SSRN Id2233367

Volatility Risk Premia and Exchange Rate Predictability y
Pasquale Della Corte Tarun Ramadorai Lucio Sarno
First version: December 2013 - Revised: July 2014
Acknowledgements: We are grateful to Bill Schwert (editor), an anonymous referee, Adrian Buss, John
Campbell, Bernard Dumas, Kenneth Froot, Federico Gavazzoni, Philippos Kassimatis, Lars Lochstoer, Stefan
Nagel, Andrea Vedolin, Adrien Verdelhan and participants at the Oxford-Man 2013 Conference on “Currency
Trading and Risk Premia”, the 2013 NBER Summer Institute in Asset Pricing, the 5th Global Alpha Forum
in New York, the 2013 Annual Conference on “Advances in the Analysis of Hedge Fund Strategies”at Imperial
College London, the 2013 Bloomberg FX Conference, the 2014 MAF Conference, the 2014 ISCEF Conference,
and seminar presentations at the City University of Hong Kong, the Einaudi Institute for Economics and
Finance, the Erasmus University Rotterdam, the Hong Kong Monetary Authority, the Hong Kong Science and
Technology, INSEAD, the Lancaster University Management School, the National Bank of Serbia, the National
Central University, the National Chung Hsing University, the Shanghai Advanced Institute of Finance, the
Singapore Management University, the Trinity College Dublin, the University of Edinburgh, the University of
Kent, the University of Manchester, the University of Nottingham, and the University of St. Gallen for helpful
conversations and suggestions. We also thank JP Morgan and Aslan Uddin for the currency implied volatility
data used in this study. Della Corte gratefully acknowledges the hospitality of the Hong Kong Institute for
Monetary Research at the Honk Kong Monetary Authority. Sarno acknowledges …nancial support from the
Economic and Social Research Council (No. RES-062-23-2340) and the gracious hospitality of the Cambridge
Endowment for Research in Finance (CERF) of the University of Cambridge. All errors remain ours.
y
Pasquale Della Corte is at Imperial College Business School, Imperial College London; email:
p.dellacorte@imperial.ac.uk. Tarun Ramadorai is at the Saïd Business School, Oxford-Man Institute,
University of Oxford and CEPR; email: tarun.ramadorai@sbs.ox.ac.uk. Lucio Sarno is at Cass Business
School, City University London and CEPR; email: lucio.sarno@city.ac.uk.
Electronic copy
Electronic copy available
available at:
at:https://ssrn.com/abstract=2233367
http://ssrn.com/abstract=2233367
Volatility Risk Premia and Exchange Rate Predictability
First version: December 2013 - Revised: July 2014
Abstract
We discover a new currency strategy with highly desirable return and diversi…cation prop-
erties, which uses the predictive capability of currency volatility risk premia for currency
returns. The volatility risk premium – the di¤erence between expected realized volatility
and model-free implied volatility – re‡ects the costs of insuring against currency volatility
‡uctuations, and the strategy sells high-insurance-cost currencies and buys low-insurance-
cost currencies. The returns to the strategy are mainly generated by movements in spot
exchange rates rather than interest rate di¤erentials, and the strategy carries a large weight
in a minimum-variance portfolio of commonly employed currency strategies. We explore al-
ternative explanations for the pro…tability of the strategy, which cannot be understood using
traditional risk factors.
Keywords: Exchange Rates; Volatility Risk Premium; Predictability, Minimum-Variance

Currency Portfolio.
JEL Classi…cation: F31; F37.
Electronic copy
Electronic copy available
available at:
at:https://ssrn.com/abstract=2233367
http://ssrn.com/abstract=2233367
1 Introduction
For decades, …nance practitioners and academics have struggled to understand and explain
currency ‡uctuations.1 More recently, the literature has focused on a closely-related ques-
tion, which is to document high returns to currency investment strategies such as carry and
momentum.2 Analogous to the di¢ culty of …nding de…nitive answers in the exchange rate
determination literature, there has been limited success in explaining these currency strategy
returns in terms of compensation for systematic risk.
In this paper, we discover a new currency strategy with high average returns, excellent
diversi…cation bene…ts relative to the set of previously discovered currency strategies, and
unusual properties that provide clues as to the underlying drivers of exchange rate movements.
The key to this new strategy is the signi…cant predictive power of the currency volatility risk
premium for changes in spot exchange rates.3 A useful summary statistic of the importance
of this new currency strategy (which we dub V RP ), is that over the 1998 to 2013 period, in a
cross-section of 10 (20) currencies, it has a large weight of 28% (26%) in the global minimum
variance portfolio of …ve well-known currency strategies, including carry and momentum.
The large weight of V RP in the currency strategy portfolio re‡ects its ability to generate
high returns per unit of risk, and is augmented by the desirable correlation properties of the
strategy relative to the other widely-studied currency strategies. This unusual low correlation
partly arises from the excellent performance of V RP during crises, and primarily from the fact
that the excess returns of V RP are almost completely obtained through prediction of spot
currency returns, rather than interest rate di¤erentials. This stands in sharp contrast with
the performance of the carry strategy, which has primarily been driven by interest di¤erentials
rather than spot currency returns.4
1
The di¢ culty of explaining and forecasting nominal exchange rates was documented early on by Meese
and Rogo¤ (1983). Over the past three decades, it has continued to be di¢ cult to …nd theoretically motivated
variables able to beat a random walk forecasting model for currencies (e.g. see Engel, Mark, and West, 2008).
2
See, for example, Lustig and Verdelhan (2007), Ang and Chen (2010), Burnside, Eichenbaum, Kleshchel-
ski, and Rebelo (2011), Lustig, Roussanov, and Verdelhan (2011), Menkho¤, Sarno, Schmeling, and Schrimpf
(2012a,b) and Barroso and Santa Clara (2014), who all build currency portfolios to study return predictability
and/or currency risk exposure.
3
To be clear from the outset, our strategy does not trade volatility products. We simply use the currency
volatility risk premium as conditioning information to sort currencies, build currency portfolios, and uncover
predictability in currency excess returns and changes in spot exchange rates.
4
We use interchangeably the terms spot currency returns and exchange rate returns to de…ne the change
in nominal exchange rates over time; similarly we use interchangeably the terms excess returns or portfolio
Electronic copy available at: https://ssrn.com/abstract=2233367

The currency volatility risk premium is the di¤erence between expected future realized
volatility, and a model-free measure of implied volatility derived from currency options. A
growing literature studies the variance or the volatility risk premium in di¤erent asset classes,
including equity, bond, and foreign exchange (FX) markets.5 In general, this literature has
shown that the volatility risk premium is on average negative: expected volatility is higher
than historical realized volatility, and since volatility is persistent, expected volatility is also
generally higher than future realized volatility. In other words, the volatility risk premium
represents compensation for providing volatility insurance. Therefore, the currency volatility
risk premium that we construct can be interpreted as the cost of insurance against volatility
‡uctuations in the underlying currency. When it is high – realized volatility is higher than
the option-implied volatility –insurance is relatively cheap, and vice versa.
We use the currency volatility risk premium to sort currencies into quintile portfolios at
the end of each month. The V RP strategy buys currencies with relatively cheap volatility
insurance, i.e., the highest volatility risk premium quintile, and sells short currencies with
relatively expensive volatility insurance, i.e., the lowest volatility risk premium quintile. We
track returns on this trading strategy over the subsequent period, meaning that these returns
are purely out-of-sample, conditioning only on information available at the time of portfolio
construction.
The performance of V RP stems virtually entirely from the predictability of spot exchange
rates rather than from interest rate di¤erentials. That is, currencies with relatively cheap
volatility insurance tend to appreciate and those with relatively more expensive volatility in-
surance tend to depreciate, over the subsequent month. The observed predictability of spot
exchange rates associated with V RP is far stronger than that arising from carry (which is per-
haps unsurprising given the well-documented fact that interest di¤erentials are the proximate
component of carry returns), and perhaps more importantly, stronger than that associated
with currency momentum or any of the other currency trading strategies that we consider.
As mentioned earlier, this is part of the reason for the diversi…cation bene…ts that the V RP
returns to refer to the returns from implementing a long-short currency trading strategy that buys and sells
currencies on the basis of some characteristic.
5
See, for example, Carr and Wu (2009), Eraker (2008), Bollerslev, Tauchen, and Zhou (2009), Todorov
(2010), Drechsler and Yaron (2010), Han and Zhou (2011), Mueller, Vedolin, and Yen (2011), Londono and
Zhou (2012) and Buraschi, Trojani, and Vedolin (2014).

strategy o¤ers in a currency portfolio.
The contribution of our paper is purely empirical, and we do not have a formal theoretical
model that links the volatility risk premium (or its determinants) to spot returns. However,
we do provide empirical evidence on three possible interpretations of our results. First, we
consider the possibility that returns from the V RP strategy re‡ect compensation for risk, and
test the pricing power of conventional risk factors for its returns using standard linear asset
pricing models. We …nd no evidence that V RP returns can be explained by various sets of
factors that have been used to explain time-series and cross-sectional variation in the returns
to trading strategies more generally, and currency strategies more speci…cally.
We then extend our search for risk-compensation to check whether V RP returns cap-
ture ‡uctuations in aversion to global volatility risk. We check the relationship between
V RP returns and global volatility risk in two ways –using cross-sectional asset pricing tests
of volatility risk premium-sorted portfolios on a global FX volatility risk factor, as well as by
estimating time-varying loadings of currency returns on various proxies for global volatility
risk and building portfolios sorted on these estimated loadings. Neither of these tests produces
evidence consistent with the proposed explanation. Indeed, the long-short strategy generated
from estimated loadings on the global volatility risk factor produces substantially lower aver-
age returns than V RP ; moreover, these returns are virtually uncorrelated with V RP returns.
In sum, the data appear to reject an explanation based on ‡uctuations in aversion to global
volatility risk and, more generally reject the hypothesis that V RP returns can be explained
by exposure to common risk factors.
A second explanation that we consider relies on limits to arbitrage, and its e¤ects on the
interaction between hedgers and speculators in the currency market. There is a growing
theoretical and empirical literature suggesting that such interactions are important in asset
return determination (see, for example, Acharya, Lochstoer, and Ramadorai, 2013; Adrian,
Etula, and Muir, 2013; and Gromb and Vayanos, 2010 for an excellent survey of the literature).
Such an explanation for our results would rely on time-variation in the amount of arbitrage
capital available to natural providers of currency volatility insurance (“speculators”), such as
…nancial institutions or hedge funds. It would also require that risk-averse natural “hedgers”
of currencies such as multinational …rms are more (less) willing to hedge and hold currencies
with relatively inexpensive (more expensive) volatility insurance. Such an explanation predicts

price impact in the spot market in response to purchases or sales of currencies based on their
relative cost of volatility insurance.
While we do not have a formal theoretical model of such a mechanism, we expect that when
funding liquidity is lower (i.e., times of high capital constraints on speculators), and demand
for volatility protection is higher (i.e., times of increased risk aversion of natural hedgers), we
should detect increases in the spread in the cost of volatility insurance across currencies, as
well as the spread in spot exchange rate returns across portfolios. We do …nd that increases
in the TED spread –a commonly used proxy for funding liquidity (see, for example, Garleanu
and Pedersen, 2011) –are associated with higher V RP returns. Fluctuations in risk aversion,
proxied by changes in the VIX, add signi…cant additional explanatory power when interacted
with the TED spread. We also measure capital ‡ows to currency and global macro hedge
funds, and …nd that when hedge fund ‡ows are high, signifying increased funding and thus
lower hedge fund capital constraints, the returns to V RP are lower and vice versa.6 In sum,
there is some evidence consistent with limits to arbitrage in the currency market constituting
part of the explanation for our results.
The third explanation we consider is that volatility risk premia predict FX returns because
investors trading in currency options markets are better informed about the value of the
underlying currency than those trading in the spot market. It is clear that our results cannot
be explained by a simple Pan and Poteshman (2006) style strategy based on informed traders
buying currency call (put) options in advance of expected appreciations (depreciations). This
is because any price pressure from increased demand for either call or put options (a la Bollen
and Whaley, 2004, and Garleanu, Pedersen, and Poteshman, 2009), will result in a lower
volatility risk premium –which stands in contrast to our empirical results, in which decreases
(increases) in the volatility risk premium predict appreciations (depreciations). While this
raises the bar for an information asymmetry-based explanation of our …ndings, it does not
necessarily rule out more complex information-driven options trading strategies as a possible
explanation.
6
Using CFTC data, we also …nd that commercial traders sell currencies which are more expensive to insure
and buy currencies which are cheaper to insure, with …nancial traders trading in the opposite direction. This
evidence also links our work to another stream of the exchange rate literature on forecasting currency returns
using currency order ‡ow. For example, Froot and Ramadorai (2005), Evans and Lyons (2005) and Rime,
Sarno, and Sojli (2010) show that order ‡ow has predictive power for exchange rate movements.

To further explore the statistical properties of volatility risk premia and underlying cur-
rency returns, we estimate a VAR-GARCH model, which is a useful tool to simultaneously
analyze bi-directional causality in both …rst and second moments of both variables (see, for
example, Bali and Hovakimian, 2009). We …nd evidence that the conditional variance of the
volatility risk premium predicts the conditional variance of the underlying currency in the
spot market. This is consistent with the possibility that there is non-directional information
in the currency options market which is relevant for future movements in the currency spot
market. While this of course does not allow us to conclude that our results are generated by
information asymmetries, the conditional variance prediction suggests a way to enhance the
performance of the V RP strategy. This might be done by adjusting the weights of currencies
in the V RP portfolio based on estimated conditional variance in the currency options market.
The results in this paper highlight intriguing similarities between the behaviour of equity
and currency options and their underlying asset markets. Several authors (see, for exam-
ple, Goyal and Saretto, 2009, Bali and Hovakimian, 2009, and Buss and Vilkov, 2012) show
that volatility risk premia have predictive power for the cross-section of stock returns. Bali
and Hovakimian (2009) study the equity market using a VAR-GARCH model, and …nd ev-
idence which they interpret as consistent with information spillovers from equity options to
underlying equities. While we do not draw the same conclusions, the similarity of the statis-
tical relationships between equity options and underlying stocks, and currency options and
underlying currencies suggests that there may be more general structural determinants of
this relationship that are common across these markets. We leave the exploration of these
important issues to future work.
The paper is structured as follows. Section 2 de…nes the volatility risk premium and
its measurement in currency markets. Section 3 describes our data and some descriptive
statistics. Section 4 presents our main empirical results on the volatility risk premium-sorted
strategy, while Section 5 investigates alternative mechanisms that could explain our …ndings.
Section 6 concludes. A separate Internet Appendix provides robustness tests and additional
supporting analyses.

2 Foreign Exchange Volatility Risk Premia
Volatility Swap. A volatility swap is a forward contract on the volatility “realized”on the
underlying asset over the life of the contract. The buyer of a volatility swap written at time
t, and maturing at time t + , receives the payo¤ (per unit of notional amount):
V Pt; = (RVt; SWt; ) (1)
where RVt; is the realized volatility of the underlying, SWt; is the volatility swap rate, and
both RVt; and SWt; are de…ned over the life of the contract from time t to time t + , and
quoted in annual terms. However, while the realized volatility is determined at the maturity
date t + , the swap rate is agreed at the start date t.
The value of a volatility swap contract is obtained as the expected present value of the
future payo¤ in a risk-neutral world. This implies, because V Pt; is expected to be 0 under
the risk-neutral measure, that the volatility swap rate equals the risk-neutral expectation of
the realized volatility over the life of the contract:
SWt; = EtQ [RVt; ] (2)

q R t+
where EtQ [ ] is the expectation under the risk-neutral measure Q, RVt; = 1
t
2 ds,
s
2
and s denotes the (stochastic) volatility of the underlying asset.
Volatility Swap Rate. We synthesize the volatility swap rate using the model-free
approach derived by Britten-Jones and Neuberger (2000), and further re…ned by Demeter…,
Derman, Kamal and Zou (1999), Jiang and Tian (2005), and Carr and Wu (2009).
Building on the pioneering work of Breeden and Litzenberger (1978), Britten-Jones and
Neuberger (2000) derive the model-free implied volatility entirely from no-arbitrage conditions
and without using any speci…c option pricing model. Speci…cally, they show that the risk-
neutral expected integrated return variance between the current date and a future date is fully
speci…ed by the set of prices of call options expiring on the future date, provided that the
price of the underlying evolves continuously with constant or stochastic volatility but without
jumps.
Demeter…, Derman, Kamal, and Zou (1999) show that the Britten-Jones and Neuberger
(2000) solution is equivalent to a portfolio that combines a dynamically rebalanced long po-
sition in the underlying, and a static short position in a portfolio of options and a forward

that together replicate the payo¤ of a “log contract.”7 The replicating portfolio strategy cap-
tures variance exactly, provided that the portfolio of options contains all strikes with the
appropriate weights to match the log payo¤. Jiang and Tian (2005) further demonstrate that
the model-free implied variance is valid even when the underlying price exhibits jumps, thus
relaxing the di¤usion assumptions of Britten-Jones and Neuberger (2000).
The annualized risk-neutral expectation of the return variance between two dates t and
t+ can be formally computed by integrating option prices expiring on these dates over an
in…nite range of strike prices:
Z Z !
Ft; 1
1 1
EtQ RVt;2 = Pt; (K)dK + Ct; (K)dK (3)
0 K2 Ft; K2
where Pt; (K) and Ct; (K) are the put and call prices at t with strike price K and maturity
date t + , Ft; is the forward price matching the maturity date of the options, St is the
price of the underlying, = 2 exp (it; ), and it; is the -period domestic riskless rate. The
risk-neutral expectation of the return variance in Equation (3) delivers the strike price of a
variance swap EtQ RVt;2 , and is referred to as the model-free implied variance.
Even though variance emerges naturally from a portfolio of options, it is volatility that
participants prefer to quote, as the payo¤ of a variance swap is convex in volatility and
large swings in volatility, as we observed during the recent …nancial crisis, are more likely to
cause large pro…ts and losses to counterparties. Therefore, our empirical analysis focuses on
volatility swaps, and we synthetically construct the strike price of this contract as
q
Et [RVt; ] = EtQ RVt;2
Q
(4)
and refer to it as model-free implied volatility.

While straightforward, this approach is subject to a convexity bias. The main complication
in valuing volatility swaps arises from the fact that the strike of a volatility swap is not
equal to the square root of the strike of a variance swap due to Jensen’s inequality, i.e.,
q
EtQ [RVt; ] EtQ RVt;2 . The convexity bias that arises from the above inequality leads
to imperfect replication when a volatility swap is replicated using a buy-and-hold strategy of
variance swaps (e.g., Broadie and Jain, 2008). Simply put, the payo¤ of variance swaps is
quadratic with respect to volatility, whereas the payo¤ of volatility swaps is linear.
7
The log contract is an option whose payo¤ is proportional to the log of the underlying at expiration
(Neuberger, 1994).

We deal with this bias in approximation in two ways. First, we measure the convexity
bias using a second-order Taylor expansion as in Brockhaus and Long (2000) and …nd that
it is empirically small.8 More importantly, when we re-do our empirical exercise with model-
free implied variances, we …nd virtually identical results. Hence the convexity bias has no
discernible e¤ect on our results and the approximation in Equation (4) works well in our
framework, which explains why it is widely used by practitioners (e.g., Knauf, 2003).
Computing model-free implied volatility requires the existence of a continuum in the cross-
section of option prices at time t with maturity date . In the FX market, over-the-counter
options are generally quoted in terms of Garman and Kohlhagen (1983) implied volatilities
at …xed deltas. Liquidity is generally spread across …ve levels of deltas. From these quotes,
we extract …ve strike prices corresponding to …ve plain vanilla options, and follow Jiang and
Tian (2005), who present a simple method to implement the model-free approach when option
prices are only available on a …nite number of strikes.
Speci…cally, we use a cubic spline around these …ve implied volatility points. This inter-
polation method is standard in the literature (e.g., Bates, 1991; Campa, Chang, and Reider,
1998; Jiang and Tian, 2005; Della Corte, Sarno, and Tsiakas, 2011) and has the advantage
that the implied volatility smile is smooth between the maximum and minimum available
strikes. We then compute the option values using the Garman and Kohlhagen (1983) valu-
ation formula,9 and use trapezoidal integration to solve the integral in Equation (3). This
method introduces two types of approximation errors: (i) the truncation errors arising from
observing a …nite number, rather than an in…nite set of strike prices, and (ii) a discretization
error resulting from numerical integration. Jiang and Tian (2005), however, show that both
errors are small, if not negligible, in most empirical settings.10
Volatility Risk Premium. In this paper we study the predictive information content
in volatility risk premia for future exchange rate returns. To this end, we work with the
q
V2
8
Brockhaus and Long (2000) show that EtQ [RVt; ] = EtQ RVt;2 8m3=2
where m and V 2 denote the
mean and variance of the futureqrealized variance, respectively, under the risk-neutral measure Q. EtQ [RVt; ] is
certainly less than or equal to EtQ RVt;2 due to Jensen’s inequality, and V 2 =8m3=2 measures the convexity
error.
9
This valuation formula can be thought of as the Black and Scholes (1973) formula adjusted for having
both domestic and foreign currency paying a continuous interest rate.
10
In the Internet Appendix (Table A.10), we present results for di¤erent interpolation methods (Castagna
and Mercurio, 2007) as well as a model-free approach that is robust to price jumps (Martin, 2012).

ex-ante payo¤ or “expected volatility premium” to a volatility swap contract. The volatility
risk premium can be thought of as the di¤erence between the physical and the risk-neutral
expectations of the future realized volatility.11 Formally, the -period volatility risk premium
at time t is de…ned as
V RPt; = EtP [RVt; ] EtQ [RVt; ] (5)
where EtP [ ] is the conditional expectation operator at time t under the physical measure
P. Following Bollerslev, Tauchen, and Zhou (2009), we proxy EtP [RVt; ] by simply using the
q P
252
lagged realized volatility, i.e., EtP [RVt; ] = RVt ; = 2
i=0 rt i , where rt is the daily
log return on the underlying security. This approach is widely used for forecasting exercises
– it makes V RPt; directly observable at time t, requires no modeling assumptions, and is
consistent with the stylized fact that realized volatility is a highly persistent process. Thus, at
time t, we measure the volatility risk premium over the [t; t + ] time interval as the ex-post
realized volatility over the [t ; t] interval and the ex-ante risk-neutral expectation of the
future realized volatility over the [t; t + ] interval, i.e., V RPt; = RVt ; EtQ [RVt; ].
For our purposes, we view currencies with high V RPt; as those which are relatively cheap
to insure at each point in time t, as their expected realized volatility under the physical
measure (i.e., the variable against which agents hedge) is lower than the cost of purchasing
option-based insurance – which is primarily driven by expected volatility under the risk-
neutral measure. Conversely, we consider those currencies with relatively low V RPt; as more
expensive to insure at time t.
3 Data and Currency Portfolios

This section describes the data and the construction of the currency portfolios employed in
our analysis. The data comprises spot and forward exchange rates, over-the-counter (OTC)
currency options, hedge fund ‡ows, and positions on currency futures and options.
Exchange Rate Data. We collect daily spot and one-month forward exchange rates (bid
and ask prices) vis-à-vis the US dollar (USD) from Barclays and Reuters via Datastream. We
11
Several papers de…ne the volatility risk premium as di¤erence between the risk-neutral and the physical
expectation. Here we follow Carr and Wu (2009) and take the opposite de…nition as it naturally arises from
the long-position in a volatility swap contract.

use monthly data by sampling end-of-month exchange rates from January 1998 to December
2013. In our empirical exercise, we build currency portfolios using two sets of countries. The
…rst sample comprises Australia, Canada, Denmark, Euro Area, Japan, New Zealand, Norway,
Sweden, Switzerland, and the United Kingdom. These 10 countries have the most traded
currencies and account for about 90 percent of the average daily turnover in FX markets
according to the Triennial Survey of the Bank for International Settlements (2013). We
refer to this sample as the “Developed” countries sample. The second sample adds the most
liquid emerging market currencies to the Developed country sample. Some currencies in this
expanded “Developed and Emerging”countries sample may be subject to capital controls and,
hence, not be tradable (in large amounts) in practice. To mitigate this concern, we follow
Menkho¤ et al. (2012b) and select the currencies for which the …nancial openess index of
Chinn and Ito (2006) index – a measure of a country’s degree of capital account openness
– is greater than or equal to zero. Ultimately, we only consider emerging market economies
for which the capital account is su¢ ciently unrestricted so that trading in this currency can
actually take place.12 The …nal expanded sample includes: Australia, Brazil, Canada, Czech
Republic, Denmark, Euro Area, Hungary, Japan, Mexico, New Zealand, Norway, Poland,
Singapore, South Africa, South Korea, Sweden, Switzerland, Taiwan, Turkey, and United
Kingdom.
Implied Volatility Data. We calculate the volatility swap rate described in Section 2
using implied volatility data on over-the-counter (OTC) currency options, obtained from JP
Morgan. We use monthly data by sampling end-of-month implied volatilities from January
1998 to December 2013.13
12
Precisely, we start from 10 emerging market currencies and apply recursively the capital account openess
index of Chinn and Ito (2006), available on Hiro Ito’s website. Data are available at yearly frequency until
2011, and we construct monthly observations by forward …lling, i.e., we keep end-of-period data constant until
a new observation becomes available. Note that the Chinn-Ito index is not available for Taiwan. In this case,
we rely on the capital account liberalization index of Kaminsky and Schmukler (2008), available on Graciela
Kaminsky’s website.
13
In a previous draft we found qualitatively identical results over a sample which began in January 1996 and
ended in August 2011. However, in our discussions with JP Morgan sta¤ we learned that volatility swaps only
became su¢ ciently liquid in FX markets in early 1998, in the aftermath of the Asian and Russian crises. This
is consistent with Carr and Lee (2009), who report that trading in both variance and volatility swaps was spo-
radic until 1998, when volatility trading took o¤ following the historically high implied volatilities experienced
that year. In early 1998, the International Swaps and Derivatives Association (ISDA), the Emerging Markets
Traders Association (EMTA), and the Foreign Exchange Committee (FXC) published the “1998 Foreign Ex-
10

The OTC currency option market is characterized by speci…c trading conventions. While
exchange traded options are quoted at …xed strike prices and have …xed calendar expiration
dates, currency options are quoted at …xed deltas and have constant maturities. More impor-
tantly, while the former are quoted in terms of option premia, the latter are quoted in terms
of Garman and Kohlhagen (1983) implied volatilities on baskets of plain vanilla options.
For a given maturity, quotes are typically available for …ve di¤erent combinations of plain-
vanilla options: at-the-money delta-neutral straddles, 10-delta and 25-delta risk-reversals, and
10-delta and 25-delta butter‡y spreads. The delta-neutral straddle combines a call and a put
option with the same delta but opposite sign such that the total delta is zero –this is the at-
the-money (ATM) implied volatility quoted in the FX market. In a risk reversal, the trader
buys an out-of-the money (OTM) call and sells an OTM put with symmetric deltas. The
butter‡y spread is built up by buying a strangle and selling a straddle, and is equivalent to
the di¤erence between the average implied volatility of an OTM call and an OTM put, and
the implied volatility of a straddle. From these data, one can recover the implied volatility
smile ranging from a 10-delta put to a 10-delta call.14 To convert deltas into strike prices, and
implied volatilities into option prices, we employ domestic and foreign interest rates, obtained
from Bloomberg.
This recovery exercise yields data on plain-vanilla European call and put for currency pairs
vis-à-vis the US dollar, with maturity of one year. Practitioner accounts suggest that natural
hedgers such as corporates prefer hedging using intermediate-horizon derivative contracts to
the more transactions-costs intensive strategy of rolling over short term positions in currency
options, and hence the one-year volatility swap is a logical contract maturity.
Hedge Fund Flows. To construct a measure of new arbitrage capital available to

hedge funds, we use data from a large cross-section of hedge funds and funds-of-funds from
January 1998 to December 2013, which is consolidated from data in the HFR, CISDM, TASS,
Morningstar, and Barclay-Hedge databases, and comprises of roughly US$ 1.5 trillion worth
of assets under management (AUM) towards the end of the sample period. Ramadorai (2013),
change and Currency Option De…nitions”providing the …rst comprehensive documentation of both deliverable
and nondeliverable cash-settled FX options transactions in both emerging market and major currencies.
14
In market jargon, a 10-delta call is a call whose delta is 0:10 whereas a 10-delta put is a put with a delta
equal to 0:10.
11

and Patton and Ramadorai (2013) provide a detailed description of the process followed to
consolidate these data.
We select a subset of 634 funds from these data, those self-reporting as currency funds or
global macro funds, and construct the net ‡ow of new assets to each fund as the change in
the fund’s AUM across successive months, adjusted for the returns accrued by the fund over
the month –this is tantamount to an assumption that ‡ows arrive at the end of the month,
following return accrual. We then normalize the …gures by dividing them by the lagged
AUM, and then value-weight them across funds to create a single aggregate time-series index
of capital ‡ows to currency and global macro funds.15
Positions on Currency Futures. We employ data from the Commitments of Traders

report issued by the Commodity Futures Trading Commission (CFTC). The report aggregates
the holdings of participants in the US futures and options markets (primarily based in Chicago
and New York). It is typically released every Friday and re‡ects the commitments of traders for
the prior Tuesday. The CFTC provides a breakdown of aggregate positions held by commercial
traders and …nancial (or non-commercial) traders. The former are merchants, foreign brokers,
clearing members or banks using the futures market primarily to hedge their business activities.
The latter are hedge funds, …nancial institutions and individual investors using the futures
market for speculative purposes. We collect data from January 1998 to December 2013 on
the Australian dollar, Brazilian real, British pound, Canadian dollar, Euro, Japanese yen,
Mexican peso, New Zealand dollar, and Swiss franc relative to the USD dollar.
In our empirical analysis, we construct the net demand of currency options and futures -
the di¤erence between long and short positions scaled by the total open interest - for both
commercial and …nancial traders. We then examine whether the buying and selling actions
of di¤erent players in the futures and options market follow the pattern implied by the V RP
strategy.
Other Data. We also collect monthly data on the VIX index, 3-month LIBOR and
3-month T-bill rate from Bloomberg, monthly data from the Federal Reserve Economic data
15
We measure the net ‡ow for each fund i as F lowti = AU Mti AU Mti 1 1 + rti , where AU Mti and rti
are assets under management and returns at time t, respectively. We then construct the AUM-weighted net
P P 1
‡ow scaled by the lagged AUM as F lowt = i wt 1 F lowti where wt = i AU Mt
i
. Finally, we winsorize
F lowt at the 1 and 99 percentile points each month.
12

website, and annual data for the purchasing power parity (PPP) spot rate from the OECD.
The latter are published every March, and we retrieve monthly data by forward …lling, i.e.,
we use the last available PPP rate until the next February.16
Currency Excess Returns. We de…ne spot and forward exchange rates at time t as
St and Ft , respectively. Exchange rates are de…ned as units of US dollars per unit of foreign
currency such that an increase in St indicates an appreciation of the foreign currency. The
excess return on buying a foreign currency in the forward market at time t and then selling it
in the spot market at time t + 1 is computed as RXt+1 = (St+1 Ft ) =St , which is equivalent
to the spot exchange rate return minus the forward premium RXt+1 = ((St+1 St ) =St )
((Ft St ) =St ). According to the CIP condition, the forward premium approximately equals
the interest rate di¤erential (Ft St ) =St ' it it , where it and it represent the domestic
and foreign riskless rates respectively, over the maturity of the forward contract. Since
CIP holds closely in the data at daily and lower frequency (e.g., Akram, Rime, and Sarno,
2008), the currency excess return is approximately equal to an exchange rate component
(i.e., the exchange rate change) minus an interest rate component (i.e., the interest rate
di¤erential): RXt+1 ' ((St+1 St ) =St ) (it it ). We construct currency excess returns
adjusted for transaction costs using bid-ask quotes on spot and forward rates. The net excess
l b
return for holding foreign currency for a month is computed as RXt+1 ' (St+1 Fta )=Sta ,
where a indicates the ask price, b the bid price, and l a long position in a foreign currency.
The net excess return accounts for the full round-trip transaction cost occurring when the
foreign currency is purchased at time t and sold at time t + 1. If the investor buys foreign
currency at time t but decides to maintain the position at time t + 1, the net excess return
l
is computed as RXt+1 ' (St+1 Fta )=Sta . Similarly, if the investor closes the position in
foreign currency at time t + 1 already existing at time t, the net excess return is de…ned as
l b
RXt+1 ' (St+1 Ft )=Stb . The net excess return for holding domestic currency for a month
s
is computed as RXt+1 ' (Ftb a
St+1 )=Stb , where s stands for a short position on a foreign
currency. In this case, our investor sells foreign currency at time t in the forward market at
the bid price Ftb and o¤sets the position in the spot market at time t + 1 using the ask price
a
St+1 . If the foreign currency leaves the strategy at time t and the short position is rolled
16
For Singapore and Taiwan, OECD’s PPP spot data are not available and we use data from the Penn
World Tables instead.
13

s
over at time t + 1, the net excess return is computed as RXt+1 ' (Ftb St+1 )=Stb . Similarly,
if the investor closes a short position on the foreign currency at time t + 1 already existing
s a
at time t, the net excess return is computed as RXt+1 ' (Ft St+1 )=Stb . In short, excess
returns are adjusted for the full round-trip transaction cost in the …rst and last month of our
sample period. The total number of currencies in our portfolios changes over time, and we
only include currencies for which we have bid and ask quotes on forward and spot exchange
rates in the current and subsequent period.
Carry Trade Portfolios. At the end of each period t, we allocate currencies to …ve
portfolios on the basis of their interest rate di¤erential relative to the US, (it it ) or forward
premia since (Ft St ) =St = (it it ) via CIP. This exercise implies that Portfolio 1 com-
prises 20% of all currencies with the highest interest rate di¤erential (lowest forward premia)
and Portfolio 5 comprises 20% of all currencies with the lowest interest rate di¤erential (high-
est forward premia), and we refer to the long-short portfolio formed by going long Portfolio 1
and short Portfolio 5 as CAR. We compute the excess return for each portfolio as an equally
weighted average of the currency excess returns within that portfolio, and individually track
both the interest rate di¤erential and the spot exchange rate component that make up these
excess returns.
Lustig, Roussanov, and Verdelhan (2011) study these currency portfolio returns using their
…rst two principal components. The …rst principal component implies an equally weighted
strategy across all long portfolios, i.e., borrowing in the US money market and investing in
foreign money markets. We refer to this zero-cost strategy as DOL. The second principal
component is equivalent to a long position in Portfolio 1 (investment currencies) and a short
position in Portfolio 5 (funding currencies), and corresponds to borrowing in the money mar-
kets of low yielding currencies and investing in the money markets of high yielding currencies.
We refer to this long/short strategy as CAR in our tables –and we use both DOL and CAR
in risk-adjustment below.
Momentum Portfolios. At the end of each period t, we form …ve portfolios based on
exchange rate returns over the previous 3-months. We assign the 20% of all currencies with
the highest lagged exchange rate returns to Portfolio 1, and the 20% of all currencies with
the lowest lagged exchange rate returns to Portfolio 5. We then compute the excess return
14

for each portfolio as an equally weighted average of the currency excess returns within that
portfolio. A strategy that is long in Portfolio 1 (winner currencies) and short in Portfolio 5
(loser currencies) is then denoted as M OM .17
Value Portfolios. At the end of each period t, we form …ve portfolios based on the
level of the real exchange rate.18 We assign the 20% of all currencies with the lowest real
exchange rate to Portfolio 1, and the 20% of all currencies with the highest real exchange
rate to Portfolio 5. We then compute the excess return for each portfolio as an equally
weighted average of the currency excess returns within that portfolio. A strategy that is long
in Portfolio 1 (undervalued currencies) and short in Portfolio 5 (overvalued currencies) is then
denoted as V AL.
Risk Reversal Portfolios. At the end of each period t, we form …ve portfolios based
on out-of-the-money options. We compute for each currency in each time period the risk
reversal, which is the implied volatility of the 10-delta call less the implied volatility of the 10-
delta put, and assign the 20% of all currencies with the lowest risk reversal to Portfolio 1, and
the 20% of all currencies with the highest risk reversal to Portfolio 5. We then compute the
excess return for each portfolio as an equally weighted average of the currency excess returns
within that portfolio. A strategy that is long in Portfolio 1 (high-skewness currencies) and
short in Portfolio 5 (low-skewness currencies) is then denoted as RR.
Volatility Risk Premia Portfolios. At the end of each period t, we group currencies
into …ve portfolios using the 1-year volatility risk premium constructed as described earlier.
We allocate 20% of all currencies with the highest expected volatility premia, i.e., those which
are cheapest to insure, to Portfolio 1, and 20% of all currencies with the lowest expected
volatility premia, i.e., those which are expensive to insure, to Portfolio 5. We then compute
the average excess return within each portfolio, and …nally calculate the portfolio return from
a strategy that is long in Portfolio 1 (cheap volatility insurance) and short in Portfolio 5
17
Consistent with the results in Menkho¤, Sarno, Schmeling, and Schrimpf (2012b), sorting on lagged
exchange rate returns or lagged currency excess returns to form momentum portfolios makes no qualitative
di¤erence to our results below. The same is true if we sort on returns with other formation periods in the
range from 1 to 12 months.
18
We compute the real exchange rate at the end of each month as RERt = St =P P Pt , where St is the
nominal exchange rate and P P Pt is the purchasing power parity rate computed using country CPI’s.
15

(expensive volatility insurance), and denote it V RP .
4 The V RP Strategy: Empirical Evidence

4.1 Summary Statistics and the Returns to V RP
Table 1 presents summary statistics for the annualized average realized volatility RVt; , syn-
thetic volatility swap rate SWt; = EtQ [RVt; ], and volatility risk premium V RPt; = RVt;
SWt; for the 1-year maturity ( = 1); in what follows, we drop the subscript, as it is always
1 year.
The table shows that, on average across developed currencies, RVt equals 10:90 percent,
with a standard deviation of 2:65 percent, and SWt equals 11:68 percent, with a standard
deviation of 2:71 percent. The average volatility risk premium V RPt across these currencies,
which is the di¤erence of these two variables, is equal to 0:78 percent, with a standard
deviation of 1:64 percent. For the full sample of developed and emerging countries, RVt
and SWt are slightly larger than for the sample of only developed currencies, and so is the
volatility risk premium, V RPt , which equals 1:15 on average. We might expect to see this
as the average price that hedgers have to pay to satisfy their demand for volatility insurance
is larger when including emerging market currencies.19
Table 2 describes the returns (net of transactions costs) generated by our short expensive-
to-insure, long cheap-to-insure currency strategy, reporting summary statistics for the …ve
portfolios that are obtained when sorting on the volatility risk premium. In this table, PL is
the long portfolio that buys the top 20% of all currencies with the cheapest volatility insurance,
P2 buys the next 20% of all currencies ranked by expected volatility premia, and so on till the
…fth portfolio, PS which is the portfolio that buys the top 20% of all currencies which are the
most expensive to insure. V RP essentially buys PL and sells PS , with equal weights, so that
V RP = PL PS .
Table 2 reveals several facts about V RP . First, there is a general tendency of portfolio
returns to decrease as we move from PL towards PS , although the decrease is not monotonic.
The V RP average return is 4.95 (4.16) for the sample of Developed (Developed and Emerging)
19
Table A.1 in the Internet Appendix reports summary statistics on the volatility risk premium for each
currency.
16

countries, and is statistically signi…cantly di¤erent from zero at least at the 5 percent level
both for the excess returns and the FX return component alone. Second, the V RP return
stems mainly from the long portfolio, PL . Third, the return from PL can almost completely
be attributed to spot rate changes. Finally, the bottom panel of Table 2 shows the transition
matrix between portfolios. This shows that there is currency rotation across quintile portfolios
such that the steady-state transition probabilities are identical. Thus the performance of the
strategy cannot simply be attributed to long-lived positions in particular currencies, a point
we analyze in greater detail later in the paper.
The returns to V RP are very robust, based on a number of checks. First, we compute
volatility risk premia using simple at-the-money implied volatility rather than the more com-
plicated model-free implied volatility. We also implement the simple variance swap formula of
Martin (2012), which allows for jumps. In both cases, results are virtually identical for devel-
oped countries, and improve for developed and emerging countries. We report these results
in Internet Appendix Table A.10. Second, in our empirical work we also experiment with
an AR(1) process for RV to form expectations of RV rather than using lagged RV over the
previous 12 months. Again, we …nd that the results are virtually identical to those reported
in Table 2. Third, in Internet Appendix Table A.7 we check whether a simple strategy based
on sorting currencies by the di¤erence between longer-term and short-term realized volatility
e¤ectively captures the returns from V RP . Using de…nitions of long-term ranging from six
to 24 months and short-term from one to six months, we …nd that while there are a few high-
return portfolios in the set, there is substantial variation in these returns across portfolios,
leading to concerns of potential data-mining. Perhaps more importantly, these returns have
low correlations with the returns of the V RP strategy, suggesting that implied volatility infor-
mation from the options market is critical to the construction of the V RP strategy. Fourth,
we show in Internet Appendix Table A.4 that the identities of the currencies most often found
in the corner V RP portfolios are not easily recognizable from other currency strategies such
as carry.
In the next section, we formalize this fourth exercise by explicitly comparing the returns
of V RP to the conventional set of currency strategies considered in the literature thus far.
17

4.2 Comparing V RP with Other Currency Strategies
In Table 3, we present the net returns to a number of long-short currency strategies computed
using only time t 1 information, to compare the predictability generated by strategies pre-
viously proposed in the literature with the new V RP strategy that we propose. We compare
CAR, M OM , V AL, and RR with our V RP strategy. We report results for both subsamples
(Developed, and Developed and Emerging) in our data.
Panel A of the table shows the results for the excess returns generated by these trading
strategies. Consistent with a vast empirical literature (e.g., Lustig, Roussanov, and Verdel-
han, 2011, Burnside, Eichenbaum, Kleshchelski, and Rebelo, 2011, and Menkho¤, Sarno,
Schmeling, and Schrimpf, 2012a), CAR delivers a sizable average excess return, especially for
the broader sample of countries analyzed. The Sharpe ratio of the carry trade is 0.38 for the
sample of developed countries, and 0.53 for the full sample. M OM generates only small, yet
positive, net excess returns, which is consistent with the recent evidence in Menkho¤, Sarno,
Schmeling, and Schrimpf (2012b) that the performance of currency momentum has weakened
substantially during the last decade. Both V AL and RR do quite well, with Sharpe ratios
between 0.28 and 0.41 for V AL, and 0.41 and 0.42 for RR. However, with the exception
of CAR for the Developed and Emerging sample and RR for the Developed sample, none
of these common currency strategies generates average returns that are statistically signi…-
cantly di¤erent from zero during the period we analyze, which includes fully the recent global
…nancial crisis and is rather short (16 years).
In contrast, the V RP strategy that we introduce generates a Sharpe ratio of 0.61 and 0.51
for the two samples of countries considered, signifying that it outperforms all strategies for
the Developed sample and is only slightly inferior to carry for the Developed and Emerging
sample of countries. It is important to note that, for both samples, the V RP returns are
clearly statistically signi…cantly di¤erent from zero. Interestingly, the V RP strategy works
better for the developed countries in our sample than for the whole sample of developed and
emerging countries. One plausible explanation for this is that there is a greater prevalence of
hedging using more sophisticated instruments such as currency options in developed markets
than in emerging markets.
While Panel A of the table suggests that the returns to the V RP strategy are comparable
18

to or better than those of the other strategies that we provide as comparison, Panel B of the
table introduces an important feature of the V RP strategy, namely that the major portion of
these returns accrue as a result of spot rate predictability. This predictability is much larger
than any competitor strategy over the sample period, generating an annualized mean spot
exchange rate return of 5.45% for the developed countries, and 5.27% for the full cross-section
of all 20 countries in our sample. In contrast, the exchange rate return from CAR is negative
for both samples, and while other strategies have relatively better performance in predicting
movements in the spot rate than CAR, the degree of predictability in any of these alternative
strategies is also substantially smaller than V RP .
Several of the other moments presented in Panel B of Table 3 are also worth highlight-
ing. First, the returns from V RP display desirable skewness properties, as its unconditional
skewness is close to zero, and the maximum drawdown is far better (i.e., smaller in absolute
size) than that of CAR. Finally, the table shows that the portfolio turnover of the V RP
strategy (measured in terms of changes in the composition of the short and long legs of the
V RP strategy, F reqS and F reqL in Table 3) is reasonable – lying in between the very low
turnover of CAR and the high turnover of M OM .20
4.3 Combining V RP with Other Currency Strategies
Panel C of Table 3 documents the correlation of the V RP strategy with the other strategies,
and …nds that the strategy tends to be mildly negatively correlated with CAR (with correla-
tions of -0.08 and -0.06 for the two samples) and mildly positively correlated with M OM (with
correlations of 0.11 and 0.15 for the two samples). The correlation with V AL for Developed
countries is higher, but at 0.19 there is substantial orthogonal information in the strategy
– indeed several of the other strategies are substantially more correlated with one another.
Apart from showing that the strategy is distinct from those already studied in the literature,
this also implies that combining V RP with CAR, M OM , V AL, and RR could yield sizable
diversi…cation bene…ts to an investor.21
20
Table A.2 in the Internet Appendix reports the same information as Table 3 for gross, rather than net
returns.
21
It is also useful to note that the correlations for the excess returns from the strategies, presented in the
table, are very close in magnitude to the correlations acquired from the exchange rate component of these
returns – in other words, it is the currency component of the returns to this strategy that is the proximate
source of the diversi…cation bene…ts.
19

Figure 1 provides a graphical illustration of the di¤erences in the performance of the
strategies highlighted in Table 2, and restricts the plot to the sample of Developed countries
to conserve space – the graph for the full sample of countries is reported in the Internet
Appendix, Figure A.1. The …gure plots the one-year rolling Sharpe ratio for these strategies,
and makes visually clear the marked di¤erence in the evolution of risk-adjusted returns of
V RP relative to the others. While there is a substantial improvement in the Sharpe ratio of
V RP during the recent crisis, the strategy is not driven entirely by the crisis period – the
Sharpe ratio appears to be no more volatile than the Sharpe ratio of CAR and M OM .
Table 4 shows the subsample performance of the currency component of these strategies as
a complement to Figure 1. Despite the inevitable attenuation of the sample period and the
attendant di¢ culty of establishing statistical signi…cance for each subperiod, the performance
of V RP does seem substantially higher during crisis and NBER recession periods. However
even outside of these recession periods, the return to V RP is still large and positive, and
higher than that of all the competitor strategies. Even if V RP were to be used primarily as
a hedge for a canonical currency strategy, it seems to exhibit desirable properties, delivering
positive returns outside of crisis periods, and very high returns within crisis periods.
Figure 2 plots the cumulative wealth generated by the strategies over the sample period
(only for the Developed Countries, the full sample graph is shown in the Internet Appendix,
Figure A.2), decomposing it into its two constituents: the exchange rate component (FX) and
the interest rate di¤erential component (yield). For CAR the yield component is the sole
positive driver of the return because the cumulative FX return component is negative. Also,
for RR the yield component vastly dominates the FX component, whereas for V AL, the yield
component and the FX component are roughly equal in size. Only for M OM is most of the
excess return driven by spot predictability, but the size of the return itself and the Sharpe
ratio are tiny during our sample. V RP returns are di¤erent in that they are made up of a
mildly negative yield component (for both samples of countries considered), and therefore the
component due to spot return predictability is in fact larger than the full portfolio return,
achieving Sharpe ratios above 0.50 in both samples of countries.
Taken together, the results from this section suggest that the V RP strategy has creditable
excess returns overall, an important tendency to deliver returns during crisis periods that
are far higher than the crashes commonly experienced with the carry trade, and far stronger
20

predictive power for exchange rate returns, which is a unique feature in the space of alternative
currency trading strategies. The importance of these features of the V RP strategy is twofold.
First, a currency investor would likely gain a great deal of diversi…cation bene…t from adding
V RP to a currency portfolio to enhance risk-adjusted returns. Second, a spot currency trader
interested in forecasting exchange rate ‡uctuations (as opposed to currency excess returns)
might value the signals provided by V RP .
To better understand the value of the V RP strategy for a currency investor, we compute
the optimal currency portfolio for an investor who uses all of the …ve strategies considered
here: CAR, V AL, RR, M OM , and V RP . Speci…cally, consider a portfolio of N assets with
covariance matrix . The global minimum volatility portfolio is the portfolio with the lowest
return volatility, and represents the solution to the following optimization problem: min w0 w
subject to the constraint that the weights sum to unity w0 = 1, where w is the N 1 vector of
portfolio weights on the risky assets, is a N 1 vector of ones, and is the N N covariance
matrix of the asset returns. The weights of the global minimum volatility portfolios are given
1
by w = 0 1 . We compute the optimal weights for both the Developed, and Developed and
Emerging countries, and report the results graphically in Figure 3.
The results show that the optimal weight assigned to the V RP strategy is high, equal to
26 and 28 percent for the two sets of countries. The Sharpe ratio of the minimum volatility
portfolio for the Developed sample, for instance, is quite impressive, at 0.69. However,
this number drops to 0.60 if the investor is not given access to the V RP strategy, and only
employs the other four currency strategies. Similarly for the Developed & Emerging sample,
the Sharpe ratio equals 0.60 when the V RP strategy is included and drops to 0.50 when it is
excluded from the menu of currency strategies. These …ndings con…rm the value of V RP in
a currency portfolio and its desirable correlation properties.
Before turning to studying possible explanations of the performance of V RP , we check
whether such predictive power is purely cross-sectional. Speci…cally, one may be concerned
whether – given the relatively short sample period of 16 years – the predictability recorded
here stems from long-lived cross-sectional di¤erences in the volatility risk premium, which
happen to be related to cross-sectional di¤erences in excess returns. To check whether this is
the case, we construct a static V RP currency strategy, which we denote V RP , which buys
(sells) the currencies with the highest (lowest) average volatility risk premia over the sample
21

period. This strategy requires no portfolio rebalancing, and its performance is informative
of the extent to which the returns to the V RP strategy are due to unconditional di¤erences
in the volatility risk premium between currencies in the cross-section. However, this strategy
does contain a look-ahead bias, since it assumes that an investor knows the unconditional
mean of the VRP for each currency rather than having to learn it over time. As a result,
the returns we compute here provide an upper bound of what a static strategy could achieve.
We also compute analogous returns CAR, M OM , V AL, and RR. These returns can be
thought of as the “static component” in the return decomposition proposed by Hassan and
Mano (2013), which is designed to measure the relative importance of cross-sectional versus
time-series predictability in FX strategies.
Table 5 shows the returns of these static strategies gross of transaction costs. Panel A
presents the overall excess return and suggests that V RP performs well, with an average return
of 3.51 (3.28) per annum for Developed (Developed and Emerging) Countries. However, Panel
B of Table 5 shows that V RP returns are virtually entirely due to cross-sectional di¤erences
in the average interest rate di¤erential, as there is basically no predictability in FX returns
– this establishes that V RP is a distinct strategy from V RP , which derives virtually all of
its performance from FX returns. Moreover, we cannot establish the statistical signi…cance of
V RP returns at conventional signi…cance levels. Finally, Panel C of Table 5 shows that these
static returns are highly correlated with one another, with V RP in particular displaying a
correlation of 0.73 with carry.22 Taken together, this table shows that time-series variation
in currency volatility risk premia is important to explain the performance of V RP .
5 Understanding V RP Returns
The empirical results reported earlier suggest that the currency volatility risk premium con-
tains powerful predictive information for currency returns that is markedly di¤erent from the
information contained in several common predictors studied in the literature. While the main
contribution of our paper is empirical and we do not have a formal theoretical model that
links the volatility risk premium (or its determinants) to spot currency returns, we suggest
three possible mechanisms that may drive our results, and provide empirical evidence on each
22
In Table A.3 in the the Internet Appendix, we examine the static, dynamic and dollar component of the
V RP returns in a similar vein to Hassan and Mano (2014).
22

of these.
5.1 Risk Premia
First, we consider the possibility that returns from the V RP strategy re‡ect compensation
for risk. We begin by testing the pricing power of conventional risk factors for V RP returns,
using standard linear asset pricing models, in both the cross-section and the time-series.
Time series tests. As a …rst step, Table 6 simply regresses the time-series of V RP
returns on a number of risk factors proposed in the literature. First, Panel A con…rms the
results found in Tables 2 and 3, by using DOL, CAR, M OM , V AL, and RR as right-hand
side variables, and shows that for both Developed and Developed and Emerging samples, there
is substantial and statistically signi…cant alpha relative to these factors. Panel B of the table
uses the three Fama-French factors and adds equity market momentum, denoted M OM E.
Again, V RP has alpha relative to these factors which is very close to that in the prior panel.
Finally, Panel C of Table 6 employs the Fung-Hsieh (2004) factor model, which has been
used in numerous previous studies; see for example, Bollen and Whaley (2009), Ramadorai
(2013), and Patton and Ramadorai (2013). The set of factors comprises the excess return
on the S&P 500 index; a small minus big factor constructed as the di¤erence between the
Wilshire small and large capitalization stock indexes; excess returns on portfolios of lookback
straddle options on currencies, commodities, and bonds, which are constructed to replicate the
maximum possible return to trend-following strategies on their respective underlying assets;
the yield spread of the US 10-year Treasury bond over the 3-month T-bill, adjusted for the
duration of the 10-year bond; and the change in the credit spread of Moody’s BAA bond over
the 10-year Treasury bond, also appropriately adjusted for duration. Yet again, the table
shows that the alpha of V RP is una¤ected by the inclusion of these factors.
Cross-Sectional Tests. Our cross-sectional tests rely on a standard stochastic discount
factor (SDF) approach (Cochrane, 2005), and we focus on a set of risk factors in our investi-
gation that are motivated by the existing asset pricing literature on the returns to currency
strategies. We begin by brie‡y reviewing the methods employed, and denote excess returns
of portfolio i in period t by RXti .
The usual no-arbitrage relation applies, so risk-adjusted currency excess returns have a
zero price and satisfy the basic Euler equation:
23

E[Mt RXti ] = 0; (6)
with a linear SDF Mt = 1 b0 (ft ), where ft denotes a vector of risk factors, b is the vector
of SDF parameters, and denotes factor means.
This speci…cation implies a beta pricing model in which expected excess returns depend
on factor risk prices , and risk quantities i, which are the regression betas of portfolio excess
returns on the risk factors for each portfolio i (see e.g., Cochrane, 2005):
0
E RX i = i (7)
The relationship between the factor risk prices in equation (7) and the SDF parameters in
equation (6) is simply given by = f b, where f is the covariance matrix of the risk factors.
Thus, factor risk prices can be easily obtained via the SDF approach, which we implement
by estimating the parameters of equation (6) via the generalized method of moments (GMM)
of Hansen (1982).23 We also present results from the more traditional two-stage procedure of
Fama and MacBeth (1973) in our empirical implementation.
In our asset pricing tests we consider a two-factor linear model that comprises DOL and one
additional risk factor, which is one of CAR and V OLF X . DOL denotes the average return
from borrowing in the US money market and equally investing in foreign money markets.
CAR is the carry portfolio described earlier. V OLF X is a global FX volatility risk factor
constructed as the innovations to global FX volatility, i.e., the residuals from an autoregressive
model applied to the average realized volatility of all currencies in our sample, as in Menkho¤,
Sarno, Schmeling, and Schrimpf (2012a). In Internet Appendix Table A.8, we also consider
innovations to global average percentage bid-ask spreads in the spot market (BASF X ) and
the option market (BASIV ), which can be seen as global proxies for the FX spot market and
the FX option market illiquidity, respectively.
In assessing our results, we are aware of the statistical problems plaguing standard asset
pricing tests, recently emphasized by Lewellen, Nagel, and Shanken (2010). Asset pricing
tests can often be highly misleading, in the sense that they can indicate strong but illusory
23
Estimation is based on a pre-speci…ed weighting matrix and we focus on unconditional moments (i.e., we
do not use instruments other than a constant vector of ones) since our interest lies in the performance of the
model to explain the cross-section of expected currency excess returns (see Cochrane, 2005; Burnside, 2011).
24

explanatory power through high cross-sectional R2 statistics, and small pricing errors, when
in fact a risk factor has weak or no pricing power. Given the relatively small cross-section of
currencies in our data, as well as the relatively short time span of our sample, these problems
can be severe in our tests. As a result, when interpreting our results, we only consider the cross-
sectional R2 and Hansen-Jagannathan (HJ) tests on the pricing errors if we can con…dently
detect a statistically signi…cant risk factor, i.e., if the estimates clearly point to a statistically
signi…cant market price of risk on a factor.
Table 7 reports GMM estimates of b, portfolio-speci…c ’s, and implied ’s, as well as
cross-sectional R2 statistics and the HJ distance measure (Hansen and Jagannathan, 1997).
In the table, standard errors are constructed as in Newey and West (1987) with optimal lag
length selection according to Andrews (1991). Besides the GMM tests, we employ traditional
Fama-MacBeth (FMB) two-pass OLS regressions (with Shanken (1992) corrected standard
errors) to estimate portfolio betas and factor risk prices. Note that we do not include a
constant in the second stage of the FMB regressions, i.e. we do not allow a common over- or
under-pricing in the cross-section of returns –however our results are virtually identical when
we replace the DOL factor with a constant in the second stage regressions.24 Since DOL
has virtually no cross-sectional relation to portfolio returns, it serves the same purpose as a
constant that allows for common mispricing.
Panels A and B of Table 7 show clearly how none of the risk factors considered enters the
SDF with a statistically signi…cant risk price , and that this is the case for both the developed
countries and the full sample. As expected, the FMB results in the table are qualitatively, and
in most cases also quantitatively identical to the one-step GMM results. The bottom part of
the panels show that there is little cross-sectional variation across the 5 portfolios sorted by
the cost of currency insurance, which is what we con…rm more formally in the asset pricing
tests. While the HJ test delivers large p-values for the null of zero pricing errors in all cases,
we attach no information to this result given the lack of clear statistical signi…cance of the
market price of risk.25
24
Also see Lustig and Verdelhan (2007) and Burnside (2011) on the issue of whether or not to include a
constant in these regressions.
25
We also carried out asset pricing tests using the same methods and risk factors in which we attempt to
price only the exchange rate component of the returns from V RP . In that exercise, the results are equally
disappointing in that all risk factors included in the various SDF speci…cations are statistically insigni…cant.
25

Aversion to Volatility Risk. Next, we investigate the possibility that the currency-
speci…c volatility risk premium captures ‡uctuations in aversion to volatility risk – i.e., a
time-varying factor loading on the global volatility risk factor. We have already ascertained
that a simple strategy allowing for static loadings on the Menkho¤, Sarno, Schmeling, and
Schrimpf (2012a) strategy fails to explain the cross-section of V RP portfolio returns, but
these tests do not account for the possibility that di¤erent currencies load di¤erently on a
global volatility shock at di¤erent points in time. There is also the possibility that market
segmentation causes expected returns on di¤erent currencies to be determined independently
– but this (remote) possibility is very di¢ cult to evaluate, and if our strategy did indeed
provide evidence of this, it would have far-reaching consequences.
To evaluate whether V RP returns can be explained by currencies exhibiting time-varying
loadings on a global volatility shock, we estimate the loadings of currency returns on various
proxies for global volatility risk, and build portfolios sorted on these estimated loadings.
Speci…cally, we estimate the following rolling regression for each currency i:
RXti = i + i GV OLt + "it ;
Here GV OL is a proxy for global volatility risk premia and we employ various measures,
including the average volatility risk premium across our currencies (with equal weights); the
…rst principal component of the currencies’ volatility risk premia; and the equity volatility
risk premium computed as the di¤erence between the time-t one-month realized volatility on
the S&P500 and the VIX index.
We estimate these regressions using rolling windows of 36 months. After obtaining esti-
mates of the i coe¢ cients, we sort currencies into …ve portfolios on the basis of these i
estimates. Finally, we construct a long-short strategy which buys currencies with low betas
and sells currencies with high betas. In essence, this strategy exploits di¤erences in exposure
of individual currencies to global measures of volatility risk premia, which is a direct test of
the above hypothesis.
The results using our three measures for GV OL are qualitatively identical and we report
in Table 8 the results for GV OL set equal to the average volatility risk-premium across the
currencies in our sample. Internet Appendix Tables A.5 and A.6 contain results for the other
two measures. The table shows that the performance of this strategy is strictly inferior to the
26

performance of the V RP strategy (in fact producing negative returns), and the correlation
between the returns from the two strategies is close to zero. On the basis of this evidence, we
conclude that there is no support for V RP returns being driven by aversion to global volatility
risk in the data. Overall, the asset pricing tests reveal that it is not possible to understand
the returns from the V RP strategy as compensation for global risk. Therefore, we turn to
examining di¤erent explanations.
5.2 Limits to Arbitrage
The second possible explanation that we consider is limits to arbitrage, in the spirit of Acharya,
Lochstoer, and Ramadorai (2013). According to this explanation, the returns to V RP arise
from the interaction between natural hedgers of FX risk, and currency market speculators.
When the risk-bearing capacity of currency-market speculators is a¤ected by shocks to the
availability of arbitrage capital, this will make currency options across the board more expen-
sive, with particular impacts on those currencies to which speculators have high exposure –
for example, currency hedge funds may reduce their outstanding short put option positions in
the currencies in which they trade (shorting put options is a favoured strategy of many hedge
funds; see Fung and Hsieh, 1997, and Agarwal and Naik, 2004).
This will result in selling pressure on expensive-to-insure currencies as natural hedgers
such as corporations sell pre-existing currency holdings, abandon expensive currency hedges,
and become more reluctant to denominate contracts in these currencies. Conversely, this
mechanism results in relatively less pressure on cheap-to-insure currencies, for which natural
hedgers are happy to hold higher inventories. This yields the positive long-short returns in
the V RP portfolio. When capital constraints loosen, we should see the opposite behavior,
i.e., a reversal in both the volatility risk premium and the spot currency position.
This explanation has several testable implications. First, for this mechanism to work
demand pressure in the option market must have an impact on option prices, as demonstrated
by Garleanu, Pedersen and Poteshman (2009) for stock options. Therefore, as a preliminary
test, we run a similar regression to Garleanu, Pedersen, and Poteshman for FX markets, in
an attempt to ascertain whether demand pressure in the FX derivatives used for hedging FX
risk generates price impact which a¤ects the volatility risk premium.
We estimate a panel regression (with …xed e¤ects) of the volatility risk premium on a
27

proxy for demand pressure in FX derivatives markets:
VRPit = i + N Demit lag + uit ; (8)
where VRPit is the 1-year volatility risk premium for currency i (i.e., the di¤erence between
the realized volatility, RVt and the synthetic volatility swap rate, SWt ),26 and N Demit denotes
the net demand of currency options and futures for end-users from the US Commodity Futures
Trading Commission (CFTC). The net demand proxy is constructed as the di¤erence between
long and short positions scaled by the total open interest, and is available for two groups of
end-users: commercial and …nancial.
For the left-hand side variable in these regressions, we employ several de…nitions of the
volatility risk premium: the de…nition used in our core analysis, where RV is calculated using
daily exchange rate returns over the previous year and SW is computed as in Britten-Jones
and Neuberger (2000) using 1-year currency option implied volatilities; in VRPsi , SW is
computed using the simple variance swap method of Martin (2012); in VRPgarch , RV is the 1-
year volatility forecast generated from the simple GARCH(1,1) applied to daily exchange rate
returns; in VRPsv , RV is the 1-year volatility forecast generated from a stochastic volatility
model for daily exchange rate returns (Della Corte, Sarno, and Tsiakas, 2009). Monthly CFTC
data are collected on the last Tuesday of every month. All other variables are measured on
the same day.
The regression results, reported in Table 9 for each of the two end-user groups, suggest that
in a contemporaneous regression (lag = 0) the net demand proxy for commercial end-users
always enters with a negative coe¢ cient that is statistically signi…cantly di¤erent from zero,
regardless of the de…nition of the VRP on the left-hand-side. This is essentially the analogue
of the result of Garleanu, Pedersen and Poteshman for the case of FX markets, and it implies
that net demand for hedging in FX markets increases the cost of volatility insurance. It is
also noticeable that this price impact is quite persistent in that the net demand proxy enters
signi…cantly also in a predictive regression (lag = 1 month). In contrast, the coe¢ cient on
…nancial end-users is positive and, in two regressions statistically signi…cantly di¤erent from
zero. Again this is consistent with the story of Garleanu, Pedersen and Poteshman, since
26
Note that this is distinct from V RP , where the italics denote the returns to the trading strategy condi-
tional on realizations of VRP, the level of the volatility risk premium.
28

…nancial users are providing volatility insurance to commercial customers, essentially acting
as market makers.
Table 10 turns from the options market to currency excess returns, testing whether time-
series variation in limits to arbitrage proxies predicts variation in V RP returns. The table
shows results from predicting the exchange rate component of V RP ; the results for excess
returns are, not surprisingly, qualitatively identical and quantitatively very similar. The …rst
column in both panels shows the univariate regression of the exchange rate component of V RP
regressed on the lagged 12-month rolling average of the TED spread. The coe¢ cient on this
variable is positive and statistically signi…cant for both sample of countries examined, which
is consistent with the limits to arbitrage explanation – when funding liquidity is lower (i.e.,
times of high capital constraints on speculators), we …nd that the expected return from V RP
increases. The second column shows that when the 12-month rolling average of changes in VIX
(a proxy for increases in the risk aversion of market participants, yielding both greater limits
to arbitrage and an increased desire to hedge) is positive, V RP returns increase (signi…cantly
for the full sample of countries), again consistent with the limits to arbitrage explanation.27
Similarly, the third column shows that a general …nancial distress indicator (FSI, constructed
by the Federal Reserve Bank of St. Louis) that captures the principal component of a variety
of liquidity and volatility indicators is positive and, for the full sample of countries, statistically
signi…cant. The fourth column of the table interacts TED with changes in VIX, and …nds
strong statistically signi…cant predictive power of this interaction for the FX returns on our
strategy in both samples of countries, suggesting that when funding liquidity is constrained
and risk aversion is high, V RP returns increase. The …nal column of the table adds in
measures of capital ‡ows into hedge funds. When aggregate capital ‡ows into hedge funds are
high, signifying that they experience fewer constraints on their ability to engage in arbitrage
transactions, we …nd that returns for the V RP strategy are lower and vice versa, although
the variable is only signi…cant for the sample of developed countries.
The …nal …ve rows of Table 10 introduce several of the variables described above simul-
taneously to test their joint and separate explanatory power. We generally include TED,
changes in VIX and the interaction separately to avoid potential collinearity in the regres-
27
This is similar to the results in Nagel (2012), who shows that a strategy of liquidity provision in equity
markets has returns which are highly correlated with VIX.
29

sions as these variables are highly correlated with one another –since they capture aggregate
variation in funding liquidity and risk aversion, it is obvious that they contain a substantial
common component. Nonetheless, we …nd that all these variables retain their signs and are
often statistically signi…cant in these multivariate predictive regressions, o¤ering some sup-
port to the limits to arbitrage explanation of our results. Table A.9 in the Internet Appendix
reports results for the same regressions using raw measures of VIX, TED and FSI rather than
rolling averages, and shows that the results are qualitatively identical.
Finally, we examine whether the observed buying and selling actions of di¤erent players
in the currency market follow the pattern implied by the limits to arbitrage explanation, i.e.,
that currencies in the high volatility-insurance portfolio are sold and those in the low volatil-
ity insurance portfolio are bought by natural hedgers, with speculators taking the opposite
position. We do so using the CFTC data on the position of commercial and …nancial traders
in FX markets, essentially taking the currencies ranked by their volatility insurance costs, and
documenting the traders’positions (cumulative net positions), rather than returns.28 We view
the CFTC position data as a proxy for cumulative order ‡ow across di¤erent segments of FX
market participants, given that there is evidence that the CFTC position data and currency
order ‡ow capture very similar information (e.g., Klitgaard and Weir, 2004).
The results of this exercise are reported in Figure 4, which plots the cumulative position
in the currencies in the V RP portfolio for …nancial and commercial traders. We …nd that the
position of commercial traders follows the pattern implied by the limits to arbitrage explana-
tion – such traders sell expensive-to-insure currencies and buy cheaper-to-insure currencies.
Financial traders display the opposite behavior, with a strongly negative position in the V RP
portfolio, which is consistent with their acting as market-makers, providing liquidity to satisfy
the buying (selling) demand for low (high)-insurance currencies.
Taken together, the results in this section lend support to a limits to arbitrage explanation
for the predictability of spot exchange rates associated with V RP . However, this evidence
should be taken with caution in the absence of a formal theoretical model in our paper. This
raises the possibility that our results in this section could be better-explained by alternative
28
To allow for meaningful cross-currency comparisons, we need to ensure that net positions are comparable
across currencies, as their absolute size di¤ers across currencies. We therefore divide net positions by their
standard deviation computed over a rolling window of 3 months.
30

mechanisms.
5.3 Information Asymmetry
The third possible explanation that we examine is whether volatility risk premia predict the
future direction of currency returns because investors trading currency derivatives are better
informed about the value of the underlying asset than those trading the underlying asset. This
mechanism is one that has been used with some success in equity and credit options markets
to explain the relationship between options and underlying assets (see, for example, Easley,
O’Hara, and Srinivas, 1998, Acharya and Johnson, 2007, Goyal and Saretto, 2009, and Buss
and Vilkov 2012). At the outset, we note that our results cannot be consistent with a simple
strategy based on informed traders buying currency call (put) options in advance of expected
currency appreciations (depreciations), for which there is some evidence in equity options
markets (see Pan and Poteshman, 2006). This is because any price pressure from increased
demand for either call or put options (a la Bollen and Whaley, 2004, and Garleanu, Pedersen,
and Poteshman, 2009) will result in increased implied volatility, and hence a lower volatility
risk premium – rather than decreases (increases) in the volatility risk premium predicting
appreciations (depreciations), as we …nd. While this makes it more di¢ cult for a simple story
based on information asymmetries between options and spot markets to explain our …ndings,
it does not rule out more complex alternatives.
Our empirical work on this explanation follows Bali and Hovakimian (2009), who propose
a VAR-GARCH model to investigate whether information asymmetries can explain the pre-
dictive power of the equity volatility risk premium for the cross-section of stock returns. We
estimate this model using the …ve portfolios used to construct the V RP strategy: yt = ct + ut ,
1=2
with ut = Ht "t where yt = (et ; vt )0 . Here et denotes the average 1-month exchange rate
return for a portfolio of currencies whereas vt is the cross-sectional average 1-year volatility
risk premium for the same portfolio of currencies. ct is the conditional mean modelled as a
Vector Autoregressive Process (VAR) of order p, and Ht is the conditional covariance of error
terms, whose elements are modelled as:
2 3 2 3 2 32 3 2 32 3
he !e e 0 ev he e 0 ev u2e
4 hh 5 = 4 ! h 5 + 4 0 h 0 5 4 hh 5 +4 0 h 0 5 4 ue uv 5 : (9)
hv t !v ve 0 v hv t 1 ve 0 v u2v t 1
31

We estimate the above speci…cation via maximum likelihood, and Table 11 reports the
coe¢ cient estimates and t-statistics of the spillover parameters ev , ve , ev and ve in the
conditional variance equations, separately for each of the …ve portfolios PL , P2 , P3 , P4 and PS .29
Non-zero values of the parameters in ev and ev imply lead-lag e¤ects from option markets
to spot markets via the conditional variance equation. Non-zero values of the parameters ve
and ve imply the reverse, with movements in spot market conditional volatility predicting
those in option markets.
Table 11 shows that the coe¢ cients in ev are statistically signi…cantly di¤erent from
zero for all portfolios except one (P4 ), and for both samples of countries examined. In other
words, lagged squared shocks to the volatility risk premium signi…cantly a¤ect the conditional
variance of FX spot returns. On the other side, ve is statistically signi…cant twice, in the
sample of Developed and Emerging markets, suggesting that the potentially less sophisticated
Emerging currency options markets are subject to some bi-directional causality.30 Overall,
there is less evidence that the conditional variance of the volatility risk premium is a¤ected
by the variance of spot returns. Interestingly, these VAR-GARCH results are qualitatively
identical to those reported by Bali and Hovakimian for equity markets.
These results are consistent with the possibility that there is non-directional information
which is manifested in the currency options market prior to movements occurring in the
underlying currency spot market. To be more speci…c, this is consistent with currency options
market participants knowing that either an appreciation or a depreciation is likely in the
subsequent period in the underlying spot, and trading based on this information. Ultimately,
while this does not allow us to conclude that information asymmetries are an explanation for
our empirical results, it does suggest a way to augment the performance of V RP –by scaling
the weights of di¤erent currencies in the portfolio using a measure of conditional volatility in
the options market.
29
We …nd that the maximum likelihood parameter estimates of the VAR-GARCH model provides no ev-
idence for signi…cant autocorrelations and cross-correlations in the conditional mean, indicating no lead-lag
e¤ects between the conditional means of FX spot and options returns or vice versa, so we do not report these
parameters in the interests of conserving space.
30
Table A.11 in the Internet Appendix repeats this exercise for individual currencies and con…rms that
for Developed markets, ve and ve are never statistically signi…cant (and ev and ev almost always are),
whereas for Emerging markets, there is more evidence of bi-directional causality, with ve and ve being
estimated to be statistically signi…cant more frequently.
32

6 Conclusions
We show that the currency volatility risk premium has substantial predictive power for the
cross-section of currency returns. Currencies with low implied volatility relative to historical
realized volatility –those with relatively cheap volatility insurance –predictably appreciate,
while currencies with relatively more expensive volatility insurance predictably depreciate.
This predictive power is speci…cally related to future variation in spot exchange rate returns,
and not to interest rate di¤erentials.
A portfolio of currencies (which we dub V RP ) constructed by going long cheap volatility
insurance currencies and short expensive volatility insurance currencies generates economically
and statistically signi…cant returns. The returns of this V RP portfolio are largely uncorrelated
with the canonical set of currency strategies, and these diversi…cation bene…ts combined with
its high returns mean that V RP has a large weight in a global minimum volatility currency
portfolio which also includes carry, currency momentum, currency value, and currency risk-
reversal strategies.
While we do not have a formal theoretical model, we do provide empirical evidence per-
taining to several possible explanations for the performance of the strategy. We …nd that
a comprehensive set of standard risk factors is unable to explain V RP returns, suggesting
that these returns are not generated on account of compensation for systematic risk. We
…nd some evidence in support of an explanation in which time-variation in limits to arbitrage
causes volatility insurance costs to ‡uctuate across time and currencies, with consequences
for the spot market as risk-averse currency hedgers become reluctant to take or hold positions
in expensive-to-insure currencies. Finally, while we do not …nd evidence that information
asymmetries between currency options and spot markets can explain our results, we do …nd
some evidence that conditional variance in the currency options market predicts conditional
variance in the underlying currency spot markets.
Overall, the results in our paper provide new insights into the predictability of exchange
rate returns, an important area, in which evidence has been di¢ cult to obtain. We also intro-
duce a new currency strategy with useful diversi…cation properties into the rapidly-expanding
research on this topic. While our empirical results point to new, powerful predictive informa-
tion for currency returns, our attempts to explain the drivers of this predictive power have
33

met with mixed success, and are limited by the absence of a formal theoretical model that
links volatility risk premia and underlying asset returns. The development of such theory is
an important avenue for future research.
34

References
Acharya, V., and T. Johnson (2007). “Insider Trading in Credit Derivatives,” Journal of
Financial Economics 84, 110-141.
Acharya, V., L.A. Lochstoer, and T. Ramadorai (2013). “Limits to Arbitrage and Hedging:
Evidence from Commodity Markets,”Journal of Financial Economics 109, 441-465.
Adrian, T., E. Etula, and T. Muir (2013). “Financial Intermediaries and the Cross-Section of
Asset Returns,”Journal of Finance, forthcoming.
Agarwal, V., and N.Y. Naik (2004). “Risks and Portfolio Decisions Involving Hedge Funds,”
Review of Financial Studies 17, 63–98.
Akram, Q.F., D. Rime, and L. Sarno (2008). “Arbitrage in the Foreign Exchange Market:
Turning on the Microscope,”Journal of International Economics 76, 237–253.
Andrews, D.W.K. (1991). “Heteroskedasticity and Autocorrelation Consistent Covariance Ma-

trix Estimation,”Econometrica 59, 817–858.
Ang, A., and J. Chen (2010). “Yield Curve Predictors of Foreign Exchange Returns,”Working
Paper, Columbia Business School.
Bali, T.G. and A. Hovakimian (2009). “Volatility Spreads and Expected Stock Returns,”
Management Science 55, 1797-1812.
Bank for International Settlements (2013). Triennial Central Bank Survey of Foreign Exchange
and Derivatives Market in 2013. Basel: Bank for International Settlements Press.
Barroso, P., and P. Santa-Clara (2014). “Beyond the Carry Trade: Optimal Currency Portfo-
lios,”Journal of Financial and Quantitative Analysis, forthcoming.
Bates, D.S. (1991). “The Crash of ’87: Was it Expected? The Evidence from Options Mar-
kets,”Journal of Finance 46, 1009–1044.
35

Black, F., and M. Scholes (1973). “The Pricing of Options and Corporate Liabilities,”Journal
of Political Economy 81, 637–654.
Bollen, N.P., and Whaley, R.E. (2004). “Does Net Buying Pressure A¤ect the Shape of Implied
Volatility Functions?”Journal of Finance 59, 711–753.
Bollen, N.P., and Whaley, R.E. (2009). “Hedge Fund Risk Dynamics: Implications for Perfor-
mance Appraisal,”Journal of Finance 64, 987–1037.
Bollerslev, T., G. Tauchen, and H. Zhou (2009). “Expected Stock Returns and Variance Risk
Premia,”Review of Financial Studies 22, 4463–4492.
Breeden, D., and R. Litzenberger (1978). “Prices of State-Contingent Claims Implicit in Op-
tion Prices,”Journal of Business 51, 621–651.
Britten-Jones, M., and A. Neuberger (2000). “Option Prices, Implied Price Processes, and
Stochastic Volatility,”Journal of Finance 55, 839–866.
Broadie, M., and A. Jain (2008). “Pricing and Hedging Volatility Derivatives,” Journal of
Derivatives 15, 7–24.
Brockhaus, O., and D. Long (2000). “Volatility Swaps Made Simple,”Risk 19, 92–95.
Buraschi, A., F. Trojani, and A. Vedolin (2014). “When Uncertainty Blows in the Orchard:
Comovement and Equilibrium Volatility Risk Premia,”Journal of Finance 69, 101-137.
Burnside, C. (2011). “The Cross Section of Foreign Currency Risk Premia and Consumption
Growth Risk: Comment,”American Economic Review 101, 3456–76.
Burnside, C., M. Eichenbaum, I. Kleshchelski, and S. Rebelo (2011). “Do Peso Problems
Explain the Returns to the Carry Trade?”Review of Financial Studies 24, 853–891.
Buss, A. and G. Vilkov (2012). “Measuring Equity Risk with Option-Implied Correlations,”
Review of Financial Studies 25, 3113-3140.
Campa, J.M., P.H.K. Chang, and R.L. Reider (1998). “Implied Exchange Rate Distributions:
Evidence from OTC Option Markets,”Journal of International Money and Finance 17, 117–
160.
36

Carr, P., and R. Lee (2009). “Volatility Derivatives,”Annual Review of Financial Economics
1, 1-21.
Carr, P., and L. Wu (2009). “Variance Risk Premiums,” Review of Financial Studies 22,
1311–1341.
Castagna, A., and F. Mercurio (2007). “The Vanna-Volga Method for Implied Volatilities,”
Risk January, 106–111.
Chinn, M.D. and H. Ito (2006). “What Matters for Financial Development? Capital Controls,
Institutions, and Interactions,”Journal of Development Economics 81, 163–192.
Cochrane, J.H. (2005). Asset Pricing. Princeton: Princeton University Press.
Della Corte, P., L. Sarno, and I. Tsiakas (2009). “An Economic Evaluation of Empirical
Exchange Rate Models,”Review of Financial Studies 22, 3491-3530.
Della Corte, P., L. Sarno, and I. Tsiakas (2011). “Spot and Forward Volatility in Foreign
Exchange,”Journal of Financial Economics 100, 496–513.
Demeter…, K., E. Derman, M. Kamal, and J. Zou (1999). “A Guide to Volatility and Variance
Swaps,”Journal of Derivatives 6, 9–32.
Drechsler, I. and A. Yaron (2011). “What’s Vol Got to Do with It,” Review of Financial
Studies 24, 1–45.
Easley, D., M. O’Hara and P. S. Srinivas, (1998) “Option Volume and Stock Prices: Evidence
on Where Informed Traders Trade,” Journal of Finance, 53, 431-465.
Engel, C., N.C. Mark, and K.D. West (2008). “Exchange Rate Models are not as Bad as you
Think,”NBER Macroeconomics Annual 2007, 381–441.
Eraker, B. (2010). “The Volatility Premium,”Working Paper, Wisconsin School of Business.
Evans, M.D.D., and R.K. Lyons (2005). “Meese-Rogo¤ Redux: Micro-Based Exchange Rate
Forecasting,” American Economic Review Papers and Proceedings 95, 405–414.
37

Fama, E.F., and J.D. MacBeth (1973). “Risk, Return, and Equilibrium: Empirical Tests,”
Journal of Political Economy 81, 607–36.
Froot, K.A., and T. Ramadorai (2005). “Currency Returns, Intrinsic Value, and Institutional
Investor Flows,”Journal of Finance 60, 1535–1566.
Fung, W., and D.A. Hsieh (1997). “Empirical Characteristics of Dynamic Trading Strategies:
the Case of Hedge Funds,”Review of Financial Studies 10, 275–302.
Fung, W., and D.A. Hsieh (2004). “Hedge Fund Benchmarks: A Risk Based Approach,”
Financial Analyst Journal 60, 65–80.
Gârleanu, N. B., and L.H. Pedersen (2011). “Margin-Based Asset Pricing and the Law of One
Price,”Review of Financial Studes 24, 1980–2022.
Gârleanu, N. B., L.H. Pedersen, and A. Poteshman (2009). “Demand-Based Option Pricing,”
Review of Financial Studies 22, 4259–4299.
Garman, M.B., and S.W. Kohlhagen (1983). “Foreign Currency Option Values,” Journal of
International Money and Finance 2, 231–237.
Goyal, A. and A. Saretto (2009). “Cross-section of Option Returns and Volatility,” Journal
of Financial Economics 94, 310-326.
Gromb, D., and D. Vayanos (2010). “Limits of Arbitrage: The State of the Theory,” Annual
Review of Financial Economics 2, 251–275.
Han B., and Y. Zhou (2011). “Variance Risk Premium and Cross-Section of Stock Returns,”
Working Paper, University of Texas at Austin.
Hansen L.P. (1982). “Large Sample Properties of Generalized Method of Moments Estima-
tors,”Econometrica 50, 1029–54.
Hansen, L.P., and R. Jagannathan (1997). “Assessing Speci…cation Errors in Stochastic Dis-
count Factor Model,”Journal of Finance 52, 557–590.
Hassan, T., and R, Mano (2014). “Forward and Spot Exchange Rates in a Multi-Currency
World,”Working Paper, University of Chicago.
38

International Swaps and Derivatives Association, Emerging Markets Traders Association, and
Foreign Exchange Committee (1998). 1998 Foreign Exchange and Currency Option De…ni-
tions. New York: International Swaps and Derivatives Association, Inc.
Jiang, G.J., and Y.S. Tian (2005). “The Model-Free Implied Volatility and Its Information
Content,”Review of Financial Studies 18, 1305–1342.
Kaminsky, G., and S. Schmukler (2008). “Short-Run Pain, Long-Run Gain: Financial Liber-
alization and Stock Market Cycles,”Review of Finance 12, 253-292.
Klitgaard, T., and L. Weir (2004). “Exchange Rate Changes and Net Positions of Speculators
in the Futures Market,” Federal Reserve Bank of New York Economic Policy Review 10,
17–28.
Knauf, S. (2003). “Making Money from FX Volatility,”Quantitative Finance 3, 48–51.
Lewellen, J., S. Nagel, and J. Shanken (2010). “A Skeptical Appraisal of Asset Pricing Tests,”
Journal of Financial Economics 96, 175–194.
Londono, J.M., and H. Zhou (2012). “Variance Risk Premiums and the Forward Premium
Puzzle,”Working Paper, Federal Reserve Board.
Lustig, H., N. Roussanov, and A. Verdelhan (2011). “Common Risk Factors in Currency
Markets,”Review of Financial Studies 24, 3731–3777.
Lustig, H., and A. Verdelhan (2007). “The Cross Section of Foreign Currency Risk Premia
and US Consumption Growth Risk,”American Economic Review 97, 89–117.
Martin, I. (2012). “Simple Variance Swaps,”Working Paper, Stanford University.
Meese, R.A., and K. Rogo¤ (1983). “Empirical Exchange Rate Models of the Seventies: Do
They Fit Out of Sample?”Journal of International Economics 14, 3–24.
Menkho¤, L., L. Sarno, M. Schmeling, and A. Schrimpf (2012a). “Carry Trades and Global
FX Volatility,”Journal of Finance 67, 681–718.
Menkho¤, L., L. Sarno, M. Schmeling, and A. Schrimpf (2012b). “Currency Momentum Strate-
gies,”Journal of Financial Economics 106, 620–684.
39

Mueller, P., Y.M. Yen, and A. Vedolin (2011). “Bond Variance Risk Premiums,” Working
Paper, London School of Economics.
Nagel, S. (2012), “Evaporating Liquidity,”Review of Financial Studies 25, 2005–2039.
Neuberger, A. (1994). “The Log Contract: A New Instrument to Hedge Volatility,” Journal
of Portfolio Management 20, 74–80.
Newey, W.K., and K.D. West (1987). “A Simple, Positive Semi-De…nite, Heteroskedasticity
and Autocorrelation Consistent Covariance Matrix,”Econometrica 55, 703–708.
Pan, J., and A. M. Poteshman (2006). “The Information in Option Volume for Future Stock
Prices,”Review of Financial Studies 19, 871–908.
Patton, A., and T. Ramadorai (2013). “On the High-frequency Dynamics of Hedge Fund Risk
Exposures,”Journal of Finance 68, 597–635.
Ramadorai, T. (2013). “Capacity Constraints, Investor Information, and Hedge Fund Re-
turns,”Journal of Financial Economics 107, 401–416.
Rime, D., L. Sarno and E. Sojli (2010). “Exchange Rates, Order Flow and Macroeconomic
Information,”Journal of International Economics 80, 72–88.
Todorov, V. (2010). “Variance Risk Premium Dynamics: The Role of Jumps,” Review of
Financial Studies 23, 345–383.
40

Table 1. Volatility Risk Premia
This table presents summary statistics for the 1-year volatility risk premium (V RPt ) de…ned as di¤erence
between the realized volatility (RVt ) and the synthetic volatility swap rate (SWt ). RVt is calculated using
daily exchange rate returns over the previous year. SWt is computed as in Britten-Jones and Neuberger
(2000) using 1-year currency option implied volatilities. Qj refers to the j th percentile. AC indicates the
1-year autocorrelation coe¢ cient. V RPt , RVt , and SWt are expressed in percent per annum, and averaged
across two sets of currencies. The sample period comprises daily data from January 1998 to December 2013.
V RPt RVt SWt V RPt RVt SWt

Developed Developed & Emerging
M ean 0:78 10:90 11:68 1:15 10:96 12:11
Sdev 1:64 2:65 2:71 1:90 2:96 3:25
Skew 0:25 2:07 1:32 0:36 2:23 1:80
Kurt 5:48 7:48 4:78 6:45 7:97 6:54
Q5 3:50 8:35 8:50 4:33 8:41 9:04
Q95 1:56 18:08 16:85 1:44 19:00 18:53
AC 0:07 0:25 0:47 0:06 0:22 0:43
41

Table 2. Volatility Risk Premia Portfolios
This table presents descriptive statistics of currency portfolios sorted on the 1-year volatility risk premia at
time t 1. The volatility risk premium is de…ned as di¤erence between the realized volatility and the synthetic
volatility swap rate both computed at time t 1. The long (short) portfolio PL (PS ) contains the top 20% of
all currencies with the highest (lowest) volatility risk premium. V RP denotes a long-short strategy that buys
PL and sells PS . The table also reports the …rst-order autocorrelation coe¢ cient (AC), the annualized Sharpe
ratio (SR), and the frequency of portfolio switches (F req). Panel A displays the overall excess return, whereas
Panel B reports the exchange rate component only. Panel C presents the transition probability from portfolio
i to portfolio j between time t and time t + 1. indicates the steady state probability. The superscripts , ,
and indicate statistical signi…cance for the mean at 10%, 5%, and 1%, respectively, based on Newey and
West (1987) and Andrews (1991). Returns are expressed in percentage per annum and adjusted for transaction
costs. The strategies are rebalanced monthly from January 1998 to December 2013.
Panel A: Excess Returns

PL P2 P3 P4 PS V RP PL P2 P3 P4 PS V RP
M ean 4:59 3:12 1:10 3:27 0:36 4:95 4:71 2:71 1:05 2:40 0:55 4:16
Sdev 9:57 9:55 9:62 10:15 10:04 8:15 10:16 8:93 9:09 10:62 8:61 8:14
Skew 0:20 0:05 0:08 0:22 0:22 0:03 0:23 0:44 0:32 0:54 0:18 0:01
Kurt 3:56 5:16 5:52 3:95 3:96 3:97 3:94 5:83 3:57 4:76 4:81 4:54
SR 0:48 0:33 0:11 0:32 0:04 0:61 0:46 0:30 0:12 0:23 0:06 0:51
AC 0:04 0:01 0:05 0:14 0:01 0:05 0:04 0:08 0:11 0:07 0:03 0:02
F req 0:29 0:48 0:54 0:52 0:35 0:35 0:28 0:43 0:50 0:49 0:27 0:27
Panel B: FX Returns
M ean 4:60 2:87 1:21 3:00 0:85 5:45 4:45 2:30 1:04 1:72 0:82 5:27
Sdev 9:58 9:52 9:53 10:08 10:00 8:12 10:17 8:90 9:00 10:53 8:61 8:20
Skew 0:25 0:10 0:10 0:23 0:24 0:03 0:29 0:49 0:35 0:59 0:27 0:09
Kurt 3:57 5:29 5:56 4:12 3:95 4:04 3:99 5:84 3:69 5:02 4:93 5:10
SR 0:48 0:30 0:13 0:30 0:08 0:67 0:44 0:26 0:12 0:16 0:10 0:64
AC 0:03 0:02 0:04 0:13 0:01 0:04 0:03 0:08 0:09 0:06 0:04 0:02
F req 0:29 0:48 0:54 0:52 0:35 0:35 0:28 0:43 0:50 0:49 0:27 0:27
Panel C: Transition Matrix
PL 0:71 0:21 0:06 0:01 0:01 0:72 0:22 0:05 0:01 0:00
P2 0:20 0:53 0:18 0:06 0:02 0:18 0:58 0:17 0:05 0:02
P3 0:05 0:20 0:46 0:20 0:08 0:02 0:21 0:51 0:22 0:04
P4 0:01 0:04 0:22 0:48 0:24 0:01 0:07 0:22 0:51 0:19
PS 0:00 0:03 0:07 0:25 0:65 0:00 0:02 0:04 0:21 0:72
0:19 0:20 0:20 0:20 0:20 0:17 0:24 0:21 0:20 0:18
42

Table 3. Currency Strategies
This table presents descriptive statistics of currency strategies formed using time t 1 information. CAR
is the carry trade strategy that buys (sells) the top 20% of all currencies with the highest (lowest) interest
rate di¤erential relative to the US dollar. Similarly, M OM is the momentum strategy that buys (sells)
currencies with the highest (lowest) past 3-month exchange rate return, V AL is the value strategy that buys
(sells) currencies with lowest (highest) real exchange rate, RR is the risk reversal strategy that buys (sells)
currencies with the lowest (highest) 1-year 10-delta risk reversal, and V RP is the volatility risk premium
strategy that buys (sells) currencies with the highest (lowest) 1-year volatility risk premium. The table also
reports …rst order autocorrelation coe¢ cient (AC), the annualized Sharpe ratio (SR), the Sortino ratio (SO),
the percentage maximum drawdown (M DD), the frequency of portfolio switches for the long (F reqL ) and
the short (F reqS ) position. Panel A displays the overall currency excess return whereas Panel B reports the
exchange rate return component only. Panel C presents the sample correlations of the currency excess returns.
The superscripts , , and indicate statistical signi…cance for the mean at 10%, 5%, and 1%, respectively,
based on Newey and West (1987) and Andrews (1991). Returns are expressed in percentage per annum and
adjusted for transaction costs. The strategies are rebalanced monthly from January 1998 to December 2013.

CAR M OM V AL RR V RP CAR M OM V AL RR V RP
M ean 4:10 0:92 3:66 5:10 4:95 4:90 0:19 2:30 4:25 4:16
Sdev 10:73 9:81 8:97 11:49 8:15 9:25 8:17 8:23 10:20 8:14
Skew 0:71 0:26 0:16 0:47 0:03 0:65 0:07 0:47 0:52 0:01
Kurt 5:25 3:75 3:71 5:41 3:97 4:21 3:81 5:08 5:26 4:54
SR 0:38 0:09 0:41 0:44 0:61 0:53 0:02 0:28 0:42 0:51
SO 0:49 0:16 0:64 0:62 0:93 0:74 0:04 0:40 0:53 0:76
M DD 37:8 22:8 15:1 35:1 17:0 28:2 18:8 15:1 31:5 24:4
AC 0:08 0:02 0:02 0:08 0:05 0:05 0:10 0:09 0:09 0:02
F reqL 0:10 0:48 0:09 0:08 0:29 0:14 0:51 0:08 0:16 0:28
F reqS 0:07 0:44 0:07 0:22 0:35 0:16 0:47 0:06 0:21 0:27
Panel B: FX Returns
M ean 0:81 0:94 2:15 1:84 5:45 1:61 0:35 0:48 0:21 5:27
Sdev 10:76 9:87 9:02 11:58 8:12 9:29 8:18 8:27 10:27 8:20
Skew 0:72 0:33 0:24 0:50 0:03 0:72 0:09 0:55 0:55 0:09
Kurt 5:43 3:94 3:76 5:63 4:04 4:35 4:06 5:29 5:65 5:10
SR 0:08 0:09 0:24 0:16 0:67 0:17 0:04 0:06 0:02 0:64
SO 0:10 0:17 0:36 0:22 1:01 0:23 0:07 0:08 0:03 0:97
M DD 43:3 23:2 22:5 40:3 14:5 37:3 18:2 20:9 38:0 17:9
AC 0:09 0:01 0:02 0:09 0:04 0:08 0:10 0:08 0:11 0:02
F reqL 0:10 0:48 0:09 0:08 0:29 0:14 0:51 0:08 0:16 0:28
F reqS 0:07 0:44 0:07 0:22 0:35 0:16 0:47 0:06 0:21 0:27
Panel C: Correlations
CAR 1:00 0:20 0:30 0:76 0:08 1:00 0:07 0:32 0:59 0:06
M OM 0:20 1:00 0:20 0:23 0:11 0:07 1:00 0:19 0:17 0:15
V AL 0:30 0:20 1:00 0:46 0:19 0:32 0:19 1:00 0:57 0:04
RR 0:76 0:23 0:46 1:00 0:10 0:59 0:17 0:57 1:00 0:09
V RP 0:08 0:11 0:19 0:10 1:00 0:06 0:15 0:04 0:09 1:00
43

Table 4. Currency Strategies: Sub-Samples
This table presents descriptive statistics of the foreign exchange return component to currency strategies
formed using time t 1 information. CAR is the carry trade strategy that buys (sells) the top 20% of all
currencies with the highest (lowest) interest rate di¤erential relative to the US dollar. Similarly, M OM is the
momentum strategy that buys (sells) currencies with the highest (lowest) past 3-month exchange rate return,
V AL is the value strategy that buys (sells) currencies with lowest (highest) real exchange rate, RR is the risk
reversal strategy that buys (sells) currencies with the lowest (highest) 1-year 10-delta risk reversal, and V RP
is the volatility risk premium strategy that buys (sells) currencies with the highest (lowest) 1-year volatility
risk premium. The table also reports …rst order autocorrelation coe¢ cient (AC) and the annualized Sharpe
ratio (SR). The superscripts , , and indicate statistical signi…cance for the mean at 10%, 5%, and
1%, respectively, based on Newey and West (1987) and Andrews (1991). Returns are expressed in percentage
per annum and adjusted for transaction costs. The strategies are rebalanced monthly from March 2001 to
November 2001, and from December 2007 to June 2009 in Panel A, from January 1998 to December 2006 in
Panel C, and from January 2007 to December 2013 in Panel D. Panel E reports the p-values in brackets of
the null hypothesis for equal means across sub-samples.
Panel A: NBER Recessions

M ean 9:63 11:01 4:58 5:96 9:32 11:72 7:83 0:06 8:78 10:65
Sdev 17:12 15:41 12:03 17:80 10:88 13:31 10:76 10:62 14:84 10:59
Skew 0:44 0:28 0:63 0:79 0:43 0:79 0:62 1:72 1:14 0:63
Kurt 3:71 2:87 3:43 4:45 3:61 2:68 4:23 7:87 4:04 5:01
SR 0:56 0:71 0:38 0:33 0:86 0:88 0:73 0:01 0:59 1:01
AC 0:35 0:12 0:09 0:38 0:03 0:44 0:03 0:24 0:38 0:05
Panel B: non-NBER Recessions
M ean 0:70 0:78 1:74 3:18 4:79 0:12 0:92 0:58 1:75 4:35
Sdev 9:26 8:56 8:45 10:18 7:57 8:37 7:64 7:84 9:26 7:73
Skew 0:54 0:06 0:08 0:06 0:09 0:37 0:28 0:05 0:10 0:22
Kurt 3:96 2:57 3:55 3:59 3:87 4:02 3:05 3:37 4:65 4:41
SR 0:08 0:09 0:21 0:31 0:63 0:01 0:12 0:07 0:19 0:56
AC 0:08 0:09 0:01 0:07 0:05 0:11 0:16 0:04 0:03 0:06
Panel C: Pre-Crisis
M ean 0:78 0:11 1:76 3:41 4:54 0:55 0:64 0:30 2:58 5:28
Sdev 8:22 7:96 9:94 9:96 7:45 8:33 7:37 8:81 9:61 7:90
Skew 0:80 0:04 0:26 0:32 0:30 0:79 0:02 0:08 0:33 0:27
Kurt 5:05 2:50 3:24 3:87 3:99 4:76 2:89 3:03 4:36 3:44
SR 0:09 0:01 0:18 0:34 0:61 0:07 0:09 0:03 0:27 0:67
AC 0:09 0:15 0:03 0:00 0:04 0:10 0:17 0:03 0:04 0:06
Panel D: Post-Crisis
M ean 2:85 2:28 2:65 0:17 6:62 4:37 1:63 0:72 2:83 5:26
Sdev 13:37 11:93 7:75 13:42 8:93 10:38 9:15 7:58 11:06 8:62
Skew 0:55 0:39 0:12 0:84 0:31 0:59 0:13 1:47 1:25 0:08
Kurt 4:27 3:58 4:65 5:53 3:95 3:79 4:40 10:27 6:00 6:53
SR 0:21 0:19 0:34 0:01 0:74 0:42 0:18 0:10 0:26 0:61
AC 0:18 0:07 0:01 0:15 0:11 0:22 0:04 0:17 0:16 0:02
44

Table 5. Static Currency Strategies
This table presents descriptive statistics of static currency strategies. CAR is the carry trade strategy
that buys (sells) the top 20% of all currencies with the highest (lowest) full sample average interest rate
di¤erential relative to the US dollar. Similarly, M OM is the momentum strategy that buys (sells) currencies
with the highest (lowest) full sample average 3-month exchange rate return, V AL is the value strategy that
buys (sells) currencies with lowest (highest) full sample average real exchange rate, RR is the risk reversal
strategy that buys (sells) currencies with the lowest (highest) full sample average 1-year 10-delta risk reversal,
and V RP is the volatility risk premium strategy that buys (sells) currencies with the highest (lowest) full
sample average 1-year volatility risk premium. The table also reports …rst order autocorrelation coe¢ cient
(AC), the annualized Sharpe ratio (SR), the Sortino ratio (SO), the percentage maximum drawdown (M DD),
the frequency of portfolio switches for the long (F reqL ) and the short (F reqS ) position. Panel A displays
the overall currency excess return whereas Panel B reports the exchange rate return component only. Panel
C presents the sample correlations of the currency excess returns. The superscripts , , and indicate
statistical signi…cance for the mean at 10%, 5%, and 1%, respectively, based on Newey and West (1987) and
Andrews (1991). Returns are expressed in percentage per annum. The sample runs monthly from January
1998 to December 2013.

M ean 4:62 2:58 1:39 4:62 3:51 4:44 3:99 1:11 4:78 3:28
Sdev 11:52 6:88 6:55 11:52 9:50 10:46 6:56 5:55 9:94 8:53
Skew 0:66 0:51 0:07 0:66 0:35 0:63 0:31 0:62 0:69 0:16
Kurt 5:03 7:33 4:24 5:03 3:07 4:00 3:87 4:44 4:54 4:47
SR 0:40 0:38 0:21 0:40 0:37 0:42 0:61 0:20 0:48 0:38
SO 0:55 0:60 0:31 0:55 0:55 0:61 1:12 0:38 0:67 0:55
M DD 0:37 0:12 0:31 0:37 0:21 0:29 0:12 0:31 0:30 0:18
AC 0:07 0:07 0:11 0:07 0:07 0:10 0:02 0:10 0:11 0:09
Panel B: FX Returns
M ean 0:25 2:48 1:51 0:25 0:65 1:80 4:81 0:88 1:07 2:92
Sdev 11:58 6:91 6:55 11:58 9:56 10:49 6:59 5:54 10:00 8:60
Skew 0:66 0:45 0:05 0:66 0:33 0:65 0:31 0:64 0:72 0:08
Kurt 5:12 7:27 4:27 5:12 3:03 4:04 3:88 4:62 4:59 4:66
SR 0:02 0:36 0:23 0:02 0:07 0:17 0:73 0:16 0:11 0:34
SO 0:03 0:56 0:34 0:03 0:10 0:24 1:35 0:30 0:15 0:49
M DD 0:43 0:12 0:31 0:43 0:27 0:41 0:11 0:23 0:36 0:20
AC 0:08 0:07 0:09 0:08 0:08 0:11 0:02 0:08 0:12 0:08
CAR 1:00 0:17 0:25 1:00 0:73 1:00 0:18 0:26 0:96 0:29
M OM 0:17 1:00 0:72 0:17 0:33 0:18 1:00 0:15 0:07 0:20
V AL 0:25 0:72 1:00 0:25 0:57 0:26 0:15 1:00 0:27 0:53
RR 1:00 0:17 0:25 1:00 0:73 0:96 0:07 0:27 1:00 0:27
V RP 0:73 0:33 0:57 0:73 1:00 0:29 0:20 0:53 0:27 1:00
45

Table 6. Risk Factors and Volatility Risk Premium Strategy: Time Series Tests
This table presents time-series regression estimates. The dependent variable is the volatility risk premium strategy (V RP ) that buys (sells) currencies
with the highest (lowest) 1-year volatility risk premium. Panel A uses the currency strategies described in Table 3 as explanatory variables. Panel B
employes the Fama and French (1992) and the equity momentum factors whereas Panel C uses the Fung and Hsieh (2004) factors. The superscripts , ,
and indicate statistical signi…cance at 10%, 5%, and 1%, respectively, based on Newey and West (1987) and Andrews (1991). Returns are annualized
and adjusted for transaction costs (except the equity and the hedge fund factors). The strategies are rebalanced monthly from January 1998 to December
2013. Fama and French (1992) factors are from French’s website whereas the Fung and Hsieh (2004) factors are from Hsieh’s website.
Panel A: Currency Factors

DOL CAR M OM V AL RR R2
Developed 0:04 0:06 0:24 0:12 0:14 0:23 0:08
Developed & Emerging 0:04 0:23 0:22 0:18 0:02 0:14 0:08
Panel B: Equity Factors
e
Rm SM B HM L M OM E R2
Developed 0:06 0:06 0:02 0:07 0:05 0:01
Developed & Emerging 0:05 0:04 0:08 0:07 0:06 0:02
46
Panel C: Hedge Fund Factors
Bond Curr Comm Equity Size Bond Credit
T rend T rend T rend M arket Spread M arket Spread R2
Developed 0:05 < :01 < :01 < :01 0:04 0:04 0:05 0:02 0:03
Developed & Emerging 0:04 0:15 0:04 0:10 0:01 0:10 0:20 0:17 0:02

Table 7. Asset Pricing Tests
This table reports asset pricing tests for a linear factor model that includes the dollar (DOL), the carry trade (CAR), and the global volatility (V OL)
factors. DOL is equivalent to a strategy that borrows in the US money market and equally invests in all foreign currencies, and serves as a constant in
the cross-section. CAR is a long-short strategy that buys (sells) the top 20% of all currencies currencies with the highest (lowest) interest rate di¤erential
relative to the US dollar. V OL is computed as the innovations to a …rst order autoregressive process applied to the average foreign exchange rate volatility.
The test assets are excess returns to …ve currency portfolios sorted on the 1-year volatility risk premium at time t 1. Panel A reports GMM and
Fama-MacBeth (FMB) estimates of the market price of risk , and the Hansen-Jagannathan distance HJ test for the null hypothesis that the pricing
errors are jointly zero. Panel B reports least-squares estimates of time series regressions and the 2 test for the null that all intercepts are jointly zero.
The superscripts , , and indicate statistical signi…cance at 10%, 5%, and 1%, respectively, based on Newey and West (1987) and Andrews (1991)
for GMM estimates, and Shanken (1992) for FMB estimates. Returns are annualized and adjusted for transaction costs. The portfolios are rebalanced
monthly from January 1998 to September 2013.
Panel A: Cross-Section
DOL CAR R2 HJ DOL CAR R2 HJ
GM M1 0:02 0:10 0:19 0:18 0:02 0:07 0:24 0:16
GM M2 0:02 0:07 0:16 0:02 0:04 0:19
47
F MB 0:02 0:10 0:19 0:02 0:07 0:24
DOL V OL R2 HJ DOL V OL R2 HJ
GM M1 0:02 0:07 0:24 0:17 0:02 0:01 0:14 0:16
GM M2 0:02 0:07 0:24 0:02 0:02 0:14
F MB 0:02 0:07 0:24 0:02 0:01 0:14
Panel B: Time-Series
2 2
DOL CAR R2 DOL CAR R2
PL 0:02 0:94 0:01 0:69 7:94 0:02 1:06 0:02 0:75 7:77
P2 0:01 0:99 0:01 0:75 0:00 0:95 0:02 0:81
P3 0:01 0:92 0:08 0:72 0:01 0:99 0:02 0:82
P4 0:01 1:14 0:14 0:82 0:00 1:19 0:09 0:84

PS 0:03 1:01 0:05 0:76 0:02 0:82 0:11 0:72
2 2
DOL V OL R2 DOL V OL R2
PL 0:02 0:95 0:13 0:69 7:90 0:02 1:05 0:04 0:75 5:60
P2 0:01 0:97 0:15 0:76 0:01 0:94 0:17 0:81
P3 0:01 0:96 0:09 0:71 0:01 0:99 0:04 0:82
P4 0:01 1:08 0:07 0:80 0:00 1:17 0:11 0:84
PS 0:03 1:02 0:15 0:75 0:01 0:86 0:06 0:71
Table 8. -Sorted Portfolios: Average Volatility Risk Premia
This table presents descriptive statistics of -sorted currency portfolios. Each is obtained by regressing
individual currency excess returns on the average volatility risk premia using a 36-month moving window. The
long (short) portfolio PL (PS ) contains the top 20% of all currencies with the lowest (highest) . The table also
reports the …rst order autocorrelation coe¢ cient (AC), the annualized Sharpe ratio (SR), and the frequency of
portfolio switches (F req). Panel A displays the overall excess return, whereas Panel B reports the exchange
rate component only. Panel C presents the pre- and post-formation s, and the pre- and post-formation
interest rate di¤erential (if ) relative to the US dollar. The superscripts , , and indicate statistical
signi…cance at 10%, 5%, and 1%, respectively, based on Newey and West (1987) and Andrews (1991). Returns
are expressed in percentage per annum and adjusted for transaction costs. The sample runs from January

PL P2 P3 P4 PS PL -PS PL P2 P3 P4 PS PL -PS
M ean 3:84 2:20 2:87 3:12 8:85 5:01 3:41 3:11 3:70 2:87 7:84 4:44
Sdev 9:13 10:51 9:28 10:34 11:94 10:69 8:16 9:62 9:64 10:26 12:23 10:82
Skew 0:35 0:11 0:65 0:27 0:50 0:87 0:08 0:19 0:50 0:55 0:84 1:06
Kurt 3:42 4:50 5:13 4:55 5:31 7:32 2:60 5:13 4:60 4:63 5:87 6:79
SR 0:42 0:21 0:31 0:30 0:74 0:47 0:42 0:32 0:38 0:28 0:64 0:41
AC 0:06 0:02 0:17 0:01 0:01 0:06 0:07 0:01 0:09 0:02 0:00 0:03
F req 0:16 0:25 0:28 0:27 0:11 0:11 0:16 0:22 0:26 0:22 0:12 0:12
Panel B: FX Returns
M ean 4:62 2:35 2:51 2:39 6:19 1:57 4:35 3:19 3:19 1:21 5:06 0:71
Sdev 9:13 10:47 9:32 10:30 11:95 10:79 8:17 9:60 9:64 10:18 12:21 10:87
Skew 0:37 0:12 0:67 0:28 0:52 0:95 0:08 0:17 0:51 0:60 0:90 1:16
Kurt 3:48 4:45 5:14 4:58 5:32 7:52 2:59 5:15 4:61 4:68 5:97 7:10
SR 0:51 0:22 0:27 0:23 0:52 0:15 0:53 0:33 0:33 0:12 0:41 0:07
AC 0:05 0:03 0:17 0:01 0:01 0:07 0:06 0:01 0:09 0:01 0:01 0:03
F req 0:16 0:25 0:28 0:27 0:11 0:11 0:16 0:22 0:26 0:22 0:12 0:12
Panel C: Portfolio Formation
pre-if 0:54 0:06 0:56 0:99 2:35 0:71 0:14 0:72 1:92 2:37
post-if 0:54 0:05 0:54 0:97 2:29 0:74 0:14 0:71 1:87 2:32
pre- 0:49 0:18 0:05 0:30 0:71 0:47 0:20 0:07 0:36 0:81
post- 0:29 0:13 0:22 0:04 0:16 0:21 0:18 0:06 0:05 0:12
48

Table 9. Net Demand Pressure and Currency Volatility Risk Premia
i
This table presents …xed e¤ects panel estimates of V RP t = i + N Demit lag + uit where V RP it is the 1-year volatility risk premium for
i
currency i whereas N Demt denotes the net demand of currency options and futures for two groups of end-users from the US Commodity Futures Trading
Commission (CFTC). The net demand is constructed as di¤erence between long and short positions scaled by the total open interest. V RP is de…ned
as the di¤erence between the realized volatility (RVt ) and the synthetic volatility swap rate (SWt ). RV is calculated using daily exchange rate returns
over the previous year whereas SW is computed as in Britten-Jones and Neuberger (2000) using 1-year currency option implied volatilities. In V RPsi ,
SW is computed using the simple variance swap method of Martin (2012). In V RPgarch , RV is the 1-year volatility forecast generated from the simple
GARCH(1,1). In V RPsv , RV is the 1-year volatility forecast generated from a stochastic volatility model. The superscripts , , and indicate statistical
signi…cance at 10%, 5%, and 1%, respectively, based on Newey and West (1987) and Andrews (1991). Monthly CFTC data are collected on the last Tuesday
of every month. All other variables are measured on the same day. The sample runs from January 1998 to December 2013.
V RP V RPsi
lag R2 R2 R2 R2
Commercial Financial Commercial Financial
0 0:011 0:016 0:025 0:011 0:021 0:025 0:010 0:012 0:017 0:009 0:016 0:016
1 0:011 0:011 0:012 0:011 0:013 0:009 0:009 0:008 0:007 0:009 0:009 0:005
49
2 0:010 0:001 < :001 0:010 0:003 < :001 0:009 < :001 < :001 0:009 0:004 0:001
V RPgarch V RPsv
lag R2 R2 R2 R2
Commercial Financial Commercial Financial
0 0:007 0:003 0:019 0:007 0:004 0:014 0:007 0:003 0:019 0:007 0:004 0:014
1 0:007 0:003 0:013 0:007 0:003 0:008 0:007 0:003 0:013 0:007 0:003 0:008
2 0:007 0:002 0:004 0:007 0:001 0:001 0:007 0:002 0:004 0:007 0:001 0:001

Table 10. Arbitrage Risk Proxies and V RP
This table presents predictive regressions estimates. The dependent variable is the exchange rate return component of the V RP strategy at time t.
This strategy is a long/short portfolio that buys (sells) the top 20% of all currencies with the highest (lowest) 1-year volatility risk premia at time t 1.
The set of predictors is measured at time t 1, and includes the T ED spread, the V IX index, the St. Louis Fed Financial Stress Index F SI, and the
F und F lows of currency and global macro funds constructed as the Asset under Management (AUM) weighted net ‡ows scaled by the lagged AUM as in
Patton and Ramadorai (2013). denotes the …rst-di¤erence operator and T ED, V IX, and F SI are averaged on a 12-month rolling. The superscripts
, , and indicate statistical signi…cance at 10%, 5%, and 1%, respectively, based on Newey and West (1987) and Andrews (1991). The exchange rate
returns are annualized. The sample runs from January 1998 to December 2013.
T ED F und T ED F und
T ED V IX F SI V IX F lows R2 T ED V IX F SI V IX F lows R2
0:01 0:12 0:02 0:02 0:14 0:02
0:06 0:03 0:01 0:06 0:07 0:03
0:06 0:24 < :01 0:06 0:50 0:03
50
0:05 0:07 0:03 0:04 0:10 0:07
0:07 1:59 0:02 0:06 1:31 0:01
0:02 0:09 1:21 0:03 0:00 0:11 0:83 0:03
0:07 0:02 1:42 0:02 0:06 0:06 0:89 0:03
0:07 0:16 1:45 0:02 0:06 0:44 0:94 0:04
0:06 0:06 1:20 0:04 0:05 0:10 0:65 0:07
0:04 0:02 0:06 0:11 1:20 0:04 0:04 0:00 0:04 0:14 0:71 0:06

Table 11. VAR-GARCH, Currency Option and Spot Markets
1=2 0
This table reports selected parameter estimates of the following VAR-GARCH speci…cation: yt = ct + ut , with ut = Ht "t where yt = (et ; vt ) , ct
is the conditional mean modelled as a Vector Autoregressive Process (VAR) of order p, and Ht is the conditional covariance whose elements are modelled
as
h 0 h :
2 3 2 3 2 32 3 2 32 3
4 hh 5 = 4 ! h 5 + 4 0 5 4 hh 5 +4 0 0 5 4 ue uv 5
We estimate the above speci…cation via maximum likelihood for each volatility risk premium portfolio described in Table 2: et denotes the average 1-month
exchange rate return whereas vt is the average 1-year volatility risk premium. The superscripts , , and indicate statistical signi…cance at 10%, 5%,
and 1%, respectively, based on Newey and West (1987) and Andrews (1991). The portfolios are rebalanced daily using t 21 information. Exchange rate
returns are annualized. The sample runs from January 1998 to December 2013:
ev ve ev ve ev ve ev ve
PL 0:12 0:05 0:35 0:04 0:15 0:09 0:31 0:10
51
P2 0:15 0:05 0:25 0:02 0:11 0:14 0:23 0:08
P3 0:25 0:24 0:29 0:05 0:03 0:19 0:09 0:02
P4 0:12 0:15 0:12 0:05 0:01 0:12 0:04 0:01
PS 0:04 0:02 0:19 0:05 0:09 0:05 0:05 0:11

52
Figure 1. Sharpe Ratios of Currency Strategies
The figure presents for developed countries the 1-year rolling Sharpe ratios of currency strategies formed using t − 1 information. CAR is the carry strategy that buys
(sells) the top 20% of all currencies with the highest (lowest) interest rate differential relative to the US dollar. Similarly, M OM is the momentum strategy that buys
(sells) currencies with the highest (lowest) past 3-month exchange rate return, V AL is the value strategy that buys (sells) currencies with lowest (highest) real exchange
rate, RR is the risk reversal strategy that buys (sells) currencies with the lowest (highest) 1-year 10-delta risk reversal, and V RP is the volatility risk premium strategy
that buys (sells) currencies with the highest (lowest) 1-year volatility risk premium. The Sharpe Ratios are computed using excess returns net of transaction costs. The
strategies are rebalanced monthly from January 1998 to December 2013. Figure A.1 in the Internet Appendix presents the 1-year rolling Sharpe ratios for developed &
emerging countries.
53
Figure 2. Currency Strategies and Payoffs
The figure presents for developed countries the cumulative wealth of the currency strategies described in Figure 1. The strategies are rebalanced monthly from January
1998 to December 2013 and adjusted for transaction costs. Figure A.2 in the Internet Appendix presents the cumulative wealth to currency strategies for developed &
emerging countries.
54
Figure 3. Global Minimum Volatility Portfolios
The figure presents the global minimum volatility portfolio (MVP) and the efficient frontier (solid line) built using the currency strategies described in Figure 1. The
portfolio weights (N × 1) are reported in parentheses and computed as w = (Σ−1 ι)/(ι0 Σ−1 ι) where Σ is the N × N covariance matrix of the strategies’ returns, ι is a
N × 1 vector of ones, and N denotes the number of strategies. The dashed line denotes the efficient frontier that excludes the volatility risk premium (VRP) strategy.
The strategies are rebalanced monthly from January 1998 to December 2013 and adjusted for transaction costs.
55
Figure 4. Net Demand and Volatility Risk Premium Strategy
The figure presents the relation between the volatility risk premium (VRP) strategy and the net demand of currency options and futures from the Commodity Futures
Trading Commission (CFTC). We sort currencies into four baskets using the volatility risk premia at time t, and then compute the average net demand of currency
options and futures at time t. Finally, we cumulate the difference between the first (currencies with the cheapest volatility insurance) and the last (currencies with the
most expensive volatility insurance) portfolio. The net demand is constructed as difference between long and short positions scaled by the total open interest for two
groups of end-users. Commercial traders use the futures market primarily to hedge their business activities whereas financial (or non-commercial) traders use the futures
market for speculative purposes. The data runs from January 1998 to December 2013 at weekly frequency (collected every Tuesday).
Internet Appendix for:
Volatility Risk Premia and Exchange Rate Predictability

(not for publication)
Pasquale Della Corte Tarun Ramadorai Lucio Sarno
July 2014
Pasquale Della Corte is at Imperial College Business School, Imperial College London; email:
p.dellacorte@imperial.ac.uk. Tarun Ramadorai is at the Saïd Business School, Oxford-Man Institute,
University of Oxford and CEPR; email: tarun.ramadorai@sbs.ox.ac.uk. Lucio Sarno is at Cass Business
School, City University London and CEPR; email: lucio.sarno@city.ac.uk.

Table A.1. Country-Speci…c Volatility Risk Premia
This table presents summary statistics for the 1-year volatility risk premium (V RPt ) de…ned as di¤erence between the realized volatility (RVt ) and
the synthetic volatility swap rate (SWt ). RVt is calculated using daily exchange rate returns over the previous year. SWt is computed as in Britten-Jones
and Neuberger (2000) using 1-year currency option implied volatilities. Qj refers to the j th percentile. AC indicates the 1-year autocorrelation coe¢ cient.
V RPt , RVt , and SWt are expressed in percent per annum. The sample period comprises daily data from January 1998 to December 2013.
AU D CAD CHF DKK EU R GBP JP Y N OK N ZD SEK BRL CZK HU F KRW M XN P LN SGD T RY TWD ZAR
Volatility Risk Premium (VRP)
M ean 0:04 0:74 0:42 1:31 1:34 1:22 0:91 0:75 0:35 0:71 2:70 0:98 2:15 2:32 3:34 0:78 1:81 3:73 2:50 2:32
Sdev 2:62 1:40 1:50 1:70 1:74 1:77 1:78 2:00 2:24 2:20 4:46 3:07 3:51 5:20 4:36 3:68 1:66 2:71 1:66 2:80
Skew 1:22 0:56 0:07 0:76 0:81 0:11 0:37 0:52 0:33 0:96 0:31 1:06 0:25 0:14 2:38 0:28 1:29 0:70 1:12 0:05
Kurt 8:13 4:96 3:33 4:68 4:54 6:14 3:47 5:13 4:99 7:01 5:01 5:00 3:81 7:42 11:76 5:19 5:40 4:08 4:89 4:43
Q5 3:87 3:11 2:82 4:44 4:67 3:73 3:63 4:06 3:65 3:89 9:56 5:55 8:64 9:22 11:83 7:03 4:76 7:56 5:15 6:63
Q95 3:84 1:33 2:15 0:62 0:61 1:01 1:30 2:06 4:20 3:34 7:18 6:37 3:45 8:85 0:40 5:50 0:22 2:13 0:32 2:09
AC 0:01 0:20 0:25 0:05 0:02 0:00 0:19 0:07 0:03 0:30 0:00 0:27 0:34 0:10 0:09 0:25 0:11 0:12 0:22 0:09
Realized Volatility (RV)

M ean 12:61 8:37 11:03 10:04 10:02 8:82 11:00 11:85 13:39 11:91 14:83 13:25 17:40 11:35 10:90 16:60 5:66 13:18 4:05 17:20
Sdev 4:38 3:21 2:19 2:12 2:11 2:72 2:75 3:27 3:53 3:65 5:63 4:59 5:24 7:54 4:57 5:91 1:72 4:63 1:02 4:31
1
Skew 2:20 1:26 0:81 0:79 0:79 2:34 1:28 1:96 1:73 2:10 1:60 0:87 0:32 1:65 0:86 0:63 0:57 0:90 0:27 1:28
Kurt 8:01 4:37 3:75 4:72 4:74 8:29 4:03 6:96 6:29 7:29 4:68 3:23 2:33 4:54 3:11 2:59 2:03 3:17 2:24 3:95
Q5 8:55 4:78 7:56 6:52 6:46 6:50 7:88 8:37 9:77 8:49 8:58 7:02 9:94 4:69 5:26 9:03 3:42 6:78 2:50 12:58
Q95 25:34 16:39 15:81 14:92 14:83 16:66 17:63 20:74 22:13 21:44 29:05 23:82 27:91 30:59 21:57 29:19 8:61 23:48 5:79 27:92
AC 0:26 0:59 0:09 0:16 0:10 0:26 0:37 0:32 0:30 0:26 0:04 0:08 0:08 0:06 0:05 0:17 0:03 0:12 0:06 0:07
Synthetic Volatility Swap Rate (SW)

M ean 12:65 9:11 11:45 11:35 11:36 10:04 11:92 12:59 13:74 12:62 17:53 14:23 19:54 13:67 14:24 17:38 7:47 16:91 6:55 19:52
Sdev 3:70 3:28 2:11 2:67 2:68 2:77 2:76 3:07 3:49 3:09 5:14 4:46 5:74 8:01 6:50 6:38 2:54 3:81 2:09 3:84
Skew 1:39 1:42 0:47 0:78 0:77 1:84 0:54 1:23 1:25 1:26 1:92 0:59 0:57 1:63 1:64 0:70 0:70 0:25 1:34 1:54
Kurt 5:01 4:77 3:90 3:82 3:77 6:69 2:84 4:22 4:30 4:17 8:32 3:55 2:80 6:96 7:03 3:30 3:25 3:69 4:89 5:70
Q5 8:22 5:87 7:69 7:18 7:17 7:38 8:07 8:88 9:81 8:94 11:57 7:09 11:44 4:60 7:26 8:46 3:81 9:88 4:17 14:96
Q95 19:99 15:96 15:50 16:06 16:30 15:61 16:87 18:66 21:01 18:73 28:37 22:66 30:75 29:84 28:03 29:76 12:72 23:68 10:56 27:74

AC 0:48 0:66 0:23 0:34 0:34 0:48 0:61 0:48 0:51 0:48 0:08 0:18 0:09 0:07 0:07 0:19 0:15 0:22 0:20 0:05
Table A.2. Currency Strategies: Gross Returns
rate di¤erential relative to the US dollar. Similarly, M OM is the momentum strategy that buys (sells)
currencies with the highest (lowest) past 3-month exchange rate return, V AL is the value strategy that buys
(sells) currencies with lowest (highest) real exchange rate, RR is the risk reversal strategy that buys (sells)
currencies with the lowest (highest) 1-year 10-delta risk reversal, and V RP is the volatility risk premium
strategy that buys (sells) currencies with the highest (lowest) 1-year volatility risk premium. The table also
reports …rst order autocorrelation coe¢ cient (AC), the annualized Sharpe ratio (SR), the Sortino ratio (SO),
the percentage maximum drawdown (M DD), the frequency of portfolio switches for the long (F reqL ) and
the short (F reqS ) position. Panel A displays the overall currency excess return whereas Panel B reports the
exchange rate return component only. Panel C presents the sample correlations of the currency excess returns.
The superscripts , , and indicate statistical signi…cance for the mean at 10%, 5%, and 1%, respectively,
based on Newey and West (1987) and Andrews (1991). Returns are expressed in percentage per annum and
not adjusted for transaction costs. The strategies are rebalanced monthly from January 1998 to December
2013.

M ean 4:82 1:61 4:39 5:81 5:63 5:71 1:02 3:18 5:14 5:01
Sdev 10:73 9:81 8:98 11:50 8:14 9:26 8:17 8:22 10:19 8:14
Skew 0:71 0:26 0:16 0:47 0:02 0:65 0:07 0:45 0:51 0:02
Kurt 5:24 3:75 3:71 5:42 3:96 4:21 3:80 5:00 5:25 4:56
SR 0:45 0:16 0:49 0:51 0:69 0:62 0:13 0:39 0:50 0:62
SO 0:57 0:29 0:77 0:71 1:06 0:87 0:20 0:56 0:64 0:92
M DD 37:0 21:8 13:9 34:5 14:3 27:8 17:3 13:9 31:0 22:6
AC 0:08 0:02 0:02 0:08 0:05 0:05 0:09 0:09 0:09 0:02
F reqL 0:10 0:48 0:09 0:08 0:29 0:14 0:51 0:08 0:16 0:28
F reqS 0:07 0:44 0:07 0:22 0:35 0:16 0:47 0:06 0:21 0:27
Panel B: FX Returns
M ean 0:73 1:31 2:22 1:97 5:69 1:43 0:83 0:58 0:40 5:54
Sdev 10:76 9:86 9:03 11:59 8:12 9:28 8:17 8:26 10:27 8:20
Skew 0:73 0:33 0:24 0:49 0:03 0:72 0:10 0:53 0:54 0:10
Kurt 5:43 3:94 3:75 5:64 4:05 4:35 4:06 5:21 5:63 5:10
SR 0:07 0:13 0:25 0:17 0:70 0:15 0:10 0:07 0:04 0:68
SO 0:09 0:24 0:37 0:23 1:06 0:21 0:17 0:10 0:05 1:02
M DD 43:2 22:7 22:3 40:3 13:6 37:0 17:5 20:6 37:6 17:2
AC 0:09 0:01 0:02 0:09 0:04 0:08 0:10 0:08 0:10 0:02
F reqL 0:10 0:48 0:09 0:08 0:29 0:14 0:51 0:08 0:16 0:28
F reqS 0:07 0:44 0:07 0:22 0:35 0:16 0:47 0:06 0:21 0:27
CAR 1:00 0:20 0:30 0:76 0:08 1:00 0:07 0:32 0:59 0:06
M OM 0:20 1:00 0:20 0:23 0:11 0:07 1:00 0:19 0:17 0:15
V AL 0:30 0:20 1:00 0:46 0:19 0:32 0:19 1:00 0:57 0:03
RR 0:76 0:23 0:46 1:00 0:10 0:59 0:17 0:57 1:00 0:09
V RP 0:08 0:11 0:19 0:10 1:00 0:06 0:15 0:03 0:09 1:00

Table A.3. Decomposition of the Volatility Risk Premium Strategy
This table presents results on the Hassan and Mano (2013) decomposition of the covariance between
volatility risk premia and future excess returns is decomposed into a ‘static’ (STA), ‘dynamic’ (DYN), and
‘dollar’(DOL) component. Combining the static and dynamic trade yields a cross-sectional currency portfolio
(CRS) which exploits persistent di¤erences in the cross-section of countries’volatility risk premia for forecasting
and portfolio formation, whereas combining the dynamic and dollar trade yields a time-series portfolio (TS)
which exploits variation in countries’volatility risk premia over time for forecasting and portfolio formation.
The table also reports …rst order autocorrelation coe¢ cient (AC), the annualized Sharpe ratio (SR), the
Sortino ratio (SO), and the percentage maximum drawdown (M DD) position. Panel A displays the overall
currency excess return whereas Panel B reports the exchange rate return component only. Panel C presents
the sample correlations of the currency excess returns. Returns are expressed in percentage per annum and
not adjusted for transaction costs. The strategies are rebalanced monthly from January 1998 to December
2013.

ST A DY N DOL CRS T M S ST A DY N DOL CRS TMS
M ean 3:51 3:82 2:59 7:33 6:41 3:28 0:49 2:54 3:77 3:03
Sdev 9:50 8:69 8:47 11:60 11:21 8:53 8:21 8:45 11:55 10:52
Skew 0:35 0:25 0:17 0:07 0:19 0:16 0:28 0:45 0:59 0:00
Kurt 3:07 3:50 3:87 4:83 3:72 4:47 4:47 4:27 6:69 4:80
SR 0:37 0:44 0:31 0:63 0:57 0:38 0:06 0:30 0:33 0:29
SO 0:55 0:76 0:47 0:92 1:02 0:55 0:10 0:42 0:50 0:43
M DD 0:21 0:22 0:22 0:23 0:27 0:18 0:34 0:24 0:25 0:31
AC 0:07 0:14 0:05 0:05 0:13 0:09 0:18 0:08 0:02 0:15
Panel B: FX Returns
M ean 0:65 4:62 2:29 5:27 6:91 2:92 1:13 1:89 4:06 3:02
Sdev 9:56 8:70 8:42 11:54 11:17 8:60 8:22 8:40 11:70 10:47
Skew 0:33 0:29 0:19 0:10 0:14 0:08 0:37 0:51 0:75 0:04
Kurt 3:03 3:62 3:89 4:85 3:76 4:66 4:74 4:40 7:46 5:11
SR 0:07 0:53 0:27 0:46 0:62 0:34 0:14 0:22 0:35 0:29
SO 0:10 0:92 0:41 0:65 1:09 0:49 0:23 0:31 0:55 0:43
M DD 0:27 0:19 0:23 0:26 0:28 0:20 0:28 0:25 0:22 0:32
AC 0:08 0:13 0:05 0:05 0:12 0:08 0:17 0:07 0:01 0:14
ST A 1:00 0:19 0:15 0:68 0:04 1:00 0:05 0:30 0:70 0:20
DY N 0:19 1:00 0:15 0:59 0:66 0:05 1:00 0:20 0:67 0:62
DOL 0:15 0:15 1:00 0:01 0:64 0:30 0:20 1:00 0:08 0:64
CRS 0:68 0:59 0:01 1:00 0:47 0:70 0:67 0:08 1:00 0:59
TMS 0:04 0:66 0:64 0:47 1:00 0:20 0:62 0:64 0:59 1:00
V RP 0:13 0:80 0:09 0:71 0:55 0:43 0:66 0:20 0:78 0:67

Table A.4. Breakdown of Volatility Risk Premia Portfolios
The table presents the number of times a given currency enters the corner portfolios of the volatility risk premia portfolios. PL denotes the long
portfolio whereas PS is the short portfolio. The strategies are rebalanced monthly from January 1998 to December 2013.
AU D CAD CHF DKK EU R GBP JP Y N OK N ZD SEK BRL CZK HU F KRW M XN P LN SGD T RY TWD ZAR
Developed
PL 92 44 53 4 7 15 58 37 54 38
PS 21 36 9 55 59 67 65 42 32 42
Developed & Emerging
PL 94 46 50 4 6 15 59 32 52 33 9 14 16 6 0 47 8 1 5 0
PS 2 13 9 26 26 46 51 34 10 35 42 3 18 36 55 9 13 39 91 0

Table A.5. -Sorted Portfolios: Principal Component of Volatility Risk Premia
individual currency excess returns on the …rst principal component of volatility risk premia using a 36-month
moving window. The long (short) portfolio PL (PS ) contains the top 20% of all currencies with the lowest
(highest) . The table also reports the …rst order autocorrelation coe¢ cient (AC), the annualized Sharpe
ratio (SR), and the frequency of portfolio switches (F req). Panel A displays the overall excess return, whereas
Panel B reports the exchange rate component only. Panel C presents the pre- and post-formation s, and
the pre- and post-formation interest rate di¤erential (if ) relative to the US dollar. The superscripts , ,
and indicate statistical signi…cance at 10%, 5%, and 1%, respectively, based on Newey and West (1987)
and Andrews (1991). Returns are expressed in percentage per annum and adjusted for transaction costs. The
sample runs from January 1998 to December 2013.

M ean 3:85 2:40 3:02 2:75 8:88 5:02 3:93 3:12 3:62 2:48 7:31 3:37
Sdev 9:14 10:59 9:23 10:38 11:91 10:67 8:12 9:55 10:03 10:12 12:16 10:97
Skew 0:36 0:09 0:67 0:26 0:51 0:86 0:20 0:22 0:42 0:59 0:87 1:00
Kurt 3:42 4:43 5:24 4:47 5:35 7:35 2:97 5:17 4:32 4:71 5:93 6:59
SR 0:42 0:23 0:33 0:27 0:75 0:47 0:48 0:33 0:36 0:24 0:60 0:31
AC1 0:06 0:02 0:17 0:01 0:01 0:06 0:10 0:01 0:05 0:01 0:02 0:05
F req 0:16 0:24 0:28 0:29 0:11 0:11 0:10 0:18 0:28 0:25 0:12 0:12
Panel B: FX Returns
M ean 4:62 2:56 2:63 2:05 6:25 1:63 4:66(( 3:10 2:90 1:46 4:55 0:11
Sdev 9:14 10:56 9:28 10:31 11:93 10:77 8:09 9:55 9:98 10:10 12:14 11:01
Skew 0:38 0:10 0:69 0:28 0:53 0:95 0:19 0:22 0:43 0:67 0:94 1:12
Kurt 3:48 4:38 5:23 4:53 5:35 7:56 2:95 5:16 4:29 4:76 6:06 6:96
SR 0:51 0:24 0:28 0:20 0:52 0:15 0:58 0:32 0:29 0:15 0:37 0:01
AC1 0:05 0:03 0:18 0:02 0:01 0:07 0:09 0:01 0:04 0:00 0:02 0:05
F req 0:16 0:24 0:28 0:29 0:11 0:11 0:10 0:18 0:28 0:25 0:12 0:12
pre-if 0:53 0:05 0:59 0:97 2:32 0:47 0:24 0:96 1:27 2:35
post-if 0:53 0:05 0:56 0:97 2:26 0:51 0:25 0:95 1:23 2:29
pre- 0:13 0:05 0:01 0:08 0:18 0:14 0:04 0:06 0:15 0:29
post- 0:07 0:03 0:06 0:01 0:04 0:08 0:05 0:02 0:01 0:04

Table A.6. -Sorted Portfolios: Equity Volatility Risk Premium
individual currency excess returns on the US equity volatility risk premium using a 36-month moving window.
The volatility risk premium is de…ned as the 1-month realized volatility on the S&P500 minus the VIX index.
The long (short) portfolio PL (PS ) contains the top 20% of all currencies with the lowest (highest) . H=L
denotes a long-short strategy that buys PL and sells PS . The table also reports the …rst order autocorrelation
coe¢ cient (AC1 ), the annualized Sharpe ratio (SR), and the frequency of portfolio switches (F req). Panel
A displays the overall excess return, whereas Panel B reports the exchange rate component only. Panel
C presents the pre- and post-formation s, and the pre- and post-formation interest rate di¤erential (if )
relative to the US dollar. The superscripts , , and indicate statistical signi…cance at 10%, 5%, and 1%,
respectively, based on Newey and West (1987) and Andrews (1991). Returns are expressed in percentage per
annum and adjusted for transaction costs. The sample runs from January 1998 to December 2013.

M ean 4:60 4:76 4:40 2:28 4:89 0:28 4:33 4:96 3:36 3:74 4:11 0:22
Sdev 10:41 10:38 10:07 10:65 8:80 9:43 10:38 9:78 10:14 10:13 8:77 8:88
Skew 0:29 0:45 0:30 0:39 0:20 1:09 0:59 0:74 0:36 0:17 0:42 1:07
Kurt 4:81 5:59 4:53 4:82 4:08 6:12 5:89 5:78 4:34 5:26 3:77 7:05
SR 0:44 0:46 0:44 0:21 0:56 0:03 0:42 0:51 0:33 0:37 0:47 0:03
AC1 0:17 0:08 0:04 0:00 0:06 0:13 0:17 0:12 0:00 0:00 0:11 0:15
F req 0:19 0:37 0:36 0:30 0:15 0:15 0:20 0:35 0:41 0:35 0:17 0:17
Panel B: FX Returns
M ean 2:99 4:00 3:90 2:17 4:99 2:00 2:43 4:12 2:61 3:80 3:86 1:43
Sdev 10:39 10:37 10:06 10:65 8:84 9:52 10:33 9:78 10:10 10:07 8:81 8:88
Skew 0:31 0:47 0:31 0:39 0:19 1:12 0:65 0:77 0:37 0:18 0:47 1:13
Kurt 4:88 5:55 4:52 4:75 4:15 6:33 6:02 5:79 4:25 5:36 3:89 7:18
SR 0:29 0:39 0:39 0:20 0:56 0:21 0:23 0:42 0:26 0:38 0:44 0:16
AC1 0:16 0:08 0:04 0:00 0:06 0:13 0:16 0:13 0:00 0:00 0:10 0:15
F req 0:19 0:37 0:36 0:30 0:15 0:15 0:20 0:35 0:41 0:35 0:17 0:17
pre-if 1:83 0:96 0:71 0:33 0:41 2:13 1:05 0:93 0:18 0:09
post-if 1:81 0:95 0:70 0:30 0:45 2:10 1:03 0:94 0:17 0:17
pre- 0:20 0:12 0:07 0:00 0:11 0:19 0:12 0:05 0:02 0:13
post- 0:02 0:01 0:01 0:01 0:01 0:03 0:00 0:02 0:02 0:00

Table A.7. Volatility Spread Strategies
This table presents selected descriptive statistics of realized volatility spread (RV SLS ) strategies formed
using time t 1 information. The strategy buys (sells) the top 20% of all currencies with the highest (lowest)
volatility spread de…ned as long-maturity (L) minus short-maturity (S) realized volatility. Realized volatilities
are constructed using daily exchange rate returns. The table reports the annualized Sharpe ratio based on
the overall excess (exchange rate) returns in Panel A (Panel B ), the sample correlation with the carry trade
(CAR) strategy in Panel C, and the sample correlation with the volatility risk premium (V RP ) strategy in
Panel D. Returns are adjusted for transaction costs. The strategies are rebalanced monthly from January
LM 6 LM 9 LM 12 LM 18 LM 24 LM 6 LM 9 LM 12 LM 18 LM 24
Developed (G10) Developed & Emerging (G20)
Sharpe Ratios: Excess Returns
SM 1 0:35 0:29 0:44 0:50 0:38 0:32 0:47 0:57 0:51 0:41
SM 2 0:48 0:49 0:47 0:42 0:41 0:37 0:44 0:45 0:35 0:37
SM 3 0:26 0:60 0:51 0:28 0:29 0:21 0:40 0:24 0:41 0:49
SM 6 0:15 0:07 0:20 0:04 0:29 0:21 0:26 0:16
Sharpe Ratios: FX Returns
SM 1 0:37 0:33 0:48 0:52 0:39 0:30 0:47 0:57 0:49 0:39
SM 2 0:50 0:51 0:48 0:44 0:42 0:39 0:43 0:43 0:35 0:37
SM 3 0:29 0:62 0:54 0:31 0:30 0:23 0:40 0:27 0:43 0:50
SM 6 0:20 0:12 0:24 0:07 0:35 0:28 0:34 0:23
Correlation with CAR: Excess Returns
SM 1 0:11 0:16 0:22 0:16 0:19 0:03 0:08 0:17 0:06 0:13
SM 2 0:08 0:10 0:08 0:08 0:19 0:04 0:09 0:07 0:05 0:16
SM 3 0:22 0:18 0:10 0:08 0:18 0:17 0:10 0:08 0:09 0:16
SM 6 0:09 0:02 0:05 0:08 0:03 0:05 0:09 0:16
Correlation with VRP: Excess Returns
SM 1 0:14 0:27 0:27 0:18 0:11 0:12 0:26 0:32 0:24 0:15
SM 2 0:27 0:28 0:29 0:20 0:06 0:27 0:31 0:34 0:27 0:21
SM 3 0:34 0:35 0:34 0:22 0:09 0:35 0:34 0:34 0:21 0:15
SM 6 0:16 0:10 0:02 0:17 0:25 0:24 0:06 0:00

Table A.8. Asset Pricing Tests: Illiquidity
This table reports asset pricing tests for a linear factor model that includes the dollar (DOL), the spot market global illiquidity (BASF X ), and the
option market global illiquidity (BASIV ) factors. DOL is equivalent to a strategy that borrows in the US money market and equally invests in all foreign
currencies, and serves as a constant in the cross-section. BASF X is computed as the innovations to a …rst order autoregressive process applied to the
average bid-ask spread of the spot exchange rate. BASIV is computed as the innovations to a …rst order autoregressive process applied to the average
bid-ask spread of the 1-year at-the-money implied volatility. The test assets are excess returns to …ve currency portfolios sorted on the 1-year volatility
risk premium at time t 1. Panel A reports GMM and Fama-MacBeth (FMB) estimates of the market price of risk , and the Hansen-Jagannathan
distance HJ test for the null hypothesis that the pricing errors are jointly zero. Panel B reports least-squares estimates of time series regressions and the
2
test for the null that all intercepts are jointly zero. The superscripts , , and indicate statistical signi…cance at 10%, 5%, and 1%, respectively,
based on Newey and West (1987) and Andrews (1991) for GMM estimates, and Shanken (1992) for FMB estimates. Returns are annualized and adjusted
for transaction costs. The portfolios are rebalanced monthly from January 1998 to September 2013.
Panel A: Cross-Section
DOL FX R2 HJ DOL FX R2 HJ
GM M1 0:03 0:17 0:36 0:18 0:03 1:78 0:62 0:13
GM M2 0:03 0:13 0:25 0:03 1:59 0:56
8
FMB 0:03 0:17 0:36 0:03 1:78 0:62
DOL IV R2 HJ DOL IV R2 HJ
GM M1 0:02 0:35 0:14 0:17 0:02 0:74 0:32 0:16
GM M2 0:02 0:35 0:19 0:02 0:44 0:20
FMB 0:02 0:35 0:14 0:02 0:74 0:32
Panel B: Time-Series
2 2
DOL FX R2 DOL FX R2
PL 0:03 0:96 0:10 0:70 8:11 0:02 1:05 0:01 0:75 9:31
P2 0:01 0:98 0:05 0:76 0:01 0:93 < :01 0:80
P3 0:01 0:91 0:02 0:70 0:01 0:98 0:01 0:82

P4 0:01 1:07 0:04 0:80 < :00 1:16 0:01 0:85
PS 0:03 1:05 0:07 0:75 0:01 0:84 < :01 0:71
2 2
DOL IV R2 DOL IV R2
PL 0:02 0:94 0:00 0:69 6:89 0:02 1:05 0:01 0:75 5:60
P2 0:01 0:98 0:00 0:75 0:01 0:95 0:01 0:81
P3 0:01 0:95 0:03 0:71 0:01 0:98 0:00 0:82
P4 0:01 1:08 0:01 0:80 0:00 1:16 0:01 0:84
PS 0:03 1:04 0:04 0:76 0:01 0:86 0:01 0:71
Table A.9. Arbitrage Risk Proxies and V RP
This table presents predictive regressions estimates. The dependent variable is the exchange rate return component of the V RP strategy at time t.
This strategy is a long/short portfolio that buys (sells) the top 20% of all currencies with the highest (lowest) 1-year volatility risk premia at time t 1.
The set of predictors is measured at time t 1, and includes the T ED spread, the V IX index, the St. Louis Fed Financial Stress Index F SI, and the
F und F lows of currency and global macro funds constructed as the Asset under Management (AUM) weighted net ‡ows scaled by the lagged AUM as
in Patton and Ramadorai (2013). denotes the …rst-di¤erence operator. The superscripts , , and indicate statistical signi…cance at 10%, 5%, and
1%, respectively, based on Newey and West (1987) and Andrews (1991). The exchange rate returns are annualized. The sample runs from January 1998
to December 2013.
T ED F und T ED F und
T ED V IX F SI V IX F lows R2 T ED V IX F SI V IX F lows R2
0:02 0:15 0:05 0:02 0:15 0:05
0:06 < :01 < :01 0:06 < :01 < :01
0:06 0:07 < :01 0:06 0:08 < :01
9
0:06 < :01 0:01 0:05 < :01 < :01
0:07 1:59 0:05 0:06 1:31 0:01
0:01 0:13 1:02 0:03 0:01 0:14 0:69 0:05
0:07 < :01 1:59 0:02 0:06 < :01 1:29 0:01
0:07 0:05 1:52 0:02 0:06 0:07 1:24 0:02
0:06 < :01 1:53 0:04 0:06 < :01 1:30 0:01
< :01 0:12 < :01 < :01 1:07 0:04 0:03 0:17 0:01 0:01 0:57 0:05

Table A.10. Currency Strategies: VRP Measures
rate di¤erential relative to the US dollar whereas V RP is the volatility risk premium strategy that buys (sells)
the top 20% of all currencies with the highest (lowest) 1-year volatility risk premium. The 1-year volatility
risk premium is de…ned as the realized volatility (RVt ) minus the synthetic volatility swap rate (SWt ). The
subscript cs indicates that SWt is computed by interpolating implied volatilities via the cubic spline method
(Jiang and Tian, 2005). This is the default approach used in the core analysis; atm indicates that SWt
is simply proxied by at-the-money implied volatility; vv indicates that SWt is constructed by interpolating
implied volatilities via the Vanna-Volga method (Castagna and Mercurio, 2007); and si indicates that SWt is
based on the simple variance swap method (Martin, 2012). The table also reports …rst order autocorrelation
coe¢ cient (AC), the annualized Sharpe ratio (SR), the Sortino ratio (SO), the percentage maximum drawdown
(M DD), the frequency of portfolio switches for the long (F reqL ) and the short (F reqS ) position. Panel A
displays the overall currency excess return whereas Panel B reports the exchange rate return component only.
Panel C presents the sample correlations of the currency excess returns. The superscripts , , and
indicate statistical signi…cance for the mean at 10%, 5%, and 1%, respectively, based on Newey and West
(1987) and Andrews (1991). Returns are expressed in percentage per annum and not adjusted for transaction

CAR V RPcs V RPatm V RPvv V RPsi CAR V RPcs V RPatm V RPvv V RPsi
M ean 4:10 4:95 5:01 4:89 4:74 4:90 4:16 5:25 3:13 4:06
Sdev 10:73 8:15 7:82 7:98 8:11 9:25 8:14 7:51 7:78 7:81
Skew 0:71 0:03 0:09 0:23 0:06 0:65 0:01 0:10 0:16 0:16
Kurt 5:25 3:97 3:58 3:68 3:53 4:21 4:54 4:08 4:10 3:98
SR 0:38 0:61 0:64 0:61 0:58 0:53 0:51 0:70 0:40 0:52
SO 0:49 0:93 1:04 1:07 0:93 0:74 0:76 1:07 0:60 0:78
M DD 0:38 0:17 0:14 0:18 0:16 0:28 0:24 0:15 0:26 0:22
AC1 0:08 0:05 0:01 0:07 0:04 0:05 0:02 0:02 0:03 0:00
F reqL 0:10 0:29 0:28 0:29 0:29 0:14 0:28 0:28 0:31 0:27
F reqS 0:07 0:35 0:34 0:36 0:38 0:16 0:27 0:29 0:29 0:27
Panel B: FX Returns
M ean 0:81 5:45 5:37 5:64 4:81 1:61 5:27 5:73 4:39 4:46
Sdev 10:76 8:12 7:77 7:93 8:06 9:29 8:20 7:53 7:78 7:78
Skew 0:72 0:03 0:06 0:24 0:02 0:72 0:09 0:11 0:12 0:17
Kurt 5:43 4:04 3:75 3:77 3:61 4:35 5:10 4:27 4:31 4:16
SR 0:08 0:67 0:69 0:71 0:60 0:17 0:64 0:76 0:56 0:57
SO 0:10 1:01 1:10 1:23 0:93 0:23 0:97 1:14 0:85 0:85
M DD 0:43 0:15 0:14 0:15 0:14 0:37 0:18 0:15 0:19 0:18
AC1 0:09 0:04 0:01 0:06 0:02 0:08 0:02 0:03 0:01 0:03
F reqL 0:10 0:29 0:28 0:29 0:29 0:14 0:28 0:28 0:31 0:27
F reqS 0:07 0:35 0:34 0:36 0:38 0:16 0:27 0:29 0:29 0:27
CAR 1:00 0:08 0:04 0:18 0:27 1:00 0:06 0:02 0:12 0:16
V RP 0:08 1:00 0:88 0:91 0:80 0:06 1:00 0:90 0:94 0:88
V RPatm 0:04 0:88 1:00 0:87 0:77 0:02 0:90 1:00 0:86 0:90
V RPvv 0:18 0:91 0:87 1:00 0:73 0:12 0:94 0:86 1:00 0:85
V RPsi 0:27 0:80 0:77 0:73 1:00 0:16 0:88 0:90 0:85 1:00
10

Table A.11. VAR-GARCH, Individual Currencies
1=2 0
This table reports selected parameter estimates of the following VAR-GARCH speci…cation: yt = ct + ut , with ut = Ht "t where yt = (et ; vt ) , ct
is the conditional mean modelled as a Vector Autoregressive Process (VAR) of order p, and Ht is the conditional covariance whose elements are modelled
as
h 0 h :
2 3 2 3 2 32 3 2 32 3
4 hh 5 = 4 ! h 5 + 4 0 5 4 hh 5 +4 0 0 5 4 ue uv 5
We estimate the above speci…cation via maximum likelihood for each currency pair described in Table A.1: et denotes the 1-month exchange rate return
whereas vt is the 1-year volatility risk premium. The superscripts , , and indicate statistical signi…cance at 10%, 5%, and 1%, respectively, based on
Newey and West (1987) and Andrews (1991). Exchange rate returns are annualized. The sample runs at daily frequency from January 1998 to December
2013:
ev ve ev ve ev ve ev ve
AUD 0:32 0:09 0:12 0:01 BRL 0:08 0:28 0:02 0:85
11
CAD 0:12 0:01 0:19 0:00 CZK 0:07 1:27 0:20 0:04
CHF 0:05 0:05 0:10 0:03 HUF 0:18 0:07 0:06 0:18
DKK 0:03 0:01 0:11 0:01 KRW 0:06 0:52 0:01 0:01
EUR 0:04 0:02 0:11 0:01 MXN 0:03 0:11 0:05 0:74
GBP 0:02 0:03 0:10 0:01 PLN 0:15 0:07 0:26 0:09
JPY 0:17 0:05 0:02 0:04 SGD 0:29 1:42 0:04 0:78

NOK 0:17 0:00 0:25 0:02 TRY 0:27 0:07 0:04 0:08
NZD 0:27 0:29 0:06 0:02 TWD 0:07 0:42 0:08 0:33
SEK 0:06 0:01 0:18 0:01 ZAR 0:36 0:01 0:14 0:01
12
Figure A.1. Sharpe Ratios of Currency Strategies
The figure presents for developed & emerging countries the 1-year rolling Sharpe ratios of currency strategies formed using t − 1 information. CAR is the carry strategy
that buys (sells) the top 20% of all currencies with the highest (lowest) interest rate differential relative to the US dollar. Similarly, M OM is the momentum strategy
that buys (sells) currencies with the highest (lowest) past 3-month exchange rate return, V AL is the value strategy that buys (sells) currencies with lowest (highest) real
exchange rate, RR is the risk reversal strategy that buys (sells) currencies with the lowest (highest) 1-year 10-delta risk reversal, and V RP is the volatility risk premium
strategy that buys (sells) currencies with the highest (lowest) 1-year volatility risk premium. The Sharpe Ratios are computed using excess returns net of transaction
13
Figure A.2. Currency Strategies and Payoffs
The figure presents for developed & emerging countries the cumulative wealth of the currency strategies described in Figure 1. The strategies are rebalanced monthly
from January 1998 to December 2013 and adjusted for transaction costs.

SSRN Id2233367

Uploaded by

Copyright:

Available Formats

SSRN Id2233367

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SSRN Id2233367

Uploaded by

Copyright:

Available Formats

Volatility Risk Premia and Exchange Rate Predictability y

Pasquale Della Corte Tarun Ramadorai Lucio Sarno

First version: December 2013 - Revised: July 2014

First version: December 2013 - Revised: July 2014

Keywords: Exchange Rates; Volatility Risk Premium; Predictability, Minimum-Variance

JEL Classi…cation: F31; F37.

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

V Pt; = (RVt; SWt; ) (1)

SWt; = EtQ [RVt; ] (2)

Electronic copy available at: https://ssrn.com/abstract=2233367

and refer to it as model-free implied volatility.

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

3 Data and Currency Portfolios

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Hedge Fund Flows. To construct a measure of new arbitrage capital available to

Electronic copy available at: https://ssrn.com/abstract=2233367

Positions on Currency Futures. We employ data from the Commitments of Traders

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

4 The V RP Strategy: Empirical Evidence

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

4.3 Combining V RP with Other Currency Strategies

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

5.1 Risk Premia

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

RXti = i + i GV OLt + "it ;

Electronic copy available at: https://ssrn.com/abstract=2233367

5.2 Limits to Arbitrage

Electronic copy available at: https://ssrn.com/abstract=2233367

VRPit = i + N Demit lag + uit ; (8)

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

5.3 Information Asymmetry

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Andrews, D.W.K. (1991). “Heteroskedasticity and Autocorrelation Consistent Covariance Ma-

Electronic copy available at: https://ssrn.com/abstract=2233367

Electronic copy available at: https://ssrn.com/abstract=2233367

Cochrane, J.H. (2005). Asset Pricing. Princeton: Princeton University Press.

Eraker, B. (2010). “The Volatility Premium,”Working Paper, Wisconsin School of Business.

Electronic copy available at: https://ssrn.com/abstract=2233367