Currency Hedging for International Portfolios
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

Author’s E-Mail Address: schmittmann.j@gmail.com

This paper examines the benefits from hedging the currency exposure of international investments in single- and multi-country equity and bond portfolios from the perspectives of German, Japanese, British and American investors. Over the period 1975 to 2009, hedging of currency risk substantially reduced the volatility of foreign investments at a quarterly investment horizon. Contrary to previous studies, the paper finds that at longer investment horizons of up to five years the case for hedging for risk reduction purposes remained strong.In addition to its impact on risk, hedging affected returns in economically meaningful magnitudes in some cases.

Abstract

This paper examines the benefits from hedging the currency exposure of international investments in single- and multi-country equity and bond portfolios from the perspectives of German, Japanese, British and American investors. Over the period 1975 to 2009, hedging of currency risk substantially reduced the volatility of foreign investments at a quarterly investment horizon. Contrary to previous studies, the paper finds that at longer investment horizons of up to five years the case for hedging for risk reduction purposes remained strong.In addition to its impact on risk, hedging affected returns in economically meaningful magnitudes in some cases.

I. Introduction

Investors can potentially improve the risk-adjusted performance of their portfolios by investing internationally. For example, asset pricing models such as the Sharpe-Lintner CAPM and multi-factor models suggest that investors should hold global portfolios. Empirically, many authors have documented the gains from international diversification of investment portfolios (see, for example, Levy and Sarnat (1970) and Ang and Bekaert (2002) among many others).

For all the apparent benefits of investing internationally, investing abroad confronts investors with the decision of how to deal with the foreign currency exposure implied by their foreign investments. Exposure to foreign currencies potentially alters the return and risk profile of international investments. International investors therefore need to decide whether to retain or to hedge the implicit currency exposure associated with investing abroad.

In this paper, we consider the impact of hedging currency exposure from the perspectives of German, Japanese, British and American investors. We analyze the impact of hedging on the risk and return of bond and equity investments in France, Germany, Japan, the U.K, and the U.S. Our dataset covers almost the entire period of free-floating exchange rates and includes the financial crisis of 2007 to 2009. We provide results for simple hedge ratios that are popular with investors as well as for optimal risk-minimizing hedge ratios that exploit the full covariance structure.2 Simple hedge ratios include no hedging, half hedging and full hedging. Many academic studies (for example, Campbell et al. 2010) advocate optimal hedge ratios but we find that correlation patterns may be time-period specific. While our methodological approach is in principle not new, our implementation and empirical results provide new insights.

We obtain our first new result by distinguishing between short and long investment horizons. Froot (1993) argued that investors with an investment horizon of several years would be naturally hedged against exchange rate fluctuations by mean-reverting real exchange rates. Consequently, only investors who are sensitive to short-term volatility over a quarter or a year should hedge currency risk. Froot’s results are limited to the perspective of a U.K. based investor investing in the U.S. Most of his data set spans the period prior to the current regime of free-floating exchange rates. Despite these limitations, Froot’s reasoning has been popular with investors. To our best knowledge, we are the first study to test whether Froot’s results carry over to the post-Bretton Woods exchange rate regime and to investors from markets other than the U.K. Over the last 35 years, we find that the case for hedging is generally not decreasing with an increasing investment horizon. While in some cases hedging becomes less effective in reducing risk at investment horizons of up to five years, there are also cases where over-hedging, i.e. shorting a currency, is the optimal risk minimizing strategy. In this context, particularly being short Yen has reduced portfolio risk as the associated carry trade profits were largely uncorrelated with bond and equity returns thus providing diversification benefits.

Most studies on currency hedging have taken the position of a U.S. Dollar based investor. We provide empirical evidence from the perspectives of investors in four major advanced economies. Our first insight in this regard is that results cannot be generalized from one base currency to another – an investor’s base currency matters significantly for drawing conclusions on a currency hedging policy.

Hedging or, more generally, currency exposure affects both return and risk of foreign investments. From a risk perspective, we find that for bond portfolios full hedging is the optimal strategy in almost all cases. This is in line with previous studies. The reason is that exchange rate volatility dominates bond return volatility. For equity investments the risk case is more complex because covariances of equities and currencies contribute much more to overall foreign investment risk than in the case of bonds. When we look at investments in one foreign country at a time we find a particularly strong positive correlation between the British Pound and equity markets and a negative correlation between the DM/Euro and equities over our sample period. Consequently, risk-minimizing investors should have hedged or even over-hedged exposure to the British Pound while maintaining some exposure to the German currency would have been optimal. In multi-country portfolios, we confirm that short positions in the British Pound and long positions in the DM/Euro would have been optimal. For the Yen and the U.S. Dollar we find that over our entire sample full hedging would have been optimal, with both over- and under-hedging being optimal in sub-periods.

From a return perspective, currency exposure can have important consequences for returns although the differences in hedged and unhedged returns are not statistically significant for the most part. This follows directly from the well-documented failure of uncovered interest rate parity which links interest rate differentials between countries to expected exchange rate movements. In line with the literature on the forward bias, we find that investors from low interest rate currencies, particularly the Japanese Yen, would have benefited from keeping the currency exposure associated with foreign investments.

In addition to differences in country coverage, our results partially differ from previous studies for two reasons. First, we improve the data quality of hedged return series by using 3-month bank deposit rates instead of the traditionally used T-bill rates which are not entirely comparable across countries.3 Second, we cover a longer time period than most studies. This is important because correlations between currencies and equities/bonds are not stable over time. In this regard the financial crisis of 2007-2009 stands out. During this time period the Euro, which has tended to move against equity markets up to the crisis, became extremely pro-cyclical falling along with equity markets. Investors who would have sought exposure to the Euro based on historical evidence would have incurred substantial currency losses in addition to losses on their other assets.4

The remainder of the paper is organized as follows. Section II briefly reviews the related literature. Section III describes our dataset. Section IV decomposes returns and variances of international investments into exchange rate and asset exposure. Section V describes our hedging approach. Section VI provides empirical results for half and full hedging. Section VII discusses risk minimizing hedge ratios. Section VIII analyzes the importance of the investment horizon for the decision to hedge and section IX concludes.

II. Literature Review

The optimal degree of currency hedging is controversial and depends on the motivation of investors’ demands for currency. Currency exposure affects portfolio risk but also affects returns to the extent that returns on foreign currency are not zero.

Based on risk considerations, full hedging of currency risk, i.e. zero demand for currencies, is optimal assuming that foreign currencies are uncorrelated with other assets (Solnik, 1974). Perold and Shulman (1988) recommend full hedging of investment related currency risk based on the assumption that currency returns are zero in the long-run and that correlations of currencies with other asset classes are close to zero on average. They proclaim currency hedging as a “free lunch” for investors arguing that it reduces risk without affecting returns. The additional risk reduction from hedging currency exposure is estimated to be as large as the gains from diversifying abroad in the first place. Similarly, Eun and Resnick (1988) show that currency risk is largely undiversifiable and that it reduces the gains from international diversification. In their study, they highlight the practical problem of estimating the right amount to hedge. That is, the return on a foreign equity investment is unknown at the time the hedge arrangement is put into place. Investors can only hedge the expected return not the actual return. This effect is often neglected, particularly in studies using a log-return representation which implies continuous hedging.

Campbell et al. (2010) find that the U.S. Dollar, the Euro, and the Swiss France have moved against world equity markets over the period 1975 to 2005. Therefore they suggest that risk-minimizing equity investors should seek exposure to these currencies. For bonds full hedging tends to be optimal in their sample, a finding that we confirm. Similarly, Glen and Jorion (1993) find that optimal currency hedging substantially reduces risk for equity investors.

Froot (1993) makes the case for not hedging exchange rate risk over long investment horizons. His argument is based on mean-reversion of real exchange rates to purchasing power parity (PPP). He tests the hypothesis that PPP provides an automatic hedge on 200 years of data for a U.K. based investor investing in the U.S. For equities, Froot finds that for investment horizons beyond two years full hedging does not reduce the variance of returns compared to no hedge. For bonds, hedging appears to be more useful as full hedging significantly reduces the variance of returns over holding horizons of up to five years.

Practioners tend to be pragmatic in determining hedge ratios. Often they use simple hedge ratios of 0, 50, and 100 percent. For example, providers of major hedged indices such as MSCI and S&P hedge each foreign currency in an index fully back into the base currency using beginning-of-period investment values. A likely reason for practioners not determining optimal hedge ratios in a portfolio context is the instability of the approach. We are sympathetic to the notion of ignoring potential correlations of currencies with equities. In our dataset we find that currency-equity correlations are unstable and fluctuate from plus 40 percent in one decade to minus 40 percent in the next decade for some currency-equity pairs. Similarly, Black (1989a) shows that, depending on the input data, hedge ratios over a very wide range of values can be optimal.

Much of the hedging literature naturally focuses on risk. However, the evidence on the failure of uncovered interest rate parity (see, for example, Fama (1984) and Engel (1996)) suggests that currency excess returns are not always zero.5 The literature finds that currencies of countries with low interest rates tend to not appreciate as much as suggested by the parity condition. The opposite holds for currencies of countries with high interest rates. This effect is behind the global currency carry trade where investors borrow in a low yielding currency and lend the proceeds in a high yield currency. Hedging currency risk associated with foreign investments removes these carry trade profits for investors from low interest rate currencies while it may enhance the returns to high interest rate currency investors. We find the effect of currency excess returns to be economically large but statistically insignificant. A second, speculative impact of currencies and hence hedging on returns results from currency returns to investors in different countries being quoted in terms of different numeraire currencies. Black (1989) points out that each party in a currency trade can simultaneously perceive positive returns. This manifestation of Jensen’s inequality is known as Siegel’s paradox and can explain symmetric speculative demands for currencies. Campbell et al. (2010) highlight that the demand for currency generated by this effect is quite small in practice given the high volatility of currencies.

III. Data

The sample data covers the period from January 1975 to December 2009. All data series are available on a monthly basis and we present results for investment horizons of up to five years. Country stock index returns are provided by Morgan Stanley Capital International (MSCI). Each of the indices is value-weighted, formed from all companies in the market that fulfill minimum requirements for size, liquidity and free-float, and adjusted for dividend payments on a daily basis. Long-term bond portfolio returns are not available prior to 1986 for all countries. We therefore use the approximation suggested by Campbell, Lo, and MacKinlay (1997) to obtain holding-period returns from bond yields.6 Government bond yields as well as spot exchange rates and Consumer Price Indices (CPI) are obtained from the IMF’s International Financial Statistics (IFS). Three-month deposit rates are obtained from IFS in the case of Japan and DataStream for Germany, France, U.K. and U.S. With the introduction of the Euro interest rate differentials between Germany and France have virtually disappeared and there are of course no more exchange rate movements. We therefore only present the German perspective in all tables following Table 1.

Table 1.

Summary Statistics – Annual 1/

article image
Sources: Author’s estimates, IMF IFS, DataStream.

Arithmetic averages and standard deviations of rolling annual changes/returns in percentages. Data coverage extends from 1975M1 to 2009M12. Data are on a monthly basis. CPIs and bond yields are obtained from the IMF’s IFS. Stock market returns are from Morgan Stanly International. Three-month interbank deposit rates are from IFS for Japan and from DataStream for the other countries.

Table 1 reports arithmetic averages and standard deviations of rolling annual changes/real returns of the Consumer Price Index (CPI), 3-month deposit rates, stock and bond returns for the full sample period from 1975 to 2009. Returns are in local currency terms and adjusted for the local CPI. The table therefore allows for the comparison of returns domestic investors can expect in their respective markets. Returns to foreigners are addressed in the next section.

Inflation, as measured by the CPI, has been highest in the U.K. with 5.6 percent per year followed by France (4.4 percent) and the U.S. (4.2 percent). Germany and Japan have experienced moderate inflation of 2.5 percent and 1.8 percent, respectively. Annualized real three-month rates on wholesale deposits with banks range from only 0.2 percent in Japan to 3.3 percent in France. Volatility of deposit rates has been low not exceeding 3.1 percent per year for any country. Real equity market returns to local investors vary substantially across countries. While a Japanese investor has only earned about 5.4 percent per year, a French investor has received 9.9 percent over the sample period. The equity premium over long-term bonds is just above 1 percent in Japan and substantially below the other markets. Equity returns are associated with substantial volatility in all countries with volatility being somewhat lower in the U.K and U.S. than in the other markets. Real returns on long-term government bonds are between 4 and 5 percent for all countries and volatilities are between 6 and 8 percent.

IV. Components of International Investment Returns

In this section, we examine the effect of currency fluctuations on the return and risk of foreign investments. After establishing some notation, we present results from the viewpoint of investors based in Germany, Japan, the U.K., and the U.S.7

Consider an investor who uses a certain base currency and is invested in a foreign currency investment. Her nominal unhedged return measured from time t - 1 to t is given by:

r˜U,t=(1+x˜t)(1+e˜t)1(1)

where x˜t is the return in foreign currency on the investment between time t - 1 and t; t is the percentage change in the base currency per unit of foreign currency over the same period. The tilde symbol identifies random variables. Equation (1) can be written as

r˜U,t=x˜t+e˜t+x˜te˜t(2)

Since the cross-product in equation (2), x˜te˜t, is small in magnitude, r˜U,t can be approximated by8

r˜U,tx˜t+e˜t(3)

Based on equation (3), the variance of foreign investment returns is approximately

var(r˜U,t)var(x˜t)+var(e˜t)+2cov(x˜te˜t)(4)

As equation (4) shows, exchange rate fluctuations contribute to the variance of unhedged foreign investment returns through their own variance and their covariance with foreign asset returns.

The preceding analysis is analog for real returns:

r˜U,t=(1+x˜t)(1+e˜t)/(1+π˜d,t)1(5)

We adjust returns for inflation in an investor’s home market, π˜d,t, as opposed to adjusting returns for the inflation in the market where returns are achieved. The reason is that inflation in her home market is the relevant measure for an investor that tries to preserve her domestic purchasing power.

Table 2 presents exchange rate gains/losses, currency excess return, and unhedged equity and bond returns on a quarterly basis for investors investing in France, Germany, Japan, the U.K. and the U.S. Currency excess returns are returns from borrowing in domestic currency for 3-months, lending the proceeds in foreign currency for the same period, and exchanging back into domestic currency after three months to repay the domestic currency loan. We assume that investors can borrow and lend at the same rate. In reality, currency excess returns to investors would be lower because of transaction costs and bid-ask spreads.

Table 2.

Quarterly Returns to International Investments 1/

article image
Sources: Author’s estimates, IMF IFS, DataStream.

Data coverage extends from 1975M1 to 2009M12. All entries are in percentages.

The exchange rate gain/loss is the change in the investor’s base currency per unit of foreign currency over one quarter.

The currency excess return is the return to an investor of borrowing in her domestic currency to invest in foreign currency deposits.

Unhedged stock and bond returns are the sum of local currency returns, exchange rate gains/losses and interaction between local currency and exchange rate returns.

The following example illustrates the table. A German investor investing in Japan would have gained 0.62 percent on average per quarter on exchange rate movements. An exchange rate gain implies a depreciation of the investor’s home currency vis-à-vis the foreign currency, so in this case the German DM/Euro has on average depreciated against the Japanese Yen. The currency excess return from borrowing in DMs/Euros and lending in Yen is, however, a negative 0.11. This implies that the exchange rate gain for the German investor is more than offset by the lower interest rate earned on the Yen deposit compared to the DM/Euro denominated loan. This is the flipside of the so-called currency carry trade.9 The gains from favorable currency movements boost the returns to a German investor investing in the Japanese stock and bond market by about 0.62 percent compared to the domestic returns of a Japanese investor. The interaction term x˜te˜t adds an additional 0.03 percent. Against other currencies, German investors generally realized exchange rate losses on foreign investments as a result of a strong home currency. The losses against investments in France all predate the introduction of the euro and indicate the depreciation of the Franc against the DM.

Japanese investors experienced exchange rate losses on investments in all countries considered in this study on the back of strong Yen appreciation. Currency excess returns from a Japanese perspective are substantial ranging from 0.35 percent for the U.S. to 0.67 percent for France on a quarterly basis. This is consistent with the fact that the Japanese Yen has been the funding currency for the global currency carry trade for many years. Positive currency excess returns imply that the Yen has not appreciated as much as suggested by uncovered interest rate parity.

The British Pound has depreciated on average against all other currencies in this study resulting in exchange rate gains on foreign investment for British investors. Similarly, U.S. dollar investors have gained from currency movements, except on their investment in the U.K.

The preceding discussion considers nominal returns. In real terms domestic inflation needs to be taken into account when comparing returns. In many cases exchange rate gains/losses compensate only partially for higher/lower domestic inflation. For example, in the case of the U.K., a country with high average inflation, domestic stock returns still exceed foreign stock returns despite substantial exchange rate gains.

Excess currency return pairs are generally above zero because percentage gains/losses are quoted in different numeraire currencies for investors from different countries, as noted earlier, an effect known as Siegel’s paradox.

Tables 3a and 3b present the breakdown of the volatility of returns to international investors into different components. Exchange rate volatility contributes between 16 and 40 percent to the volatility of investing in foreign stock markets.10 For bond portfolios, exchange rate risk dominates overall volatility contributing up to 95 percent of total unhedged return volatility. The larger relative importance of exchange rate risk for bond portfolios compared to equity portfolios explains why practioners tend to view hedging exchange risk in the case of bonds as much more important. The covariance of currency returns with bond and equity returns matters generally a lot less for overall investment volatility than currency volatility itself. We also find covariance structures to be unstable over time in many cases.

Table 3a.

Variance Decomposition of Quarterly Returns 1/

article image
Sources: Author’s estimates, IMF IFS, DataStream.

Data coverage extends from 1975M1 to 2009M12. All entries are in percentages. Column (1) contains the variance of local currency returns, column (2) the variance of exchange rate gains/losses, column (3) the covariance and column (4) the correlation of local currency and exchange rate returns. Column (5) shows the overall variance of unhedged returns. Columns (6) through (9) show the percentage contributions of variance components to the overall variance of unhedged returns. Additional terms in column (9) include the variance of (x˜t * e˜t) the covariance of (x˜t,x˜t  * e˜t) and the covariance of (e˜t,x˜t  * e˜t).

Table 3b.

Variance Decomposition of Quarterly Returns 1/

article image
Sources: Author’s estimates, IMF IFS, DataStream.

Data coverage extends from 1975M1 to 2009M12. All entries are in percentages. Column (1) contains the variance of local currency returns, column (2) the variance of exchange rate gains/losses, column (3) the covariance and column (4) the correlation of local currency and exchange rate returns. Column (5) shows the overall variance of unhedged returns. Columns (6) through (9) show the percentage contributions of variance components to the overall variance of unhedged returns. Additional terms in column (9) include the variance of (x˜t  *  e˜t) the covariance of (x˜t,x˜t  * e˜t) and the covariance(e˜t,x˜t  * e˜t).

The covariance between local currency stock market returns and exchange rate movements is positive in all cases for a German investor. Exchange rate movements are thus found to reinforce, rather than offset, the stock market movements in this case. From a Japanese perspective covariances are positive except for the French stock market. For the U.K., with the exception of the Japanese stock market, and for the U.S. covariances between exchange rate changes and stock market returns are negative thus offsetting some of the stock market movement.

In the case of bond markets, German and U.S. investors have benefited from negative co-movement between local currency and exchange rate returns except for Japanese bonds. Yen-based investors have generally benefited from risk reduction through a negative covariance between bond and exchange rate returns, while British investors have experienced positive covariances.

Eun and Resnick (1988) extend the preceding analysis to a portfolio context. As they show, in the multi-currency case overall portfolio risk of foreign investment depends on (a) the covariances among stock market returns, (b) the covariances among the exchange rate changes, and (c) the cross-covariances among the stock market returns and the exchange rate changes.

V. Hedging Currency Risk

The previous section has shown the substantial contribution of exchange rate risk to the overall risk of international investments. It is therefore natural for investors to consider hedging exchange rate exposure. In this section we develop a framework for calculating hedged returns. Sections V and VI present empirical evidence.

A. Hedging methodology and notation

One way to implement a currency hedge involves short-term borrowing in foreign currency and lending the proceeds in the investor’s base currency. A fully hedged investor would borrow the present value of the expected foreign investment proceeds, i.e. [1+E(x˜t)]/(1+if,t1), where if,t–1 represents the foreign interest rate, and exchange the proceeds at the spot invest at the domestic interest rate id,t–1. At maturity the investor would repay the foreign currency loan valued 1+E(x˜t) with the expected proceeds on the foreign investment. This hedging strategy is imperfect to the extent that the realization of the return on the foreign investment deviates from its expectation at time t-1. For example, consider a U.S. Dollar 10 million investment for a Japanese investor. Selling U.S. Dollar 10 million to buy Yen perfectly hedges the exchange rate exposure for as long as the value of the investment remains U.S. Dollar 10 million. However, any movement in the U.S. dollar asset value will reduce the effectiveness of the hedge. For instance, if the value of the Yen-hedged investment increases to U.S. dollar 12.5 million, the investment remains hedged only for the original U.S. dollar 10 million. The differential of U.S. dollar 2.5 million is fully exposed to currency movements. The quality of the hedge depends on the predictability of the underlying asset’s value which is, inter alia, a function of the investment’s volatility and the hedge horizon.

The preceding discussion shows that due to estimation risk it is impossible to obtain ex ante exactly the desired target hedge ratio, i.e. the proportion of an investment’s currency exposure that is hedged. Eun and Resnick (1988) discuss and test several approaches to estimating E(x˜t) in the context of currency hedging. Practioners, however, often simply hedge the beginning-of-period value of their investments, in effect setting E(x˜t)=0.11 We find that for quarterly returns this approach is sensible given the difficulties associated with forecasting returns. The data support this view – the average quarterly return due to the unhedged currency exposure of the difference between beginning- and end-of-period investment values is below 0.07 percent for all base currency/foreign investment combinations considered in this paper. In the empirical section, we therefore proceed by only hedging beginning-of-period investment balances. As we will discuss in section VIII, estimation risk can, however, have a very large impact on returns over long periods.

The hedge ratio can be varied to arrive at investment portfolios that are over- or under-hedged to varying degrees. Investors may seek to take active currency risks based on their views on future currency movements. Many studies have also pointed out that hedging 100 percent of currency exposure is not optimal from a risk minimization standpoint when currencies and equities/bonds are correlated.

The domestic currency return on the borrowing/lending hedge over the period t-1 to t is given by

h˜t=1(1+e˜t)1+id,t11+if,t11(6)

Let Φt be the hedge ratio. The return on a hedged investment is then a combination of the proportion of the expected investment value the investor chooses to hedge, the proportion of the expected investment value left unhedged, and the unexpected return on the investment which is exposed to currency risk:

r˜H,t=Φt[1+E(x˜t)](1+e˜t)(1+h˜t)+[1Φt][1+E(x˜t)](1+e˜t)+[(x˜tE(x˜t)](1+e˜t)1=Φt[1+E(x˜t)]1+id,t1+if,t+[1Φt][1+E(x˜t)](1+e˜t)+[(x˜tE(x˜t)](1+e˜t)1(7)

Proceeding by setting E(x˜t)=0, equation (7) simplifies to

r˜H,t=Φt1+id,t11+if,t1+[1Φt](1+e˜t)+x˜t(1+e˜t)1(8)

The same hedged result can be achieved with lower transaction costs by employing currency forward contracts.12,13 An investor would sell the proportion of expected foreign currency proceeds that she wishes to hedge in the forward market capturing the forward exchange premium/discount ft, equal to Ft-1/St-1 -1, where Ft-1 and St-1 are, respectively, the forward and spot exchange rates in domestic currency equivalents. This hedging practice, of course, also leaves residual foreign exchange exposure through unexpected foreign currency proceeds. The hedged return based on a forward hedge is therefore given by

r˜H,tk=Φt[1+E(x˜t)](1+ft)+[1Φt][1+E(x˜t)](1+e˜t) +[(x˜tE(x˜t)](1+e˜t)1(9)

To see that hedging using borrowing/lending and hedging using forwards yields equivalent results if covered interest rate parity (CIP) holds, note that this arbitrage condition links the forward premium to interest rates:

1+id,t11+if,t1=1+ft(10)

For this relationship to hold, both interest rates must be based on instruments with identical default risk, maturity, and liquidity. Given equation (10), equation (9) is identical with equation (7). In the absence of investment barriers, CIP must hold to preclude arbitrage opportunities. Empirical research generally finds strong evidence of CIP.14

Using equation (10) in equation (8) yields

r˜H,t=Φt[1+ft]+[1Φt](1+e˜t)+x˜t(1+e˜t)1  =x˜t+e˜t+x˜te˜t+Φt[fte˜t](11)

A comparison of equation (11) with the unhedged return on an international investment in equation (2) shows that by hedging exchange rate risk an investor replaces the stochastic gain or loss on the exchange rate, t, with the forward premium/discount, ft, which is known at the time of the investment. If the investor hedges 100 percent of the beginning-of-period exchange rate exposure, equation (11) becomes

r˜H,t=x˜t+ft+x˜te˜t(12)

B. Impact of hedging on returns

Currency hedging is sometimes described as a “free lunch” (Perold and Schulman, 1988) based on the argument that currencies add only volatility but have zero expected returns. In the preceding notation, currency hedging affects returns if the unconditional expectation of h˜t is different from zero.

From a theoretical perspective, if investors are risk neutral and have rational expectations, then E(h˜t)=0, a relationship known as uncovered interest parity (UIP). UIP implies that the interest differential between a domestic and a foreign market is an estimate of the future exchange rate changes. UIP, however, is not a pure arbitrage condition. To see this, suppose the 3-month U.S. interest rate is 5 percent and the 3-month Euro interest rate is 3 percent.15 Risk neutral, rational investors must expect the U.S. Dollar to depreciate by about 2 percent over the next 3 months to make both investments equally attractive. If, for example, risk neutral and rational investors would expect a smaller U.S. Dollar depreciation of 1 percent, they would borrow in Euros and lend in U.S. Dollars, thus driving up Euro rates and down U.S. Dollar rates until the interest differential is also equal to 1 percent. This is clearly not a riskless arbitrage opportunity as exchange rates may not move in line with the parity condition.

Indeed, a large body of empirical literature finds that UIP does not hold; a failure often referred to as the forward discount bias.16 Empirically, low interest rate currencies tend to not appreciate as much as the interest rate differential and high interest rate currencies do not depreciate as much as the interest rate differential.17 The failure of UIP suggests that in some cases hedging affects expected mean returns of foreign investments.

C. Impact of hedging on volatility

For most investors, hedging currency exposure is about reducing the volatility of foreign investments. In subsection A. it was shown that hedging replaces the stochastic exchange rate gain/loss with the ex ante known forward premium/discount. The volatility of a hedged return series compared to the equivalent unhedged return series thus depends on the volatility of t versus the volatility of ft. In a preview of the findings presented in the empirical sections, we find that the quarterly volatility of ft is only about 7 to 16 percent of the volatility of t.18 Hedging, therefore, has the potential to reduce volatility substantially at least at short investment horizons. Mean-reverting properties of exchange rate movements could potentially change this result for longer horizons, an issue addressed in section VIII.

In addition to the volatility of foreign exchange, the correlation of currencies with other assets matters for the risk properties of investment-related currency exposure. For example, a foreign currency that tends to depreciate/appreciate relative to the investor’s domestic currency when the foreign equity market increases/decreases offsets some of the risk of the underlying investment. Investors should ideally retain some exposure to such a currency. On the other hand, a currency that is expected to reinforce asset market movements should be over-hedged, i.e. sold short.

D. Calculating the forward premium in practice

The calculation of hedged returns requires data on interest rates in the investor’s base currency and in the foreign currency. To be comparable across countries, interest rates should be based on instruments with the same maturity, credit risk and liquidity.

We consider 3-month deposit rates and 3-month T-bill rates as candidate rates that are available across countries. A problem with T-Bill rates is that no 3-month paper is issued by the German government and that France only started issuing 3-month paper in 1989. For Japan, 3-month government paper was relatively illiquid before 1999 and therefore, the Bank of Japan deemed the interest rate on these financing bills as not representative of market conditions in Japan (IMF, 2000). In support of this conclusion, we find that the interest rate on Japanese T-bills is often stale before 1999, sometimes not changing for up to two years. For 3-month deposit rates, comparability across France, Germany, Japan, the U.S., and the U.K. is better than for T-Bills. Rates starting in 1975 are available in DataStream for all countries except Japan. For Japan the IMF’s International Financial Statistics provide the relevant deposit time series.

We check the comparability of interest rates across countries by comparing interest-rate-based forward premia to forward premia derived from forward and spot exchange rates for the period 1990 to 2009. By covered interest rate parity both calculation approaches should yield the same result. Any systemic deviation would suggest that the employed interest rates are not comparable across countries. Table 4 presents the comparison of forward premia calculated from exchange rates and forward premia based on deposits and T-Bills for the U.S. dollar. The presented differences are for quarterly premia. Forward premia derived from deposit rates are generally closer to “true” exchange-rate-based forward premia. The improvement is particularly large for France, Germany, and Japan. This is consistent with the French T-Bill rate being partially and the German T-Bill rate being entirely based on 12-month maturity rates. For Japan, as mentioned above, the problem is likely to be the absence of a liquid secondary market for T-Bills until 1999. We conclude that deposit rates provide a more accurate approximation of the forward premium and continue by using deposit rates in our calculations.

Table 4.

Forward Premia versus the U.S. Dollar Derived from Deposit Rates and T-Bills Compared to Exchange Rate based Premia (1990M1-2009M12) 1/

article image
Sources: Author’s estimates.

This table provides a comparison of forward premia against the U.S. dollar derived from forward and spot exchange rates with forward premia derived from interest rate differentials.

2/ Panel A contains forward premia based on differentials in three-month bank deposit rates compared to exchange rate based forward rates.3/ Panel B contains forward premia derived from T-Bill rate differentials compared to exchange rate based forward rates.

VI. Simple Hedge Ratios

In this section, we present empirical evidence for the impact of currency hedging for the full sample period 1975 to 2009. The academic literature has pointed out that hedge ratios deviating from 100 percent can be optimal in the presence of correlation between exchange rate and asset movements. A survey of the hedging policies of institutional investors in major markets in 2004 by Russell/Mellon suggests however that a majority of investors chooses to hedge 0, 50, or 100 percent of foreign currency exposure. The reluctance of practioners to calculate optimal hedge ratios and to treat currencies like other assets in a portfolio optimization framework may be partially attributable to the instability of hedge ratios (see, for example, Black 1989a). We proceed by presenting results for unhedged, fully hedged, and 50 percent hedged portfolios. Optimal hedge ratios are addressed in the next section.

A. Single-country portfolios

Tables 5a and 5b show returns on unhedged and fully hedged single-country portfolios on a quarterly basis. Returns are additive, therefore, with the results for no hedging and 100 percent hedging, results for any other hedge ratio can be obtained. In almost all cases the null hypothesis of equal means of hedged and unhedged quarterly returns cannot be rejected at conventional levels. The substantial sample variance of the return series, especially for equities, makes it difficult to find statistically significant differences. In economic terms, many of the return differentials between hedged and unhedged portfolios are, however, substantial. For example, a Japanese investor in the French stock market would have earned quarterly returns of 2.75 percent without hedging currency risk and only 2.1 percent on a hedged basis. The difference is not statistically significant but an approximate annual return differential of 2.6 percent over the last 34 years is very relevant to investors.

Table 5a.

Quarterly Returns on Hedged and Unhedged Portfolios 1/

article image
Sources: Author’s estimates.

Quarterly returns on unhedged and fully hedged stock and bond portfolios from the perspectives of German and Japanese investors. Hedged returns are based on rolling quarterly hedges of beginning-of-period balances. Reported T-statistics are based on the null hypothesis of equal returns.

Table 5b.

Quarterly Returns on Hedged and Unhedged Portfolios 1/

article image
Sources: Author’s estimates.

Quarterly returns on unhedged and fully hedged stock and bond portfolios from the perspectives of British and American investors. Hedged returns are based on rolling quarterly hedges of beginning-of-period balances. Reported T-statistics are based on the null hypothesis of equal returns.

For German investors returns are generally higher on an unhedged basis than on a hedged basis except in the case of investments in Japan. Excluding investments in France, British investors have yielded higher returns if they chose to hedge currency risk. Japanese and American investors have experienced generally lower returns if hedged.

The, in some cases, economically substantial return differentials between hedged and unhedged returns point to a failure of UIP. Differences are especially large in the case of Yen-based investors who would generally have yielded higher returns without hedging. Going back to Table 2, this finding may surprise given that Japanese investors would have experienced exchange rate losses against all other currencies. The explanation is that for the Yen the forward premium, ft, is generally even more negative than the exchange rate loss, t. Interest rate differentials have thus predicted an even larger Yen appreciation than actually materialized. Japanese investors who chose to remain unhedged on their international investments in effect engaged in a carry trade speculating that the Yen will not appreciate as much as suggested by UIP. The Japanese experience also highlights that currency hedging does not allow international investors to access local asset returns as sometimes stated.

We now turn to the potential of currency hedging to reduce risk. Tables 6a and 6b present standard deviations for unhedged, half hedged and fully hedged bond and equity portfolios. Hedging currency risk reduces the risk of international investments in almost all cases significantly statistically as well as economically. The case for hedging is particularly apparent for bond portfolios. For bond portfolios hedging 100 percent of currency exposure is the dominant strategy from a risk reduction standpoint. Hedging is more effective for bonds because, as table 3 shows, currency risk makes up a large portion of the overall risk of international bond portfolios. Full hedging reduces risk more than half hedging in all cases except the French and German stock markets from a U.K. investor perspective and the German stock market from a U.S. perspective. In these cases it is optimal for risk-minimizing investors to retain some currency exposure because the foreign currency has on average moved against the foreign stock market thus providing hedging benefits. Specifically, the Euro in terms of British Pounds and U.S. Dollars tended to appreciate when the French and German stock markets have fallen. For the Euro and DM against the British Pound this effect is present during the entire sample period. It is particularly strong over the last twenty years and even more during the financial crisis of 2007 to 2009. For investments in the French stock market, the effect is entirely due to the period after the Euro introduction. Against the U.S. dollar the Euro and the DM have moved against the German equity market during the period 1990 to 2007. During the financial crisis this pattern has dramatically changed as the Euro fell against the U.S. Dollar along with equity markets.19

Table 6a.

Quarterly Standard Deviations of Hedged and Unhedged Portfolios 1/

article image
Sources: Author’s estimates.

Quarterly standard deviations of returns on unhedged, half hedged and fully hedged stock and bond portfolios from the perspectives of German and Japanese investors. Hedged returns are based on rolling quarterly hedges of beginning-of-period balances. Reported F-statistics are based on the null hypothesis of equal variances.

Table 6b.

Quarterly Standard Deviations of Hedged and Unhedged Portfolios 1/

article image
Sources: Author’s estimates.

Quarterly standard deviations of returns on unhedged, half hedged and fully hedged stock and bond portfolios from the perspectives of British and American investors. Hedged returns are based on rolling quarterly hedges of beginning-of-period balances. Reported F-statistics are based on the null hypothesis of equal variances.

We conclude that for bonds hedging unequivocally reduces risk at quarterly horizons but, depending on an investor’s base currency, the risk reduction may come at the price of lower returns. For equities there is also strong evidence for the effectiveness of hedging to reduce quarterly return volatility. However, the empirical results to this point also indicate that hedge ratios other than 100 percent are optimal for equities in some cases where correlations of currencies with equities are large. We provide more evidence on this in section VII.

B. Multi-country portfolios

In the previous section, we have presented the effect of hedging currency exposure for investors invested in a single foreign market. We now turn to the impact of hedging on multi-country portfolios. Returns on multi-country portfolios are simply weighted averages of single-country portfolios. For portfolio risk, however, results in the multi-country context depend on the covariances among the stock/bond market returns, the covariances among the exchange rate changes, and the cross-covariances among the stock/bond market returns and the exchange rate changes.

We form portfolios by equally weighting the French, German, Japanese, British, and U.S. stock and bond markets.20 Table 7 presents quarterly unhedged and hedged returns. Similar to the single-country analysis in tables 5a and 5b, the differences between unhedged and hedged returns are statistically not significant in most cases. In economic terms however, Japanese investors and to a lesser extent U.S. Dollar based investors would have yielded substantially higher returns on unhedged investments as opposed to hedged investments.

Table 7.

Quarterly Returns on Hedged and Unhedged Equal-Weighted Portfolios 1/

article image
Source: Author’s estimates, DataStream, IMF IFS.

Data coverage extends from 1975M1 to 2009M12.

Global stock portfolios include the MSCI country indices for France, Germany, Japan, the U.K. and the U.S. in equal proportions.

Global bond portfolios include returns on long-term government bonds for France, Germany, Japan, the U.K. and the U.S. in equal proportions.

We present quarterly standard deviations of unhedged and hedged returns for multi-country portfolios in Table 8. Hedging currency exposure results in economically and statistically significant risk reduction in almost all cases. The ratio of variances between unhedged and hedged returns for diversified portfolios is comparable to the single-country cases presented in Tables 6a and 6b. We conclude that currency risk is largely undiversifiable.

Table 8.

Quarterly Standard Deviations of Hedged and Unhedged Equal-Weighted Portfolios 1/

article image
Source: Author’s estimates, DataStream, IMF IFS.

Data coverage extends from 1975M1 to 2009M12.

Global stock portfolios include the MSCI country indices for France, Germany, Japan, the U.K. and the U.S. in equal proportions.

Global bond portfolios include returns on long-term government bonds for France, Germany, Japan, the U.K. and the U.S. in equal proportions.

VII. Optimal Hedge Ratios

To this point we have only considered no, half, and full hedging of currency risk, which are by far the most popular hedging strategies with institutional investors. Optimal hedge ratios, however, are usually defined as the hedge resulting in the greatest risk reduction.

A. Single-country portfolios

We estimate optimal hedge ratios for German, Japanese, British and American investors investing in foreign equity and bond markets. From equation (11) it follows that minimizing the variance of a hedged return with respect to the hedge ratio, Φ, is equal to

minΦVar(x˜t+e˜t+x˜te˜tΦt[e˜ft])(13)

The first three terms in equation (13) are equal to the unhedged return. In order to find the risk minimizing hedge ratio we perform an OLS estimation of the following equation:

r˜U,t=α+β[e˜tft]+εt(14)

where the estimate of β is the estimate of the minimum-variance hedge ratio.

We present estimated minimum-variance hedge ratios and associated Newey-West standard errors to correct for autocorrelation due to overlapping return intervals in Table 9. Optimal hedge ratios for investments in foreign bond portfolios are essentially one for investors in all base currencies – this is consistent with results for bonds in section VI. Since correlations between bond returns and exchange rate movements are in some cases not insignificant, the reason must be that bond volatility is dominated by exchange rate volatility. For equities the case is more interesting because the volatility of this asset class is higher so that equity market – exchange rate correlations matter.

Table 9.

Estimated Minimum Variance Hedge Ratios

article image
Source: Author’s estimates, DataStream, IMF IFS.

Minimum-variance hedge ratios for quarterly returns are obtained by regressing the unhedged return on the row stock and bond markets on the associated exchange rate gain minus the forward premium. All regressions include an intercept. We run monthly regressions on overlapping quarterly returns.

Standard errors are corrected for autocorrelation due to overlapping intervals using the Newey-West procedure.

From a German perspective the risk minimizing hedge strategy over the sample period would have been to hedge about 100 percent of currency exposure in all cases except for investments in the UK stock market. For the UK stock market German investors should have hedged 140 percent of currency exposure, i.e. they should have taken a short position in the British pound. The reason for this is the large positive correlation of 16 percent between the UK stock market in local currency terms and the DM/euro exchange rate versus the British pound. The UK stock market has tended to do well/bad when the Pound has appreciated/depreciated against the German currency. The exchange rate movements have therefore magnified the stock market movements. A possible explanation is that economic problems in the U.K as proxied by falling stock prices lead to capital outflows into Germany and thus a falling pound versus the DM/Euro. Correlations of the German exchange rate versus the Yen and the U.S. dollar with these countries’ respective stock markets are very close to zero over the entire period. In the first half of the sample period the correlations are large and positive, similar to the UK, but this is offset by large negative correlations in the latter part of the sample.

For Japanese investors risk minimizing hedge ratios are statistically indistinguishable from one in all cases. In sub-periods there are strong positive and negative correlations between the Yen exchange rate and foreign stock markets but overall there is no consistent effect so that correlations for the entire sample period are close to zero.

As a mirror image to German investors, U.K. investors should have retained some exposure to the German currency. As mentioned before, the German stock market has tended to do well/bad when the British Pound has appreciated/depreciated against the German currency. A similar effect exists for investments in the French stock market but this is entirely due to the second half of the sample period after the Euro introduction. The DM/euro thus has been a “safe haven” currency for British investors – it has done well during falling stock markets.

Similar to British investors, but to a lesser extent, risk-minimizing U.S. investors should have slightly under-hedged their stock market investments in Germany and France. For investments in Japan and the U.K. the optimal hedge ratio is indistinguishable from one.

B. Multi-country portfolios

The simultaneous estimation of minimum variance hedge ratios for portfolios containing investments in several countries allows investors to achieve optimal results by exploiting the full covariance structure. A potential danger is over-fitting to the sample. This would be a problem if covariances are time period specific. For this reason we present results for the first and second halves of our sample period in addition to full sample results.

Our estimation approach for multi-country portfolios is an extension of the single-country case. The hedged return on a portfolio of N different countries with weight ωi,t for country i at time t can be written as

r˜H,tP=i=1Nωi,tr˜U,i,ti=1Nωi,tΦi,t[e˜i,tfirt](15)

We perform an OLS estimation of the following equation to find the hedge ratios that minimize portfolio variance

i=1Nωi,tr˜U,i,t=α+β1[e˜1,tf1,t]+β2[e˜2,tf2,t]++βN[e˜N,tfN,t]+εt(16)

Estimates of the optimal hedge ratio for currency exposure associated with investing in country i are then obtained from Φi = βii.

Table 10 contains estimated hedge ratios and associated standard errors for stock and bond portfolios formed by equally weighting the markets of Germany, Japan, the U.K. and the U.S over the entire sample period. Table 11 presents results for the first half of the sample ranging from 1975M1 to 1992M7, and Table 12 presents the second sub-period from 1992M8 to 2009M12. As in the single country case, hedge ratios below one imply that a risk-minimizing investor would retain some exposure to the foreign currency whereas hedge ratios above one would imply that the investor should over-hedge, i.e. short, the foreign currency. Negative hedge ratios indicate that it is optimal for risk minimizing investors to seek active exposure to these currencies beyond the exposure associated with the unhedged foreign bond/equity investment.

Table 10.

Estimated Minimum Variance Hedge Ratios for Multi-Country Portfolios – Full Sample 1975M1-2009M12 1/

article image
Source: Author’s estimates.

For investors from each base currency perspective, minimum-variance hedge ratios for quarterly returns are obtained by estimating equation 16. All regressions include an intercept. We run monthly regressions on overlapping returns. Standard errors are corrected for autocorrelation due to overlapping intervals using the Newey-West procedure.

Global stock portfolios include the MSCI country indices for Germany, Japan, the U.K. and the U.S. in equal proportions.

Global bond portfolios include returns on long-term government bonds for Germany, Japan, the U.K. and the U.S. in equal proportions.

Table 11.

Estimated Minimum Variance Hedge Ratios for Multi-Country Portfolios – First Half 1975M1 – 1992M7 1/

article image
Source: Author’s estimates.For footnotes see table 10.
Table 12.

Estimated Minimum Variance Hedge Ratios for Multi-Country Portfolios – Second Half 1992M8 – 2009M12 1/

article image
Source: Author’s estimates.For footnotes see table 10.

Over the full sample, risk-minimizing equity investors should have over-hedged exposure to the British Pound. This result is in line with our finding that the British Pound tends to be pro-cyclical. This effect is, however, driven entirely by the second half of our sample. In the first half, full hedging of exposure to the Pound would have been optimal. There is strong evidence that exposure to the DM and its successor, the Euro, is an optimal strategy for risk minimizing investors. Negative coefficients on DM/Euro exposure indicate that investors should not only not have hedged but sought additional exposure to the DM/Euro. This result is particularly strong for the second half of our sample. For the Yen and the U.S. Dollar, optimal hedge ratios are not statistically different from one at conventional levels of significance. There is some indication that over-hedging both currencies was optimal in the first half of our sample, while under-hedging was optimal in the second half. In unreported results for the financial crisis of 2007 to 2009, we find that the Yen and the U.S. Dollar have moved against equity markets with correlations jumping to the range of negative 50 to 70 percent. The Euro did not provide the hedging benefits it has exhibited over our full sample during the crisis. On the contrary, the Euro was extremely pro-cyclical falling against Yen and U.S. dollar along with stock markets. This experience provides a caveat that currency correlations as well as asset market – currency cross-correlations are unstable and may break down when investors need diversification most.

For investors in global bond portfolios, estimates of optimal hedge ratios are more precise as standard errors are much smaller. Similar to the single-country case, hedging currency exposure fully is optimal with the exception of the British Pound to which risk-minimizing bond investors should have retained some exposure.

VIII. Hedging and The Investment Horizon

Thus far our analysis has been based on quarterly returns and their associated variances. We demonstrated empirically that hedging in almost all cases reduces risk at a quarterly return horizon. In this section, we turn to the question of whether the preceding results apply at longer investment horizons, an issue of relevance for long-term investors such as endowments. In doing so, we consider investment horizons of up to 5 years while continuing to hedge returns using three-month interest rates.

At investment horizons longer than one quarter, results on the efficacy of currency hedging for reducing the risk of a foreign investment are potentially different depending on the properties of exchange rates over longer horizons as compared to short horizons. At relatively short horizons exchange rate fluctuations are dominated by changes in real exchange rates. However, Purchasing Power Parity (PPP) suggests that real exchange rates are mean-reverting over long horizons.

There is a vast literature on whether PPP holds but some consensus appears to have emerged that real exchange rates mean revert over long horizons.21 A problem of traditional empirical tests is lack of power to reject the random walk hypothesis for exchange rates. One approach to circumvent this is by using very long sample periods (100 to 200 years) – these studies find support for PPP.22 Recently, studies that incorporate nominal price rigidities, transaction costs, and non-linear adjustments are able to detect evidence in favor of PPP over shorter sample periods.

Froot (1993) applies the insights from research on real exchange rate mean reversion using long-term data sets to currency hedging. Based on empirical evidence over 200 years from the perspective of a British investor investing in the United States, Froot argues that for long-term investors mean reversion towards PPP provides a “natural hedge”. Specifically, he finds that for horizons of more than two years, the volatility of a hedged portfolio of stocks exceeds the volatility of the equivalent unhedged portfolio. For bond portfolios the crossover point is about seven years.23 In Froot’s data set the risk reduction potential of currency hedging decreases almost monotonically with an investor’s time horizon, leading him to conclude that “no hedging at all is likely to be best for those who care primarily about long-horizon moments.”

Although, to our best knowledge, there are no further studies substantiating Froot’s findings, his analysis has been influential with practioners. Froot’s empirical analysis is limited to the case of a U.K. based investor investing only in the U.S. An additional caveat pertains to the 200 year dataset which includes periods with very different exchange rate regimes.

We proceed by testing the proposition that hedging is less effective at long investment horizons on our free-floating exchange rate data set covering the perspectives of German, Japanese, British and American investors. A potential problem with our 35 year data set is that we do not have sufficient independent observations. For instance, in the case of a 5 year investment horizon we have only seven independent return intervals. This could limit the statistical significance of our results at long horizons. We acknowledge that the use of rolling returns implies overweighting of the observations in the middle of the sample. We maintain our quarterly hedging strategy and calculate hedged returns over k-periods as the product of quarterly returns: r˜H,t k=[i=0k1(1+r˜H,t+i 1)]1.

Table 13 presents the ratio of the variance of unhedged returns to the variance of hedged returns at investment horizons ranging from one quarter to 5 years. We do not provide p-values for the F-Stats because of the autocorrelation due to overlapping returns.

Table 13.

Variance Ratios of Unhedged and Hedged Returns over Different Horizons 1/

article image
Source: Author’s estimates.

The ratio of the variance of unhedged and fully hedged returns. Variances are calculated over rolling return intervals ranging from one quarter to five years. Hedged returns are based on rolling quarterly hedges of beginning-of-period balances.