Description
We find that order flows are highly informative about future exchange rates and provide significant economic value for the few large dealers who have access to these flows. Moreover, customer groups systematically engage in risk sharing with each other and differ markedly in their predictive ability, trading styles, and risk exposure.
BIS Working Papers
No 405
Information Flows in Dark
Markets: Dissecting Customer
Currency Trades
by Lukas Menkhoff, Lucio Sarno, Maik Schmeling and
Andreas Schrimpf
Monetary and Economic Department
March 2013
JEL classification: F31, G12, G15.
Keywords: Order Flow, Foreign Exchange Risk Premia,
Heterogeneous Information, Carry Trades, Hedge
Funds.
BIS Working Papers are written by members of the Monetary and Economic
Department of the Bank for International Settlements, and from time to time by
other economists, and are published by the Bank. The papers are on subjects of
topical interest and are technical in character. The views expressed in them are
those of their authors and not necessarily the views of the BIS.
This publication is available on the BIS website (www.bis.org).
© Bank for International Settlements 2013. All rights reserved. Brief excerpts may be
reproduced or translated provided the source is stated.
ISSN 1020-0959 (print)
ISBN 1682-7678 (online)
Information Flows in Dark Markets:
Dissecting Customer Currency Trades
?
Lukas Menkho?
†
Lucio Sarno
‡
Maik Schmeling
??
Andreas Schrimpf
§
This version: March 5, 2013
Abstract
We study the information in order ?ows of di?erent customer segments in the world’s
largest over-the-counter market, the foreign exchange market. The analysis draws on
a unique dataset covering a broad cross-section of currency pairs and distinguishing
trades by key types of foreign exchange end-users. We ?nd that order ?ows are highly
informative about future exchange rates and provide signi?cant economic value for the
few large dealers who have access to these ?ows. Moreover, customer groups system-
atically engage in risk sharing with each other and di?er markedly in their predictive
ability, trading styles, and risk exposure.
JEL Classi?cation: F31, G12, G15.
Keywords: Order Flow, Foreign Exchange Risk Premia, Heterogeneous Information, Carry Trades,
Hedge Funds.
?
We would like to thank an anonymous Referee, Alessandro Beber, Claudio Borio, Geir Bjønnes, Michael
Brandt, Steve Cecchetti, Jacob Gyntelberg, Hendrik Hakenes, Campbell Harvey, Joel Hasbrouck, Terrence
Hendershott, Søren Hvidkjær, Gur Huberman, Alex Kostakis, Jeremy Large, Albert Menkveld, Roel Oomen,
Richard Payne, Alberto Plazzi, Lasse Pedersen, Tarun Ramadorai, Jesper Rangvid, Paul S¨oderlind, Christian
Upper, Adrien Verdelhan, Michel van der Wel, as well as participants at several conferences, workshops and
seminars for helpful comments and suggestions. We are very grateful to Gareth Berry, Geo?rey Kendrick
and UBS for providing us with the proprietary data used in this study, and for numerous conversations
on the institutional details of foreign exchange trading at UBS. Sarno acknowledges ?nancial support from
the Economic and Social Research Council (No. RES-062-23-2340) and Menkho? and Schmeling gratefully
acknowledge ?nancial support by the German Research Foundation (DFG). The views expressed in this paper
are those of the authors and do not necessarily re?ect those of the Bank for International Settlements.
†
Kiel Institute for the World Economy and Department of Economics, Leibniz Universit¨at Hannover,
K¨onigsworther Platz 1, 30167 Hannover, Germany, Tel: +49 511 7624552, Email: [email protected]
hannover.de.
‡
Cass Business School and Centre for Economic Policy Research (CEPR). Corresponding author: Faculty
of Finance, Cass Business School, City University London, 106 Bunhill Row, London EC1Y 8TZ, UK, Tel:
+44 20 7040 8772, Fax: +44 20 7040 8881, Email: [email protected].
??
Faculty of Finance, Cass Business School, City University London, 106 Bunhill Row, London EC1Y 8TZ,
UK, Email: [email protected].
§
Bank for International Settlements and CREATES, Centralbahnplatz 2, 4002 Basel, Switzerland. Tel:
+41 61 280 8942. Email: [email protected].
The foreign exchange (FX) market is the largest ?nancial market in the world with a
daily trading volume of about four trillion U.S. dollars (BIS, 2010). Also, the FX market is
largely organized as an over-the-counter (OTC) market, meaning that there is no centralized
exchange and that market participants can have only partial knowledge about trades of other
market participants and available liquidity in di?erent market segments. Hence, despite its
size and sophistication, the FX market is fairly opaque and decentralized because of its
market structure. Adding to this lack of transparency, various trading platforms have been
introduced and market concentration has risen dramatically over the last decade with a
handful of large dealers nowadays controlling the lion’s share of FX market turnover (see,
e.g., King, Osler, and Rime, 2012). The FX market can thus be characterized as a fairly
“dark” market.
1
This paper addresses several related questions that arise in this opaque market setting.
First, do large dealers have an informational advantage from seeing a large portion of cus-
tomer trades, that is, do customer trades carry economic value for the dealer? Answering
this question is relevant for regulators and useful for understanding the implications of the
observed shift in market concentration. Second, how does risk sharing take place in the FX
market? Do customers systematically trade in opposite directions to each other or is their
trading positively correlated and unloaded onto dealers (as in, e.g., Lyons, 1997)? Answering
these questions is highly relevant to provide a better understanding of the working of the
FX market and, more generally, the functioning of OTC markets. Third, what characterizes
di?erent customer groups’ FX trading, e.g., do they speculate on trends or are they con-
trarian investors? In which way are they exposed to or do they hedge against market risk?
Answering these questions allows for a better grasp of what ultimately drives the demand for
currencies from di?erent types of end-users and enhances the knowledge about the ecology
of the world’s largest ?nancial market.
We empirically tackle these questions by means of a unique data set covering more than
ten years of daily end-user order ?ow for up to ?fteen currencies from one of the top FX
1
Du?e (2012) provides a general overview of how opaqueness and market structure impact price discovery
and trading in “dark markets”, that is, OTC markets.
1
dealers, UBS. The data are disaggregated into four di?erent groups of ?nancial (asset man-
agers and hedge funds) and non-?nancial (corporate and private clients) end-users of foreign
exchange. We therefore cover the trading behavior of various segments of end-users that are
quite heterogenous in their motives of market participation, informedness and sophistication.
Putting these data to work, we ?nd that: (i) Order ?ow by end-users is highly informative for
future exchange rate changes and carries substantial economic value for the dealer observing
these ?ows; (ii) there is clear evidence that di?erent end-user segments actively share risks
with each other; and (iii) end-user groups follow very heterogeneous trading styles and strate-
gies and di?er in their exposures to risk and hedge factors. This heterogeneity across players
is crucial for risk sharing and helps explain the vast di?erences in the predictive content of
?ows across end-user segments that we document in this paper.
To gauge the impact of order ?ow on currency excess returns, we rely on a simple portfolio
approach. This multi-currency framework allows for a straightforward measurement of the
economic value of the predictive content of order ?ow and is a pure out-of-sample approach
in that it only conditions on past information. Speci?cally, we sort currencies into portfolios
to obtain a cross-section of currency excess returns, which mimics the returns to customer
trading behavior and incorporates the information contained in (lagged) ?ows.
2
The infor-
mation contained in customer trades is highly valuable from an economic perspective: We
?nd that currencies with the highest lagged total order ?ows (that is, the strongest net buy-
ing pressure across all customer groups against the U.S. dollar) outperform currencies with
the lowest lagged ?ows (that is, the strongest net selling pressure across all customer groups
against the U.S. dollar) by about 10% per annum (p.a.).
For portfolios based on disaggregated customer order ?ow, this spread in excess returns
is even more striking. A zero-cost long-short portfolio that mimics asset managers’ trading
behavior yields an average excess return of 15% p.a., while conditioning on hedge funds’ ?ows
leads to a spread of about 10% p.a. Flows by corporate customers basically generate no spread
in returns, whereas private customers’ ?ows even lead to a highly negative spread (about -14%
p.a.). In sum, we ?nd that order ?ow contains signi?cant economic value for a dealer with
2
Lustig and Verdelhan (2007) were the ?rst to build cross-sections of currency portfolios.
2
access to such information. Hence, the trend towards more market concentration observed
in FX markets over recent years clearly bene?ts large ?nancial institutions acting as dealers
and potentially trading on this information in the inter-dealer market. These informational
advantages of dealers are further enhanced by the non-anonymous nature of transactions in
OTC markets, as trades by di?erent categories of customers convey fundamentally di?erent
information for price movements.
What drives the predictive content in ?ows? We investigate three main channels. First,
order ?ow could be related to the processing of information by market participants via the
process of “price discovery”. According to this view, order ?ow acts as the key vehicle that
impounds views about (economic) fundamentals into exchange rates.
3
If order ?ow contains
private information, its e?ect on exchange rates is likely to be persistent. Second, there
could be a price pressure (liquidity) e?ect due to downward-sloping demand curves (e.g.,
Froot and Ramadorai, 2005). If a mechanism like this is at play, we are likely to observe a
positive correlation between ?ows and prices for some limited time, followed by a subsequent
reversal as prices revert to fundamental values.
4
Third, we consider the possibility that
order ?ow is linked to returns due to the di?erent risk sharing motives and risk exposures of
market participants. For example, order ?ow could re?ect portfolio rebalancing of investors
tilting their portfolios towards currencies that command a higher risk premium. Related to
this, risk sharing could lead to the observed predictability pattern if non-?nancial customers
are primarily concerned about laying o? currency risk and implicitly paying an insurance
premium, whereas institutional investors are willing to take on that risk.
Discriminating between alternative explanations for the predictive content of order ?ow,
we ?nd clear di?erences across the four segments of end-users. Asset managers’ ?ows are
associated with permanent shifts in future exchange rates, suggesting that their order ?ow is
3
See, e.g., Payne (2003), Love and Payne (2008), Evans and Lyons (2002a, 2007, 2008), Evans (2010),
and Rime, Sarno, and Sojli (2010). Other papers relate order ?ow in a structural way to volatility (Berger,
Chaboud, and Hjalmarsson, 2009) or directly to exchange rate fundamentals (Chinn and Moore, 2011).
4
Several studies explore the underlying mechanism for the impact of order ?ow and discuss the evidence
in terms of information versus liquidity e?ects (e.g. Berger, Chaboud, Chernenko, Howorka, and Wright,
2008; Fan and Lyons, 2003; Marsh and O’Rourke, 2005; Osler, Mende, and Menkho?, 2011; Menkho? and
Schmeling, 2010; Phylaktis and Chen, 2010; Moore and Payne, 2011; Ito, Lyons, and Melvin, 1998).
3
related to superior processing of fundamental information.
5
In contrast, hedge funds’ ?ows
are merely associated with transitory exchange rate movements, that is, the impact of their
trades on future exchange rates is far less persistent. This result is more in line with short-
term liquidity e?ects but not with fundamental information processing. Corporate customers’
and private clients’ ?ows, however, seem to re?ect largely uninformed trading.
Our results also point to a substantial heterogeneity across customers in their trading
styles and risk exposures, giving rise to di?erent motives for risk sharing. First, we ?nd
that the trades of various end-user groups react quite di?erently to past returns. Asset
managers tend to be“trend-followers”(positive feedback traders) with regard to past currency
returns. By contrast, private clients tend to be “contrarians” (negative feedback traders).
The latter ?nding squares well with recent ?ndings for equity markets by Kaniel, Saar, and
Titman (2008) who show that individual equity investors behave as contrarians, e?ectively
providing liquidity for institutional investors. Di?erent from their results, however, private
clients do not directly bene?t from serving as (implicit) counterparties of ?nancial customers
in FX markets. Second, the ?ows of most customer groups are negatively correlated over
short to intermediate horizons, suggesting that di?erent groups of end-users in FX markets
engage in active risk sharing among each other. It is thus not just via the inter-dealer
market that risk is shared in FX markets, as documented by Lyons (1997), but a signi?cant
proportion of risk is shared among end-users in the customer-dealer segment. Third, we
?nd substantial heterogeneity in the exposure to risk and hedge factors across customer
segments. Asset managers’ trading does not leave them exposed adversely to systematic risk,
which suggests that the information in their ?ows is not due to risk taking but likely re?ects
superior information. Hedge funds, by contrast, are signi?cantly exposed to systematic risk
such as volatility, liquidity, and credit risk. This lends credence to the view that hedge funds
earn positive returns in FX markets by e?ectively providing liquidity and selling insurance
to other market participants. For non-?nancial customers there is some evidence of hedging
but it is not strong enough to fully explain their negative forecast performance arising from
5
This information processing can come in di?erent ways, e.g., a more accurate and/or faster interpretation
of macroeconomic news releases, and better forecasting of market fundamentals such as liquidity and hedging
demands of other market participants.
4
poor short-term market timing.
Our paper is related to prior work on the microstructure approach to exchange rates
(e.g., Evans and Lyons, 2002a,b), which suggests that order ?ow is crucial for understanding
how information is incorporated into exchange rates. It is well known from the literature
that order ?ow is positively associated with contemporaneous returns in basically all asset
classes; see, e.g., Hasbrouck (1991a,b) for stock markets, and Brandt and Kavajecz (2004)
for U.S. bonds. This is a stylized fact which also holds in FX markets, as shown by Evans
and Lyons (2002a) and many subsequent studies. There is less clear evidence, however, on
whether order ?ow predicts exchange rates. A few papers have shown that FX order ?ow
contains information about future currency returns but tend to disagree on the source of this
predictive power (e.g., Evans and Lyons, 2005; Froot and Ramadorai, 2005; Rime, Sarno,
and Sojli, 2010).
6
Some other papers fail to ?nd robust predictive power of exchange rates
by order ?ow in the ?rst place (see, e.g., Sager and Taylor, 2008). Our work is also related
to a di?erent strand of recent literature that analyzes the returns to currency portfolios by
investigating the predictive power of currency characteristics, such as carry or lagged returns,
and the role of risk premia in currency markets.
7
Overall, we contribute to the literature in the following ways. We are the ?rst to show
that order ?ow forecasts currency returns in an out-of-sample forecasting setting by forming
order ?ow portfolios. This multi-currency investment approach provides an intuitive mea-
sure of the economic value of order ?ow for the few large dealers observing these ?ows. This
seems important as earlier papers either did not consider out-of-sample forecasting at all
or relied on purely statistical performance measures derived from time-series forecasts of a
limited number of currency pairs (e.g., Evans and Lyons, 2005, who study the DEM/USD
and JPY/USD crosses). Time-series forecasts are a?ected by trends in exchange rates, most
notably the U.S. dollar. Our portfolio procedure, by contrast, studies exchange rate pre-
6
There is also evidence that marketwide private information extracted from equity order ?ow is useful for
forecasting currency returns (Albuquerque, de Francisco, and Marques, 2008).
7
Lustig and Verdelhan (2007), Farhi, Fraiberger, Gabaix, Ranciere, and Verdelhan (2009), Ang and Chen
(2010), Burnside, Eichenbaum, Kleshchelski, and Rebelo (2011), Lustig, Roussanov, and Verdelhan (2011),
Barroso and Santa-Clara (2011) and Menkho?, Sarno, Schmeling, and Schrimpf (2012a,b) all build currency
portfolios to study return predictability and/or currency risk exposure.
5
dictability in dollar-neutral long-short portfolios, and it does so out-of-sample over very long
time spans compared to the extant FX microstructure literature. Moreover, we are the ?rst
to test whether risk exposure drives the information in customer order ?ows. We show how
di?erent key FX market players trade, e.g., to which extent they rely on trend-following or
behave as contrarians, and in which ways they are exposed to systematic risk. We ?nd strong
evidence of heterogeneity in the exposures and trading behavior across di?erent groups of
market participants. These ?ndings indicate that there is signi?cant risk sharing between ?-
nancial and non-?nancial customers as well as between di?erent groups of ?nancial customers
(leveraged versus real money managers).
Taken together, these results have implications for our general understanding of informa-
tion ?ows in dark markets and how large dealers in OTC markets bene?t from observing a
large proportion of the order ?ow. These results also add to our general understanding of
how risk is shared in ?nancial markets due to di?erent motives for trade and trading styles
across end-user segments.
The rest of the paper is structured as follows. Section I describes our data, Section
II presents empirical results on the predictive power of order ?ow, Section III empirically
investigates alternative underlying reasons for why order ?ow forecasts FX excess returns,
and Section IV presents results of robustness tests. Section V concludes.
I. Data
We employ a unique dataset based on daily customer order ?ows for up to 15 currency pairs
over a sample period from January 2, 2001 to May 27, 2011, for a total of 2,664 trading
days. In contrast to much of the earlier literature, we employ order ?ow from the end-
user segment of the FX market and not from the inter-dealer market. This is important
since microstructure models suggest that the information in ?ows stems from trading with
customers and not from inter-dealer trading (e.g. Evans and Lyons, 2002a). Order ?ows in
our sample are measured as net buying pressure against the U.S. dollar (USD), that is, the
U.S. dollar volume of buyer-initiated minus seller-initiated trades of a currency against the
6
USD. The data cover all trades of customers (end-users) with UBS during our sample period.
A positive number indicates net buying pressure in the foreign currency relative to the USD.
Order ?ow therefore does not measure trading volume but net buying (or selling) pressure, as
mentioned above. Our order ?ow data are available both in aggregated form and at a higher
level of granularity allowing for a di?erentiation across end-user groups.
Aggregate order ?ow. Aggregate order ?ows, that is, aggregated across customers (re-
gardless of their type), are available for the following 15 currencies: Australia (AUD), Brazil
(BRL), Canada (CAD), the Euro (EUR), Hong Kong (HKD), Japan (JPY), Sweden (SEK),
Mexico (MXN), New Zealand (NZD), Norway (NOK), Singapore (SGD), South Africa (ZAR),
South Korea (KRW), Switzerland (CHF), and the United Kingdom (GBP). In the following,
we refer to these ?ows as “total ?ows” since they are aggregated across all customers of UBS.
A natural question is whether ?ows by customers of UBS are generally representative
of end-user currency demands in the FX market. While this question cannot be answered
without knowledge of the customer ?ows of all other dealers, there are good reasons to believe
that the ?ows employed in our paper are highly correlated with a large portion of end-user
order ?ows. First, UBS is among the largest dealers in the FX market and their average
market share (according to the Euromoney FX Survey) over our sample period amounts to
about 13%. Over most of our sample period, UBS was ranked among the top three of all
FX dealers (with Deutsche Bank, Barclays, and Citi usually being the closest competitors).
Thus, UBS clearly is one of the most important FX dealers with a signi?cant portion of the
market solely on its own.
8
Second, a handful of top dealers in the FX market account for
more than 50% of total market share (e.g., King, Osler, and Rime, 2012) and all of these large
dealers essentially have access to the same set of large customers. Hence, it seems very likely
that UBS ?ows’ are highly correlated with ?ows observed at, e.g., Deutsche Bank, Barclays,
Citi, or JP Morgan, which in turn implies that our order ?ows are representative of the top
end of customer trading in the FX market.
8
Note that most UBS FX customers are in fact big players and include other banks, many large asset
management ?rms and hedge funds, and a large fraction of wealthy private clients. According to the Eu-
romoney survey, UBS has a particularly high market share in FX business with ?nancial customers (banks,
real money and leveraged funds).
7
Table I about here
Table I shows descriptive statistics for total ?ows (in billion USD). Daily order ?ows are
largest on average for EUR, JPY, and CHF. The pair with the largest average imbalance (in
absolute value) between buyer- and seller-initiated trading volume is the EUR/USD, where
customers (on a net basis) sold on average 63 million EUR against USD per trading day over
our sample period. Hence, average order ?ows are fairly small relative to gross daily trading
volume in FX markets.
9
Flows are fairly volatile, however, which means that order ?ow
imbalances can frequently be very large. Daily ?ows tend to be positively autocorrelated,
but the degree of autocorrelation is very small albeit sometimes statistically signi?cant. There
is also a clear pattern in standard deviations. Major currencies, such as the EUR, CHF, JPY,
GBP, have much larger variation in order ?ows and, hence, a larger absolute size of order
?ows compared to other currencies and especially emerging markets. This is intuitive as
there is much more trading in major currencies, but it also suggests that one cannot easily
compare order ?ows across currencies and that some form of standardization is needed to
make sensible comparisons.
10
We take this into account in our empirical analysis below.
Finally, aggregate order ?ows display a high kurtosis (especially the British pound), which is
largely driven by some days with extremely high (in absolute value) order ?ows. Eliminating
these few outliers does not change our results reported below.
Disaggregated order ?ow. We also obtain order ?ows disaggregated by customer groups
for the same sample period, albeit only for a subset of nine major currencies.
11
There are four
customer groups for which ?ows are available: Asset Managers (AM), Hedge Funds (HF),
Corporate Clients (CC), and Private Clients (PC). The segment of asset managers comprises
“real money investors”, such as mutual funds and pension funds. Highly leveraged traders and
short-term oriented asset managers not included in the asset managers segment are classi?ed
9
To provide a benchmark, daily gross spot turnover in the Euro/USD pair in April 2010 amounted to USD
469 billion according to the most recent FX triennial survey (BIS, 2010). These (gross) ?gures for both the
customer-dealer segment and the inter-dealer market are based on data collected from about 4,000 reporting
dealers worldwide.
10
In addition, the volatility of ?ows also varies over time and ?ows tend to become increasingly volatile
towards the end of the sample. Also for this reason, some form of standardization is necessary.
11
The nine currencies are: AUD, CAD, EUR, JPY, SEK, NZD, NOK, CHF, and GBP.
8
as hedge funds. Hedge funds are unregulated entities, whereas asset managers are regulated.
The corporate segment includes non-?nancial corporations that import or export products
and services around the world or have an international supply chain. Corporates also include
the treasury units of large non-?nancial corporations, with the exception of those pursuing a
highly leveraged investment strategy, which are classi?ed by UBS as hedge funds. The last
segment, private clients, includes wealthy clients with investable liquid assets in excess of 3
million U.S. dollars. Private clients trade primarily for ?nancial reasons and with their own
money. Hence, there is substantial heterogeneity in the motives for market participation by
these four customer types, and the groups are likely to di?er considerably in the degree of
informedness and sophistication. One of the key features of OTC markets, that is, the non-
anonymous nature of transactions can thus further enhance the informational advantages of
dealers in dark markets.
The order ?ow data are assembled as follows. Each transaction booked in the UBS
execution system at any of its world-wide o?ces is tagged with a client type. At the end of
each business day, global transactions are aggregated for each customer group. Order ?ow is
measured as the di?erence between the dollar value of purchase and sale orders for foreign
currency initiated by a particular UBS customer group. The transaction is recorded with
a positive sign if the initiator of the transaction (the non-quoting counterparty) is buying
foreign currency and vice versa.
12
Summary statistics for the disaggregated order ?ow data
are reported in Table A.1 of the Internet Appendix.
Exchange rate returns and excess returns. For our empirical analysis below, we com-
plement these order ?ow data with daily spot exchange and forward rates from Reuters
(available from Datastream). We denote log changes of spot exchange rates as ‘exchange rate
12
Our data are raw order ?ow data with ?ltering limited to the most obvious cases. For instance, data
are adjusted for large merger and acquisition deals which are announced well in advance. Cross-border
mergers and acquisitions involve large purchases of foreign currency by the acquiring company to pay the
cash component of the deal. These transactions are generally well-publicized and thus are anticipated by
market participants. Finally, FX reserve managers, UBS proprietary traders and small banks not participating
in the inter-dealer market are excluded from the data. Flows from FX reserve managers are stripped out
due to con?dentiality issues, ?ows from proprietary traders because they trade with UBS’ own money, while
small banks represent small customers less concerned about the FX market.
9
returns’
?s
t+1
= s
t+1
? s
t
, (1)
where lowercase letters refer to logs and all exchange rates are quoted as the USD price
of foreign currency, so that positive exchange rate returns correspond to an appreciation of
the foreign currency. Hence, a positive correlation of order ?ows and exchange rate returns
means that net buying pressure in the foreign currency (against the USD) is associated with
an appreciation of the foreign currency (against the USD) and vice versa.
We also compute currency excess returns which account for the interest rate di?erential
in a foreign currency position. Hence, currency excess returns rx are given by
rx
t+1
= s
t+1
? s
t
+ (i
t
? i
t
), (2)
where i
denotes the foreign interest rate and i
t
denotes the U.S. interest rate. Since we
are working at the daily frequency in our main analysis, we need to obtain daily interest
rates for all 15 countries (plus the U.S. interest rate). However, since one-day interest rates
are not directly available for all countries in our sample, we employ information in forward
rates to infer interest rate di?erentials. Interest rate di?erentials for horizon k are commonly
approximated by i
k,t
?i
k,t
? s
t
?f
k,t
where f
k,t
denotes the log forward rate for horizon k of
a given currency.
13
II. The Value of Information in Customer Flows
A. Portfolios Conditioning on Aggregate Order Flow
We rely on a portfolio approach, mimicking the returns to customer FX trading by condi-
tioning on lagged order ?ow. This provides a straightforward and intuitive assessment of the
13
This approximation is exact if covered interest rate parity (CIP) holds, which tends to be the case at
daily or even shorter horizons in normal times (Akram, Rime, and Sarno, 2008). There have been violations
of this no-arbitrage relation over the recent ?nancial crisis. As we show below, the results in this paper are
entirely driven by changes in spot rates, whereas interest rate di?erentials only play a negligible role. Thus,
the results do not depend on whether CIP holds or not.
10
economic value of order ?ow in predicting currency excess returns.
As a benchmark test, we ?rst sort currencies into portfolios based on (lagged) total order
?ows for each currency. Speci?cally, we sort currencies into ?ve portfolios (P
1
, P
2
, ..., P
5
)
depending on their total order ?ow on day t and compute portfolio excess returns (or spot
exchange rate changes) for the following day. In this basic setup, portfolios are rebalanced
at the end of each trading day. Note that these portfolios are computed from the viewpoint
of a U.S. investor as each individual portfolio consists of a short position in USD and a long
position in a basket of foreign currencies. Taking the return di?erence between any two
portfolios P
j
?P
i
thus gives the return of a portfolio short in the basket of foreign currencies
in P
i
and long in the basket of currencies in P
j
, so that the USD component cancels out and
the long-short portfolio is dollar-neutral by construction.
Standardizing order ?ows. Before sorting currencies into portfolios, we need to make
sure that order ?ows are comparable across currencies. As the absolute size of order ?ows
di?ers across currencies (as shown above in Table I) it is not sensible to sort currencies based
on raw order ?ows. To allow for meaningful cross-currency comparisons, it is necessary to
standardize ?ows. We do this by dividing ?ows by their standard deviation to remove the
di?erence in absolute order ?ow sizes across currencies
x
R
j,t
=
x
j,t
?(x
j,t?59;t
)
, (3)
where x
R
j,t
denotes order ?ow standardized over a rolling window and x
j,t
denotes the raw
order ?ow. In our baseline results, we compute the standard deviation of ?ows via a rolling
scheme over a 60-day rolling window. Robustness tests based on alternative approaches to
standardize ?ows are reported in a separate Internet Appendix.
14
Portfolio excess returns. Table II shows average annualized excess returns for order ?ow
14
In these robustness exercises, we also report results with longer rolling windows of up to three years as
well as for an expanding window. Furthermore, we provide tests where we standardize both with respect to
volatility as well as the mean. Finally, we also consider a standardization scheme based on gross FX turnover
data for di?erent currencies drawing on data from the BIS FX triennial survey. These tests, reported in the
separate Internet Appendix to conserve space, show that our results are not sensitive with regard to the way
?ows are standardized.
11
portfolios (P
1
, P
2
, ..., P
5
), where P
1
contains the three currencies with the lowest lagged stan-
dardized order ?ow and P
5
contains the three currencies with the highest lagged standardized
order ?ow. Hence, P
5
can be thought of as a portfolio of currencies with the highest buying
pressure, whereas P
1
refers to a portfolio with the strongest selling pressure. Column “Av.”
shows average returns across all currencies in the cross-section and column “BMS” denotes a
portfolio which is long in P
5
and short in P
1
(“Buying Minus Selling” pressure). We report
returns for the full sample period from January 2001 to May 2011.
15
To get started, Panel A of Table II reports results for the sample of all 15 markets (T15)
as well as for the sub-sample of 9 developed markets (T9); for the T9 sub-sample we only
form four portfolios rather than ?ve to ensure we always have two currencies in the corner
portfolios. We observe a strong increase in average excess returns as we move from the
portfolio of currencies with low buying pressure P
1
to the one with high buying pressure P
5
(or P
4
for the T9 sample). The spread in excess returns between the high buying pressure
and the low buying pressure portfolio, that is, the excess return of the BMS portfolio, is
economically large (10.31% and 12.43% p.a., respectively) and statistically highly signi?cant.
Similarly, the Sharpe Ratios (p.a.) of the two BMS portfolios of 1.26 and 1.45 are large and
also point toward high economic signi?cance. Thus, order ?ows carry signi?cant information
for future currency excess returns, as captured by our dollar-neutral out-of-sample trading
strategy which only conditions on real-time information. These results demonstrate the
economic value for the owner of this (private) information, that is, the few large FX dealer
banks which observe a signi?cant share of end-user order ?ow and are able to trade on this
information in the inter-dealer market.
Table II about here
Table A.3 in the Internet Appendix shows results for the other standardization schemes
and for sub-samples. We ?nd that our results are equally strong in various sub-periods.
Likewise, Table A.4 in the Internet Appendix shows the same exercise for exchange rate
15
Sub-sample tests for a pre-crisis subperiod from January 2001 to June 2007, and a crisis/post-crisis
subperiod from July 2007 to May 2011 are reported in the Internet Appendix.
12
changes instead of excess returns. Results in that table clearly show that the patterns in
average spot exchange rate changes across portfolios are at least as pronounced as for average
excess returns or, if anything, even more pronounced. Hence, order ?ow is informative about
future spot rates and not about interest rate di?erentials.
Tests for return monotonicity. The last three columns “MR” and “Up” in Table II report
tests for return monotonicity (Patton and Timmermann, 2010), that is, whether there is a
signi?cantly increasing or decreasing pattern of average excess returns when moving from the
portfolio of low buying pressure (P
1
) to the one with high buying pressure (P
5
).
16
These tests
go beyond the standard t-test of a zero BMS portfolio return since they take into account
the whole cross-sectional pattern. This is interesting since one would intuitively expect an
increasing pattern of average portfolio excess returns when moving from P
1
to P
5
if order ?ow
is truly informative about future excess returns. This prediction is signi?cantly borne out in
the data for both the T15 and T9 sample of countries and for both the “MR” and “Up” test.
Hence, there is strong evidence for a signi?cant relationship between order ?ow and future
excess returns.
Excess returns over time. Finally, we plot cumulative excess returns for the T15 and T9
BMS portfolios in the upper left and right panel of Figure 1. As can be seen, excess returns
are quite striking and stable for most of the sample period, although somewhat more volatile
at the beginning and towards the end of the sample.
Figure 1 about here
16
The MR statistic tests for a monotonically increasing return pattern, whereas the Up (Down) test is
somewhat less restrictive and simply tests for a generally increasing (decreasing) pattern without requiring
monotonicity in average portfolio returns. Speci?cally, the MR test requires that the return pattern is
monotonically increasing P
1
< P
2
< ... < P
5
and formulates the null hypothesis as H
0
: ? ? 0 and the
alternative hypothesis as H
a
: min
i=1,...,4
i
> 0, where ? is a vector of di?erences in adjacent average
portfolio excess returns (P
2
? P
1
, P
3
? P
2
, P
4
? P
3
, P
5
? P
4
) and
i
is element i of this vector. The
Up test formulates the null hypothesis of a ?at pattern H
0
: ? = 0 and the alternative hypothesis as
H
a
:
4
n=1
|
i
|1{
i
> 0} > 0, so that the test is less restrictive and also takes into account the size and
magnitude of deviations from a ?at return pattern. The Down test follows in an analogous way.
13
B. Portfolios Conditioning on Disaggregated Order Flow
If superior information processing or genuine forecasting ability drive our results above, one
expects clear di?erences in the forecasting power of di?erent customers’ ?ows, depending on
the groups’ characteristics (see, e.g., Fan and Lyons, 2003; Evans and Lyons, 2007, among
others). Speci?cally, one would expect to see superior information in ?ows of ?nancial cus-
tomers, given that non-?nancial players do not specialize in FX trading as their core activity.
To investigate this, we now build portfolios based on our disaggregated data for customer
?ows. We closely follow the earlier approach with the exception that we only build four
portfolios (rather than ?ve) here since we only have disaggregated ?ows for nine currencies
and want to have a minimum of two currencies per portfolio.
Table II, Panel B, reports results for the four customer groups: Asset Managers (AM),
Hedge Funds (HF), Corporate Clients (CC), and Private Clients (PC). Results are clear-
cut. Asset managers’ net buying or selling pressure of currencies is the most informative
about subsequent exchange rate behavior. Conditioning on asset managers’ ?ows generates a
cross-sectional spread in excess returns of 15% p.a., followed by hedge funds with a spread of
about 10%. In stark contrast, corporate clients’ and private clients’ ?ows actually generate
a negative spread in portfolio excess returns of about ?4% and ?14%, respectively.
17
The
results point towards substantial di?erences in the customers’ predictive information and
provide a quantitative summary of the value of this information in economic terms. The
latter is underscored by the large spread in (annualized) Sharpe Ratios of BMS portfolios
across customer groups. The asset managers’ BMS portfolio yields a Sharpe Ratio of 1.79,
whereas the private clients’ BMS portfolio has a Sharpe Ratio of -1.55.
18
As above, we also present p-values for tests of return monotonicity. Since order ?ow of
17
Table A.5 shows results for spot rate changes instead of excess returns, which display no qualitative
di?erences.
18
Table A.6 in the Internet Appendix also shows that excess returns to the BMS portfolios based on
di?erent customers’ ?ows are not highly correlated. Hence, the information contained in the di?erent ?ows
appears to stem from di?erent sources. In practice, this also means that BMS portfolios could be combined
to obtain even higher Sharpe Ratios. For example, a combined portfolio long in the asset managers’ BMS
portfolio and short in the private clients’ BMS portfolio yields an annualized Sharpe Ratio of 2.19, which is
substantially higher than the individual Sharpe Ratios.
14
corporate and private customers negatively forecasts returns, we modify the MR test in these
cases to test for a monotonically decreasing pattern. Results from these tests corroborate
the simple t-tests for the BMS portfolios. There is a monotonically increasing pattern in
average excess returns for portfolios based on asset managers’ and hedge funds’ ?ows which
is highly signi?cant. By contrast, we ?nd a monotonically decreasing pattern in average
excess returns for portfolios based on private customers’ ?ows, and marginally signi?cant
evidence for a decreasing pattern in portfolios based on corporate ?ows.
Hence, it is not the case that all order ?ow is equal in terms of its information content for
exchange rates. Instead, ?nancial customers’ ?ows (asset managers and hedge funds) account
for the positive relation between lagged ?ows and future exchange rate returns uncovered in
the previous section. Flows by corporates are more or less uninformative, whereas private
clients’ ?ows even forecast returns in the wrong direction. Using total end-user order ?ow,
which is likely to be dominated by ?nancial customers due to their higher trading volume
19
,
masks these di?erences and might even lead to wrong inference about the link between ?ows
and returns. In a nutshell, what matters for the relation between end-user order ?ows and
future returns is disaggregated data since the information content of ?ows for future returns
varies markedly across customer groups.
The middle and lower panel of Figure 1 shows cumulative returns for all four customer
groups. It can directly be seen that returns are very di?erent across customer groups, even
when comparing, for example, asset managers and hedge funds. Both groups’ BMS portfo-
lios generate signi?cant excess returns but returns for hedge funds are much more volatile
than those of asset managers. Hence, we will investigate possible sources of these di?erent
behaviors of returns below.
C. Marginal Predictive Content of Flows at Longer Horizons
Our analysis so far has been concerned with the relation between order ?ows and returns over
the subsequent trading day. An interesting question, however, is whether the information
19
This is especially true for the order ?ow employed in this paper since UBS is one of the largest dealers
in FX and has a high proportion of ?nancial customers (relative to corporate clients).
15
contained in order ?ow quickly decays or whether it is useful for forecasting returns over more
than one trading day.
We examine the marginal predictive content of ?ows by forming portfolios as in the
analysis above, but we now allow for a longer lag between the order ?ow signal and portfolio
formation. Table III contains the results for di?erent lags of 0, 1, 2, . . . , 9 days. To be more
speci?c, a lag of 0 days means that ?ows of trading day t are used to predict returns of day
t +1 (and thus reproduces BMS returns from Tables II above), whereas a lag of, e.g., 2 days
means that ?ows of day t are used to forecast returns of trading day t + 3.
Table III about here
Results in Table III show that order ?ow appears to be most informative for the ?rst
two to three days after portfolio formation and that the information in ?ows becomes in-
signi?cant afterwards. Hence, the information contained in daily ?ows is fairly short-lived
and is impounded relatively quickly into exchange rates, especially considering that the order
?ows employed here are private information that is available to only a small number of large
FX dealers and that could not realistically be incorporated into prices without some lag.
This ?nding is in contrast to, e.g., Evans and Lyons (2005) who study a shorter and smaller
sample and ?nd that times-series predictability of returns by order ?ow increases at longer
horizons when judged from statistical metrics of forecast evaluation. This contrast in results
also highlights the importance to assess the predictive power of order ?ow using measures of
economic value as opposed to purely statistical ones, as statistical evidence of exchange rate
predictability in itself does not guarantee that an investor can earn pro?ts from a trading
strategy that exploits this predictability.
D. Order Flow vs. Carry and Momentum
To further learn about the predictive content of customer order ?ow for future FX returns, we
run panel regressions, which allows us to control for other possible determinants of currency
excess returns as well as cross-sectional and time ?xed-e?ects. For example, it could be the
16
case that asset managers’ and hedge funds’ order ?ow mimicking portfolios simply reproduce
a carry trade (Burnside, Eichenbaum, Kleshchelski, and Rebelo, 2011; Lustig, Roussanov,
and Verdelhan, 2011; Menkho?, Sarno, Schmeling, and Schrimpf, 2012a) or that their order
?ow just picks up momentum e?ects in currency returns (Menkho?, Sarno, Schmeling, and
Schrimpf, 2012b).
Speci?cally, we run panel regressions of the general form
rx
j,t+1
= ?
c
OF
c
t
+ ?
1
(i
j,t
? i
t
) + ?
2
rx
j,t
+ ?
3
rx
j,t?60;t?1
+ ?
j,t+1
(4)
where j (1, ..., N) indexes currencies, rx denotes currency excess returns, OF
c
denotes order
?ow of customer group c, (i
j,t
? i
t
) denotes interest rate di?erentials (carry), and rx
t
and
rx
t?60;t?1
denotes lagged excess returns over the prior trading day and the average over the
past 60 trading days, respectively.
20
The error term is given by ?
j,t+1
= e
t+1
+u
j
+
j,t+1
and
thus captures time and cross-sectional ?xed-e?ects (we also report results without ?xed-e?ects
below). Standard errors are clustered by currency pair. Note that these panel regressions
employ non-standardized order ?ows and are based on individual currency returns and not
on portfolio returns.
Results are shown in Table IV and corroborate our ?ndings based on our portfolio ap-
proach above, namely that order ?ows of ?nancials positively predict future excess returns,
whereas ?ows by non-?nancial end-users negatively forecast returns. Based on speci?cation
(vi) in Table IV, the coe?cients on lagged order ?ow imply that a positive order ?ow of USD
1 billion forecasts a four basis point (b.p.) higher excess return on the following day for asset
managers’ ?ows, a one b.p. higher return for hedge funds, a minus one b.p. lower return for
corporates, and a minus two b.p. lower return for private clients. The magnitude of these
e?ects seems reasonable given the deep liquidity of the FX market.
Table IV about here
More important, however, is the fact that the predictive relation between lagged order
20
Using other windows of less or more than 60 trading days does not yield qualitatively di?erent results.
17
?ow and future FX excess returns remains very strong when controlling for two common
predictors of returns in FX markets, interest rate di?erentials and (short-term) momentum.
Carry shows up with a positive sign, that is, high interest rate currencies deliver high excess
returns on average in line with the large literature on the forward discount bias (e.g., Fama,
1984). Interest rate di?erentials, however, do not drive out the information contained in
order ?ows and they become insigni?cant once we include cross-sectional ?xed-e?ects in the
regression. In our panel regressions, lagged currency returns do not have consistent predictive
content beyond order ?ow and carry, which is in line with recent evidence in Menkho?, Sarno,
Schmeling, and Schrimpf (2012b), who show that FX momentum strategies are not pro?table
for major exchange rates over the last decade.
III. What Drives the Predictive Power of Flows?
A. Permanent vs. Transitory Forecast Power of Flows
To better understand the driving forces behind our results above, we next investigate whether
order ?ow forecasts returns because it signals permanent shifts in spot exchange rates or
whether it merely forecasts temporary movements which are eventually reversed after some
time. The question whether order ?ow has a permanent or transitory e?ect in prices is a
central theme in the earlier microstructure literature (see Hasbrouck, 1991a,b). A transitory
movement is interpreted as suggesting that order ?ow e?ects are merely due to short-term
liquidity or price pressure e?ects which eventually die out, whereas a permanent movement
in spot rates would indicate that order ?ow conveys information about fundamentals.
21
More
speci?cally, a permanent price impact would most probably indicate that order ?ow is related
to changes in expectations about fundamentals given the daily frequency we are working
on. This question is relevant for our analysis since we ?nd substantial heterogeneity with
21
One strand of literature argues that order ?ow is the conduit by which information about fundamentals
is impounded in prices and therefore has permanent e?ect on exchange rates (e.g., Evans and Lyons, 2002a;
Brandt and Kavajecz, 2004; Evans and Lyons, 2007, 2008). Another strand of the literature suggests that
order ?ow matters due to downward sloping demand curves or “illiquidity” and, hence, that order ?ow only
has a transitory impact on prices (e.g., Froot and Ramadorai, 2005).
18
regard to the forecasting power of di?erent customer groups’ order ?ows. Therefore it is
particularly interesting to ?nd out if all (or some) customers’ ?ows signal information relevant
for permanent changes in FX rates or whether some customer groups’ order ?ow simply exerts
price pressure and liquidity e?ects.
To this end, we apply our portfolio sorts framework as above but now track cumulative
exchange rate returns to BMS portfolios for overlapping periods of 30 trading days after
portfolio formation. This approach yields a direct estimate of how spot rates move after
experiencing intensive buying or selling pressure by customers.
Figure 2 illustrates the persistence of the predictive content of order ?ow. The solid lines
show the cumulative excess returns (in basis points), whereas the shaded areas show 95%
con?dence intervals based on a moving-block bootstrap with 1,000 repetitions. Total ?ows
for all 15 currencies (T15) forecast a permanent change in spot rates which is statistically
signi?cantly di?erent from zero. Exchange rates with the highest net buying (selling) pressure
appreciate (depreciate) against the USD for approximately three days. Currency returns on
the BMS portfolios increase by about 15 basis points over this period, and afterwards the
e?ect of the order ?ow signal levels out. Importantly, these ?ndings suggest that order ?ow
conveys information and its impact on exchange rates is not reversed.
Figure 2 about here
This picture changes when looking only at the nine developed currencies. Here, we ob-
serve the same increasing pattern initially, followed by a subsequent partial reversal. After
approximately 25 ?30 trading days, about one half of the initial impact of 15 basis points is
reversed and the con?dence interval includes zero. Hence, there is much less evidence that
order ?ow conveys information about fundamentals when only looking at major developed
markets. This ?nding makes sense, however, since the major currency markets are most
probably more researched and more e?cient than smaller currency markets so that the scope
for superior information processing is reduced.
22
22
This may be interpreted in the context of the adaptive markets hypothesis (see e.g. Neely, Weller, and
Ulrich, 2009, for an analysis in FX markets).
19
As a natural next step, we also examine the same question separately for disaggregated
order ?ows (lower panels of Figure 2). Results are clear-cut. The only end-user group with a
statistically signi?cant permanent price impact is asset managers. Hedge funds’ trading has
a positive but transitory impact in line with an interpretation that they provide liquidity.
Corporate clients have no impact at all, and private clients have a transitory negative impact.
Given our ?nding for total ?ows of the nine major currencies above, it is interesting to see
that asset managers’ ?ows are indeed associated with permanent spot rate changes. Hence,
order ?ow of asset managers seems to be related to the processing of fundamental information
whereas hedge funds’ order ?ow corresponds to short-lived information unrelated to funda-
mental information. Similarly, it seems reasonable that the negative relation between private
clients’ ?ows and future spot rates eventually dies out over time.
These ?ndings are novel in the literature and suggest that order ?ows by di?erent end-
user groups – even by the two ?nancial customer groups – embed di?erent information for
future exchange rates. These di?erences can arise either because they are based on di?erent
mechanisms to process information or because of di?erent trading motives and hedging needs.
To explore this further, we investigate the drivers of order ?ow in more detail and shed light
on the observed di?erences in end-user order ?ows.
B. Risk Sharing Among Foreign Exchange End-Users
The analysis above suggests that asset managers’ ?ows are related to the processing of fun-
damental information that is quickly, but permanently, impounded into prices whereas the
other customer groups’ ?ows are not. A potential explanation is that risk sharing among
market participants drives (part of) our results. For instance, private clients’ negative BMS
returns could be explained by their possible need for hedging FX risk, whereas the positive
returns of hedge funds might implicitly re?ect a compensation for taking on such risks. While
these are just examples, a risk sharing story in general implies that we observe customers sys-
tematically trading in opposite directions and that their portfolios load on di?erent sources
of systematic risk. We investigate these issues below.
20
Portfolio returns in event time. We ?rst provide a more detailed look at the return be-
havior around portfolio formation dates to better understand di?erences in customer groups.
Figure 3 shows the average annualized BMS excess return for the ?ve days prior to portfo-
lio formation (?5, ?4, ..., ?1), the day of portfolio formation 0, and the ?rst ten days after
portfolio formation (1, 2, ..., 10). Shaded areas correspond to 95% con?dence intervals based
on Newey and West (1987) standard errors. Note that these returns, unlike Figure 2, are not
cumulative.
Figure 3 about here
Two results stand out. First, asset managers tend to be trend-followers in that they exert
buying (selling) pressure in currencies that recently appreciated (depreciated). Conversely,
private clients tend to trade against the trend, that is, they react upon past returns in a con-
trarian fashion. The pattern for hedge funds and corporates is less clear. Second, formation
day returns (day 0) are signi?cantly di?erent from zero for all four customer groups. How-
ever, hedge funds (positive) and private clients (negative) have the largest contemporaneous
returns in absolute value, indicating that their trading either heavily drives exchange rates
or is heavily triggered by returns (e.g., via stop-loss and stop-buy orders). The latter expla-
nation seems more reasonable especially for private clients who do not trade large enough
volumes to move prices in FX markets.
Overall, these ?ndings suggest that customer groups trading positions at least partly
o?set each other, as asset managers and private clients clearly di?er in terms of their trend-
following behavior. This ?nding is di?erent from equity markets where Kaniel, Saar, and
Titman (2008) ?nd that individual investors also tend to be contrarian traders but that
they experience subsequent positive returns, presumably due to implicitly providing liquidity
to institutional investors. In our data, we ?nd a similar contrarian behavior of individual
investors, but this trading behavior does not yield positive returns on average.
Flow correlations over longer horizons. Given these ?ndings, we next look at the corre-
lation of customer groups’ ?ows directly. While there is little contemporaneous correlation in
?ows, as noted above (see Table A.2 in the Internet Appendix), it is nevertheless interesting
21
to look at ?ows over longer horizons to ?nd out if customer groups tend to trade in the same
or in opposite directions. For a risk sharing explanation to make sense, we would expect to
see negative ?ow correlations between customer groups at some horizons.
Figure 4 plots contemporaneous correlations between ?ows of all four customer groups for
horizons of one to 60 days (using overlapping observations) where the shaded areas correspond
to 95% bootstrap con?dence intervals. For the two ?nancial customer groups, there is a small
and short-term negative ?ow correlation which turns positive after three days. Hence, asset
managers and hedge funds tend to trade in opposite directions over very short horizons but
in the same direction over the longer run. Moreover, all correlations between ?nancial and
non-?nancial customers are signi?cantly negative at all horizons while there is no signi?cant
correlation between ?ows of the non-?nancial customer groups. These results are generally
in line with a risk sharing story where ?nancial players trade in the opposite direction of
non-?nancial market participants. This ?nding is interesting because the perception in the
literature is that risk sharing takes place in the inter-dealer market (see, e.g., Lyons, 1997)
where dealers quickly lay o? their accumulated inventory from customer orders. However,
the high concentration in today’s FX market implies that large dealers can match customer
trades to a large extent internally, allowing them to manage their inventory more e?ciently.
Given the negative correlation of ?ows we observe in the data, there clearly seems to be scope
for such warehousing of inventory risk (also see King, Osler, and Rime, 2012, on this topic).
Figure 4 about here
Drivers of ?ows. As a natural next step we seek to provide a better understanding of the
drivers of end-user order ?ows and shed light on the source of the negative ?ow correlations
discussed above. First, we examine whether the ?ows of some customer groups systematically
lead the ?ows of other groups. Second, we study whether customers’ ?ows di?er in their
response to lagged asset returns in other key asset classes. In this context we are interested
in the possible e?ects of portfolio re-balancing on the end-user demand for currencies (Hau
and Rey, 2004). To investigate this, we run panel regressions of order ?ows on lagged ?ows
and further explanatory variables, such as interest rate di?erentials (i
t
?i
t
), lagged exchange
22
rate changes over one and 20 days (?s
t
, ?s
t?1;t?20
), lagged stock returns (r
eq
t
, r
eq
t?1;t?20
), and
lagged bond returns (r
b
t
, r
b
t?1;t?20
)
OF
c
j,t+1
= ? + ?
AM
OF
AM
j,t
+ ?
HF
OF
HF
j,t
+ ?
CC
OF
CC
j,t
+ ?
PC
OF
PC
j,t
+?
1
(i
j,t
? i
t
) + ?
2
?s
j,t
+ ?
3
?s
j,t?1;t?20
(5)
+?
4
r
eq
j,t
+ ?
5
r
eq
j,t?1;t?20
+ ?
6
r
b
j,t
+ ?
7
r
b
j,t?1;t?20
+ ?
j,t+1
,
where c denotes one of the four customer groups, j denotes currencies/countries, and ?
j,t+1
=
e
t+1
+ u
j
+
j,t+1
includes both cross-sectional and time ?xed-e?ects. Standard errors are
clustered by currency pair. We use benchmark 10-year government bonds and country equity
indices from Datastream for bond and stock returns. The frequency is daily.
Results from these regressions are shown in Table V. For each customer group we report
one speci?cation which only includes lagged ?ows and one which additionally includes interest
rate di?erentials and lagged returns.
23
Looking ?rst at the speci?cations which only include
lagged ?ows, we ?nd that the ?ows of asset managers are signi?cantly related to the ?ows
by the other groups. These results (akin to simple Granger causality tests) indicate again
that asset managers trade in the opposite direction of non-?nancial customers. Flows by
hedge funds, on the other hand, do not load signi?cantly on lagged ?ows of any group, which
shows that asset managers and hedge funds show a quite di?erent behavior. Corporate
?ows are positively driven by own lagged ?ows and those of private clients, whereas ?ows by
private clients are signi?cantly negatively related to lagged hedge funds’ ?ows and signi?cantly
positively autocorrelated. In sum, there is a wealth of interrelationships between customer
?ows and their lags although it seems overambitious to interpret them in any structural way.
Table V about here
When including lagged returns as additional regressors, we ?nd that asset managers trade
against the interest rate di?erential, whereas corporate customers trade with the interest
23
Using more than one lag of ?ows in the regressions generally yields insigni?cant coe?cient estimates so
we restrict the regressions to include one lag of ?ows.
23
rate di?erential. Surprisingly, ?ows by hedge funds (and private clients) are not a?ected
by the interest di?erential suggesting that, on average, carry trading is not a dominant
driver of their ?ows over our sample.
24
Results for lagged exchange rates indicate that
asset managers are trend-followers (positive feedback traders), whereas private clients can
be described as contrarians (negative feedback traders). Asset managers’ ?ows also react
signi?cantly positively to lagged equity returns, whereas private clients’ ?ows are positively
driven by lagged bond returns. Hence, investors tend to increase their position in a currency
(against the USD) when the country’s stock market return has been high (asset managers)
or when government bond prices went up (private clients). These results do not suggest that
order ?ows are driven by portfolio rebalancing in the sense that investors sell a currency in
response to rising equity or bond prices in the country (see, e.g. the mechanism described in
Hau and Rey, 2004). However, the results strongly support the notion that ?ows of di?erent
groups are to some extent driven by the returns of other asset classes, although the factors
that in?uence ?ows clearly di?er across end-user groups.
Finally, it seems worthwhile mentioning that the estimated overall constant in the panel
regression is signi?cantly negative for hedge funds and corporates but signi?cantly positive
for private clients. Hence, hedge funds and corporates have been net sellers of the U.S. dollar,
whereas private clients have been net buyers of U.S. dollar. Given the large current account
de?cit of the U.S. over the last decade, the negative coe?cient for corporate clients is not
surprising. Also, the strong in?ow of foreign savings into U.S. capital markets over the sample
period makes sense of the positive coe?cient for private clients.
24
While this result may be surprising, it is worth bearing in mind that it relates to the aggregate hedge
funds community and on average over the full sample. It is therefore entirely possible, or even likely, that
there is variation across hedge funds and across time: For example, it may well be the case that some hedge
funds follow carry trade strategies and some follow uncovered interest parity (anti-carry trade) strategies, or
that for a particular hedge fund carry trades were implemented for the ?rst part of the sample and deleveraged
during the second part of the sample which is characterized by the recent global crisis. Thus, our result is
not intended to detect the individual behavior of hedge funds.
24
C. Di?erences in Risk Exposures
Finally, we investigate if di?erences in risk exposures can account for BMS return patterns
across FX end-users. A risk channel could explain the observed BMS excess returns if asset
managers and hedge funds tilt their portfolios towards risky currencies and earn a risk pre-
mium whereas corporate and private clients tilt their portfolios towards safe currencies and,
hence, earn low or even negative returns.
Since there are many possible sources of systematic risk that might be relevant in our case,
we consider an augmented version of the Fung and Hsieh (2002, 2004) multi-factor model as
the basis for these risk adjustments. The Fung-Hsieh model has served as the workhorse for
understanding risk exposures in the hedge funds literature (see, e.g. Patton and Ramadorai,
2013). The model relies on various U.S. equity-market and bond-market factors and also
includes the returns to trend-following strategies to capture exposure to non-linear option-
like payo?s that are quite typical of hedge funds. The trend-following factors are constructed
from portfolios of lookback straddles in various asset classes. We modify the model to make
it amenable to an analysis focused on the FX market and to allow for conditional exposures
(e.g. Ferson and Schadt, 1996; Patton and Ramadorai, 2013). The regression which serves as
the basis of these tests takes the following form
rx
p;t
= ? +
K
k=1
?
k
F
k;t
+
J
j=1
?
j
r
m;t
· z
j;t?1
+
t
. (6)
The set of factors F
t
includes the excess return on the U.S. equity market (r
m
), the change
in the yield spread of U.S. long-term bonds (?TS), and changes in credit spreads (?DF).
It further includes returns on portfolios of lookback straddles for FX futures and interest
rate futures, denoted by PTFS
FX
and PTFS
IR
respectively. We augment this sub-set of
factors from Fung and Hsieh (2004) by additional factors that are intended to capture FX-
related risk. We include the Dollar risk factor (DOL) and the carry factor (HML
FX
) by
Lustig, Roussanov, and Verdelhan (2011) as well as a factor-mimicking portfolio of global FX
volatility (V OL
FX
) (Menkho?, Sarno, Schmeling, and Schrimpf, 2012a). Following Patton
25
and Ramadorai (2013), we also allow for conditional risk exposures by interacting the equity
market risk factor r
m;t
with lagged conditioning variables z
j;t?1
. We consider (a) changes in
the TED spread (Brunnermeier, Nagel, and Pedersen, 2009), (b) changes in the VIX (Whaley,
2000), and (c) the change in the 3-month T-Bill rate.
To keep the analysis tractable and to avoid over?tting, we perform model selection of
the space of risk factors. Ideally, we want to explore the same set of factors for each of the
customer segments to be able to compare the exposures across customers and learn about
di?erences which might explain the variation in BMS excess returns. However, as ?nancial
and non-?nancial customers are likely to be very di?erent, we focus on asset managers versus
hedge funds in the ?rst set of results and private clients versus corporates in the second
set of results. More speci?cally, we perform model selection over a two-equation seemingly
unrelated regression (SUR) for asset managers’ and hedge funds’ BMS returns, and a separate
model selection for a SUR for corporate and private clients.
Results for asset managers and hedge funds are shown in Table VI. Panel A shows results
for linear models, whereas Panel B allows for conditional market exposures. We report the
four best performing models with a maximum of three factors included in the regression.
The best linear model in Panel A picks global FX volatility (V OL
FX
) as the single factor.
Other model speci?cations which also perform well tend to incorporate the trend-following
factors as well as term spread and default spread changes. Interestingly, when comparing
asset managers’ and hedge funds’ exposures to these factors, we ?nd that the signs are always
opposite. While asset managers’ BMS returns load positively on FX volatility shocks, trend-
following factors, and changes in the default spread, hedge funds load negatively on these
factors. Hence, asset managers’ FX trading positions tend to perform well in periods of
market-wide stress and when there are large returns to trend-following (which happens to be
in volatile periods, when markets trend more). Hedge funds’ FX trading positions, however,
are adversely exposed to systematic risk and market distress. These results are quite striking
as they indicate that asset managers show a very di?erent FX trading behavior and exposure
26
to systematic risk than hedge funds.
25
Table VI about here
Allowing for conditional exposures by adding interaction terms of market returns (r
m
)
with lagged changes in TED spreads and the VIX (Table VI, Panel B) leaves the main
factors chosen largely unchanged but tends to improve the model ?t. The results reported
in Panel B corroborate the previous results. The equity market exposure of asset managers
tends to decrease when the lagged TED spread and VIX increase, and vice versa for hedge
funds exposures. This is further evidence that the trading by asset managers and hedge
funds is very di?erent and that their FX trading positions are exposed di?erently to market
stress. It should be noted, though, that the alphas of asset managers and hedge funds are
signi?cantly di?erent from zero and still quite large, that is, exposure to risk does not drive
the information in ?ows for excess returns to zero.
26
Table VII about here
We repeat the analysis above for corporate and private clients’ BMS portfolios as well,
and results are shown in Table VII. However, as might be expected, risk exposures do not
matter as much for non-?nancial customers. Still, we ?nd a negative equity market exposure
for both groups (Panel A), which increases (decreases) following increases in the TED spread
for private (corporate) clients. Moreover, there is some evidence that the private clients’
BMS portfolio has positive exposure to changes in credit spreads.
IV. Additional Tests and Robustness
We provide extensive robustness checks to all our main results. Below, we ?rst examine the
e?ect of transaction costs and then turn to a brief discussion of various other robustness tests
25
Additional evidence is provided in the Internet Appendix. Table A.17 summarizes exposures to equity
factors, Table A.18 considers FX factors, Table A.19 focuses on the Fung and Hsieh (2002) factors, and Table
A.20 reports results for the BMS portfolio based on total ?ows for completeness.
26
Table A.16 reports pricing errors for the cross-section of order ?ow portfolios. Speci?cally, we report the
Gibbons, Ross, and Shanken (1989) test for the null that the alphas are jointly equal to zero. Corroborating
the time-series regressions in Tables VI and VII, the test always rejects the null of zero alphas.
27
for which results can be found in a separate Internet Appendix to conserve space.
A. Transaction Costs
Our analysis above is intentionally quiet on questions of exploitability of order ?ow infor-
mation for trading strategies or the e?ects of transaction costs. This is because our data on
customer order ?ow are not available to participants in the broader market and thus cannot
form the basis for a trading strategy, except for UBS itself or for one of the other few large
dealers with access to similar customer ?ows. However, an interesting issue is whether owners
of this type of private information, that is, large FX dealer banks with a large concentration
of informed customers, could potentially employ this information by simply piggy-backing
the order ?ow of their customers.
27
To examine this question, we compute net excess returns for BMS portfolios by adjusting
for bid-ask spreads.
28
We investigate returns to strategies with varying portfolio re-balancing
frequencies to balance the e?ects of transaction costs and using the most recent information.
Figure 5 presents the results for re-balancing frequencies from 1 to 10 days. The dashed
lines give average excess returns (p.a.) and 95% con?dence intervals for excess returns before
transaction costs to show the e?ect of di?erent re-balancing periods. The solid line and
shaded area show average net excess returns (p.a.) and 95% con?dence intervals when taking
transaction costs into account.
Figure 5 about here
We ?nd that exploiting the information in ?ows should in practice be feasible for a dealer.
This holds for both sets of currencies T15 and T9. Average excess returns are signi?cantly
27
Obviously the data should be used in respect of clients’ con?dentiality and the speci?c compliance agree-
ments governing customers’ transactions.
28
The bid-ask spread data available is for quoted spreads and not e?ective spreads. It is well known that
quoted spreads are much higher than e?ective spreads (Lyons, 2001). We therefore follow earlier work, e.g.,
Goyal and Saretto (2009), and employ 50% of the quoted bid-ask spread as the actual spread. Even this
number seems conservative, though. First, banks with access to this kind of customer order ?ow data are
big dealers and pay very low spreads since they are key market makers. Second, Gilmore and Hayashi (2011)
?nd in a recent study that transaction costs due to bid-ask spreads are likely to be much lower than our 50%
rule. This ?nding was corroborated by our own conversations with UBS dealers.
28
di?erent from zero for all re-balancing horizons and economically attractive even for short
frequencies. These results clearly demonstrate the potential value of being able to observe
order ?ow by customers, especially the one by informed customers such as asset managers or
leveraged funds.
B. Further Robustness Checks
We check whether our results are robust to other sensible choices of standardizing ?ows. First,
we check whether standardizing ?ows over longer horizons of one and three years produce
similar results (see Tables A.7 and A.8 in the Internet Appendix). They do. Second, we
measure ?ows relative to total currency trading volume (obtained from the BIS FX triennial
surveys).
29
Table A.9 shows the results, which also indicate signi?cant predictability of
returns by order ?ows. Third, we standardize ?ows by additionally demeaning ?ows over
the rolling window (Table A.10). As a ?nal step, we form portfolios based on ?ows for all
currencies via the following procedure: We cross-sectionally standardize order ?ows for day
t (we subtract the cross-sectional mean and divide by the standard deviation), rescale these
standardized ?ows to sum to two in absolute value, and then use these as weights to form a
portfolio held from day t to t + 1 (Table A.11). Our results remain robust.
We also check if order ?ows forecast returns at longer horizons. To this end, we use
an exponential moving average to sum order ?ows into the past and then use these lower
frequency ?ows to build BMS portfolios which we rebalance every 2, 3, 4, 5, 10, 20, 60 trading
days. We report results for two di?erent decay parameters (0.25 and 0.75) in the exponential
moving average in Table A.12. We ?nd that predictability dies out fairly quickly and that
only asset managers’ ?ows have some predictive power over longer horizons of up to one
month (20 trading days).
To rule out that a simple liquidity story drives our predictability results, we also look at
a sub-sample of the four most liquid currency pairs in our sample: EUR/USD, JPY/USD,
29
We linearly interpolate between the data of the BIS survey to obtain a daily time-series of trading volumes
in USD for the nine developed currencies and then use the ratio of customer ?ows to total trading volumes
as our sorting variable.
29
GBP/USD, and CHF/USD. Table A.13 reports results for BMS portfolio returns and Figure
A.1 shows results for BMS returns in event time (similar to Figure 3 in the main text). We
?nd that our main results remain qualitatively unchanged.
We next explore whether a speci?c currency is driving the pro?tability of the order ?ow
portfolios. To investigate this issue, we rely on a cross-validation setting in which we form
portfolios as before but in each case delete one of the available currencies. For example, we
exclude the EUR/USD pair and compute BMS portfolio returns for the remaining 14 (total
order ?ows) or 8 currency pairs (disaggregated order ?ows). Table A.14 summarizes the
results from this exercise. We always ?nd the same general return pattern, that is, our main
?ndings do not depend strongly on any particular currency.
V. Conclusion
This paper empirically addresses three related questions to improve our understanding of
the ecology of the world’s largest ?nancial market, the FX market. First, given that the
FX market is fairly opaque with a large concentration of market making in the hands of
a few large intermediaries, how valuable is it for dealers to observe a large proportion of
the market’s order ?ow? Second, do FX end-users share risks among themselves, or is their
trading highly correlated and unloaded to the dealers and the inter-dealer market? Third,
how can we understand the trading behavior, trading styles and risk exposures of various key
players in FX markets, and how is this linked to risk sharing?
We ?nd that observing customer order ?ows in a dark market is highly valuable from
the dealer’s perspective. Currency excess returns to portfolios mimicking aggregate customer
order ?ows in real-time are about 10% p.a. and highly signi?cant. In addition, trading in FX
markets (as in other OTC markets) is anonymous, meaning that dealers know the identity
of their clients. Incorporating this feature into our setup, we ?nd excess returns as high as
15% p.a., that is, non-anonymity further increases the informational advantage of dealers.
The ?ows by asset managers have the strongest predictive power for exchange rates, likely
re?ecting the processing of fundamental information. Their ?ows have a permanent forecast
30
power, whereas ?ows originating from the other groups only predict transitory changes of
exchange rates. All this suggests that dealers have a strong incentive to gain large market
shares (besides other reasons such as economies of scale in the provision of trading infrastruc-
ture, for example) and to set up trading in a way that reveals end-users’ identities. These
?ndings about strong information asymmetries and incentives should be useful to inform
policy discussions on the appropriate framework for OTC markets.
We also show that the main segments of end-users di?er markedly in their trading strate-
gies and hedging demands. Asset managers, for instance, tend to be trend-followers, whereas
individual investors behave as contrarians. Hedge funds (on aggregate) do not seem to fall
in any of these two categories. Moreover, ?ows of di?erent end-user segments tend to be
negatively correlated over longer horizons. These ?ndings suggest that risk sharing also takes
place among end-users and not only via the inter-dealer market as suggested by previous FX
microstructure research.
Taken together, these results bring some light into one of the main dark ?nancial mar-
kets. Our ?ndings suggest that the FX market is populated by quite heterogeneous market
participants and that we gain valuable economic insights from observing their transactions
and learning about their di?erent predictive ability, trading motives, trading styles, and risk
exposures.
31
References
Akram, Q. Farooq, Dag?nn Rime, and Lucio Sarno, 2008, Arbitrage in the foreign exchange
market: Turning on the microscope, Journal of International Economics 76, 237–253.
Albuquerque, Rui, Eva de Francisco, and Luis Marques, 2008, Marketwide private informa-
tion in stocks: Forecasting currency returns, Journal of Finance 63, 2297–2343.
Ang, Andrew, and Joseph Chen, 2010, Yield curve predictors of foreign exchange returns,
Working Paper, Columbia University.
Barroso, Pedro, and Pedro Santa-Clara, 2011, Beyond the carry trade: Optimal currency
portfolios, Working Paper, NOVA School of Business and Economics.
Berger, D.W., A.P. Chaboud, S.V. Chernenko, E. Howorka, and J.H. Wright, 2008, Order ?ow
and exchange rate dynamics in electronic brokerage system data, Journal of International
Economics 75, 93–109.
Berger, D.W., A.P. Chaboud, and E. Hjalmarsson, 2009, What drives volatility persistence
in the foreign exchange market?, Journal of Financial Economics 94, 192–213.
BIS, 2010, Triennial central bank survey - report on global foreign exchange market activity
in 2010, Bank for International Settlements, Basel.
Brandt, Michael W., and Kenneth A. Kavajecz, 2004, Price discovery in the U.S. treasury
market: The impact of order?ow and liquidity on the yield curve, Journal of Finance 59,
2623–2654.
Brunnermeier, Markus, Stefan Nagel, and Lasse Pedersen, 2009, Carry trades and currency
crashes, NBER Macroeconomics Annual 2008 23, 313–347.
Burnside, Craig, Martin Eichenbaum, Isaac Kleshchelski, and Sergio Rebelo, 2011, Do Peso
problems explain the returns to the carry trade?, Review of Financial Studies 24, 853–891.
Carhart, Mark M., 1997, On persistence in mutual fund performance, Journal of Finance 52,
57–82.
Chinn, Menzie D., and Michael Moore, 2011, Order ?ow and the monetary model of exchange
rates: Evidence from a novel data set, Journal of Money, Credit and Banking 43, 1599–
1624.
Du?e, Darrell, 2012, Dark Markets: Asset Pricing and Information Transmission in Over-
the-Counter Markets (Princeton University Press).
32
Evans, Martin D., 2010, Order ?ows and the exchange rate disconnect puzzle, Journal of
International Economics 80, 58–71.
, and Richard K. Lyons, 2002a, Order ?ow and exchange rate dynamics, Journal of
Political Economy 110, 170–180.
, 2002b, Time-varying liquidity in foreign exchange, Journal of Monetary Economics
49, 1025–1051.
, 2005, Meese-Rogo? redux: Micro-based exchange rate forecasting, American Eco-
nomic Review, Papers and Proceedings 95, 405–414.
, 2007, Exchange rate fundamentals and order ?ow, Working Paper, Georgetown
University.
, 2008, How is macro news transmitted to exchange rates?, Journal of Financial
Economics 88, 26–50.
Fama, Eugene F., 1984, Forward and spot exchange rates, Journal of Monetary Economics
14, 319–338.
Fan, M., and Richard K. Lyons, 2003, Customer trades and extreme events in foreign ex-
change, in P. Mizen, ed.: Monetary History, Exchange Rates and Financial Markets: Es-
says in Honour of Charles Goodhart . pp. 160–179 (Edward Elgar: Northampton).
Farhi, Emmanuel, Samuel P. Fraiberger, Xavier Gabaix, Romain Ranciere, and Adrien
Verdelhan, 2009, Crash risk in currency markets, Working Paper, MIT Sloan.
Ferson, Wayne E., and R. Schadt, 1996, Measuring fund strategy and performance in changing
economic conditions, Journal of Finance 51, 425–462.
Froot, Kenneth A., and Tarum Ramadorai, 2005, Currency returns, intrinsic value, and
institutional-investor ?ows, Journal of Finance 60, 1535–1566.
Fung, William, and David Hsieh, 2001, The risk in hedge fund strategies: Theory and evidence
from trend followers, Review of Financial Studies 14, 313–341.
Fung, William, and David A. Hsieh, 2002, Asset-based style factors for hedge funds, Financial
Analysts Journal 58, 16–27.
, 2004, Hedge fund benchmarks: A risk based approach, Financial Analysts Journal
60, 65–80.
Gibbons, Michael R., Stephen A. Ross, and Jay Shanken, 1989, A test of the e?ciency of a
given portfolio, Econometrica 57, 1121–1152.
33
Gilmore, Stephen, and Fumio Hayashi, 2011, Emerging market currency excess returns,
American Economic Journal: Macroeconomics 3, 85–111.
Goyal, Amit, and Alessio Saretto, 2009, Cross-section of option returns and volatility, Journal
of Financial Economics 94, 310–326.
Hasbrouck, Joel, 1991a, Measuring the information content of stock trades, Journal of Fi-
nance 46, 179–207.
, 1991b, The summary informativeness of stock trades: An econometric analysis,
Review of Financial Studies 4, 571–595.
Hau, Harald, and Helene Rey, 2004, Can portfolio rebalancing explain the dynamics of eq-
uity returns, equity ?ows, and exchange rates?, American Economic Review, Papers and
Proceedings 94, 126–133.
Ito, Takatoshi, Richard K. Lyons, and Michael T. Melvin, 1998, Is there private information
in the FX market?, Journal of Finance 53, 1111–1130.
Kaniel, Ron, Gideon Saar, and Sheridan Titman, 2008, Individual investor trading and stock
returns, Journal of Finance 63, 273–310.
King, Michael R., Carol Osler, and Dag?nn Rime, 2012, Foreign exchange market structure,
players and evolution, in Jessica James, Ian Marsh, and Lucio Sarno, ed.: Handbook of
Exchange Rates . pp. 3–44 (Wiley).
Love, Ryan, and Richard Payne, 2008, Macroeconomic news, order ?ows, and exchange rates,
Journal of Financial and Quantitative Analysis 43, 467–488.
Lustig, Hanno, Nikolai Roussanov, and Adrien Verdelhan, 2011, Common risk factors in
currency markets, Review of Financial Studies 24, 3731–3777.
Lustig, Hanno, and Adrien Verdelhan, 2007, The cross section of foreign currency risk premia
and consumption growth risk, American Economic Review 97, 89–117.
Lyons, Richard K., 1997, A simultaneous trade model of the foreign exchange hot potato,
Journal of International Economics 42, 275–298.
, 2001, The Microstructure Approach to Exchange Rates (MIT Press: Cambridge).
Marsh, Ian W., and C. O’Rourke, 2005, Customer order ?ow and exchange rate movements:
Is there really information content?, Working Paper, Cass Business School.
Menkho?, Lukas, Lucio Sarno, Maik Schmeling, and Andreas Schrimpf, 2012a, Carry trades
and global foreign exchange volatility, Journal of Finance 67, 681–718.
34
, 2012b, Currency momentum strategies, Journal of Financial Economics 106, 660–
684.
Menkho?, Lukas, and Maik Schmeling, 2010, Whose trades convey information? Evidence
from a cross-section of traders, Journal of Financial Markets 13, 101–128.
Moore, Michael, and Richard Payne, 2011, On the existence of private information in FX
markets, Journal of Banking and Finance 35, 1250–1262.
Neely, Christopher, Paul Weller, and Joshua M. Ulrich, 2009, The adaptive markets hypoth-
esis: Evidence from the foreign exchange market, Journal of Financial and Quantitative
Analysis 44, 467–488.
Newey, Whitney K., and Kenneth D. West, 1987, A simple, positive semi-de?nite, het-
eroskedasticity and autocorrelation consistent covariance matrix, Econometrica 55, 703–
708.
Osler, Carol L., Alexander Mende, and Lukas Menkho?, 2011, Price discovery in foreign
exchange markets, Journal of International Money and Finance 30, 1696–1718.
Patton, Andrew J., and Tarun Ramadorai, 2013, On the high-frequency dynamics of hedge
fund risk exposures, Journal of Finance forthcoming.
Patton, Andrew J., and Allan Timmermann, 2010, Monotonicity in asset returns: New
tests with applications to the term structure, the CAPM, and portfolio sorts, Journal of
Financial Economics 98, 605–625.
Payne, Richard, 2003, Informed trade in spot foreign exchange markets: An empirical inves-
tigation, Journal of International Economics 61, 307–329.
Phylaktis, Kate, and L. Chen, 2010, Asymmetric information, price discovery and macroe-
conomic announcements in FX market: Do top trading banks know more?, International
Journal of Finance and Economics 15, 228–246.
Rime, Dag?nn, Lucio Sarno, and Elvira Sojli, 2010, Exchange rates, order ?ow and macroe-
conomic information, Journal of International Economics 80, 72–88.
Sager, Michael, and Mark P. Taylor, 2008, Commercially available order ?ow data and ex-
change rate movements: Caveat emptor, Journal of Money, Credit and Banking 40, 583–
625.
Whaley, R.E., 2000, The investor fear gauge, Journal of Portfolio Management 26, 12–17.
35
Table I. Descriptive Statistics for FX Customer Order Flows
This table shows descriptive statistics for total order ?ows for the 15 currencies in our sample.
Flows are measured in billions USD and all currencies are against the USD. A positive
(negative) ?ow means that there is net buying (selling) pressure for the respective currency.
The frequency is daily and the sample is from January 2001 to May 2011. Currencies included
are the Australian Dollar (AUD), Canadian Dollar (CAD), Swiss Franc (CHF), Euro (EUR),
Great Britain Pound (GBP), Japanese Yen (JPY), Norwegian Krone (NOK), New Zealand
Dollar (NZD), Swedish Krona (SEK), Brazilian Real (BRL), Hong Kong Dollar (HKD),
(South) Korean Won (KRW), Mexican Peso (MXN), Singapore Dollar (SGD), and South
African Rand (ZAR).
Mean Median Std Skew Kurt AC(1) p-val.
Panel A. Developed Markets
AUD -0.003 -0.001 0.197 -2.69 55.89 -0.01 (0.15)
CAD 0.007 0.003 0.169 1.98 43.02 0.00 (1.00)
CHF 0.020 0.012 0.324 0.11 74.88 0.02 (0.00)
EUR -0.063 -0.041 0.656 -3.31 79.53 0.03 (0.00)
GBP -0.001 -0.002 0.484 -5.82 270.70 0.01 (0.03)
JPY 0.027 0.019 0.412 1.88 55.87 0.03 (0.00)
NOK 0.003 0.000 0.067 0.60 49.20 0.08 (0.00)
NZD -0.002 0.000 0.070 -1.78 51.30 0.14 (0.00)
SEK 0.001 0.000 0.070 1.60 39.84 0.01 (0.04)
Panel B. Emerging Markets
BRL -0.004 0.000 0.068 -1.15 30.50 0.03 (0.00)
HKD 0.006 0.000 0.079 2.32 35.39 0.01 (0.02)
KRW -0.005 0.000 0.070 -0.15 59.45 0.05 (0.00)
MXN -0.002 0.000 0.049 -0.67 27.78 0.06 (0.00)
SGD 0.000 0.000 0.068 -4.39 110.15 0.06 (0.00)
ZAR 0.002 0.000 0.068 -0.91 36.38 0.16 (0.00)
36
Table II. Order Flow Portfolios: Excess Returns
This table shows average annualized portfolio excess returns for currency portfolios sorted on
lagged order ?ow. We standardize order ?ow over a rolling window of 60 trading days prior to
the order ?ow signal as outlined in the text. Column“Av”shows average excess returns across
all currencies, column “BMS” (bought minus sold) reports average excess returns for long-
short portfolios in currencies with the highest versus lowest order ?ow. Numbers in brackets
are t-statistics based on Newey-West standard errors whereas numbers in parentheses show
(annualized) Sharpe Ratios. Columns ‘MR’, ‘Up’, and ‘Down’ report p-values for tests of
return monotonicity. The frequency is daily and the sample is from January 2001 to May
2011. Panel A shows results for total order ?ows and all 15 markets (T15) as well as for total
order ?ows and the subsample of nine developed markets (T9). Panel B reports results for
order ?ows disaggregated by customer type (asset managers AM, hedge funds HF, corporate
clients CC, private clients PC).
Panel A. Total Order Flows
P
1
P
2
P
3
P
4
P
5
Av. BMS MR Up Down
T15 0.82 1.05 6.15 6.77 11.13 5.18 10.31 0.00 0.00 –
[0.29] [0.37] [2.23] [2.40] [4.04] [2.20] [4.05]
(0.09) (0.11) (0.71) (0.77) (1.21) (0.69) (1.26)
T9 0.34 2.24 8.21 12.76 5.89 12.43 0.00 0.00 –
[0.10] [0.74] [2.60] [4.17] [2.15] [4.68]
(0.03) (0.23) (0.80) (1.23) (0.66) (1.45)
Panel B. Disaggregated Order Flows
AM -1.13 3.75 6.30 14.31 15.43 0.00 0.00 –
[-0.35] [1.24] [2.04] [4.63] [5.72]
(-0.11) (0.38) (0.62) (1.38) (1.79)
HF -0.32 6.05 6.26 9.78 10.09 0.04 0.00 –
[-0.10] [2.04] [1.94] [3.02] [3.94]
(-0.03) (0.61) (0.59) (0.94) (1.20)
CO 6.90 5.27 7.02 2.61 -4.29 0.35 – 0.09
[2.15] [1.73] [2.16] [0.84] [-1.66]
(0.67) (0.53) (0.66) (0.26) (-0.51)
PC 12.71 6.69 2.90 -1.30 -14.01 0.00 – 0.00
[4.06] [2.18] [0.93] [-0.41] [-5.20]
(1.23) (0.67) (0.28) (-0.13) (-1.55)
37
Table III. Order Flow Portfolios: Marginal Forecast Performance for Longer Horizons
This table shows average excess returns (p.a.) for BMS portfolios sorted on lagged order ?ow
as in Table II. t-statistics based on Newey-West standard errors are reported in brackets.
However, we do not only sort on order ?ow of the previous day but also allow for longer
lags of up to nine days between order ?ow signals and portfolio formation. Portfolios are
rebalanced daily. T15 denotes portfolios sorts on total order ?ows and the sample of all 15
currencies, and T9 denotes portfolios sorts on total order ?ows and the sample of 9 developed
currencies; AM, HF, CC, and PC denote portfolios sorts on asset managers’, hedge funds’,
corporate clients’, and private clients’ order ?ows, respectively.
Lags between order ?ow signal and portfolio formation (days)
1 2 3 4 5 6 7 8 9 10
T15 10.31 24.63 10.22 -1.11 3.02 0.20 0.31 1.93 -2.32 -0.43
[4.05] [8.94] [4.38] [-0.44] [1.28] [0.09] [0.13] [0.84] [-0.95] [-0.19]
T9 12.43 24.27 7.44 -4.17 5.39 -1.55 2.28 1.33 -1.08 -1.75
[4.68] [8.73] [2.99] [-1.61] [2.00] [-0.61] [0.90] [0.51] [-0.42] [-0.71]
AM 15.43 24.86 8.27 -1.29 2.17 0.62 -0.20 3.37 2.26 -2.79
[5.72] [8.80] [3.03] [-0.47] [0.87] [0.23] [-0.07] [1.22] [0.82] [-0.97]
HF 10.09 28.22 2.05 -2.94 0.14 -6.19 2.84 -0.29 -4.66 -1.05
[3.94] [9.26] [0.79] [-1.15] [0.05] [-2.39] [1.12] [-0.10] [-1.77] [-0.40]
CC -4.29 -8.13 -1.47 2.25 -4.98 1.91 -0.01 1.40 -0.33 2.80
[-1.66] [-2.86] [-0.49] [0.88] [-1.93] [0.74] [0.00] [0.56] [-0.12] [1.08]
PC -14.01 -33.77 3.21 1.82 -3.29 -0.77 2.27 -1.35 0.65 2.10
[-5.20] [-10.80] [1.24] [0.67] [-1.15] [-0.27] [0.86] [-0.52] [0.24] [0.78]
38
Table IV. Panel Regressions of Currency Returns on Lagged Order Flow
This table reports results for panel regressions of currency excess returns (rx
t+1
) on lagged
customer order ?ow (OF
t
) and control variables (the interest rate di?erential i
j,t
?i
t
, lagged
excess returns over the previous day rx
t
, lagged excess returns over the prior 60 days
rx
t?1;t?60
). Order ?ow is measured in billion USD. T15 and T9 refer to total order ?ow
for all 15 currencies and the sample of developed market currencies, respectively. The re-
gressions in (v) and (vi) also include disaggregated order ?ow for asset managers, AM, hedge
funds, HF, corporate clients, CC, and private clients, CC). In each speci?cation, we show
results both for pooled regressions (pooling over all currency pairs) and for speci?cations
with currency pair- and time-?xed e?ects. t-statistics based on clustered standard errors (by
currency pair) are reported in brackets.
(i) (ii) (iii) (iv) (v) (vi)
const. 0.015 -0.010 0.020 -0.015 0.020 -0.015
[4.92] [-3.56] [6.81] [-2.77] [6.41] [-2.72]
OF
T15
t
0.025 0.023
[3.72] [3.49]
OF
T9
t
0.023 0.021
[3.42] [3.10]
OF
AM
t
0.043 0.038
[4.94] [4.29]
OF
HF
t
0.011 0.010
[2.12] [1.95]
OF
CC
t
-0.017 -0.014
[-2.63] [-2.23]
OF
PC
t
-0.028 -0.024
[-3.43] [-3.00]
i
j,t
? i
t
1.036 1.720 0.897 0.108 0.936 0.348
[7.74] [2.45] [2.59] [0.11] [2.75] [0.36]
rx
t
0.002 -0.010 0.001 -0.013 -0.006 -0.018
[0.27] [-1.21] [0.12] [-1.37] [0.57] [-1.86]
rx
t?1;t?60
0.000 -0.010 0.000 -0.015 0.000 -0.015
[-0.79] [-3.56] [-1.37] [-2.77] [-0.96] [-2.72]
Country dummies NO YES NO YES NO YES
Time dummies NO YES NO YES NO YES
R
2
0.002 0.024 0.001 0.030 0.006 0.045
obs 37,936 37,936 23,436 23,436 23,436 23,436
39
Table V. Drivers of Customer FX Order Flow: Panel Regressions
This table reports results for panel regressions of customer order ?ows (OF) on lagged cus-
tomer order ?ow (OF
t
for asset managers, AM, hedge funds, HF, corporate clients, CC, and
private clients, CC). The regressions also consider lagged returns on various asset classes
as additional regressors (the interest rate di?erential i
j,t
? i
t
, lagged exchange rate changes
over the previous day ?s
t
and over the prior 20 trading days ?s
t?1,t?20
, lagged country-level
equity returns over the previous trading day r
eq
t
and over the prior 20 trading days r
eq
t?1;t?20
),
and lagged country-level government bond returns r
b
t
(10-year maturity benchmark bonds).
t-statistics based on clustered standard errors (by currency pair) are reported in brackets and
we account for currency pair- and time-?xed e?ects.
Dependent variable: Customer order ?ows
OF
AM
t+1
OF
HF
t+1
OF
CC
t+1
OF
PC
t+1
OF
AM
t
0.035 0.033 0.013 0.012 -0.010 -0.009 -0.005 -0.003
[4.46] [4.22] [1.79] [1.73] [-1.13] [-1.08] [-0.61] [-0.38]
OF
HF
t
0.034 0.031 0.008 0.007 -0.009 -0.008 -0.037 -0.350
[2.75] [2.66] [0.57] [0.50] [-1.70] [-1.55] [-2.59] [-2.56]
OF
CC
t
-0.017 -0.016 0.000 0.000 0.035 0.034 -0.012 -0.013
[-2.58] [-2.53] [0.02] [0.05] [2.93] [2.88] [-1.47] [-1.62]
OF
PC
t
-0.026 -0.025 -0.005 -0.004 0.025 0.024 0.027 0.025
[-2.10] [-2.05] [-0.67] [-0.61] [2.47] [2.46] [2.21] [2.02]
i
j,t
? i
t
-0.150 0.102 0.413 0.185
[-2.05] [0.80] [2.51] [1.02]
?s
t
3.541 1.769 -1.312 -4.187
[4.97] [1.47] [-1.15] [-2.52]
?s
t?1,t?20
1.012 0.612 -0.741 -2.187
[1.97] [0.50] [-0.45] [-1.52]
r
eq
t
1.251 0.399 -1.164 -0.226
[2.56] [0.41] [-2.37] [-0.34]
r
eq
t?1;t?20
0.347 -0.113 0.205 -0.225
[1.44] [-0.52] [1.17] [-0.34]
r
b
t
-3.730 -5.170 -1.135 10.145
[-1.56] [-1.26] [-0.57] [2.68]
r
b
t?1;t?20
-0.019 0.278 0.626 1.151
[-0.03] [-0.55] [1.04] [2.03]
const. 0.008 -0.002 -0.078 -0.089 -0.320 -0.295 0.039 0.076
[0.71] [0.03] [-4.42] [-3.47] [-7.27] [-6.01] [4.77] [4.41]
Country dummies YES YES YES YES YES YES YES YES
Time dummies YES YES YES YES YES YES YES YES
R
2
0.013 0.015 0.011 0.011 0.029 0.030 0.015 0.018
obs 23,796 23,796 23,796 23,796 23,796 23,796 23,796 23,796
40
Table VI. Risk Exposures of Financial End-Users
This table reports regression results for the risk exposures of the BMS portfolios of ?nancial FX
market end-users, that is, asset managers and hedge funds. The methodological framework in Panel
A is a modi?ed linear Fung-Hsieh model with eight factors as outlined in the main text. Panel
B also accounts for conditional equity market exposures by including additional interaction terms.
The three conditioning variables are ?rst di?erences of the TED spread, the VIX and the 3-month
T-Bill rate. The Table shows results for four parsimonious model speci?cations where the factors
are selected according to the Schwarz criterion as outlined in the main text. Results for the other
factors are not reported. We further report the estimated intercept ?, the adjusted R
2
and the
BIC computed for the two-equation system (Sys-BIC). Below the regression coe?cients, t-statistics
based on HAC standard errors are reported (in brackets).
A. Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
PTFS
FX
2.35 -2.68
[2.65] [-2.51]
PTFS
IR
3.07 2.18 -1.33 -1.16
[4.03] [2.86] [-1.85] [-1.67]
?TS -2.03 0.38
[-2.06] [0.59]
?DF 3.15 3.65 -3.58 -3.67
[2.83] [2.87] [-2.61] [-2.69]
VOL
FX
0.07 0.06 -0.07 -0.05
[2.44] [2.13] [-2.50] [-2.09]
? 1.46 1.40 1.26 1.23 0.71 0.78 0.89 0.90
[5.32] [5.45] [5.68] [5.25] [3.10] [3.49] [4.01] [3.97]
¯
R
2
0.10 0.12 0.21 0.15 0.11 0.14 0.10 0.10
Sys-BIC 3.53 3.53 3.54 3.54 3.53 3.53 3.54 3.54
B. Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
r
m
·?VIX(t-1) -0.25 -0.26 -0.26 0.15 0.17 0.16
[-2.95] [-2.69] [-2.47] [2.30] [3.08] [3.32]
r
m
·?TED(t-1) -0.18 -0.19 -0.20 -0.20 0.32 0.37 0.33 0.34
[-2.44] [-2.82] [-3.06] [-2.48] [4.14] [5.74] [4.58] [4.63]
PTFS
FX
2.52 -2.54
[2.31] [-2.48]
PTFS
IR
2.11 -0.87
[2.79] [-1.35]
VOL
FX
0.05 0.06 -0.05 -0.05
[2.08] [2.15] [-2.10] [-2.29]
? 1.35 1.18 1.20 1.44 0.72 0.85 0.86 0.67
[5.57] [5.55] [5.38] [5.31] [3.50] [3.89] [4.05] [3.21]
¯
R
2
0.17 0.18 0.15 0.12 0.21 0.18 0.21 0.19
Sys-BIC 3.45 3.46 3.47 3.47 3.45 3.46 3.47 3.47
41
Table VII. Risk Exposures of Private and Corporate Clients
This table reports regression results for the risk exposures of the BMS portfolios computed from
the ?ows of corporate clients (CC) or private clients (PC). The methodological framework in Panel
A is a modi?ed linear Fung-Hsieh model with eight factors as outlined in the main text. Panel
B also accounts for conditional equity market exposures by including additional interaction terms.
The three conditioning variables are ?rst di?erences of the TED spread, the VIX and the 3-month
T-Bill rate. The Table shows results for four parsimonious model speci?cations where the factors
are selected according to the Schwarz criterion as outlined in the main text. Results for the other
factors are not reported. We further report the estimated intercept ?, the adjusted R
2
and the
BIC computed for the two-equation system (Sys-BIC). Below the regression coe?cients, t-statistics
based on HAC standard errors are reported (in brackets).
A. Corporate Clients Private Clients
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
r
m
-0.14 -0.11 -0.14 -0.07 -0.09 -0.10
[-2.32] [-1.63] [-1.94] [-1.56] [-2.27] [-2.43]
PTFS
IR
-1.83 -0.70
[-1.04] [-0.52]
?DF -3.43 -2.20 2.57 3.18
[-1.38] [-0.99] [2.81] [3.51]
? -0.30 -0.31 -0.37 -0.25 -1.16 -1.15 -1.19 -1.12
[-1.49] [-1.5] [-1.73] [-1.42] [-4.27] [-4.13] [-4.34] [-4.04]
¯
R
2
0.07 0.04 0.01 0.07 0.06 0.03 0.04 0.03
Sys-BIC 3.93 3.94 3.94 3.96 3.93 3.94 3.94 3.96
B. Corporate Clients Private Clients
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
r
m
-0.08 -0.10 -0.11 -0.13
[-1.30] [-1.50] [-2.48] [-3.05]
r
m
·? TED(t-1) 0.47 0.44 0.40 0.49 -0.20 -0.24 -0.31 -0.12
[2.95] [2.29] [2.1] [2.56] [-2.53] [-2.27] [-2.68] [-0.99]
PTFS
IR
-0.94 -1.38
[-0.8] [-1.45]
?DF 0.39 2.51
[0.25] [1.71]
? -0.46 -0.41 -0.37 -0.46 -1.18 -1.12 -1.06 -1.19
[-2.04] [-1.91] [-1.85] [-2.02] [-4.25] [-4.27] [-4.20] [-4.46]
¯
R
2
0.13 0.14 0.15 0.12 0.02 0.07 0.09 0.04
Sys-BIC 3.82 3.83 3.87 3.88 3.82 3.83 3.87 3.88
42
Figure 1. Cumulative Excess Returns on BMS Portfolios
01 02 03 04 05 06 07 08 09 10 11
-20
0
20
40
60
80
100
120
140
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
All countries
01 02 03 04 05 06 07 08 09 10 11
-20
0
20
40
60
80
100
120
140
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
Developed countries
01 02 03 04 05 06 07 08 09 10 11
0
20
40
60
80
100
120
140
160
180
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
Asset Managers
01 02 03 04 05 06 07 08 09 10 11
-20
0
20
40
60
80
100
120
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
Hedge Funds
01 02 03 04 05 06 07 08 09 10 11
-60
-50
-40
-30
-20
-10
0
10
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
Corporate Clients
01 02 03 04 05 06 07 08 09 10 11
-150
-100
-50
0
50
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
Private Clients
This ?gure shows cumulative log excess returns for a long-short portfolio based on total order
?ows and all countries (T15), total ?ows and developed markets (T9), asset manager ?ows
(AM), hedge fund ?ows (HF), corporate clients’ ?ows (CC), and private clients’ ?ows (PC).
The sample period is daily from January 2001 – May 2011.
43
Figure 2. Cumulative Post-Formation Exchange Rate Changes
This ?gure shows average cumulative spot exchange rate changes for BMS portfolios based
on total ?ows and disaggregated ?ows over the ?rst 30 days after portfolio formation. We
use daily data so that post-formation periods overlap. Shaded areas correspond to a 95%
con?dence interval obtained from a moving-block bootstrap with 1,000 repetitions.
44
Figure 3. BMS Excess Returns in Event Time
This ?gure shows BMS portfolio excess returns (solid lines) in event time, from ?ve days prior to portfolio
formation (t = ?5, the day of portfolio formation (t = 0), up to ten days after portfolio formation (t = 10).
BMS excess returns are annualized and in percent. The shaded areas correspond to 95% con?dence intervals
based on Newey/West standard errors. The frequency is daily and the sample is from January 2001 – May
2011.
45
Figure 4. Correlation of Customer Order Flows Over Longer Horizons
This ?gure shows average correlation coe?cients between customer order ?ows (left panel) for horizons of
1, 2, ..., 60 trading days. Average correlations between ?ows are based on the average correlation across all nine
currency pairs. A horizon of one day corresponds to (non-overlapping) daily observations, whereas correlations
for longer horizons are based on (overlapping) sums of daily observations. Shaded areas correspond to
bootstrapped 95% con?dence intervals based on a moving-block bootstrap with 1,000 repetitions. The sample
period is January 2001 – May 2011.
46
Figure 5. Rebalancing Frequency and Net Excess Returns
This ?gure shows average annualized excess returns for BMS portfolios based on total and
disaggregated order ?ows for di?erent rebalancing frequencies ranging from one to ten days.
The dotted lines show excess returns and a 95% con?dence interval based on Newey-West
standard errors before transaction costs whereas the solid line and shaded area show net
excess returns and a 95% con?dence interval based on Newey-West standard errors after
transaction costs.
47
Internet Appendix to accompany
Information Flows in Dark Markets:
Dissecting Customer Currency Trades
(Not for Publication)
48
Table A.1. Descriptive Statistics for Disaggregated Customer Order Flows
This table shows descriptive statistics for customer order ?ows which are available for the
nine major markets in our sample, that is, the Australian Dollar (AUD), Canadian Dollar
(CAD), Swiss Franc (CHF), Euro (EUR), Great Britain Pound (GBP), Japanese Yen (JPY),
Norwegian Krone (NOK), New Zealand Dollar (NZD), Swedish Krona (SEK). Flows are
measured in billions (in USD) and all currencies are against the USD. A positive (negative)
?ow means that there is net buying (selling) pressure in the foreign currency against the
USD. We report means, medians, standard deviations, skewness, kurtosis, and ?rst-order
autocorrelation coe?cients (AC(1)) for all four customer groups’ ?ows and, for comparison,
for total order ?ow in the nine currencies (T9). The ?rst number in each cell corresponds
to the cross-sectional mean across currencies (e.g., the mean across time-series standard
deviations of all nine currencies), whereas the second (parentheses) and third (brackets)
number correspond to the 5% and 95% percentile of the cross-sectional distribution (across
currencies), respectively. The frequency is daily and the sample is from January 2001 to May
2011.
Mean Median Std Skew Kurt AC(1)
Panel A. Asset Managers
-0.001 -0.001 0.272 -0.827 80.0 0.034
(-0.063) (-0.041) (0.067) (-5.820) (39.8) (-0.009)
[0.027] [0.019] [0.656] [1.978] [270.7] [0.140]
Panel B. Hedge Funds
0.002 0.001 0.205 -0.738 125.1 0.032
(-0.004) (-0.002) (0.054) (-7.977) (17.5) (-0.117)
[0.009] [0.005] [0.494] [4.810] [271.0] [0.128]
Panel C. Corporate Clients
-0.003 -0.001 0.171 -1.091 176.8 0.004
(-0.028) (-0.022) (0.036) (-23.273) (11.0) (-0.107)
[0.012] [0.010] [0.387] [12.143] [898.1] [0.091]
Panel D. Private Clients
-0.003 -0.003 0.068 -0.137 208.8 0.072
(-0.049) (-0.038) (0.009) (-17.616) (22.2) (-0.025)
[0.007] [0.006] [0.165] [10.063] [638.8] [0.192]
Panel E. Total Flows (T9)
0.003 0.002 0.091 -2.857 225.0 0.024
(-0.001) (0.000) (0.009) (-30.643) (16.2) (-0.106)
[0.014] [0.012] [0.265] [5.212] [1,385.8] [0.075]
49
Table A.2. Correlation Between Customer Groups’ Order Flows
This table reports correlation coe?cients between ?ows of customer groups for nine major
currencies and for a pooled sample over all currencies.
Correlation coe?cients
AM/HF AM/CC AM/PC HF/CC HF/PC CC/PC
EUR -0.04 -0.05 -0.05 -0.10 -0.20 -0.05
JPY 0.05 -0.05 -0.12 -0.02 -0.20 -0.05
GBP -0.03 0.02 -0.11 -0.02 -0.17 0.02
CHF 0.01 -0.09 -0.08 -0.07 -0.20 -0.09
AUD 0.00 0.03 -0.02 -0.06 -0.07 0.03
NZD 0.00 -0.05 -0.06 0.01 -0.03 -0.05
CAD -0.04 -0.08 -0.05 -0.01 -0.15 -0.08
SEK -0.03 -0.01 -0.01 -0.02 0.04 -0.01
NOK -0.02 -0.04 -0.01 -0.03 0.03 -0.04
Pooled -0.01 -0.04 -0.05 -0.03 -0.10 -0.04
50
Table A.3. Order Flow Portfolios: Di?erent Standardization Schemes and Sub-Samples
The setup of this table is identical to Table II, Panel A, in the main text but shows results for
rolling (Panel A), recursive (Panel B) ,and in-sample standardization (Panel C) of customer
order ?ow and for three di?erent sample periods as opposed to the rolling standardization
scheme employed in Table II.
Panel A. Rolling Window
P
1
P
2
P
3
P
4
P
5
Av. BMS MR Up
2001/01 – 2011/05 0.82 1.05 6.15 6.77 11.13 5.18 10.31 0.01 0.01
[0.29] [0.37] [2.23] [2.40] [4.04] [2.20] [4.05]
2001/01 – 2007/06 2.14 4.21 5.06 6.02 11.84 5.85 9.69 0.00 0.04
[0.71] [1.41] [1.79] [2.23] [4.14] [2.55] [3.45]
2007/07 – 2011/06 -1.18 -3.70 7.79 7.90 10.07 4.18 11.25 0.18 0.05
[-0.21] [-0.67] [1.44] [1.37] [1.87] [0.87] [2.36]
Panel B. Recursive Window
P
1
P
2
P
3
P
4
P
5
BMS MR Up
2001/01 – 2011/05 -0.42 2.35 5.68 6.73 11.74 12.16 0.00 0.00
[-0.14] [0.83] [2.13] [2.40] [4.19] [4.97]
2001/01 – 2007/06 0.56 5.82 3.86 7.62 11.68 11.12 0.19 0.00
[0.18] [2.07] [1.39] [2.83] [3.91] [4.00]
2007/07 – 2011/06 -1.89 -2.87 8.41 5.4 11.83 13.72 0.02 0.01
[-0.34] [-0.51] [1.62] [0.94] [2.20] [3.07]
Panel C. In-Sample
P
1
P
2
P
3
P
4
P
5
BMS MR Up
2001/01 – 2011/05 0 1.91 7.16 6.09 10.98 10.98 0.11 0.00
[0.00] [0.68] [2.58] [2.14] [4.00] [4.65]
2001/01 – 2007/06 1.86 4.47 6.54 6.4 10.36 8.5 0.01 0.07
[0.63] [1.52] [2.18] [2.31] [3.65] [3.26]
2007/07 – 2011/06 -2.79 -1.92 8.09 5.61 11.91 14.7 0.15 0.01
[-0.51] [-0.35] [1.53] [0.97] [2.21] [3.34]
51
Table A.4. Order Flow Portfolios: Exchange Rate Changes
This table shows average portfolio exchange rate changes for ?ve portfolios (P
1
, ..., P
5
) sorted
on lagged order ?ow. Sorting is done based on standardized total ?ows of all customers.
Column “Av” shows average excess returns across all currencies, column “BMS” (bought
minus sold) reports average excess returns to investing in P
5
and shorting P
1
. Panel B
reports the same information for spot exchange rate changes instead of excess returns. Flows
are standardized by their standard deviation (i) using a rolling window over the previous 60
trading days (Panel A), (ii) using a recursive scheme with 60 days initialization horizon (Panel
B), and (iii) their in-sample standard deviation. Average spot rate changes are annualized
(assuming 252 trading days per year). Numbers in brackets are t-statistics based on Newey-
West standard errors. The frequency is daily and the sample is from January 2001 – May
2011.
Panel A. Rolling Window
P
1
P
2
P
3
P
4
P
5
Av. BMS
Jan 2001 – May 2011 -1.28 -0.64 4.01 4.13 10.20 3.28 11.48
[-0.45] [-0.22] [1.47] [1.41] [3.72] [1.40] [4.57]
Jan 2001 – Jun 2007 -0.24 2.56 2.70 2.73 11.35 3.82 11.59
[-0.08] [0.86] [0.98] [0.91] [4.02] [1.68] [4.25]
Jul 2007 – May 2011 -2.85 -5.45 5.99 6.23 8.46 2.48 11.31
[-0.52] [-0.98] [1.11] [1.08] [1.57] [0.52] [2.37]
Panel B. Recursive Window
P
1
P
2
P
3
P
4
P
5
BMS
Jan 2001 – May 2011 -2.40 0.71 3.40 4.32 10.33 12.73
[-0.83] [0.25] [1.27] [1.50] [3.70] [5.17]
Jan 2001 – Jun 2007 -1.42 4.17 1.02 4.62 10.61 12.03
[-0.46] [1.46] [0.36] [1.60] [3.58] [4.28]
Jul 2007 – May 2011 -3.87 -4.49 6.97 3.87 9.90 13.77
[-0.69] [-0.80] [1.34] [0.67] [1.84] [3.08]
Panel C. In-Sample
P
1
P
2
P
3
P
4
P
5
BMS
Jan 2001 – May 2011 -1.17 -0.98 4.63 4.26 9.68 10.85
[-0.41] [-0.34] [1.68] [1.49] [3.51] [4.57]
Jan 2001 – Jun 2007 1.16 0.78 3.34 4.44 9.42 8.27
[0.40] [0.26] [1.13] [1.55] [3.30] [3.14]
Jul 2007 – May 2011 -4.67 -3.61 6.58 4.01 10.07 14.74
[-0.84] [-0.65] [1.25] [0.70] [1.87] [3.34]
52
Table A.5. Order Flow Portfolios: Customer Groups and Exchange Rate Changes
This table is similar to Panel B of Table II but here we report results for spot exchange rate
changes (and not excess returns).
P
1
P
2
P
3
P
4
Av. BMS
AM -1.65 2.97 5.62 13.86 15.52
[-0.51] [0.98] [1.81] [4.49] [5.75]
HF -0.90 5.32 5.70 9.25 10.15
[-0.29] [1.80] [1.77] [2.85] [3.96]
CO 6.30 4.47 6.37 2.26 -4.04
[1.97] [1.47] [1.96] [0.73] [-1.56]
PC 12.08 5.99 2.28 -1.84 -13.91
[3.85] [1.96] [0.73] [-0.57] [-5.16]
T9 -0.31 1.54 7.58 12.24 5.27 12.55
[-0.09] [0.51] [2.40] [4.00] [1.93] [4.72]
53
Table A.6. Correlation of Excess Returns
This table reports correlation coe?cients between excess returns of di?erent BMS portfolios
based on (i) lagged total ?ows of all 15 currency pairs (T15), (ii) lagged total ?ows of nine
developed countries (T9), (iii) lagged ?ows of asset managers (AM), (iv) lagged ?ows of hedge
funds, (v) lagged ?ows of corporate clients (CC), and lagged ?ows of private clients (PC). All
?ows are standardized by their lagged volatility over a 60-day rolling window. The frequency
is daily and the sample period is January 2001 to Ma 2011.
T15 T9 AM HF CC
T15 1.00
T9 0.63 1.00
AM 0.27 0.42 1.00
HF 0.30 0.42 0.06 1.00
CC 0.00 -0.06 -0.08 -0.13 1.00
PC -0.04 -0.02 -0.07 0.03 0.01
54
Table A.7. Order Flow Portfolios: Standardizing Flows (One Year)
This table shows average annualized portfolio excess returns for ?ve (or four) portfolios (P
1
,
..., P
5
) sorted on lagged order ?ow. Sorting is done based on standardized total ?ows and
standardized customer ?ows. Column“Av”shows average excess returns across all currencies,
column “BMS” (buying minus selling pressure) reports average excess returns to investing in
P
5
(or P
4
) and shorting P
1
. Flows are standardized by their standard deviation using a rolling
window over the previous 252 trading days (that is, roughly one year). We form 5 portfolios
for total ?ows of all 15 currency pairs (T15) and 4 portfolios for total ?ows of the nine
currencies for which we have disaggregated ?ows available (T9), for asset managers’ ?ows
(AM), hedge funds’ ?ows (HF), corporate clients’ ?ows (CC), and private clients’ ?ows (PC).
Numbers in brackets are t-statistics based on Newey-West standard errors. The frequency is
daily and the sample is from January 2001 to May 2011.
Panel A. Excess Returns
P
1
P
2
P
3
P
4
P
5
Av. BMS
T15 0.71 2.61 7.91 5.95 12.60 5.96 11.89
[0.23] [0.87] [2.77] [2.00] [4.29] [2.38] [4.59]
T9 0.66 2.56 8.44 13.81 6.37 13.15
[0.19] [0.82] [2.54] [4.24] [2.19] [4.81]
AM -1.15 3.79 6.66 16.10 17.25
[-0.33] [1.19] [2.02] [4.98] [6.23]
HF -0.63 7.35 5.62 10.75 11.38
[-0.19] [2.32] [1.65] [3.13] [4.23]
CC 7.81 6.09 8.55 1.10 -6.71
[2.36] [1.86] [2.48] [0.34] [-2.39]
PC 16.30 5.55 3.36 -1.48 -17.79
[4.96] [1.65] [0.99] [-0.45] [-6.49]
Panel B. Exchange Rate Changes
P
1
P
2
P
3
P
4
P
5
Av. BMS
T15 -1.42 0.66 6.26 3.18 11.20 3.97 12.62
[-0.47] [0.22] [2.20] [1.04] [3.80] [1.59] [4.79]
T9 -0.05 1.93 7.86 13.23 5.74 13.27
[-0.01] [0.62] [2.36] [4.06] [1.98] [4.85]
AM -1.68 3.03 5.98 15.65 17.33
[-0.49] [0.96] [1.82] [4.84] [6.25]
HF -1.27 6.68 5.10 10.13 11.41
[-0.39] [2.11] [1.49] [2.95] [4.24]
CC 7.18 5.31 7.92 0.76 -6.42
[2.16] [1.62] [2.30] [0.23] [-2.28]
PC 15.67 4.87 2.77 -2.05 -17.72
[4.77] [1.45] [0.82] [-0.63] [-6.47]
55
Table A.8. Order Flow Portfolios: Standardizing Flows (Three Years)
This table shows average annualized portfolio excess returns for ?ve (or four) portfolios (P
1
,
..., P
5
) sorted on lagged order ?ow. Sorting is done based on standardized total ?ows and
standardized customer ?ows. Column“Av”shows average excess returns across all currencies,
column “BMS” (buying minus selling pressure) reports average excess returns to investing in
P
5
(or P
4
) and shorting P
1
. Flows are standardized by their standard deviation using a
rolling window over the previous 750 trading days (that is, roughly three years). We form 5
portfolios for total ?ows of all 15 currency pairs (T15) and 4 portfolios for total ?ows of the
nine currencies for which we have disaggregated ?ows available (T9), for asset managers’ ?ows
(AM), hedge funds’ ?ows (HF), corporate clients’ ?ows (CC), and private clients’ ?ows (PC).
Numbers in brackets are t-statistics based on Newey-West standard errors. The frequency is
daily and the sample is from January 2001 to May 2011.
Panel A. Excess Returns
P
1
P
2
P
3
P
4
P
5
Av. BMS
T15 -2.21 -0.55 5.97 6.28 10.25 3.95 12.46
[-0.61] [-0.16] [1.75] [1.74] [3.04] [1.31] [4.30]
T9 -3.02 -1.86 7.29 13.28 3.92 16.30
[-0.75] [-0.50] [1.91] [3.40] [1.14] [5.01]
AM -4.89 1.15 3.20 15.39 20.28
[-1.22] [0.30] [0.84] [3.99] [6.30]
HF -2.44 2.31 3.97 9.92 12.36
[-0.64] [0.61] [0.99] [2.42] [3.85]
CC 6.48 2.16 7.25 -2.37 -8.85
[1.65] [0.55] [1.78] [-0.63] [-2.70]
PC 13.26 4.32 0.98 -6.05 -19.31
[3.47] [1.09] [0.24] [-1.54] [-5.78]
Panel B. Exchange Rate Changes
P
1
P
2
P
3
P
4
P
5
Av. BMS
T15 -3.89 -2.04 5.17 4.95 8.60 2.56 12.49
[-1.08] [-0.58] [1.52] [1.37] [2.55] [0.85] [4.28]
T9 -3.51 -2.09 7.05 12.92 3.59 16.43
[-0.86] [-0.56] [1.84] [3.31] [1.05] [5.04]
AM -5.16 0.75 2.88 15.14 20.30
[-1.29] [0.20] [0.76] [3.93] [6.31]
HF -2.93 2.06 3.82 9.50 12.43
[-0.77] [0.54] [0.95] [2.32] [3.87]
CC 6.15 1.71 6.93 -2.47 -8.62
[1.57] [0.44] [1.70] [-0.66] [-2.62]
PC 12.76 4.06 0.79 -6.40 -19.16
[3.34] [1.02] [0.19] [-1.63] [-5.73]
56
Table A.9. Order Flow Portfolios: Order Flows Scaled By Currency Trading Volume
This table is identical to Table II but here we do not standardize order ?ows by rolling
windows of the previous 60 trading days volatility but by total currency trading volume
(from the BIS FX triennial surveys for 2001, 2004, 2007, 2010). We linearly interpolate
between the turnover ?gures to obtain a daily measure of total trading volume for each of
the ?fteen currencies in our sample.
Panel A. Total order ?ows
P
1
P
2
P
3
P
4
P
5
Av. BMS MR Up Down
T15 -1.14 3.18 3.19 6.73 11.23 5.01 12.37 0.02 0.00 –
[-0.38] [1.19] [1.26] [2.37] [3.99] [2.00] [5.05]
(-0.12) (0.37) (0.39) (0.76) (1.19) (0.66) (1.53)
T9 -0.73 2.35 6.09 12.46 5.45 13.19 0.00 0.00 –
[-0.22] [0.83] [1.98] [4.01] [2.11] [5.09]
(-0.07) (0.25) (0.61) (1.17) (0.63) (1.55)
Panel B. Disaggregated order ?ows
AM -0.97 2.40 5.98 13.23 14.19 0.00 0.00 –
[-0.29] [0.85] [2.02] [4.19] [5.33]
(-0.09) (0.25) (0.61) (1.24) (1.66)
HF 0.14 4.05 5.94 9.43 9.29 0.00 0.04 –
[0.04] [1.40] [1.95] [2.87] [3.50]
(0.01) (0.42) (0.59) (0.87) (1.10)
CO 5.94 5.16 4.25 3.07 -2.88 0.01 – 0.65
[1.88] [1.72] [1.34] [1.01] [-1.15]
(0.58) (0.52) (0.41) (0.31) (-0.35)
PC 14.06 5.01 0.69 -1.14 -15.20 0.00 – 0.00
[4.52] [1.67] [0.22] [-0.36] [-5.68]
(1.35) (0.51) (0.07) (-0.11) (-1.71)
57
Table A.10. Order Flow Portfolios: Demeaning Flows
This table is similar to Table II but here we present results for total ?ows (T15 and T9) and
customer groups’ ?ows and we standardize order ?ows by subtracting the rolling mean and
dividing by the rolling standard deviation. The frequency is daily and the sample is from
January 2001 – May 2011.
P1 P2 P3 P4 P5 Av BMS
T15 -0.68 5.02 5.54 5.67 11.16 5.34 11.84
[-0.24] [1.86] [1.95] [2.07] [3.88] [2.26] [4.89]
T9 0.11 3.21 6.12 13.60 13.49
[0.03] [1.08] [1.91] [4.40] [5.25]
AM -0.28 4.64 3.45 14.50 14.77
[-0.08] [1.56] [1.07] [4.77] [5.44]
HF -0.04 6.29 5.26 9.95 9.99
[-0.01] [2.09] [1.69] [3.02] [3.88]
CO 7.91 5.41 4.85 3.75 -4.17
[2.48] [1.82] [1.54] [1.17] [-1.50]
PC 12.11 6.37 5.05 -2.09 -14.19
[3.77] [2.15] [1.60] [-0.63] [-5.00]
58
Table A.11. Order Flow-Weighted BMS Portfolios
This table shows returns for BMS portfolios based on total ?ows and customer ?ows but
here we employ portfolio weights directly based on lagged order ?ows. For each trading day
t, we cross-sectionally standardize order ?ows, rescale these standardized ?ows so that they
sum to two in absolute value, and then use these rescaled and standardized ?ows as portfolio
weights for day t to t + 1. Numbers in squared brackets are based on Newey/West standard
errors. The frequency is daily and the sample is from January 2001 – May 2011.
T15 T9 AM HF CO PC
Mean 12.70 11.12 13.94 6.61 -1.94 -9.77
t-Stat. 5.07 4.70 5.86 2.91 -0.89 -4.27
St. Dev. 8.38 7.60 7.72 7.46 7.04 7.12
Sharpe Ratio 1.52 1.46 1.81 0.89 -0.28 -1.37
Skewness 0.25 1.42 1.76 0.09 -0.12 -0.20
Kurtosis 19.63 33.82 28.54 11.75 12.87 11.05
Maximum 6.57 7.79 7.42 3.95 2.96 3.21
Minimum -5.33 -3.68 -2.46 -3.69 -3.95 -4.12
59
Table A.12. BMS Portfolios: Longer Horizons
This table shows average annualized BMS portfolio excess returns for longer forecast horizons
of 1, 2, ..., 5, 10, 20, 40, and 60 trading days. We use an exponential moving average (EMA)
with a decay parameter of 0.25 (Panel A) and 0.75 (Panel B) for lagged order ?ows to consider
longer histories of order ?ows for forecasting. BMS portfolios are based on 5 portfolios
for total ?ows of all 15 currency pairs (T15) and 4 portfolios for total ?ows of the nine
developed markets for which disaggregated ?ows are available (T9), for asset managers’ ?ows
(AM), hedge funds’ ?ows (HF), corporate clients’ ?ows (CC), and private clients’ ?ows (PC).
Numbers in brackets are t-statistics based on Newey-West standard errors. The frequency is
daily and the sample is from January 2001 to May 2011.
Rebalancing Frequency (Trading Days)
2 3 4 5 10 20 40 60
EMA with Decay Parameter 0.25
T15 17.20 12.31 9.82 7.68 4.57 4.15 2.83 3.60
[8.22] [5.68] [4.76] [3.66] [2.46] [2.40] [1.92] [1.68]
T9 20.74 14.25 11.27 7.98 5.19 4.84 2.14 -0.41
[7.99] [5.32] [4.46] [3.11] [2.99] [2.13] [1.76] [-0.15]
AM 20.28 16.30 11.99 6.63 5.72 5.12 2.99 1.56
[7.39] [5.91] [4.75] [2.61] [2.70] [2.08] [1.42] [0.59]
HF 15.78 7.60 3.78 2.31 -1.03 -0.72 1.50 3.14
[5.30] [2.67] [1.30] [0.81] [-0.36] [-0.26] [0.58] [1.09]
CC -4.74 -3.53 -2.85 -2.61 -0.61 -1.03 0.13 -3.51
[-1.84] [-1.32] [-1.12] [-1.03] [-0.25] [-0.42] [0.05] [-1.53]
PC -16.65 -8.58 -6.44 -2.42 -2.66 0.23 -1.78 -1.77
[-5.73] [-3.33] [-2.31] [-0.87] [-0.99] [0.08] [-0.58] [-0.71]
EMA with Decay Parameter 0.75
T15 18.97 16.21 9.48 10.67 5.78 4.27 0.63 3.09
[7.92] [7.46] [4.40] [5.36] [2.94] [1.94] [0.32] [1.58]
T9 23.05 18.09 12.03 10.14 7.83 4.90 1.14 3.87
[8.00] [6.76] [4.62] [4.12] [3.28] [2.22] [0.46] [1.54]
AM 23.75 20.92 16.19 11.74 7.27 5.98 2.14 0.82
[8.30] [7.68] [5.89] [4.32] [3.61] [2.54] [0.87] [0.31]
HF 20.66 11.34 6.83 8.15 6.84 1.46 1.37 3.71
[7.58] [3.73] [2.51] [2.91] [2.67] [0.55] [0.56] [1.41]
CC -3.61 -3.47 -1.54 -4.52 -1.01 -2.20 -2.67 -1.76
[-1.31] [-1.29] [-0.62] [-1.70] [-0.42] [-0.96] [-1.10] [-0.76]
PC -24.34 -14.51 -9.43 -7.12 -3.94 1.10 -0.78 0.72
[-7.96] [-5.25] [-3.51] [-2.61] [-1.62] [0.45] [-0.29] [0.29]
60
Table A.13. Order Flow Portfolios: Four Liquid Currencies
This table is similar to Table II in the main text but here we only include EUR/USD,
JPY/USD, GBP/USD, and CHF/USD in our sample of currencies and only form two port-
folios. T4 denotes portfolios based on total order ?ows.
Excess Returns Exchange Rate Changes
P
1
P
2
Av. BMS P
1
P
2
Av. BMS
T4 1.55 5.48 3.51 3.93 1.95 6.20 4.07 4.25
[0.56] [2.00] [1.40] [1.70] [0.70] [2.24] [1.62] [1.81]
AM -1.31 8.34 9.65 -0.79 8.94 9.73
[-0.46] [3.11] [4.23] [-0.28] [3.34] [4.27]
HF 1.68 5.35 3.67 2.18 5.97 3.79
[0.61] [1.93] [1.55] [0.79] [2.15] [1.60]
CC 5.16 1.90 -3.25 5.61 2.57 -3.05
[1.83] [0.70] [-1.41] [2.00] [0.95] [-1.32]
PC 7.59 -0.56 -8.15 8.14 0.00 -8.14
[2.75] [-0.20] [-3.47] [2.95] [0.00] [-3.47]
61
Table A.14. Order Flow Portfolios: Sensitivity to Individual Currencies
This table show average annualized excess returns to BMS portfolios based on total ?ows,
asset managers (AM) ?ows, hedge fund (HF) ?ows, corporate clients (CC) ?ows, and private
clients (PC) ?ows for a cross-validation setting in which we discard one of the available
currencies in our sample. We do this for each available currency and the ?rst column indicates
which currency is left out when computing returns to the BMS portfolio. Hence, BMS returns
for total ?ows are based on 14 currencies and BMS returns for customer ?ows are based on
8 currencies instead of 15 and 9 currencies, respectively. t-statistics in brackets are based on
Newey/West standard errors.
Total ?ows AM HF CC PC
rx t rx t rx t rx t rx t
EUR 10.79 [4.12] 15.40 [5.35] 8.86 [3.10] -2.15 [-0.78] -14.62 [-5.23]
JPY 8.26 [3.34] 9.62 [3.95] 11.30 [4.73] -3.13 [-1.33] -9.12 [-3.87]
GBP 10.18 [3.99] 15.29 [5.65] 8.95 [3.35] -3.04 [-1.17] -13.65 [-4.93]
CHF 10.29 [4.09] 13.58 [5.03] 11.49 [4.51] -2.08 [-0.79] -16.31 [-5.90]
AUD 10.39 [4.18] 10.39 [4.03] 9.78 [3.83] -5.58 [-2.19] -10.77 [-4.26]
NZD 9.35 [3.67] 14.75 [5.52] 8.16 [3.15] -2.22 [-0.87] -11.93 [-4.45]
CAD 11.98 [4.70] 14.70 [5.34] 9.65 [3.81] -4.07 [-1.61] -10.98 [-4.12]
SEK 9.33 [3.70] 15.51 [5.76] 6.32 [2.28] -4.46 [-1.76] -14.91 [-5.45]
NOK 10.04 [3.91] 16.52 [6.00] 9.38 [3.72] -3.37 [-1.24] -16.20 [-5.84]
MXN 10.10 [4.06]
BRL 11.22 [4.56]
ZAR 6.71 [2.79]
KRW 11.75 [4.78]
SGD 10.84 [4.27]
HKD 11.68 [4.75]
62
T
a
b
l
e
A
.
1
5
.
O
r
d
e
r
?
o
w
p
o
r
t
f
o
l
i
o
r
e
t
u
r
n
s
a
r
o
u
n
d
c
u
s
t
o
m
e
r
t
r
a
d
i
n
g
T
h
i
s
t
a
b
l
e
s
h
o
w
s
a
v
e
r
a
g
e
a
n
n
u
a
l
i
z
e
d
p
o
r
t
f
o
l
i
o
e
x
c
e
s
s
r
e
t
u
r
n
s
f
o
r
c
u
r
r
e
n
c
y
p
o
r
t
f
o
l
i
o
s
s
o
r
t
e
d
o
n
l
a
g
g
e
d
d
i
s
a
g
g
r
e
g
a
t
e
d
c
u
s
t
o
m
e
r
o
r
d
e
r
?
o
w
i
n
e
v
e
n
t
t
i
m
e
a
r
o
u
n
d
p
o
r
t
f
o
l
i
o
f
o
r
m
a
t
i
o
n
(
t
=
?
5
,
?
4
,
,
0
,
1
,
2
,
,
1
0
d
a
y
s
w
h
e
r
e
t
=
0
i
s
t
h
e
d
a
y
o
f
p
o
r
t
f
o
l
i
o
f
o
r
m
a
t
i
o
n
)
f
o
r
t
h
e
h
i
g
h
(
P
4
)
a
n
d
l
o
w
(
P
1
)
p
o
r
t
f
o
l
i
o
s
.
E
v
e
n
t
t
i
m
e
(
d
a
y
)
?
5
?
4
?
3
?
2
?
1
0
1
2
3
4
5
6
7
8
9
1
0
A
M
P
4
2
.
4
1
5
.
6
8
6
.
9
0
9
.
6
1
1
1
.
3
7
1
3
.
6
0
1
4
.
3
1
1
8
.
6
7
7
.
7
5
5
.
1
2
6
.
6
2
4
.
5
2
5
.
6
5
9
.
2
0
7
.
2
5
3
.
2
2
[
0
.
7
4
]
[
1
.
7
8
]
[
2
.
0
2
]
[
3
.
1
0
]
[
3
.
6
2
]
[
4
.
2
7
]
[
4
.
6
5
]
[
5
.
9
0
]
[
2
.
5
0
]
[
1
.
5
1
]
[
2
.
1
5
]
[
1
.
4
3
]
[
1
.
7
4
]
[
2
.
9
5
]
[
2
.
2
3
]
[
1
.
0
0
]
P
1
7
.
3
1
4
.
4
7
3
.
6
1
3
.
9
9
1
.
4
3
-
1
.
2
0
-
1
.
1
3
-
6
.
1
8
-
0
.
5
2
6
.
4
1
4
.
4
4
3
.
9
0
5
.
8
6
5
.
8
2
4
.
9
9
6
.
0
1
[
2
.
3
5
]
[
1
.
4
7
]
[
1
.
1
2
]
[
1
.
2
8
]
[
0
.
4
8
]
[
-
0
.
3
7
]
[
-
0
.
3
4
]
[
-
1
.
8
9
]
[
-
0
.
1
6
]
[
2
.
1
2
]
[
1
.
4
7
]
[
1
.
2
0
]
[
1
.
8
4
]
[
1
.
8
1
]
[
1
.
5
6
]
[
1
.
8
8
]
H
F
P
4
0
.
7
4
1
.
7
8
2
.
0
2
3
.
1
0
3
.
6
2
4
.
2
7
4
.
6
5
5
.
9
0
2
.
5
0
1
.
5
1
2
.
1
5
1
.
4
3
1
.
7
4
2
.
9
5
2
.
2
3
1
.
0
0
[
1
0
.
1
0
]
[
7
.
4
0
]
[
5
.
4
8
]
[
7
.
2
7
]
[
8
.
6
0
]
[
1
7
.
9
0
]
[
9
.
7
8
]
[
1
9
.
6
7
]
[
6
.
6
7
]
[
4
.
1
5
]
[
4
.
5
3
]
[
1
.
9
2
]
[
7
.
0
4
]
[
5
.
6
0
]
[
2
.
9
6
]
[
4
.
9
0
]
P
1
5
.
3
0
3
.
0
3
6
.
2
7
-
0
.
7
4
0
.
2
0
-
9
.
1
1
-
0
.
3
2
-
8
.
5
5
4
.
6
1
7
.
0
8
4
.
3
9
8
.
1
1
4
.
2
0
5
.
8
9
7
.
6
1
5
.
9
5
[
1
.
7
4
]
[
0
.
9
7
]
[
2
.
0
7
]
[
-
0
.
2
4
]
[
0
.
0
6
]
[
-
2
.
7
7
]
[
-
0
.
1
0
]
[
-
2
.
4
7
]
[
1
.
5
2
]
[
2
.
3
3
]
[
1
.
5
0
]
[
2
.
6
2
]
[
1
.
4
1
]
[
1
.
9
3
]
[
2
.
5
1
]
[
1
.
9
6
]
C
C
P
4
4
.
8
7
3
.
7
3
2
.
0
6
6
.
9
0
1
.
5
1
0
.
4
9
2
.
6
1
2
.
8
0
4
.
3
8
5
.
8
9
3
.
4
0
8
.
0
9
5
.
6
0
6
.
5
7
5
.
3
6
7
.
6
5
[
1
.
6
2
]
[
1
.
2
4
]
[
0
.
6
8
]
[
2
.
1
9
]
[
0
.
4
9
]
[
0
.
1
6
]
[
0
.
8
4
]
[
0
.
9
1
]
[
1
.
3
5
]
[
1
.
9
4
]
[
1
.
0
3
]
[
2
.
6
8
]
[
1
.
8
3
]
[
2
.
1
2
]
[
1
.
6
9
]
[
2
.
5
2
]
P
1
5
.
6
7
7
.
1
3
7
.
2
7
4
.
5
3
7
.
2
7
8
.
5
2
6
.
9
0
1
0
.
9
3
5
.
8
4
3
.
6
4
8
.
3
8
6
.
1
8
5
.
6
1
5
.
1
6
5
.
6
9
4
.
8
6
[
1
.
8
3
]
[
2
.
2
7
]
[
2
.
3
6
]
[
1
.
4
7
]
[
2
.
2
4
]
[
2
.
4
9
]
[
2
.
1
5
]
[
3
.
4
1
]
[
1
.
9
1
]
[
1
.
1
5
]
[
2
.
7
7
]
[
2
.
0
1
]
[
1
.
7
8
]
[
1
.
6
8
]
[
1
.
8
1
]
[
1
.
5
8
]
P
C
P
4
2
.
9
3
3
.
1
6
2
.
7
5
1
.
7
6
-
3
.
6
8
-
9
.
9
3
-
1
.
3
0
-
1
0
.
6
6
6
.
8
9
5
.
8
2
4
.
3
4
6
.
7
7
5
.
4
9
5
.
2
2
4
.
9
4
6
.
3
1
[
0
.
9
7
]
[
1
.
0
2
]
[
0
.
8
7
]
[
0
.
5
7
]
[
-
1
.
1
8
]
[
-
3
.
0
0
]
[
-
0
.
4
0
]
[
-
2
.
9
9
]
[
2
.
1
4
]
[
1
.
9
2
]
[
1
.
3
3
]
[
2
.
0
6
]
[
1
.
7
5
]
[
1
.
6
2
]
[
1
.
5
5
]
[
2
.
0
5
]
P
1
7
.
3
1
8
.
1
1
8
.
7
7
9
.
1
9
1
3
.
8
5
1
8
.
2
4
1
2
.
7
1
2
3
.
1
1
3
.
6
7
4
.
0
0
7
.
6
3
7
.
5
4
3
.
2
2
6
.
5
7
4
.
2
9
4
.
2
1
[
2
.
2
2
]
[
2
.
4
5
]
[
2
.
7
0
]
[
2
.
8
6
]
[
4
.
3
7
]
[
5
.
6
1
]
[
4
.
0
5
]
[
7
.
3
1
]
[
1
.
2
0
]
[
1
.
1
8
]
[
2
.
3
6
]
[
2
.
3
5
]
[
0
.
9
8
]
[
2
.
1
2
]
[
1
.
3
6
]
[
1
.
2
6
]
63
Table A.16. Pricing Error Statistics For The Cross-Section
This table reports pricing error statistics based on estimating the models of Table VI for the broader
cross-section of the order ?ow mimicking portfolios of ?nancial end-users. GRS is the test statistic
by Gibbons, Ross, and Shanken (1989). We compute the joint test for zero alphas for the entire
cross section of eight portfolios of Asset Managers and Hedge Funds. We also compute the GRS-
statistic separately for the portfolios (P1-P4) of each group. Model speci?cations (i) - to (iv) follow
the setup of table VI.
GRS p-val. GRS AM p-val. GRS HF p-val.
A. Linear Exposures
(i) 6.87 0.00 10.37 0.00 3.79 0.01
(ii) 6.49 0.00 9.51 0.00 4.00 0.00
(iii) 7.61 0.00 10.77 0.00 6.69 0.00
(iv) 7.14 0.00 9.51 0.00 6.75 0.00
B. Conditional Exposures
(i) 6.19 0.00 9.46 0.00 4.14 0.00
(ii) 6.51 0.00 9.04 0.00 6.31 0.00
(iii) 6.53 0.00 9.34 0.00 5.81 0.00
(iv) 6.58 0.00 10.24 0.00 3.79 0.01
64
Table A.17. Risk Exposures: Equity Factors
This table reports regression results for the risk exposures of the BMS portfolios of ?nancial FX
market end-users, that is, asset managers and hedge funds. The risk factors include the excess return
on the market portfolio (r
m
), and the Fama-French size (SMB) and value (HML) factors. UMD
denotes the return on Carhart’s momentum factor. The Table shows results for four parsimonious
model speci?cations (i-iv) where the factors are selected according to the Schwarz criterion (joint
estimation of the equation for asset managers’ and hedge funds’ BMS returns). Speci?cation (v)
includes all factors jointly. We further report the estimated intercept ?, the adjusted R
2
and the
BIC computed for the two-equation system (Sys-BIC). Below the regression coe?cients, t-statistics
based on HAC standard errors are reported (in brackets).
Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (v) (i) (ii) (iii) (iv) (v)
r
m
-0.02 -0.03 0.03 0.03
[-0.28] [-0.38] [0.61] [0.52]
SMB 0.05 0.07 0.06 0.05
[-0.66] [1.01] [0.46] [0.39]
HML -0.05 -0.04 0.03 0.02
[-0.47] [-0.47] [0.46] [0.3]
UMD 0.01 0.00 0.00 0.02
[0.16] [-0.13] [-0.07] [0.37]
? 1.26 1.29 1.30 1.28 1.28 0.84 0.86 0.86 0.87 0.83
[4.57] [4.85] [4.74] [4.77] [4.72] [3.38] [3.69] [3.7] [3.77] [3.45]
¯
R
2
-0.01 -0.01 -0.01 -0.01 -0.03 0.00 0.00 -0.01 -0.01 -0.03
Sys-BIC 3.74 3.74 3.74 3.75 3.97 3.74 3.74 3.74 3.75 3.97
65
Table A.18. Risk Exposures: FX Factors
This table reports regression results for the risk exposures of the BMS portfolios of ?nancial FX market
end-users, that is, asset managers and hedge funds. The FX factors include the excess Dollar risk factor and
the carry risk factor by Lustig, Roussanov, and Verdelhan (2011). VOL
FX
is the global FX volatility risk
factor (factor mimicking portfolio). The Table shows results for four parsimonious model speci?cations (i-iv)
where the factors are selected according to the Schwarz criterion (joint estimation of the equation for asset
managers’ and hedge funds’ BMS returns). Speci?cation (v) includes all factors jointly. We further report
the estimated intercept ?, the adjusted R
2
and the BIC computed for the two-equation system (Sys-BIC).
Below the regression coe?cients, t-statistics based on HAC standard errors are reported (in brackets).
Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (v) (i) (ii) (iii) (iv) (v)
DOL 0.04 0.04 -0.11 -0.11
[0.28] [0.28] [-0.99] [-0.98]
HML
FX
-0.21 -0.04 -0.03 0.27 0.19 0.19
[-1.75] [-0.33] [-0.32] [2.63] [1.66] [1.62]
VOL
FX
0.07 0.07 0.08 0.07 -0.07 -0.03 -0.08 -0.05
[2.44] [2.34] [2.24] [2.55] [-2.5] [-1.30] [-2.58] [-1.46]
? 1.46 1.43 1.47 1.45 1.46 0.71 0.69 0.67 0.73 0.69
[5.32] [4.75] [5.15] [5.43] [5.26] [3.1] [2.95] [2.93] [3.11] [2.96]
¯
R
2
0.10 0.06 0.09 0.09 0.09 0.11 0.13 0.13 0.11 0.13
Sys-BIC 3.53 3.55 3.57 3.60 3.64 3.53 3.55 3.57 3.60 3.64
66
Table A.19. Risk Exposures: Fung-Hsieh Factors
This table reports regression results for the risk exposures of the BMS portfolios of ?nancial FX market end-
users, that is, asset managers and hedge funds. The options-based factors are intended to capture non-linear
payo? features that are typical of hedge fund returns (Fung and Hsieh, 2001). Panel A considers the ?ve
market timing-factors for various asset classes (BN - Bonds, FX, CM - commodities, Eq - equities, and IR -
short term interest rates) as the starting point. Panel B uses the Fung-Hsieh 7-factor model as the starting
point, as outlined in the main text. The Table shows results for four parsimonious model speci?cations (i-iv)
where the factors are selected according to the Schwarz criterion (joint estimation of the equation for asset
managers’ and hedge funds’ BMS returns). Speci?cation (v) includes all factors jointly. We further report
the estimated intercept ?, the adjusted R
2
and the BIC computed for the two-equation system (Sys-BIC).
Below the regression coe?cients, t-statistics based on HAC standard errors are reported (in brackets).
A. Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (v) (i) (ii) (iii) (iv) (v)
PTFS
BD
1.19 1.20 -1.95 -1.96
[0.54] [0.54] [-1.24] [-1.30]
PTFS
FX
2.69 3.67 2.50 3.00 -3.37 -3.87 -3.06 -3.49
[2.67] [2.70] [2.78] [2.78] [-2.97] [-2.62] [-2.70] [-3.13]
PTFS
CM
-1.88 1.63
[-1.24] [1.07]
PTFS
IR
2.28 2.70 2.20 2.50 -1.16 -1.69 -1.04 -1.29
[3.45] [3.61] [3.47] [3.78] [-1.32] [-1.49] [-1.12] [-1.78]
PTFS
EQ
-0.45 0.36
[-0.23] [0.17]
? 1.20 1.25 1.22 1.24 1.20 0.93 0.91 0.91 0.86 0.90
[5.51] [5.18] [5.4] [4.8] [4.11] [4.25] [4.24] [4.02] [4.22] [4.15]
¯
R
2
0.14 0.07 0.11 0.13 0.13 0.11 0.09 0.04 0.12 0.11
Sys-BIC 3.55 3.58 3.59 3.61 3.75 3.55 3.58 3.59 3.61 3.75
B. Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (v) (i) (ii) (iii) (iv) (v)
rm 0.03 -0.03
[0.53] [-1.10]
SMB 0.07 0.06
[1.13] [0.52]
PTFS
FX
2.48 3.67 2.13 -2.89 -3.87 -3.05
[2.39] [2.70] [1.77] [-2.85] [-2.62] [-2.64]
PTFS
CM
0.46 0.96
[0.34] [0.61]
PTFS
BD
1.68 0.98 -2.31 -1.95
[0.82] [0.54] [-1.60] [-1.38]
?TS -0.90 -0.24
[-1.1] [-0.37]
?DF 3.72 4.75 4.38 3.99 -3.06 -4.26 -3.75 -2.79
[2.05] [2.49] [2.48] [2.02] [-1.86] [-2.28] [-1.88] [-1.56]
? 1.26 1.29 1.25 1.34 1.27 0.90 0.87 0.91 0.79 0.83
[5.11] [4.97] [5.18] [4.58] [4.86] [4.17] [3.91] [4.24] [3.78] [4.03]
¯
R
2
0.12 0.09 0.07 0.09 0.11 0.13 0.09 0.09 0.10 0.11
Sys-BIC 3.55 3.56 3.58 3.61 3.89 3.55 3.56 3.58 3.61 3.89
67
Table A.20. Risk Exposures T15/T9
This table reports regression results for the risk exposures of the BMS portfolios computed from the total
?ows based on either 15 (T15) or 9 currencies. The methodological framework in Panel A is a modi?ed linear
Fung-Hsieh model with eight factors as outlined in the main text. Panel B also accounts for the conditional
exposure to stock market returns by including additional interaction terms of market returns. The three
conditioning variables are ?rst di?erences of the 3-month T-Bill rate, the VIX and the TED spread. The
Table shows results for four parsimonious model speci?cations where the factors are selected according to
the Schwarz criterion (joint estimation of the equation for T15 and T9 BMS returns). Results for the other
factors are not reported. We further report the estimated intercept ?, the adjusted R
2
and the BIC computed
for the two-equation system (Sys-BIC). Below the regression coe?cients, t-statistics based on HAC standard
errors are reported (in brackets).
A. T15 T9
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
r
m
-0.04 -0.09
[-0.79] [-2.11]
DOL -0.08 -0.28
[-0.54] [-2.74]
?TS -0.96 -0.60
[-1.45] [-1.18]
?DF -4.72 -5.20 -5.06 -4.74 1.37 -0.22 0.59 1.35
[-2.17] [-2.17] [-2.11] [-2.41] [0.97] [-0.17] [0.48] [0.94]
? 0.88 0.92 0.90 0.91 1.05 1.17 1.09 1.06
[4.87] [4.51] [4.97] [4.85] [4.46] [4.94] [4.48] [4.41]
¯
R
2
0.10 0.10 0.10 0.11 0.00 0.05 0.03 0.00
Sys-BIC 3.19 3.21 3.23 3.25 3.19 3.21 3.23 3.25
B. T15 T9
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
DOL -0.07 -0.27
[-0.42] [-2.56]
r
m
·? TED(t-1) 0.26 0.26
[2.73] [2.03]
r
m
·? TB(t-1) -0.36 0.05
[-2.46] [0.19]
?DF -4.79 -3.70 -3.36 -5.16 1.30 1.15 2.71 -0.20
[-2.24] [-2.03] [-2.05] [-2.19] [0.92] [0.76] [1.15] [-0.15]
? 0.86 0.85 0.82 0.89 1.03 1.03 0.99 1.15
[4.63] [4.67] [4.53] [4.22] [4.3] [4.3] [4.13] [4.73]
¯
R
2
0.11 0.13 0.14 0.10 0.00 -0.01 0.04 0.04
Sys-BIC 3.19 3.20 3.21 3.21 3.19 3.20 3.21 3.21
68
Table A.21. Marginal Forecast Performance: Four-Factor Adjusted Excess Returns
This table shows excess returns for BMS portfolios sorted on lagged order ?ow as in Table
III. We do not only sort on order ?ow of the previous day but also allow for longer lags of
up to nine days between order ?ow signals and portfolio formation. Portfolios are rebalanced
daily. T15 denotes portfolios sorts on total order ?ows and the sample of all 15 currencies,
and T9 denotes portfolios sorts on total order ?ows and the sample of 9 developed currencies;
AM, HF, CC, and PC denote portfolios sorts on asset managers’, hedge funds’, corporate
clients’, and private clients’ order ?ows, respectively. Compared to Table III which reports
unadjusted excess returns, we report adjusted excess returns based on the Carhart (1997)
four-factor model.
Controlling for MKTRF, HML, SMB, and UMD
T15 10.31 24.87 10.55 -1.11 3.31 0.33 0.27 2.21 -2.15 -0.72
[3.92] [8.90] [4.50] [-0.44] [1.38] [0.14] [0.11] [0.94] [-0.86] [-0.31]
T9 12.52 24.59 7.83 -4.10 6.20 -1.76 2.26 1.45 -0.97 -1.93
[4.53] [8.89] [3.10] [-1.56] [2.29] [-0.68] [0.88] [0.54] [-0.37] [-0.78]
AM 16.06 25.30 8.85 -1.55 2.68 0.23 -0.06 3.67 2.31 -2.98
[5.74] [8.88] [3.15] [-0.55] [1.06] [0.09] [-0.02] [1.29] [0.82] [-1.03]
HF 9.95 28.37 1.44 -2.90 0.05 -6.09 2.88 -0.20 -4.69 -0.72
[3.83] [9.42] [0.53] [-1.14] [0.02] [-2.34] [1.11] [-0.07] [-1.78] [-0.26]
CC -4.24 -8.30 -1.82 2.48 -5.23 2.13 -0.52 1.48 -0.09 3.41
[-1.54] [-2.88] [-0.59] [0.96] [-2.01] [0.82] [-0.20] [0.57] [-0.03] [1.32]
PC -14.35 -34.38 3.45 2.29 -3.37 -1.08 2.33 -1.65 0.80 1.83
[-5.09] [-10.97] [1.29] [0.84] [-1.13] [-0.36] [0.87] [-0.61] [0.30] [0.68]
69
Figure A.1. BMS Excess Returns in Event Time: Four Liquid Currency Pairs
This ?gure is similar to Figure 3 but here BMS returns are based on sorting four liquid currencies (EUR/USD,
JPY/USD, GBP/USD, CHF/USD) into two portfolios.
70
doc_447151224.pdf
We find that order flows are highly informative about future exchange rates and provide significant economic value for the few large dealers who have access to these flows. Moreover, customer groups systematically engage in risk sharing with each other and differ markedly in their predictive ability, trading styles, and risk exposure.
BIS Working Papers
No 405
Information Flows in Dark
Markets: Dissecting Customer
Currency Trades
by Lukas Menkhoff, Lucio Sarno, Maik Schmeling and
Andreas Schrimpf
Monetary and Economic Department
March 2013
JEL classification: F31, G12, G15.
Keywords: Order Flow, Foreign Exchange Risk Premia,
Heterogeneous Information, Carry Trades, Hedge
Funds.
BIS Working Papers are written by members of the Monetary and Economic
Department of the Bank for International Settlements, and from time to time by
other economists, and are published by the Bank. The papers are on subjects of
topical interest and are technical in character. The views expressed in them are
those of their authors and not necessarily the views of the BIS.
This publication is available on the BIS website (www.bis.org).
© Bank for International Settlements 2013. All rights reserved. Brief excerpts may be
reproduced or translated provided the source is stated.
ISSN 1020-0959 (print)
ISBN 1682-7678 (online)
Information Flows in Dark Markets:
Dissecting Customer Currency Trades
?
Lukas Menkho?
†
Lucio Sarno
‡
Maik Schmeling
??
Andreas Schrimpf
§
This version: March 5, 2013
Abstract
We study the information in order ?ows of di?erent customer segments in the world’s
largest over-the-counter market, the foreign exchange market. The analysis draws on
a unique dataset covering a broad cross-section of currency pairs and distinguishing
trades by key types of foreign exchange end-users. We ?nd that order ?ows are highly
informative about future exchange rates and provide signi?cant economic value for the
few large dealers who have access to these ?ows. Moreover, customer groups system-
atically engage in risk sharing with each other and di?er markedly in their predictive
ability, trading styles, and risk exposure.
JEL Classi?cation: F31, G12, G15.
Keywords: Order Flow, Foreign Exchange Risk Premia, Heterogeneous Information, Carry Trades,
Hedge Funds.
?
We would like to thank an anonymous Referee, Alessandro Beber, Claudio Borio, Geir Bjønnes, Michael
Brandt, Steve Cecchetti, Jacob Gyntelberg, Hendrik Hakenes, Campbell Harvey, Joel Hasbrouck, Terrence
Hendershott, Søren Hvidkjær, Gur Huberman, Alex Kostakis, Jeremy Large, Albert Menkveld, Roel Oomen,
Richard Payne, Alberto Plazzi, Lasse Pedersen, Tarun Ramadorai, Jesper Rangvid, Paul S¨oderlind, Christian
Upper, Adrien Verdelhan, Michel van der Wel, as well as participants at several conferences, workshops and
seminars for helpful comments and suggestions. We are very grateful to Gareth Berry, Geo?rey Kendrick
and UBS for providing us with the proprietary data used in this study, and for numerous conversations
on the institutional details of foreign exchange trading at UBS. Sarno acknowledges ?nancial support from
the Economic and Social Research Council (No. RES-062-23-2340) and Menkho? and Schmeling gratefully
acknowledge ?nancial support by the German Research Foundation (DFG). The views expressed in this paper
are those of the authors and do not necessarily re?ect those of the Bank for International Settlements.
†
Kiel Institute for the World Economy and Department of Economics, Leibniz Universit¨at Hannover,
K¨onigsworther Platz 1, 30167 Hannover, Germany, Tel: +49 511 7624552, Email: [email protected]
hannover.de.
‡
Cass Business School and Centre for Economic Policy Research (CEPR). Corresponding author: Faculty
of Finance, Cass Business School, City University London, 106 Bunhill Row, London EC1Y 8TZ, UK, Tel:
+44 20 7040 8772, Fax: +44 20 7040 8881, Email: [email protected].
??
Faculty of Finance, Cass Business School, City University London, 106 Bunhill Row, London EC1Y 8TZ,
UK, Email: [email protected].
§
Bank for International Settlements and CREATES, Centralbahnplatz 2, 4002 Basel, Switzerland. Tel:
+41 61 280 8942. Email: [email protected].
The foreign exchange (FX) market is the largest ?nancial market in the world with a
daily trading volume of about four trillion U.S. dollars (BIS, 2010). Also, the FX market is
largely organized as an over-the-counter (OTC) market, meaning that there is no centralized
exchange and that market participants can have only partial knowledge about trades of other
market participants and available liquidity in di?erent market segments. Hence, despite its
size and sophistication, the FX market is fairly opaque and decentralized because of its
market structure. Adding to this lack of transparency, various trading platforms have been
introduced and market concentration has risen dramatically over the last decade with a
handful of large dealers nowadays controlling the lion’s share of FX market turnover (see,
e.g., King, Osler, and Rime, 2012). The FX market can thus be characterized as a fairly
“dark” market.
1
This paper addresses several related questions that arise in this opaque market setting.
First, do large dealers have an informational advantage from seeing a large portion of cus-
tomer trades, that is, do customer trades carry economic value for the dealer? Answering
this question is relevant for regulators and useful for understanding the implications of the
observed shift in market concentration. Second, how does risk sharing take place in the FX
market? Do customers systematically trade in opposite directions to each other or is their
trading positively correlated and unloaded onto dealers (as in, e.g., Lyons, 1997)? Answering
these questions is highly relevant to provide a better understanding of the working of the
FX market and, more generally, the functioning of OTC markets. Third, what characterizes
di?erent customer groups’ FX trading, e.g., do they speculate on trends or are they con-
trarian investors? In which way are they exposed to or do they hedge against market risk?
Answering these questions allows for a better grasp of what ultimately drives the demand for
currencies from di?erent types of end-users and enhances the knowledge about the ecology
of the world’s largest ?nancial market.
We empirically tackle these questions by means of a unique data set covering more than
ten years of daily end-user order ?ow for up to ?fteen currencies from one of the top FX
1
Du?e (2012) provides a general overview of how opaqueness and market structure impact price discovery
and trading in “dark markets”, that is, OTC markets.
1
dealers, UBS. The data are disaggregated into four di?erent groups of ?nancial (asset man-
agers and hedge funds) and non-?nancial (corporate and private clients) end-users of foreign
exchange. We therefore cover the trading behavior of various segments of end-users that are
quite heterogenous in their motives of market participation, informedness and sophistication.
Putting these data to work, we ?nd that: (i) Order ?ow by end-users is highly informative for
future exchange rate changes and carries substantial economic value for the dealer observing
these ?ows; (ii) there is clear evidence that di?erent end-user segments actively share risks
with each other; and (iii) end-user groups follow very heterogeneous trading styles and strate-
gies and di?er in their exposures to risk and hedge factors. This heterogeneity across players
is crucial for risk sharing and helps explain the vast di?erences in the predictive content of
?ows across end-user segments that we document in this paper.
To gauge the impact of order ?ow on currency excess returns, we rely on a simple portfolio
approach. This multi-currency framework allows for a straightforward measurement of the
economic value of the predictive content of order ?ow and is a pure out-of-sample approach
in that it only conditions on past information. Speci?cally, we sort currencies into portfolios
to obtain a cross-section of currency excess returns, which mimics the returns to customer
trading behavior and incorporates the information contained in (lagged) ?ows.
2
The infor-
mation contained in customer trades is highly valuable from an economic perspective: We
?nd that currencies with the highest lagged total order ?ows (that is, the strongest net buy-
ing pressure across all customer groups against the U.S. dollar) outperform currencies with
the lowest lagged ?ows (that is, the strongest net selling pressure across all customer groups
against the U.S. dollar) by about 10% per annum (p.a.).
For portfolios based on disaggregated customer order ?ow, this spread in excess returns
is even more striking. A zero-cost long-short portfolio that mimics asset managers’ trading
behavior yields an average excess return of 15% p.a., while conditioning on hedge funds’ ?ows
leads to a spread of about 10% p.a. Flows by corporate customers basically generate no spread
in returns, whereas private customers’ ?ows even lead to a highly negative spread (about -14%
p.a.). In sum, we ?nd that order ?ow contains signi?cant economic value for a dealer with
2
Lustig and Verdelhan (2007) were the ?rst to build cross-sections of currency portfolios.
2
access to such information. Hence, the trend towards more market concentration observed
in FX markets over recent years clearly bene?ts large ?nancial institutions acting as dealers
and potentially trading on this information in the inter-dealer market. These informational
advantages of dealers are further enhanced by the non-anonymous nature of transactions in
OTC markets, as trades by di?erent categories of customers convey fundamentally di?erent
information for price movements.
What drives the predictive content in ?ows? We investigate three main channels. First,
order ?ow could be related to the processing of information by market participants via the
process of “price discovery”. According to this view, order ?ow acts as the key vehicle that
impounds views about (economic) fundamentals into exchange rates.
3
If order ?ow contains
private information, its e?ect on exchange rates is likely to be persistent. Second, there
could be a price pressure (liquidity) e?ect due to downward-sloping demand curves (e.g.,
Froot and Ramadorai, 2005). If a mechanism like this is at play, we are likely to observe a
positive correlation between ?ows and prices for some limited time, followed by a subsequent
reversal as prices revert to fundamental values.
4
Third, we consider the possibility that
order ?ow is linked to returns due to the di?erent risk sharing motives and risk exposures of
market participants. For example, order ?ow could re?ect portfolio rebalancing of investors
tilting their portfolios towards currencies that command a higher risk premium. Related to
this, risk sharing could lead to the observed predictability pattern if non-?nancial customers
are primarily concerned about laying o? currency risk and implicitly paying an insurance
premium, whereas institutional investors are willing to take on that risk.
Discriminating between alternative explanations for the predictive content of order ?ow,
we ?nd clear di?erences across the four segments of end-users. Asset managers’ ?ows are
associated with permanent shifts in future exchange rates, suggesting that their order ?ow is
3
See, e.g., Payne (2003), Love and Payne (2008), Evans and Lyons (2002a, 2007, 2008), Evans (2010),
and Rime, Sarno, and Sojli (2010). Other papers relate order ?ow in a structural way to volatility (Berger,
Chaboud, and Hjalmarsson, 2009) or directly to exchange rate fundamentals (Chinn and Moore, 2011).
4
Several studies explore the underlying mechanism for the impact of order ?ow and discuss the evidence
in terms of information versus liquidity e?ects (e.g. Berger, Chaboud, Chernenko, Howorka, and Wright,
2008; Fan and Lyons, 2003; Marsh and O’Rourke, 2005; Osler, Mende, and Menkho?, 2011; Menkho? and
Schmeling, 2010; Phylaktis and Chen, 2010; Moore and Payne, 2011; Ito, Lyons, and Melvin, 1998).
3
related to superior processing of fundamental information.
5
In contrast, hedge funds’ ?ows
are merely associated with transitory exchange rate movements, that is, the impact of their
trades on future exchange rates is far less persistent. This result is more in line with short-
term liquidity e?ects but not with fundamental information processing. Corporate customers’
and private clients’ ?ows, however, seem to re?ect largely uninformed trading.
Our results also point to a substantial heterogeneity across customers in their trading
styles and risk exposures, giving rise to di?erent motives for risk sharing. First, we ?nd
that the trades of various end-user groups react quite di?erently to past returns. Asset
managers tend to be“trend-followers”(positive feedback traders) with regard to past currency
returns. By contrast, private clients tend to be “contrarians” (negative feedback traders).
The latter ?nding squares well with recent ?ndings for equity markets by Kaniel, Saar, and
Titman (2008) who show that individual equity investors behave as contrarians, e?ectively
providing liquidity for institutional investors. Di?erent from their results, however, private
clients do not directly bene?t from serving as (implicit) counterparties of ?nancial customers
in FX markets. Second, the ?ows of most customer groups are negatively correlated over
short to intermediate horizons, suggesting that di?erent groups of end-users in FX markets
engage in active risk sharing among each other. It is thus not just via the inter-dealer
market that risk is shared in FX markets, as documented by Lyons (1997), but a signi?cant
proportion of risk is shared among end-users in the customer-dealer segment. Third, we
?nd substantial heterogeneity in the exposure to risk and hedge factors across customer
segments. Asset managers’ trading does not leave them exposed adversely to systematic risk,
which suggests that the information in their ?ows is not due to risk taking but likely re?ects
superior information. Hedge funds, by contrast, are signi?cantly exposed to systematic risk
such as volatility, liquidity, and credit risk. This lends credence to the view that hedge funds
earn positive returns in FX markets by e?ectively providing liquidity and selling insurance
to other market participants. For non-?nancial customers there is some evidence of hedging
but it is not strong enough to fully explain their negative forecast performance arising from
5
This information processing can come in di?erent ways, e.g., a more accurate and/or faster interpretation
of macroeconomic news releases, and better forecasting of market fundamentals such as liquidity and hedging
demands of other market participants.
4
poor short-term market timing.
Our paper is related to prior work on the microstructure approach to exchange rates
(e.g., Evans and Lyons, 2002a,b), which suggests that order ?ow is crucial for understanding
how information is incorporated into exchange rates. It is well known from the literature
that order ?ow is positively associated with contemporaneous returns in basically all asset
classes; see, e.g., Hasbrouck (1991a,b) for stock markets, and Brandt and Kavajecz (2004)
for U.S. bonds. This is a stylized fact which also holds in FX markets, as shown by Evans
and Lyons (2002a) and many subsequent studies. There is less clear evidence, however, on
whether order ?ow predicts exchange rates. A few papers have shown that FX order ?ow
contains information about future currency returns but tend to disagree on the source of this
predictive power (e.g., Evans and Lyons, 2005; Froot and Ramadorai, 2005; Rime, Sarno,
and Sojli, 2010).
6
Some other papers fail to ?nd robust predictive power of exchange rates
by order ?ow in the ?rst place (see, e.g., Sager and Taylor, 2008). Our work is also related
to a di?erent strand of recent literature that analyzes the returns to currency portfolios by
investigating the predictive power of currency characteristics, such as carry or lagged returns,
and the role of risk premia in currency markets.
7
Overall, we contribute to the literature in the following ways. We are the ?rst to show
that order ?ow forecasts currency returns in an out-of-sample forecasting setting by forming
order ?ow portfolios. This multi-currency investment approach provides an intuitive mea-
sure of the economic value of order ?ow for the few large dealers observing these ?ows. This
seems important as earlier papers either did not consider out-of-sample forecasting at all
or relied on purely statistical performance measures derived from time-series forecasts of a
limited number of currency pairs (e.g., Evans and Lyons, 2005, who study the DEM/USD
and JPY/USD crosses). Time-series forecasts are a?ected by trends in exchange rates, most
notably the U.S. dollar. Our portfolio procedure, by contrast, studies exchange rate pre-
6
There is also evidence that marketwide private information extracted from equity order ?ow is useful for
forecasting currency returns (Albuquerque, de Francisco, and Marques, 2008).
7
Lustig and Verdelhan (2007), Farhi, Fraiberger, Gabaix, Ranciere, and Verdelhan (2009), Ang and Chen
(2010), Burnside, Eichenbaum, Kleshchelski, and Rebelo (2011), Lustig, Roussanov, and Verdelhan (2011),
Barroso and Santa-Clara (2011) and Menkho?, Sarno, Schmeling, and Schrimpf (2012a,b) all build currency
portfolios to study return predictability and/or currency risk exposure.
5
dictability in dollar-neutral long-short portfolios, and it does so out-of-sample over very long
time spans compared to the extant FX microstructure literature. Moreover, we are the ?rst
to test whether risk exposure drives the information in customer order ?ows. We show how
di?erent key FX market players trade, e.g., to which extent they rely on trend-following or
behave as contrarians, and in which ways they are exposed to systematic risk. We ?nd strong
evidence of heterogeneity in the exposures and trading behavior across di?erent groups of
market participants. These ?ndings indicate that there is signi?cant risk sharing between ?-
nancial and non-?nancial customers as well as between di?erent groups of ?nancial customers
(leveraged versus real money managers).
Taken together, these results have implications for our general understanding of informa-
tion ?ows in dark markets and how large dealers in OTC markets bene?t from observing a
large proportion of the order ?ow. These results also add to our general understanding of
how risk is shared in ?nancial markets due to di?erent motives for trade and trading styles
across end-user segments.
The rest of the paper is structured as follows. Section I describes our data, Section
II presents empirical results on the predictive power of order ?ow, Section III empirically
investigates alternative underlying reasons for why order ?ow forecasts FX excess returns,
and Section IV presents results of robustness tests. Section V concludes.
I. Data
We employ a unique dataset based on daily customer order ?ows for up to 15 currency pairs
over a sample period from January 2, 2001 to May 27, 2011, for a total of 2,664 trading
days. In contrast to much of the earlier literature, we employ order ?ow from the end-
user segment of the FX market and not from the inter-dealer market. This is important
since microstructure models suggest that the information in ?ows stems from trading with
customers and not from inter-dealer trading (e.g. Evans and Lyons, 2002a). Order ?ows in
our sample are measured as net buying pressure against the U.S. dollar (USD), that is, the
U.S. dollar volume of buyer-initiated minus seller-initiated trades of a currency against the
6
USD. The data cover all trades of customers (end-users) with UBS during our sample period.
A positive number indicates net buying pressure in the foreign currency relative to the USD.
Order ?ow therefore does not measure trading volume but net buying (or selling) pressure, as
mentioned above. Our order ?ow data are available both in aggregated form and at a higher
level of granularity allowing for a di?erentiation across end-user groups.
Aggregate order ?ow. Aggregate order ?ows, that is, aggregated across customers (re-
gardless of their type), are available for the following 15 currencies: Australia (AUD), Brazil
(BRL), Canada (CAD), the Euro (EUR), Hong Kong (HKD), Japan (JPY), Sweden (SEK),
Mexico (MXN), New Zealand (NZD), Norway (NOK), Singapore (SGD), South Africa (ZAR),
South Korea (KRW), Switzerland (CHF), and the United Kingdom (GBP). In the following,
we refer to these ?ows as “total ?ows” since they are aggregated across all customers of UBS.
A natural question is whether ?ows by customers of UBS are generally representative
of end-user currency demands in the FX market. While this question cannot be answered
without knowledge of the customer ?ows of all other dealers, there are good reasons to believe
that the ?ows employed in our paper are highly correlated with a large portion of end-user
order ?ows. First, UBS is among the largest dealers in the FX market and their average
market share (according to the Euromoney FX Survey) over our sample period amounts to
about 13%. Over most of our sample period, UBS was ranked among the top three of all
FX dealers (with Deutsche Bank, Barclays, and Citi usually being the closest competitors).
Thus, UBS clearly is one of the most important FX dealers with a signi?cant portion of the
market solely on its own.
8
Second, a handful of top dealers in the FX market account for
more than 50% of total market share (e.g., King, Osler, and Rime, 2012) and all of these large
dealers essentially have access to the same set of large customers. Hence, it seems very likely
that UBS ?ows’ are highly correlated with ?ows observed at, e.g., Deutsche Bank, Barclays,
Citi, or JP Morgan, which in turn implies that our order ?ows are representative of the top
end of customer trading in the FX market.
8
Note that most UBS FX customers are in fact big players and include other banks, many large asset
management ?rms and hedge funds, and a large fraction of wealthy private clients. According to the Eu-
romoney survey, UBS has a particularly high market share in FX business with ?nancial customers (banks,
real money and leveraged funds).
7
Table I about here
Table I shows descriptive statistics for total ?ows (in billion USD). Daily order ?ows are
largest on average for EUR, JPY, and CHF. The pair with the largest average imbalance (in
absolute value) between buyer- and seller-initiated trading volume is the EUR/USD, where
customers (on a net basis) sold on average 63 million EUR against USD per trading day over
our sample period. Hence, average order ?ows are fairly small relative to gross daily trading
volume in FX markets.
9
Flows are fairly volatile, however, which means that order ?ow
imbalances can frequently be very large. Daily ?ows tend to be positively autocorrelated,
but the degree of autocorrelation is very small albeit sometimes statistically signi?cant. There
is also a clear pattern in standard deviations. Major currencies, such as the EUR, CHF, JPY,
GBP, have much larger variation in order ?ows and, hence, a larger absolute size of order
?ows compared to other currencies and especially emerging markets. This is intuitive as
there is much more trading in major currencies, but it also suggests that one cannot easily
compare order ?ows across currencies and that some form of standardization is needed to
make sensible comparisons.
10
We take this into account in our empirical analysis below.
Finally, aggregate order ?ows display a high kurtosis (especially the British pound), which is
largely driven by some days with extremely high (in absolute value) order ?ows. Eliminating
these few outliers does not change our results reported below.
Disaggregated order ?ow. We also obtain order ?ows disaggregated by customer groups
for the same sample period, albeit only for a subset of nine major currencies.
11
There are four
customer groups for which ?ows are available: Asset Managers (AM), Hedge Funds (HF),
Corporate Clients (CC), and Private Clients (PC). The segment of asset managers comprises
“real money investors”, such as mutual funds and pension funds. Highly leveraged traders and
short-term oriented asset managers not included in the asset managers segment are classi?ed
9
To provide a benchmark, daily gross spot turnover in the Euro/USD pair in April 2010 amounted to USD
469 billion according to the most recent FX triennial survey (BIS, 2010). These (gross) ?gures for both the
customer-dealer segment and the inter-dealer market are based on data collected from about 4,000 reporting
dealers worldwide.
10
In addition, the volatility of ?ows also varies over time and ?ows tend to become increasingly volatile
towards the end of the sample. Also for this reason, some form of standardization is necessary.
11
The nine currencies are: AUD, CAD, EUR, JPY, SEK, NZD, NOK, CHF, and GBP.
8
as hedge funds. Hedge funds are unregulated entities, whereas asset managers are regulated.
The corporate segment includes non-?nancial corporations that import or export products
and services around the world or have an international supply chain. Corporates also include
the treasury units of large non-?nancial corporations, with the exception of those pursuing a
highly leveraged investment strategy, which are classi?ed by UBS as hedge funds. The last
segment, private clients, includes wealthy clients with investable liquid assets in excess of 3
million U.S. dollars. Private clients trade primarily for ?nancial reasons and with their own
money. Hence, there is substantial heterogeneity in the motives for market participation by
these four customer types, and the groups are likely to di?er considerably in the degree of
informedness and sophistication. One of the key features of OTC markets, that is, the non-
anonymous nature of transactions can thus further enhance the informational advantages of
dealers in dark markets.
The order ?ow data are assembled as follows. Each transaction booked in the UBS
execution system at any of its world-wide o?ces is tagged with a client type. At the end of
each business day, global transactions are aggregated for each customer group. Order ?ow is
measured as the di?erence between the dollar value of purchase and sale orders for foreign
currency initiated by a particular UBS customer group. The transaction is recorded with
a positive sign if the initiator of the transaction (the non-quoting counterparty) is buying
foreign currency and vice versa.
12
Summary statistics for the disaggregated order ?ow data
are reported in Table A.1 of the Internet Appendix.
Exchange rate returns and excess returns. For our empirical analysis below, we com-
plement these order ?ow data with daily spot exchange and forward rates from Reuters
(available from Datastream). We denote log changes of spot exchange rates as ‘exchange rate
12
Our data are raw order ?ow data with ?ltering limited to the most obvious cases. For instance, data
are adjusted for large merger and acquisition deals which are announced well in advance. Cross-border
mergers and acquisitions involve large purchases of foreign currency by the acquiring company to pay the
cash component of the deal. These transactions are generally well-publicized and thus are anticipated by
market participants. Finally, FX reserve managers, UBS proprietary traders and small banks not participating
in the inter-dealer market are excluded from the data. Flows from FX reserve managers are stripped out
due to con?dentiality issues, ?ows from proprietary traders because they trade with UBS’ own money, while
small banks represent small customers less concerned about the FX market.
9
returns’
?s
t+1
= s
t+1
? s
t
, (1)
where lowercase letters refer to logs and all exchange rates are quoted as the USD price
of foreign currency, so that positive exchange rate returns correspond to an appreciation of
the foreign currency. Hence, a positive correlation of order ?ows and exchange rate returns
means that net buying pressure in the foreign currency (against the USD) is associated with
an appreciation of the foreign currency (against the USD) and vice versa.
We also compute currency excess returns which account for the interest rate di?erential
in a foreign currency position. Hence, currency excess returns rx are given by
rx
t+1
= s
t+1
? s
t
+ (i
t
? i
t
), (2)
where i
denotes the foreign interest rate and i
t
denotes the U.S. interest rate. Since we
are working at the daily frequency in our main analysis, we need to obtain daily interest
rates for all 15 countries (plus the U.S. interest rate). However, since one-day interest rates
are not directly available for all countries in our sample, we employ information in forward
rates to infer interest rate di?erentials. Interest rate di?erentials for horizon k are commonly
approximated by i
k,t
?i
k,t
? s
t
?f
k,t
where f
k,t
denotes the log forward rate for horizon k of
a given currency.
13
II. The Value of Information in Customer Flows
A. Portfolios Conditioning on Aggregate Order Flow
We rely on a portfolio approach, mimicking the returns to customer FX trading by condi-
tioning on lagged order ?ow. This provides a straightforward and intuitive assessment of the
13
This approximation is exact if covered interest rate parity (CIP) holds, which tends to be the case at
daily or even shorter horizons in normal times (Akram, Rime, and Sarno, 2008). There have been violations
of this no-arbitrage relation over the recent ?nancial crisis. As we show below, the results in this paper are
entirely driven by changes in spot rates, whereas interest rate di?erentials only play a negligible role. Thus,
the results do not depend on whether CIP holds or not.
10
economic value of order ?ow in predicting currency excess returns.
As a benchmark test, we ?rst sort currencies into portfolios based on (lagged) total order
?ows for each currency. Speci?cally, we sort currencies into ?ve portfolios (P
1
, P
2
, ..., P
5
)
depending on their total order ?ow on day t and compute portfolio excess returns (or spot
exchange rate changes) for the following day. In this basic setup, portfolios are rebalanced
at the end of each trading day. Note that these portfolios are computed from the viewpoint
of a U.S. investor as each individual portfolio consists of a short position in USD and a long
position in a basket of foreign currencies. Taking the return di?erence between any two
portfolios P
j
?P
i
thus gives the return of a portfolio short in the basket of foreign currencies
in P
i
and long in the basket of currencies in P
j
, so that the USD component cancels out and
the long-short portfolio is dollar-neutral by construction.
Standardizing order ?ows. Before sorting currencies into portfolios, we need to make
sure that order ?ows are comparable across currencies. As the absolute size of order ?ows
di?ers across currencies (as shown above in Table I) it is not sensible to sort currencies based
on raw order ?ows. To allow for meaningful cross-currency comparisons, it is necessary to
standardize ?ows. We do this by dividing ?ows by their standard deviation to remove the
di?erence in absolute order ?ow sizes across currencies
x
R
j,t
=
x
j,t
?(x
j,t?59;t
)
, (3)
where x
R
j,t
denotes order ?ow standardized over a rolling window and x
j,t
denotes the raw
order ?ow. In our baseline results, we compute the standard deviation of ?ows via a rolling
scheme over a 60-day rolling window. Robustness tests based on alternative approaches to
standardize ?ows are reported in a separate Internet Appendix.
14
Portfolio excess returns. Table II shows average annualized excess returns for order ?ow
14
In these robustness exercises, we also report results with longer rolling windows of up to three years as
well as for an expanding window. Furthermore, we provide tests where we standardize both with respect to
volatility as well as the mean. Finally, we also consider a standardization scheme based on gross FX turnover
data for di?erent currencies drawing on data from the BIS FX triennial survey. These tests, reported in the
separate Internet Appendix to conserve space, show that our results are not sensitive with regard to the way
?ows are standardized.
11
portfolios (P
1
, P
2
, ..., P
5
), where P
1
contains the three currencies with the lowest lagged stan-
dardized order ?ow and P
5
contains the three currencies with the highest lagged standardized
order ?ow. Hence, P
5
can be thought of as a portfolio of currencies with the highest buying
pressure, whereas P
1
refers to a portfolio with the strongest selling pressure. Column “Av.”
shows average returns across all currencies in the cross-section and column “BMS” denotes a
portfolio which is long in P
5
and short in P
1
(“Buying Minus Selling” pressure). We report
returns for the full sample period from January 2001 to May 2011.
15
To get started, Panel A of Table II reports results for the sample of all 15 markets (T15)
as well as for the sub-sample of 9 developed markets (T9); for the T9 sub-sample we only
form four portfolios rather than ?ve to ensure we always have two currencies in the corner
portfolios. We observe a strong increase in average excess returns as we move from the
portfolio of currencies with low buying pressure P
1
to the one with high buying pressure P
5
(or P
4
for the T9 sample). The spread in excess returns between the high buying pressure
and the low buying pressure portfolio, that is, the excess return of the BMS portfolio, is
economically large (10.31% and 12.43% p.a., respectively) and statistically highly signi?cant.
Similarly, the Sharpe Ratios (p.a.) of the two BMS portfolios of 1.26 and 1.45 are large and
also point toward high economic signi?cance. Thus, order ?ows carry signi?cant information
for future currency excess returns, as captured by our dollar-neutral out-of-sample trading
strategy which only conditions on real-time information. These results demonstrate the
economic value for the owner of this (private) information, that is, the few large FX dealer
banks which observe a signi?cant share of end-user order ?ow and are able to trade on this
information in the inter-dealer market.
Table II about here
Table A.3 in the Internet Appendix shows results for the other standardization schemes
and for sub-samples. We ?nd that our results are equally strong in various sub-periods.
Likewise, Table A.4 in the Internet Appendix shows the same exercise for exchange rate
15
Sub-sample tests for a pre-crisis subperiod from January 2001 to June 2007, and a crisis/post-crisis
subperiod from July 2007 to May 2011 are reported in the Internet Appendix.
12
changes instead of excess returns. Results in that table clearly show that the patterns in
average spot exchange rate changes across portfolios are at least as pronounced as for average
excess returns or, if anything, even more pronounced. Hence, order ?ow is informative about
future spot rates and not about interest rate di?erentials.
Tests for return monotonicity. The last three columns “MR” and “Up” in Table II report
tests for return monotonicity (Patton and Timmermann, 2010), that is, whether there is a
signi?cantly increasing or decreasing pattern of average excess returns when moving from the
portfolio of low buying pressure (P
1
) to the one with high buying pressure (P
5
).
16
These tests
go beyond the standard t-test of a zero BMS portfolio return since they take into account
the whole cross-sectional pattern. This is interesting since one would intuitively expect an
increasing pattern of average portfolio excess returns when moving from P
1
to P
5
if order ?ow
is truly informative about future excess returns. This prediction is signi?cantly borne out in
the data for both the T15 and T9 sample of countries and for both the “MR” and “Up” test.
Hence, there is strong evidence for a signi?cant relationship between order ?ow and future
excess returns.
Excess returns over time. Finally, we plot cumulative excess returns for the T15 and T9
BMS portfolios in the upper left and right panel of Figure 1. As can be seen, excess returns
are quite striking and stable for most of the sample period, although somewhat more volatile
at the beginning and towards the end of the sample.
Figure 1 about here
16
The MR statistic tests for a monotonically increasing return pattern, whereas the Up (Down) test is
somewhat less restrictive and simply tests for a generally increasing (decreasing) pattern without requiring
monotonicity in average portfolio returns. Speci?cally, the MR test requires that the return pattern is
monotonically increasing P
1
< P
2
< ... < P
5
and formulates the null hypothesis as H
0
: ? ? 0 and the
alternative hypothesis as H
a
: min
i=1,...,4
i
> 0, where ? is a vector of di?erences in adjacent average
portfolio excess returns (P
2
? P
1
, P
3
? P
2
, P
4
? P
3
, P
5
? P
4
) and
i
is element i of this vector. The
Up test formulates the null hypothesis of a ?at pattern H
0
: ? = 0 and the alternative hypothesis as
H
a
:
4
n=1
|
i
|1{
i
> 0} > 0, so that the test is less restrictive and also takes into account the size and
magnitude of deviations from a ?at return pattern. The Down test follows in an analogous way.
13
B. Portfolios Conditioning on Disaggregated Order Flow
If superior information processing or genuine forecasting ability drive our results above, one
expects clear di?erences in the forecasting power of di?erent customers’ ?ows, depending on
the groups’ characteristics (see, e.g., Fan and Lyons, 2003; Evans and Lyons, 2007, among
others). Speci?cally, one would expect to see superior information in ?ows of ?nancial cus-
tomers, given that non-?nancial players do not specialize in FX trading as their core activity.
To investigate this, we now build portfolios based on our disaggregated data for customer
?ows. We closely follow the earlier approach with the exception that we only build four
portfolios (rather than ?ve) here since we only have disaggregated ?ows for nine currencies
and want to have a minimum of two currencies per portfolio.
Table II, Panel B, reports results for the four customer groups: Asset Managers (AM),
Hedge Funds (HF), Corporate Clients (CC), and Private Clients (PC). Results are clear-
cut. Asset managers’ net buying or selling pressure of currencies is the most informative
about subsequent exchange rate behavior. Conditioning on asset managers’ ?ows generates a
cross-sectional spread in excess returns of 15% p.a., followed by hedge funds with a spread of
about 10%. In stark contrast, corporate clients’ and private clients’ ?ows actually generate
a negative spread in portfolio excess returns of about ?4% and ?14%, respectively.
17
The
results point towards substantial di?erences in the customers’ predictive information and
provide a quantitative summary of the value of this information in economic terms. The
latter is underscored by the large spread in (annualized) Sharpe Ratios of BMS portfolios
across customer groups. The asset managers’ BMS portfolio yields a Sharpe Ratio of 1.79,
whereas the private clients’ BMS portfolio has a Sharpe Ratio of -1.55.
18
As above, we also present p-values for tests of return monotonicity. Since order ?ow of
17
Table A.5 shows results for spot rate changes instead of excess returns, which display no qualitative
di?erences.
18
Table A.6 in the Internet Appendix also shows that excess returns to the BMS portfolios based on
di?erent customers’ ?ows are not highly correlated. Hence, the information contained in the di?erent ?ows
appears to stem from di?erent sources. In practice, this also means that BMS portfolios could be combined
to obtain even higher Sharpe Ratios. For example, a combined portfolio long in the asset managers’ BMS
portfolio and short in the private clients’ BMS portfolio yields an annualized Sharpe Ratio of 2.19, which is
substantially higher than the individual Sharpe Ratios.
14
corporate and private customers negatively forecasts returns, we modify the MR test in these
cases to test for a monotonically decreasing pattern. Results from these tests corroborate
the simple t-tests for the BMS portfolios. There is a monotonically increasing pattern in
average excess returns for portfolios based on asset managers’ and hedge funds’ ?ows which
is highly signi?cant. By contrast, we ?nd a monotonically decreasing pattern in average
excess returns for portfolios based on private customers’ ?ows, and marginally signi?cant
evidence for a decreasing pattern in portfolios based on corporate ?ows.
Hence, it is not the case that all order ?ow is equal in terms of its information content for
exchange rates. Instead, ?nancial customers’ ?ows (asset managers and hedge funds) account
for the positive relation between lagged ?ows and future exchange rate returns uncovered in
the previous section. Flows by corporates are more or less uninformative, whereas private
clients’ ?ows even forecast returns in the wrong direction. Using total end-user order ?ow,
which is likely to be dominated by ?nancial customers due to their higher trading volume
19
,
masks these di?erences and might even lead to wrong inference about the link between ?ows
and returns. In a nutshell, what matters for the relation between end-user order ?ows and
future returns is disaggregated data since the information content of ?ows for future returns
varies markedly across customer groups.
The middle and lower panel of Figure 1 shows cumulative returns for all four customer
groups. It can directly be seen that returns are very di?erent across customer groups, even
when comparing, for example, asset managers and hedge funds. Both groups’ BMS portfo-
lios generate signi?cant excess returns but returns for hedge funds are much more volatile
than those of asset managers. Hence, we will investigate possible sources of these di?erent
behaviors of returns below.
C. Marginal Predictive Content of Flows at Longer Horizons
Our analysis so far has been concerned with the relation between order ?ows and returns over
the subsequent trading day. An interesting question, however, is whether the information
19
This is especially true for the order ?ow employed in this paper since UBS is one of the largest dealers
in FX and has a high proportion of ?nancial customers (relative to corporate clients).
15
contained in order ?ow quickly decays or whether it is useful for forecasting returns over more
than one trading day.
We examine the marginal predictive content of ?ows by forming portfolios as in the
analysis above, but we now allow for a longer lag between the order ?ow signal and portfolio
formation. Table III contains the results for di?erent lags of 0, 1, 2, . . . , 9 days. To be more
speci?c, a lag of 0 days means that ?ows of trading day t are used to predict returns of day
t +1 (and thus reproduces BMS returns from Tables II above), whereas a lag of, e.g., 2 days
means that ?ows of day t are used to forecast returns of trading day t + 3.
Table III about here
Results in Table III show that order ?ow appears to be most informative for the ?rst
two to three days after portfolio formation and that the information in ?ows becomes in-
signi?cant afterwards. Hence, the information contained in daily ?ows is fairly short-lived
and is impounded relatively quickly into exchange rates, especially considering that the order
?ows employed here are private information that is available to only a small number of large
FX dealers and that could not realistically be incorporated into prices without some lag.
This ?nding is in contrast to, e.g., Evans and Lyons (2005) who study a shorter and smaller
sample and ?nd that times-series predictability of returns by order ?ow increases at longer
horizons when judged from statistical metrics of forecast evaluation. This contrast in results
also highlights the importance to assess the predictive power of order ?ow using measures of
economic value as opposed to purely statistical ones, as statistical evidence of exchange rate
predictability in itself does not guarantee that an investor can earn pro?ts from a trading
strategy that exploits this predictability.
D. Order Flow vs. Carry and Momentum
To further learn about the predictive content of customer order ?ow for future FX returns, we
run panel regressions, which allows us to control for other possible determinants of currency
excess returns as well as cross-sectional and time ?xed-e?ects. For example, it could be the
16
case that asset managers’ and hedge funds’ order ?ow mimicking portfolios simply reproduce
a carry trade (Burnside, Eichenbaum, Kleshchelski, and Rebelo, 2011; Lustig, Roussanov,
and Verdelhan, 2011; Menkho?, Sarno, Schmeling, and Schrimpf, 2012a) or that their order
?ow just picks up momentum e?ects in currency returns (Menkho?, Sarno, Schmeling, and
Schrimpf, 2012b).
Speci?cally, we run panel regressions of the general form
rx
j,t+1
= ?
c
OF
c
t
+ ?
1
(i
j,t
? i
t
) + ?
2
rx
j,t
+ ?
3
rx
j,t?60;t?1
+ ?
j,t+1
(4)
where j (1, ..., N) indexes currencies, rx denotes currency excess returns, OF
c
denotes order
?ow of customer group c, (i
j,t
? i
t
) denotes interest rate di?erentials (carry), and rx
t
and
rx
t?60;t?1
denotes lagged excess returns over the prior trading day and the average over the
past 60 trading days, respectively.
20
The error term is given by ?
j,t+1
= e
t+1
+u
j
+
j,t+1
and
thus captures time and cross-sectional ?xed-e?ects (we also report results without ?xed-e?ects
below). Standard errors are clustered by currency pair. Note that these panel regressions
employ non-standardized order ?ows and are based on individual currency returns and not
on portfolio returns.
Results are shown in Table IV and corroborate our ?ndings based on our portfolio ap-
proach above, namely that order ?ows of ?nancials positively predict future excess returns,
whereas ?ows by non-?nancial end-users negatively forecast returns. Based on speci?cation
(vi) in Table IV, the coe?cients on lagged order ?ow imply that a positive order ?ow of USD
1 billion forecasts a four basis point (b.p.) higher excess return on the following day for asset
managers’ ?ows, a one b.p. higher return for hedge funds, a minus one b.p. lower return for
corporates, and a minus two b.p. lower return for private clients. The magnitude of these
e?ects seems reasonable given the deep liquidity of the FX market.
Table IV about here
More important, however, is the fact that the predictive relation between lagged order
20
Using other windows of less or more than 60 trading days does not yield qualitatively di?erent results.
17
?ow and future FX excess returns remains very strong when controlling for two common
predictors of returns in FX markets, interest rate di?erentials and (short-term) momentum.
Carry shows up with a positive sign, that is, high interest rate currencies deliver high excess
returns on average in line with the large literature on the forward discount bias (e.g., Fama,
1984). Interest rate di?erentials, however, do not drive out the information contained in
order ?ows and they become insigni?cant once we include cross-sectional ?xed-e?ects in the
regression. In our panel regressions, lagged currency returns do not have consistent predictive
content beyond order ?ow and carry, which is in line with recent evidence in Menkho?, Sarno,
Schmeling, and Schrimpf (2012b), who show that FX momentum strategies are not pro?table
for major exchange rates over the last decade.
III. What Drives the Predictive Power of Flows?
A. Permanent vs. Transitory Forecast Power of Flows
To better understand the driving forces behind our results above, we next investigate whether
order ?ow forecasts returns because it signals permanent shifts in spot exchange rates or
whether it merely forecasts temporary movements which are eventually reversed after some
time. The question whether order ?ow has a permanent or transitory e?ect in prices is a
central theme in the earlier microstructure literature (see Hasbrouck, 1991a,b). A transitory
movement is interpreted as suggesting that order ?ow e?ects are merely due to short-term
liquidity or price pressure e?ects which eventually die out, whereas a permanent movement
in spot rates would indicate that order ?ow conveys information about fundamentals.
21
More
speci?cally, a permanent price impact would most probably indicate that order ?ow is related
to changes in expectations about fundamentals given the daily frequency we are working
on. This question is relevant for our analysis since we ?nd substantial heterogeneity with
21
One strand of literature argues that order ?ow is the conduit by which information about fundamentals
is impounded in prices and therefore has permanent e?ect on exchange rates (e.g., Evans and Lyons, 2002a;
Brandt and Kavajecz, 2004; Evans and Lyons, 2007, 2008). Another strand of the literature suggests that
order ?ow matters due to downward sloping demand curves or “illiquidity” and, hence, that order ?ow only
has a transitory impact on prices (e.g., Froot and Ramadorai, 2005).
18
regard to the forecasting power of di?erent customer groups’ order ?ows. Therefore it is
particularly interesting to ?nd out if all (or some) customers’ ?ows signal information relevant
for permanent changes in FX rates or whether some customer groups’ order ?ow simply exerts
price pressure and liquidity e?ects.
To this end, we apply our portfolio sorts framework as above but now track cumulative
exchange rate returns to BMS portfolios for overlapping periods of 30 trading days after
portfolio formation. This approach yields a direct estimate of how spot rates move after
experiencing intensive buying or selling pressure by customers.
Figure 2 illustrates the persistence of the predictive content of order ?ow. The solid lines
show the cumulative excess returns (in basis points), whereas the shaded areas show 95%
con?dence intervals based on a moving-block bootstrap with 1,000 repetitions. Total ?ows
for all 15 currencies (T15) forecast a permanent change in spot rates which is statistically
signi?cantly di?erent from zero. Exchange rates with the highest net buying (selling) pressure
appreciate (depreciate) against the USD for approximately three days. Currency returns on
the BMS portfolios increase by about 15 basis points over this period, and afterwards the
e?ect of the order ?ow signal levels out. Importantly, these ?ndings suggest that order ?ow
conveys information and its impact on exchange rates is not reversed.
Figure 2 about here
This picture changes when looking only at the nine developed currencies. Here, we ob-
serve the same increasing pattern initially, followed by a subsequent partial reversal. After
approximately 25 ?30 trading days, about one half of the initial impact of 15 basis points is
reversed and the con?dence interval includes zero. Hence, there is much less evidence that
order ?ow conveys information about fundamentals when only looking at major developed
markets. This ?nding makes sense, however, since the major currency markets are most
probably more researched and more e?cient than smaller currency markets so that the scope
for superior information processing is reduced.
22
22
This may be interpreted in the context of the adaptive markets hypothesis (see e.g. Neely, Weller, and
Ulrich, 2009, for an analysis in FX markets).
19
As a natural next step, we also examine the same question separately for disaggregated
order ?ows (lower panels of Figure 2). Results are clear-cut. The only end-user group with a
statistically signi?cant permanent price impact is asset managers. Hedge funds’ trading has
a positive but transitory impact in line with an interpretation that they provide liquidity.
Corporate clients have no impact at all, and private clients have a transitory negative impact.
Given our ?nding for total ?ows of the nine major currencies above, it is interesting to see
that asset managers’ ?ows are indeed associated with permanent spot rate changes. Hence,
order ?ow of asset managers seems to be related to the processing of fundamental information
whereas hedge funds’ order ?ow corresponds to short-lived information unrelated to funda-
mental information. Similarly, it seems reasonable that the negative relation between private
clients’ ?ows and future spot rates eventually dies out over time.
These ?ndings are novel in the literature and suggest that order ?ows by di?erent end-
user groups – even by the two ?nancial customer groups – embed di?erent information for
future exchange rates. These di?erences can arise either because they are based on di?erent
mechanisms to process information or because of di?erent trading motives and hedging needs.
To explore this further, we investigate the drivers of order ?ow in more detail and shed light
on the observed di?erences in end-user order ?ows.
B. Risk Sharing Among Foreign Exchange End-Users
The analysis above suggests that asset managers’ ?ows are related to the processing of fun-
damental information that is quickly, but permanently, impounded into prices whereas the
other customer groups’ ?ows are not. A potential explanation is that risk sharing among
market participants drives (part of) our results. For instance, private clients’ negative BMS
returns could be explained by their possible need for hedging FX risk, whereas the positive
returns of hedge funds might implicitly re?ect a compensation for taking on such risks. While
these are just examples, a risk sharing story in general implies that we observe customers sys-
tematically trading in opposite directions and that their portfolios load on di?erent sources
of systematic risk. We investigate these issues below.
20
Portfolio returns in event time. We ?rst provide a more detailed look at the return be-
havior around portfolio formation dates to better understand di?erences in customer groups.
Figure 3 shows the average annualized BMS excess return for the ?ve days prior to portfo-
lio formation (?5, ?4, ..., ?1), the day of portfolio formation 0, and the ?rst ten days after
portfolio formation (1, 2, ..., 10). Shaded areas correspond to 95% con?dence intervals based
on Newey and West (1987) standard errors. Note that these returns, unlike Figure 2, are not
cumulative.
Figure 3 about here
Two results stand out. First, asset managers tend to be trend-followers in that they exert
buying (selling) pressure in currencies that recently appreciated (depreciated). Conversely,
private clients tend to trade against the trend, that is, they react upon past returns in a con-
trarian fashion. The pattern for hedge funds and corporates is less clear. Second, formation
day returns (day 0) are signi?cantly di?erent from zero for all four customer groups. How-
ever, hedge funds (positive) and private clients (negative) have the largest contemporaneous
returns in absolute value, indicating that their trading either heavily drives exchange rates
or is heavily triggered by returns (e.g., via stop-loss and stop-buy orders). The latter expla-
nation seems more reasonable especially for private clients who do not trade large enough
volumes to move prices in FX markets.
Overall, these ?ndings suggest that customer groups trading positions at least partly
o?set each other, as asset managers and private clients clearly di?er in terms of their trend-
following behavior. This ?nding is di?erent from equity markets where Kaniel, Saar, and
Titman (2008) ?nd that individual investors also tend to be contrarian traders but that
they experience subsequent positive returns, presumably due to implicitly providing liquidity
to institutional investors. In our data, we ?nd a similar contrarian behavior of individual
investors, but this trading behavior does not yield positive returns on average.
Flow correlations over longer horizons. Given these ?ndings, we next look at the corre-
lation of customer groups’ ?ows directly. While there is little contemporaneous correlation in
?ows, as noted above (see Table A.2 in the Internet Appendix), it is nevertheless interesting
21
to look at ?ows over longer horizons to ?nd out if customer groups tend to trade in the same
or in opposite directions. For a risk sharing explanation to make sense, we would expect to
see negative ?ow correlations between customer groups at some horizons.
Figure 4 plots contemporaneous correlations between ?ows of all four customer groups for
horizons of one to 60 days (using overlapping observations) where the shaded areas correspond
to 95% bootstrap con?dence intervals. For the two ?nancial customer groups, there is a small
and short-term negative ?ow correlation which turns positive after three days. Hence, asset
managers and hedge funds tend to trade in opposite directions over very short horizons but
in the same direction over the longer run. Moreover, all correlations between ?nancial and
non-?nancial customers are signi?cantly negative at all horizons while there is no signi?cant
correlation between ?ows of the non-?nancial customer groups. These results are generally
in line with a risk sharing story where ?nancial players trade in the opposite direction of
non-?nancial market participants. This ?nding is interesting because the perception in the
literature is that risk sharing takes place in the inter-dealer market (see, e.g., Lyons, 1997)
where dealers quickly lay o? their accumulated inventory from customer orders. However,
the high concentration in today’s FX market implies that large dealers can match customer
trades to a large extent internally, allowing them to manage their inventory more e?ciently.
Given the negative correlation of ?ows we observe in the data, there clearly seems to be scope
for such warehousing of inventory risk (also see King, Osler, and Rime, 2012, on this topic).
Figure 4 about here
Drivers of ?ows. As a natural next step we seek to provide a better understanding of the
drivers of end-user order ?ows and shed light on the source of the negative ?ow correlations
discussed above. First, we examine whether the ?ows of some customer groups systematically
lead the ?ows of other groups. Second, we study whether customers’ ?ows di?er in their
response to lagged asset returns in other key asset classes. In this context we are interested
in the possible e?ects of portfolio re-balancing on the end-user demand for currencies (Hau
and Rey, 2004). To investigate this, we run panel regressions of order ?ows on lagged ?ows
and further explanatory variables, such as interest rate di?erentials (i
t
?i
t
), lagged exchange
22
rate changes over one and 20 days (?s
t
, ?s
t?1;t?20
), lagged stock returns (r
eq
t
, r
eq
t?1;t?20
), and
lagged bond returns (r
b
t
, r
b
t?1;t?20
)
OF
c
j,t+1
= ? + ?
AM
OF
AM
j,t
+ ?
HF
OF
HF
j,t
+ ?
CC
OF
CC
j,t
+ ?
PC
OF
PC
j,t
+?
1
(i
j,t
? i
t
) + ?
2
?s
j,t
+ ?
3
?s
j,t?1;t?20
(5)
+?
4
r
eq
j,t
+ ?
5
r
eq
j,t?1;t?20
+ ?
6
r
b
j,t
+ ?
7
r
b
j,t?1;t?20
+ ?
j,t+1
,
where c denotes one of the four customer groups, j denotes currencies/countries, and ?
j,t+1
=
e
t+1
+ u
j
+
j,t+1
includes both cross-sectional and time ?xed-e?ects. Standard errors are
clustered by currency pair. We use benchmark 10-year government bonds and country equity
indices from Datastream for bond and stock returns. The frequency is daily.
Results from these regressions are shown in Table V. For each customer group we report
one speci?cation which only includes lagged ?ows and one which additionally includes interest
rate di?erentials and lagged returns.
23
Looking ?rst at the speci?cations which only include
lagged ?ows, we ?nd that the ?ows of asset managers are signi?cantly related to the ?ows
by the other groups. These results (akin to simple Granger causality tests) indicate again
that asset managers trade in the opposite direction of non-?nancial customers. Flows by
hedge funds, on the other hand, do not load signi?cantly on lagged ?ows of any group, which
shows that asset managers and hedge funds show a quite di?erent behavior. Corporate
?ows are positively driven by own lagged ?ows and those of private clients, whereas ?ows by
private clients are signi?cantly negatively related to lagged hedge funds’ ?ows and signi?cantly
positively autocorrelated. In sum, there is a wealth of interrelationships between customer
?ows and their lags although it seems overambitious to interpret them in any structural way.
Table V about here
When including lagged returns as additional regressors, we ?nd that asset managers trade
against the interest rate di?erential, whereas corporate customers trade with the interest
23
Using more than one lag of ?ows in the regressions generally yields insigni?cant coe?cient estimates so
we restrict the regressions to include one lag of ?ows.
23
rate di?erential. Surprisingly, ?ows by hedge funds (and private clients) are not a?ected
by the interest di?erential suggesting that, on average, carry trading is not a dominant
driver of their ?ows over our sample.
24
Results for lagged exchange rates indicate that
asset managers are trend-followers (positive feedback traders), whereas private clients can
be described as contrarians (negative feedback traders). Asset managers’ ?ows also react
signi?cantly positively to lagged equity returns, whereas private clients’ ?ows are positively
driven by lagged bond returns. Hence, investors tend to increase their position in a currency
(against the USD) when the country’s stock market return has been high (asset managers)
or when government bond prices went up (private clients). These results do not suggest that
order ?ows are driven by portfolio rebalancing in the sense that investors sell a currency in
response to rising equity or bond prices in the country (see, e.g. the mechanism described in
Hau and Rey, 2004). However, the results strongly support the notion that ?ows of di?erent
groups are to some extent driven by the returns of other asset classes, although the factors
that in?uence ?ows clearly di?er across end-user groups.
Finally, it seems worthwhile mentioning that the estimated overall constant in the panel
regression is signi?cantly negative for hedge funds and corporates but signi?cantly positive
for private clients. Hence, hedge funds and corporates have been net sellers of the U.S. dollar,
whereas private clients have been net buyers of U.S. dollar. Given the large current account
de?cit of the U.S. over the last decade, the negative coe?cient for corporate clients is not
surprising. Also, the strong in?ow of foreign savings into U.S. capital markets over the sample
period makes sense of the positive coe?cient for private clients.
24
While this result may be surprising, it is worth bearing in mind that it relates to the aggregate hedge
funds community and on average over the full sample. It is therefore entirely possible, or even likely, that
there is variation across hedge funds and across time: For example, it may well be the case that some hedge
funds follow carry trade strategies and some follow uncovered interest parity (anti-carry trade) strategies, or
that for a particular hedge fund carry trades were implemented for the ?rst part of the sample and deleveraged
during the second part of the sample which is characterized by the recent global crisis. Thus, our result is
not intended to detect the individual behavior of hedge funds.
24
C. Di?erences in Risk Exposures
Finally, we investigate if di?erences in risk exposures can account for BMS return patterns
across FX end-users. A risk channel could explain the observed BMS excess returns if asset
managers and hedge funds tilt their portfolios towards risky currencies and earn a risk pre-
mium whereas corporate and private clients tilt their portfolios towards safe currencies and,
hence, earn low or even negative returns.
Since there are many possible sources of systematic risk that might be relevant in our case,
we consider an augmented version of the Fung and Hsieh (2002, 2004) multi-factor model as
the basis for these risk adjustments. The Fung-Hsieh model has served as the workhorse for
understanding risk exposures in the hedge funds literature (see, e.g. Patton and Ramadorai,
2013). The model relies on various U.S. equity-market and bond-market factors and also
includes the returns to trend-following strategies to capture exposure to non-linear option-
like payo?s that are quite typical of hedge funds. The trend-following factors are constructed
from portfolios of lookback straddles in various asset classes. We modify the model to make
it amenable to an analysis focused on the FX market and to allow for conditional exposures
(e.g. Ferson and Schadt, 1996; Patton and Ramadorai, 2013). The regression which serves as
the basis of these tests takes the following form
rx
p;t
= ? +
K
k=1
?
k
F
k;t
+
J
j=1
?
j
r
m;t
· z
j;t?1
+
t
. (6)
The set of factors F
t
includes the excess return on the U.S. equity market (r
m
), the change
in the yield spread of U.S. long-term bonds (?TS), and changes in credit spreads (?DF).
It further includes returns on portfolios of lookback straddles for FX futures and interest
rate futures, denoted by PTFS
FX
and PTFS
IR
respectively. We augment this sub-set of
factors from Fung and Hsieh (2004) by additional factors that are intended to capture FX-
related risk. We include the Dollar risk factor (DOL) and the carry factor (HML
FX
) by
Lustig, Roussanov, and Verdelhan (2011) as well as a factor-mimicking portfolio of global FX
volatility (V OL
FX
) (Menkho?, Sarno, Schmeling, and Schrimpf, 2012a). Following Patton
25
and Ramadorai (2013), we also allow for conditional risk exposures by interacting the equity
market risk factor r
m;t
with lagged conditioning variables z
j;t?1
. We consider (a) changes in
the TED spread (Brunnermeier, Nagel, and Pedersen, 2009), (b) changes in the VIX (Whaley,
2000), and (c) the change in the 3-month T-Bill rate.
To keep the analysis tractable and to avoid over?tting, we perform model selection of
the space of risk factors. Ideally, we want to explore the same set of factors for each of the
customer segments to be able to compare the exposures across customers and learn about
di?erences which might explain the variation in BMS excess returns. However, as ?nancial
and non-?nancial customers are likely to be very di?erent, we focus on asset managers versus
hedge funds in the ?rst set of results and private clients versus corporates in the second
set of results. More speci?cally, we perform model selection over a two-equation seemingly
unrelated regression (SUR) for asset managers’ and hedge funds’ BMS returns, and a separate
model selection for a SUR for corporate and private clients.
Results for asset managers and hedge funds are shown in Table VI. Panel A shows results
for linear models, whereas Panel B allows for conditional market exposures. We report the
four best performing models with a maximum of three factors included in the regression.
The best linear model in Panel A picks global FX volatility (V OL
FX
) as the single factor.
Other model speci?cations which also perform well tend to incorporate the trend-following
factors as well as term spread and default spread changes. Interestingly, when comparing
asset managers’ and hedge funds’ exposures to these factors, we ?nd that the signs are always
opposite. While asset managers’ BMS returns load positively on FX volatility shocks, trend-
following factors, and changes in the default spread, hedge funds load negatively on these
factors. Hence, asset managers’ FX trading positions tend to perform well in periods of
market-wide stress and when there are large returns to trend-following (which happens to be
in volatile periods, when markets trend more). Hedge funds’ FX trading positions, however,
are adversely exposed to systematic risk and market distress. These results are quite striking
as they indicate that asset managers show a very di?erent FX trading behavior and exposure
26
to systematic risk than hedge funds.
25
Table VI about here
Allowing for conditional exposures by adding interaction terms of market returns (r
m
)
with lagged changes in TED spreads and the VIX (Table VI, Panel B) leaves the main
factors chosen largely unchanged but tends to improve the model ?t. The results reported
in Panel B corroborate the previous results. The equity market exposure of asset managers
tends to decrease when the lagged TED spread and VIX increase, and vice versa for hedge
funds exposures. This is further evidence that the trading by asset managers and hedge
funds is very di?erent and that their FX trading positions are exposed di?erently to market
stress. It should be noted, though, that the alphas of asset managers and hedge funds are
signi?cantly di?erent from zero and still quite large, that is, exposure to risk does not drive
the information in ?ows for excess returns to zero.
26
Table VII about here
We repeat the analysis above for corporate and private clients’ BMS portfolios as well,
and results are shown in Table VII. However, as might be expected, risk exposures do not
matter as much for non-?nancial customers. Still, we ?nd a negative equity market exposure
for both groups (Panel A), which increases (decreases) following increases in the TED spread
for private (corporate) clients. Moreover, there is some evidence that the private clients’
BMS portfolio has positive exposure to changes in credit spreads.
IV. Additional Tests and Robustness
We provide extensive robustness checks to all our main results. Below, we ?rst examine the
e?ect of transaction costs and then turn to a brief discussion of various other robustness tests
25
Additional evidence is provided in the Internet Appendix. Table A.17 summarizes exposures to equity
factors, Table A.18 considers FX factors, Table A.19 focuses on the Fung and Hsieh (2002) factors, and Table
A.20 reports results for the BMS portfolio based on total ?ows for completeness.
26
Table A.16 reports pricing errors for the cross-section of order ?ow portfolios. Speci?cally, we report the
Gibbons, Ross, and Shanken (1989) test for the null that the alphas are jointly equal to zero. Corroborating
the time-series regressions in Tables VI and VII, the test always rejects the null of zero alphas.
27
for which results can be found in a separate Internet Appendix to conserve space.
A. Transaction Costs
Our analysis above is intentionally quiet on questions of exploitability of order ?ow infor-
mation for trading strategies or the e?ects of transaction costs. This is because our data on
customer order ?ow are not available to participants in the broader market and thus cannot
form the basis for a trading strategy, except for UBS itself or for one of the other few large
dealers with access to similar customer ?ows. However, an interesting issue is whether owners
of this type of private information, that is, large FX dealer banks with a large concentration
of informed customers, could potentially employ this information by simply piggy-backing
the order ?ow of their customers.
27
To examine this question, we compute net excess returns for BMS portfolios by adjusting
for bid-ask spreads.
28
We investigate returns to strategies with varying portfolio re-balancing
frequencies to balance the e?ects of transaction costs and using the most recent information.
Figure 5 presents the results for re-balancing frequencies from 1 to 10 days. The dashed
lines give average excess returns (p.a.) and 95% con?dence intervals for excess returns before
transaction costs to show the e?ect of di?erent re-balancing periods. The solid line and
shaded area show average net excess returns (p.a.) and 95% con?dence intervals when taking
transaction costs into account.
Figure 5 about here
We ?nd that exploiting the information in ?ows should in practice be feasible for a dealer.
This holds for both sets of currencies T15 and T9. Average excess returns are signi?cantly
27
Obviously the data should be used in respect of clients’ con?dentiality and the speci?c compliance agree-
ments governing customers’ transactions.
28
The bid-ask spread data available is for quoted spreads and not e?ective spreads. It is well known that
quoted spreads are much higher than e?ective spreads (Lyons, 2001). We therefore follow earlier work, e.g.,
Goyal and Saretto (2009), and employ 50% of the quoted bid-ask spread as the actual spread. Even this
number seems conservative, though. First, banks with access to this kind of customer order ?ow data are
big dealers and pay very low spreads since they are key market makers. Second, Gilmore and Hayashi (2011)
?nd in a recent study that transaction costs due to bid-ask spreads are likely to be much lower than our 50%
rule. This ?nding was corroborated by our own conversations with UBS dealers.
28
di?erent from zero for all re-balancing horizons and economically attractive even for short
frequencies. These results clearly demonstrate the potential value of being able to observe
order ?ow by customers, especially the one by informed customers such as asset managers or
leveraged funds.
B. Further Robustness Checks
We check whether our results are robust to other sensible choices of standardizing ?ows. First,
we check whether standardizing ?ows over longer horizons of one and three years produce
similar results (see Tables A.7 and A.8 in the Internet Appendix). They do. Second, we
measure ?ows relative to total currency trading volume (obtained from the BIS FX triennial
surveys).
29
Table A.9 shows the results, which also indicate signi?cant predictability of
returns by order ?ows. Third, we standardize ?ows by additionally demeaning ?ows over
the rolling window (Table A.10). As a ?nal step, we form portfolios based on ?ows for all
currencies via the following procedure: We cross-sectionally standardize order ?ows for day
t (we subtract the cross-sectional mean and divide by the standard deviation), rescale these
standardized ?ows to sum to two in absolute value, and then use these as weights to form a
portfolio held from day t to t + 1 (Table A.11). Our results remain robust.
We also check if order ?ows forecast returns at longer horizons. To this end, we use
an exponential moving average to sum order ?ows into the past and then use these lower
frequency ?ows to build BMS portfolios which we rebalance every 2, 3, 4, 5, 10, 20, 60 trading
days. We report results for two di?erent decay parameters (0.25 and 0.75) in the exponential
moving average in Table A.12. We ?nd that predictability dies out fairly quickly and that
only asset managers’ ?ows have some predictive power over longer horizons of up to one
month (20 trading days).
To rule out that a simple liquidity story drives our predictability results, we also look at
a sub-sample of the four most liquid currency pairs in our sample: EUR/USD, JPY/USD,
29
We linearly interpolate between the data of the BIS survey to obtain a daily time-series of trading volumes
in USD for the nine developed currencies and then use the ratio of customer ?ows to total trading volumes
as our sorting variable.
29
GBP/USD, and CHF/USD. Table A.13 reports results for BMS portfolio returns and Figure
A.1 shows results for BMS returns in event time (similar to Figure 3 in the main text). We
?nd that our main results remain qualitatively unchanged.
We next explore whether a speci?c currency is driving the pro?tability of the order ?ow
portfolios. To investigate this issue, we rely on a cross-validation setting in which we form
portfolios as before but in each case delete one of the available currencies. For example, we
exclude the EUR/USD pair and compute BMS portfolio returns for the remaining 14 (total
order ?ows) or 8 currency pairs (disaggregated order ?ows). Table A.14 summarizes the
results from this exercise. We always ?nd the same general return pattern, that is, our main
?ndings do not depend strongly on any particular currency.
V. Conclusion
This paper empirically addresses three related questions to improve our understanding of
the ecology of the world’s largest ?nancial market, the FX market. First, given that the
FX market is fairly opaque with a large concentration of market making in the hands of
a few large intermediaries, how valuable is it for dealers to observe a large proportion of
the market’s order ?ow? Second, do FX end-users share risks among themselves, or is their
trading highly correlated and unloaded to the dealers and the inter-dealer market? Third,
how can we understand the trading behavior, trading styles and risk exposures of various key
players in FX markets, and how is this linked to risk sharing?
We ?nd that observing customer order ?ows in a dark market is highly valuable from
the dealer’s perspective. Currency excess returns to portfolios mimicking aggregate customer
order ?ows in real-time are about 10% p.a. and highly signi?cant. In addition, trading in FX
markets (as in other OTC markets) is anonymous, meaning that dealers know the identity
of their clients. Incorporating this feature into our setup, we ?nd excess returns as high as
15% p.a., that is, non-anonymity further increases the informational advantage of dealers.
The ?ows by asset managers have the strongest predictive power for exchange rates, likely
re?ecting the processing of fundamental information. Their ?ows have a permanent forecast
30
power, whereas ?ows originating from the other groups only predict transitory changes of
exchange rates. All this suggests that dealers have a strong incentive to gain large market
shares (besides other reasons such as economies of scale in the provision of trading infrastruc-
ture, for example) and to set up trading in a way that reveals end-users’ identities. These
?ndings about strong information asymmetries and incentives should be useful to inform
policy discussions on the appropriate framework for OTC markets.
We also show that the main segments of end-users di?er markedly in their trading strate-
gies and hedging demands. Asset managers, for instance, tend to be trend-followers, whereas
individual investors behave as contrarians. Hedge funds (on aggregate) do not seem to fall
in any of these two categories. Moreover, ?ows of di?erent end-user segments tend to be
negatively correlated over longer horizons. These ?ndings suggest that risk sharing also takes
place among end-users and not only via the inter-dealer market as suggested by previous FX
microstructure research.
Taken together, these results bring some light into one of the main dark ?nancial mar-
kets. Our ?ndings suggest that the FX market is populated by quite heterogeneous market
participants and that we gain valuable economic insights from observing their transactions
and learning about their di?erent predictive ability, trading motives, trading styles, and risk
exposures.
31
References
Akram, Q. Farooq, Dag?nn Rime, and Lucio Sarno, 2008, Arbitrage in the foreign exchange
market: Turning on the microscope, Journal of International Economics 76, 237–253.
Albuquerque, Rui, Eva de Francisco, and Luis Marques, 2008, Marketwide private informa-
tion in stocks: Forecasting currency returns, Journal of Finance 63, 2297–2343.
Ang, Andrew, and Joseph Chen, 2010, Yield curve predictors of foreign exchange returns,
Working Paper, Columbia University.
Barroso, Pedro, and Pedro Santa-Clara, 2011, Beyond the carry trade: Optimal currency
portfolios, Working Paper, NOVA School of Business and Economics.
Berger, D.W., A.P. Chaboud, S.V. Chernenko, E. Howorka, and J.H. Wright, 2008, Order ?ow
and exchange rate dynamics in electronic brokerage system data, Journal of International
Economics 75, 93–109.
Berger, D.W., A.P. Chaboud, and E. Hjalmarsson, 2009, What drives volatility persistence
in the foreign exchange market?, Journal of Financial Economics 94, 192–213.
BIS, 2010, Triennial central bank survey - report on global foreign exchange market activity
in 2010, Bank for International Settlements, Basel.
Brandt, Michael W., and Kenneth A. Kavajecz, 2004, Price discovery in the U.S. treasury
market: The impact of order?ow and liquidity on the yield curve, Journal of Finance 59,
2623–2654.
Brunnermeier, Markus, Stefan Nagel, and Lasse Pedersen, 2009, Carry trades and currency
crashes, NBER Macroeconomics Annual 2008 23, 313–347.
Burnside, Craig, Martin Eichenbaum, Isaac Kleshchelski, and Sergio Rebelo, 2011, Do Peso
problems explain the returns to the carry trade?, Review of Financial Studies 24, 853–891.
Carhart, Mark M., 1997, On persistence in mutual fund performance, Journal of Finance 52,
57–82.
Chinn, Menzie D., and Michael Moore, 2011, Order ?ow and the monetary model of exchange
rates: Evidence from a novel data set, Journal of Money, Credit and Banking 43, 1599–
1624.
Du?e, Darrell, 2012, Dark Markets: Asset Pricing and Information Transmission in Over-
the-Counter Markets (Princeton University Press).
32
Evans, Martin D., 2010, Order ?ows and the exchange rate disconnect puzzle, Journal of
International Economics 80, 58–71.
, and Richard K. Lyons, 2002a, Order ?ow and exchange rate dynamics, Journal of
Political Economy 110, 170–180.
, 2002b, Time-varying liquidity in foreign exchange, Journal of Monetary Economics
49, 1025–1051.
, 2005, Meese-Rogo? redux: Micro-based exchange rate forecasting, American Eco-
nomic Review, Papers and Proceedings 95, 405–414.
, 2007, Exchange rate fundamentals and order ?ow, Working Paper, Georgetown
University.
, 2008, How is macro news transmitted to exchange rates?, Journal of Financial
Economics 88, 26–50.
Fama, Eugene F., 1984, Forward and spot exchange rates, Journal of Monetary Economics
14, 319–338.
Fan, M., and Richard K. Lyons, 2003, Customer trades and extreme events in foreign ex-
change, in P. Mizen, ed.: Monetary History, Exchange Rates and Financial Markets: Es-
says in Honour of Charles Goodhart . pp. 160–179 (Edward Elgar: Northampton).
Farhi, Emmanuel, Samuel P. Fraiberger, Xavier Gabaix, Romain Ranciere, and Adrien
Verdelhan, 2009, Crash risk in currency markets, Working Paper, MIT Sloan.
Ferson, Wayne E., and R. Schadt, 1996, Measuring fund strategy and performance in changing
economic conditions, Journal of Finance 51, 425–462.
Froot, Kenneth A., and Tarum Ramadorai, 2005, Currency returns, intrinsic value, and
institutional-investor ?ows, Journal of Finance 60, 1535–1566.
Fung, William, and David Hsieh, 2001, The risk in hedge fund strategies: Theory and evidence
from trend followers, Review of Financial Studies 14, 313–341.
Fung, William, and David A. Hsieh, 2002, Asset-based style factors for hedge funds, Financial
Analysts Journal 58, 16–27.
, 2004, Hedge fund benchmarks: A risk based approach, Financial Analysts Journal
60, 65–80.
Gibbons, Michael R., Stephen A. Ross, and Jay Shanken, 1989, A test of the e?ciency of a
given portfolio, Econometrica 57, 1121–1152.
33
Gilmore, Stephen, and Fumio Hayashi, 2011, Emerging market currency excess returns,
American Economic Journal: Macroeconomics 3, 85–111.
Goyal, Amit, and Alessio Saretto, 2009, Cross-section of option returns and volatility, Journal
of Financial Economics 94, 310–326.
Hasbrouck, Joel, 1991a, Measuring the information content of stock trades, Journal of Fi-
nance 46, 179–207.
, 1991b, The summary informativeness of stock trades: An econometric analysis,
Review of Financial Studies 4, 571–595.
Hau, Harald, and Helene Rey, 2004, Can portfolio rebalancing explain the dynamics of eq-
uity returns, equity ?ows, and exchange rates?, American Economic Review, Papers and
Proceedings 94, 126–133.
Ito, Takatoshi, Richard K. Lyons, and Michael T. Melvin, 1998, Is there private information
in the FX market?, Journal of Finance 53, 1111–1130.
Kaniel, Ron, Gideon Saar, and Sheridan Titman, 2008, Individual investor trading and stock
returns, Journal of Finance 63, 273–310.
King, Michael R., Carol Osler, and Dag?nn Rime, 2012, Foreign exchange market structure,
players and evolution, in Jessica James, Ian Marsh, and Lucio Sarno, ed.: Handbook of
Exchange Rates . pp. 3–44 (Wiley).
Love, Ryan, and Richard Payne, 2008, Macroeconomic news, order ?ows, and exchange rates,
Journal of Financial and Quantitative Analysis 43, 467–488.
Lustig, Hanno, Nikolai Roussanov, and Adrien Verdelhan, 2011, Common risk factors in
currency markets, Review of Financial Studies 24, 3731–3777.
Lustig, Hanno, and Adrien Verdelhan, 2007, The cross section of foreign currency risk premia
and consumption growth risk, American Economic Review 97, 89–117.
Lyons, Richard K., 1997, A simultaneous trade model of the foreign exchange hot potato,
Journal of International Economics 42, 275–298.
, 2001, The Microstructure Approach to Exchange Rates (MIT Press: Cambridge).
Marsh, Ian W., and C. O’Rourke, 2005, Customer order ?ow and exchange rate movements:
Is there really information content?, Working Paper, Cass Business School.
Menkho?, Lukas, Lucio Sarno, Maik Schmeling, and Andreas Schrimpf, 2012a, Carry trades
and global foreign exchange volatility, Journal of Finance 67, 681–718.
34
, 2012b, Currency momentum strategies, Journal of Financial Economics 106, 660–
684.
Menkho?, Lukas, and Maik Schmeling, 2010, Whose trades convey information? Evidence
from a cross-section of traders, Journal of Financial Markets 13, 101–128.
Moore, Michael, and Richard Payne, 2011, On the existence of private information in FX
markets, Journal of Banking and Finance 35, 1250–1262.
Neely, Christopher, Paul Weller, and Joshua M. Ulrich, 2009, The adaptive markets hypoth-
esis: Evidence from the foreign exchange market, Journal of Financial and Quantitative
Analysis 44, 467–488.
Newey, Whitney K., and Kenneth D. West, 1987, A simple, positive semi-de?nite, het-
eroskedasticity and autocorrelation consistent covariance matrix, Econometrica 55, 703–
708.
Osler, Carol L., Alexander Mende, and Lukas Menkho?, 2011, Price discovery in foreign
exchange markets, Journal of International Money and Finance 30, 1696–1718.
Patton, Andrew J., and Tarun Ramadorai, 2013, On the high-frequency dynamics of hedge
fund risk exposures, Journal of Finance forthcoming.
Patton, Andrew J., and Allan Timmermann, 2010, Monotonicity in asset returns: New
tests with applications to the term structure, the CAPM, and portfolio sorts, Journal of
Financial Economics 98, 605–625.
Payne, Richard, 2003, Informed trade in spot foreign exchange markets: An empirical inves-
tigation, Journal of International Economics 61, 307–329.
Phylaktis, Kate, and L. Chen, 2010, Asymmetric information, price discovery and macroe-
conomic announcements in FX market: Do top trading banks know more?, International
Journal of Finance and Economics 15, 228–246.
Rime, Dag?nn, Lucio Sarno, and Elvira Sojli, 2010, Exchange rates, order ?ow and macroe-
conomic information, Journal of International Economics 80, 72–88.
Sager, Michael, and Mark P. Taylor, 2008, Commercially available order ?ow data and ex-
change rate movements: Caveat emptor, Journal of Money, Credit and Banking 40, 583–
625.
Whaley, R.E., 2000, The investor fear gauge, Journal of Portfolio Management 26, 12–17.
35
Table I. Descriptive Statistics for FX Customer Order Flows
This table shows descriptive statistics for total order ?ows for the 15 currencies in our sample.
Flows are measured in billions USD and all currencies are against the USD. A positive
(negative) ?ow means that there is net buying (selling) pressure for the respective currency.
The frequency is daily and the sample is from January 2001 to May 2011. Currencies included
are the Australian Dollar (AUD), Canadian Dollar (CAD), Swiss Franc (CHF), Euro (EUR),
Great Britain Pound (GBP), Japanese Yen (JPY), Norwegian Krone (NOK), New Zealand
Dollar (NZD), Swedish Krona (SEK), Brazilian Real (BRL), Hong Kong Dollar (HKD),
(South) Korean Won (KRW), Mexican Peso (MXN), Singapore Dollar (SGD), and South
African Rand (ZAR).
Mean Median Std Skew Kurt AC(1) p-val.
Panel A. Developed Markets
AUD -0.003 -0.001 0.197 -2.69 55.89 -0.01 (0.15)
CAD 0.007 0.003 0.169 1.98 43.02 0.00 (1.00)
CHF 0.020 0.012 0.324 0.11 74.88 0.02 (0.00)
EUR -0.063 -0.041 0.656 -3.31 79.53 0.03 (0.00)
GBP -0.001 -0.002 0.484 -5.82 270.70 0.01 (0.03)
JPY 0.027 0.019 0.412 1.88 55.87 0.03 (0.00)
NOK 0.003 0.000 0.067 0.60 49.20 0.08 (0.00)
NZD -0.002 0.000 0.070 -1.78 51.30 0.14 (0.00)
SEK 0.001 0.000 0.070 1.60 39.84 0.01 (0.04)
Panel B. Emerging Markets
BRL -0.004 0.000 0.068 -1.15 30.50 0.03 (0.00)
HKD 0.006 0.000 0.079 2.32 35.39 0.01 (0.02)
KRW -0.005 0.000 0.070 -0.15 59.45 0.05 (0.00)
MXN -0.002 0.000 0.049 -0.67 27.78 0.06 (0.00)
SGD 0.000 0.000 0.068 -4.39 110.15 0.06 (0.00)
ZAR 0.002 0.000 0.068 -0.91 36.38 0.16 (0.00)
36
Table II. Order Flow Portfolios: Excess Returns
This table shows average annualized portfolio excess returns for currency portfolios sorted on
lagged order ?ow. We standardize order ?ow over a rolling window of 60 trading days prior to
the order ?ow signal as outlined in the text. Column“Av”shows average excess returns across
all currencies, column “BMS” (bought minus sold) reports average excess returns for long-
short portfolios in currencies with the highest versus lowest order ?ow. Numbers in brackets
are t-statistics based on Newey-West standard errors whereas numbers in parentheses show
(annualized) Sharpe Ratios. Columns ‘MR’, ‘Up’, and ‘Down’ report p-values for tests of
return monotonicity. The frequency is daily and the sample is from January 2001 to May
2011. Panel A shows results for total order ?ows and all 15 markets (T15) as well as for total
order ?ows and the subsample of nine developed markets (T9). Panel B reports results for
order ?ows disaggregated by customer type (asset managers AM, hedge funds HF, corporate
clients CC, private clients PC).
Panel A. Total Order Flows
P
1
P
2
P
3
P
4
P
5
Av. BMS MR Up Down
T15 0.82 1.05 6.15 6.77 11.13 5.18 10.31 0.00 0.00 –
[0.29] [0.37] [2.23] [2.40] [4.04] [2.20] [4.05]
(0.09) (0.11) (0.71) (0.77) (1.21) (0.69) (1.26)
T9 0.34 2.24 8.21 12.76 5.89 12.43 0.00 0.00 –
[0.10] [0.74] [2.60] [4.17] [2.15] [4.68]
(0.03) (0.23) (0.80) (1.23) (0.66) (1.45)
Panel B. Disaggregated Order Flows
AM -1.13 3.75 6.30 14.31 15.43 0.00 0.00 –
[-0.35] [1.24] [2.04] [4.63] [5.72]
(-0.11) (0.38) (0.62) (1.38) (1.79)
HF -0.32 6.05 6.26 9.78 10.09 0.04 0.00 –
[-0.10] [2.04] [1.94] [3.02] [3.94]
(-0.03) (0.61) (0.59) (0.94) (1.20)
CO 6.90 5.27 7.02 2.61 -4.29 0.35 – 0.09
[2.15] [1.73] [2.16] [0.84] [-1.66]
(0.67) (0.53) (0.66) (0.26) (-0.51)
PC 12.71 6.69 2.90 -1.30 -14.01 0.00 – 0.00
[4.06] [2.18] [0.93] [-0.41] [-5.20]
(1.23) (0.67) (0.28) (-0.13) (-1.55)
37
Table III. Order Flow Portfolios: Marginal Forecast Performance for Longer Horizons
This table shows average excess returns (p.a.) for BMS portfolios sorted on lagged order ?ow
as in Table II. t-statistics based on Newey-West standard errors are reported in brackets.
However, we do not only sort on order ?ow of the previous day but also allow for longer
lags of up to nine days between order ?ow signals and portfolio formation. Portfolios are
rebalanced daily. T15 denotes portfolios sorts on total order ?ows and the sample of all 15
currencies, and T9 denotes portfolios sorts on total order ?ows and the sample of 9 developed
currencies; AM, HF, CC, and PC denote portfolios sorts on asset managers’, hedge funds’,
corporate clients’, and private clients’ order ?ows, respectively.
Lags between order ?ow signal and portfolio formation (days)
1 2 3 4 5 6 7 8 9 10
T15 10.31 24.63 10.22 -1.11 3.02 0.20 0.31 1.93 -2.32 -0.43
[4.05] [8.94] [4.38] [-0.44] [1.28] [0.09] [0.13] [0.84] [-0.95] [-0.19]
T9 12.43 24.27 7.44 -4.17 5.39 -1.55 2.28 1.33 -1.08 -1.75
[4.68] [8.73] [2.99] [-1.61] [2.00] [-0.61] [0.90] [0.51] [-0.42] [-0.71]
AM 15.43 24.86 8.27 -1.29 2.17 0.62 -0.20 3.37 2.26 -2.79
[5.72] [8.80] [3.03] [-0.47] [0.87] [0.23] [-0.07] [1.22] [0.82] [-0.97]
HF 10.09 28.22 2.05 -2.94 0.14 -6.19 2.84 -0.29 -4.66 -1.05
[3.94] [9.26] [0.79] [-1.15] [0.05] [-2.39] [1.12] [-0.10] [-1.77] [-0.40]
CC -4.29 -8.13 -1.47 2.25 -4.98 1.91 -0.01 1.40 -0.33 2.80
[-1.66] [-2.86] [-0.49] [0.88] [-1.93] [0.74] [0.00] [0.56] [-0.12] [1.08]
PC -14.01 -33.77 3.21 1.82 -3.29 -0.77 2.27 -1.35 0.65 2.10
[-5.20] [-10.80] [1.24] [0.67] [-1.15] [-0.27] [0.86] [-0.52] [0.24] [0.78]
38
Table IV. Panel Regressions of Currency Returns on Lagged Order Flow
This table reports results for panel regressions of currency excess returns (rx
t+1
) on lagged
customer order ?ow (OF
t
) and control variables (the interest rate di?erential i
j,t
?i
t
, lagged
excess returns over the previous day rx
t
, lagged excess returns over the prior 60 days
rx
t?1;t?60
). Order ?ow is measured in billion USD. T15 and T9 refer to total order ?ow
for all 15 currencies and the sample of developed market currencies, respectively. The re-
gressions in (v) and (vi) also include disaggregated order ?ow for asset managers, AM, hedge
funds, HF, corporate clients, CC, and private clients, CC). In each speci?cation, we show
results both for pooled regressions (pooling over all currency pairs) and for speci?cations
with currency pair- and time-?xed e?ects. t-statistics based on clustered standard errors (by
currency pair) are reported in brackets.
(i) (ii) (iii) (iv) (v) (vi)
const. 0.015 -0.010 0.020 -0.015 0.020 -0.015
[4.92] [-3.56] [6.81] [-2.77] [6.41] [-2.72]
OF
T15
t
0.025 0.023
[3.72] [3.49]
OF
T9
t
0.023 0.021
[3.42] [3.10]
OF
AM
t
0.043 0.038
[4.94] [4.29]
OF
HF
t
0.011 0.010
[2.12] [1.95]
OF
CC
t
-0.017 -0.014
[-2.63] [-2.23]
OF
PC
t
-0.028 -0.024
[-3.43] [-3.00]
i
j,t
? i
t
1.036 1.720 0.897 0.108 0.936 0.348
[7.74] [2.45] [2.59] [0.11] [2.75] [0.36]
rx
t
0.002 -0.010 0.001 -0.013 -0.006 -0.018
[0.27] [-1.21] [0.12] [-1.37] [0.57] [-1.86]
rx
t?1;t?60
0.000 -0.010 0.000 -0.015 0.000 -0.015
[-0.79] [-3.56] [-1.37] [-2.77] [-0.96] [-2.72]
Country dummies NO YES NO YES NO YES
Time dummies NO YES NO YES NO YES
R
2
0.002 0.024 0.001 0.030 0.006 0.045
obs 37,936 37,936 23,436 23,436 23,436 23,436
39
Table V. Drivers of Customer FX Order Flow: Panel Regressions
This table reports results for panel regressions of customer order ?ows (OF) on lagged cus-
tomer order ?ow (OF
t
for asset managers, AM, hedge funds, HF, corporate clients, CC, and
private clients, CC). The regressions also consider lagged returns on various asset classes
as additional regressors (the interest rate di?erential i
j,t
? i
t
, lagged exchange rate changes
over the previous day ?s
t
and over the prior 20 trading days ?s
t?1,t?20
, lagged country-level
equity returns over the previous trading day r
eq
t
and over the prior 20 trading days r
eq
t?1;t?20
),
and lagged country-level government bond returns r
b
t
(10-year maturity benchmark bonds).
t-statistics based on clustered standard errors (by currency pair) are reported in brackets and
we account for currency pair- and time-?xed e?ects.
Dependent variable: Customer order ?ows
OF
AM
t+1
OF
HF
t+1
OF
CC
t+1
OF
PC
t+1
OF
AM
t
0.035 0.033 0.013 0.012 -0.010 -0.009 -0.005 -0.003
[4.46] [4.22] [1.79] [1.73] [-1.13] [-1.08] [-0.61] [-0.38]
OF
HF
t
0.034 0.031 0.008 0.007 -0.009 -0.008 -0.037 -0.350
[2.75] [2.66] [0.57] [0.50] [-1.70] [-1.55] [-2.59] [-2.56]
OF
CC
t
-0.017 -0.016 0.000 0.000 0.035 0.034 -0.012 -0.013
[-2.58] [-2.53] [0.02] [0.05] [2.93] [2.88] [-1.47] [-1.62]
OF
PC
t
-0.026 -0.025 -0.005 -0.004 0.025 0.024 0.027 0.025
[-2.10] [-2.05] [-0.67] [-0.61] [2.47] [2.46] [2.21] [2.02]
i
j,t
? i
t
-0.150 0.102 0.413 0.185
[-2.05] [0.80] [2.51] [1.02]
?s
t
3.541 1.769 -1.312 -4.187
[4.97] [1.47] [-1.15] [-2.52]
?s
t?1,t?20
1.012 0.612 -0.741 -2.187
[1.97] [0.50] [-0.45] [-1.52]
r
eq
t
1.251 0.399 -1.164 -0.226
[2.56] [0.41] [-2.37] [-0.34]
r
eq
t?1;t?20
0.347 -0.113 0.205 -0.225
[1.44] [-0.52] [1.17] [-0.34]
r
b
t
-3.730 -5.170 -1.135 10.145
[-1.56] [-1.26] [-0.57] [2.68]
r
b
t?1;t?20
-0.019 0.278 0.626 1.151
[-0.03] [-0.55] [1.04] [2.03]
const. 0.008 -0.002 -0.078 -0.089 -0.320 -0.295 0.039 0.076
[0.71] [0.03] [-4.42] [-3.47] [-7.27] [-6.01] [4.77] [4.41]
Country dummies YES YES YES YES YES YES YES YES
Time dummies YES YES YES YES YES YES YES YES
R
2
0.013 0.015 0.011 0.011 0.029 0.030 0.015 0.018
obs 23,796 23,796 23,796 23,796 23,796 23,796 23,796 23,796
40
Table VI. Risk Exposures of Financial End-Users
This table reports regression results for the risk exposures of the BMS portfolios of ?nancial FX
market end-users, that is, asset managers and hedge funds. The methodological framework in Panel
A is a modi?ed linear Fung-Hsieh model with eight factors as outlined in the main text. Panel
B also accounts for conditional equity market exposures by including additional interaction terms.
The three conditioning variables are ?rst di?erences of the TED spread, the VIX and the 3-month
T-Bill rate. The Table shows results for four parsimonious model speci?cations where the factors
are selected according to the Schwarz criterion as outlined in the main text. Results for the other
factors are not reported. We further report the estimated intercept ?, the adjusted R
2
and the
BIC computed for the two-equation system (Sys-BIC). Below the regression coe?cients, t-statistics
based on HAC standard errors are reported (in brackets).
A. Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
PTFS
FX
2.35 -2.68
[2.65] [-2.51]
PTFS
IR
3.07 2.18 -1.33 -1.16
[4.03] [2.86] [-1.85] [-1.67]
?TS -2.03 0.38
[-2.06] [0.59]
?DF 3.15 3.65 -3.58 -3.67
[2.83] [2.87] [-2.61] [-2.69]
VOL
FX
0.07 0.06 -0.07 -0.05
[2.44] [2.13] [-2.50] [-2.09]
? 1.46 1.40 1.26 1.23 0.71 0.78 0.89 0.90
[5.32] [5.45] [5.68] [5.25] [3.10] [3.49] [4.01] [3.97]
¯
R
2
0.10 0.12 0.21 0.15 0.11 0.14 0.10 0.10
Sys-BIC 3.53 3.53 3.54 3.54 3.53 3.53 3.54 3.54
B. Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
r
m
·?VIX(t-1) -0.25 -0.26 -0.26 0.15 0.17 0.16
[-2.95] [-2.69] [-2.47] [2.30] [3.08] [3.32]
r
m
·?TED(t-1) -0.18 -0.19 -0.20 -0.20 0.32 0.37 0.33 0.34
[-2.44] [-2.82] [-3.06] [-2.48] [4.14] [5.74] [4.58] [4.63]
PTFS
FX
2.52 -2.54
[2.31] [-2.48]
PTFS
IR
2.11 -0.87
[2.79] [-1.35]
VOL
FX
0.05 0.06 -0.05 -0.05
[2.08] [2.15] [-2.10] [-2.29]
? 1.35 1.18 1.20 1.44 0.72 0.85 0.86 0.67
[5.57] [5.55] [5.38] [5.31] [3.50] [3.89] [4.05] [3.21]
¯
R
2
0.17 0.18 0.15 0.12 0.21 0.18 0.21 0.19
Sys-BIC 3.45 3.46 3.47 3.47 3.45 3.46 3.47 3.47
41
Table VII. Risk Exposures of Private and Corporate Clients
This table reports regression results for the risk exposures of the BMS portfolios computed from
the ?ows of corporate clients (CC) or private clients (PC). The methodological framework in Panel
A is a modi?ed linear Fung-Hsieh model with eight factors as outlined in the main text. Panel
B also accounts for conditional equity market exposures by including additional interaction terms.
The three conditioning variables are ?rst di?erences of the TED spread, the VIX and the 3-month
T-Bill rate. The Table shows results for four parsimonious model speci?cations where the factors
are selected according to the Schwarz criterion as outlined in the main text. Results for the other
factors are not reported. We further report the estimated intercept ?, the adjusted R
2
and the
BIC computed for the two-equation system (Sys-BIC). Below the regression coe?cients, t-statistics
based on HAC standard errors are reported (in brackets).
A. Corporate Clients Private Clients
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
r
m
-0.14 -0.11 -0.14 -0.07 -0.09 -0.10
[-2.32] [-1.63] [-1.94] [-1.56] [-2.27] [-2.43]
PTFS
IR
-1.83 -0.70
[-1.04] [-0.52]
?DF -3.43 -2.20 2.57 3.18
[-1.38] [-0.99] [2.81] [3.51]
? -0.30 -0.31 -0.37 -0.25 -1.16 -1.15 -1.19 -1.12
[-1.49] [-1.5] [-1.73] [-1.42] [-4.27] [-4.13] [-4.34] [-4.04]
¯
R
2
0.07 0.04 0.01 0.07 0.06 0.03 0.04 0.03
Sys-BIC 3.93 3.94 3.94 3.96 3.93 3.94 3.94 3.96
B. Corporate Clients Private Clients
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
r
m
-0.08 -0.10 -0.11 -0.13
[-1.30] [-1.50] [-2.48] [-3.05]
r
m
·? TED(t-1) 0.47 0.44 0.40 0.49 -0.20 -0.24 -0.31 -0.12
[2.95] [2.29] [2.1] [2.56] [-2.53] [-2.27] [-2.68] [-0.99]
PTFS
IR
-0.94 -1.38
[-0.8] [-1.45]
?DF 0.39 2.51
[0.25] [1.71]
? -0.46 -0.41 -0.37 -0.46 -1.18 -1.12 -1.06 -1.19
[-2.04] [-1.91] [-1.85] [-2.02] [-4.25] [-4.27] [-4.20] [-4.46]
¯
R
2
0.13 0.14 0.15 0.12 0.02 0.07 0.09 0.04
Sys-BIC 3.82 3.83 3.87 3.88 3.82 3.83 3.87 3.88
42
Figure 1. Cumulative Excess Returns on BMS Portfolios
01 02 03 04 05 06 07 08 09 10 11
-20
0
20
40
60
80
100
120
140
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
All countries
01 02 03 04 05 06 07 08 09 10 11
-20
0
20
40
60
80
100
120
140
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
Developed countries
01 02 03 04 05 06 07 08 09 10 11
0
20
40
60
80
100
120
140
160
180
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
Asset Managers
01 02 03 04 05 06 07 08 09 10 11
-20
0
20
40
60
80
100
120
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
Hedge Funds
01 02 03 04 05 06 07 08 09 10 11
-60
-50
-40
-30
-20
-10
0
10
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
Corporate Clients
01 02 03 04 05 06 07 08 09 10 11
-150
-100
-50
0
50
C
u
m
u
l
a
t
i
v
e
l
o
g
r
e
t
u
r
n
(
i
n
%
)
Private Clients
This ?gure shows cumulative log excess returns for a long-short portfolio based on total order
?ows and all countries (T15), total ?ows and developed markets (T9), asset manager ?ows
(AM), hedge fund ?ows (HF), corporate clients’ ?ows (CC), and private clients’ ?ows (PC).
The sample period is daily from January 2001 – May 2011.
43
Figure 2. Cumulative Post-Formation Exchange Rate Changes
This ?gure shows average cumulative spot exchange rate changes for BMS portfolios based
on total ?ows and disaggregated ?ows over the ?rst 30 days after portfolio formation. We
use daily data so that post-formation periods overlap. Shaded areas correspond to a 95%
con?dence interval obtained from a moving-block bootstrap with 1,000 repetitions.
44
Figure 3. BMS Excess Returns in Event Time
This ?gure shows BMS portfolio excess returns (solid lines) in event time, from ?ve days prior to portfolio
formation (t = ?5, the day of portfolio formation (t = 0), up to ten days after portfolio formation (t = 10).
BMS excess returns are annualized and in percent. The shaded areas correspond to 95% con?dence intervals
based on Newey/West standard errors. The frequency is daily and the sample is from January 2001 – May
2011.
45
Figure 4. Correlation of Customer Order Flows Over Longer Horizons
This ?gure shows average correlation coe?cients between customer order ?ows (left panel) for horizons of
1, 2, ..., 60 trading days. Average correlations between ?ows are based on the average correlation across all nine
currency pairs. A horizon of one day corresponds to (non-overlapping) daily observations, whereas correlations
for longer horizons are based on (overlapping) sums of daily observations. Shaded areas correspond to
bootstrapped 95% con?dence intervals based on a moving-block bootstrap with 1,000 repetitions. The sample
period is January 2001 – May 2011.
46
Figure 5. Rebalancing Frequency and Net Excess Returns
This ?gure shows average annualized excess returns for BMS portfolios based on total and
disaggregated order ?ows for di?erent rebalancing frequencies ranging from one to ten days.
The dotted lines show excess returns and a 95% con?dence interval based on Newey-West
standard errors before transaction costs whereas the solid line and shaded area show net
excess returns and a 95% con?dence interval based on Newey-West standard errors after
transaction costs.
47
Internet Appendix to accompany
Information Flows in Dark Markets:
Dissecting Customer Currency Trades
(Not for Publication)
48
Table A.1. Descriptive Statistics for Disaggregated Customer Order Flows
This table shows descriptive statistics for customer order ?ows which are available for the
nine major markets in our sample, that is, the Australian Dollar (AUD), Canadian Dollar
(CAD), Swiss Franc (CHF), Euro (EUR), Great Britain Pound (GBP), Japanese Yen (JPY),
Norwegian Krone (NOK), New Zealand Dollar (NZD), Swedish Krona (SEK). Flows are
measured in billions (in USD) and all currencies are against the USD. A positive (negative)
?ow means that there is net buying (selling) pressure in the foreign currency against the
USD. We report means, medians, standard deviations, skewness, kurtosis, and ?rst-order
autocorrelation coe?cients (AC(1)) for all four customer groups’ ?ows and, for comparison,
for total order ?ow in the nine currencies (T9). The ?rst number in each cell corresponds
to the cross-sectional mean across currencies (e.g., the mean across time-series standard
deviations of all nine currencies), whereas the second (parentheses) and third (brackets)
number correspond to the 5% and 95% percentile of the cross-sectional distribution (across
currencies), respectively. The frequency is daily and the sample is from January 2001 to May
2011.
Mean Median Std Skew Kurt AC(1)
Panel A. Asset Managers
-0.001 -0.001 0.272 -0.827 80.0 0.034
(-0.063) (-0.041) (0.067) (-5.820) (39.8) (-0.009)
[0.027] [0.019] [0.656] [1.978] [270.7] [0.140]
Panel B. Hedge Funds
0.002 0.001 0.205 -0.738 125.1 0.032
(-0.004) (-0.002) (0.054) (-7.977) (17.5) (-0.117)
[0.009] [0.005] [0.494] [4.810] [271.0] [0.128]
Panel C. Corporate Clients
-0.003 -0.001 0.171 -1.091 176.8 0.004
(-0.028) (-0.022) (0.036) (-23.273) (11.0) (-0.107)
[0.012] [0.010] [0.387] [12.143] [898.1] [0.091]
Panel D. Private Clients
-0.003 -0.003 0.068 -0.137 208.8 0.072
(-0.049) (-0.038) (0.009) (-17.616) (22.2) (-0.025)
[0.007] [0.006] [0.165] [10.063] [638.8] [0.192]
Panel E. Total Flows (T9)
0.003 0.002 0.091 -2.857 225.0 0.024
(-0.001) (0.000) (0.009) (-30.643) (16.2) (-0.106)
[0.014] [0.012] [0.265] [5.212] [1,385.8] [0.075]
49
Table A.2. Correlation Between Customer Groups’ Order Flows
This table reports correlation coe?cients between ?ows of customer groups for nine major
currencies and for a pooled sample over all currencies.
Correlation coe?cients
AM/HF AM/CC AM/PC HF/CC HF/PC CC/PC
EUR -0.04 -0.05 -0.05 -0.10 -0.20 -0.05
JPY 0.05 -0.05 -0.12 -0.02 -0.20 -0.05
GBP -0.03 0.02 -0.11 -0.02 -0.17 0.02
CHF 0.01 -0.09 -0.08 -0.07 -0.20 -0.09
AUD 0.00 0.03 -0.02 -0.06 -0.07 0.03
NZD 0.00 -0.05 -0.06 0.01 -0.03 -0.05
CAD -0.04 -0.08 -0.05 -0.01 -0.15 -0.08
SEK -0.03 -0.01 -0.01 -0.02 0.04 -0.01
NOK -0.02 -0.04 -0.01 -0.03 0.03 -0.04
Pooled -0.01 -0.04 -0.05 -0.03 -0.10 -0.04
50
Table A.3. Order Flow Portfolios: Di?erent Standardization Schemes and Sub-Samples
The setup of this table is identical to Table II, Panel A, in the main text but shows results for
rolling (Panel A), recursive (Panel B) ,and in-sample standardization (Panel C) of customer
order ?ow and for three di?erent sample periods as opposed to the rolling standardization
scheme employed in Table II.
Panel A. Rolling Window
P
1
P
2
P
3
P
4
P
5
Av. BMS MR Up
2001/01 – 2011/05 0.82 1.05 6.15 6.77 11.13 5.18 10.31 0.01 0.01
[0.29] [0.37] [2.23] [2.40] [4.04] [2.20] [4.05]
2001/01 – 2007/06 2.14 4.21 5.06 6.02 11.84 5.85 9.69 0.00 0.04
[0.71] [1.41] [1.79] [2.23] [4.14] [2.55] [3.45]
2007/07 – 2011/06 -1.18 -3.70 7.79 7.90 10.07 4.18 11.25 0.18 0.05
[-0.21] [-0.67] [1.44] [1.37] [1.87] [0.87] [2.36]
Panel B. Recursive Window
P
1
P
2
P
3
P
4
P
5
BMS MR Up
2001/01 – 2011/05 -0.42 2.35 5.68 6.73 11.74 12.16 0.00 0.00
[-0.14] [0.83] [2.13] [2.40] [4.19] [4.97]
2001/01 – 2007/06 0.56 5.82 3.86 7.62 11.68 11.12 0.19 0.00
[0.18] [2.07] [1.39] [2.83] [3.91] [4.00]
2007/07 – 2011/06 -1.89 -2.87 8.41 5.4 11.83 13.72 0.02 0.01
[-0.34] [-0.51] [1.62] [0.94] [2.20] [3.07]
Panel C. In-Sample
P
1
P
2
P
3
P
4
P
5
BMS MR Up
2001/01 – 2011/05 0 1.91 7.16 6.09 10.98 10.98 0.11 0.00
[0.00] [0.68] [2.58] [2.14] [4.00] [4.65]
2001/01 – 2007/06 1.86 4.47 6.54 6.4 10.36 8.5 0.01 0.07
[0.63] [1.52] [2.18] [2.31] [3.65] [3.26]
2007/07 – 2011/06 -2.79 -1.92 8.09 5.61 11.91 14.7 0.15 0.01
[-0.51] [-0.35] [1.53] [0.97] [2.21] [3.34]
51
Table A.4. Order Flow Portfolios: Exchange Rate Changes
This table shows average portfolio exchange rate changes for ?ve portfolios (P
1
, ..., P
5
) sorted
on lagged order ?ow. Sorting is done based on standardized total ?ows of all customers.
Column “Av” shows average excess returns across all currencies, column “BMS” (bought
minus sold) reports average excess returns to investing in P
5
and shorting P
1
. Panel B
reports the same information for spot exchange rate changes instead of excess returns. Flows
are standardized by their standard deviation (i) using a rolling window over the previous 60
trading days (Panel A), (ii) using a recursive scheme with 60 days initialization horizon (Panel
B), and (iii) their in-sample standard deviation. Average spot rate changes are annualized
(assuming 252 trading days per year). Numbers in brackets are t-statistics based on Newey-
West standard errors. The frequency is daily and the sample is from January 2001 – May
2011.
Panel A. Rolling Window
P
1
P
2
P
3
P
4
P
5
Av. BMS
Jan 2001 – May 2011 -1.28 -0.64 4.01 4.13 10.20 3.28 11.48
[-0.45] [-0.22] [1.47] [1.41] [3.72] [1.40] [4.57]
Jan 2001 – Jun 2007 -0.24 2.56 2.70 2.73 11.35 3.82 11.59
[-0.08] [0.86] [0.98] [0.91] [4.02] [1.68] [4.25]
Jul 2007 – May 2011 -2.85 -5.45 5.99 6.23 8.46 2.48 11.31
[-0.52] [-0.98] [1.11] [1.08] [1.57] [0.52] [2.37]
Panel B. Recursive Window
P
1
P
2
P
3
P
4
P
5
BMS
Jan 2001 – May 2011 -2.40 0.71 3.40 4.32 10.33 12.73
[-0.83] [0.25] [1.27] [1.50] [3.70] [5.17]
Jan 2001 – Jun 2007 -1.42 4.17 1.02 4.62 10.61 12.03
[-0.46] [1.46] [0.36] [1.60] [3.58] [4.28]
Jul 2007 – May 2011 -3.87 -4.49 6.97 3.87 9.90 13.77
[-0.69] [-0.80] [1.34] [0.67] [1.84] [3.08]
Panel C. In-Sample
P
1
P
2
P
3
P
4
P
5
BMS
Jan 2001 – May 2011 -1.17 -0.98 4.63 4.26 9.68 10.85
[-0.41] [-0.34] [1.68] [1.49] [3.51] [4.57]
Jan 2001 – Jun 2007 1.16 0.78 3.34 4.44 9.42 8.27
[0.40] [0.26] [1.13] [1.55] [3.30] [3.14]
Jul 2007 – May 2011 -4.67 -3.61 6.58 4.01 10.07 14.74
[-0.84] [-0.65] [1.25] [0.70] [1.87] [3.34]
52
Table A.5. Order Flow Portfolios: Customer Groups and Exchange Rate Changes
This table is similar to Panel B of Table II but here we report results for spot exchange rate
changes (and not excess returns).
P
1
P
2
P
3
P
4
Av. BMS
AM -1.65 2.97 5.62 13.86 15.52
[-0.51] [0.98] [1.81] [4.49] [5.75]
HF -0.90 5.32 5.70 9.25 10.15
[-0.29] [1.80] [1.77] [2.85] [3.96]
CO 6.30 4.47 6.37 2.26 -4.04
[1.97] [1.47] [1.96] [0.73] [-1.56]
PC 12.08 5.99 2.28 -1.84 -13.91
[3.85] [1.96] [0.73] [-0.57] [-5.16]
T9 -0.31 1.54 7.58 12.24 5.27 12.55
[-0.09] [0.51] [2.40] [4.00] [1.93] [4.72]
53
Table A.6. Correlation of Excess Returns
This table reports correlation coe?cients between excess returns of di?erent BMS portfolios
based on (i) lagged total ?ows of all 15 currency pairs (T15), (ii) lagged total ?ows of nine
developed countries (T9), (iii) lagged ?ows of asset managers (AM), (iv) lagged ?ows of hedge
funds, (v) lagged ?ows of corporate clients (CC), and lagged ?ows of private clients (PC). All
?ows are standardized by their lagged volatility over a 60-day rolling window. The frequency
is daily and the sample period is January 2001 to Ma 2011.
T15 T9 AM HF CC
T15 1.00
T9 0.63 1.00
AM 0.27 0.42 1.00
HF 0.30 0.42 0.06 1.00
CC 0.00 -0.06 -0.08 -0.13 1.00
PC -0.04 -0.02 -0.07 0.03 0.01
54
Table A.7. Order Flow Portfolios: Standardizing Flows (One Year)
This table shows average annualized portfolio excess returns for ?ve (or four) portfolios (P
1
,
..., P
5
) sorted on lagged order ?ow. Sorting is done based on standardized total ?ows and
standardized customer ?ows. Column“Av”shows average excess returns across all currencies,
column “BMS” (buying minus selling pressure) reports average excess returns to investing in
P
5
(or P
4
) and shorting P
1
. Flows are standardized by their standard deviation using a rolling
window over the previous 252 trading days (that is, roughly one year). We form 5 portfolios
for total ?ows of all 15 currency pairs (T15) and 4 portfolios for total ?ows of the nine
currencies for which we have disaggregated ?ows available (T9), for asset managers’ ?ows
(AM), hedge funds’ ?ows (HF), corporate clients’ ?ows (CC), and private clients’ ?ows (PC).
Numbers in brackets are t-statistics based on Newey-West standard errors. The frequency is
daily and the sample is from January 2001 to May 2011.
Panel A. Excess Returns
P
1
P
2
P
3
P
4
P
5
Av. BMS
T15 0.71 2.61 7.91 5.95 12.60 5.96 11.89
[0.23] [0.87] [2.77] [2.00] [4.29] [2.38] [4.59]
T9 0.66 2.56 8.44 13.81 6.37 13.15
[0.19] [0.82] [2.54] [4.24] [2.19] [4.81]
AM -1.15 3.79 6.66 16.10 17.25
[-0.33] [1.19] [2.02] [4.98] [6.23]
HF -0.63 7.35 5.62 10.75 11.38
[-0.19] [2.32] [1.65] [3.13] [4.23]
CC 7.81 6.09 8.55 1.10 -6.71
[2.36] [1.86] [2.48] [0.34] [-2.39]
PC 16.30 5.55 3.36 -1.48 -17.79
[4.96] [1.65] [0.99] [-0.45] [-6.49]
Panel B. Exchange Rate Changes
P
1
P
2
P
3
P
4
P
5
Av. BMS
T15 -1.42 0.66 6.26 3.18 11.20 3.97 12.62
[-0.47] [0.22] [2.20] [1.04] [3.80] [1.59] [4.79]
T9 -0.05 1.93 7.86 13.23 5.74 13.27
[-0.01] [0.62] [2.36] [4.06] [1.98] [4.85]
AM -1.68 3.03 5.98 15.65 17.33
[-0.49] [0.96] [1.82] [4.84] [6.25]
HF -1.27 6.68 5.10 10.13 11.41
[-0.39] [2.11] [1.49] [2.95] [4.24]
CC 7.18 5.31 7.92 0.76 -6.42
[2.16] [1.62] [2.30] [0.23] [-2.28]
PC 15.67 4.87 2.77 -2.05 -17.72
[4.77] [1.45] [0.82] [-0.63] [-6.47]
55
Table A.8. Order Flow Portfolios: Standardizing Flows (Three Years)
This table shows average annualized portfolio excess returns for ?ve (or four) portfolios (P
1
,
..., P
5
) sorted on lagged order ?ow. Sorting is done based on standardized total ?ows and
standardized customer ?ows. Column“Av”shows average excess returns across all currencies,
column “BMS” (buying minus selling pressure) reports average excess returns to investing in
P
5
(or P
4
) and shorting P
1
. Flows are standardized by their standard deviation using a
rolling window over the previous 750 trading days (that is, roughly three years). We form 5
portfolios for total ?ows of all 15 currency pairs (T15) and 4 portfolios for total ?ows of the
nine currencies for which we have disaggregated ?ows available (T9), for asset managers’ ?ows
(AM), hedge funds’ ?ows (HF), corporate clients’ ?ows (CC), and private clients’ ?ows (PC).
Numbers in brackets are t-statistics based on Newey-West standard errors. The frequency is
daily and the sample is from January 2001 to May 2011.
Panel A. Excess Returns
P
1
P
2
P
3
P
4
P
5
Av. BMS
T15 -2.21 -0.55 5.97 6.28 10.25 3.95 12.46
[-0.61] [-0.16] [1.75] [1.74] [3.04] [1.31] [4.30]
T9 -3.02 -1.86 7.29 13.28 3.92 16.30
[-0.75] [-0.50] [1.91] [3.40] [1.14] [5.01]
AM -4.89 1.15 3.20 15.39 20.28
[-1.22] [0.30] [0.84] [3.99] [6.30]
HF -2.44 2.31 3.97 9.92 12.36
[-0.64] [0.61] [0.99] [2.42] [3.85]
CC 6.48 2.16 7.25 -2.37 -8.85
[1.65] [0.55] [1.78] [-0.63] [-2.70]
PC 13.26 4.32 0.98 -6.05 -19.31
[3.47] [1.09] [0.24] [-1.54] [-5.78]
Panel B. Exchange Rate Changes
P
1
P
2
P
3
P
4
P
5
Av. BMS
T15 -3.89 -2.04 5.17 4.95 8.60 2.56 12.49
[-1.08] [-0.58] [1.52] [1.37] [2.55] [0.85] [4.28]
T9 -3.51 -2.09 7.05 12.92 3.59 16.43
[-0.86] [-0.56] [1.84] [3.31] [1.05] [5.04]
AM -5.16 0.75 2.88 15.14 20.30
[-1.29] [0.20] [0.76] [3.93] [6.31]
HF -2.93 2.06 3.82 9.50 12.43
[-0.77] [0.54] [0.95] [2.32] [3.87]
CC 6.15 1.71 6.93 -2.47 -8.62
[1.57] [0.44] [1.70] [-0.66] [-2.62]
PC 12.76 4.06 0.79 -6.40 -19.16
[3.34] [1.02] [0.19] [-1.63] [-5.73]
56
Table A.9. Order Flow Portfolios: Order Flows Scaled By Currency Trading Volume
This table is identical to Table II but here we do not standardize order ?ows by rolling
windows of the previous 60 trading days volatility but by total currency trading volume
(from the BIS FX triennial surveys for 2001, 2004, 2007, 2010). We linearly interpolate
between the turnover ?gures to obtain a daily measure of total trading volume for each of
the ?fteen currencies in our sample.
Panel A. Total order ?ows
P
1
P
2
P
3
P
4
P
5
Av. BMS MR Up Down
T15 -1.14 3.18 3.19 6.73 11.23 5.01 12.37 0.02 0.00 –
[-0.38] [1.19] [1.26] [2.37] [3.99] [2.00] [5.05]
(-0.12) (0.37) (0.39) (0.76) (1.19) (0.66) (1.53)
T9 -0.73 2.35 6.09 12.46 5.45 13.19 0.00 0.00 –
[-0.22] [0.83] [1.98] [4.01] [2.11] [5.09]
(-0.07) (0.25) (0.61) (1.17) (0.63) (1.55)
Panel B. Disaggregated order ?ows
AM -0.97 2.40 5.98 13.23 14.19 0.00 0.00 –
[-0.29] [0.85] [2.02] [4.19] [5.33]
(-0.09) (0.25) (0.61) (1.24) (1.66)
HF 0.14 4.05 5.94 9.43 9.29 0.00 0.04 –
[0.04] [1.40] [1.95] [2.87] [3.50]
(0.01) (0.42) (0.59) (0.87) (1.10)
CO 5.94 5.16 4.25 3.07 -2.88 0.01 – 0.65
[1.88] [1.72] [1.34] [1.01] [-1.15]
(0.58) (0.52) (0.41) (0.31) (-0.35)
PC 14.06 5.01 0.69 -1.14 -15.20 0.00 – 0.00
[4.52] [1.67] [0.22] [-0.36] [-5.68]
(1.35) (0.51) (0.07) (-0.11) (-1.71)
57
Table A.10. Order Flow Portfolios: Demeaning Flows
This table is similar to Table II but here we present results for total ?ows (T15 and T9) and
customer groups’ ?ows and we standardize order ?ows by subtracting the rolling mean and
dividing by the rolling standard deviation. The frequency is daily and the sample is from
January 2001 – May 2011.
P1 P2 P3 P4 P5 Av BMS
T15 -0.68 5.02 5.54 5.67 11.16 5.34 11.84
[-0.24] [1.86] [1.95] [2.07] [3.88] [2.26] [4.89]
T9 0.11 3.21 6.12 13.60 13.49
[0.03] [1.08] [1.91] [4.40] [5.25]
AM -0.28 4.64 3.45 14.50 14.77
[-0.08] [1.56] [1.07] [4.77] [5.44]
HF -0.04 6.29 5.26 9.95 9.99
[-0.01] [2.09] [1.69] [3.02] [3.88]
CO 7.91 5.41 4.85 3.75 -4.17
[2.48] [1.82] [1.54] [1.17] [-1.50]
PC 12.11 6.37 5.05 -2.09 -14.19
[3.77] [2.15] [1.60] [-0.63] [-5.00]
58
Table A.11. Order Flow-Weighted BMS Portfolios
This table shows returns for BMS portfolios based on total ?ows and customer ?ows but
here we employ portfolio weights directly based on lagged order ?ows. For each trading day
t, we cross-sectionally standardize order ?ows, rescale these standardized ?ows so that they
sum to two in absolute value, and then use these rescaled and standardized ?ows as portfolio
weights for day t to t + 1. Numbers in squared brackets are based on Newey/West standard
errors. The frequency is daily and the sample is from January 2001 – May 2011.
T15 T9 AM HF CO PC
Mean 12.70 11.12 13.94 6.61 -1.94 -9.77
t-Stat. 5.07 4.70 5.86 2.91 -0.89 -4.27
St. Dev. 8.38 7.60 7.72 7.46 7.04 7.12
Sharpe Ratio 1.52 1.46 1.81 0.89 -0.28 -1.37
Skewness 0.25 1.42 1.76 0.09 -0.12 -0.20
Kurtosis 19.63 33.82 28.54 11.75 12.87 11.05
Maximum 6.57 7.79 7.42 3.95 2.96 3.21
Minimum -5.33 -3.68 -2.46 -3.69 -3.95 -4.12
59
Table A.12. BMS Portfolios: Longer Horizons
This table shows average annualized BMS portfolio excess returns for longer forecast horizons
of 1, 2, ..., 5, 10, 20, 40, and 60 trading days. We use an exponential moving average (EMA)
with a decay parameter of 0.25 (Panel A) and 0.75 (Panel B) for lagged order ?ows to consider
longer histories of order ?ows for forecasting. BMS portfolios are based on 5 portfolios
for total ?ows of all 15 currency pairs (T15) and 4 portfolios for total ?ows of the nine
developed markets for which disaggregated ?ows are available (T9), for asset managers’ ?ows
(AM), hedge funds’ ?ows (HF), corporate clients’ ?ows (CC), and private clients’ ?ows (PC).
Numbers in brackets are t-statistics based on Newey-West standard errors. The frequency is
daily and the sample is from January 2001 to May 2011.
Rebalancing Frequency (Trading Days)
2 3 4 5 10 20 40 60
EMA with Decay Parameter 0.25
T15 17.20 12.31 9.82 7.68 4.57 4.15 2.83 3.60
[8.22] [5.68] [4.76] [3.66] [2.46] [2.40] [1.92] [1.68]
T9 20.74 14.25 11.27 7.98 5.19 4.84 2.14 -0.41
[7.99] [5.32] [4.46] [3.11] [2.99] [2.13] [1.76] [-0.15]
AM 20.28 16.30 11.99 6.63 5.72 5.12 2.99 1.56
[7.39] [5.91] [4.75] [2.61] [2.70] [2.08] [1.42] [0.59]
HF 15.78 7.60 3.78 2.31 -1.03 -0.72 1.50 3.14
[5.30] [2.67] [1.30] [0.81] [-0.36] [-0.26] [0.58] [1.09]
CC -4.74 -3.53 -2.85 -2.61 -0.61 -1.03 0.13 -3.51
[-1.84] [-1.32] [-1.12] [-1.03] [-0.25] [-0.42] [0.05] [-1.53]
PC -16.65 -8.58 -6.44 -2.42 -2.66 0.23 -1.78 -1.77
[-5.73] [-3.33] [-2.31] [-0.87] [-0.99] [0.08] [-0.58] [-0.71]
EMA with Decay Parameter 0.75
T15 18.97 16.21 9.48 10.67 5.78 4.27 0.63 3.09
[7.92] [7.46] [4.40] [5.36] [2.94] [1.94] [0.32] [1.58]
T9 23.05 18.09 12.03 10.14 7.83 4.90 1.14 3.87
[8.00] [6.76] [4.62] [4.12] [3.28] [2.22] [0.46] [1.54]
AM 23.75 20.92 16.19 11.74 7.27 5.98 2.14 0.82
[8.30] [7.68] [5.89] [4.32] [3.61] [2.54] [0.87] [0.31]
HF 20.66 11.34 6.83 8.15 6.84 1.46 1.37 3.71
[7.58] [3.73] [2.51] [2.91] [2.67] [0.55] [0.56] [1.41]
CC -3.61 -3.47 -1.54 -4.52 -1.01 -2.20 -2.67 -1.76
[-1.31] [-1.29] [-0.62] [-1.70] [-0.42] [-0.96] [-1.10] [-0.76]
PC -24.34 -14.51 -9.43 -7.12 -3.94 1.10 -0.78 0.72
[-7.96] [-5.25] [-3.51] [-2.61] [-1.62] [0.45] [-0.29] [0.29]
60
Table A.13. Order Flow Portfolios: Four Liquid Currencies
This table is similar to Table II in the main text but here we only include EUR/USD,
JPY/USD, GBP/USD, and CHF/USD in our sample of currencies and only form two port-
folios. T4 denotes portfolios based on total order ?ows.
Excess Returns Exchange Rate Changes
P
1
P
2
Av. BMS P
1
P
2
Av. BMS
T4 1.55 5.48 3.51 3.93 1.95 6.20 4.07 4.25
[0.56] [2.00] [1.40] [1.70] [0.70] [2.24] [1.62] [1.81]
AM -1.31 8.34 9.65 -0.79 8.94 9.73
[-0.46] [3.11] [4.23] [-0.28] [3.34] [4.27]
HF 1.68 5.35 3.67 2.18 5.97 3.79
[0.61] [1.93] [1.55] [0.79] [2.15] [1.60]
CC 5.16 1.90 -3.25 5.61 2.57 -3.05
[1.83] [0.70] [-1.41] [2.00] [0.95] [-1.32]
PC 7.59 -0.56 -8.15 8.14 0.00 -8.14
[2.75] [-0.20] [-3.47] [2.95] [0.00] [-3.47]
61
Table A.14. Order Flow Portfolios: Sensitivity to Individual Currencies
This table show average annualized excess returns to BMS portfolios based on total ?ows,
asset managers (AM) ?ows, hedge fund (HF) ?ows, corporate clients (CC) ?ows, and private
clients (PC) ?ows for a cross-validation setting in which we discard one of the available
currencies in our sample. We do this for each available currency and the ?rst column indicates
which currency is left out when computing returns to the BMS portfolio. Hence, BMS returns
for total ?ows are based on 14 currencies and BMS returns for customer ?ows are based on
8 currencies instead of 15 and 9 currencies, respectively. t-statistics in brackets are based on
Newey/West standard errors.
Total ?ows AM HF CC PC
rx t rx t rx t rx t rx t
EUR 10.79 [4.12] 15.40 [5.35] 8.86 [3.10] -2.15 [-0.78] -14.62 [-5.23]
JPY 8.26 [3.34] 9.62 [3.95] 11.30 [4.73] -3.13 [-1.33] -9.12 [-3.87]
GBP 10.18 [3.99] 15.29 [5.65] 8.95 [3.35] -3.04 [-1.17] -13.65 [-4.93]
CHF 10.29 [4.09] 13.58 [5.03] 11.49 [4.51] -2.08 [-0.79] -16.31 [-5.90]
AUD 10.39 [4.18] 10.39 [4.03] 9.78 [3.83] -5.58 [-2.19] -10.77 [-4.26]
NZD 9.35 [3.67] 14.75 [5.52] 8.16 [3.15] -2.22 [-0.87] -11.93 [-4.45]
CAD 11.98 [4.70] 14.70 [5.34] 9.65 [3.81] -4.07 [-1.61] -10.98 [-4.12]
SEK 9.33 [3.70] 15.51 [5.76] 6.32 [2.28] -4.46 [-1.76] -14.91 [-5.45]
NOK 10.04 [3.91] 16.52 [6.00] 9.38 [3.72] -3.37 [-1.24] -16.20 [-5.84]
MXN 10.10 [4.06]
BRL 11.22 [4.56]
ZAR 6.71 [2.79]
KRW 11.75 [4.78]
SGD 10.84 [4.27]
HKD 11.68 [4.75]
62
T
a
b
l
e
A
.
1
5
.
O
r
d
e
r
?
o
w
p
o
r
t
f
o
l
i
o
r
e
t
u
r
n
s
a
r
o
u
n
d
c
u
s
t
o
m
e
r
t
r
a
d
i
n
g
T
h
i
s
t
a
b
l
e
s
h
o
w
s
a
v
e
r
a
g
e
a
n
n
u
a
l
i
z
e
d
p
o
r
t
f
o
l
i
o
e
x
c
e
s
s
r
e
t
u
r
n
s
f
o
r
c
u
r
r
e
n
c
y
p
o
r
t
f
o
l
i
o
s
s
o
r
t
e
d
o
n
l
a
g
g
e
d
d
i
s
a
g
g
r
e
g
a
t
e
d
c
u
s
t
o
m
e
r
o
r
d
e
r
?
o
w
i
n
e
v
e
n
t
t
i
m
e
a
r
o
u
n
d
p
o
r
t
f
o
l
i
o
f
o
r
m
a
t
i
o
n
(
t
=
?
5
,
?
4
,
,
0
,
1
,
2
,
,
1
0
d
a
y
s
w
h
e
r
e
t
=
0
i
s
t
h
e
d
a
y
o
f
p
o
r
t
f
o
l
i
o
f
o
r
m
a
t
i
o
n
)
f
o
r
t
h
e
h
i
g
h
(
P
4
)
a
n
d
l
o
w
(
P
1
)
p
o
r
t
f
o
l
i
o
s
.
E
v
e
n
t
t
i
m
e
(
d
a
y
)
?
5
?
4
?
3
?
2
?
1
0
1
2
3
4
5
6
7
8
9
1
0
A
M
P
4
2
.
4
1
5
.
6
8
6
.
9
0
9
.
6
1
1
1
.
3
7
1
3
.
6
0
1
4
.
3
1
1
8
.
6
7
7
.
7
5
5
.
1
2
6
.
6
2
4
.
5
2
5
.
6
5
9
.
2
0
7
.
2
5
3
.
2
2
[
0
.
7
4
]
[
1
.
7
8
]
[
2
.
0
2
]
[
3
.
1
0
]
[
3
.
6
2
]
[
4
.
2
7
]
[
4
.
6
5
]
[
5
.
9
0
]
[
2
.
5
0
]
[
1
.
5
1
]
[
2
.
1
5
]
[
1
.
4
3
]
[
1
.
7
4
]
[
2
.
9
5
]
[
2
.
2
3
]
[
1
.
0
0
]
P
1
7
.
3
1
4
.
4
7
3
.
6
1
3
.
9
9
1
.
4
3
-
1
.
2
0
-
1
.
1
3
-
6
.
1
8
-
0
.
5
2
6
.
4
1
4
.
4
4
3
.
9
0
5
.
8
6
5
.
8
2
4
.
9
9
6
.
0
1
[
2
.
3
5
]
[
1
.
4
7
]
[
1
.
1
2
]
[
1
.
2
8
]
[
0
.
4
8
]
[
-
0
.
3
7
]
[
-
0
.
3
4
]
[
-
1
.
8
9
]
[
-
0
.
1
6
]
[
2
.
1
2
]
[
1
.
4
7
]
[
1
.
2
0
]
[
1
.
8
4
]
[
1
.
8
1
]
[
1
.
5
6
]
[
1
.
8
8
]
H
F
P
4
0
.
7
4
1
.
7
8
2
.
0
2
3
.
1
0
3
.
6
2
4
.
2
7
4
.
6
5
5
.
9
0
2
.
5
0
1
.
5
1
2
.
1
5
1
.
4
3
1
.
7
4
2
.
9
5
2
.
2
3
1
.
0
0
[
1
0
.
1
0
]
[
7
.
4
0
]
[
5
.
4
8
]
[
7
.
2
7
]
[
8
.
6
0
]
[
1
7
.
9
0
]
[
9
.
7
8
]
[
1
9
.
6
7
]
[
6
.
6
7
]
[
4
.
1
5
]
[
4
.
5
3
]
[
1
.
9
2
]
[
7
.
0
4
]
[
5
.
6
0
]
[
2
.
9
6
]
[
4
.
9
0
]
P
1
5
.
3
0
3
.
0
3
6
.
2
7
-
0
.
7
4
0
.
2
0
-
9
.
1
1
-
0
.
3
2
-
8
.
5
5
4
.
6
1
7
.
0
8
4
.
3
9
8
.
1
1
4
.
2
0
5
.
8
9
7
.
6
1
5
.
9
5
[
1
.
7
4
]
[
0
.
9
7
]
[
2
.
0
7
]
[
-
0
.
2
4
]
[
0
.
0
6
]
[
-
2
.
7
7
]
[
-
0
.
1
0
]
[
-
2
.
4
7
]
[
1
.
5
2
]
[
2
.
3
3
]
[
1
.
5
0
]
[
2
.
6
2
]
[
1
.
4
1
]
[
1
.
9
3
]
[
2
.
5
1
]
[
1
.
9
6
]
C
C
P
4
4
.
8
7
3
.
7
3
2
.
0
6
6
.
9
0
1
.
5
1
0
.
4
9
2
.
6
1
2
.
8
0
4
.
3
8
5
.
8
9
3
.
4
0
8
.
0
9
5
.
6
0
6
.
5
7
5
.
3
6
7
.
6
5
[
1
.
6
2
]
[
1
.
2
4
]
[
0
.
6
8
]
[
2
.
1
9
]
[
0
.
4
9
]
[
0
.
1
6
]
[
0
.
8
4
]
[
0
.
9
1
]
[
1
.
3
5
]
[
1
.
9
4
]
[
1
.
0
3
]
[
2
.
6
8
]
[
1
.
8
3
]
[
2
.
1
2
]
[
1
.
6
9
]
[
2
.
5
2
]
P
1
5
.
6
7
7
.
1
3
7
.
2
7
4
.
5
3
7
.
2
7
8
.
5
2
6
.
9
0
1
0
.
9
3
5
.
8
4
3
.
6
4
8
.
3
8
6
.
1
8
5
.
6
1
5
.
1
6
5
.
6
9
4
.
8
6
[
1
.
8
3
]
[
2
.
2
7
]
[
2
.
3
6
]
[
1
.
4
7
]
[
2
.
2
4
]
[
2
.
4
9
]
[
2
.
1
5
]
[
3
.
4
1
]
[
1
.
9
1
]
[
1
.
1
5
]
[
2
.
7
7
]
[
2
.
0
1
]
[
1
.
7
8
]
[
1
.
6
8
]
[
1
.
8
1
]
[
1
.
5
8
]
P
C
P
4
2
.
9
3
3
.
1
6
2
.
7
5
1
.
7
6
-
3
.
6
8
-
9
.
9
3
-
1
.
3
0
-
1
0
.
6
6
6
.
8
9
5
.
8
2
4
.
3
4
6
.
7
7
5
.
4
9
5
.
2
2
4
.
9
4
6
.
3
1
[
0
.
9
7
]
[
1
.
0
2
]
[
0
.
8
7
]
[
0
.
5
7
]
[
-
1
.
1
8
]
[
-
3
.
0
0
]
[
-
0
.
4
0
]
[
-
2
.
9
9
]
[
2
.
1
4
]
[
1
.
9
2
]
[
1
.
3
3
]
[
2
.
0
6
]
[
1
.
7
5
]
[
1
.
6
2
]
[
1
.
5
5
]
[
2
.
0
5
]
P
1
7
.
3
1
8
.
1
1
8
.
7
7
9
.
1
9
1
3
.
8
5
1
8
.
2
4
1
2
.
7
1
2
3
.
1
1
3
.
6
7
4
.
0
0
7
.
6
3
7
.
5
4
3
.
2
2
6
.
5
7
4
.
2
9
4
.
2
1
[
2
.
2
2
]
[
2
.
4
5
]
[
2
.
7
0
]
[
2
.
8
6
]
[
4
.
3
7
]
[
5
.
6
1
]
[
4
.
0
5
]
[
7
.
3
1
]
[
1
.
2
0
]
[
1
.
1
8
]
[
2
.
3
6
]
[
2
.
3
5
]
[
0
.
9
8
]
[
2
.
1
2
]
[
1
.
3
6
]
[
1
.
2
6
]
63
Table A.16. Pricing Error Statistics For The Cross-Section
This table reports pricing error statistics based on estimating the models of Table VI for the broader
cross-section of the order ?ow mimicking portfolios of ?nancial end-users. GRS is the test statistic
by Gibbons, Ross, and Shanken (1989). We compute the joint test for zero alphas for the entire
cross section of eight portfolios of Asset Managers and Hedge Funds. We also compute the GRS-
statistic separately for the portfolios (P1-P4) of each group. Model speci?cations (i) - to (iv) follow
the setup of table VI.
GRS p-val. GRS AM p-val. GRS HF p-val.
A. Linear Exposures
(i) 6.87 0.00 10.37 0.00 3.79 0.01
(ii) 6.49 0.00 9.51 0.00 4.00 0.00
(iii) 7.61 0.00 10.77 0.00 6.69 0.00
(iv) 7.14 0.00 9.51 0.00 6.75 0.00
B. Conditional Exposures
(i) 6.19 0.00 9.46 0.00 4.14 0.00
(ii) 6.51 0.00 9.04 0.00 6.31 0.00
(iii) 6.53 0.00 9.34 0.00 5.81 0.00
(iv) 6.58 0.00 10.24 0.00 3.79 0.01
64
Table A.17. Risk Exposures: Equity Factors
This table reports regression results for the risk exposures of the BMS portfolios of ?nancial FX
market end-users, that is, asset managers and hedge funds. The risk factors include the excess return
on the market portfolio (r
m
), and the Fama-French size (SMB) and value (HML) factors. UMD
denotes the return on Carhart’s momentum factor. The Table shows results for four parsimonious
model speci?cations (i-iv) where the factors are selected according to the Schwarz criterion (joint
estimation of the equation for asset managers’ and hedge funds’ BMS returns). Speci?cation (v)
includes all factors jointly. We further report the estimated intercept ?, the adjusted R
2
and the
BIC computed for the two-equation system (Sys-BIC). Below the regression coe?cients, t-statistics
based on HAC standard errors are reported (in brackets).
Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (v) (i) (ii) (iii) (iv) (v)
r
m
-0.02 -0.03 0.03 0.03
[-0.28] [-0.38] [0.61] [0.52]
SMB 0.05 0.07 0.06 0.05
[-0.66] [1.01] [0.46] [0.39]
HML -0.05 -0.04 0.03 0.02
[-0.47] [-0.47] [0.46] [0.3]
UMD 0.01 0.00 0.00 0.02
[0.16] [-0.13] [-0.07] [0.37]
? 1.26 1.29 1.30 1.28 1.28 0.84 0.86 0.86 0.87 0.83
[4.57] [4.85] [4.74] [4.77] [4.72] [3.38] [3.69] [3.7] [3.77] [3.45]
¯
R
2
-0.01 -0.01 -0.01 -0.01 -0.03 0.00 0.00 -0.01 -0.01 -0.03
Sys-BIC 3.74 3.74 3.74 3.75 3.97 3.74 3.74 3.74 3.75 3.97
65
Table A.18. Risk Exposures: FX Factors
This table reports regression results for the risk exposures of the BMS portfolios of ?nancial FX market
end-users, that is, asset managers and hedge funds. The FX factors include the excess Dollar risk factor and
the carry risk factor by Lustig, Roussanov, and Verdelhan (2011). VOL
FX
is the global FX volatility risk
factor (factor mimicking portfolio). The Table shows results for four parsimonious model speci?cations (i-iv)
where the factors are selected according to the Schwarz criterion (joint estimation of the equation for asset
managers’ and hedge funds’ BMS returns). Speci?cation (v) includes all factors jointly. We further report
the estimated intercept ?, the adjusted R
2
and the BIC computed for the two-equation system (Sys-BIC).
Below the regression coe?cients, t-statistics based on HAC standard errors are reported (in brackets).
Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (v) (i) (ii) (iii) (iv) (v)
DOL 0.04 0.04 -0.11 -0.11
[0.28] [0.28] [-0.99] [-0.98]
HML
FX
-0.21 -0.04 -0.03 0.27 0.19 0.19
[-1.75] [-0.33] [-0.32] [2.63] [1.66] [1.62]
VOL
FX
0.07 0.07 0.08 0.07 -0.07 -0.03 -0.08 -0.05
[2.44] [2.34] [2.24] [2.55] [-2.5] [-1.30] [-2.58] [-1.46]
? 1.46 1.43 1.47 1.45 1.46 0.71 0.69 0.67 0.73 0.69
[5.32] [4.75] [5.15] [5.43] [5.26] [3.1] [2.95] [2.93] [3.11] [2.96]
¯
R
2
0.10 0.06 0.09 0.09 0.09 0.11 0.13 0.13 0.11 0.13
Sys-BIC 3.53 3.55 3.57 3.60 3.64 3.53 3.55 3.57 3.60 3.64
66
Table A.19. Risk Exposures: Fung-Hsieh Factors
This table reports regression results for the risk exposures of the BMS portfolios of ?nancial FX market end-
users, that is, asset managers and hedge funds. The options-based factors are intended to capture non-linear
payo? features that are typical of hedge fund returns (Fung and Hsieh, 2001). Panel A considers the ?ve
market timing-factors for various asset classes (BN - Bonds, FX, CM - commodities, Eq - equities, and IR -
short term interest rates) as the starting point. Panel B uses the Fung-Hsieh 7-factor model as the starting
point, as outlined in the main text. The Table shows results for four parsimonious model speci?cations (i-iv)
where the factors are selected according to the Schwarz criterion (joint estimation of the equation for asset
managers’ and hedge funds’ BMS returns). Speci?cation (v) includes all factors jointly. We further report
the estimated intercept ?, the adjusted R
2
and the BIC computed for the two-equation system (Sys-BIC).
Below the regression coe?cients, t-statistics based on HAC standard errors are reported (in brackets).
A. Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (v) (i) (ii) (iii) (iv) (v)
PTFS
BD
1.19 1.20 -1.95 -1.96
[0.54] [0.54] [-1.24] [-1.30]
PTFS
FX
2.69 3.67 2.50 3.00 -3.37 -3.87 -3.06 -3.49
[2.67] [2.70] [2.78] [2.78] [-2.97] [-2.62] [-2.70] [-3.13]
PTFS
CM
-1.88 1.63
[-1.24] [1.07]
PTFS
IR
2.28 2.70 2.20 2.50 -1.16 -1.69 -1.04 -1.29
[3.45] [3.61] [3.47] [3.78] [-1.32] [-1.49] [-1.12] [-1.78]
PTFS
EQ
-0.45 0.36
[-0.23] [0.17]
? 1.20 1.25 1.22 1.24 1.20 0.93 0.91 0.91 0.86 0.90
[5.51] [5.18] [5.4] [4.8] [4.11] [4.25] [4.24] [4.02] [4.22] [4.15]
¯
R
2
0.14 0.07 0.11 0.13 0.13 0.11 0.09 0.04 0.12 0.11
Sys-BIC 3.55 3.58 3.59 3.61 3.75 3.55 3.58 3.59 3.61 3.75
B. Asset Managers Hedge Funds
(i) (ii) (iii) (iv) (v) (i) (ii) (iii) (iv) (v)
rm 0.03 -0.03
[0.53] [-1.10]
SMB 0.07 0.06
[1.13] [0.52]
PTFS
FX
2.48 3.67 2.13 -2.89 -3.87 -3.05
[2.39] [2.70] [1.77] [-2.85] [-2.62] [-2.64]
PTFS
CM
0.46 0.96
[0.34] [0.61]
PTFS
BD
1.68 0.98 -2.31 -1.95
[0.82] [0.54] [-1.60] [-1.38]
?TS -0.90 -0.24
[-1.1] [-0.37]
?DF 3.72 4.75 4.38 3.99 -3.06 -4.26 -3.75 -2.79
[2.05] [2.49] [2.48] [2.02] [-1.86] [-2.28] [-1.88] [-1.56]
? 1.26 1.29 1.25 1.34 1.27 0.90 0.87 0.91 0.79 0.83
[5.11] [4.97] [5.18] [4.58] [4.86] [4.17] [3.91] [4.24] [3.78] [4.03]
¯
R
2
0.12 0.09 0.07 0.09 0.11 0.13 0.09 0.09 0.10 0.11
Sys-BIC 3.55 3.56 3.58 3.61 3.89 3.55 3.56 3.58 3.61 3.89
67
Table A.20. Risk Exposures T15/T9
This table reports regression results for the risk exposures of the BMS portfolios computed from the total
?ows based on either 15 (T15) or 9 currencies. The methodological framework in Panel A is a modi?ed linear
Fung-Hsieh model with eight factors as outlined in the main text. Panel B also accounts for the conditional
exposure to stock market returns by including additional interaction terms of market returns. The three
conditioning variables are ?rst di?erences of the 3-month T-Bill rate, the VIX and the TED spread. The
Table shows results for four parsimonious model speci?cations where the factors are selected according to
the Schwarz criterion (joint estimation of the equation for T15 and T9 BMS returns). Results for the other
factors are not reported. We further report the estimated intercept ?, the adjusted R
2
and the BIC computed
for the two-equation system (Sys-BIC). Below the regression coe?cients, t-statistics based on HAC standard
errors are reported (in brackets).
A. T15 T9
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
r
m
-0.04 -0.09
[-0.79] [-2.11]
DOL -0.08 -0.28
[-0.54] [-2.74]
?TS -0.96 -0.60
[-1.45] [-1.18]
?DF -4.72 -5.20 -5.06 -4.74 1.37 -0.22 0.59 1.35
[-2.17] [-2.17] [-2.11] [-2.41] [0.97] [-0.17] [0.48] [0.94]
? 0.88 0.92 0.90 0.91 1.05 1.17 1.09 1.06
[4.87] [4.51] [4.97] [4.85] [4.46] [4.94] [4.48] [4.41]
¯
R
2
0.10 0.10 0.10 0.11 0.00 0.05 0.03 0.00
Sys-BIC 3.19 3.21 3.23 3.25 3.19 3.21 3.23 3.25
B. T15 T9
(i) (ii) (iii) (iv) (i) (ii) (iii) (iv)
DOL -0.07 -0.27
[-0.42] [-2.56]
r
m
·? TED(t-1) 0.26 0.26
[2.73] [2.03]
r
m
·? TB(t-1) -0.36 0.05
[-2.46] [0.19]
?DF -4.79 -3.70 -3.36 -5.16 1.30 1.15 2.71 -0.20
[-2.24] [-2.03] [-2.05] [-2.19] [0.92] [0.76] [1.15] [-0.15]
? 0.86 0.85 0.82 0.89 1.03 1.03 0.99 1.15
[4.63] [4.67] [4.53] [4.22] [4.3] [4.3] [4.13] [4.73]
¯
R
2
0.11 0.13 0.14 0.10 0.00 -0.01 0.04 0.04
Sys-BIC 3.19 3.20 3.21 3.21 3.19 3.20 3.21 3.21
68
Table A.21. Marginal Forecast Performance: Four-Factor Adjusted Excess Returns
This table shows excess returns for BMS portfolios sorted on lagged order ?ow as in Table
III. We do not only sort on order ?ow of the previous day but also allow for longer lags of
up to nine days between order ?ow signals and portfolio formation. Portfolios are rebalanced
daily. T15 denotes portfolios sorts on total order ?ows and the sample of all 15 currencies,
and T9 denotes portfolios sorts on total order ?ows and the sample of 9 developed currencies;
AM, HF, CC, and PC denote portfolios sorts on asset managers’, hedge funds’, corporate
clients’, and private clients’ order ?ows, respectively. Compared to Table III which reports
unadjusted excess returns, we report adjusted excess returns based on the Carhart (1997)
four-factor model.
Controlling for MKTRF, HML, SMB, and UMD
T15 10.31 24.87 10.55 -1.11 3.31 0.33 0.27 2.21 -2.15 -0.72
[3.92] [8.90] [4.50] [-0.44] [1.38] [0.14] [0.11] [0.94] [-0.86] [-0.31]
T9 12.52 24.59 7.83 -4.10 6.20 -1.76 2.26 1.45 -0.97 -1.93
[4.53] [8.89] [3.10] [-1.56] [2.29] [-0.68] [0.88] [0.54] [-0.37] [-0.78]
AM 16.06 25.30 8.85 -1.55 2.68 0.23 -0.06 3.67 2.31 -2.98
[5.74] [8.88] [3.15] [-0.55] [1.06] [0.09] [-0.02] [1.29] [0.82] [-1.03]
HF 9.95 28.37 1.44 -2.90 0.05 -6.09 2.88 -0.20 -4.69 -0.72
[3.83] [9.42] [0.53] [-1.14] [0.02] [-2.34] [1.11] [-0.07] [-1.78] [-0.26]
CC -4.24 -8.30 -1.82 2.48 -5.23 2.13 -0.52 1.48 -0.09 3.41
[-1.54] [-2.88] [-0.59] [0.96] [-2.01] [0.82] [-0.20] [0.57] [-0.03] [1.32]
PC -14.35 -34.38 3.45 2.29 -3.37 -1.08 2.33 -1.65 0.80 1.83
[-5.09] [-10.97] [1.29] [0.84] [-1.13] [-0.36] [0.87] [-0.61] [0.30] [0.68]
69
Figure A.1. BMS Excess Returns in Event Time: Four Liquid Currency Pairs
This ?gure is similar to Figure 3 but here BMS returns are based on sorting four liquid currencies (EUR/USD,
JPY/USD, GBP/USD, CHF/USD) into two portfolios.
70
doc_447151224.pdf