New empirical research provides guidance as to how to use real exchange rates for currency strategies. First, real exchange rates can serve as a basis for value-based strategies, but only if they are adjusted for key secular structural factors, such as productivity growth and product quality. Second, real exchange rates in conjunction with macroeconomic indicators can serve as indicators for the risk premia paid on currency positions.

Menkhoff Lukas., Lucio Sarno, Maik Schmeling and Andreas Schrimpf (2016 Forthcoming), “Currency Value”, Review of Financial Studies.

The below are excerpts from the paper. Headings and cursive text have been added.

The real exchange rate value issue

“We show that countries with a weak real exchange rate against the U.S. dollar (that is, the currency is cheap in real terms compared to the dollar) have higher currency excess returns going forward than countries with a strong real exchange rate. A natural interpretation might be that currencies with a low valuation level appreciate in the future and thus deliver high excess returns as suggested by standard purchasing power parity reasoning. Contrary to this intuition, however, we find that low valuation levels actually predict a significant relative depreciation of the currency, which is just the opposite of the intuition for currency value strategies. In fact, the high excess returns to currencies with low valuations are driven by the carry component (higher interest rates). This means that, while real exchange rate levels contain predictive information for currency excess returns in the cross section of currencies, this predictability is not consistent with the notion of currency value as the returns are generated by buying (selling) high (low) interest rate currencies rather than predicting future spot returns.”

“Some countries feature persistently high valuation levels and strong real exchange rates; currencies of such countries exhibit low risk premia and tend to experience further nominal appreciations going forward. Other countries, by contrast, have persistently low valuation levels, their nominal exchange rates tend to depreciate on average, and investors demand a positive risk premium for holding such currencies.”

Adjusting real exchange rates

“Real exchange rates that differ from unity…capture the deviation of a currency’s value from purchasing power parity…Relaxing the assumption of perfect substitutability – for example due to differences in quality of a country’s traded goods – introduces a wedge in prices such that countries producing higher-quality goods experience stronger real exchange rates… it is clear that a lower quality of tradable goods abroad will be generally accompanied by a weaker real exchange rate…[Also] the tendency of high income and productivity countries to have persistently higher price levels than low income and productivity countries…is typically explained by a Harrod-Balassa-Samuelson effect, that is, productivity differentials between tradable and non-tradable goods.”

“Our empirical analysis suggests that observable structural differences across countries can indeed account for large parts of the predictability pattern. We adjust real exchange rates across countries by considering two key variables that account for persistent differences in price levels: (a) Harrod-Balassa-Samuelson effects, and (b) the quality of a country’s exports. In this way, we remove from the real exchange rate the impact of country characteristics that can prevent goods market arbitrage from restoring purchasing power parity, thus delivering a more accurate measure of fair value. We find that the Harrod-Balassa-Samuelson effect (captured by differences in real per capita GDP) and differences in export quality across countries explain a large share of the cross-country variation in real exchange rates.”

“Regarding the export quality index we rely on data constructed by the International Monetary Fund…The basis for the index construction is an extension of the UN-NBER dataset covering bilateral trade…The quality index of export goods is constructed bottom up from a very disaggregated level of 851 product categories.”

Accounting for such structural cross-country differences eliminates the puzzling link between real exchange rates and future currency appreciations and depreciations. In fact, using the Harrod-Balassa-Samuelson- and quality- adjusted measures of real exchange rates, we find that an undervaluation relative to fair value predicts a high excess return going forward, which is almost fully driven by an appreciation of the currency, in line with standard intuition for currency value strategies.”

Real exchange rates and risk premia

“The real exchange rate can be decomposed into [i] expected future excess returns, i.e., currency risk premia [ii] the current and future expected interest rate differential, and [iii] expected future inflation differentials…We use the decompositions…to measure the relevance of macroeconomic fundamentals (including interest rates, expected inflation, output gaps) in explaining the predictive power of the real exchange rate for the cross-section of currency (excess) returns.”

“Our evidence…suggests that currency risk premia can be regarded as fairly persistent and mostly driven by strong asymmetries across countries. Longer-term trends of appreciation tend to be followed by even more appreciation and signal low risk premia, while periods of high inflation tend to be followed by depreciations and signal high risk premia.”

“We find that output gaps account for the lion’s share of risk premium predictability whereas inflation and interest rate differentials matter most for exchange rate predictability by the real exchange rate. This finding squares well with standard theory and earlier results where measures of the business cycle (output gaps) matter for time-variation in risk premia and where nominal fundamentals (interest rates, inflation) matter for exchange rates. We thus identify a link between macro fundamentals and currency risk premia in the cross-section of countries which operates via the connection between macro fundamentals and real exchange rates.”

Notes on the empirical research

“Our exchange rate data are taken from the Global Financial Database and cover a long sample period from 1970Q1 to 2014Q1 at a quarterly frequency. Exchange rates are end-of-quarter values. The empirical analysis is based on exchange rate changes and returns starting in the first quarter of 1976, i.e. shortly after the fall of Bretton Woods…We collect data on a total of 23 developed and emerging countries. Since all exchange rates are quoted against USD, we have a total of 22 exchange rates.”

“Our real exchange rate measure is based on real exchange rates normalized to unity in 1970Q1 for all countries. We can proceed this way as we have a balanced sample of consumer price inflation indices (CPI)…and exchange rates spanning the entire sample period for all countries.”

“The excess return measures the return accruing to a U.S. investor who borrows at the US interest rate and uses the funds to hold a position in foreign currency for one quarter, earning the foreign interest rate and then converting the proceeds back to dollars.”

“Time-series analysis is affected by trends in exchange rates, most notably the U.S. dollar. Our portfolio procedure, by contrast, studies exchange rate predictability in dollar-neutral long-short portfolios.”