Вы находитесь на странице: 1из 34

Real options approach to coal power plant

investment

Jurica Brajkovic

Dresden, April 16, 2010


Introduction Investment problem Application of real options Model Selection

Outline


Question: How to analyze investment in coal fired power plant in liberalized
market?

Structure of the presentation:


• Motivation with an example
• Short history of real options (RO)
• Description of investment problem
• Application of RO
Introduction Investment problem Application of real options Model Selection

Motivation
How to value large capital investments? Utilities mostly use NPV.
Introduction Investment problem Application of real options Model Selection

Motivation
How to value large capital investments? Utilities mostly use NPV.
This was good when:
• Markets monopolized (one vertically integrated company, no
competition)
• Certain electricity prices (regulation)
• Costs could be passed on to the final consumer
• State support
• In essence, no business risk implying certain cash flows
Introduction Investment problem Application of real options Model Selection

Motivation
How to value large capital investments? Utilities mostly use NPV.
This was good when:
• Markets monopolized (one vertically integrated company, no
competition)
• Certain electricity prices (regulation)
• Costs could be passed on to the final consumer
• State support
• In essence, no business risk implying certain cash flows
But electricity markets have changed and utilities face different
conditions:
• Competition in generation
• Generators face plethora of risks: uncertain revenues and costs,
demand risk, construction, financial risk...
Introduction Investment problem Application of real options Model Selection

Drawbacks of discounted cash flow approach

 
Question: What’s wrong with NPV?
 
Introduction Investment problem Application of real options Model Selection

Drawbacks of discounted cash flow approach

 
Question: What’s wrong with NPV?
 
It does not recognize the following:
• Mostly, investments are not ’now or never’ decisions
• Investments are irreversible
• Underlying variables follow stochastic processes (price, cost,
demand..)
• There are options in the project (option to shut down, suspend,
expand...)
• Managers act to eliminate downsize and explore upside
Good investment tool should address these issues.
Introduction Investment problem Application of real options Model Selection

How does NPV work?

Simple example: a plant costs e100 to build instantaneously. It


produces 1 unit of output per year forever at the cost of e40. During the
first year the revenue is e50 per unit. From second year on, expected
revenue is e50: 50% chance that it will be e70 and 50% chance that it
will be e30. Assume discount rate of 10 %. To invest or not to invest?
Introduction Investment problem Application of real options Model Selection

Example cont.

∞ ∞
X [E (pt ) − c] · qt X 50e − 40e
NPV = −I = − 100e = 10e
t=1
(1 + r )t t=1
(1.1)t

• NPV rule suggests to invest because NPV > 0


• But - opportunity cost of immediate action. Waiting for year 2 will
reveal the actual future price. Real options would suggest to wait.
• What is the value of the plant if investor waits for a year?
Introduction Investment problem Application of real options Model Selection

Example cont.

∞ ∞
X [E (pt ) − c] · qt X 50e − 40e
NPV = −I = − 100e = 10e
t=1
(1 + r )t t=1
(1.1)t

• NPV rule suggests to invest because NPV > 0


• But - opportunity cost of immediate action. Waiting for year 2 will
reveal the actual future price. Real options would suggest to wait.
• What is the value of the plant if investor waits for a year?
Price=e70

X 70 − 40
NVP = − 100 = e200
t=1
(1.1)t
Introduction Investment problem Application of real options Model Selection

Example cont.

∞ ∞
X [E (pt ) − c] · qt X 50e − 40e
NPV = −I = − 100e = 10e
t=1
(1 + r )t t=1
(1.1)t

• NPV rule suggests to invest because NPV > 0


• But - opportunity cost of immediate action. Waiting for year 2 will
reveal the actual future price. Real options would suggest to wait.
• What is the value of the plant if investor waits for a year?
Price=e70 Price=e 30
∞ ∞
X 70 − 40 X 30 − 40
NVP = − 100 = e200 NVP = − 100 = −e200
t=1
(1.1)t t=1
(1.1)t
• Obviously NPV is wrong and waiting gives superior results.
Introduction Investment problem Application of real options Model Selection

Financial vs real options


Value of the project consists of:
• Intrinsic value (NPV)
• Option to invest (opportunity costs)
• Options embedded in the project (managerial and operational
flexibility)
How are financial and real options related?
Real option Variable Financial option
Investment cost X Exercise price
Value of the project S Stock price
Amount of time decision can be deferred t Time to maturity
Riskiness of underlying variable σ2 Variance of the stock
Time value of money r Risk free rate of return

Table: Link between financial and real options

Optimal stopping problem: when to exercise option to invest and to


receive the project?
Introduction Investment problem Application of real options Model Selection

NPV vs Real Options


Introduction Investment problem Application of real options Model Selection

Plant characteristics
Assumptions:
• Analyze investment decision in a base load coal fired power plant
• Investor does not need to worry about emission permits
• Focus: what is the most appropriate stochastic process and how
does the choice affect investment decision
Assume following parameters:
Parameter Value
Project life 40 years
Capacity 500 MW
On line % 85%
Productive hours 7446
Annual production 3,723,000 MWh
Investment cost per kW 1,500 e
Discount rate 10%
Table: Investment parameters
Introduction Investment problem Application of real options Model Selection

Dark Spread
Dark spread=Price of electricity [e/MWh] - Price of coal [e/MWh]
Introduction Investment problem Application of real options Model Selection

NPV value

Figure: NPV calculation

Break even occurs at 20.6 e/MWh.


Introduction Investment problem Application of real options Model Selection

Stochastic processes

The following stochastic processes were considered:


• Arithmetic Brownian motion (ABM)
• Ornstein Uhlenbeck (OU)
• Cox Ingersoll Ross (CIR)
• Schwartz one factor model
I evaluate the processes using two criteria:
• 10 000 Monte Carlo simulations - distributional properties
• Root mean squared error (RMSE) for in and out of sample
Introduction Investment problem Application of real options Model Selection

Arithmetic Brownian motion (ABM)


• Originated in financial analysis
• Popular in RO literature due to analytical tractability

Change in dark spread is given by:

dp = at + σdz (1)

Table: Estimated weekly parameters for dark spread prices


Weekly values
Drift (a) -0.006
Volatility (σ) 10.16

Table: Summary statistics for observed and 10 000 simulated price trajectories
min. median mean max. se. skew. kurtosis
Simulations -807.65 17.29 16.61 811.51 141.38 -0.03 4.01
Observed prices 1.09 20.74 24.25 121.39 14.45 1.86 9.33
Introduction Investment problem Application of real options Model Selection

A few simulated trajectories:


Introduction Investment problem Application of real options Model Selection

ABM - Optimal investment decision


Value of the project (V) is given by:
Z T
V (p) = qpt e −rt dt (2)
0
p0 q(1 − e −rT ) aq(1 − e −rT
(1 + Tr ))
V (p) = + (3)
r r2
Option value given by:

E [df ] = rfdt (4)


1
Fpp σ 2 + aFp − rF = 0 (5)
2
f (p) = Ae β1 p (6)

Investment rule:

f (p) = V (p) − I value mathching (7)


fp = Vp smooth pasting (8)
Introduction Investment problem Application of real options Model Selection

Optimal exercise occurs at price of 186 e/MWh.

Figure: Optimal exercise for arithmetic Brownian motion


Introduction Investment problem Application of real options Model Selection

Ornstein Uhlenbeck - OU
The process is given by:

dpt = k[µ − pt ]dt + σdzt (9)

Table: Estimated weekly parameters for dark spread prices

Weekly values
Mean reversion (k) 0.28
Long run price (µ) 24.24
Volatility (σ) 10.9
Standard error (ζ) 9.52

Table: Summary statistics for observed and 10 000 simulated price trajectories
for OU process
min. median mean max. se. skew. kurtosis
Simulations -46.11 24.13 24.17 104.79 14.44 0.01 3.01
Observed prices 1.09 20.74 24.25 121.39 14.45 1.86 9.33
Introduction Investment problem Application of real options Model Selection

A simulated sample path:

Figure: A sample path of dark spread following OU process


Introduction Investment problem Application of real options Model Selection

Optimal investment decision

Value of the project is given by:


Z T
−rt
V (p) = qpt e dt (10)
0
p0 (1 − e −(k+r )T ) m(1 − e −rT ) m(e −(k+r )T
 
− 1)
V (p) = q + + (11)
k +r r k +r

Option value is given by:


1
Fpp σ 2 + fp k(m − p) + ft − rf = 0 (12)
2
Unfortunately no closed form solution - estimate numerically using
implicit finite difference scheme.
Introduction Investment problem Application of real options Model Selection

Optimal exercise occurs at 40.9e/MWh.

Figure: Exercise boundary for OU process


Introduction Investment problem Application of real options Model Selection

Cox Ingersoll Ross (CIR) model


CIR process is given by:

dp = k(m − p)dt + σ pdz (13)

When 2km > σ 2 variable always stays positive.


Difference between OU and CIR process:
• CIR admits only positive prices
• CIR allows for non constant variance - depends on price level
σ 2 −kt σ2
Var (pt ) = pt−1 (e − e −2kt ) + µ (1 − e −kt )2
k 2k
Table: Estimated weekly parameters for dark spread prices

Weekly values
Mean reversion (k) 0.29
Long run price (µ) 24.05
Volatility (σ) 1.94
Introduction Investment problem Application of real options Model Selection

min. median mean max. se. skew. kurtosis


Simulations 0.16 21.82 23.98 134.43 12.50 1.06 4.69
Observed prices 1.09 20.74 24.25 121.39 14.45 1.86 9.33

Table: Summary statistics for observed and simulated prices

Figure: A sample path of simulated CIR process


Introduction Investment problem Application of real options Model Selection

Optimal investment decision


Investment threshold is at 38.5 e/MWh.

Figure: Optimal exercise price


Introduction Investment problem Application of real options Model Selection

Schwartz one factor model

dp = k(a − lnp)pdt + σpdz (14)


First, a logarithm transform is performed giving (where x = lnp):

dxt = k(m − xt )dt + σdzt (15)

Table: Estimated parameters for log of dark spread

Weekly values
Mean reversion (k) 0.365
Long run price (a) 3.367
Standard error (ζ) 0.418
Volatility (σ) 0.497
Introduction Investment problem Application of real options Model Selection

min. median mean max. se. skew. kurtosis


Simulations 1.17 20.61 24.44 507.60 15.49 2.19 12.88
Observed prices 1.09 20.74 24.25 121.39 14.45 1.86 9.33

Table: Summary statistics for observed and simulated prices

Figure: A sample path of simulated Schwartz one factor process


Introduction Investment problem Application of real options Model Selection

Optimal investment threshold is at 39.5 e/MWh.

Figure: Exercise boundary for Schwartz one factor model as a function of time
Introduction Investment problem Application of real options Model Selection

Model Selection
min. median mean max. se. skew. kurtosis
Observed 1.09 20.74 24.25 121.39 14.45 1.86 9.33
ABM -807.65 17.29 16.61 811.51 141.38 -0.03 4.01
OU -46.11 24.13 24.17 104.79 14.44 0.01 3.01
CIR 0.16 21.82 23.98 134.43 12.50 1.06 4.69
Schwartz 1.17 20.61 24.44 507.60 15.49 2.19 12.88

Table: Summary statistics for observed and simulated prices

ABM OU CIR Schwartz


In sample RMSE 127.07 20.36 19.04 21.08
(68.57) (1.30) (1.32) (1.85)

Out of sample RMSE 37.16 14.51 12.08 13.78


(20.03) (3.33) (4.08) (5.30)

Table: Mean values of RMSE for different processes with standard errors in
brackets
Introduction Investment problem Application of real options Model Selection

Impact of model selection on investment threshold

Process Value (e/MWh)


NPV 20.6
ABM 186
OU 40.9
CIR 38.5
Schwartz 39.5
Table: Threshold prices for different stochastic processes

Choose CIR because of:


• Analytical solution
• Performs better than other processes when looking at MC
simulations and RMSE.
Introduction Investment problem Application of real options Model Selection

Thank you on your attention

Вам также может понравиться