EVOLVING SEASONAL PATTERNS IN UK ENERGY SERIES

UK Energy series exhibit pronounced regular but not necessarily fixed seasonal patterns. Failure to reflect such changing patterns in econometric models of energy use can result both in misleading estimates of elasticities and policy responses and in forecasts which under- and over-predict seasonal peaks and troughs. Structural Times Series models permit the formulation, estimation and testing of models which allow for evolving stochastic seasonal components and reflect changing patterns of economic behaviour. Moreover such components can be incorporated into causal regression equations to permit greater flexibility in modelling the seasonal variation than is possible using ordinary dummy variables. By estimating suitable dynamic models which allow for evolving seasonal effects and then nesting the fixed effects models, we compare estimated elasticities and test the restriction of fixed seasonal effects.