Bad beta means market exposure that is expensive to hedge. Good beta is market exposure that is cheap to hedge. Distinguishing between these is crucial for FX trading strategies. The market sensitivity of FX positions can be decomposed into a risk premium beta (‘bad beta’) and a real rate beta (‘good beta’). FX positions with risk premium betas are associated with a positive price of risk that increases in crisis periods. FX positions with real rate beta are hedges, whose value increases in crisis times. Many conventional currency trading strategies carry either excessive ‘bad beta’ or too little ‘good beta’ and, thus, fail to produce true investor value.

The below summary is based on:
I-Hsuan Ethan Chiang and Xi Nancy Mo, “A ‘Bad Beta, Good Beta’ Anatomy of Currency Risk Premiums and Trading.”

Emphasis and cursive text have been added.

Good and bad betas

“Good betas [are] betas associated with relatively low price of risk and bad betas [are] betas associated with relatively high price of risk.”

“Assets with positive risk premium betas are risky, because their returns are lower when domestic investors become more risk averse. Hence…risk-premium beta is associated with counter-cyclical and unconditionally positive price of risk. On the other hand, assets with positive real-rate betas are hedges because their returns are higher when the domestic inflation is high or when the domestic nominal interest rate is low…Low interest rates are associated with a bad state of economy in the U.S… Investors with preference for counter-cyclical risk premiums should be willing to pay more to hold the hedge in bad times…Hence… real-rate beta is associated with pro-cyclical and unconditionally negative price of risk…In relative sense, risk-premium beta is ‘bad beta,’ while real-rate beta is ‘good beta’.”

The two FX betas

“This paper tests a new two-beta intertemporal capital asset pricing model (ICAPM) for currencies, and uses it to analyze currency trading strategies.”

“[Academic] literature in currency risk premiums has identified the predominant common factor in currency returns being the ‘dollar factor’, an equally-weighted portfolio of floating exchange rate currencies…While the dollar factor resembles the market portfolio in capital asset pricing models, the relation between currency returns and currency betas with respect to the dollar factor, called ‘dollar betas, is too flat to explain the cross-sectional variation in expected returns.”

“We decompose the conventional dollar beta into two betas: the beta with respect to currency market risk premium news (‘risk-premium beta’) and the beta with respect to market-wide real interest rate news (‘real-rate beta’), motivated by the decomposition of real exchange rates into risk premium and real rate components… we reserve the term ‘news’ for the above updates of long-term expectations… As the two news components are not highly correlated, and the two new betas capture distinct components of currency returns, they can explain more cross sectional variations in currency risk premiums.”

“This framework extends the two-beta model [of equity market research], featuring stock return co-movements with equity-market discount rate news (‘discount-rate beta’) and cash flow news (‘cash-flow beta’), motivated by the decomposition of dividend-price ratio into discount-rate and cash- flow components.”

“[The figure below] plots the risk-premium news and real-rate news of the dollar factor. The risk premium news clearly peaks during the Asian financial crisis, internet bubble burst, and the global financial crisis. The real-rate news is particularly high during the early 1990s recession, and hit the trough during the global financial crisis.”

“We need both risk-premium and real-rate betas to explain the cross section of currency premiums…Higher risk-premium beta currencies are systematically riskier, and higher real-rate beta currencies are hedges.”

“With the two new currency betas, we test the two-beta intertemporal capital asset pricing model for currency returns, featuring time-varying betas and time-varying prices of beta risks…Modelling time-variation in risk measures and risk premiums is crucial in currencies, because a vast of empirical literature rejects the uncovered interest rate parity, suggesting time-variation in risk premiums…”

Empirical evidence on the two FX betas

“A simple no-arbitrage model…suggests that the cross-sectional variation in betas are due to various exposures to country-specific or global shocks, and high risk-premium-beta currencies depreciate in bad times, while high real-rate beta currencies appreciate in bad times. The model motivates time-variation in betas and prices of risks, which is driven by U.S. and global volatilities, and can be potentially explained by average forward discounts or real exchange rates.”

“Based on end-of-month observations for the period January 1986 to December 2017 for 34 countries…our empirical tests of the two-beta currency international capital asset pricing mode robustly produce evidence consistent with theoretical predictions. Risk-premium betas are associated with significantly positive unconditional price of risk of 2.52% per year, and real-rate betas have negative price of risk. Furthermore, we document even stronger and more robust conditional price of risk results. Using the counter-cyclical indicator average forward discount to represent economic conditions, we find that the price of risk-premium-beta risk is counter-cyclical: risk-premium beta becomes relatively ‘better’ in good times (i.e., lower price of risk), and ‘worse’ in bad times (i.e., higher price of risk). The price of risk-premium-beta risk is 4.32% per annum higher when the average forward discount rises by one standard deviation.
On the other hand, the price of real-rate-beta risk is pro-cyclical: investors are willing to pay more to hold a hedge during bad times, and pay less in good times… it is estimated at -0.02% per month and statistically indistinguishable from zero.”

“We then utilize our estimation results to analyze the performance of several notable recent currency trading strategies, including the dollar carry trade, high-minus-low carry trade, country-level carry trade, purchasing power parity deviation, momentum, and value strategies. We find that most currency trading strategies either have excessive ‘bad beta’ or too little ‘good beta,’ failing to deliver abnormal performance.”