A global historical analysis of FX carry trades shows positive long-term performance but a negative skew of returns. Large drawdowns have been associated with global financial stress. This supports the view that FX carry returns are to some extent a premium for undiversifiable risk. FX carry trade crashes have been diverse in duration and size, exceeding 2 years and 30% in extreme episodes. Historically, high carry and positive valuation metrics have shortened the duration of sell-offs in FX carry portfolios. Financial stress in the developed world has prolonged the duration of drawdowns of developed market FX carry trades.
The post ties in with the lecture on the basics of setback risks in trading strategies on this site.
The below are excerpts from the paper. Emphasis and cursive text have been added.
Measuring “the” FX carry trade
“A carry trade is a speculative strategy of buying currencies of countries with high interest rates funded with the sale of currencies of countries with low interest rates.”
“The developed market [currencies considered for FX carry trades] consists of Australia (AUD), Canada (CAD), euro area (EUR), Japan (JPY), New Zealand (NZD), Norway (NOK), Sweden (SEK), Switzerland (CHF) and the United Kingdom (GBP). Prior to the launch of the EUR in 1999, the legacy national currencies were used in determining carry portfolios. The emerging market universe consists of Brazil (BRL), Chile (CLP), Colombia (COP), Czech Republic (CZK), Hungary (HUF), India (INR), Indonesia (IDR), Malaysia (MYR), Mexico (MXN), Peru (PEN), Philippines (PHP), Poland (PLN), Romania (RON), Russia (RUB), Singapore (SGD), South Africa (ZAR), South Korea (KRW), Taiwan (TWD), Thailand (THB) and Turkey (TRY). The DM data sample covers the period of December 1983 – August 2013 and the EM sample is the period February 1997 – August 2013.”
“Portfolios are constructed by ordering all currencies in the investment universe based upon interest rates. At each time currencies are ranked from high to low based upon interest differentials versus the US dollar [separately for EM currencies and DM currencies and a joint global portfolio of currencies]…In all cases, the strategy is…taking long positions in the three highest interest rate currencies funded by short positions in the three lowest interest rate currencies. Overall…the carry trade was a good performer, with only a few significant setbacks until the time of the financial crisis when the drawdowns became larger and more frequent.”
“A carry crash occurs in a de-risking or de-leveraging event… Over the sample period studied, that generally means that positions in relatively high-interest rate currencies like the Australian and New Zealand dollar are sold to close out long positions, while relatively low-interest rate currencies like Japanese yen and Swiss franc are bought to close out short positions.”
The history of carry crashes
“Recent research on the topic has pointed to carry trade returns representing a risk premium… While carry trades have offered positive returns over the long-run, it is well known that they are subject to tail risk of large drawdowns. The figure below provides histogram and summary statistics for daily carry portfolio returns.”
“Digging deeper into the left tail of carry returns, in this section we examine the worst drawdowns in recent experience. Drawdowns are calculated using the following methodology: a peak is determined as the highest point up to date in the cumulative return series. A trough is the lowest point following a peak before a new peak is established. The magnitude of the drawdown is the cumulative fall from peak to trough.”
“[The table below] lists the top 10 drawdowns for the different portfolios of currencies. For each episode, the table lists the magnitude, length, and beginning and ending dates of the drawdowns. DM currencies experienced top-10 drawdowns ranging from 31.99 percent and 399 days to 7.15 percent and 30 days. The worst drawdown began in late-July 2007 and ended in early February 2009…The worst drawdown of 16.90 percent for EM currencies began mid-February 1998 and lasted 92 days, until mid-June. This drawdown was associated with Indonesia and the lingering effects of the Asian financial crisis.”
Explaining and predicting the duration of sell-offs
“Duration analysis… is important to understand in order to assess whether to cut positions once the drawdown has begun… The duration of carry drawdowns varies considerably, from 1 to 689 days for DM, and from 1 to 623 days for EM.”
“A hazard function measures the probability that a duration event lasting until time t ends in the next short interval of time following t. A…downward sloping hazard function is said to have negative duration dependence, where the likelihood of the event ending right after t, conditional upon duration lasting up to t is decreasing in t [as is the case for DM and EM carry drawdowns].”
“We introduce three variables hypothesized to be related to the duration of carry drawdowns.
- A Financial Stress Index (FSI) created by Melvin and Taylor (2009)…measures…financial conditions…The input variables are  the 1 year rolling beta of bank stocks… the average TED spread [the difference between the 3-month interbank rate and 3-month treasury yield]…in the G10 developed market universe… the inverted slope of the yield curve. [between three months and ten years]… the average monthly equity return over the G10 universe… he 780 day half-life exponentially weighted moving standard deviation…across the G10 equity market indexes… the 780 day half-life exponentially weighted moving standard deviation…across the G10 currencies… the spread of… the BofA Merrill Lynch US Corporate A yield…above government bonds… The hypothesis is that the greater the stress in financial markets, the longer the duration of the drawdown event.
- A carry opportunity variable, measured by the average of the interest rates of the long currencies in the portfolio minus the average of the interest rates on the short side of the carry portfolio… The greater the interest differentials between currencies, the greater the opportunity for carry profits, other things equal… The hypothesis is that the greater the interest differentials at the start of the drawdown, the greater carry positioning is likely to be so that there is more incentive to cut positions quickly (don’t be last out the door).
- A measure of spot exchange rate valuation… The value signal is measured by deviations of spot exchange rates from IMF annual purchasing power parities (PPPs) for individual countries… A positive value…means that the carry long currencies are more overvalued than the carry shorts… In the case where the long carry currencies are undervalued in a PPP sense…the valuation signal reinforces the positioning of the carry portfolios…Consistent with the crowded-trades argument of the Carry variable, the more crowded the currency positioning, the faster the exit from the carry trade.”
“We estimate models of the duration of DM and EM carry unwinds as a function of the three variables described…
- The hazard rate [probability of drawdown ending] is decreasing in the Financial Stress Index for the DM and global samples and increasing for the EM sample. So for DM events, duration is expected to be longer, the greater the degree of financial crisis, as measured by the FSI…[However] for EM, duration is expected to be shorter the greater the FSI value…since the FSI variable was created using DM financial conditions data.
- The results suggest that the hazard rate [probability of drawdown ending] is increasing in the size of the carry opportunity at the start of the drawdown for the DM and global samples but decreasing in carry for the EM sample…It is not surprising that larger interest differentials result in shorter duration of carry drawdowns as positions tend to be exited via mass sell-offs as investors race to be among the first to exit…The lower liquidity and higher trading costs associated with EM currencies does not allow as quick an exit from positions as is possible for the DM currencies.
- The greater the undervaluation of the carry-long currencies relative to the carry-short currencies…the shorter the duration of carry drawdowns…for all currency universes. The interpretation is similar to that of the carry opportunity variable in that when the value signal positioning aligns with the carry positioning, overall positioning will be more crowded and active investors in both trades will exit once losses are realized and the realization of both types of positions being unwound shortens the duration of losses on similarly positioned carry trades.”