Beating Popular FX Trading Strategies With Bond Flows

Summary

• A new BIS working paper explores the extent to which international bond flows predict exchange rate returns on EM currencies.
• It finds a long-short portfolio, which goes long in currencies with the largest weekly inflows and short in those with the largest outflows, delivers large excess returns.
• The strategy outperforms more traditional FX strategies, such as carry, momentum and valuation, and can be explained by a ‘flow risk premium’.
• This premium is a compensation for the risk that countries with large inflows today face a higher probability of significant financial tightening tomorrow.

Introduction

There are three widely known and followed approaches to capture currency returns: carry, momentum and valuation. Their popularity lies in their ability to predict well-known patterns in exchange rates – a risk premium, a time trend, and a long-term fair value.

Yet, many variables determine exchange rates in the short-, medium- and long-run, leaving the door open to several different strategies. One such determinant could be international bond flows. A new BIS working paper explores the extent to which these flows can affect EM exchange rate movements. It turns out, international bond flows have strong predictive power over currency returns in emerging market economies (EMEs).

Specifically, currencies that experience large bond outflows in the preceding two weeks of portfolio formation, on average, depreciate in the subsequent week. Meanwhile, those that experience inflows appreciate. A strategy that goes long in the portfolio with the largest inflows and short in the one with the largest outflows generates annualised excess returns of around 9% – beating the three most popular FX trading strategies.

Why Do International Bond Flows Matter?

Theoretically, bond markets can influence exchange rates in many ways. Beyond interest rate differentials, which are a key medium-term determinant of currency values, bond markets provide real-time insights into financial and macroeconomic conditions. They can be a predictor of growth, currency crises, monetary policy and much more. This is particularly evident in advanced economies, where bond markets are deeper.

However, EME sovereign bond markets, especially those in the local currency segment, have grown significantly in recent decades. Accordingly, so have international bond portfolio flows. The AUM of global funds investing in EME bond markets increased from $11bn in 2004 to $383bn in 2020. The sheer size of these markets now makes them potentially important in EME exchange rate dynamics.

By looking at international bond flows, the authors can hone in on any potential links between bond prices and exchange rate returns, on the one hand, and portfolio flows and macroeconomic fundamentals on the other. If these flows matter for currency valuations, portfolios formed on the basis of these bond flows should yield economically significant returns.

Data

The authors use bond country flows data reported by Emerging Portfolio Fund Research (EPFR)^[1]. The EPFR combines individual fund flow data and fund allocation data to put together country flow statistics that track the aggregate flows in and out of countries.

The data is reported each Wednesday and covers the total flows that occurred during the week before. To reduce the volatility of weekly flow data, they compute aggregate flows over the past two weeks.

In total, the authors use flow data on 24 EMEs (includes China) and 10 advanced economies (AEs) spanning 14 years, from 2006 to 2019. They match this with exchange rate data, consisting of bilateral spot rates against the US dollar, sourced from Bloomberg, taken end-of-day each Wednesday. They also collect forward exchange rates to calculate excess currency returns.

Portfolio Formation and Trading Rules

The paper uses a portfolio approach. In the literature, these approaches are used to analyse the returns to currency portfolios formed by sorting on characteristics that may have predictive power for currency returns, such as lagged returns, carry or currency order flow. If the returns are significant, the characteristics are relevant predictors.

In the paper’s case, currency portfolios are formed based on lagged bond flows. They call them flow-sorted portfolios because currencies are sorted into four portfolios based on the size of their bond flows. To be sorted, bond flows over the past two weeks need to be ‘large’. That is, one standard deviation greater than the average over the past two years.

Then, once a currency has experienced ‘large’ flows, they are put into one of four portfolios based on whether these flows were inflows or outflows. Portfolio one contains the currencies with the largest bond outflows. Portfolio four contains the currencies with the largest bond inflows.

The excess returns of these portfolios are then calculated over the subsequent week, assuming equal investment into each of the currencies in the respective portfolios. The returns of these portfolios correspond to going long in the currencies included in each portfolio and going short the US dollar. The authors also consider a long-short portfolio, where they go long the largest inflow portfolio and short the largest outflow portfolio – this is a dollar-neutral portfolio that they call ‘FLOW’.

Results

The largest positive excess returns of going long are for the fourth portfolio (Chart 1). That is the one that contains currencies with the largest bond inflows over the past two weeks. Consistent with this, using a long-short strategy yields even higher annualised returns – 8.9% with a SR of 1.08. The cumulative returns of the FLOW portfolio are 60% over 14 years.

Why do currencies that experience the largest bond inflows yield the largest positive returns? Well, according to the authors, it has nothing to do with interest differentials. In other words, the currency appreciation that occurred in the week after the portfolio formation could not be explained by differences in interest rates.

One variable that can explain the predictability of returns is the bond flows themselves. That is, past inflows predict future inflows, and vice versa for outflows. So, investors could buy the currencies of economies where inflows are predicted to take place in anticipation of appreciation pressures that would go along with such inflows.

However, if you considered only past flows, i.e., created a portfolio only based on where bond flows were predicted to go, you would not be able to match the annualised returns from above. That means there is an additional element driving the predictability of returns above.

A Flow Risk Premium

The missing element that explains the predictability of currency returns is, according to the authors, related to a flow risk premium. That is, when comparing the results against other common risk premia, such carry, volatility, sovereign risk and liquidity, the premium that explains the returns most is a ‘flow factor’.

This flow factor is, based on the paper’s finding, a compensation for the risk that countries experiencing large portfolio inflows today could be facing a future tightening of their aggregate financial conditions.

How do they know this? The authors take the mean difference in financial conditions indices (FCI) between portfolio four countries (largest inflows) and portfolio one countries (largest outflows) and see how this difference changes over time (Chart 2). On average, the difference increases, meaning outflow countries are typically much more likely to experience significantly tighter financial conditions than outflow countries in the subsequent two years.

Beating Benchmark Trading Strategies

The authors compare the performance of their flow-sorted portfolios against our three most popular FX strategies – carry, momentum and value.

The carry portfolios are based on currencies sorted by the interest rate differential at the time of portfolio formation. Momentum sorts currencies by past mean excess returns in the month or year prior to portfolio formation. The value portfolio sorts currencies by the negative of the past 1-year exchange rate return minus the difference in foreign CPI inflation relative to that in the US over the same period.

The results are interesting. Correlations between the flow-based strategy and the alternatives tend to be relatively high, ranging from 0.62 to 0.91. But, in the case of the long-short portfolio (FLOW portfolio), the correlations are much smaller, suggesting potential diversification benefits from combining FLOW with the others.

The FLOW portfolio has the highest mean excess return at 8.9% and Sharpe ratio (1.08) among all portfolios. This is followed by the carry portfolio with a mean excess return of 6.8% and a Sharpe ratio of 0.93. The FLOW portfolio also has the highest Sortino ratio and the lowest maximum drawdown (-11.5% vs between -16.2% and -65.5%).

Bottom Line

Flows and positioning data is often very valuable for short-term trading strategies. Recently, we noted how a machine learning algorithm picked ‘mutual fund flows’ as the best predictor of fund performance, beating 59 other stock- and fund-level characteristics. Consequently, that bond flows have strong predictive power over short-term EM exchange rate movements is unsurprising.

Moreover, in the potentially daunting world of EM FX, the paper’s promising results suggest augmenting traditional strategies with a flow-based measure may be both fruitful and have some diversification benefits for investors.

[1] EPFR tracks flows for over 135,000 individual investment funds domiciled globally, with more than $48 trillion in total assets. EPFR also gathers data on fund manager allocations which provide country and industry weightings along with funds’ equity and bond holdings. This data is all sourced directly from fund managers or administrators.