2.1. Materials
2.1.1. Data and Discussion on Dynamics of Each Market
The focus lies on Austria and the Netherlands, as each exemplifies a diverse evolutionary phase in the European wholesale gas markets liberalization on the one hand, and each is exposed to a different indigenous gas production and gas supply portfolio on the other hand.
2.1.2. Austria
The Austrian gas transmission network is composed of three market zones: Tyrol, East and Vorarlberg. The market zones of Tyrol and Vorarlberg are only associated with the German transportation network, have no physical link to Austria, and have neither storage nor indigenous gas production at their disposal. The vast majority of the Austrian gas demand is located in the Eastern market area [
28]. Within the Eastern market area, the import stations are connected to the border points by the domestic distribution system, and major transit pipelines exist.
Since 2013, within the new “entry-exit” model, the Eastern market area forms one entry-exit zone with one central Virtual Trading Point (VTP). Settlement at the VTP is carried out by the Central European Gas Hub (CEGH). This creates trading possibilities at the wholesale level, which is a major shift for a market that used to be governed by long-term oil-indexed gas contracts. Additionally, with the adoption of the transmission Network Code, new traders wishing to exchange gas titles experience reduced congestion in the transmission networks [
29].
Austria imports more than two-thirds of its inland consumption. Most of the gas entering Austria is delivered from Russia through Gazprom, arriving at the key entry point Baumgarten. The remainder originates from Germany through the West-Austria-Gasleitung (WAG) pipeline, the physical connection at the border with Germany. The indigenous production is undertaken by two companies, OMV and RAG.
In addition to local producers, as seen in
Figure 1, the Austrian market is supplied from the east on the one hand, where a single company is the only supplier, and from Germany on the other hand, where several big gas suppliers are active.
2.1.3. The Netherlands
The Dutch gas company Gasunie is operating an entry-exit tariff system for the gas transmission network in the Netherlands, similarly to the Austrian market. The Title Transfer Facility (TTF) is the virtual point where operating traders exchange gas titles in the Netherlands.
Gas is exported and imported by means of connections that are located near the borders to Germany and Belgium. Gas is also imported from the United Kingdom via Belgium through the bi-directional interconnector. Nonetheless, gas can only be directly imported via the connection with Norway in Emden, northern Germany.
Gas consumed in the Netherlands mainly comes from local production, specifically the Groningen field, in addition to further onshore and offshore Dutch fields. The Netherlands is the principal gas producer in the EU.
Additionally, as per
Figure 2, other sources feed the local market, such as imports from Russia, Norway, and LNG. More than two thirds of the total gas produced within the country originates from the Groningen field (Giant and biggest European gas field located in Netherland) and other onshore fields. The remainder is produced in 150 fields located offshore in the North Sea [
30].
2.1.4. Key Performance Indicators of Both Gas Hubs
Another set of data that is important for this study is the wholesale gas price in each market.
Figure 3 shows the time series of the natural gas prices of both the Dutch market, represented by the TTF hub, and the Austrian market, represented by the CEGH. All prices are denominated in Euros per MWh. The dataset consists of monthly values recorded between January 2013 and December 2018. As shown in the graph, both price trends are, to some extent, positively correlated, albeit the price of gas sold on the Austrian side is at most times higher than that on the Dutch market.
This price and volume behavior should usually be justified and interpreted by a solid understanding of the economics of both market structures and evaluated by means of a quantitative analysis, which is the subject of this research.
Some reports presented by think tanks such as [
31,
32,
33] conducted an analytical analysis that can also provide an insight into market structure and prior formation of these gas markets. The results of these analytical studies are summarized below:
The number of firms trading at a hub indicates the willingness of traders to be involved and the ease of participation; a larger number of active participants is a sign of increased competition and consequently less market manipulation.
According to
Table 1, the low quantity of gas traded in the CEGH is an indication of low liquidity in Austria, as compared with the Netherlands. On the other hand,
Table 2 shows the churn rate, which is the traded volume as compared with the size of the market. It clearly highlights the fact that the Dutch market scores a high churn ratio, which is attractive for traders and financial players. However, this is not the case for the CEGH; the low churn rate and the low quantities of traded volumes indicate lower liquidity in Austria, an unattractive environment for traders and financial players. The churn ratio is a main factor in determining the success of a trading platform according to [
10].
Based on the presented numbers, it can be directly concluded that the Dutch TTF hub is dynamic and less concentrated because of the liquidity and wide range of participants in the gas trade, unlike the CEGH in Austria.
These studies show the positive impact of the liberalization process on the gas price mechanism in Europe. More specifically, the northwestern part of Europe now reflects gas market supply and demand fundamentals. However, this is not yet fully the case in Central Europe, i.e., the CEGH. In other words, the gas-on-gas competition in Central Europe has increased from zero in 2005 (this means that prior to 2005, the contracts in the eastern part of Europe were wholly governed by long term contracts indexed on oil) to over 56% in 2015, with changes accelerating from 2013 onwards [
32]. The gas indexation level in the TTF has reached higher norms.
By 2013, most of the Dutch and Norwegian long-term contracts had moved to hub prices. In fact, companies like Statoil, Gasterra, and UK producers shifted their gas contracts to hub indexation. Moreover, Russia and some of its customers agreed to revise the pricing formula by reducing the base price by about 10% and by reducing the take or pay commitment from as high as 90% to 60%. The only supply source that is still resistant to change is located in Algeria, where the gas supply contracts are still indexed on oil. Gazprom, Sonatrach, other key producers and several LNG exporting companies prefer long-term bilateral contracting with a higher presence of oil-price indexation [
34].
As stated previously, the new regime foresees that any gas transported through the Austrian and Dutch networks is traded at the VTP where both bilateral and exchange trades are possible. Therefore, the large suppliers will sell their gas at both trading hubs. This means that both markets are interdependent, and that game theory is indeed a convenient model to analyze the strategic behavior of such companies. The following section elaborates on and presents the methods used in this analysis.
2.2. Non-Parametric Method
Translated into modeling, a country importing gas from
different suppliers and
the number of available observations is considered. For each
and
, the gas price at period
is denoted by
and the quantity of gas supplied by supplier
at period t by
. The monthly data related to supply and prices are illustrated in
Figure 1,
Figure 2 and
Figure 3.
The classical Cournot equilibrium test, which is designated as the parametric method in this study, takes into account several assumptions on cost and demand functions and yields an optimization problem. The simplest way to solve such an optimization problem is by applying the Krush-Kuhn-Tucker (KKT) principle to solve the following objective function
where the first derivative of the inverse demand function is
,
is the marginal cost of the gas, and where the total quantity supplied to the relevant markets is,
The latter objective function is the profit maximization strategy of each gas supplier, and the success of solving this problem is not constrained by mathematical theory. As long as there is a convex problem and the conditions of the KKT are satisfied, the optimal solution can be found. However, the problem lies in the assumptions of such empirical models. At the end of this section, the parametric method is further explained, and the different assumptions are enumerated.
As previously explained, a practical non-parametric theory to test if a data set with convex cost functions belongs to Cournot is developed. In the following paragraph, it is explained how this theory can be customized and applied to the natural gas market.
is defined as the marginal cost of supplier
at period
. The first-order condition of firm
optimization problem previously defined in Equation (1) now states that there is
contained in
such that
The set of observations respects the Cournot equilibrium if both conditions mentioned below are met (these conditions are derived from Equation (3):
The observed price in each period should belong to the inverse demand function; it follows that the array
must obey the following condition
At each period, firm
i′s quantity level
maximizes its profit subject to the quantity of the competing firms. This property can be stated as the following inequality
At first, it is essential to know and compare the values of marginal costs for each firm at each time . In other words, if firm is producing a quantity that is higher than the quantity produced by firm , this simply tells us that the marginal cost of the latter firm should in theory be higher than the former.
Condition 2 is then applied to extend the analysis from several firm output at a specific time , to a whole range of firm outputs at different time intervals. It will then be possible to compare the marginal costs of different firms and at different times. So, if firm produces a quantity at time t, and then changes its strategic outcome to a lower level , Equation (5) tells us that must be lower than . The same logic is repeated for firm j. Finally, this will lead us to combine the results of both condition 1 and 2, and then order the marginal costs for each firm and at each time in increasing order. Once this is completed, we can then check if this order respects Equation (4).
Since the constraints of a specific inverse demand function and a specific cost function, both of which are defined in
Section 3, are lifted, prices and volumes can be tested in an algorithm to verify whether or not each observation
respects the Cournot equilibrium. This algorithm is based on the results of the previous statements. It starts with an assumption of the highest possible marginal cost (Upper bound) of firm
at
, which is equal to the price
(this is typical in a fully competitive market, where the price of any commodity (i.e., gas in our case) should be equal to its delivery cost.), and verifies if conditions 1 and 2 are met at each change in marginal cost level.
The Cournot acceptance rate is then calculated, which is the ratio of cases where conditions 1 and 2 are met to the total number of observations. A higher ratio means that the firms are competing under the umbrella of the a Cournot competition, each trying to behave strategically and willing to maximize its profit, taking into account what it believes to be a strategic output of its rival
Finally, in order to check if the changes in the concentration of suppliers affect the movement of gas prices in each market, the correlation coefficient
between the prices
. and the concentration index (the Herfindahl–Hirschman Index is used to calculate the concentration in each market) in each market is solved as follows,
The results of the non-parametric model applied for the first time to natural gas markets are then compared with the results of the parametric models that are used extensively in the literature. In order to highlight the complexity of making parametric assumptions, a brief description of the parametric method and its assumptions is presented below.
2.3. Parametric Method
Firstly, with regard to the cost assumptions, the production cost function of each supplier, in addition to its transportation costs, in this study follows the form proposed by the following articles: [
14,
35]. Equation (7) gives the capacity utilization marginal cost function
:
where
is the maximum production capacity of each supplier. The parameters
are adjusted for inflation in order to fit the purpose of the study. The production cost functions are assumed to be convex and monotonically increasing.
As shown in
Figure 1, the Austrian market is supplied by three main parties (OMV and RAG are counted as one indigenous supplier, since both are assumed to have the same marginal cost function. In addition, the indigenous gas producers in Austria are assumed to be price takers and therefore cannot influence the market price of the gas).
The cost assumptions (Equation (7) for the gas are shown in
Table 3.
Secondly, assumptions on the price demand function are made. To this end, the price and volume data are fitted to a classical linear price function, .
The assumption of a simple linear regression model results in a low correlation coefficient. The authors acknowledge that there are many other explicative variables affecting the gas price and that it is more accurate to estimate the demand function using econometric models of high complexity level (i.e., vector autoregressive analysis using several supply and demand variables, machine learning, etc.). Due to the limited set of available data points, the fact that modeling the price function exceeds the scope of this paper, and the fact that the use of the parametric method is only limited to comparison purposes, the assumption of a simple linear regression model is made for the demand function.
The results and the use of simple linear regression and the least square principle are assumed to be indifferent to the possibility of endogeneity and correlation with error terms in this particular context. The market inverse demand function
is a linear function with ay downward slope. Additional assumptions, such as price elasticity of demand, are made in order to obtain more accurate numbers (the price elasticity of demand is assumed to be −0.2 for both markets, Austria and the Netherlands. This is an assumption, and different values that range between −0.1 to −0.8 are found in the literature for oil and gas markets [
37,
38].). The final step of the optimization is to solve the system of nonlinear equations and apply the KKT principle. The number of equations depends on the number of suppliers that are competing in each market.