A model of imperfect dynamic competition in the Nordic power market

Abstract

This dissertation presents a framework for testing for market power in storable-good markets. The framework is applied to the Nordic wholesale electricity market, in which the storable commodity is hydroelectricity. The marginal cost of a unit of hydro output arises from the opportunity cost of not being able to sell the unit in the future. Thus, to measure price-cost margins, the economist must evaluate the value of the water at the state of the market where the production decision is made. This value depends on the hydro producers' expectations about the future market conditions. The Nordic power market presents a unique opportunity for testing the nature and degree of market power in storage behavior, because of the availability of precise data on market fundamentals, which determine the expectations about the future value of water.

The thesis first develops a model of socially optimal hydro allocation. This competitive benchmark is modeled as an aggregative single agent stochastic dynamic programming problem, and is solved numerically on the computer. The key inputs of the model are estimated from actual market data. The model can be used to construct distributions of the expected val- ues of the key market outcomes, such as storage levels, prices and outputs. The expected price of electricity is shown to exhibit features that are typical for both exhaustible resources and for storable goods. The results from the benchmark model also suggest that the observed market behavior in 2000-05 was markedly different from the social optimum. This inefficient allocation of the hydro resource is estimated to have lead to a welfare loss of 621 million euros.

To study whether the welfare loss can be attributed to market power, the thesis next develops an explicit model of dynamic imperfect competition. The model maps the primitive distributions to market outcomes as a function of the market structure. Empirical models of dynamic imperfect competition where the product market equilibrium is connected to the dynamics of the state of the market are very scarce in the literature. The model presented here is built upon a dominant firm approach, which greatly facilitates the computation of the model. Apart from the change in the market structure, the model is unchanged from the model of competitive behavior. The computational tractability enables the estimation of the market structure that best explains the observed market behavior.

It is shown that a model, where 30 per cent of total hydro resources is controlled by a single firm, and the rest by competitive producers, provides the best fit with the historical market outcomes. Market power is shown to lead to higher expected storage levels, prices and price risk. However, the expected welfare loss from the estimated level of market power is very small. The estimated, relatively large welfare loss in 2000-05 is shown to have arised from an exceptional in?ow shortage in 2002, which enabled the strategic hydro firms to reduce output profitably. Finally, the thesis studies the possibility that the pattern attributed to market power could also be explained by some mismeasured or unobserved factors. However, the main results are shown to be robust to several reparameterizations of the model of competitive hydro allocation.