The macroeconomic balance approach to exchange rate assessments consists of three steps. First, an equilibrium relationship between current account balances and a set of fundamentals is estimated with panel econometric techniques. Second, for each country, equilibrium current accounts (“current account norms”) are computed from this relationship as a function of the levels of fundamentals projected to prevail in the medium term. Third, the real exchange rate adjustment that would close the gap between the estimated current account norm and the underlying current account balance (i.e., the current account balance that would emerge at a zero output gap both domestically and in partner countries) is computed for each country.
The current account norms are estimated using a panel data set of 54 advanced and emerging market economies over 1973–2004.3 This large sample of countries is likely to be helpful in achieving greater precision in the estimation of the equilibrium relationship between current account balances and the set of fundamentals; the sample period extends two previous studies by IMF staff (Debelle and Faruqee, 1996, and Chinn and Prasad, 2003), which used data through the mid-1990s.
The first subsection discusses the theoretical basis for the empirical investigation and defines the variables. We then address estimation issues, presenting some representative results; describe the current account norms obtained from the econometric estimates; and explain how the real exchange rate adjustment that closes the gap between current account norms and the underlying current account is derived.
Theoretical Background and Variable Definitions
Economic theory underscores how in open economies national saving may exceed or fall short of domestic investment, thus allowing consumption to be smoothed and investment to reflect rate of return opportunities, rather than just the availability of domestic saving (Obstfeld and Rogoff, 1996, and Obstfeld, 2004). The substantial body of literature on the subject has guided the empirical investigation below and led to the identification of the following robust determinants of the current account balance over the medium term:4
Fiscal balance. A higher government budget balance raises national saving and thereby increases the current account balance (Ahmed, 1986, and Chinn, 2005). Only in the particular case of full Ricardian equivalence, where private saving fully offsets changes in public saving, would there be no link between government budget balances and current account balances.5 The measure of fiscal balance used below is the ratio of the general government budget balance to GDP in deviation from the average budget balance of trading partners: if the government budget balance improved in all countries, there would be a world-wide macroeconomic effect but little expected effect on the current account balance of each country.
Demographics. A higher share of the economically inactive dependent population reduces national saving and decreases the current account balance (Higgins, 1998, and Federal Reserve Bank of Kansas City, 2004). To proxy for this, the model includes an old-age dependency ratio as well as the population growth rate (which captures the share of economically dependent young people). Both variables, measured in deviation from trading-partner averages, are expected to decrease the current account balance.6
Net foreign assets (NFA). The level of NFA can affect the current account in opposite directions. On the one hand, economies with relatively high NFA can afford to run trade deficits on an extended basis and still remain solvent, potentially leading to a negative association between NFA and the current account. On the other hand, economies with high NFA benefit from higher net foreign income flows, which tend to create a positive association between NFA and current account balances. Standard open economy macroeconomic models predict that this second effect is stronger.7 The “initial” NFA position used in the empirical model is measured before the period of reference for the current account balance, to avoid capturing a reverse link from the current account balance to NFA.8
Oil balance. Higher oil prices increase the current account balance of oil-exporting countries and decrease the balance of oil-importing countries (International Monetary Fund, 2006). The variable used here (the oil balance as a ratio to GDP) allows the effect of oil prices to differ in sign and magnitude across countries.
Economic growth. Economies that are in the early stages of economic development have a greater need for investment and are likely to finance investment through external borrowing (Obstfeld and Rogoff, 1996). As they develop and approach the income levels of advanced economies, their current account balances should improve. Among countries at a similar initial stage of development, the stronger economic growth is relative to trading partners, the lower the current account is likely to be. The ratio of GDP per capita in purchasing power parity terms to the U.S. level—hereafter referred to as relative income—is taken to measure the relative stage of economic development, while the deviation of the real per capita GDP growth rate from its trading-partner average is the variable used to capture relative economic growth.9 The current account balance is expected to increase with relative income but to decline with relative growth.
Economic crises. During economic crises, sharp current account adjustments occur as a by-product of macroeconomic contraction, the reduced availability of international financing, or the attempt to reduce net external liabilities. The empirical evidence suggests that crises have an effect even after controlling for other macroeconomic factors. This is particularly true in the case of the Asian crisis, where a dummy variable remains highly significant even after other plausible determinants of the current account are controlled for. An indicator of banking crisis episodes (Demirgüç-Kunt and Detragiache, 2005, and Gruber and Kamin, 2005) also helps to explain current account behavior.
Financial center. Economies that serve as hubs for international financial flows have tended to run substantial current account surpluses and net creditor positions. This effect is captured by a dummy variable that represents the following financial centers: Belgium, Hong Kong SAR, Luxembourg, the Netherlands, Singapore, and Switzerland.
Estimation Results
The estimation database consists of nonoverlapping four-year averages for 54 economies over 1973–2004. There are thus eight observations for most countries and three observations for transition economies (where the data begin in the early 1990s). The 54 countries in the sample were selected because of their significance in global trade, on the grounds that an economy with a larger global presence will have greater multilateral effects on the exchange rates of other countries. This country coverage enables one to exploit the substantial cross-country variation among the advanced and emerging market economies in the sample.10
High-frequency fluctuations are filtered out by taking four-year averages of the data; this enables the specification to uncover the medium-term relationship between the current account and macroeconomic determinants. Recent studies, including Chinn and Prasad (2003), Chinn and Ito (2005), and Gruber and Kamin (2005), have used similar methods. Cointegration methods are not appropriate here because the current account balance (in percent of GDP) is a stationary series in most countries during most sample periods. Moreover, the current account needs to be stationary for the intertemporal budget constraint to hold (Ghosh and Ostry, 1997; Taylor, 2001; and Lee and Chinn, 2007). To reflect the significant persistence of the current account series, a specification including a lagged current account term has also been estimated. Under this specification, the initial NFA term is excluded because its effect becomes statistically indistinguishable from that of the lagged current account term.
Three representative estimates are reported below: two pooled estimates, the first with no country-specific constant terms and the second with a very limited number thereof, and one fixed-effects estimate that allows country-specific constant terms for all countries. These results are supported by several robustness checks across different variable definitions, samples, and specifications, as briefly discussed in Appendix 2.1.
The pooled estimation results include no or few country-specific constants and therefore use the variables in the regression to explain both the cross-sectional and time-series (within-country) variation in the data. Estimation biases can arise if there are important factors explaining the cross-country variation in the data that are not captured in the specification but are correlated with the other variables. While fixed-effects estimation controls for this possibility by including country-specific constants, the resulting estimates of country effects may be unduly influenced by historical realizations of the dependent variable—especially for countries with a short sample—or may end up accounting for the bulk of the cross-country variation when data change little over time. For this reason, estimation results are presented below for both the pooled and fixed-effects models.11
The two left-hand-side columns of Table 1 report the results of the pooled estimation. The first column (pooled estimation) uses initial NFA and contains no country-specific constant terms, while the second column (hybrid pooled estimation) uses lagged current account and allows for a few country-specific constant terms that were found to have very high statistical significance (see Appendix 2.1 for details). Estimated coefficients are statistically significant and have expected signs and plausible magnitudes:
The coefficients on the fiscal balance imply that a 1 percentage point increase in the government budget balance (relative to trading partners) leads to a 0.2 percentage point increase in the current account balance in percent of GDP. This result is broadly consistent with previous estimates, which mostly ranged between 0.2 and 0.5.
A higher dependency ratio reduces the current account balance. The coefficients on population growth imply that a 1 percentage point increase in the population growth rate relative to trading partners—a very large change given the cross-country variation in the data—deteriorates the current account balance by 1 to 1.25 percent of GDP.
The 0.02 coefficient on initial NFA implies that an increase in NFA of 10 percent of GDP raises the medium-term current account balance by about 0.2 percent of GDP. Although the sign of the coefficient is theoretically ambiguous as discussed above, the positive sign estimated here is consistent with previous empirical findings, including those of Lane and Milesi-Ferretti (2002) and Chinn and Prasad (2003). The estimated size of the coefficient—which is below the average interest rate on external assets and liabilities—indicates that countries with larger initial NFA positions tend to run a smaller trade balance, offsetting part of the positive effect on the current account from higher investment income.
When the lagged current account term is used, the NFA term (which is strongly correlated with past current account balances) loses statistical significance, and the coefficient on the lagged current account is 0.37. The estimated coefficient indicates the gradual nature of current account adjustment: for example, nearly 30 percent of the effect of a shock to the current account would remain five years after the event.
The coefficients on the oil balance are about 0.2, reflecting the cross-country variation in the effect of oil price changes. Oil exporters have large oil surpluses, amounting to tens of percent of GDP, but spend a large part of them on imports of goods and services, leading to a much smaller current account surplus.12 Oil importers compress other imports as oil prices increase. The oil balance coefficient is larger in the fixed-effects model than in the pooled models, reflecting a negative correlation between oil balances and country-specific factors.
An increase in relative income raises the current account balance while higher relative output growth lowers it. The coefficient of 0.02 on relative income implies that, other things being equal, a country whose income is half the U.S. level will have on average a current account balance that is 1 percentage point of GDP smaller than that of the United States. The coefficient estimate of about 0.2 on relative output growth implies that a 1 percentage point increase in real GDP growth of an emerging market economy (compared with the trading-partner average) reduces the current account balance by 0.2 percent of GDP.
The banking crisis and Asian crisis dummy variables lead to a higher current account balance, by 1 and 4–6 percent of GDP respectively, confirming that the macroeconomic contraction and reduced availability of international financing associated with crises tend to increase current account balances temporarily (while the crisis prevails), other things being equal.
The current account balances of financial centers are found to be about 3 percent of GDP larger than those of other countries.
Macroeconomic Balance Approach: Current Account Regressions
Macroeconomic Balance Approach: Current Account Regressions
| Pooled Estimation | Hybrid Pooled Estimation | Fixed Effects Estimation | ||
|---|---|---|---|---|
| Fiscal balance | 0.20*** | 0.19*** | 0.32*** | |
| Old-age dependency | −0.14** | −0.12** | −0.23** | |
| Population growth | −1.21*** | −1.03** | −0.47 | |
| Initial net foreign | 0.02*** | … | … | |
| assets (NFA) | ||||
| Lagged current | … | 0.37*** | … | |
| account | ||||
| Oil balance | 0.23*** | 0.17*** | 0.31*** | |
| Output growth | −0.21** | −0.16* | −0.27 | |
| Relative income | 0.02* | 0.02* | … | |
| Banking crisis | 0.01* | 0.01 | … | |
| Asian crisis | 0.06*** | 0.04*** | 0.07*** | |
| Financial center | 0.03*** | 0.03*** | … | |
| Adjusted R2 | 0.52 | 0.62 | 0.56 | |
Macroeconomic Balance Approach: Current Account Regressions
| Pooled Estimation | Hybrid Pooled Estimation | Fixed Effects Estimation | ||
|---|---|---|---|---|
| Fiscal balance | 0.20*** | 0.19*** | 0.32*** | |
| Old-age dependency | −0.14** | −0.12** | −0.23** | |
| Population growth | −1.21*** | −1.03** | −0.47 | |
| Initial net foreign | 0.02*** | … | … | |
| assets (NFA) | ||||
| Lagged current | … | 0.37*** | … | |
| account | ||||
| Oil balance | 0.23*** | 0.17*** | 0.31*** | |
| Output growth | −0.21** | −0.16* | −0.27 | |
| Relative income | 0.02* | 0.02* | … | |
| Banking crisis | 0.01* | 0.01 | … | |
| Asian crisis | 0.06*** | 0.04*** | 0.07*** | |
| Financial center | 0.03*** | 0.03*** | … | |
| Adjusted R2 | 0.52 | 0.62 | 0.56 | |
Compared with the results of the pooled estimation, the fixed-effect estimates (right-hand-side column of Table 1) have the same signs but somewhat different magnitudes. On the one hand, the fiscal balance, old-age dependency ratio, and oil balance have weaker effects in the cross-sectional dimension, reflecting the fact that the impact of these variables on the current account may be weakened by country-specific factors (such as a different retirement age across countries). For this reason the fixed-effect estimates, which capture mainly the time-series correlations with the current account, are larger (in absolute value). On the other hand, population growth has a stronger economic and statistical effect across countries than over time, reflecting the very gradual change in population growth within countries; this shows up as a smaller (in absolute value) coefficient in the fixed-effects specification. Other variables whose economic and statistical significance is mostly captured by country-specific constants—initial NFA, relative income, the financial center dummy variable, and the banking crisis dummy variable—are excluded from the regression.
While the estimates of Table 1 capture medium-term tendencies in the co-movement of the current account balance with the underlying fundamentals, they are unavoidably subject to significant uncertainty, reflecting the large variation in current account balances across countries and over time and the limits of the common specification imposed across a diverse set of countries. The standard errors of the in-sample current account forecast are in the range of 2–3.5 percent of GDP, with the standard errors for the emerging markets at the higher end of the range.
Current Account Norms
Illustrative current account norms13 can be calculated by applying the coefficient estimates in Table 1 to the medium-term values of the regressors. In computing the norms, medium-term values of the fiscal balance, oil balance, output growth, and relative income are drawn from the World Economic Outlook (WEO) database (projections for 2012), while demographic variables are obtained from the United Nations (UN) database under the assumption of a constant fertility rate. Lagged current accounts are calculated by the average current account–to-GDP ratio over the most recent five-year period, from 2002 to 2006, whereas the average over a longer period is used for a few countries for which current account developments over 2002–06 were substantially different from historical trends. The effect of crises is excluded from the norm calculations because they can be expected to wane over the medium term.
Table 2 presents current account balances and the illustrative current account norms for six country groups (defined in Appendix 2.1): European advanced economies, other advanced economies, and four groups of emerging markets (emerging Asian countries, Central and Eastern European countries, Latin American countries, and other countries). The first two columns of Table 2 report the actual and projected current account balances, in GDP-weighted averages for each country group. The 2012 projection is taken to be the underlying current account—the level reached after lagged exchange rate effects have worked themselves out and output gaps have closed.14 The last column of Table 2 reports the group-wide current account norms, defined as the GDP-weighted averages of the individual-country norms that were calculated by using the hybrid pooled estimates of Table 1. Of course, going forward, exchange rate assessments under the MB approach will make use of estimates of the individual country norms, and will reflect a judgment about the relative informational value of statistical estimates in particular country cases.
Macroeconomic Balance Approach: Illustrative Current Account Norms
(In percent of GDP)
Based on the September 2007 World Economic Outlook database.
Calculated from hybrid pooled estimates.
Macroeconomic Balance Approach: Illustrative Current Account Norms
(In percent of GDP)
| Current Account1 | |||||
|---|---|---|---|---|---|
| Country | Observed 2006 | Medium term 2012 | Current Account Norm2 | ||
| Advanced countries | |||||
| Europe | 0.3 | −0.2 | 0.3 | ||
| Other | −3.4 | −3.3 | −1.9 | ||
| Emerging markets | |||||
| Asia | 6.3 | 7.3 | 1.3 | ||
| Latin America | 1.8 | −0.8 | −0.3 | ||
| Central and Eastern | −3.8 | −3.9 | −2.8 | ||
| European countries | |||||
| Other | 3.7 | −0.2 | 1.1 | ||
Based on the September 2007 World Economic Outlook database.
Calculated from hybrid pooled estimates.
Macroeconomic Balance Approach: Illustrative Current Account Norms
(In percent of GDP)
| Current Account1 | |||||
|---|---|---|---|---|---|
| Country | Observed 2006 | Medium term 2012 | Current Account Norm2 | ||
| Advanced countries | |||||
| Europe | 0.3 | −0.2 | 0.3 | ||
| Other | −3.4 | −3.3 | −1.9 | ||
| Emerging markets | |||||
| Asia | 6.3 | 7.3 | 1.3 | ||
| Latin America | 1.8 | −0.8 | −0.3 | ||
| Central and Eastern | −3.8 | −3.9 | −2.8 | ||
| European countries | |||||
| Other | 3.7 | −0.2 | 1.1 | ||
Based on the September 2007 World Economic Outlook database.
Calculated from hybrid pooled estimates.
The gap between group-wide current account norms and underlying current account balances in Table 2 varies significantly across country groups. For the group of European advanced economies, the current account norm is a surplus of 0.3 percent of their combined GDP, close to the broadly balanced position of their underlying current account. In contrast, there are larger gaps between the current account norm and the underlying current account for other areas, including the non-European advanced economies and emerging Asia. For the non-European advanced economies, the current account norm is a deficit of 1.9 percent, somewhat lower than their underlying current account deficit of 3.3 percent of GDP—which reflects primarily the large projected U.S. deficit. For the group of 10 Asian emerging market economies, the current account norm is a surplus of 1.3 percent of their combined GDP, substantially below the underlying current account surplus of 7.3 percent.
Exchange Rate Assessments
The last step of the MB approach consists of computing the real exchange rate adjustment that would close the gap between the estimated current account norm and the underlying current account of each economy.15 The magnitude of the exchange rate adjustment is derived by applying the elasticity of the current account balance to the real exchange rate. The current account elasticity is calculated as (export elasticity) ×(export-to-GDP ratio) — (import elasticity — 1) × (import-to-GDP ratio): for a given response of export and import volumes to the real exchange rate, the impact on the trade balance and the current account will be roughly proportional to trade openness. Therefore, a country more open to trade will be able to close the current account gap with less exchange rate adjustment.
Once exchange rate adjustments are calculated for all countries, a final correction is made to ensure that they are mutually consistent. This multilateral consistency is required by the fact that there can only be n1 independent exchange rates among n currencies. The correction consists of adjusting all exchange rate misalignments equally or proportionately, to preserve their relative ranking.16
Appendix 2.1. MB Approach: Data and Methodology
Data Description
The estimation sample includes 54 economies and the euro area, for the period from 1973 to 2004, and four-year averages are used in the estimation. The main data sources are International Financial Statistics (IFS) and World Development Indicator (WDI), with World Economic Outlook (WEO) data used to fill in some missing values. Data for the euro area were obtained from the Euro Area Business Cycle Network Real Time Database and the European Centre for Advanced Research in Economics and Statistics. Data for Taiwan Province of China come primarily from national sources. Demographic data come from the United Nations Population Database (Population Prospects: The 2004 Revision), except for data for Taiwan Province of China, which were obtained from the U.S. Census Bureau International Database.
Definitions of each variable are as follows. The following four variables are calculated as deviations from the averages for trading partners.
Fiscal balance is measured as the ratio of the general government balance to GDP. Exceptions include Algeria and Korea, for which the central government balance was used instead of the general government balance.
Old-age dependency ratio relative to the prime age population is measured as the ratio of the population above 65 to the population between 30 and 64.
Population growth rate is the annual population growth rate of each country.
Growth rate of real per capita GDP is included only for emerging market economies.
The remaining variables are not calculated as deviations from the averages of trading partners, either because this is already implicit in their measure (NFA and oil balance) or because it turned out to be statistically redundant (crisis variables).
Initial NFA is measured as the ratio of NFA to GDP prevailing at the beginning of each four-year period, using the NFA data from Lane and Milesi-Ferretti (2007b).
Oil balance is measured as a ratio to GDP.
Dummy variable for Asian crisis is included for Asian emerging markets for 1997–2004: China, Hong Kong SAR, India, Indonesia, Korea, Malaysia, the Philippines, Singapore, Taiwan Province of China, and Thailand.
Dummy variable for banking crisis is obtained from Demirgüç-Kunt and Detragiache (2005), and Gruber and Kamin (2005).
Relative income is measured as the ratio of per capita PPP income to the U.S. level, both in constant 2000 international dollars.
The 54 sample countries are as follows.
Current CGER countries: Australia, Canada, Denmark, Japan, New Zealand, Norway, Sweden, Switzerland, United Kingdom, United States; and 12 euro area countries (comprising Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, and Spain).
Newly industrialized or emerging markets: Algeria, Argentina, Brazil, Chile, China, Colombia, Croatia, Czech Republic, Egypt, Hong Kong SAR, Hungary, India, Indonesia, Israel, Korea, Malaysia, Mexico, Morocco, Pakistan, Peru, the Philippines, Poland, Russia, Singapore, Slovak Republic, Slovenia, South Africa, Taiwan Province of China, Thailand, Tunisia, Turkey, and Venezuela.
Country groupings for Tables 2 and 4 are defined as follows.
Advanced countries, Europe: the euro area, Denmark, Norway, Sweden, Switzerland, and the United Kingdom.
Advanced countries, other: Australia, Canada, Japan, New Zealand, and the United States.
Emerging markets, Asia: China, Hong Kong SAR, India, Indonesia, Korea, Malaysia, the Philippines, Singapore, Taiwan Province of China, and Thailand.
Emerging markets, Latin America: Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela.
Emerging markets, Central and Eastern Europe: Croatia, Czech Republic, Hungary, Poland, Slovak Republic, and Slovenia.
Emerging markets, Other: Algeria, Egypt, Israel, Morocco, Pakistan, Russia, South Africa, Tunisia, and Turkey.
Econometric Methodology
In the hybrid pooled estimation of Table 1, country-specific constant terms are included for six medium-sized economies—Australia, Chile, Israel, New Zealand, Sweden, and Thailand—for which the hypothesis that the constant term is zero is strongly rejected (at the 0.5 percent critical value) in this as well as alternative specifications that were estimated to check robustness; and the Asian crisis dummy is included for four countries with statistically significant coefficients—Korea, Malaysia, Singapore, and Thailand. The three estimates reported in Table 1 are confirmed by several robustness checks. For example, similar patterns are found in estimates over 1980–2004. When estimated over samples consisting of advanced economies and emerging markets separately, coefficients are comparable but less statistically significant, reflecting the more limited variation in samples half as large. The conventional old-age dependency ratio based on the working age population (between ages 15 and 64) had a statistically weaker effect than the dependency ratio based on the prime age population. Finally, several measures of financial development—such as capital account liberalization and financial depth—were found to have economically and statistically less robust effects than the variables included in Table 1.