Explaining the global pattern of current account imbalances

Explaining the global pattern of current account imbalances

Journal of International Money and Finance 26 (2007) 500e522 www.elsevier.com/locate/jimf Explaining the global pattern of current account imbalances...

794KB Sizes 0 Downloads 9 Views

Journal of International Money and Finance 26 (2007) 500e522 www.elsevier.com/locate/jimf

Explaining the global pattern of current account imbalances Joseph W. Gruber*, Steven B. Kamin Board of Governors of the Federal Reserve System, International Finance Division, Washington DC 20551, USA

Abstract We assess some of the explanations that have been put forward for the global pattern of current account imbalances that has emerged in recent years, particularly the large U.S. current account deficit and the large surpluses of the Asian developing economies. Adopting a panel-regression approach, as in Chinn and Prasad (2003. Medium-term determinants of current accounts in industrial and developing countries: an empirical exploration. Journal of International Economics 59, 47e76), we find that the Asian surpluses are well explained by a model that incorporates, in addition to standard determinants, the impact of financial crises on current accounts. However, our model, even when augmented by measures of institutional quality, fails to explain the large U.S. current account deficit. Published by Elsevier Ltd. JEL classification: F21; F32 Keywords: Current account; Financial crisis; Capital flows

1. Introduction The pattern of global current account imbalances has received considerable attention in recent years. Most prominently, the U.S. current account deficit has widened to over 6% of GDP, and this deficit is mirrored in some equally marked surpluses on the part of some of the United States’ trading partners. The current account balances of major developing East Asian * Corresponding author. Tel.: þ1 202 452 3931; fax: þ1 202 736 5638. E-mail address: [email protected] (J.W. Gruber). 0261-5606/$ - see front matter Published by Elsevier Ltd. doi:10.1016/j.jimonfin.2007.03.003

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

501

economies (China, Hong Kong, Indonesia, Korea, Malaysia, Philippines, Singapore, Taiwan, and Thailand) have moved from a small aggregate deficit in 1995 to a surplus of over 5% of GDP in recent years. More generally, the aggregate current account balance of the developing countries moved into surplus starting in 2000. The present pattern of external imbalances appears inconsistent with the standard view that mature industrial economies should be exporting capital to developing countries. Because developing countries have higher labor/capital ratios, they should in principle have higher marginal productivities of capital and thus attract capital from relatively labor-scarce industrial economies. Moreover, if developing countries can expect faster income growth as they catch up to industrial countries, their prospects for a steep upward-sloping earnings’ profile provide an incentive for them to borrow against their higher future income, also leading to current account deficits. There is no consensus explanation for the current pattern of international capital flows, and many hypotheses have been put forward: U.S. fiscal deficits; declines in U.S. private saving; the surge in U.S. productivity growth; increases in global financial intermediation; a global savings glut; a rash of emerging market financial crises; and exchange rate pegs by our trading partners. However, many of these factors are quite amorphous, and it has been difficult to muster support for one explanation against another. By the same token, it has been even more difficult to assess all of the proposed factors jointly and compare their separate contributions to the international pattern of current account imbalances. Ferguson (2005) uses the Federal Reserve staff’s open economy macroeconomic simulation model to gauge the contribution of different shocks to the U.S. trade deficit; the analysis identifies a role for the rise in U.S. productivity growth, a fall in the risk premium on dollar assets, and the weakening of foreign domestic demand. Nevertheless, the simulation model is unlikely to precisely capture the relationships determining the external balance, and identifying the magnitude of the shocks affecting the trade deficit is both difficult and subjective. In this paper, we adopt a more empirical but less structured approach to explaining the recent emergence of large current account imbalances. Our approach jumps off from the influential research of Chinn and Prasad (2003).1 Chinn and Prasad developed a large multi-country database for the period 1971e1995 and estimated a battery of cross-sectional and panel regressions relating current account/GDP ratios to a wide range of potential determinants, including, among others, the fiscal deficit, net foreign asset position, per capita income, output growth, demographic variables, terms of trade volatility, and openness to trade. Based on a similar approach, but with altered specifications and data extending through 2003, we estimate panel regression models to explain the current account/GDP ratios of 61 countries. We use these models to assess preliminary hypotheses regarding the determinants of current account balances, and we also assess the extent to which our models can explain the international pattern of current account balances that has emerged in recent years, particularly the U.S. deficit and developing Asian surpluses.2

1 This research has been updated in Chinn and Ito (2005), in work conducted at the same time, but independently of ours. 2 Our model imposes no long-run solvency condition. As such, there is no guarantee that the model is consistent with a long-run equilibrium. Thus, we are implicitly assuming that long-run solvency concerns play no role in the determination of current account balances.

502

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

To summarize our key results, we find that a regression model comprising the standard determinants of the current account identified in the literaturedper capita income, output growth, fiscal balances, net foreign assets, economic openness, and demographic variablesdcan explain neither the large U.S. current account deficit of recent years nor the large developing Asian surpluses. But when this standard model is augmented by a variable representing the incidence of financial crisis, it does a good job of explaining the emergence of developing Asia’s current account surpluses. However, the model remains unable to explain the recent large U.S. deficits, even when it is further augmented by several explanatory variables designed to capture the special attractiveness of U.S. financial markets: (1) a measure of institutional quality (which enters significantly in the regressions), and (2) a measure of financial development (which does not). In sum, our results lend support to the view that the financial crises of the late 1990s contributed to Asian current account surpluses and to what Bernanke (2005) has referred to as a ‘‘global saving glut’’, but it remains an open question why those surplus savings ended up in the United States rather than being spread more evenly throughout the world. The plan of this paper is as follows. Section 2 reviews different explanations that have been put forward for the emergence of large external imbalances in recent years. Section 3 describes the methodology and data used in our panel regression models of the current account/GDP ratio. The estimation results and the ability of the basic model to predict current account balances are addressed in Section 4, while Section 5 focuses on the role of government institutions and financial development in the determination of the current account. Section 6 decomposes our model’s predictions of current account balances into the contributions of the various explanatory variables, while Section 7 addresses the robustness of our estimates to alternative specifications. Section 8 concludes. 2. Explanations for the pattern of large external imbalances The proximate causes of the widening of the U.S. current account deficit and corresponding widening of our trading partners’ deficits are reasonably obvious: the rise in the dollar between 1995 and early 2002 (which has given up only part of its gains since then); the pickup in U.S. real GDP growth relative to that of its trading partners; and the higher elasticity of U.S. imports with respect to income than of U.S. exports with respect to foreign income (the Houthakkere Magee effect).3 Yet, listing these proximate factors leaves open the more fundamental causes of the deficit. What caused the appreciation of the dollar? Why has U.S. growth translated into so much more imports than has foreign growth? This section reviews some of the answers to these deeper questions. 2.1. Expansion of the fiscal deficit The simultaneous emergence of fiscal and current account deficits in the United States in the mid-1980s gave rise to the ‘‘twin deficits’’ hypothesis. (See, among others, Truman, 2004, and Gramlich, 2004.) Some observers see the U.S. budget deficit as an important factor in the U.S. economy’s more recent move into external imbalance (Cline, 2005; Chinn, 2005; Chinn and Ito, 2005). However, analysis using simulation models suggests that the budget deficit may not have played a central role (Erceg et al., 2005a; Ferguson, 2005). 3

See, among others, Chinn (2004), Mann (1999, 2002, 2004), and Roubini and Setser (2004).

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

503

2.2. Declines in private saving rates Along with the slide in public saving rates, the decline in U.S. private saving since the mid1990s could help to explain the widening of the United States’ current account deficit. However, it is not clear whether the decline in saving has been autonomous, perhaps reflecting financial innovations that have made it easier for Americans to borrow, or the endogenous response to other developments; for example, Bernanke (2005) argues that an excess of saving abroad has simultaneously boosted the U.S. current account deficit and depressed U.S. private saving. 2.3. U.S. productivity surge The rise in the growth rate of U.S. labor productivity since the mid-1990s likely boosted perceived rates of return on U.S. assets, generating capital inflows and buoying the dollar. Expectations of higher rates of return may also have motivated greater domestic investment, and consumption may have been supported by increases in stock prices and perceived long-run income. As shown in simulation experiments described in Erceg et al. (2005b) and Ferguson (2005), all these developments may have contributed to larger trade deficits. 2.4. Expanding global financial intermediation In recent years, the correlation between national saving and investment ratesdthe so-called FeldsteineHorioka paradoxdhas declined, suggesting that national savings are being used to finance investment to a lesser extent than in the past (Blanchard and Giavazzi, 2002; Gruber, 2004). Additionally, there is considerable evidence that the extent of home bias in portfolio allocationdthat is, the tendency for portfolios to be overweight domestic assetsdis declining (Ahearne et al., 2004). Greenspan (2003) has suggested that these trends signal improvements in international financial intermediation which allow larger external imbalances to be financed than in the past. Analyses by Blanchard et al. (2005) and Edwards (2005) also accord a role to reduced home bias in the widening of U.S. deficits. However, the increased ability of the international financial system to move capital across borders does not, by itself, mean that it is the United States that would be expected to exploit that improved ability. Observers point to various strengths of the U.S. economydits favorable investment climate, deep and sophisticated financial markets, protections of investor rights, and solid rates of returndas making it likely that the United States would attract international capital once it became available. 2.5. Global savings glut/emerging market financial crises Bernanke (2005) argues that the large U.S. current account deficit owes importantly to a surge in the availability of saving from overseas. He notes that much of the increased flow of foreign saving has come from developing countries, a development he attributes in large part to the series of financial crises experienced in the past decade. Emerging market financial crises may generate current account surpluses (or lower deficits) through several channels: the economy may lose access to foreign credit; financial intermediation within the economy may become obstructed, causing a credit crunch; balance sheet problems among firms and consumers may restrain domestic spending; and authorities may respond to the weakness in

504

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

domestic demand by taking actions to keep the exchange rate competitive so as to maintain external demand. As discussed in Kamin (2005), all of those factors, to greater and lesser degrees, were involved in the Asian developing countries’ swing into surplus after 1997. Of course, as in the increased global financial intermediation story, the increased flows of capital from developing countries cannot, by themselves, explain the rise in the U.S. current account deficitdin principle, other industrial economies might also have increased their net imports of capital. Again, to explain the rise in U.S. deficits, one must explain why investors found the United States to be a particularly attractive target for their funds. 2.6. Developing Asian exchange rate intervention Whether described as the ‘‘revived Bretton Woods system’’ (Dooley et al., 2003, 2004) or ‘‘co-dependency’’ (Mann, 2004), one explanation put forward for the large U.S. current account deficits is, simply, that some of our trading partners (mainly Asian) have been intervening to keep their currencies competitive and promote their own growth. In principle, this explanation does not differ greatly from the savings glut/financial crisis story in Section 2.5, above. Dooley et al., however, argue that intervention to keep exchange rates competitive and produce current account surpluses is in the interest of all developing countries, with the implication that such surpluses may be with us for some time to come. 2.7. Rising oil prices This factor is not, in some sense, as ‘‘fundamental’’ as the other explanations for the U.S. deficit reviewed above. Nonetheless, owing mainly to higher oil prices, between 1996 and 2004, U.S. imports of oil rose by nearly $110 billion, or about 1% of GDP. 3. Empirical methodology and data Our empirical research will not shed light on all of the explanations for the large U.S. current account deficit surveyed above. It should, however, bear on at least some of those stories. Our sample consists of observations on 61 countries over the period between 1982 and 2003. Since our primary interest is the medium and long term determinants of current account balances, we consider multi-year averages of annual observations as in Chinn and Prasad (2003).4 Averages were constructed over 1982e1986, 1987e1991, 1992e1996, and 1997e 2003, giving us four period observations for each of our 61 cross-sections. For some series a lagged observation was created based on the 1977e1981 average. In cases of missing annual observations, averages were calculated based on the remaining years in a period, allowing a larger sample of developing countries than would have otherwise been possible. Our primary method of analysis consists of panel regressions with the ratio of the current account balance to GDP as the dependent variable. For all of the regressions we include a period fixed effect, thereby allowing the average current account balance to GDP ratio across the 4

The consideration of multi-year averages allows us to abstract somewhat from current account dynamics driven by the business cycle. By focusing on the medium-term, we are able to better identify current account dynamics associated with changes in fundamental determinants rather than dynamics that might result from various frictions and related adjustment processes.

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

505

cross-section to vary from period to period. We generally do not include country fixed effects, since controlling for country-specific means would remove much of the cross-country differences in current accounts that we are interested in explaining. (Section 7 describes estimates that incorporate country fixed effects.) To calculate the current account to GDP ratio, we use the current account measured in U.S. dollars and we convert GDP to dollars using the annual average market exchange rate. We briefly review the explanatory variables in our models below. In that the current account is inherently a relative measure, most of the independent variables were calculated as ratios to sample averages. 3.1. Per capita income As noted above, economic theory predicts that labor-intensive capital-poor developing countries should be net importers of capital, and hence run current account deficits, while capitalrich developed countries should export capital and run current surpluses. To capture this dynamic, we include the ratio of real per capita income to its sample mean in our regressions. 3.2. Changes in growth rates An increase in the growth rate of productivity relative to other countries should be associated with a more negative current account balance, as an increase in the return on capital increases investment and the potential for higher future income decreases saving. In our model, changes in productivity growth are proxied by the change in the growth rate of real per capita income; this measure is available for many countries, although it does not control for changes in labor force participation or hours worked. The growth rates were constructed as the multi-year average of the difference of annual growth from the GDP-weighted sample mean. In some regressions, we control for changes in labor force participation by using per employee GDP rather than per capita GDP. 3.3. Fiscal balance The fiscal balance was calculated as a ratio to GDP and expressed as the difference from the GDP-weighted mean ratio for the sample. 3.4. Net foreign assets A country’s net foreign asset (NFA) position directly affects its net investment income, and therefore its current account balance. Because the NFA position is, in effect, the accumulation of past current account balances, its lagged value, expressed as a ratio to GDP, is entered into the regressions to avoid correlation with the dependent variable. 3.5. Demographics The life-cycle theory of consumption and saving predicts that young households borrow, middle-age households save for retirement, and households in retirement dissave. Therefore relatively young and relatively old countries are both more likely to run current account deficits. To capture these effects, we included both the youth dependency ratio and the old-age

506

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

dependency ratio in the regressions. The youth dependency ratio was defined as the ratio of the population ages 0e14 to the working age population (ages 15e64), while the old-age dependency ratio was defined as the ratio of the population 65 and older to the working age population. Both measures were entered as deviations from a GDP-weighted sample mean. 3.6. Openness Openness was identified by Chinn and Prasad (2003) as a potential determinant of the current account. We define openness as the sum of imports and exports relative to GDP. 3.7. Oil balance To capture the effect on current accounts of terms-of-trade shocks related to oil prices, we included the nominal oil balance to GDP ratio. Although the oil balance is a component of the overall current account balance, changes in the oil balance will not have a one-for-one impact on the current account if the non-oil balance also responds to oil price shocks. 3.8. Financial crises Our financial crisis variable is based on a list of systemic banking crises developed by Caprio and Klingebiel (2003), who identify a banking crisis as systemic by the extent to which bank capital in a country is exhausted.5 To construct our financial crises variable, we first construct a dummy variable that takes on the value of one if a country is in crisis and zero otherwise. As with many of the explanatory variables we consider, however, a financial crisis should affect a country’s current account only when measured in relative terms; if all countries are in crisis simultaneously, there should be no impact on current account balances. Accordingly, we transform our crisis dummies into relative measures by subtracting from the value of each country’s annual zero/one crisis dummy a GDP-weighted average of these dummies, aggregated across all the countries in our sample. The GDP-weighted annual averages thus represent the percentage of the aggregate sample GDP in crisis in each year, represented in Fig. 1. The aggregate series rises in the early 1980s with the Latin American debt crisis and again in 1991 with the Japanese banking crisis. The aggregate index is relatively flat through the 1990s as several Nordic and Latin American crises fade around the time that the Asian financial crisis begins. In addition to the financial crisis variable, we also include as an explanatory variable in our model an interaction term, the financial crisis variable multiplied by the openness variable. The rationale for this is that more open economies are likely to have larger tradable goods sectors and thus be able to adjust their external balances more flexibly in response to financial crisis.6 Finally, it is worth emphasizing that the financial crises in the Caprio and Klingebiel data set are measures of banking crises, not currency crises such as those identified by Frankel and Rose 5

Our one deviation from the Caprio and Klingebiel list is that their list identified China having a banking crisis in the ‘‘1990s’’, but given that the country’s banking problems have not been resolved, we extend that crisis through the present. 6 For example, consider two economies, both with trade deficits of 5% of GDP, and both of whom experience declines in imports and increases in exports of 10% following a financial crisis. In a very open economydsay, with imports of 60% of GDP and exports of 55%dimports would fall to 54% of GDP and exports would rise to 60.5% of GDP, causing the trade balance to swing to a surplus of 6.5% of GDP. Conversely, in a more closed economydsay, with imports of 20% of GDP and exports of 15%dthe trade deficit would narrow but not close entirely.

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

507

0.3

0.25

0.2

0.15

0.1

0.05

19 7 19 6 77 19 7 19 8 7 19 9 80 19 8 19 1 82 19 8 19 3 84 19 8 19 5 8 19 6 87 19 8 19 8 8 19 9 90 19 9 19 1 92 19 93 19 94 19 9 19 5 9 19 6 97 19 9 19 8 9 20 9 00 20 0 20 1 02 20 03

0

Fig. 1. GDP-weighted financial crisis dummy.

(1996) or Kaminsky and Reinhart (1999). Although banking crises frequently give rise to currency crises (Kaminsky and Reinhart, 1999; Glick and Hutchison, 1999), it would be inappropriate to use currency crises per se in our model. Such measures often signal a sharp exchange rate depreciation, thus representing a proximate determinant of the current account rather than a more fundamental factor. Moreover, currency crises often reflect sharp reversals of capital flows, but such reversals are the capital-account counterparts to swings of the current account, not an independent cause of them. The fact that banking and currency crises often come together raises the question of whether it is possible econometrically to identify the effects of the former rather than the latter. We used the methodology defined by Kaminsky and Reinhart (1999) to identify the currency crises that occurred in our sample countries and compared their timing with that of our banking crises.7 We found that of 1403 total annual observations during the 1982e2003 period in our sample, 148 were associated with banking crises alone, 73 with currency crises alone, and 38 with banking and currency crises simultaneously. Accordingly, there appears to be sufficient independent variation in banking and currency crises to identify their separate effects on current account balances, as evidenced by estimation of an alternative specification described in Section 7 below. 3.9. Quality of government institutions As noted above, one explanation for the attractiveness of the United States to foreign investors is the quality of U.S. institutions, which both protects investor rights and enhances rates of return. In order to capture the importance of institutions to capital flows, we include in a subset 7 For countries analyzed in Kaminsky and Reinhart (1999) and Kaminsky (2003), we used the actual dates presented in those papers. For other countries in our sample, we identified crises by applying their methodology to IFS reserves and exchange rate data.

508

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

of our regressions a set of indicators of the quality of government institutions described in Kaufmann et al. (2005).8 The Kaufmann et al. data set consists of four annual observations (1996, 1998, 2000, and 2002) for each country, estimated using an unobserved components model with data from a wide array of sources. For each country we matched the average of the 1996 and 1998 indicators to the 1992e1996 period of the panel and the average of the 2000 and 2002 values to our 1997e2003 period of the panel. 3.10. Level of financial development Another explanation for the attractiveness of the United States to foreign investors is the depth and breadth of U.S. financial markets, which offers greater liquidity, security, and opportunities for diversification than most economies. By the same token, analysts have suggested that lack of financial development in other, less mature economies limits investment opportunities and hence encourages capital outflows. (See Prasad et al., 2006; Ju and Wei, 2006.) As in Chinn and Ito (2005) and Aghion et al. (2006), we use the ratio of private credit to GDP, expressed as a deviation from its GDP-weighted sample mean, as a proxy for financial development. (The data are drawn from Beck et al., 2000, revised 2006.) 4. Basic estimation results Before proceeding to the econometric estimates, Table 1 provides a brief snapshot of the data. Note that the data are expressed as unweighted sample averagesdthus, ratios of the current account to GDP may average positive or negative, even though GDP-weighted global averages would have to equal zero (abstracting from the global current account statistical discrepancy). The four right-most columns divide the sample into large deficit and surplus countries, based on current accounts in the 1997e2003 period. The data line up in accord with most of the hypotheses described in Section 2. Large surpluses are associated with: higher income levels, larger declines in GDP growth, smaller budget deficits, and more financial crises. However, at least in these bilateral comparisons, large surpluses are also associated with better government institutions and higher levels of private credit, contradicting views that these factors help explain the large U.S. deficit. Table 2 presents the econometric estimation results for several variants of our model. As described above, the model is estimated over data organized into period-averages for four different periods: 1982e1986, 1987e1991, 1992e1996, and 1997e2003. Both a constant and separate time dummies for each period are included in the model but not shown. For each explanatory variable, the first row represents the coefficient estimate and the second row represents the t-statistic; coefficients that are statistically different from zero at the 90% level or higher are indicated in bold. In the regressions shown in Table 2, we have excluded Singapore from the sample, as it is an extreme outlier in terms of current account history (an average surplus of roughly 20% of GDP in the 1997e2003 period) and net foreign assets (reaching 155% of GDP in 2003), and it also is unusually influential in some specifications of our model; the robustness of our results to this exclusion is discussed below. 8 Introducing institutional measures into a cross-country model of current accounts is novel; the only other research we are aware of that does this is Chinn and Ito (2005). They use different measures of institutional quality and draw from the International Country Risk Guide (ICRG).

Total sample

Unweighted sample averages Current account (% of GDP) Per capita income ($) DGrowth per capita Government deficit (% of GDP) Lagged NFA (% of GDP) Elderly ratio Youth ratio Openness Oil balance (% of GDP) Private credit (% of GDP) Government institutionsc Number of financial crises a b c

Large CA surplusesa

Large CA deficitsb

1982e1986

1987e1990

1991e1996

1997e2003

1991e1996

1997e2003

1991e1996

1997e2003

2.3 9574 0.2 4.8 16 0.12 0.56 0.67 0.4 50

1.0 10 807 1.1 2.7 23 0.12 0.52 0.69 0.7 56

1.5 11 770 0.3 2.6 21 0.13 0.48 0.73 0.7 62 0.59

0.4 13 466 0.9 1.7 21 0.14 0.44 0.81 0.3 73 0.58

0.4 16 504 0.0 0.9 6 0.13 0.42 1.02 2.3 97 0.80

5.8 18 752 1.8 0.3 4 0.14 0.38 1.14 1.7 99 0.78

3.5 7872 1.0 2.6 45 0.12 0.52 0.63 1.4 47 0.48

4.1 9126 0.6 2.2 42 0.13 0.48 0.67 1.7 65 0.52

13

17

11

15

4

4

5

8

Large surplus defined as over 2% of GDP in 1997e2003 period. Sample includes 19 countries. Large deficit defined as over 2% of GDP in 1997e2003 period. Sample includes 19 countries. Average of Government Institution Indicators (ranging between 2.5 and 2.5, with a higher value indicating superior institutions).

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

Table 1 Summary statistics

509

510

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

Table 2 Main results

Per capita GDP DGrowth Fiscal balance NFA Youth ratio Elderly ratio Openness Oil balance

1

2

3

4

5

0.012 3.073 0.095 1.211 0.113 1.688 0.018 1.733 0.032 2.623 0.135 2.294 0.008 1.308 0.089 5.301

0.013 2.830 0.012 0.167 0.115 2.220 0.021 2.392 0.014 0.679 0.096 1.870 0.003 0.410 0.085 5.131

0.012 3.072 0.020 0.280 0.119 1.998 0.024 2.413 0.022 1.364 0.114 2.351 0.012 3.442 0.081 4.630 0.021 2.253 0.069 4.764

0.014 2.718 0.003 0.040 0.116 2.383 0.024 2.599 0.015 0.758 0.111 2.343 0.010 2.655 0.080 4.449 0.027 2.598 0.070 2.555 0.032 1.597 0.033 0.874 0.032 1.823 0.001 0.059 0.009 0.177 0.026 1.618 0.024 0.297 0.034 0.929 0.050 2.883

0.013 2.969 0.018 0.240 0.119 2.077 0.024 2.416 0.021 1.388 0.120 2.538 0.012 3.183 0.079 4.653 0.022 2.501 0.070 4.901

234 0.348 0.029

234 0.357 0.029

Financial crisis Financial crisis*Openness China (1997e2003)

0.041 2.149 0.016 0.378 0.053 3.754 0.026 1.216 0.098 3.169 0.062 10.606 0.036 0.428 0.082 2.706 0.045 2.724

Hong Kong (1997e2003) Indonesia (1997e2003) Korea (1997e2003) Malaysia (1997e2003) Philippines (1997e2003) Taiwan (1997e2003) Thailand (1997e2003) U.S. (1997e2003) #Obs R2 SER

234 0.286 0.030

234 0.339 0.029

234 0.351 0.029

0.051 3.102

Panel regression with unreported constant and period fixed effects. 59 cross-sections and 4 periods. t-statistic reported underneath coefficient. Bold text indicates significance at the 10% level.

Column 1 indicates the results from the simplest model, which contains standard variables posited by the literature as helping to determine the current account balance. All of the coefficients have the expected sign, and most of them are significantly different from zero or nearly so. Consistent with the averages shown in Table 1, larger current account balances (where

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

511

positive indicates a surplus) are associated with higher per capita incomes, declines in growth, higher fiscal balances, higher net foreign asset positions, lower shares of youth and elderly in the population, greater degrees of economic openness, and a positive nominal oil balance. The coefficient on the fiscal balance variable is quite low, at 0.11; however, our estimate is not all that far from many of the coefficients estimated by Chinn and Prasad (2003), Chinn and Ito (2005) or Bussiere et al. (2004, 2005), which range from less than 0.1 to a little over 0.4. The coefficient on per capita output growth also is small (and statistically insignificant), but this is consistent with Chinn and Prasad (2003) and Chinn and Ito (2005), who use a slightly different specification. Bussiere et al. (2005) find a larger and significant negative impact on the current account, but using productivity growth rather than output growth and using data only for industrial economies. (See, also, Glick and Rogoff, 1995.) As discussed in Section 7 below, using a more specific measure of productivitydoutput per employeeddoes not improve these results and requires shrinking the number of observations. The coefficient on the nominal oil balance is positive, as expected, but is far smaller than unity. Apparently, variations in the oil balance are largely offset by variations in the non-oil balance, so current account balances remain relatively unaffected. How well does the model shown in Column 1 capture the most salient aspects of the recent international pattern of current account imbalances? Column 2 presents the same model, but with separate dummy variables for the United States and for each of the major developing Asian countries in the sample during the 1997e2003 period. These separate dummies have the effect of essentially excluding these observations from the regression, while the coefficients on these dummies represent the difference between their actual current account balances and the model’s prediction. These coefficients indicate that for China, Indonesia, Malaysia, the Philippines, and Thailand, current account surpluses during 1997e2003 were significantly above the model prediction. For the United States, the current account balance was significantly below the model prediction, with the miss being more than 4 percentage points of GDP. Thus, the standard factors identified by the literature cannot explain the recent global pattern of current account imbalances. Column 3 represents the same basic model as in Column 1, but with the addition of the financial crisis and (financial crisis)*(openness) variables described in Section 3. The coefficients on both variables are statistically significant, but have opposite signs. Their values imply that for countries with an openness ratio of 0.3 or higher, a financial crisis raises the current account balance; in our sample, 37 of the 40 developing countries had openness ratios exceeding this level, including all of the Asian countries. Compared with the model shown in Column 1, the addition of the financial crisis variables raises the adjusted R2 from 0.29 to 0.35 and reduces the standard error of the regression slightly as well. The results shown in Column 3 represent prima facie evidence that financial crises have a systematic positive effect on current account balances, and this may help explain developing Asia’s recent surpluses. However, an alternative possibility is that the significant coefficients on the financial crisis variables are being driven exclusively by the behavior of the developing Asian nations in the 1997e2003 period. It is possible that, absent those observations, financial crises would not have been estimated to have pronounced effects on the current account. Column 4 presents the estimation results when, as in Column 2, we add separate dummy variables for each of the developing Asian economies (as well as the United States) in the 1997e2003 period, thus effectively removing them from the estimation sample. The resultant coefficients on the financial crisis variable are remarkably similar to those in Column 3. The coefficient on the (financial crisis)*(openness) variable declines in significance, as would be expected by the

512

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

removal of a large number of observations combining both financial crises and large current account surpluses, but remains significantly different from zero at the 95% level. Thus, the regression supports our portrayal of financial crises as a general factor boosting current account balances, not something specific to the developing Asian countries in recent years. Moreover, the coefficients on the Asian dummy variables are now generally smaller than they were in the model shown in Column 2dwhich does not include financial crisis variablesdand insignificant, indicating that the model now predicts the large developing Asian surpluses of the 1997e2003 period. This represents significant evidence that those surpluses were importantly associated with the on-going effects of the financial crisis that started in 1997. Importantly, while the models in Columns 3 and 4 explain much of the Asian current account surpluses in recent years, they offer no improved ability to predict the U.S. deficit; as with the models that do not include the financial crises variables, they continue to predict U.S. balances that are about 5 percentage points of GDP too high. This is confirmed in Column 5, where only a dummy variable for the United States in the 1997e2003 period is included. 5. Institutional quality, financial development, and the U.S. current account balance To improve the model’s ability to capture the factors underlying foreign investors’ strong attraction to the United States in recent years, we first added to the model the composite indicator of the quality of government institutions outlined in Kaufmann et al. (2005). This measure is an average of six separate sub-indexes: voice and accountability; political stability; government effectiveness; regulatory burden; rule of law; and control of corruption. As noted above, this measure is only available biennially for 1996 onwards, and thus requires restricting estimation of the model to the last two panels: 1991e1995 and 1997e2003. To provide a baseline against which to compare results, Column 1 of Table 3 replicates the model shown in Column 5 of Table 2, but estimated only over the 1992e1996 and 1997e2003 periods; this model includes the financial crisis variables but not the government institutions index. By and large, restricting the sample to the last two periods does not greatly change the estimates, although the negative coefficient on output growth becomes substantially larger and significant. Column 2 introduces the government institutions variable to the model. In contrast to the unconditional positive correlation suggested in Table 1, the coefficient on this variable is negative and significant at the 99% level (increases in the index indicate better institutional quality); this is consistent with the hypothesis that more market-friendly and stable government institutions attract foreign capital, lowering the current account balance.9 The introduction of the government institutions variable does not, however, explain the U.S. current account deficitdthe coefficient on the dummy variable remains large and significant. Because only some of the sub-indexes in the composite indicator of government institutions may affect the current account balance, in Column 3 the sub-indexes are entered separately. Only two of the six indicators affect the current account balance negatively, and these two indicatorsdregulatory burden (measured so that increases in the index represent decreases in burden) and rule of lawdare statistically significant. These results are telling, as low regulatory 9

Of course, it is also possible that capital inflows help improve institutions. Chinn and Ito (2005) find similar results to hold, mainly for the high-income subset of their country data set.

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

513

Table 3 The model with governance indicators and financial development

Per capita GDP DGrowth Fiscal balance NFA Youth ratio Elderly ratio Openness Oil balance Financial crisis Financial crisis*Openness

1

2

3

4

5

6

0.019 3.228 L0.179 2.013 0.157 3.984 0.019 2.827 0.008 0.365 0.092 1.230 0.007 1.781 0.072 2.700 L0.015 2.568 0.066 5.351

0.027 3.365 0.153 1.399 0.173 3.555 0.013 1.990 0.043 2.581 0.071 0.875 0.013 3.006 0.087 3.681 0.025 4.373 0.072 6.181 0.018 4.057

0.022 3.209 0.222 1.662 0.099 1.551 0.014 2.332 0.041 1.994 0.108 2.050 0.015 2.936 0.068 2.078 0.018 2.407 0.061 3.119

0.022 3.623 0.212 1.634 0.129 2.212 0.014 2.316 L0.038 2.523 0.074 1.187 0.014 3.028 0.068 2.172 L0.019 6.159 0.065 4.330

0.026 3.637 0.164 1.461 0.176 3.660 0.014 2.955 L0.046 2.005 0.062 0.852 0.015 3.099 0.082 2.502 L0.021 4.411 0.070 6.216 L0.017 4.174

0.027 3.426 0.153 1.274 0.173 3.829 0.013 3.005 L0.043 1.954 0.071 0.882 0.013 2.719 0.087 2.623 L0.025 5.443 0.072 6.109 L0.018 3.698

0.006 0.620

0.000 0.004 L0.055 2.437

117 0.476 0.026

117 0.489 0.026

Government institutions Voice

0.002 0.660 0.004 0.699 0.014 1.649 L0.027 6.033 L0.012 1.698 0.002 0.183

Stability Government effectiveness Regulatory Rule of law Corruption

L0.020 7.679 0.003 1.104

Financial development U.S. (1997e2003)

L0.058 3.562

0.055 3.595

0.049 6.246

L0.050 4.342

#Obs R2 SER

117 0.473 0.027

117 0.494 0.026

117 0.497 0.026

117 0.506 0.026

Panel regression with unreported constant and period fixed effects. 59 cross-sections and 2 periods. t-statistic reported underneath coefficient. Bold text indicates significance at the 10% level.

burdens are often cited as the category in which the U.S. economy stands out, not only relative to developing countries, but relative to many industrial economies as well. Nevertheless, as indicated by the coefficient on the U.S. dummy in Column 3, even with the government institutions variable disaggregated into its component parts, the model still substantially

514

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

overpredicts the U.S. current account deficit (although the error is less than without any governance indicators, as shown in Table 3, Column 1). Column 4 removes all of the sub-indexes except for the measures of regulatory burden and role of law, but this has little effect on the model. To assess the role of another factor that has been proposed to explain the particular attraction of the United States to foreign investorsdthe depth and breadth of U.S. financial marketsd Column 5 presents estimates of a model that also includes the ratio of private credit to GDP, a proxy for financial development. The coefficient on the financial development variable is negative, as expected, but very small and insignificant. In Column 6, addition of a dummy variable for the United States in the 1997e2003 period turns the coefficient on the financial development variable to zero, and also shows that the large U.S. deficit remains unexplained.10 We do not view our results as conclusive evidence against the role of financial development in determining current account balances, but clearly more research would be needed to establish this connection. (Chinn and Ito (2005) find the private credit variable to negatively affect the current account balance, but not in all specifications nor with a large coefficient.) 6. Decomposition of model predictions Based on the model shown in Table 3, Column 5, Fig. 2 performs two functions: First, the lines connecting the squares and diamonds compare the model’s prediction of the path of the U.S. current account with its actual values for all four periods; the model-fitted predictions are represented by black squares and the actual values by blue diamonds. (The results for the 1982e1986 and 1987e1991 periods are based on the coefficients estimated over the 1992e1996 and 1997e2003 periods, with the institutions measure fixed at its 1992e1996 period level and the period fixed effect set to zero.) Second, the multi-colored and textured columns decompose the model’s prediction into the contributions of the different explanatory variables. Each bar in a column represents the value of an explanatory variable multiplied by its respective coefficient, with positive contributions plotted above the zero line and negative contributions below it. The period fixed effect is included in the constant term, indicated by the green checkerboard bars. The effects of the financial crisis variable and its interaction term have been combined into a single contribution (blue cross-hatch), as have the effects of the youth and elderly dependency ratios (light blue horizontal lines). The sum of the contributions is equal to the model’s predicted value, as positive and negative contributions net out. The chart confirms that the model’s predictions track neither the level of the U.S. deficit in the 1997e2003 period nor its deterioration from previous periods. The columns of contributions offer some insight into why the model cannot explain the deterioration of the U.S. deficit in the 1997e2003 period. First, the NFA (gray), government institutions (wavy fuschia lines), and financial development (purple) variables do pull down the model’s prediction of the current account balance, but only by about 3 percentage points of GDP; moreover, these variables change relatively little between 1992e1996 and 1997e2003. Second, the fiscal variable (peach dots) actually imparts a slight positive contribution to the predicted current account during the 10 Estimates of the model that do not include the government institutions variable, and therefore which can be estimated over all four periods (as in Table 2), also show small and insignificant coefficients on the financial development variable. See the working paper version of this article, Gruber and Kamin (2005).

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522 Financial Development Financial Crisis Institutions Oil

Openness Demographics NFA

Fiscal Balance Growth Per Capita Income

515

Constant Predicted Actual

0.08

0.06

0.04

0.02

0

-0.02

-0.04

-0.06

-0.08

1982 - 1986*

1987 - 1991*

1992 - 1996

1997 - 2003

*1982 - 1986 and 1987 - 1991 backcast using Table 3 - Regression 5 coefficients, the 1992-1996 value for the institution variable, and no period fixed effect.

Fig. 2. United Statesdcontribution to predicted current account/GDP (Table 3dRegression 5).

1997e2003 period as the average U.S. fiscal balance at that time slightly exceeded the GDPweighted sample average; deficits in the latter portion of the period were largely offset by surpluses in the earlier part. Finally, the relative per capita income variable (solid pale blue) adds over 5 percentage points to the model’s current account prediction, reflecting the U.S. position as a mature industrial economy. Fig. 3 repeats the same analysis as in Fig. 2, but for a GDP-weighted average of the key Asian developing economies (excluding Singapore). The figure confirms that the model (again, Table 3, Column 5) not only predicts the level of the developing Asian current account balance reasonably well in 1997e2003, but also tracks its upswing from the 1992e1996 period. As might be expected, the financial crisis variables account for much of that rise, as does the decline in the relative growth performance (green vertical lines) of the Asian economies. Fig. 4 extends the analysis across the individual Asian countries in the final 1997e2003 period. The model’s predictions for the individual Asian countries are quite good, with the actual and predicted values often quite close to each other. The financial crisis variable contributes positively to the predicted balance in all cases. The contribution of the institutions variable is positive in China and Indonesia, suggesting that the poor institutional environment in these countries supports capital outflows and therefore current account surpluses. The last two columns of Fig. 4 compare the predicted and actual current accounts for the Asian aggregate and the United States in 1997e2003. The model predicts a slight surplus for

516

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

Fig. 3. GDP-weighted Asian aggregatedcontributions to predicted current account/GDP (Table 3dRegression 5).

the United States instead of the large deficit that actually occurred. Nevertheless, the model successfully predicts a weaker current account balance for the United States than for the Asian aggregate, despite the United States’ much higher per capita income. This result is driven, among other factors, by the positive impact of slowing growth and the financial crisis on the Asian balances, as well as the negative impact of institutional quality on the U.S. balance. 7. Robustness to alternative specifications We estimated the model under several alternative specifications to gauge the robustness of our results. In the first two columns of Table 4, we test whether our financial crisis variable, which is based on identified banking crises, might be proxying for currency crises. This could be problematic, insofar as currency crises represent a more proximate cause of the current account balance, and could indeed be interpreted as endogenous with respect to the current account. In principle, it would be desirable to use instrumental variables estimation to correct for any endogeneity of the financial crisis variable. In practice, the literature on this topic (Kaminsky and Reinhart, 1999; Glick and Hutchison, 1999) suggests it would be difficult to find instruments that are correlated with banking crises but uncorrelated with currency crises. Instead, we assess the estimated effects of financial crises (using our banking crisis proxy) once we control for currency crises. Column 1 indicates that, by itself, the currency crisis variable

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

517

Fig. 4. 1997e2003 contributions to predicted current account/GDP (Table 3dRegression 5).

(interacted with openness) does boost the current account surplus. However, as shown in Column 2, this effect disappears once the financial crisis variables are added to the model; moreover, the coefficients on the financial crisis variables themselves are little changed by inclusion of the currency crisis variables. Column 3 presents estimates of our basic model that include country fixed effects in addition to period fixed effects. With the country fixed effects, the estimated coefficients on several variables become smaller and insignificant, including per capita GDP and the fiscal balance; these are variables that likely show more variation across countries than over time, so that introduction of the country fixed effects diminishes their residual correlation with current account balances. Conversely, coefficients on other variablesdthe change in growth, the oil balanced become larger in absolute magnitude, suggesting that period-to-period variations in these terms have a greater influence on the current account than do differences in average levels of these variables across countries. Similar logic may explain the change in sign of the coefficient on the net foreign assets (NFA) variable from positive to negative. Countries with higher NFA likely have stronger current account balances than countries with lower NFA, both because they earn more investment income and because their NFA reflects past strong current accounts; with country fixed effects included, however, only variations over time in NFA matter, and increases in NFA may be associated with increases in wealth that lead subsequently to higher imports and lower current accounts. Notwithstanding these differences from the model without country fixed effects, the coefficients on the financial crises variables in Column 3 are little changed and still quite significant.

518

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

Table 4 Alternative specifications

Per capita GDP DGrowth

1

2

3

4

0.012 2.820 0.062 0.985

0.013 3.013 0.016 0.222

0.024 1.314 L0.094 2.209

0.022 0.941 0.071 1.289

DGrowth per employee Fiscal balance NFA Youth ratio Elderly ratio Openness Oil balance

0.111 1.811 0.019 1.822 0.046 2.803 0.156 2.820 0.004 0.760 0.080 4.566

Financial crisis Financial crisis*Openness Currency crisis Currency crisis*Openness

0.028 0.938 0.093 1.661

0.111 1.839 0.023 2.311 L0.033 2.014 L0.130 2.820 0.015 3.310 0.078 4.525 L0.025 2.223 0.077 3.683 0.032 1.084 0.041 0.767

5

6 0.014 4.687

0.015 3.975

0.060 0.630 0.074 0.943 0.042 2.590 0.007 0.625 0.172 3.741 0.014 2.346 0.159 4.844 0.019 1.903 0.062 3.269

0.105 1.096 0.064 0.960 0.042 2.730 0.003 0.217 L0.157 3.577 0.008 1.305 0.169 4.186 0.013 0.946 0.030 0.706

0.008 0.135 L0.034 2.550 0.070 1.022 L0.437 4.002 0.042 2.590 0.297 3.004 L0.025 3.884 0.065 5.901

0.008 0.129 L0.035 2.305 0.098 1.369 L0.451 2.777 0.025 1.173 0.302 3.091 0.017 1.375 0.031 1.363

No

0.037 2.363 0.004 0.080 0.048 1.763 0.017 0.587 0.048 0.676 0.043 1.966 0.014 0.190 0.072 1.695 L0.052 2.538 No

186 0.423 0.027

186 0.427 0.027

China (1997e2003)

Country fixed effects

No

No

Yes

0.010 0.995 0.014 0.399 0.056 2.387 0.036 2.151 0.056 1.235 0.042 2.082 0.026 0.696 0.081 3.445 L0.041 9.438 Yes

#Obs R2 SER

230 0.312 0.030

230 0.359 0.029

234 0.663 0.021

234 0.678 0.020

Hong Kong (1997e2003) Indonesia (1997e2003) Korea (1997e2003) Malaysia (1997e2003) Philippines (1997e2003) Taiwan (1997e2003) Thailand (1997e2003) U.S. (1997e2003)

Panel regression with unreported constant and period fixed effects. t-statistic reported underneath coefficient. Bold text indicates significance at the 10% level.

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

519

This suggests that the results shown in Tables 2e3 are not merely reflective of differences across countries, but capture the effects of financial crises in individual countries over time, as well. Column 4 indicates that when the specification includes both country fixed effects and dummies for the developing Asian economies during 1997e2003, the estimated effect of financial crises is halved and the model explains much less of the latter’s surpluses in that period. Our view is that by controlling both for cross-country variation and the significant recent experiences in Asia, too much information is extracted from the data set to allow the effects of financial crises to be estimated reliably. In Columns 5 and 6, the variable representing the change in output growth is replaced by the change in the growth of labor productivity (output per employee). The coefficient on this variable does not come in with the expected negative sign, and the model fails to explain the large U.S. current account deficit in 1997e2003.11 We estimated several additional alternative specifications, the results of which are shown in Gruber and Kamin (2005). First, the net foreign asset variable was replaced by the current account balance in the prior period; this had little effect on the coefficients on the financial crisis variables or their statistical significance. Second, we added Singapore to the data set. In our standard model, this sharply reduced the size and significance of the financial crisis variables, likely reflecting Singapore’s extremely large current account surpluses over the past 1½ decades, even as it stayed out of crisis; however, in a specification where we replaced the net foreign asset variable with the lagged current account balance, the financial crisis variables were again significant. Finally, we estimated a model in which the oil balance variable was replaced by a dummy variable that equaled one for net oil exporters and zero otherwise, but this did not materially affect the results.

8. Conclusion Our empirical work sheds light on several of the explanations for the global pattern of current account imbalances reviewed in Section 2. To begin with, our work provides support for at least half of the global savings glut hypothesis (Section 2.5 above): the half suggesting that the U.S. deficit owes partly to a rise in the savings made available to the United States by its trading partners. Our estimation results indicate that financial crises systematically are associated with higher current account imbalances, and that the wave of crises that swept the developing world in the late 1990s, especially in East Asia, apparently contributed to this shift into surplus. It may seem curious that the financial crises that hit East Asia in 1997e98 could still be affecting the region’s current accounts half a decade later, particularly as these economies regained access to international capital markets reasonably quickly. However, financial crises may affect the saving/investment balance not only through their effects on international financial intermediation, but also through their effects on domestic intermediation and balance sheets. As Japan’s recent experience indicates, such effects may be long-lastingdin fact, our

11

Gruber and Kamin (2005) describe a version of this model that includes the composite government institutions index and is estimated over just the last two periods. In this specification, the growth of output per employee becomes significant and the fit of the model improves a bit. However, the large U.S. deficit remains unexplained, and the number of observations is substantially reduced owing to lack of data on employment for many countries.

520

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

measure of financial crisis indicates continued banking-sector distress in the region throughout the 1997e2003 period. Our finding that financial crises help explain the large East Asian external surpluses tends to undercut the view that those surpluses reflect a rational development strategy and can be expected to persist as part of a ‘‘revived Bretton Woods system.’’ If currency manipulation to achieve surpluses is good development strategy, why did the East Asian economies wait until the late 1990s to start implementing it, why did this implementation coincide so neatly with their financial crises, and why haven’t other regions also adopted this strategy? Our sense is that much of developing Asia’s surpluses will likely dissipate as adjustments following its earlier financial crises are completed, restoring the net flow of capital from industrial economies back to developing ones. For the global saving glut story to explain the large U.S. current account deficit, however, it must also explain why the increase in global savings was tapped primarily by the United States. Our research provides mixed support for several explanations. The U.S. budget deficit (Section 2.1, above) does not appear to explain the U.S. current account deficit, at least for the 1997e 2003 period; not only is the estimated pass-through of the fiscal balance to the current account quite small, but the average budget balance during this period was relatively positive by international standards. There was also little evidence of a significant role for the level of financial development (discussed in Section 2.4). However, we found that output growth generally exerted a significant negative effect on the current account (Section 2.3), as did the quality of government institutions (also Section 2.4), particularly the regulatory regime. The U.S. economy compares favorably to other economies along these dimensions, and this contributes to our model predicting a smaller current account surplus in the 1997e2003 period for the United States than for developing Asia. Even so, however, our model cannot explain the emergence of the large U.S. deficit. How do we interpret our model’s inability to explain the widening of the U.S. external imbalance? First, it is possible that our estimates of the relationships linking fiscal balances, economic growth, and institutional quality to the current account balance, which are based on the experiences of a wide range of countries over two decades, may understate the strength of these linkages in the United States in recent years. Second, the widening of the U.S. deficit may owe to other factors not included in the models. For example, the model does not contain a proxy for autonomous movements in consumption, as it is difficult to think of a quantifiable variable that could capture such movements and which is available for a large number of countries. Thus, more research on this topic is clearly indicated.

Acknowledgements The views expressed in this paper are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or of any person associated with the Federal Reserve System. We would like to thank Menzie Chinn, Marcel Fratzscher, Joseph Gagnon, Philipp Hartmann, Jane Ihrig, Eswar Prasad, John Rogers, Mark Taylor, Rob Vigfusson and seminar participants at the Federal Reserve Board, the IMF, the University of Wisconsin’s April 2006 conference on ‘‘Current Account Sustainability,’’ and the ECB’s summer 2006 conference on ‘‘Financial Globalization and Integration’’ for helpful comments and advice. Benjamin Vallis provided able research assistance.

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

521

Appendix. Data Series

Source

Current account, nominal GDP, imports, and exports Exchange rates, reserves Private credit Per capita income Employment

IFS, OECD

Fiscal balance Net foreign asset position Population data Indicators of institutional quality Financial crisis indicators Oil: Imports, exports, and prices

IFS Beck et al. (2000, revised 2006) World Development Indicators Groningen Growth and Development Centre, OECD GFS, OECD, Asian Development Bank, MOF (Taiwan) Lane and Milesi-Ferretti (2006) UN Kaufmann et al. (2005) Caprio and Klingebiel (2003) Energy Information Agency d Department of Energy

Sample: Argentina, Australia, Austria, Bahrain, Bangladesh, Belgium, Bolivia, Brazil, Canada, Chile, China, Colombia, Costa Rica, Cyprus, Denmark, Dominican Republic, Egypt, El Salvador, Finland, France, Germany, Greece, Guatemala, Haiti, Honduras, Hong Kong, Iceland, India, Indonesia, Ireland, Israel, Italy, Japan, Jordan, South Korea, Malaysia, Mexico, Netherlands, New Zealand, Norway, Oman, Pakistan, Panama, Paraguay, Peru, Philippines, Portugal, Singapore, South Africa, Spain, Sri Lanka, Sweden, Switzerland, Syria, Taiwan, Thailand, Turkey, United Kingdom, United States, Venezuela.

References Aghion, P., Bacchetta, P., Ranciere, R., Rogoff, K., 2006. Exchange rate volatility and productivity growth: the role of financial development. NBER Working Paper # 12117. Ahearne, A., Griever, W., Warnock, F., 2004. Information costs and home bias. Journal of International Economics 62, 313e336. Beck, T., Demirgu¨c¸-Kunt, A., Levine, R., 2000 revised 2006. A new database on financial development and structure. World Bank Economic Review 14, 567e605. Bernanke, B., 2005. The global saving glut and the U.S. current account deficit. Homer Jones Lecture, St. Louis, Missouri. Available from: . Blanchard, O., Giavazzi, F., 2002. Current account deficits in the euro area: the end of the FeldsteineHorioka puzzle? Brookings Papers on Economic Activity 2, 147e209. Blanchard, O., Giavazzi, F., Sa, F., 2005. The U.S. current account and the dollar. NBER Working Paper No. 11137. Bussiere, M., Fratzscher, M., Muller, G.J., 2004. Current account dynamics in OECD and EU acceding countriesdan intertemporal approach. European Central Bank Working Paper Series No. 311. Bussiere, M., Fratzscher, M., Muller, G.J., 2005. Productivity shocks, budget deficits and the current account. European Central Bank Working Paper Series No. 509. Caprio, G., Klingebiel, D., 2003. Episodes of systemic and borderline financial crisis. World Bank Working Paper. Chinn, M.D., 2004. Incomes, exchange rates and the U.S. trade deficit, once again. International Finance 7 (3), 451e469. Chinn, M.D., 2005. Getting serious about the twin deficits. Council on Foreign Relations Special Report 10. Chinn, M.D., Ito, H., 2005. Current account balances, financial development, and institutions: assaying the world ‘savings glut’. La Follette School Working Paper No. 2005-023. Chinn, M.D., Prasad, E.S., 2003. Medium-term determinants of current accounts in industrial and developing countries: an empirical exploration. Journal of International Economics 59, 47e76. Cline, W.R., 2005. The United States as a Debtor Nation: Risks and Policy Reform. Institute for International Economics, Washington, D.C.

522

J.W. Gruber, S.B. Kamin / Journal of International Money and Finance 26 (2007) 500e522

Dooley, M.P., Folkerts-Landau, D., Garber, P., 2003. An essay on the revived Bretton Woods System. NBER Working Paper 9971. Dooley, M.P., Folkerts-Landau, D., Garber, P., 2004. Direct investment, rising real wages and the absorption of excess labor in the periphery. NBER Working Paper 10626. Erceg, C.J., Guerrieri, L., Gust, C., 2005a. Expansionary fiscal shocks and the current account. Board of Governors of the Federal Reserve System, International Finance Discussion Papers No. 825. Erceg, C.J., Guerrieri, L., Gust, C., 2005b. SIGMA: a new open economy model for policy analysis. Board of Governors of the Federal Reserve System, International Finance Discussion Papers No. 835. Edwards, S., 2005. Is the U.S. current account deficit sustainable? And if not, how costly is adjustment likely to be? NBER Working Paper 11541. Ferguson, R.W., 2005. U.S. current account deficit: causes and consequences. Speech to Economics Club of the University of North Carolina, Chapel Hill, North Carolina. Available from: . Frankel, J., Rose, A., 1996. Currency crashes in emerging markets: an empirical treatment. Journal of International Economics 41, 351e366. Glick, R., Hutchison, M., 1999. Banking and currency crises: how common are twins? Federal Reserve Bank of San Francisco, Pacific Basin Working Paper No. PB99-07. Glick, R., Rogoff, K., 1995. Global versus country-specific productivity shocks and the current account. Journal of Monetary Economics 35, 159e192. Gramlich, E.M., 2004. Budget and trade deficits: linked, both worrisome in the long run, but not twins. Speech to National Association for Business Economics, Los Angeles, California. Available from: . Greenspan, A., 2003. Current account. Speech to 21st Annual Monetary Policy Conference, Cato Institute, Washington D.C. Available from: . Gruber, J.W., 2004. Increased current account dispersion: differential growth, demographic dispersion, or greater financial integration? Mimeo, Board of Governors of the Federal Reserve System. Gruber, J.W., Kamin, S.B., 2005. Explaining the global pattern of current account imbalances. Board of Governors of the Federal Reserve System, International Finance Discussion Papers No. 846. Ju, J., Wei, S-J., 2006. A solution to two paradoxes of international capital flows. NBER Working Paper 12668. Kamin, S.B., 2005. The revived Bretton Woods System: does it explain developments in non-China developing Asia? Paper presented at Federal Reserve Bank of San Francisco conference, ‘‘The Revived Bretton Woods System: A New Paradigm for Asian Development?’’ Available from: . Kaminsky, G., 2003. Varieties of currency crises. NBER Working Paper 10193. Kaminsky, G., Reinhart, C., 1999. The twin crises: the causes of banking and balance of payments problems. American Economic Review 89 (3), 473e500. Kaufmann, D., Kraay, A., Mastruzzi, M., 2005. Governance matters IV: governance indicators for 1996e2004. World Bank Policy Research Working Paper 3630. Lane, P., Milesi-Ferretti, G., 2006. The external wealth of nations mark II: revised and extended estimates of foreign assets and liabilities, 1970e2004. Institute for International Integration Studies, Working Paper No. 126. Mann, C.L., 1999. Is the U.S. trade deficit sustainable? Institute for International Economics, Washington, D.C. Mann, C.L., 2002. Perspectives on the U.S. current account deficit and sustainability. Journal of Economic Perspectives 16, 131e152. Mann, C.L., 2004. Managing exchange rates: achievement of global re-balancing or evidence of global co-dependency? Business Economics 39, 20e29. Prasad, E.S., Rajan, R., Subramanian, A., 2006. Foreign capital and economic growth. Paper presented at Federal Reserve Bank of Kansas City Annual Economic Symposium. Roubini, N., Setser, B., 2004. The US as a net debtor: the sustainability of the US external imbalances. Unpublished Working Paper. Truman, E.M., 2004. Budget and external deficits: not twins but the same family. Paper presented at the Federal Reserve Bank of Boston Annual Research Conference.