Computing the Market Price of Volatility Risk

in the Energy Commodity Markets

James S. Doran
Department of Finance
Florida State University
and
Ehud I. Ronn
Department of Finance
University of Texas at Austin
Forthcoming Journal of Banking and Finance
Revised: January 21, 2008

Acknowledgments: The authors acknowledge the helpful comments and suggestions of Ty-
rone Callahan, Dave Chapman, Li Gan, Ron Kaniel, Ramesh Rao, Laura Starks, Sheridan
Titman, Stathis Tompaidis and Hong Yan, and of seminar participants at the University
of Texas at Austin, Florida State University, Federal Reserve Bank of Chicago, Southern
Methodist University, 13th Annual Derivatives Securities Conference, and the Commodities
and Finance Centre Conference. The authors would additionally like to thank Vince Kamin-
ski and Brandon Peters for providing data in support of this project. Special thanks are due
the anonymous referee whose suggestions significantly improved the paper.

Communications
Author: Ehud I. Ronn

Address: Department of Finance

McCombs School of Business
University of Texas-Austin
Austin, TX. 78757

Tel.: (512) 471-5853 (Office)

E-mail: eronn@mail.utexas.edu

Electronic copy available at: http://ssrn.com/abstract=628562

 Abstract

In this paper we demonstrate the need for a negative market price of volatility risk
to recover the difference between Black-Scholes (1973)/Black (1976) implied volatility and
realized term volatility. Initially, using quasi-Monte Carlo simulation, we demonstrate nu-
merically that a negative market price of volatility risk is the key risk premium in explaining
the disparity between risk-neutral and statistical volatility in both equity and commodity-
energy markets. This is robust to multiple specifications that also incorporate jumps. Next,
using futures and options data from natural gas, heating oil and crude oil contracts over a
ten year period, we estimate the volatility risk premium and demonstrate that the premium
is negative and significant for all three commodities. Additionally, there appear distinct
seasonality patterns for natural gas and heating oil, where winter/withdrawal months have
higher volatility risk premiums. Computing such a negative market price of volatility risk
highlights the importance of volatility risk in understanding priced volatility in these financial
markets.

Electronic copy available at: http://ssrn.com/abstract=628562

 Computing the Market Price of Volatility Risk
in the Energy Commodity Markets

1 Introduction
The importance of energy markets has increased with the development of futures and options
markets, and with the always-important impact of energy on the economy. In this paper,
we seek an in-depth understanding of priced volatility in the energy markets, as well as
quantitatively displaying the mirror-image aspect of the energy and equity markets: Whereas
prices and volatilities are negatively correlated in the equity markets, they tend to display a
positive correlation in the energy markets. Because of this positive correlation, computing a
negative market price of volatility risk in energy markets may imply a negative market price
of commodity risk in the energy markets and consequently an upward bias of energy futures
contracts prices relative to expected spot prices.
Financial research has made numerous advances in testing the sensitivity of a given data-
generating process to changes in the instantaneous parameters that govern it. Currently there
are a wealth of parametric models attempting to explain stock price movement. The most
notable extensions to the Black-Scholes (1972) model are the inclusion of stochastic volatility
`a la Heston (1993), and the inclusion of jumps by Bates (1996) and others. Recent additions
such as volatility jumps introduced by Duffie, Pan, and Singleton (2000) and Bates crash
risk are further advancements to the well-tested Black-Scholes model. A problem for current
researchers is the ability to reconcile time-series and cross-sectional differences in spot and
option prices by fitting a given underlying model to capture the distributions of both returns.
The parameter of particular interest is the market price of volatility risk. While extensive
research has focused on stochastic volatility models, there is conflicting information on the
impact of the market price of volatility risk. Recently, findings in Bakshi and Kapadia
(2003), Coval and Shumway (2001), Pan (2002), and Doran (2007) address the direction and
magnitude of the market price of volatility risk, but with contradictory conclusions. Our
hypothesis is that the market price of volatility risk is indelibly linked to the bias in Black-
Scholes implied volatility. If the market price of volatility risk is significant and negative, this
potentially explains the upward-bias observed in Black-Scholes implied volatility (henceforth,
BSIV) as well as contributing to the Bates (1996) finding that out-of-the money (OTM) puts
are expensive relative to other options (the so-called volatility skew). Additionally, Eraker
(2004) and Bates (2000) have documented that selling options results in Sharpe ratios that
are significantly higher than the Sharpe ratio of traditional equity portfolios. These results
suggest large premiums for exposure to volatility risk, and lend justification for option traders

Electronic copy available at: http://ssrn.com/abstract=628562

tendency to be short options.
Another strand of the literature has focused on the appropriate econometrics to estimate
continuous-time models. Chernov and Ghysels (2002) use the Gallant and Tauchen (1998)
EMM technique, Pan (2002) applied an IS-GMM framework, and Jones (2003) and Eraker
(2003) have used Bayesian analysis to arrive at their estimates. These works have improved
our understanding of equity market pricing dynamics and the risk premium within these
markets by combining spot and options data on the S&P index. By comparison, the work
in energy markets incorporating options is mostly unexplored. As Broadie, Chernov, and
Johannes (2007) point out, it is very difficult to arrive at precise parameter estimates for
multiple risk premia, especially when one is the volatility risk premium. Noting this issue, we
attempt to estimate the volatility risk premium in energy markets by combining both implied
and realized volatility in two-step estimation procedure. Using Black (1976) implied volatil-
ity (BIV) and implementing the relationship between expected volatility and instantaneous
volatility as given in Ait Sahalia (1996), we infer the instantaneous risk-neutral parameter es-
timates from the discrete-time analogue. The market price of volatility risk is then deduced
via calibration from these instantaneous parameters and the difference between expected
realized volatility and the actual realized volatility whilst avoiding the under-identification
problem.
The findings suggest a significant negative market price of volatility risk for three energy
commodities natural gas, crude-oil and heating oil. Additionally, there appears to be a
strong seasonal component to the volatility risk premium for natural gas and mild seasonality
in heating oil. For robustness, in conjunctions with the volatility risk premium, the market
price of risk and the correlation between the price and volatility innovations are estimated in
the Hansen (1982) GMM framework, in a test for model misspecification and stability of the
parameter estimates. While we are unable to make a definitive statement on the commodity
price risk premium, the volatility risk premium remained negative and significant using this
alternative specification.
The remainder of the paper is organized as follows. Section 2 will review the current
findings on the market price of volatility risk. Section 3 will introduce the simulation and
the findings for the various parametric models tested. Section 4 will detail the estimation
procedure and its application to the market price of volatility risk. Section 5 concludes.

2 The Volatility Risk Premium

The current state of the literature on volatility risk premium has focused on equity markets,
in part due the prevalence of option data on indices such as the S&P 500. Much of the

2
attention paid to energy markets focuses on modeling and estimating convenience yields.1
The general findings suggest that convenience yields are positive, since futures prices are
generally below spot, are negatively correlated with inventories, and tend to be time-varying.
While estimation of the convenience yield is typically conducted using spot and futures
prices, estimating the market price of volatility risk needs be derived from information in
option prices. With increased access to energy option data, estimation of the volatility risk
premium is now more reliable. By using the realized volatility of the futures contract price,
and implied volatility from the options on those futures, the volatility risk premium can be
inferred.
For equities, the consensus of works such as Bakshi and Kapadia (2003), Carr and Wu
(2004) and Coval and Shumway (2001) find that the market price of volatility risk is negative.
Several authors, such as Pan (2002) and more recently, Broadie, Chernov, and Johannes
(2007), disagree with the marginal impact of this risk parameter, noting that empirically
disentangling multiple risk premiums is problematic. Our intuition regarding the sign of the
market price of volatility risk is informed in part by the notion that options are purchased as
hedges against significant declines in the equity market, and buyers of the options are willing
to pay a premium for such downside protection.2 In addition to the high Sharpe ratios in
trading option, pointed out by Bates (2000) and Eraker (2003), Jackwerth and Rubinstein
(1996) have also suggested that at-the-money (ATM) implied volatilities are systematically
higher than realized volatilities, a phenonmenon that clearly be explained by a negative
volatility risk premium.3
While implied volatilities are higher than realized volatilities in energy markets, too, the
dynamics of the energy markets differ from those of equity markets in specific ways:
1. Higher market prices tend to coincide with higher volatility.

2. Beta coefficients tend to be negative for commodities markets.

3. There is a significant term structure of volatility and (at least in some markets) sea-
sonality for energy prices.
In the presence of these major differences we show that the market price of volatility
risk for energy contracts is also negative and significant. This is consistent with the Doran
1 Inparticular, see Schwartz (1997), Hilliard and Reis (1998), Schwartz and Smith (2000),
Routledge, Seppi, and Spatt (2000), Casassus and Collin-Dufresne (2005) and many others.
2 This would have the appearance of buying market volatility, since high volatility coincides

with falling market prices, as pointed out by French, Schwert and Stambaugh (1987) and
Nelson (1991).
3 Jackwerth and Rubinstein (1996) demonstrate this by recovering the probability distri-

butions from option prices.

3
(2007) finding that option prices incorporate volatility risk in natural gas. First, simulation
evidence will highlight that only the market price of volatility risk can explain the difference
between implied and realized volatility, even in the presence of jumps. Second, we will
estimate the volatility risk premium in three important energy commodities by incorporating
the volatility difference between implied and realized and thus confirming the simulated
numerical evidence.

3 Monte Carlo Simulation

To demonstrate the need for a negative market price of volatility risk in energy, we adopt
a quasi-Monte Carlo simulation to test several parametric-model candidates. The choice
of possible model candidates is almost boundless, ranging from Gaussian, non-Gaussian,
continuous and discrete and including the Duffie, Pan, and Singleton (2000) double-
jump model. For the sake of brevity we will focus on the most common, i.e., Bates (1996)
stochastic volatility with jumps.4 Our purpose here is to demonstrate that the volatility risk
premium is negative and significant, and is the only parameter that can explain volatility
differences between implied and realized volatility, even in the presence of jumps and jump
premium.
We have chosen to model both the jump size risk as well as the jump intensity risk as
proportional to the level of volatility, analogous to what is typically done in equities. To test
for this proportional relationship in energy markets an ordered Probit test is conducted on
the frequency of the jumps relative to volatility using crude oil futures. The independent
variable chosen is beginning-of-month volatility, which is then sampled over the next twenty-
two days for jump frequency and jump size. Daily futures price movements of 3% and 5%
are selected to signify a jump in a given futures contract. If a month had one jump of 3% or
5%, the dependent variable is set equal to one; if there are two jumps the dependent variable
is set equal to two, and so forth. The results of the ordered probit regression reveal positive
t-statistics of 3.47 and 5.23 for absolute value changes of index on volatility. This suggests
a proportional relationship between jump intensity and volatility in energy. For magnitude
of the jumps, the absolute values of the magnitude of the price movements above 3% are
summed within the twenty-two day periods, and tested on volatility using an OLS regression
with corrected standard errors for jump size on volatility level. Additional regressions are
run examining only positive upward spikes as well as only negative jumps. The t-statistics
are 3.48 for the absolute value jumps, 5.46 for the positive upward spikes, and 4.12 for the
4 For
a good discussion on option pricing model performance, see Bakshi, Cao and Chen
(1997), and Christoffersen, Jacobs and Minouni (2006).

4
negative jumps. These results suggest modeling the jumps proportional to the underlying
volatility/variance level for energy commodities.
For completeness, we present the Bates (1996) stochastic volatility model with jumps,
henceforth the SVJ model, for the futures process in equations (1) (4) below: The price
process follows Geometric Brownian Motion, while the volatility process is Hestons (1993)
mean-reverting square-root process. The price jumps are captured by the jump-process ,
which is conditional on volatility and has an arrival rate which is Poisson.

Risk-Neutral Process:

dFt = dt t Ft dt + Ft dzs (1)

   q 
dt2 = t2 dt + t dzs + 1 2 dz (2)

The transformation from risk-neutral to real-world results in:

Real-World Process:

dFt = [f t + t (1 ) ( )] Ft dt + dt t Ft dt + t Ft dzs (3)

h   i  q 
dt2 = t2 + t2 dt + t dzs + 1 2 dz (4)

where

f = market of price of risk

= market price of volatility risk
= market price of jump intensity risk
= market price of jump size risk
= the volatility of volatility
= speed of mean-reversion
= long-run mean
= correlation between the price and volatility processes
, and = the jump intensity (arrival rate), the risk-neutral and real-world
jump size, respectively.
dzs and dz = geometric Brownian motions under the real-world measure P
dzs and dz = Brownian motions transformed under the risk-neutral measure
Q

5
 The term t Ft dt is the compensation for the instantaneous change in returns as a
result of jump-process . The market prices of risk are introduced in the transformation
from risk-neutral to real-world densities using the Girsanov Theorem. Whereas in the risk-
neutral world the process is governed by two Wiener processes dz , dz under the Q-measure,
the real-world processes dz, dz are governed under the transformed P -measure.5 For the
simulation the jumps are drawn from a N ( , 2 ) for the risk-neutral distribution and
N ( , 2 ) for the real-world distribution, where 2 is the variance of jump size. The
state-dependent arrival rate for the risk-neutral jump is given by t while the real-world
jump has an arrival rate of (1 )t . This incorporates a premium for the uncertainty in
the jump arrival.
Two variance reduction control techniques are implemented to help reduce the standard
error of the option value and improve the efficiency of the results.6 For each sample path,

random shocks are drawn from N (0, 4t) at 4t intervals over the life of the option.
For a 1-month to expiration option, 22 random shocks are drawn for the price process for 22
trading days. This procedure is replicated for the volatility process governed by separate,
but correlated Brownian motions. Since it is necessary to generate both risk-neutral and
real-world paths, a total of four random draws are needed for the evolution of one day. The
option value for a call and put are then calculated as the average value across all n paths:

Cnt = er(T t) max {FnT K, 0} (5)

Pnt = er(T t) max {FnT + K, 0} (6)

under the risk neutral measure, where FT is the futures price at maturity, K is the strike
price, n represents the particular path, and r is the risk-free rate. This process is repeated for
30,000 runs. While slightly excessive, it is important to reach an efficient estimator for the
call value since inferring the correct volatility, especially for OTM options, relies on precise
estimates within fractions of a cent.7 The final call and put values are then:

n
1X
Ct = Cit (7)
n i=1
5 See Pan (2002) for a discussion of both risk-neutral and real-world processes.
6 We have used the antithetic variable and control variant techniques using Black-Scholes
as the known analytical solution. In addition, we have run quasi Monte-Carlo simulations
using the Sobol sequence to generate results; this is done to determine the number of runs
to achieve efficiency.
7 As noted in Hentschel (2003) and Christoffersen and Jacobs (2004), pricing and implied

volatility errors can be large when call prices are measured inaccurately, or the loss function
mis-specified.

6
 n
1X
Pt = Pit (8)
n i=1

By finding C, P and given the starting value for the futures price, risk-free rate, time
to expiration and strike price, the Black formula can be inverted and the estimate for BIV
found:

h  i
C t = ert Ft N (d1 ) KN d1 iv, t T t (9)
i2
ln (Ft /K) + (T t)
d1 =
iv, t T t

where N () is the cumulative Normal distribution, T t is the time to maturity, and iv

the Black implied volatility.
Applying the Black formula while using a data generating process that incorporates
jumps and stochastic volatility may appear contradictory, but is entirely consistent with
determining the effect of the various risk premiums. Since implied volatility is inverted from
the Black-Scholes/Black formulas, it is necessary to create simulated values that are proxies
for such data. In calculating the BIV, our intent is to examine the disparity between BIV
and realized volatility.
In our simulations, the option is considered European, to avoid the problems of early
exercise. Whereas our empirical estimation will take full account of the American-style
options traded on the NYMEX, our simulation is designed to highlight volatility differences
between implied and realized volatility, and the lack of an early exercise feature for ATM
options is of little consequence on this difference. It is important that the time interval for
sampling the random shocks be small, otherwise it could lead to an instantaneous shock to
the volatility process resulting in a negative variance. For current purposes, the shocks are
bounded so that there are zero negative-variance realizations. Nevertheless, even without
this restriction the variance process infrequently dips below zero.8
To solve for the realized-term volatility, brv , two methods are adopted. The first is to
sample the returns of the futures contract over the remaining life of the option:
v
t
1 TX
u
(ri r)2 ,
u
brv, t =t (10)
T t i=1

where
8 The larger is, the larger the changes in variance. This can be countered by a smaller
sampling interval. This still presents potential problems when is large and may result in
higher implied variances due to truncation.

7
 ri = return on day i on the futures contract
r = the average daily return of the futures contract over the options life
Additionally, this daily volatility estimate is annualized to make an easy comparison with
the BIV. This estimate of realized volatility has been used by Christiansen and Prabhala
(1998) and Doran and Ronn (2006).
The second measure is
sP
[ ln (FiT /Ft ) r ]2
n
i=1
brv, t = (11)
n
where the term variance is calculated, in contrast to equation (10)s daily variance within
the period. Calculating the mean return, r, requires the transformation from Normal to
LogNormal, with the mean return adjusted from to 2 /2. In turn, this requires ex-
ante knowledge of the standard deviation of the distribution, , which is unknown prior
to computing r. We first transform the initial normal mean, and calculate 2 from this
initial estimate of r. From this first estimate of 2 , the initial mean is then adjusted to a
LogNormal estimate. The process to find 2 with this transformed estimate of r is repeated
until convergence is achieved for both values, such that the 2 is exactly the same for the
standard deviation and the LogNormal adjustment for r.9
Both estimates of volatility are calculated for every path generated, and averaged to
arrive at a final estimate of realized-term volatility. The average realized volatility is then
compared to the implied volatility to determine the effect of the instantaneous parame-
ters on the resulting volatility difference. This process is repeated with small changes in
the instantaneous parameters, where new paths are generated, and new estimates for im-
plied and realized volatility are calculated. By adopting this methodology, the individual
continuous-time parameters effect on the level of implied and realized volatility difference
can be isolated.

3.1 Simulation Results

While the simulation focuses on the Bates (1996) SVJ model, it is easy to extend the model
to other specifications such as a constant elasticity of variance model (CEV), or the nested
Heston (1993) stochastic volatility model. This allows us to test a variety of specification to
test if the results are model dependent.10 To generate volatility differences, all the instanta-
neous parameters are allowed to vary over a range of different values. In each case, only one
9 No implementation required more than four iterations to achieve convergence.
10 These models are available upon request but excluded from the paper as there is little
variation in the results. Additional tests included perfect and zero correlation cases along
with using the spot price as the underlying.

8
parameter is allowed to vary on any given simulation, while all others where held constant
at pre-determined values. Since our knowledge of many instantaneous parameter estimates
for energy models is limited with much of the past focus dedicated to spot- rather than
the futures-price process, and constant rather than stochastic volatility initial values are
given by the empirical findings of Schwatz and Smith (2000), Casassus and Collin-Dufresne
(2005), and the equity findings in Pan (2002).
The results presented in Table 1 highlight the volatility difference over a given range of
instantaneous parameter values. The results are shown for two specific models, SVJ and
SVJ0 (stochastic volatility with jump but = 0), using 30-day (22-trading days) ATM
options for the implied volatility and the underlying futures contract for realized volatility.
The key insight gained from this table is how the volatility risk premium is the key parameter
to generate the positive difference between implied and realized volatility. When = 2 the
volatility difference ranges from 2.39% to 3.12% regardless of the instantaneous parameter
values, except for . Since the risk premium is multiplied by , higher levels of result in
greater volatility differences. As shown in the table, when = .7 the volatility difference
is 4.71%. However, in the SVJ0 model different values of have no effect on volatility
difference. Over all parameter values in the SVJ0 case, the volatility difference ranges from
0.03% to 0.62%. This provides numerical support for the ability of a negative volatility risk
premium to help explain why Black implied volatility is higher than realized volatility in
energy markets.
An additional objective is the need to capture some of the empirical regularities observed
in the cross-section of option prices in energy markets. Typically, as is the case in equities,
volatility skew patterns can be explained by correlation , and jump parameters , ,
and 2 . For the S&P 500, the volatility skew tends to be negative and exhibits significant
kurtosis.11 Models that fit this behavior result in negative values for the correlation and
jump-size parameters. In energy markets, the volatility skew tends to be positive, so in the
simulation and are set to positive values, to generate the appropriate pattern. However,
while generating a positive volatility skew is a necessary feature for reasonable parameter
estimates, our main concern is examining the effect of changing risk premium on the shape
of this skew and the difference in BIV and realized volatility. This is done in particular to
highlight the importance of the volatility risk premium, and to demonstrate how jump risk
premium alone cannot explain the difference in implied and realized volatility. As a result,
we have re-run the simulation using a strike/futures price ratio between 0.8 to 1.2 under four
criteria: 1) No risk premium; 2) = 2, with all other risk premiums equal to zero; 3)
11 The additional benefit of using a double-jump model appears to be in capturing excess
kurtosis see Bakshi and Cao (2004).

9
 = .5, all other risk premiums equal to zero; and 4) = .09, all other risk premiums
equal to zero. The instantaneous parameters are kept the same as in the base case as shown
in Table 1. The results of this simulation are shown in Figure 1.
The results highlight two particular effects. First, in the case where there is no risk
premium, a positive skew is generated by specifying a positive coefficient for . This same
pattern is observable for the case when = 2, with the difference between the two
lines approximately constant (2%) across moneyness. For the ATM point (strike/futures
ratio = 1), there is almost no volatility difference between Black implied volatility and
realized volatility for the case of no risk premium, while for the case of = 2, the
difference is 1.79%. This highlights the need for a negative volatility risk premium, but
also suggests there is little to no cross-sectional differences. The second effect shows how
jump size risk premium can enhance the positive skew. As the moneyness level increases,
the volatility difference increases from an ATM difference of almost zero, to an out-of-the
money (strike/futures ratio = 1.2 for a call option) of 4.10%. This in turn demonstrates
the importance of a jump premium in explaining volatility skew properties, where out-of-the
money (OTM) calls and in-the-money (ITM) puts are volatility-expensive relative to other
options in energy. However, in contrast to the volatility risk premium, there is almost no
volatility difference between BIV and realized volatility with a positive jump premium at
the ATM point.12 This suggests the volatility risk premium has an impact on all option
values, while jump risk premium impacts OTM calls/ITM puts. This is further consistent
with the Doran, Peterson and Tarrant (2007) findings that jump premiums are contained
in the OTM portion of the equity volatility skew. Consequently, inferring risk premium
using ATM options should contain most-to-all information on the volatility risk premium,
and negligible-to-no information on jump risk premium. To avoid the problems noted in
Broadie, Chernov and Johannes (2007), we thus estimate the volatility risk premium by
restricting our attention to implied volatility from close-to-ATM options.
This evidence lends strong numerical justification for the negativity of the volatility risk
premium and its necessity in energy markets. These same conclusions mirror those found
12 Results for longer maturities show similar behavior but with more muted skews. This
is confirmed by running a fixed-effects regression over all simulated volatility differences for
the stochastic volatility and stochastic volatility with jumps models on the market premiums
while controlling for the strike price and other instantaneous parameters. Regardless of
model specification, has significant and negative impact on volatility difference. Jump
size risk premium is significant for OTM call options only. For longer maturities, the sign
and significance of is unchanged while the jump risk parameter became insignificant.
This result is intuitive and consistent with the Das and Sundaram (1999) finding of greater
short-term impact of jumps, and the long-term effects of stochastic volatility in explaining
implied return distributions.

10
for equity markets. Given the findings of Doran and Ronn (2006) of a large positive bias in
Black implied volatility in natural-gas and crude-oil markets, but mixed results for heating
oil, our expectation is of a large negative market price of risk for the natural gas and crude
oil, but smaller or negligible volatility risk premium for heating oil.

4 Estimation
The numerical evidence presented suggested the volatility risk premium should be negative
and significant. To test empirically for the existence of this risk premium in energy markets,
we make use of the information in implied and realized volatilities. By using both volatility
measures, we can effectively estimate the volatility risk premium with a two-step proce-
dure. First, we estimate the instantaneous risk-neutral parameters that govern the volatility
process by using the relationship between instantaneous variance and average variance. Sec-
ond, we infer the risk premium by minimizing the difference between actual realized volatility
and an estimated realized volatility using the risk-neutral parameters found in step one. For
robustness, we incorporate the price process to check for the stability of the volatility risk
premium estimate, and in addition, attempt to estimate the price risk premium.

4.1 Data
To solve for the volatility risk premium for energy commodities, ten years of natural gas,
crude oil and heating oil futures and option contracts are used. For each commodity, daily
observations are collected for both the futures and options prices for all contracts starting
with a contract that expired January 1995 and ending with the contract that expired De-
cember 2005. The data spans prices from January 1, 1994 through April 30, 2004. On
any given day, crude oil monthly contracts out to 5 years are available for both futures and
options. For natural gas there are 72 consecutive monthly contracts traded while for heating
oil there 18 consecutive monthly contracts. Since most of the volume and open interest is
on the near-term contracts, we focus on only prices from contracts with a year or less until
expiration.13
Since the primary goal is on the estimation of the volatility risk premium, we retain
information on only close to ATM options, since the prior simulation suggested the effect
of the jump premium for ATM options is negligible. Both options and futures data come
from Bloomberg and NYMEX. The energy contracts term-implied volatility comes from a
13 One
futures contract is worth 1,000 U.S barrels, 10,000 mmBtus, 1,000 U.S barrels for
crude oil, natural gas, and heating oil respectively.

11
weighted average of binomial approximations of actual contracts, explicitly accounting for the
early-exercise feature of the American options.14 This weighted average of implied volatilities
across different strike prices results in limiting the potential model-misspecification measure-
ment error in the implied volatility estimate. These implied volatility estimates are term
implied-volatilities, incorporating a term-structure of volatility (TSOV) which must be ac-
counted for in estimation. As in Doran and Ronn (2006), a reciprocal specification is used
to capture the typical ramping up of volatility as each contract approaches maturity.
Within the sample there are 23,408, 22,483 and 22,406 contract-days for the natural gas,
heating oil and crude oil futures and option contracts, respectively. For example, on February
10, 2003, there were twelve traded futures contracts (March 2003 February 2004) for each
commodity, as well as options on many of those contracts, resulting in twelve observations.
Table 2 provides the descriptive statistics for both implied and realized term volatility for
each energy commodity, and by individual monthly contracts. Consistent with findings of
Jackwerth and Rubinstein (1996), the sample clearly indicates implied volatility has been
higher than realized volatility for each commodity. This is especially true in the natural gas
market, and in particular the October-through-March winter-withdrawal contracts, where
this volatility difference averaged 12.56%. The infusion months have a difference volatility
difference of 8.30%. By comparison there is small seasonal variation over the same months
for heating oil, 4.73% versus 3.60%, and almost none for crude oil, 4.38% versus 4.18%.
By comparison, the average volatility difference for the S&P 100 and S&P 500 over the
same period is 5.18% and 4.66%, respectively. While these differences are similar to those
for crude and heating oil, Figure 2 demonstrates how the level of implied volatility for each
near-term energy commodity is consistently higher than the VIX index over the sample
period. For natural gas, the level of implied volatility has exceeded 100% multiple times,
and is almost double the level of its equity counterpart. Since the volatility risk premium
is proportional to the level of volatility, and given the volatility differences between implied
and realized volatility, we expect volatility risk premiums in energy to be negative and highly
significant.

4.2 Solving for the Risk-Neutral Parameters

To infer the volatility risk premium we adopt a similar, albeit somewhat modified approach
established in Jones (2003).15 We implement a two-step procedure, which estimates the
14 The approximation does not account for jumps and stochastic volatility, and only con-
siders the potential early exercise of the option.
15 We are indebted to an anonymous referee for detailed approach regardging this estimation

approach.

12
risk-neutral parameters first, and then uses these estimates in conjunction with realized
volatility to solve for the volatility risk premium. However, estimating the risk-neutral
volatility parameters of equation (2) requires knowledge of the latent volatility factor .
Fortunately, as shown in Ait Sahalia (1996), average variance can be expressed as a function
of the instantaneous variance such that,

Vt = A (t, T ) + B (t, T ) t2 (12)

where t2 is the instantaneous variance, Vt is the average expected variance, and the expec-
tation is taken under the risk-neutral measure. Moreover,
1 h i
B (t, T ) = 1 e(T t)
(T t)
A (t, T ) = [1 B (t, T )]

Applying Itos Lemma to (12) and substituting from (2) results in

! !
B B
dVt = B dt + t2 dt + Bt dz, t (13)
t t

B
Substituting for t2 from (12) and implies
t
s
Vt A
dVt = t ( Vt ) dt + B dz, t (14)
B
where
1
t = (15)
1 e(T t) T t
To estimate (14), henceforth the SV model, we employ non-linear maximum likelihood

estimation, using dz,t s Normally distribution. The maximization of the likelihood function
of SV, with the respect to , and , requires the additional constraint that A > Vt at all
times. Similar to Jones (2003), we set Vt equal to implied variance. A caveat in making
a direct substitution for expected average variance using implied variance is that implied
volatility can be measured with error. It is advisable to add an error term for implied
volatility; we forgo this extra step since our implied volatility is a weighted average of implied
volatilities. As Hentschel (2003) shows, using a weighted average for volatility across multiple
options, similar to the methodology for the VIX index, has little to no measurement error
for ATM options.
Prior to the estimation of (14), we modify the specification to check for the robustness of
the parameter estimates. First, we incorporate a control for the term-structure of volatility

13
(TSOV). To capture the increase in volatility as a contract approaches expiration we allow
to vary through time:

T t = + (16)
(T t)2
where is the long-run variance and is the time-dependent component. As Doran and
Ronn (2006) show, the reciprocal specification is merely an extension of the one-parameter
Schwartz (1997) to a two-factor model, with significant improvement for in-sample fit. Sub-
stituting eq. (16) into the SV model results in
" # s
Vt A
dVt = t + 2 Vt dt + B dz, t (17)
(T t) B

Eq. (17), henceforth the SVT model, is designed to capture the significant empirical
implications of an increasing TSOV that asymptotes to a positive, non-zero constant. This
procedure requires the estimation of one additional parameter, , in addition to , and .
The second addendum allows for the effect of the correlation between the price and
volatility innovations. This allows us to incorporate information contained in the futures
prices as well as implied volatility. To capture the effect of the correlation between price and
volatility, eq. (12) is substituted into (1), such that

s
Vt A
dFt = Ft dzst (18)
B
" # s
Vt A q  

dVt = t + Vt dt + B dz st + 1 2 dz, t (19)
(T t)2 B

where is the correlation between price and volatility innovations. We have also removed
the effects of jumps from eqs. (18) (19), henceforth the SVT-P model, since we are not
concerned with cross-sectional prices and focus on ATM implied volatilities. Additionally,
excluding jumps provides the benefit of estimating the SVT parameters plus with a similar
maximum likelihood approach since the joint distribution of changes in price and volatility
is bivariate Normal.
The results of each estimation for the three commodities is presented in Table 3. Addi-
tionally reported are the values for the likelihood function for each three models, and the
Akaike information criterion (AIC) and Schwartz information criterion (SIC) which assess
model performance. For each estimation for each commodity, the parameters are statisti-
cally significant at the 1% level except for the correlation parameter, which is insignificant for
crude and heating oil. For each commodity there is significant model improvement with the

14
inclusion of the TSOV variable, and small but insignificant improvement when incorporating
correlation. This suggests that the inclusion of TSOV is a critical component in explaining
the volatility of these energy commodities.
There are significant differences in the parameter estimates across commodities. The
level to which volatility reverts to is higher in natural gas contracts, = .335, than the
other commodities, while also having a more economically significant TSOV parameter,
= 1.96E 3 . This is not surprising given the higher levels of volatility present in Table 2
and Figure 2. The speed of mean-reversion is significantly higher for crude oil, = 27.74,
than the other two commodities, as well as the volatility in the variance process, = 0.449.
A high value for and appear consistent for the time-series behavior for crude oil, also
shown in Figure 2.
The relatively low values for are not surprising since there is a strong TSOV presence
for each commodity, which is especially true for natural gas. Noted earlier, excluding the
TSOV parameter will result in a downward-bias estimated of . With the inclusion of the
TSOV factor, has increased by 45% for natural gas, or the equivalent of a half-life increase
of 150 trading days.16 By comparison, the crude oil has a half-life increase of three trading
days with the inclusion of the TSOV parameter.
These results highlight the significant differences between these three energy commodities.
While it is not clear what potential impact these parameters have for volatility risk premium,
if the volatility risk premium is proportional to the level of volatility as is the case in equity
markets, then natural gas should exhibit the highest market price of volatility risk for the
three energy commodities.

4.3 Estimating the Volatility Risk Premium

To estimate the volatility risk premium, we incorporate the risk-neutral parameter estimates
found from the prior section and the realized volatility calculated from equation (10). The
real-world analog to equation (12) is,

VtP = AP (t, T ) + B P (t, T ) t2 (20)

where VtP is the average expected variance under the real-world (physical) measure, and AP
and B P are, respectively,
1 h
(T t)
i
BP = 1 e
(T t)
 
AP = 1 B P
16 The half-life is calculated as ln 2/.

15
with

=

=

where is the market price of volatility risk. Substituting (12) into (20) results in

BP
VtP = AP + (Vt A) (21)
B
To solve for , we minimize the difference between (21) and the observed realized vari-
ances from the data in a style similar to that of Bates (2000) and Bakshi, Cao and Chen
(1997): v
u
u T
X 2
RMSE = t min (brv, t VtP ) (22)
u
{ } t=1
Since , , , and are known from the prior section, and Vt is given as the ob-
served implied volatility, the non-linear estimation is straightforward. We use the parameter
estimates from the SVT-P model, even though is not incorporated in the estimation of
eq. (22). The estimation alternately uses daily observations over the entire sample as well
as monthly sub-samples. The parameter estimates for along with significance levels are
reported in Table 4.
The results for for each commodity is negative and highly significant, consistent with
the simulation evidence presented earlier. Of the three commodities, natural gas has the
highest volatility risk premium of 1.435. Additionally reported in Table 4 are the monthly
estimates for each commodity. In each case, separate monthly risk-neutral parameters are
found using the SVT-P model prior to solving eq. (22).17 The results show that the volatility
risk premium is negative and significant for all months and for all three commodities.
Natural Gas and heating oil exhibits seasonal patterns, where withdrawal/winter months
tend to have higher volatility risk premiums. This is especially true for natural gas, which
has a premium of 2.281 in January as compared to 0.851 in April. This appears highly
related to whether a month is a withdrawal or infusion month. The combined volatility risk
premium for the withdrawal and infusion months is 1.73 and 1.00 respectively, and this
difference is significant at the 1% level. Intuitively, this makes sense given the extra demand
for natural gas in the winter months, especially in the North-east states. The seasonality
17 Additionally, the monthly volatility risk premiums are calculated assuming that each
month had the same risk-neutral parameters as given in Table 3. Since the risk-neutral
parameters are quite stable across months, the results for the volatility risk premiums are
not significantly different from the results presented.

16
that is present for heating oil is reasonable as it is considered a substitute for natural gas.
By comparison there is little obvious pattern for crude oil. Since crude oil tends to have
constant demand regardless of season, there is little reason to suggest that there should be
a seasonal pattern.

4.4 GMM Estimation

These results are limited in that we have placed strong restrictions on the data generating
process by assuming that the correlation between the volatility and price process, , and
the market price of risk, f , are both equal to zero. To test the validity of the findings, we
adopt an approach similar to that of Broadie, Chernov and Johannes (2007), incorporating
estimation of the price risk premium and the correlation of the processes with the volatility
risk premium, to test for the stability of the parameter estimates. We re-estimate the
correlation to allow for any additional effect incorporating f has on . After estimation, we
set each risk parameter equal to zero and check the overall model fit to determine if there
are significant differences between the restricted and unrestricted model. Our expectation is
that the market price of volatility risk will have a significant impact on model fit, but the
price risk premium will not. This is due to the relatively short time series, and the possibility
that the price risk premium is close to zero. While we are inherently ignoring the effects of
jumps and potential jump premium, we feel that this has limited impact on the results since
we are using only ATM implied volatilities.18
Estimating both price and volatility risk premium requires incorporating the real-world
price process, given in equation (3) assuming = 0, with both the risk-neutral and real-world
volatility processes. This is best estimated using Hansen (1982) GMM procedure, where the
parameter vector, , incorporates the risk-neutral parameters given from the SVT model
and f , , and defined in equations (1) (4). In particular, we define,
! s
Ft Vt A
R1
t = ln f
Ft1 B
s
Vt A
R2
t = t2
B
" # !

Vt 1 2 2
= iv, t iv, t1 t + Vt dt
(T t)2
s
Vt A 2
 2 
Vt 2 = E Vt 1
B
18 Incorporatingall options will allow for a more comprehensive test for the jump parame-
ters given in equation (1). However, this goes beyond the scope of this paper.

17
 BP
" #
RV
t
1 2
= iv, t A + P
(Vt A)
B

where R1 R2
t and t are the error-terms with respect to the difference between the actual and
expected first two moments of return. Vt 1 and Vt 2 are the error-terms with respect to the
difference between the actual and expected first two moments of implied variance, and RV t
1

is the error term with respect to actual and estimated variance. ft () is defined as:

R1
t


R2
t


V1

ft ()
t
(23)
Vt 2



RV
t
1

Solving for f , , and requires setting the E [ft ()] = 0. This is done by minimizing
the expression
Jt () = gT ()0 WT () gT () , (24)
where gT () = (1/T ) Tt=1 ft () and WTh () is the ioptimal positive-definite symmetric
P

weighting matrix equal to the inverse of E ft ()0 ft () as defined by Hansen (1982).

Equation (24) is minimized for all three commodities with respect to f , , and . The
results are shown in Table 5. Comparing the market price of volatility risk to that of the prior
estimation shows no difference in the sign and significance of the parameters, although the
natural gas estimate has fallen from 1.435 to 1.095. The strong statistical significance for
the volatility risk premium, even after accounting for the price risk premium and correlation,
supports the notion of negative volatility risk aversion. The values for the correlation suggest
positive correlation between price and volatility innovations for natural gas, while there is
no correlation for heating and crude oil. This is similar to the findings in Table 3, except
that the correlation for natural gas has increased. Positive correlation between price and
volatility supports the notion that energy consumers are concerned, and typically try to
hedge, natural gas price increases. The insignificant correlation for heating and crude oil
may be due to examining only ATM options rather than the entire cross-section of option
prices. The values for the price risk premium are not significant for each commodity.
To test for the consistency of the risk premiums across months, equation (24) is re-
estimated using monthly sub-samples. The volatility risk premium is significant and negative
for all the individual estimations on monthly contracts for all commodities, and consistent
with the prior results. Additionally, the seasonality is persistent in the withdrawal/winter
months for natural gas and heating oil, while crude oil demonstrates no seasonality. The
correlation coefficients are consistent across months. Although only four out of twelve months

18
are statistically significant for natural gas, all of the months show positive correlation. The
price risk premium is insignificant for each monthly contract for each commodity.
To check for the robustness of our findings, equation (24) is re-estimated on the full
and monthly sub-samples with the restriction that = 0. The results of minimization
are reported in Jt1 () in Table 5. An additional test is done by restricting that f =
0, and is reported in Jt2 (). Using the J-statistics of the normalized difference between
the unrestricted Jt () and the restricted models [Jt1 (), Jt2 ()] , where this test statistic is
distributed 2 , reveals that we cannot reject f = 0, but we can reject = 0, both for the
full and monthly sub-samples. This is the case for all three commodities using the entire
sample and confirms the notion of a negative volatility risk premium in energy.
The fact that we could not reject f = 0 has at least three possible explanations. First,
the time sample is too short to get conclusive statistical estimates. As noted by Pindyck
(1999), a long-time series is needed to fully reject the existence of a unit root given the mean-
reverting nature of prices. Second, we have intentionally ignored modeling or estimating
any jump process, causing possible measurement errors in the risk-premium or correlation
parameter estimates. Clearly, we could have incorporated all options on the futures contracts
to estimate the jump parameters in addition to volatility parameters, but as Brodie, Chernov
and Johannes (2006) discuss, the computation expense is staggering. Additionally, this would
have removed the focus of the paper away from the volatility risk premium and led more
to discussion on appropriate models and model fit for energy markets. This is best left for
future research. Finally, it possible that the price risk premium for energy markets might
be zero. While our evidence does not suggest any significance or direction for the price risk
premium, it is clear that further work needs to be done. As the energy markets grow in
liquidity and are open to a wider range of investors, the implications for understanding the
risk premiums become even more paramount.

5 Conclusions
As Rubinstein and Jackwerth (1996) pointed out, implied volatility is higher on average than
realized volatility. In the realm of energy markets, we explain the difference between Black-
Scholes (1973)/Black (1976) implied volatility and realized term volatility by modeling and
computing a negative market price of volatility risk.
We use Monte Carlo simulation to demonstrate numerically that a negative market price
of volatility risk is the key risk premium in explaining this disparity in both the equity and
commodity-energy markets. Across several models, the market price of volatility risk is the
critical parameter that generates the differences in risk-neutral (Black implied volatility)

19
and real-world volatility (realized-term volatility). A negative market price of volatility risk
translates to a BIV greater than realized volatility. Our results for energy markets mirror
those found for equities by Coval and Shumway (2001) and Bakshi and Kapadia (2003). This
has implications for the market price of risk for energy, which is more difficult to estimate
(due to limited and more volatile data) than that of equities.
By controlling such variables as correlation between the processes, underlying volatility,
and jump size, inferences were made on the impact each parameter had on the difference
between implied and realized volatility. Additionally, it was determined that was the key
parameter, even in the presence of a jump process, that determined this difference. This
finding suggests that jump models, in energy or equities, fail to account for a negative market
price of volatility risk are incomplete.
The resulting empirical tests for revealed that the factor is negative and significant
for all three energy commodities. In addition, there appears to be strong seasonality in
volatility risk for natural gas and heating oil, while crude oil volatility risk is consistent across
months. Arriving at these estimates required a two-step procedure, where the risk-neutral
parameters were first estimated using implied volatility, and then calibrated in conjunction
with the volatility risk premium to estimate the realized volatility. The results were robust
to alternative specifications. The finding that is negative explains why option traders
tend to be short options: They write options at the higher implied volatilities and reap the
risk premium by delta-hedging their price exposures.

20
 References
1. At-Sahalia, Y., 1996. Nonparametric pricing of interest rate derivative securities.
Econometrica 64, 527 560.

2. Bakshi, G., Cao, C., Chen, Z., 1997. Empirical Performance of Alternative Option
Pricing Models. Journal of Finance 52, 2003-2049.

3. Bakshi, G., and N. Kapadia, 2003. Delta-hedged Gains and the Negative Market
Volatility Risk Premium. Review of Financial Studies 16, 527-566.

4. Bates, D., 1996. Jumps and Stochastic Volatility: Exchange Rate Processes Implicit
in Deutsche Mark Options. Review of Financial Studies 9, 69-107.

5. Bates, D., 2000. Post 87 crash fears in S&P 500 futures options. Journal of Econo-
metrics 94, 181-238.

6. Black, F., 1976. Studies of stock price volatility changes. Proceedings of the 1976
Meetings of the Business and Economics Statistics Section, American Statistical Asso-
ciation, pp. 177-181.

7. Black, F., and M. Scholes, 1973. The pricing of options and corporate liabilities.
Journal of Political Economy 81, 637-654.

8. Broadie, M., Chernov, M., and M. Johannes, 2007, Model Specification and Risk
Premia: Evidence from Futures Options, Journal of Finance 62, 1453-1490.

9. Carr, P. and W. Liuren, 2004, Variance Risk Premia, Working Paper.

10. Casassus, J., and P. Collin-Dufresne, 2005, Convenience Yields Implied from Interest
Rates and Commodity Futures, Journal of Finance 60, 2283-2331.

11. Chernov, M. and E. Ghysels, 2000. A study towards a unified approach to the joint
estimation of objective and risk neutral measures for the purpose of options valuation.
Journal of Financial Economics 56, 407-458.

12. Christensen, B and N. Prabhala, 1998, The relationship between implied and realized
volatility, Journal of Financial Economics 50, 125-150.

13. Christoffersen, P., and K. Jacobs, 2004, The Importance of the Loss Function in
Option Pricing, Journal of Financial Economics 72, 291-318.

21
14. Christoffersen, P., Jacobs K., and K. Mimouni, 2006, An Empirical Comparison of
Affine and Non-Affine Models for Equity Index Options, Manuscript, McGill Univer-
sity.

15. Coval, J., and T. Shumway, 2001, Expected Option Returns, Journal of Finance 56,
3, 983-1009.

16. Das, S., and R. Sundaram, 1999, Of Smiles and Smirks: A Term-Structure Perspec-
tive, Journal of Financial and Quantitative Analysis 34, pp. 211-239.

17. Dennis, P., Mayhew, S., and C. Stivers. 2006, Stock Returns, Implied Volatility
Innovations, and the Asymmetric Volatility Phenomenon. Journal of Financial and
Quantitative Analysis 41, 381-406.

18. Dincerler, C., and E. Ronn, 2001. Risk Premia and Price Dynamics in Electric Power
Markets. University of Texas working paper.

19. Doran, J., 2007, The The Influence of Tracking Error on Volatility Premium Estima-
tion. Journal of Risk Vol 9, No.3 1-36.

20. Doran, J., and E. Ronn, 2006. The Bias in Black-Scholes/Black Implied Volatility: An
Analysis of Equity and Energy Markets, Review of Derivatives Research 8, 177-198.

21. Doran, J., Peterson, D., and B. Tarrant, 2007. Is there Information in the Volatility
Skew? Journal of Futures Markets forthcoming.

22. Duffie, D., Pan, J., and K. Singleton, 2000, Transform Analysis and Asset Pricing for
Affine Jump-Diffusions, Econometrica 68, 1343-1376.

23. Eraker, B.m 2004. Do Stock Prices and Volatility Jump? Reconciling Evidence from
Spot and Option Prices, Journal of Finance, 59, p. 1367-1403.

24. Gallant, A. R. and G. Tauchen, 1998. Reprojecting partially observed systems with
application to interest rate diffusions. Journal of American Statistical Association
93,10-24.

25. Heston, S., 1993. A closed-form solution of options with stochastic volatility with
applications to bond and currency options. Review of Financial Studies 6, 327-343.

26. Hentschel, L., 2003. Errors in Implied Volatility Estimation. Journal of Financial
and Quantitative Analysis, 38, 779-810.

22
27. Hilliard, J., and J. Reis, 1998, Valuation of Commodity Futures and Options un-
der Stochastic Convenience Yields, Interest Rates, and Jump Diffusions in the Spot,
Journal of Financial and Quantitative Analysis 33, 61-86.

28. Jackwerth, J. C. and M. Rubinstein, 1996. Recovering probability distributions from

option prices. Journal of Finance 51, 1611-1631.

29. Jones, C., 2003. The Dynamics of Stochastic Volatility: Evidence from the underlying
and options markets. Journal of Econometrics 116, 181-224.

30. Pan, J., 2002, The Jump-Risk Premia Implicit in Options: Evidence from an Inte-
grated Time-Series Study, Journal of Financial Economics, 63(1), 3-50.

31. Pindyck, R., 1999,The Long-Run Evolution of Energy Prices, The Energy Journal,
April.

32. Routledge, B., D. Seppi, and C. Spatt, 2000, Equilibrium Forward Curves for Com-
modities, Journal of Finance 55, 1297-1338.

33. Schwartz, E., 1997. The Stochastic Behavior of Commodity Prices: Implications for
Valuation and Hedging. Journal of Finance 52, 923-973.

34. Schwartz, E., and J. Smith, 2000, Short-Term Variations and Long-Term Dynamics
in Commodity Prices, Management Science 46, 893-911.

23
 Table 1: Simulated Volatility Difference

This table highlights the simulated percentage differences between Black im-
plied volatility (BIV) and realized volatility using the SVJ (stochastic volatil-
ity with jumps) and the SVJ0 (stochastic volatility with jump but no market
price of volatility risk) models. In each case the simulation is conducted over
30 days (22-trading days) using the given futures contract and a correspond-
ing ATM option. The base case for the model uses parameter values equal to:
= 7, = .09, = .3, = .3, = .01, = 1 and 2 = .025. The market
prices of risk are s = .1, = 2, = .5 and = .01. For the base
case SVJ0 model, = 0. For the low and high cases, the parameters are:
= (1, 100) , = (.01, .36) , = (.1, .7) , = (.8, .8) , = (.1, .1) , =
(0.1, 3) and 2 = (.01, .1). The simulation is conducted over 30,000 runs.

SVJ SVJ0
Base Case 2.75% 0.20%

SVJ SVJ0
Low High Low High
2.45% 3.12% 0.35% 0.47%
2.39% 2.81% 0.26% 0.25%
1.56% 4.71% 0.03% 0.62%
2.62% 2.28% 0.59% 0.07%
2.45% 3.08% 0.03% 0.34%
2.68% 2.83% 0.56% 0.37%
2 2.54% 2.39% 0.10% 0.13%

24
 Table 2: Descriptive Statistics

This table reports the mean, standard deviation (given in parenthesis), and number of ob-
servations for implied and realized volatility levels for natural gas, crude oil, and heating oil.
The observations are annualized daily volatility from monthly contracts with yearly maturi-
ties between 1995 through 2005 over the period from January 1994 through April 2004. The
implied volatility is representative of the implied volatility from an at-the-money option. It
is calculated using a weighted-average of binomial approximations to account for the early
exercise feature of the option. The realized volatility is calculated from daily returns on the
futures contracts as given in equation (10).

Natural Gas Heating Oil Crude Oil

iv rv OBS iv rv OBS iv rv OBS
All Months 41.15 30.67 23408 29.66 25.37 22483 28.30 24.01 22406
(14.5) (16.8) (8.1) (9.1) (8.9) (9.5)
Jan 46.06 31.44 2131 30.59 25.69 1981 28.82 23.94 1975
(18.0) (18.6) (7.9) (9.3) (8.8) (9.8)
Feb 48.47 34.25 2091 31.78 25.46 2085 28.60 24.33 1966
(19.2) (23.0) (9.4) (9.9) (9.6) (10.2)
Mar 48.10 35.49 1887 30.64 24.47 2367 28.73 23.89 2095
(17.2) (21.8) (8.8) (9.2) (9.2) (9.8)
Apr 40.32 31.17 1957 31.27 25.55 2035 29.02 24.29 1840
(13.1) (17.9) (9.1) (10.6) (9.8) (9.8)
May 36.91 28.76 1924 30.18 26.23 1929 29.89 25.51 1819
(10.4) (14.8) (8.6) (10.5) (9.8) (10.8)
Jun 35.84 27.76 1984 27.69 24.77 1872 28.36 23.63 1885
(9.9) (13.4) (8.6) (9.6) (9.1) (10.4)
Jul 36.18 28.25 1875 28.38 25.73 1652 29.41 25.38 1730
(10.0) (14.1) (7.5) (8.7) (8.1) (10.0)
Aug 37.45 29.18 1806 29.49 26.36 1534 28.83 25.51 1625
(10.8) (13.2) (6.4) (7.6) (8.1) (9.3)
Sep 38.88 30.62 1929 28.49 25.24 1784 28.42 24.56 1887
(11.5) (14.3) (6.6) (7.4) (7.3) (8.1)
Oct 40.18 31.16 1931 28.63 25.91 1787 27.36 23.60 1847
(13.0) (15.5) (7.5) (9.2) (7.5) (7.4)
Nov 41.77 30.39 1901 29.39 25.39 1854 29.23 24.42 1858
(13.8) (14.9) (7.2) (8.3) (10.2) (8.7)
Dec 42.73 29.22 1992 28.15 23.83 1603 28.11 24.38 1879
(14.5) (14.6) (7.2) (7.1) (6.8) (6.7)

25
 Table 3: Estimation of Instantaneous Risk-Neutral Parameter Estimates

This table reports the maximum likelihood estimates of the risk-neutral volatility parameters estimates for natural gas, crude oil, and
heating oil. is the speed of mean-reversion, () is the level to which variance (long-run variance) reverts to, is the variance of
the variance process, represents the term-structure of volatility (TSOV) parameter, and is the correlation between the price and
the variance process. The daily for prices and implied volatility observations come from monthly contracts with yearly maturities
between 1995 through 2005 over the period from January 1994 through April 2004. Each commodity is estimated over three models:
(i) SV model expressed in eq. (14), (ii) SVT model expressed in equation (17), and (iii) SVT-P expressed in eqs. (18) (19), where
the SVT-P model incorporates both price and volatility estimates. LL is the value for the likelihood function, AIC is the Akaike
information criterion, and SIC is the Schwarz information criterion. Absolute values of t-statistics are in parenthesis.

Natural Gas Heating Oil Crude Oil

SV SVT SVT-P SV SVT SVT-P SV SVT SVT-P
1.638 2.377 2.372 3.157 3.480 3.476 24.694 27.739 27.636
(3.33)** (4.69)** (4.66)** (8.45)** (9.03)** (9.82)** (31.39)** (34.87)** (34.46)**

26
() 0.489 0.335 0.331 0.129 0.116 0.118 0.094 0.082 0.077
(4.28)** (6.25)** (6.84)** (14.99)** (14.41)** (14.35)** (44.45)** (42.22)** (42.85)**
0.138 0.146 0.130 0.096 0.102 0.103 0.402 0.449 0.443
(14.27)** (13.27)** (13.14)** (22.86)** (25.90)** (25.45)** (9.96)** (13.379)** (13.86)**
1.90E-03 1.96E-03 7.80E-04 9.04E-04 6.70E-04 6.45E-04
(3.63)** (3.36)** (2.74)** (2.61)** (17.00)** (17.96)**
0.110 -0.091 0.023
(2.10)* (1.98)* (1.12)

OBS 23,408 23,408 23,408 22,483 22,483 22,483 22,406 22,406 22,406
LL -17819.32 -2813.72 -2788.40 -637.31 -273.03 -265.39 -2687.23 -2376.31 -2370.21
AIC -5.40E+03 -4.72E+04 -4.74E+04 -7.49E+04 -9.30E+04 -9.36E+04 -4.74E+04 -5.02E+04 -5.02E+04
SIC -0.24 -2.08 -2.09 -3.51 -4.36 -4.38 -2.12 -2.24 -2.24

Robust t statistics in parentheses

* significant at 5%; ** significant at 1%
 Table 4: Volatility Risk Premium Estimates

This table reports the estimates for the volatility risk premium ( ) for natural gas, crude oil, and heating oil. The results are
generated by minimizing the difference between the actual realized variance, calculated in equation (10), and the estimated
variance, as given in equation (21). The daily volatility observations come from monthly contracts with yearly maturities
between 1995 through 2005 over the period from January 1994 through April 2004. The estimated volatility uses the risk-
neutral parameters found in the SVT-P model. RMSE is the value of the minimization of equation (22) which is done on
the whole sample and individual monthly contracts. Absolute values of t-statistics are in parenthesis.
Natural Gas All Months Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
-1.435 -2.281 -1.473 -1.260 -0.851 -0.952 -1.019 -1.079 -1.126 -1.007 -1.096 -1.787 -2.493
(5.38)** (6.75)** (3.23)* (4.83)** (2.49)* (2.48)* (2.34)* (2.39)* (3.08)* (3.33)** (4.82)** (4.87)** (4.99)**
OBS 23408 2131 2091 1887 1957 1924 1984 1875 1806 1929 1931 1901 1992
RMSE 213.78 18.19 59.89 30.67 37.01 15.07 9.23 7.78 6.79 8.84 9.69 6.65 5.14

Heating Oil
-0.385 -0.480 -0.596 -0.591 -0.532 -0.336 -0.261 -0.244 -0.229 -0.210 -0.203 -0.329 -0.405
(11.92)** (6.24)** (9.24)** (10.20)** (7.10)** (4.34)** (2.83)* (3.72)** (8.89)** (6.44)** (4.06)** (6.75)** (5.82)**
Observations 22483 1981 2085 2367 2035 1929 1872 1652 1534 1784 1787 1854 1603

27
RMSE 31.28 2.78 2.93 2.67 2.66 3.35 2.56 1.72 1.70 1.65 3.49 3.37 1.45

Crude Oil
-0.910 -1.036 -0.736 -0.949 -1.024 -0.826 -0.934 -0.753 -0.714 -0.989 -1.050 -1.039 -1.334
(6.10)** (4.35)** (3.43)** (3.57)** (3.26)** (4.40)** (4.29)** (4.11)** (5.44)** (6.39)** (4.97)** (1.820) (6.46)**
Observations 22406 1975 1966 2095 1840 1819 1885 1730 1625 1887 1847 1858 1879
RMSE 38.39 6.76 4.07 3.63 3.23 4.67 4.09 3.75 2.59 2.57 1.76 8.02 1.07

Robust t statistics in parentheses

* significant at 5%; ** significant at 1%
 Table 5: Risk Premium Estimates

This table reports the GMM estimates for the price risk premium (f ), volatility risk premium ( ), and correlation () between the price and volatility process for natural
gas, crude oil, and heating oil. The results are generated by minimizing equation (24). The daily price and volatility observations come from monthly contracts with yearly
maturities between 1995 through 2005 over the period from January 1994 through April 2004. J( ) is the value of the minimization of equation (24). J 1 () is the value of
the minimization of equation (24) while imposing the restriction that = 0. J 2 () is the value of the minimization of equation (24) while imposing the restriction that
f = 0. UThe normalized difference between the unrestricted Jt () and the restricted models (Jt1 (), Jt2 ()) is distributed 2 . Absolute values of the t-statistics are in
parenthesis.

All Months Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Natural Gas
f 0.1600 0.1174 0.1217 0.2350 0.1476 0.1762 0.2210 0.2849 0.2088 0.0276 0.1978 0.0954 0.1143
(1.09) (0.86) (0.86) (1.53) (1.00) (1.22) (1.59) (1.9) (1.31) (0.18) (1.31) (0.64) (0.80)
-1.0950 -2.0783 -1.2733 -1.1756 -0.9992 -0.8913 -0.8084 -0.8719 -1.0943 -1.1875 -1.3674 -1.1855 -1.9795
(7.93)** (20.36)** (11.19)** (8.77)** (6.38)** (5.50)** (6.22)** (6.5)** (7.24)** (7.68)** (7.25)** (10.48)** (18.37)**
0.2533 0.2890 0.3334 0.2923 0.2746 0.2560 0.1814 0.2601 0.1838 0.1503 0.1679 0.2947 0.3463
(2.96)** (3.91)** (2.81)** (1.43) (2.35)* (1.79) (0.48) (1.52) (0.58) (1.49) (0.92) (2.84)** (1.82)

J() 84.61 5.19 5.83 5.72 9.91 10.97 12.19 9.44 6.45 6.23 5.23 3.82 3.64
J 1 () 1931.71 270.59 477.29 263.14 241.35 103.43 68.9 60.79 57.12 71.19 83.52 96.81 137.59
J 2 () 93.19 6.21 7.11 6.50 10.79 12.04 13.33 10.21 7.14 6.44 5.10 4.06 4.26

28
obs 23408 2131 2091 1887 1957 1924 1984 1875 1806 1929 1931 1901 1992
Heating Oil
f -0.1088 -0.2282 -0.1406 -0.0536 0.0343 0.0438 -0.0128 0.0070 -0.2297 -0.1457 -0.1958 -0.1978 -0.2538
(0.63) (1.37) (0.90) (0.36) (0.21) (0.25) (0.07) (0.04) (1.11) (0.79) (1.07) (1.13) (1.40)
-0.4749 -0.4859 -0.5982 -0.6011 -0.5270 -0.3562 -0.5317 -0.4687 -0.2866 -0.3832 -0.3387 -0.4089 -0.5617
(5.47)** (6.93)** (9.40)** (10.07)** (7.73)** (4.39)** (3.94)** (2.93)** (3.98)** (5.52)** (4.97)** (5.17)** (6.30)**
-0.0088 -0.0317 -0.0283 -0.0253 -0.0151 0.0561 -0.0053 -0.0005 -0.0066 0.0059 -0.0141 -0.0035 -0.0385
(0.96) (1.04) (0.98) (0.94) (0.50) (1.75) (0.16) (0.02) (0.17) (0.18) (0.42) (0.11) (1.13)

J() 19.65 1.64 1.70 2.54 1.83 1.41 1.68 1.55 1.04 1.53 1.51 1.49 1.73
J 1 () 700.92 68.19 86.82 85.25 76.49 70.02 48.62 33.34 33.53 32.22 63.42 67.46 35.56
J 2 () 27.91 2.05 2.21 2.92 2.38 1.99 2.37 2.22 1.93 2.06 2.58 2.83 2.37
obs 22483 1981 2085 2367 2035 1929 1872 1652 1534 1784 1787 1854 1603
Crude Oil
f 0.1997 0.2076 0.3042 0.2493 0.2854 0.1352 0.1736 0.1330 0.0795 0.1978 0.1031 0.3262 0.1637
(0.69) (0.73) (1.06) (0.92) (0.97) (0.45) (0.64) (0.44) (0.25) (0.68) (0.34) (1.08) (0.67)
-0.5470 -0.5728 -0.5128 -0.5823 -0.5514 -0.5064 -0.5713 -0.5801 -0.5290 -0.5534 -0.5677 -0.5667 -0.4718
(9.83)** (10.29)** (9.61)** (11.56)** (9.31)** (8.39)** (9.95)** (8.59)** (8.37)** (11.07)** (11.51)** (10.51)** (9.63)**
-0.0233 0.0164 0.0020 -0.0196 -0.0326 -0.0321 -0.0333 -0.0079 -0.0396 -0.0478 -0.0593 -0.0119 -0.0287
(1.40) (0.71) (0.08) (0.87) (1.36) (1.33) (1.41) (0.31) (1.55) (2.02)* (2.46)* (0.50) (1.22)

J() 52.87 3.92 3.75 4.54 3.56 3.43 4.01 4.08 3.98 4.33 4.20 7.00 6.07
J 1 () 834.97 130.56 78.97 82.37 77.22 93.01 85.12 72.01 50.97 59.23 45.28 29.22 31.01
J 2 () 66.37 4.89 5.05 5.74 4.68 4.59 5.13 4.89 4.36 5.37 5.28 8.96 7.46
obs 22406 1975 1966 2095 1840 1819 1885 1730 1625 1887 1847 1858 1879

Robust t statistics in parentheses

* significant at 5%; ** significant at 1%
significant at 1% for the 2 distribution
 Figure 1: Volatility Skews across Risk Premium
Figure 1 demonstrates the simulated volatility difference between implied and realized volatility across the cross-section of option
prices using the SVJ model. The four samples presented are: 1. No Premia (all risk premiums set equal to zero); 2. Volatility Risk
= 2 ( = 2 ; all other risk premiums equal to zero); 3. Jump Intensity Risk = .5 ( = .5 ; all other risk premiums equal to zero);
4. Jump Size Risk = .09 ( = .09 ; all other risk premiums equal to zero).

!" "
#$

29
 Figure 2: Implied Volatility in Energy Markets
Figure 2 demonstrates the implied volatility of the near-term contract for the three energy commodities and the VIX index from January
1994 through April 2004.The implied volatility for each energy commodity comes from the contract that is closest to maturity up
until 10-days to expiration. The second near-term monthly contract is then used instead. HO is the implied volatility from heating
oil contracts, CO is from crude oil contracts, and NG is from natural gas contracts. The period is from January 1994 through April
2004.

30