[downloadable]
Summary
- Building on Mang (2014), I use the different return forecasting factors of Cochrane and Piazzesi (2005, 2008) to understand the stochastic properties of currency excess returns.
- Real-time consistent forecasts of future currency excess returns inherit the statistical properties from the recursive return forecasting spreads in that they tend to have a larger persistent component. Alternatively, return forecasting factor spreads that implicitly assume all-access to historical data treat transitory and persistent shocks appropriately. Quantitatively, these all-access return forecasting factor spreads have a large transitory component. The difference between the all-access return forecasting factor spread and its recursive counterpart, the "FX forecast difference" also has a large transitory component. I use these predictor variables, along with the commonly used interest differential, to quantitatively measure the size of the predictable component in currencies.
- I exclusively examine industrialized currencies in relation to the U.S. where departures from uncovered interest parity are most apparent. For bilateral exchange rates, empirical tests of predictable currency excess returns confirm recent research findings rejecting the null that currency excess returns are unpredictable. The statistical significance and the amount of variability explained by country-specific and global predictors vary considerably; Japan and Singapore are highly predictable whereas England and Canada have a relatively small predictable component. For a portfolio of industrialized currencies, the empirical regressions comfortably reject the hypothesis that currency excess returns are unpredictable.
- Using the same set of predictor variables, I re-examine the classical forward premium regressions that have been extensively documented in the literature. I confirm the evidence documented in the literature but modify the empirical regressions in two ways: (1) To acknowledge the global nature of financial markets, I use global variables in the empirical tests, particularly variables that carry information about the market's expectations of future interest rates; (2) To account for the large degree of speculative activity in foreign exchange markets, I construct a global portfolio of currencies to average out idiosyncratic factors.
- It turns out, these two modifications to the standard empirical test resolves the forward premium anomaly in terms of having the right sign and the right magnitude. These modifications successfully account for the forward premium anomaly at short-horizons as well as over longer horizons. I employ cross-sectional data to verify the predictions of the empirical forecasts at longer horizons and find overwhelming empirical support.
Abstract
- I analyze foreign exchange market data in relation to the literature using results from Mang (2012, 2014). For industrialized currencies, the spread in the return forecasting factor of Cochrane and Piazzesi (2005, 2008) and Mang (2012) forecasts currency excess returns and exchange rate changes. Return forecasting factor spreads can be viewed as a systematic reaction to expected future changes in interest rate differentials; market participants are anticipating and reacting to the uncertainty of the duration of these spreads. This proposed explanation not only resolves the departure from uncovered interest parity once these systematic factors are ex-post identified and accounted for at short-term frequencies, exchange rate fluctuations are consistent with its predictions at longer horizons.
JEL Classification: C2, D83, D84, E2, E37, E44, F31, F44, G14
Keywords: adaptive learning; expectations formation; forward premium anomaly; foreign exchange markets; return predictability