|
Article Excerpt
1. INTRODUCTION
In 2005, natural gas consumption in the European Union (EU) states was approximately 530 billion cubic meters per year (bcm/y) (EU, 2004). Presently, this demand is fairly evenly divided between industry, power generation, and residential consumers. Figure 1 indicates that there are a dozen producing regions that potentially can sell gas to this market. However, recent events, for instance, where Russia disrupted European supply due to a conflict with Ukraine in January 2006, have highlighted the vulnerability of the EU market to the exercise of market power. Production in the EU countries can only meet about half of their own demand; meanwhile, the EU's production capacity is less than what just three major suppliers to the region (Norway, Russia, and Algeria) can devote to exports to the EU. Demand growth, in part spurred by the EU CO2 Emissions Trading System, would increase this vulnerability. Consumption could potentially increase by more than 60% (or 2%/yr) by 2030, with power taking an increasing share. The growth rate in the eastern EU countries that were formerly in the Warsaw Pact is anticipated to be almost twice as high. Meanwhile, EU production capability is likely to fall from the present level of about 260 bcm/yr to two-thirds of that level by 2030 (EU, 2004). Although imports of LNG from elsewhere in the world could meet much of the growing gap between production and consumption in the EU, Norway, Russia, and Algeria will continue to play dominant roles in the EU natural gas market.
[FIGURE 1 OMITTED]
The purpose of this paper is to use the multiperiod version of GAS-TALE, an equilibrium model of the EU gas market, to explore the potential for exercise of market power by gas producers in the region between 2005 and 2030 and to illustrate the use of equilibrium models for that purpose. Figure 1 shows the study region, and Table 1 lists the producing and consuming regions considered. After providing a brief overview of methods for projecting market power in natural gas markets (Section 2), we summarize the structure and assumptions of GASTALE (Section 3). Three cases of market power are presented in Section 4: perfect competition, pure Cournot competition among gas producers, and an intermediate base case in which the strategic behavior of producing region's actions is partially constrained by pre-existing contracts. A set of conclusions (Section 5) closes the paper.
2. METHODS FOR PROJECTING MARKET POWER IN GAS MARKETS
There are several methods for characterizing and projecting market power in energy markets such as the EU gas market. We categorize them into statistical empirical models, competitiveness indices, experimental economics, and simulation models. Models can be further divided into agent-based and equilibrium models, the latter being the approach we adopt here. No method is completely satisfactory by itself; each has advantages that complement the others.
The statistical approach uses market outcomes to estimate the extent to which market power has been exercised in the past. For instance, Murry and Zhen (2008) identified dynamic price behavior at US gas hubs that was consistent with the exercise of market power. However, such estimates lose relevance to the extent that the market structure changes due to, for example, demand growth, reorganization of the industry, shifts in world markets, or alterations to gas transport and storage infrastructure. The many recent changes in EU gas markets mean that there is relatively little data that could be used to build statistical models for projecting market power in that region. However, statistical analysis can still help validate simulation models.
Competitiveness indices are simple summaries, such as the Hirschman-Herfindahl Index or whether the largest supplier is pivotal in a market. Such indices are commonly used in regulatory proceedings. For instance, the US Federal Energy Regulatory Commission evaluates applications by pipelines for market-based rates by assessing whether potential substitute pipelines have spare capacity that equals or exceeds the applicant's capacity (McAfee and Reny, 2007). However, indices may fail to capture aspects of markets, such as transmission limits, that have important impacts on the ability to exercise market power; for that reason, Borenstein et al. (1999) recommend use of simulation models.
The experimental economic approach, which uses live subjects, has the potential to identify likely modes of behavior among large market players because it allows for learning and suboptimal decision making. Further, it can capture features of market rules that are difficult to represent mathematically. An early evaluation of the efficiency of gas auctions on a network is by McCabe et al. (1990), but we have found no other experiments in a natural gas context. Agent-based mathematical models have a similar objective: to simulate imperfect and dynamic decision making by market players in the face of complex market environments, but using computerized instead of live agents. Without considering network constraints, Barrot and Tchung-Ming (2008) simulated the interaction of flexible contracts and spot markets in natural gas, considering how the former may amplify market power. Several agent-based efforts are reported to be underway to model market power in gas networks, but no actual applications have been reported (e.g., Tatara et al., 2007). Unfortunately, live experiments are expensive and both live and agent-based experiments tend to be difficult to replicate, so results are difficult to generalize.
The last approach we consider for projecting market power in gas markets is equilibrium modelling, the basis of our model GASTALE. Equilibrium models formulate the optimisation problems facing producers, transporters, traders, storage, and consumers of gas and then solve them simultaneously while imposing market clearing conditions. The results will depend both on market structure and on behavioural assumptions, for instance concerning conjectural variations or the degree of forward contracting (Gabriel and Smeers, 2006). The ability to accommodate different structural and behavioural assumptions is both an advantage and disadvantage. The advantage is that the effect of structural changes can be explored in a way not possible with statistical models--for...
|
