Journal of Policy Modeling 28 (2006) 629643

The U.S. current account decit: Gradual correction or abrupt adjustment?

Abstract In this paper I use a large multi-country data set to analyze the determinants of abrupt and large current account reversals. FS 2006 Society for Policy Modeling. Published by Elsevier Inc. All rights reserved.
Keywords: Current account; Exchange rates; U.S. dollar; Real exchange rate; Net international investment position; Current account reversals; Sudden stops JEL classication: F02, F40, O11

1. Introduction In 1991, and after 8 years of running a decit, the U.S. posted a current account surplus of 0.7% of GDP. That was the last time the current account balance was positive. Since then the U.S. external imbalance has grown steadily. In 2005 the decit was 6.4% of GDP, and it is expected to grow even further during 2006. A number of analysts have become increasingly alarmed by these very large external imbalances. Some authors have argued that by relying on foreign central banks purchases of government securities, the U.S. has become vulnerable to changes in expectations and economic sentiments (Feldstein, 2006). A number of analysts have argued that at over 6% of GDP the U.S. current account decit is clearly unsustainable, and that in the next few years it will have to be cut, approximately, in half.1 Of course, from a global perspective, a reduction in the U.S. decit implies a decline in the rest of the worlds collective current account surpluses.
1

Tel.: +1 310 206 6797; fax: +1 310 206 5825. E-mail address: [email protected]. Obstfeld and Rogoff (2005), Blanchard, Giavazzi, and Sa (2005) and Mussa (2004).

0161-8938/$ see front matter 2006 Society for Policy Modeling. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.jpolmod.2006.06.012

630

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

That is, adjustment is required both in decit and surplus nations. According to the International Monetary Funds World Economic Outlook (2005), the issue is not whether, but how they will adjust (p. 67). A key question is what will be the effect of a major reduction in the U.S. external decit on economic activity. More specically, the question is whether there will be an orderly adjustment process, with no signicant impact on growth and employment, or, whether adjustment will create havoc and large dislocations. The answer to this question depends on how rapid adjustment takes place. If adjustment is gradual and distributed through a number of years, neither GDP growth nor employment are likely to be affected in a signicant way. If, on the other hand, adjustment is sudden and abrupt, the U.S. and the world economies are likely to suffer. Recently, The Financial Times described the likely consequences of an abrupt U.S. external adjustment as follows2 : If willing lenders to the US were to dry up [suddenly], the consequences would be painful: a plunge in the dollar, a spike in interest rates, higher import prices. Recent evidence presented in Calvo, Izquierdo, and Mejia (2004), Edwards (2004, 2005a, 2005b) and Frankel and Cavallo (2004) suggests that countries that have experienced sudden declines in capital inows and/or abrupt current account reversals have suffered signicant reduction in the rate of economic growth. According to Edwards (2005b), industrial countries that experienced current account reversals that exceeded 4% of GDP in 1 year, suffered a short-term decline in GDP per capita growth in the order of 2.9%3.9%, relative to trend. In this paper I use a large multi-country data set to analyze the determinants of abrupt and large current account reversals. More specically, I use a panel random-effect probit model to estimate which macroeconomic variables affect the probability that a country experiences an abrupt current account reversal. This analysis is important for assessing the likelihood that the U.S. will be subject to a sudden, large and disruptive current account correction sometime in the next few years. It is important to stress at the outset that the results reported in this paper should be interpreted with care. This is for two reasons: rst, the U.S. plays a unique and central role in the international nancial system. Second, in modern times a very large country such as the U.S. has never run persistent and very large current account decits (Edwards, 2005b). That is, there are no historical precedents for the current large global imbalances. The rest of the paper is organized as follows: in Section 2 I discuss the U.S. current account decit from a historical perspective. I analyze trends and other countries experiences. I look at both the nancing of the decit and at the evolution of the U.S. net international investment position (NIIP). In Section 3 I provide a summary of the international evidence on large and signicant current account reversals. Section 4 contains new econometric results on the probability of countries experiencing abrupt and large current account reversals. In this Section I discuss how the probability of the U.S. experiencing a reversal has evolved during the last few years. Finally Section 5 contains the conclusions. 2. The U.S. current account imbalance in international perspective In this paper, Section 1 provide some background information on the evolution of the U.S. current account during the last three decades. In Fig. 1 I present quarterly data for the U.S. current
2

Economists expect to escape a disorderly correction, Financial Times, January 25, 2006, p. 3.

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

631

Fig. 1. Current account balance and real exchange rate.

account balance as percentage of GDP, for the period 19732004.3 I also include data on the evolution of the Federal Reserves trade-weighted index of the U.S. dollar real exchange rate (an increase in the RER index represents a real exchange rate appreciation).4 This gure shows that decits have become increasingly large since 1992. Fig. 1 also shows that during the period under consideration the RER index experienced signicant gyrations. Finally, Fig. 1 shows a pattern of negative correlation between the trade-weighted real value of the dollar and the current account balance. In Table 1 I present data on the current account as a percentage of GDP, and its nancing for the period 19902004. As may be seen, during the last few years the nature of external nancing has changed signicantly. Since 2002 net foreign direct investment (FDI) ows have been negative; this contrasts with the 19972001 period when FDI ows contributed in an important way to decit nancing. Also, after 4 years on net positive equity ows (19982002), these became negative in 20032004. As the gures in Table 1 show, during 2003 and 2004 the U.S. current account decit was fully nanced through net xed income ows, and in particular through ofcial foreign purchases of government securities.5 In Fig. 2 I present the evolution of the U.S. net international investment position (NIIP) as percentage of GDP. As may be seen, this has become increasingly negative: in 2004 U.S. net international liabilities reached 29% of GDP. An important feature of the NIIP is that gross U.S. international assets and gross U.S. international liabilities are held in different currencies. While more than 70% of gross foreign assets held by U.S. nationals are denominated in foreign currency, approximately 95% of gross U.S. liabilities in hands of foreigners are denominated in U.S. dollars. This means that net liabilities as a percentage of GDP are subject to valuation effects stemming from changes in the value of the dollar. Dollar depreciation reduces the value of net liabilities; a
Parts of this section draw on Edwards (2005a). This is the Federal Reserve RER index. 5 See, for example, Martin Wolfs October 1, 2003 article in the Financial Times, Funding Americas recovery is a very dangerous game, (p. 15).
4 3

632

Table 1 U.S. net nancial ows: 19902004 ($ billion) 1990 Reserves (net) Foreign private purchases of U.S. treasuries Currency Securities (net) Debt securities Equity securities FDI (net) Claims reported by non-banks (net) Claims reported by banks (net) Net nancing Current account decit 31.8 2.5 18.8 27.2 11.3 17.3 8.6 58.0 79.0 1991 23.2 18.8 15.4 10.5 14.7 8.0 3.4 43.5 3.7 1992 44.4 37.1 13.4 19.1 28.4 13.2 37.4 97.9 48.0 1993 70.4 24.4 18.9 66.2 32.6 11.3 55.7 81.8 82.0 1994 44.9 34.3 23.4 6.2 34.0 35.0 100.1 127.4 118.0 1995 100.1 91.5 12.3 45.1 41.0 14.4 44.9 87.3 109.5 1996 133.4 147.0 17.4 46.0 13.0 36.8 5.4 32.6 75.1 138.7 120.2 1997 18.0 130.4 24.8 44.6 84.2 24.7 0.8 5.2 7.9 221.3 136.0 1998 26.7 28.6 16.6 32.1 145.5 30.3 36.4 15.1 4.2 76.2 209.6 1999 52.3 44.5 22.4 182.6 104.2 84.5 64.5 21.5 22.0 233.8 296.8 2000 42.5 70.0 5.3 338.0 267.7 93.0 162.1 31.9 31.7 478.0 413.4 2001 23.1 14.4 23.8 309.2 300.3 12.6 24.7 57.6 7.5 416.6 385.7 2002 110.3 100.4 21.5 301.4 269.8 37.5 62.4 32.6 66.1 569.9 473.9 2003 250.1 113.4 16.6 178.6 241.8 63.2 133.9 55.1 65.2 542.7 530.7 2004 358.1 108.1 14.8 323.2 360.1 36.8 133.0 41.5 15.6 614.0 665.9

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

Source: BEA, U.S. International Transactions and International Investment Position.

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

633

Fig. 2. U.S. net international investment position, 19762004.

dollar appreciation, on the other hand, increases the dollar value of U.S. net liabilities. Because of this valuation effect, the deterioration of the U.S. NIIP during 20022004 was signicantly smaller than the accumulated current account decit during those 2 years. An important policy question refers to the reasonable long-run equilibrium value of the ratio of U.S. net international liabilities to GDP; the higher this ratio, the higher will be the sustainable current account decit. According to some authors, the current ratio of almost 30% of GDP is excessive, while others believe that a NIIP to GDP ratio of up to 50% would be reasonable.6 From an accounting point of view, the current account is the difference between savings and investment. A number of authors have argued that a worsening of a current account balance that stems from an increase in investment is very different from one that results from a decline in national savings. Some have gone as far as arguing that very large decits in the current account dont matter, as long as they are the result of higher (private sector) investment (Corden, 1994). The recent deterioration of the U.S. current account has largely been the result of a decline in national savings, and in particular of public and household savings. A simple implication of this trend and one that is emphasized by most authors is that an improvement in the U.S. current account situation will not only imply a RER adjustment; it will also require an increase in the national savings ratio, and in particular in household savings. Symmetrically, a correction of current global imbalances will also require a decline in Europes and Japans savings rates and/or an increase in their investment rates.7 In Table 2 I present data on the distribution of current account decits in the world economy, as well as in six groups of nations Industrial, Latin America, Asia, Middle East, Africa and Eastern Europe for the period 19702001. As may be seen, at almost 6% of GDP the U.S. decit is very large from a historical and comparative perspective. It is in the top decile of decits distribution for all industrial countries in the rst 30 years of oating exchange rates. Since 1970 the U.S. has been the only large industrial country that has run current account decits in excess of 5%. This reects the unique position that the U.S. has in the international
See Obstfeld and Rogoff (2004), Edwards (2005a), and Mussa (2004). That is, the global savings glut identied by Bernanke (2005) would have to be reversed. See Greenspans Speech to the International Monetary Conference in Beijing, June 6, 2005.
7 6

634

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

Table 2 Distribution of current account decits by region: 19702001 Region Industrialized countries Latin America and Caribbean Asia Africa Middle East Eastern Europe Mean 0.6 5.4 3.0 6.3 0.0 3.9 Median 0.7 4.1 2.7 5.3 1.4 3.0 1st Perc. 3.8 2.5 7.1 3.4 18.8 2.4 1st Quartile 1.6 1.1 0.6 1.2 5.0 0.3 3rd Quartile 3.0 8.0 6.3 9.9 6.4 6.1 9th Perc. 4.8 16.9 11.3 16.9 13.6 10.7

Source: Authors elaboration based on World Development Indicators. A negative number indicates a current account surplus. Table 3 Net sock of liabilities: U.S and other industrial countries: selected years (percent of GDP) Country Australia Canada Denmark Finland Iceland New Zealand Sweden United States 1980 34.7 14.6 12.9 1985 36.3 19.0 20.9 1.3 1990 47.4 38.0 29.2 48.2 88.7 26.6 4.2 1995 55.1 42.4 26.5 42.3 49.8 76.6 41.9 6.2 2000 65.2 30.6 21.5 58.2 55.5 120.8 36.7 14.1 2003 59.1 20.6 13.0 35.9 66.0 131.0 26.5 22.1

Source: Bureau of Economic Analysis and Lane and Milesi-Ferretti (2001).

nancial system, where its assets have been in high demand, allowing it to run high and persistent decits. On the other hand, this fact also suggests that the U.S. is moving into uncharted waters. As Obstfeld and Rogoff (2004, 2005), among others, have pointed out, if the decit continues at its current level, in 25 years the U.S. net international liabilities will surpass the levels observed by any country in modern times. During the last 30 years only small industrial countries have had current account decits in excess of 5% of GDP: Australia, Austria, Denmark, Finland, Greece, Iceland, Ireland, Malta, New Zealand, Norway and Portugal. What is even more striking is that very few countries either industrial or emerging have had persistently high current account decits for more than 5 years (Edwards, 2005b). In Table 3 I present data on net international liabilities as a percentage of GDP for a group of advanced countries that have historically had a large negative NIIP position.8 The picture that emerges from this table is quite different than what one nds when analyzing current account decits across countries. Indeed, a number of advanced nations have had and continue to have a signicantly larger net international liabilities position than the U.S. This suggests that, at least in principle, the U.S. NIIP could continue to deteriorate for some time into the future. However, even if this does happen, at some point this process would have to end, and the U.S. net international liabilities position as percentage of GDP would have to stabilize. It

8 For the U.S. the data are from the Bureau of Economic Analysis. For the other countries the data are, until 1997, from the Lane and Milesi-Ferretti (2001) data set. I have updated them using current account balance data. Notice that the updated gures should be interpreted with a grain of salt, as I have not corrected them for valuation effects.

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

635

makes a big difference, however, at what level U.S. net international liabilities do stabilize. For example, if in the steady state foreigners are willing to hold the equivalent of 35% of U.S. GDP in the form of net U.S. assets, the U.S. could sustain a current account decit of (only) 2.1% of GDP.9 If, on the other hand, foreigners net demand for U.S. assets grows to 60% of GDP which, as shown in Table 3, is approximately the level of (net) foreign holdings of Australian assets the U.S. sustainable current account decit would be 3.6% of GDP. Moreover, if foreigners are willing to hold (net) U.S. assets for the equivalent of 100% of GDP a gure that Mussa (2004) considers implausible the sustainable U.S. current account decit can be as high as 6% of GDP approximately its current level. Since there are no historical precedents for a large advanced nation running persistently large decits, it is extremely difcult to have a clear idea on what will be the actual evolution of foreigners demand for U.S. assets. It is worth noting that an analysis for a longer period of time conrms the view that the recent magnitude of the current account decit has no historical precedent in the U.S. According to Backus and Lambert (2005) the U.S. ran a current account decit of 5% of GDP in 1815, and a somewhat smaller but persistent decit during the 1830s and 1870s. Greenspan (2004, p. 6) has pointed out that the large decits during the 19th century were nanced with capital ows related to specic major development projects (such as railroads). 3. Current account reversals in the world economy: empirical evidence A number of authors have concluded that the U.S. current account decit is unsustainable in the long-run. Even under an optimistic scenario, where foreigners demand for U.S. securities doubles from its current level, there would have to be a signicant decline in the decit. For example, if the (negative) NIIP were to go from its current level of 30% of GDP to 60% of GDP, the sustainable current account decit would be 3.6%. This is almost three percentage points below its current level. In reality, however, the adjustment is likely to be even larger. The reason for this is that in order for the NIIP to go from 30% to 60% of GDP in a reasonable period of time, the current account decit needs to overshoot its steady state level by a signicant margin (Edwards, 2005a). A key question is what will be the nature of this adjustment process? In particular, will the adjustment be abrupt, and thus costly, or will it be gradual? In this Section I summarize the international experience with abrupt and large current account reversals in the period 19702001. Although the U.S. case is unique both because of the size of its economy and because the dollar is the main vehicle currency in the world , an analysis of the international experience will provide some light on the likely nature of its adjustment. In a recent study, Frankel and Cavallo (2004) concluded that sudden stops of capital inows (a phenomenon closely related to reversals) have resulted in growth slowdown. Croke, Kamin, and Leduc (2005), on the other hand, argue that there is no evidence suggesting that reversals have historically been associated with a decline in the rate of growth.10 Their denition of reversal, however, is of a rather small turnaround in the current account balance. In a series of papers (Edwards, 2002, 2004, 2005a), I have analyzed the impact of current account reversals on economic performance. I have found that countries that experience large and abrupt reversals have experienced drastic reductions in investment and in GDP growth. I also found that if the current

9 10

This calculation assumes a 6% rate of growth of nominal GDP going forward. See also Milesi-Ferretti and Razin (2000) and Edwards (2002).

636

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

Table 4 Incidence of current account reversals: 19702001 (percentages) Region Type of reversal Reversal (4%) Industrial countries Latin American and Caribbean Asia Africa Middle East Eastern Europe Total Pearson Uncorrected 2 (5) Design-based F(5, 12500) P-value 1.3 5.5 8.2 8.8 10.4 5.9 6.5 33.8 6.8 0.00 Reversal (2%) 3.3 9.4 10.7 11.9 14.9 7.3 9.4 33.7 6.7 0.00

Source: Authors elaboration based on World Development Indicators.

account adjustment is orderly and gradual, it does not disrupt economic activity in a signicant way. 3.1. The incidence of reversals: international evidence In this study I dene a current account reversal (CAR) episode as a reduction in the current account decit of at least 4% of GDP in a 1 year period, that accumulates to a reduction of at least 5% of GDP over 3 years. In Table 4 I present data on the incidence of current account reversals for six groups of countries. As may be seen, for the overall sample the incidence of reversals is 6.5%. The incidence of reversals among the industrial countries is much smaller however, at 1.3%. The advanced countries that have experienced current account reversals during the period under study are: Greece (1986), Italy (1975), Malta (1997), New Zealand (1975), Norway (1978, 1989), and Portugal (1982, 1983, 1985).11 With the exception of Italy, all of these countries are quite small; this underlies the point that there are no historical precedents of large countries, such as the U.S., undergoing profound current account adjustments. An alternative way of dividing the sample and one that is particularly relevant for the discussion of possible lessons for the U.S. is by country size. I dene large countries as those having a GDP in the top 25% of the distribution in 1995 (according to this criterion there are 44 large countries in the sample); the incidence of abrupt current account reversals among large countries is 3.9% for 19712001. 3.2. The U.S. current account Reversal of 19871991 Between 1987 and 1991 the U.S. current account decit experienced a signicant current account correction. In the third quarter of 1987 the decit stood at 3.7%, a gure that was then considered to be exceptionally high. During the next 3 years the decit declined gradually, and in the fourth quarter of 1990 it was 1% of GDP. During the next two quarters, and as a result of foreign countries contributions to the nancing of the Gulf War, the current account briey
11

See Edwards (2005b) for alternative denitions of reversals.

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

637

posted a surplus of 0.8% of GDP. The 19871991 adjustment process was accompanied by a major depreciation of the U.S. dollar. The dollar began to loose value in the second quarter of 1985, almost 2 years before the current account decit began its turnaround.12 In the period immediately preceding the adjustment process (19851987) the U.S. dollar depreciated signicantly in real terms; this weakening of the U.S. dollar continued, although at a slower pace during 19871991. Between the second quarter of 1985 and the second quarter of 1991 the dollar lost 30% of its value in real trade-weighted terms. Although this episode does not qualify as a reversal, in the sense dened in this paper, it is the closest to a major current account adjustment that the U.S. has experienced in modern times. During the early part of this episode there was no decline in GDP, nor was there an increase in unemployment. However, during the latter part of the adjustment starting in the second quarter of 1990 there was a decline in GDP and a marked increase in unemployment. Indeed, GDP stayed below its stochastic trend well into 1993; unemployment was above its own trend until early 1994. According to the National Bureau of Economic Research in August of 1990 the U.S. entered into a recession that lasted until March of 1991.13 The Federal Funds interest rate increased signicantly during the rst part of this adjustment episode. In October 1986 the Federal Funds rate was 5.85%; by March 1989 it had increased by 400 basis points, to 9.85%. In June 1989 the Fed cut rates by 25 basis points, and began a period of interest rate reduction. By the end of the adjustment, in June 1991, the Federal Funds rate stood at 5.9%. The yield on the 10-year Treasury Note increased signicantly in the months preceding the actual current account adjustment. The 10 year Note yield went from 7.1% in January 1987, to 9.4% in September of that yearan increase of 230 basis points. From that time and until March 1989, the yield on the 10-year Note moved between 9% and 9.4%. Starting in April 1989, long term interest rates began to fall, reaching 8% in April 1991. In June 1993, 2 years after the current account adjustment had ended, the long tem interest rate was 6%. Also, the yield curve became inverted in January 1989, and stayed inverted until January 1990. The 19871991 current account adjustment in the U.S. was signicant, but gradual. And although the episode does not qualify as a current account reversal, as dened in this paper, it does provide some useful information. This adjustment was not characterized by a traumatic collapse in output. However, the 19871991 adjustment episode in the U.S. was characterized by: (a) a steep depreciation of the U.S. dollar. (b) An increase in ination. (c) Higher interest rates; the Fed Funds rate increased through the rst half of the adjustment, while the 10 years rate increased in the months prior to the beginning of the actual adjustment. (d) A decline in GDP below trend towards the latter part of the adjustment. In fact, the U.S. entered into a recession while the adjustment was taking place. (e) An increase in the rate of unemployment above trend, during the nal quarters of the adjustment. 4. Abrupt current account reversals: empirical determinants In order to understand further the forces behind abrupt external adjustments, I estimated a number of equations on the probability of large countries experiencing a major current account

12 This two-year lag coincides with the conventional wisdom of the time it takes a dollar depreciation to affect the current account. 13 I am not necessarily implying causality in this description of the data.

638

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

reversal. The empirical model is a variance component probit, and is given by Eqs. (1) and (2): tj = 1, 0,
>0 if tj

otherwise

(1) (2)

tj = tj + tj

Variable tj is a dummy variable that takes a value of one if country j in period t experienced a current account reversal (as dened above), and zero if the country did not experience a reversal. According to Eq. (2), whether the country experiences a current account reversal is assumed to . , in turn, is assumed to depend linearly on be the result of an unobserved latent variabletj tj vector tj . The error term tj is given by given by a variance component model: tj = vj + tj . vj 2 ; is normally distributed with zero mean and variance is iid with zero mean and variance tj 2 = 1. The data set used covers 44 countries, for the 19702001 period; not every country has data for every year, however. See Edwards (2005b) for exact data denition and data sources. In determining the specication of this probit model I followed the literature on external crises, and I included the following covariates14 : (a) the ratio of the current account decit to GDP, lagged one period. (b) The lagged ratio of the countrys scal decit relative to GDP. (c) An index that measures the relative occurrence of sudden stops in the countrys region (excluding the country itself). This variable captures the effect of regional contagion, and I expect its coefcient to be positive. (d) Change in the logarithm of the terms of trade (dened as the ratio of export prices to import prices), with a 1 year lag. (f) The countrys initial GDP per capita (in logs). This measures the degree of development of the country in question. If more advanced countries are less likely to experience a reversal, its coefcient would be negative. (e) The 1-year lagged rate of growth of domestic credit. This is a measure of the monetary policy stance. (g) A dummy variable that takes the value of one if that particular country had a exible exchange rate regime, and zero otherwise. 4.1. Basic results In Table 5 I present the results obtained from the estimation of this variance-component probit model for a sample of large countries. As before, I have dened a country as being large if in the year 1995 its GDP was in the top 25% of the global GDP distribution. As may be seen in columns (5.1) and (5.2), the coefcients of both the current account decit and the scal decit are signicantly positive, indicating that an increase in these imbalances increases the probability of the country in question experiencing an abrupt current account reversal. All the other regressors in columns (5.1) and (5.2) are signicantly estimated, and have the expected signs. The results indicate that there is a regional contagion effect, and that a deterioration in the terms of trade increases the probability of a reversal. These results also indicate that countries with a higher (log of) GDP per capita have a lower probability of a reversal. In columns (5.1) and (5.2) the scal and current account decits variables were introduced separately in the estimation. In column (5.3) I present estimates when both variables are included in the same probit equation. As may be seen, in this case the coefcient of the (lagged) current account decit continues to be positive and signicant. However, the coefcient of the scal decit ceases to be statistically signicant. This result is rather intuitive: higher scal imbalances that are not associated with a deterioration of the external accounts, do not affect in a signicant way
14

See, for example, Frankel and Rose (1996), Milesi-Ferretti and Razin (2000) and Edwards (2002).

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

Table 5 Current account reversals: random effects probit modelunbalanced panel large countries Variable Current-account decit to GDP Fiscal decit to GDP Sudden stops in region Changes in terms of trade Domestic credit growth Flexible exchange rate GDP per capita Observations Countries (5.1) 0.165 (7.51)* 2.335 (3.15)* 0.013 (2.30)** 0.127 (2.18)** 881 42 (5.2) 0.035 (2.07)** 2.731 (3.84)* 0.019 (3.50)* 0.180 (2.71)* 822 36 (5.3) 0.174 (7.20)* 0.003 (0.21) 2.094 (2.73)* 0.013 (2.33)** 0.140 (2.36)** 822 40 (5.4) 0.165 (6.43)* 0.002 (0.10) 2.327 (2.70)* 0.013 (2.08)** 0.379(2.00)* 0.104 (1.69)*** 694 40 (5.5) 0.153 (5.59)* 0.009 (0.54) 2.261 (2.50)** 0.014 (1.92)*** 0.0001 (1.36) 0.298 (1.62)*** 0.127 (1.71)*** 608 36

Absolute values of z-statistics are reported in parentheses; explanatory variables are one-period lagged variable; country-specic dummies are included, but not reported. * Signicant at 1%. ** Signicant at 5%. *** Signicant at 10%.

639

640

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

Table 6 Current account reversals: marginal effects and predicted probabilitya,b (computed from the estimates in column (5.3)) Variable Current-Account decit to GDP Fiscal decit to GDP Sudden stops in region Changes in terms of trade GDP per capita Predicted Probability
a b

(6.1) 0.011 (5.70)* 0.000 (0.21) 0.128 (2.43)** 0.001 (2.20)** 0.009 (2.40)** 0.026

(6.2) 0.042 (4.53)* 0.001 (0.21) 0.509 (2.75)* 0.003 (2.38)** 0.034 (2.44)** 0.160

(6.3) 0.007 (2.69)* 0.000 (0.20) 0.090 (2.09)** 0.001 (2.01)** 0.006 (2.95)* 0.017

(6.4) 0.040 (3.75)* 0.001 (0.21) 0.435 (2.92)* 0.003 (1.87)*** 0.033 (2.81)* 0.149

For details on the computations in each column, see the text. Absolute values of z-statistics are reported in parentheses. For (6.1) sample means are 1.567 for current account decit to GDP, 4.074 for scal decit to GDP, 0.092 for sudden stops in region, 5.459 for changes in terms of trade, and 9.744 for log of GDP per capita. * Signicant at 1%. ** Signicant at 5%. *** Signicant at 10%.

the probability of an abrupt current account reversal.15 Finally, the results in columns (5.4) and (5.5) suggest that countries with exible exchange rates have been less likely to experience an abrupt current account reversal, and that a more expansive monetary policy has had a positive although statistically marginal effect on the probability of a sudden current account reversal. All the estimated models presented in Table 5 performed quite well; the pseudo-R2 ranged between 0.41 and 0.29. 4.2. Marginal effects of current account decits on the probability of reversals An important question in U.S. policy debates is whether the higher current account decits experienced during the last few years have resulted in a higher likelihood of the country facing an abrupt adjustment sometime in the future. This issue can be investigated by computing the marginal effect (and standard error) of the current account decit on the probability of an abrupt reversal. For a probit model these marginal effects are estimated as the derivatives of the cumulative normal distribution with respect to the corresponding regressor. These derivatives are then evaluated for given values of the independent variables. An important property of probit models is that marginal effects are highly nonlinear and are conditional on the values of all covariates. If the value of any of the independent variables changes, the marginal effect of any of them on the probability of the outcome variable will also change. In Table 6 I present a series of marginal effects using the coefcient estimates from column (5.3) in Table 5; the z-statistics of these marginal effects are in parenthesis. The last row in this table corresponds to the predicted probability of occurrence of an abrupt reversal. This predicted probability is computed for the same values of the independent variables that were used to compute the marginal effects reported in that particular column. The marginal effects in column (6.1) have been evaluated at the sample means of the independent variables (See the footnote in Table 6 for the means values.) As may be seen, a marginal increase in the current account decit raises the probability of a reversal in 1.1 percentage points.
15 The signicant positive coefcient of the scal decit in column (5.2) is picking up the effect of the omitted current account variable.

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

641

The last row shows that, when evaluated at the sample means, the predicted probability of a reversal is a relatively low 2.6%. Column (6.2) presents the marginal effects assuming that the initial current account decit is 7% of GDP a gure close to the estimated U.S. decit for 2006 and that all other regressors remain at the sample means values. As may be seen, the marginal effect of the current account decit increases to 4.2%. Moreover, as the last row in Table 6 shows, as a result of the much higher current account decit (7% as opposed to 1.6%), the predicted probability of a current account reversal has increased to 16%. The next two columns in Table 6 are an attempt to analyze the case of the U.S. in greater detail. Column (6.3) presents the marginal effects and predicted probability evaluated for the values of the independent variables corresponding to the U.S. in 1999. The corresponding values for these variables are: current account decit = 2.5; scal decit = 0.62; contagion ratio = 0.048; change in the terms of trade = 10.1; log of initial GDP per capita = 9.744. What is particularly interesting about these 1999 values is that at that time the U.S. current account decit was 2.5%, close to what many analysts consider a sustainable level. Also, in 1999 the U.S. was running a small scal surplus. As may be seen from column (6.3), when the 1999 U.S. values are used, the marginal effect of the current account decit is less than 1% (0.0075). Moreover, as the last row in the table shows, in 1999 the predicted probability of a current account reversal in the U.S. was a very low 1.7%. At that time, the predicted probability of the U.S. facing an abrupt and large change in its current account balance was negligible. The last column in Table 6 column (6.4) evaluates the marginal effects and predicted probably using values for the independent variables that reect the U.S. situation in mid 2006. In computing the marginal effects reported in column (6.4) I use the following values: current account decit = 7.0; scal decit = 3.0; contagion ratio = 0.048; change in the terms of trade = 12; log of initial per capita GDP = 9.744.16 Two interesting results emerge from the computation reported in column (6.4): rst, the marginal effect of the current account decit on the probability of an abrupt reversal has increased signicantly, relative to the estimates using the 1999 U.S. values (see column (6.3)). Indeed, this marginal effect is now 0.04; in column (6.3) it was only 0.0075. This suggests that further increases in the U.S. external imbalance will increase signicantly the likelihood of an abrupt reversal. Second, the last column in Table 6 indicates that the predicted probability of a current account reversal has increased substantially in 2006 relative to 1999. In column (6.3) in 1999 the predicted probability of the U.S. experiencing a reversal was a mere 1.7%; by 2006 this predicted probability had jumped to 14.9%. Although in absolute terms this is still a rather small number, its rate of increase has been quite remarkable. 5. Concluding remarks Never in the history of modern economics has a large industrial country run persistent current account decits of the magnitude posted by the U.S. since 2000. Most analysts have argued that the U.S. cannot continue to run these large decits for much longer.17 At some point the decit will have to decline to a more sustainable level. An important question is whether there will be an orderly adjustment process, with no signicant impact on growth and employment, or whether the adjustment will be sudden and abrupt, creating havoc and large dislocations.
16 The deterioration in the terms of trade is the result of the hike in the price of oil in 2005. I also assume that the contagion ratio stays at the same level as in 1999. 17 For an alternative view see Caballero, Farhi, and Gourinchas (2006).

642

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

In order to have an idea of the possible nature of this adjustment process, in this paper I have analyzed the international evidence on abrupt and signicant current account reversals. The results from the estimation of a number of variance-component probit equations indicate that the probability of experiencing a major current account reversal is positively affected by larger current account decits, a deterioration in terms of trade, and expansive monetary policies. On the other hand, this probability has been lower for more advanced countries and for countries with exible exchange rates. Overall, these results indicate that some of the recent developments in the U.S. economy have increased the probability of the country experiencing a current account reversal. An analysis of the marginal effects of current account decits and of the predicted probability of reversal indicates that both have increased signicantly for the U.S. since 1999. Using values of the key variables, I estimated that the predicted probability of a current account reversal in the U.S. has increased from 1.7% in 1999, to 14.9% in 2006. Although the absolute value of this probability continues to be on the low side, its rate of increase has been signicant and fast. It is important to point out once again, that the U.S. case is unique, and the results reported in this paper should be interpreted carefully. This is for several reasons. First, and as pointed out above, there are no historical precedents of a very large industrial country running persistent and very large decits. This means that there have been no historical experiments that guide us in the analysis, or in the estimation. Second, the U.S. plays a fundamental role as the center of the international nancial system. This means that the global demand for U.S. issued securities is high, and may continue to expand. What we do not know is for how long the expansion in the demand for U.S. debt will continue to grow. And third, the U.S. decit is enormous in absolute terms in 2005 it exceeded USD 800 billion and is being nanced by a very large percentage of world savings. Acknowledgements I thank Ed Leamer for helpful discussions, and Roberto Alvarez for his excellent assistance. References
Backus, D., & Lambert, F. (2005). Current account fact and ction. Stern School, New York University, 2005. Bernanke, B. S. (2005). The global saving glut and the U.S. current account decit. Remarks at the Sandridge Lecture, Virginia Association of Economics, Richmond, Virginia, March 10. Blanchard, O., Giavazzi, F., & Sa, F. (2005). The U.S. current account and the dollar. NBER working paper no. 11137, February. Caballero, R. J., Farhi, E., & Gourinchas, P. O. (2006). An equilibrium model of Global Imbalances and low interest rates. NBER working paper no. 11996, February. Calvo, G. A., Izquierdo, A., & Mejia, L. F. (2004). On the empirics of sudden stops: The relevance of balance-sheet effects. NBER working paper no. 10520, May. Croke, H., Kamin, S. B., & Leduc, S. (2005). Financial market developments and economic activity during current account adjustments in industrial economies. International nance discussion papers 827, Board of Governors of the Federal Reserve System. Corden, W. M. (1994). Economic policy, exchange rates, and the international system. Oxford and Chicago: Oxford University Press; The University of Chicago Press. Edwards, S. (2002). Does the current account matter? In S. Edwards & J. A. Frankel (Eds.), Preventing currency crises in emerging markets (pp. 2169). The University of Chicago Press. Edwards, S. (2004). Thirty years of current account imbalances, current account reversals and sudden stops, IMF Staff Papers, Vol. 61, Special Issue: 149 International Monetary Fund. Edwards, S. (2005a). Is the U.S. current account decit sustainable? And if not, how costly is adjustment likely to be? Brookings Papers on Economic Activity, 0(1), 211288.

S. Edwards / Journal of Policy Modeling 28 (2006) 629643

643

Edwards, S. (2005b). The end of large current account decits, 19702002: Are there lessons for the United States? In The Greenspan era: Lessons for the future. The Federal Reserve Bank of Kansas City, pp. 205268. Feldstein, M. (2006). Why Uncle Sams bonanza might not be all that it seems. The Financial Times, January 10, 2006. Frankel, J. A., & Rose, A. K. (1996). Currency crashes in emerging markets: An empirical treatment. Journal of International Economics, 41(34), 351366. Frankel, J. A., & Cavallo, E. A. (2004). Does openness to trade make countries more vulnerable to sudden stops, or less? Using gravity to establish causality. NBER working paper no. 10957, December. Greenspan, A. (2004). The evolving U.S. payments imbalance and its impact on Europe and the rest of the World. CATO Journal, 27(Spring/Summer (12)). Greenspan, A. (2005). Remarks at the international monetary conference. Beijing, Peoples Republic of China, June 6. Lane, P., & Milesi-Ferretti, G. M. (2001). The external wealth of nations: Measures of foreign assets and liabilities for industrial and developing countries. Journal of International Economics, 55(2), 263294. Milesi-Ferretti, G. M., & Razin, A. (2000). Current account reversals and currency crises: Empirical regularities. In P. Krugman (Ed.), Currency crises. University of Chicago Press. Mussa, M. (November 2004). Exchange rate adjustments needed to reduce global payments imbalance. In C. F. Bergsten & J. Williamson (Eds.), Dollar adjustment: How far? Against what? Washington, DC: Institute for International Economics. Obstfeld, M., & Rogoff, K. (2004). The unsustainable US current account position revisited. NBER working paper 10869, November. Obstfeld, M., & Rogoff, K. (2005). Global exchange rate adjustments and global current account imbalances. Brookings Papers on Economic Activity, 0(1), 67146.