Covered interest arbitrage profits: The role of liquidity and credit risk

Journal of Banking & Finance 34 (2010) 1098–1107
Contents lists available at ScienceDirect
Journal of Banking & Finance
journal homepage: www.elsevier.com/locate/jbf
Covered interest arbitrage profits: The role of liquidity and credit risk
Wai-Ming Fong a, Giorgio Valente b,*, Joseph K.W. Fung c
a
The Chinese University of Hong Kong, Shatin, N.T., Hong Kong
University of Leicester, University Road, Leicester LE1 7RH, United Kingdom
c
Hong Kong Baptist University, Kowloon Tong, Kowloon, Hong Kong
b
a r t i c l e
i n f o
Article history:
Received 20 July 2009
Accepted 6 November 2009
Available online 11 November 2009
JEL classification:
F31
F41
G14
G15
a b s t r a c t
We study the profitability of Covered Interest Parity (CIP) arbitrage violations and their relationship with
market liquidity and credit risk using a novel and unique dataset of tick-by-tick firm quotes for all financial instruments involved in the arbitrage strategy. The empirical analysis shows that positive CIP arbitrage deviations include a compensation for liquidity and credit risk. Once these risk premia are taken
into account, small arbitrage profits only accrue to traders who are able to negotiate low trading costs.
The results are robust to stale pricing and the nonsynchronous trading occurring in the markets involved
in the arbitrage strategy.
Ó 2009 Elsevier B.V. All rights reserved.
Keywords:
Exchange rates
Arbitrage
Covered interest rate parity
Foreign exchange microstructure
1. Introduction
The Covered Interest Parity (henceforth CIP) theorem states that
the covered interest rate differential between two identical riskfree securities denominated in two different currencies should be
equal to zero. Since Keynes (1923) and Einzig (1937) a vast body
of empirical and theoretical literature investigated the joint behavior of foreign exchange (henceforth FX) and money markets and
the occurrence of arbitrage opportunities. The general consensus
is that ‘‘CIP is a reasonably mild assumption, given the extensive
empirical evidence suggesting that CIP holds” (Sarno, 2005, p.
675) especially when transaction costs are taken into account.1
However, some studies pointed out that currency markets were
and still are characterized by a substantial number of instances in
which CIP deviations exceed the transaction cost band, implying
profitable arbitrage opportunities (see, inter alia, Balke and Wohar,
1998; Peel and Taylor, 2002; Akram et al., 2008, 2009; Marshall et
al., 2008; Mancini Griffoli and Ranaldo, 2009 and the references
* Corresponding author. Tel.: +44 116 2522629; fax: +44 116 2525351.
E-mail addresses: wmfong@cuhk.edu.hk (W.-M. Fong), giorgio.valente@le.ac.uk
(G. Valente), jfung@hkbu.edu.hk (J.K.W. Fung).
1
A selective list of early studies investigating the validity of CIP include, among
others, Frenkel and Levich (1975, 1977), Deardorff (1979), Taylor (1987, 1989), Rhee
and Chang (1992) and the references therein.
0378-4266/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jbankfin.2009.11.008
therein). Although these findings could be explained by means of
slow and infrequent communications between London and New
York banks in the early 1920s (Einzig, 1937, p. 57; Peel and Taylor,
2002), it is more difficult to rationalize them in presence of modern
FX dealing rooms, where prices can be obtained within seconds
and orders can be carried out in an automated and synchronized
fashion.
CIP arbitrage profits are generally assumed to be riskless2 and
attainable without capital outlay. However, real-world arbitrage
activities are subject to impediments which prevent arbitrageurs
to fully exploit arbitrage opportunities and move asset prices to their
equilibrium no-arbitrage values (see, inter alia, Shleifer and Vishny,
1997 and the references therein). Two of these impediments are par-
2
In the spirit of Deardorff (1979), a different test of CIP may arise from considering
one-way arbitrage opportunities in the form of owner’s arbitrage (OA) and borrower’s
arbitrage (BA). However, it is important to point out that round-trip and one-way
arbitrage conditions differ in that violations of the latter do not necessarily prove the
existence of riskless profits. This is due to the fact that one-way arbitrage
opportunities require an excess supply or demand of funds, while round-trip
arbitrage does not require funds to be lent or borrowed. They only indicate the
presence of price differentials that are due to different pricing practices, market
segmentation and/or different demand/supply conditions in all of the markets
involved in the CIP arbitrage. For these reasons we consider CIP (or round-trip)
arbitrage as the relevant arbitrage condition throughout the paper. See also Akram
et al. (2009).
W.-M. Fong et al. / Journal of Banking & Finance 34 (2010) 1098–1107
ticularly important in the context of CIP arbitrage: market liquidity
and credit risk.
It is well known that the Law of One Price (LOP) holds when
markets have few frictions associated with transacting, especially
illiquidity indicators such as the bid–ask spread. In these cases,
few arbitrage opportunities arise in real time and the violations
of LOP are generally not profitable if all transaction costs are taken
into account. However, LOP does not hold when markets are illiquid. Several studies have documented that aspects of the market
microstructure may cause the temporary deviation of prices from
their no-arbitrage values (Stoll, 1978; O’Hara and Oldfield, 1986;
Gârleanu and Pedersen, 2009) and financial market liquidity plays
a key role in moving prices to eliminate arbitrage opportunities
(Kumar and Seppi, 1994; Roll et al., 2007 and the references
therein).
The other important aspect that has been surprisingly neglected
in the literature on CIP is the fact that CIP arbitrage opportunities
are generally computed using interbank deposit rates which are
not risk-free since they are subject to different levels of credit risk.
Arbitrage activities can eliminate the market risk associated with
the movements of exchange rates and interest rates, but they cannot eliminate the credit risk associated with the interbank deposits
and the counterparty risk in the forward contracts. Furthermore,
when interbank markets are highly concentrated, the banks that
have an extensive retail deposit-taking network have also a high
degree of market power that is exercised by means of a very selective assessment of counterparty credit-worthiness and the imposition of tight credit limits (i.e. trading volumes).
All of these arguments suggest that some of the arbitrage profits
recorded in recent studies may reflect premia demanded by market
participants as compensation for the liquidity and credit risk faced
during arbitrage transactions.3 In this paper we test this conjecture
by investigating CIP arbitrage profits for a single currency market
(i.e. Hong Kong dollar, henceforth HKD) which is among the 10 largest currency markets in terms of daily average turnover (Bank for
International Settlement, BIS 2007) but its overall liquidity is
poorer than in major currency markets and its banking market is
highly concentrated. Using a novel and unique dataset of tick-bytick tradable (firm) spot and forward quotes for HK dollar vis-àvis US dollar (USD), as well as tradable HKD- and USD-denominated deposit rates over different short-term maturity tenors, we
carry out an extensive analysis of CIP deviations and their relationship with market liquidity and credit risk.
Our main findings are as follows. First, we find that HKD/USD FX
market is characterized by a large number of CIP deviations and
most of these deviations exceed the transaction cost band implied
by the bid–ask spreads of the individual financial instruments. Second, positive CIP deviations are clustered at the bid side of the market and their economic value is sizable across the maturity
spectrum. Third, positive CIP deviations are associated with market
illiquidity at very short maturities and with differences in credit
risk between the Hong Kong and US financial institutions across
the full maturity spectrum. Fourth, small residual arbitrage profits
are still present even after taking into account liquidity and credit
risk. However, since additional trading costs (i.e. brokerage fees
and settlement costs) are generally not included in quoted prices,
their explicit consideration is likely to offset any profits that arise
from arbitrage activities. Given the existence of heterogeneous
additional trading costs (Mavrides, 1992), only the traders who
are able to negotiate low trading costs will be able to reap genuine
arbitrage profits as compensation for their activity (Deardorff,
1979; Grossman and Stiglitz, 1980).
This study differs and improves upon previous contributions in
several respects. The dataset employed in this study, to the best of
our knowledge, is the first to record tradable quotes at tick-by-tick
frequency for all financial instruments involved in the CIP arbitrage
over a period of eight months. Taylor (1987), a landmark study in the
literature on CIP arbitrage, employed interest rates and exchange
rates data that are recorded by phoning several London brokers at
10 min frequency during the most active hours (9:00–16:30 GMT)
over three days in 1985. Other studies re-examining Taylor’s results,
employ datasets which exhibit various limitations. One of the most
relevant limitations is that all or some of the prices used to compute
arbitrage deviations are purely indicative, hence evidence of arbitrage opportunities recorded using these datasets may not necessarily imply glaring profitable opportunities. The dataset employed in
this paper is therefore unique and allows us to carry out a thorough
economic assessment of CIP arbitrage since contemporaneous tradable quotes of domestic and foreign deposit rates and spot and forward exchange rates are crucial to establish whether an apparent
deviation from the no-arbitrage conditions in the FX market represents a genuine profitable arbitrage opportunity.
Our work builds on Cheung and Chan (1994) and Akram et al.
(2008, 2009) who investigate CIP arbitrage violations in both
round-trip and one-way arbitrage in the Hong Kong market and
for the major currencies respectively. Our work differs from theirs
in several important ways. First, Cheung and Chan (1994) employ
daily indicative quotes over a shorter period of time (3 months) recorded over an interval of 30 min (between 10:45 am and 11:15
am Hong Kong time) while Akram et al. (2008, 2009) employ a
tick-by-tick dataset comprising tradable quotes for spot FX rates
and indicative quotes of FX swaps and currency deposit rates. Second, Cheung and Chan (1994) data are collected from only two
commercial banks (i.e. Tokyo Forex and Tullett and Hua Chiao
Commercial Bank Limited), while our tradable quotes are provided
by ICAP, the world’s largest voice and electronic interdealer broker
which exhibits a substantial market share of the spot and forward
FX trading in Hong Kong (ICAP, 2007).4 Third, both Cheung and
Chan (1994) and Akram et al. (2008, 2009), along the lines of earlier
studies, do not consider the role of market liquidity and credit risk in
affecting CIP arbitrage profits, which is the main goal of our paper.
Other related papers are Baba and Packer (2009), Genberg et al.
(2009), Mancini Griffoli and Ranaldo (2009) and McAndrews and
Sarkar (2009) who investigate the spillover effects of the money
market turbulence in 2007–2008 on short-term CIP deviations between the US dollar and the major currencies and the effects of the
Federal Reserve’s responses to the crisis on credit and liquidity risk.
Our work differs from theirs in that we do not focus on the behavior of the FX and money markets during turbulent periods, such as
the recent 2007–2008 financial crisis, but we try to understand the
role of market liquidity and credit risk on arbitrage activities under
normal market conditions. Furthermore, we assess the impact of
market liquidity and credit risk on CIP arbitrage profits computed
by taking into account features which reflect very closely realistic
conditions faced by arbitrageurs, while most of the other studies
look at CIP deviations computed without taking into account transaction costs or use interest rate spreads.5
4
3
Another way of rationalizing the effect of liquidity on arbitrage profits relates to
the intimate relationship between market liquidity and funding liquidity. If market
liquidity is correlated with funding liquidity (as in models with liquidity spirals such
as Brunnermeier and Pedersen, 2009) the effect we document in this paper may not
be necessarily interpreted as a liquidity risk, but as a higher cost of arbitrage capital.
See also Mancini Griffoli and Ranaldo (2009).
1099
Further details on the market coverage of our dataset are discussed in Section 2.2.
Mancini Griffoli and Ranaldo (2009) also investigate deviations from CIP for the
major currencies incorporating realistic features of the FX market. However, our work
differs from theirs since we use tick-by-tick firm (transactable) quotes provided by
the world’s largest electronic and voice interdealer broker (ICAP) while Mancini
Griffoli and Ranaldo (2009) use indicative quotes provided by one single intermediary
(Tullet Prebon) recorded four times per day.
5
1100
W.-M. Fong et al. / Journal of Banking & Finance 34 (2010) 1098–1107
The remainder of this paper is organized as follows. In Section 2
we discuss the theoretical underpinnings, our dataset and the main
institutional features of the HKD FX market. Section 3 reports preliminary summary statistics and in Section 4 the results of empirical analysis are discussed. Section 5 contains some robustness
checks and a final section concludes.
2. Computations, data and institutional issues
2.1. CIP deviations
Deviations from CIP are routinely expressed as
ð1 þ id;k Þ F k
ð1 þ if ;k Þ S
ð1Þ
where id and if are the domestic and foreign interest rates on similar
assets of a certain maturity tenor k, S is the prevailing spot exchange
rate and Fk is the forward exchange rate with maturity tenor k.
Any deviations from Eq. (1) would represent a risk-free arbitrage opportunity in a frictionless world. If we introduce bid and
ask prices to Eq. (1), any deviations from CIP are profitable if either
of the following inequalities hold6:
a
F bk ð1 þ id;k Þ
>0
Sa ð1 þ ibf;k Þ
a
Sb ð1 þ if ;k Þ
>0
F ak ð1 þ ibd;k Þ
ð2Þ
where superscript a and b denote the ask and bid prices,
respectively.
Since the quotes in the deposit markets are reported as annualized rates and HKD deposits and USD deposits have a different day
count convention, throughout the paper we compute CIP deviations as follows:
2
3
D
a
F bk 4ð1 þ id;k Þ365 5
> 0 or
CIP ðbidÞ : a D
b
S
ð1 þ if ;k Þ360
2
3
D
a
Sb 4ð1 þ if ;k Þ360 5
CIP ðaskÞ : a >0
D
b
Fk
ð1 þ id;k Þ365
ð3Þ
where D is the number of days to maturity of the forward and deposit contracts over the maturity tenor k,7 and superscripts a and
b denote the ask and bid prices respectively. Eq. (3) indicate that
only positive deviations are profitable.
2.2. Data
The empirical analysis performed in the paper employs data for
the HKD FX market. The BIS triennial survey on FX market activity
shows that, in April 2007, the HKD market is the 10th-largest currency market in terms of percentage share of the daily world average turnover, together with the Swedish krona (SEK) and just
above the Norwegian krone (NOK), two currencies that have been
investigated in recent empirical studies (Bjønees and Rime, 2005).
6
We assume that CIP arbitrage is conducted by trading four instruments: spot and
forward exchange rates and domestic and foreign deposit rates. An alternative
characterization of the arbitrage strategy would imply trading swap points rather
than spot and outright forward rates (Cheung and Chan, 1994; Akram et al., 2008,
2009 and the references therein). Although we do not consider this alternative
characterization of the CIP strategy in this paper, we note that the use of swaps may
affect the number of arbitrage opportunities which could be somewhat different than
the ones we report in Section 4.
7
In Eq. (3) the days to maturity D are computed as the actual number of business
days between the spot date (or value date) and the maturity date after taking into
account bank holidays in Hong Kong (see Akram et al., 2008, Appendix A).
Our dataset is a collection of tick-by-tick data obtained from ICAP
for the sample period ranging between May 17th, 2005 and
December 31st, 2005.8 ICAP is the world’s largest voice and electronic interdealer broker; in 2006 it covered 65% of the worldwide
FX spot voice market and 35% of the FX voice forward market.
Although recent market trends have exhibited a furious shift from
voice broking to electronic broking, the voice-broking FX market is
still very active and in certain geographical areas it is the dominant
trading venue. In fact, in 2006, FX voice broking contributed a hefty
69% to the ICAP group’s overall profits. In the context of emerging
markets (especially the HKD market), the above percentages can be
considered conservative, as ICAP covers more than half of the market share in emerging market securities trading (ICAP, 2007). The
market coverage of our dataset, computed as in Bjønees and Rime
(2005), is about 40% of the overall HKD spot and forward FX market
as reported in BIS (2007).9
The dataset comprises all of the best ask and bid prices for the
HKD/USD spot exchange rate, the HKD/USD outright forward rate
and all of the best ask and bid prices for the HKD- and USD-denominated deposit rates. The forward exchange rates and both the
domestic- and foreign-currency denominated deposit rates are relative to three different maturity tenors: one week, four weeks, and
12 weeks. A particular novelty of this unique dataset is that all of
the ask and bid prices are firm (hence, directly tradable), which
is different from most of the previous studies in which all or some
of the quotes are indicative. All of the quotes in our dataset are retrieved from ICAP voice-broking record tapes.
2.3. Institutional issues
Two important aspects characterize the HKD FX market. The
first is that the Hong Kong monetary policy is carried out by means
of a currency board. As a consequence the HKD has been tied to the
USD since 1983 and in the last two and one-half decades, the value
of the HKD/USD spot exchange rate has not moved substantially
away from the imposed parity (7.8 HKD per USD), even under
the turbulent market conditions of 1997–1998. It is instructive to
note that this peculiar monetary regime does not hinge on arbitrage practices. In fact, arbitrage pricing in the FX market is independent of the currency regime. The arbitrage conditions tested
in this paper should offer, in absence of frictions, riskless profits
if the deviations are large enough to cover transaction costs. There
is no role for expectations of exchange rates in these conditions as
the degree of violation is known at any point in time and any rational trader would exploit such violation (size of trades allowed
by the market) regardless of what the central bank is doing.10
The second important aspect is that the banking market in Hong
Kong is highly concentrated. In fact, although there are about 155
8
The sample period is chosen because of data availability (firm quotes are not
available from ICAP before May 17th, 2005).
9
We thank Dagfinn Rime for suggesting this to us.
10
However, as a robustness check we have tested for the possibility that the activity
of the Hong Kong Monetary Authority (henceforth HKMA) might have affected
arbitrage activities and the size of arbitrage opportunities over the sample period.
More specifically we have considered the case of FX intervention dates and dates
when policy rates are changed. As for the first case, the ex-post records of the HKMA’s
FX market activities indicate that during the sample period investigated in this paper,
no FX intervention has taken place. In fact official market operations stopped in April
2005 and resumed only at the end of 2007. When considering policy rate changes, it is
important to emphasize that, given the specific currency regime, the HKMA does not
set interest rates but passively follows any federal fund rate changes set by Federal
Open Market Committee (FOMC) of the US Federal Reserve. During our sample period
5 FOMC meetings took place which revised US interest rates. In order to check for the
possibility that our results may be due to anomalies related to dates where policy
interest rates are set, we have carried out a fraction of the empirical analysis using
data where the FOMC are removed. The results, not reported to save space, are
qualitatively and quantitatively similar to the ones discussed in Section 4.
W.-M. Fong et al. / Journal of Banking & Finance 34 (2010) 1098–1107
licensed banks, only a small number of them are retail banks (i.e.
banks that have a network of branches to carry on both retail
and wholesale business in the domestic market). In addition, the
top three banking groups account for over a half of total customer
deposits in Hong Kong (Carse, 2001). As a result, the banks that
have an extensive retail deposit-taking network have also a high
degree of market power that is exercised by means of a very selective assessment of counterparty credit-worthiness and the imposition of tight credit limits (i.e. trading volumes). This peculiar aspect
of the HKD market, associated with its relative lower liquidity
(compared to major currencies markets), makes it an interesting
case study to investigate how liquidity and credit risk affect the
formation of CIP arbitrage deviations.
3. Preliminary analysis
In this section measures of market liquidity and credit risk are
computed. For each trading day, the measures described below
are computed using daily or intraday data. In the latter case, as
in previous empirical studies a daily average is used as input
for the subsequent estimations (see Roll et al., 2007 and the references therein).
Market liquidity has many aspects (immediacy, costs, depth
and resiliency) and no empirical measure is able to reflect them
all. Our dataset only comprises best ask and bid prices for all
instruments involved in the CIP arbitrage, hence no information
is provided on transaction prices, limit order books and trading
volumes. As consequence, we proxy liquidity by means of quoted
bid–ask spreads, daily number of quotations and effective bid–
ask spreads estimated using the procedure suggested by Corwin
and Schultz (2008). The liquidity measures employed in the
empirical investigation are routinely used in existing studies on
market liquidity (see, inter alia, Roll et al., 2007; Corwin and
Schultz, 2008; Goyenko et al., 2009 and the references therein).
However, they are not immune from criticisms. In fact, bid–ask
spreads capture the implicit cost of trading (see Harris, 2003;
Stoll, 2000 and the references therein) but they do not consider
other aspects of liquidity. Similarly, the daily number of quotations is usually employed in empirical works on the basis of
the notion that liquidity ought to be higher in more active markets (Demsetz, 1968). However, the relationship between liquidity and market activity has been questioned in recent empirical
studies (Johnson, 2008).
To test for the effect of credit risk on CIP arbitrage deviations
we use a measure of the difference in credit risk perception
between Hong Kong and US financial institutions. As in Genberg
et al. (2009) the measure is computed as the difference
between the HKD-denominated LIBOR-OIS spread and the equivalent LIBOR-OIS spread denominated in USD.11 A positive difference denotes a situation where Hong Kong financial institutions
are perceived to be riskier than US financial institutions. This, in
turn, generates a credit risk premium that adds to the funding
rates in HKD since Hong Kong banks, having to preserve funds
on hand, became more cautious in lending to their US
counterparties.
The use of LIBOR-OIS as reflecting credit risk is currently debated and recent papers use other measures to proxy for credit risk
(see, inter alia, Baba and Packer, 2009; McAndrews and Sarkar,
2009; Coffey et al., 2009; Mancini Griffoli and Ranaldo, 2009 and
the references therein). In fact, LIBOR-OIS spreads change over
time because of changes in credit risk (counterparty risk) exhibited
by the banks members of the LIBOR panel and the overall funding
liquidity conditions (Giavazzi, 2008; Kotomin et al., 2008; Bru11
LIBOR and Overnight Index Swaps (OIS) daily data are obtained from Bloomberg.
1101
nnermeier and Pedersen, 2009). A cleaner measure of credit risk
would be represented by the differences in CDS spreads between
the Hong Kong and US financial institutions. However, although
there are indices of CDS spreads for US financial institution, an
equivalent index for Hong Kong financial institutions is not available over the sample period.12,13
Table 1 reports preliminary statistics based on calculations of
CIP arbitrage deviations (Panel A), credit risk measures (Panel B)
and liquidity measures (Panel C). A preliminary look at the intraday data suggests the need for a synchronization of the quotes in
different markets to compute CIP deviations, as the HKD/USD forward and the two deposit markets are comparatively less active
than the HKD/USD spot market. To obtain a time series of contemporaneous quotes for the different instruments, we construct our
synchronized data as follows: First, we exclude days with few
observations (such as weekends and public holidays). After these
adjustments the number of trading days range between 157 (four
weeks maturity) and 159 (one and 12 weeks maturity). Second, we
retain the active time period within each day (i.e. between 7:00
and 17:00 Hong Kong time).14 Third, for each instrument, we generate a 15-second interval time series of prevailing quotes. Then,
for each maturity tenor, we consolidate all of the 15-second interval time series to form a synchronized sample. Finally, to mitigate a
possible problem of stale quotes, we delete all observations in
which either the spot or forward prevailing bid and ask were
quoted more than 5 min previously, or in which either the HKD
or USD deposit prevailing bid and ask prices were quoted more
than one hour previously.15 Using this synchronized sample, we
then compute the daily average of these deviations.16
The summary statistics of daily CIP deviations are reported in
Table 1, Panel A for all maturity tenors. All figures are expressed
in pips. In line with Taylor (1987), per-period average CIP
deviations increase with maturity tenors. Furthermore average
12
In fact, the CDS market in Hong Kong is relatively small (Genberg et al., 2009) and
over the sample period investigated in this paper only CDS spreads for 9 non-financial
companies are available. A simple comparison between the US financial CDS index
and the simple average of the 9 Hong Kong CDS spreads would be inappropriate since
the average rating of the 9 Hong Kong companies is generally higher than the rating of
the average company included in the US index. Hence, the difference between the
levels of the two series would not be zero on average and this would induce a bias in
subsequent estimations. We thank Mico Loretan and Eli Remolona for useful
discussions on the CDS market in Hong Kong.
13
Another way to interpret the effect of counterparty risk on arbitrage profits is
suggested in Gârleanu and Pedersen (2009). In a theoretical setting, they demonstrate
that CIP may fail because it requires capital (margin) to trade and profit from CIP
deviations. Our interpretation is also consistent with this line of reasoning since a
higher counterparty risk is likely to induce higher and binding margin requirements
which, in turn, may prevent arbitrage activity and allow positive CIP deviations to
persist over time. See also Darvas (2009).
14
It is worth noting that the market we analyze, during the sample period is mostly
voice-intermediated. Therefore trading does not necessarily occur over 24 h as in
electronic-intermediated markets and it makes sense in this context to define an
opening and closing time.
15
Overall, the number of intraday observations obtained by applying this synchronization scheme is larger than 12,000 for all maturity tenors, with a maximum of
about 28,000 observations at the 12 weeks maturity. In order to check for the
robustness of our results with respect to the selected synchronization scheme, in
Section 5 we compute CIP deviations using other data synchronization filters and
adjustments.
16
It is instructive to note that, on the basis of various conversations with several
proprietary traders and desk managers, it is possible to conduct arbitrage trades
through the voice broker. In fact ICAP does act as one of the main intermediaries in
both foreign exchange (spot and forward) and deposit market in Hong Kong. Hence,
when a positive CIP deviation is detected, an arbitrageur may complete a set of
arbitrage trades by contacting the relevant voice-brokers responsible for different
desks. However these trades do not occur often since, as we detail in Section 4,
additional aspects must be taken into account (such as effective brokerage fees and
settlements costs and credit limits imposed on various traders). In order to check our
results with respect to the effect of stale prices or nonsynchronous trading in the four
markets involved in CIP arbitrage strategies, we carry out some robustness checks in
Section 5.
1102
W.-M. Fong et al. / Journal of Banking & Finance 34 (2010) 1098–1107
CIP deviations are different when computed at the bid or the ask
side of the market. In fact, for all three maturities, CIP deviations
are always positive, on average, at the bid side while they are always negative at the ask side. In all cases the figures are significant
at the 5% statistical level.17 However, it is important to note that
negative average CIP deviations do not imply that CIP arbitrage is
prevented in continuous time. In fact, as in Akram et al. (2008),
the maximum of the distribution of intraday CIP deviations, even
at the ask side of the market, is positive suggesting that some profitable arbitrage opportunities may be present.
Table 1 Panel B reports summary statistics of the credit risk
measure. Average daily LIBOR-OIS spreads are positive for both
currencies and their value increases with maturity tenors. HKDdenominated spreads are higher than USD-denominated spreads.
This suggests that, over the sample period, Hong Kong financial
institutions are perceived to be riskier than US financial institutions. The differences between the two spreads, reported at the
bottom of Panel C, are all significant at conventional statistical level and close to 0.15% per annum across the three maturities
investigated.
The summary statistics of liquidity measures are reported in Table 1 Panel C. As we are interested in measures of liquidity conditions of the four markets, an aggregation of the individual markets’
liquidity measures is carried out. More specifically, aggregate measures are computed as simple arithmetic average (EW) or weighted
average (VW) of the individual markets’ liquidity measures where,
in the latter case, the weights assigned to each market are computed in terms of their daily average turnover as reported in BIS
(2007). Daily quoted bid–ask spreads and effective bid–ask spreads
computed as in Corwin and Schultz (2008) are expressed in pips
and exhibit an increasing pattern as maturity increases. The daily
number of quotations is instead fairly similar across maturities.
The aggregation scheme does not affect the summary statistics
since the differences between EW and VW variables, with the
exception of the number of quotations, are also small in size.
4. CIP arbitrage profits, liquidity, credit risk and additional
trading costs
4.1. CIP arbitrage profits
In this section, we compute positive CIP arbitrage opportunities
that satisfy Eq. (3) and associate them with the measures of liquidity and credit risk discussed in the previous section.
Table 2 shows the percentage of positive CIP deviations out of
all of the observations obtained at intraday level. Across maturity
tenors, we find that the average percentage of positive arbitrage
deviations ranges between 8% and 84%. Furthermore, arbitrage at
the bid price of CIP is much more frequent (73% on average across
maturities) than the arbitrage at the ask price (16%). This finding is
not novel in the literature. In fact Akram et al. (2008) reports for
the Japanese yen (JPY)/USD CIP arbitrage a similar pattern. This
phenomenon can be explained in different ways. One possible
explanation suggested in Akram et al. (2008) relies on the quoting
practices of both currencies. In fact, for both JPY/USD and HKD/USD
the USD is not the quoting currency.
A different and perhaps more interesting explanation can be
provided by looking at the concentration of the banking system.
In fact, looking at both Table 2 and Panel A, the existence of a large
number of arbitrage opportunities which seems to be positive on
average indicates that some frictions are in place and prevent arbi17
As pointed out in Akram et al. (2008) the existence of statistically significant
negative average CIP deviations may be due to the fact that market makers may price
instrument more conservatively than CIP conditions would imply, fearing that some
of the instruments may move in way to generate arbitrage opportunities.
Table 1
Summary statistics.
Mean
Median
Standard deviation
Panel A: CIP deviations
1 week
bid
6.32**
ask
11.76**
6.55**
11.78**
13.30
13.48
4 weeks
bid
ask
14.84**
30.87**
21.91**
37.40**
37.43
37.87
78.47**
121.00**
90.87**
134.24**
74.47
77.81
12 weeks
bid
ask
Panel B: credit risk measures
USD spreads
1 week
0.019**
4 weeks
0.026**
12 weeks
0.037**
0.017**
0.023**
0.038**
0.054
0.034
0.020
HKD spreads
1 week
4 weeks
12 weeks
0.166**
0.182**
0.184**
0.145**
0.165**
0.165**
0.112
0.070
0.086
Spread (HKD–USD)
1 week
4 weeks
12 weeks
0.147**
0.156**
0.146**
0.126**
0.142**
0.132**
0.114
0.082
0.092
Panel C: aggregate liquidity measures
Quoted bid–ask spreads
1 week
VW
5.34**
EW
3.55**
4 weeks
VW
5.64**
EW
4.96**
12 weeks
VW
6.08**
EW
7.89**
Number of quotations
1 week
VW
EW
4 weeks
VW
EW
12 weeks
VW
EW
CS liquidity measure
1 week
VW
EW
4 weeks
VW
EW
12 weeks
VW
EW
1211.1**
343.8**
1207.8**
343.7**
1211.4**
345.3**
3.99**
2.48**
6.67**
5.04**
8.80**
8.93**
5.24**
3.46**
5.61**
4.91**
6.03**
7.80**
1042.9**
295.0**
1042.9**
294.8**
1043.1**
296.1**
3.13**
1.99**
5.36**
4.34**
7.92**
8.27**
1.13
0.66
0.85
0.55
0.85
0.79
837.6
235.6
840.9
236.8
837.6
235.8
3.05
1.77
4.64
3.12
5.05
4.32
The table reports summary statistics of daily CIP deviations, credit risk measures
and liquidity measures computed over the sample period May 17th, 2005 to
December 30th, 2005. Panel A: CIP deviations are daily averages of intraday CIP
deviations computed by synchronizing the time series of quotes of the four markets
from the raw tick data, as described in the text (Section 3). CIP deviations are
expressed in pips. Panel B: credit risk measures over the three maturity tenors are
HKD- and USD-denominated Libor-OIS spreads computed using daily mid-quotes.
Spread (HKD–USD) denotes the difference between HKD-denominated Libor-OIS
spread minus the USD-denominated Libor-OIS spread. Credit risk spreads are
expressed in percentage per annum. Panel C: quoted bid–ask spreads are daily
average of differences between intraday best bid and ask prices for each instrument
involved in the arbitrage strategy. Number of quotations is the daily number of
quotations recorded for each instrument involved in the arbitrage strategy. CS
liquidity measure is the Corwin and Schultz’s (2008) measure computed for each
instrument using intraday highest and lowest quotations. Bid–ask spreads and CS
measures are expressed in pips. All aggregate liquidity measures are computed
assigning either equal weights (EW) to all markets or weights computed in terms of
daily average market turnover (VW).
**
Denote significant at 5% statistical level.
trage forces to operate. If access to liquidity in the HKD market is
restricted by the selectivity of the major Hong Kong retail banks,
positive arbitrage deviations at the bid side (which require borrowing in the HKD deposits market, selling HKD for USD in the spot FX
1103
W.-M. Fong et al. / Journal of Banking & Finance 34 (2010) 1098–1107
Table 2
Frequency of positive arbitrage opportunities.
Table 3
Average positive arbitrage opportunities.
bid
ask
1 week
Total dev
Profitable dev
% share
12,861
9341
72
12,861
2313
17
4 weeks
Total dev
Profitable dev
% share
22,277
14,403
64
22,277
5560
24
12 weeks
Total dev
Profitable dev
% share
28,709
24,299
84
28,709
2310
8
73
16
Average % share
This table shows the frequency of intraday positive CIP arbitrage deviations computed as in Eq. (3) in the main text. Total dev represents the number of all intraday
deviations (including non-positive) computed using the synchronization scheme
described in Section 3 of the main text. Profitable dev records the number of
profitable intraday deviations. % share are the profitable deviations as the percentage of all deviations. See notes to Table 1.
market, lending in the USD deposits market and selling USD for
HKD in the forward market) would survive for a long time since
arbitrageurs would not be able to exploit them because of the credit limits (trading volumes) present in the HKD deposit market.18
However, the unexpectedly high frequency of positive deviations reported in Table 2 is not sufficient evidence of substantial
profitable opportunities, as deviations may have a skewed distribution, i.e. positive deviations may be frequent and small in size.
To better quantify the economic value of these deviations, we report the mean values of the positive CIP deviations in Table 3. Average arbitrage profits are computed as per-period returns and
annualized returns.19 At face value, the average profits from CIP
arbitrage are handsome, and as in Taylor (1987) per-period returns
increase with maturity. Annualized rates do not share this increasing
pattern and range between 0.187% (12 weeks at the ask side) and
0.742% (1 week at the bid side) per annum.
4.2. The role of market liquidity and credit risk
The figures reported in Table 3 take into account transaction
costs (i.e. bid–ask spreads) but they do not tell us whether they
may be just reflecting liquidity and credit risk premia demanded
by market participants to carry out transactions in the HKD market. In order to test this conjecture, in the spirit of Baba and Packer
(2009), we estimate the following linear factor model:
PCDit;k ¼ a þ bLIQ t;k þ cCRDt;k þ et;k
ð4Þ
where PCDit;k denotes the daily average of positive CIP deviations at
the maturity tenor k computed at the i = bid, ask side of the market
on the trading day t, LIQt and CRDt are aggregate liquidity measures
and the HKD–USD LIBOR-OIS spread (credit risk measure) computed on the same day t respectively, and et is an error term.
The parameters b and c denote the sensitivity of positive CIP
arbitrage deviations to market liquidity and credit risk variations.
When liquidity is poorer, positive arbitrage deviations increase because of an increasing liquidity premium. This would imply a b > 0
18
This finding is also consistent with a situation where funding liquidity conditions
are tighter in the HKD deposit market.
19
Annualized returns are computed using the assumptions that 1 pip represents
HKD100 per USD1 million traded, 1 USD is exchanged at the rate of 7.8 HKD and a 365
day-a-year convention is employed.
Per-period returns
Annualized returns
bid
ask
bid
ask
1 week
11.13
(0.99)
7.06
(1.20)
0.742
(0.0006)
0.471
(0.0008)
4 weeks
33.47
(2.30)
28.31
(2.76)
0.559
(0.0004)
0.472
(0.0005)
12 weeks
94.40
(5.24)
33.76
(6.64)
0.525
(0.0003)
0.187
(0.0004)
This table shows the average of all of positive CIP deviations. Per-period returns are
reported in pips. Annualized returns are computed using the assumptions that 1 pip
represents HKD100 per USD1 million traded, 1 USD is exchanged at the rate of 7.8
HKD and a 365 day-a-year convention is employed. Annualized returns are reported
in percentages per annum. The values in parenthesis are asymptotic standard errors
calculated using autocorrelation and heteroskedasticity variance–covariance
matrices (Newey and West, 1987).
when liquidity is proxied by bid–ask spreads and b < 0 when
liquidity is proxied by the daily number of quotations. When credit
risk increases against Hong Kong financial institutions, credit risk
premia are demanded by market participants. This affects the cost
of HKD funds which, in turn, causes positive CIP deviations to increase. This would imply the parameter c to be positive.
The results of the estimation are reported in Table 4. Panel A–C
reports the value of the parameter estimates when liquidity measures are proxied by quoted bid–ask spreads, number of quotations
and effective bid–ask spreads respectively. Some interesting patterns can be retrieved: First, the sign of the estimated parameters
b and c is consistent with our initial prior. In fact, the poorer the
market liquidity and the higher the perceived credit risk against
Hong Kong financial institutions, the larger the positive CIP deviations. Second, across maturity tenors and at both the bid and ask
side of the market, credit risk measures are found to affect positive
CIP deviations significantly at conventional statistical level. Third,
market liquidity seems to affect positive CIP deviations only at very
short maturities (i.e. 1 week). Fourth, the results are insensitive to
the different aggregation schemes used to construct aggregate
liquidity measures.
With respect to the economic significance of the estimated
parameters, if the HKD–USD LIBOR-OIS spread increases by 1 bp,
on average across maturities, positive CIP deviations increase by
0.6 pip at the bid side and 0.8 pip at the ask side. If liquidity decreases, for example when the effective bid–ask spread increases
by 1 pip, positive CIP deviations increase by an average of 0.6 pip
at the bid side and 1.1 pip at the ask side.
These results are consistent with the notion that CIP deviations
are not only due to transaction costs but they also incorporate risk
premia demanded by market participants to provide liquidity in an
illiquid market and carry out transactions with less credit-worthy
counterparties.
Eq. (4) also provides us with some additional insights with regards to residual CIP arbitrage profits which may accrue to arbitrageurs once liquidity and credit risk premia are also explicitly
taken into account. In fact, these residual arbitrage profits can be
measured by the estimated parameter a in Eq. (4), which may be
interpreted as a proxy of daily average positive CIP deviation adjusted for market liquidity and credit risk.
Table 5 reports the estimated residual CIP arbitrage profits expressed as per-period and annualized returns. It is interesting to
note that, on average across different specifications and maturity
tenors, residual arbitrage profits are lower than the positive CIP
deviations reported in Table 3. This reinforces the previous finding
that the positive CIP deviations computed by taking into account
bid and ask prices, contain liquidity and credit risk premia. Once
1104
W.-M. Fong et al. / Journal of Banking & Finance 34 (2010) 1098–1107
Table 4
Arbitrage profits: liquidity or credit risk?
Liquidity
Table 5
Residual arbitrage profits.
R2
Credit risk
Panel A: liquidity = quoted bid–ask spread
bid
(0.325)
1 week
VW
0.7326**
(0.510)
EW
1.1487**
**
0.0071
0.0071**
(0.001)
(0.001)
0.27
0.27
4 weeks
VW
EW
0.9629*
1.1137
(0.577)
(0.704)
0.0062**
0.0062**
(0.002)
(0.002)
0.26
0.26
12 weeks
VW
EW
0.0019
0.0742
(0.754)
(0.646)
0.0097**
0.0097**
(0.003)
(0.003)
0.27
0.27
VW
EW
**
1.3280
2.0478**
(0.687)
(1.000)
**
0.0069
0.0069**
(0.001)
(0.001)
0.24
0.24
VW
EW
0.3996
0.3756
(0.626)
(0.711)
0.0090**
0.0090**
(0.001)
(0.001)
0.27
0.27
(1.008)
(0.802)
**
(0.005)
(0.005)
0.24
0.24
ask
1 week
4 weeks
12 weeks
VW
EW
0.0743
0.3183
7
Liquidity (10 )
Panel B: liquidity = number of quotations
bid
(0.916)
1 week
VW
1.521*
(3.334)
EW
5.536*
0.0102
0.0102**
2
Credit risk
R
0.0070**
0.0071**
(0.001)
(0.001)
0.24
0.24
4 weeks
VW
EW
0.060
0.380
(1.868)
(6.638)
0.0061**
0.0061**
(0.002)
(0.002)
0.25
0.25
12 weeks
VW
EW
5.631
19.455
(3.929)
(11.933)
0.0094**
0.0098**
(0.003)
(0.003)
0.28
0.28
VW
EW
3.294**
11.732**
(1.469)
(5.118)
0.0069**
0.0069**
(0.001)
(0.001)
0.23
0.22
4 weeks
VW
EW
0.151
0.653
(1.950)
(6.884)
0.0092**
0.0091**
(0.001)
(0.001)
0.27
0.27
12 weeks
VW
EW
2.553
8.788
(3.281)
(11.753)
0.0107**
0.0102**
(0.001)
(0.005)
0.24
0.24
ask
1 week
Liquidity
CS
Average
Panel A: CIP deviations (bid)
Per-period returns
1 week
3.4
[2.4, 4.4]
4 weeks
6.1
[7.1,5.1]
12 weeks 63.0
[64.0, 62.0]
1.3
[0.1,2.5]
11.6
[129, 10.3]
63.5
[64.8, 62.2]
0.7
[1.9, 0.5]
17.8
[18.9, 16.6]
72.7
[73.9, 71.5]
1.3
Annualized returns
1 week
0.225
[0.158, 0.291]
4 weeks
0.103
[0.119,0.086]
12 weeks 0.351
[0.373, 0.328]
0.083
[0.002, 0.169]
0.194
[0.210, 0.177]
0.353
[0.376, 0.331]
0.049
[0.128, 0.030]
0.297
[0.313, 0.280]
0.405
[0.427, 0.382]
0.086
Panel B: CIP deviations (ask)
Per-period returns
1 week
16.2
[15.2, 17.2]
4 weeks
11.8
[12.8, 10.8]
12 weeks 21.3
[22.3, 20.3]
14.8
[13.5, 16.1]
11.4
[12.7, 10.1]
20.7
[22.0, 19.4]
12.8
[11.6, 14.0]
12.9
[14.1, 11.7]
23.6
[24.8, 22.4]
14.6
Annualized returns
1 week
1.080
[0.913, 1.247]
4 weeks
0.197
[0.231, 0.164]
12 weeks 0.119
[0.152, 0.085]
0.987
[0.813, 1.160]
0.190
[0.224, 0.157]
0.115
[0.149, 0.082]
0.853
[0.693, 1.013]
0.215
[0.249, 0.182]
0.131
[0.135, 0.128]
0.973
11.8
66.4
0.198
0.369
12.0
21.9
0.201
0.122
This table shows residual profit computed as the intercept a in Eq. (4) reported in
the main text. BA, NQ and CS denote parameter estimates obtained using aggregate
bid–ask spreads, number of quotations and Corwin and Schultz’s (2008) measures
of liquidity respectively. Average denotes the arithmetic average of the parameter
estimates across the three different specifications. See also notes to Table 3.
0.0055**
0.0055**
(0.001)
(0.001)
0.31
0.30
4.3. Additional trading costs
The results reported so far indicate that small residual arbitrage
profits are available to arbitrageurs even after adjusting for bid–ask
prices and liquidity and credit risk premia. A logical question to ask
is then whether the profits reported in Table 5 represent the genuine compensation provided to arbitrageurs for their activity
(Deardorff, 1979; Grossman and Stiglitz, 1980). Those figures, albeit incorporating the effect of the bid and ask prices and liquidity
and credit risk, do not include additional trading costs (i.e. brokerage fees and settlement costs). It is well known among market
practitioners that voice-broking intermediated FX markets are
characterized by a large variability of trading costs and estimates
for broker’s fees are very hard to get (Mavrides, 1992, p. 15). These
costs are applied with different magnitudes to different market
participants with different credit-worthiness and, in some cases,
even to the same market participants for different financial instruments or different market conditions.20 Thus, it is difficult to pin
down a unique number that unambiguously represents the
amount of additional trading costs in this market and ready estimates are not available.
In order to provide an educated guess of the additional roundtrip arbitrage costs faced by market participants in the HKD FX
market, we borrow two estimates computed in a similar context
by Cheung and Chan (1994, p. 25) and Akram et al. (2008, p.
4 weeks
VW
EW
0.0608
0.0369
(0.259)
(0.395)
0.0026*
0.0026*
(0.001)
(0.001)
0.33
0.33
12 weeks
VW
EW
0.4149
0.3527
(0.430)
(0.461)
0.0036*
0.0038*
(0.001)
(0.002)
0.31
0.31
VW
EW
0.857*
1.470**
(0.487)
(0.770)
0.0064**
0.0064**
(0.001)
(0.001)
0.27
0.26
4 weeks
VW
EW
0.0009
0.0203
(0.295)
(0.431)
0.0084**
0.0082**
(0.001)
(0.001)
0.23
0.23
12 weeks
VW
EW
0.1910
0.0843
(0.784)
(0.703)
0.0083*
0.0078*
(0.005)
(0.005)
0.27
0.27
ask
1 week
NQ
R2
Credit risk
Panel C: liquidity = CS liquidity measure
bid
(0.284)
1 week
VW
0.473*
(0.523)
EW
0.869*
BA
The table reports the results of the estimation of Eq. (4). The variable used to
measure credit risk in all Panels is the HKD–USD LIBOR-OIS spread constructed as
discussed in Table 1 while estimations using the three proxies for aggregate
liquidity are reported in Panels A, B and C respectively. Values in parenthesis are
asymptotic standard errors calculated using autocorrelation and heteroskedasticity
variance–covariance matrices (Newey and West, 1987).
*
Denotes parameter estimates significant at 10% level.
**
Denotes parameter estimates significant at 5% level.
these premia are taken into account the amount of residual profits
available to arbitrageurs is smaller. More specifically residual profits are not available at the 1 week maturity, as the estimates are
negative or statistically insignificant, and the positive ones range
between 0.12% (12 weeks at the ask side) and 0.36% (12 weeks at
the bid side) per annum.
20
This is also corroborated by the anecdotal evidence provided by several FX chief
dealers in Hong Kong during informal conversations.
1105
W.-M. Fong et al. / Journal of Banking & Finance 34 (2010) 1098–1107
251–252). The two estimates are imperfect substitutes of the true
additional trading costs. In fact, although Cheung and Chan (1994)
investigate the HKD voice-intermediated FX market, their estimates are representative of the trading costs faced by market participants in 1990s. Differently, Akram et al. (2008) reports trading
costs on the Reuters D3000 platform for the three major currencies
vis-à-vis the USD. Although these estimates are relative to a more
recent period, the markets they analyze and the trading venue
(i.e. electronic) are different from the one investigated in this
paper.
Nevertheless, albeit imperfect, the above estimates can provide
us with two extreme cases within which the current voice-broking
trading costs are located. Residual CIP arbitrage profits and the
estimates of additional trading costs are reported in Table 6. Overall, across maturities and on both the bid and ask side of the market, additional trading costs do not offset the residual arbitrage
profits. In fact, if the largest estimates of trading costs are used,
the residual profits net of trading costs range between 0.088%
and 0.332% per annum on average. These results suggest that when
trading costs deriving from the specific voice-broking environment
are incorporated, only a subset of traders – those with high creditworthiness or longer and established business relationship with
their counterparts – will be able to reap genuine arbitrage profits
as a compensation for their activity.
5. Robustness
In this section we report some robustness checks carried out to
assess the sensitivity of our baseline results discussed in Section 4.
Our first check relates to the possible bias induced by the synchronization rule applied to intraday data discussed in Section 3. More
in particular we assess how the percentage share of positive CIP
deviations, their average and the parameter estimates of Eq. (4)
change when CIP deviations are computed using a different, and
more stringent, synchronization rule.
We construct a different sample using the synchronization rule
so that CIP deviations are only computed when all instruments are
quoted within a minute interval. The results computed using as
liquidity measure the value-weighted aggregate quoted bid–ask
spread21 and reported in Table 7 Panel A and Table 8 Panel A indicate
that the baseline results are robust to this alternative synchronization scheme, although the number of usable intraday observations
drops substantially. In fact, the percentage shares of positive CIP
deviations at both the bid and ask side of the market and their average are consistent with the figures reported in Tables 2 and 3. Similarly the parameter estimates reported in Table 8 Panel A are in line
with the estimates reported in Tables 4 and 5.
A second robustness check on the computation of positive CIP
deviations involves the effects of possible nonsynchronous trading in the four markets involved in the CIP arbitrage. To address
this issue we adopt the methodology proposed by Jokivuolle’s
(1995) which allows us to estimate the true value of an index
when some of its components are subject to nonsynchronous
trading. The results are reported in Table 7 Panel B and Table 8
Panel B. Once again the results are in line with the ones reported
in Tables 2–5.
Overall, the robustness exercises reported in this section indicate that the key results discussed in Sections 3 and 4 are robust.
These baseline results are not affected by stale pricing, the nonsynchronous trading occurring in the markets involved in the arbitrage strategy.
21
The results computed using the remaining two liquidity measures and the EW
aggregation scheme are not reported to save space, but available upon request. In all
cases the results are consistent with the baseline estimates reported in Tables 2–5.
Table 6
Additional trading costs.
Residual profits
(% annualized)
Additional trading costs (% annualized)
Akram et al.
(2008)
Cheung and Chan
(1994)
bid
4 weeks
12 weeks
0.198
0.369
0.013
0.004
0.110
0.037
ask
4 weeks
12 weeks
0.201
0.122
0.013
0.004
0.110
0.037
This table reports the average residual profits from CIP arbitrage as in Table 5
together with estimates of brokerage fees and settlement costs per round-trip
transaction. Estimates of brokerage fees and settlement costs are computed using
the figures reported in Akram et al. (2008, Appendix A.B, pp. 251–252) and Cheung
and Chan (1994, Table 2, p. 25). All figures are annualized rates of return. See notes
to Tables 3 and 5.
Table 7
Data issues and arbitrage opportunities.
CIP deviations
bid
ask
Panel A: alternative synchronization scheme
1 week
Total dev
266
% share
77
Mean (% p.a.)
0.827
266
10
0.387
4 weeks
Total dev
% share
Mean (% p.a.)
268
70
0.626
268
20
0.503
12 weeks
Total dev
% share
Mean (% p.a.)
278
78
0.569
278
9
0.165
Average % share
75
13
Panel B: CIP deviations corrected as in Jokivuolle (1995)
1 week
Total dev
12,861
% share
73
Mean (% p.a.)
0.740
12,861
18
0.573
4 weeks
Total dev
% share
Mean (% p.a.)
22,277
65
0.540
22,277
25
0.453
12 weeks
Total dev
% share
Mean (% p.a.)
28,709
85
0.483
28,709
37
0.209
Average % share
74
26
This table shows the total number of synchronized CIP deviations (Total dev), the
frequency of profitable deviations from round-trip arbitrage in the FX market (%
share) and the average positive CIP deviation (Mean). The figures are computed
using a different synchronization scheme involving the four financial instruments,
as discussed in Section 5 of the main text (Panel A) and the correction for asynchronous trading as in Jokivuolle (1995) (Panel B). Average positive CIP deviations
are expressed as annualized rate of returns. See Notes to Table 3.
6. Conclusions
In this paper we study the profitability of CIP arbitrage violations and their relationship with market liquidity and credit risk.
This analysis is motivated by the growing evidence documenting
that FX markets are often characterized by a substantial number
of instances in which CIP deviations exceed the transaction cost
band coupled with the assumption that such arbitrage deviations
1106
W.-M. Fong et al. / Journal of Banking & Finance 34 (2010) 1098–1107
Table 8
Data issues and parameter estimates.
Liquidity
Credit risk
Residual profits (% p.a.)
Panel A: alternative synchronization scheme
bid
1 week
0.8113**
4 weeks
0.8815
12 weeks
0.0020
(0.325)
(0.626)
(0.754)
0.0064**
0.0060**
0.0098**
(0.001)
(0.002)
(0.003)
0.079
0.124
0.301
[0.026, 0.183]
[0.168, 0.087]
[0.322, 0.274]
ask
1 week
4 weeks
12 weeks
(0.687)
(0.705)
(0.901)
0.0075**
0.0098**
0.0099**
(0.001)
(0.001)
(0.005)
1.047
0.252
0.119
[0.942, 1.151]
[0.303, 0.202]
[0.159, 0.078]
Panel B: CIP deviations corrected as in Jokivuolle (1995)
(0.325)
1 week
0.7327**
(0.577)
4 weeks
0.9629*
12 weeks
0.0020
(0.754)
0.0070**
0.0062**
0.0098**
(0.001)
(0.002)
(0.003)
0.225
0.103
0.351
[0.158, 0.291]
[0.131, 0.075]
[0.373, 0.328]
ask
1 week
4 weeks
12 weeks
0.0069**
0.0090**
0.0102**
(0.001)
(0.001)
(0.005)
1.080
0.197
0.119
[1.013, 1.147]
[0.239, 0.155]
[0.153, 0.085]
1.0116**
0.0580
0.0092
1.3819*
0.3995
0.0742
(0.687)
(0.626)
(1.008)
This table shows the results of the estimation of Eq. (4) when a different synchronization scheme involving the four financial instruments (Panel A) and the correction for
asynchronous trading as in Jokivuolle (1995) (Panel B) are employed. The results are reported for the case of aggregate liquidity = quoted bid–ask spreads where the
aggregation is carried out using weights computed in terms of daily average market turnover (VW). See notes to Tables 1, 4, 5 and 7.
*
Denotes parameter estimates significant at 10% level.
**
Denotes parameter estimates significant at 5% level.
are risk-free profits. We take explicitly into account that real-world
arbitrage activities are subject to impediments which prevent
arbitrageurs to fully exploit arbitrage opportunities and market
liquidity and credit risk assessments are particularly important in
affecting arbitrage activities in the FX market.
Using a novel and unique dataset of tick-by-tick tradable
(firm) spot and forward quotes for HK dollar vis-à-vis US dollar
(USD), as well as tradable HKD- and USD-denominated deposit
rates over different short-term maturity tenors we test the conjecture that CIP arbitrage deviations reflect liquidity and credit
risk premia.
We find a host of interesting results. First, the HKD/USD FX
market is characterized by a large number of CIP deviations
and most of these deviations exceed the transaction cost band
implied by the bid–ask spreads of the individual financial instruments. Second, positive CIP deviations are clustered at the bid
side of the market and their economic value is sizable across
the maturity spectrum. Third, positive CIP deviations are positively correlated with the illiquidity of the market and the differences in credit risk between the Hong Kong and US financial
institutions. Fourth, small residual arbitrage profits are still present even after taking into account liquidity and credit risk. However, since additional trading costs (i.e. brokerage fees and
settlement costs) are generally not included in quoted prices,
their explicit consideration is likely to offset any profits that arise
from arbitrage activities. Given the existence of heterogeneous
additional trading costs, only the traders who are able to negotiate low trading costs will be able to reap genuine arbitrage profits as compensation for their activity. The empirical results
reported in this paper are robust to stale pricing and the nonsynchronous trading occurring in the markets involved in the arbitrage strategy.
Acknowledgments
The authors are grateful to Ike Mathur (Editor) and an anonymous referee for constructive comments. They also would like to
thank Bruno Biais, John Cotter, Andrew Filardo, Hans Genberg,
Leo Goodstadt, Jacob Gyntelberg, Mico Loretan, Anthony Neuberger, Richard Payne, Eli Remolona, Mark Salmon, Lucio Sarno, Alex
Stremme, Andy Rose, Dagfinn Rime, Mark Taylor for useful conversations on an earlier draft of this paper and to the participants in
presentation to the 5th Conference of the Asia-Pacific Association
of Derivatives, Busan, Korea; the 6th HKIMR Summer Workshop,
Hong Kong, Norges Bank, University of Bristol, Cass Business
School, the Chinese University of Hong Kong, Rotterdam School
of Management, University College Dublin, University of Toulouse,
University of Warwick, Trinity College Dublin. Part of this research
has been carried out when both Fung and Valente were visiting the
Hong Kong Institute for Monetary Research (HKIMR). Both authors
are grateful for the hospitality received and the generous financial
support. The authors alone are responsible for the views expressed
in the paper and for any errors that may remain.
References
Akram, F.Q., Rime, D., Sarno, L., 2008. Arbitrage in the foreign exchange market:
Turning on the microscope. Journal of International Economics 76, 237–253.
Akram, F.Q., Rime, D., Sarno, L., 2009. Does the Law of One Price hold in
international financial markets? Evidence from tick data. Journal of Banking
and Finance 33, 1741–1754.
Baba, N., Packer, F., 2009. Interpreting deviations from Covered Interest Parity
during the financial market turmoil of 2007–2008. Journal of Banking and
Finance 33, 1953–1962.
Balke, N.S., Wohar, M.E., 1998. Nonlinear dynamics and Covered Interest Parity.
Empirical Economics 23, 535–559.
Bank for International Settlements, 2007. Triennial central bank survey of foreign
exchange and derivatives market activity.
Bjønees, G.H., Rime, D., 2005. Dealer behaviour and trading systems in foreign
exchange markets. Journal of Financial Economics 75, 571–605.
Brunnermeier, M., Pedersen, H.L., 2009. Market liquidity and funding liquidity.
Review of Financial Studies 22, 2201–2238.
Carse, D., 2001. The banking industry: Competition, consolidation and systemic
stability: The Hong Kong experience. Bank for International Settlements papers
N. 4.
Cheung, K.-Y., Chan, D.P.-M., 1994. Covered interest arbitrage with transaction
costs: An evidence from the Hong Kong foreign exchange market. Asia Pacific
Journal of Management 11, 21–32.
Coffey, N., Hrung, W., Sarkar, A., 2009. Capital constraints, counterparty risk and
deviations from Covered Interest Parity. Working paper, Federal Reserve Bank
of New York.
Corwin, S.A., Schultz, P., 2008. A simple way to estimate bid–ask spreads from daily
high and low prices. Working paper, University of Notre Dame.
Darvas, Z., 2009. Leveraged carry trade portfolios. Journal of Banking and Finance
33, 944–957.
Deardorff, A.V., 1979. One-way arbitrage and its implications for the foreign
exchange markets. Journal of Political Economy 87, 351–364.
W.-M. Fong et al. / Journal of Banking & Finance 34 (2010) 1098–1107
Demsetz, H., 1968. The cost of transacting. Quarterly Journal of Economics 82, 33–
53.
Einzig, P., 1937. The Theory of Forward Exchange. Macmillan, London.
Frenkel, J.A., Levich, R.M., 1975. Covered interest arbitrage: Unexploited profits.
Journal of Political Economy 83, 325–338.
Frenkel, J.A., Levich, R.M., 1977. Transaction costs and interest Arbitrage: Tranquil
versus turbulent periods. Journal of Political Economy 85, 1209–1226.
Gârleanu, N., Pedersen, L.H., 2009. Margin-Based asset pricing and deviations from
the Law of One Price. Working paper, University of California Berkeley and New
York University.
Genberg, H., Hui, C.-H., Wong, A., Chung, T.-K., 2009. The link between FX swaps and
currency strength during the credit crisis of 2007–2008. Hong Kong Monetary
Authority Research Note 01/2009.
Giavazzi, F., 2008. Why does the spread between LIBOR and expected future policy
rates persist, and should central banks do something about it? Vox column
(www.voxeu.org) June 2nd, 2008.
Goyenko, R., Holden, C.W., Trzcinka, C.A., 2009. Do liquidity measures measure
liquidity? Journal of Financial Economics 92, 153–181.
Grossman, S.J., Stiglitz, J.E., 1980. On the impossibility of informationally efficient
markets. American Economic Review 70, 393–408.
Harris, L., 2003. Trading and Exchanges: Market Microstructure for Practitioners.
Oxford University Press, Oxford.
ICAP, 2007. Annual report.
Johnson, T.C., 2008. Volume, liquidity and liquidity risk. Journal of Financial
Economics 87, 388–417.
Jokivuolle, E., 1995. Measuring true stock index value in the presence of infrequent
trading. Journal of Financial and Quantitative Analysis 30, 455–464.
Keynes, J.M., 1923. A Tract on Monetary Reform. Macmillan, London.
Kumar, P., Seppi, D.J., 1994. Information and index arbitrage. Journal of Business 67,
481–509.
Kotomin, V., Smith, S.D., Winters, D.B., 2008. Preferred habitat for liquidity in
international short term interest rates. Journal of Banking and Finance 32, 240–
250.
1107
Mancini Griffoli, T., Ranaldo, A., 2009. Deviations from Covered Interest Parity
during the crisis: A story of funding liquidity constraint. Working Paper, Swiss
National Bank.
Marshall, B.R., Treepongkaruma, S., Young, M., 2008. What level of compensation is
available to arbitrageurs in the foreign exchange market. Working paper,
Massey University and Australian National University.
Mavrides, M., 1992. Triangular Arbitrage in the Foreign Exchange Market:
Inefficiencies, Technology and Investment Opportunities. Greenwood
Publishing Group.
McAndrews, J., Sarkar, A., 2009. Credit risk, liquidity risk and the Federal Reserve’s
response to the crisis. Working paper, Federal Reserve Bank of New York.
Newey, W.K., West, K.D., 1987. A simple, positive semi-definite, heteroskedasticity
and autocorrelation consistent covariance matrix. Econometrica 55, 703–708.
O’Hara, M., Oldfield, G.S., 1986. The microeconomics of market making. Journal of
Financial and Quantitative Analysis 21, 361–376.
Peel, D.A., Taylor, M.P., 2002. Covered interest rate arbitrage in the interwar period
and the Keynes-Einzig conjecture. Journal of Money Credit and Banking 34, 52–
85.
Rhee, S.G., Chang, R.P., 1992. Intraday arbitrage opportunities in foreign exchange
and Eurocurrency markets. Journal of Finance 47, 363–379.
Roll, R., Schwartz, E., Subrahmanyam, A., 2007. Liquidity and the Law of One Price:
The case of future-cash basis. Journal of Finance 62, 2201–2234.
Sarno, L., 2005. Towards a solution to the puzzles in exchange rate economics:
Where do we stand? Canadian Journal of Economics 38, 673–708.
Shleifer, A., Vishny, R.W., 1997. The limits to arbitrage. Journal of Finance 52, 35–55.
Stoll, H., 1978. The supply of dealer services in securities markets. Journal of Finance
33, 1133–1151.
Stoll, H., 2000. Frictions. Journal of Finance 55, 1479–1514.
Taylor, M.P., 1987. Covered Interest Parity: A high-frequency, high-quality study.
Economica 54, 429–438.
Taylor, M.P., 1989. Covered interest arbitrage and market turbulence. The Economic
Journal 99, 376–391.