16. Capital Flows and the Yen-U.S. Dollar Exchange Rate

Alessandro Zanello, and Daniel Citrin
Published Date:
November 2008
  • ShareShare
Show Summary Details
Papa N’Diaye


A weak yen has fueled the perception that its evolution is disconnected from the economic fundamentals that determine its long-term value. The yen has continued to depreciate against major currencies since 2006. In 2006 the yen lost 5½ percent of its value against the U.S. dollar and 9¼ percent in real effective terms. This trend has continued through the first half of 2007. Yet, Japan’s fundamentals have strengthened markedly in recent years.1

The economy has expanded at a healthy pace since 2002, exceeding potential growth in the last three years. Activity has been supported by rapid gains in labor productivity, with the largest advances in the tradable sector (which maintains its secular positive productivity gap over the nontradable sector). At the same time, Japan’s external position has remained favorable: the current account surplus reached just under 4 percent of GDP in 2006, further raising the stock of net foreign assets (NFA).

Against this background, this chapter tries to shed some light on possible sources of the disconnect between the current and long-term value of the yen-U.S. dollar exchange rate. It does so by assessing the dynamics of the yen–dollar exchange rate in response to cyclical and structural factors, including developments in Japanese investors’ behaviors, global risks appetite, demographic trends, and ongoing structural reforms. The main conclusion is that, although the yen could be expected to appreciate over the medium to longer run, nontrade factors are likely to delay the adjustment, barring a sudden change in investors’ sentiment.

Long-term value of the yen

The yen is undervalued relative to its long-term level in many assessments. Estimates of the long-term value of a currency can be obtained in a variety of ways—none of which is without shortcomings or pitfalls (Appendix 16.1). There are four popular approaches to estimating equilibrium exchange rates: the reduced-form equilibrium real exchange rate (ERER) approach, the macroeconomic balance (MB) approach, the external sustainability (ES) approach, and the global general equilibrium model (GGEM) approach.2 Most existing measures of Japan’s equilibrium exchange rate based on these approaches suggest that the yen is below its long-term level against major currencies and in real effective terms (Table 16.1). The range of estimated undervaluations is large.

Table 16.1Recent Estimates of Japan’s Exchange Rate Misalignment
AuthorApproachType of Exchange RatePeriodEstimation TechniqueAssumption% Undervaluation
Benassy and othersERERyen/US$2003Panel data/cointegration14–22
ConsensusPPPyen/US$Nov 0618
Courdet-CouhardeMBCPI-based REERa2002–03Panel dataS-I norm 1.916–20
Courdet-CouhardeMByen/US$2002–03Panel dataS-I norm 1.936–37
Deutsche BankERERyen/euroAug 06Time series20
IMFMBCPI-based REERaMar 07Panel data, 4 year averagesS-I norm 1.8Significant
IMFMByen/US$Mar 07Panel dataS-I norm 1.8Significant
IMFMByen/euroMar 07Panel dataS-I norm 1.8Significant
IMFERERCPI-based REERaMar 07Panel data/cointegrationfundamentals at 2012Significant
IMFERERyen/US$Mar 07Panel data/cointegrationSignificant
IMFERERyen/euroMar 07Panel data/cointegrationSignificant
IMFESCPI-based REERaMar 07Panel dataStable NFA at 2005Significant
IMFESyen/US$Panel dataStable NFA at 2005Significant
Morgan StanleyEREERbyen/US$Oct 06Time series20
Mussayen/US$2004Global imbalances35
ObstfeldGGEMCPI-based REERaEnd 2005SimulationsaNarrow Japan’s current

account surplus from 3.5%

of GDP to zero

Real Effective Exchange Rate.

Equilibrium Real Effective Exchange Rate.

Real Effective Exchange Rate.

Equilibrium Real Effective Exchange Rate.

Estimates of the equilibrium value of the yen are subject to large statistical uncertainty and reflect the interplay of factors often pointing in different directions. Uncertainties relate to data definition, model specification, and the restrictiveness of the underlying assumptions to capture multilateral consistency in estimated equations of equilibrium exchange rates.3 The influence of offsetting factors is most apparent in the case of productivity differentials and demographics, which are found to have a significant explanatory power on the long-term value of the yen (Isard and Faruqee, 1998). In particular:

  • Productivity differentials suggest a long-term value of the yen closer to the lower bound of the range of estimates in Table 16.1. The differential in labor productivity growth between the tradable and nontradable sector (productivity gap) in the United States has surpassed that in Japan since 2000 (Figure 16.1).4 This is mainly because the pick-up in productivity in the tradable sector has been more rapid in the United States than in Japan. This shift in productivity gap differentials has been accompanied with a trend real depreciation thereafter, in line with the predictions of the Balassa-Samuelson effect.5 This trend could be exacerbated by productivity-boosting reforms in the nontradable sector in Japan and lead to a weaker real yen in the longer term.6 (By contrast, reforms that unlock productivity gains in all sectors would tend to increase wages, demand, and prices, strengthening the yen in real terms over time.)

  • By contrast, demographic factors suggest estimates of the long-term equilibrium of the yen closer to the higher end of the range of estimates in Table 16.1. The effects of demographics are generally captured through the evolution of NFA, which increase with the old-age dependency ratio. Higher long-run NFA will in the transition be associated with a weaker yen (to generate the current account surplus needed to raise NFA) but eventually call for a more appreciated currency.7 On this reasoning, Japan’s strong accumulation of assets since the mid-1990s points to a large undervaluation of the yen relative to its longer-term value (Figure 16.2).

Figure 16.1Sectoral Labor Productivity Gaps (ten-year moving average): United States vs. Japana

a Productivity gap defined as the labor productivity growth differential between the traded and the nontraded sectors, with the former including Agriculture, Mining, Manufacturing, and Finance and Insurance.

Source: IMF staff estimates.

Figure 16.2NFA and Yen-US$ Real Exchange Rate

a Data on intervention prior to 1991 are approximated using changes in official reserves net of interests. Official figures are publicly available starting in 1991.

Source: IMF staff estimates.

Be that as it may, conventional analyses of misalignment provide little clues on how, if, or when an exchange rate adjustment will take place. This is in part because estimates of equilibrium exchange rates are not forecasts. The assessment that a currency is misaligned relative to its long-term value does not necessarily imply that an imminent adjustment is likely (Isard and others, 2001). At best, some empirical studies on the dynamics of exchange rates indicate a tendency for currencies to revert to their equilibrium gradually (perhaps in three to five years), but this convergence is conditional on the absence of any disturbance to fundamentals. A gradual appreciation seems consistent with current markets expectations. For example, the May 2007 foreign exchange consensus forecast indicates that on average analysts expect the yen to rise by about 7 percent by the first half of 2008, although there seem to be a great deal of uncertainty on the magnitude of such a rise (Figure 16.3).

Figure 16.3Distribution of Expected Change in the Yen

Note: Probability (on y-axis) possible % changes in the Yen-US $ exchange rate.

Negative value (on x-axis) implies appreciation (May 2007 Consensus Forecast).

Transition to a longer-term equilibrium

The process of adjustment of exchange rates to their long-term level is influenced by capital flows. With an emphasis on cross-border trade in goods and services, many existing models of equilibrium exchange rates do not explicitly account for capital flows.8 These flows have become important determinants of the supply and demand conditions in currency markets (Figure 16.4). In the case of Japan, capital flows dwarf trade flows.9 The average daily turnover in yen foreign exchange markets worldwide (about US$360 billion in dollar equivalent) exceeds the annual dollar value of net trade in goods and services by about a factor of two. An alternative modeling strategy would bring to the fore a possible role for capital flows and changing investor’s appetite for a country’s assets at least in the transition to a long-term equilibrium pinned down by other economic fundamentals.

Figure 16.4Share of Foreign Securities in Total Assets


Source: IMF staff estimates.

In the case of Japan, ongoing capital flows represent an adjustment to secular and cyclical forces, such as a decline in the home bias, large interest rate differentials against yen assets, and globally low volatility in asset markets.

  • Secular decline in home bias and portfolio diversification. The stock of overseas investment has doubled since 2000 to 121 percent of GDP as at end 2006.10 Japanese investors (particularly retail investors) are increasing their holdings of foreign securities, reflecting a decline in their preference for domestic assets and deregulation in the banking sector. This portfolio rebalancing is probably supported by population aging as retirees seek higher returns abroad.

  • Large interest rate differentials have also played a part in generating capital outflows. Some of these outflows, which are more speculative in nature, are carry trades. The magnitude of these carry trades is however difficult to gauge as there is no standard definition and the underlying transactions can be off-balance sheet.

  • Global search for yields. The global environment of low volatility, ample liquidity, record corporate profits, and financial innovation has contributed to a search for yields and spurred capital flows. This environment has also created strong risk appetite and new linkages between different asset classes and segments of capital markets with shifts in portfolio choices and investment decisions. For example, low volatility (combined with large interest rates differentials) has boosted the risk-adjusted return (Sharpe ratio) for investing outside Japan (Figure 16.5).

Figure 16.5Sharpe Ratio

Source: IMF staff calculations.

Capturing the effects of capital flows on the yen-dollar exchange rate

The modeling strategy in this chapter is to estimate reduced-form equations that link changes in the yen-dollar rate to capital flows as well as changes in conventional determinants of the long-run equilibrium exchange rate of the yen against the dollar. The protypical specification takes the form of the following error-correction model.

This equation is estimated using quarterly data starting in 1990 by the Generalized Method of Moments with a set of instruments defined in Appendix 16.2.

In this specification:

  • Long-term determinants of the bilateral yen-U.S. dollar include: the relative price levels in Japan and the United States, the relative labor productivity, and the ratio of Japan’s NFA to GDP. These variables are used to estimate an “augmented purchasing power parity (PPP)” relationship that allows to compute an equilibrium exchange rate (Table 16.A2.1). The estimated equilibrium exchange rate provides information on the degree of undervaluation of the yen-dollar rate. For example, using the estimated long-term relationship and NFA values based on the cumulated projections of the current account surplus (produced in the context of the IMF’s World Economic Outlook exercise) together with forecast for labor productivity growth indicate a significant undervaluation in real terms.

  • The deviation from the equilibrium rate is the difference between the actual yen-dollar exchange rate and the long-term equilibrium rate determined through the augmented PPP relationship, as described above.

  • The effects of capital flows in the adjustment process are captured through the short-term interest rate differential, the long-term interest rate differential, the global volatility index, and through high-frequency changes in NFA, which capture portfolio rebalancing.11

Estimates of alternative specifications of Equation 1 are reported in Table 16.2. The estimated coefficients indicate the percentage change in the yen-dollar rate in response to changes in the variables described above. The results indicate a statistically significant impact of interest rates differentials, NFA, and volatility on the adjustment process of the yen-dollar exchange rate with significance levels (p-values) generally below 5 percent. For example:

Table 16.2Yen-US$ Bilateral Exchange Rate and Macroeconomic Variables
Model 1Model 2Model 3Model 4
Inflation differential-1.9000.000-1.9420.000-1.3350.000-1.3220.000
Relative productivity-0.2360.207-0.9720.000-0.9530.000
Long-term interest rates differential-0.0010.752
Short-term interest rates differential-0.0100.001-0.0090.005
Net foreign assets2.3490.0002.2660.0002.2930.0002.2930.000
Correction to long-term equilibrium-0.0240.030-0.0260.018-0.0210.141-0.0220.110
Sharpe ratio-0.0390.085-0.0380.077
Sharpe ratio squared0.1190.0720.1200.064
Adjusted R-squared0.4630.4840.4330.446
S.E. of regression0.0350.0350.0360.036
Note: Models 2 and 4 exclude the variables with the highest p-values in the previous regression.Source: IMF staff estimates.
Note: Models 2 and 4 exclude the variables with the highest p-values in the previous regression.Source: IMF staff estimates.
  • A narrowing of the interest rates differential between Japan and the United States (currently against yen assets) leads to an appreciation of the yen relative to the dollar.12

  • An accumulation of NFA leads to a depreciation of the yen in the short run, but to an appreciation of the yen in the longer term (as discussed above).

  • Higher volatility is associated with an appreciation of the yen relative to the dollar, supporting the view of reversal of short positions in the wake of a volatility shock.13

  • The coefficient of the deviation from the long-term equilibrium suggests that only about 8–10 percent of the yen-U.S. dollar adjustment of the exchange rate toward its longer-term value takes place every year, other things being equal, suggesting a half-life of undervaluation of about five years.

The estimated models fit the data relatively well. The regressors explain between 40 percent and 50 percent of the changes in the yen-dollar rate. Within-sample dynamic simulations using Model 1 above suggest that, taken together, these variables predict relatively well changes in the yen-dollar exchange rate (Figure 16.6).14

Figure 16.6Yen-US$ Exchange Rate: Actual versus Model Prediction I

(simulations from 2002Q1 and 2004Q1 using actual values of exogenous variables)

A counterfactual exercise based on these estimates allows a first assessment of the impact of the recent pattern of capital flows on the external value of the yen. A comparison between the actual yen-dollar exchange rates and those predicted by Model 1 without the contributions of variables related to the capital account (e.g. interest rates, volatility, and high-frequency changes in NFA) suggests that in the absence of these factors the yen-dollar rate would have been about 9 percent more appreciated than the actual rate at the end of 2006 (Figure 16.7). This result needs to be interpreted with care, but gives a first benchmark of the downward pressure from the pattern of capital flows in recent years.

Figure 16.7Yen-US$ Exchange Rate: Actual versus Model Prediction II

(simulations from 2002Q1 using actual values of exogenous variables)

Although single-equation estimation gives some insights, it is not without weaknesses. For example, the estimated impact of the inflation differential between Japan and the United States is incorrectly signed. This might be due to the fact that relative inflation captures information related to productivity differential in the absence of other restrictions that could be imposed in a multivariate model.

Thus, a multivariate model is used to cross-check the main results. The evolution of the bilateral exchange rate is modeled in a structural vector error correction model (SVECM) that captures the interactions between the variables of interests, while ensuring the consistency between their short-run and long-run dynamics. The SVECM framework decomposes each variable into “factors,” some with temporary effects and others with long-lasting ones. The factors that have long-lasting effects explain both the short-run and long-run movements in the variables—i.e. they determine the trends and movements around these trends.

The results from this multivariate approach are qualitatively similar to those in Table 16.2. For example:

  • An increase in volatility, whether temporary or permanent, appreciates the yen in the short run. In the case of the temporary increase in volatility, the exchange rate appreciates for about five quarters, with some undershooting before returning to its initial level (Figure 16.8). This response suggests that a global flight from risk or volatility shock could lead to a rapid appreciation of the yen against the dollar.

Figure 16.8Effects of Increases in Volatility

Source: IMF staff estimates.

  • A temporary increase in interest rates also leads to an initial appreciation of the yen, consistent with an overshooting model of exchange rate determination (Figure 16.9).

  • A permanent gain in (aggregate) productivity appreciates the yen both in the short run and the long run, while temporary gains generate appreciation pressures only in the short run.

  • Higher prices in Japan than in the United States lead to a depreciation of the yen in the short run for temporary shocks, and in both the short run and the long run if the shock is permanent, consistent with the estimated augmented PPP relationship in Appendix 16.2. This result, which was not fully validated in the single equation approach discussed earlier, provides support for a system approach.

  • Finally, an alternative set of identification restrictions to analyze the effects of permanently higher NFA indicates that higher NFA would appreciate the yen in the long run, consistent with the predictions of many standard models.

Figure 16.9Temporary Increase in the Short-term Interest Rate

Source: IMF staff estimates.

Conclusions and policy implications

Shifts in capital flows appear to play an important part in the adjustment process of the yen-dollar exchange rate to its longer-term value. To the extent that the underlying drivers of capital outflows from Japan prove to be persistent, the adjustment of the yen to its longer-term equilibrium value (linked to real factors such as demographics and productivity differentials) may be slowed, although there is always the risk of a sudden change in investors’ sentiment. In fact, there are reasons to believe that these outflows will persist for some time, as discussed in Chapter 9.

Although capital flows play an important role in determining the evolution of the value of the yen, the role of other factors should not be forgotten. Ultimately, the transitional dynamics of exchange rates are the reflection of offsetting forces, the relative importance of which is hard to predict. For example, structural reforms that unlock economywide productivity gains would strengthen the yen in the long run, although the initial effects could be in the opposite direction if those gains are concentrated in the nontradable sector. The strengthening of the yen could be amplified by greater capital inflows as return on capital in Japan rises.

Appendix 16.1: overview of existing methodologies

There are four broad approaches to estimating equilibrium exchange rates: the reduced-form equilibrium real exchange rate (ERER) approach, the macroeconomic balance (MB) approach, the external sustainability (ES) approach, and the global general equilibrium model (GGEM) approach.

  • The ERER approach uses the (augmented) PPP concept and involves the estimation of a single equation for the exchange rate as a function of key medium- to longer-term determinants. The determinants generally include factors that have been identified as major drivers of medium- to longer-term movements of the real exchange rates: (relative) productivity growth in the tradable and nontradable sectors (Balassa-Samuelson effect), the net foreign assets position, terms of trade, openness, fiscal balance, real interest rates differential (Behavioral Equilibrium Exchange Rate (BEER) concept by Clark and McDonald, 1998), and demographics. Based on criteria that sometimes involve pre-filtering techniques, long-term values of the key explanatory variables are used to derive the equilibrium real exchange rate. The difference between the actual value of the real exchange rate and its predicted equilibrium value indicates the degree of undervaluation or overvaluation. The most recent estimate by the IMF Consultative Group on Exchange Rate (CGER) indicates an undervaluation of the yen in real effective terms.

  • The MB approach is based on the fundamental equilibrium exchange rate concept (Williamson, 1994), which is the exchange rate consistent with internal and external balance. In this approach, the equilibrium exchange rate equalizes a country’s sustainable saving-investment balance with its underlying current account balance (UCUR), when all economies are producing at potential output and the lagged effects of past exchange rate changes have been fully realized.15 The approach involves three steps (Figure 16.A1.1). In step 1, an equilibrium relationship linking the current account balance to a set of fundamentals is estimated. In step 2, an underlying current account balance is computed from the WEO medium-term current account projections (which assume that all economies are producing at potential output and the lagged effects of past exchange rate changes have been fully realized, point A in Figure 16.A1.1). In step 3, the two previous steps are combined to derive an equilibrium real exchange rate (point Q* in Figure 16.A1.1), which intersects the sustainable savings investment balance schedule with the underlying current account balance schedule. The equilibrium real exchange rate varies with sustained shifts in a country’s national savings, investment, or underlying current account balance. In addition to many of the factors enumerated above as major drivers of long-run real exchange rates, the level of development, economic crises, and the level of a country’s financial development could also shift these variables on a sustained basis. For Japan, the saving investment norm is estimated at 1.8 percent of GDP implying a significant undervaluation in real terms.

Figure 16.A1.1Medium-Run Fundamentals

  • The ES approach considers the equilibrium real exchange rate as the real exchange rate that equalizes the current account balance to the level that stabilizes a country’s NFA position to some benchmark level. As the previous approach, it involves three steps. In step 1, the current account balance that stabilizes the country’s NFA position to a given benchmark level is determined. In step 2, the projected medium-term current account estimates obtained in the MB approach are used. In step 3, the change in the exchange rate required to equalize the medium-term current account with its NFA-stabilizing level is determined. In addition to the information needed to obtain the medium-term current account, the approach requires assumptions on the country’s potential growth rate, inflation rate, and rates of return on external assets and liabilities. In the case of Japan, the benchmark level of NFA is that at the end of 2005 (about 35¾ percent of GDP), indicating a significant undervaluation in real terms.

  • The GGEM approach. In the context of the heightened risks of a disorderly unwinding of global imbalances, recent studies have focused on the needed adjustment of major currencies to facilitate the correction of domestic and external imbalances. Most of these studies use a GGEM à la Obstfeld and Rogoff (2005a, 2005b) and the IMF Global Economic Model (GEM) where relative prices clear the world markets for traded goods as well as the domestic markets for nontraded goods. While most studies beside those of the IMF focus on the possible U.S. dollar adjustments that would facilitate the reduction of the ballooning U.S. current account deficit, a recent study by Obstfeld (2006) suggests that to narrow Japan’s actual current account surplus down to zero, the yen would need to appreciate in real terms by 19–38 percent. This implies an appreciation of over 10 percent for every 1 percent of GDP reduction in the current account surplus. In the GGEM framework, the extent of undervaluation or overvaluation depends critically on deep parameters such as the elasticity of substitution between traded and nontraded goods. The lower the elasticity, the sharper the price changes—hence real exchange rate changes—that are needed. Empirical evidence on these parameters is however limited in the case of Japan, weakening the assessment of required yen real appreciation. Nevertheless, a back-of the-envelope calculation that assumes a 10 percent appreciation for every 1 percent of GDP decline in the current account surplus suggests that to bring Japan’s 4 percent of GDP current account surplus to its CGER medium-term norm of 1.8 percent could require a real appreciation of the yen of as much as 22 percent. Put differently, in real terms the yen is estimated to be about 22 percent below its equilibrium value.

As shown above, estimates of the yen equilibrium real effective or bilateral value vary widely depending on the approach used. This wide range of estimates reflects the uncertainty inherent to estimating equilibrium exchange rates, which occurs in practice because of the following possible factors:

  • Inherent “conceptual” differences: the ERER approach ensures the long-run consistency between the exchange rate and the set of fundamentals considered, the MB approach focuses on the flow of current account balance over the medium term, and the ES approach ensures consistency between the stock of NFA and the flows of current account balances.

  • Other factors such as data availability, definition, and measurement, as well as estimation and filtering techniques not only explain the differences between the approaches estimates of a country’s equilibrium real exchange rate, but also cause large variations in the estimates from any given methodology (Dunaway and others, 2006).

Appendix 16.2: framework

The strategy involves estimating a set of single-equations models in a partial equilibrium setting and a structural vector error correction model. The data used cover the period 1990Q1 to 2006Q4 and include the bilateral yen-U.S. dollar exchange rate, the relative price levels between Japan and the United States, the relative labor productivity, the ratio of Japan’s NFA to GDP, the short-term interest rate differential, the long-term interest rate differential, the global volatility index, the ratio of foreign exchange intervention to GDP, and the Sharpe ratio.

Stationarity and cointegration

Unit root tests were performed on all variables and the null hypothesis of a unit root could not be rejected in all cases except for the intervention series that was used in the single equation models only for parsimony reasons. A test of cointegration indicated four cointegration relationships, which were identified by imposing restrictions as follows: (i) an augmented PPP relating the nominal bilateral exchange rate to the price differential (with a coefficient of one), relative productivity (with a negative coefficient), and the ratio of NFA to GDP (higher NFA appreciates the currency); (ii) a term structure equation relating the short-term interest rate differential to long-term interest differential and volatility (high volatility raises the premium); (iii) a relationship that links net foreign assets to productivity and long-term interest rates (wider long-term interest rate differential reduces NFA); and (iv) a volatility equation that links volatility to the short-term interest rate differential and the level of the exchange rate (Table 16.A2.1).

Table 16.A2.1Cointegrating Equations


Net Foreign

Exchange rate1.069.1
Price differential-1.0
Relative productivity7.93.7
Net foreign assets1.01.0
Short-term interest rate1.0-2.9
Long-term interest rate-2.00.04
LR test for binding restrictions
(rank = 4):
LR test for binding restrictions,
Note: Figures in brackets are t-statistics.Source: IMF staff estimates.
Note: Figures in brackets are t-statistics.Source: IMF staff estimates.

Single equation or partial equilibrium

The equations that are displayed in Table 16.2 relate, depending on the specification, exchange rate changes to current changes in NFA, the change in long-term bond yield, the one-quarter lagged change in the short-term interest rate, the change in the volatility index, an error correction term from a long-run relationship estimated above (lagged by two quarters), the change in relative productivity, and the amount of intervention in relation to GDP, the change in the Sharpe ratio, and the change in the Sharpe ratio squared.

The equations were estimated using the Generalized Method of Moments estimator with all current regressors treated as endogenous variables (price differential, relative productivity, long-term interest rate differential, volatility, net foreign assets, and intervention). The set of instruments for Model 1 and 2 include four lags of: the change in exchange rate, the change in the NFA-to-GDP ratio, the change in long-term interest rates, the change in volatility, the change in relative productivity, and the relative prices; the second-quarter lag of the error correction term; and two lags of the intervention-to-GDP ratio. For Models 3 and 4, two lags of the change in the Sharpe ratio and three lags of the change in the Sharpe ratio squared were added to the previous set of instruments. After adjustments the number of observations in all models was 63.

Structural vector error correction model

The system considered included seven variables (the bilateral yen-U.S. dollar exchange rate, the relative prices between Japan and the United States, the relative labor productivity, the ratio of Japan’s NFA to GDP, the short-term interest rate differential, the long-term interest rate differential, and the global volatility index). With four cointegration relations among these variables, the system admits three common stochastic trends. That is, there are four disturbances that have only transitory effects on the variables and three other disturbances that have permanent effects and explain the trends that are displayed by the variables. These different disturbances are identified using three restrictions for the permanent shocks and six restrictions for the transitory shocks.

The identifying restrictions on the permanent shocks are: (i) the first permanent shock (a permanent volatility shock) has no long-run impact on the relative productivity and price differential; and (ii) the second permanent shock (a permanent increase in relative prices) has no long-run impact on productivity. With regard to the transitory shocks, the following restrictions were imposed: the first transitory shock (a temporary productivity shock) initially leaves unchanged relative prices, short-term interest rates, and volatility; the second shock (a temporary inflation shock) initially leaves unchanged the short-term interest rate and volatility; the third shock (an interest rate shock) initially leaves volatility unchanged.

Data description

The data are from the IMF International Financial Statistics database, the Nomura database, the IMF External Wealth of Nations database, the CEIC database, and the Ministry of Finance of Japan. Most variables were expressed in deviation from their U.S. counterparts. In particular:

  • Labor productivity is defined as relative output per man-hours.

  • Price differential is the logarithm of the relative cpi indexes.

  • Short-term interest rate is the difference between the discount rates.

  • The long-term interest rate is the difference in the ten-year government bond yields.

  • Volatility is the global volatility index (VIX) from the Chicago Board of Options Exchange.

  • The Sharpe ratio is defined as the ratio of the short-term interest rate differential and the three-month moving average of the actual volatility using one year of monthly exchange rate data.

  • The net foreign assets data are expressed in relation to GDP and have been transformed in quarterly frequency from annual data.

  • The intervention data are expressed in relation to GDP and from the Ministry of Finance after 1991. Positive data indicate yen selling and dollar buying.


    Bénassy-QuéréAgnésPascaleDuran-VigneronAminaLaréche-Révil and ValérieMignon2004Burden Sharing and Exchange Rate Misalignments Within the Group of Twentyin Dollar Adjustment: How Far? Against What? ed. C.Fred Bergsten and JohnWilliamson (Washington, D.C.: Institute for International Economics).

    Bank for International Settlements (BIS)2007Quarterly ReviewMarch.

    ClarkPeter and RonaldMcDonald1998Exchange Rates and Economic Fundamentals: A Methodological Comparison of BEER and FEERsIMF Working Paper 98/00.

    CoudertVirginie and CecileCouharde2005Real Equilibrium Exchange Rate in ChinaCEPII Working Paper No. 2005–01.

    DunawayStevenLaminLeigh and XiangmingLi2006How Robust are Estimates of Equilibrium Exchange Rates: The Case of ChinaIMF Working Paper 06/220.

    FaruqeeHamid1995Long-Run Determinants of the Real Exchange Rate: A Stock-Flow PerspectiveIMF Staff PapersInternational Monetary Fund Vol. 42 (March).

    GagnonJoseph1996Net Foreign Assets and Equilibrium Exchange Rates: Panel EvidenceFederal Reserve Board International Finance Discussion Paper No. 574 (December).

    IsardPeter and HamidFaruqee1998Exchange Rate Assessment: Extension of the Macroeconomic Balance ApproachIMF Occasional Paper No. 167.

    IsardPeterHamidFaruqeeRussellKincaid and Martin J.Fetherston2001Methodology for Current Account and Exchange Rate AssessmentsIMF Occasional Paper No. 209.

    LanePhilip and Gian MariaMilesi-Ferretti2002External Wealth, The Trade Balance, and the Real Exchange RateEuropean Economic Review (June).

    LanePhilip and Gian MariaMilesi-Ferretti2004The Transfer Problem Revisited: Net Foreign Assets and Real Exchange RatesReview of Economics and Statistics Vol. No. 86 (November).

    LaxtonDouglasPaoloPesenti and PapaN’Diaye2006Deflationary Shocks and Monetary Policy Rules: An Open Economy Scenario AnalysisNBER Working Paper No. 12703.

    ObstfeldMaurice2006Implications for the Yen of Japanese Current Account AdjustmentIMES Discussion Paper No. 2006-E-15.

    ObstfeldMaurice and KennethRogoff2005aThe Unsustainable U.S. Current Account Position Revisitedpaper presented at the NBER conference on G7 Current Account Global Imbalances Newport Rhode Island; revised versionNovember2005.

    ObstfeldMaurice and KennethRogoff2005bGlobal Current Account Imbalances and Exchange Rate AdjustmentBrookings Papers on Economic Activity 1:2005b pp. 67146.

    SteinJerome L.1994The Natural Real Exchange Rate of the U.S. Dollar and Determinants of Capital Flowsin Estimating Equilibrium Exchange Rates ed. JohnWilliamson (Washington, D.C.: Institute for International Economics).

    SteinJerome L.2002The Equilibrium Real Exchange Rate of the Euro: An Evaluation of ResearchIfo Studien Vol. 48 No. 3 pp. 34981.

    TilleCédricNicolasStoffels and OlgaGorbachev2001To What Extent Does Productivity Drive The DollarFederal Reserve Bank of New York Current Issues in Economics and Finance Vol. 7 No. 8.

    WilliamsonJohn1994Estimating Equilibrium Exchange Rates (Washington, D.C.: Institute of International Economics).

Theory and empirical research identify a long list of “fundamentals” or longer-term determinants of exchange rates. These include in comparative terms: (relative) productivity growth in the tradable and nontradable sectors (Balassa-Samuelson effect), the net foreign assets position, the current account, the commodity terms of trade, openness, the fiscal balance, real interest rate differentials, and demographics.

The uncertainty that surrounds and limitations of estimates of equilibrium exchange rates (including those of the IMF Consultative Group on Exchange Rate (CGER)) have been discussed in many studies (see, for example, Coudert and others, 2005; Independent Evaluation Office (IEO) report, 2007; Dunaway and others, 2006).

From the perspective of the Balassa-Samuelson effect, productivity gains that are concentrated in the tradables sector should lead to a real appreciation, while productivity gains equally spread across the tradables and nontradables sector would leave the real exchange rate unchanged.

See Tille and others (2001) for an analysis of labor productivity gaps and the dollar exchange rate.

Simulations using the IMF Global Economy Model suggest that product market reforms in Japan that aim at raising productivity in the nontradables sector would lead to a real depreciation of the yen, if the gains are concentrated only in that sector (see Laxton and others, 2006).

Heuristically, countries with relatively high NFA (which could reflect demographics) can “afford” more appreciated real exchange rates—and the associated trade deficits—while still remaining solvent (see Faruqee, 1995; Gagnon, 1996; Lane and Milesi-Ferretti, 2002, 2004).

Some exceptions may be constituted by reduced form models of equilibrium exchange rates, such as the Behavioral Equilibrium Exchange Rate (BEER) or the NATREX (Natural Real Exchange Rate).

Some models of equilibrium exchange rate (such as the macroeconomic balance approach) that rely on the saving-investment balance as a measure of net saving outflows could be considered as models of capital account balance from a medium-to long-term perspective (Isard and Faruqee, 1998). In these models, short-term changes in capital flows (e.g. changes in portfolio flows) affect the short-run levels of exchange rates and not their longer-term value as long-term capital flows are ultimately determined by the same economic fundamentals that determine saving outflows and trade (e.g. relative productivity).

See Chapters 9 and 10 of this volume.

See Appendix 16.2 for details.

This result runs against uncovered interest parity which, however, has been shown not to hold at short horizons.

For the impact of volatility on exchange rates see BIS (2007).

For the simulations. Model 1 was written in level terms. The actual value of the yen-dollar rate in the quarter before the beginning of the simulation is used as a starting value for the one-period lagged exchange rate. The simulations use for every period the calculated value of the yen-dollar exchange rate and the actual values of the variables listed at p. 259.

Variants of the MB approach include the Natural Real Exchange Rate (NATREX) of Stein (1994, 2002).

    Other Resources Citing This Publication