Testing for integrated electricity series – A formalized synthesis of known problems

Abstract

This paper reviews and discusses methodological choices and statistical inferences relevant to the literature that has worked at identifying long-memory behaviours within the stochastic features of electricity series (consumption and prices). We argue that persisting divergences reported in this literature can be linked to the (non)-treatment of key identification challenges in time-series econometrics. We synthetized those problems in a formalized fashion focusing on the followings: presence of structural breaks in the series (i); inferences affecting panel data structure such as cross-sectional dependence among the series, heterogeneous slope coefficients as well as the problem of missing observations (ii); the issue of integrated time-series observed with noise and some implications for cointegrated relations (iii).