determinants of forward premia in electricity markets: a ... · determinants of forward premia in...
TRANSCRIPT
Determinants of forward premia in electricity markets:
A taxonomic empirical analysis
Christian Redl1,a, Derek W. Bunnb aEnergy Economics Group, Vienna University of Technology
bEnergy Markets Group, London Business School
Abstract
A taxonomy of electricity forward premia determinants is introduced. Preliminary empirical
models give insights into the corresponding propositions on the forward premium. The risk
attitude of participants in the electricity market is strongly influenced by the agents’
assessment of energy commodities, which serve as fuel input or are of sentimental importance
for energy markets in general. Market participants react sensitively on volatility in the
electricity market itself and, additionally, on a tightening excess supply in the spot market
during the trading period of forward contracts. Furthermore, statistically significant demand
shock proxies underpin adaptive expectations of the actors in the forward market. Finally,
spot price mark-ups contribute to increased forward premia.
JEL classification
Q40; C10; G13
Keywords
Forward markets, predictive power, risk premium
1 Corresponding author. Energy Economics Group, Vienna University of Technology Gusshausstrasse 25-29/373-2, 1040 Vienna, Austria; E-mail: [email protected]; Tel.: +43 1 58801 37361; Fax: +43 1 58801 37397
2
1. Introduction
The goal of various worldwide liberalisation efforts of the electricity supply industry was the
introduction of competition as precondition for an efficient energy supply. Wholesale markets
were established and the prices are a result of the market forces.
In this new institutional environment risks emerged for market participants unknown
in the previous regulated area. Still, exchange traded futures or OTC traded forward contracts
allow for a management of the price risk by locking in a fixed price. Therefore uncertain
future spot prices can be avoided. In fact, electricity spot prices are characterised by high
volatility and occasional spikes (for a detailed analysis see e.g. Lucia and Schwartz (2002),
Burger et al. (2004), Huisman et al. (2007), Kanamura and Ohashi (2008), Karakatsani and
Bunn (2008), Bowden and Payne (2008), Higgs and Worthington (2008)). This is caused by
convex supply curves and a – in the short term – price inelastic demand. Supply or demand
shocks lead, in turn, to sudden rises in spot market prices. Moreover, as electricity cannot be
stored economically in suitable quantities, dampening effects of stocks are lacking. Hence,
these characteristics give rise to an increased demand for contracts.
The economic theory provides two alternative approaches for pricing these forward
contracts. The most common one is the theory of storage (Kaldor, 1939) which, however,
cannot straightforward be applied to electricity forwards since electricity cannot be stored in
economically suitable quantities. Instead, a second approach in economic theory considers
equilibrium relationships for forward pricing (Keynes, 1930): The (current) forward price can
be split up into a forecast of the future spot price at delivery and a risk premium. Hence, the
ex post forward premium Ft,T-ST is the key variable assessed in the (empirical) literature:
, , , , (1)
Equation (1) shows that the ex post forward premium equals the ex ante premium plus a
random error of the (rational) spot price forecast due to supply and demand shocks.
3
We argue that it is this difference which creates the importance of forward markets for
market actors and policy makers alike. As mentioned, the market participants are faced with a
forecasting problem and, depending on the spot price distribution and the attitude towards
risk, either demand a compensation for contracting, are willing to pay a corresponding
premium or to accept a discount to eliminate the risk of uncertain future cash flows. This
brings about important policy implications since it is the forward market, due to this economic
reasoning, which determines investments and welfare. Hence, it is necessary to understand the
components of the risk premium which warrants the attention of policy makers. In turn,
understanding the drivers of the forward premium allows a better regulation of electricity
markets and design of corresponding market rules. Nevertheless, due to the strong
interrelation between current events on the forward and spot markets (and current spot and
forward prices), the short term markets have to be integrated in the analysis of long term
markets.
Our analysis will focus on month-ahead futures – for several reasons: First, most price
data is available for futures with monthly delivery periods. Second, due to the shorter and
subsequent delivery period, forecast errors – in this analysis modelled via supply and demand
shocks – are expected to be lowest in the case of month-ahead futures. More specifically,
prices on the last trading day are considered since monthly averaging of futures prices yields
autocorrelation in the residuals. Finally, considering the full history of prices of a specific
contract, results may not be robust due to the increased time to delivery – and lack of trading.
Specifically, this analysis aims to unravel the components of the ex post forward
premium in a comprehensive taxonomic manner with the help of a structural empirical
modelling approach, whereas the literature on forward premia modelling typically focuses on
risk aversion measured by higher central moments (up to the third) and shocks in
inframarginal generation (typically hydro power) and demand. In our analysis econometric
models are used to assess forward pricing at the biggest regional European power market: the
4
Western European Power market with its leading power exchange, the European Energy
Exchange (EEX) based in Leipzig, Germany.
The article proceeds as follows: The next section summarises related literature in the field
of forward markets and compares our approach with existing literature. Section 3 introduces
the market setting. In section 4 the realised forward premia are quantified. Section 5 gives
propositions on the forward premia determinants and section 6 presents econometric models
of the baseload and peak load premia at the EEX. Finally, section 7 concludes.
2. Literature overview
Following Keynes (1930), futures prices are related to expected spot prices. This forward
pricing theory has extended to a broad stream of empirical literature, the most relevant for our
analysis is summarised below.
Findings on (ex post) forward premia
Gjolberg and Johnsen (2001) and Botterud et al. (2009) identify positive forecast errors
respectively positive risk premiums in the Nordic market. Gjolberg and Johnsen (2001) argue
that due to the identified size, differences cannot be explained by risk premiums only but
would indicate informational inefficiencies or the exercise of market power because of the
high concentration of suppliers. Weron (2008) determines the market price of risk in the Nord
Pool futures market using stochastic models. He finds increasing risk premiums with
decreasing time to maturity – this is equivalent to decreasing forward premia over time. Bunn
(2006) identifies positive risk premiums for peak hours when comparing the UK day ahead
and prompt market and the week ahead and day ahead market. He argues, that during peak
hours the demand side has a higher willingness to pay day ahead in order to avoid high
volatility in the intra-day market. Similarly, Longstaff and Wang (2002), Hadsell and Shawky
(2006), Diko et al. (2006), Lucia and Torro (2008), Douglas and Popova (2008), Redl et al.
(2009), Daskalakis and Markellos (2009), and Furio and Meneu (2010) find significant risk
5
premiums in long-term electricity markets. Bessembinder and Lemmon (2002) present an
equilibrium model where the forward premium (i.e. the difference between forward and
expected spot prices) is a function of the variance and skewness of spot prices. In turn, these
moments of the spot price distribution serve as risk assessment of market participants.
Douglas and Popova (2008) confirm the theoretical result of Bessembinder and Lemmon
(2002) for the PJM market. Moreover, they propose an augmented model including, among
others, gas storage inventories. Similarly, Lucia and Torro (2008), Redl et al. (2009) and
Furio and Meneu (2010) also confirm (at least partly) the results of Bessembinder and
Lemmon (2002). Additionally, supply and demand shock parameters are included in these
analyses to allow for a better capturing of risk versus forecast forward premium components.
In conclusion, the existing literature on empirical forward premia modelling typically
focuses on risk aversion (measured by higher central moments (up to the third) of the spot
price distribution) and shocks in inframarginal generation (typically hydro power) and
demand. However, our analysis aims to unravel the structural components of the ex post
forward premium in a comprehensive taxonomic manner with the help of an empirical
modelling approach. Therefore, in our evaluation, besides conventional risk assessment
measures and the representation of supply and demand shocks, additional variables are
introduced, which shall capture further sources of the futures-spot price bias. Before we
discuss these propositions on the forward premium components in detail, we will, however,
shortly examine the relevant market setting.
3. Market setting and corresponding price data
The European electricity market is still characterised by several price areas. Reasons for this
price divergence can be found, among others, in limited cross-border transmission capacities
(EC, 2005). However, quite a few regional electricity markets have emerged within the
European Union as some countries are not separated by cross-border transmission capacity
6
bottlenecks. One of these is the Central/Western European market comprising Austria,
Germany, France and, to a certain extent, Switzerland forming the biggest market in
Continental Europe. The EEX is the leading exchange in this sub market. In early 2007
implicit auctions between France, Belgium and the Netherlands have been introduced leading
to a coupling of these markets thereby effectively removing the market separation in North
Western Europe and extending the Central European market. Finally, wholesale prices in the
Czech Republic as well as Poland have reached the EEX level. Figure 1 depicts these price
developments and an increasing convergence over time.
Figure 1. Wholesale electricity prices for the considered regional electricity market. Source: Various power exchanges
Fig. 2 shows the price evolution of monthly averages of spot peak load prices as well as
month-ahead peak load prices, noted on the last trading day for delivery during the plotted
month, at the EEX from October 2003 to January 20102. Spot and forward prices were rising
continuously until early 2006. Since fossil fuelled power plants constitute the price setting
technologies in the EEX market, increasing power prices reflected rising primary energy
prices. The highest increases could be observed during 2005 due to the commencement of the
2 In figure 2 the depicted forward price at, e.g. October 2003, was the settlement price of the month-ahead peak load futures on 30 September 2003 for a delivery during peak hours in October 2003.
0
10
20
30
40
50
60
70
80
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
[€/M
Wh]
APX
EXAA
EEX
Powernext
Belpex
OTE
PolPX
7
European Emission Trading Scheme (EU-ETS). Spot and month-ahead prices were falling,
with a short exception, from March 2006 onwards mainly due to a massive drop-off in
emission allowance prices. In 2008, EEX spot and month-ahead prices again started rising due
to a price jump on the spot market for CO2 allowances when the second EU-ETS period
started in January 2008. Prices have been falling since the fourth quarter of 2008 due to
substantial price decreases in the oil and gas and, correspondingly, the CO2 markets.
Figure 2. Evolution of monthly averages of peak load spot prices (red) and peak load month-ahead futures prices on the last trading day (blue) at the EEX from October 2003 to January 2010. Source: EEX
Figure 3 presents descriptive statistics for daily EEX base and peak load spot prices from
October 2003 to January 2010. The price series are non-normal, positively skewed and show a
high kurtosis. Moreover, the peak load price series is more volatile than baseload series.
Figure 3. Descriptive statistics of daily base (left) and peak load spot prices (right) at the EEX from October 2003 to January 2010.
0
20
40
60
80
100
120
140
Okt.03
Feb.04
Jun.04
Okt.04
Feb.05
Jun.05
Okt.05
Feb.06
Jun.06
Okt.06
Feb.07
Jun.07
Okt.07
Feb.08
Jun.08
Okt.08
Feb.09
Jun.09
Okt.09
[€/M
Wh]
Month ahead futures price
Spot price
8
It is this distribution which faces the market participants with a forecasting problem of future
spot prices. Moreover, risk-averse agents have an incentive to reduce their risk exposure by
trading on the forward market. As will be shown in the following section, the willingness to
pay in order to reduce this risk exposure is significant.
4. Realised forward premia
First, for each monthly contract the relative ex-post difference between the forward price in
the trading period and spot price in the delivery period is determined:
∆ , (2)
where ΔT is the relative difference between the forward and spot price, FT-1,T is either the
average futures price in month T-1 for delivery in T or the settlement price on the last trading
day in month T-1 for delivery in T and ST is the spot price average in month T.
The differences between forward and corresponding spot prices are significant (see
figure 4). Table 1 summarises some additional statistics. On a monthly average, base load
contracts were traded 9% above actual spot prices in the delivery periods of the futures at
EEX. Month-ahead peak load futures were traded even 12% above spot prices in the delivery
period. The identified differences are significantly different from zero for a double-sided test.
Moreover, errors for base load and peak load are significantly larger than zero. If one looks at
each contract separately, the absolute value of the relative difference ΔT for peak load is
greater than for base load for almost every contract. Due to a higher slope of the supply curve
in peak load (unforeseen) variations in supply and demand induce greater price differences
between forward and spot prices in peak load which is confirmed by the results in Table 1.
Using futures prices on the last trading day instead of monthly averages for determination of
the relative differences ΔT still yields significant positive errors although the magnitude is
lower. On the last trading day, base load contracts were traded 5% above actual spot prices in
9
the delivery periods. Peak load futures were traded on the last trading day 7% above spot
prices in the delivery period (see Table 1).
Figure 4. Relative differences of month-ahead peak load futures prices (noted on the last trading day) with respect to the actual spot price during the delivery period at the EEX. Source: EEX, own calculations
Table 1. Summary statistics of forecast errors for monthly averages and for prices on the last trading day of month-ahead futures with delivery from October 2003 to January 2019 traded at EEX.
The above analysis does not consider seasonalities in the (relative) forward premium. Figure 5
shows a seasonal graph of the relative differences for peak load. Without further elaborating
this matter at this point, especially with respect to statistical inference, we note, from visual
inspection, that there is evidence of a seasonal pattern in the forward premium being highest
in January and lowest in the mid seasons April and September.
‐0,6
‐0,4
‐0,2
0
0,2
0,4
0,6
0,8
Okt.03
Feb.04
Jun.04
Okt.04
Feb.05
Jun.05
Okt.05
Feb.06
Jun.06
Okt.06
Feb.07
Jun.07
Okt.07
Feb.08
Jun.08
Okt.08
Feb.09
Jun.09
Okt.09
ΔT
Monthly average Last trading day Monthly average Last trading day
Mean 9% 5% 12% 7%
Standard dev. 21% 15% 26% 20%
Minimum -38% -38% -50% -50%
Maximum 87% 65% 98% 72%
t-statistic 3.66* 2.96* 4.04* 3.16*
EEXBase load Peak load
10
Figure 5. Seasonal graph of realised monthly percentage forward premia.
In the following section a categorisation of forward premia determinants is proposed which
shall mirror the risk and market assessment of the market participants and, moreover,
comprehensively describe the structural supply and demand characteristics and its effects on
the market outcome. Within each category several explanatory variables are described which
give further insights on the propositions on the electricity forward premium.
5. Propositions on electricity forward premia and a corresponding taxonomy of
structural components
Our aim is to extend established concepts of forward premia in electricity markets. We have
organised these components into a taxonomy of fundamental influences, behavioural effects,
market conduct, dynamic effects and shock effects. Specifically we provide insights on the
following propositions:
Fundamental influences
Fuels and their risk premia: Proposition: An increase in the gas forward premium is
expected to increase the electricity forward premium, whereas the effect is anticipated
to be more pronounced in peak load compared to base load.
11
Given the high importance of fossil fuelled generation technologies in the EEX
market, the risk premium prevailing in the electricity contract market is directly
influenced by the risk premium in the gas market. For our analysis, this influence can
be motivated by either risk management considerations, since – assuming gas fired
price setting technologies – the realised spark spread constitutes the risk exposure of a
generator having contracted gas, or forecast errors of the respective market participant.
An empirical comparison of these categories, still, is challenging. The oil market,
although oil fired power plants are rarely dispatched, given its “sentimental”
importance is also to be included in this analysis (see below discussion on market
sentiments and behavioural issues).
Electricity system fundamentals: Proposition: A positive relationship between scarcity
and the forward premium is expected.
The definition of the forward premium according to equation (1) assumes that market
actors can correctly anticipate the general fundamental drivers and deviations between
forward and spot prices occur due to risk assessments. Moreover, any deviations of the
fundamentals from the expected ones’ are the result of shocks (see below). However, a
sluggish reaction of the market to fundamentals (specifically the margin) is
hypothesised in this analysis and corresponding variables are defined (supply/demand
ratios). We now turn to the hypothesis of the market participants’ expectation
formation in detail.
Behavioural effects
Proposition: We postulate pronounced adaptive expectation formation with respect to the
risk assessments of the market participants. Among others, this is motivated by high
correlations between current spot and forward electricity prices. Furthermore, we argue that
spot price forecasts for a delivery period comprising one month ahead prove to be elusive (for
12
research aiming to model expectation behaviour and market participants alike). In this sense,
in our model the realisation of the relevant assessment parameters in the spot market of the
same delivery month a year ago as well as the realisation in the trading month of the forward
contract are used as proxies for the anticipated spot distribution realisations in the delivery
month.
Higher moments: Proposition: Central moments of the spot price distribution beyond
variance and skewness are of importance for the risk assessment of market actors.
Specifically, the kurtosis of spot prices is introduced additionally to capture the effect
of rare extreme deviations from the mean on the forward premium. Generally, a
positive influence of the kurtosis of spot prices on the premium is expected (given
positively skewed spot prices).
Spikes: Proposition: The forward premium increases due to the occurrence of spikes
in the spot market.
To test the reaction of the market to price spikes occurring in the spot market during
the trading period of the futures contracts above an average measure as the kurtosis
does, dummy variables which account for the occurrence of spikes are created.
Different degrees of spikiness respectively thresholds are defined (mean plus one, two
and three standard deviations) (Weron, 2006). The relevant spot price aggregation
level for estimating spikes at the EEX is the daily base or peak load spot price average
(Phelix Base, Phelix Peak), since the underlying of monthly futures contracts is the
monthly average of the Phelix day indices.
Price trends: Asymmetric effects are tested by dividing the whole sample period into
sub samples of underlying price increases/decreases and stable phases. Taking the
rockets and feathers theory as a foundation we would expect a delayed reaction of the
forward premium when spot price trends reverse from increasing to decreasing (see,
e.g., Borenstein et al. (1997) or Zachmann and v. Hirschhausen (2008) for an
13
application in the energy sector). Whether the forward market participants adapt to
lasting periods of price trends remains to be empirically assessed.
Oil market volatility: Proposition: Increased volatility in the oil market increases the
electricity forward premium.
Due to the dominance of underlying oil prices for energy commodities in general,
besides the above discussion on risk premia spilling over to electricity markets and
behavioural effects in the expectation formation, price volatility in oil spot markets can
contribute to an increased premium in electricity contracts.
Conduct
Market power: Proposition: The exercise of market power in the spot market positively
influences the forward premium. Different (theoretical) modelling approaches
analysing the competitive effects of the introduction of futures markets in electricity
markets with a highly concentrated supply side yield contrary results (e.g. Allaz and
Vila (1993) vs. Robinson and Baniak (2002)). The estimation of reliable forward
market concentration proxy variables, which would allow empirical insights, is,
however, not doable. On the other hand, estimated base load and peak load price mark
ups for the spot market are available (own research – see figure 6 below which depicts
the evolution of electricity prices (EEX) and estimated monthly averages of system
marginal costs in the regional EU-4-market from 1999-2009). This variable –
especially its relative pattern compared to observed spot prices – can be an indicator of
the abuse of market power of the dominant producers.3 We argue that producers who
can increase spot market prices demand a higher premium to contract forward. On the
3 We note that the marginal cost estimate is an average monthly value, which is compared to the average base or peak load price index. Start up costs or other opportunity cost considerations are, hence, not part of the monthly average cost estimate. On the other hand, brief downward excursions in the day ahead price (e.g. negative daily prices on certain days in 2009) can cause average monthly prices to decrease, which, however, is not reflected in the average SRMC estimate. Therefore, observed market prices can, at certain months, also be below the SRMC estimate.
14
other hand the buyers price generator market power as risk which increases the
willingness to pay forward. Yet, the linking of the mark up to the forward premium,
and the corresponding supply vs. demand side causes, is challenging from a
(empirical) modelling point of view.
Figure 6. Evolution of electricity prices (average baseload price at the EEX) and system marginal costs in the regional EU-4-market from 1999-2008. Source: EEX, BAFA, UCTE, own calculations
Dynamic effects
It has to be assessed whether there are any additional seasonal effects, above those which can
be fundamentally modelled (e.g. margin), in the forward premium. In a first step, a winter
dummy variable is intended to capture these additional effects. Still, further seasonal
representations, motivated by figure 4, need to be included in a comprehensive analysis.
Shocks
To be able to account for supply and demand shocks between forward trades and future spot
trades consumption and generation shock indices are introduced. The consumption shock
index is the ratio between actual electricity consumption in the relevant regional market and
the long-term average of the corresponding consumption in the specific month. Similarly, the
generation shock index of hydro and nuclear generation is the ratio between actual generation
0
10
20
30
40
50
60
70
80
90
100
Okt. 03
Feb. 04
Jun. 04
Okt. 04
Feb. 05
Jun. 05
Okt. 05
Feb. 06
Jun. 06
Okt. 06
Feb. 07
Jun. 07
Okt. 07
Feb. 08
Jun. 08
Okt. 08
Feb. 09
Jun. 09
[€/M
Wh]
Phelix Base
SRMC
15
and the corresponding long-term average of generation.4 Proposition: Consumption shocks
negatively influence the forward premium, whereas the influence of generation shocks is
positive. If, ceteris paribus, consumption is unexpectedly high in the delivery month spot
prices should exceed forward prices. On the other hand, if, ceteris paribus, inframarginal
generation in hydro and nuclear plants rises unexpectedly spot prices should fall below
forward prices since the supply curve is shifted to the right.
The following table 2 qualitatively summarises the above introduced propositions on the
effects of forward premia components and respective proxy variables.
Table 2. Summary of forward premia determinants.
Effect on forward
premium Proxy variable
Fundamentals
Premia in fuels + Month ahead gas forward premium
Supply margin - Ratio generation/consumption in the regional market Behavioural effects
Variance ~ Coefficient of variation of spot price
Skewness + Skewness of spot price
Kurtosis + Kurtosis of spot price
Spikes + Count spikes outside 1, 2, 3 standard deviations of mean spot
Trends + Dummy to account for 1, 2, 3 month continuous spot increase
Oil volatility + Coefficient of variation of Brent oil spot price
Conduct Spot market
power + Fundamental cost mark up estimate for regional spot market
Dynamics
Winter seasonals + Dummy to account for winter months
Shocks
Supply shocks + Dummy to account for high inframarginal generation in delivery
month
Demand shocks - Dummy to account for high consumption in delivery month
In the following section structural models aiming to give insights on the above propositions
are presented.
4 The scope of the shock supply variables can be refined by considering other inframarginal generation besides hydro and nuclear power.
16
6. A model of ex post forward premia
6.1 Adaptive expectation formation
A model is presented where market participants form myopic expectations. In this sense, they
are influenced by current and historic events on the spot market. These events, in turn,
contribute to the risk and market assessment of the agents and, hence, to the forward
premium. All parameters are observable for the market participants on the last trading day of
month t.
Base load
Sequentially minimising the AIC criterion and excluding the non-significant coefficients from
the comprehensive model – characterised by all parameters discussed in section 5– yields the
following equation for the ex post baseload forward premium:
, , , (3)
where Ft,T-ST is the ex post forward premium, Ft,T is the futures price on the last
trading day in month t for delivery in month T, ST is the spot price average in month T, cv(St)
is the coefficient of variation of daily spot prices in month t, cv(Brentt) is the coefficient of
variation of daily Brent spot prices in month t, FPGas t-1,t is the realised gas forward premium
of a month ahead futures for month t, and ConsT is the consumption index in month T. Results
for the corresponding model are shown in table 3.
The significant positive influence of volatility in the oil market confirms the
“sentimental” importance of the oil market for energy commodities in general. Interestingly,
its influence is as important as the influence of the volatility on the electricity market itself.
The volatility of electricity spot prices positively influences the futures price and, hence, the
forward premium. The influence of the spot price volatility on the forward premium is in
agreement with the literature (Bessembinder and Lemmon (2002) and the corresponding
empirical literature cited above), however, the sign of this measure seems to be indeterminate
17
as in our case it is positive.5 Realised premia in the gas market influence, as expected, the
electricity premia significantly positive, which shows the importance of gas fired power plants
also in baseload. Finally, the consumption shock coefficient gives the expected sign and is
statistically significant. Therefore, we consider this variable as being important for assessing
misjudgements of future demand conditions.
Table 3. Results of regression analysis (3) for ex post forward premia of month-ahead baseload futures at EEX for monthly delivery periods (t-statistics in brackets). All tests are based on heteroscedasticity consistent standard errors. Results are shown for premia determined by futures prices on the last trading day.
Coefficient Variable Base load
b1 Constant -3.25 (-1.09)
b2 Coeff of Var. Spot t 21.74 (3.18)
b3 Coeff of Var. Brent t 74.59 (2.17)
b4 Forward premium gas t 0.33 (1.92)
b5 Cons T -4.85 (-2.77)
R2 (R2corr) 0.21 (0.16)
DW 1.98
F-statistic 4.37
Serial correlation χ212 (p-value) 0.301
Functional form χ21 (p-value) 0.9065
Normality JB (p-value) 0.000
Heteroscedasticity χ24 (p-value) 0.744
Observations 71; 11/03-09/09
Peak load
A similar procedure to the above described one yields the following equation for the ex post
peak load forward premium:
, , ,
(4)
where Ft,T-ST is the ex post forward premium, Ft,T is the peak load futures price on the
last trading day in month t for delivery during peak hours in month T, ST is the peak load spot
price average in month T, Margint is the ratio of regional generation and demand in month t,
5 Note that in our analysis volatility is measured via the coefficient of variation – and not via variance. Among others, this is motivated by allowing a better comparision between different “informational sources” of volatility for market actors (i.e. oil and power market volatility).
18
FPGas t-1,t is the realised gas forward premium of a month ahead futures for month t, Spot
market powert is the ratio of the spot price in month t and the fundamental marginal cost
estimate for month t, and ConsT is the consumption index in month T. Results for the
corresponding model are shown in table 4.
Realised premia in the gas market influence, as expected, the electricity peak load
premia significantly positive. Generally, the price setting technologies in peak load hours are,
in fact, gas fired power plants. The significant positive influence gas market confirms the
importance of these generation technologies. Interestingly, the forward premium is positively
influenced by the market power estimate. In fact, spot price mark ups yield increases in the
forward premium. This can be caused by a higher willingness to pay of the buyers, which
price generator market power as a risk factor, a compensation demanded by dominant
producers to be willing to sell forward – and hence loose incentives to exercise their market
power in the spot market due to the contracted generation (Newbery, 1998), or a combination
of both. This result suggests, that the (positive) competitive effect of forward markets is, in
fact, limited.
If market participants perceive an increasing margin (or, correspondingly, a decreasing
scarcity) in the spot market, the forward premium tends to decrease (significant on a 10%
level). A decreasing margin is related to the increased likelihood of spikes occurring in the
spot market and, due to the convex supply curve, an increased skewness of spot prices. This
should increase the willingness to pay of risk averse buyers and represent opportunity costs of
producers having sold forward. Closely related to the spot price distribution is, in turn, the
scarcity of the system.
Due to this interrelationship statistical inference is difficult to obtain if parameters and
higher moments characterising the spot price distribution interfere with a fundamental
equivalent (e.g., in our case, the margin). Hence, the price distribution parameters are not part
of the model (4). Clearly, peak load prices are very sensitive to changes in a variety of
19
parameters and statistically significant inference generally proves to be elusive. Finally, the
consumption shock coefficient shows the expected sign and is statistically significant.
Table 4. Results of regression analysis (4) for ex post forward premia of month-ahead peak load futures at EEX for monthly delivery periods (t-statistics in brackets). Results are shown for premia determined by futures prices on the last trading day.
Coefficient Variable Peak load
b1 Constant 204.02 (1.63)
b2 Margin t -185.02 (-1.58)
b3 Forward premium gas t 0.93 (1.89)
b4 Market power spot t 12.65 (1.94)
b5 Cons T -10.28 (-2.41)
R2 (R2corr) 0.18 (0.13)
DW 2.04
F-statistic 3.51
Serial correlation χ212 (p-value) 0.767
Functional form χ21 (p-value) 0.090
Normality JB (p-value) 0.000
Heteroscedasticity χ21 (p-value) 0.419
Observations 71; 11/03-09/09
7. Conclusions and Outlook
We have introduced a taxonomy of electricity forward premia determinants in this paper.
Moreover, preliminary empirical models have been presented to give insights into the
corresponding propositions on the forward premium, which suggest the latter to be affected
by fundamental, behavioural, dynamic, conduct and unexpected components.
In fact, the risk attitude of participants in the electricity market is strongly influenced
by the agents’ assessment of commodities, which serve as fuel input (e.g. gas) or are of
sentimental importance for energy commodities in general (e.g. perception of the oil market
and its volatility). Moreover, market participants react sensitively on volatility in the
electricity market itself and on extreme events occurring in the spot market during the trading
period of forward contracts. This is mirrored in a significant influence of the scarcity of the
system. Interestingly, parameters indicating price mark ups and, correspondingly, the exercise
of market power contribute to increased forward premia which questions the competitive
20
effect of long term markets. Finally, demand shock measures contribute to the forward
premium. This is in line with our assumption on the adaptiveness of the market participants.
The results have to be interpreted with due care. First, they arise of a rather short data
set. As in any empirical analysis, the passing of time will allow a more complete investigation
of the market and an assessment of our propositions. In any case, due to the complex
interactions of market forces and its drivers, empirical models alone cannot encompass all
causal market relationships. Hence, as a next step, besides increasing the quality and
complexity of the empirical models – which will comprise a more subtle representation of the
evaluation of trend effects – we intend to provide further insights by developing a simple
equilibrium model of the market, which shall focus on issues which interact rather strongly
with each other: Market power and price mark ups, risk aversion, supply and demand shocks
and fuel price uncertainty. This research can build upon seminal work by Allaz (1991), Allaz
and Vila (1993), Newbery (1998), Green (1999) and Bessembinder and Lemmon (2002).
References
Allaz, B., 1992. Oligopoly, uncertainty and strategic forward transactions. International Journal of Industrial Organization 10, 297-308. Allaz, B., Vila, J.L., 1993. Cournot Competition, Forward Markets and Efficiency. Journal of Economic Theory 59, 1-16. Bessembinder, H., Lemmon, M.L., 2002. Equilibrium Pricing and Optimal Hedging in Electricity Forward Markets. The Journal of Finance 57(3), 1347-1382. Borenstein, S., Cameron, A.C., Gilbert, R., 1997. Do Gasoline Prices Respond Asymmetrically to Crude Oil Price Changes? The Quarterly Journal of Economics 112(1), 305-339. Botterud, A., Kristiansen, T., Ilic, M., 2009. The relationship between spot and futures prices in the Nord Pool electricity market. Energy Economics, In Press. Bowden, N., Payne, J.E., 2008. Short Term Forecasting of Electricity Prices for MISO Hubs: Evidence from ARIMA-EGARCH Models. Energy Economics, In Press, Accepted Manuscript, Available online 3 July 2008. Bunn, D.W., 2006. Risk and electricity price dynamics. Platts.com News Feature. Burger, M., Klar, B., Müller, A., Schindlmayr, G., 2004. A spot market model for pricing derivatives in electricity markets. Quantitative Finance 4(1), 109-122. Daskalakis, G., Markellos, R., N., 2009. Are electricity risk premia affected by emission allowance prices? Evidence from the EEX, Nord Pool and Powernext. Energy Policy (37). 2594-2604.
21
Diko, P., Lawford, S., Limpens, V., 2006. Risk Premia in Electricity Forward Markets. Studies in Nonlinear Dynamics & Econometrics 10(3), Article 7. Douglas, S., Popova, J., 2008. Storage and the electricity forward premium. Energy Economics 30(4), 1712-1727. EC, 2005. Report on progress in creating the internal gas and electricity market. Communication from the Commission to the Council and the European Parliament, COM(2005) 568 final, Brussels. Furio, D., Meneu, V., 2010. Expectations and forward risk premium in the Spanish deregulated power market. Energy Policy 38, 784-793. Gjolberg, O., Johnsen, T., 2001. Electricity Futures: Inventories and Price Relationships at Nord Pool. Discussion Paper #D-16/2001. Green, R., 1999. The electricity contract market in England and Wales. The Journal of Industrial Economics 47(1), 107-124. Hadsell, L., Shawky, 2006. Electricity Price Volatility and the Marginal Cost of Congestion: An Empirical Study of Peak Hours on the NYISO Market, 2001-2004. The Energy Journal 27-2, 157-179. Higgs, H., Worthington, A., 2008. Stochastic price modelling of high volatility, mean-reverting, spike-prone commodities: The Australian wholesale spot electricity market. Energy Economics, doi:10.1016/j.eneco.2008.04.006. Huisman, R., Huurman, C., Mahieu, R., 2007. Hourly electricity prices in day-ahead markets. Energy Economics 29(2), 240-248. Kaldor, N., 1939. Speculation and economic stability. The Review of Economic Studies 7, 1-27. Kanamura, T., Ohashi, K., 2008. On transition probabilities of regime switching in electricity prices. Energy Economics 30(3), 1158-1172. Karakatsani, N.V., Bunn, D.W., 2008. Intra-day and regime-switching dynamics in electricity price formation. Energy Economics 30(4), 1776-1797. Keynes, J. M., 1930. A Treatise on Money. London: Macmillan. Longstaff, F.A., Wang, A., 2002. Electricity Forward Prices: A High-Frequency Empirical Analysis. Working paper, University of California Los Angeles. Lucia, J.J., Schwartz, E.S., 2002. Electricity Prices and Power Derivatives. Evidence from the Nordic Power Exchange. Review of Derivatives Research 5(1), 5-50. Lucia, J.J, Torro, H., 2008. Short-term electricity futures prices: Evidence on the time-varying risk premium. Working paper. Newbery, D.M., 1998. Competition, contracts, and entry in the electricity spot market. RAND Journal of Economics 29(4), 726-749. Redl, C., Haas, R., Huber, C., Böhm, B., 2009. Price formation in electricity forward markets and the relevance of systematic forecast errors, Energy Economics 31(3), 356-364. Robinson, T., Baniak, A., 2002. The volatility of prices in the English and Welsh electricity pool. Applied Economics 34, 1487-1495. Weron, R., 2006. Modeling and forecasting electricity loads and prices: a statistical approach. Wiley, Chichester, ISBN 047005753X. Weron, R., 2008. Market price of risk implied by Asian-style electricity options and futures. Energy Economics 30(3), 1098-1115. Zachmann, G., v. Hirschhausen, C., 2008. First evidence of asymmetric cost pass-through of EU emission allowances: Examining wholesale electricity prices in Germany. Economics Letters 99, 465-469.