Risk Premium in the Nordic Power Market: A Quantile Regression Analysis

Abstract

Electricity is a highly volatile commodity and the risk premium affects marketparticipants who use financial contracts. In this thesis, the risk premium in theNordic electricity market is investigated using data from 2009 to early 2013. Thepurpose is to determine the probability distribution of the risk premium and to investigatehow the determinants impact various quantiles of the risk premium distribution.Quantile regression is used to examine the risk premium. To the authors? knowledge,quantile regression has never been applied to the risk premium in electricity markets.Several determinants are selected and incorporated into the quantile regression modelbased on their assumed impact on the risk premium distribution. Nine differentestimates are made, one for each respective quantile model. These models areevaluated using a goodness of fit test Pseudo-R2, and unconditional and conditionalcoverage probability for backtesting. The determinants are examined by evaluatingtheir coefficients and t-statistics for the various estimates. Based on Pseudo-R2, themodels for the two most extreme quantiles are found to have the best fit. Of theevaluated explanatory variables spot price variance, seasonal adjusted demand anddemand variance has a significant coefficient over multiple quantiles. For several ofthe evaluated determinants, there are no significant coefficients. Hence, the impactof these determinants remain unanswered.From the results, it can be concluded that the impact of the determinants on thevarious quantiles differ. Moreover, the computed models capture a part of thedistribution to the potential risk premium. The results can be of special interest tomarket participants with a VaR mindset. Considering that the model, using quantileregression estimation, can be used to predict the probability that the risk premiumwill be above or below a given level.