Working PaperPDF Available

Liquidity and Risk Premia in the New Zealand Electricity Futures Market

Authors:
1
Liquidity and Risk Premia in the New Zealand
Electricity Futures Market
Fergus Bevin-McCrimmon
Department of Accountancy and Finance, University of Otago, Dunedin, New Zealand
Ivan Diaz-Raineyŧ
Department of Accountancy and Finance, University of Otago, Dunedin, New Zealand
Greg Sise
Energy Link Ltd., Dunedin, New Zealand
May 2016
Abstract
Despite being one of the early adopters of electricity markets and an example of ‘textbook’
reform, to date, no academic research on the liquidity or risk premia of electricity futures
markets is discernible for New Zealand. Using data from October 2009 to December 2015 we
address this gap in the literature. We find that liquidity has been gradually increasing and that
a policy intervention to impose a maximum bid-offer spread was associated with a liquidity-
enhancing structural breaks, but this was evident only in the nearest-to-maturity futures
contracts. Further, we develop models to explain risk premia that include a range of risk
factors which we categorise as either statistical, physical market, production cost, investor
behaviour and liquidity variables. From this analysis we document significant time varying
premia which are driven by potentially inefficient behaviour.
JEL Classification: D46; G12; G13; L94; Q41; Q48
KEYWORDS: Electricity markets, electricity prices, electricity futures, liquidity, risk premium
ACKNOWLEDGEMENTS: This paper was possible due to two scholarship received by Fergus
Bevin-McCrimmon, namely a Otago Business School Kick-Start Summer Research Scholarship
(2015/2016) and the Energy Link Limited Energy Markets Scholarship (2016). We thank the
wider Energy Link team (in particular Mark Nelson) for their assistance in data collection and
their insights into New Zealand electricity market. The usual disclaimer applies.
ŧCorresponding author: Tel. +64 3 479 8117; Email: ivan.diaz-rainey@otago.ac.nz
2
1. Introduction
Electricity markets have unique characteristics, for example the fact that electricity cannot be
stored economically at the grid level, and have evolved rapidly. Ever since a rise of
deregulation and privatisation in the 1990s, various electricity markets models have emerged.
The New Zealand (NZ) electricity market (NZEM) is an early example of a deregulated market
structure, and is considered by some to be a ‘textbook’ reform (Joskow, 2006). A
characteristic inherent in the nature of electricity markets is the high degree of volatility in
spot prices. As in any market, this creates challenges for both market participants and
regulators alike, and in the NZEM volatility of spot prices is exacerbated by the predominance
of hydro-electric generation which has limited storage and exhibits a high degree of volatility
in the inflows into storage lakes. In order for these participants, whether they be generators,
retailers or large consumers, to successfully operate in this sector effective hedge instruments
need to be available to mitigate risk. One such tool available in the NZEM are futures
contracts.
A consultation by the Electricity Authority (EA), the regulatory body of the electricity
market in NZ, suggested that the hedge market should provide effective means of managing
spot price risk whilst giving a transparent view of future prices (EA, 2014). The EA added that
the “more liquid the hedge market is, the better these ends will meet” (EA, 2014;p.11).
Unfortunately, responses from market participants to the EA consultation indicated that the
futures market may not be supporting risk management and price transparency as it might.
These concerns primarily stem from the illiquidity of the futures markets as reflected by low
depth of volume and large spreads (EA, 2014;p18-19). This is despite EA interventions that
attempted to promote liquidity and stimulate trading activity (most prominently on 5th
January 2010 the ASX introduced market making for the four largest generators and after
work commissioned by the EA recommended reducing the maximum bid-ask spread, on the
3rd October 2011 the market-making moved to a maximum bid-offer spread of 5% in the
futures market).
Futures contracts allow participants to manage their overall risk profile and exposure
to the underlying spot market. The ability to do so has inherent value. Such value will be
reflected in the risk premium present in these futures contracts. Previous research has found
significant premia in electricity futures which are relatively large in comparison to those found
in other commodities (Shawky, Marathe, & Barrett, 2003). Indeed, within the NZ context
3
there have been concerns that large positive risk premia exist and that these may potentially
be indicative of inefficiencies in the market (EL, 2014;p.1).
Despite being one of the early adopters of electricity markets and an example of
‘textbook’ reform, to date, no academic research on the liquidity or risk premia of electricity
futures markets is discernible for New Zealand. Accordingly we provide the first detailed
examination of both liquidity and risk premia in the New Zealand electricity futures market.
In doing so we address two research questions:
Q1: How has the liquidity of the NZ electricity futures market evolved over time, and have
efforts to increase liquidity succeeded?
Q2: What drives risk premia in the NZ electricity futures markets and is illiquidity a factor
in these premia?
These questions are addressed by using data for the period 2nd October 2009 to 31st
December 2015. We employ from the literature two measures of liquidity/illiquidity) on
which we run structural break tests (Bai and Perron, 2003). Further we develop models to
explain risk premia that include a range of risk factors which we categorise as either statistical,
physical market, production cost, investor behaviour or liquidity variables.
The rest of the paper is structured as follows. Section 2 provides more background on
the NZEM and the related futures market. Section 3 reviews respectively the relevant
literatures on liquidity and on risk premia in electricity futures markets. Section 4 outlines the
data and models we employ. Section 5 presents our results and section 7 provides some
conclusions.
2. The New Zealand Electricity Market: An Overview
The NZEM utilises locational marginal pricing, or nodal pricing, with spot prices set half hourly.
This mechanism simply refers to the fact that there is no single market price and that spot
prices vary across the country, reflecting the marginal cost of supplying electricity at that
particular location. Another characteristic of the NZEM is the use of a mandatory pool,
meaning that apart from a small portion of supply generated and consumed entirely within a
consumer’s site, all electricity must be traded through the spot market. This differs from, for
example, Great Britain, where the majority of wholesale electricity is contracted prior to real-
time and only a small portion is traded through a balancing market at real-time.
4
Due to the natural resources NZ is endowed with, in particular voluminous rivers and
lakes, generation in the NZEM comes primarily in the form of hydro currently around 57%
of total energy generated (MBIE, 2015). In this sense the NZEM is similar to the Nord Pool
market, which has received considerable academic attention (See, for example, Botterud et
al., 2010; Fleten et al., 2015; Weron & Zator, 2014). As a result, the approach we employ in
this paper in the context of the NZEM draws from the literature on Nord Pool.
Though much of the generation comes from hydrological sources, there is still a
significant portion which comes from alternative sources. Geothermal, gas and coal comprise
approximately 18, 13 and 4 percent, respectively (MBIE, 2015). The decisions to utilise these
peaking plants will not only be influenced by the wholesale spot price of electricity, but also
fossil fuel prices such as oil, gas and coal. Another relevant cost of production in New Zealand
is the price of carbon. Since its introduction in 2008, the New Zealand Emissions Trading
Scheme (NZ ETS) has applied to electricity generators (see Diaz-Rainey & Tulloch, 2016).
Trading in futures contracts in the NZEM commenced July 2009 and is referenced to
two nodes, or locations, on the power grid: Benmore in the South Island and Otahuhu in the
North Island. Initially both quarterly and “strip” contracts were traded, while monthly
contracts were not introduced until late in 2015. This paper focuses on the quarterly
contracts.
3. Literature on Liquidity and Electricity Risk Premia
3.1 Liquidity Literature
Liquid markets have long been considered to exhibit certain characteristics. Kyle (1985),
building Black (1971)’s description of a liquid market, developed terms to describe the degree
of liquidity in a market, for example; tightness, depth and resiliency. As we are concerned
with examining the liquidity of electricity markets in this paper, it follows that we must
determine how to measure or capture these characteristics.
A commonly used liquidity measure is simply trading volume. It has been shown that,
all else being equal, volume is an increasing function of a market’s liquidity
1
. The use of trading
volume as a measure of liquidity can be justified by the argument of Brunnermeier and
Pedersen (2009) that trading itself can be an important liquidity indicator. However, as noted
1
For example, see Amihud and Mendelson (1986a), Brennan and Subrahmanyam (1995), Brennan, Chordia,
and Subrahmanyam (1998) and Glosten and Harris (1988).
5
by Fleming (2003), increased trading volume is also associated with greater volatility, which
is considered to negatively impact liquidity. Therefore, it is not always clear what an increase
in trading volume is actually reflecting. All things considered, its ease of calculation and
interpretation makes this an attractive measure.
Tightness and transaction costs are interrelated, so the level to which market
participants are exposed to transactions costs will be reflected by the degree of tightness in
a market. The bid-ask spread can be considered one of these costs, as it is the price paid for
the provision of immediacy (Demsetz, 1968). Furthermore, it has been shown to display a
strong negative correlation with several aspects that reflect liquidity, such as trading volume
and the number of market makers. It is therefore appropriate to consider the bid-ask spread
a natural measure of liquidity (Amihud & Mendelson, 1986b).
Though spreads can be effective estimates of transaction costs incurred by market
participants, and by proxy liquidity, its use is not always a viable option since it requires high-
frequency data that is not always easily available. In response, several authors have
developed spread proxies which attempt to capture these costs with low-frequency data. For
example, Roll (1984) developed the Roll Spread, Lesmond et al. (1999) formulated a “Zeros”
measure, while Corwin and Schultz (2012) defined the High Low Spread. Though such
measures have been demonstrably effective, some are relatively computationally intensive,
which is to be expected considering the task they are carrying out.
In terms of the characteristics of depth and breadth, liquid markets will be better able
to absorb large market orders without a significant impact on the market price than less liquid
markets, ceteris paribus. Should markets be unable to efficiently do so, participants will suffer
price impact costs as they “walk up the book”. This price impact is another potential
transaction cost faced by investors and one which can also be captured through proxies.
Again there are measures which can effectively capture this characteristic of liquidity,
but at the cost of a need for high-frequency data. As was the case with spread proxies, several
authors have attempted to overcome such issues. The most notable of these is the Illiquidity
Ratio developed by Amihud (2002). This is simply the ratio of absolute daily return to daily
volume and can be interpreted as the “percentage price change per dollar of daily trading
volume, or the daily price impact of the order flow” (Amihud, 2002; p.34). Alternatively, it can
be considered “a measure of consensus belief about new information” (Marcelo and Quiros,
2006; p.258). As the name implies, a higher value represents a greater degree of illiquidity.
6
In terms of measuring liquidity in an electricity market context, the literature is thin.
Frestad (2012) examined the liquidity of the Nord Pool swaps market, utilising trading volume,
open interest, bid-ask spread and trading volatility as their chosen measures. Hagemann and
Weber (2013) also used volume and spread, as well as a high-low difference, to measure the
liquidity of the German intraday electricity market.
3.2 Risk Premia in Electricity Markets
Within the literature examining risk premia in electricity markets, there are essentially two
methods which can be employed. Researchers can either examine the ex-post, or realised
premia, or the ex-ante premia. Those who examine the ex-ante premia must develop an
estimate of the expected future spot price and, as a result, will have results which are at least
partially determined by their modelling methods. To overcome such issues, a majority of the
literature focuses on the ex-post premia which uses actual realised market prices.
Significant premia have been found in electricity markets all over the world: Longstaff
and Wang (2004) in PJM (in the USA), Kolos and Ronn (2008) in EEX (covers Central Europe)
and Botterdud et al. (2010) in Nord Pool (covers Nordic and Baltic markets). However, despite
this, there exists no consensus with regards to the sign of this premia. For example, Furió and
Meneu (2010) found both significantly negative and positive premia in the Spanish market.
Viehmann (2011) also found mixed results in the German day-ahead market. This may be in
some part due to the seasonal and diurnal nature of electricity markets. For example,
Karakatsani and Bunn (2005) found the sign flipped throughout the day, while Lucia and Torró
(2011) found that it was dependent on the season in which the contract matures.
Another strand of the literature examines the determinants of premia, which are in
turn influenced by market participants hedging decisions. The seminal paper in this area was
the equilibrium hedging model developed by Bessembinder and Lemmon (2002) (hereafter
the “B&L Model”), which considered the statistical risk factors of the underlying spot price to
be important determinants. They express the forward premium as being linearly determined
by the variance of the spot price and the standardised skewness of the spot price. That is:
  
(1)
7
Based upon their equilibrium model,   and   , meaning that the forward premium
should be positively related to the expected skewness of the spot price and negatively related
to the variance of the spot price. These hypothesised relationships and coefficient signs have
been given empirical support by (amongst others) Lucia and Torró (2011), Douglas and
Popova (2008) and Longstaff and Wang (2004) - albeit only partial support in some cases.
However, there have also been several conflicting pieces of work. For example, Bunn and
Chen (2013) found a significant negative relationship between risk premia and the skewness
of the underlying spot price, as well a significant positive relationship with the volatility. This
led them to suggest that the mixed results in the literature are due to under specifications
and that “exogenous and endogenous variables are often more significant than the statistical
measures (of variance and skewness)” (Bunn and Chen, 2013;p.183) and omitting them can
have a significant impact on analyses.
These other variables mentioned typically concern physical market factors. Douglas
and Popova (2008) considered these in their development on the B&L Model and believed
the original model “(did) not address the question of what factors cause the variance and
skewness […] to vary” (p.1716). As their study was conducted in the context of the PJM
Interconnection, the physical market factors they considered were related to gas storage
inventories. Clearly the physical market variables used in analyses will need to reflect the
conditions of the underlying physical market.
A more appropriate model to draw upon for the NZEM is that used by Botterud et al.
(2010) for NordPool. In a similar augmentation of the B&L model, they suggested that the
forward premium be related to the availability of hydro generation. Empirical analysis
confirmed this relationship and was supported later by Lucia and Torró (2011). Both papers
also gave evidence supporting a time-varying premia, which suggested that one should not
only consider physical market factors, but also the effects of prior premia.
The third area which may influence participants hedging decisions is the underlying
cost of production. Bunn and Chen (2013) deviated from examining the effects of the physical
capabilities of the systems and used factors related to generators costs of production, such
as gas, coal and carbon prices. Similar costs were also used in the model of Fleten et al. (2015).
8
4. Methodology
4.1 Data
We obtained NZEM data from Energy Link Limited. This consisted of trading volumes and daily
closing prices of quarterly futures contracts, as well as hydrological inflows and storage and
market demand. The contracts under consideration began trading from the 14th July 2009.
However, the initial months of trading saw unusual trading behaviour. To remove the impact
of this period, out sample period begins 2nd October 2009 and extends to the 31st December
2015.
Financial market data was sourced from Bloomberg. This included daily prices for
carbon (New Zealand Units (NZUs) from New Zealand’s Emissions Trading Scheme), oil (Dubai
Fateh Oil) and a general local stock market index (NZX50 Gross Index). Attempts were made
to source data for both coal and gas prices, however unfortunately no reliable data series with
an appropriate frequency were available.
For the electricity futures data, at any point in time, there are contracts traded for
every quarter up to three years ahead. This paper considers three of these contract “forms”.
The first of these is what we term the “front end” contract. This simply refers to the contract
with the closest maturity. We also consider the contracts which mature in the current quarter
one year ahead and two years ahead, respectively. For example, at the start of our dataset in
the fourth quarter of 2009, we were considering the Q4 2009 contract, the Q4 2010 contract
and the Q4 2011 contract.
Table 1 presents a description and the basic descriptive statics for the variables used
in this study, this includes our two measures of liquidity which we describe below.
[INSERT TABLE 1 ABOUT HERE]
4.2 Liquidity Measures and Anticipated Structural Breaks
The first measure we employ is simply trading volume. Though we are aware of potential
issues around its use, the simplicity of both calculation and interpretation overcome these.
We define this as simply the number of contracts traded each trading day. In November 2015,
the size of the NZEM futures contracts was reduced from 1 megawatt (MW) to 0.1MW i.e.
decreased by a factor of ten. In order to correct for this and allow for accurate comparisons,
the trading volume used from the 1st November 2015 is given by:
9


(2)
The second measure utilised is the Illiquidity Ratio developed by Amihud (2002). This is
defined as:


(3)
where  is the logarithmic return
2
on futures contract on day   is the trading
volume of futures contract on day . As mentioned in Section 3.1, there are several liquidity
proxies present in the literature. The use of this particular measure stems from the findings
of Marshall et al. (2012), who found it was the best and most consistent measure when
compared to liquidity benchmarks. Though other conflicting results exist, this was carried out
in the context of commodity futures and, while not directly considering electricity, is the most
relevant. These conflicting results are likely due to the fact that one single measure cannot
possibly capture all aspects of liquidity (Amihud, 2002).
There is one issue with the use of this Illiquidity Ratio in this context which likely hasn’t been
present for its previous users. The ratio can only be calculated on days where trading actually
takes place, that is, when volume is positive. Due to the lack of activity present in the NZEM
futures market, there are many days which do not meet this criteria. Such an issue would
obviously very rarely be present in the equity markets it was initially developed for.
In order to determine the effectiveness of interventions which aimed to enhance
liquidity we carry out structural break tests (Bai and Perron, 2003) and identify changes in
structure using global information criteria. We do so on our two measures of liquidity and
there are interventions that sought to increase liquidity that are of interest. The first occurred
in January of 2010 when the mandatory market making was introduced involving the four
largest generators. This involved introducing a daily window during which the four largest
market participants must post bid and ask prices and be willing to trade at these quotes. The
second was in October of 2011 and involved a reduction of the maximum bid-ask spread from
2
Where the logarithmic return on day is defined as 
, where and  are the closing prices on
days and   , respectively.
10
10% to 5%. As already mentioned, a reduction in the bid-ask spread is a typical indicator of
improving liquidity.
4.3 Modelling Risk Premia
As with much of the prior literature we examine the ex-post, or realised risk premia. We define
this as:
  
(4)
Where  is the risk premia at time for a contract which matures at time .  is the
closing futures price at time for a contract which matures at time , while is the spot
price at time . NZEM futures contracts settle relative to the arithmetic average spot price
over the settlement quarter, so is calculated appropriately. That is, those who purchase
these contracts pay/receive the differences between the future’s price they entered the
contract at, and the average half-hourly wholesale spot price over the settlement quarter.
We determine the drivers of premia using several models of varying complexity. The
first of these considers the statistical risk measures first suggested in the B&L model. This
model is defined as:
   


(5)
where  and  are the variance and skewness of the spot price, ,at node
, at time . In order to effectively capture the market environment which participants
hedging decisions are based upon, these are calculated based upon the past seven days for
front end contracts and past thirty days for the other two forms. In each of our models, we
include a set of quarterly dummies (). This is to control for the significant seasonality
displayed by risk premia. We also control for the effects of the premia in previous periods to
overcome any heteroscedasticity or autocorrelation issues
3
.
From Bessembinder and Lemmon (2002), we anticipate that the coefficient for
skewness will be positive. Greater skewness in the spot market represents a greater chance
of price spikes in the market. In order to hedge against this risk, we expect that market
3
The choice regarding the number of lags included is based upon the Durbin-Watson and Breusch-Godfrey
regression diagnostics. That is, lags are included until the regression “quality” ceases to increase.
11
participants will demand a greater quantity of these futures contracts and, as a result, drive
the price and premia up. The model developed by Bessembinder and Lemmon (2002) also
suggests that there should be a negative relationship between premia and the variance of the
spot market. However, intuition suggests that this should instead be a positive relationship.
Greater variance brings with it greater spot price risk for both generators and retailers. Hence,
the same argument can be made as for the relationship between skewness and premia. It
should be noted, however, that there is no consensus in the empirical work done with regards
to the signs of these coefficients.
We then include variables which reflect the physical state of the market. Due to the
hydro dominated generation system present in NZ, these variables are similar to those
utilised by Botterud et al. (2010). We define this model as:
   



(6)
,  and  represent deviations from the historical average value
during that week of the year from hydrological storage, hydrological inflows and electricity
demand (See Table 1). For each week of the year     over the sample period, an
average is calculated for each variable. To calculate the series of deviations, we subtract the
average for the particular week from the realised value. We anticipate that coefficients for
both storage and inflow deviations will be negative. When hydrological storage and inflows
are below expectation i.e. negative deviations, hydro generators capacity to supply electricity
is reduced. This increases the potential for spikes in the spot market (Botterud et al., 2010).
Both Botterud et al. (2010) and Weron and Zator (2014) find significantly positive
relationships between demand deviations and premia in the Nord Pool electricity market. We
anticipate the same. When demand is higher than expected, spot prices will likely increase in
response and increase the incentive for purchasers to hedge such exposure.
Drawing upon the argument of Redl and Bunn (2013) that premia will be related to
the underlying fuel source, we also include production costs as these are relevant to New
Zealand (See Section 2). These are carbon and oil prices and create the model:
12
   
   



(7)
As outlined by Fleten et al. (2015), increases in fossil fuel prices will lead to higher electricity
spot prices due to increased production costs. This, in turn, increases demand for futures
contracts and results in greater forward premia. Hence, we anticipate the coefficient for 
to be positive. Since the introduction of the New Zealand Emissions Trading Scheme (NZ ETS)
in July 2010, the price of NZUs is also a relevant cost of production for electricity generators
and retailers (See Section 2). In the same manner that we would expect fossil fuel prices to
positively affect forward premia, the coefficient for  would be expected to be
positive. For both fuel variables we use the logarithmic return to overcome issues of non-
stationarity. Following Fleten et al. (2015), we also include the logarithmic return of the
NZX50 Gross Index. We do so to capture ‘speculative’ influences of investor sentiment and
risk aversion.
As mentioned previously, we also investigate whether a liquidity premia exists in these
futures contracts. In line with this, the final addition to the model is the Illiquidity Ratio
outlined in the previous section. This final model is defined as:
(8)
As will be evident in the results, we also examine other models with slight deviations from
those stated here. For the sake of brevity we have only defined those which form the basis
of any alterations.
5. Results
5.1 Liquidity Results
Figure 1 displays both daily trading volume and illiquidity ratios for all three contract forms.
We report the average between the Benmore and Otahuhu nodes. Volumes in the first year
13
of trading highlight the illiquidity concerns discussed in Section 1, since only a handful of
contracts traded each day.
[INSERT FIGURE 1 ABOUT HERE]
As outlined in section 5.1, we carry out structural break tests (Bai and Perron, 2003)
to determine the existence of any significant changes. The results of these are presented
empirically in Table 2 and graphically in Figure 2. Had the interventions achieved their desired
outcome, we would expect to find breaks around January 2010 (mandatory market making)
and October 2011 (reduction of bid-ask spread). What we actually observe is no statistically
significant change around the introduction of mandatory market making, implying that this
policy intervention failed to achieve its objective of enhanced liquidity. This may suggest that
the market makers simply posted quotes with the maximum spread allowed by the exchange
(10%), carrying out their duty yet making no contribution to improving market liquidity.
However, it appears the reduction of the maximum bid-ask spread from 10% to 5%
did have an impact. We see that front end volumes increased in October 2011 and January
2012 for Benmore and Otahuhu, respectively. Front end trading at both nodes also
experienced a decrease in the Illiquidity Ratio in October 2011. These results suggest that
imposition of a smaller spread appears to have achieved its desired outcome. However this
only occurred in the front end contracts. While there were some changes further out, this
likely reflected the decreased sample size for the Illiquidity Ratio for these contracts. One
would expect the front end contracts to be the most liquid, however anecdotal evidence
suggests that there is hedging activity taking place one and two years ahead, so the lower
liquidity of these contracts should still be of concern.
[INSERT TABLE 2 AND FIGURE 2 ABOUT HERE]
5.2 Risk Premia Results
Table 3 displays the average risk premia (eq. 4) for both nodes, across all contract
forms and by quarter. As expected, there is evidence of significant premia at both nodes, for
all contract forms. Though the overall premia for front end contracts is insignificant at both
Benmore and Otahuhu, when broken down by quarter there is quite clearly significant
premia. This premia generally increases further out the curve i.e. premia are greater the
longer the time to maturity.
[INSERT TABLE 3 ABOUT HERE]
14
The premia for all six contracts examined exhibit a clear time variant nature. This is
illustrated by figure 3. We see that the premia are consistently higher in Q2 and Q3 contracts.
Aside from the front end Benmore contract, we see positive premia which is significant at the
1% level. By contrast, Q1 and Q4 contracts, in general, display significantly negative premia.
[INSERT FIGURE 3 ABOUT HERE]
These findings are consistent with those of Lucia and Torró (2011) for Nord Pool; they
found that the premia is dependent on the season in which the contract matures. Q2 and Q3
represent the driest months of the year, when demand is at a peak while hydro storage/ and
inflows are lowest. During these times the potential for price spikes is at its highest, so those
market participants with downside risk from price increases, i.e. those who purchase
electricity, have a greater incentive to hedge this exposure. In order to reduce their exposure
to large and frequent price spikes during a dry period, they may be willing to pay a greater
price than would be realised were they not hedged. On the other hand, during Q1 and Q4 the
opposite is the case. There is a lull in demand, while hydro generators have excess capacity
to meet demand and storage lakes are either filling (Q4) or at or near their annual peak
storage (Q1). In this scenario it is likely those with downside risk from low prices, i.e. those
who sell electricity, will have greater incentive to hedge and may accept a lower price than
would be realised in the spot market.
These alternative arguments may be related to the balance between the relative
hedging of generators and retailers, as found in Fleten et al. (2015). During Q2 and Q3 it may
be those retailers and large consumers who utilise these futures contracts, while Q1 and Q4
sees hedging activity by generators taking place. There is potential for an investigation into
this using participant data from the EA, however this is beyond the scope of this paper.
As outlined in Section 4.3, next we seek to explain this premia through a variety of
models. Table 4 shows the results for our “basic” model (eq. 5). These results exhibit several
surprising elements. The first of these is that we find little support for the B&L model. Though
the variance and skewness coefficients for front end Benmore contracts are what is predicted
by the B&L model, they are only significant at the 10% and 5% level, respectively.
Furthermore, the variance coefficient changes sign and is significant at the 1% level at the
Otahuhu node. This change in sign could potentially be explained by the fact that volatility
incentivises both producers and consumers to hedge their exposure. As demand for futures
contracts is being spurred from both sides with “competing premia” (producers will accept
15
lower prices, whilst consumers will accept higher prices), this may reflect an almost zero-sum
game i.e. a close to zero premia. Were this to be the case, it may explain the lack of agreement
with regards to the sign of variance coefficients in previous empirical research. The
explanatory power of skewness and variance decrease as you move further out the curve,
which is to be expected if market participants behave efficiently.
The second surprising result is the significance of the current level of the spot price,
as far as two years ahead. Fortunately, all the coefficients are of the same sign in this case.
This result will be discussed in depth later. We also find significant lagged dependent
variables, supporting the findings of Lucia and Torró (2011). As expected due to the seasonal
nature of premia, we also find the quarterly dummies to be significant. These reported results
for the statistical risk measures are robust when we drop the spot price as a variable.
4
[INSERT TABLE 4 ABOUT HERE]
We then include the physical market factors of demand, inflows and storage. The
results of this are reported in Table 5 (Panel A). We again find little support for the B&L
hypothesised relationships. Only two coefficients are significant at the 1% level, and are of
the opposite sign to that expected. One of these occurs in the two years ahead contract at
the Benmore node.
In terms of the physical market factors, we find significant coefficients for both front
end contracts for the hydrological variables. All four are significant at the 1% level.
Significantly negative coefficients for the inflow variables imply that when inflows are lower
than expected, realised premia increases. This is related to the potential for price spikes:
when there are below average inflows, the ability for generators to meet demand decreases
and there is greater potential for price spikes due to the reliance on hydro generation in the
NZEM. Thus, the incentive to hedge increases.
One would also expect this to be the same for the storage variable. In fact, both are
significantly positive. These conflicting results suggest that when lake levels are lower than
expected, premia decrease, however if the water which flows into these lakes is lower than
expected, premia increase. As reported in Table 5 (Panel B), the significance of the physical
market variables disappears when the spot price is not included in the model. This suggests
4
These additional results are available upon request
16
that there are potential issues around endogeneity between these variables. In terms of
demand, it appears to have no explanatory power in this model.
[INSERT TABLE 5 ABOUT HERE]
As is expected, these physical market variables have no explanatory power one and
two years ahead. The spot price again has a surprising amount of explanatory power up to
one year ahead. This is puzzling considering the minimal correlation between spot price levels
one year ahead. Figure 4 illustrates this further as it shows the correlation between spot
prices now and in future periods, as well as between future’s price changes now and in the
future. We see that the spot price correlation dissipates beyond two quarters. That is, the
correlation between the spot price now and only three quarters ahead is essentially zero.
However, it appears that market participants are pricing the current spot price into future’s
prices far more than would be expected. For example, inflows into hydro lakes are highly
volatile, so if inflows are low and prices are high now, this could all change only weeks or
months into the future. The future’s price correlations show that a change in the future’s
price for front end contracts is met with a similar price movement for contracts as far as two
years ahead. That is, if front end futures prices increase in response to current market factors,
two years ahead contracts price experience upwards movements as well. This, coupled with
the explanatory power of the spot price in our regressions, suggests inefficient behaviour by
market participants.
[INSERT FIGURE 4 ABOUT HERE]
[INSERT TABLE 6 ABOUT HERE]
It appears that the underlying costs of production, which Redl and Bunn (2013)
suggested should explain premia, have no explanatory power in a New Zealand context. Table
6 presents the results from equation 7. We see that the significant explanatory power in the
front end for our physical market factors remain robust, however the coefficients for oil and
carbon are generally insignificant. We do observe a significantly positive coefficient for
Benmore two years ahead contracts, however as such significance is not mirrored in the
Otahuhu contract a robust interpretation of this is difficult. There is also no evidence of
speculative investor behaviour in these markets, as the stock coefficient is largely
insignificant. This suggests that there is no trade-off between potential returns in the stock
market or electricity futures market being made by NZ investors.
17
Our final model considers the effect of liquidity. We adapt equation 8, dropping the
costs of production variables due to their limited explanatory power. The results of this are
presented in Table 7 (Panel A). Due to the previously mentioned issues around the calculation
of the Illiquidity Ratio, the statistical power of these regressions is limited. Though we find
some significant illiquidity coefficients, these come from regressions of sample sizes of 83, 37
and 33, respectively.
In an attempt to overcome this, we replace the Illiquidity Ratio with trading volume
which provides far greater sample sizes. The results of this are reported in Table 7 (Panel B).
We find some evidence of trading volume having explanatory power with regards to premia.
Though the coefficient for front end Benmore contracts is significant at the 5% level, it is the
opposite sign to what would be expected. A positive coefficient suggests that greater trading
volume, and thus liquidity, increases premia. This may seem counterintuitive but may reflect
the fact that high volume at points may reflect high volatility (See Section 3.1). This volume-
volatility effect is likely to be present in the front-end futures since these should more closely
reflect underlying spot conditions (from which the volatility is derived). Consistent with this
interpretation, we do see significant negative coefficients two years ahead, however only one
of these is significant beyond the 10% level.
[INSERT TABLE 7 ABOUT HERE]
6. Conclusions
In this paper we have documented increasing (decreasing) liquidity (illiquidity) for contracts
of all maturities, a result which is encouraging for a relatively new market. The results of our
structural breaks tests demonstrate that the imposition of mandatory market making did not
improve liquidity, yet the reduction in the maximum bid-ask spread did, albeit at the front
end only.
We also document significant premia, both positive and negative, in all three contract
maturities. Premia are consistently higher during the winter quarters, yet significantly
negative during the summer quarters. Regression estimations show limited support for the
seminal work of Bessembinder and Lemmon (2002) and suggest that physical market factors
play a role in driving front end premia. Most surprising is the significant explanatory power of
the current spot price, leading to our suggestion of inefficient behaviour of market
participants.
18
In terms of the initial goal of this paper of determining whether a liquidity premium
exists in these futures contracts, the evidence is mixed. While there are several significant
coefficients, we have signs opposite to that which would be expected, as well as concerns
around statistical power of our regressions.
The findings of this paper may have several potentially important implications for both
market participants and regulators. For regulators, this may provide support for a future
reduction of the maximum bid-ask spread, however any potential liquidity increase must be
traded off against the potential additional risk imposed on market makers. Once should also
be wary of the fact that the relationship between bid-ask spreads and liquidity is potentially
non-linear. Should further reductions in the maximum allowable spread be made in the
future, the returns may be diminishing. The documented inefficient behaviour may also be
attractive for speculative investors who wish to take advantage of such anomalous price
movements. However, due to the relative illiquidity of this market, exiting positions and
realising any profits may prove difficult.
19
References
Amihud, Y. (2002). Illiquidity and stock returns: cross-section and time-series effects. Journal
of Financial Markets, 5(1), 31-56.
Amihud, Y., & Mendelson, H. (1986a). Asset pricing and the bid-ask spread. Journal of
Financial Economics, 17(2), 223-249.
Amihud, Y., & Mendelson, H. (1986b). Liquidity and Stock Returns. Financial Analysts
Journal, 42(3), 43-48.
Bessembinder, H., & Lemmon, M. L. (2002). Equilibrium Pricing and Optimal Hedging in
Electricity Forward Markets. Journal of Finance, 57(3), 1347.
Black, F. (1971). Towards a Fully Automated Stock Exchange. Financial Analysts Journal,
27(6), 24-87.
Botterud, A., Kristiansen, T., & Ilic, M. D. (2010). The relationship between spot and futures
prices in the Nord Pool electricity market. Energy Economics, 32(5), 967-978.
Brennan, M. J., Chordia, T., & Subrahmanyam, A. (1998). Alternative factor specifications,
security characteristics, and the cross-section of expected stock returns. Journal of
Financial Economics, 49(3), 345-373.
Brennan, M. J., & Subrahmanyam, A. (1995). Investment analysis and price formation in
securities markets. Journal of Financial Economics, 38(3), 361-381.
Brunnermeier, M. K., & Pedersen, L. H. (2009). Market Liquidity and Funding Liquidity.
Review of Financial Studies, 22(6), 2201-2238.
Bunn, D. W., & Chen, D. (2013). The forward premium in electricity futures. Journal of
Empirical Finance, 23, 173-186.
Corwin, S. A., & Schultz, P. (2012). A Simple Way to Estimate Bid-Ask Spreads from Daily
High and Low Prices. Journal of Finance, 67(2), 719-760.
Daskalakis, G., Symeonidis, L., & Markellos, R. N. (2015). Electricity Futures Prices in an
Emissions Constrained Economy: Evidence from European Power Markets. Energy
Journal, 36(3), 1-33.
Demsetz, H. (1968). The cost of transacting. The Quarterly Journal of Economics, 33-53.
Diaz-Rainey, I & Tulloch, D (2016) ‘Carbon pricing in New Zealand’s Emissions Trading
Scheme’, USAEE/IAEE Working Paper No. 16-255, Available at SSRN
Douglas, S., & Popova, J. (2008). Storage and the electricity forward premium. Energy
Economics, 30(4), 1712-1727.
20
EA (2014). Hedge Market Development: A WAG Discussion Paper. Electricity Authority,
Wellington, New Zealand. Available from https://www.ea.govt.nzEL (2014).
Assessment of Deltas in Futures Prices’, Report prepared by Energy Link for the
Electricity Authority, in Appendices to (Appendix K) to Hedge Market Development: A
WAG Discussion Paper. Electricity Authority, Wellington, New Zealand. Available
from https://www.ea.govt.nz
Fleming, M. J. (2003). Measuring Treasury Market Liquidity. Working Paper. Available from
http://papers.ssrn.com
Fleten, S.-E., Hagen, L. A., Nygård, M. T., Smith-Sivertsen, R., & Sollie, J. M. (2015). The
overnight risk premium in electricity forward contracts. Energy Economics, 49, 293-
300.
Frestad, D. (2012). Liquidity and dirty hedging in the Nordic electricity market. Energy
Economics, 34(5), 1341-1355.
Furió, D., & Meneu, V. (2010). Expectations and forward risk premium in the Spanish
deregulated power market. Energy Policy, 38(2).
Glosten, L. R., & Harris, L. E. (1988). Estimating the Components of the Bid/Ask Spread.
Journal of Financial Economics, 21(1), 123-142.
Hagemann, S., & Weber, C. (2013). An Empirical Analysis of Liquidity and its Determinants in
the German Intraday Market for Electricity. Working Paper. Available from
http://papers.ssrn.com
Joskow, P.L. (2006). ‘Introduction to Electricity Sector Liberalization: Lessons Learned from
Cross-Country Studies’, in Sioshansi, F.P. and Pfaffenberger, W. (eds.), Electricity
Market Reform: An International Perspective, Oxford: Elsevier.
Karakatsani, N. V., & Bunn, D. W. (2005). Diurnal Reversals of Electricity Forward Premia.
Energy Markets Group Working Paper, London Business School. Available from
https://www.london.edu
Kolos, S. P., & Ronn, E. I. (2008). Estimating the commodity market price of risk for energy
prices. Energy Economics, 30(2), 621-641.
Kyle, A. S. (1985). Continuous Auctions and Insider Trading. Econometrica, 53(6), 1315-1335.
Lesmond, D. A. (2005). Liquidity of emerging markets. Journal of Financial Economics, 77(2),
411-452.
21
Lesmond, D. A., Ogden, J. P., & Trzcinka, C. A. (1999). A new estimate of transaction costs.
Review of Financial Studies, 12(5), 11131141.
Longstaff, F. A., & Wang, A. W. (2004). Electricity Forward Prices: A High-Frequency
Empirical Analysis. Journal of Finance, 59(4), 1877-1900.
Lucia, J. J., & Torró, H. (2011). On the risk premium in Nordic electricity futures prices.
International Review of Economics & Finance, 20(4), 750-763.
Marshall, B. R., Nguyen, N. H., & Visaltanachoti, N. (2012). Commodity Liquidity
Measurement and Transaction Costs. Review of Financial Studies, 25(2), 599-638.
MBIE. (2015). Electricity graph and data tables [Data file, electricity generation by source].
Retrieved May 12, 2016, from http://www.mbie.govt.nz/info-services/sectors-
industries/energy/energy-data-modelling/statistics/electricity
Redl, C., & Bunn, D. (2013). Determinants of the premium in forward contracts. Journal of
Regulatory Economics, 43(1), 90-111.
Roll, R. (1984). A Simple Implicit Measure of the Effective Bid-Ask Spread in an Efficient
Market. Journal of Finance, 39(4), 1127-1139.
Shawky, H. A., Marathe, A., & Barrett, C. L. (2003). A First Look at the Empirical Relationship
Between Spot and Futures Electricity Prices in the United States, Journal of Futures
Markets, 23(10), 931-955.
Viehmann, J. (2011). Risk premiums in the German day-ahead Electricity Market. Energy
Policy, 39(1), 386-394.
Weron, R., & Zator, M. (2014). Revisiting the relationship between spot and futures prices in
the Nord Pool electricity market. Energy Economics, 44, 178-190.
22
Figure 1. Evolution of liquidity
0
5
10
15
20
25
30
35
2/10/2009 2/10/2010 2/10/2011 2/10/2012 2/10/2013 2/10/2014 2/10/2015
Trading Volume
Front End Volume Average
0
5
10
15
20
25
30
35
2/10/2009 2/10/2010 2/10/2011 2/10/2012 2/10/2013 2/10/2014 2/10/2015
Trading Volume
1 Year Ahead Volume Average
0
5
10
15
20
25
30
35
2/10/2009 2/10/2010 2/10/2011 2/10/2012 2/10/2013 2/10/2014 2/10/2015
Trading Volume
2 Years Ahead Volume Average
23
Figure 1. Evolution of liquidity (continued)
0
0.05
0.1
0.15
0.2
0.25
2/10/2009 2/10/2010 2/10/2011 2/10/2012 2/10/2013 2/10/2014 2/10/2015
Illiquidity Ratio
Front End Illiquidity Average
0
0.05
0.1
0.15
0.2
0.25
2/10/2009 2/10/2010 2/10/2011 2/10/2012 2/10/2013 2/10/2014 2/10/2015
Illiquidity Ratio
1 Year Ahead Illiquidity Average
0
0.05
0.1
0.15
0.2
0.25
2/10/2009 2/10/2010 2/10/2011 2/10/2012 2/10/2013 2/10/2014 2/10/2015
Illiquidity Ratio
2 Years Ahead Illiquidity Average
24
Figure 2. Structural Breaks in Liquidity Measures (selected: front end only)
Front End Trading Volume (Benmore) Front End Illiquidity (Benmore)
Front End Trading Volume (Otahuhu) Front End Illiquidity (Otahuhu)
-20
0
20
40
60 0
10
20
30
40
50
60
2009 2010 2011 2012 2013 2014 2015
Residual Actual Fitted
-.1
.0
.1
.2
.3
.4 .0
.1
.2
.3
.4
2009 2011 2012 2013 2014 2015
Residual Actual Fitted
-5
0
5
10
15
20
25 0
4
8
12
16
20
24
28
2009 2010 2011 2012 2013 2014 2015
Residual Actual Fitted
-.05
.00
.05
.10
.15
.20 .00
.05
.10
.15
.20
.25
2010 2011 2012 2013 2014 2015
Residual Actual Fitted
25
Figure 3. Evolution of Risk Premia
-40.00
-30.00
-20.00
-10.00
0.00
10.00
20.00
30.00
40.00
50.00
60.00
Q1 Q2 Q3 Q4
Realised Risk Premia
Benmore
Front End 1 Year Ahead 2 Years Ahead
-30.00
-20.00
-10.00
0.00
10.00
20.00
30.00
40.00
Q1 Q2 Q3 Q4
Realised Risk Premia
Otahuhu
Front End 1 Year Ahead 2 Years Ahead
26
Figure 4. Correlations in Movements in BEN Futures Prices
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1 qtr ahead 2 qtr ahead 3 qtr ahead 4 qtr ahead 5 qtr ahead 6 qtr ahead 7 qtr ahead 8 qtr ahead
Q1 Q2 Q3 Q4 All Quarters BEN Spot
27
Table 1. Definition of Variables and Descriptive Statistics
Variable
Description
Source
Mean
Std. Dev.
Minimum
Maximum
RP (Risk
Premia)
Front End (Benmore)
Difference between the current
futures price and the average
wholesale spot price over the
settlement quarter
EL
-0.606
19.322
-52.233
71.167
Front End (Otahuhu)
0.099
14.466
-40.245
56.413
1 Year Ahead (Benmore)
13.151
33.846
-54.244
72.136
1 Year Ahead (Otahuhu)
6.903
23.766
-43.245
58.917
2 Years Ahead (Benmore)
13.180
37.451
-51.294
71.636
2 Years Ahead (Otahuhu)
7.560
24.379
-30.745
39.203
Trading
Volume
Front End (Benmore)
Daily trading volume in the
number of contracts
EL
1.45
3.33
0.00
55.00
Front End (Otahuhu)
1.26
3.09
0.00
26.00
1 Year Ahead (Benmore)
0.60
2.04
0.00
26.00
1 Year Ahead (Otahuhu)
0.51
1.78
0.00
20.00
2 Years Ahead (Benmore)
0.32
1.36
0.00
16.00
2 Years Ahead (Otahuhu)
0.31
1.34
0.00
13.00
Illiquidity
Front End (Benmore)
Amihud’s Illiquidity measure
EL
0.02
0.03
0.00
0.32
Front End (Otahuhu)
0.01
0.02
0.00
0.21
1 Year Ahead (Benmore)
0.01
0.01
0.00
0.06
1 Year Ahead (Otahuhu)
0.01
0.01
0.00
0.11
2 Years Ahead (Benmore)
0.01
0.01
0.00
0.06
2 Years Ahead (Otahuhu)
0.01
0.01
0.00
0.06
Open
Interest
Front End (Benmore)
Number of contracts outstanding
in the market
EL
70.70
50.90
0.00
171.00
Front End (Otahuhu)
41.21
32.37
1.00
116.00
1 Year Ahead (Benmore)
21.31
16.57
1.00
62.00
1 Year Ahead (Otahuhu)
86.00
68.63
0.00
234.00
2 Years Ahead (Benmore)
37.03
34.49
0.00
112.00
2 Years Ahead (Otahuhu)
19.06
15.33
0.00
53.00
Skew(7)
Benmore
Skewness of wholesale spot price
over the past seven days
EL
0.19
0.85
-2.49
2.64
Otahuhu
0.32
0.92
-2.38
2.65
Skew(30)
Benmore
Skewness of wholesale spot price
over the past thirty days
EL
0.56
1.02
-2.55
5.12
Otahuhu
0.98
1.27
-1.34
5.12
Var(7)
Benmore
Variance of the wholesale spot
price over the past seven days
EL
334.81
1060.83
2.13
13426.56
Otahuhu
1165.15
9257.73
0.92
158056.52
Var(30)
Benmore
Variance of the wholesale spot
price over the past thirty days
EL
634.87
1188.74
18.79
7526.74
Otahuhu
1390.22
5372.01
19.62
50725.58
S(Quarter)
Benmore
Average wholesale spot price
during the current quarter
EL
61.44
29.09
0.94
163.55
Otahuhu
67.53
18.76
7.99
136.52
S(30)
Benmore
Average wholesale spot price over
the past thirty days
EL
64.22
35.00
9.13
200.28
Otahuhu
70.56
23.42
11.70
147.66
Demand
Demand(1)
Deviation from historical average
demand over the past week
EL
0.00
18.86
-56.14
54.92
Demand(4)
Deviations from historical average
demand over the past four weeks
EL
0.00
14.55
-42.86
32.37
Key: EL = Energy Link Ltd; Bl =Bloomberg
28
Table 1. Definition of Variables and Descriptive Statistics (continued)
Variable
Description
Source
Mean
Std. Dev.
Minimum
Maximum
Inflow
Inflow(1)
Deviation from historical average
hydrological inflows over the past
week
EL
2.01
177.21
-434.64
773.52
Inflow(4)
Deviations from historical average
hydrological inflows over the past
four weeks
EL
0.00
120.50
-309.62
389.47
Storage
Storage(1)
Deviation from historical average
hydrological storage over the past
week
EL
6.98
501.82
-1130.12
1224.60
Storage(4)
Deviations from historical average
hydrological storage over the past
four weeks
EL
0.00
482.92
-1113.25
1023.05
Carbon
Carbon (1)
Logarithmic return of NZUs over
the past week
Bl
0.00
0.07
-0.41
0.50
Carbon(4)
Logarithmic return of NZUs over
the past four weeks
Bl
-0.01
0.14
-0.49
0.76
Stock
Stock(1)
Logarithmic return of the NZX50
Gross Index over the past week
Bl
0.00
0.01
-0.06
0.05
Stock(4)
Logarithmic return of NZX50
Gross Index over the past four
weeks
Bl
0.01
0.03
-0.11
0.11
Oil
Oil(1)
Logarithmic return of Dubai Fateh
oil over the past week
Bl
0.00
0.04
-0.15
0.19
Oil(4)
Logarithmic return of Dubai Fateh
oil over the past four weeks
Bl
0.00
0.09
-0.31
0.29
Key: EL = Energy Link Ltd; Bl =Bloomberg
29
Table 2. Structural Breaks in Liquidity Measures
Maturity
Variable
Breaks
Variable
Breaks
Front End Benmore
Volume
17/10/2011 & 20/11/2014
Illiquidity
4/10/2011
Front End Otahuhu
Volume
5/01/2012
Illiquidity
25/10/2011
1 Year Ahead Benmore
Volume
20/01/2012
Illiquidity
1 Year Ahead Otahuhu
Volume
7/09/2012
Illiquidity
2 Years Ahead Benmore
Volume
5/09/2012
Illiquidity
3/07/2012
2 Years Ahead Otahuhu
Volume
5/09/2012
Illiquidity
30
Table 3. Average Risk Premia 2/10/2009-31/12/2015
Contract Form
Front End
One Year Ahead
Two Years Ahead
Node
Benmore
Otahuhu
Benmore
Otahuhu
Benmore
Otahuhu
Overall
-0.61
0.10
13.15
***
6.90
***
13.18
***
7.56
***
Q1
-9.56
***
-6.63
***
-16.59
***
-18.66
***
-30.43
***
-23.20
***
Q2
12.34
***
12.73
***
26.46
***
27.72
***
25.90
***
23.71
***
Q3
1.81
2.00
***
42.04
***
26.12
***
54.93
***
34.64
***
Q4
-6.29
***
-6.72
***
1.39
-6.12
***
2.54
*
-3.69
***
*** represents statistical significance at the 1% level
** represents significance at the 5% level
* represents significance at the 10% level
31
Table 4. Time Series Regression Results for Equation 5
Contract Form
Front End
One Year Ahead
Two Years Ahead
Node
Benmore
Otahuhu
Benmore
Otahuhu
Benmore
Otahuhu
Constant
-1.939
***
-3.004
***
-2.314
***
-3.249
***
-3.843
***
-5.211
***
Skewness
0.324
***
0.091
-0.013
0.020
0.344
-0.016
Variance
0.000
0.000
0.000
**
0.000
0.000
0.000
Spot Price
0.024
***
0.035
***
0.024
***
0.020
***
0.012
*
0.016
**
Lagged RP
1.041
***
0.974
***
0.957
***
0.917
***
0.928
***
0.847
***
Lagged RP 2
-0.094
***
-0.043
Q2
0.708
1.088
**
2.333
***
4.698
***
5.299
***
7.997
***
Q3
0.660
*
0.854
**
3.076
***
4.104
***
7.308
***
9.527
***
Q4
0.523
0.423
0.896
**
1.142
***
2.810
***
3.267
***
Durbin Watson
1.99
1.934
1.938
1.867
1.861
1.737
Chi-square heteroscedasticity
0.0777
0.0683
0.8939
0.499
0.892
<0.0001
Adjusted R^2
0.9429
0.9416
0.9757
0.9725
0.977
0.9742
n
1589
1589
1336
1336
1081
1081
Notes:
*** represents statistical significance at the 1% level, ** represents significance at the 5% level, * represents significance at the 10% level
Front end use Skew(7), Var(7) and S(Q), while the others use Skew(30), Var(30) and S(30).
Regressions run on daily data
32
Table 5 (Panel A): Regression results for Equation 6
Contract Form
Front End
One Year Ahead
Two Years Ahead
Node
Benmore
Otahuhu
Benmore
Otahuhu
Benmore
Otahuhu
Constant
-13.962
***
-17.502
***
-12.184
***
-15.423
***
-12.939
***
-13.478
***
Skewness
0.612
0.155
-0.459
-0.163
2.573
***
0.601
Variance
0.000
0.000
***
-0.002
*
0.000
0.000
0.000
*
Spot Price
0.185
***
0.208
***
0.145
***
0.116
***
-0.001
-0.007
Demand Deviations
0.017
0.023
-0.099
-0.022
0.022
0.079
**
Inflow Deviations
-0.013
***
-0.008
***
-0.003
0.000
-0.010
0.004
Storage Deviations
0.006
***
0.004
**
0.003
0.003
**
-0.003
-0.005
**
Lagged RP
0.730
***
0.776
***
0.808
***
0.649
***
0.689
***
0.489
***
Lagged RP 2
-0.122
0.013
0.051
0.013
Q2
2.857
6.316
**
9.113
***
18.040
***
21.557
***
26.605
***
Q3
3.759
**
4.826
**
13.565
***
16.138
***
28.914
***
30.366
***
Q4
3.716
**
2.888
*
3.902
*
4.900
***
11.148
***
10.782
***
Durbin Watson
1.788
1.828
1.808
1.541
1.492
1.236
Chi-square heteroscedasticity
0.8967
0.0674
0.2891
0.032
0.0652
>0.0001
Adjusted R^2
0.7039
0.7533
0.8929
0.8947
0.9115
0.923
n
325
324
272
272
218
219
Notes:
*** represents statistical significance at the 1% level, ** represents significance at the 5% level, * represents significance at the 10% level
Front end use Skew(7), Var(7) and S(Q), while the others use Skew(30), Var(30) and S(30)
Front end use Demand(1), Inflow(1) and Storage(1), while the others use Demand(4), Inflow(4) and Storage(4)
Regressions run on weekly data
33
Table 5 (Panel B): Regression results for Equation 6 (Excluding price)
Contract Form
Front End
One Year Ahead
Two Years Ahead
Node
Benmore
Otahuhu
Benmore
Otahuhu
Benmore
Otahuhu
Constant
-2.902
**
-2.572
***
-3.185
-6.377
***
-12.858
***
-13.971
***
Skewness
0.429
0.112
-0.709
-0.164
2.543
**
0.605
Variance
0.001
0.000
***
0.000
0.000
0.000
0.000
**
Demand Deviations
0.061
0.052
*
-0.006
0.034
-0.001
0.075
**
Inflow Deviations
-0.005
-0.004
0.002
0.003
0.023
0.003
Storage Deviations
-0.002
-0.002
-0.004
**
-0.001
-0.009
**
-0.004
***
Lagged RP
0.789
***
0.762
***
0.831
***
0.694
***
-0.003
***
0.492
***
Lagged RP 2
0.703
Q2
4.852
**
5.041
***
9.738
***
17.289
***
21.479
***
26.546
***
Q3
2.650
2.629
*
11.143
***
14.434
***
28.815
***
30.361
***
Q4
1.748
0.991
1.836
3.317
**
11.052
***
10.811
***
Durbin Watson
1.844
1.732
1.791
1.566
1.504
1.24
Chi-square heteroscedasticity
0.9807
0.6398
0.0186
0.0037
0.0433
<0.0001
Adjusted R^2
0.6866
0.7214
0.8888
0.8889
0.9125
0.9265
n
325
325
273
273
219
219
Notes:
*** represents statistical significance at the 1% level, ** represents significance at the 5% level, * represents significance at the 10% level
Front end use Skew(7) and Var(7), while the others use Skew(30) and Var(30).
Front end use Demand(1), Inflow(1) and Storage(1), while the others use Demand(4), Inflow(4) and Storage(4)
Regressions run on weekly data
34
Table 6: Regression Results for Equation 7
Contract Form
Front End
One Year Ahead
Two Years Ahead
Node
Benmore
Otahuhu
Benmore
Otahuhu
Benmore
Otahuhu
Constant
-14.756
***
-17.768
***
-13.111
***
-16.414
***
-15.848
***
-13.989
***
Skewness
0.574
0.024
-0.314
-0.059
2.947
***
0.616
Variance
0.000
0.000
***
-0.002
**
0.000
0.001
0.000
*
Spot Price
0.191
***
0.210
***
0.157
***
0.123
***
0.016
0.000
Demand Deviations
0.008
0.025
-0.101
-0.021
-0.017
0.072
*
Inflow Deviations
-0.012
***
-0.008
***
-0.002
0.001
-0.013
0.005
Storage Deviations
0.007
***
0.004
***
0.003
0.003
*
-0.002
-0.005
**
Oil
13.757
15.609
10.837
9.465
39.429
***
15.633
*
Carbon
5.362
5.069
-7.398
-7.162
**
-8.633
-1.452
Stock
62.764
16.274
-33.221
-18.594
-5.635
-33.348
*
Lagged RP
0.776
***
0.765
***
0.810
***
0.686
***
0.703
***
0.494
***
Lagged RP 2
-0.073
-0.117
**
Q2
3.817
*
6.652
***
9.596
***
18.868
***
23.336
***
26.849
***
Q3
4.222
**
4.994
***
14.565
***
17.136
***
30.054
***
30.332
***
Q4
4.076
**
3.111
**
4.529
**
5.501
***
11.598
***
10.742
***
Durbin Watson
1.899
1.819
1.812
1.621
1.523
1.236
Chi-square heteroscedasticity
0.9998
0.4648
0.3083
0.4394
0.3363
0.1117
Adjusted R^2
0.7032
0.7535
0.8942
0.8965
0.9145
0.9236
n
324
324
273
273
219
219
Notes:
*** represents statistical significance at the 1% level, ** represents significance at the 5% level, * represents significance at the 10% level
Front end use Skew(7), Var(7) and S(Q), while the others use Skew(30), Var(30) and S(30).
Front end use Demand(1), Inflow(1) and Storage(1), while the others use Demand(4), Inflow(4) and Storage(4)
Front end use Oil(1), Carbon(1) and Stock (1), while others use Oil(4), Carbon(4) and Stock(4).
35
Table 7 (Panel A): Regression results for equation 8 (Illiquidity measure)
Contract Form
Front End
One Year Ahead
Two Years Ahead
Node
Benmore
Otahuhu
Benmore
Otahuhu
Benmore
Otahuhu
Constant
-5.932
-5.067
-13.764
***
-11.666
**
-27.015
***
-17.817
***
Skewness
0.934
0.630
-0.752
-0.298
1.616
2.916
***
Variance
0.000
0.000
***
0.002
0.000
0.002
0.000
Spot Price
0.096
*
0.060
0.053
0.059
0.035
-0.079
Demand
0.014
0.012
-0.184
*
-0.115
0.032
0.119
Inflows
-0.012
**
-0.009
**
0.014
0.003
-0.002
-0.001
Storage
0.000
0.001
-0.008
*
-0.001
-0.005
-0.002
Illiquidity
6.486
6.679
-167.382
296.169
***
282.359
**
99.607
*
Lagged RP
0.736
***
0.732
***
0.566
***
0.778
***
0.348
***
0.198
***
Lagged RP 2
Q2
-0.245
3.286
**
21.782
***
11.862
***
47.413
***
42.774
***
Q3
0.568
0.346
27.872
***
11.201
**
51.618
***
47.070
***
Q4
-0.264
-0.650
13.389
***
4.605
19.347
***
8.033
***
Durbin Watson
1.809
1.884
1.699
2.015
1.965
1.673
Chi-square heteroscedasticity
0.9609
0.9677
0.5334
0.3778
0.8485
0.5267
Adjusted R^2
0.7099
37873
0.9085
0.9174
0.9232
0.9893
n
197
191
96
83
37
33
Notes:
*** represents statistical significance at the 1% level, ** represents significance at the 5% level, * represents significance at the 10% level
Front end use Skew(7), Var(7) and S(Q), while the others use Skew(30), Var(30) and S(30).
Front end use Demand(1), Inflow(1) and Storage(1), while the others use Demand(4), Inflow(4) and Storage(4)Regressions run on weekly data
36
Table 7 (Panel B): Regression results for Equation 8 (Volume)
Contract Form
Front End
One Year Ahead
Two Years Ahead
Node
Benmore
Otahuhu
Benmore
Otahuhu
Benmore
Otahuhu
Constant
-14.895
***
-17.716
***
-12.615
***
-14.988
***
-12.466
***
-13.074
***
Skewness
0.545
0.143
-0.542
-0.166
2.554
***
0.526
Variance
0.000
0.000
***
-0.001
0.000
0.000
0.000
**
Spot Price
0.181
***
0.208
***
0.143
***
0.112
***
-0.008
-0.009
Demand
0.030
0.025
-0.090
-0.021
0.016
0.069
*
Inflows
-0.013
***
-0.008
***
-0.004
0.000
-0.008
0.005
Storage
0.007
***
0.004
**
0.004
0.003
**
-0.004
-0.005
**
Volume
0.141
**
0.027
0.141
-0.054
-0.270
*
-0.393
***
Lagged RP
0.725
***
0.775
***
0.821
***
0.695
***
0.700
***
0.481
***
Lagged RP 2
-0.121
*
Q2
3.279
*
6.319
***
9.356
***
17.954
***
21.742
***
26.706
***
Q3
3.874
**
4.831
***
13.831
***
16.211
***
29.214
***
30.761
***
Q4
4.021
**
2.939
*
4.247
4.982
***
11.331
***
10.917
***
Durbin Watson
1.801
1.828
1.823
1.63
1.524
1.27
Heteroscedasticity (p)
0.8952
0.0873
0.1226
<0.0001
0.0955
<0.0001
Adjusted R^2
0.7072
0.7528
0.8935
0.8944
0.9122
0.9238
n
325
324
273
273
219
219
Notes
*** represents statistical significance at the 1% level, ** represents significance at the 5% level, * represents significance at the 10% level
Front end use Skew(7), Var(7) and S(Q), while the others use Skew(30), Var(30) and S(30).
Front end use Demand(1), Inflow(1) and Storage(1), while the others use Demand(4), Inflow(4) and Storage(4)
Regressions run on weekly data.
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
We investigate the economic factors that drive electricity risk premia in the European emissions constrained economy. Our analysis is undertaken for monthly baseload electricity futures for delivery in the Nordic, French and British power markets. We find that electricity risk premia are significantly related to the volatility of electricity spot prices, demand and revenues, and the price volatility of the carbon dioxide (CO 2) futures traded under the EU Emissions Trading Scheme (EU ETS). This finding has significant implications for the pricing of electricity futures since it highlights for the first time the role of carbon market uncertainties as a main determinant of the relationship between spot and futures electricity prices in Europe. Our results also suggest that for the electricity markets under scrutiny futures prices are determined rationally by risk-averse economic agents. ___________________________ We would like to thank Nicolas Koch, Chris Brooks, Yiannis Karavias and Apostolos Kourtis for their useful comments and suggestions on earlier versions of this paper. We are also grateful for feedback received from the participants of the 31 st USAEE/IAEE North American conference "Transition to a Sustainable Era: Opportunities & Challenges" (Austin, Texas, 2012). Finally, we are grateful to three anonymous referees for providing valuable suggestions. The usual disclaimer applies.
Article
The New Zealand Emissions Trading Scheme (NZ ETS) is an intensity-based system and the second oldest national ETS. It is unique in that it is highly international (with unlimited use of Kyoto allowances) and it incorporates forestry. We provide the first empirical analysis of the determinants of allowance prices on the NZ ETS. Our results indicate that imports of offsets rather than fundamentals have been the major price determinant. Moreover, the pricing of New Zealand units (NZUs) can be placed into three distinct periods, delineated by two structural breaks. In the first period, the system is largely autarkic; in the second period, as international offset prices drop below the NZU price, the system becomes a ‘price taker’; in the final period, following some policy interventions, the system regains some independence. The case of the NZ ETS shows both the power of linking ETSs and the dangers of doing so.
Article
This paper presents a theoretical and empirical analysis of liquidity in the German intraday market for electricity. Two models that aim at explaining intraday liquidity are developed. The first model considers the fundamental merit-order and intraday adjustment needs as the drivers of liquidity in a perfectly competitive market. The second model relaxes the assumption of perfect competition in the intraday market and assumes that the trading behavior of profit maximizing market participants influences the liquidity provision. The relevance of commonly used liquidity indicators like the bid ask-spread, resiliency, market depth, price variance, delay and search costs as well as trading volume and the number of trades are analyzed with respect to both models of liquidity. The empirical findings indicate that liquidity in the German intraday market can be explained by the trading model while the purely fundamental model is rejected.
Article
We analyze the risk premium on electricity forward contracts traded for the Nordic and German/Austrian electricity markets. We argue that finding risk premiums by analyzing overnight returns is more relevant than the frequently used ex post approach. The derivatives in these markets can be characterized as trading products and hedging products. Each contract shows a clear increase in trading volume and liquidity when approaching maturity. We link this to a testable hypothesis where financial traders are compensated for holding price risk, and where the sign and magnitude of the risk premium changes depending on the hedging pattern of producers and retailers. Incorporating this in regressions we find that there are higher risk premiums in the period before the forwards become front products, compared to the risk premiums in the front period. Quarterly and monthly contracts show the most significant results.
Article
Analyses of electricity market prices have presented contradictory perspectives on the sign and economic properties of the day-ahead forward premia. By classifying half-hourly trading periods into two homogenous clusters, peak and off-peak, we uncover a systematic diurnal reversal in the sign of forward premium in the British market. This can be explained by the asymmetric positions of generators and suppliers towards risk and its intra-day variation, which is induced both by the intra-day heterogeneity of plant technical characteristics as well as aspects of market design. The sign reversal substantially explains controversial findings in previous research. The analysis is extended to explore the time-varying responses of forward premium to economic, strategic and risk variables. Subsequently, ex-ante forward premia are derived based on spot price predictions and assumptions on agent learning. This procedure allows the derivation of predictive intervals for price risk.
Article
This work discusses potential pitfalls of applying linear regression models for explaining the relationship between spot and futures prices in electricity markets. In particular, the bias coming from the simultaneity problem, the effect of correlated measurement errors and the impact of seasonality on the regression results. Studying a 13-year long (1998-2010) price series of spot and futures prices at Nord Pool and employing regression models with GARCH residuals, we show that the impact of the water reservoir level on the risk premium is positive, which is to be expected, but contradicts the results of [6]. We also show that after taking into account the seasonality of the water level, the storage cost theory proposed by [6] to explain the behavior of convenience yield has only limited support in the data.
Article
Understanding the nature of the forward premium is particularly crucial, but rather elusive, for a non-storable commodity such as wholesale electricity. Whilst forward prices emerge as the expectation of spot plus, or minus, an ex ante premium for risk, the manifestation and empirical analysis must focus upon realised ex post premiums. This presents modelling requirements to control for shocks to the spot expectation as well as the endogeneity of ex post premia with spot price outcomes. In addition, because electricity is a derived commodity in the sense that market prices are often set by technologies that convert gas or coal into power, it is an open question whether much of the premia in power may actually be a pass-through of the premia in gas (or coal). Using a four dimensional VAR model we are able to distinguish fundamental and behavioural aspects of price formation in both the daily and monthly forward premia from the British market. We present new evidence on daily and seasonal sign reversals, associated with demand cycles, the greater importance of behavioural adaptations in the risk premia than fundamental or spot market risk measures, and the substantial fuel risk pass-through. We also show the value of a nonlinear specification in this context.