exploiting volatility movements in the sydney futures exchange's bank bill contract
TRANSCRIPT
EXPLOITING VOLATILITY MOVEMENTS IN
THE SYDNEY FUTURES EXCHANGE’S BANK
BILL CONTRACT
B. F. HUNT, R. BHAR
ABSTRACT
An appropriate stochastic model was fitted to one year of data on the implied volatility of options
on 90 day bank accepted bill futures contracts traded in the Sydney Futures Exchange. The
model used was ARIMA augmented with day of the week variables, an option time to maturity
variable, and recent values of historic volatility. The high ex-post predictive accuracy of the
model was then employed as the central element of a strategy of buy low/sell high volatility.
We employed two trading schemes with suitably constructed Delta neutral portfolios com-
prising bill futures and call and put options on those futures over a period of six months, to test
whether speculative trading profit could be earned. The existence of trading profits before
transaction costs validated the potential of the buy low/sell high volatility strategies to generate
speculative profits. The absence of any such trading profits after transaction costs however,
showed that the market pricing of these securities is such that the dependencies within implied
volatility cannot be profitably exploited.
This result may be interpreted as evidence supporting an hypothesis of a semi-strong form
of market efficiency.
Direct all correspondence to: R. Bhar, Faculty of Business andTechnology, University of Western Sydney, Box 555, Can&&own, N.S.W., 2560, AUSTRALIA. l B. F. Hunt, University of Technology- Sydney, AUSTRALIA.
International Review of Economics and Finance, 2(4) 403-415 Copyright 0 1993 by JAI Press, Inc. ISSN: 1059-0560 All rights of remoduction in any form rese.rved
403
404 B. F. HUNT and R. BHAR
INTRODUCTION
This paper examines the efficiency of the bank Accepted Bill Futures market as traded in
the Sydney Futures Exchange @FE). The definition of efficiency applied here relates to the
possibility of obtaining abnormal profits by the appropriate use of available information.
The proposition of semi-strong market efficiency is that the market disseminates publicly
available information so rapidly that the prices already reflect this and no abnormal profit
can be earned. Thus, any publicly available information which can be used to produce
abnormal profits invalidates the proposition of semi-strong form market efficiency. Our
focus here is upon the information pertaining to volatility.
The growth of trading in Australian interest rate contracts has been rapid since their
introduction in 1979. While the Australian market is small in relation to world markets the
development of an Australian financial futures market has mirrored that of other countries.
Market practice in the SFE is closely related to that in the U.S. futures markets. While the
SFE was chosen as the focus of this study, the techniques used here, and perhaps the adduced
results, are directly applicable to other markets.
The SFE is the principal Australian interest rate hedge market. While there are two other
interest rate futures contracts, the Bank Accepted Bills contract is the most traded and liquid
contract.’ The SFE Bank Bill contract is for the delivery of 90 day Bank Accepted Bills to
the face value of AUD500,OOO. The contracts are for delivery on the second Friday of March,
June, September, and December. Both call and put options are traded on these futures
contracts. The prices of the nearest to the money call option on the closest to maturity futures
contract was used to construct a daily implied volatility series.
A forecasting model was fitted to the implied volatility series. We were encouraged in the
search for an efficacious forecasting model by the existence of previous studies, including French,
Schautz, and Stambaugh (1987) for equities and Bessembinder and Seguin (1990) for futures,
which had shown that volatilities showed persistent deviation from a random walk.*
Having obtained a suitable model, the forecasts of implied volatility were utilized in the
construction of delta neutral trading portfolios. These portfolios were designed to profit from
forecast changes in option implied volatility. Cumulative profit/loss from the implementa-
tion of the trading strategy over a six month period, both before and after transaction costs,
provided the measure of the strategy’s performance and thus evidence of the efficiency, or
otherwise, of the market.
Kemna (1989) and Wilson and Fung (1990) have previously examined the informational
content of implied volatility in equities and futures markets respectively. Both studies
examined the proposition that implied volatilities predicted future volatility. Our study goes
further in that we attempt to use the prediction of future volatility to predict and profit from
future changes in option prices.
The vast majority of studies of market efficiency have examined markets for the existence
or otherwise of profitable opportunities in terms of the price of the traded security. Our study
differs in that we did not examine market prices directly but rather indirectly through an
examination of one of the determinants of option market price, the option volatility. The
SFE Bank Bill Contract 405
90 Day Bank Bill Futures Yields 16.0%-
Jan 66 Apr 66 July 66 od66 Dee 66
Option Implied Volatility
0.90%
1 0.60%
0.70%
0.60%
0.50%
0.40%
0.30% m Jan 66 Apr 66 July 66 act 66 Dee 88
Figure 1. Sample period implied volatility and bill rates. The data points represent daily
values for the period January 1988-December 1988. The implied volatility series is com-
puted from the nearest to the money call option on the futures contract closest to maturity
using Black’s Model. The bill rates are derived as 100 minus the price of the futures contract
closest to maturity.
406 B. F. HUNT and R. BHAR
only other study which to our knowledge, has concentrated on implied volatility is the
Harvey and Whaley (1992) paper on the S&P 500 index.
If option volatility is forecastable it means that arbitrageurs can consistently generate
profits by forecasting future volatility and hence forecasting future option prices. It follows
that the existence of any structure other than a random walk in an option volatility series is
potentially indicative of market inefficiency if the non-randomness admits exploitable profit
opportunities after allowing for transaction costs. The SFE bank bill market is characterized
by many traders generating large trading volumes with market information displayed and
disseminated rapidly. One would not expect a priori a market like that for SFE bank bills to
be inefficient.
The fact that volatility, unlike price, is not directly observable may make the existence of
patterns in volatility more likely. One of the main contributions of this paper is that it provides
evidence that implied volatilities of bank bill futures options are forecastable and that this
forecastability implies trading strategies that are a profitable gross of transaction costs. The
fact that transactions costs eliminate any such profit infers that the market for bank bill futures
options is semi-strong form efficient.
METHODOLOGY
The study required data on the nearest bank bill futures contract and at the money call and
put options on this contract. The market prices used here are the closing (or settlement) prices
for the particular day. The data has been broken in to two periods. The in-sample model
estimation period runs from January 1988 to December 1988. Data from January 1989 to
June 1989 was used in an out-sample phase to test trading strategies and hence to make
inferences regarding market efficiency. Figure 1 sets out bank bill yield and implied
volatilities for the estimation sample period.
The principal unobserved variable in options pricing is the volatility of the underlying
security. As this volatility is unobservable market traders often focus upon the volatility
which is implied by the current market price of an option. This variable is termed the implied
volatility. Of course, the volatility implied by any pricing model is dependent upon the nature
of the pricing model. In this study we have used Black’s formula for options on futures.3
Support for the use of this model is given by Scott and Tucker (1989).4 There is of course
an implied volatility for each traded option series. A single implied volatility series was
constructed from the nearest to the money call option on the futures contract closest to
maturity. Implied volatility was computed using the Newton-Raphson algorithm applied to
Black’s model.
In the search for non-random patterns in the implied volatility series we used parametric
tests of correlation and non-parametric runs tests. Subsequently, the discovered nonrandom
patterns were modelled using a time series ARIMA model augmented with other instrumen-
tal variables. More specifically, variability in bill futures prices in the immediate past, option
length, and the day of the week were added to an ARIMA model to explain movements in
implied volatility. The model was developed in order to provide an implementable predictor
SFE Bank Bill Contract 407
of future volatility. This model was used to forecast future volatility and to act as the basis
of a volatility trading strategy. Like all trading strategies, the aim of a volatility strategy is
to profit by buying low and selling high.
Since volatility is only one of a number of influences on the level of option prices, it is
necessary, in order to successfully implement a buy low/sell high volatility strategy, to
insulate any market position taken on the basis of forecast volatility from the effect of
changes in the value of other influential variables. In particular, it is necessary to isolate the
trading position from the effects of a change in the price of the underlying 90 day bills futures.
Here, this was achieved by constructing market positions which were sensitive to changes
in volatility but, by virtue of being delta neutral, were insensitive to changes in the bill
futures price. Two delta neutral, volatility sensitive market portfolios were designed to profit
by buying low/selling high volatility.
The first of these delta neutral portfolios were constructed using bill futures and call
options on those futures. The second, used call and put options on bill futures. A futures
contract is insensitive to volatility changes and is thus used in the first strategy simply to
render that strategy delta neutral. The value of both put and call options on bill futures is
positively related to the level of volatility. A delta neutral portfolio constructed from calls
and puts is thus more sensitive to volatility changes than one constructed from calls and
futures.
RESULTS
Model Estimation
Prior to modelling the levels implied volatility series was first differenced to produce a
stationary series. The first phase of our investigation involved a search for predictable
dependencies in the constructed implied volatility series. Autocorrelations were calculated
for 20 lags. Each of these correlations were significantly different from zero at the five
percent level. Further, the Box-Pierce Q-statistic for the 20 lags of 1771.9 was highly
significant. While these parametric statistics indicate that the implied volatility series did not
follow a random walk they are dependent upon distributional assumptions. Nonparametric
run tests are more robust tests of the random walk hypothesis.5
The expected total number of runs of all signs and the variance of this statistic can be
estimated (Mood, 1940). This provided the basis for an extremely robust test of inde-
pendence. In our sample the total observed number of runs was 155 while the expected
number of runs of all signs was 117.33. The standardized normal Z-statistic was 4.88, which
was significant at the 0.10 percent level. The results of analysis of the total number of runs
suggest a lack of independence in the volatility series. A Chi-squared test indicated that the
series contained significantly more (at the 1 percent level) short runs than would be expected
under an assumption of random implied volatility changes. This result of excessive short
runs is indicative of a reversal pattern in the implied volatility series. Such patterns in respect
of prices in volatile markets have been previously found to occur.6
B. F. HUNT and R. BHAR
The results of the correlation and runs analysis suggest that the implied volatility series
contains a degree of dependence. On that basis, the next step was to identify a suitable model
to account for this dependence. Our emphasis was upon modelling the implied volatility
using a stochastic ARIMA time series model augmented with appropriate instrumental
variables.7 The instrumental variables tested to provide a further explanation of the implied
volatility series were:
1) the time to the expiration of the option;
2) the day of the week; and
3) a measure of daily volatility of the underlying futures contract.
Our reason for using time to maturity as an explanatory variable was the existence of a
number of studies which show that the Black model’s theoretical prices have been at
systematic variance with market prices.8 Thus, we hypothesized that the implied volatilities
derived from market prices using the Black model might well vary systematically with time
to maturity. The effect of the influence of time to maturity was included in the ARIMA model
through the use of a square root of days to maturity explanatory variable.’
The existence of a day of the week effect was tested using daily dummy variables. The
rationale for the inclusion of these variables was provided by the many studies which have
shown that volatile markets often have systematic day of the week effect. (See for example
French (1980) and Gibbons and Hess (1981)). Although these studies have concentrated on
returns it is reasonable to extend the argument to the possibility of day of the week effect on
volatilities.
A measure of daily volatility in the underlying futures contractlo was used in an attempt
to explain some of the variation in the implied volatility series.’ ’ Reason would suggest that
both implied volatility and historical volatility ought to be predictors of future volatility. On
that premise, increases in market volatility, as indicated by large changes in the price of the
underlying futures, ought to be reflected in both the level of the implied volatility and the
most recent historical volatility as given by measures of daily futures volatility. If the implied
volatility series were to react instantly to changes in market volatility there would be no
relationship between implied volatility and past daily observed volatility. However, where
there is a delayed adjustment or overreaction of implied volatility to actual volatility, implied
volatility will be statistically related to the lagged values of the daily futures volatility. In
that case the lagged daily volatility will be of assistance in forecasting future changes in
implied volatility.
One approach to the selection of a best forecasting model is to include all variables which
improve the forecast without justification as to the variables’ theoretical appropriateness. Our
approach, however has been to include only those variables which can be justified in terms of
the market structure and to select a preferred model on the statistical diagnostic criteria.
The preferred model set out in Table 1 was ARIMA( 1 ,l ,1) augmented with lagged values
of daily futures price volatility, square root of days to option maturity and day of the week
dummy variables.
SFE Bank Bill Contract 409
Table 1. Model Estimation. The implied volatility model was estimated, using TSP 6.0, from a sample of daily implied volatilities over the period January 1988 to December 1988.
In order to produce a stationary series the implied volatility was first differenced. This differencing required that a number of the explanatory variables were also first differenced.
Model Structure
The estimated model consisted of an ARIMA (l,l,l) augmented with three classes of instrumental variables.
1) Day of the week dummies.
2) Time to Maturity.
3) Measure of Daily Futures.
Volatility
Five dummies were included to identify any day of the week effects.
This variable was included to capture the influence, if any, of the
length of the option on implied volatilities.
A series of daily volatility in the underlying contract was induded to
proxy for the effect of recent dhanges in historical volatility on
implied volatility.
R-Squared 0.26
Model Parameters Coef t-Stat F-test
Ma(l) -0.560 -3.12**
Ar(l) 0.150 0.92
Days 0.004 1.68
Monday 0.012 2.33**
Tuesday -0.008 -1.54
Wednesday -0.001 -0.31
Thursday -0.113 -2.17
Friday 0.009 0.17
Dvol(0) 0.020 3.56**
Dvol(-1) 0.016 2.14**
Dvol(-2) 0.008 0.72
Dvol(-3) 0.003 0.44
Dvol(4) 0.007 1.34
*represents significance at the 5% level, ** represents signifi-
cance at the 1% level
F-statistic 6.44
Where, days to maturity. = fust difference of the square root of the time
Dvol(i) = The first difference of the daily futures price volatility lagged i days (see Note 7)
MA(l) = First order moving average.
AR(l) = First order autocorrelation.
Autocorrelations
The series were checked for stationatity using the Q statistic computed over 20 lags
Implied Volatility series 1771.85
First differenced Implied volatility 49.87
Mode1 residuals 18.49
Residuals in the forecast 32.50
Forecast Correlation
The correlation of the forecast implied volatility with actual volatility was 0.8755 The forecast mean absolute percentage error was 4.49%.
-
410 B. F. HUNT and R. BHAR
The preferred model contains the contemporaneous daily volatility variable, dvol(0). This
variable, while included for model estimation purposes, was set equal to zero for the forecast
period as it is not observable at the time of forecast. Although the model was selected on the
basis of its fit over the sample period, of interest is its fit over the forecast period. Figure 2
plots the actual and model forecast implied volatility, for both one and two day forecasts,
over the forecast period January to June 1989. The correlation of the one day forecasts with
the actual implied volatility was 0.8733 and the mean absolute percentage error was 2.81
percent. The same statistics for the two day forecasts were 0.7569 and 3.97 percent
respectively. One deficiency of correlation statistics in the analysis of forecasts is that they
do not identify consistent bias in forecasts. For example, a forecast which consistently
over- or under-estimates a target series may well have a high correlation with the target
series. In contrast, Theil’s U statistic, inter aliu, takes account of forecast bias. Theils’s U
statistics, which are presented in Table 2, indicates that the model has some power.
Market Strategy
Two distinct trading portfolios were constructed and used in an attempt to profit from the
preferred model’s forecast of implied volatility. Both portfolios were constructed to be delta
neutral. The first portfolio (call/futures) was constructed from call options and futures. The
second portfolio (call/put) consisted of call options and put options. The two portfolios
differed in their sensitivity to volatility. The call/put was approximately twice as sensitive
to volatility. As the call/put portfolio was more highly geared to volatility changes it was
expected to generate greater trading profits but with greater risk/error than the call/futures
portfolio.
The basis of our volatility trading strategy was to buy volatility when the forecast model
indicated that volatility was relatively cheap and to sell volatility when the model indicated
volatility was relatively dear. For example, if the model forecast that volatility would
increase in the future, i.e., that the forecast future volatility exceeded the current level of
Table 2. Measurement of Forecast Accuracy by Theil’s U Statistic
Statistic
U Urn
US
UC
I-Day Forecast
0.0394
0.0065
0.0006
0.9929
2-Day Forecast
0.0550
0.0398
0.0008
0.9594
Theil’s U statistic can be decomposed into three. component parts Urn, US, UC. Where,
Urn is the percentage of error due to bias, ideally 0%,
US is the percentage of error due to variance, ideally 0%
UC is the percentage of error due to covariance, ideally 100%.
SFE Bank Bill Contract 411
Implied Volatility Observed vs 1 Day Forecast
- Observed l 1 Day Forecast
Observed vs 2 Day Forecast
0.9
0.8
0.7
0.6
0.5
0.4
Jan 89 Feb 69 Mar 89 Apr 89 Apr 69 May 89 June 89
- Observed l 2 Day Forecast
Figure 2. Actual and forecast implied volatility. The forecast implied volatility was
derived from an augmented ARIMA model. These series and the actual implied volatility is
plotted against time for the forecast period January 1989-June 1989. Theil’s U-statistic was
0.0394 and 0.0550 for the one day and the two day forecast respectively.
412 B. F. HUNT and R. BHAR
option implied volatility we would “buy” volatility through the purchase of a delta neutral
portfolio whose value was positively correlated to volatility changes. Such a portfolio was
said to be long volatility. Specifically we chose either a long volatility future/call portfolio
(sold future, bought call) or a long volatility call/put portfolio (bought call, bought put
option). If the model forecast of an increase in volatility was vindicated by the passage of
time then the value of the long volatility portfolio would increase thus yielding a profit for
the trading strategy. Conversely if the forecast model indicated that volatility was to fall we
would sell volatility by taking a short volatility portfolio, either a short futures/call portfolio
(bought futures, sold call) or a short call/put (sold call, sold put). The specific details of the
long and short volatility positions are set out in Table 3. Here it can be seen in that each
position is rendered delta neutral using the computed call option delta from Black’s model.
The trading strategy involved the taking of a position on the basis of forecast volatility on
one day and the liquidation of that position at a later date. In the first instance the strategy
involved the taking of a position at the close of trade on one day and the liquidation of that
position on the following day. However such a strategy was open to the legitimate criticism
that it was not implementable. Here we assumed we could use the closing price on a particular
day for both the generation of the forecast and as the entry price for trading. Accordingly we
designed another strategy which was implementable. Namely, we used a two day forecast
to take a position based on forecast produced yesterday in order to undertake a one day
trading strategy entered into today and liquidated tomorrow. Here there is a one day lag
between the observation of prices and the selection of positions and the resulting trades. We
report the results for both the one day and the two day forecast strategies.
The trading processes generated daily profit/loss series which were accumulated to give
an overall profit/loss for the entire forecast period. In order to account for transaction costs
the net change in futures and options positions from one day to the next was calculated. The
transaction costs in a deregulated market like the SFE depend upon the nature of transactor.
We have used the per unit transaction cost, supplied by Barclays de Zoete Wedd, applicable
to a large institutional client.
The success of the market strategy was dependant upon both the ability of the model to
forecast changes in the implied volatility and upon the ability of the constructed portfolio to
be insensitive to changes in variables other than volatility. In particular, if changes in the
Table 3. Market Strategy Positions
Portfolio Call/Future CalVPut
Long Volatility Buy 1 Call Buy 1 Call
Sell d Futures BUY d/(1-d) Puts
Short Volatility Sell Call Sell 1 Call
Buy d Futures Sell d/(1-d) Puts
Now Each of the strategies are standardized to 1 call option. “d” is the Black
model delta of a call option.
SFE Bank Bill Contract 413
Table 4. Trading Profits
Panel A One Day Forecast Horizon
Portfolio Perjiect Foresight Forecast
Without Transaction Costs
Call/Puture
Call/Put
With Transaction Costa
Call/Puture
call/Put
Panel B Two Day Forecast
Portfolio
Without Transaction Costs
ChWFUtUE
Call/Put
With Transaction Costs
Call/Ptmue
Call/Put
162,632 (-0.66%)
367,208 (-3.22%)
71,968 (-22.71%)
222,804 (-18.93%)
Perfect Foresight
18,157 (-2.66%)
55,152 (-12.68%)
-41,756 (-24.78%)
-48,464 (-28.40%)
56,328 (-0.24%)
87,844 (1.63%)
-28,732 (-20.56%)
47,512 (-14.21%)
Forecast
19,936 (-0.32%)
34,066 (4.35%)
-37,279 (-20.04%)
65,961 (-11.37%)
Note: The values in the body of the table represent the cumulative profit resulting from trading
in portfolios standardized to include 100 calls. The specific transaction cost used in
computing the trading profit/loss was $4.34 per trade. The percentage figures in paren-
theses associated with the two day forecasts with transaction cost included refer to the mean of the daily return on funds invested in the implementing the call/future portfolio.
price of bank bills futures, unaccompanied by a change in volatility, were able to significantly
alter the value of our call/futures or call/put portfolios this would obscure the results of
volatility trading.
Our market trading strategy was predicated on an ability to forecast future implied
volatility and to profit from these forecasts using a specific volatility strategy. A logical
benchmark for measuring the success, if any, of our strategy was the result of a similar
strategy based on a perfect foresight prediction model. In a perfect foresight strategy future
volatility changes are anticipated without fail. While the results of this strategy are clearly
unattainable, they nevertheless provide an upper limit against which one can judge the results
of other attainable strategies. That is, the results of the perfect foresight model provided an
upper limit to the trading profits possible under our strategy of buying low volatility selling
high volatility. Other simple benchmark strategies such as buying and holding either calls,
puts, or futures are inappropriate in the context of volatility forecasting as they are not delta
neutral and are consequently effected by factors other than volatility.
It was of course important that the perfect foresight volatility strategy be able to produce
positive profits after allowing for transaction costs. If this were not so then the inability of
414 B. F. HUNT and R. BHAR
a strategy of this type to produce profits would indicate the inappropriateness of our chosen
strategy. It would not necessarily indicate market efficiency.
Table 4 sets out the profit and loss for both strategies. As stated earlier we employed both
a one day and a two day forecast strategy. It can be seen that ignoring transaction costs both
forecast portfolios based upon predicted implied volatility have positive returns. The trading
profits generated by these portfolios however, are much less than those which would be
generated by the Perfect Foresight Model. The call/future strategy returned some 35 percent
of the profits available to this trading strategy (as given by the perfect foresight model). The
put/call strategy while generating a greater absolute level of profit provides a smaller
percentage, 23 percent of the available profits.
When transaction costs are accounted for it renders both of our forecast implied volatility
strategies unprofitable. The perfect foresight profits remain however, strongly positive. The
after transaction results support a proposition of semi-strong form efficiency.
The two day forecast trading produced different results. Generally the profit or loss
attached to the two day forecast strategy is less than the one day equivalent strategy. This is
not unexpected as we would expect two day forecasts to be less accurate than the one day
forecasts. While the one day forecast strategy can be criticized as being difficult to implement
the two day forecast strategy can be criticized as being unrealistic. No trader would wait, as
the two day forecast strategy requires, until the close of trading next day to undertake a
transaction based on a buy or sell signal generated 24 hours earlier.
CONCLUSION
The objective of our study was to test the proposition that the market for bank bill futures
options was semi-strong form efficient with respect to the information contained within
option implied volatilities. The methodology employed showed that implied volatilities were
non-random and predictable at least to some extent. Thus, the volatility trading strategies
employed in the study had the potential to generate profits (as shown by the perfect foresight
results) however, the implied volatility series did not contain sufficient dependencies, after
accounting for transaction costs, to allow for profitable exploitation.
NOTES
1. The SFE trades three interest rate contracts. These are the three and 10 year 12 percent coupon Treasury Bonds contract and the 90 day bank accepted discount bills contract.
2. Predicting volatility is also central to studies under the heading of ARCWGARCH models. 3. See Hull (1989, p. 147). The fact that the SFE use Black’s model to compute theoretical prices
and option deltas provides further support for the use of the model in this study. 4. Scott Tucker (1989) found in investigating currency options that a stationary parameter models,
like Black’s futures model, were as accurate as more sophisticated models such as C.E.V. models.
5. A run in a data series is defined as the number of successive changes of the same sign.
6. See Cootner, Alexander, and Alexander in Cootner (1964)
SFE Bank Bill Contract 415
7. Various ARIMA models, augmented with instrumental variables, were estimated and diagnos-
tically checked using the Micro TSP 6.0 package.
8. See for example Kemna (1989).
9. Arguments contained in Milonas (1986) would justify the use a linear measure of days to expiry.
However we chose to use a non-linear (square root) measure on the ground that this was more consistent
with the way in which time is incorporated into the standard option pricing models.
10. Daily volatility of the underlying futures contract is constructed as the standard deviation of
a single day’s futures return. The use of standard deviation rather than variance in measuring volatility
is consistent with our definition of implied volatility, which is more accurately described as implied
standard deviation. The daily volatility of futures series was constructed as the absolute value of the
logarithm of the ratio of today’s futures price to yesterday’s futures price, multiplied by the square
root of 250 to annualize the measure.
11. The inclusion of daily volatility as an explanatory variable was expected to account for the
reversal pattern in the implied volatility series.
REFERENCES
Bessembinder, Hendrik, Seguin and Paul J. “Futures Trading, Liquidity and Stock Price Volatility.”
Unpublished Manuscript, Arizona State University (1990).
Cootner, P.H. The Random Character of Stock Market Prices. M.I.T. Press (1964).
French, Kenneth R., Schautz, William G., and Stambaugh, Robert F. “Expected Stock Returns and
Volatility.” Journal of Financial Economics 19 (1964): 3-30. French, K.R. “Stock Returns and the Weekend Effect.” Journal of Financial Economics (March
1980).
Gibbons, M.R., and Hess, P. “Day of the Week Effects and Asset Returns.” Journal of Business (October 198 1).
Harvey, CR. and Whaley, R.E. “Market Volatility Prediction and the Efficiency of the S&P 100 Index
Option Market.” Journal of Financial Economics (1992). Hull, J. Options, Futures, and Other Derivative Securities. Prentice Hall International (1989).
Jensen, M.C. “Some Anomalous Evidence Regarding Market Efficiency.” Journal of Financial Economics 6 (1978), 95-101, as quoted in “What Do We Know about Stock Market Efficiency?.”
Ball, ray, NATO ASI Series, Vol. F54. Springer-Verlag (1989): 25-55.
Kemna, Angelien. “An Empirical Test of the OPM Based on EOE-Transactions Data.” NATO ASI
Series, Vol. F54. Springer-Verlag (1989): 745-768.
Mood, A.M. “The Distribution Theory of Runs.” Annals of Mathematical Statistics 11 (1940):
367-392.
Scott, Elton and Tucker, Alan L. “Predicting Currency Return Volatilities.” Journal of Banking and Finance, 13 North-Holland (1989): 839-85 1.
Wilson, William W., Fung Hung Gay. “Information Content of Volatilities Implied by Option
Premiums in Grain Futures Markets.” The Journal of Futures Markets, 10 (1) (1990): 13-27.