Global currency hedging with common risk factors

Abstract

We develop a novel method to dynamically hedge foreign exchange exposure in international equity and bond portfolios. The method exploits the time-series predictability of currency returns, which we show emerges from exploiting a forecastable component in global factor returns. The hedging strategy outperforms leading alternative approaches to currency hedging across a large set of performance metrics. Moreover, we find that exploiting currency return predictability via an independent currency portfolio delivers a high risk-adjusted return and provides superior diversification gains to global equity and bond investors relative to currency carry, value, and momentum investment strategies.

Introduction

International investment positions are vast and have experienced rapid growth over recent decades. In the United States alone, investors’ holdings of foreign debt and equity have grown at over ten times the speed of aggregate US output, with the majority of those holdings denominated in foreign currency (see Fig. 1).¹ The currency exposure arising from international investments can substantially affect the underlying assets’ overall risk-return profile once redenominated in the investors’ home currency, raising the crucial question: how should investors manage their foreign exchange (FX) exposure?

Currency concerns have a first-order impact on investors’ portfolio choices. Burger et al. (2018) and Maggiori et al. (2019) have found that US investors avoid currency exposure where possible, holding the majority of their foreign debt securities in US dollars. Indeed, the “home bias” in international investments is largely a bias toward home currency. But prices in most asset classes are denominated in local currency, forcing international investors to enact a currency management strategy. One natural approach is to eliminate all currency exposure via hedging in forward markets. Yet doing so is expensive, and it eliminates the potential gains from currency returns (Glen and Jorion, 1993) and the natural hedging benefits (against underlying asset risk) arising from currency exposure (Campbell et al., 2010), suggesting that a more comprehensive approach to managing FX exposure could prove optimal.²

The primary approach to managing FX exposure is via mean-variance optimization—a process that determines optimal currency hedge positions. The approach is theoretically appealing and naturally incorporates both risk management and speculative hedging demands. However, this approach, when applied out of sample, suffers from acute estimation error in currency returns due to the well-known difficulty in predicting exchange rates (Meese and Rogoff, 1983), leading to poor overall currency hedging performance (Gardner, Stone, 1995, Larsen Jr., Resnick, 2000).³

In this paper we devise a novel method to dynamically determine currency hedge positions that we refer to as dynamic currency factor (DCF) hedging. The approach exploits the predictability of currency returns that we find emerges from a forecastable component in global risk factors. Recent breakthroughs in international macro-finance by Lustig et al. (2011) and Verdelhan (2018) have found that the cross-section of currency returns can be explained as compensation for risk, in a linear factor model that includes two global currency risk factors: dollar and carry. The dollar factor corresponds to the average return of a basket of currencies against the US dollar, while the carry factor reflects the returns on a currency carry trade. The factors also account for a large proportion of time-series exchange rate behavior in contemporaneous regressions, and thus if factor returns are forecastable, it implies currency returns are also partially predictable. We show that exploiting a forecastable component in both global factors helps to mitigate the estimation error that typically hinders mean-variance currency hedging and delivers significant economic investment gains.

In the empirical analysis, we take the perspective of a US investor who invests in a portfolio of G10 developed economies.⁴ The investor is assumed to have a predetermined long position in either foreign equities or bonds and desires to manage the FX exposure by forming optimal hedge positions within a mean-variance framework.⁵ To construct optimal hedge positions, monthly estimates of currency returns are initially formed by estimating currencies’ exposures to dollar and carry factors (i.e., factor betas). Factor returns are then forecasted using variables that have been theoretically motivated to drive either one or both factor returns, including FX volatility (Merton, 1973, Menkhoff, Sarno, Schmeling, Schrimpf, 2012, Cenedese, Sarno, Tsiakas, 2014), the average forward discount (Lustig et al., 2014), the TED spread (Brunnermeier, Pedersen, 2009, Brunnermeier, Markus, Nagel, Pedersen, 2009), and commodity returns (Ready et al., 2017). These predicted global factor returns are combined with the estimated factor betas to form out-of-sample (OOS) forecasts of currency returns. The predicted currency returns are incorporated within a mean-variance optimizer to produce optimal, currency-specific, hedge positions.

We evaluate the performance of DCF hedging, over a 20-year OOS period, against nine leading alternative approaches ranging from naive solutions in which FX exposure is either fully hedged or never hedged through to the most sophisticated techniques that also adopt mean-variance optimization. We find DCF hedging generates systematically superior OOS performance compared to all alternative approaches across a range of statistical and economic performance measures for both international equity and bond portfolios. As a preview, Fig. 2 plots the cumulative payoff to a $1 investment in international equity and bond portfolios beginning in January 1997. When adopting DCF hedging, the $1 investment grows to over $5 by July 2017 for the global equity portfolio and to almost $4 for the global bond portfolio. These values contrast with $2 and $1.5, which a US investor would have obtained if the FX exposure in the equity or bond portfolios remained unhedged.

While the performance of DCF hedging is evaluated across a large set of performance measures, special attention is paid to the OOS Sharpe ratio and certainty-equivalent (CEQ) return that capture the utility preferences of a mean-variance investor. We find an equal-weighted global equity portfolio has a statistically higher Sharpe ratio under DCF hedging relative to all nine alternative frameworks. The improvement is over 90% relative to an unhedged portfolio and over 40% relative to a portfolio that fully hedges FX exposure, which is particularly surprising given the difficulty in generating statistically superior Sharpe ratios relative to unhedged or passively hedged strategies (Glen and Jorion, 1993). Furthermore, fully hedging FX exposure is often viewed as the optimal approach to managing global bond portfolios (Campbell et al., 2010); yet when evaluating a global bond portfolio, DCF hedging is found to deliver a Sharpe ratio that is over 40% higher than the fully hedged portfolio. The relative performance of the strategy is also impressive once evaluated using the CEQ return that accounts for investor risk aversion. In each comparison, the CEQ return is found to be statistically significantly higher under DCF hedging, and thus a mean-variance global investor would always choose to adopt the approach relative to each alternative. The same exercise is performed using GDP-weighted equity and bond portfolios and delivers qualitatively identical results. In particular, the Sharpe ratio and CEQ return are consistently higher under DCF hedging relative to each alternative framework.

The core results are confirmed in a battery of alternative statistical and economic performance metrics. The average return to the international equity and bond portfolios increases significantly under DCF hedging but without negatively impacting portfolio variance or skewness. In fact, the DCF-hedged equity portfolio has the least negative skewness and one of the lowest maximum drawdowns, indicating that the superior performance of DCF hedging is not due to increased “crash” risk. The strategy thus generates the strongest performance across measures designed to penalize negatively skewed return distributions, including the Sortino ratio (Sortino and van der Meer, 1991) and the manipulation-proof “theta” measure (Ingersoll et al., 2007). DCF hedging also provides the largest information ratio, highlighting the persistence of the outperformance over time. Moreover, the performance fee a mean-variance investor would pay to switch to DCF hedging is found to be economically large: the investor would pay 3% per annum, for example, to switch to DCF hedging from portfolios that are unhedged against FX exposure.

We extend the analysis in various ways. First, we explore if refinements to the measurement of covariance terms can further enhance DCF hedging but find that even perfect foresight of the following month’s returns—and hence employing the actual realized covariance structure—has little effect on the overall investment performance. Second, we take the perspective of investors situated in all other G10 economies and show that DCF hedging has broad applicability, delivering either the highest or second highest Sharpe ratio and CEQ return in 95% of tests performed. Third, we explore the source of the gains arising from DCF hedging and find it stems almost entirely from timing exchange rate movements and capturing the predictable component of factor returns rather than from hedging low yielding currencies or incorporating time-varying factor betas. Fourth, we present a series of additional tests that highlight the robustness of DCF currency hedging to different OOS periods, alternative expanding and rolling windows, excluding crises periods, and incorporating higher transaction costs.

Finally, we examine an alternative approach to managing FX exposure that exploits the currency return predictions developed in the baseline DCF approach. The method involves the construction of an independent DCF currency trading portfolio that is subsequently combined with fully hedged global equity and bond portfolios to generate diversification gains. This “separate” currency investment approach is often unavailable to fund managers, who are usually mandated to hedge only existing FX exposure, but for managers with broader mandates, the option has the advantage of placing less constraint on currency positions and has been found to deliver strong investment gains in earlier studies (Asness, Moskowitz, Pedersen, 2013, Kroencke, Schindler, Schrimpf, 2014, Barroso, Santa-Clara, 2015). The DCF portfolio goes long in currencies with positive expected returns and short in currencies with negative expected returns. We find the strategy returns are positively skewed and deliver a high Sharpe ratio of 0.78. Moreover, the DCF strategy generates returns largely uncorrelated with existing currency strategies and offers diversification benefits to international investors beyond those provided by currency carry, value, and momentum strategies.

In sum, we contribute to the literature by developing a novel method for hedging FX exposure. In contrast to previous evidence showing mean-variance currency hedging fails OOS due to estimation error in currency returns, we show that exploiting a forecastable component in global factor returns provides a means for forming superior estimates of currency returns, which substantially improves the performance of currency hedging in international asset portfolios. The method provides a high benchmark when assessing the performance of currency hedging strategies and raises the prospect that further breakthroughs in our understanding of the factors driving currency returns could generate even larger investment gains from currency management. In addition, for managers with a broad mandate to invest in a separate currency portfolio, we propose a new multi-currency investment strategy. The strategy generates high risk-adjusted returns and offers substantial diversification benefits to global equity and bond investors.

The remainder of the paper is organized as follows: Section 2 outlines the related literature. Section 3 explains DCF hedging and the alternative currency hedging frameworks. Section 4 describes the data. Section 5 presents the core results on the performance of DCF hedging for international equity and bond portfolios. Section 6 shows results on various extensions to DCF hedging. Section 7 presents results on an independent DCF trading strategy. Section 8 reports results on various additional analyses. Section 9 concludes. An Internet Appendix can be found on the JFE web page.