Predicting exchange rates

Robert Czech, Pasquale Della Corte, Shiyang Huang and Tianyu Wang

Can investors predict future foreign exchange (FX) rates? Many economists would say that this is an incredibly difficult task, given the weak link between exchange rate fluctuations and the state of an economy – a phenomenon also known as the ‘exchange rate disconnect puzzle’. In a recent paper, we show that some investors in the ‘FX option market’ are indeed able to accurately forecast exchange rate returns, particularly in periods with strong demand for the US dollar. These informed trades primarily take place on days with macroeconomic announcements and in options with higher embedded leverage. We also find that two groups of investors – hedge funds and real money investors – have superior skills in predicting exchange rates.


But let’s take a step back. According to the Efficient Markets Hypothesis (EMH), it should be impossible to predict future returns with past market information (for example, trading volumes and past returns). However, if markets are inefficient, then informed investors are at times able to predict future returns due to their superior skills in collecting and processing trade-relevant information. In doing so, these investors incorporate information into prices and hence accelerate the price discovery process.

Previously, due to a lack of granular trading data, it remained unclear whether and how FX option investors contribute to the price discovery process in the currency market. In other words, it’s uncertain whether investors trading in the FX option market possess value-relevant information on future exchange rate fluctuations. This is despite the fact that the FX option market is one of the world’s largest and most liquid derivative markets, with an average daily volume that exceeds $250 billion and an outstanding notional close to $12 trillion.

Our data and methodology

To fill this important gap, we use the EMIR Trade Repository Data to obtain trade-level information on European-style FX options, which are mainly traded over-the-counter. Our data cover the period from November 2014 to December 2016, and we observe all trades submitted to the DTCC Derivatives Repository – the largest trade repository in terms of market share at the time – in which at least one of the counterparties is a UK-regulated entity. Consistent with London’s role as the largest trading hub for FX instruments, our data cover 42% of the global trading activity in terms of average daily volume.

We obtain option data on twenty different currencies against the dollar. Taking a closer look at the different currency pairs, we find that the lion’s share of trading volume is concentrated in options on the euro (36%), yen (25.4%), and pound sterling (7.6%) against the dollar (see Figure 1). On the sectoral level, we discover that interdealer trades account for more than three quarters of the total trading volume, while 23% of the volume can be attributed to dealer-client trades (eg a dealer trading with a hedge fund). Using a subset of our data with more granular reporting on trading directions, we also find that the volume of put options (expecting a dollar appreciation) is almost twice as high as the volume of call options (expecting an appreciation of the foreign currency). To clarify, we conveniently call all non-dollar currencies ‘foreign’, and we use the traditional approach of defining exchange rates as units of dollars per unit of foreign currency.

Figure 1: FX option volume – currency pairs

Note: The data are collected from the DTCC Derivatives Repository and our sample covers the period between November 2014 and December 2016.

Having introduced our data, we now turn towards our core analysis. The main hypothesis we put forward is that higher trading volumes in FX options today predict a foreign currency depreciation (ie a dollar appreciation) tomorrow. Our intuition is as follows: investors typically seek a positive exposure to the dollar due to liquidity and safety reasons. Informed investors may then implement their views in the option market based on certain trading signals, which, for example, could be based on their superior analysis of currency fundamentals. Importantly, when informed traders receive a positive trading signal for the dollar (or, equivalently, a negative signal for the foreign currency), they further increase their exposure to the US dollar by buying put options or selling call options. Similarly, when investors obtain a negative signal for the dollar, they decrease their exposure to the dollar – but they avoid to offset their positive dollar exposures entirely due to the dollar’s safe-haven characteristics. Put differently, FX option volume reflects more positive than negative signals for the dollar (ie more negative than positive signals for the foreign currency).

We use a portfolio sorting approach to test this hypothesis. More precisely, we construct a strategy that buys currencies with low option volume and sells currencies with high option volume. To do so, we first calculate the given currency’s volume across all options on each trading day. Next, we sort currencies into four buckets based on their FX option trading volume, and then construct equal-weighted portfolios of the currencies within each bucket. The portfolios are rebalanced on a daily basis. We then test whether the group of currencies with low option volume provides higher exchange rate returns than the group with high option volume on the following trading day.

We also use this portfolio sorting approach – as well as ordinary panel regressions – to run a battery of additional tests to confirm our informed trading hypothesis. For example, we test whether the effect is more pronounced for trades of more sophisticated investors, around macro announcements, or when using options with higher embedded leverage. Importantly, we conduct our analyses separately for all twenty currencies in our sample, as well as for a restricted group of the seven major currencies against the dollar (AUD, CAD, CHF, EUR, GBP, JPY and NZD).

What we find

We find strong evidence that FX option volume negatively predicts future exchange rate returns, especially for the seven major currency pairs. In other words, higher option volume observed today indeed predicts a non-dollar currency depreciation (ie a US dollar appreciation) tomorrow. Specifically, our strategy that buys major currencies with low option volume and sells major currencies with high option volume delivers a return of more than 14% per year, with an annualized Sharpe ratio of 1.69. Importantly, the effect is largely unrelated to existing currency strategies and robust to controlling for interest rate differentials, currency volatility and liquidity.

Consistent with the existence of informed trading in FX options, we further show that clients’ option volume is a more powerful predictor than interdealer volume for future exchange rate fluctuations. Moreover, taking a closer look at the client sector, we find that the trading of generally better informed hedge funds and real money investors (eg asset managers, pension funds, insurers) considerably outperforms the trading of less informed clients such as corporates and non-dealer banks.

Next, we show that the exchange rate predictability is largely concentrated around US macro announcements (eg announcements on inflation or GDP). Such macro announcements provide lucrative opportunities for informed investors to capitalize on their superior skills to relate economic fundamentals to exchange rate fluctuations. We also find that the effect is stronger for options with higher embedded leverage (ie short-maturity and out-of-the-money options), which offer informed investors more ‘bang for the buck’.

As a reminder, the link between option volumes and exchange rates may reflect investors’ demand for dollar assets, driven by liquidity and safety concerns. Importantly, this link should be more pronounced when investors’ initial demand for dollars is higher. To test this, we identify periods with high dollar demand using two different proxies: the US Treasury premium (the yield gap between US government bonds and currency-hedged foreign government bonds) and the VXY index (a measure of the expected volatility of FX rates). Consistent with our main hypothesis, we indeed find that the effect is stronger during periods with high demand for dollars. Last but not least, we also show that our results remain robust when using public data from Bloomberg on aggregate FX option volumes for an extended sample period (March 2013–December 2020).

Implications for policymakers

Our findings have important implications. Hedge funds and real money investors both appear to have a significant advantage in collecting and processing trade-relevant information in the FX market, which enables them to predict future exchange rate fluctuations. In doing so, both groups incorporate information into major exchange rates and ‘pull’ prices towards fundamentals. Therefore, these informed traders help to expedite the price discovery process in this important financial market.

From a policy perspective, our methodology could be employed as an early warning indicator for exchange rate fluctuations, with potentially important implications for central bank swap lines. More precisely, monitoring FX option volumes would enable policymakers to anticipate periods of significant volatility in their domestic exchange rate, which could be particularly useful when trying to predict dollar demand spikes in crisis periods. The analysis of FX option volumes would therefore not only enhance our understanding of the price discovery process in FX markets, but could also help policymakers to identify if and when investors may need to draw on central bank swap lines.

Robert Czech works in the Bank’s Research Hub, Pasquale Della Corte works for Imperial College and CEPR, Shiyang Huang works for Hong Kong University and Tianyu Wang works for Tsinghua University.

If you want to get in touch, please email us at or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

2 thoughts on “Predicting exchange rates

  1. Hi, thanks for the work.
    I’d like to download data you mentioned in this paper but i can’t find it at the link you provided, you have another link to share? Thanks

  2. Dear sirs,
    Good morning.
    Thank you very much for providing your suggestive research paper.
    Reading it, I am considering why this strategy delivered a return of more than 14% per year.
    My understanding at this moment is as follows.

    (1) In the period from November 2014 to December 2016, same as recent market circumstances, USD had appreciated against other major currencies. Under the circumstance, it is feasible that corporates, such as manufacturers and trade companies, wanted to buy more USD-call options (other currencies put options) than USD-put options (other currencies call options).
    So, it is understandable that the volume of USD-call options in the period became almost twice as high as the volume of USD-put options in the period.
    (2) As a feasible pre-condition, I believe that most of the FX options are sold by financial institutions and bought by corporates. Financial institutions must adjust delta positions of their sold options continuously, especially at the timing of hectic rate movements. On the contrary, I believe that almost all corporates don’t adjust the delta positions of their bought options continuously.
    (3) In addition, it may be feasible that corporates buy their USD-call options at out of the money (OTM) striking prices, because their purposes are buying insurances from their concerns of more USD appreciations, and options at the money (ATM) and in the money (ITM) are expensive. If macroeconomic announcement timings are not far from option contract dates, we may assume that high percentages of USD-call options remain OTMs at the announcement timing.
    (4) It may be a feasible assumption that FX rates are moved mainly by delta hedges of options in hectic rate movements, such as timings immediately after the US macroeconomic announcement.
    (5) Before the macroeconomic announcement, I assume that one dealer buys currencies with low option volume (currency-A) and sells currencies with high options volume (currency-B).
    (6) If the FX market has an enormous net USD-call option position, financial institutions are obliged to adjust their delta positions as option sellers, in the hectic rate movement after the announcement.
    (7) After the announcement, in case the USD is appreciated against other major currencies, financial institutions must buy USD (must sell other currency). They must sell much more amount of currency-B than the amount of currency-A because currency-B’s option position is bigger than currency-A’s one. As a result, currency-B will be depreciated more than currency-A, and the dealer will be able to gain profit by the position of buying currency-A and selling currency-B.
    (8) In case the USD is depreciated against other major currencies, they must sell USD (must buy other currency). In this case, they must buy much more amount of currency-B than the amount of currency-A, and currency-B will be appreciated more than currency-A, and the dealer will lose money by the position of buying currency-A and buying selling currency-B.
    (9) However, on average, FX rate movements of USD depreciation would be much calmer than the case of USD appreciation, taking the assumption (3) above.
    The FX market has a net USD-call (OTM) option position. Option seller’s USD trading amount for adjusting delta hedging would be much smaller in case of USD depreciation.
    (10) In conclusion from (1) to (9), if the USD is appreciated after the announcement, the dealer would be able to gain profit. If USD is depreciated, the dealer would lose but the loss would be smaller than the profit. If this is true, even if we assume that the probability of USD-appreciation/USD-depreciation is 50%-50% in accordance with the Efficient Markets Hypothesis (EMH), the dealers would be gainers, by trying many announcement events.
    (11) The strategy in (5) above may be applied only to market circumstances where the volume of USD-call options is much larger than the volume of USD-put options.
    In market circumstances where the USD has depreciated against other currencies, it is feasible that the volume of USD-put options is much larger than the volume of USD- call options. In these circumstances, the dealer will adopt the strategy of selling currencies with low option volume (currency-A) and buying currencies with high options volume (currency-B), for gaining profit.

    By the above consideration, it seems to me that the same conclusion could be reached by applying the Efficient Markets Hypothesis (EMH).
    Best regards,
    Takuya Matsuo

Comments are closed.