Carry strategies are core to cross-sectional currency trading; buy the high-yielding currencies and short the low-yielding currencies. This strategy works but has been subject to negative skew and underperforms during crises or periods of high volatility. For researchers, the concept of currency carry is in conflict with uncovered interest rate parity and has been a puzzle which is only deepened without good pricing models to explain the risk premia. This puzzle has only gotten murkier with a paper called "Good Carry, Bad Carry".
There are prototypical G10 carry trade currencies (AUD, CHF, and JPY) with JPY and CHF usually used as funding currencies with low rates and AUD often serving as a high yield currency. For the longest time, the sort of carry currencies would have showed same exposures for decades. A carry portfolio will always hold the rate differential extremes.
However, it was found that if you exclude these three currencies from a G10 currency carry sort, you will get portfolios with higher Sharpe ratios and lower skew. Limiting the set currencies to exclude specific currencies will actually improve your portfolio efficiency. This is not supposed to be how a cross-sectional carry sort should work. Placing restrictions on currencies included in a portfolio will create good and bad currency mixes with the bad portfolios having lower Sharpe ratios and negative skew.
The authors tested a number of variations for good and bad and get the same results. Hypotheses based on current carry thinking were tested for this anomaly but were not able to draw any definitive conclusions. Good carry portfolios reduce tail risk and have better return to risk. Investors need to think beyond simple sorts or look at risk through different criteria.