Aizhan Nurzhanova
Zhanar Tajikulova
AN EMPIRICAL
STUDY OF RISK AND RETURN CHARACTERISTICS OF SWAP SPREAD ARBITRAGE
Fixed-income
arbitrage refers to the market-neutral strategies from a variety of financial
instruments and strategies used by hedge funds and designed to exploit pricing
inefficiencies between the fixed income securities and their derivatives. By
taking offsetting long and short position in the securities, an arbitrageur
reduces exposure to certain market risk and bears the idiosyncratic risk for
which he or she expects to be rewarded.
Swap
spread arbitrage is the simplest, and thereby, the most popular type of
fixed-income arbitrage strategy. According to Gregoriou (2009) “the trading
strategy combines entering a par swap with a fixed-for-floating rate payments
while at the same time shorting a par Treasury bond with the same maturity as
the swap and investing the proceeds at the repo rate”. In essence, an
arbitrageur earns a difference between the long-term swap spread and the spread
of short-term Libor over the repo rates.
Risks
that are relevant to a swap agreement, and therefore, also to the strategy
include the counterparty default risks and risks related to the underlying
floating rate, that is, the Libor rate. While the former has been diminished
significantly by the swap market methods such as using of notional principal,
imposing collateral, and a marking to market, the latter is directly associated
with the average default probability of the bank system which is substantial as has been evidenced
during the previous and especially during the current financial crisis (Li, 2006). In particular, rapid expand of
securitization during the last two decades, which had become a source of solid
profits on the financial market, turned to be the major reason that dried up
liquidity on it during the 2007-2008 turmoil. Lack of external funding and
erosion of bank capital resulted in the imbalance between banks’ assets and
liabilities. Hence, concerned for their own project banks were forced to raise
“precautionary hoarding” which, in chain reaction, led to the troubles in the
interbank lending, and as a sequence, skyrocketed the Libor rate (Brunnermeier,
2009). For instance, the Libor-repo spread was relatively small during the
period prior to the crisis, fluctuating between 10-30 basis points both for the
US and the UK markets. However, during the crisis it widened dramatically,
reaching 230 and 360 basis points for the US and the UK markets respectively.
The highest volatility occurred in September 2008 just after bankruptcy of
Lehman Brothers (see Figure 1).
For
these reasons, Liu and Longstaff (2004b) emphasized that “the fundamental risk
inherent in the swap spread strategy is that of a major financial event or
crisis that causes the Libor-repo spread to widen significantly”. Another risk
noted in the study was erosion of the trade positions as a result of
mark-to-market. This, in turn, triggered margin calls and “haircuts” due to
lack of funding liquidity which was usual during financial recessions.
Nevertheless, they
Figure 1. The Libor rate and General Collateral
Repo rate spread.
This figure plots spreads between a
three-month Libor rate and a three-month general collateral repo rate for the
US and the UK markets data for the period from January 2002 to May 2010. Source: DataStream
study showed that swap spread arbitrage had
sizable excess returns (1-6% per year depending on the horizon of the strategy)
with positive skewness. This contradicted the concept of “arbitrage” that
enjoys small profit most of the time but, occasionally, suffers large losses.
Moreover, these excess returns were sensitive to the equity, credit, and bond
market risk factors, i.e. solely comprised a premium for bearing the financial
market risk. Authors, though, explained these remarkable results by the fact
that the Libor-repo spread was small and stable during the research time
horizon.
The main objective of my dissertation was
to conduct an empirical study of swap spread strategy using both the US and the
UK swap markets data for the period from January 2002 to May 2010. This work
contributes to the existing studies of fixed-income arbitrage strategies by
using, for the first time, the UK market data in addition to the traditional US
market data. The choice of markets was based on the fact that both the US and
the UK had been leading centers for the global hedge fund industry with 68% and
20% of the total at the end of 2009[1].
Moreover, the analysis was applied on new data set that included periods of the
industry rapid growth (2002-2006) as well as its break-neck fall during the
financial crisis (2007-2010).
Thereby, in this work, we challenged to construct the swap
spread arbitrage strategy by replicating the methodology described in Liu and
Longstaff (2004b) because this working paper was a single fundamental study of
the swap spread arbitrage which, later on, had been published as a part of a
comprehensive research about five popular fixed-income arbitrage strategies
conducted by Duarte et al. (2006).
Further, to examine return and risk characteristics of
the swap spread arbitrage strategy we constructed a strategy return index which
was simply series of monthly returns on an equally-weighted portfolio of swap
spread trades. Analysis of the excess returns for index showed that the swap
spread arbitrage is an arbitrage in the textbook sense. In other words, all
strategies over different time horizons resulted in the negative excess returns
ranged from –2% and –3% for a two-year horizon to –12% and –52% for a five-year
horizon, on average, for the US and the UK markets respectively. The excess
return skewness, though, was positive for the US market data and negative for
the UK market one. These outcomes were consistent with obtained results for
actual hedge fund fixed-income arbitrage index returns.
To assess an exposure to probable risk factors, the
excess return for the strategy return indexes were regressed on two groups of
risk factors – financial market factors (equity market risk,
bank sector risk, and bond market risk) and strategy-specific factors
(liquidity risk and default risk). Application of the model to the actual hedge
fund series showed high level of fit (
) for the US
and the UK markets, 97% and 96% respectively, and revealed low explanatory power
of the bank sector risk factor. The latter was also true for the excess returns
on the swap spread arbitrage strategies which, in addition, had no correlation
with the excess returns on the equity market. However, bond market risk factors
as well as the strategy-specific factors were significant for both the US and
the UK market outcomes. It could be explained by the fact that the swap spread
arbitrage strategy is like most of the fixed-income arbitrage strategies viewed
as market neutral; however, it might be exposed to intrinsic risks (Fung and Hsieh, 2004).
To examine the issue another test was conducted. The
theoretical framework is based on the model applied by Mitchell and Pulvino
(2001), where dummy variables were used to assess correlation between risk
arbitrage returns and market returns in normal market conditions and market
downturns. This test indicated that in flat market the beta of the swap spread arbitrage
strategies was insignificant, i.e. equal to zero, suggesting “neutrality” of the
swap spread arbitrage to the market movements. In down market, however, the
beta became statistically significant that supported an aforementioned
suggestion of strategy-specific risk factors inherent in the swap spread
arbitrage.
Overall, these results are controversial in comparison
with findings in recent academic studies in the area. Possible criticism of my
findings relates to the fact that, first; it was difficult to obtain relevant
hedge fund performance data, especially regarding to the UK market data, in
order to control estimated returns on the strategies.[2]
Second, we assumed a linear relationship between swap spread arbitrage returns
and market risk factors that might be inappropriate for a strategy that implies
financial instruments with a non-linear pricing model such as swaps and bonds,
and the last; we used monthly strategy returns, and thereby, could not capture
an exposure to changes that occurred within a month.[3]
On the other hand, no studies have yet been done to
examine performance of fixed-income arbitrage strategies during the credit
crunch. Therefore, more tests have to be conducted to understand in full an
economics of the swap spread arbitrage strategy.
This work provides an empirical test of risk and return characteristics of
the swap spread arbitrage using data from the UK and the US markets. Before
conducting the test, we constructed the index of returns that was obtained as a
result of simulations of the trades involved into the arbitrage.
Thus, in contrast to existing studies of fixed-income arbitrage, current
research showed that the swap spread arbitrage might be very risky strategy
that generates negative excess return during sever financial crisis as in
2007-2008. Such
a poor performance of the swap spread arbitrage strategy during research
horizon, first of all, referred to unpredictable and extremely volatile swap
and Libor-repo spreads. It might be crucial for the strategy that relies on
quantitative models to place trades. Hence, inability of the model to forecast market
“shocks” leads to “unwinding” of trade positions, and as a sequence, to large
losses.
In addition, regression analysis showed that the swap spread
arbitrage strategies was similarly exposed to idiosyncratic risks and neutral
to systematic risk on both the UK and the US markets.
On the other hand, analysis of the strategy return
versus market return illustrated that the swap spread
arbitrage strategy cannot be absolutely “market neutral” during
recessions. Furthermore, significance of liquidity and credit risks for all
strategies together with a consistency of these results with those for the
hedge fund indexes might relate to the fact that in recent time systematic risk
extended as a result of complex network connection both in a hedge fund
industry specifically, and in the financial market in whole.
Overall obtained results revealed “loops” in understanding of the trade
mechanism in the swap spread arbitrage which, in turn, should arise more
research in the future.