position is the major beneficiary, because U.S.-owned foreign assets are concentrated in European currencies. Europe is thus hammered both by a sharp decline in its competitiveness and by a loss on its net foreign assets of about $2 trillion.

The third scenario reported in tables 3 through 6 is a muted version of the Bretton Woods II scenario. Here, instead of maintaining its dollar currency peg, Asia maintains its current account surplus unchanged in the face of U.S. adjustment to a balanced position. That is, rather than increasing its current account surplus, it allows enough exchange rate adjustment to keep the surplus constant. In this case, as table 5 shows, Europe's real effective exchange rate rises by much less than in the Bretton Woods II scenario (31.7 percent versus 58.9 percent), and the Asian currencies experience a real effective depreciation of only 2.9 percent, versus 29.8 percent in Bretton Woods II. Still, because the U.S. current account balance improves dramatically while Asia's holds steady, the Asian currencies rise in real terms by 19.4 percent against the dollar (table 3). This exercise reveals a fallacy in the argument that Asia cannot allow its dollar peg to move without losing the ability to absorb its surplus labor. To the extent that European demand increases, Asia can retain its external surplus while releasing its dollar peg.

In table 7 we revisit the global rebalancing scenario but vary the critical substitution elasticities in the model. (Only real exchange rate changes, which equal nominal changes under CPI inflation targeting, are listed.) In the first column we assume an elasticity of substitution between traded and nontraded goods, [theta], of 2 instead of 1. As we have already argued, the limited evidence in the empirical macroeconomics literature suggests that this estimate is well on the high side, but it allows us to incorporate a more conservative range of potential exchange rate adjustments alongside our baseline estimates. Under this assumption the real dollar exchange rate with the European currencies rises by only 19.3 percent, instead of 28.6 percent as in the first column of table 3, and the Asian currencies rise against the dollar by 22.5 percent instead of 35.2 percent. The dollar falls in real effective terms (results not shown) by 21.5 percent rather than 33 percent. These calculations show that, even with a relatively high value for [theta], the required adjustment of exchange rates is quite significant even if, as here, prices are flexible.

The second column in table 7 examines the case in which [theta] = 1 but [eta] = 100, so that the various countries' tradable outputs are almost perfect substitutes. This exercise, which essentially eliminates terms-of-trade adjustments as a factor in moving real exchange rates, allows us to see how much of the change in exchange rates is due to within-country substitution between traded and nontraded goods. This variation mutes the exchange rate changes by an amount roughly similar to those found in the previous exercise. The real effective dollar exchange rate (again not shown) falls by 21 percent. According to this calibration, roughly two-thirds of the needed dollar adjustment is driven by substitution between traded and nontraded goods, and only one-third is driven by the terms-of-trade channel typically emphasized in the literature. This should not be surprising, given that (according to our previously cited calibration) roughly 75 percent of GDP is nontraded. With more conservative assumptions about international trade, however (either greater home bias in consumption or lower substitutability of countries' traded outputs, such that [eta] = 1), the terms-of-trade channel would become more important.

At present the United States is absorbing traded goods (domestic and foreign) equivalent to roughly 30 percent of its GDP. This demand needs to adjust downward while avoiding a reduction in nontraded goods absorption if full employment is to be maintained; such a shift will therefore require a significant change in the relative price of nontraded goods. Still, terms-of-trade changes do account for about one-third of the overall adjustment, a proportion slightly larger than that found in our two-country model, where we did not allow for trade or terms-of-trade adjustments between non-U.S. economies.

Given the United States' leveraged international portfolio, with gross debts mostly in dollars and assets significantly in foreign currencies, an unexpected dollar depreciation reduces the U.S. net foreign debt. The first two columns of table 8 report the results of simulations, within the global rebalancing scenario, that illustrate the quantitative importance of such asset valuation effects. Gourinchas and Rey have recently estimated that nearly one-third of the settlement of the U.S. net foreign debt has historically been effected by valuation changes, with the remaining two-thirds covered by higher net exports. (42) The first column in table 8 shows results for the global rebalancing scenario with valuation effects taken into account (identical to the first column in table 3). The second column shows the changes in bilateral exchange rates that would be required if there were no valuation effects (or, equivalently, if exchange rate changes were accurately anticipated and nominal returns adjusted fully to compensate). All relative price changes against the United States are larger in this case, because the United States does not get the benefit of a sharp reduction in its net dollar liabilities. Correspondingly, the U.S. trade balance needs to adjust more for any given adjustment in the current account deficit. The real exchange rate between the dollar and the European currencies needs to move by 33.7 percent, rather than 28.6 percent when valuation effects are taken into account, and the real value of the Asian currencies needs to rise by 40.7 percent against the dollar instead of 35.2 percent. The real effective dollar exchange rate falls by 37.8 percent instead of 33.0 percent (results not shown). According to these numbers, asset revaluation effects will mute the required movement in exchange rates as the U.S. current account closes up, but the trade balance has to do the heavy lifting, since 87 percent (33.0 / 37.8) of the necessary real exchange rate adjustment remains. That valuation effects have only a secondary effect on equilibrium relative price changes is not surprising: big valuation effects can only come from big exchange rate movements.

Our calculations so far do not take into account the likelihood of an accompanying rise in global interest rates, which would hurt the United States (a net debtor) and help Asia (a net creditor). A broad range of scenarios are possible here; we examine only a single very simple one. (Appendix A gives details of the calculation.) In the third column of table 8, we assume that annual interest rates on short-term U.S. debt rise from 3.75 percent to 5 percent, the same level assumed for all other liabilities. In other words, perhaps because of heightened risk perceptions, the United States simply loses its historical low borrowing rate and is put on a par with other debtors. This change wipes out a good deal of the effect of the valuation changes (and would wipe out even more if it applied to all U.S. external liabilities, not just the roughly 30 percent consisting of short-maturity debt). As our introductory discussion suggested, the United States, as an important issuer of bonds relative to equity, is extremely vulnerable to increases in interest rates, even when all global bond rates rise together.

Until now we have been concentrating on demand shocks. Productivity shocks may make the adjustment process more or less difficult, depending on their source. Higher productivity in foreign traded goods production can actually result in an even greater real depreciation of the dollar as equilibrium is reestablished in world markets. If, on the other hand, it is nontraded goods productivity in Asia and Europe that rises, the exchange rate effects of global rebalancing will be muted. As table 9 illustrates, a 20 percent rise in nontraded goods productivity outside the United States implies notably smaller real exchange rate changes, although the terms-of-trade shifts are similar. A large rise in U.S. traded goods productivity would also facilitate a softer landing. In this case, however, although the extent of real dollar depreciation is somewhat reduced, the U.S. terms of trade fall much more sharply (results not reported).

Some Further Considerations

We believe our model offers many useful insights, but of course there are many caveats to its interpretation. Some of these suggest that our results understate the dollar's potential decline, and some that they overstate it.

Intersectoral Factor Mobility

A critical implicit assumption of our model is that capital and labor cannot quickly migrate across sectors, so that prices rather than quantities must bear the burden of adjustment in response to any sudden change in relative demands for different goods. This assumption seems entirely reasonable if global current account adjustment (full or partial) takes place moderately quickly, say, over one to two years. In the short run, workers cannot change location easily, worker retraining is expensive, and a great deal of capital is sector-specific. Over much longer periods, however (say, ten to twelve years), factor mobility is considerable. If, for example, prices rise dramatically in the U.S. traded goods sector, new investment will be skewed toward that sector, as will new employment. Thus, in principle, a gradual closing of the U.S. current account deficit would facilitate much smoother adjustment with less exchange rate volatility. Unfortunately, our model is not explicitly dynamic. (43) One can, however, artificially approximate gradual current account adjustment by allowing for progressively higher elasticities of substitution. We do this in table 10, where we reconsider our central scenario (which assumed [theta] = 1 and [eta] = 2) by comparing it with two cases in which substitution elasticities are much higher. As the table shows, in the case with "gradual" unwinding (proxied by [theta] = 2 and [eta] = 4), which we loosely take to capture a five- to seven-year adjustment horizon, the bilateral exchange rate changes involving the dollar are only about half as big as in our central global rebalancing scenario. For a "very gradual" unwinding (which we take to occur over ten to twelve years, with [theta] = 4 and [eta] = 8), the same real exchange rate changes are less than a quarter as large as in the central scenario.

Sticky Prices

Factor mobility kicks in to smooth current account adjustment if the adjustment is slow and relatively well anticipated. If, on the other hand, current account imbalances have to close up very quickly (say, because of a collapse in U.S. housing prices), the bias in our estimates would point in the other direction. Nominal rigidities in prices would then play a large role, and actual exchange rate movements would likely be two or more times as large as in our central scenario, for several reasons. (44)

For one thing, our model assumes that the law of one price holds for traded goods, whereas in fact at most half of an exchange rate adjustment typically passes through to traded goods prices even after one year. (45) Thus, in order to balance supply and demand for the different categories of goods while maintaining full employment, central banks would have to allow much larger exchange rate movements--possibly double those suggested by the model. These larger movements would be "overshoots" in the sense that they would unwind over time as domestic prices adjust.

The nominal prices of nontraded goods are typically even stickier than those of traded goods; this further amplifies the overshooting effect. In general, both sticky prices and slow factor mobility point toward the likelihood that a slow unwinding of the U.S. current account deficit will lead to smaller changes in real exchange rates than would a relatively abrupt correction.

Rising U.S. Interest Rates and the Dollar

Another qualification to our results is that our model does not account for financial factors, and in particular for the possibility of temporarily high real interest rates in the United States muting the dollar's decline. Using the Federal Reserve's macroeconomic model, David Reifschneider, Robert Tetlow, and John Williams estimate that a 1-percentage-point rise in the federal funds rate (presumably unmatched by the rest of the world) leads to a 2.2 percent appreciation of the dollar after one year, and a 4.9 percent appreciation after two years. (46) Therefore the fact that, over the past year, U.S. short-term interest rates have been rising relative to Europe's is a countervailing consideration to those discussed above (although our calculations suggest that it is likely to be far less important quantitatively). In addition, Europe and Asia can always choose to lower their interest rates to further mute the dollar's decline. Of course, interest rate policy can only affect the dollar's real value temporarily, and so long-term global rebalancing will still require a combination of real exchange rate adjustment and factor reallocation across sectors.

The Fundamental Unpredictability of Exchange Rates

Our model suggests that the gaping U.S. current account deficit is a very large negative factor in assessing the future prospects of the dollar. It is well known, however, that it is extremely difficult to explain exchange rate swings between major currencies, much less forecast them, at least at horizons up to eighteen months. (47) Although a number of small qualifications must be made to this result, (48) it remains broadly true. How, then, can one be concerned about a medium-term dollar decline if a rise is equally likely? There are two broad answers to this question. First, even the most cheery U.S. current account optimist would have to concede that an abrupt reversal is a potential risk, particularly while federal government deficits remain less than fully tamed. Reversal need not result from what Guillermo Calvo, in the context of emerging markets, has called a "sudden stop" of capital inflows; (49) as we have noted, it could follow, for example, from a rise in U.S. saving due to a purely domestic asset price collapse. Our calibrations are useful in laying out the exchange rate consequences and in illuminating how the burden of adjustment might be shared among the major economies.

Second, and more fundamentally, there is some evidence that nonlinearities are also important, so that, when exchange rates are particularly far out of line with one or more fundamentals, some predictability emerges. Obstfeld and Alan Taylor, for example, argue that convergence to purchasing power parity is much more important quantitatively when a currency is relatively heavily over- or undervalued compared with its long-term real exchange rate. (50) Gourinchas and Rey argue that, contrary to the canonical Meese-Rogoff result, there is a forecastable component to trade-weighted dollar exchange rate movements when net foreign assets or debts are large relative to the United States' net export base. (51) Their work supports much earlier work by Peter Hooper and John Morton suggesting that net foreign assets may be important in explaining dollar movements. (52) As we argued in the introduction, the U.S. current account deficit today is so large and unprecedented that it is difficult to project its future path and the consequences thereof simply by extrapolating from past data.

Conclusions

We have developed a simple stylized model that can be used to calibrate exchange rate changes in response to various scenarios under which the U.S. current account deficit might be reduced from its unprecedented current level. Aside from its quantitative predictions, the model yields a number of important qualitative insights.

First, Asia' s greater openness to trade implies that the requisite exchange rate adjustments for that region are not all that much greater than Europe' s. This appears true despite the fact that Asia starts from a much larger current account surplus than Europe.

Second, we find that, if Asia tries to stick to its dollar peg in the face of, say, a rise in the U.S. saving rate that closes up the U.S. current account gap even partly, Asia will actually have to run significantly larger surpluses than it does now. Europe would bear the brunt of this policy, ending up with a current account deficit even larger than that of the United States today, while at the same time suffering a huge loss on its net foreign assets.

Third, although dollar depreciation does tend to improve the U.S. net foreign asset position (because virtually all of its gross foreign liabilities, but less than half of its gross foreign assets, are denominated in dollars), this effect only slightly mitigates the requisite exchange rate change. Valuation effects will not rescue the United States from a huge trade balance adjustment. Indeed, if relative interest rates on U.S. short-term debt rise even moderately during the adjustment process, this adverse effect could easily cancel out any gain due to valuation effects.

Fourth, our model suggests that the need for deficit countries to shift demand toward nontraded goods (and for surplus countries to shift demand away from them) is roughly twice as important quantitatively as the much more commonly stressed terms-of-trade channel (which involves substitution between the traded goods produced by different countries). The importance of the terms of trade would be greater with lower international trade elasticities than we have assumed, or with a greater degree of home bias in consumption.

We have only scratched the surface of the possible questions that can be asked within our framework. To that end, we have tried to make our approach as transparent as possible so that other researchers can easily investigate alternative scenarios using the model. Clearly, it would be interesting to extend the model in many dimensions, in particular to allow for sticky prices and for dynamic adjustments, such as factor movement across sectors. It would also be interesting to extend the framework to allow for more regions of the world economy, for example, oil producers, non-Asian emerging markets, and Asian subregions. Nonetheless, in a literature that is often long on polemics and short on analysis, we hope it is useful to have a concrete model on which to base policy evaluation.

APPENDIX A

Equilibrium Prices, Revaluation Effects, and Interest Rate Effects

Equilibrium Prices

Here we show how real exchange rates depend on equilibrium relative prices, and we spell out the relevant equilibrium conditions for our three-region world economy. By definition, real exchange rates depend on relative international prices for both traded and nontraded goods. We take up relative traded goods prices first.

As the text noted, notwithstanding the law of one price, the assumed internationally asymmetric preferences over tradables permit relative regional price indexes for tradable consumption to vary over time. Instead of being fixed at unity, these ratios are given in our model by

(A1) [P.sup.E.sub.T]/[P.sup.A.sub.T] = [[[alpha][[tau].sup.1-[eta].sub.U,E] + ([beta] - [alpha]) + (1 - [beta])][[tau].sup.1-[eta].sub.U,A]].sup.1/ 1-[eta]] / [[[alpha] + ([beta] - [alpha])[[tau].sup.1-[eta].sub.U,E] + (1 - [beta])][[tau].sup.1-[eta].sub.U,A]].sup.1/1-[eta]]

[P.sup.A.sub.T]/[P.sup.U.sub.T] = [[[delta][[tau].sup.1-[eta].sub.U,A] + (1 - [delta]/2) + (1 - [delta]/2) [[tau].sup.1-[eta].sub.U,E]].sup.1/1-[eta]] / [[[alpha] + ([beta] - [alpha])[[tau].sup.1-[eta].sub.U,E] + (1 - [beta])] [[tau].sup.1-[eta].sub.U,A]].sup.1/1-[eta]]

[P.sup.A.sub.T]/[P.sup.E.sub.T] = [[[delta][[tau].sup.1-[eta].sub.U,A] + (1 - [delta]/2) + (1 - [delta]/2) [[tau].sup.1-[eta].sub.U,E]].sup.1/1-[eta]] / [[[alpha][[tau].sup.1-[eta].sub.U,E] + ([beta] - [alpha]) + (1 - [beta]) [[tau].sup.1-[eta].sub.U,A]].sup.1/1-[eta]]

Thus shifts in interregional real exchange rates q reflect both shifts in the relative prices of traded and nontraded goods and shifts in the relative prices of exports and imports:

(A2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.].

Having defined relative price indexes, one can easily derive global market-clearing conditions for each region's tradable output, again using very standard techniques for constant elasticity of substitution models such as the one we have here. (53) For real U.S. tradable goods output, the market-clearing condition is given by

(A3) [Y.sup.U.sub.T] = [gamma][alpha][([P.sub.U]/ [P.sup.U.sub.T]).sup.-[eta]][([P.sup.U.sub.T]/[P.sup.U.sub.C]).sup.-[theta]] [C.sup.U] + [gamma]([beta] - [alpha])[([P.sub.U]/[P.sup.E.sub.T]).sup.-[eta]] [([P.sup.E.sub.T]/[P.sup.E.sub.C]).sup.-[theta]] [C.sup.E] + [gamma] (1 - [delta]/2)[([P.sub.U]/[P.sup.A.sub.T]).sup.-[eta]] [([P.sup.A.sub.T]/ [P.sup.A.sub.C]).sup.-[theta]] [C.sup.A],

and that for real European traded goods output is given by

(A4) [Y.sup.E.sub.T] = [gamma][alpha][([P.sub.E]/ [P.sup.E.sub.T]).sup.-[eta]] [([P.sup.E.sub.T]/[P.sup.E.sub.C]).sup.-[theta]] [C.sup.U] + [gamma](1 - [delta]/2)[([P.sub.E]/[P.sup.A.sub.T]).sup.-[eta]] [([P.sup.A.sub.T]/[P.sup.A.sub.C]).sup.-[theta]] [C.sup.A].

Walras' Law implies that the condition for Asian traded goods equilibrium is superfluous, given the two others. One can similarly derive the market-clearing condition for U.S. nontraded goods as

(A5) [Y.sup.U.sub.N] = (1 - [gamma])[([P.sup.U.sub.N]/ [P.sup.U.sub.C]).sup.-[theta]] [C.sup.U]

(which depends, of course, only on U.S. demand), as well as the two corresponding conditions for European and Asian nontraded goods.

We take output endowments as given, and we then use the market-equilibrium conditions just stated to solve for relative prices as functions of current account balances and initial net foreign asset positions. (In our simulations we allow for currency revaluation effects on foreign assets and liabilities, and for the feedback to trade balances needed to sustain any given constellation of current accounts.)

To proceed, we first rewrite the equilibrium condition for the U.S. export good's market as

(A6) [Y.sup.U.sub.T] = [alpha][([P.sub.U]/[P.sup.U.sub.T]).sup.[eta]] [C.sup.U.sub.T] + ([beta] + [alpha])[([P.sub.U]/[P.sup.E.sub.T]).sup.-[eta]] [C.sup.E.sub.T] + (1 - [delta]/2)[([P.sub.U]/[P.sup.A.sub.T]).sup.-[eta]] [C.sup.A.sub.T],

or, in nominal terms, as

(A7) [P.sub.U][Y.sup.U.sub.T] = [alpha][([P.sub.U]/ [P.sup.U.sub.T]).sup.1-[eta]] [P.sup.U.sub.T][C.sup.U.sub.T] + ([beta] + [alpha])[([P.sub.U]/[P.sup.E.sub.T]).sup.1-[eta]] [P.sup.E.sub.T][C.sup.E.sub.T] + (1 - [delta]/2)[([P.sub.U]/ [P.sup.A.sub.T]).sup.1-[eta]] [P.sup.A.sub.T][C.sup.A.sub.T]

If trade were balanced and international debts zero, then, of course, the value of U.S. traded goods consumption would have to equal that of U.S. traded goods production. Here we want to allow for international debt as well as for trade and current account imbalances (which are the same in the model except for net factor payments). The U.S. current account surplus in dollars is given by

(A8) C[A.sup.U] = [P.sub.U][Y.sup.U.sub.T] + r[F.sup.U] - [P.sup.U.sub.T] [C.sup.U.sub.T],

where [F.sup.U] is the stock of U.S. net foreign assets (in dollars) and r is the nominal (dollar) rate of interest. Similarly, for Europe (and again measuring in dollars),

(A9) C[A.sup.E] = [P.sup.E][Y.sup.E.sub.T] + r[F.sup.E] - [P.sup.E.sub.T] [C.sup.E.sub.T].

In the aggregate, of course (in theory if not in the actual data),

(A10) C[A.sup.U] + C[A.sup.E] + C[A.sup.A] = 0.

Similarly,

(A11) [F.sup.U] + [F.sup.E] + [F.sup.A] = 0.

Thus,

(A12) C[A.sup.A] = -(C[A.sup.U] + C[A.sup.E]) = [P.sub.A][Y.sup.A.sub.T] - r([F.sup.U] + [F.sup.E]) - [P.sup.A.sub.T][C.sup.A.sub.T].

In this framework one can consider the effects of a variety of shocks that change the current nexus of global current account imbalances into one where, say, C[A.sup.U] = 0. (Other external balance benchmarks can be analyzed just as easily.)

To do so, we use the above current account equations (and the implied trade balances) to substitute for dollar values of consumption of traded goods in the goods-market equilibrium conditions. The results are

(A13) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.].

Critically, current account imbalances also spill over into relative prices for nontraded goods, to a degree that depends on the elasticity of substitution between traded and nontraded goods. For the three nontraded goods markets, one can show that

(A14) [P.sup.U.sub.N][Y.sup.U.sub.N] = 1 - [gamma]/[gamma] [([P.sup.U.sub.N]/(P.sup.U.sub.T]).sup.1-[theta]] [P.sup.U.sub.T][C.sup.U.sub.T]

= 1 - [gamma]/[gamma][([P.sup.U.sub.N]/[P.sup.U.sub.T]).sup.1-[theta]] [P.sup.U][Y.sup.U.sub.T] + r[F.sup.U] - C[A.sup.U]

[P.sup.E.sub.N][Y.sup.E.sub.N] = 1 - [gamma]/[gamma][([P.sup.E.sub.N]/ [P.sup.E.sub.T]).sup.1-[theta]][[P.sub.E][Y.sup.E.sub.T] - r([F.sup.E] + C[A.sup.E])

[P.sup.A.sub.N][Y.sup.A.sub.N] = 1 - [gamma]/[gamma][([P.sup.A.sub.N]/ [P.sup.A.sub.T]).sup.1-[theta]][[P.sub.A][Y.sup.A.sub.T] - r([F.sup.U] + [F.sup.E]) + C(A.sup.U) + C[A.sup.E]].

Revaluation of Gross Asset Stocks through Exchange Rate Changes

A key variable in the simulation analysis is [f.sup.i], which is the ratio of net foreign assets (in dollars), [F.sup.i], divided by the dollar traded goods income of the United States, [P.sub.U][Y.sup.U.sub.T]. In reality, a country's gross assets and liabilities are often denominated in different currencies, so that focusing only on the net position misses important revaluation effects that can occur as the exchange rate changes. Here we show how we have modified our simulation analysis to take into account both the normalization of dollar net foreign assets and the revaluation effects of exchange rate changes. (54)

Let [H.sup.i] equal the gross assets of country i and [L.sup.i] its gross liabilities, measured in dollars. Then

(A15) [F.sup.i] = [H.sup.i] - [L.sup.i] and

(A16) [f.sup.i] = [H.sup.i] - [L.sup.i]/[P.sub.U][Y.sup.U.sub.T]

One can show that, under a monetary policy that targets the GDP deflator,

(A17) [P.sub.U] - [([P.sup.U.sub.N]/[P.sup.U.sub.T]).sup.[gamma] - 1] [[[alpha] + ([beta] - [alpha])[[tau].sup.1 - [eta].sub.U,E] + (1 - [beta]) [[tau].sup.1 - [eta].sub.U,A]].sup.[gamma] - 1/1 - [eta]]

The first step is to substitute this formula for [P.sub.U] into the denominators of [f.sup.U], [f.sup.E], and [f.sup.A]. The second step is to consider how exchange rate changes affect the numerators.

Let [[omega].sup.i.sub.j] be the share of region i gross foreign assets denominated in the currency of region j, j = U, E, A, where the European and (especially) the Asian regional currencies are composites. Similarly, define the portfolio currency shares [[lambda].sup.i.sub.j] on the liability side. We will assume that central banks target GDP deflators and that [E.sub.U, j] denotes the (nominal) dollar price of currency j (j = E, A) under the monetary rule. Then, after a change in exchange rates, the new dollar values of net foreign assets (with values after the change denoted by primes) are

(A18) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.].

Note that the following two constraints must hold in a closed system:

(A19) [[omega].sup.U.sub.E][H.sup.U] + [[omega].sup.E.sub.E][H.sup.E] + [[omega].sup.A.sub.E][H.sup.A] = [[lambda].sup.U.sub.E][L.sub.U] + [[lambda].sup.E.sub.E][L.sub.E] + [[lambda].sup.A.sub.E][L.sub.A]

[[omega].sup.U.sub.A][H.sup.U] + [[omega].sup.E.sub.A][H.sup.E] + [[omega].sup.A.sub.A][H.sup.A] = [[lambda].sup.U.sub.A][L.sub.U] + [[lambda].sup.E.sub.A][L.sub.E] + [[lambda].sup.A.sub.A][L.sub.A]

So we can eliminate the European asset shares by writing the preceding as post-change net asset values:

(A20) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.].

We also know that

(A21) [H.sup.U] + [H.sup.E] + [H.sup.A] = [L.sup.U] + [L.sup.E] + [L.sup.A].

For our numerical findings we must posit estimated values for nominal assets and liabilities. Given the well-known measurement problems, any numbers are bound to be loose approximations at best. For the United States, the numbers we use are for end-2003 (from the 2005 Economic Report of the President) and show foreign-owned assets in the United States to be $10.5 trillion and U.S.-owned assets abroad to be $7.9 trillion. We take the current values to be $11 trillion and $8.25 trillion, respectively, for purposes of our simulations. To a first approximation, essentially all U.S. foreign liabilities are denominated in dollars, but only about 40 percent of U.S. foreign assets are. (In principle, foreign assets such as stocks and land are real, but in practice the dollar returns on these assets are highly correlated with dollar exchange rate movements.) Of the remaining 60 percent, we take 41 percent to be in European currencies and 19 percent in Asian currencies. Following Tille (2004), and including Canada, the United Kingdom, and Switzerland in region E, the United States does have a very small share of its liabilities in foreign currencies. The exact portfolio weights that we assume for the United States are

(A22) [[omega].sup.U.sub.E] = 0.405, [[omega].sup.U.sub.A] = 0.193,

[[lambda].sup.U.sub.E] = 0.03, [[lambda].sup.U.sub.A] = 0.006.

Drawing on the work of Lane and Milesi-Ferretti (but taking into account the adding-up constraints that need to hold in our theoretical model), we take Asia's assets to be $11 trillion and its liabilities to be $8.25 trillion. (55) As for portfolio shares, on the asset side, data from the International Monetary Fund's 2001 Coordinated Portfolio Investment Survey suggest that most Asian countries hold predominantly U.S. dollars (and some yen), but that Japan's foreign assets are more evenly balanced between dollar and euro holdings. If we assume that Japan owns about 40 percent of the region's gross foreign assets, we have the following approximation:

(A23) [[omega].sup.A.sub.E] = 0.2, [[omega].sup.A.sub.A] = 0.

On the liabilities side, Japan borrows in yen, but the other Asian economies have equity liabilities (including foreign direct investment) in local currencies, and extraregional debt liabilities predominantly in dollars and euros (or sterling). We assume that

(A24) [[lambda].sup.A.sub.E] = 0.33, [[lambda].sup.A.sub.A] = 0.33.

We again base our portfolio estimates for the E zone in our model on the latest data from Lane and Milesi-Ferretti, which indicate that assets and liabilities at the end of 2003 were both approximately $11 trillion. Thus we take [H.sup.E] = [L.sup.E] = $11 trillion. Most of greater Europe' s liabilities are in domestic currencies; here we assume the share is 80 percent. We take the remaining 20 percent to be entirely in U.S. dollars. On the asset side, however, we derive from equation A19 that 32 percent of Europe's holdings are in dollar assets, and 11 percent in assets denominated in Asian currencies, with the remaining 57 percent in assets denominated in European currencies. (56)

In our simulations we take [P.sub.U][Y.sup.U.sub.T] = $(11/4) trillion, based on Obstfeld and Rogoff (2000b), who argue that roughly one-quarter of U.S. GDP may be regarded as traded.

Given our assumptions on each region's gross assets and liabilities and their currencies of denomination, our analysis will also tell us how net foreign assets change across various scenarios for the current account and the exchange rate, as well as allow for the feedback effect on interest payments. We will see that, given the large size of gross stocks, large changes in exchange rates can translate into large changes in net foreign asset positions. Indeed, for many short-run and medium-run issues, knowing the gross asset and liability positions is at least as important as understanding the net positions. This conclusion is very much in line with the empirical findings of Gourinchas and Rey (2005a) for the United States.

Effects of Changing Interest Rates

It seems plausible that, in the process of U.S. current account adjustment, global interest rates will shift. Such changes could come about simply as a result of the reequilibration of the global capital market, or they could also reflect a shift in the portfolio preferences of foreign investors such that, given the exchange rate of the dollar, higher dollar interest rates are necessary to persuade them to maintain their existing dollar-denominated portfolio shares. We adopt the latter perspective, allowing the interest rate on U.S. short-term debt liabilities to rise as the dollar adjusts, without a corresponding increase in the earnings on U.S. foreign assets. Capital market shifts of this nature are likely to be quantitatively more important for the dollar than more generalized, synchronized increases in world interest rates (although the United States, as a debtor, would naturally lose while its creditors would gain).

To illustrate this channel, we first, for simplicity, abstract from the effects of nominal exchange rate changes on asset stocks (for the purpose of our simulations, this case is only a computation check). We focus on the scenario under which, as the United States adjusts, it faces a sharp increase in its borrowing rates. Thus there are two interest rates in the world economy: the rate [r.sup.U] that the United States pays on its liabilities, and the rate [r.sup.W] > [r.sup.U] that all other countries pay on their liabilities and that all countries, including the United States, earn on assets outside the United States. We focus on the implications of [r.sup.U] rising when the United States adjusts; the increase in [r.sup.U] may itself have an effect on U.S. adjustment, although that possibility does not affect our calculation.

There is also a long-run versus short-run distinction: in the short run only U.S. short-term liabilities will pay higher interest (as these are rolled over). According to U.S. Treasury data for September 2004 (from www.treas. gov/tic/debta904.html), U.S. short-term liabilities were about 30 percent of total liabilities (and thus about 30 percent of U.S. GDP). If the United States were required to pay, for example, 200 basis points more on this liability base, the result would be an additional drain of about 0.02 x 0.3 = 0.6 percent of total GDP.

Let [[omega].sup.i.sub.j] represent the share of country i gross foreign assets invested in country j.

To make the previous modeling consistent, we replace r[F.sup.i] everywhere (for the United States, Europe, and Asia, respectively) by

(A25) [r.sup.W] [H.sup.U] - [r.sup.U][L.sup.U]

[[[omega].sup.E.sub.U][r.sup.U] + (1 - [[omega].sup.E.sub.U])[r.sup.W]] [H.sup.E] - [r.sup.W][L.sup.E]

[[[omega].sup.A.sub.U][r.sup.U] + (1 - [[omega].sup.A.sub.U])[r.sup.W]] [H.sup.A] - [r.sup.W][L.sup.A].

From estimates described in the last subsection, we have the dollar values of [H.sup.i] and [L.sup.i]. Asian currency shares probably exceed the Asian country shares, because of Asian claims on offshore Eurodollars; we might assume that [[omega].sup.A.sub.U] = 0.6. Since total U.S. liabilities equal the claims on the United States of Europe and Asia,

(A26) [[omega].sup.E.sub.U][H.sup.U] + [[omega].sup.A.sub.U][H.sup.A] = [L.sup.U],

and so, with [H.sup.E], [H.sup.A], and [L.sup.U] each equal to $11 trillion, we must have [[omega].sup.E.sub.U] = 0.4.

We now turn to the calibration of interest rates (or, rather, nominal rates of return on asset and liability portfolios). We know that, for the United States currently, [r.sup.W][H.sup.U] - [r.sup.U][L.sup.U] [approximately equal to] 0. Since, also, [H.sup.U]/[L.sup.U] [approximately equal to] 0.75, [r.sup.U]/[r.sup.W] [approximately equal to] 0.75.

So we take [r.sup.U] = 3.75 percent initially, (57) but we maintain the earlier baseline assumption that [r.sup.W] = 5 percent. We ultimately wish to consider alternative increases in [r.sup.U], for example, of 125 basis points or more. These possibilities range from a scenario in which the United States simply loses its privilege of borrowing at a favorable rate, to some in which there is an element of loss of confidence in U.S. solvency absent ongoing dollar depreciation. We will also assume that only the interest rate on short-term liabilities rises in the short run. Suppose the share [sigma] of short-term liabilities in total U.S. foreign liabilities is 30 percent, or [sigma] = 0.3. Then the investment income account of the United States and the other two regions would change as follows:

(A27) [r.sup.W][H.sup.U] - [r.sup.U][L.sup.U] [right arrow] [r.sup.W] [H.sup.U] - ([r.sup.U] + [sigma][DELTA][r.sup.U])[L.sup.U]

[[[omega].sup.E.sub.U][r.sup.U] + (1 - [[omega].sup.E.sub.U])[r.sup.W]] [H.sup.E] - [r.sup.W][L.sup.E] [right arrow] [[[omega].sup.E.sub.U]([r.sup.U] + [sigma][DELTA][r.sup.U]) + (1 - [[omega].sup.E.sub.U])[r.sup.W]][H.sup.E] - [r.sup.W][L.sup.E]

[[[omega].sup.A.sub.U][r.sup.U] + (1 - [[omega].sup.A.sub.U])[r.sup.W]] [H.sup.A] - [r.sup.W][L.sup.A] [right arrow] [[[omega].sup.A.sub.U]([r.sup.U] + [sigma][DELTA][r.sup.U]) + (1 - [[omega].sup.A.sub.U])[r.sup.W]][H.sup.A] - [r.sup.W][L.sup.A].

The last two changes assume that, empirically, [[omega].sup.E.sub.U] + [[omega].sup.A.sub.U] = 1 and that Europe and Asia hold equal proportions of short-term U.S. liabilities.

One might also consider a formulation where [DELTA][r.sup.U] = f([DELTA]C[A.sup.U]), f' > 0. In this case adjustment could be quite painful if the f function is too rapidly increasing, [L.sup.U] is too big, or [sigma] is too big (or any combination of these three). We leave this possibility for future research.

Synthesis of Interest Rate Changes and Asset Revaluations

We are now ready to illustrate the techniques used to calculate the results in the third column in table 8, in which asset revaluations and interest rate changes occur simultaneously and interactively. We proceed as in the last section but add the following equations:

(A28) [H.sup.U]' = [H.sup.U] + ([E'.sub.U,E] - [E.sub.U,E]/[E.sub.U,E]) [[omega].sup.U.sub.E][H.sup.U] + ([E'.sub.U,A] - [E.sub.U,A]/[E.sub.U,A]) [[omega].sup.U.sub.A][H.sup.U]

[H.sup.E]' = [H.sup.E] + ([E'.sub.U,E] - [E.sub.U,E]/[E.sub.U,E]) [[omega].sup.E.sub.E][H.sup.E] + ([E'.sub.U,A] - [E.sub.U,A]/[E.sub.U,A]) [[omega].sup.E.sub.A][H.sup.E]

[H.sup.A]' = [H.sup.A] + ([E'.sub.U,E] - [E.sub.U,E]/[E.sub.U,E]) [[omega].sup.A.sub.E][H.sup.A] + ([E'.sub.U,A] - [E.sub.U,A]/[E.sub.U,A]) [[omega].sup.A.sub.A][H.sup.A]

and

(A29) [L.sup.U]' = [L.sup.U] + ([E'.sub.U,E] - [E.sub.U,E]/[E.sub.U,E]) [[lambda].sup.U.sub.E][L.sup.U] + ([E'.sub.U,A] - [E.sub.U,A]/[E.sub.U,A]) [[lambda].sup.U.sub.A][L.sup.U]

[L.sup.E]' = [L.sup.E] + ([E'.sub.U,E] - [E.sub.U,E]/[E.sub.U,E]) [[lambda].sup.E.sub.E][L.sup.E] + ([E'.sub.U,A] - [E.sub.U,A]/[E.sub.U,A]) [[lambda].sup.E.sub.A][L.sup.E]

[L.sup.A]' = [L.sup.A] + ([E'.sub.U,E] - [E.sub.U,E]/[E.sub.U,E]) [[lambda].sup.A.sub.E][L.sup.A] + ([E'.sub.U,A] - [E.sub.U,A]/[E.sub.U,A]) [[lambda].sup.A.sub.A][L.sup.A].

These equations, rather than the equations for net positions used in the simpler revaluation exercise in which interest rates do not change, become necessary because assets and liabilities can now pay different rates of interest and therefore must be tracked separately.

Comments and Discussion

Richard N. Cooper: This paper by Maurice Obstfeld and Kenneth Rogoff is very much a "what if" exercise. What if demand behaves according to constant elasticity of substitution functions? What if consumption is fixed, apart from terms-of-trade effects? What if the U.S. current account deficit is eliminated (or, in one of the authors' simulations, halved) by an appropriate increase in the U.S. saving rate? Then we learn from the authors' model what the implied changes in exchange rates and the terms of trade must be.

The authors' calibration is necessarily arbitrary, but it seems reasonable. A major contribution of their model is to provide a general equilibrium framework that includes two stylized regions outside the United States. It therefore permits an exploration of differing effects by region, and that seems very useful. The model also includes asset revaluation effects and not just trade effects.

The model assumes flexible prices. As the authors note, it probably understates the exchange rate changes that would be required in a sticky-price regime. The authors also usefully note the potentially ambiguous impact of faster European or Japanese economic growth, which many people see as a potential partial solution to the correction of the U.S. current account deficit. The source of the growth makes a difference, and one should not assume that more-rapid growth abroad will ease the problem. Productivity increases in traded goods, which is where such increases have typically occurred, especially in Japan, could aggravate rather than mitigate the imbalances.

No doubt it is interesting to see how large are the exchange rate changes required to close the U.S. current account deficit, but all the adjustment here is done through prices, including the asset revaluation effect. Whether that is helpful to policy is not at all clear to me. There is no explicit treatment of output or employment in the model. The simulations are driven by a postulated increase in U.S. saving and a corresponding decline in or, in most of the simulations, elimination of the U.S. current account deficit, so that presumably U.S. output is unaffected.

The paper's simulations are, however, unclear about what is happening to saving in the other two regions of the world as the U.S. external deficit is reduced. Their current account surpluses at the outset reflect excessive saving in those two regions, and they disappear in the simulation in which all three current accounts go to zero. But how do they disappear?

Consumption is assumed to be fixed, except for the terms-of-trade effect. The fall in saving in the paper's simulated Europe and Asia therefore implies a corresponding fall in output and income in order to get these results. But a decline in output will surely affect employment, hence income, hence consumption and saving, raising the question of how the initial level of consumption is sustained in Asia and Europe. That in turn leaves me wondering whether the paper provides any useful lessons for addressing the issue of global imbalances in, say, the coming decade.

Let me offer, in sketchy terms, my own view of the issue. The discussion of the U.S. current account deficit has focused largely on how much adjustment must occur in the United States. The authors' model is properly a general equilibrium model, and so it includes the rest of the world. I will focus on adjustment in the rest of the world.

In 2004 the large current account surpluses in the world were in Japan, at $172 billion, and in Germany and the Netherlands (which can be considered a satellite of the German economy) at $116 billion. Thus these three countries together account for nearly half the U.S. current account deficit. Russia and China together add another $130 billion. These are where the big numbers are. Adding up all of the rest of East Asia accounts for another $110 billion, and OPEC another $100 billion. Then there is the statistical discrepancy, which has grown to about $200 billion.

There is, as always in ex post accounting, a problem of attribution, but if we stipulate that the U.S. current account deficit is to be eliminated, we need to ask where the impact will fall in the rest of the world, and it has to fall largely where the big surpluses are. (The present surpluses of emerging economies could also shrink, or their deficits grow, but that would require a willingness on the part of savers around the world to invest in those countries on the required scale, which is not a given. If all the adjustment occurred there, their current account deficits would have to rise far in excess of generally accepted levels.)

Like the authors, I will put oil to one side. The increase in oil prices over the past two years has added roughly $100 billion to the U.S. current account deficit. I assume that, one way or another, either through a decline in oil prices or through an increase in absorption by the oil-exporting countries, their surpluses will decline significantly in the next few years. Instead I will focus on Germany and Japan, which, again, are where the really big surpluses are, and then I will comment on China.

Germany and Japan are rapidly aging, high-saving societies with limited domestic investment. Saving rates have declined in Japan, but saving in the corporate sector remains quite high. What has fallen in Japan is investment, which remains low even after a revival in the last year.

A big absorber of capital in rich countries is the residential sector. Investing in housing does not look very attractive in rapidly aging societies, with very low birth rates and low new household formation, which is the case in both of these countries. If anything, Germany and Japan have a surplus of housing in the aggregate, although it may not all be in quite the fight places. Housing construction is down essentially to replacement plus a little bit to allow for mobility. Meanwhile rates of return on industrial investment are low and, of course, very sensitive to what is happening to the export sector.

I will now make some sweeping (perhaps too sweeping) national generalizations. For reasons having to do with their defeat in World War II, a key question for the Germans and the Japanese was how to rebuild their national self-esteem. Both countries built it on export performance. That legacy continues sixty years later. The national psyche in both Germany and Japan is heavily influenced by export performance. If exports are not doing well, people feel badly about the economy and society. In my view, that influences their saving behavior. If the economy is not performing well, precautionary saving rises in these now-rich countries.

Given the aging of their society, as the Japanese have been saying for some years, Japanese saving should decline and eventually become negative. That may be so, but it has been a much slower process than the life cycle advocates forecast twenty years ago. Saving remains remarkably high given Japan's demographic structure, and the same is true of Germany.

That syndrome, in which German and Japanese saving is sensitive to perceived economic performance, which in turn is remarkably sensitive to export performance, is important when it comes to correcting the U.S. current account. If, as Obstfeld and Rogoff suggest, there will be big changes in exchange rates and big declines in the export competitiveness of key surplus countries, we are likely to see an increase, not a reduction, in the propensity to save in those countries. Whether that increase gets translated into actual additional saving depends, of course, on what happens to income. The conditions just described are, after all, the conditions under which a recession could occur. An increase in the propensity to save with no obvious vehicle for that saving leads to a fall in output and income.

In the textbooks the adjustment mechanism in this process is the interest rate, which is assumed to reconcile ex ante differences in saving and investment. Suppose the long-term nominal interest rate is only 2 percent, as it has been in Japan for several years, and not much higher in Germany. The question then becomes, What sort of investment in Japan and Germany will be stimulated by a 2 percent interest rate, given the demographics, in the presence or even with the prospect of a significantly stronger yen and euro? The sector most responsive to low interest rates in rich economies generally is the housing sector, not industrial investment. Firms will not invest in increased capacity if they see poor sales prospects, no matter how low the interest rate. Yet, for the demographic reasons already noted, demand for housing will be limited, even at low long-term interest rates.

Hence I do not see the interest rate as being an effective adjuster here. With a large appreciation of these surplus countries' currencies, the adjuster is more likely to be economic activity. Economic activity will decline, except insofar as the authorities become so concerned about it that the Europeans break all the rules they have imposed on themselves, through the Stability and Growth Pact's constraints on fiscal policy and the European Central Bank's primary focus on price stability, and pursue an aggressively stimulative policy.

I therefore see a big problem with substantial current account adjustment, mainly for Europe but also for Japan. Both already have large budget deficits: the Japanese budget deficit is roughly 7 percent of GDP, and France, Germany, and Italy, the core of Europe, have fiscal deficits expected to exceed the 3 percent limit under the Pact.

Excess saving in these big rich countries manifests itself in budget deficits and current account surpluses. Savers directly or indirectly buy claims on their governments or claims on foreigners. In my judgment, further reductions in the long-term interest rate are not going to produce enough domestic investment to substitute for those two, particularly in the face of a decline in competitiveness brought about through large appreciations of the currency.

Japan and Germany are perhaps unusual because of their peculiar history and their dependence on export performance. But I do not see how the currency appreciations that the paper simulates will produce the changes in saving required to eliminate, or even greatly reduce, the current account surpluses of Asia and Europe. A decline in Asian and European output in turn is likely to reduce the value of U.S. equity claims on those countries, weakening and possibly even reversing the valuation effect arising from dollar depreciation.

China is more complicated, and I will not discuss it in detail. Although I do not subscribe to the whole of the Dooley-Garber thesis, I am sympathetic to one of its main thrusts. China is a very-high-saving, high-investment country. I believe there is a pent-up, latent demand in China for foreign assets, which cannot be realized because violation of the foreign currency rules, especially those regarding the export of resident funds, is severely punished. China has a very weak capital market, and the central bank of China is, in effect, making the foreign investments that the public is prevented from making. Tighter fiscal policy, which might be called for at the moment on domestic grounds, would actually increase national saving in China. That moves in the wrong direction.

To conclude, I believe that the United States has comparative advantage at producing marketable assets. We sell these marketable assets to the rest of the world. As long as Americans use the proceeds of the sale of those marketable assets productively--and that is an important qualification, beating on the desirability of the fiscal deficit--I do not see why that process cannot go on indefinitely.

I am not saying the U.S. current account deficit will go on forever. My crystal ball fades rapidly after fifteen or twenty years. But, for the next decade, I do not see why the process whereby the United States generates marketable assets and sells them to foreigners who are eager to buy them cannot continue on the current scale, that is, roughly half a trillion dollars a year.

Will there be a dollar crisis? I don't have any idea. It depends in large measure on how markets react to debates such as the one we are having here. Expectations in financial markets can be very fragile. Currency markets could start to run away rapidly from the dollar. My main point is that there need not be a crisis.

T. N. Srinivasan: (1) Had Yogi Berra, the great baseball player and savant, been the discussant of this paper, he likely would have begun with his famous words, "It is deja vu all over again!" Fifteen years ago Stefan Gerlach and Peter Petri published a collection of essays with a pompous title, The Economics of the Dollar Cycle. (2) They viewed the movements of the external value of the dollar as cyclical: having appreciated by more than 40 percent between late 1979 and February 1985, the dollar had then collapsed to a new low by 1987, only to stabilize and fluctuate narrowly around the bottom of the range experienced during the 1980s. Gerlach and Petri also made the following astounding claim:

   Unlike narrowly focused studies in a technical specialty, this
   book explores the subject simultaneously from the viewpoints of
   exchange rate economics, empirical trade analysis, the economics of
   international financial markets, and macroeconomic policy-making in
   the United States, Japan, Europe, and the developing countries. (3)

Here we are, fifteen years later, exploring the same issues, except that the buzzword of that day--"newly industrializing countries" or NICs--has been replaced by another, "emerging markets." Evsey Domar, when asked by a graduate student at MIT why the questions in the macroeconomics examination do not change from year to year, is said to have replied, "Ab--but the answers do!" Domar did not claim that the answers got better over time, but one can hope that the papers in this volume will provide better answers to the same questions covered in the Gerlach-Petri volume.

In preparing this comment, I found particularly useful the contributions to that volume by my late colleague James Tobin, and the comments on Tobin's paper by his Yale student Ralph Bryant and my dear departed friend Rudiger Dornbusch. In 1988, as now, the U.S. current account was in deficit, albeit at a little more than 2 percent of GDP rather than 6 percent as now. Then as now, the United States was running a fiscal deficit of around 3.7 percent of GDP, similar to today's 4 percent. (4) In 1988 nominal interest rates in the United States had fallen from the dizzy Volckerian heights of over 19 percent in the early 1980s to a low of less than 7 percent and had again begun to rise. Nominal rates again reached a low of around 1 percent (in terms of the federal funds rate) in mid-2004 and again have begun to rise.

On the financing of the U.S. current account deficit in those days, Tobin remarked, "We are being warned incessantly that we depend on foreigners--mainly Japanese banks, insurance companies and pension funds--to buy US Treasury bonds and other dollar assets.... Should they decide not to buy dollar securities, we are told, [the result] would be calamitous." (5) Now, besides the Japanese, the financiers are the Asian central banks, particularly those of China, India, and Korea. (Even then, as Dornbusch noted in his comment, "Central banks rather than private savers have been financing the US current account.") (6) But the dire warnings are being repeated. Then, as Dornbusch put it, there was a "sharp shift in trade with the NICs. The United States has experienced a $60 billion shift in its manufactures trade with these countries since 1980." (7) Now China figures prominently in U.S. and world manufacturing trade, not to mention the prominence of India and other countries in services trade through offshoring. Related to this shift in trade was the issue of the domestic price consequences of dollar depreciation. To quote Tobin again, with any further depreciation of the dollar, "certainly imports from Japan and Europe will be more costly in dollars. So will imports from Asian 'NICs' if we induce them to let their currencies rise against the dollar." (8) Today the inducement takes the form of a demand by the secretary of the Treasury of the United States that the Chinese revalue their renminbi by at least 10 percent.

At that time some held that the dollar was correctly valued, and hence there was no need for policy-induced corrections. (9) This view was rationalized in three ways. First was the J-curve story: the response of the economy to the fall in the value of the dollar that had already taken place had yet to manifest itself fully and would surely do so soon. Second, U.S. current account deficits can be financed indefinitely by selling U.S. assets, since, in the portfolio of the rest of the world, the share of U.S. assets was probably below its long-run equilibrium value. The contemporary version of this argument holds that China (and probably other Asian countries, including India, as well) undervalues its currency to sustain its export-led growth, so as to provide employment to its huge stock of underemployed labor. Until this stock is exhausted, undervaluation of its currency may continue, with the result that any possible appreciation from capital inflows is prevented by accumulation of reserves held in dollar-denominated assets. This argument is forcefully put forward by Dooley, Folkerts-Landau, and Garber. (10) Third, the United States can wipe out the value of dollar-denominated assets held by the rest of the world through inflation. Then as now, the question was raised whether whatever transitional adjustments (policy induced or otherwise) would inevitably take place to restore the dollar to its long-run equilibrium would be orderly and smooth (a soft landing), or delayed such that, when they eventually do take place, they would be abrupt and large (a hard landing).

Brookings also held a workshop in that earlier period, in January 1987, to discuss these issues, just as it is doing now. Then, however, it assembled a group of multicountry-macroeconometric modelers and gave them alternative, but commonly specified, scenarios for the future. Starting from the common actual history of U.S. and foreign prices and exchange rates, the modelers (five in all) were asked to investigate the causes of the burgeoning external deficit of the United States during 1980-86 and to study the likely path of the deficit for the period 1987-91.

Bryant, in his comment on Tobin's paper in the Gerlach-Petri volume, reported on his updating of the conclusions of that workshop. His qualitative and quantitative findings regarding policy are not that much different from those of Obstfeld and Rogoff in the paper under discussion here, once one adjusts for differences in initial conditions. Briefly, he found that, first, in the short run (that is, until 1989), it was plausible to expect a substantial reduction in the U.S. external deficit. (11) Second, in the medium and long run (1990 and after), the prospects were much less encouraging, with the improvement in the constant-price deficit leveling off after 1989 and the current-price deficit (the deficit measured in nominal dollars) ceasing to improve well before it reached an acceptably low level. Third, a fall in the dollar's real exchange value can play a powerful role in reducing the external deficit. Fourth, the deficit in nominal dollar terms was unlikely to decline to an acceptable level ($30 billion to $40 billion, or around 0.75 percent of GDP--one-fifth of its value in 1987) without either a somewhat further depreciation of the dollar or markedly slower growth in the United States than abroad.

Bryant's own estimate of the real dollar depreciation needed to bring the deficit down to an acceptable rate was between 7 percent and 15 percent; he viewed a 20 percent depreciation as excessive. Obstfeld and Rogoff estimate the depreciation in terms of the real trade-weighted exchange rate required in order to eliminate the present U.S. deficit of about 5 percent of GDP to be between 19 and 28 percent, in a scenario in which Asia neither adjusts its exchange rate nor reduces its current account surplus (see their table 4). Bringing the deficit down to 1 percent of GDP, or one-fifth of its initial value (if one can make a linear interpolation from the authors' estimates), would call for a dollar depreciation in the range of 15 to 22 percent, very close to the estimate of 20 percent, for a similar proportional reduction of the deficit in 1987, that Bryant cited but found excessive.

Tobin's contribution to the Gerlach-Petri volume is aptly titled, "Eight Myths about the Dollar." He encountered these myths regarding what should be done to eliminate the U.S. external deficit "all too often in contemporary public and, yes, professional discussion." (12) In exploding the myths, he also provided an analytical framework for thinking about policies for eliminating the deficit. Some of the myths seem to be still going around, and, more important, Tobin's analytical framework remains relevant today. I therefore briefly summarize Tobin's contribution in the next section.

TOBIN'S ANALYTICAL FRAMEWORK AND THE EIGHT MYTHS. Unsurprisingly, Tobin found that, "just as [Hicks's] IS-LM [model], for all the hard knocks it has received from theorists, remains a good general first approximation, so its international application, Mundell-Fleming, has been a good guide." (13) In the basic, two-country version of this model, there are two goods, and each country specializes in producing and exporting part of its output of one of the goods. There are two real assets, consisting of the real money stocks of the two countries. Each country holds some of the other country's asset. The simplest way of incorporating a trade (or current account) deficit in equilibrium in this model is through a capital transfer from one country to another that allows the receiving country to spend more on the two goods than its income, by the amount of the transfer. With one country's good as the numeraire, the price of the second country's good (that is, its relative price in terms of the first country's good) is the real exchange rate in the model. Given the amount of the transfer in numeraire terms, goods market equilibrium determines the real exchange rate. The income of each country includes, besides the value of its goods output, its asset income, which in this simple framework equals the interest income on the part of its domestic real money stock held at home at its domestic interest rate and the interest income (in numeraire terms) on the foreign asset held by domestic residents at the foreign interest rate. Asset market equilibrium requires that, given the portfolio choices (in which capital transfers from one country to the other are incorporated), the demand for each country's asset equal its exogenous supply. With free capital mobility, asset market equilibrium implies that the difference between the domestic and the foreign interest rate satisfies the uncovered interest party condition: in other words, that it equals the rate of expected real depreciation. Under perfect foresight (rational expectations) the expected rate of depreciation is zero in equilibrium, so that the model solves for two prices: the real exchange rate and the common interest rate.

Obviously, this real model cannot determine the nominal exchange rate or any other nominal variables. Trivially, one could introduce nominal variables by viewing each country's asset as its nominal currency stock. By choosing units of measurement of the two goods such that the price of each country's good in its own currency is unity, nominal and real exchange rates can be made to equal each other. Any other price normalization rule will lead to a different nominal exchange rate corresponding to a given real exchange rate, but has no consequence for the determination of equilibrium real values.

Clearly, away from equilibrium and assuming away unwanted inventory accumulation (for example, assuming that the two commodities are perishable), the excess of expenditure over income by one country will equal the capital transfer from the other country ex post--this is the identity by which the current account deficit is matched by an equivalent capital inflow. However, the excess expenditure and the corresponding capital inflows are not equilibrium amounts. Thus, as Rachel McCulloch correctly pointed out, from the ex post identities that national dissaving (the excess of expenditure over income or production) equals capital surplus in the balance of payments, which in turn equals the current account deficit, no causal relationship among the variables can be inferred. (4) Put another way, no inference about the policy or behavioral changes needed to restore equilibrium can be drawn from ex post identities. Indeed, some of Tobin's eight myths illustrate this proposition very clearly.

Simply put, policies that claim to restore equilibrium in both markets by operating only in one market cannot, in general, succeed. In other words, both the real exchange rate and the interest rate will have to change to restore equilibrium. Tobin's first myth, that "eliminating the federal budget deficit will automatically eliminate the deficit in the U.S. external current account," and the second myth, its corollary, that "correction of the federal budget would solve the problem of external imbalance without further depreciation of the dollar," (5) illustrate this proposition. I cannot resist referring also to Tobin's sixth myth: "depreciation will be counterproductive for the United States because it will cause recessions in Europe and Japan and diminish their demands for US goods and services." (6) As Tobin rightly said, even an undergraduate should be ashamed to fall for the chain of arguments that lead to this myth, yet "it has been advanced with straight faces by high financial officials" (17)--a statement that remains true today. This palpably false reasoning follows, in effect, by reading causality from ex post accounting identities. Myths three, four, five, and seven illustrate other but similarly faulty reasoning. Myth eight, on the naivete of faith in macroeconomic policy coordination among major economies, lives on. At every meeting of the governors of the International Monetary Fund or the summit of the G-7 or G-8, such coordination for addressing "global imbalances" is advocated. Tobin was right: attempts at coordination on a possibly wrong policy can be very harmful, and such an event cannot be ruled out as unlikely, given the differences among analysts as to what constitutes the right policy in the first place.

THE OBSTFELD-ROGOFF CALIBRATED MODEL AND ITS SIMULATIONS. The Obstfeld-Rogoff model is, in some aspects, a more elaborate version of the basic real model described above, and less elaborate in others. Each country produces two goods, an internationally traded and a nontraded good, instead of one in the basic model; each country consumes three goods--the traded good that it produces, that produced by the other country, and, of course, its own nontraded good--instead of two. However, theirs is a static endowment model: each country has an exogenously determined endowment of its two goods; there is no production (and there are no factors of production) or investment, only consumption and trade. There are no factor or asset markets. Appearances to the contrary, in the Obstfeld-Rogoff model nominal exchange rates, the currency denomination of assets held by U.S. and foreign residents, their valuation, and interest rate effects are all add-ons: they do not form part of the behavioral specification of the model. The model determines the equilibrium real exchange rates from which Obstfeld and Rogoff then derive nominal exchange rates under alternative assumptions of the policy objectives of the central bank: to stabilize the CPI deflator, the GDP deflator, or a bilateral exchange rate. These assumptions, being unrelated to the behavioral variables of the model, do not influence its equilibrium determination. Investment decisions, interest rates, and portfolio choices obviously do not arise in an endowments model. Obstfeld and Rogoff add the return from net asset holdings at an exogenously specified interest rate to the value of endowment on the income side. Although asset valuation effects and interest rates do influence the equilibrium through their income effects, they are still add-ons, since gross values of assets and liabilities, portfolio weights, and interest rates are not endogenously determined by the model, and the fact that gross values are in nominal terms does not matter, since nominal exchange rates are mechanically linked to real rates, given the assumed behavior of the central bank.

When I say that add-on assumptions are not part of the behavioral structure model, I do not mean to imply that the results of the authors' numerical simulations, such as the magnitudes for nominal exchange rates, are not plausible. They might well be, but their plausibility or otherwise cannot be inferred from the plausibility of the assumptions alone. I will not, therefore, comment on the paper's numerical results.

Obstfeld and Rogoff simulate the impact of specified changes in the international pattern of external imbalances, those changes being the consequence of unspecified shocks to demand. In one scenario these shocks are assumed to bring down initial current account deficits and surpluses in all three regions to zero. In another the U.S. current account is set at zero, Asia pegs its nominal exchange rate to the dollar, and so on. The model then determines the comparative-static changes in equilibrium real exchange rates and other endogenous variables associated with the change in external imbalances. Since, in an endowment model, the only way to influence international balances is by affecting demand, analysis of changes in international balances brought about through shocks to supply, investment, or saving behavior is ruled out.

Policies are not explicitly modeled--they are whatever is needed to induce the unspecified shocks to demand. For example, all fiscal policy combinations that, through shocks to consumption, result in the specified changes in external imbalances are equivalent from the perspective of the model. The focus of the analysis is entirely on the real exchange rate implications of the specified changes in the pattern of initial external imbalances, and not on the policies that are behind those changes in imbalances. Those policy changes are outside the model and cannot be evaluated through the model. Thus the various policy proposals currently being made, including a policy of neglect, benign or otherwise, of the initial imbalances, cannot be evaluated. Also, the policy-relevant question of whether the prevailing U.S. current account deficit of 6 percent of GDP is indefinitely sustainable cannot be meaningfully posed, let alone answered, by the model. The implications for U.S. interest rates of a shift away from dollar-denominated assets in the portfolios of foreigners, including central banks, or a reduction in saving propensities abroad, or a rise in saving propensities in the United States, cannot be examined. Such questions as whether the end to global imbalances will come smoothly, predictably, and at a modest cost, or abruptly, unexpectedly, and at a heavy cost, cannot be analyzed either.

The analytics of the model are easily illustrated in a slightly simpler version in which there are two countries, home and foreign, each producing two traded goods. Each traded good produced by one country is a perfect substitute for the corresponding good produced by the other country. This simplification, of course, rules out home bias, as in the Obstfeld-Rogoff model, in the consumption of traded goods, and it rules out having more than one real exchange rate. However, it goes beyond the Obstfeld-Rogoff model by allowing a production (supply) response to changes in relative prices. Allowing both demand and supply responses to changes in global balances will, in general, attenuate the change in relative prices required to restore equilibrium. By distinguishing the short from the long run, such that flexibility to shift resources from producing one good to producing the other is limited in the short run relative to the long run, one can allow for possible overshooting in equilibrium relative prices in the short run without having to appeal to price rigidities. The model collapses to the endowment model if there is complete inflexibility in the short run. In a special case of the model, there is no change in the equilibrium real exchange rate from its initial value as the economy adjusts to the elimination of the trade deficit.

The model is static, and there are no investment or capital flows. The current account deficit is modeled as a pure income transfer from one country to the other. For simplicity, preferences over three commodities (the two traded goods and the nontraded good) in each country are assumed to be homothetic. For simplicity, assume that the economy running a trade deficit financed by an income transfer is a small open economy trading with a large rest of the world, so that it is a price taker in the world market for its traded goods. (This is the so-called dependent economy of W. Salter and T. Swan, (18) in which the relative price of one traded good in terms of the other is set by the world market.) This means that the two traded goods can be aggregated into a single composite traded good for the small open economy, as long as conditions in the world market do not change.

Suppose there are two factors of production (say, capital and labor), and suppose the technology of production of all three goods exhibits constant returns to scale and factor intensity (capital-labor ratio) nonreversal, so that the ranking of the goods in terms of cost-minimizing factor intensities is independent of factor prices. Then, given the relative price of traded goods, the factor prices can be uniquely solved from the two expressions equating price to unit cost, if both goods are produced in positive amounts. Given the unique factor prices, the minimum unit cost of production of the nontraded good and hence its price, given that a positive amount of it is produced, are determined. Suppose the factor endowments are such that the production possibility frontier (PPF) includes a nonempty subset S in which all three goods are produced in positive amounts. In S the marginal rate of transformation between the traded composite good (T) and the nontraded good is constant.

In figure 1 above, the PPF is depicted as ABC, where the linear stretch AB (except at points A and B) corresponds to S. The consumption possibility frontier (CPF), given a transfer AD (in units of the traded composite good) from the rest of the world, is depicted by DEFC. It is simply a vertical shift of the PPF by the distance AD. Assume that preferences are represented by a homothetic, quasi-concave utility function. Maximizing utility subject to the CPF leads to the initial equilibrium consumption at [Q.sub.0] and production at [P.sub.0] vertically below it. At [Q.sub.0] an indifference curve touches the PPF so that the common slope of the two is the slope of DE, which equals the relative price of the traded composite good, that is, the real exchange rate.

Suppose now that the transfer is withdrawn so that the CPF coincides with the PPF. If the real exchange rate does not change, by virtue of homothetic preferences, consumption shifts to [Q'.sub.0] on AB, where the ray O[Q.sub.0] from the origin intersects AB. If production remains at P0, there will be an excess supply of the nontraded good. However, since the marginal rate of transformation between traded and nontraded goods is constant along AB, the production point shifts to [Q'.sub.0] and the excess supply is eliminated. Thus the economy adjusts to the elimination of the trade deficit by a pure quantity adjustment with no change in the real exchange rate.

One could depict an endowment economy by reinterpreting the PPF as just a single point [P.sub.0]. With income transfer [P.sub.0][Q.sub.0], the economy consumes at [Q.sub.0]. Let DE be the tangent to the indifference curve at [Q.sub.0], so that it is the real exchange rate from a consumption perspective. The withdrawal of the transfers requires that the consumption point move to [P.sub.0]. But if the consumption real exchange rate does change, it will move to [Q'.sub.0], once again creating an excess demand for traded goods. To eliminate this excess demand, the real exchange rate for consumption has to change to the slope of the indifference curve through [P.sub.0]. This means (under the standard assumption of convexity of preferences and both goods being normal) there has to be a rise in the relative price of the traded good in terms of the nontraded good, or a real depreciation, since at [P.sub.0] and [Q.sub.0] the consumption of the nontraded good is the same, whereas that of the traded good composite is lower at [P.sub.0].

Returning to the production economy, what if the initial consumption point is on EF, such as [Q.sub.1]? By construction, because each point on EF is at the same vertical distance (of AD) above the point on the stretch BC vertically below it, the slope of the CPF at [Q.sub.1] is the same as the slope of the PPF at [P.sub.1], and their common slope is equal to the equilibrium real exchange rate from both a consumption and a production perspective. Now, with the withdrawal of the income transfer, the CPF coincides with the PPF, and production will remain at [P.sub.1]. At unchanged real exchange rates, consumption will move to a point (not shown) to the left of [P.sub.1] on the straight line that is tangent to the PPF at [P.sub.1], thus creating excess demand for traded goods. To eliminate this excess demand, a real depreciation has to occur, with the new equilibrium point lying to the left of [P.sub.1] on the PPF, where an indifference curve touches the PPF (not shown).

Short-run inflexibility and long-run flexibility in shifting resources starting from the production point [P.sub.1] can be easily illustrated. The short-run PPF touches the long-run PPF at the initial production point PI but is below it otherwise, with the vertical distance between the two increasing as the production of the nontraded good increasingly deviates from its level at [P.sub.1] in either direction. Under these assumptions, which seem natural for depicting short-run inflexibility, it is clear that the short-run equilibrium point where an indifference curve touches the short-run PPF will imply a larger depreciation in terms of the real exchange rate than its long-run value at the point where an indifference curve touches the long-run PPF. In other words there is overshooting of the real exchange rate in the short run. If we add on a dynamic adjustment of the short-run PPF to the long-run PPF over time, it follows that, after overshooting, the real exchange rate will converge to its long-run value.

The essential features of the adjustment will remain in its extension to a multicountry general equilibrium setup. However, its diagrammatic exposition will not. The reason is that the convenient device of a Hicksian composite traded good depends on the relative price of traded goods not changing. This cannot hold in general in the general equilibrium setup, because the relative prices are endogenous. Although the Obstfeld-Rogoff model is a three-country general equilibrium model, it replicates the essential qualitative conclusion of adjustment in a small open economy, namely, that a real depreciation is generally (though not necessarily always) needed to eliminate global imbalances.

For their purpose, which is to arrive at a quantitative estimate of the extent of the real depreciation needed to eliminate global imbalances, Obstfeld and Rogoff have had to calibrate their general equilibrium model. Let me therefore conclude with a couple of comments on the calibration. (19) Obstfeld and Rogoff relate their choice of values for the two crucial parameters, [theta] (the elasticity of substitution in consumption between the traded aggregate and the nontraded good) and [eta] (the so-called Armington elasticity of substitution between the domestic and foreign traded goods in the traded goods aggregate), to econometric estimates in the literature. There are several problems with this procedure. First, although [theta] is arguably a "deep" parameter in the Lucas sense, since it relates to preferences, [eta] is not. As such, any estimate of [eta] will depend on the trade policy regime and therefore cannot be stably estimated econometrically. Second, setting aside the policy dependence of parameter values, since the Obstfeld-Rogoff model involves aggregates, alternative schemes of aggregation will influence parameter values, and whether the estimates in the literature are all comparable and correspond to the implied aggregation of the Obstfeld-Rogoff model is not obvious. To be fair, the authors are certainly aware of these issues, and their simulations cover a range of values for the two parameters. Perhaps they should cover an even broader range of values, particularly for [eta].

General discussion: William Nordhaus agreed with the authors that productivity's effect on the trade deficit depends critically on whether the change in productivity occurs in the traded goods or the nontraded goods sector. He reported his own recent findings showing that, whereas the productivity slowdown in the United States during the 1970s had occurred in both sectors, the acceleration of the 1990s was mainly in traded goods. Although some estimates attribute almost all the acceleration to computers and associated industries, Nordhaus estimated that only somewhere between a half and two-thirds came from that source. The most recent data also suggest that productivity has accelerated, but he cautioned that this may in part reflect mismeasurement of productivity in the retail sector. Noting that Jack Triplett had found the United States to be on the frontier of improved measurement techniques that have tended to raise estimates of productivity growth, Nordhaus speculated that Europe's productivity performance might look more like the United States' if the same measurement techniques were used for both.

Gian Maria Milesi-Ferretti noted that the April 2005 issue of the International Monetary Fund's World Economic Outlook (WEO) contains a paper using a four-region global economic model that is similar to the authors' three-region model but allows for changes in production. As one would expect, the WEO model finds that allowing for a production response leads to a smaller, but still quite substantial, real depreciation of the dollar. Sebastian Edwards observed that, in the authors' Bretton Woods II scenario, Asia's surplus increases from 15 percent of U.S. traded-goods GDP to 25 percent. The authors' real model is unable to consider the monetary consequences of this increase, but Edwards suggested that in the real world it would create enormous pressure to expand the money supply in China and the other Asian countries, requiring an extraordinary amount of sterilization to avoid inflation. Indeed, the latest data already show an increase in inflation in China. Edwards also observed that the results of the authors' global rebalancing scenario do not differ significantly from those of an earlier two-region model of theirs; the introduction of the third region does not appear to make a significant difference to the results.

Edmund Phelps reminded the panel of a paper he had co-written in 1986, which argued that the expansionary U.S. fiscal policy of that era would result in a boom in the United States while causing world real interest rates to increase, leading to a recession in Europe. This analysis suggests that if the United States were now to adopt fiscal austerity, world real interest rates would decrease. This in turn would likely lead to an increase in asset values in Europe and Asia, and thus an increase, rather than a decrease as Richard Cooper had predicted in his comment, in output in those countries. In the United States the shadow prices of business assets would fall, causing a decline in production and investment.

Peter Garber argued that although Japan's and Germany's populations are aging, and their populations growing slowly, there is significant underemployment in their economies, especially in the nontradable services industries. Underemployment is the reason that Japan has dramatically increased its monetary base in order to maintain a high yen-dollar exchange rate. Garber suggested that Germany is facing the same problem but is incapable of making a similar intervention, and its difficulties are exerting pressure on the European Central Bank to lower interest rates.

(1.) I thank Benjamin Friedman and Robert Solow for their comments.

(2.) Interestingly, some of the contributors to the Gerlach-Petri volume were participants at this Brookings Panel meeting.

(3.) Gerlach and Petri (1990, p.2).

(4.) See the authors' figure 4.

(5.) Tobin (1990, p. 34).

(6.) Dornbusch (1990, p. 53).

(7.) Dornbusch (1990, p. 53).

(8.) Tobin (1990, p. 30, emphasis added).

(9.) In the Gerlach-Petri volume, these views are summarized by Dornbusch (1990, pp. 53-54).

(10.) Dooley, Folkerts-Landau, and Garber (2004a, 2004b).

(11.) Bryant (1990, pp. 42-43).

(12.) Tobin (1990, p. 28).

(13.) Tobin (1990, p. 28).

(14.) McCulloch (1990).

(15.) Tobin (1990, pp. 28-29).

(16.) Tobin (1990, p. 33).

(17.) Tobin (1990, p. 33).

(18.) Salter (1959); Swan (1963).

(19.) For a more detailed discussion, see Dawkins. Srinivasan, and Whalley (2001).

References

Broda, Christian, and David E. Weinstein. 2004. "Globalization and the Gains from Variety." Working Paper 10314. Cambridge, Mass.: National Bureau of Economic Research (February).

Bryant, Ralph C. 1990. "Comment on 'Eight Myths About the Dollar.' " In The Economics of the Dollar Cycle, edited by Stefan Gerlach and Peter A. Petri. MIT Press.

Calvo, Guillermo A. 1998. "Capital Flows and Capital-Market Crises: The Simple Economics of Sudden Stops." Journal of Applied Economics 1: 35-54.

Campa, Jose Manuel, and Linda S. Goldberg. 2002. "Exchange Rate Pass-Through into Import Prices: A Macro or a Micro Phenomenon?" Working Paper 8934. Cambridge, Mass.: National Bureau of Economic Research (May).

Cline, William R. Forthcoming. The United States as a Debtor Nation: Risks and Policy Reform. Washington: Institute for International Economics.

Cooper, Richard N. 2001. "Is the U.S. Current Account Sustainable? Will It Be Sustained?" BPEA, no. 1: 217-26.

Croke, Hilary, Steven B. Kamin, and Sylvain Leduc. 2005. "Financial Market Developments and Economic Activity during Current Account Adjustments in Industrial Economies." International Finance Discussion Papers 827. Washington: Board of Governors of the Federal Reserve System (February).

Dawkins, Christina, T. N. Srinivasan, and John Whalley. 2001. "Calibration." In Handbook of Econometrics, vol. 5, edited by James J. Heckman and Edward Leamer. Amsterdam: Elsevier.

Dooley, Michael P., David Folkerts-Landau, and Peter M. Garber. 2004a. "The Revived Bretton Woods System: The Effects of Periphery Intervention and Reserve Management on Interest Rates and Exchange Rates in Center Countries." Working Paper 10332. Cambridge, Mass.: National Bureau of Economic Research (March).

--. 2004b. "Direct Investment, Rising Real Wages, and the Absorption of Excess Labor in the Periphery." Working Paper 10626. Cambridge, Mass.: National Bureau of Economic Research (September).

Dornbusch, Rudiger. 1976. "Expectations and Exchange Rate Dynamics." Journal of Political Economy 84, no. 6:1161-76.

--. 1990. Comment on Tobin, "Eight Myths about the Dollar." In The Economics of the Dollar Cycle, edited by Stefan Gerlach and Peter A. Petri. MIT Press.

Edwards, Sebastian. 2004. "Thirty Years of Current Account Imbalances, Current Account Reversals, and Sudden Stops." International Monetary Fund Staff Papers 51 (special issue): 1-49.

Feenstra, Robert C. 1994. "New Product Varieties and the Measurement of International Prices." American Economic Review 84, no. 1: 157-77.

Frankel, Jeffrey A., and Andrew K. Rose. 1995. "Empirical Research on Nominal Exchange Rates." In Handbook of International Economics, vol. 3, edited by Gene M. Grossman and Kenneth Rogoff. Amsterdam: Elsevier.

Freund, Caroline, and Frank Warnock. 2005. "Current Account Deficits in Industrial Countries: The Bigger They Are, The Harder They Fall?" World Bank and Darden Graduate School of Business, University of Virginia (May).

Gagnon, Joseph E. 2003. "Productive Capacity, Product Varieties, and the Elasticities Approach to the Trade Balance." International Finance Discussion Paper 781. Washington: Board of Governors of the Federal Reserve System (October).

Gerlach, Stefan, and Peter A. Petri. 1990. The Economics of the Dollar Cycle. MIT Press.

Goldberg, Pinelopi Koujianou, and Michael M. Knetter. 1997. "Goods Prices and Exchange Rates: What Have We Learned?" Journal of Economic Literature 35, no. 3: 1243-72.

Gourinchas, Pierre-Olivier, and Helene Rey. 2005a. "International Financial Adjustment." Working Paper 11155. Cambridge, Mass.: National Bureau of Economic Research (February).

--. 2005b. "U.S. External Adjustment: The Exorbitant Privilege." University of California, Berkeley, and Princeton University (April).

Greenspan, Alan. 2004. "The Evolving U.S. Payments Imbalance and Its Impact on Europe and the Rest of the World." Cato Journal 24 (Spring-Summer): 1-11.

Grubert, Harry, Timothy Goodspeed, and Deborah Swenson. 1993. "Explaining the Low Taxable Income of Foreign-Controlled Companies in the United States." In Studies in International Taxation, edited by Alberto Giovannini, R. Glenn Hubbard, and Joel Slemrod. University of Chicago Press.

Harris, David, Randall Morck, Joel Slemrod, and Bernard Yeung. 1993. "Income Shifting in U.S. Multinational Corporations." In Studies in International Taxation, edited by Alberto Giovannini, R. Glenn Hubbard, and Joel Slemrod. University of Chicago Press.

Hooper, Peter, Karen Johnson, and Jaime Marquez. 2000. Trade Elasticities for the G-7 Countries. Princeton Studies in International Economics 87. Princeton University (August).

Hooper, Peter, and John Morton. 1982. "Fluctuations in the Dollar: A Model of Nominal and Real Exchange Rate Determination." Journal of International Money and Finance 1, no. 1: 39-56.

Krugman, Paul R. 1985. "Is the Strong Dollar Sustainable?" In The U.S. Dollar--Recent Developments, Outlook, and Policy Options. Kansas City, Mo.: Federal Reserve Bank of Kansas City.

--. 1989. "Differences in Income Elasticities and Trends in Real Exchange Rates." European Economic Review 33, no. 5:1031-54.

--. 1991. Has the Adjustment Process Worked? Policy Analyses in International Economics 34. Washington: Institute for International Economics.

Lane, Philip R., and Gian Maria Milesi-Ferretti. 2004. "The Transfer Problem Revisited: Net Foreign Assets and Real Exchange Rates." Review of Economics and Statistics 86, no. 4: 841-57.

--.2005a. "Financial Globalization and Exchange Rates." Working Paper 05/3 (January). Washington: International Monetary Fund.

--. 2005b. "A Global Perspective on External Positions." Trinity College, Dublin, and International Monetary Fund (May).

--. Forthcoming. "The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 1970-2003." Trinity College, Dublin, and International Monetary Fund.

Mann, Catherine L. 2005. "Breaking Up Is Hard to Do: Global Co-Dependency, Collective Action, and the Challenges of Global Adjustment." CESifo Forum 6, no. 1: 16-23.

Marazzi, Mario, and others. 2005. "Exchange Rate Pass-through to U.S. Import Prices: Some New Evidence." International Finance Discussion Paper 2005-833. Federal Reserve Board (April).

McCulloch, Rachel. 1990. "Macroeconomic Policy, Trade and the Dollar: Remarkable Events of the 1980s." In The Economics of the Dollar Cycle, edited by Stefan Gerlach and Peter A. Petri. MIT Press.

Meese, Richard, and Kenneth Rogoff. 1983. "Empirical Exchange Rate Models of the Seventies: Do They Fit Out of Sample?" Journal of International Economics 14 (February): 3-24.

Mendoza, Enrique G. 1991. "Real Business Cycles in a Small Open Economy." American Economic Review 81, no. 4: 797-818.

Mussa, Michael. 2005. "Sustaining Global Growth while Reducing External Imbalances." In The United States and the World Economy: Foreign and Economic Policy for the Next Decade, edited by C. Fred Bergsten. Washington: Institute for International Economics.

Obstfeld, Maurice. 2004. "External Adjustment." Review of World Economics 140, no. 4: 541-68.

Obstfeld, Maurice, and Kenneth Rogoff. 1996. Foundations of International Macroeconomics. MIT Press.

--. 2000a. "Perspectives on OECD Economic Integration: Implications for U.S. Current Account Adjustment." In Global Economic Integration: Opportunities and Challenges. Kansas City, Mo.: Federal Reserve Bank of Kansas City (August).

--. 2000b. "The Six Major Puzzles in International Macroeconomics: Is There a Common Cause?" In NBER Macroeconomics Annual 2000, edited by Ben S. Bernanke and Kenneth Rogoff. MIT Press.

--. 2004. "The Unsustainable US Current Account Position Revisited." Working Paper 10869. Cambridge, Mass.: National Bureau of Economic Research (November).

Obstfeld, Maurice, and Alan M. Taylor. 1997. "Nonlinear Aspects of Goods-Market Arbitrage and Adjustment: Heckscher's Commodity Points Revisited." Journal of the Japanese and International Economies 11, no. 4: 441-79.

--. Forthcoming. "Sources of America's 'Exorbitant Privilege.'" University of California.

Orcutt, Guy H. 1950. "Measurement of Price Elasticities in International Trade." Review of Economics and Statistics 32, no. 2:117-32.

Ostry, Jonathan D., and Carmen M. Reinhart. 1992. "Private Saving and Terms of Trade Shocks: Evidence from Developing Countries." International Monetary Fund Staff Papers 39, no. 3: 495-517.

Porter, Richard D., and Ruth A. Judson. 1996. "The Location of U.S. Currency: How Much Is Abroad?" Federal Reserve Bulletin 82 (October): 883-903.

Reifschneider, David, Robert Tetlow, and John Williams. 1999. "Aggregate Disturbances, Monetary Policy, and the Macroeconomy: The FRB/US Perspective." Federal Reserve Bulletin 85, no. 1: 1-19.

Roubini, Nouriel, and Brad Setser. 2004. "The U.S. as a Net Debtor: The Sustainability of U.S. External Imbalances." Stern School of Business, New York University, and University College, Oxford (August).

Ruhl, Kim. 2003. "Solving the Elasticity Puzzle in International Economics." Federal Reserve Bank of Minneapolis (November).

Salter, W. E. G. 1959. "Internal and External Balance--The Role of Price and Expenditure Effects." Economic Record 35: 226-38.

Stockman, Alan C., and Linda L. Tesar. 1995. "Tastes and Technology in a TwoCountry Model of the Business Cycle: Explaining International Comovements." American Economic Review 85, no. 1: 168-85.

Swan, T. 1963. "Longer Run Problems of the Balance of Payments." In The Australian Economy, edited by H.W. Arndt and W.M. Corden. Melbourne: Cheshire.

Tille, Cedric. 2004. "Financial Integration and the Wealth Effect of Exchange Rate Fluctuations." Federal Reserve Bank of New York (August).

Tobin, James. 1990. "Eight Myths about the Dollar." In The Economics of the Dollar Cycle, edited by Stefan Gerlach and Peter A. Petri. MIT Press.

Ventura, Jaume. 2001. "A Portfolio View of the U.S. Current Account Deficit." BPEA, no. 1: 241-53.

Warnock, Francis E. 2003. "Exchange Rate Dynamics and the Welfare Effects of Monetary Policy in a Two-Country Model with Home-Product Bias." Journal of International Money and Finance 22, no. 3 (June): 343-63.

Eyal Dvir, Jose Antonio Rodriguez-Lopez, and Jon Steinsson provided dedicated and excellent research assistance, for which we are extremely grateful. We also thank Philip Lane and Gian Maria Milesi-Ferretti for discussions and data. Jane Trahan's technical support was outstanding.

(1.) See Obstfeld and Rogoff (2000a, 2004). From an accounting perspective, a country's current account balance essentially adds net interest and dividend payments to its trade balance. As we discuss below, the United States presently receives about the same amount of income on its foreign assets as it pays out to foreign creditors. Hence, for the United States (and indeed many countries), the current account balance and the trade balance are quantitatively very similar. As we later emphasize, however, the current account does not include capital gains and losses on existing wealth. Thus the overall change in a country's net foreign asset position can, in principle, be less than or greater than its current account deficit or surplus.

(2.) Calculated from the World Economic Outlook database of the International Monetary Fund, using current account data from 2004.

(3.) Roubini and Setser (2004).

(4.) Blanchard, Giavazzi, and Sa (this volume).

(5.) Cline (forthcoming). Mann (2005), although not alarmist, also points to risks in the adjustment process. Of course, similar discussions accompanied earlier U.S. adjustment episodes, but the present situation is quite different in both scale and setting. Krugman (1985, 1991) takes as dim a view as anyone on the sustainability of long-term twin (fiscal and current account) deficits. His views on the 1980s experience would seem to apply with even greater force to the current scene.

(6.) Greenspan (2004).

(7.) Obstfeld and Rogoff (2000a, 2000b).

(8.) Croke, Kamin, and Leduc (2005). Freund and Warnock (2005) survey current account adjustment in industrial countries and find that deficits tend to be associated with real depreciations, which are larger for consumption-driven deficits.

(9.) See especially Lane and Milesi-Ferretti (2005a, 2005b). In line with this development, Cooper (2001) identifies ongoing international portfolio diversification as a driving force behind the U.S. deficit. Diversification does not, however, require any net capital flows: even with a balanced current account, foreigners and U.S. residents can still swap assets. According to preliminary estimates by the Bureau of Economic Analysis, for example, private foreign investors added $1.1 trillion in U.S. assets to their portfolios in 2004, far more than that year's U.S. current account deficit of $666 billion.

(10.) Tille (2004); Gourinchas and Rey (2005a, 2005b).

(11.) Obviously, if one measures saving taking into account capital gains and losses on wealth, the trend decline in saving is much less, although housing wealth is largely not internationally tradable and both housing and securities wealth can evaporate quickly.

(12.) Ventura (2001).

(13.) See also Lane and Milesi-Ferretti (2005a, 2005b) and Obstfeld (2004). The BEA applies market valuation to foreign direct investment holdings starting only in 1982. Gourinchas and Rey (2005b) construct U.S. international position data going back to 1952. In 1976, with foreign direct investment valued at current cost rather than at market value, U.S. gross foreign assets amounted to 25 percent of GDP, and gross foreign liabilities were 12.6 percent of GDP.

(14.) Gourinchas and Rey (2005b) present a similar graph covering a much longer period. The estimates in the text are consistent with those found by Obstfeld and Taylor (forthcoming) using a different methodology. For a complementary discussion of returns on foreign assets and liabilities, see Lane and Milesi-Ferretti (2005b).

(15.) The broad rate-of-return measures for gross assets and liabilities are constructed by adding to the investment income flow the total capital gain on the previous end-of-period assets (or liabilities) and then dividing this total return by the previous value of assets (or liabilities). Thus, in 2003, a year in which the dollar depreciated, the rate of return on U.S. foreign assets was 19 percent, and that on liabilities 8.4 percent. Total capital gains are calculated by subtracting the change in U.S.-owned assets abroad (change in foreign-owned assets in the United States), as reported in the financial account, from the change in U.S. foreign assets (liabilities) at market value, as reported in the BEA international position data.

(16.) See the survey in Central Banking, "The Rise of Reserve Management," March 2005, p. 14.

(17.) See Porter and Judson (1996).

(18.) Of course, multinationals' practice of income shifting in response to differing national tax rates on profits distorts reported investment income flows, making an accurate picture of the true flows difficult to obtain. See, for example, Grubert, Goodspeed, and Swenson (1993) and Harris and others (1993). The expansion of gross international positions over the past decade may have worsened this problem.

(19.) Obstfeld and Rogoff (2000a).

(20.) See Obstfeld and Rogoff (2000a and 2004, respectively).

(21.) Obstfeld and Rogoff (1996).

(22.) Warnock (2003) also takes this approach.

(23.) Obstfeld and Rogoff (2000b).

(24.) See, for example, Obstfeld and Rogoff (1996).

(25.) The methodology is specified in appendix A and further online at www.economics. harvard.edu/faculty/rogoff/papers/BPEA2005.pdf.

(26.) Details can be found in appendix A and online at www.economics.harvard.edu/ faculty/rogoff/papers/BPEA2005.pdf.

(27.) As noted above, this effect has recently been emphasized by Tille (2004), Lane and Milesi-Ferretti (2005a, 2005b), and Gourinchas and Rey (2005a, 2005b).

(28.) Mendoza's (1991) point estimate is 0.74, Ostry and Reinhart (1992) report estimates in the range 0.66 to 1.28 for a sample of developing countries, and Stockman and Tesar (1995) use an estimate of 0.44. Using a different approach, Lane and Milesi-Ferretti (2004) derive estimates as low as 0.5. Indeed, for larger and relatively closed economies (such as the United States, Europe, and Japan), they suggest that the value should be even lower.

(29.) Examples are the estimates of Feenstra (1994) and the more recent figures of Broda and Weinstein (2004).

(30.) Orcutt (1950).

(31.) For an excellent example of this bias in action, see Hooper, Johnson, and Marquez (2000), who report that, because oil and tourism demand are relatively price-inelastic, trade equations based on aggregates that include oil and services imply apparently much lower price elasticities than equations for nonoil manufactures only. For the Group of Seven countries, Hooper, Johnson, and Marquez report short-run price elasticities for imports and exports (including oil and services) that in most cases do not satisfy the Marshall-Lerner condition. We view the elasticities implied even by aggregated estimates that exclude oil and services as unreasonably low; but, if they are accurate, they imply larger terms-of-trade and real exchange rate effects of international spending shifts.

(32.) See Ruhl (2003). Our model omits not only dynamics of the type suggested by Ruhl, but also those resulting from the introduction of new product varieties, which would act over the longer run to dampen the extent to which a rise in a country's relative productivity lowers its terms of trade. See, for example, Krugman (1989) and Gagnon (2003).

(33.) Obstfeld and Rogoff (2000a).

(34.) Obstfeld and Rogoff (2000b).

(35.) The phrase "exorbitant privilege" is commonly but wrongly attributed to French president Charles de Gaulle. For its true origin, see the interesting historical note provided by Gourinchas and Rey (2005b).

(36.) As noted earlier, we estimate tradable GDP to be at most 25 percent of total GDP.

(37.) It would be interesting and useful to extend the model to include emerging markets and OPEC as a composite fourth region, as suggested by our discussant T. N. Srinivasan.

(38.) Mussa (2005).

(39.) We provide a detailed account of nominal exchange rate determination under GDP deflator targeting at www.economics.harvard.edu/faculty/rogoff/ papers/BPEA2005.pdf.

(40.) Indeed, if one recalibrates the model so that [beta] = 0.85 (in which case all countries' preferences are completely symmetric, so that Europeans and Americans no longer prefer Asian goods to each other's), then, in the global rebalancing scenario, Asia's currency appreciates in real terms against the dollar by 37.8 percent and against European currencies by 12.2 percent. These numbers exceed the 35.2 percent and 6.7 percent reported in table 3.

(41.) Dooley, Folkerts-Landau, and Garber (2004a, 2000b).

(42.) Gourinchas and Rey (2005a).

(43.) For an example of a dynamic approach see the small-country q-model analysis in Obstfeld and Rogoff (1996, chapter 4).

(44.) See the discussion in Obstfeld and Rogoff (2000a).

(45.) P. Goldberg and Knener (1997); Campa and L. Goldberg (2002). For recent evidence suggesting a substantial decline in pass-through to U.S. import prices, see Marazzi and others (2005).

(46.) Reifschneider, Tetlow, and Williams (1999). A back-of-the-envelope calculation based on the Dornbusch overshooting model (Dornbusch, 1976) yields a similar result.

(47.) Meese and Rogoff (1983).

(48.) See the survey in Frankel and Rose (1995), for example.

(49.) Calvo (1998).

(50.) Obstfeld and Taylor (1997).

(51.) Gourinchas and Rey (2005a).

(52.) Hooper and Morton (1982).

(53.) As illustrated, for example, in Obstfeld and Rogoff (1996).

(54.) Details can be found online at www.economics.harvard.edu/faculty/ rogoff/papers/BPEA2005.pdf.

(55.) Lane and Milesi-Ferretti (forthcoming).

(56.) The European position assumptions are not needed to implement equation A20, but they are necessary for assessing the effects of interest rate changes below.

(57.) This number is in line with the estimate given above of the excess return of U.S. foreign assets over U.S. liabilities to foreigners.

MAURICE OBSTFELD

University of California, Berkeley

KENNETH S. ROGOFF

Harvard University

Table 1. International Investment Positions of Selected Industrial
Countries, 2003

Percent of GDP (a)

Country          Gross assets   Gross liabilities   Net position

Canada                75                93               -18
Euro area            107               118               -10
France               179               172                 7
Germany              148               141                 6
Italy                 95               100                -5
Japan                 87                48                39
Switzerland          503               367               135
United Kingdom       326               329                -2

Source: International Monetary Fund. International Financial
Statistics.

(a.) Gross assets may differ from the sum of gross liabilities and
the net position because of rounding.

Table 2. Recent and Two-Year-Average Exchange Rates of Selected
Currencies

Currency units per dollar except where noted otherwise

                                         Exchange rate

Currency                   As of June 1, 2005    Two-year average

U.K. pound sterling (a)             1.81                 1.79
Canadian dollar                     1.25                 1.23
Euro (a)                            1.22                 1.23
Korean won                      1,010                1,129
New Taiwan dollar                  31.30                33.21
Singapore dollar                    1.67                 1.69
Japanese yen                      108.4                109.3

Source: Federal Reserve data.

(a.) In dollars per indicated currency unit.

Table 3. Changes in Real Exchange Rates and Terms of Trade Following
U.S. Current Account Adjustment under Baseline Assumptions (a)

Log change x 100

                                      Adjustment scenario

                                                            Europe and
                                                              United
Real exchange rate          Global           Bretton       States trade
or terms of trade       rebalancing (b)    Woods II (c)     places (d)

Real exchange rate
United States/Europe         28.6              58.5            44.6
United States/Asia           35.2              -0.5            19.4
Europe/Asia                   6.7             -59.0           -25.2
Terms of trade
United States/Europe         14.0              29.4            22.0
United States/Asia           14.5               7.2            11.1
Europe/Asia                   0.5             -22.2           -10.8

Source: Federal Reserve data.

(a.) Exchange rates are defined such that an increase represents a real
depreciation of the first region's currency against the second's; terms
of trade are defined such that an increase represents a deterioration
for the first region (that is, a fall in the price of the first
region's export good against the second). Assumed parameter values are
as follows; substitution elasticity between traded and nontraded goods
[theta] = 1; substitution elasticity between traded goods of different
regions [eta] = 2; share of traded goods in total consumption
[gamma] = 0.25.

(b.) Current account balances of all three regions go to zero.

(c.) Asia's current account surplus rises to keep its exchange rate
with the dollar fixed. Europe's current account absorbs all changes in
the U.S. and Asian current accounts.

(d.) Europe absorbs the entire improvement in the U.S. current account
balance while Asia's current account balance remains unchanged.

Table 4. Changes in Nominal Exchange Rates Following U.S. Current
Account Adjustment under Alternative Inflation Target (a)

Log change x 100

                                   Adjustment scenario

Nominal                   Global       Bretton      Europe and United
exchange rate           rebalancing    Woods II    States trade places

Target is consumer
  price index (b)
United States/Europe       28.6          58.5              44.6
United States/Asia         35.2          -0.5              19.4
Europe/Asia                 6.7         -59.0             -25.2
Target is GDP
  deflator
United States/Europe       30.0          61.4              46.8
United States/Asia         36.9           0.0              20.6
Europe/Asia                 6.9         -61.4             -26.3

Source: Federal Reserve data.

(a.) See table 3 for definitions of exchange rates. scenarios, and
parameter assumptions.

(b.) With flexible prices and CPI targeting by central banks, nominal
exchange rate changes are equal to the real exchange rate changes
reported in table 3.

Table 5. Changes in Real and Nominal Effective (Trade-Weighted)
Exchange Rates Following U.S. Current Account Adjustment under Baseline
Assumptions (a)

Log change x 100

                               Adjustment scenario

Effective              Global       Bretton      Europe and United
exchange rate (b)    rebalancing    Woods II    States trade places

U.S. real              -33.0         -19.1             -27.8
U.S. nominal           -34.6         -20.5             -29.3
Europe real              5.1          58.9              31.7
Europe nominal           5.4          61.4              33.1
Asia real               20.9         -29.8              -2.9
Asia nominal            21.9         -30.7              -2.9

(a.) See table 3 for definitions of scenarios and parameter
assumptions. An increase is an appreciation of the indicated currency
against foreign currencies.

(b.) Nominal exchange rate changes are calculated under the assumption
of GDP deflator targeting; see appendix A for details.

Source: Authors' calculations using model described in the text.

Table 6. Net Foreign Assets by Region Following U.S. Current Account
Adjustment (a)

Ratio to value of U.S. traded goods output

                                         Adjustment scenario (b)

                                                               Europe
                                                             and United
                 Baseline net                                  States
                 foreign asset      Global       Bretton       trade
Region             position       rebalancing    Woods II      places

United States        -1.0            -0.3          -0.1         -0.2
Europe                0.0            -0.1          -0.7         -0.4
Asia                  1.0             0.4           0.8          0.6

Source: Federal Reserve data.

(a.) See table 3 for definitions of exchange rates, scenarios, and
parameter assumptions.

(b.) Net asset positions taking into account valuation effects of
changes in nominal exchange rates under GDP deflator targeting.

Table 7. Changes in Real Exchange Rates and Terms of Trade in the
Global Rebalancing Scenario under Alternative Calibrations (a)

Log change x 100

                                                 Verv high elasticity
                        Higher elasticity of       of substitution
                        substitution between       between regions'
                        traded and nontraded         traded goods
Real exchange rate       goods ([theta] = 2,        ([theta] = 1,
or terms of trade            [eta] = 2)              [eta] = 100)

Real exchange rate
United States/Europe            19.3                     16.5
United States/Asia              22.5                     23.5
Europe/Asia                      3.3                      7.0
Terms of trade
United States/Europe            14.6                      0.0
United States/Asia              15.1                      0.0
Europe/Asia                      0.5                      0.0

Source: Federal Reserve data.

(a.) In the global rebalancing scenario all regions' current account
balances go to zero. See table 3 for definitions of exchange rates and
other parameter assumptions.

Table 8. Changes in Real Exchange Rates and Terms of Trade in the
Global Rebalancing Scenario with and without Valuation and Interest
Rate Effects (a)

Log change x 100

                            With
                          valuation        Without       With valuation
                         effects and      valuation       effects and
                           without       effects and        interest
Real exchange rate      interest rate      interest       rate effects
or terms of trade        effects (b)     rate effects         (c)

Real exchange rate
United States/Europe        28.6             33.7             30.1
United States/Asia          35.2             40.7             37.2
Europe/Asia                  6.7              7.0              6.3
Terms of trade
United States/Europe        14.0             16.5             15.1
United States/Asia          14.5             16.5             15.3
Europe/Asia                  0.5              0.0              0.2

Source: Federal Reserve data.

(a.) In the global rebalancing scenario all regions' current account
balances go to zero. See table 3 for definitions of exchange rates and
other parameter assumptions.

(b.) Same as the baseline scenario reported in first colmnn of table 3.

(c.) Interest rates on U.S. short-term liabilities held by foreigners
are assumed to rise 1.25 percentage points, to the same level as the
return earned by U.S. residents abroad.

Table 9. Changes in Real Exchange Rates and Terms of Trade in Global
Rebalancing Scenario with Higher Productivity in Non-U.S. Nontraded
Goods (a)

Log change x 100

                                               With 20 percent increase
                                                  in productivity in
Real exchange rate       Without increase        European and Asian
or terms of trade       in productivity (b)         ontraded goods

Real exchange rate
United States/Europe           28.6                      17.0
United States/Asia             35.2                      23.6
Europe/Asia                     6.7                       6.6
Terms of trade
United States/Europe           14.0                      15.0
United States/Asia             14.5                      15.3
Europe/Asia                     0.5                       0.2

Source: Federal Reserve data.

(a.) In the global rebalancing scenario all regions' current account
balances go to zero. See table 3 for definitions of exchange rates and
other parameter assumptions.

(b.) Same as the baseline scenario reported in the first column of
table 3.

Table 10. Changes in Real Exchange Rates under Alternative Assumed
Speeds of Global Rebalancing.

Log change x 100

                                           Speed  (b)

                           Moderate          Gradual      Very gradual
Real exchange rate      (1-2 years) (c)    (5-7 Years)    (10-12 years)

United States/Europe         28.6             13.4             6.5
United States/Asia           35.2             17.3             8.5
Europe/Asia                   6.7              3.9             2.0

Source: Federal Reserve data.

(a.) In the global rebalancing scenario all regions' current account
balances go to zero. See table 3 for definitions of exchange rates and
other parameter assumptions.

(b.) Proxied by varying elasticities of substitution: moderate,
[theta] = 1, [eta] = 2: gradual, [theta] = [eta] = 4; very gradual,
[theta] = 4, [eta] = 8.

(c.) Same as the baseline scenario reported in the first column of
table 3.