It is widely accepted that in liberalized electricity markets log-returns display fat-tailed den sities. Besides qualitative assessments, so far precise characterizations of the shape of the distribution have been seldom provided. In this work, we propose the Subbotin family as a possible candidate. More specifically, we fit a Subbotin to the empirical density of NordPool daily log- returns between 1996 and 1999, for each of the 24 hours, as well as to the underlying shocks, un covered by washing away the linear autocorrelation structure of the data. Log-return densities a re well described by a symmetric Subbotin with shape parameter of 0.5, while the distribution of shocks is approximately Laplace (Subbotin with shape parameter equal to 1). This is somewhat counter-intuitive, in that a linear autocorrelation structure, if applied to an i.i.d fat-tailed shock, should yield a less leptokurtic series. Moreover, both distributions are stable across hours, despite the wide cross-hour heterogeneity in linear autocorrelation structures. These results hint at correlations between bidding at different hours, as well as non-linear dyn amic dependence structures, as possible key mechanisms behind the outcomes of electricity markets.
Some Statistical Investigations on the Nature and Dynamics of Electricity Prices
SECCHI, ANGELO
2005-01-01
Abstract
It is widely accepted that in liberalized electricity markets log-returns display fat-tailed den sities. Besides qualitative assessments, so far precise characterizations of the shape of the distribution have been seldom provided. In this work, we propose the Subbotin family as a possible candidate. More specifically, we fit a Subbotin to the empirical density of NordPool daily log- returns between 1996 and 1999, for each of the 24 hours, as well as to the underlying shocks, un covered by washing away the linear autocorrelation structure of the data. Log-return densities a re well described by a symmetric Subbotin with shape parameter of 0.5, while the distribution of shocks is approximately Laplace (Subbotin with shape parameter equal to 1). This is somewhat counter-intuitive, in that a linear autocorrelation structure, if applied to an i.i.d fat-tailed shock, should yield a less leptokurtic series. Moreover, both distributions are stable across hours, despite the wide cross-hour heterogeneity in linear autocorrelation structures. These results hint at correlations between bidding at different hours, as well as non-linear dyn amic dependence structures, as possible key mechanisms behind the outcomes of electricity markets.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
