Professional Documents
Culture Documents
Jurnal 5
Jurnal 5
Kingsley Fong
David R. Gallagher
Aaron Ng*
ABSTRACT
ABSTRACT
2
1. INTRODUCTION
Recent decades have witnessed a dramatic shift in the nature of risk in global
financial markets, in which the volatility of many asset classes has increased.1 In an
risks, derivatives have become a valuable tool used in the risk management practices
markets have experienced significant growth in both the size and scope of the
securities available to traders.2 These innovations have provided users with a wide
Despite the potential for derivatives to effectively manage the risks faced by
institutions, public opinion concerning these instruments suggests the general public
view derivative securities as inherently dangerous (McGough (1995)). This view has
been argued to have arisen due to the recent high-profile corporate losses that have
been attributed to derivative use, as well as recent episodes of rogue trading scandals.
Barings Bank in the U.K., and the recent foreign exchange episode at National
Australia Bank. These anecdotes highlight the potential hazards of derivative use, and
1
Campbell, Lettau, Malkiel, and Xu (2001) document the upward trend in idiosyncratic volatility in
equity markets. Gross (1997) also notes the increase in the volatility of U.S. bond yields.
2
Seminal papers include Black and Scholes (1973), the binomial option pricing model by Cox, Ross,
and Rubinstein (1979), American option pricing methods by Baroni-Adesi and Whaley (1987), pricing
of compound options by Geske (1979), and the constant elasticity of variance (CEV) model by Cox
(1975).
3
Accounting regulators have formulated standards which require the disclosure of nominal contract
exposures and fair value of derivative instruments in financial accounting statements.
3
Regulation of derivative use in the Australian investment management industry has
not required quantitative disclosure, but rather the design and implementation of risk
Prudential Regulation Authority (APRA) prescribes that derivatives may not be used
for speculative purposes, and funds may not hold uncovered positions in derivatives.
Derivative use must also be conducted within the confines of the trust deed.5
benefits, in the form of reduced transaction costs and improved risk control (Merton
(1995)). Despite these alleged benefits, several studies from both the corporate and
investment management literature (Hentschel and Kothari (2001), Koski and Pontiff
(1999)) have found that the risk and return characteristics of derivative users are
indeed similar to that of non-users. This gives rise to an apparent puzzle of why
observable benefit in terms of either portfolio performance or risk, then one may
question what the purpose of derivatives really is. This intuition also raises further
derivative transactions.
are used by institutional investment managers. The extant literature on derivative use
differences from the use of derivative instruments within the investment management
4
For a detailed discussion of risk management practices in the Australian investment management
industry, see Brown, Gallagher, Steenbeek, and Swan (2005). The governance mechanism is most well
known among superannuation funds, which are required to maintain a Part B Risk Management
Statement (RMS) outlining their response to how risks are monitored.
5
APRA Circular (No. II.D.7)
4
industry.6 There are only a few studies which directly analyse the use of derivatives
by investment managers. Koski and Pontiff (1999) examine the distribution of mutual
fund returns partitioned according to whether derivatives are used or not, and they
Pinnuck (2004) shows that Australian equity managers utilise options to both increase
or decrease exposure to underlying stocks held in the fund, however Pinnuck (2004)
does not attempt to infer the motivation for derivative use by portfolio managers. The
empirical literature has therefore been unable to characterise how derivatives are used
by investment managers, and our study therefore approaches this issue with respect to
equity manager transactions and holdings. Prior studies examining U.S. funds have
end periods, where this data only includes equity holdings and not other instruments
and Pontiff (1999) conduct telephone interviews to determine whether or not mutual
funds are permitted to trade derivatives, however this binary approach is unable to
capture the size of derivative exposures held by funds. While the data used by
Pinnuck (2004) includes the month-end holdings of options, the sample did not
include data showing holdings of futures contracts. Our study offers a significant
improvement to Pinnuck (2004) given our database contains the month-end portfolio
holdings and daily trades of all securities, including derivatives, and also investigates
6
While the motivation to use derivatives as a part of a risk management strategy still exists for
investment managers, derivatives also represent a direct alternative to investment in underlying asset
markets.
7
The portfolio holdings of U.S. institutional investors have been derived from 13F disclosure forms
reported on a quarterly basis to the Securities & Exchange Commission (SEC).
5
In evaluating the impact of derivatives on the returns of active equity
managers, tests are employed to measure the differences in the distributions across a
range of widely used performance and risk measures (see Koski and Pontiff (1999)).
performance or risk between derivative users and non-users. We attribute this result to
the low level of derivative exposure relative to the total size of fund portfolios.
on stocks whose future price movements are expected to follow momentum patterns.
Investment managers are shown to uniformly avoid trading options on stocks with
relatively poor past returns, while some funds also trade disproportionately heavily in
options on stocks with relatively strong past price performance. This provides
empirical support that investment managers indeed use options to capture momentum
returns. We also examine the information content of option trades by active equity
managers and document no abnormal price movements in the underlying stocks over
short-run event windows, both immediately prior and subsequent to option trades. We
interpret these findings as evidence that investment managers do not execute informed
Overall, this study provides evidence that the trading strategies that are widely
known to be implemented in stock markets are also executed using option securities.
instruments, our results confirm previous academic research that the use of derivatives
6
The remainder of the paper is structured as follows. Section 2 summarises the
relevant extant literature and formulates the hypotheses that are addressed in this
hypotheses. Section 4 presents the results and provides a discussion of the key
2. HYPOTHESES
explicit costs of trading (in imperfect markets) and implicit costs such as liquidity-
motivated trading. Conversely, the potential costs of derivative use arise from the
the actual costs and benefits that arise from the use of derivatives.
the underlying share price, they may be used as alternatives in implementing equity
trading strategies. In particular, this study focuses on the momentum strategy outlined
by Jegadeesh and Titman (1993). Previous studies have shown that mutual funds do
sale of poorly performing stocks triggers a capital loss, which may be undesirable if
8
Brown, Harlow and Starks (1996) and Chevalier and Ellison (1997) find that fund flows are
significantly driven by past fund performance, which creates incentives for investment managers to
adjust portfolio risk in order to maximise expected funds-under-management.
9
See Grinblatt, Titman and Wermers (1995) and Carhart (1997).
7
the fund is forced to offset the loss against long-term capital gains.10 Regulations
investment. Also, funds are often mandated to invest a minimum proportion of fund
assets in specified markets.11 This hinders the ability of a fund to sell stocks
Investment managers, through their skills in stock selection and market timing,
are theoretically able to generate private information which it may use to trade and
gain abnormal returns. Prior studies have found mixed evidence of informed trading
desirable to an informed trader. Further, Share Price Index futures provide traders
informed trading.
3. EMPIRICAL METHODOLOGY
3.1 Data
10
In Australia, capital gains tax is only paid on 50% of the total long-term capital gains amount.
Capital losses are offset against total capital gains, which erodes tax benefits from long-term holdings.
11
This floor is commonly set at 90% of a fund’s total net assets.
12
Easley, O’Hara, and Srinivas (1998) show that option volumes lead stock prices, whilst Chakravarty,
Gulen and Mayhew (2004) provide evidence of price discovery in options markets.
8
We study Australian equity funds sourced from the Portfolio Analytics
Database. This unique database, containing month-end portfolio holdings and daily
trading data, was formed on an invitation basis to the largest investment managers
operating in Australia.13 The sample period for this study is 1 January 1993 to 31
December 2003. Further details concerning the construction of the database can be
obtained from Brown et al. (2005), Gallagher and Looi (2005) and Foster et al.
(2005). These studies also show that the sample of funds drawn from the database is
In addition to the managed fund data, daily stock price data and details of
dividends paid are collected from the SEATS system, sourced from the Securities
was used as the market proxy, whilst the Reserve Bank of Australia short interest rate
The initial sample consists of 48 equity funds, across all fund styles. We
sample only actively managed funds that have the discretion to trade in derivatives.
There are no exchange traded derivatives listed on shares of companies outside of the
largest 100. Therefore we exclude index funds, enhanced passive funds and small-cap
equity funds. Of the remaining 34 funds, 50% are classed as derivative users.15 Value
funds accounted for 11 of the total sample funds, of which 3 were derivative users.
security type across the sample of active equity funds. We convert all derivative
13
The size of investment managers was measured by total funds under management.
14
The daily compounded RBA cash rate was converted to a continuously compounded annual rate.
15
A derivative user was defined as a fund that had holdings of derivatives any time during the sample
period.
9
contracts into dollar exposures before computing a fund’s percentage holdings. In
computing the mean values we first calculate, for each fund, the mean number of a
particular type of securities held across month ends, and then we take the mean of
these fund specific values. For instance, in Panel A the low number of call options
relative to the number of stock holdings shows that managers do not hold diversified
portfolios of options. N less than 1 shows that there are months in which funds have
Panel B and Panel C contrast the holdings of derivatives users and non-users
and shows that whilst derivative users still hold large exposures to stocks for which
there are options, the size of exposures to these stocks taken by non-users are smaller
as a proportion of the fund. This suggests that derivative users utilise options as a
direct alternative to taking exposures in stock markets. At any given time, derivative
users hold positions in a small number of different futures contracts, suggesting that
holdings in futures are confined to single maturities. This may also imply that since
futures holdings are limited to few maturities, contracts are rolled-over as they
mature. Across the entire sample, funds trade only in Share Price Index futures, and
do not trade in individual share futures. Funds are on average net short in call options
when measured at the market price, providing some evidence that they are writing call
Table II reports descriptive statistics of the daily trading data for the sample
across all securities. This suggests that managers occasionally trade in significantly
large parcels in stock, futures, and options markets. The number of trades per month
exhibits similar trends to the average trade value, being highly positively skewed,
10
stock, futures, and options markets. The distribution of time to maturity of derivatives
is relatively symmetric. Degree of moneyness shows that option trades are relatively
the median put option trade is slightly in-the-money. The high standard deviation of
addition, funds may be writing out-of-the-money call options, which have a low
income.
These metrics include the variables examined by Koski and Pontiff (1999) and
relative to the fund’s mean return. The first four moments of the fund return
risk relative to market factors. The alpha and beta of each fund are derived from a
single factor model, and idiosyncratic risk calculated as the standard deviation of the
managers to utilise portfolio insurance to protect the portfolio against adverse market
conditions.
16
The single-factor model is estimated from the equation rt − r f ,t = α + β (rMkt ,t − r f ,t ) + ε t , where rt is
the return for month t, r f ,t is the risk-free rate at the end of month t, and rMkt ,t is the ex-post return on
the market during month t. A continuously compounded RBA cash rate is used to proxy for the risk-
free rate, and the logarithmic change in the All Ordinaries Index used as a proxy for the market return.
11
Sortino (2001) stresses the importance of considering the risk of falling below
return.17
Titman (1993) show that stocks exhibit persistent returns based on past 3-to-12 month
performance. Carhart (1997) shows that the persistence of mutual fund performance is
attributable to the 12-month momentum effect. For this reason, this study defines
and Wahal (2002), continue to use momentum windows ranging from 3-to-12 months.
are taken at each month-end of the sample period. Taking these constituents at each
month-end captures the entry of new firms into the market, and removes firms as they
are delisted from the ASX. The 12-month buy-and-hold return is calculated for each
stock at each month-end date, which is then used as the basis of ranking.
1
Downside risk is defined by the equation (R − MAR )2 Pr , where Rt is the return of the fund
17 MAR 2
∑ t t
−∞
during month t , MARt is the minimal acceptable return for month t , and Pr is the probability
associated with observing that level of return. The MAR for each month is set at the return on the All
Ordinaries Accumulation Index
12
Two alternative classifications of momentum are defined. Jegadeesh and
Titman (1993) construct a zero-cost trading strategy to profit from the momentum
effect, which involves buying a portfolio of stocks consisting of the top decile firms
based on past performance, and selling a portfolio of stocks consisting of the bottom
decile of firms. Therefore, stocks in the “Loser” portfolio are firstly defined as firms
in the bottom decile of the performance ranking, while stocks in the “Winner”
portfolio are defined as firms in the top decile of the performance ranking. As an
alternative measure, the “Loser” portfolio is also defined as the bottom quintile of
firms, and the “Winner” portfolio defined as the top quintile of firms.
Individual options trades are then classified as either loser momentum trades,
options on any underlying stock that fell within the loser portfolio as at the previous
month-end date. Similarly, winner momentum trades are option trades on any
underlying stock that fell within the winner portfolio as at the previous month-end
date. Non-momentum trades are all remaining trades which were not considered
from upside exposure to the winner portfolio, and downside exposure to the loser
portfolio. Options provide linear price exposure to either upside or downside price
price exposures. Therefore, the sample of option transactions are separated by both
call option and put option trades, and by buy trades and sell trades. This yields a total
of four trade types. In addition to this, two trade types are defined in order to capture
13
1. Delta increasing trades are transactions that profit from upside price changes
in the underlying stock, and consist of call option buy trades and put option
2. Delta decreasing trades are transactions that profit from downside price
changes in the underlying stock, and consist of call option sell trades and put
The total number of option transactions for each trade type in the loser, winner
and non-momentum categories are calculated for each fund over the entire sample
period. These totals form the basis of the tests of proportions. For each fund, the
distribution of trades over the three categories is compared against the expected
For the goodness-of-fit tests over decile-based momentum trades, the expected
number of trades in winner and loser categories are each equal to 10% of the total
number of option trades, while the expected number of non-momentum trades is equal
to 80% of total option trades. Similarly for the goodness-of-fit tests over quintile-
based momentum trades, the expected number of trades in winner and loser categories
are each equal to 20% of the total number of option trades, while the expected number
Tests of the nominal number of trades will not incorporate relative differences
in trade size, and the level of exposure generated by each trade. For this reason, a
14
The exposure of each trade is calculated as the equivalent ordinary share exposure
The literature approaches the use of options for informed trading at a market
level, by comparing aggregate variables such a prices and volumes between stock and
option markets. Much of this literature has utilised high-frequency intraday data to
establish lead/lag relationships between option and stock markets, finding mixed
support for informed trading using options. Easley, O’Hara and Srinivas (1998) show
that price changes in stock markets lead option volumes with a lag of 20 to 30
minutes, whilst option volumes affect stock price changes more rapidly. Jarnecic
(1999) shows that stock volumes on the ASX lead option volumes on the ASX
Options market. Cairney and Swisher (2004) represent one of few studies which
abnormal volume in Chicago Board Options Exchange (CBOE) options for three days
This study benefits from high granular data comprising the daily transactions
evaluating informed trading are not confined to market level analysis, and therefore
market participants. Importantly, prior studies have found evidence that mutual funds
18
In order to ensure that the chi-squared distribution adequately approximates the discrete sampling
distribution, the goodness-of-fit tests must conform to the rule of five. As a consequence, funds with an
insufficient number of trades were excluded from these tests (less than 50 trades for decile-based tests,
and less than 25 for quintile-based tests). Across all option trade types, 8-to-9 funds were excluded
from the decile based tests, and between 3-to-8 funds were excluded from the quintile based tests.
15
Wermers (2000)). Therefore, this study provides an evaluation of the decision to
examining the behaviour of the share price of the underlying stock, both before and
measure the significance of share price movements around a specific period in time,
Standard event study methodologies typically pool abnormal returns across all
observed events, and test the hypothesis that the mean abnormal return is equal to
zero (Kritzman (1994)). This approach is inappropriate in this study for several
reasons. Firstly, there are many factors influencing the decision by investment
on public information.19 Thus, not all option trades will be motivated by acting upon
private information. Pooling option trades will create significant noise in the event
study model by including trades initiated by different motivations, and may distort the
underlying share price impacts of option trades. Secondly, since options are short-
term trading instruments (given that they are financial contracts with expiry dates),
investment managers are likely implement further trades to reverse initial trading
positions. The timing of this reversal trade will be influenced by factors other than
information motives, such as managing the loss of time value of the option exposure.
This effect also prevents a pooled analysis of investment manager option trades. To
overcome these issues we study the underlying share price behaviour around option
16
To implement the event study, daily abnormal returns are calculated from 10
days prior to the option trade, to 10 days subsequent to the option transaction. For
each day, the market-adjusted abnormal return is calculated as the difference between
the daily simple compounded return on the stock minus the contemporaneous daily
return on the All Ordinaries Accumulation Index.20 Cumulative abnormal returns for
the post-event windows are then formed by summing the abnormal returns across
k
CARt ,t + k −1 = ∑ ( Rt + j −1 − RMkt ,t + j −1 ) (2),
j =1
k = 1, 3, 5, 10.
Cumulative abnormal returns for the pre-event windows are formed exclusive of the
k
CARt −k ,t −1 = ∑ ( Rt − j − RMkt ,t − j ) (3),
j =1
k = 1, 3, 5, 10.
For each event date, we estimate the standard deviation of returns as the
square root of the sample variance of returns from t − 100 to t − 11 , where t is the
20
Daily stock returns are adjusted for dividends and capital distributions.
17
day of the option trade. The estimation of the standard deviation prior to the event
window ensures that estimates of the standard error are not biased by the effects of the
event. The standard error for each event window is then calculated as:
where s.e.(CARt ,k ) = the standard error on day t for a k -day event window.
3, 5, and 10.
For option trade, the test statistics for post-event windows are calculated as:
CARt ,t + k −1
Z= (5),
s.e.(CARt ,k )
CARt − k ,t −1
Z= (6).
s.e.(CARt ,k )
This test statistic follows a standard normal distribution, and tests the null hypothesis
4. RESULTS
18
[Insert Table III here]
Table III presents the results for the difference in cross-sectional mean tests
for the distributional moments and risk measures. Broadly, these tests do not identify
any significant differences in the performance or risk between funds that use
Panel A of Table III shows that during the sample period, active equity funds
yield positive mean returns that are negatively skewed and leptokurtic. This is
consistent with previous studies that have found similar non-normality in the return
markets (Bird and Gallagher (2002)). Across the entire sample of funds, derivative
users display insignificantly higher mean returns, and insignificantly lower standard
deviation, skewness and kurtosis. In the subsets of fund styles, Growth and Growth-
At-a-Reasonable Price (GARP) funds show the same trends as the full sample. Value
funds using derivatives yield insignificantly lower mean returns than those that did
not use derivatives, and otherwise follow the trends found in the full sample. The
behaviour of Style Neutral funds opposed the trends of the full sample, with
standard deviation, and were more negatively skewed and platykurtic. Therefore,
whilst these results are of the correct sign to suggest that derivatives improve
performance and reduce risk, they are of insufficient magnitude to conclude these
Panel B of Table III shows that on average, funds display positive alphas.
Growth and GARP funds that did not use derivatives are the only group to derive
negative alphas. There is weak evidence indicating a difference in the mean alpha
between derivative users and non-users for Growth and GARP-style funds, with the
19
Wilcoxon rank-sum test significant at the 10% level. For Value and Style Neutral
funds, derivative users are documented as having insignificantly lower alphas. Higher
betas (measuring systematic risk) are found for Growth, GARP, and Style Neutral
funds that used derivatives, and lower mean beta for Value funds using derivatives.
Idiosyncratic risk is insignificantly lower for derivative users across all fund styles,
with the exception of Style Neutral funds. Timing beta is found to be negative across
coefficients. GARP funds that did not use derivatives displayed a positive mean
timing beta, though insignificantly different to derivative users. Style Neutral funds
that use derivatives also display a positive mean timing beta, which is also
smaller for derivative users across all fund styles except the Style Neutral category.
evidence that the dispersion of the performance and risk parameters between
portfolios that include and exclude derivative securities are different. Panel A of
Table IV reveals the standard deviation of all distributional moments are lower for
derivative users across all funds. The standard deviation of the mean is significantly
different in the test of all equity funds, however, the difference is only weakly
significant for GARP funds and insignificant for all other fund styles. This suggests
that the observed differences across the aggregate sample are attributable to cross-
sectional differences in fund style. The dispersion of the mean is significantly smaller
21
Cross-sectional standard deviation is computed by taking the standard deviation of the moment of
interest. Note that the standard deviation of mean here is different from the mean standard deviation in
Table III.
20
across all funds, which is largely driven by a lower dispersion of derivative users in
the Value funds category. GARP and Style Neutral funds also suggest evidence of
lower dispersion. The standard deviation of skewness is lower for derivative users
across all funds, but similar to the dispersion of the mean, is likely caused by
differences in fund style rather than derivative use. The dispersion of kurtosis is lower
lower for derivative users across Growth, GARP and Value funds, and insignificantly
high for Style Neutral derivative users. The significant difference in the standard
deviation of all funds is not robust after controlling for fund style. The standard
deviation of beta is mixed across fund styles, being lower for derivative users in
Growth and Style Neutral funds, and higher in GARP and Value funds. The
dispersion of idiosyncratic risk is significantly lower for All Funds and Growth funds
that use derivatives, and significantly higher for derivative users in the Style Neutral
category. Timing beta exhibits lower standard deviation for all fund styles except for
Style Neutral funds. Downside risk shows a generally lower dispersion for derivative
distributional moments and risk measures. This result is consistent with the findings
of Koski and Pontiff (1999), who conclude that derivative use is not cross-sectionally
related to risk exposure and the higher moments of fund return distributions. We find
differences in the means of the parameters across fund style, but find no evidence of
significant differences between derivative users and non-users after controlling for
21
fund styles. We conclude that derivative use, on average, has no observed impact on
The discussion for the descriptive statistics of fund holdings in section 4.1.3
observes that the size of derivative exposures is small as a proportion of the total
users, and mean options exposures contributing to less than 1% of portfolio exposure.
This presents a likely reason for the lack of observed impact of derivative use on
performance and risk. Since investment managers are not engaging heavily in
derivative exposures, the effects of derivative use are masked by analysis of aggregate
The results of Table IV add depth to the inference drawn from the analysis of
the mean values of performance and risk parameters in Table III. The use of
derivatives reduces the level of cross-sectional dispersion in each of the first four
moments of the distribution of returns. This suggests that active investment managers
are using derivatives for common purposes, causing the returns of these funds to
employing derivatives for hedging. Derivative use causes fund returns to be more
dispersion of timing beta for derivative users. This indicates that the use of derivatives
allows funds to respond to changes in the market, and consequently, positive market
movements have less impact on the fund’s market timing ability. As an alternative
explanation, the literature has established that fund flows initiate liquidity-motivated
trading, which creates a drag on fund performance (Edelen (1999)). During periods of
strong stock market performance, funds would experience large inflows. The use of
22
derivatives may assist in equitizing fund flows, which would reduce the impact of
lower dispersion of downside risk indicates that growth funds may be using
Table V presents the results of the chi-squared goodness-of-fit tests over the
frequently in options over stocks in loser portfolios. There is evidence of some funds
Panel A reports the empirical findings where momentum portfolios are defined
by upper and lower return deciles of the All Ordinaries Index. Across all option trade
types, there is evidence that active equity funds trade disproportionately in option
funds across all option trade types. This shows that the observed number of trades in
particular, active equity managers avoid exposure to the loser portfolio, as shown by
the shortfall in the observed proportion of trades in this category below the expected
proportion. The median proportion of trades in the loser portfolio is less than 2% for
all trade types. For call option trades, the median proportion of option trades in the
23
winner portfolio was less than the uniform expectation, suggesting that overall funds
are also avoiding upward momentum exposure. However, the maximum proportion of
buy trades in call options over the winner portfolio was 18.4%, signifying that some
funds use options to generate exposures to winner momentum stocks. In contrast, the
small proportion of put option trades over winner stocks shows that these options are
Panel B presents the results of the momentum portfolios which are defined by
upper and lower quintiles. These results provide stronger evidence that active equity
of-fit tests are significant for substantially all of the sample funds. This is largely due
to funds trading lower proportions of options over loser stocks. In contrast to the
decile based results, the majority of funds trade disproportionately in call options over
winner stocks, representing around 30% of the median fund’s trades. Furthermore, at
the maximum, funds place close to half of their call option trades over stocks in the
winner portfolio. The same phenomenon is less acute in put option securities, where
the median proportion of fund trades is close to the expected 20%, and the maximum
Table VI reports the results of the chi-squared goodness-of-fit tests over the
portfolios. These tests incorporate the relative size of each option trade, measuring the
value of each transaction as the equivalent ordinary share exposure. These results
24
confirm the findings of the equally-weighted momentum tests in Table V.
categories.
Panel A of Table VI shows that the value of option trades over stocks in loser
portfolios is substantially lower than expected, and to an even greater degree than for
the results of the equally-weighted proportions. Some funds are shown to completely
across call option trades, and for put option sell trades. The median value-weighted
proportion of trades over winner stocks is lower than the equally-weighted proportion,
providing further evidence that funds are, on average, also avoiding options exposure
to winner stocks. Further, there is evidence that some funds acquire exposure to the
winner portfolio using option securities, with the maximum proportion of trades over
winner stocks exceeding the expected value of 10%. Funds unanimously trade
disproportionately less in put options over winner stocks. This is consistent with the
findings of Pinnuck (2004), which show that funds prefer to hold stocks that are past
winners. Therefore, investment managers would not require the right or the obligation
to sell the stocks in their portfolio with strong past performance, making put options
equally-weighted trade proportions, but are much more pronounced. For call option
trades, funds both under-invest in options over loser stocks, and over-invest in options
over winner stocks. The maximum proportion of loser trades is 10%, while the
maximum proportion of winner trades is 60-80%. The table clearly shows that funds
trade disproportionately higher volume in winner stocks. Put option trades show no
such effects, with the median fund trading close to the expected values.
25
4.2.3. Interpretation of the results
Overall, the goodness-of-fit tests reveal that funds trade relatively infrequently
in options over stocks with poor past performance, and relatively frequently in options
over stocks with strong past performance. These results may be interpreted as
showing that equity funds use options to gain exposure to “winner” momentum
with a momentum trading strategy which requires taking long exposure in stocks with
does not show signs of the opposing side of the momentum strategy that prescribes
taking short exposure in stocks with poor past performance. That is, the option trades
Although these results are consistent with momentum trading, they may also
be symptomatic of other forms of trading behaviour. As such, whilst the results are a
One aspect which is not captured by these tests is the timing of option trades.
The timing of option trades is important in indicating the purpose of a particular trade.
That is, an option trade of opposite direction to previous trades in the same option
security may be performed to reverse out an exposure that the fund currently holds.
Since multiple trades are required to conduct a round trip into and out of an exposure
to an option security, this would inflate the relative proportion of trades over
particular stocks.
26
Another aspect that is not captured by the level of trading activity by fund
managers is the relative level of exposure to a particular stock which is currently held
in the portfolio. Jegadeesh and Titman (1993) show that the momentum anomaly is
to the momentum portfolios, and not trade continuously across these stocks. In these
circumstances, a test of trading activity will not consider the ongoing level of
clearly identify a systematic trend that the options trades of investment managers are
Table VII presents the results of the event study examining the impact of
investment manager option trades on underlying stock returns. These results show that
the options trades of investment managers are not as a response to, or do not trigger
abnormal share price reaction in the underlying stock. This suggests that the market
The results shown in Table VII are the percentiles of the t-statistics for the
cumulative abnormal returns (CARs) in the underlying stock calculated for each
the information will quickly flow from the option price through to the price of the
27
underlying stock, driving abnormal stock returns around an option transaction. The
results in Table VII do not show such reaction. Whilst many of the 1st percentile and
99th percentile of the t-statistics are significant at the 1% level, this is not indicative of
abnormal share price reaction. The significance of the t-statistic reveals that the
only occurs for less than 1% of the observations, the effects of an option transaction
are equivalent to a random stock price change, therefore this does not signify an
abnormal stock return. To demonstrate that some proportion of option trades contain
of option trades substantially greater than the level of significance. We do not find
such evidence.
the sign of CARs for particular option trade types. Call option buy trades exhibit a
consistent positive mean CAR prior to an option trade, and a negative mean CAR
subsequent to an option trade. Call option sell trades display consistently positive
mean CARs during both pre-event and post-event windows. Put option buy trades
show a general trend of positive CARs around option trades, with only negative CARs
seen 3-to-5 days before the event. Put option sell trades show a consistently negative
CAR for all event windows, except for the 10 day post-event window. CARs around
Delta Increasing trades are negative immediately prior to and following and option
trade, and positive otherwise. CARs around Delta Decreasing trades are consistently
The results of the event study suggest that there is no informed trading in the
options trades by active equity managers. This is a lack of observable abnormal share
28
price reaction around option trades. This result is consistent across all option trade
types.
At first instance, it appears that this result conflicts with some of the results
found in the prior literature (Easley, O’Hara, and Srinivas (1998), Chakravarty,
Gulen, and Mayhew (2004)). Previous studies have provided mixed support for the
operation of informed trading in options markets. Whilst it should be noted that the
issue of whether informed trading occurs in options markets is still unresolved, the
bulk of this research has been conducted on a market level. To our knowledge, this is
institutional trading data. Thus, while some degree of informed trading may be
performed using options, the results of this study suggest that investment managers do
implies that the informed trading in options markets found by prior studies is
conducted by traders other than active equity managers. Therefore, the findings of this
study complement the findings of the market level informed trading research (Easley,
O’Hara, and Srinivas (1998), Chakravarty, Gulen, and Mayhew (2004)) by identifying
a specific group of institutions that do not use options for informed trading.
The sign of CARs provides further insight into the trading behaviour of
investment managers. For call option buy trades, the mean CARs are steadily
decreasing from 10 days prior until the day of the event. The abnormal share price
return on the day of the call purchase is the largest negative mean return around the
period. The CARs steadily increase from the day after the trade. This suggests that
funds may be able to locate profitable opportunities to purchase call options during
days in which the share price has fallen. The behaviour after the trade may suggest a
29
stable correction in the market. Likewise, the mean abnormal return on the day of a
call option sell trade is positive, and largest in the days around the event. Investment
managers may be selling call options following a rise in the share price during the
day, since it would be seen as less likely to rise further and into a loss position for the
fund. The mean contemporaneous CAR for put option buy trades is positive, which
would be consistent with funds buying put options following a share price rise during
the day on the expectation of a future share price fall. Put option sell trades show an
opposite effect in the underlying share price, and is therefore consistent with selling
put options in expectation of a future share price rise. Overall, this contrarian
in participating in the option market than simply a pure risk management or informed
trading explanation.
5. CONCLUSION
This study uses a unique and confidential database comprising the monthly
holdings and daily trades of investment managers to study derivatives use by active
risk across derivative users and non-users. This suggests that derivatives use by
investment managers does not achieve higher returns, or reduce portfolio risk relative
and risk measures provides evidence that derivatives improve managers’ ability to
30
control fund risk, and leads to more homogeneity in fund returns. The small size of
derivative exposure relative to fund size limits the impact of derivatives on aggregate
fund portfolios.
options on stocks with relatively poor past price performance. Some funds are also
found to trade heavily in options on stocks with relatively strong past performance.
with prior research on fund manager stock trades (Grinblatt, Titman, and Wermers
(1995)).
term windows around the option trades of active investment managers. The lack of
market reaction suggests that options are not used to execute informed trades.
However this result is consistent with managers using option trading, particularly call
31
REFERENCES
Bird, R., Gallagher, D. R., 2002. The evaluation of active manager returns in a non-
Black, F., Scholes, M., 1973. The pricing of options and corporate liabilities. Journal
Brown, S. J., Gallagher, D. R., Steenbeek, O., Swan, P. L., 2005. Double of nothing:
Brown, K. C., Harlow, W. V., Starks, L. T., 1996. Of tournaments and temptations:
Cairney, T., Swisher, J., 2004. The role of the options market in the dissemination of
1041.
Campbell, J. Y., Lettau, M., Malkiel, B. G., Xu, Y., 2001. Have individual stocks
52, 57-82.
Chakravarty, S., Gulen, H., Mayhew, S., 2004. Informed trading in stock and option
32
Chevalier, J., Ellison, G., 1997. Risk taking behaviour by mutual funds as a response
Cox, J. C., 1975. Notes on option pricing 1: Constant elasticity of variance diffusions.
Cox, J. C., Ross, S. A., Rubenstein, M., 1979. Option pricing: A simplified approach.
Daniel, K., Grinblatt, M., Titman, S., Wermers, R., 1997. Measuring mutual fund
1035-1058.
Demir, I., Muthuswamy, J., Walter, T., 2004. Momentum returns in Australian
equities: The influences of size, risk, liquidity and return computation. Pacific-
Easley, D., O’Hara, M., Srinivas, P. S., 1998. Option volume and stock prices:
Edelen, R. M., 1999. Investor flows and the assessed performance of open-end mutual
Foster, F. D., Gallagher, D. R., Looi, A, 2005. Institutional trading and share returns,
Gallagher, D. R.; Looi, A., 2005. Trading behaviour and the performance of daily
Geske, R., 1979. The valuation of compound options. Journal of Financial Economics
7, 63-81.
Grinblatt, M., Titman, S., Wermers, R., 1995. Momentum investment strategies,
33
Gross, William H., 1997, Bill Gross on investing, John Wiley & Sons, Inc., New
York.
Hentschel, L., Kothari, S. P., 2001. Are corporations reducing or taking risks with
Jarnecic, E., 1999. Trading volume lead/lag relations between the ASX and ASX
Jegadeesh, N., Titman, S., 1993. Returns to buying winners and selling losers:
Koski, J. L., Pontiff, J., 1999. How are derivatives used? Evidence from the mutual
Kritzman, M. P., 1994. About event studies. Financial Analysts Journal 50, 17-20.
McGough, R. C., 1995. SEC seeks aid in devising yardsticks for funds’ risk. The Wall
Merton, R. C., 1995. Financial innovation and the management and regulation of
Moskowitz, T. J., 2000. Discussion of Wermers. 2000, Journal of Finance 55, 1695-
1703.
Pinnuck, M., 2004. Stock preferences and derivative activities of Australian fund
Sortino, F. A., 2001. From alpha to omega, in: Sortino, F. A., Satchell S. E. (Eds.),
stock-picking talent, style, transactions costs and expenses. Journal of Finance 55,
1655-1695.
34
Table I
Descriptive Statistics of Fund Holdings
This table provides descriptive statistics of the month-end portfolio holdings sourced from Portfolio Analytics Database for
the period 1 January 1993 to 31 December 2003. Exposures are reported as a proportion of fund size. Contract exposure is
measured as the number of securities multiplied by the market price of the securities. The measurement of delta exposure
differs by security. For stocks, delta exposure is equal to contract exposure. For futures, delta exposure is equal to (number
of contracts x futures price x delta of the future). For options, delta exposure is measured as (number of option contracts x
number of shares per contract x spot stock price of the underlying asset x delta of the option). Figures are reported for net
holdings, long only, and short only positions. Stocks (All) refers to exposures to all holdings of stocks. Stocks (sub) refers
to the subset of stocks over which options were traded. Panel A reports statistics for the full sample of funds, Panel B
reports statistics for funds that use derivatives, and Panel C reports statistics for funds that do not use derivatives.
35
Table II
Descriptive Statistics of Fund Trades
This table provides descriptive statistics of the daily trades sourced from the Portfolio Analytics Database
for the period 1 January 1995 to 31 December 2002. For stocks and options, trade value refers to the
number of securities traded x the price per security. For futures, trade value refers to the number of
contracts trade x the SPI futures value x $25. Degree of moneyness is measured as the spot price of the
underlying asset minus the strike price of the option, divided by the spot price of the underlying asset.
Panel A - Stocks
Mean Std. Dev. Median
Average trade value ($) 564,667 2,405,018 185,429
Number of trades per month 117 125 81
Proportion of buy trades 57.01% 20.74% 55.56%
Panel B - Futures
Mean Std. Dev. Median
Average trade value ($) 757,198 3,721,481 39,720
Number of trades per month 33 54 19
Proportion of buy trades 51.29% 24.39% 50.00%
Time to maturity (Days) 35.63 23.74 33.42
36
Table III
Location of Distributional Moments and Risk Measures
This table reports the mean of distributional moments and risk measures of fund returns. Panel A reports the mean, standard
deviation, skewness, and kurtosis for the full sample of funds, then separated by derivative users and non-users. Panel B reports the
alpha, beta, idiosyncratic risk, timing beta, and downside risk of fund returns. Tests of Differences tests the null hypothesis that the
mean estimates are equal for nonusers and users of derivatives. The t test is a parametric test, whilst the Wilcoxon rank sum test is a
37
Table IV
Dispersion of Distributional Moments and Risk Measures
This table reports the standard deviation of distributional moments and risk measures of fund returns. Panel A reports the
mean, standard deviation, skewness, and kurtosis for the full sample of funds, then separated by derivative users and non-
users. Panel B reports the alpha, beta, idiosyncratic risk, timing beta, and downside risk of fund returns. F-Test is a
comparison of variance of the parameters between nonusers and users of derivatives. Results for each parameter are also
38
Table V
Momentum by Proportion of Trades
This table reports descriptive statistics on the proportions of investment manager trades in options over stocks
exhibiting momentum and non-momentum behaviour. Chi-squared is a goodness-of-fit test on the number of option
trades in the Loser, Non-momentum, and Winner categories. In Panel A, loser and winner stock portfolios are defined
as the lower and upper deciles based on past 12 month buy-and-hold returns. The expected proportions of loser, non-
momentum, and winner trades are 10%, 80% and 10% respectively. In Panel B, loser and winner stock portfolios are
defined as the lower and upper quintiles based on past 12 month buy-and-hold returns. The expected proportion of
loser, non-momentum, and winner trades are 20%, 60%, and 20% respectively. Tests are performed by fund.
39
Panel B - Momentum on Upper and Lower Quintiles by Performance
% Non-
No. of Trades % Loser Momentum % Winner Chi-squared p-value
Call Option Buy Trades
Mean 358.87 5.81% 64.72% 29.47% 69.49 0.07
Median 119.00 3.68% 66.96% 30.80% 26.72 0.00
Min 25.00 0.00% 47.37% 10.00% 0.08 0.00
Max 2950.00 18.75% 80.00% 50.00% 532.89 0.96
No. significant @ 10% 14
No. significant @ 5% 14
Total funds in sample 15
Call Option Sell Trades
Mean 431.67 6.18% 63.35% 30.47% 78.51 0.01
Median 200.00 2.56% 65.04% 31.67% 28.77 0.00
Min 26.00 1.09% 38.46% 17.39% 5.03 0.00
Max 3044.00 30.77% 76.09% 45.71% 512.22 0.08
No. significant @ 10% 12
No. significant @ 5% 11
Total funds in sample 12
Put Option Buy Trades
Mean 260.00 7.89% 74.18% 17.92% 37.54 0.06
Median 113.00 4.53% 74.39% 19.86% 15.16 0.00
Min 39.00 0.00% 61.95% 7.69% 1.97 0.00
Max 972.00 23.01% 92.31% 26.92% 155.88 0.37
No. significant @ 10% 6
No. significant @ 5% 6
Total funds in sample 7
Put Option Sell Trades
Mean 321.33 5.43% 73.47% 21.10% 53.98 0.02
Median 166.50 3.11% 73.24% 21.38% 22.44 0.00
Min 59.00 1.69% 69.31% 13.86% 3.87 0.00
Max 1218.00 16.83% 79.31% 27.78% 209.21 0.14
No. significant @ 10% 5
No. significant @ 5% 5
Total funds in sample 6
Delta Increasing Trades
Mean 490.73 5.42% 64.51% 30.07% 91.06 0.07
Median 178.00 3.51% 67.57% 26.89% 29.40 0.00
Min 27.00 0.00% 44.90% 14.05% 0.18 0.00
Max 4168.00 18.18% 78.57% 52.54% 730.92 0.91
No. significant @ 10% 13
No. significant @ 5% 13
Total funds in sample 15
Delta Decreasing Trades
Mean 506.29 6.32% 65.47% 28.21% 83.60 0.02
Median 170.50 3.35% 67.53% 27.58% 29.17 0.00
Min 27.00 0.00% 40.74% 7.69% 4.17 0.00
Max 4016.00 29.63% 92.31% 51.11% 634.09 0.12
No. significant @ 10% 12
No. significant @ 5% 12
Total funds in sample 14
40
Table VI
Momentum by Value-weighted Proportion of Trades
This table reports descriptive statistics on the value-weighted proportions of investment manager trades in options over
stocks exhibiting momentum and non-momentum behaviour. The number of trades are weighted by the value of
equivalent ordinary share exposure of the trade. Chi-squared is a goodness-of-fit test on the value-weighted number of
option trades in the Loser, Non-momentum, and Winner categories. In Panel A, loser and winner stock portfolios are
defined as the lower and upper deciles based on past 12 month buy-and-hold returns. The expected proportions of loser,
non-momentum, and winner trades are 10%, 80% and 10% respectively. In Panel B, loser and winner stock portfolios
are defined as the lower and upper quintiles based on past 12 month buy-and-hold returns. The expected proportion of
loser, non-momentum, and winner trades are 20%, 60%, and 20% respectively. Tests are performed by fund.
41
Panel B - Momentum on Upper and Lower Quintiles by Performance
Total Value-
Weighted
Number of % Non-
Trades % Loser Momentum % Winner Chi-squared p-value
Call Option Buy Trades
Mean 358.87 3.84% 60.59% 35.57% 104.40 0.00
Median 119.00 2.94% 62.14% 33.90% 26.43 0.00
Min 25.00 0.00% 39.16% 8.80% 8.71 0.00
Max 2950.00 10.49% 86.97% 58.43% 519.97 0.01
No. significant @ 10% 11
No. significant @ 5% 11
Total funds in sample 11
Call Option Sell Trades
Mean 431.67 3.40% 61.14% 35.46% 102.67 0.00
Median 200.00 3.27% 66.08% 32.57% 57.63 0.00
Min 26.00 0.00% 20.14% 10.20% 7.73 0.00
Max 3044.00 10.35% 85.72% 79.86% 541.64 0.02
No. significant @ 10% 11
No. significant @ 5% 11
Total funds in sample 11
Put Option Buy Trades
Mean 260.00 6.57% 78.77% 14.65% 48.32 0.01
Median 113.00 3.06% 78.07% 16.56% 18.85 0.00
Min 39.00 0.00% 69.61% 4.08% 6.17 0.00
Max 972.00 24.47% 95.92% 25.01% 177.48 0.05
No. significant @ 10% 6
No. significant @ 5% 6
Total funds in sample 6
Put Option Sell Trades
Mean 321.33 6.55% 73.64% 19.81% 62.69 0.18
Median 166.50 3.18% 76.96% 19.86% 32.84 0.00
Min 59.00 2.56% 60.78% 12.88% 0.17 0.00
Max 1218.00 18.37% 84.04% 28.06% 222.63 0.92
No. significant @ 10% 4
No. significant @ 5% 4
Total funds in sample 5
Delta Increasing Trades
Mean 490.73 0.87% 84.36% 14.77% 131.92 0.00
Median 178.00 0.22% 89.47% 10.31% 39.81 0.00
Min 27.00 0.00% 42.26% 0.00% 8.41 0.00
Max 4168.00 4.68% 100.00% 55.05% 719.51 0.01
No. significant @ 10% 11
No. significant @ 5% 11
Total funds in sample 11
Delta Decreasing Trades
Mean 506.29 0.56% 88.27% 11.16% 114.77 0.00
Median 170.50 0.21% 93.47% 6.28% 54.11 0.00
Min 27.00 0.00% 41.32% 0.00% 8.43 0.00
Max 4016.00 2.89% 100.00% 58.68% 671.38 0.01
No. significant @ 10% 12
No. significant @ 5% 12
Total funds in sample 12
42
Table VII
Event Study of Informed Trading
This table reports the mean and percentiles of the t -statistics of the cumulative abnormal returns for underlying stocks in event
windows around investment manager option trades. For cumulative abnormal returns, the numbers in the brackets represent
the period of the event window. The first number in the brackets is the day in which the event window begins, and the second
number is the day in which the event window ends. These days are relative to an option trade event, set at day 0.
43
APPENDIX – CALCULATION OF EQUIVALENT ORDINARY SHARE
EXPOSURES
Since derivative exposures contain leverage, the cost of the option does not
represent the level of exposure to changes in the underlying stock price. Thus,
measuring the level of exposure by the market price of the derivative securities will
significantly underestimate the true level of exposure to the underlying stock. Due to
the availability of data from accounting disclosures, research into corporate derivative
use has typically measured the level of exposure by the notional principal of the
contracts (see Geczy, Minton and Schrand (1997), Hentschel and Kothari (2001)).
Whilst this measure incorporates the leverage of derivative securities, it does not
capture the price sensitivity relative to the underlying asset, and is therefore an
involves replacing the value of options exposures with the equivalent holding of
ordinary shares which would yield the same level of share price exposure. In this
paper, a similar approach is taken and expanded across both options and futures
holdings.
underlying stock that would result in the same change in value following a price
change in the underlying stock. For derivatives, this price sensitivity is measured by
delta, the rate of change of the derivative price with respect to a change in the
∂D
∆=
∂P
44
∂D = an instantaneous change in the price of the derivative
Futures contracts
price of the underlying asset results in an immediate gain to the futures holder.
Consequently, Hull (2003) shows that the delta of a futures is equal to:
∆ FUTURE = e (r − q )T
T = Time to maturity
The funds in this sample confine their trading in futures to Share Price Index (SPI)
futures, and do not trade futures over individual shares. Thus, the delta of these
futures contracts is e rT , since the SPI does not pay a yield. The equivalent ordinary
Option contracts
∆ CALL = N (d1 )
∆ PUT = N (d1 ) − 1
22
Share Price Index futures on the Sydney Futures Exchange (SFE) are priced at $25 per point. The
futures price is quoted as an index value.
45
ln (S 0 / K ) + (r + σ 2 / 2)T
where d 1 =
σ T
T = Time to maturity
market are American style exercise. Furthermore, stocks on the ASX typically pay
relatively high dividend yields.23 As a result, the Black-Scholes solutions for delta
may no longer be applied. Whilst methods have been found to adjust the Black-
Scholes model for early exercise (Black (1975)), no closed form solution can be found
for pricing American put options. These options can be valued using numerical
procedures. In this paper, we employ a 50-step binomial option pricing model from
Cox, Ross and Rubenstein (1979). The value of the ith node of the tree is adjusted for
S 0*u j d i − j + De − r (τ −iδt ) , j = 0, 1, …, i
S 0*u j d i − j , j = 0, 1, …, i
23
In a study of gross dividends per share across major stocks markets around the world, Australia
ranked ninth with an average dividend yield of 3.4% per annum. In comparison, the United States
ranked 40th, with an average dividend yield of 1.4%. Source: Wren Research, August 2002.
46
D = the dollar amount of the dividend
u = eσ δt
d = 1
u
In this model, both the timing and amount of dividends are assumed to be
known with perfect foresight. Furthermore, the volatility of the underlying stock is
approximated by annualising the standard deviation of the daily stock returns for the
past 30 trading days.24 Finally, delta is calculated from this model as:
Pu ,1 − Pd ,1
∆ OPTION =
S 0*u − S 0* d
where Pu ,1 = Value of the option on the “up” change of the first node
calculated as:
No. of options contracts × No. of shares per contract × Underlying stock price × ∆ OPTION 25
delta measures the price sensitivity of the option price at the current spot stock price.
The value of delta will change over time, and with changes in the spot stock price.
24
It should be noted that the correct measure of volatility used to value options is expected volatility.
The market consensus of expected volatility is observable through calculating implied volatility.
However, this measure is only observable at the precise time in which a trade occurs, and only applies
at the current stock price. Since options are thinly traded securities, and we require valuation as at each
month-end in the sample period, implied volatility is unable to be computed. The implied volatilities
computed as at each month-end were suffer significant misspecification due to stale option prices.
Hence, we utilise a measure of historical volatility.
25
Exchange-traded options on the ASXD are written over 1000 shares, however this is adjusted in the
event of a capital distribution.
47
Secondly, delta is an instantaneous measure of price sensitivity. The level of exposure
computed by delta will only apply to small changes in the stock price.
48