How Much Does It Cost To Short Sale A Stock Index ?

1. Introduction
In a comprehensive study of the borrowing/lending market for stocks, D’Avolio (2002)
showed that constituents of the S&P 500 are almost always general collateral. Such a
result means that the rebate rate received on the collateral deposit by the borrower of a
stock pertaining to the S&P 500 is very close to the prevailing market interest rates
when the short is undertaken. Using a different sample, Ofek et al. (2003) also showed
that more than 70% of the stocks in their sample have a null rebate rate spread, that is
the difference between (i) the rebate rate and (ii) the standard rebate rate on the majority
of stocks, which is known as the “cold” rate, is zero. Geczy et al. (2002) found that in
their sample, Large and Medium general collaterals have rebate rates that are 8 bp
(0.08%) and 15 bp only below the Federal Funds Effective Rate. Therefore, it may
safely be concluded that the cost of shorting the S&P 500 Index is likely to be
negligible. In addition to the empirical findings cited above, there exist a very liquid
market for S&P 500 futures, and the market for S&P Deposit Receipt (SPDR) recorded
a constant growth of activity since its launching in 1993, two facts that are likely to ease
even more short selling of the S&P 500.
The evidence about rebate rates relative to market interest rates have been
supplemented by analyses of the short sale cost that is implicit in options prices, notably
by Ofek et al. (2003), Evans et al. (2003) and Lamont and Thaler (2003). This approach
builds upon the well-documented evidence that the existence of options helps short
selling of the underlying stocks (see Figlewski and Webb (1993), Sorescu (2000) and

*
We are very grateful to Raphael Franck for extensive Editorial Assistance and especially to
Patrice Poncet for very detailed comments on previous drafts of this paper. The first author
acknowledges financial support from The Maurice Falk Institute for Economic Research in Israel.
2

Danielsen and Sorescu (2001)). The methodology used is a simple test of the Call-Put
Parity (CPP) in which options prices are used to calculate the implicit stock price. The
latter is then compared to the prevailing stock price. Because of short sale restrictions in
the market, the average implicit stock price should be lower than the prevailing stock
price in the market. The (algebraic) difference is indeed what the short seller is willing
to forsake for short selling the stock through the options market. Empirical evidence in
Ofek et Al. (2003) showed that the CPP disparities are asymmetric in the direction of
short sale constraints. They find that on the average the implicit stock price is lower that
the prevailing stock price by 30 bp (0.30%) of the stock price in their sample. Similar
findings are reported in Evans et al. (2003) who show that the average disparity is 16 bp
of the stock price on the average. They all use American stock options and the CPP has
been adjusted accordingly for future dividend and early exercise premium.
Our objective in this paper is to assess the cost of short selling a stock index
using traded stock index options. However, the standard approach in the literature
should be amended for the following reason: it ignores the fact that in the same way that
options could be used for shorting synthetically the underlying, traders may also use
options to buy synthetically the underlying through the options market. This is
especially true for a stock index where CPP disparities reflect both possibilities. Thus
one should disentangle these disparities to assess the cost of buying synthetically the
underlying through the options market from the cost of short selling it synthetically.
One may expect the latter to be higher than the former, the increment representing the
additional cost for the short sale.
To achieve this goal, we suggest the following approach. First, we define the
spread as the difference between the stock index value implicit in options prices and the
prevailing stock index value. If the buyers of the stock index on the options market are
the majority, we expect the spread to be positive. If short sellers of the stock index on
the options market are the majority, we expect the spread to be negative. We can thus
build two sub-samples of CPP disparities, one of positive spreads and another of
 3

negative spreads. The first sub-sample assesses the cost of buying the underlying
synthetically and the second one assesses the cost of short selling it the underlying
synthetically. Although the latter sub-sample is likely to represent a large fraction of the
total sample, we expect the sample of positive spreads not to be small. Around one third
of the cases analyzed by Ofek et al. (2003) had positive spreads.
If our decomposition is sustainable empirically, it means that the previous
measures of short selling costs through options markets used in the literature
underestimated the real cost of short selling. To illustrate our argument, consider two
successive observations in the options market. In the first one, the implicit stock price is
lower than the prevailing stock price (negative spread) by 1.20 %. This negative spread
suggests that if options have been used as substitute for direct trading in the underlying,
it was for shorting it this day. In the next observation, the implicit stock price is higher
by 1 % than the prevailing stock price. This positive spread indicates that the market
was buying the underlying through the options market. Averaging these two
observations yields an average CPP disparity of 20 bp. This is obviously far below the
real cost of shorting which worth 120 bp.
One possible way to assess whether our decomposition is sustainable
empirically is as follows. In days when the spread is positive, the market is a “net
buyer” and thus optimistic. One widely used market indicator for such a situation is the
ratio of call trading volume over put trading volume. When this ratio is larger than one,
calls are more traded than puts and thus, the market is optimistic. When it is lower than
one, the market is pessimistic: puts are more traded than calls. Therefore, in days where
the spread is positive, trading volume on calls should be higher than trading volume on
puts whereas in days where the spread is negative, the reverse should be true.
We implement our methodology in the case of a Stock Index. In addition to this
methodological contribution, the database we use allows us to provide an empirical
contribution to the literature. We consider Stock Index options on the TA 25, a stock
index from the Tel Aviv Stock Exchange (TASE). Whereas most papers dealing with
4

short sale costs focused on individual stocks only, we provide evidence on the cost of
short selling a stock index. We also give an estimation of the cost of buying the stock
index through the options markets. As to the implicit costs of short selling in options
prices, Kamara and Miller (1995) showed that deviations from the CPP are much less
frequent and smaller with European options than with American options. One possible
reason for this lies in the necessity to estimate the early exercise premium in the case of
American options, which may affect the results. In the case of the TASE, the options are
European, nevertheless it is likely that CPP deviations are significant since:
i) easy and cheap substitutes to the underlying do not have a liquid market. Although
Futures contract exist on the TASE, their trading volume is close to zero. Calls with
an exercise price of 1 index point may also be traded but their trading volume is
negligible. A security that is equivalent to the S&P Deposit Receipt has only been
introduced recently and did not exist during our sample period.1
ii) the market for short sales of individual stocks is very limited, if not inexistent.

In addition, the TASE has several properties that render inference using the CPP
easy to implement and probably more accurate since:
i) the stock index is dividend protected. This eliminates another source of error when
testing for CPP deviations. Indeed, for individual stocks, the dividend yield to be
taken into account in the CPP must be estimated. The same is true for most index
options. In our case, there is no need for such estimation since the index is calculated
with all dividends reinvested.
ii) there are no designated market makers on this market. This is a pure supply and
demand market, a feature that makes the empirical findings comparable to the market
for shorting stocks directly. Empirical evidence suggest that the bid and ask spreads

1
A first study by Ackert and Tian (2001) suggests that these securities may increase
substantially pricing efficiency.
 5

posted by market makers may have important sensitivity to market structure and thus
affect inference based on the midpoint between the bid and the ask (Mayhew (2002)).

We show that the average CPP deviation for the buyer of the stock index
through the options market is 28 bp while the average CPP deviations for the short
seller of the stock index is 57 bp. Therefore, the incremental cost for short selling is 29
bp. Hence, annual deviations for the buyer and the short seller respectively amount to
5.35% and 7.01%, i.e. an incremental cost for short sale of 1.66%.2 These numbers may
seem relatively low, but they must be put into perspective: our sample is restricted to at-
the-money options with moneyness of 98% to 102%, compared to the 90% to 110%
usually retained. In addition, the deviations in dollars from the CPP were negligible for
individual stocks. For example, in Ofek et al. (2003), the average stock price and the
average CPP deviations are respectively $32 and 30 bp, thus a dollar deviation of 10
cents. The dollar deviation in our sample is around $23. Following our decomposition,
it costs the buyer around $23 to buy the stock index and it costs the short seller around
$44 to short sale the stock index, thus an incremental cost for short selling of $21 per
contract.
The remainder of the paper is planned as follows. In the next section we give the
institutional background of the TA 25 options market and provide descriptive statistics.
In section 3, we give two perspectives on the implicit costs in the options markets,
through the spread and through the Implied Volatilities. In section 4, we isolate
variables deemed to explain the spread level. The last section contains some concluding
remarks.

2
One should be aware that the conversion on an annual basis should be weighted by the
time to maturity of the options, details are given in section 3.
6

2. Data
Stocks included into the TA 25 Stock Index are those with the highest market
capitalization and are among the 75 shares with the highest average daily turnover. The
Index is updated twice a year, on January and July. The stock index is dividend
protected, so that its value is obtained assuming the reinvestment of any dividend into
the stock distributing the dividend. European options on the stock index started to be
traded on August 1993. Until 2000, options were traded with expiration date every two
months. Since 2000, a new serie of options is introduced each month with three months
to expiration so that each month one traded serie expires. The average daily number of
contracts traded has known a constant growth for the past years, reaching more than 100
000 contracts per day. In terms of the underlying (weighted by the deltas), options
trading represents more than 800% of the turnover in the TA 25 shares. Futures
contracts on the stock index have been introduced on October 1995 with three months
to maturity. There are no official market makers on the TASE and this could explain the
lack of trading in those futures contracts.3
Following Evans et al. (2003) and Ofek et al. (2003), we use closing prices for
TA 25 options. Our sample covers 6 years of data, from January 1, 1996 to December
31, 2001. Our data set, provided by the TASE, include daily options closing price,
trading volume, open interest, number of transactions. Based on this information, we
filtered the data according to the following steps. We first dropped all the options for
which trading volume was zero. Since there are no market makers, we cannot afford to
keep options within a wide range of moneyness. Therefore, we limited ourselves to
options with moneyness (exercise price over index value) ranging from 98% to 102%.
For the sake of comparison, Ofek et al. (2003) used moneyness that ranged from 90% to
110%. For each option, we calculated the implied volatility by inverting the standard

3
On June 17, 2003, TASE announced the introduction of market makers on both derivatives
markets and primitive asset markets.
 7

Black and Scholes formula. The risk free interest rate used in this study has also been
provided by the TASE. It is based on the average of the yield of the Treasury Bills with
maturity from 60 to 120 days. The mean risk free rate for the period was 10% with a
standard deviation of 3%. Following previous studies, all options with implied volatility
of more than 100% have been dropped. Options with the same exercise price and the
same maturity were organized by pairs. At the end of this filtering process, we were left
with 4350 pairs of options. For each day, we had between 1 to 4 pairs. Time to maturity
range from 1 day to 120 days. Since options with maturity inferior to 7 days (which
represents 7% of the data) have characteristics which are very close to the other options,
we did not drop them.

Table I. Summary Statistics for the TA 25 daily return from 1/1/96 to 31/12/01.

Mean 0.03 %
Median 0.05 %
Standard Deviation 1.19 %
Kurtosis 5.97
Skewness -0.54
Minimum -9.41 %
Maximum 7.40 %
Number of Observations 1,423

In Table I, we report the summary statistics of the TA 25 for our sample period. The
daily average return for the TA 25 was 0.03%, that is 12.41% on an annual basis, with a
daily standard deviation of 1.19%. Thus, the index was very volatile during this period
which was characterized by an asymmetric return distribution towards positive daily
returns.
8

In addition, on the average, calls were more expensive than puts as shown in the
following table.

Table II
Summary Statistics for the Options Prices as a fraction of the stock index value from 1/1/96 to
31/12/01. The figures are for options with moneyness (exercise price over underlying value)
between 98% and 102%.

Call Put
Mean 3.7 % 2.7 %
Median 3.7 % 2.7 %
Standard Deviation 1.9 % 1.2 %
Kurtosis -0.584 -0.126
Skewness 0.211 0.178
Minimum 0.0 % 0.0 %
Maximum 10.5 % 7.6 %
Number of Observations 4,350 4,350

While the value of calls was on the average 3.7 % of the underlying, it was only 2.7%
for puts for the same period. These numbers should be kept in mind when assessing the
economic (quantitative) significance of the short sale cost implied in option prices
computed in the next section. Calls were also more traded than puts during this period.
 9

Table III
Summary Statistics for TA 25 Calls and Puts Trading Volume in number of contracts traded and
in dollars (turnover). The sample period ranges from 1/1/96 to 31/12/01. The options have a
moneyness (exercise price over the underlying value) between 98% and 102%. The exchange
rate used for converting NIS turnover in dollars is 4.5 NIS per dollar.

Trading Volume in Trading Volume in Dollar

Number of Contracts (Turnover)

Call Put Call Put

Mean 3,705 3,067 732,777 530,452
Median 1,560 1,462 323,183 240,137
Maximum 54,992 37,906 6,405,767 6,665,878
Minimum 1 1 3.33 17.78
Standard Deviation 5,448.84 4,280.43 974,605.1 725,105.8
Skewness 2.43 2.38 1.99 2.31
Kurtosis 10.56 10.10 7.23 9.836736
Number of Observations 4,350 4,350 4,350 4,350

For our moneyness between 98% and 102%, the average daily trading volume was of
3705 calls and of 3067 puts per serie of options. Given that we have 1 to 4 series per
day, the total daily trading volume for the options with moneyness between 98% and
102% is from 3705 to 14820 contracts for calls, on the average, and for puts, the figures
are from 3067 to 12268 contracts. An interesting feature of the data concerns the high
correlations between call and put trading volume and call and put number of
transactions, which respectively amount to 0.80 and 0.84. This feature of the data stems
from the behavior of some market participants, usually large institutions who can trade
heavily on the underlying, and act unofficially as market makers. They have two
strategies at their disposal: They may sell the underlying on the options market by
selling expensive calls, buy the corresponding puts, and hedge themselves by buying the
underlying on the spot market. Alternatively, they may sell expensive puts, buy the
10

corresponding calls, and hedge themselves by selling the underlying on the spot market.
Since trades are not simultaneous, there exists a residual risk, which is probably less
severe than the residual risk of a market maker who dynamically hedges his position.
There is no dynamic hedging on the TASE market and this may explain the significant
lack of liquidity for the futures contract.
Even for a small market like the TASE, the turnover in the options segment is
significant. The average daily turnover is $732,777 for calls and $530,452 for puts.
Parts of these amounts are related to frictions in the market, and it will be interesting to
assess their economic importance.
With this background in mind, we turn now to the issue of the level of short sale
costs for the TA 25 stock index implicit in the sample we have described.

3. The Cost of Short Selling the TA 25 Index

In this section, we report the results on the CPP deviations. As explained in the
introduction, we follow a standard methodology in the literature that consists in
comparing the stock index value implied in the options prices to the prevailing index
value in the market. When markets are arbitrage free and frictionless, European call and
put options on a stock index with the same exercise price (K) and the same time to
maturity (T-t) should satisfy at each time t until the maturity of the options the
following relation:

C(t) – P(t) = S(t) – Ke-R(T-t) (1)

where S(t) is the stock index value at time t and R is the risk free rate prevailing in the
economy. This relation holds when the underlying does not distribute dividend or
alternatively, when it is dividend protected. Since this is the case for the TA 25, we need
not estimate future dividends which is very valuable for the robustness of our results.
 11

Table IV
Summary Statistics for Call-Put Parity deviations. Column I gives the implied index obtained
from the Call – Put parity minus the closing index; Column II gives the implied index obtained
from the Call – Put parity minus the closing index in percentage of the Closing Index and
Column III gives the implied index obtained from the Call – Put parity minus the closing index
in percentage of the Closing Index on an annual basis. The annualisation is obtained by
multiplying the deviation in percentage by 365 and then dividing by the time to maturity of the
options. The exchange rate used to convert NIS into dollars is 4.5 NIS per dollar. Panel A
contains all the option pairs in our sample, Panel B contains option pairs for which the implied
index is higher than the closing index (positive spread) and Panel C contains option pairs for
which the implied index is less than the closing index (negative spread).

I (in $) II (in %) III (in %)

Panel A
Mean -23.6 -0.32 -3.30
Median -18.4 -0.21 -1.97
Standard Deviation 46.3 0.62 14.15
Kurtosis 2.09 2.70 263.38
Skewness -0.73 -1.11 -4.45
Minimum -310.2 -3.72 -365.53
Maximum 154.1 1.92 342.78
Number of Observations 4,350 4,350 4,350
Panel B
Mean 24 0.28 5.35
Median 18 0.20 2.33
Standard Deviation 22.2 0.27 12.59
Kurtosis 5.08 5.39 400.34
Skewness 1.95 1.99 16.09
Minimum 0.01 0.00 0.00
Maximum 154.1 1.92 342.78
Number of Observations 1,304 1,304 1,304
Panel C
Mean -43.9 -0.57 -7.01
Median -32.8 -0.41 -3.66
Standard Deviation 38.3 0.55 13.13
Kurtosis 3.73 3.67 331.69
Skewness -1.63 -1.77 -14.06
Minimum -310.2 -3.72 -365.53
Maximum -0.01 0.00 0.00
Number of Observations 3,046 3,046 3,046
12

Table IV reports the CPP deviations in our sample. Panel A is the outcome for the
whole sample, where no distinction is made between a positive spread and a negative
spread. On the average, the implicit stock index is less than the prevailing stock index
by 109 NIS (New Israeli Shekels) that represents around $23.6. Relative to the stock
index value, it represents a cost of 32 bp and on an annual basis this cost is 3.30%.
Caution is required when converting the daily result to an annual basis. Indeed, the
options offer the possibility to short sale or to buy the underlying up to the options
maturity. The cost is incurred at the time the trader enters the market up to the options
maturity. Therefore the right way to convert the cost on an annual basis is by weighting
the daily cost by the time to maturity, that is:

365
Annual Spread = Daily Spread ×
Time to Maturity

The standard interpretation of this finding is that a short seller of the stock index
through the options market incurs a cost of $23.6 per contract. Indeed, looking at the
data at this aggregate level leads us to lose important information and eventually to
under- estimate the real cost of short selling the underlying through options. In Panel B,
we report all the cases in which the spread was positive while Panel C reports all the
cases in which it was negative. Panel B is intended to provide information on the
convenience yield paid by the trader who uses options as a device for trading the stock
index. This convenience yield is to be incurred by a short seller who also pays for the
short sell possibility offered by the options. The difference between the two spreads
informs us on this incremental cost for short selling.
As shown by table IV, when the spread is positive, the implied index is larger
than the prevailing index by 107.8 NIS, thus around $24. Hence for each contract traded
through the option market, the convenience cost is $24. It represents 0.3% of the
underlying or an annual cost of 5.35%. Compared to the average call and put prices for
our sample period, this cost is still high. As to the cost incurred by the short seller, it
 13

amounts to 197.53 NIS on the average or around $44. This represents an additional cost
for shorting of around $20 per contract. Relative to the underlying, it amounts to an
annual cost of 7% and thus, an additional cost of 1.65% for short selling. Indeed, several
discussions with the professionals in the TASE confirmed this figure which is evaluated
to be around 1.5%. To sum up, the total cost of short selling the TA 25 Index through
the options is 7%, which is more than twice as much as the 3.3% obtained by simply
averaging the whole sample.
An important remark is in order at this stage. Standard tests applied to Panel B
and C would have led to a difference which is not statistically significantly different
from 0. This would have made further analysis irrelevant. However, this would be the
wrong way to test the significance of our results, especially since the data are highly
skewed. This is why we performed a bootstrap simulation of our data and the results are
given in the following Table.
14

Table V
Bootstrap tests for the mean deviations of the call-put parity. The bootstrap simulations were
performed with replacement and the basis sample was simulated 1000 times. Column I gives
the implied index obtained from the Call – Put parity minus the closing index; Column II gives
the implied index obtained from the Call – Put parity minus the closing index in percentage of
the Closing Index and Column III gives the implied index obtained from the Call – Put parity
minus the closing index in percentage of the Closing Index on an annual basis. The
annualisation is obtained by multiplying the deviation in percentage by 365 and then dividing
by the time to maturity of the options. The exchange rate used to convert NIS into dollars is 4.5
NIS per Dollar. Panel A contains all the option pairs in our sample, Panel B contains option
pairs for which the implied index is higher than the closing index (positive spread) and Panel C
contains option pairs for which the implied index is less than the closing index (negative
spread).

Panel A
I II III
Simulated Mean - $23.6 -0.32% -3.30%
t Statistic -35 -35 -15.26
95% Conf. Int. - $24.9; - $22.2 -0.34%; -0.3% -3.73%; -2.88%
Panel B
I II III
Simulated Mean $24 0.28% 5.35%
t Statistic 39 38 16
95% Conf. Int. $22.8; $25.2 0.27%; 0.3% 4.68%; 6.02%
Panel C
I II III
Simulated Mean - $43.9 -0.57% -7%
t Statistic -64 -57 30
95% Conf. Int. - $45.2; - $42.6 -0.59%; -0.55% -7.46%; -6.55%

This Table reports results from a bootstrap simulation of our sample data with
replacement to compute the distribution of our mean estimates. We performed the
simulation 1000 times for each Panel. The simulated means are almost equal to the
means of our sample appearing in Table IV and statistically significant at almost any
 15

level. Therefore, the means in Panels B and C are statistically different from zero, and
their differences represent the additional cost for short selling.
Overall, the average short sell cost of the TA 25 Index is $20. Given that the
average put daily trading volume is less than the average call daily trading volume, we
take this number as an approximation of the number of contracts synthetically traded in
the market. The average put daily trading volume is 3067 times $20, that is $61,340 per
option series. Since for our moneyness range, we had 1 to 4 options pairs, it turns out
that the total short sell cost was from $61,340 to $245,360. The average turnover of the
put was $530,452 per option serie and per day. Therefore, as a fraction of the turnover,
the short sell cost in the market represents around 12 % of this turnover! Thus, market
frictions do matter for asset pricing in general, and derivatives pricing in particular.
An equivalent perspective on the short sale cost implicit in options prices may
be obtained by looking at the implied volatilities in the options. This is probably a more
accurate way than the implied stock index value although the later approach is the one
retained by the literature. It is well documented, notably by Figlewski and Green
(1999), that traders mark up the implied volatility to account for market imperfections.
Therefore, since market participants are aware that options are used as a substitute to
direct trading in the underlying, they will charge some premium in the options prices
through a mark up of the implied volatilities. When the spread is positive, the market is
“buyer” and thus the implied volatility of a call will be higher than the corresponding
put’s implied volatility. In days where the spread was negative, the implied volatility of
a put will be higher than the implied volatility of the corresponding call.
16

Table VI
Summary Statistics for TA 25 Calls and Puts implied volatilities and their ratio. The sample
period ranges from 1/1/96 to 31/12/01. The options have a moneyness (exercise price over the
underlying value) between 98% and 102%. Panel A contains all the option pairs in our sample,
Panel B contains option pairs for which the implied index is higher than the closing index
(positive spread) and Panel C contains option pairs for which the implied index is less than the
closing index (negative spread).

Call Put Ratio

Panel A
Mean 22 % 24.5 % 1.15
Median 21.8 % 24.3 % 1.09
Standard Deviation 5% 5.3 % 0.33
Kurtosis 5.7 7.19 34.66
Skewness 0.43 0.69 3.57
Minimum 3.1 % 1.3 % 0.27
Maximum 73.3 % 81.7 % 6.45
Number of Observations 4,350 4,350 4,350
Panel B
Mean 24.7 % 22 % 0.89
Median 24.7 % 21.9 % 0.92
Standard Deviation 4.9 % 4.7 % 0.11
Kurtosis 10.29 8.34 8.34
Skewness 1.02 0.06 -1.94
Minimum 3.1 % 1.3 % 0.22
Maximum 73.3 % 48.2 % 1.05
Number of Observations 1,304 1,304 1,304
Panel C
Mean 20.9 % 25.6 % 1.27
Median 20.6 % 25.2 % 1.16
Standard Deviation 4.6 % 5.1 % 0.33
Kurtosis 3.03 8.76 42.71
Skewness 0.07 1.00 4.41
Minimum 3.8 % 5.9 % 0.41
Maximum 3.82 % 81.7 % 6.45
Number of Observations 3,046 3,046 3,046
 17

Table VI provides evidence as to the implied volatilities of call and puts and their ratio
according to the spread sign. When the spread is positive, puts’ volatilities are less than
their corresponding calls’ implied volatility by almost 12%. However, when the spread
is negative, put volatilities are higher than their corresponding calls volatility by 26.5%.
These numbers may seem excessive given the standard findings in the empirical
literature that when options are at the money, call and put implied volatilities are very
close to each other. To check the robustness of the findings, we performed a bootstrap
simulation of our data. Results are reported in Table VII.

Table VII
Bootstrap tests for the mean of the implied volatility of TA 25 Calls and Puts, and their ratio.
The bootstrap simulations were performed with replacement and the basis sample was simulated
1000 times. The sample period ranges from 1/1/96 to 31/12/01. The figures are for options with
moneyness (exercise price over underlying value) between 98% and 102%. Panel A contains all
the option pairs in our sample, Panel B contains option pairs for which the implied index is
higher than the closing index (positive spread) and Panel C contains option pairs for which the
implied index is less than the closing index (negative spread).

Call Put Ratio

Panel A
Simulated Mean 22.02 % 24.55 % 1.15
t Statistic 299 305 226
95% Conf. Int. 21.87 % ; 22.17 % 24.39 % ; 24.70 % 1.14 ; 1.16
Panel B
Simulated Mean 24.80 % 22.09 % 0.89
t Statistic 160 147 305
95% Conf. Int. 24.5 % ; 25.11 % 21.79 % ; 22.39 % 0.89 ; 0.9
Panel C
Simulated Mean 20.87 % 25.63 % 1.27
t Statistic 254 278 211
95% Conf. Int. 20.70 % ; 21.03 % 25.45 % ; 25.81 % 1.25 ;1.28

These results show that differences between the call and put implied volatilities are
significant at almost any level. A natural question is whether volatility differences are
18

function of the moneyness, in other words, whether there is a smile or a smirk in this
short range.
Figure 1

We report the call and put implied volatilities of the TA 25 as a function of their moneyness.
Panel A contain all the option pairs in our sample, Panel B contains option pairs for which the
implied index is higher than the closing index (positive spread) and Panel C contains option
pairs for which the implied index is less than the closing index (negative spread).

Panel A

0.27

0.25
Implied Volatility

0.23
Call
0.21
Put
0.19
0.17

0.15
25

0. 0
75

0. 0

0. 5
50

1. 5

1. 0

1. 5

1. 0

1. 5

1. 0

1. 5

1. 0

1. 5
00
5

7
98

98

98

99

99

99

99

00

00

00

00

01

01

01

01

02
0.

0.

0.

0.

Moneyness

Panel B

0.29
0.27
Implied Volatility

0.25
0.23 Call
0.21 Put
0.19
0.17
0.15
50
75

25

75
00

50

00
25

75
25

00

50

25

75

50

00
98
98
98
99
99
99
99
00
00
00
00
01
01
01
01
02
0.

0.

0.

1.

1.

1.

1.
0.
0.

0.

0.
1.

1.

1.
1.

1.

Moneyness
 19

Panel C

0.27

0.25
Implied Volatility

0.23
Call
0.21
Put
0.19

0.17

0.15
25

50

75

00

25

50

75

00

25

50

75

00

25

50

75

00
98

98

98

99

99

99

99

00

00

00

00

01

01

01

01

02
0.

0.

0.

0.

0.

0.

0.

1.

1.

1.

1.

1.

1.

1.

1.

1.
Moneyness

As shown in Figure 1, implied volatilities are almost flat for calls and puts relative to
options’ moneyness meaning that there is no smile effect.
Some empirical support for our decomposition can be found in the behavior of
the ratio of call trading volume (in units or dollars) to the corresponding put trading
activity. This is a widely used indicator on financial markets which assesses market
optimism. When markets are optimistic, this ratio is greater than one. When the market
is pessimistic, it is below one.
20

Table VIII
Summary Statistics for the trading volume in units and the turnover of TA 25 Calls and Puts.
The sample period ranges from 1/1/96 to 31/12/01. The options have a moneyness (exercise
price over the underlying value) between 98% and 102%. The ratio is the trading volume of the
call over the trading volume of the put. Panel A contains all the option pairs in our sample,
Panel B contains option pairs for which the implied index is higher than the closing index
(positive spread) and Panel C contains option pairs for which the implied index is less than the
closing index (negative index).

Ratio of Trading volume Ratio of Trading Volume in

in number of contracts dollars (Turnover)
Panel A Panel B Panel C Panel A Panel B Panel C
Mean 2.14 2.27 2.08 2.84 3.89 2.39
Median 1.09 1.04 1.11 1.40 1.42 1.39
Standard Deviation 752.0 752.0 665.50 953.5 953.5 232.76
Kurtosis 0.007 0.014 0.07 0.00 0.016 0.00
Skewness 15.84 21.4 12.75 20.47 35.46 7.74
Minimum 41.57 33.23 46.72 37.78 23.69 20.31
Maximum 1,862 1,158.55 2,408.52 1,610.27 591.89 513.02
Number of Observations 4,350 1,304 3,046 4,350 1,304 3,046

Evidence on the relative trading volume of the call and its corresponding put is given in
Table VIII. Before interpreting the data in table VIII, it must be remembered that in the
period covered by our sample, and on the average, calls were always traded more than
puts. As shown in Panel A, the trading volume of the call was 214% more important
than the trading volume of the put in terms of the number of traded contracts, and 284%
more important than the turnover of the put. However, when the spread was positive
calls trading volume was respectively 227% and 389% the corresponding trading
volume of the put. When the spread was negative, calls were traded more than the
corresponding puts by only 208% and 239%. Such evidence about the market activity
lead to the conclusion that positive spreads usually appear when the market is optimistic
 21

and thus, is a buyer of the index through the options market. Negative spreads appear
when the market is a seller of the stock index.

4. Factors that explain the Spread level

We now focus on the relevant variables explaining the level of CPP deviations. For
individual stocks, the natural candidate is the rebate rate as discussed by Ofek et al.
(2003). However, in the case of a stock index, such a variable does not exist and other
explanations must be given. Natural candidates for explaining CPP deviations are:

- Stock Index Return: If the market is optimistic, it is legitimate to expect the return on
to be positive. And the reverse if the market is pessimistic. Therefore, the stock index
return may be expected to have a positive impact on the spread;
- Time To Maturity: Options offer the possibility to trade the underlying but by their
own nature, this possibility is limited in time. Therefore, the longer the time to maturity,
the higher the expected spread should be;
- Risk Free Rate: the options offer, in addition to a convenient way to trade the
underlying, a leverage that allows buying the stock index by paying a small amount of
cash. Therefore, the higher the risk free rate is, the more valuable the service and the
larger the spread;
- Implied Standard Deviation: it is the standard measure of expensiveness or cheapness
on a given day. It is hard to predict the impact of this variable of the options prices. On
the one hand, a high standard deviation implies high options prices and therefore,
investors may be reluctant to pay some additional premia. This implies a negative
impact on the spread. On the other hand, high uncertainty leads the market to charge
high implied volatility. This requires a high spread to make people accept to trade the
derivatives.
22

We do not expect liquidity variables to have any impact on the spread for the simple
reason that we restricted ourselves to very liquid options. Nevertheless, we tested such a
conjecture.

We thus performed the following regressions for each of our three panels:

Spread ti = λ 0 + λ 1 Index Re turn t + λ 2 Time To Maturity ti + λ 3 Risk Free Rate t

+ λ 4 Im plied S tan dard Deviation

and
Spread ti = β 0 +β1 Index Re turn t + β 2 Time To Maturity ti + β 3 Risk Free Rate t
+β 4 Im plied S tan dard Deviation + β 5 Ln (Trading Volume In Units)
+β 6 Ln (TRading Volume In Dollars)

In the first regression, Regression 1, no liquidity variable has been introduced. In the
second regression, Regression 2, we introduced variables related to trading activity in
the options market.
For the Implied Standard Deviation of each option pair, we took the average of
the call and the put implied standard deviation. As to the trading volume variables, we
also took the average of the call and the put trading volume both for the units and the
turnover. The main motivation for this choice is the high correlations between call and
put data. The implied standard deviation of the call and the put have a positive
correlation of 0.6, trading volume in units of 0.8 and turnover of 0.72. Therefore, to
avoid multicolinearities when using data for each option, we took for each pair the mid
point between put and call data. Since the spread is in percentage and the trading
volume in different units of measure, we took the logarithm of the trading volume for
the coefficients of the regression to be meaningful. The results from these regressions is
given in Table IX.
 23

Table IX
We report the results of two regressions where the dependent variable is the spread (, i.e., the
implicit stock index obtained from option prices using the call put parity and the closing index.
In Regression 1, the explanatory variables are the stock index return, the time to maturity, the
risk free rate and the implied standard deviation (which is the average of the implied volatility
of the call and of the implied volatility of the corresponding put). In Regression 2, in addition to
the explanatory variables in Regression 1, we added the traded volume in number of contracts
and in dollars (turnover). The sample period ranges from 1/1/96 to 31/12/01. The figures are for
options with moneyness (exercise price over underlying value) between 98% and 102%. Panel
A contains all the option pairs in our sample, Panel B contains option pairs for which the
implied index is higher than the closing index (positive spread) and Panel C contains options
pairs for which the implied index is less than the closing index (negative spread).

Regression Regression
1 2
Panel A Panel B Panel C Panel A Panel B Panel C
0.59 -0.12 0.84 0.14 0.08 -0.23
Constant (9.56) (-2.62) (13.52) (0.83) (0.63) (-1.35)
Stock Index -1.15 -0.43 -0.26 -1.35 -0.46 -0.71
Return (-1.87) (-0.9) (-0.42) (-2.18) (-0.95) (-1.15)
Time To -0.004 0.001 -0.005 -0.004 0.003 -0.008
Maturity (-12.46) (4.11) (-15.74) (-6.16) (5.57) (-12.47)
Risk Free -3.99 1.84 -5.7 -3.57 1.13 -3.63
Rate (-13.09) (7.4) (-19) (-8.69) (3.6) (-8.75)
Implied -1.31 0.73 -2.48 -1.66 1.12 -3.66
Standard (-6.23) (4.58) (-11.8) (-6.28) (5.57) (-13.9)
Deviation
Turnover 0.069 -0.07 0.24
(2.17) (-3.1) (7.33)
Trading -0.05 0.08 -0.24
Volume (-1.63) (3.66) (-7.2)
(Contracts)

Adj. R2 0.10 0.09 0.22 0.10 0.10 0.24

Before interpreting the findings, some caveats are in order. An important property of the
options prices is that they have non linear relations with several of the variables used in
the regressions. As a consequence, all OLS may not be an adequate estimation
24

procedure. We performed some GMM estimations of the preceding equations in lieu of

the OLS and the parameters estimate was very close to those obtained in the case of the
OLS.
A general word of caution is that, while sometimes the liquidity variables
(turnover and trading volume) are statistically significant, they only add marginally to
the explanatory power of our results. Therefore, our restriction of the moneyness to 2%
mitigated the potential impact of market liquidity that may be crucial due to the lack of
market makers.
We first focus on the findings of Regression 1. In this regression, we employed
standard variables deemed to impact options prices and tested their effect on the spread.
In Panels B and C, it turns out that these variables have the same impact on the spread
whether it is negative or positive. The stock index return impact is not statistically
significant. This is a desirable property of the market since it is likely that the rebate rate
for individual stocks is only marginally, if at all, affected by the stock return.
Concerning the Time to Maturity, the risk free rate and the Implied Standard Deviation,
they have the expected impact. The higher the time to maturity is, the higher the cost
both for the buyer and the seller. It must be remembered that a negative impact on a
negative variable (negative spread in Panel C) is equivalent to a positive impact. The
same is true when the risk free rate is changing. Negative spread is however much more
sensitive to the above variables than positive spreads reflecting the additional impact of
short sale constraints.
Regression 2 provides additional insights as to the parameters affecting the
spread. The additional explanatory power of liquidity variables is marginal. This means
that our findings are not likely to result from some liquidity premium in the market. The
moneyness range used here is certainly at the origin of this result. The impact of the
other variables is not changed compared to Regression 1.
 25

5. Concluding Remarks
Market frictions have long been a concern to practitioners and academicians as to their
impact on primitive and derivatives assets pricing. No less worrisome was the
perception by the market of such imperfections. While theoretical analysis of such
questions is well advanced, only few opportunities to assess these costs empirically are
available. In the TASE market where there are no market makers, and where European
options are written on a dividend protected index, we have an interesting opportunity to
quantify the importance of such market frictions. Our findings show that these frictions
are quantitatively economically significant, thereby justifying the researchers’ concern.
26

References
Ackert, L. and Y. Tian, 2001, Efficiency in index options markets and trading in stock
baskets, Journal of Banking and Finance 25, 1607-1634.
Danielsen, B. and S. Sorescu, 2001, Why do option introductions depress stock prices?
A study of diminishing short sales constraints, Journal of Financial and
Quantitative Analysis 36(4), 451-484.
D`Avolio, G., 2002, The Market for Borrowing Stock, Journal of Financial Economics
66(2-3), 271-306.
Figlewski, S. and G. Webb, 1993, Options, short sales and market completeness,
Journal of Finance 48(2), 761-777.
Figlewski, S. and T. Green, 1999, Market risk and model risk for a financial institution
writing options, Journal of Finance 54(4), 1465-1499.
Geczy, C., D. Musto and A. Reed, 2002, Stocks are special too: an analysis of the equity
lending market, Journal of Financial Economics 66(2-3), 241-269.
Evans, R., Geczy, C., D. Musto and A. Reed, 2003, Failure is an option: Impediments to
short selling and options prices, Working Paper.
Kamara. A. and T. Miller, 1995, Daily and Intra daily tests of European Put-Call Parity,
Journal of Financial and Quantitative Analysis 30(4), 519-539.
Lamont, O. and R. Thaler, 2003, Can the Market Add and Subtract ? Mispricing in Tech
Stock Carve-outs, Journal of Political Economy 111(2), 227-268.
Mayhew, S., 2002, Competition, Market Structure, and Bid Ask Spreads in Stock
Option Markets, Journal of Finance 57(2), 931-958.
Ofek, E., M. Richardson and R. Whitelaw, 2003, Limited arbitrage and short sales
restrictions: evidence from options markets, Journal of Financial Economics,
forthcoming.
Sorescu, S., 2000, The effects of options on stock prices: 1973 to 1995, Journal of
Finance 55(1), 487-514.