The Natural Number of Forward Markets for Electricity

1
The Natural Number of Forward Markets for
Electricity

• 9th Annual POWER Conference on Electricity
  Industry Restructuring
• March 19, 2004
• Hiroaki Suenaga and Jeffrey Williams
• Department of Agricultural and Resource Economics
• University of California, Davis
• suenaga_at_primal.ucdavis.edu, williams_at_primal.ucdavi
  s.edu

2
Common observations about electricity

• (1) Extremely volatile prices in spot wholesale
  markets
• short-run capacity constraints
• retail prices inflexible
• pronounced seasonality in demand
• short-run weather shocks
• electricity not storable
• (2) Underdeveloped forward wholesale markets
• most efforts by exchanges have failed
• California PX restrained to one-day-ahead
• generally, private bilateral trades

3
Our propositions

• Because of electricitys very properties,
  long-dated forward markets for electricity are
  essentially redundant.
• The NYMEX natural gas futures market duplicates
  an electricity futures market.

4
How to demonstrate that some price is
redundantif it cannot be observed?

• An idealized world for trading electricity
• full profile of forward prices
• forward prices are best possible forecasts by
  construction
• companion forward prices for a fuel
• An analogy with corn
• considerable price variation recently
• well developed futures market

5
Progressions of Prices of Corn Futures Contracts
6
Progressions of Prices of Corn Futures Contracts
7
Profiles of Corn Futures Prices in Mid June
8
Profiles of Corn Futures Prices in Mid June
9
Profiles of Corn Futures Prices in Mid June
10
Idealized Market Model (Spot Market)

• Generating and retailing firms trade wholesale
  electricity for a full constellation of delivery
  hours and days, far into the future.
• All firms are competitive and risk-neutral.
• Aggregate supply
• PSt b wt Qc-1(1 ?MC e1,t) e1,t ?MC
  e1,t -1 u1,t
• where wt price of primary input (fuel)
• b, c, ?MC, ?MC parameters
• u1,t iid N(0,1)
• Retail demand is exogenously determined (QAt).
• ? Equilibrium spot price in any hour Pt b wt
  QAtc-1(1 ?MC e1,t)

11
Exogenous Variables

• Demand (load)
• QAt QDT(t a)(1 ?QA e2, t) e2,t ?QA e2,
  t-1 u2, t
• Fuel Price
• wd w0, d ?w vd e3,d e3,d ?w e3, d-1
  u3, d
• w0, d w0(d b)
• vd v(d g)
• u2,t, u3,t iid N(0,1)
• A total of 25 parameters with 3 stochastic
  factors (a shock to the load, a shock to the fuel
  price, and a shock to the cost of generation).

12
Seasonal and Diurnal Variations in Deterministic
Load QDTt
13
Seasonal cycles in fuel price and price variance
14
Simulated data - 3 price relationships

• (1) Examine forward profiles For each delivery
  hour t, generate as many forward prices, Ft,t-k,
  as the number of k. Each is the best, unbiased
  forecast by construction (Ft,t-k Et-kPt).
• ? If the profiles consistently attenuate to a
  stable price, forward prices beyond that time
  ahead are redundant.
• (2) Examine spreads
• ? If the spread between the forward prices of two
  distinct delivery hours is stable, one price can
  be deduced from the other.
• (3) Compare the forecasting ability of the
  forward price of primary input (wt,t-k) with that
  of the forward electricity price (Ft,t-k).
• ? If the price movements of the two commodities
  are highly correlated, one forward price can be
  deduced from the other.

15
Representative time series of simulated spot
prices
16
Representative time series of simulated spot
prices
17
Representative time series of simulated spot
prices
18
Representative time series of simulated spot
prices
19
(1) Progressions of an electricity forward price
20
(1) Progressions of an electricity forward price
21
(1) Progressions of an electricity forward price
22
(1) Progressions of an electricity forward price
23
Variation across realizations in a forward price
24
Variation across realizations in a forward price
25
(2) Spreads among forward prices for three
distinct delivery periods (Hour 18, Aug. 1, 2,
and 8) - Base parameter case
26
(3) Forecasting ability

• Regression Models
• (1) ln Pt a0 b0 ln Ft,t-k e0,t
• (2) ln Pt a2 b2 ln wt,t-k c2 ln QFt,t-k
  e2,t
• Load forecast, QFt,t-k, in (2) allows the market
  heat rate to be non-constant and vary by season.
• If the R2 for (2) is close to the R2 for (1), the
  forward price of fuel predicts the spot
  electricity price as accurately as the
  electricity forward.
• ? If so, the benefit from a separate forward
  market for electricity would be small.

27
R-squared for regressions explaining the
electricity spot priceRegressor Electricity
forward
28
R-squared for regressions explaining the
electricity spot price - Comparison
29
R-squared for regressions explaining the
electricity spot price - Sensitivity
30
One-Month-Ahead Forecasting Ability of Corn
Futures Contracts (1996-2001)
31
Six-Month-Ahead Forecasting Ability of Corn
Futures Contracts (1996-2001)
32
Eighteen-Month-Ahead Forecasting Ability of Corn
Futures Contracts (1996-2001)
33
Thirty-Month-Ahead Forecasting Ability of Corn
Futures Contracts (1996-2001)
34
Conclusions

• Forecasting ability of electricity forward prices
  inevitably low.
• Local electricity forward price profiles well
  represented by
• Local spot markets plus forwards perhaps as far
  as a week ahead.
• Regional month-ahead energy forward market, such
  as natural gas.
• National benchmark long-dated energy forward
  market, such as the NYMEX natural gas.
• Complex varieties of contracting likely
• Local long-dated forward basis agreements.