LNG Pricing Differences across the Atlantic,
a Comparison between the United States
and Europe
Micaela Ponce1
Virginie Krone
Universität Potsdam2
Universität Potsdam1
[email protected]
[email protected]
WORK IN PROGRESS
Abstract
Projections for the energy market worldwide tend to coincide in one aspect: the increasing importance
of natural gas as a source of energy. Although natural gas seems to be a key player in the future world
energy mix, there are significant differences in liquefied natural gas (LNG) import prices in each of the
main natural gas markets. Our analysis focuses on understanding the price difference between two of the
major natural gas markets: The United States and Europe. Therefore we apply a cointegration analysis to
derive the long-run main determinants of LNG prices and apply an SVAR analysis to study how the same
determinants have different impacts in both regions.
JEL-Clasification: L95, C32, Q40
Keywords: LNG import prices, natural gas, structural VAR, price determinants
1 Thanks
are due to Sebastian Nick and Dr. Till Strohsal (Universität Potsdam) for their helpful discussions and advice, and
the participants of the IAEE 2014 in New York for their feedback.
of Economics and Social Sciences, Chair for Economic Policy, August-Bebel-Str. 89, D-14482 Potsdam, Germany.
2 Faculty
1
1.
Introduction
Projections and future developments in energy markets are of major importance to economists
and policy makers worldwide. Each of the World Energy Outlook (IEA), the International Energy
Outlook (EIA) and the Energy Outlook (BP), emphasize the strong increase in energy demand
over the coming decades, driven by a steep population rise especially in emerging economies such
as India, Brazil and China. The worldwide increase in primary energy consumption directs our
attention towards natural gas. Although substantial changes in the worldwide fossil fuel mix are
less realistic in a nearby time horizon, and given the lack of new energy sources, mild changes on
the relevance of the constitution of the energy mix are possible.
According to the IEO by the EIA, renewables provide the biggest growth in the 2040 energy mix.
Fossil fuels, on the one side, are affected by stricter climate policies which cause a reduction in the
shares of coal and oil liquids in the energy mix. Consequently, natural gas is the only fossil fuel
that increases its share as a primary source of energy, given its low environmental impact and the
availability of unconventional gas. Natural gas becomes the world’s fastest-growing fossil fuel
and its importance increases in the world’s energy mix of 2040 reflecting an approximate equal
share as coal and oil. A similar development is observed in BP’s Energy Outlook, where natural
gas is the fastest growing fossil fuel and its worldwide consumption increases at an annual rate of
2%. It is in this context of special significance for emerging and import dependent countries, that
natural gas markets and their pricing mechanisms need to be well understood.
Contrary to the oil market, natural gas markets have not gained as much attention from research.
Historically, coal and oil have played a bigger role in the energy mix. The linkage between oil and
economic growth is a well-studied one throughout the energy economics literature (see Mork et al.
(1994), Papapetrou (2001), Barsky and Kilian (2004), Jiménez-Rodríguez and Sánchez (2005), Lardic
and Mignon (2008) as well as the factors altering the oil price Hamilton (2003), Hamilton (2008),
Breitenfellner et al. (2009), ECB (2008) or OECD (2004)). Acknowledging the increasing relevance
of natural gas over the future decades and the importance of energy as a source of economic
growth3 , we focus on understanding how LNG import prices are determined in different world
regions. Therefore we analyze two of the main natural gas markets, the Unites States and the
European Union4 , and empirically determine what impact the same determinants have on LNG
import prices on both regions. For this purpose, we use time series data and apply a structural
3A
vast number of studies focus on the linkage between economic activity/growth and energy. Some of the latest references
are: Belke et al. (2009), Lee (2005), Lee and Chang (2008), Narayan and Smyth (2008), and Apergis and Payne (2009).
the term European Union, Europe and EU is used as a synonym in this entire paper. For the purpose of this
4 Where
exercise, the European Union region is composed out of the EU-27 countries.
2
vector autoregressive model (SVAR). Although the Atlantic basin is not considered an interesting
trade market now a days, due to the fact that the volumes being traded are low in comparison to
the Pacific basin, we believe that the recent shale gas developments are changing this point of view.
Comparing Europe to the US seems a natural exercise considering that the model for restructured
gas markets in Europe is based on the positive experiences of the US natural gas markets.
In section 2 of this paper, we identify theoretically the determinants of LNG import prices in
each of the regions, section 3 presents the empirical model and is followed by section 4 where we
present the applied data. Section 5 presents the results and section 6 concludes.
2.
Identifying Potential Determinants
The characteristics of global gas markets have experienced substantial changes over the past three
decades marking a clear difference in the pricing mechanisms in every regional market. While the
North American market developed towards a gas on gas competition (GOG) mechanism based on
its liberalized market and the advanced development of natural gas hubs, the European natural
gas markets are a mix of oil price escalation (OPE) 80% in 2005 to 42% in 2013 5 and gas on gas
competition (from 15% in 2005 to over 53% in 2010). In the past 10 years a development towards
GOG and away from OPE is observed particularly due to the recent introduction of spot price
indexation in long term contracts (IGU, 2014). This observation can be extended worldwide. In
contrast to the existence of an international market for crude oil with similar price patterns across
regions, regional natural gas prices, in particular for LNG differ substantially. Figure 1 presents
LNG landed prices around the world in US Dollars per MMBtu. Here the price difference across
regions seems evident with prices in the United States (around US$ 3/MMBtu) three times lower
than in Europe (US$ 10 / MMBtu) and five times lower than in Asia (US$ 15/MMBtu) and Latin
America.
This regional price disparity motivates our paper. We utilize the oil market literature on price
determinants as a starting point for the determination of explanatory variables for LNG import
prices in natural gas markets. The link between the two resources and the oil price building
mechanism, are examined by several authors (i.e. Villar and Joutz (2006),ECS (2007) Brown
and Yücel (2008)). Most of the studies in the oil price literature analyze time spans where oil
prices show high volatility and an abrupt, unexpected change in levels (i.e.OECD (2004), Möbert
(2007)) with a strong focus on market structures and the role of OPEC. Additionally, more general
market fundamentals on the supply and demand side are included. Price movements in the crude
5 OPE
is a mechanism where the price is linked to a base price plus an escalation clause to other competing fossil fuels like
oil, heating oil, etc.
3
Figure 1: Worldwide LNG Landed Prices as of July 2014.
Source: FERC (2014)
oil market are analyzed in some studies considering financial market factors such as financial
futures contracts and exchange rates (Hamilton (2008), Möbert (2007) and Breitenfellner et al.
(2009)). As part of our approach we transfer some of the oil price determinants to the natural gas
market considering that both fossil fuel markets have similarities and differences. Natural gas
is heterogeneously located worldwide and much more geographically diversified than oil. It is
used mostly for electricity generation, industrial purposes as well as residential and commercial
purposes. In the past years it has found an increasing use in transportation especially in emerging
markets (OECD/IEA, 2013).
One of the main differences along the value chain of natural gas in comparison to oil, is its
transportation due to its physical properties. Natural gas can be transported on mainland through
pipeline and across longer distances over seas by LNG tankers. In both cases, infrastructure plays
an important role but in the LNG case importing and exporting terminals have to be built and in
operation to be able to trade natural gas. At the importing terminal the liquefied gas has to be
regasified for its further transportation downstream.
4
2.1.
Fundamental Market determinants
Based on previous work in natural gas markets and the oil market literature (i.e. Nick and Thoenes
(2013), The Energy Journal Vol. 34, No. 3 and previously mentioned Breitenfellner et al. (2009),
Möbert (2007)), we focus on four main groups of variables for two LNG importing regions in the
Atlantic basin. The first two are fundamental determinants of the price building process such as
supply and demand. The third group is composed of financial market factors and the fourth is related
to external events. Consequently we describe the determinants in each of the above identified
groups.
2.1.1
Supply
The application and combination of horizontal drilling and hydraulic fracturing altered significantly the natural gas supply side in the US at the beginning of the 21st century . A country
highly dependent on imports, started traveling down the road of self sufficiency based on what is
called today the "shale gas boom". Indigenous production increased from 2004 to 2013 at an average
rate of 3% per year and a total of 28%. The increase in indigenous production reduced transport,
and regasification costs significantly, compared to LNG because of the shorter way of distribution
to the end customer. Unconventional shale gas production and technological advances are still
expected to lead to higher supply rates in the United States.
In Europe only a few production locations are currently active and indigenous production in this
region comes mainly from the Netherlands and the United Kingdom, covering in 2012 around 40%
of the natural gas imports. In addition, Europe relies strongly on pipeline imports from Norway
and Russia, and LNG imports from Qatar, Algeria and Nigeria and shows a similar trend as in the
US of source diversification (See Figure 4 and 5) (BP, 2013).
During the past two decades, the United States have diversified its LNG importing sources moving
away from a strong dependency on Algerian gas during the 1990’s. Although today more players
are available to import from (i.e. Qatar, Yemen, Norway and Nigeria), the US relies strongly on
Trinidad and Tobago importing more than 70% of the LNG. In 2012 the US relied on only 5 LNG
suppliers. Given the substantial indigenous production only a minor share if imports are required
to meet domestic demand. Historically the US relied only on two pipeline natural gas suppliers:
Canada and Mexico. The number of natural gas suppliers (international suppliers and domestic
producers) within a country offers valuable information for market structures. An increasing
number of suppliers leads to a lower market concentration and towards a more competitive
market. In the oil market literature the high market concentration of the OPEC members plays an
important role for the price determination (i.e. Hamilton (2008)).
5
2013
2005
1997
0%
Trinidad and Tobago
10%
20%
Algeria
30%
Qatar
40%
Australia
50%
Egypt
60%
Malaysia
70%
Nigeria
80%
Norway
90%
Oman
100%
UAE
Yemen
Figure 2: US LNG Imports by country of origin
Source: EIA
On the other side of the Atlantic, reality is different. The low volume of domestic production
makes Europe dependent on foreign natural gas. In 2012 Europe relied on 12 pipeline suppliers
(the largest suppliers being Russia and Norway), 6 main LNG suppliers (Algeria, Qatar, Nigeria,
Egypt, Trinidad and Tobago, Peru) and a very small share covered by other OECD countries (See
Figure 4).
Storage capacities and levels play a strategic role along the natural gas supply chain (Albrecht
et al., 2014) as these help modulate the flow of supply to the network system. It is the volatility
characteristics of natural gas markets, that make storage capacities and levels important for price
determination and therefore storage levels are particularly interesting for price developments.
2.1.2
Demand
The demand of LNG in the United States has decreased over the past six years, since the application
of drilling technologies such as hydraulic fracturing, which unlocked unconventional gas deposits
that had been viewed as non-profitable in the past. As a result, the increased domestic natural gas
6
MMcm
US Production
700,000
600,000
500,000
EU Imports
400,000
300,000
EU Production
200,000
US Imports
100,000
0
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Figure 3: Natural Gas Imports vs. Indigenous Production in the US and Europe
supply has led to a decrease on imports volumes on LNG and pipelines (See Figure 3 above). The
oversupply of natural gas in the market drove domestic prices down to a low of 1.83 US$/MMBtu
on September 4th 2009 (EIA database 6 ).
The "shale gas revolution" and the resulting decrease in prices, triggered also an increase in
consumption after 2009, especially in the industrial sector (EIA, 2013). The use of natural gas in the
electricity generation accounts for the highest share of demand in the US. Together with industrial
consumption they accounted for 60% of the natural gas end use in 2013 (EIA Database7 )The
consumption level reflects a regular and repeating pattern throughout the year showing higher
consumption levels during winter and lower levels throughout the summer. This is consistent with
the consumption of natural gas as heating source especially in the commercial and the residential
sectors.
The abundance of shale gas has also changed future perspectives on the import of LNG. Fourteen
regasification plants were planned in the US. Since 2009, twelve of the planned projects have been
canceled, and instead liquefation plants are under construction or in the planning process. This
illustrates the intention of the US to base itself on the shale gas boom in order to become a natural
gas exporting country. In Europe, on the other side, although projects have been canceled during
the past years no structural change is expected. Europe has no liquefaction plants and currently
6 Natural
7 Natural
Gas Spot and Futures Prices (NYMEX) accessed at: http://www.eia.gov/dnav/ng/ng_pri_fut_s1_d.htm
gas Consumption by End Use. http://www.eia.gov/dnav/ng/ng_cons_sum_dcu_nus_m.htm
7
2012
2009
2005
1997
0%
Algeria
10%
Egypt
20%
Lybia
30%
Qatar
Malaysia
40%
Nigeria
50%
Other OECD
60%
Oman
70%
Peru
80%
Trinidad and Tobago
90%
100%
United Arab Emirates
Figure 4: European LNG Imports by country of origin
Source: IEA
21 regasification plants are operational.
The import capacity can be considered a potential determinant because the amount of LNG volume
delivered is constrained by the liquefaction/regasification capacities of the involved countries.
Other aspects of interest on the demand side are environmental factors. The EU Emissions Trading
System (ETS) trades carbon dioxide emission allowances by cap and trade to reduce greenhouse
gases through financial enforcement for industries. Therefore, the CO2 price affects the cost scheme
of the industry due to the use of fossil fuels.
A comparable system for the US is the Acid Rain Program aiming to reduce sulfur dioxide (SO2 )
and nitrogen oxides (NOx) by trading allowances imposed nationwide in 1990.8 The cap and
trade system leads to higher costs for the emitters and triggers an increase in demand due to
environmental advantages of natural gas given its substitution properties as fossil fuel. So both of
these systems could potentially influence natural gas and LNG demand.
8 U.S.
Environmental Protection Agency, Retrieved: 11 May 2014 from
http://www.epa.gov/airmarkets/progsregs/arp/s02.html
8
2012
2009
2005
1997
0%
10%
20%
Russian Federation
Denmark
Algeria
30%
Former USSR
France
Azerbaijan
40%
50%
Norway
Germany
Iran
60%
70%
80%
Belgium
United Kingdom
Other
90%
100%
Netherlands
Other OECD
Figure 5: European Pipeline Imports by country of origin
Source: IEA
2.2.
Financial Market
Gas to gas competition based on spot market pricing is a key feature for pricing natural gas in the
US. The European agenda to develop competitive markets also foresees the development of liquid
trading places producing price signals to market participants. The flexibility of "self contracted"
LNG volumes can produce arbitrage potential to market participants based on regional spot
market price differences.
Futures stay in tight relationship to spot prices. In so called term markets a limited number of
standardized futures contracts are traded. They reflect the average risk level of the active market
participants at the moment of transaction for a time period in the future (Bros, 2012). Future
contracts and the development of prices, are supposed to have an impact on behavior and decisions
for import and export volumes and therefore trade. Therefore futures could also be interesting
determinants for LNG import prices.
The long-term relationship between natural gas prices and crude oil prices was empirically well
9
MMcm
US$/MMBtu
120,000
9.00
2008
7.76
Pipeline Imports
8.00
100,000
7.00
80,000
6.00
5.00
60,000
4.00
40,000
2009
3.57
Wellhead price 3.00
(right axis)
LNG imports
20,000
2012
2.59
2.00
1.00
0
0.00
1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
Figure 6: US prices at the wellhead vs. Import volumes
Source: EIA
studied in the past years based on cointegration analysis (see i.e. Villar and Joutz (2006), Brown
and Yücel (2008), Asche et al. (2013), Hartley and Medlock-III (2014) ). Oil-indexed long-term
contracts, common in continental Europe, are a reason for including crude oil spot prices as a
determinant of LNG prices. Another reason is the substitution relationship of oil and natural gas
in the residential heating sector.
2.3.
External Factors
Sudden events of political nature like the reduction of natural gas supply or unstable political
conditions in the exporting country, can influence negatively LNG import prices. The UkraineCrimea-crises or the Arab Spring are just two of the most recent events and rather unpredictable.
Other uncertain events can take place in form of acts of nature. Hurricanes near gas extracting
facilities threaten the production and can lead consequently to high prices due to a supply disruption. Although climate centers predict weather phenomena with forecast methods, uncertainty
in damages and aftermath in matters of prices is always possible. Another weather related
determinant is heating degree days. Natural gas is used in the residential and commercial sectors
mostly for heating purposes, in the industrial sector for dehumidification and cooling purposes,
and in the transport sector as fuel (EIA, 2012). In the US, indigenous production increases at full
10
capacity during winter, because of the higher demand triggered to the heating period (Albrecht
et al., 2014). Between September and December, higher prices for natural gas are observable due
to the impact of seasonal demand in Europe.
3.
Empirical Strategy
Before focusing on the main methodology approach of this paper (the SVAR) an upfront analysis
should provide empirical evidence for the inclusion of some of the variables. Based on a cointegration analysis we deliver a framework to identify a a long-term relationship among the variables.
This is based on Johansen test for cointegration (Johansen, 1991) which allows us to proceed to the
actual analysis of the paper.
3.1.
Structural VAR Model
In order to conduct the proposed analysis, the general structural vector autoregressive (SVAR)
model is introduced. This method allows us to understand how shocks in each of the prevoiusly
proposed determinants affect the LNG import price in each of the analyzed regions. Therefore,
for both studied regions (EU/USA) a SVAR analysis is equally conducted. The results of the
SVAR are later on represented in impulse response functions (IRFs) and compared to each other
in order to explain the different effects that the determinants have on the LNG import price in
either region. The dynamic relationship between each determinant and the LNG import price, can
be estimated and interpreted by a structural representation. The SVAR methodology allows us
also to implement endogenous or exogenous variables. Vector autoregressive (VAR) models, first
proposed by Sims (1980), were developed for empirical macroeconomic analysis. Since then they
have been widely used with economic time series data.
A VAR is an n-equation, n-variable linear model, where each variable is explained by its own
lagged values, additionally to past and present values of the other included variables (Stock and
Watson, 2001). An extension of this model is the SVAR, which allows for economic considerations
and causal interpretation due to a system of structural equations. Following Sims (1986) and
Blanchard (1990), Lütkepohl (2011) leads to a formal scheme to first specify and then estimate
the reduced form SVAR model. He shows how to check for model adequacy and estimate the
structural analysis via impulse response functions (IRFs). In our analysis, the restrictions are
imposed as short-term factors because the LNG import price is flexible and economic relationships
are evaluated as short impulses. The shock is therefore temporary in our estimation.
In order to compare the influence of determinants on the European and US LNG import price,
at first the time series data has to be controlled for covariance stationarity to fulfill the SVAR
11
requirement. Testing each identified plausible determinant for unit roots is necessary to make
sure that the data is in the end stationary and lead to meaningful interpretation results. However,
based on (Kilian and Murphy, 2014), incorrectly differencing a variable would cause the impulse
response estimates to be inconsistent. Therefore even though stationarity is a requirement in
SVARs, in some cases as this one, we could specify our model in levels although not all variables
are stationary. Possible tests for detecting unit roots are the augmented Dickey-Fuller (ADF), the
Philipps-Perron (PP) and the Kwiatkowski-Phillips-Schmidt-Shin (KPSS) test. See section 4.1 for
the test results of the considered determinants.
Identification Scheme
As stated previously a VAR is a K dimension stationary process in which each K variable is
explained by its past values and in some cases by additional exogenous variables. A VAR( p)
without exogenous variables can be formally written as:
yt = v + A1 yt−1 + ... + A p yt− p + et
(1)
Where yt is a (Kx1) vector of observable time series variables. The A’s are (KxK) coefficient matrices and et a (Kx1) white noise matrix with et ∼ (0, ∑e ), hence a mean zero serially uncorrelated
error term.9 In such a model: E(et ) = 0; E(et , et0 ) = ∑e and E(et , es0 ) = 0
Where ∑e is the white noise covariance matrix which contains all the information about contemporaneous correlations among the K variables in yt . Any stable VAR process10 , can be expressed as a
vector moving average where yt is expressed in terms of past and present shocks such as:
∞
yt = µ + ∑ φi et−i
(2)
i =0
Where µ is the stationary mean of yt , and


 IK
φi =

 ∑i φ A
j =1 i − j j
if i = 0
(3)
if i > 0
The φi term is determined recursively and presents the impulse response functions. The drawback
of a VAR model is that it cannot provide causal interpretation of φ terms, since the error terms are
contemporaneously correlated, making impossible to hold all other impacts constant. To overcome
this flaw, equation 2 needs to be rewritten in terms of mutually uncorrelated shocks.
∞
y t = µ + ∑ θi wt −i
i =0
9 In
further context also named structural shock or structural innovation.
is stable when its variables are covariance stationary and and none of the autocorrelations are too high.
10 VAR
12
(4)
Where θi = φi P; wt = P−1 et and ∑ = PP0 . With such a matrix P, wk would be mutually orthogonal
achieving so the causal relation holding all other factors constant. One way to find the P matrix
is through estimating the triangular SVAR proposed by Sims (1980). In other words restricting
∑e to be a diagonal matrix. This would set the diagonal of the covariance matrix to 1, so that the
elements of ut are uncorrelated. The triangular VAR model can be formally represented as:
yt − v + A1 yt−1 + ... + A p yt− p = et
(5)
A(yt − v − A1 yt−1 − ... − A p yt− p ) = Aet = But
(6)
or as:
where A follows a triangular form (Sims, 1980). A and B are non-singular matrices of parameters
to be estimated, et is a (Kx1) vector of disturbances with ut ∼ N (0, IK ). Sufficient constraints
have to be placed on A and B so that P is identified. The A matrix sets restrictions on the
directly contemporaneous relations among the observable variables while the B restricts the
contemporaneous structural shocks. The number of restrictions required for identification in the
matrices is determined by the type of model chosen (A-Model, B-Model or AB Model). For A
and B Models the number is determined by K (K + 1)/2, in AB-Models the number of restrictions
is given by: 2K2 − 1/2K (K + 1) (Lütkepohl, 2007). Based on the restrictions, the IRFs are now
orthogonalized and they can provide causal interpretation. The short-run SVAR approach chooses
b−1 B
b to identify the casual IRFs. In the short-term models the constraints are applied directly
P=A
to the A and B matrices and then A and B interact with the estimated parameters of the underlying
b̄ −1 B.
b̄ −1 is the matrix
b= A
b Where A
VAR (StataCorp, 2011). The long-run SVAR chooses P = C
of estimated long-run effects and in such a model the constraints are placed on the long term
accumulated effects of the innovations.
The results for the US data and the EU data are expected to differ in their intensity of reactions and
a comparison of the different estimations is important towards the examination of LNG import
price differences.
The impulse response functions measure the effect of a shock to an endogenous variable on itself
or on another endogenous variable over a certain time path.
4.
11
Data
Following section 2, where the variables of interest were identified, we proceed to describe the
utilized data set for our estimation purposes. The applied monthly data starts in January 1997 and
11 For
more detailed information see Lütkepohl (2007).
13
ends in December 2012 and comprehends 192 observations; all of which are publicly available.
The data used in the US model was provided by the US Energy Information Administration (EIA),
the International Energy Agency (IEA), the US National Climatic Data Center (NCDC) and the
Federal Reserve Economic Database (FRED). The European data is available from the EIA, IEA
and the European Commission (Eurostat). The European data comprises the European Union as a
whole12 .
The data set consists of the LNG import price, natural gas gross withdrawals as an indicator
of domestic production, the total natural gas consumption, the industrial production index, the
Henry Hub natural gas spot price, an average of the European spot prices, indicators for the crude
oil spot prices (WTI, and Brent) and heating degree days. All the variables are gathered for the US
and the EU and all of them in natural logarithms.
Table 1 summarizes all variables used for the estimation including their sources, units and expected
impacts on the particular LNG import price, while Table 2 shows the descriptive statistics for the
utilized variables. The interpretation of the expected signs is as follows, a ’+’ means an increase of
the determinant and reflects an increase in the LNG import price; ’-’ has a reverse effect on the
LNG import price. Determinants in monetary units are given in US-dollars13 .
The dependent variables to be explained by our models are LNG import prices for the US the
EU. As defined by the EIA, import prices are typically reported as ’landed’ or ’tailgate’. Landed
LNG prices refer to received at the terminal and tailgate is the price called after the regasification
process. Consequently, the reported LNG import price differs in point of location and is calculated
as average of both(EIA, 2012). Indigenous production, as an LNG import price determinant
of supply, is interesting particularly in the US due to the shale gas production that increased
significantly domestic natural gas production and triggered a structural change in importing LNG.
Indigenous production plays a minor role in Europe due to the fact that natural gas reserves
have been reduced over the past 50 years. The macroeconomic variable, industrial production
index, serves as proxy for industrial natural gas consumption and therefore economic growth. The
growth factors for the US and EU are rather stable but the LNG trade is predicted as increasing,
so it is interesting to include the IPI to see whether the impact is equal or significantly different in
both areas. In the US, it measures the amount of output from the manufacturing, mining, electric
and natural gas sectors. The reference year for the index is 2002 at a level of 100. The EU data is
working-day and seasonally adjusted. Calculations are made with 2010 as the base year.
The last determinant in the fundamental factor group is natural gas consumption. It reflects the
demand best and is important for the price building process. Limitation in data availability for
12 The
EU consisted from January 1997 until April 2004 of 15 member states, until 2006 of 25 and until June 2013 of 27.
in other units i.e. Euros are converted by the average of the monthly exchange rate.
13 Prices
14
Table 1: Determinants
Variable
Description
Source
Periodicity
Expected
Impact
Unit of
measurement
Dependent Variable: LNG Import Price
PLNGt,US
US LNG Import Price
IEA
monthly
US$/MMBtu
PLNGt,EU
EU LNG Import Price
IEA
monthly
US$/MMBtu
Covariates: Supply (St,i )
Prodt,US
Production: Natural Gas Gross
EIA
monthly
−
MMcm
Prodt,EU
Withdrawals
IEA
monthly
−
MMcm
Demand (Dt,i )
Const,US
US Natural Gas Total Consumption
IEA
monthly
+
MMcm
Const,EU
EUNatural Gas Total Consumption
IEA
monthly
+
MMcm
IPIt,US
US Industrial Production Index
FRED
monthly
+
Index
IPIt,EU
EU Industrial Production Index
Eurostat
monthly
+
Index
Financial Market (FMt,i )
SPt,US
Henry Hub Natural Gas Spot Price
EIA
monthly
+
US$/MMBtu
SPt,EU
Average European Natural Gas
EIA
monthly
+
US$/MMBtu
EIA
monthly
+
US$/barrel
EIA
monthly
+
US$/barrel
Spot Price (NBP, ZEE, TTF)
WTIt
Western Texas Intermediate
Spot Price FOB
Brentt
European Brent
Spot Price FOB
Seasonality (Set,i )
HDDt,US
US Heating Degree Days
NCDC
monthly
+
days
HDDt,EU
EU Heating Degree Days
Eurostat
monthly
+
days
LNG consumption in total for the European market forces us to use consumption for natural gas
in general. Consumption reflects all natural gas consuming sectors.
The financial market determinant US Henry Hub (the biggest spot market in the US) is set as
reference for US natural gas spot prices. The EU-spot price is created from data for the United
Kingdom (UK), Belgium (BE) and the Netherlands (NL). The three spot markets are important
for financial price building and build the core within the European market. The variable SPeu is
based on own calculations and consists of the volume weighted average values of the three hubs
NBP (UK), ZEE (BE) and TTF (NL). The long term relationship among the variables is founded in
the cointegration relationship as examined i.e. by Stern and Rogers (2011),Siliverstovs et al. (2005)
15
Table 2: Descriptive statistics
Variable
Obs.
Mean
Std. Dev.
Min.
Max.
US variables
logprice
192
1.508
.435
.712
2.545
logprod
192
14.54
.074
14.384
14.759
logcons
192
7.540
.184
7.221
7.968
logipi
192
4.514
.058
4.346
4.613
loghh
192
1.444
.487
.542
2.596
logwti
192
3.766
.634
2.429
4.896
loghdd
192
5.015
1.701
.693
6.895
logprice
192
1.548
.503
.587
2.408
logprod
192
10.084
.219
9.678
10.466
logcons
192
10.570
.327
9.912
11.189
logipi
192
4.609
.059
4.471
4.742
logspot − eu
192
1.452
.641
.298
2.522
logbrent
192
3.744
.696
2.284
4.888
loghdd
192
5.131
1.103
1.945
6.436
EU variables
14 .
The number of observations for each of the three European spots variables is different due to
the time the hub was established. The monthly European spot price is estimated as the mean of
the three spot variables, for the NBP (established 1996) 192 values are used, the ZEE (since 2000)
has 166 values and the TTF (since 2003) 122 (Stern and Rogers, 2011).
Oil prices that are taken into account, are related to previous research for cointegration between
natural gas and oil prices for both markets. We utilize the Brent as the reference oil price in Europe
and the West Texas Intermediate as the reference crude oil price for the US market.
Since LNG is used also for heating purposes HDD is used here as a proxy for heating periods.
The quantity of days covers the whole US and the EU-27 states respectively. The determinant
HDD captures the seasonality within other determinants in the estimation. The interpretation for
heating degree days is different in both analyzed regions due to statistical calculations. In the US,
heating degree days are estimated from the National Climatic Data Center (NCDC) as follows: the
base value is 65◦ F and one degree day is defined as the difference between the base value and the
average daily temperature. The values are added throughout a month.
14
See the graphical overlay of the three time series in annex A figure 13.
16
In the EU-27 (Eurostat) heating degree days are estimated as: the severity of the cold in a specific
time period taking into consideration outdoor temperature and room temperature. The calculation
is: (18◦ C − Tm)∗ days= 0 with Tm as average temperature and 15◦ C as heating threshold.
4.1.
Unit Root Tests
Table 4 shows the results of the unit root test for the US and EU determinants. The null for the
existence of unit roots is tested by the Augmented Dickey-Fuller test (ADF) using the Schwert
criterion for lag length selection. A graphical analysis of each time series provides a hint for
adding a trend term or not to the tests (see annex C for detailed figures). For those cases where a
deterministic trend is not empirically proven but still a trend could be identified on the variable
plot, a stochastic trend is included in the unit root test. Table 4 indicates that for the US model, the
log of natural gas production (logprod) and the log of WTI (logwti) follow a stochastic trend. For
the EU model, stochastic trends where included for: the log LNG import price (logprice), the log of
the average European natural gas spot price (logspot-eu) and for the log of Brent prices (logbrent).
Overall, the results suggest that the null cannot be rejected (at the 5% significance level )for the
following variables in the US: logprice, logprod, loghh, logwti and logipi. Conducting the same
test on its first differences, we conclude these variables are non-stationary unit root processes
I(1). The I(1) variables from the US model are tested for cointegration following Johansen (1991)
with the exception of logprod, which from plotting the data we can conclude has no unit root 15 .
The Johansen test on cointegration finds one cointegration relation among all the I(1) variables
indicating a common long-run relationship among the LNG import price, the Henry Hub price,
the WTI price and the industrial production index. This result confirms the inclusion of the
variables on the SVAR.
On the right side of the table, the same unit root test is applied for the European model variables.
The results on the European model vary strongly from those of the US. All variables seem to be
I(0) with only the exception of logipi.As a consequence no cointegration analysis is conducted for
the EU model. Whereas the results for the US data are intuitive, European data is less stringent.
Thus could be a consequence of data quality issues or aggregation methodologies.
5.
Empirical Results
The model presented in section 3 is specified using the data in section 4. The number of lags
included in the model was determined by minimizing the Schwarz’s Bayesian information criterion
15 Throughout
the unit root literature it is well known that unit root test have low power when processes are near-unit roots.
See Elliott (1998)
17
Table 3: Results from Unit Root Tests
US variables
trend
lags
ADF
EU variables
trend
lags
ADF
logprice
-
1
-2.041
logprice
+
3
-3.58**
logprice
-
3
-0.720
D.logprice
-
0
-20.39***
D.logprice
-
2
-5.716***
logprod
+
6
-0.527
logprod
-
5
-10.675***
D.logprod
-
5
-8.764***
D.logprod
logprodsa
+
6
-0.527
logprodsa
D.logprodsa
-
5
-8.764***
D.logprodsa
logconssa
-
5
-3.760***
logconssa
–
D.logconssa
D.logconssa
–
-
6
-3.895***
–
-
6
-3.144**
–
logipi
-
4
-2.774*
logipi
-
3
-2.762*
D.logipi
-
3
-3.331**
D.logip
-
2
-4.519***
loghh
-
1
-2.286
logspot − eu
+
4
-4.001***
D.loghh
-
0
-13.207***
D.logspot − eu
logwti
+
1
-3.244*
logbrent
+
D.logwti
-
0
-10.770***
D.logbrent
-
loghdd
-
6
-14.258***
loghdd
-
–
1
-3.426**
–
6
-14.213***
*, **, *** indicate significance at the 10%, 5% and 1% level.
D. indicates the variable is on first difference.
+ indicates a trend is added.
(SBIC). The SBIC is minimized at two lags for the US and at one lag for the European model. The
identification of the number of lags included in the VAR analysis is important for the inferences
afterwards (Becketti, 2013).
5.1.
Identification of Short-term Restrictions
One of the advantages of structural VARs is the inclusion of short-term restrictions manually.
Economic theory helps to restrict the identification matrix. The structural model is non-recursive
and overidentified by two restrictions. The identification matrix is the same for the European and
for the US LNG natural gas market. In spite of the restriction, the effects of each determinant
should be different in either region. A different outcome for the interaction between two variables
is expected for example for production and LNG import prices in Europe because of the market
structure and dependency on fewer suppliers and because of the fact that the USA is the largest
18
natural gas producer worldwide. Each row in the SVAR stands for an explanatory equation. The
heating degree days variable hdd is ranked first because of its exogenous character. Therefore the
influence of every other determinant in that equation is restricted to zero. Indigenous production
prod is not instantaneously influenced by consumption or the overall production index ipi.
The vector of structural innovations is et = eth etpd etp etsp eto etipi etc ’.
pd
p
where eth is the shock in heating degree days, et , stays for production et , a shock in the LNG
p
sp
import price itself is et , in the spot price et , for the respective crude oil spot price eto , for the
ipi
industrial production index et and one time shock for consumption etc .

1


α21,t


 0


α41,t


α51,t


α61,t

α71,t
0
0
0
0
0
1
α23,t
α24,t
α25,t
0
α32,t
1
α34,t
0
α36,t
0
0
1
0
0
0
α53,t
α54,t
1
0
α62,t
α63,t
α64,t
0
1
α72,t
α73,t
α74,t
0
α76,t
0
 
hdd


hddt−1/2


eth


 
 



 pd 


 
 et 
 prodt−1/2 
0   prod 

 
 



 sp 


 
 et 





spt−1/2
sp
α37,t
 


 


 


 
0  .  oil  = At  oilt−1/2  + ∑  eto 
 


 

t  ipi 



 
 et 





ipit−1/2
ipi
α57,t
 


 


 c


 
 et 
 const−1/2 
α67,t   cons 
 



 
p
et
pricet−1/2
1
price
(7)
The α’s denote the estimation of a parameter from the data and allows instantaneous relationships.
The zeros indicate the restriction to zero for the parameter.
In order to derive impulse response functions out of the SVAR the model has to be tested for
stability. The companion matrix is therefore checked if the eigenvalues lie inside the unit circle.
This condition is met for our two SVAR models, and so we can proceed to analyze the IRFs
16 .
The IRFs help to interpret how the LNG import price would respond in a certain time period to a
shock of a certain determinant in the first period. We have selected 20 periods since a span over a
year is expected to capture the whole short-term response.
5.2.
Structural Analysis
The orthogonal IRFs for the US and EU models, are derived from 4. Although we have a 7x7
matrix we only present here the responses on LNG import prices reducing the amount of results
to six in each model. Their graphical interpretation show in general the expected reaction, that
determinants of LNG import prices react differently in both regions.
16 A
graphical representation of the eigenvalues can be found in annex B figure 14
19
Supply
A one time supply shock in domestic production in the US seems to have a delayed negative impact
on the LNG import price, which starts after the first quarter and lasts for about one year. This
result is not consistent with economic theory. A plausible explanation behind this result could
be that the market is well prepared for a domestic supply shock with vast storage supplies that
could overcome the one time supply shock in one year. In the European model a shock in the
indigenous production does not lead to statistical significant results implying that the response is
not different than zero.
Production shock
Production shock
.02
.02
0
.01
-.02
0
-.04
-.01
-.06
-.02
0
2
4
6
8
10
12
14
16
18
20
0
2
4
6
8
10
12
14
months
months
(a) US production shock
(b) US production shock
16
18
20
Figure 7: IRF: production shocks on LNG import price
Demand
On the demand side we have identified a general demand shock by the general natural gas total
consumption variable and an industrial demand shock that can be interpreted from shocking the
industrial production variable. The general demand shock in the US seems to lead to increasing
LNG import prices but only after at least 10 periods. The response to the shock is rather short (5
periods)and an absorption of the general demand shock is seen after 16 periods. In the EU there
is an immediate reaction of LNG prices to a consumption shock pushing prices down. But only
after one period prices start to increase following the theoretical expectations. The response to the
shock is very short and it seems to be absorbed in less than 3 periods.
The industrial production shock has a more significant impact on LNG prices in the US than the
general consumption shock. On the one side it lasts much longer and even after 20 periods it is
not absorbed, and on the other hand, it shows a brief delay of about four months but afterwards
20
Consumption shock
Consumption shock
.04
.02
.01
.02
0
0
-.01
-.02
-.02
0
2
4
6
8
10
12
14
16
18
20
0
2
4
6
8
10
12
14
months
step
(a) US consumption shock
(b) EU consumption shock
16
18
20
Figure 8: IRF: consumption shocks on LNG import price
prices increase according to our expectations. In the EU the industrial demand shock does not
lead to a statistical significant reaction implying that the response is not different than zero.
Industrial production shock
Industrial production shock
.06
.04
.04
.02
.02
0
0
-.02
-.02
0
2
4
6
8
10
12
14
16
18
20
0
2
4
6
8
10
12
14
16
months
months
(a) US industrial production shock
(b) EU industrial production shock
18
20
Figure 9: IRF: industrial production shocks on LNG import price
Financial factors
The financial market determinant Henry hub spot price triggers a strong immediate and positive
response on LNG import prices. This means that a one-time structural innovation on the Henry
Hub spot price leads to an increase in LNG import prices. Afterwards the shock is absorbed in
around one year. A different picture is drawn for the average EU spot price. The LNG import
price reacts also increasing to the spot price, which is consistent with our expectations but the
21
shock is absorbed much quicker and the intensity does not seem to be as high as in the US.
Spot price shock
Henry hub price shock
.1
.04
.05
.02
0
0
-.05
-.02
0
2
4
6
8
10
12
14
16
18
20
0
2
4
6
8
10
12
14
months
months
(a) US Henry hub price shock
(b) EU spot price shock
16
18
20
Figure 10: IRF: spot price shocks on LNG import price
Although a structural innovation crude oil prices generates a positive reaction in both regions, the
pattern the impulse response functions in both regions is quite different. Firstly the US seems to
absorb, although late (around 20 periods),the one-time shock. In Europe the shock persists much
longer, which can be interpreted as an indicator of the oil price escalation type of LTC contract
linked to crude oil prices. The relationship between oil prices and LNG prices stays confirmed for
both regions. Another interesting observation is the significance of the results on the US model
after the first quarter, where in the EU responses are significant right after the shock.
Brent price shock
WTI price shock
.1
.1
.05
.05
0
0
-.05
0
2
4
6
8
10
12
14
16
18
20
0
2
4
6
8
10
12
14
months
months
(a) US WTI price shock
(b) EU Brent price shock
Figure 11: IRF: crude oil price shocks on LNG import price
22
16
18
20
External factors
A one-time shock in heating degree days implies a colder month and additional unexpected heating.
The LNG import price in the US responded to a structural shock in hdd with a delayed upward
increase after about two months. This effect is very short and last for about a month. In contrast
to that, in the EU temperature shocks lead to immediate responses on LNG import prices, that
last for about a quarter. In comparison temperature shocks affect Europe longer than the US.
Heating degree days shock
Heating degree days shock
.04
.02
.02
0
0
-.02
-.02
-.04
-.04
0
2
4
6
8
10
12
14
16
18
20
0
2
4
6
8
10
12
months
months
(a) US HDD shock
(b) EU HDD shock
14
16
18
20
Figure 12: IRF: HDD shocks on LNG import price
After conducting the structural vector autoregressive analysis on both sides of the Atlantic our
results imply the following: The LNG import price in the US tends to react in most of the cases
with a time delay only with the exception of the Henry Hub spot price, which is the only variable
that makes LNG prices reacts immediately to a shock. IRF pattern differences are visible in the
response of LNG prices on: Heating degree days and natural gas spot prices. Although the
responses are delayed in the US they seem to last for as long as in the EU. Grouping the variables
by the response length -short (1-6 months), medium (6-12 month) and long (more than 12) - we
obtain the following:
6.
Conclusion
We have analyzed the impact of the same determinants on LNG import prices in the Atlantic
basin, applying a structural VAR model which allows for short-term restrictions. We have used
publicly available data in order to estimate the impulse response functions and find that the price
differences across the regions can be explained by the diverse impact of the proposed determinants.
23
Table 4: Responces by absorption length
US Variables
Response
EU Variables
length
Response
length
SHORT
General consumption (lconssa )
5
General consumption (lconssa )
2
Heating degree days (lhdd)
2
Heating degree days(lhdd)
3
Henry Hub spot price(lhh)
11
EU average spot price(lspoteu)
10
Natural gas production (l prod)
10
MEDIUM
LONG
Crude oil spot price WTI(lwti)
17
Industrial natural gas consumption (lipi)
15
Crude oil spot price BRENT (lbrent)
20+
Our analysis shows that LNG prices react with different IRF patterns especially for natural gas
spot prices and heating degree days. The difference in the response of LNG import prices on
heating degree days is given by the different calculation methods of HDDs in Europe and in
the US (See section 4). The difference in spot prices is much more interesting: on the one side,
the Henry Hub spot price, seems to be the only variable that makes LNG import prices react
immediately in the US. On the other side although on both sides of the Atlantic the reactions
are positive, the IRFs look quite different. This could be a sign of the lower share of natural gas
trade and contracts based on spot pricing in Europe compared to the US. For nearly all potential
determinants the length of the periods the response last is almost the same with the exception of
the crude oil price. Thus, the results indicate a much longer absorption time of changes in crude
oil prices in the European markets for natural gas. This points toward the relevance of oil price
linkages in the European LNG pricing mechanism.
We are currently implementing a variance decomposition analysis in order to draw conclusions on
the strength of the responses of each variable. In addition we still need to calculate the confidence
intervals of the IRFs applying a bootstrapping approach instead of our current asymptotic method.
Further research could also take into account political and extraordinary events i.e. the current
Ukraine-Crimea crisis, the Arab Spring or extreme weather events. Taken together our analysis
gained additional insight into the LNG market and took a step forward the understanding of
regional import price differences although further investigations are needed.
24
References
Albrecht, U., M. Altmann, J. Zerhusen, T. Raksha, and P. Maio (2014, March). The Impact of the
Oil Price on EU Energy Prices. European Union: Energy Study PE 518.747, LBSt and Hinicio
and CEPS, Brussels.
Apergis and Payne (2009). Energy consumption and economic growth: Evidence from the
Commonwealth of Independent States. Energy Economics 31-5, 641–647.
Asche, F., B. Misund, and M. Sikveland (2013). The relationship between spot and contract gas
prices in Europe. Energy Economics 38, 212–217.
Barsky, R. B. and L. Kilian (2004). Oil and The Macroeconomy Since The 1970s. Journal of Economic
Perspectives, 115–134.
Becketti, S. (2013). Introduction to Time Series Using Stata, Volume 1. Texas, United States of America:
Stata Press.
Belke, C. Dreger, and F. de Haan (2009). Energy consumption and Economic Growth: New Insights
into the Cointegration Relationship. Technical report, Ruhr Economic Papers No.190.
Blanchard, O. J. (1990). Empirical Structural Evidence on Wages, Prices and Employment in the
US. NBER Working Paper Series, National Bureau of Economic Research (2044).
BP (2013). BP Statistical Review of World Energy. Technical report.
Breitenfellner, A., J. C. Cuaresma, and C. Keppel (2009). Determinants of Crude Oil Prices: Supply,
Demand, Cartel or Speculation? Monetary Policy & The Economy Q4, 111–136.
Bros, T. (2012). After the US Shale Gas Revolution. Paris, France: Editions Technip.
Brown, S. P. A. and M. K. Yücel (2008). What drives natural gas prices? The Energy Journal,
International Association for Energy Economics 0(2), 45–60.
ECB (2008). Assessing the Factors Behind Oil Price Changes. Technical report, European Central
Bank, Working Paper Series No. 855.
ECS (2007). Putting a Price on Energy - International Pricing Mechanisms for Oil and Gas.
Technical report, Energy Charter Secretariat.
EIA (2012). Annual Energy Review 2011. Technical report, U.S. Energy Information Administration.
25
EIA (2013). Natural Gas Weakly Update -for week ending September 4, 2013. Technical report,
EIA.
Elliott, G. (1998). On the Robustness of Cointegration Methods When Regressors Almost Have
Unit Roots. Econometrica 66(1), 149–158.
FERC (2014). Natural Gas Market Oversight. Technical report, FERC.
Hamilton, J. (2003). What Is an Oil Shock? Journal of Econometrics, 113(2), 363–98.
Hamilton, J. D. (2008). Understanding Crude Oil Prices. NBER Working Paper Series, National
Bureau of Economic Research (14492).
Hartley, P. R. and K. B. Medlock-III (2014). The Relationship between Crude Oil and Natural Gas
Prices: The Role of the Exchange Rate. The Energy Journal, by IAEE 35(2), 25–44.
IGU (2014). World LNG Report - 2014 Edition 1. Technical report, International Gas Union.
Jiménez-Rodríguez, R. and Sánchez (2005). Oil Price Shocks and Real GDP Growth: Empirical
Evidence for some OECD Countries. Applied Economics 37, 201–228.
Johansen, S. (1991). Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian
Vector Autoregressive Models. Econometrica 59, 1551–1580.
Kilian, L. and D. P. Murphy (2014). The Role of Inventories and Speculative Trading in the Global
Market for Crude Oil. Journal of Applied Econometrics 29, 454–478.
Lardic, S. and V. Mignon (2008). Oil prices and economic activity: An asymmetric cointegration
approach. Energy Economics 30(3), 847–855.
Lee, C. (2005). Energy consumption and GDP in developing countries: A cointegrated panel
analysis. Energy Economics 27(3), 415–427.
Lee, C. and C. Chang (2008). Energy consumption and economic growth in Asian economies: A
more comprehensive analysis using panel data. Resource and Energy Economics 30(1), 50–65.
Lütkepohl, H. (2007). New Introduction to Multiple Time Series Analysis (2nd ed.). Berlin, Heidelberg:
Springer-Verlag.
Lütkepohl, H. (2011, October). Vector Autoregressive Models. European University Institute WP
ECO (30), 1 – 28.
Möbert, J. (2007). Crude Oil Price Determinants. Darmstadt Discussion Papers in Economics (186),
1–35.
26
Mork, K., O. Olsen, and H. Terje Mysen (1994). Macroeconomic Responses to Oil Price Increases
and Decreases in Seven OECD Countries. Energy Journal 15(4), 19–35.
Narayan, P. K. and R. Smyth (2008). Energy consumption and real GDP in G7 countries: New
evidence from panel cointegration with structural breaks. Energy Economics 30(5), 2331–2341.
Nick, S. and S. Thoenes (2013, January). What Drives Natural Gas Prices? - A Structural VAR
Approach. EWI Working Paper (3/02).
OECD (2004). Oil Price Developments: Drivers, Economic Consequences and Policy Responses.
Technical report, Organisation for Economic Co-operation and Development, Economics Department, Working Papers No. 412.
OECD/IEA (2013). World Energy Outlook 2013. Technical report, International Energy Agency,
Paris, France.
Papapetrou, E. (2001). Oil prices and economic activity in Greece. Economic Change and Restructuring 46(4), 385–397.
Siliverstovs, B., G. L’Hégaret, A. Neumann, and C. von Hirschhausen (2005). International market
integration for natural gas? A Cointegration Analysis of prices in Europe, North America, and
Japan. Energy Economics 27(4), 603 – 615.
Sims, C. A. (1980). Macroeconomics and Reality. Econometrica 48(1), 1 – 48.
Sims, C. A. (1986). Are forecasting models usable for policy analysis? Federal Reserve Bank of
Minneapolis, Quarterly Review 10, 2 – 16.
StataCorp (2011). Time Series Reference Manual. In Stata Quick Reference and Index, Release 12.
Stern, J. and H. Rogers (2011, March). The Transition to Hub-Based Gas Pricing in Continental
Europe. Oxford Institute for Energy Studies NG 49.
Stock, J. H. and M. W. Watson (2001). Vector Autoregressions. Journal of Economic Perspectives 15(4),
101 – 115.
Villar, J. A. and F. L. Joutz (2006, October). The Relationship Between Crude Oil and Natural Gas
Prices. Energy Information Administration, Office of Oil and Gas, 1 – 43.
27
Annex 1:
0
1
2
3
A.
Jan-96
Jan-98
Jan-00
Jan-02
Jan-04
log of ZEE Spot Price
Jan-06
Jan-08
Jan-10
Jan-12
log of TTF Spot Price
log of NBP Spot Price
Figure 13: EU spot prices: NBP, ZEE and TTF hub
Annex 2:
.5
Imaginary
0
-.5
-1
-.5
Imaginary
0
.5
1
Roots of the companion matrix
1
Roots of the companion matrix
-1
B.
-1
-.5
0
Real
.5
1
-1
(a) US, two lags
-.5
0
Real
(b) EU, one lag
Figure 14: Eigenvalues of the companion matrix with seven variables
28
.5
1
Annex 3:
US LNG Import Price*
.5
4.3 4.4 4.5 4.6
1 1.5 2 2.5
US Industrial Production Index*
Henry Hub Price*
0
14.4
1
14.6
2
3
14.8
US NG Production*
Heating Degree Days*
0
2
4
6
8
2.5 3 3.5 4 4.5 5
Western Texas Intermediate*
NG consumption*
Seasonaly adjusted NG Consumption*
-.4-.2 0 .2 .4
10.5 11 11.5
C.
Jan-96
Jan-98
Jan-00
Jan-02
Jan-04
Jan-06
months
Jan-08
Jan-10
Jan-12
Jan-96
Jan-98
Figure 15: Plots of US logged variables
29
Jan-00
Jan-02
Jan-04
Jan-06
months
Jan-08
Jan-10
Jan-12
.5
4.4
4.6
4.8
Industrial Production Index
1 1.5 2 2.5
LNG Import Price
Natural Gas Spot Price
0
9.5
1
10
2
3
10.5
Production
Seasonally adjusted Natural Gas Consumption
-.5
10
0
10.5
11
.5
Natural Gas Consumption
Heating Degree Days
2
2
3
4
4
6
5
Spot Price, Brent
Jan-96 Jan-98 Jan-00 Jan-02 Jan-04 Jan-06 Jan-08 Jan-10 Jan-12
Jan-96 Jan-98 Jan-00 Jan-02 Jan-04 Jan-06 Jan-08 Jan-10 Jan-12
Figure 16: Plots of EU logged variables
30