ENERGY PRICING

Energy Price Processes

Used for Derivatives Pricing &
Risk Management
In this first of three articles, we will describe the most commonly used process,
Geometric Brownian Motion, and in the second and third pieces, we will
introduce two processes that are gaining rapid acceptance for a wide
range of applications involving commodity derivatives: Mean Reversion and
Jump Diffusion. We will talk about the main uses of each of these processes
and some of the pitfalls that practitioners should take into account when
using these processes for pricing and managing the risk of various energy
derivatives structures. By Carlos Blanco, Sue Choi & David Soronow, Financial
Engineering Associates.
his is the first article in a three-part than other prices set in financial markets. Due

T series that will focus on the main

processes used to model commodity
spot and forward prices. Stochastic processes
to short term supply and demand imbalances,
prices for short-term delivery of the
commodity - or spot prices - tend to exhibit
form the basis of derivatives pricing and risk significantly different behaviour than prices
management models, as they allow us to for delivery of the commodity in the future, or
model possible price evolution through time, forward prices.
and assign probabilities to possible future
prices as a function of current spot and Spot Prices: There are several important
forward prices and a set of parameters that properties associated with the volatility of spot
describe the possible variability of those prices energy prices, principal among them being:
over time, (hopefully matching empirical
patterns). - Seasonal Effects: - In response to
cyclical fluctuations in supply and
Characteristics of Commodity Prices demand mostly due to weather and
Before starting with the description of the climate changes, energy prices tend to
mathematical models, it is important to keep exhibit strong seasonal patterns.
in mind the actual behaviour of commodity - Mean-Reversion: - Prices tend to
prices that we are trying to model. fluctuate around and drift over time to
Commodity prices are somewhat different values determined by the cost of
production and the level of demand.
Henry Hub NYMEX Front Month Weekly Prices - Occasional Price Spikes: - Large
January 1992 - February 2001
changes in price attributed to major
shocks (e.g. generation/transmission
10
outages, unanticipated political events …).

8 Forward Prices: The forward markets

provide the ‘best guess’ about the future spot
price for different maturities. As we can see in
$/MMBTU

6
the adjacent chart, there can be dramatic
differences between prices for delivery of
4 electricity in successive months.
Energy prices typically display seasonal
2 variations in volatility, occasional price
spikes, and a tendency to quickly revert to
the average cost of production. Stochastic
0
Processes used to model Commodity
Jan 92

Jan 93

Jan 94

Jan 95

Jan 96

Jan 97

Jan 98

Jan 99

Jan 00

Jan 01

prices should capture the specific

characteristics of the commodity.

74 Commodities Now, March 2001

 ENERGY PRICING

In addition to seasonality and mean California Oregon Border Electricity

reversion, forward prices are characterised by (Forward Price Curve for January 8th, 2001)
exhibiting a different behaviour depending on
160
the time to maturity. As contracts get closer to
their maturity date, the volatility usually 140
increases, (Samuelson’s Hypothesis). 120
In order to capture energy markets reality -
regardless of the price process being used - we 100

should aim to incorporate, as the model’s 80

essential building blocks, the information
60
contained in the forward price term
structures(i.e. expected prices for delivery at 40
different times), and the forward volatility term 20
structure (i.e. expected variability of prices at
0
different points in time). In the next two Feb 01 Mar 01 Apr 01 May 01 Jun 01 Jul 01 Aug 01 Sep 01 Oct 01 Nov 01
articles of this series, we will analyze the more Futures Expiration
sophisticated models, which also incorporate Source: NYMEX

into the analysis the mean reversion and

price spikes observed in many Main Stochastic Processes Used to Value Commodity Derivatives
commodity prices. Geometric Brownian Motion Lognormal price diffusion
Derivatives pricing and Risk Mean-Reversion Lognormal price diffusion with mean reversion
Management models are based on some
Mean Reverting Jump Diffusion Same assumptions as the mean-reversion processes
type of assumption about the price
and adds a separate distribution for price jumps.
evolution of the underlying commodity.
Each of the processes that we are going to Since then, Brownian Motion has been
present have a set of advantages and used in multiple fields, including in finance, to
disadvantages. The more simple ones may model the behaviour of security prices. Over
provide a simplistic view of reality, but allow time, in the finance literature, Brownian
us to characterise the multiple sources of risk motion came to be called ‘Random Walk’, in
in a very limited number of parameters, and reference of the path of a drunk after leaving
therefore are easier to interpret and calibrate the bar on his way home.
from market prices. The more complex
processes provide the ability to incorporate The main properties of random walk
more information about the possible price process are:
changes, but at the cost of having to estimate - Price changes are independent of
many more parameters and increasing the each other (no memory).
probability of model errors. - Price changes have a constant mean
and volatility.
Brownian Motion, Random Walks &
Black-Scholes Brownian Motion With & Without Drift
The most known price process is ‘Brownian (Forward Price Curve for January 8th, 2001)
Motion’, which takes the name from Scottish
55
botanist Robert Brown, who in 1827 found
that particles within water-suspended pollen
grains followed a particular random Drift Rate Determined by
movement that resembled a zigzag path. Even Commodity Yield
particles in pollen grains that had been stored 45
for a century moved in the same way.
Price ($)

If we are using the spot price only, we need

to add a drift to reflect the expected change No Drift
35
in price to converge to the observed forward
price. The commodity yield plays a large role
in determining that drift.

When we are simulating forward prices, it 25

is common to assume that the drift is zero,
Source: NYMEX
and use the current forward price as the
starting point.

Commodities Now, March 2001 75

 ENERGY PRICING

Drifting Brownian Motion model - which was based on the extension of

Brownian motion contributed to the explosive
Price Change Process = Drift effect (Non-random) + Volatility effect (Random)
growth in trading of derivatives.
For most commodities, it is possible to The original Black-Scholes model allowed
observe or build a forward price curve that only for pricing options on a non-dividend
provides the expected level of paying stock. Extensions to the Black-Scholes
For most commodities, it is prices for different delivery dates. model, such as the Garman-Kohlhagen and
possible to observe or build Within the forward price curve, Black(1976) model, allowed for pricing global
the relation between different commodity options, and options on futures
a forward price curve commodity prices is governed by respectively.
the commodity yield. In order to Geometric Brownian motion implies that
incorporate the term structure of commodity returns have a lognormal distribution, meaning
yields, it is necessary to add a drift component that the logarithmic returns, which are simply
to the Brownian motion. continuously compounded returns follow the

Price Evolution From ti to ti+1

S(ti+1) = S(ti) exp [r - q - 1 σ2 ] (ti+1 - ti) + σ ε i+1 S ε√(ti+1 - ti) √

2
↵
Price Price Drift Random
at ti+1 at ti Component Component

r = Risk free rate

S (ti+1) q = Commodity yield
σ = Annualised standard
deviation of returns
S (ti) (ti+1 - ti) = Time step in years
ε i+1 is the random shock to
price from t to t+1
ti ti+1 Time
exp = the base of the natural logarithm

Geometric Brownian Motion & Black- normal (bell shaped) distribution. Consistent
Scholes with reality, the lognormal distribution
In 1973, Fischer Black and Myron Scholes restricts prices from falling below zero (eg. the
published their seminal paper on options maximum negative return is 100%).
pricing. The Black-Scholes option pricing If S is the price, the change in price can be
approximated by:
Geometric Brownian Motion - Sample Price Paths
Forward Price = $30 - Volatility = 50% ∆ S = µ S∆t + σ S ε√∆t
70

60
Over a short period of time, the
logarithmic change in price is assumed to be
50 normally distributed with:

40 - Mean or Drift µ S∆t

Price

σ2
30
with µ=r-q- 1
2
20
- Standard Deviation σ S ε√∆t
10

0
We can see randomly
31/3/01

30/4/01

30/5/01

29/6/01

29/7/01

28/8/01

27/9/01

25/1/02

24/2/02

26/3/02
27/10/01

26/11/01

26/12/01

simulated price paths for daily

steps during a year.
Source: @Energy. Financial Engineering Associates

76 Commodities Now, March 2001

ENERGY PRICING

Price Histogram at Expiration - The lognormal distribution allows for

commodity prices between zero and
Geometric Brownian Motion
Forward Price = $30 - Volatility = 50%; 360 days
infinity and has an upward bias
(representing the fact that commodity
prices can only drop 100% but can
rise by more than 100%).

Under geometric Brownian Motion,

volatilities are proportional to the square root
of time. The process of converting volatilities
between different time horizons is known as
$4 $7 $10 $14 $17 $21 $24 $28 $31 $35 $38 $42 $45 $49 $52 $56 $59 $63 $67 the square-root-of-time rule. This rule allows
Source: @Energy. Financial Engineering Associates
us to annualise hourly, daily, weekly, monthly
or any other volatilities.
GBM Parameters: Volatility
In order to simulate possible future prices 4. Using the square-root-of-time rule, we
following GBM, we just need to know the annualise to obtain s* (cell B15) assuming
current asset price, and the expected variability that time is measured in trading days and there
of the asset (volatility). are 250 trading days per year. The formulas
There are several methods to estimate are:
‘expected’ volatility.
Some traders prefer to use estimates based B14: =SQRT(250)
on historical prices (‘historical volatilities’), B15: =B13*B14
while others prefer to use the volatilities
implied by option market prices (‘implied • Example 1: Suppose we have computed
volatilities’). the daily volatility as being 1.26%. Then the
annualized volatility would be equal to 19.98%
How to Calculate Historical Daily & ( 1.26% x √ 250 ).
Annual Volatility in Excel The square root of time rule is commonly
1. Choose a particular price series. Prices used in Risk Management to convert certain
could be hourly, daily, weekly, etc. Column A Value at Risk measures for different holding
(rows 2–11) contains historical prices for asset periods.
x over a ten-day period.
2. Calculate the logarithmic price changes. Introducing ‘Seasonality’ in the Simulation
The continuous returns are the natural logs of with the Forward Price & Volatility Curve
the price relatives. For example, cell B3’s Prices and volatilities have a strong
formula is =LN(A3/A2). Column B (rows seasonal component that should be taken into
3–11) shows the daily returns, denoted u. account at the time of describing the possible
A B evolution of prices through time.
1 X u We can replace the constant volatility
2 14.40 parameter by a time dependent one (see figure
3 14.20 -0.0140 opposite).
4 14.25 0.0035 In this figure, we can observe simulated
5 14.00 -0.0177 paths for a Electricity Forward Price Curve
6 13.75 -0.0180 with strong price seasonality, and volatilities
7 13.80 0.0036 of 50% during each summer month (July,
8 13.60 -0.0146 August, September) and 20% for all other
9 13.75 0.0110 months. We can observe the strong seasonal
10 13.70 -0.0036 component in price variations given to
11 13.90 0.0145 seasonal price differences.
12
13 STDEV(u) 0.0126 Pitfalls of Using Geometric Brownian
14 SQRT(250) 15.8114 Motion to Model Commodity Prices
15 s* 0.1998 1. Energy prices are not ‘exactly’ lognormally
distributed.
3. Use Excel’s built-in STDEV (sample Several empirical studies have shown that
standard deviation) function to calculate the energy prices experience significant variations
daily volatility (cell B13). from lognormality. In the next two articles, we

79 Commodities Now, March 2001

 ENERGY PRICING

FEA @Energy: Monte Carlo Simulation of Price Processes

Market Inputs Value
Forward Price Curve
Spot Prices 2/4/05 2/5/05 2/6/05 2/7/05 2/8/05 2/9/05 2/10/05 2/11/05
Interpolation 1 30 29 29 46 63 58 30 30
Futures Contract Volatility Curve
2/4/05 2/5/05 2/6/05 2/7/05 2/8/05 2/9/05 2/10/05 2/11/05
Simulation Inputs Value 20% 20% 20% 50% 50% 50% 20% 20%

Simulation sate date 2/4/05

Simulation end date 2/12/05 Monte Carlo Sample Price Paths - Lognormal Process
Sampling Frequency d 100
90
Model Inputs Value
Model Black Scholes 80
Mean-Revision rate 222
Cap on Price 70
Floor on Price 60
Price $

Yearly jump frequency 7

Expected jump size 1 50
St. Dev. of jump 0 40

Other Inputs 30
Value Date* 2/5/05 20
Interest rate 5.0096
Seed* 10
0
Histogram Inputs

1/10/01

1/11/01

1/12/01
1/4/01

1/5/01

1/6/01

1/7/01

1/8/01

1/9/01
Simul* 10000
Min-Max* 0

Source: @Energy. Financial Engineering Associates

will introduce Mean Reversion and Jump high volatilities can be very different than what
Diffusion processes that better characterise the most traders have in mind at the time of using
behaviour of certain commodities, especially that process. The technical explanation is that
electricity. when volatility is significantly large, the drift
component starts to dominate the price
2. Extreme Price changes may be evolution. For assets with very high volatilities
underestimated by the lognormal distribution and mean reversion (e.g. power), it is highly
A direct consequence of the previous recommended to use other processes that
pitfall is that Geometric Brownian motion better describe the evolution of the underlying,
does not capture extreme price changes such as mean reversion or jump diffusion.
accurately. This pitfall is particularly
important for Risk Management and the Conclusion
pricing of certain exotic options. The development of general diffusion
models contributed to the development of the
3. Volatilities are not known. options markets. Today, these models are still
The only unknown parameter in the the most commonly used by market
Geometric Brownian motion is the volatility practitioners, largely due to the relative
of the underlying. The ideal volatility to use simplicity of estimating input parameters.
for modelling purposes would be the ‘future However, as we will point out in the next
volatility’, but by definition, it is not possible articles, these models fail to capture many of
to know the ‘future volatility’ until we know the key characteristics of commodity prices.
what has happened in the market, and by that In particular, commodity prices tend to
time has become ‘historical volatility’. fluctuate around and drift over time to values
Therefore, the volatility curve used as an input determined by the cost of production, and
should be our "best estimate" of future often experience large changes in price due to
volatility, and reflect our expectations shock events. The mean reversion model and
regarding the variability of the asset price over jump diffusion model aim to modify the
the period of time under consideration. general diffusion price process in order to
capture these additional market realities. ■
4. Volatilities are not constant.
Any trader knows that volatilities change Carlos Blanco is Manager of Support &
through time, and the assumption of constant Educational Services at Financial
volatilities may not be very realistic. More Engineering Associates.
complex processes incorporate time varying Sue Choi and David Soronow are
volatility, and some Risk Management models derivatives specialists at FEA.The
assume that volatilities fluctuate just as asset authors would like to thank the FEA
prices do. Financial Engineering Team, especially
Angelo Barbieri, for helpful comments
5. Beware of the model results for very high and excellent software assistance.
volatilities (e.g. above 300%). Email: carlos@fea.com
Price paths generated with GBM with very www fea.com

Commodities Now, March 2001 80