How to tell if options

are cheap
The use of volatility cones to determine whether options lire
cheap or dear.

Galen Biirghardt and Morton Lane

It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.
The Journal of Portfolio Management 1990.16.2:72-78. Downloaded from www.iijournals.com by COLUMBIA UNIVERSITY on 06/30/12.

he single greatest challenge in trading op- volatility of any traded commodity bounces around
72 tions is determining whether options are cheap or from period to period. Further, the market continu-
expensive. ously revises its estimates of how voIatile rnarkets are

-8
m
P
A trader who buys or sells an option takes a
position on the volatility of whatever underlies the
likely to be.
Both of these volatility traits are illustrated in
option. The price of the option is directly related to Figures 1 and 2. Figure 1, for example, shows a run-
the market’s perception of the underlying’s volatility. ning history of two measures of IZurodollar futures
If the market expects high volatility, option prices will volatility for the year from June 1987 to June 1988.
be high. If the market expects low volatility, option The first is the thirty-day historical volatility of the
prices will be low. June 1988 futures contract (dashed line). The second
The buyer of an option can profit if either of is the level of volatility implied by at-the-money op-
two things happens: tions on the June 1988 futures contract (solid line).
1. The realized volatility of the underlying turns out Figure 2 provides a similar comparison for September
to be greater than the volatility implied by the price
paid for the option,’ or FIGURE 1
2. The market revises upward its perception of how EURODOLLAR JUNE 1988 IMPLIED VEJRSUS HISTORICAL
volatile it believes the underlying is likely to be. (30 TRADING DAYS) VOLATILITIES
By the same token, an option seller can profit (Period: June 1, 1987 to June 13, 1988)
from realized volatility that is lower than implied vol-
atility or from a decrease in the market’s perception
of volatility.
We propose to show how one can use infor- 10 -

mation about the maturity structure of historical vol-

atilities, which we represent with volatility cones, to
determine whether options are cheap or dear.
VOLATILITY IS NOT CONSTANT

What makes option trading a challenging ac-

ou
tivity i s that the world of options does not conform i ,:.....
to the world described by Fischer Black and Myron
Scholes in their seminal work on option pricing. In
JUN87 JUL87 AUG87 SEP87 OC187 NOV87 DEC87 JAN88 ,.::
FEE88 MAR88 APR88 M, A. Y,8.8, JUNE8 JI
that world, volatility is both known and constant. DATE
I8

Instead, what we find is that the observed or historical SERIES -Implied Voi -......--..30-1) H i l l Yo1

GALEN BURGHARDT is First Vice President of Research at Discount Corporation of New York Futures in Chicago (IL
60606). MORTON LANE is President of the same firm. The authors would like to acknowledge th’e empirical work by
Terry Belton and Geoffrey Luce that went into testing the usefulness of volatility cones.
 FIGURE 2 rations are spaced three months apart. In currency
JAPANESE YEN SEPTEMBER 1988 options, both cash and futures, options are spaced
IMPLIED VERSUS HISTORICAL one month apart, at least for the first three months.
(30 TRADING DAYS) VOLATILITIES Thus, one can find implied volatilities that provide
(Period: Jan. 1, 1988 to Aug. 16, 1988)
VOL
..,
am-
. .
volatility forecasts over substantially different hori-
20- i, '..._...
... , zons.
$9-

18- <...
;
.,;
Further, these horizons grow smaller by one
17- ~
day with each day that passes. For example, bond
16. : futures options that trade today might have expira-
15- ~

14- : . . .. tions of three weeks, three months and three weeks,

13 - ,..
i: . and six months and three weeks. Tomorrow, each of
12.
these options will be one day closer'to expiration, and
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.
The Journal of Portfolio Management 1990.16.2:72-78. Downloaded from www.iijournals.com by COLUMBIA UNIVERSITY on 06/30/12.

11-

10 - the corresponding volatility forecast from each of the

9-

e -
options will cover a horizon that is one day smaller.
7-

l- : : FINDING THE APPROPRIATE

. . ..
5- . ...i .'
1.
HISTORICAL HORIZON
JAN88 FEBBB MAR88 APRBB MAYBe
DATE
JUN88 Jut88 AUGB8 SEP88
Because an option has a specific time horizon
SERIES - Imp,i C d "01 ....----..
30-0 Hist VOI - its time to expiration - any comparison of its im-
plied volatility with a fixed-period historical volatility
Japanese yen futures and options from January 1988 is inappropriate. What we need is a tool for comparing
to August 16, 1988. an option's volatility forecast with historical volatility
experience over the same or similar horizons. That is,
HOW CAN YOU TELL IF AN OPTION IS CHEAP?
implied volatilities from four-month options should
In the face of such variability, how can the be compared with four-month historical volatilities,
option trader decide if an option is cheap or expen- while implied volatilities from one-month options
sive? should be compared with one-month historical vol-
Our own experience indicates that traders atilities. Put differently, the only option for which a
place considerable weight on a comparison of implied thirty-trading-day historical volatility is an appropri-
volatility with a fixed-horizon historical volatility. For ate standard is an option with thirty trading days
example, historical volatilities frequently are esti- remaining to expiration.
mated over ten-, twenty-, and thirty-trading day pe-
VOLATILITY CONES
riods. Such periods are short enough for the observed
volatility information to be fairly current and long Our answer to this need is the volatility cone.
enough for the standard deviation estimates to be free Examples are shown in Figures 3 and 4 for Eurodollar
of really wild errors. and Japanese yen futures and options.
Do such comparisons provide a reliable guide The purpose of a volatility cone is to illustrate
to whether an option is cheap or expensive? We think the ranges of volatility experience for different trading
not. horizons. For example, a one-year span can be broken
To see why, consider this problem, An option up into one-day, one-week, and one-month periods.
has several key characteristics, including its exercise This would give us roughly 250 one-day volatility es-
price and its time to expiration. Thus, any reading on timates, fifty-two one-week volatility estimates, and
implied volatility taken from an option is a reading twelve one-month volatility estimates.
on option traders' forecasts of the volatility of the For the cones shown in Figures 3 and 4, we
underlying over the remaining life of the option. Im- started with two years of trading data and estimated
plied volatility taken from an option with two weeks historical volatilities over periods of one month, three
to expiration is a two-week forecast of the underly- months, six months, and a year. Because we moved
ing's volatility. Implied volatility taken from an option forward in one-month increments, this procedure
with two months to expiration is a two-month forecast produced twenty-four one-month historical volatili-
of volatility. ties, twenty-two three-month historicals, nineteen
On any given day, one can find options on six-month historicals, and thirteen one-year histori-
essentially the same underlying but with expirations cals. We could have used daily rather than one-month
that are anywhere from one to three months apart. increments, of course, but the value of the additional
In stock options and in most futures options, expi- information would have been small.
 FIGURE 3 TABLE 1
MATURITY STRUCTURE OF JUNE 1988 Eurodollar Lead Contract Historical Volati3ties
EURODOLLAR VOLATILITY:
Period
IMPLIED VOLATILITY VERSUS 30-DAY Ending" 1-Month 3-Month 6-Month l-Year 2-Year
HISTORICAL VOLATILITY
VOL VOL 1985
I" I I-- November 10.338
December 16.248

1986
January 18.067 15.921
February 13.416 15.998
March 13.957 15.286
April 24.683 18.114 16.7.41
18.872 18.796 17.6.57
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.
The Journal of Portfolio Management 1990.16.2:72-78. Downloaded from www.iijournals.com by COLUMBIA UNIVERSITY on 06/30/12.

May
June 22.809 22.469 19.0153
July 17.923 19.965 19.0.34
August 17.132 19.015 19.2834
September 23.444 19.660 20.810
0 30 60 90 120 150 180 210 240 270 300 330 360
Days t o Option Expiration
October 14.146 20.050 19.6.56 18.343
November 32.047 18.203 18.671 18.363
VDlaLllltleS - IrnPhd 3oO-D"i.tori..l trrkt cone I December 11.351 12.767 16.984 17.669
Time Period f o r Hatortcnl Cones N o v - 8 5 through OcL-87
Source DCNIF Research G r o u ~
1987
74 January 12.402 12.335 16.3.55 17.707

z
6
I
MATURITY STRUCTURE OF SEPTEMBER 1988
JAPANESE YEN VOLATILITY:
IMPLIED VOLATILITY VERSUS 30-DAY
FIGURE 4 February
March
April
May
16.320
14.417
20.313
19.113
13.677
14.392
17.390
18.035
16.112
14.349
14.733
15.9112
17.694
17.727
17.391
37.296
B
0 June 19.957 20.899 17.7'37 17.417
HISTORICAL VOLATILITY 13.456 19.176 18.1.51 17.266
July
VOL VOI
25 August 11.476 15.408 17.5:37 16.679
September 19.014 15.281 18.269 16.107
October 57.415 35.415 28.098 22.528 20.536

Maximum 57.415 35.415 28.098 22.528 20.536

Minimum 10.338 12.335 14.349 16.107 20.536

*Period ending on last trading day of each month

of volatilities was a full twenty-four percentage points

narrower. Note, too, that the range of six-month his-
0 30 60 90 I20 150 180 210 240 270 300 330 360
toricals was smaller still.
Days t o Option Expiration From a trader's perspective, however, the most
important use of the volatility cone is in deciding
Tuns Poriod b r Htstoncal Cone- Llay-88 through Apr-88
Source DCNYF Research Group
whether an option is cheap or expensive. To show
how this can be done, we have overlaid two volatility
The historical volatility estimates used to con- series on each of the cones. Foir Eurodollars, the
struct the cones are shown in Table 1(for Eurodollars) dashed line in Figure 3 represents the thirty-trading-
and Table 2 (forJapanese yen). In the first data column day historical volatility shown in Table 1, and the solid
of Table 1, we find that the highest one-month his- line represents the implied volatility taken from at-
torical volatility was 57% during - not surprisingly the-money June 1988 Eurodollar options. For yen, the
-- October 1987. The quietest month was November solid line in Figure 4 represents the implied volatility
11385, when historical volatility was only 10%. taken from at-the-money September 1988 options.
The chief difference here, holwever, is that time
What Can Be Learned From Volatility Cones?
runs from right to left. For example, in September
One of the most important lessons to be 1987, the June Eurodollar option hlad ten months re-
learned from volatility cones is that short-term his- maining to expiration, and its imp1:ied volatility is rec-
torical volatilities are far more variable than long-term orded in the cone chart at 300 days. In the middle of
volatilities. For example, while one-month Eurodollar December, the June Eurodollar option had six months
historicals ranged from 10% to 57%, three-month his- left to expiration, and its implied volatility is recorded
toricals ranged from 12% to 35%. That is, the range at 180 days.
 TABLE 2 Note, however, that the implied volatility path
Japanese Yen Lead Contract Historical Volatilities lies outside the volatility cone. What does this mean?
Period The June 1988 Eurodollar options had anywhere from
Ending 1-Month 3-Month 6-Month I-Year 2-Year
eight to nine months remaining to expiration during
1986 this time. During the two years leading up to the
May 17.193 episode, the highest level of volatility recorded for a
June 12.600 nine-month period was somewhere between 24% and
July 12.842 13.430
August 10.676 11.353 29%.At 30% and above, June Eurodollar options were
September 9.914 10.870 trading at volatilities higher than the highest nine-
October 11.587 10.501 11.950 month historical volatility recorded any time during
November 10.258 10.227 10.806
December 7.438 9.495 10.097
the two years up to and including October 1987.
Thus, we would count the June 1988Eurodollar
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.
The Journal of Portfolio Management 1990.16.2:72-78. Downloaded from www.iijournals.com by COLUMBIA UNIVERSITY on 06/30/12.

1987 options as expensive, even though implied volatilities

January 12.768 9.982 10.230
February 7.928 9.693 9.822
from these options were well below thirty-day his-
March 8.458 9.880 9.608 torical volatilities. The correctness of this view is
April 11.447 9.430 9.613 10.845 borne out in Figure 5, which shows how both implied
May 10.393 10.276 9.818 10.305
June 9.879 11.120 10.436 10.252 FIGURE 5
July 11.894 11.062 10.231 10.182 MATURITY STRUCTURE OF JUNE 1988
August 12.778 11.372 11.068 10.447 EURODOLLAR VOLATILITY:
September 9.104 11.188 11.117 10.400 IMPLIED VOLATILITY VERSUS 30-DAY
October 13.926 12.075 11.439 10.627 HISTORICAL VOLATILITY
November 10.134 11.129 11.169 10.706
December 13.968 12.974 11.909 11.240 YOL

1988
7 4 IOLI

January 21.603 15.952 13.987 12.274

February 8.159 15.582 13.423 12.302
March 10.050 11.474 13.564 12.400
April 9.781 9.139 12.971 12.229 11.539

Maximum 21.603 15.952 13.987 12.400 11.539

Minimum 7.438 9.139 9.608 10.182 11.539

*Period ending on last trading day of each month

54 ~ " " ' " " " , ' ' ' ' ' ' ' I " ' ~ " ' ' 1 5
0 30 60 90 120 150 180 210 240 270 300 330 360
Where Do the Cone Charts Make a Difference? Days t o Option Expiration

To see how the cone charts put a different slant

on the question of whether options are cheap or dear, h m c Period for Historical Cones Nov-05 Lhrough Ocl-07
Source DCNYF Research Group

consider two different situations:

0 Eurodollar options in late October and early No- and historical Eurodollar volatility played out during
vember of 1987 the months following the Crash. Selling Eurodollar
Japanese yen options in late May and early June of options after the Crash would have been highly profit-
1988 able. Buying Eurodollar options would have been an
Following the market explosion surrounding extraordinarily costly mistake.
the stock market crash on October 19, interest rates A second example is provided in Figure 4. In
(for a fairly brief period, as it happened) became con- late May and early June, observed volatility in Sep-
siderably more volatile. The effect on Eurodollar vol- tember 1988Japanese yen futures was very low. Sim-
atility is apparent in Figure 3. Option traders ilarly, September options on those futures were
responded with substantial increases in the prices trading at very low levels of implied volatility. What
they were willing to pay for options. we can see, however, is that while short-term histor-
The big question is this. During the weeks fol- ical volatilities were very low during this time, Sep-
lowing October 19, were Eurodollar options cheap or tember yen options - which had three months left
dear? The standard comparison would have sug- to expiration - were trading at volatilities that were
gested that they were cheap. Even at 30% volatility, far lower than any three-month historical volatility
options were trading at levels well below the thirty- observed during the two years leading up to this pe-
trading day historical volatilities that were observed riod.
at the time. Thus, we would count the September 1988Jap-
 anese yen options as cheap even though they were realized volatilities. These are
trading at levels higher than thirty-trading day his- The efficient market (all relevani information) hy-
torical volatilities. Figure 6 shows that both implied pothesis, under which the conditional distribution
FIGURE 6
of subsequently realized volatilities is centered
around implied volatility
MATURITY STRUCTURE OF SEPTEMBER 1988
JAPANESE YEN VOLATILITY: The volatility cone hypothesis, under which the dis-
IMPLIED VOLATILITY VERSUS 30-DAY tribution of historical volatilities can be c.sed to im-
HISTORICAL VOLATILITY prove forecasts of subsequently realized volatilities
The efficient market hypothesis is illustrated
by panels A and B of Figure 7. Panel A shows the
distribution of N-day historical volatilities, where N
is the number of days left in the option's life. Suppose,
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.
The Journal of Portfolio Management 1990.16.2:72-78. Downloaded from www.iijournals.com by COLUMBIA UNIVERSITY on 06/30/12.

now, that the level of volatility implied by the prices

of options with N days remaining to expiration is
toward the upper end of the distribution, which puts
it in the upper reaches of the volatility cone.
If the efficient market hypothesis is a reason-
able characterization of the world, then the distri-
bution of realized volatilities would look like the
76 0 3c1 60 90 120 150 180 210
Days to Option Expiration
240 270 300 330 360
distribution drawn in Panel B. In an efficient market
the probability of profiting from a dlelta-balanced sale
of options would be roughly half, as shown by the
shaded area in Panel B.
In contrast, if the volatility cone hypothesis is
and historical yen volatilities came back inside the a better representation of the world, then the distri-
cone well before the September yen options expired. bution of realized volatilities should look like the dis-
Those who bought yen options during late May and tribution drawn in Panel c. In this case, the location
early June reaped handsome profits. Those who were
FIGURE 7
selling yen options suffered substantial losses.
COMPETING VOLATILITY HYPOTHESES
How Useful Are the Cones as a Guide to Cheapness?
A. DISTRIBUTION OF HISTORICAL VOLATILITIES
For those who are steeped in the efficient mar-
ket tradition, a natural question is whether the po-
sition of implied volatility relative to a volatility cone
provides any useful information about whether op-
tions are cheap or dear. Our understanding of effi-
cient markets is that prices embody all relevant
information, including information about the history
of prices. In the world of options, efficiency in this B. CONDITIONAL DISTRIBUTION OF REALIZED
VOLATILITIES UNDER EFFICIENT MARKET HYPOTHESIS
sense would mean that implied volatilities, which are
derived from option prices, should reflect all relevant
information.
If implied volatilities do contain all relevant in-
formation about the future course of volatility, there
are twlo empirical consequences. First, implied vola-
tilities should be unbiased forecasts of subsequently
realized volatilities. Second, information about the C. CONDITIONAL DISTRIBUTION OF REALIZED
VOLATILITIES UNDER VOLATILITY CONE HYPOTHESIS
distribution of historical volatilities, or, what is the
same thing, information about the position of implied
volatility relative to an historical volatility cone should
provide no information about the subsequent distri-
bu tion of realized volatilities.
What we have, then, are two competing hy-
pothes,esabout the relationship between implied and
 of implied volatility in the cone is a predictor. More- volatilities and delta-balanced them to expiration,
over, the probability of profiting from a delta-bal- your average volatility gain would have been 1.71
anced sale of options would be substantially larger percentage points. Your worst loss would have been
than under the efficient market hypothesis. -0.2 percentage points. Again, the average gain is
To test the efficient market hypothesis, we un- different from zero at any reasonable level of statistical
dertook the following exercise with Japanese yen for significance.
April 1986through July 1988. First, we looked for days What these results seem to show is that options
when implied volatilities in yen options were trading markets are not yet entirely efficient, at least when
either in the top decile or bottom decile of implied implied volatilities are trading at very high or very
volatilities over the period. Second, for options with low levels in markets for long-dated options. Put dif-
three to six months to expiration (that is, compara- ferently, the position of implied volatility relative to
tively long-dated options), we calculated realized vol- the volatility cone seems to provide useful and reliable
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.
The Journal of Portfolio Management 1990.16.2:72-78. Downloaded from www.iijournals.com by COLUMBIA UNIVERSITY on 06/30/12.

atilities over the actual days left in each option’s life information about whether the options are cheap or
and took the difference between the resulting vola- dear.
tility and the option’s implied volatility.
How Robust Are the Cones?
The results are shown in Table 3. For days
when implied volatility was in the upper range of the How much data does one need to construct a
volatility cone, we found that implied volatility fore- reliable volatility cone? The answer depends on the
casts of realized volatility were systematically too market you wish to examine, but two years or more
high. The largest difference between realized and im- of data would seem to be adequate, at least if there
plied volatilities was -4.39%. Had you sold those has been no fundamental change in the forces driving
the market.
TABLE 3 Figure 8 shows the history of three-month his-
Difference Between Realized and Implied torical volatilities for Japanese yen futures. How well
Volatility for Japanese Yen Options
would a cone constructed on January 1, 1987, using
With More Than 3 Months Left To Expiration
(April 1986 through July 1988) data for 1985 and 1986, have held up in 1987? The
Implied Volatility Implied Volatility highest three-month historical volatility over 1985and
Summary at High End of at Low End of 1986 was 15.7%; the lowest historical volatility was
Statistic Volatility Cone‘ Volatility Cone**
5.2%. Thus, the upper and lower bounds of the vol-
Maximum -4.39 +4.83 atility cone at three months to expiration would have
Minimum +0.82 -0.20 been 15.7%and 5.2%, respectively.
Mean -2.93 + 1.71 How well did the cone hold up in 1987? Very
Standard Deviation
of Mean 0.19 0.32 well. There were no new highs and no new lows, and
T-Value - 15.40 5.34
Number of Observations 46 20
FIGURE 8
*Implied Volatility in Top Decile
**Implied Volatility in Bottom Decile LEAD JAPANESE YEN FUTURES 3-MONTH
HISTORICAL VOLATILITY
(Period: January 1, 1985 to April 18, 1989)
options and delta-balanced the position to expiration,
you would have made a profit consistent with selling
volatility 4.39 percentage points higher than it proved
to be. The worst case was an error on the high side
of 0.82 percentage points. On average, you would
have sold volatility 2.93 percentage points higher than
realized volatilities. And, as shown in the bottom
panel of the table, the average gain was not only large,
but it was also different from zero at any reasonable
level of statistical significance.
For days when implied volatility was in the
lower range of the volatility cone, the results are just
the opposite. Implied volatility forecasts proved to be
biased downward, with a maximum error of 4.83 per- QUARTER

centage points, and an average error of 1.71 percent-

Lead Series Snilcher on Lhe firs1 day of lhe expiring lead conlract month
age points. Had you bought options at these implied Source : DCNYF Research Croup
 the distribution of volatilities stayed near the middle the option’s life. One-month options should be
of the range. Further, it is easy to see that a two-year judged using one-month historical vdatilities. Three-
cone constructed with data from 1986 and 1987would month options should be judged using three-month
have served quite well through 1988. To be sure, there historical volatilities. This differentiation is important,
were both a new high (16.9% versus 15.3%)and a because the range of observed one-month historical
new low (8.3%versus 8.9%),but the differences were volatilities is considerably larger than the range of
comparatively small. observed three-month historical volatilities.
Second, the market for longer-dated options
CAVEAT
does not appear to be entirely efficient.’ At least in
We have shown that volatility cones can be those instances when implied volaltilities ,ire trading
useful in determining whether options are cheap or at extremely high or extremely low levels, we find
dear. Even so, whenever the cones suggest that op- that implied volatilities are biased forecasts of realized
It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.
The Journal of Portfolio Management 1990.16.2:72-78. Downloaded from www.iijournals.com by COLUMBIA UNIVERSITY on 06/30/12.

tions are unusually cheap or dear, there is merit in volatilities, and that there is valuable inforrnation con-
taking a hard look at the world to see whether any- tained in the distribution of historical vo1a‘:ilities. An-
thing fundamental has changed to affect the long-run other way of putting this is that longer-term volatility
outlook for volatility. forecasts, as embedded in the prices of longer-dated
For example, when the Federal Reserve put a options, appear to be overly sensitive to short-term
reserves targeting policy into place in the fall of 1979, volatility developments. Consequently, volatility
there was a dramatic upward shift in the distribution cones provide a useful and potentially profitable
78 of interest rate volatility. And, when the Fed aban- guide as to whether options are cheap or dear.
3
z doned that policy at the end of 1982, the distribution
3
n shifted downward again.
Thus, whenever the volatility cone standard
Realized volatility and historical volatility are both standard
suggests either a sale or purchase of volatility, it is deviations of percentage changes in the price of the un-
certainly worth one’s while to review changes in eco- derlying security or futures price. Percentage changes most
nomic policy (or in the market for stock options, the often are calculated using daily closing prices, although
capital structure of the firm) that would have a bearing there are several variants and refinements ‘:hat can be
brought to bear on volatility estimation.
on the volatility generating machine that drives the
options market. We first introduced the concept of volatiiity cones in a May
CONCLUSIONS 4,1988, memorandum, ”Is Currency Volatility Low Enough
to Buy?‘ and in an August 29, 1988, follow-up, ”Results of
We have shown two things here. First, the ap- Recommended May 4th Purchase of Currency Volatility.’’
propriate standard of comparison for gauging Since then, we have encountered an October 1988 piece by
Steven Pomerantz, “An Information-Based Model of Vol-
whether implied volatilities are high or low is an his- atility” (MorganStanley), which deals with what the author
tarical volatility that corresponds to the time left in calls a volatilitv time curve.