International Review of Financial Analysis 34 (2014) 222–234
Contents lists available at ScienceDirect
International Review of Financial Analysis
A microstructure analysis of the carbon finance market☆
Don Bredin a, Stuart Hyde b,⁎, Cal Muckley a
a
b
University College Dublin, Ireland
University of Manchester, United Kingdom
a r t i c l e
i n f o
Article history:
Received 18 October 2013
Received in revised form 10 February 2014
Accepted 12 March 2014
Available online 21 March 2014
JEL classification:
G13
G14
G19
a b s t r a c t
The European Union Emissions Trading Scheme is the key policy instrument of the European Commission's
Climate Change Program aimed at reducing greenhouse gas emissions to 8% below 1990 levels by 2012. The
key asset traded under the scheme is the European Union allowance (EUA). This article examines ultra high frequency data to assess the extent of the development in the futures market of the EU Emissions Trading Scheme.
Our results indicate significant developments consistent with sequential information arrival. They also indicate a
negative contemporaneous relationship between volume and volatility for all contracts. The implication is that
liquidity traders dominate any role played by informed traders. Incorporating the duration between trades in
our analysis has significant impact suggesting that any empirical investigation of the intra-day volume–volatility
relationship needs to actively account for the impact of time elapse between trades.
© 2014 Elsevier Inc. All rights reserved.
Keywords:
CO2 emission allowances
Futures
Emission trading
Energy
Kyoto Protocol
Market microstructure
1. Introduction
In January 2005 the European Union (EU) introduced formally the
emission trading scheme (ETS). The scheme forms a central part of the
EU agreement to cut worldwide emissions of carbon dioxide (CO2).1
Under the Kyoto agreement, the EU is committed to reduce greenhouse
gas (GHG) emissions by 8% (relative to 1990 levels) by 2012.2 The aim
☆ The authors would like to thank Michael Brennan, John McConnell, Maureen O'Hara
and Matthew Spiegel for valuable comments and participants at the 2009 Global Finance
Academy Conference at University College Dublin, participants at the 2009 Finsia/MCFS
Conference, Melbourne, the International Association for Energy Economists annual meeting 2012 in Venice and at the European Association of Environmental and Resource
Economists annual meeting 2012 in Prague. This work has emanated from research conducted with the financial support of the Science Foundation Ireland under Grant
Number 08/SRC/FM1389. Ike Johnson and Robert Bradley provided research assistance.
⁎ Corresponding author at: The University of Manchester, Manchester Business School,
United Kingdom.
E-mail addresses: don.bredin@ucd.ie (D. Bredin), stuart.hyde@mbs.ac.uk (S. Hyde),
cal.muckley@ucd.ie (C. Muckley).
1
Member states of the European Union ratified the Kyoto agreement in May, 2002. The
agreement became legally binding on February 16, 2005.
2
Incidentally, the European Environmental Agency (October, 2012) has indicated that
the European Union as a whole will over-deliver on its Kyoto target. Ellerman and
Buchner (2006) find that in 2005 data from the Pilot Phase of the ETS, the scheme has been
responsible for reducing CO2 emissions by between 50 and 200 million tonnes.
http://dx.doi.org/10.1016/j.irfa.2014.03.003
1057-5219/© 2014 Elsevier Inc. All rights reserved.
of the EU to reduce emissions by 2012 through the EU ETS is to be
achieved via two phases. The first Phase (Pilot) of trading is 2005–2007
and the second one (Kyoto), which coincides with the first compliance
period of the Kyoto Protocol, is 2008–2012. After these two initial phases,
the third European trading Phase will commence in 2013.
The emission trading scheme issues a restricted amount of emission
allowances to firms on an annual basis. At the end of each year firms
must hold the required amount of emission permits to meet their emissions of CO2 over that year.3 The ETS facilitates trades for firms that
are either long or short on emission permits. Non-compliance with
the commitments will result in a penalty of € 40 (€ 100) per tonne of
CO2 produced during the first (second) commitment period. The aim
of the ETS is to generate a price signal that will encourage firms to reduce their emissions. Paolella and Taschini (2008) emphasize that
the ultimate aim of this scheme (as well as the US CAAA-Title IV
scheme) must be to create an environment where there is scarcity of allowances which will lead to an upward trend in prices. There has been a
3
Firms are required to submit a report verifying their emissions in any year by March
31st of the following year.
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
considerable amount of uncertainty associated with the price of CO2
emissions over its short life to date.4
This paper examines the performance of the EU ETS from a market
microstructure perspective with the aim of understanding the information assimilation process, liquidity and the degree of market efficiency in the ETS5 during both the Pilot and the Kyoto Phase.6 To
commence our study, we examine the relevance of the mixture of distributions hypothesis (MDH) and the sequential information arrival hypothesis (SIAH) in the emission trading scheme. The MDH
asserts a joint dependence of volume and volatility on the underlying information flow (see, inter alia, Andersen, 1996; Clark, 1973;
Tauchen & Pitts, 1983). Alternatively, while the MDH does not allow
for serial dependence between volume and volatility — it is a contemporaneous relationship, the sequential information arrival hypothesis (SIAH) (Copeland, 1976) posits that traders both receive
information and act on it in a sequential manner, hence a new equilibrium is not established instantaneously and the potential for a lead-lag
relationship between volume and volatility exists. Yet while microstructure models (see Diamond & Verrecchia, 1987; Easley & O'Hara, 1992)
suggest trade arrival times convey information neither the MDH nor
the SIAH hypotheses explicitly account for the effect of time duration,
the elapsed time between consecutive trades, in measuring the information content of trades. In this paper, our key contribution is to take
account of duration effects on the information assimilation process in
the EU ETS.
The microstructure literature acknowledges that trading activity
in asset markets is induced by both revisions to investor expectations
of asset prices, following the arrival of new information, and by the requirements of investors for liquidity (Blume, Easley, & O'Hara, 1994;
O'Hara, 1987). Recent research into the analysis of trading volume as a
measure and signal of investor information and liquidity needs raises
important issues concerning its interaction with other key variables of
the trade process, in particular asset price volatility and trade duration
(Manganelli, 2005; Xu, Chen, & Wu, 2006). However, prior empirical research customarily focuses upon the price impact of one or two specific
trade process variables in isolation, rather than their combined effects.
In order to understand the full dynamic interactions among the three
trade process variables they should be modeled simultaneously. We
adopt such an approach in this paper. We initially adopt the duration
based model proposed by Xu et al. (2006) to examine the relationship
between time consistent volume and volatility, while controlling for exogenous duration effects. In addition, we adopt an alternative model
specification, in the vein of Manganelli (2005), which allows duration
to be endogenous and to be modeled simultaneously together with
4
In April 2006, by way of an example, coincident to the unofficial release of the 2005
emission data by some of the EU member states, the EUA's price declined sharply. EU
ETS spot prices had reached a high of € 30.50 prior to April 2006. Following the official release by the EU commission on the 15th May 2006, showing a larger than expected surplus
in the market, the spot price fell to € 15.63 on the 17th May 2006. Given that banking EUAs
was prohibited between phases, the price eventually converged to close to zero at the end
of Phase 1. In Phase 1, it would appear that the cap placed on emissions was far too lax and
so downward pressure on the price continued.
5
Our interpretation of these concepts draws on the work of Dacorogna et al. (2001).
6
An examination of both the Pilot as well as the Kyoto Phase is important. An understanding of the initial setting up period of the exchange may have relevant and significant
implications for future exchanges or trading schemes, as well as subsequent periods of the
EU ETS. For example, emission trading systems are to be established in six regions in China
by 2013 and nationwide by 2015 to reduce China's emission intensity relative to its GDP.
In addition, Convery and Redmond (2007) highlight that the main objective of the Pilot
Phase was to get the scheme up and running and specifically that it would be fully operational by 2008, the start of Phase 2 (Kyoto Phase). As spot and futures markets, registries
and monitoring, reporting and verification have been established, this principal aim of
Phase 1 has been achieved. The adopted microstructure approach, in this paper, provides
the first thorough indication, using volume, volatility and trade duration (time between
consecutive trades), of the initial performance of the EU ETS.
223
volume and volatility. Our analysis is carried out on Phase 1 and 2 EU
ETS futures and we focus on futures data provided by the ECX.7
Our key results indicate significant market developments, with reduced market frictions suggesting evidence of greater levels of efficiency present in the market. The evidence of reduced market frictions is
reflected in the form of reduced durations between trades, higher volumes traded and lower volatilities. As bi-directional Granger causality
is found in the trade variables, the empirical evidence is consistent
with the sequential information arrival hypothesis (Copeland, 1976)
rather than the predictions of the mixture of distribution hypothesis
(Clark, 1973; Tauchen & Pitts, 1983). This evidence in favor of the sequential information arrival hypothesis is consistent with recent results
on the EU ETS reported by Conrad, Rittler, and Rotfuß (2012) and
Chevallier and Sevi (2011). Turning to the duration related information,
we find that trades associated with longer durations have a higher impact on volatility. Finally, our results indicate a negative contemporaneous relationship, accounting for duration effects, between volume and
volatility. The implication is that liquidity traders dominate any role
played by informed traders (Easley & O'Hara, 1992).
The remainder of the paper is set out as follows: Section 2 discusses
the emergent literature in carbon finance with specific reference to microstructure issues. Section 3 introduces the empirical methodology,
outlining the adopted approach to study the information assimilation
process, liquidity and the degree of market efficiency. Section 4 presents
the data and empirical results. Section 5 provides some concluding
comments.
2. Recent developments in carbon finance
Futures contracts were introduced on the European Climate Exchange (ECX) in March 2005, with options trading following in October
2006. Trades on the ECX are cleared through the Intercontinental Exchange (ICE), which also hosts the electronic marketplaces for the
Chicago Climate Exchange. In March 2008, the ECX introduced certified
emission reduction (CER) futures and options and in October 2008 it introduced an EUA-CER spread trade.
There have been several studies which examine the pricing behavior
and developments in both spot and futures markets. Redmond and
Convery (2006) examine the EU ETS using daily data over the period
1st December 2004 to 31st July 2006 focusing on the behavior of
the price of carbon in relation to energy commodities, meteorological
factors and a number of other variables including dummy variables to
take account of policy and regulatory issues. Mansanet Bataller et al.
(2007) report that energy sources and extreme temperatures are the
principal factors in the determination of CO2 price levels in 2005. Similarly, Alberola, Chevallier, and Cheze (2008) and Bredin and Muckley
(2011) further document the role of fundamentals on daily spot and
futures prices respectively across phases of the ETS. Chevallier (2011a)
assesses the transmission of international shocks from the macroeconomy, commodities and the financial markets to the carbon market.
The results show that carbon prices tend to respond negatively to an exogenous recessionary shock on economic indicators.
While this latter literature examines low frequency price behavior, a
relatively small number of studies investigate microstructure issues,
e.g., price discovery, liquidity and intra-day volatility issues, using high
frequency data. The first such study, Benz and Hengelbrock (2009)
7
Futures contracts account for 76% of the volume of trades in EUAs. Of these contracts,
the vast majority are traded on the ECX which was acquired by the Intercontinental Exchange in July 2010 (Mizrach, 2010). In late April 2010, the Intercontinental Exchange
(ICE) agreed to pay over US $ 600 million in cash to acquire Climate Exchange (owner
of the ECX) for which it previously cleared trades. The remainder of the EUA futures contracts are largely traded on the European Energy Exchange (EEX) and Nordpool (see
Mansanet-Bataller & Pardo, 2008). Both spot and futures contracts are traded on the
Norwegian platform Nord Pool, which has a market share of 12.5% on the futures market.
In 2010 NASDAQ QMX acquired all the shares in Nord Pool International, Nord Pool clearing ASA and Nord Pool Consulting (Mizrach, 2010).
224
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
finds evidence of increased liquidity during Phase 1 for both ECX and
Nord pool, with particularly strong evidence for the considerably larger
ECX market. Further, these authors indicate, in terms of price impact
and price discovery, that the ECX represents a price leader especially
for more recently traded contracts (December 2007 and 2008). Extending this research on price discovery, Mizrach and Otsubo (2014) examine a number of issues relating to trading volumes, transaction costs
(spreads), price impact and return predictability for EUA and CER contracts during 2009 (Phase 2). Consistent with Benz and Hengelbrock
(2009), Mizrach and Otsubo (2014), examining Granger-Gonzalo and
Hasbrouck information shares, find evidence that the ECX futures market provides 90% of the price discovery. This microstructure research establishes the primacy of the ECX platform for investigating the behavior
of EUA futures.8
Another strand of the carbon microstructure emergent literature
empirically examines the extent of intra-day volatility. Rotfuß (2009)
provides an analysis of volatility characteristics for the first and second
trading periods while Chevallier and Sevi (2011) document the superior
ability of realized variance measures at forecasting volatility for EUA
futures.9 Rittler (2009) shows evidence of volatility transmission from
the futures market to the spot market and confirms that futures provide
the leading role in price discovery. Focusing on the initial period of
Phase 2, Conrad et al. (2012) demonstrate that regulatory announcements have statistically significant effects on intra-day volatility, while
there is a relatively marginal influence from unexpected positive aspects of scheduled macroeconomic announcements in Germany and
the United States.
In this article, we extend the outlined microstructure related findings with regard to market liquidity, volumes and intra-day volatility
on the EU's ETS. Specifically, we further examine the information assimilation process and the degree of market efficiency, in the ETS, during both the Pilot and Kyoto Phases of the Scheme. We examine the
empirical evidence in regard to the sequential information arrival hypothesis (Copeland, 1976) and the mixture of distributions hypothesis
(Clark, 1973; Tauchen & Pitts, 1983). In addition, unlike the previously
discussed microstructure studies, we simultaneously model the three
key trade process variables: duration (i.e. the time elapsed between
consecutive trades), trade volume and asset price volatility with a
view to accounting for the dynamic interactions involved and identifying the class of traders principally present in the market, whether informed or liquidity traders. Finally, we conduct a range of sensitivity
tests, across model specifications, contract expirations and near run to
expiration contract tests, and we compare our Pilot Phase findings
with those of the Kyoto Phase.
Table 1
Summary statistics for EU ETS futures.
Instrument
Dec. 05
Dec. 06
Dec. 07
Dec. 08
Dec. 09
Dec. 10
Dec. 11
Dec. 12
Market microstructure research must confront the problems generated by the fact that transaction data arrives at irregularly spaced intervals in time. Nevertheless, many studies simply ignore the time
variation between successive trades, and analyze the data utilizing
econometric techniques more appropriate for equally spaced observations. We maintain that this approach may result in a significant loss
of information, since the elapsed time between consecutive trades
(duration) may convey important information about the state of the
market.
8
A comparison of price impacts for both EUAs and certified emission reductions (CERs)
is reported by Mizrach and Otsubo (2014). The authors find that the median impact for an
EUA trade is 0.0106 euro and 0.0145 for the CER trade. Mizrach and Otsubo (2014) indicate that the larger price impact of a CER trade can be explained by the larger bid-ask
spreads.
9
Chevallier and Sevi (2011) also find the distribution of standardized returns to be nonnormal, contrary to the mixture of distribution hypothesis (MDH).
Duration
3135
15,831
5118
126,722
248,214
261,161
301,064
262,228
Volume
Mean
Median
Mean
Median
21.63
9.99
23.15
2.32
0.59
0.70
0.69
0.38
421.76
191.00
392.00
27.47
0.05
0.05
0.05
0.01
10.32
12.07
13.22
7.04
4.78
7.57
8.25
22.99
10.00
10.00
10.00
10.00
1.00
3.00
3.00
0.76
Notes. The summary statistics are reported for the full samples of available data. The sample
periods for our seven contracts are as follows: December 2005 (22nd April 2005 to 19th
December 2005), December 2006 (16th June 2005 to 18th December 2006), December
2007 (3rd June 2005 to 17th December 2007) and December 2008 (28th September
2005 to 15th December 2008), December 2009 (2nd January 2009 to 14th December
2009), December 2010 (4th January 2010 to 10th December 2010), December 2011 (4th
January 2011 to 19th December 2011) and December 2012 (3rd January 2012 to 17th
December 2012). The duration is defined as time elapsed between consecutive trades
and the mean and median figures above are measured in minutes.
Two approaches have been proposed to accommodate this concern.
First, Engle (2000) argues that variables in empirical microstructure
models should be adjusted for duration in order to obtain timeconsistent parameter estimates. In this vein, Xu et al. (2006) incorporate
a time factor by standardizing both volume and volatility with respect to
elapsed time (trade duration) thereby creating time consistent measures for empirical tests. We begin our analysis, in this spirit, by utilizing
a time consistent structural vector autoregression (VAR) model to characterize the dynamic relationship between our duration-based volatility
and volume measures. Second, Engle and Russell (1998) and Dufour
and Engle (2000) explicitly incorporate the duration between trades
as an endogenous variable in a microstructure model. The inclusion of
a duration variable not only addresses the estimation problems associated with irregularly occurring observations, but it also permits the
information conveyed by duration to be explicitly incorporated in the
modeling framework. In order to better understand the dynamic relationship between duration, volume, and volatility, we further extend
the analysis by specifying a trivariate VAR framework in which we explicitly model duration.
3.2. The VAR models
Initially, we formulate a bivariate VAR model where volume and volatility are standardized per unit of time, thereby accommodating the
scale effect of duration on volatility and trade volume,
3. Econometric methodology
3.1. Empirical issues
No. Obs
zd;t ¼ μ z þ
p
X
i¼1
vd;t ¼ μ v þ
p
X
i¼1
azi zd;t−i þ
q
X
ðbzi þ czi τt−i Þvd;t−i þ ut
i¼0
q
X
avi zd;t−i þ
ðbvi þ cvi τt−i Þvd;t−i þ ϵt
ð1Þ
i¼1
where zd,t and vd,t are duration-based volatility and volume, respectively.10 Specifically, if dt = Tt − Tt − 1 is duration, the elapsed time between
h i
consecutive trades at Tt and Tt − 1, then zd;t ¼ ln Zdt is the log volatility
t
h i
per unit of time, and vd;t ¼ ln Vd t is the log volume per unit of time.
t
Here, we measure volatility, Zt, by the absolute value of returns, |rt|
and the volume variable, Vt, by the trade size at calendar time t. The orders of the lag length for the duration-based volatility and volume variables, (p,q), are determined on the basis of the Akaike information
criterion (AIC). Following Xu et al. (2006), we include an interaction
10
Both duration-based volatility and volume are per unit time and in log terms to ensure
that the variables are positive.
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
225
Table 2
Causality tests.
H0 no GC
Dec. 05
Dec. 06
Dec. 07
Dec. 08
Dec. 09
Dec. 10
Dec. 11
Dec. 12
Volm. to dur.
2.528
(0.000)
3.088
(0.000)
1.496
(0.072)
0.943
(0.531)
1.657
(0.033)
0.756
(0.769)
11.767
(0.000)
9.138
(0.000)
1.753
(0.020)
1.869
(0.011)
2.526
(0.000)
3.076
(0.000)
4.309
(0.000)
4.122
(0.000)
1.745
(0.021)
1.674
(0.030)
3.132
(0.000)
2.678
(0.000)
62.923
(0.000)
88.127
(0.000)
9.964
(0.000)
2.803
(0.000)
1.831
(0.013)
9.889
(0.000)
152.19
(0.000)
158.32
(0.000)
11.27
(0.000)
1.16
(0.274)
14.84
(0.000)
9.18
(0.000)
58.05
(0.000)
9.54
(0.000)
102.19
(0.000)
2.658
(0.000)
3.99
(0.000)
44.43
(0.000)
129.74
(0.000)
214.75
(0.000)
54.73
(0.000)
14.19
(0.000)
7.77
(0.000)
7.18
(0.000)
60.63
(0.000)
256.468
(0.000)
69.034
(0.000)
7.338
(0.000)
14.758
(0.000)
4.929
(0.000)
Dur. to volm.
Voly. to dur.
Dur. to voly.
Voly. to volm.
Volm. to voly.
Notes. This table reports Granger causality (GC) tests on duration (dur.), volume (volm.) and volatility (voly.) derived from the tick by tick data for the December 2005, 2006, 2007, 2008,
2009, 2010 and 2011 contracts. The sample periods for our seven contracts are as follows: December 2005 (22nd April 2005 to 19th December 2005), December 2006 (16th June 2005 to
18th December 2006), December 2007 (3rd June 2005 to 17th December 2007) and December 2008 (28th September 2005 to 15th December 2008), December 2009 (2nd January 2009 to
14th December 2009), December 2010 (4th January 2010 to 10th December 2010), December 2011 (4th January 2011 to 19th December 2011) and December 2012 (3rd January 2012 to
17th December 2012). F statistics are reported, with levels of significance in brackets.
term τt − i(τt = ln(dt)), that allows the volume to vary with duration in
both equations.11
Given duration is informative about an asset's true value (Diamond
& Verrecchia, 1987; Easley & O'Hara, 1992), it follows that duration
should interact with volume to influence the characteristics of asset
prices. The bivariate VAR specification provides an insightful base
from which to analyze our subsequent trivariate findings, which allow
for the possible exogeneity of duration effects. Specifically, the detailed
bivariate VAR model, which specifies an interaction term between
volume and duration in the determination of future volumes and volatility, serves to investigate, in a preliminary manner, the importance of
accounting for duration effects. The shock, ut, represents the informed
shock as it allows for a contemporaneous relationship between
duration–volume through to volatility. Easley and O'Hara (1992) highlight that market makers will adjust their quoted prices as a result of
the trading of informed traders. The shock, ϵt, represents the uninformed shock as it allows for lagged relationships.
In the trivariate model, the assumption of exogenous duration variation is relaxed and duration is included as an endogenous term. This allows for interaction effects from duration to volume and to volatility
and vice versa. The inclusion of both model specifications, disallowing
and allowing for endogenous duration effects, is of particular interest.
To capture these interactions, we extend the VAR analysis to include
time as an endogenous variable. The VAR for duration, dt, volume, Vt,
and volatility, Zt, is:
dt
¼
μd þ
p1
X
γid dt−i þ
q1
X
i¼1
Vt ¼ μv þ
Zt ¼ μ z þ
p2
X
γiV dt−i þ
ρid V t−i þ
i¼1
q2
X
ρiV V t−i þ
r2
X
i¼0
i¼1
i¼0
q3
X
r3
X
γ iZ dt−i þ
i¼0
ρiZ V t−i þ
δid Z t−i þ εt
Z t ¼ K t þ εt
ð3Þ
where Kt = ln k1/2
t , kt is the latent information flow and εt is the disturbance term. In the traditional MDH model, return volatility is related to
the latent information flow kt concurrently, and the disturbance term is
serially independent. The volatility processes, specified in Eqs. (1) and
(2), generalize the MDH volatility process in Eq. (3) in several ways.
The presented equations accommodate a process of dynamic adjustment or a learning effect in relation to the new information, so that return volatility is affected by possibly lagged information. For example,
they permit a lagged dependence in volatility, Zt.12 If there is an absence
of serial dependence, the values of the lagged coefficients in the VAR
systems of equations will be insignificantly different from zero. The
models simply degenerate to the traditional MDH volatility process. In
contrast, significant coefficients on the lagged terms are indicative of
the SIAH.13
Typically, empirical tests of the MDH model at the intraday level use
transaction data aggregated up to fixed intraday intervals. Naturally,
there is a loss of information in this aggregation. This loss is justified
partly due to the large number of zeros, making econometric analysis
relatively complex if the intervals are small. In contrast, our VAR
model specifications are directly based on irregularly spaced transaction data, which is free from the problem of temporal aggregation. At
the same time they account for transaction time, by the inclusion of
duration.
i¼1
p3
X
i¼0
r1
X
information and a random disturbance term. The return volatility process in the MDH framework can be written as:
δiV Z t−i þ ηt
ð2Þ
δiZ Z t−i þ ζ t
i¼1
where the lag orders are determined by the Akaike information criterion (AIC).
3.2.1. Sequential and mixed distribution information assimilation
The return volatility processes, specified in Eqs. (1) and (2), are generalizations of the return process in the traditional MDH model. The
MDH model postulates that return volatility depends on the flow of
11
It is then possible to examine whether duration between trades affects price adjustment to trades in the volatility equation. In the volume equation, the interaction term allows us to observe the correlation between current and past volume, while accounting for
the effect of elapsed time on this relation.
3.2.2. Prevalence of informed or liquidity trading
Bagehot (1974) assumes that in the market there exist traders with
superior information who try to transform their informational advantage into net profits. In this framework, prices are semi-strong form
efficient, they reflect all public information, but not necessarily the private information which may be in the process of being traded into
prices. Informed traders are defined as traders that use the fundamentals to form reliable opinions as to whether an instrument is under or
12
This generalization is quite appropriate given that empirical evidence suggests strong
volatility persistence, at the microstructure level in liquid asset markets, even after controlling for the effect of information flow. The cause of volatility persistence is usually attributed to one or a combination of microstructure imperfections including, inventory
control, exchange-mandated price smoothing, and/or lagged adjustment to information
among others.
13
It is worthwhile noting that the greater the extent of the sequential information arrival
dynamics over time, the more slowly information is accounted for in market price. In this
scenario, it is less likely that there is a prevalence of informed trading in the market.
226
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
Table 3
Bivariate volume–volatility VAR (December 2005 contract).
Lagged volatility
Current volume
Lagged volume
Current dur. ∗ volume
Lagged dur. ∗ volume
a1
a2
a3
a4
a5
a6
a10
b0
b1
b2
b3
b4
c0
c1
Table 5
Bivariate volume–volatility VAR (December 2008 contract — full sample).
Volatility equation
R2 = 0.162
Volume equation
R2 = 0.121
Coefficient
t-Stats
Coefficient
t-Stats
0.206
0.051
0.044
0.074
0.052
0.072
0.037
−0.346
11.133
2.715
2.314
3.909
2.767
3.803
2.001
4.446
−0.009
−0.017
1.952
3.661
0.191
0.113
0.074
0.058
−0.111
Lagged volatility
9.858
5.750
3.755
2.933
4.247
−0.017
2.681
Notes. This table reports the results of the bivariate volume–volatility VAR. The variables
are standardized with respect to duration and are in log-form to ensure that they remain
positive. Estimation is based on the tick-by-tick data for the December 2006 expiry EUA
futures contract traded on the ECX. The sample is 22nd April 2005 to 19th December
2005. Only coefficients statistically significant at a significance level of 5% are reported
along with absolute t-statistics.
Current volume
Lagged volume
Table 4
A bivariate volume volatility VAR summary for December 2005–07 contracts.
Period
b0
c0
2007
(t-value)
2006
(t-value)
2005
(t-value)
−0.49
(7.64)
−0.37
(9.82)
−0.35
(4.45)
−0.15
(7.34)
−0.22
(17.32)
−0.11
(4.25)
Notes. This table reports the summary results of the bivariate volume–volatility VARs in
respect to the Phase I contracts which expire December in each year 2005–07. In particular, it reports the contemporaneous volatility–volume coefficient, b0, and the contemporaneous duration–volume interaction coefficient, c0, in the volatility equation of these
bivariate model specifications. The variables are standardized with respect to duration
and are in log-form to ensure that they remain positive. Estimation is based on the tickby-tick data for EUA futures contracts traded on the ECX. The December 2005 contract is
sampled 22nd April 2005 to 19th December 2005. The December 2006 contract is sampled
from 16th June 2005 until 18th December 2006. The December 2007 contract is sampled
from 3rd June 2005 until 17th December 2007, respectively. Only coefficients statistically
significant at a significance level of 5% are reported along with absolute t-statistics.
overvalued and then act on these opinions. In contrast, liquidity traders
are uninformed traders that do not know the fundamental value of the
instrument when forming their trading decisions (Harris, 2003). Liquidity traders are motivated by consumption needs and portfolio strategies,
e.g., their trading is a reflection of their particular price sensitivities or
particular trading rules (see Easley & O'Hara, 1992).
Easley and O'Hara (1992) and, more recently, Manganelli (2005)
highlight that although an informed trader may prefer to exploit his informational advantage by trading a larger amount of the asset, initiating
this strategy would reveal his inherent informational advantage and the
market maker would adjust quotes accordingly. Hence, in order to disguise his investor type, an informed trader may segment his order
flow into a sequence of smaller trades separated in time.14 Consequently, given that informed traders trade to exploit their informational advantage, subsequent to the arrival of price relevant information,
14
The underlying assumption is that liquidity (noise) agents, trade with constant
intensity.
Current dur. ∗ volume
Lagged dur. ∗ volume
a1
a2
a3
a4
a5
a6
a7
a8
a9
a10
a11
a12
a13
a14
a15
a16
a17
a19
a20
b0
b1
b2
b3
b4
b5
b6
b7
b8
b9
b10
b11
b12
b13
b14
b15
b16
b17
b18
b19
b20
c0
c1
c2
c3
c4
c5
c6
c7
c8
c10
c13
c14
c15
c18
c20
Volatility equation
Volume equation
R2 = 0.070
R2 = 0.119
Coefficient
t-Stats
Coefficient
t-Stats
0.169
0.052
0.031
0.023
0.020
0.016
0.011
0.017
0.009
0.012
0.008
59.869
18.332
10.797
7.899
7.067
5.446
3.932
6.101
3.214
4.027
2.722
−0.015
20.622
0.002
2.111
0.004
5.350
0.002
0.002
0.003
2.572
2.457
4.003
0.002
2.104
0.012
4.239
0.012
4.088
0.002
0.002
2.640
2.878
0.009
0.012
0.007
−0.340
0.043
0.022
0.034
0.037
3.087
4.060
2.354
31.753
3.839
1.983
3.028
3.314
0.224
0.082
0.044
0.034
0.024
0.024
0.020
0.013
0.015
0.012
0.012
0.014
0.011
0.017
0.011
0.011
0.013
0.009
0.007
0.010
79.234
28.379
14.944
11.669
8.217
8.335
6.867
4.527
5.132
4.192
4.139
4.632
3.791
5.897
3.603
3.779
4.350
3.000
2.567
3.374
−0.033
−0.033
−0.026
−0.018
−0.012
−0.011
7.511
7.457
5.910
4.024
2.748
2.530
−0.008
−0.003
6.711
2.315
−0.002
2.084
−0.012
−0.010
−0.012
2.811
2.224
2.776
−0.003
2.173
−0.012
−0.012
−0.013
−0.012
2.271
2.713
2.834
2.753
−0.004
3.172
Notes. This table reports the results of the bivariate volume–volatility VAR. The variables
are standardized with respect to duration and are in log-form to ensure that they remain
positive. Estimation is based on the tick-by-tick data for the December 2008 expiry EUA
futures contract traded on the ECX. The sample is 28th September 2005 to 15th December
2008. Only coefficients statistically significant at a significance level of 5% are reported
along with absolute t-statistics.
expected duration will fall and expected volumes will rise, making
trade more likely.15 As a result of this rationale, the informed traders
are associated with negative contemporaneous duration–volume and
15
Similarly, in Kyle (1985), insider (informed) traders act strategically. They rationally anticipate the impact of orders on prices, by conditioning on the behavior of other market participants (namely, the market maker and uninformed traders). As new private information
arrives, informed traders review the existing price quotes of a specialist market maker and
select the amount of shares which maximizes their informational advantage to trade.
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
227
Table 6
Trivariate duration–volume–volatility VAR (December 2008 contract).
Duration equation
Coefficient
Current duration
Lagged duration
Current volume
Lagged volume
Current volatility
Lagged volatility
γ0
γ1
γ2
γ3
γ4
γ5
γ6
γ7
γ8
γ9
γ10
γ11
γ12
γ13
γ14
γ15
γ16
ρ0
ρ1
ρ2
ρ3
ρ4
ρ5
ρ6
ρ7
ρ8
ρ9
ρ10
ρ11
ρ12
ρ13
ρ14
ρ15
ρ16
δ0
δ1
δ2
δ3
δ4
δ5
δ6
δ7
δ8
δ9
δ10
δ11
δ12
δ13
δ14
δ15
δ16
Volume equation
t-Stats
0.197
0.106
0.061
0.043
0.032
0.024
0.015
0.021
0.018
0.013
0.014
0.007
0.014
0.011
0.014
0.017
59.191
30.918
17.712
12.312
9.215
6.728
4.236
5.994
5.185
3.894
4.000
2.112
4.037
3.095
4.133
5.043
0.011
0.044
0.015
2.927
11.927
3.988
0.055
0.012
20.961
5.133
−0.006
2.321
−0.006
−0.007
2.213
−2.435
−0.007
2.754
Volatility equation
Coefficient
t-Stats
Coefficient
t-Stats
0.103
0.034
0.014
35.039
11.378
4.720
−0.008
2.686
0.399
−0.086
−0.038
−0.029
−0.016
−0.015
−0.018
99.939
20.668
9.127
6.941
3.723
3.552
4.161
−0.018
−0.009
−0.015
−0.013
4.185
2.076
3.530
3.165
−0.009
2.092
−0.114
0.018
0.016
0.013
25.889
4.097
3.568
3.000
0.205
0.067
0.044
0.026
0.033
0.018
0.017
0.021
0.027
0.018
0.012
0.015
0.016
0.014
0.013
0.019
64.123
20.441
13.391
7.916
10.019
5.542
5.062
6.301
8.105
5.493
3.772
4.489
4.771
4.283
3.961
5.816
−0.007
0.186
0.110
0.061
0.038
0.023
0.019
0.015
0.016
0.009
0.014
0.016
0.012
0.013
0.011
0.016
0.020
−0.060
0.023
2.212
58.191
33.872
18.681
11.514
7.093
5.804
4.661
4.770
2.922
4.273
4.842
3.686
3.962
3.423
4.994
6.304
25.889
9.637
Notes. This table reports the results of the trivariate duration–volume–volatility VAR with 16 lags. Estimation is based on the tick-by-tick data for the December 2008 expiry EUA
futures contract traded on the ECX. The sample is 2nd January 2008 to 15th December 2008. Only coefficients statistically significant at a significance level of 5% are reported along
with absolute t-statistics.
duration–volatility relations and a positive contemporaneous volume–
volatility relation.
We adopt a causal ordering, consistent with the outlined rationale,
that runs from duration, a variable with evident information content,
to volume, a variable that also pertains to information content. In turn,
either or both of these information laden variables may induce trades,
which may move prices.16 Hence, the ordering of the variables in the bivariate system (and in the trivariate VAR) is theoretically informed.17
16
This is consistent with Manganelli (2005, p.382) who highlights ‘(i) the duration between trades can affect prices, (ii) duration should be correlated with volume, and
(iii) both duration and volume should be correlated with transaction price variances.’
17
Additionally, as a result of this rationale, it can be seen from Eq. (1), scaled volatility, zd,t and, volume, vd,t, are driven by two uncorrelated random shocks; an informed
traded shock, ut and an uninformed traded shock, ϵt.
4. Data and empirical results
4.1. Data and contractual issues
We adopt intra-day transaction data for Phase 1 and 2 EUA futures
contracts with a December 2005–12 expiry.18 One futures contract corresponds to 1000 EUAs, and each EUA represents the right to emit
1000 tonnes of CO2 equivalent greenhouse gases. The settlement is
three days after the last trading day. The data is taken from the continuous trading session, which runs from 7 am to 5 pm (UK local time). The
18
Contracts with a quarterly expiry exist (March, June, September and December), however the December expiry contracts are significantly more liquid and hence we focus on
these instruments. For the remainder of the paper the contracts will be referred to by
the year of expiration.
228
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
Table 7
Trivariate duration–volume–volatility VAR (December 2008 contract — nearby contract).
Duration equation
Current duration
Lagged duration
Current volume
Lagged volume
Current volatility
Lagged volatility
γ0
γ1
γ2
γ3
γ4
γ5
γ6
γ7
γ8
γ9
γ10
γ11
γ12
γ13
γ14
γ15
ρ0
ρ1
ρ2
ρ3
ρ4
ρ5
ρ6
ρ7
ρ8
ρ9
ρ10
ρ11
ρ12
ρ13
ρ14
ρ15
ρ16
δ0
δ1
δ2
δ3
δ4
δ5
δ6
δ7
δ8
δ9
δ10
δ11
δ12
δ13
δ14
δ15
δ16
Volume equation
Volatility equation
Coefficient
t-Stats
0.176
0.087
0.053
0.036
0.036
0.017
0.016
0.017
0.015
0.012
0.019
32.174
15.668
9.529
6.441
6.534
3.027
2.858
2.975
2.637
2.131
3.397
0.022
4.028
0.005
3.092
0.055
0.141
0.060
2.839
7.117
2.998
0.219
0.110
0.060
0.047
0.017
0.028
0.016
0.018
42.705
20.893
11.446
8.890
3.123
5.293
3.014
3.422
0.012
0.021
0.013
0.011
0.018
0.015
0.011
−1.091
0.300
2.326
4.020
2.444
2.066
3.351
2.812
2.111
25.615
6.809
0.091
2.085
0.105
2.417
2.001
12.365
Coefficient
t-Stats
Coefficient
0.022
0.004
0.005
15.494
2.457
3.462
0.011
−0.002
−0.001
−0.001
70.189
11.763
3.827
5.615
−0.001
2.495
−0.003
2.044
−0.001
2.390
−0.001
2.802
−0.016
0.005
−25.615
7.212
0.001
2.315
0.200
0.050
0.048
0.025
0.018
0.027
0.012
38.914
9.902
9.160
4.804
3.521
5.122
2.325
0.016
0.010
0.030
0.011
3.102
1.960
5.900
2.235
0.020
0.015
3.954
2.926
0.099
−0.448
2.308
t-Stats
2.791
Notes. This table reports the results of the tri-variate duration–volume–volatility VAR with 16 lags. Estimation is based on the tick-by-tick data for the December 2008 expiry EUA futures
contract traded on the ECX. The sample is observed in November 2008. Only coefficients statistically significant at a significance level of 5% are reported along with absolute t-statistics.
ECX has currently over 100 members trading EUAs.19 The minimum tick
size is € 0.01 per tonne of CO2, while a range of order types are available
including limit, market and stop orders.20 The annual fee for ECX members is € 2500, while the trading and clearing fee per contract is currently € 3.50. There are currently (introduced in December 2008) three
market makers for the ECX (Five Rings Capital — formerly Jane Street
Global Trading, Saxon Finance and RNK Capital). The market maker is
required to provide bid and ask quotes for at least 85% of the trading
time between 7 am and 5 pm (UK local time) and are required to respond to quote requests within 5 min.
Our dataset includes comprehensive information relating to all
transactions, maturities, trade prices, opening prices and trading volume for eight different futures instruments, 2005–12, traded on Phases
19
See www.ecx.eu for details.
A dummy variable in respect to the intercept is adopted to take account of the change
in the tick size, from € 0.05 to € 0.01, on the 27th March 2007 has been included.
20
1 and 2 of the EU ETS. To prepare the data for the analysis, we first sum
the trading volumes when we detect a series of equal recorded prices
for the same contract maturity which occur at the same time.21 The
summation of trading volumes can be motivated by the practice in
the management of the electronic book. This consists of utilizing the
existing liquidity that is available in the sense that a trader can choose
to split a large transaction into a series of smaller transactions and trading with different counter parties. We then compute the duration between trades, treating the overnight period as if it did not exist.22
21
We adjusted the data to remove diurnal patterns, which were found principally in the
Phase II data, in line with the deterministic time filter of Tsay (2009).
22
As our futures contracts are quite thinly traded, following Manganelli (2005), we prefer to retain the overnight return. In respect to duration, we treat the overnight period as if
it didn't exist. By way of contrast, eliminating absolute returns during the overnight period, e.g., by adopting a dummy variable, would have possibly caused the loss of important
data for these quite illiquid futures contracts.
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
229
Table 8
Trivariate duration–volume–volatility VAR (December 2008 contract — signed volume).
Duration equation
Current duration
Lagged duration
Current volume
Lagged volume
Current volatility
Lagged volatility
γ0
γ1
γ2
γ3
γ4
γ5
γ6
γ7
γ8
γ9
γ10
γ11
γ12
γ13
γ14
γ15
ρ0
ρ1
ρ2
ρ3
ρ4
ρ5
ρ6
ρ7
ρ8
ρ9
ρ10
ρ11
ρ12
ρ13
ρ14
ρ15
ρ16
δ0
δ1
δ2
δ3
δ4
δ5
δ6
δ7
δ8
δ9
δ10
δ11
δ12
δ13
δ14
δ15
δ16
Volume equation
Coefficient
t-Stats
0.179
0.091
0.056
0.038
0.038
0.017
0.016
0.017
0.015
0.012
0.020
32.907
16.522
10.157
6.851
6.745
3.075
2.930
3.028
2.703
2.111
3.521
0.023
4.140
−0.078
5.195
1.935
12.054
Coefficient
Volatility equation
t-Stats
0.306
0.105
0.057
0.044
0.025
0.024
0.018
0.022
59.626
19.622
10.481
8.087
4.633
4.492
3.280
4.005
0.019
3.528
0.014
0.013
2.693
2.474
Coefficient
t-Stats
0.011
−0.002
−0.001
−0.001
68.669
12.092
4.429
5.676
−0.001
2.345
−0.001
2.431
−0.001
2.779
0.001
1.930
0.197
0.052
0.048
0.026
0.019
0.019
0.013
38.413
9.902
9.131
4.933
3.664
3.664
2.434
0.018
0.010
0.029
0.010
3.500
1.969
5.756
2.059
0.020
0.014
3.877
2.911
Notes. This table reports the results of the trivariate duration–volume–volatility VAR with 16 lags. Estimation is based on the tick-by-tick data for the December 2008 expiry EUA
futures contract traded on the ECX. The sample is 2nd January 2008 to 15th December 2008. Only coefficients statistically significant at a significance level of 5% are reported along
with absolute t-statistics.
4.2. Empirical results
In Table 1, we report the EUA futures contracts' descriptive statistics.
The data examined for each contract varies considerably; with approximately 9, 19, 31 and 40 months available for the December 2005, 2006,
2007 and 2008 respectively. In respect to the 2009–12 contracts, we examine the calendar year of expiration exclusively, as these time periods
relate to the majority of the volumes traded. The sample periods for our
seven contracts are as follows: December 2005 (22nd April 2005 to 19th
December 2005), December 2006 (16th June 2005 to 18th December
2006), December 2007 (3rd June 2005 to 17th December 2007) and
December 2008 (28th September 2005 to 15th December 2008),
December 2009 (2nd January 2009 to 14th December 2009), December
2010 (4th January 2010 to 10th December 2010), December 2011 (4th
January 2011 to 19th December 2011) and December 2012 (3rd January
2012 to 17th December 2012). Indicative of the development of the
market, particularly in the Kyoto Phase of the ETS, the number of transactions and the volume of trades grow approximately monotonically
over time and in accordance the duration declines. Furthermore, the
volatility in the Kyoto Phase is generally at a lower level than that in
the Pilot Phase, with the exception of the heightened volatility present
in the 2010 period.
In Table 2, we examine the interaction over time between the variables in the VAR specifications and whether the theoretical justification
for the ordering of the variables in the VAR can be rejected empirically
using Granger causality tests. The distinction between the December
2005 contract and all other contracts (either Phase 1 or 2) is evident.
We fail to reject a number of hypotheses for the December 2005 contract, for instance that volatility does not Granger cause duration and
that either duration or volume does not Granger cause volatility, at
230
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
the standard 5% level of significance.23 However, for the remaining
contracts the results indicate strong inter-relationships between the
three variables. Such strong evidence of inter-relationships and bidirectional causality between all variables does not provide evidence
to counter our theoretical basis for ordering the VAR. Moreover, the
bi-directional Granger causality points to evidence in favor of SIAH,
and is inconsistent with the MDH, for the EU ETS futures market. Finally,
these initial findings motivate further analyses, adopting VAR specifications, with regard to the interactions between trade duration, volume
and volatility.
Following Alberola et al. (2008) we take account of recognized structural breaks in the data, e.g., the compliance break in April 2006 and
the break associated with the announcement of stricter validations of
National Allocation Plans in October 2006, by including dummy variables to capture the potential shifts in the intercepts due to these breaks
in the estimation of the 2006, 2007 and 2008 contracts. In addition, following Rotfuß (2009), we adopt a dummy variable in the intercepts to
take account of a structural break March 27th, 2007 when the Intercontinental Exchange (ICE) reduced the minimum tick size for all EUA
futures contracts from € 0.05 to € 0.01. It has been included in the
model for the December 2007 and the December 2008 contract. We
adopt an identical set of dummy variables across VAR specifications,
where the sampled data contains the potential structural break. This latter dummy variable is statistically significant in all cases for the 2008
contract and there are mixed levels of significance for the 2007 contract.
However, our results show that the impact of the breaks on volume and
volatility is mainly confined to the 2006 contract. In each case the VAR
model specification is estimated up to the optimal lag length identified
by the AIC (akaike information criterion). The tables report significant
coefficients only.
4.3. Phase I vector autoregression findings
The estimates of the bivariate, duration standardized Pilot Phase,
VAR specification represented in Eq. (1) are reported in Tables 3
and 4. In Table 3 the bivariate model for December 2005 is reported,
while in Table 4 a summary of the Phase 1 results using the bivariate
model is reported.24 The results (coefficient b0) reported in Table 4 indicate that there is a negative contemporaneous relationship between
volume and volatility for all contracts. Moreover, the coefficient is highly significant in each case. A similar finding has been reported by Karpoff
(1988) for the case of financial futures contracts, bonds and commercial paper. This relationship between volume and volatility is also
contingent on the time duration between trades. The negative volume–volatility relationship is strengthened as the duration between
trades increases since the contemporaneous interaction term between
duration and volume (see coefficient c0 in Table 3) is negative and
significant.25 This suggests that any empirical investigation of the
intra-day volume–volatility relationship needs to actively account for
the impact of time elapse between trades. The implication here, arising
both from the negative contemporaneous relation between volume and
volatility, as well as the strengthening of this relation, with the time
elapse between trades, is that liquidity traders are dominating any informed trading behavior.
23
Notwithstanding, it is worthwhile highlighting that, despite a relatively small sample
size as detailed in Table 1, the Dec 05 contract is associated with granger causality in three
of the six bivariate relations studied [at the 5% level] and in four of the bivariate relations
studied [at the 10% level]. Albeit this evidence isn't as convincing as the causality evidence
found in respect to the 2006–11 contracts, it nevertheless is quite persuasive.
24
In the interests of conserving space, we have not reported the individual bivariate
model results for all Phase 1 contracts. The individual contract results are qualitatively
consistent and are available from the authors upon request.
25
The identified positive contemporaneous relation between duration and volatility,
when controlling for volume, is indicative of the prevalence of liquidity traders in the market (Easley & O'Hara, 1992).
The volatility is quite persistent, as evidenced by the positive significant coefficients on lagged volatility in Table 3. There is also consistent
evidence for all contracts of positive autocorrelation in lagged volume.
In addition, there is some evidence to indicate that lagged volume increases volatility, when accounting for duration effects. The estimated
interactions over time between volatility and our lagged information
variables, duration and volume, are consistent with the SIAH. Overall,
the results are consistent across the Pilot Phase contracts. They indicate
the relative prevalence of liquidity traders, considerable market development and the results also indicate the importance of the sequential
information arrival hypothesis in the Pilot phase.
4.4. Phase II vector autoregression findings
The estimates of the bivariate VAR specification represented by
Eq. (1) for the Kyoto Phase are reported in Table 5 for the December
2008 contract. The estimated trivariate VAR specifications, represented
in Eq. (2) are reported in respect to the December 2008 contract in
Tables 6 to 8. The results reported in Table 5 are consistent with the
SIAH and those results reported for Phase 1. There is a negative contemporaneous volume and volatility relationship which is highly significant.
Table 9
A trivariate duration volume volatility VAR summary for December 2008–12 phase II contracts.
Period
12 month VAR
Nearby VAR
Signed volume VAR
Panel A: contemporaneous volatility–volume relation
2012
−0.01
−0.01
(t-value)
(17.57)
(4.62)
2011
−0.06
−0.02
(t-value)
(36.13)
(3.43)
2010
−0.10
−0.07
(t-value)
(64.34)
(11.54)
2009
−0.17
−0.08
(t-value)
(72.53)
(10.98)
2008
−0.11
−0.02
(t-value)
(25.89)
(25.62)
−0.01
(4.56)
−0.05
(13.37)
0.01
(7.38)
0.00
(3.24)
–
–
Panel B: contemporaneous volatility–duration relation
2012
0.06
0.06
(t-value)
(247.48)
(75.49)
2011
0.13
0.14
(t-value)
(237.04)
(79.31)
2010
0.13
0.11
(t-value)
(245.81)
(58.14)
2009
0.14
0.14
(t-value)
(217.00)
(73.21)
2008
0.40
0.01
(t-value)
(99.94)
(70.19)
0.04
(41.87)
0.13
(235.61)
0.13
(242.11)
0.14
(210.9)
0.01
(68.67)
Panel C: first order autocorrelation in volume
2012
0.23
0.22
(t-value)
(116.63)
(35.74)
2011
0.24
0.26
(t-value)
(129.10)
(47.30)
2010
0.18
0.22
(t-value)
(92.11)
(29.12)
2009
0.18
0.13
(t-value)
(90.43)
(5.38)
2008
0.19
0.22
(t-value)
(58.19)
(42.71)
0.26
(26.02)
0.43
(233.28)
0.16
(84.00)
0.43
(215.00)
0.31
(59.63)
Notes. This table reports the summary results of the trivariate volume–volatility–duration
VARs in respect to the Phase II contracts which expire December in each year 2008–11. In
particular, it reports the contemporaneous volatility–volume coefficient, ρ0, the contemporaneous volatility–duration coefficient, γ0, in the volatility equation of these trivariate
model specifications and the first-order autoregressive coefficient, δ0. The variables are
standardized with respect to duration and are in log-form to ensure that they remain
positive. Estimation is based on the tick-by-tick data for EUA futures contracts traded on
the ECX. The December 2008 and 2009 contracts are sampled 2nd January each contract
year to 15th December 2008 and 14th December 2009, respectively. The December 10
and 11 contracts are sampled from the 4th January of each contract year until 20th
December 2010 and 19th December 2011, respectively. Finally, December 2012 is sampled 3rd January 2012 to 17th December 2012. Only coefficients statistically significant
at a significance level of 5% are reported along with absolute t-statistics.
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
231
Fig. 1. Prices and trading volumes for Kyoto phase contracts. Notes. The figure presents EUA futures prices and cumulative trading volumes in respect to the December 2008, December
2009, December 2010, December 2011 and December 2012 expiration futures contracts.
Further, this finding is strengthened as the duration between trades
increases. The contemporaneous impact of the interaction term, duration by volume, on volatility is negative and significant across Kyoto
Phase contracts. Finally, the bivariate VAR findings show lagged effects
of information variables, volume and duration, on volatility.26 The
main trivariate VAR specification findings for the December 2008–12
contracts are reported in Table 9.27,28
By treating duration as an endogenous variable, we are able to obtain
a complementary understanding, relative to the bivariate specifications,
of the interaction between time elapsed between trades, volume and
volatility. Our results reported in Table 6 further highlight the level of
consistency relative to the duration based bivariate VAR results. Volume
continues to have a significantly negative impact on volatility contemporaneously. Moreover, the contemporaneous duration has a positive
and statistically significant impact on volatility. As per the previous discussion, the positive relationship is also economically significant as it indicates that liquidity traders prevail in the EU ETS market rather than
informed traders.
In the trivariate model specifications, there is also a positive relationship between lagged volume and volatility which is consistent with
the SIAH (see, Conrad et al., 2012). Again, consistent with the bivariate
results as well as the SIAH, lagged duration has a negative and significant impact on current volatility. Trade arrival times, durations, convey
private information in a similar manner to volume. A rational market
maker will interpret high volumes in combination with short durations
as evidence of trading by informed agents and will adjust quoted prices
26
To conserve space and given that the main emphasis is on the trivariate representation
we do not report the bivariate results for the complete Kyoto phase. Full details of bivariate
results for the Kyoto Phase contracts are available from the authors on request.
27
The Kyoto Phase estimates are for the year expiring samples. As can be seen from
Fig. 1, the vast majority of the trading activity occurs in the contract year of expiration.
Hence, all our Kyoto Phase analysis focuses on transaction data specific to the sample year
for the particular contract, e.g. a January 2008 to December 2008 sample is adopted for the
December 2008 contract.
28
We conduct similar analyses for each of the Pilot Phase contracts. The results are qualitatively similar to those obtained using the bivariate VARs in the Pilot Phase. Hence, to
conserve space, we refrain from reporting the results here. The results are available from
the authors on request.
accordingly (Easley & O'Hara, 1992). As a result, within the SIAH framework, the lagged information variable, duration, is expected to be economically and statistically significant in the volatility equation of the
trivariate VAR system. Given the significant evidence of lagged relationships, our results are at odds with the predictions of the MDH and are
better characterized by the SIAH.
In Fig. 2 we report the impulse response functions for the Kyoto
Phase (Phase 2) 2008 December expiration contract for 25 lags (trades),
including standard errors. The triangular decomposition (Cholesky)
is adopted to achieve orthogonalization. The diagonal graphs indicate
the effect of the shock on duration, volatility and volume on each of
the respective terms. There is some persistence in observed variables,
which is consistent with results in mature asset markets. For the offdiagonal entries, we find further evidence that duration does have an
impact on volume and volatility, and vice versa, but generally over
short time intervals.29
4.5. Robustness tests
As a robustness test in respect to findings in the trivariate VAR model
specifications, we estimate these models using data relevant for nearby
months. Nearby contracts are defined in this market context as those
trading in the month prior to their maturity, i.e. November for each contract. Our exclusion of observations on contract trading in the maturity
month avoids biases associated with unusual market activity characteristic of times approaching maturity as traders unwind or rollover positions. Given the fact that a substantial proportion of the trading
activity generally takes place in the contract specified on the nearby
month, the analysis of data taken only from this contract data may be
useful to further investigate our findings. This focus on the nearby
month confines the data analysis to the most actively traded period
29
The results for the impulse response functions, in respect to the December 2005 contract, for 25 lags (trades), including standard errors, is reported in Fig. 3. Findings for the
2006 and 2007 contracts are available from the authors on request. The impulse response
functions for Phases 1 and 2 are consistent, although the standard error bands are relatively larger for the earlier Phase 1 contracts given the relatively smaller number of
observations.
232
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
Fig. 2. Impulse response functions of the December 2008 contract. Notes. The figure presents impulse response functions, as well as the corresponding standard errors, for up to 25 lags for
the December 2008 maturity futures contract. The sample period is 28th September 2005 to 15th December 2008. The graphs presented on the top left to bottom right diagonal of the
figure trace out the time path of the effect of the shocks on duration, volatility and volume on each of the respective terms. The off-diagonal graphs trace out the time path of the effect
of shocks across duration, volatility and volume.
for each contract. The findings are reported in Table 7 in respect to the
nearby months for the December 2008 contract, with the summarized
findings presented in Table 8 for the complete set of Phase 2 contracts.
In particular, volume has a contemporaneous significant negative impact on volatility and duration is contemporaneously positively and significantly associated with volatility. There is also some evidence that
lagged volume increases volatility. Again, consistent with the bivariate
results and the full blown trivariate results, lagged duration has a negative and significant impact on current volatility. As discussed above, taking these findings together they support the SIAH and they indicate that
liquidity traders prevail in the current EU ETS market rather than informed traders.
A second robustness test returns to the larger samples studied previously, however with signed volume substituted for volume. The results
for the December 2008 contract are reported in Table 8, with the summary for all Phase 2 contracts reported in Table 9. The signed volume indicates whether the corresponding trade is a buyer initiated (+) or a
seller initiated (−) trade while the conventional volume measurement
which we adopt in the other specifications is always non-negative and
does not convey this information. The contemporaneous negative
relation between volume and volatility is no longer present. Notwithstanding, the contemporaneous duration has a statistically significant
positive impact on volatility. The positive contemporaneous relation between duration and volatility is economically significant as it indicates
that liquidity traders prevail in the current EU ETS market rather than
informed traders. In addition, consistent with the bivariate results and
the full blown trivariate VAR results, lagged duration has a negative
and significant impact on current volatility. Finally, as is established
by, inter alia, Lyons (1997) and Berger et al. (2005) for high frequency
exchange rate data, pronounced positive signed autocorrelations of
signed volume (i.e. order flow) results are evident at the tick-by-tick
level in our study. This high level of autocorrelation may indicate the
passing of a large trade among traders or it may reflect order splitting,
whereby a large trade is executed sequentially. Given the significant evidence of lagged relationships, our results are once again at odds with
the predictions of the MDH and are better characterized by the SIAH.
The results reported in Tables 6–9 do not support MDH and they indicate the prevalence of liquidity traders rather than informed traders in
the EU ETS market. Overall, the findings in both Pilot and Kyoto Phases
of the ETS suggest, across contracts, sampled data and accounting for
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
233
Fig. 3. Impulse response functions of the December 2005 contract. Notes. The figure presents impulse response functions, as well as the corresponding standard errors, for up to 25 lags for
the December 2005 maturity futures contract. The sample period is 22nd April 2005 to 19th December 2005. The graphs presented on the top left to bottom right diagonal of the figure
trace out the time path of the effect of the shocks on duration, volatility and volume on each of the respective terms. The off-diagonal graphs trace out the time path of the effect of shocks
across duration, volatility and volume.
signed volume, the prevalence of liquidity traders and the importance of
the SIAH.
liquidity trades are dominating any informed trader behavior. Informed
traders, as is often the case, may have a preference to trade on the over
the counter (OTC) market.
4.6. Carbon finance market implications
5. Conclusions
Our detailed high frequency examination of the EU ETS extends the
large number of recent studies examining carbon finance market developments using low frequency data. Exchange traded market activity,
and EUA futures in particular, have grown dramatically with the move
from Pilot to the Kyoto Phase. Besides the spike in 2006, mean duration
and volumes for the contracts examined have been reflective of a
market that is developing year on year. When we take account of
trade initiation using signed volume our results continue to indicate a
market being driven by buyer initiated trades rather than seller initiated
trades. Our conclusion is supported by findings in a recent study by
Kalaitzoglou and Ibrahim (2013). The authors highlight that the EU
ETS can be viewed as a buyer orientated market, where a greater proportion (in terms of volume) of trades are buyer rather seller initiated.
Our empirical results for duration and its relation with both volume
and volatility are consistent with this view. The implications are that
This paper investigates the initial stages of the European Union (EU)
emission trading scheme (ETS) introduced formally in January 2005
from a market microstructure perspective. The scheme instigated as
part of the EU agreement to cut worldwide emissions of carbon dioxide
(CO2) issues a restricted amount of emission allowances to firms on an
annual basis and subsequently allows firms to trade the emission permits they hold. This study covers the first phase of trading from 2005
to 2007 and the period from 2008 to 2012 of the second phase, which
coincides with the first compliance period of the Kyoto Protocol. The
main motivations for making inferences is to better understand the information assimilation process, liquidity and the degree of market efficiency during the phases of the emission trading scheme. We examine
the microstructure and trading behavior, in the ETS, in the context of
these objectives.
234
D. Bredin et al. / International Review of Financial Analysis 34 (2014) 222–234
Our key result is that significant market developments have taken
place, with reduced market frictions suggesting evidence of greater
levels of efficiency along the lines discussed in Dacorogna, Gencay,
Muller, Olsen, and Pictet (2001). This is reflected in the form of reduced
durations between trades, higher volumes traded and lower volatilities.
Secondly, bi-directional Granger causality is found in the trade variables, the empirical evidence is consistent with the sequential information arrival hypothesis (Copeland, 1976) rather than the predictions of
the mixture of distribution hypothesis (Clark, 1973; Tauchen & Pitts,
1983). Ours is the only study to take account of duration in relation to
the sequential information arrival versus the mixture of distribution hypothesis for the EU ETS. In addition, our evidence in favor of the sequential information arrival hypothesis is consistent with recent results on
the EU ETS reported by Conrad et al. (2012) and Chevallier and Sevi
(2011). Finally, our results indicate a negative contemporaneous relationship, accounting for duration effects, between volume and volatility.
The implication is that liquidity traders dominate any role played by informed traders (Easley & O'Hara, 1992). Notwithstanding the significant
developments, this result is not surprising for a market that is dominated by hedgers, see Kalaitzoglou and Ibrahim (2013).
A common finding from the VAR models indicates that there is a
negative contemporaneous relationship between volume and volatility
for all contracts. The implication here is that liquidity traders are dominating any informed trading behavior. This result is consistent with the
market view that informed traders, as is often the case, may have a preference to trade on the OTC market. If informed traders are active in the
OTC market, is this due to greater levels of depth, breadth and overall
market liquidity? We leave this question to be examined by future
research.
References
Alberola, E., Chevallier, J., & Cheze, B. (2008). Price drivers and structural breaks in
European carbon prices 2005–2007. Energy Policy, 36(2), 787–797.
Andersen, T. (1996). Return volatility and trading volume: An information flow interpretation of stochastic volatility. Journal of Finance, 50, 169–204.
Bagehot, W. (1974). The only game in town. Financial Analysts Journal, 27, 12–14 (22).
Benz, E., & Hengelbrock, J. (2009, May 5). Liquidity and price discovery in the European CO2
futures market: An intraday analysis. Carbon Markets Workshop, LSE.
Berger, D. W., Chaboud, A. P., Chernenko, S. V., Howorka, E., Iyer, R. S. K., Liu, D., et al.
(2005). Order flow and exchange rate dynamics in electronic brokerage system data.
IFDP 830. Board of Governors of the Federal Reserve System.
Blume, L., Easley, D., & O'Hara, M. (1994). Market statistics and technical analysis: The role
of volume. Journal of Finance, 49(1), 153–181.
Bredin, D., & Muckley, C. (2011). An emerging equilibrium in the EU Emissions Trading
Scheme. Energy Economics, 33(2), 353–362.
Chevallier, J. (2011). Macroeconomics, finance, commodities: Interactions with carbon
markets in a data-rich model. Economic Modelling, 28(1–2), 557–567.
Chevallier, J., & Sevi, B. (2011). On the realized volatility of the ECX CO2 emissions 2008
futures contract: Distribution, dynamics and forecasting. Annals of Finance, 7, 1–29.
Clark, P. K. (1973). A subordinated stochastic process model with finite variance for speculative prices. Econometrica, 41(1), 135–155.
Conrad, C., Rittler, D., & Rotfuß, W. (2012). Modelling and explaining the dynamics of
European Union allowance prices at high frequency. Energy Economics, 316–326.
Convery, F. J., & Redmond, L. (2007). Market and price developments in the European
Union Emissions Trading Scheme. Review of Environmental Economics and Policy,
1(1), 88–111.
Copeland, T. E. (1976). A model of asset trading under the assumption of sequential information arrival. Journal of Finance, 31, 1149–1168.
Dacorogna, M. M., Gencay, R., Muller, U., Olsen, R. B., & Pictet, O. V. (2001). An introduction
to high-frequency finance. USA: Academic Press.
Diamond, D., & Verrecchia, R. E. (1987). Constraints on short-selling and asset price adjustment to private information. Journal of Financial Economics, 18(2), 277–311.
Dufour, A., & Engle, R. F. (2000). Time and the price impact of a trade. Journal of Finance,
55(6), 2467–2498.
Easley, D., & O'Hara, M. (1992). Time and the process of security price adjustment. Journal
of Finance, 47, 557–605.
Ellerman, A.D., & Buchner, B. (2006). Over-allocation or abatement? A preliminary analysis
of the EU ETS based on the 2005 emissions data. Nota di Lavoro 139.2006. FEEM (also
listed as WP-2006-016 at MIT CEEPR).
Engle, R. F. (2000). The econometrics of ultra-high frequency data. Econometrica, 68, 1–22.
European Environmental Agency (2012, October). Greenhouse gas emission trends and
projections in Europe 2012.
Harris, L. (2003). Trading & exchanges: Market microstructure for practitioners. Oxford University Press.
Kalaitzoglou, I., & Ibrahim, B.M. (2013). Does order flow in the European carbon allowances market reveal information? Journal of Financial Markets, 16, 604–635.
Karpoff, J. M. (1988). Costly short sales and the correlation of returns with volume. Journal
of Financial Research, 11, 173–188.
Kyle, A. S. (1985). Continuous auctions and insider trading. Econometrica, 53, 1315–1335.
Lyons, R. (1997). A simultaneous trade model of the foreign exchange hot potato. Journal
of International Economics, 42, 275–298.
Manganelli, S. (2005). Duration, volume, and the price impact of trades. Journal of
Financial Markets, 8, 377–399.
Mansanet-Bataller, M., Pardo, A., & Valor, E. (2007). CO2 Prices, Energy and Weather. The
Energy Journal, 28, 73–92.
Mansanet-Bataller, M., & Pardo, A. (2008). CO2 prices and portfolio management. University of Valencia working paper (http://ssrn.com/abstract=1156975).
Mizrach, B. (2010). Integration of the global emissions trading markets. Rutgers University
working paper (http://ssrn.com/abstract=1621340).
Mizrach, B., & Otsubo, Y. (2014). The market microstructure of the European climate exchange. Journal of Banking and Finance, 39, 107–116.
O'Hara, M. (1987). Trading mechanisms and stock returns: An empirical investigation:
Discussion. Journal of Finance, 42, 554–555.
Paolella, M. S., & Taschini, L. (2008). An econometric analysis of emission trading allowances. Journal of Banking and Finance, 32, 2022–2032.
Redmond, L., & Convery, F. J. (2006). Determining the price of carbon in the EU ETS. Planning and environmental policy research series (PEP) working paper 06/09. School of Geography, Planning and Environmental Policy, University College Dublin.
Rittler, D. (2009). Price discovery, causality and volatility spillovers in European Union allowances phase II: A high frequency analysis. Discussion paper no. 492. Department of
Economics, University of Heidelberg.
Rotfuß, W. (2009). Intraday price formation and volatility in the European Emissions
Trading Scheme: A first analysis. Working paper. Mannheim: ZEW.
Tauchen, G. E., & Pitts, M. (1983). The price variability–volume relationship on speculative
markets. Econometrica, 51, 485–505.
Tsay, R. (2009). Conditional duration models. In T. C. Mills, & K. Patterson (Eds.),
Handbook in econometrics, vol. 2. (pp. 1004–1024). Palgrave (Chapter 21).
Xu, X., Chen, P., & Wu, C. C. (2006). Time and dynamic volume–volatility relation. Journal
of Banking and Finance, 30, 1535–1558.