THEFINANCIAL

REVIEW VOL.30 No. 2 MAY 1995 PP. 211-241

Empirical Tests of the Pricing of

Nikkei Put Warrants

Jason Z. Wei*

Abstract
The purpose of this study is to empirically examine
the pricing of Nikkei put warrants, which are long-term
put options written on the Nikkei 225 index. Using war-
rants traded on the Toronto Stock Exchange, this study
performs various tests on the pricing models proposed by
Dravid, Richardson, and Sun [ll],Reiner [24],and Wei
[311. It is found that the models tend to overprice the
warrants. The overpricing, possibly caused by the omis-
sion of the credit risk and the Extraordinary Event
Clause, is found to be positively related to the degree to
which the warrants are in the money, the volatility level,
and the trading volume.

Introduction
Recently cross-currency derivative securities have
gained increasing popularity. Major exchanges in North
America have listed options on such foreign stock in-
dexes as the Nikkei 225, FT-SE 100, and CAC 40.
Among the listed foreign index warrants, the most popu-
lar is the Nikkei Put Warrant (NPW hereinafter). For
instance, the American Stock Exchange (AMEX) listed
six different issues of NPWs in early 1990. The Toronto
Stock Exchange (TSE) also listed six different issues of
NPWs (see the appendix for details). Despite the in-
creasing popularity of the foreign index warrants, the
literature has been lacking in formal treatments on

*University of Saskatchewan, Saskatoon, Saskatchewan, Canada S7N 5A7

This paper is extracted from the author's doctoral dissertation completed at
the University of Toronto. The author would like to thank John Hull and Alan
White for their guidance, and Marlene Puffer and Paul Potvin for discussions.
211
2 12 Wei

these instruments in terms of both theoretical pricing

and empirical testing. There are a few exceptions. Rum-
sey [261 discusses the estimation of volatilities for a par-
ticular class of cross-currency options.' Reiner [24]
shows in a heuristic fashion how certain cross-currency
options can be priced. Wei [311 discusses specific issues
for the pricing of NPWs. Wei [321 also examined the
pricing of cross-currency options in a broader setting. A
study that is closely related to the current paper is by
Dravid, Richardson, and Sun [ l l l . Working with NPWs
traded on the American Stock Exchange, they compare
the trading prices with the model prices and conclude
that the pricing models perform reasonably well. Other
related papers are by Derman, Karasinski, and Wecker
[lo] and Gruca and Ritchken [161. Except for Dravid,
Richardson, and Sun [ l l l and Wei [321, the above cited
papers concentrate on the description and theoretical
pricing of cross-currency products. No attempt is made
to empirically assess the pricing models.
The purpose of this study is to empirically examine
the pricing models for NPWs developed in Dravid,
Richardson, and Sun [ll],Reiner [241, and Wei [31]. As
an attempt in the literature t o empirically investigate
the NPW market, this study will shed light on the per-
formance of the models and related issues. This should
constitute the major contribution of the study. NPWs
traded on the Toronto Stock Exchange are used to per-
form the tests. This choice is dictated by the fact that
there are more varieties on the TSE than on the AMEX.'
The remainder of the paper first briefly describes the
NPWs and introduces the pricing models. It then de-
scribes the data and the tests, followed by discussions on
empirical results and the conclusion.

The Nikkei Put Warrants and Their Valuation

The Nikkei Put Warrants

An NPW is a put option written on the Nikkei 225
index. The key difference between an NPW and a con-
ventional put option is that an NPW is traded in dollars
Nikkei Put Warrants 2 13

while the underlying asset (i.e., the Nikkei 225 index) is

denominated in yen. Depending on how the yen payoff is
converted into dollars, there are four possible payoff
specifications. Notation:
S = Nikkei 225 index level in yen;
X = current US$/JapY (or Cdn$/JapP) exchange
rate;
X , = pre-specified, fixed exchange rate;
K = exercise price in yen;
r = domestic riskfree interest rate (constant);
r f = Japanese riskfree interest rate (constant);
0, = annual volatility of the Nikkei index;
0, = annual volatility of the exchange rate;
p = correlation coefficient between the index and
the exchange rate;
q = continuous dividend yield on the Nikkei
index;
t = current time;
T = maturitydate.
The possible specifications can then be summarized as
follows:3

Category Pavoff dDon Exercise in Dollars

Category I NPWs: Max[O, X A K - S,)l
Category I1 NPWs: Max[O, X,(K - S,)1
Category I11 NPWs: Max[O, X& - X,S,I
Category IV NPWs: Max[O, X& - X&I
For the first category, the yen payoff of the warrant is
converted into dollars at the prevailing exchange rate,
X,. For the second category, a pre-specified exchange
rate, X,, is used to convert the yen payoff. The fixed
change rate, X,, is specified at the time of issue and re-
mains the same throughout the life of the warrant. The
payoff conversion is slightly complicated for the last two
categories. When a Category I11 NPW is exercised, the
exercise price is converted at the prevailing exchange
rate X,, while the terminal Nikkei index level is con-
verted at a pre-specified exchange rate X,. It is the oppo-
2 14 Wei

site for a Category IV NPW. Here, the exercise price is

converted up-front at the fixed exchange rate X,,, and
hence stays constant in dollars, whereas the Nikkei in-
dex is converted at the spot exchange rate. At any point
in time, the quantity X S is simply the dollar price of the
Nikkei index. Except for Category 111, all the other cate-
gories of NPWs have appeared on either the American
Stock Exchange or the Toronto Stock Exchange, or both.
All listed NPWs are American style options in that they
can be exercised anytime before maturity.
Due to the time zone difference, the exercise of a n
NPW can not be settled immediately. A simultaneous
quote for the Nikkei index does not exist when the North
American markets are open. As a result, NPW issuers
typically specify a one- or two- (business) day exercise
delay. On the TSE, Trilon Financial Corporation NPWs
have a two-day exercise delay, while the other warrants
have a one-day exercise delay. For instance, if a n exer-
cise notice is delivered on Thursday for a Trilon Finan-
cial Corporation NPW, the issuer will wait and use the
following Monday’s close of the Nikkei and the exchange
rate to settle the exercise. For other warrants, Friday’s
closing data will be used for the exercise settlement.
Y

Valuation of Nikkei Put Warrants4

Assuming joint geometric Brownian motions for the
Nikkei index and the exchange rate in a Black-Scholes
environment, Dravid, Richardson, and Sun [111, Reiner
1241, and Wei [311 independently show that all categories
of ordinary cross-currency options can be priced within a
one-state-variable framework despite the presence of
two underlying variables (i.e. the foreign index and the
exchange rate). Wei [311 specifically shows that the one-
state-variable framework also applies to NPWs. The di-
mension reduction enables Wei [31] to derive
closed-form pricing formulas for all categories of Euro-
pean NPWs. For American style NPWs, closed-form for-
mulas do not exist. Numerical procedures such as a
binomial tree must be used. Generally speaking, a three-
dimension tree is required to handle a two-state-variable
American option. However, Wei [31] shows that the di-
Nikkei Put Warrants 2 15

mension reduction with European NPWs also extends to

American NPWS.~Therefore, only a two-dimension tree
(e.g., [9]>is needed. Specifically, Wei [31] suggests using
either a binomial [9] or a trinomial [4] lattice, the frame-
work of which can be summarized as follows.6
Current
Value of
the State Strike Dividend Discount
Variable Price Drift Rate Variance Yield Rate
2
Category1 SX KX rf 0s 4 rf
2
Category I1 SXo KXo rf- 0s 4 r
2posox

CategoryIV SX KXo r ff:+2ppx 4 r

+ ox

When pricing an NPW using the above framework,

one proceeds as if an ordinary lattice tree is being built to
price an ordinary put option. The only difference is that
different inputs are used for the tree construction. For ex-
ample, to price a Category I1 NPW, a trinomial lattice is
built for a state variable with SX, being its current value,
0,its volatility, and (rf- 2p0,0,) its drift. (In contrast, in a
standard lattice, the risk-neutral drift is r.) Then one
works backwards along the tree to solve for the American
NPWs value with exercise price no. (See Wei [311 for a
detailed example of implementing the above framework.)
It should be pointed out that two approximation as-
sumptions are necessary in order to apply the pricing
framework to American NPWs. First, it is assumed that
the closing level of the index from the previous trading
session in Tokyo is the current value of S (i.e., prices are
assumed to be simultaneous). Second, the exercise delay
can be ignored, and the exercise of a warrant will be set-
tled immediately. Obviously, pricing errors are intro-
duced when making these two assumptions. However, as
shown in Wei [311, the pricing errors are quite small.
Therefore, for practical purposes, the approximations
are well warranted.
In this study the above pricing framework is tested.
Specifically, a 100-step trinomial lattice [4] in conjunc-
tion with the control variate technique [17] is used to
2 16 Wei

price American NPWs traded on the Toronto Stock Ex-

change. The model prices are then compared with the
market prices. The European counterpart of each war-
rant is used as the control variable when implementing
the control variate technique. Closed-form pricing for-
mulas for European NPWs are given in Wei [311.

Test Design

The Data
The data used in this study come from four sources:
the Nikkei Telecom-Japan News & Retrieval, RBC Do-
minion Securities Inc., Nomura Inc. Canada, and the fi-
nancial newspapers. Specifically, the warrant data set
was originally supplied by the RBC Dominion Securities
Inc. and was updatedkompleted using daily quotations
from the Globe and Mail daily newspaper. The data set
contains daily high, low, close, and volume for the six
warrants listed on the Toronto Stock Exchange. The AB
Svensk Exp. Corp. warrant is deleted from the data set
due to thin trading. The sample period ends on May 31,
1991, for all five remaining warrants. The starting date
varies. The sample period starts on June 14, 1989, for
BT Bank of Canada Series I and I1 NPWs, and April 10,
1990, for the other two series. For Trilon Financial Cor-
poration NPWs, the data start on February 23,1990.
The daily open, high, low and close quotes for the
Nikkei 225 are retrieved from the Nikkei TelecomAapan
News & Retrieval. Daily quotations for the Cdn$/JapY ex-
change rate in the same time period are taken from the
Globe and Mail newspaper.
Yields on government bonds with maturity dates
close to those of the warrants are used in lieu of the risk-
free interest rates. Because the warrants expire on dif-
ferent dates, bonds with different maturities are used.
Daily Canadian government bond prices are obtained
from the Financial Post. Their Japanese counterparts
are obtained from Nikkei TelecomAapan News & re-
trieval. Continuously compounded yields are calculated
from bond prices.'
Nikkei Put Warrants 217

Finally, monthly dividend yield data, starting in

January 1980 and ending in December 1990, are ob-
tained from Nomura Inc. Canada. The yields in this pe-
riod are used to obtain a simple moving average forecast
for the next two years, which is 0.43 percent per annum.
It is assumed that a continuous dividend yield of 0.43
percent will prevail during the data period.’

The Tests
The major objective is to test the predicting power
of the proposed valuation models reviewed earlier.
Broadly speaking, there are two classes of tests in the
literature that investigate the predicting power of option
models. Tests in the first class make direct comparisons
of actual prices and model prices. If the model is valid,
then the predicted model price should not deviate sys-
tematically from the actual price (e.g., [3, 20, 21, 301).
Tests in the second class are based on volatilities im-
plied from actual optiodstock prices, using the tested
pricing model. If the model is valid (and if the market is
efficient, and the other parameter inputs are accurate),
then the implied volatilities should behave as dictated
by the model. For example, if the Black-Scholes model is
valid, then the implied volatility should be stationary
over time, across maturities and exercise prices (e.g., [l,
5, 6, 25, 281). The methodology employed in this study
falls into the first class.
Two questions can be asked: 1) Are the market
prices of the NPWs significantly different from those
predicted by the pricing models? 2) If the answer to
question 1) is “yes”, are there any systematic relation-
ships between the deviations and the model parameters,
such as time to maturity? A deviation test will be carried
out to answer the first question, and regression tests to
answer the second question. Because model prices are
needed to complete the tests, it is necessary t o illustrate
how the unobservable parameters are estimated.
Estimation of the Unobservable Parameter
Values. It can be seen in the previous section that the
unobservable parameters include the index volatility o,,
218 Wei

the exchange rate volatility ox,and the correlation coeffi-

cient between the index and the exchange rate p. For
Category I NPWs, only the index volatility enters the
pricing model. For Category IV NPWs, the three pa-
rameters enter the pricing model as a single input
u2 = o,”+20s,+o,2(q,= pose,). For Category I1 NPWs,
the three parameters enter the pricing model as two in-
puts, osand q,.There are two alternative approaches to
estimating these parameters. The first approach in-
volves calculating historical volatilities and correlation
coefficient and using them as the required parameter es-
timates. The second approach involves computing the
parameter values from the pricing models using values
of the observable variabledparameters. The well known
drawback of the first approach is its backward-looking
nature; i.e. , it can not incorporate investors’ expectations
about the future volatilities (and the correlation). Many
researchers have used this approach for conventional op-
t i o n ~Because
.~ all NPWs have a relatively longer time to
maturity, there is a unique problem which is less of a
concern for ordinary options with shorter maturities:
How far back should the data go in order to calculate a
historical parameter value that will be used as a for-
ward-looking estimate applicable to the next two (or
three) years? If the data do not go back far enough (say,
using data for the past three months), then there is the
risk of getting a totally wrong estimate;” if, however, the
data go back too far (say, using data for the past three
years), there is the problem of putting too much weight
on the remote observations that are less relevant for pre-
dicting the future. In light of this dilemma, the second
approach is used in this study, whose major advantage
is the ability to incorporate investors’ expectations. The
drawback, of course, is its reliance on a particular pric-
ing model. If the model is the one under study, then this
approach is subject to the error of “using the model to
test the model.” The commonly adopted technique to
overcome this problem is to use the previous day’s im-
plied volatility as the current day’s volatility estimate.
Moreover, multiple options are used so that a weighted
average of implied volatility can be calculated as the
volatility forecast.
Nikkei Put Warrants 219

In this study, the implied volatilities and covariance

are calculated in the following steps.
a) Compute the volatilities and covariances for day
t - 1based on the observable variables on day t - [Link]-
cifically, for Category I NPWs, 0,is imputed; for Cate-
gory IV NPWs, v = do,"+ 2o, + o: is imputed as a single
parameter; for Category I1 NPWs, o, is imputed given a
historical estimate of
b) Calculate a weighted average of the implied in-
dex volatilities from Category I and Category I1 NPWs,
with the weight being the degree to which the NPWs are
in the money.13 Specifically, the weighted average, de-
noted by WGISD,? is calculated as,

where i denotes the NPWs used to compute the individ-

ual o,,and K, is the strike price of warrant i. (There are
one Category I and three Category I1 NPWs). The verti-
cal bars stand for a n absolute value operator. In the
above procedures the warrant prices are the midpoint
averages of high and low prices. The main rationale for
using the high-low average in favor of the closing price is
to achieve a better match between the index level and
the warrant price. Recall that the Nikkei close from the
previous trading session in Tokyo will be used as the
model input. The averaging will hopefully even out the
impact of inaccurate index input.
Deviation Tests. The first question posed before
can be answered by calculating the absolute and per-
centage deviations between the market prices and the
model prices. If the models are correct, then the devia-
tions should not be significantly different from zero. The
testing procedures are as follows.
220 Wei

a) Using the implied volatility and other model in-

puts, calculate the model price, Pyd,for each warrant on
day t. A time series of model prices is obtained for each
warrant.
b) For each warrant on day t , calculate the absolute
deviation Pykt- rd and the percentage deviation
100%*(ckt- P r d )/[Link] warrant market price, Prkt,
is the daily closing price.
c) Repeat b) with ckt being replaced by the mid-
point average of the high and low prices.

Regression Tests, Bias Analysis. The second

question concerns the relationship between the predic-
tion errors and the (potential) systematic effects of the
model inputs. Many authors have tried to relate pricing
biases to such model inputs as time to maturity, the de-
gree to which the option is in the money, and the volatil-
ity.14 In this study the relative deviations are regressed
on similar model inputs. Specifically, the following time-
series regressions are run for each warrant:

where PD, = (ek' - Pyd)/ PYod(Prktis the midpoint aver-

age of high and low prices), and Y, is the independent
variable. For NPWs in categories I and 11, Y, is
(K - S ) / K (the degree to which the warrants are in-the-
money), z (time to maturity), 0,(volatility), and Zn(uoZ-
ume) (log of the trading volume). Therefore, there are
four separate regressions for each warrant. For Category
IV NPWs, Y, is (no - SXj/K&, z, u (volatility), and
Zn(uoZume). There are also four separate regressions for
each warrant.
If the models perfectly predict the market prices,
then the two coefficients, yo and yl,should not be signifi-
cantly different from zero for each regression. On the
other hand, if the models produce consistent prediction
biases, but the biases are not systematically related to
the model inputs, then for each regression the intercept
'yo should be a non-zero number, while the coefficient y1
should not be statistically different from zero.
Nikkei Put Warrants 22 1

The Empirical Results

Deviation Tests
The results for deviation tests are shown in Table 1
and Table 2. By examining the average deviations (both
absolute and percentage), it can be seen that the models
tend to overprice NPWs. (The BT Bank of Canada Series
I11 NPW is an exception.) This is true for both sets of
market prices (i.e., the closing prices and high-low aver-
age prices). The average overpricing ranges from 7.5q to
42.1g per warrant, and the t-values for the absolute de-
viations are all statistically significant. Although the
models seem to underprice BT Bank of Canada Series
I11 NPWs, the t-values are not significant for the abso-
lute average deviations.
The average deviations can not tell the whole story.
In a particular time-series of pricing deviations, if the
number and magnitude of the negative and positive de-
viations are such that the two types of deviations cancel
each other, then the overall average deviation is zero, in-
dicating that the model values the warrants correctly.
This is, of course, a misleading inference. To address
this issue, deciles are provided for the deviations for
each warrant. In addition, the proportions and averages
of negative and positive deviations are calculated for
each warrant. If, for example, the proportion of negative
deviations is high and their absolute average is bigger
than that of the positive deviations, then it can be con-
cluded that the model tends to overprice warrants most
of the time. The numbers in Table 1 and Table 2 gener-
ally support this conclusion. It can be seen that, when
closing prices are used to conduct the tests, all median
deviations are negative. When the high-low average
prices are used, all median deviations are again nega-
tive, with only one exception: the BT Bank of Canada
Series I11 NPWs. Moreover, for each warrant, the aver-
age size of negative (absolute) deviations is bigger than
that of the positive deviations. This should come as no
surprise, since the overall averages are negative with
significant t-values.
 to
to
to
TABLE1

Results of Absolute Deviation Tests

This table contains the absolute deviations measured in dollars. The numbers with the percentage
symbol are the proportions of negative and positive deviations. The numbers in the parentheses are
averages of the negative or positive category. For example, in Panel A, for BTI NPWs, 58 percent
(285/490 = 0.58) of the observed deviations are negative, while 42 percent (205/490) are positive de-
viations. The averages of the negative and positive deviations are -0.245 and 0.156, respectively.

BTI BTII BTIII BTIV TFC

Panel A Absolute Deviations Based on Closing Prices of NFWsa
Minimum -1.734 -1.744 -1.610 -2.206 -1.350
10th Percentile -0.484 -0.568 -0.659 -1.007 -0.889
20th Percentile -0.251 -0.329 -0.413 -0.765 -0.531
30th Percentile -0.138 -0.164 -0.243 -0.572 -0.450
40th Percentile -0.066 -0.087 -0.163 -0.390 -0.320
Median -0.030 -0.045 -0.005 -0.281 -0.159
60th Percentile 0.007 0.000 0.194 -0.216 -0.035
70th Percentile 0.050 0.025 0.417 -0.133 0.062
80th Percentile 0.103 0.068 0.553 -0.053 0.308
90th Percentile 0.233 0.183 0.860 0.097 0.596
Maximum 0.959 1.207 1.559 0.507 1.673
Average -0.075 -0.120 0.067 -0.400 -0.154
t-valueb -5.463*** -8.084*** 1.168 -14.407*** -3.643***
(-) Deviation 58% (-0.245) 60% (-0.285) 50% (-0.408) 84% (-0.511) 65% (-0.463)
(+) Deviation 42% (0.156) 40% (0.125) 50% (0.541) 16%(0.173) 35% (0.410)
# of obs. 490 488 102 285 161
 Panel B: Absolute Deviations Based on Averages of High-Low Prices of NPWsa 2:
F
Minimum -1.734 -1.595 -1.360 -2.605 -1.350 K
10th Percentile -0.412 -0.497 -0.668 -1.004 -0.803
20th Percentile -0.258 -0.297 -0.397 -0.794 -0.611
30th Percentile -0.136 -0.185 -0.288 -0.557 -0.465
40th Percentile -0.068 -0.097 -0.163 -0.437 -0.334
Median -0.028 -0.048 0.019 -0.323 -0.177
60th Percentile 0.003 -0.011 0.176 -0.252 -0.083
70th Percentile 0.029 0.016 0.419 -0.176 0.039
80th Percentile 0.071 0.049 0.536 -0.087 0.212
90th Percentile 0.160 0.099 0.860 0.029 0.439
Maximum 0.885 1.020 1.059 0.444 1.173
Average -0.076 -0.121 0.049 -0.421 -0.180
t-value -6.747*** -9.522*** 0.894 -16.881*** -4.673***
(-) Deviation 59% (-0.212) 63% (-0.254) 49% (-0.423) 88% (-0.495) 65% (-0.467)
(+) Deviation 41% (0.121) 37% (0.103) 51% (0.505) 12% (0.120) 35% (0.343)
# of obs. 490 488 102 285 161
a Column headings: BTI, BTII, BTIII, and BTIV stand for BT Bank of Canada Series I, Series 11, Se-
ries 111, and Series IV NPWs, respectively; TFC stands for Trilon Financial Corp. NPWs.
*: significant at 10 percent level for a two-tail test; **: significant at 5 percent level for a two-tail
test; ***: sigdicant at 1percent level for a two-tail test.
 to
TABLE2 to
rp

Results of Percentage Deviation Tests

This table contains deviations measured in percentage. The lines headed by “(-1 Deviation” and “(+)
Deviation” report the proportions of negative and positive deviations. The numbers in the parenthe-
ses are averages of the negative or positive category. For example, in Panel A, for BTI NPWs, 58 per-
cent (285/490 = 0.58) of the observed deviations are negative, while 42 percent (205/490) are positive
deviations. The averages of the negative and positive percentage deviations are 4 . 2 8 percent and
4.98 percent, respectively.

BTI BTII BTIII BTIV TFC

Panel A Percentage Deviations Based on Closing Prices of NPWsa

Minimum -25.64% -22.43 -12.28 -32.60 -21.82
10th Percentile -7.34% -9.30 -6.74 -16.82 -9.25
20th Percentile 4.68% -6.51 -4.80 -14.00 -6.04
30th Percentile -3.04% -4.42 -2.92 -11.32 -4.97
40th Percentile -1.89% -2.80 -1.70 -8.89 -3.34
Median -0.86% -1.36 -0.07 -6.71 -1.65
60th Percentile 0.24% -0.11 1.98 -5.26 -0.52
70th Percentile 1.42% 0.86 4.30 -3.46 0.92
80th Percentile 3.21% 2.34 9.30 -1.20 2.77
90th Percentile 6.80% 5.26 13.25 1.89 6.96
Maximum 37.75% 27.19 21.63 11.04 26.44
Average -0.41% -1.81 1.92 -7.52 -1.39
t-valueb -1.386 -6.439*** 2.611*** -17.100*** -2.525**
(-) Deviation 58% (-4.28%) 60% (-5.42%) 50% (-4.04%) 84% (-9.54%) 65% (-5.23%)
(+) Deviation 42% (4.98%) 40% (3.77%) 50% (7.71%) 16% (3.00%) 35% (5.63%)
# of obs. 490 488 102 285 161
 Panel B: Percentage Deviations Based on Averages of High-Low Prices of NPWsa
Minimum -18.33% -17.84 -10.38 -28.70 -19.91
10th Percentile -6.11% -8.67 -6.74 -16.52 -8.86 b
20th Percentile 4.33% -5.57 4.33 -13.88 -6.48 E
30th Percentile -2.98% -4.16 -2.86 -10.83 4.90
40th Percentile -1.98% -2.67 -2.05 -9.41 -3.56
Median -0.90% -1.61 0.16 -7.08 -1.89
60th Percentile 0.09% -0.48 2.10 -6.03 -0.90
70th Percentile 0.96% 0.52 4.24 -4.34 0.51
80th Percentile 2.06% 1.90 7.96 -2.41 2.77
90th Percentile 5.08% 3.82 13.25 0.46 6.36
Maximum 22.17% 17.60 19.23 7.49 18.54
Average -0.54% -1.84 1.77 -7.99 -1.86
t-vaheb -2.237** -7.851*** 2.510** -20.308*** -3.591***
(-) Deviation 59% (-7.98%) 63% (4.76%) 49% (-4.12%) 88% (-9.36%) 65% (-5.49%)
(+) Deviation 41% (4.53%) 37% (3.07%) 51% (7.44%) 12% (2.15%) 35% (4.76%)
# of obs. 490 488 102 285 161
a Column headings: BTI, BTII, BTIII, and BTIV stand for BT Bank of Canada Series I, Series 11, Se-
ries 111, and Series IV NPWs, respectively; TFC stands for Trilon Financial Cow. NPWs.
*: signifcant at 10percent level for a two-tail test; **: significant at 5 percent level for a two-tail
test; ***: significant at 1percent level for a two-tail test.
226 Wei

Finally, although almost all the percentage devia-

tions for all warrants are statistically significantly dif-
ferent from zero, the average size is not extremely large.
The biggest average percentage deviation is -7.99 per-
cent associated with BT Bank of Canada Series IV
NPWs (Table 2, Panel B). Most average deviations are
negative and smaller than 2 percent (in absolute terms).
In terms of dispersion, it can be seen that within the
twentieth and eightieth percentiles, most deviations are
smaller than 7 percent. Note that the t-values for the
positive percentage deviations for BT Bank of Canada
Series I11 NPWs are significant, which means overall
underpricing. However, relative to other warrants, this
particular warrant is the most thinly traded in terms of
both the trading volume and the number of trading
days, and hence the results should be taken with a grain
of salt. Overall, it can be concluded that the models gen-
erally tend to overprice warrants,

Regression Tests, Bias Analysis

Since the answer to the first question is “yes,” the
second question naturally follows: Is the overpricing sys-
tematically related to the model inputs? The regression
test results are summarized in Table 3 through Table 7.
The effects of each independent variable is examined
separately.
(K - S)JK (or (lix,- SX)JKX,), The Degree to
Which the Warrants are in the Money. As shown in
Panel A of Tables 3 through 7, the F-values are all sig-
nificant (except for Trilon Financial Corporation NPW),
which means that the relative deviations are significant.
Examining F-values alone can not reveal whether the
deviations are systematically related to the degree to
which the warrants are in the money. It is necessary to
look at the regression coefficient of the independent vari-
able. It can be seen that all coefficients (a,)are negative.
(BT Bank of Canada Series IV NPW is an exception.)
The above findings seem to indicate that the models
tend to overprice (underprice) in-the-money (out-of-the-
money) warrants. The conclusion of underpricing out-of-
 TABLE3

Results of Regression Tests-Bias Analysis ha

Bankers Trust Bank of Canada Series I NPWs ( N u m b e r of Observations: 490) E
This table reports regression results for Bankers Trust Bank of Canada Series I NPWs. Percentage deviations are regressed on A) degree 3
7
of being in the money, B) time to maturity, C) Nikkei index volatility, and D) natural log of trading volume. The numbers in parentheses 3
are t-values. *: significant at 10 percent level for a two-tail test; **: significant at 5 percent level for a two-tail test; ***: significant at 1
percent level for a two-tail test. For all regressions, the dependent variable PD is defined as PD = (Pkt - Pd) I F o dwhere
, Pktand Pd $
are market and model prices of NPWs. The degree of freedom for the F-test is (1.488). #: significant at 5 percent level; ##: significant at 1
percent level.

K-S
Panel A PD = a0 + al -+ E Panel B: PD = bo + bl.t + E
K
A A
a0 a1 R2 F-value 60 61 R2 F-value
0.00026 -0.09663 0.080 42.338## -0.04505 0.02323 0.059 30.862##
(0.005) (-6.507)*** (-0.863) (5.555)***

Panel C:PD = co + C ~ O , + E Panel D: PD = eo + elln(uoZume)+ E

A A A h
co c1 R2 F-value eo el R2 F-value
0.07392 -0.29147 0.083 44.369## 0.01053 -0.00148 0.001 0.647
(1.434) (-6.661)*** (0.196) (-0.804)
 4
TABLE

Results of Regression Tests-Bias Analysis

Bankers Trust Bank of Canada Series I1 NPWs (Number of Observations: 488)
This table reports regression results for Bankers Trust Bank of Canada Series I1 NPWs. Percentage deviations are regressed on A) de-
gree of being i n the money, B) time to maturity, C) aggregate volatility of the Nikkei index and the exchange rate, and D) natural log of
trading volume. The numbers in parentheses are t-values. *: significant at 10 percent level for a two-tail test; **: significant at 5 percent
level for a two-tail test; ***: significant at 1 percent level for a two-tail test. For all regressions, the dependent variable PD is defined as
P D = ( P k tP-d ) / P n d ,where Pkt and Pdare market and model prices of NPWs. The degree of freedom for the F-test is (1.488). #: sig-
nificant at 5 percent level; ##: significant at 1 percent level.

Irx,- sx
P a n e l A P D = ao+al- +E Panel B: PD = bo + blz + E
Irx,
A A
ao a1 R2 F-value 60 61 R2 F-value
4.0071 -0.1362 0.137 76.996## -0.0647 0.0228 0.062 32.267##
(-0.148) (-8.775)*** (-1.287) (5.680)***

Panel C: PD = co + C ~ V ,+ E Panel D: PD = eo + elln(uoZume)+ E

A h A A
CO C1 R2 F-value eo el R2 F-value
0.0276 -0.1537 0.037 18.661## 0.0149 -0.0031 0.010 4.664#
(1.434) (-6.661)*** (0.288) (-2.160)**
 TABLE5

Results of Regression Tests-Bias Analysis

Bankers Trust Bank of Canada Series I11 NPWs (Number of Observations: 102)
d
This table reports regression results for Bankers Trust Bank of Canada Series 111 NPWs. Percentage deviations are regressed on A) de- $
gree of being in the money, B) time to maturity, C) volatility of the Nikkei index, and D) natural log of trading volume. The numbers in 3
parentheses are t-values. *: significant at 10 percent level for a two-tail test; **: significant at 5 percent level for a two-tail test; ***: sig-
2
nificant at 1 percent level for a two-tail test. For all regressions, the dependent variable PD is defined as PD = (Pkt-Ppm”d)/Pnod, where o,
Pmk‘ and Pdare market and model prices of NPWs. The degree of freedom for the F-test is (1.488). #: significant a t 5 percent level; ##:
significant at 1 percent level.

K-S
Panel A PD = a0 + al-+ & Panel B: PD = bo + b,z +E
K
A A
ao a1 R2 F-value 60 6, R2 F-value
0.1178 -0.3855 0.268 36.670## 4.0414 0.0228 0.005 0.524
(1.910)* (-6.056)*** (-0.576) (0.724)

Panel C: PD = co + c p , +E Panel D: PD = eo + elln(voZurne)+ E

A h A h
co c1 R2 F-value eo el R2 F-value
0.0480 -0.1176 0.011 1.145 0.0571 -0.0047 0.008 0.811
(0.670) (-1.070) (0.794) (-0.90 1)
 p.3
w
0
TABLE6

Results of Regression Tests-Bias Analysis

Bankers Trust Bank of Canada Series IV NPWs (Number of Observations: 285)
This table reports regression results for Bankers “rust Bank of Canada Series IV NPWs. Percentage deviations are regressed on A) de-
gree of being in the money, B) time to maturity, C)volatility of the Nikkei index, and D) natural log of trading volume. The numbers in
parentheses are t-values. *: significant at 10 percent level for a two-tail test; **: significant at 5 percent level for a two-tail test; ***: sig-
nificant at 1 percent level for a two-tail test. For all regressions, the dependent variable PD is defined as PD = ( P k-fPmd)/Pd, where
Pktand P”” are market and model prices of NPWs. The degree of freedom for the F-test is (1.488). #: significant at 5 percent level; ##:
significant at 1 percent level.

K-S
Panel A PD = a0 + al ~ +& Panel B: PD = bo + b,.r + E
K
A A
a0 a1 R2 F-value 60 61 R2 F-value
-0.0933 0.1282 0.047 13.874## 0.1109 -0.0783 0.153 51.217##
(-1.434) (3.725)*** (1.809)* (-7.157)***

Panel C:PD = co + clos+ E Panel D: PD = eo + elln(volume) + E

A A A A
co c1 R2 F-value eo el R2 F-value
-0.0793 4.0021 0.000 0.001 0.0457 4.0109 0.046 13.628##
(-1.190) (-0.033) (0.702) (-3.692)***
 TABLE7

Results of Regression Tests-Bias Analysis b

Trilon Financial Corp. NPWs (Number of Observations: 161) E
This table reports regression results for Trilon Financial Corp. NPWs. Percentage deviations are regressed on A) degree of being in the
3
Y
money, €3) time to maturity, C) volatility of the Nikkei index, and D) natural log of trading volume. The numbers in parentheses are t- 3
values. *: significant at 10 percent level for a two-tail test; **: significant at 5 percent level for a two-tail test; ***: significant at 1 percent
level for a two-tail test. For all regressions, the dependent variable PD is defined as PD = ( P k tPd) - IF"'"', where Pktand P"' are mar- 3
ket and model prices of NPWs. "he degree of freedom for the F-test is (1.488). #: significant at 5 percent level; ##: significant a t l percent
level.

K-S
Panel A PD = uo+ al -+ & Panel B: PD = bo + bl.r + E
K
A A
ao a1 R2 F-value 60 6, R2 F-value
-0.0156 -0.0129 0.000 0.064 0.2329 -0.0954 0.116 20.772##
(-0.236) (-0.252) (3.747)*** (-4.558)***

Panel C:PD = co + clo, +E Panel D: PD = eo + elln(uoZume) + E

A A A h
co C1 R2 F-value eo el R2 F-value
-0.0425 0.0956 0.009 1.489 0.1396 -0.0178 0.279 61.379##
(-0.646) (1.220) (2.486) (-7.835)***
232 Wei

the-money warrants should be taken with a grain of salt,

since the warrants are in the money most of the time
within the sample period. Until further empirical results
are presented, the conclusion has to be treated as tenta-
tive. However, given the negative sign of the regression
coefficient, it is safe to infer that the overpricing tends to
be severe for deep in-the-moneywarrants.
z, Time to Maturity. It is well known that when a
new security is introduced it takes some time for inves-
tors to “learn,”or for the security’s price to become “well
behaved.” This “market learning” would be reflected in
the decreasing of pricing deviations over time. This mar-
ket learning hypothesis can be captured by the regres-
sion Coefficient. Specifically, if there is a market learning
effect, the intercept should be close to zero while the
slope coefficient should be non-zero. As shown in Panel
B of Tables 3 through 7, the F-values are all significant
except for BT Bank of Canada Series I11 NPWs (Table
5). The null hypothesis of market learning is confirmed
in Tables 3 and 4 (BT Bank of Canada Series I and Se-
ries I1 NPWs), where the intercept is not statistically
different from zero but the slope is. For the last two war-
rants (in Tables 6 and 7) the intercept is also statisti-
cally different from zero. This is somewhat disturbing. A
non-zero intercept means that even if the warrant is ap-
proaching its maturity, there still is a pricing deviation.
Finally, no systematic relation between relative devia-
tions and the time to maturity has been found for BT
Bank of Canada Series I11 NPWs (Table 5).
0, (or v = d<+2o,+d), the Volatility. As
shown in Panel C of Tables 3 through 7, the regression
coefficient is all negative, except for Trilon Financial
Corporation NPWs (Table 7). For Series I and I1 BT
Bank of Canada NPWs, the negative coefficient is sig-
nificantly different from zero. Thus it could be inferred
that the models tend to overprice the warrants and the
overpricing is more manifest when the volatility is high.
Zn(uoZume), (log of) the Trading Volume. It can
be seen from Panel D of Tables 3 through 7 that the re-
Nikkei Put Warrants 233

gression intercepts are all close to zero (statistically) but

the regression coefficients are all negative, with three
(out of five) having statistically significant t-values. This
implies that the models over-price the NPWs, and a big-
ger overpricing tends to be related to high trading vol-
umes. Although the coefficients in Tables 3 and 5 are not
significant, the signs of the coefficients are in agreement
with those of other regressions.
Overall, the deviation tests reveal that the models
tend to overprice Nikkei put warrants. But the average
size of the mispricing is generally small (less than 2 per-
cent). The regression tests detect some systematic rela-
tionships between the mispricing and various model
inputs. Specifically, the overpricing is more manifest
when a) the warrants are deep in the money, b) the vola-
tility is high, and c) the trading volume is high.
So far, it has been found that the models overprice
warrants and the overpricing is systematically related to
some model inputs. A natural question that follows is:
What causes the overpricing? The systematic links be-
tween the mispricing and the model parameter inputs do
not necessarily imply that the model inputs actually
cause the mispricing. Instead, it is likely that the mis-
pricing is due to some other unmeasurable factors which
are reflected in the model parameters. There are many
possible factors. A straightforward one is the omission of
credit risk. Unlike conventional options that are guaran-
teed by the exchanges, Nikkei put warrants are guaran-
teed only by the issuers. This will put a downward
pressure on the warrant prices. Another possible factor
is the so-called “Extraordinary Event Clause” applicable
to all Nikkei put warrants. An Extraordinary Event
Clause is specified in the prospectus of an NPW that
would prevent the exercise of NPWs upon the occurrence
of certain abnormal events. It is therefore a protector for
the issuer against undesirable market conditions. The
detailed specifications of an Extraordinary Event Clause
vary across warrants. But the common “events” gener-
ally include the following:

a) suspension or material limitation of trading in

securities on the Tokyo Stock Exchange;
234 Wei

b) suspension or material limitation of trading in

Nikkei 225 futures contracts on both the Sin-
gapore International Monetary Exchange and
the Osaka Stock Exchange; and
c) any outbreak or escalation of national or inter-
national calamity or crisis.

Upon the occurrence of an extraordinary event, the is-

suer will either prevent any exercise of the warrants or
settle an exercise at a lower value.
It is obvious that an Extraordinary Event Clause
would lower the market price of NPWs in order for war-
rant holders to be compensated for the commensurate
risk. The pricing models being tested above do not incor-
porate the effect of this Extraordinary Event Clause, so
the model prices are biased upwards. This is exactly
what has been observed. It is also observed that the
overpricing is more severe when the warrants are deep
in the money or when the volatility is high. This can also
be explained by the omission of the Extraordinary Event
Clause. When the warrants are deep in the money, the
early exercise possibility increases, which makes the Ex-
traordinary Event Clause more relevant. (When the
warrants are deep out of the money, warrant holders
would care (relatively) less about the Extraordinary
Event Clause.) A high volatility makes trading suspen-
sion or limitation more possible; therefore, investors
would require a higher risk premium (hence lower
price), ceteris paribus. It is easy to see that the above
reasoning also applies to the credit risk. (When the war-
rants are out of the money or when the volatility is low,
credit risk is less of an issue.) It is therefore apparent
that the true factors responsible for the overpricing are
the credit risk and the Extraordinary Event Clause,
which are reflected in the related model inputs.15
Exactly how investors price the Extraordinary
Event Clause is a difficult question to answer. Theoreti-
cally, the model prices could be adjusted downwards by
incorporating the probabilities of the extraordinary
events occurring. But the probability estimation will in-
evitably be subjective. A more serious difficulty is that
Nikkei Put Warrants 235

the specifications of the clause are not uniform across

warrants. Also, it is believed that issuers, notwithstand-
ing their desire to protect themselves, are reluctant to
exercise the clause. The main reason is the concern for
goodwill. If an issuer (a bank, e.g.) strives to exercise the
Extraordinary Event Clause, then it may find making
further issues very difficult. This is especially true when
there are many issuers and some of them are lenient on
the clause.
It should be noted that the discussions here are
only speculative and suggestive. Although credit risk
and the provision of the Extraordinary Event Clause
would intuitively justify positive risk premium, the ex-
actly amount of that premium is unknown. Moreover,
there are other potential factors that could cause the ob-
served pricing errors. For instance, the tested models as-
sume constant volatilities and interest rates. To the
author’s best knowledge, no studies exist in the litera-
ture that incorporate stochastic volatilities into the pric-
ing of cross-currency options. Therefore, it is difficult to
precisely assess the effect in our context. However, the
implied volatilities, which are updated daily, are used in
the tests. To the extent that the stochastic nature of
volatilities is partly reflected in daily changes, the strat-
egy of updating the implied volatilities should mitigate
any potential pricing biases (due to assuming constant
volatilities). Nevertheless, until formal empirical results
come into existence, it is not known for sure what effect
stochastic volatilities will have on the warrant prices.
As for the interest rates, it should be realized that,
in general, a constant interest rate pricing model omits
two effects of a stochastic interest rate. The first can be
called the “yield curve effect.” When the interest rate is
stochastic, a pricing model using the spot rate as the in-
put for the constant interest rate will miss the effect of
the non-flat term structure. The second effect is the vola-
tility of the interest rate.16 Choi and Hauser [7] have
shown that the yield curve effect is very strong for cur-
rency 0pti0ns.l~On the other hand, Wei [33] introduces
stochastic interest rates into the pricing of long-term
cross-currency options, and examines the pricing errors
236 Wei

caused by assuming constant interest rates. Specifically,

Wei [331 first corrects for the yield curve effect by using
the discount bond yields (rather than spot rates) as the
constant interest rates, and then tests the volatility ef-
fect. It is found that the constant interest rate models
(with bond yields as interest rate inputs) tend to under-
price cross-currency options, but the pricing errors are
generally very small. In this study, as noted earlier, the
bond yields (as opposed to the spot rates) are used as
proxies for long term interest rates. Therefore, the re-
sults are not subject to the yield curve effect. The only
effect that is omitted is the volatility effect, which is
small anyway. Therefore it is unlikely that the assump-
tion of constant interest rates is responsible for the ob-
served overpricing.

Conclusions
One of the recent financial innovations in the mar-
ket place is the formal listinghrading of foreign index
warrants. Many exchanges have listed long-term options
written on foreign stock indexes. The most popular is the
Nikkei Put Warrant. Despite the ever increasing popu-
larity of foreign index warrants, the literature has been
lacking in formal treatment on these instruments in
terms of both theoretical pricing and empirical testing.
This study is an attempt to empirically examine the
pricing of Nikkei put warrants.
Using data of Nikkei put warrants traded on the
Toronto Stock Exchange, this paper empirically tests
the pricing models developed in Dravid, Richardson,
and Sun [lll,Reiner [241, and Wei [311. It is found that
the models tend to overprice Nikkei put warrants. The
overpricing becomes more severe in the following situ-
ations: 1)the warrants are deep in the money; 2) the in-
dedexchange rate volatilities are high; and 3) the
trading volume is high. It is suggested that the major
reason for the overpricing is the omission of credit risk
and the existence of Extraordinary Event Clauses. Ex-
actly how much risk premium is attached to the war-
rant price is unknown, and it is the subject of further
research.
Nikkei Put Warrants 237

Notes
1. In this paper, “cross-currency options” is a general term for options
on foreign assets. An NPW is a particular type of cross-currency option.
2. See the following section for a classification of NPWs, NPWs on
the AMEX cover only Categories I and 11, while those on the TSE Cover
categories I, 11, and IV.
3. Most of the materials in this section are from Wei [311, which con-
tains more detailed descriptions.
4. Since a Category I11 NPW does not exist, only the pricing of the re-
maining three categories will be discussed.
5. Dravid, Richardson, and Sun [ l l ] are Reiner [241 make the same
argument in a similar setting.
6. The intuition behind the dimension reduction lies in the nature of
the second state-variable, the exchange rate. It is a special variable in the
sense that it serves only as a “medium” between the domestic and the for-
eign economies. For example, SX can be treated as a single variable be-
cause it is simply the dollar price of the foreign asset.
7. Unlike the case in the US, long term discount bonds do not exist in
Japan or Canada. The calculated yields are only approximations of the true
discount bond yields. Factors such as taxation may cause coupon-bearing
bond yields to be different from the pure discount bond yields. But the dif-
ference, if any, should be small as far as warrant pricing is concerned.
8. See Wei [311 for a detailed discussion about this assumption.
9. See, for example, [181 and [271.
10. An extreme example would be to use three-month data around
the 1987 market crash to estimate a forward-looking three-year volatility.
11. pa, is estimated using the past 250 observations. The choice of 250,
the number of trading days in a year, hopefully will balance the two sides of
the aforementioned dilemma. Alternatively, pa, could also be imputed from
warrant prices. This is not done in this study, because the magnitude of PO,
is small, and any potential pricing bias caused by an inaccurate historic es-
timate of po, is likely to be negligible. “he computing costs do not justify the
marginal gain. Of course, one may correctly argue that pox can be ignored
altogether since it is small. Obviously, the choice here is suboptimal.
12. A secant method [23] is employed for the iterative procedures
when calculating implied volatilities.
13. In the Black-Scholes model context, many different ad hoc weigh-
ing schemes have been used in estimating the implied volatility. Latane
and Rendleman [191 use as the weights the partial derivatives of the op-
tion price with respect to the standard deviation. Whaley [341 employs a
procedure that minimizes the residual sum of squares between the model
and the market option prices when imputing an implied volatility. Choi
and Hauser [7] take as the weights the partial derivatives of the option
price with respect to the time to maturity. The scheme used in this study
is similar to that in Sterk [291, which in turn is based on the findings in
Black [2], MacBeth and Merville [20], and MacBeth and Merville [21].
14. See, for example, [71, [El, [251, [271, and [341.
15. The trading volume can be considered as a n indirect identifier of
the two factors. More specifically, a n empirical check reveals that trading
volumes are positively correlated with the implied volatilities.
16. In Meton’s stochastic interest rate model [22], the “yield curve ef-
fect” is reflected in the discount bond price, and the volatility effect is cap-
tured in the overall volatility term.
17. As explicitly noted in their paper, Choi and Hauser did not study
the volatility effect.
 APPENDIX

Nikkei Put Warrants Listed on The Toronto Stock Exchange (TSE)

Expiration Issue Siz- Exercise Fixed

Warrants Issue Date Date #wts Exercise Price Multipleb Exchange Rate Category

AE3 Svensk Exp.

Corp. I (SERWT)" Dee. 1, 1989 Nov. 16,1992 2,366,181 Y35963.74 0.11680 NIA Category I
AE3 Svensk Exp.
Corp. I1 (SERWT)" Feb. 7, 1990 Nov. 16, 1992 1,726,651 Y35963.74 0.11680 NIA Category I
BT Bank of Canada
Series I ([Link]) Feb. 17, 1989 Feb. 17, 1992 9,100,000 Y32174.00 0.11680 NIA Category I
BT Bank of Canada
Series I1 ([Link].A) Jun. 15,1989 Jun. 15,1992 12,375,000 Y33403.00 0.10311 Xa = 11123.47 Category IV
BT Bank of Canada
Series I11 ([Link].B) Feb. 16, 1990 Mar. 16, 1993 4,800,000 Y37460.32 0.00092 NIA" Category I1
BT Bank of Canada
Series IV ([Link].C) Mar. 22, 1990 Apr. 12, 1993 6,000,000 Y29843.34 0.00116 NIA" Category I1
Trilon Financial
Corp. ([Link].N) Feb. 22, 1990 Feb. 22, 1993 3,734,900 Y37460.32 0.00105 N/A' Category I1
aAlthough these two series are issued at different times, they are traded on the AMEX as a single issue, due to the same specifications of
terms.
The exercise multiple is used by issuers to rescale the payoff so that the warrants can be traded with small denomination. For instance,
if a BT Bank of Canada Series I1 NPW is exercised when the index and the exchange rate are at Y25000 and 0.008 Cdn$/Y respectively,
then the payoff will be 0.10311 * (33403D23.47 - 25000 * 0.008) = $7.27 (Cdn).
w.
3
'The fixed change rate is not independently specified. It is reflected in the exercise multiple.
Nikkei Put Warrants 239

References
[l] Beckers, Stan. “Standard Deviations Implied in Option Prices as
Predictors of Future Stock Price Variability.” Journal of Banking
and Finance 5(1981):363-381.
[2] Black, Fischer. “Fact and Fantasy in the Use of Options.” Financial
Analysts Journal 31(July/August 1975):36-72.
[3] Blomeyer, Edward C., and Robert C. Klemkosky. “Tests for Market
Efficiency for American Call Options.” In Option Pricing Theory
and Applications, edited by M. Brenner. Lexington, MA: D.C.
Heath and Co., 1983.
[4] Boyle, Phelim. “A Lattice Framework for Option Pricing with Two
State Variables.” Journal of Financial and Quantitative Analysis
(March 1988): 1-12.
[51 Brenner, Menachem, and Dan Galai. “On Measuring the Risk of
Common Stocks Implied by Options Prices: A Note.” Journal of Fi-
nancial and Quantitative Analysis 19(1984):403-412.
[61 Charis, Donald P., and Steven Manaster. “The Information Content
of Option Prices and a Test of Market Efficiency.” Journal of Fi-
nancial Economics 6(1978):213-234.
[71 Choi, Jongmoo J., and Shmuel Hauser. “The Effects of Domestic
and Foreign Yield Curves on The Value of Currency American Call
Options.” Journal of Banking and Finance 14(1990):41-53.
[81 Choi, J.Y., and Kuldeep Shastri. “Bid-Ask Spreads and Volatility
Estimates.” Journal of Banking and Finance 13(1989):207-219.
[91 Cox, John C., Stephen A. Ross, and Mark Rubinstein. “Option Pric-
ing: A Simplified Approach.” Journal of Financial Economics
7( 1979):229-264.

[lo] Derman, Emanual, Piotr Karasinski, and Jeffrey S. Wecker. “Un-

derstanding Guaranteed Exchange-Rate Contracts in Foreign
Stock Investments.” International Equity Strategies, Goldman
Sachs, 1990.
[111 Dravid, Ajay, Matthew Richardson, and Tong-Sheng Sun. “Pricing
Foreign Index Contingent Claims: An Application to Nikkei Index
Warrants.” Journal of Derivatives l(Fal1 1993):33-51.
[121 Fama, Eugene. “The Behaviour of Stock Prices.” Journal of Busi-
ness 38(1965):34-105. ~

[131 French, Ken R. “Stock Returns and the Weekend Effect.” Journal
of Financial Economics 8(1980):55-69.

[141 Geske, Robert, and Richard Roll. “On Valuing American Call Op-
tions with the Black-Scholes European Formula.” Journal of Fi-
nance 39(June 1984):443-455.
240 Wei

[15] Geske, Robert, and Richard Roll. “Isolating the Observed Biases in
American Call Option Pricing: An Alternative Variance Estimator.”
Working paper, University of California, Los Angeles, 1984.
[16] Gruca, Edward, and Peter Ritchken. “Exchange Traded Foreign
Warrants”. Advances i n Futures and Options Research 6(1933):53-
66.
[17] Hull, John, and Alan White. “The Use of the Control Variate Tech-
nique in Option Pricing.” Journal of Financial and Quantitative
Analysis 23(1988): 237-51.
[18] Jarrow, Robert A. and Maureen OHara. “Primes and Scores: An
Essay on Market Imperfections.” Journal of Finance 44(1989):
1263-1287.
[19] Latane, Henry A., and Richard J. Rendleman. “Standard deviations
of Stock Price Ratios Implied in Option Prices.” Journal of Finance
31(1976):369-381.
[201 MacBeth, James D., and Larry J. Merville. “An Empirical Exami-
nation of the Black-Scholes Call Option Pricing Model.” Journal of
Finance 34(1979): 1173-1186.
[21l MacBeth, James D., and Larry J. Merville. “Tests of Black-Scholes
and Cox Call Option Valuation Models.” Journal of Finance
35(1980): 285-303.

[22] Merton, Robert. “Theory of Rational Option Pricing.” Bell Journal

of Economics and Management Science (Spring 1973): 141-183.

[23] Ortega, James M., and Werner C. Rheinboldt. “Iterative Solution of

Nonlinear Equations in Several Variables.” New York: Academic
PressJ970.
[24] Reiner, Eric. “Quanto Mechanics.” Risk (March 1991): 59-63.
[25] Rubinstein, Mark. “Non-parametric Tests of Alternative Option
Pricing Models Using All Reported Trades and Quotes on the 30
Most Active CBOE Option Classes from 23 Aug. 1976 through 31
Aug. 1978.”Journal of Finance 40(June 1985):455-480.
[261 Rumsey, John. “Pricing Cross-Currency Options.” The Journal of
Futures Markets ll(1991):89-93.
[27] Shastri, Kuldeep, and Kishore Tandon. “Valuation of American Op-
tions on Foreign Currency.” Journal of Banking and Finance
ll(1987):245-269.

[28] Schmalensee, Richard, and Robert R. Trippi. “Common Stock Vola-

tility Expectations Implied by Option Premia.” Journal of Finance
32(1978): 129-147.
[29] Sterk, William E. “Comparative Performance of the Black-Scholes
and Roll-Geske-Whaley Option Pricing Models.” Journal of Finan-
cial and Quantitative Analysis (September 1983):345-354.
Nikkei Put Warrants 241

[301 Trippi, Robert. “A Test of Option Market Efficiency Using a Ran-

dom-Walk Valuation Model.” Journal of Economics and Business
29( 1977): 93-98.
[31] Wei, Jason Z. “Pricing Nikkei Put Warrants.” Journal of Multina-
tional Financial Management 2(November 1992): 45-75.
[321 Wei, Jason Z. “Pricing Forward Contracts and Options on Foreign
Assets-Theories and Empirical Tests.” Doctoral Dissertation, Uni-
versity of Toronto, 1992.
[331 Wei, Jason Z. “Pricing Options on Foreign Assets When Interest
Rates are Stochastic.” Advances in Znternational Banking and Fi-
nance 1(1995):67-89.
[34] Whaley, Robert. “Valuation of American Call Options on Dividend
Paying Stocks: Empirical Tests.” Journal of Financial Economics
lO(March 1982): 29-58.