In the last two decades many countries in the world have implemented across the board reforms seeking to increase the reliability and efficiency of their electric power sectors. These reforms have often included the introduction of hourly/half-hourly multiple- generation-units electricity auctions as the core electricity supply and demand clearing mechanism. Thorough economic analyses indicate that by and large these market mechanisms have yielded significant efficiency improvements, however the experiences have been quite varied. In particular, the level of competitiveness among electricity generators that they are able to induce seems to be very sensitive to 1) The precise details of the mechanism 2) The technological composition of the country’s energy sector. These concerns have spawned individual country studies to measure in detail the market power of the participants, the precise ways in which they exercise that market power and the features of the economic environment and the market’s design which explain them. The standard models of these markets, represent them as sequences of independent Cournot competition encounters between the electricity generating firms. That is, a key assumption of the model is that the markets do not exhibit any forms of inter-temporally sustained tacit collusion. However, the details of these market mechanisms seem to match very well the broad conditions which the theory of repeated games has identified as conducive towards tacit collusion: 1) The interaction is very frequent and occurs at regular intervals. 2) Information about competitors’ fundamentals is public and easily available 3) There are a number predictable, naturally focal variables in the environment which can serve as correlating devices. This paper is the first part of a project studying the Colombian wholesale electricity market. The main objective of the project is to formulate an explicit model of tacit collusion for this market with three specific goals in mind: 1) That it matches the data better than the standard Cournot models 2) That it can be used to explicitly test hypotheses asserting that generating firms are tacitly colluding 3) That it can provide concrete policy recommendations to increase this market’s efficiency by reducing the amounts of tacit collusion. Concretely, I have in mind a very simple kind of tacit collusion: The weather (specifically the amount of rainfall) serves as an intertemporal coordination mechanism, indexing how aggressive each participant “can be” at any given time. Given the transparency of the environment and visibility of the actions of all participants, no “punishments phases” ever occur on the equilibrium path. This means that in the data, the behaviors of the generators should be fully driven by a (small) finite partition of the periodic rainfall levels. In this paper I begin this exploration by studying the conjecture that the behavior of participants in this particular market, is driven by a small partition of a set of publicly observable variables, not including past behaviors.
Read full abstract