Today we are fortunate to present a guest post written by Michal Rubaszek (SGH Warsaw School of Economics), Joscha Beckmann (FernUniversität Hagen and Kiel Institute for the World Economy) Michele Ca’ Zorzi (ECB), and Marek Kwas (SGH Warsaw School of Economics). The views expressed in this paper are those of the authors and not necessarily those of the institutions they are affiliated with.
We have released a new ECB Working Paper (No. 2731) entitled “Boosting carry with equilibrium exchange rate estimates”. The title suggests that it may be possible to boost the performance of FX trading strategies departing from the assumption that exchange rates move randomly.
It is a widely held view in economics that predicting future movements of exchange rates is almost impossible (Rossi, 2013). Two pieces of evidence are often put forward to support this thesis. The first comes from the (time series) FX literature on currency forecasting, arguing that it is preferable to assume that exchange rates follow a “random walk” and make a prediction of no change than forecasting exchange rates using a macro model. The second, which prevails in the FX literature on currency portfolios, points to the success of an investment strategy known as “carry trade”, consisting in borrowing in low-yield and investing in high-yield currencies (Lustig et al., 2011; Koijen et al., 2018). Profitability of carry trades contradicts the uncovered interest rate parity and implicitly assumes that exchange rates behave as random walks. For several decades there has been a parallel quest in these two strands of the FX literature, the first aiming at forecast accuracy and the second at portfolio profitability, to outperform the respective benchmarks, the random walk, and the carry trade strategy. In both cases the approach is the same in spirit, i.e. to set up a model that attempts to recover either theoretically or empirically a link between exchange rates and economic fundamentals (Menkhoff et al., 2017; Cheung et al., 2019; Colacito et al., 2020).
The assumption of random exchange rates is difficult to accept as it is at odds with economic theory. In Econbrowser blog “Exchange rate forecasting on a napkin” we argued that a gradual process of convergence of the exchange rate towards Purchasing Power Parity (PPP) tends to outperform the random walk in exchange rate forecasting. In the article “The reliability of equilibrium exchange rate models: A forecasting perspective” we generalized this result, showing that it also holds for equilibrium measures based on the Behavioral Equilibrium Exchange Rate (BEER) model (and not only PPP). Such predictive power contradicts the first piece of evidence in favor of the random walk hypothesis. In this blog we address instead the second piece of evidence in favor of such hypothesis, namely the success of carry trade strategies, and whether their strong past performance is evidence in favor of the random walk hypothesis.
To understand if one can exploit (time series) exchange rate predictability to design a competitive currency portfolio, we collected quarterly data for the G10 currencies from 1975 to 2020. We then calculated two sets of equilibrium exchange rate estimates using both the PPP and BEER models and verified our claim that there has been over this time horizon at least some time series predictability at the one-quarter horizon. We finally employed the tools and methods of the FX trading literature to show that investors could have built FX portfolios with competitive risk-return characteristics, exploiting evidence of exchange rate misalignments.
For that purpose, we evaluated:
- three benchmark strategies: momentum (M), value (V) and carry (C);
- strategies based on FX misalignments alone (EqER);
- strategies based on the assumption that exchange rates gradually return to their equilibria – postulating that half of the adjustment is completed in a fixed number of years, for example 3 (HL3) or ten (HL10).
The key performance statistics of the above strategies and their cumulative returns are shown respectively in Table 1 and Figure 1. The results suggests that strategies based on FX misalignments (EqER) over the horizon 1975 to 2020 would have been profitable as shown in terms of mean returns and Sharpe ratios (see rows 1 and 4 in Table 1). This is consistent with the insight that it is possible to exploit evidence of under or overvaluation given a large enough pool of currencies. However, their performance would have been inferior to that of a naïve carry trade strategy. We interpret this result as being consistent with the evidence that exchange rates initially adjust very slowly toward their equilibria – hence the component which is predictable is insufficient to outweigh the gains derived from knowing with certainty the prevailing configuration of interest rate differentials across countries.
The lesson one can draw is that it is preferable to rely on the alternative hypothesis that exchange rates adjust only gradually toward their equilibria. We show that indeed both HL strategies would have been very competitive compared to other benchmarks by exploiting simultaneously the time series predictability of exchange rates but also extracting forward premia (see rows 2 and 3 in Table 1). Among them, the HL10 strategy would have generated higher expected returns and Sharpe ratios than those of the naïve carry trade strategy, both in the case of PPP and BEER models. This result, which is robust for a wide range of half-lives, shows how departing from the random walk hypothesis can be instrumental for boosting a carry-based FX trading strategy. The key requirement is to assume a sufficiently slow adjustment process. Besides the issue of performance, HL strategies profoundly change the nature of expected returns since a significant component comes now from the modeler’s ability to extract the predictability of spot exchange rates (Table 1 and bottom panels of Figure 1).
Figure 1: FX portfolio returns
Notes: The upper panels present cumulated rate of returns for EqER-based and benchmark strategies. The bottom panels present excess return decomposition into spot rate predictability and forward premium.
The main message of this blog is not to dispute the evidence that carry trades performed well in recent decades. To the contrary, HL strategies are not dissimilar from carry trade strategies in practice. What we dispute is that the success of carry trade strategies automatically imply that exchange rates are random. The gradual adjustment of exchange rates to close existing misalignments seems instead a preferable assumption both from an economic theory and portfolio investors’ perspective.
References
Cheung, Y.-W., Chinn, M. D., Pascual, A. G., and Zhang, Y. (2019). Exchange rate prediction redux: New models, new data, new currencies. Journal of International Money and Finance, 95:332–336.
Colacito, R., Riddiough, S. J., and Sarno, L. (2020). Business cycles and currency returns. Journal of Financial Economics, 137(3):659–678.
Koijen, R. S., Moskowitz, T. J., Pedersen, L. H., and Vrugt, E. B. (2018). Carry. Journal of Financial Economics, 127(2):197–225.
Lustig, H., Roussanov, N., and Verdelhan, A. (2011). Common risk factors in currency markets. Review of Financial Studies, 24(11):3731–3777.
Menkhoff, L., Sarno, L., Schmeling, M., and Schrimpf, A. (2017). Currency value. Review of Financial Studies, 30(2):416–441.
Rossi, B. (2013). Exchange rate predictability. Journal of Economic Literature, 51(4):1063–1119.
This post written Michal Rubaszek, Joscha Beckmann, Michele Ca’ Zorzi and Marek Kwas.
