An empirical time series analysis of energy consumption in Cyprus
Journal
Economic Analysis Paper
Date Issued
March 2006
Author(s)
Abstract
This report presents the first econometric analysis of energy consumption in Cyprus. Time
series analysed were those of residential, commercial, industrial and agricultural electricity use,
gasoline consumption as well as the aggregate non-electricity and total energy consumption
using annual data from 1960 to 2004. The dynamic interaction between the corresponding
energy form and appropriate income, price and weather variables was examined through the
application of widely used time series analysis techniques such as unit root and cointegration
tests, Vector Error Correction (VEC) models, Granger causality tests and impulse response
functions. Because of power and size problems associated with these methods in small
samples, single-equation autoregressive distributed lag (ARDL) models were also employed
for each energy variable. The validity of inferences made with such models has been reestablished
in the late 1990s thanks to the work of Pesaran-Shin-Smith.
Results from cointegration tests and VEC models show that a long-run equilibrium relationship
between energy, income and prices exists for most energy uses. The long-term impact of
income and prices on energy use is significant, with elasticities similar to those reported for
other countries (above unity for income, less than 0.5 for prices in absolute terms). Weather
fluctuations seem to be the most significant cause of short-term variation in electricity use
(albeit with small elasticity values). Granger causality tests indicate that energy prices can be
treated as purely exogenous, income and prices often Granger-cause energy use, and there is
bidirectional causality between most energy forms and income or economic activity.
ARDL test results have to be interpreted in conjunction with those of the cointegration tests. In
the cases of residential and commercial electricity consumption, elasticities are found to be
similar with those of the VEC model, whereas results of the two methods are different for the
long-term elasticities of gasoline and industrial electricity consumption.
Despite the quite small sample size, which poses limitations on the analysis, the evidence from
both the VEC and ARDL models shows that results are meaningful and robust for residential,
commercial and industrial electricity as well as gasoline consumption. This finding is important
as it allows the corresponding income, price and weather elasticities to be used for forecasting
purposes and policy analyses.
series analysed were those of residential, commercial, industrial and agricultural electricity use,
gasoline consumption as well as the aggregate non-electricity and total energy consumption
using annual data from 1960 to 2004. The dynamic interaction between the corresponding
energy form and appropriate income, price and weather variables was examined through the
application of widely used time series analysis techniques such as unit root and cointegration
tests, Vector Error Correction (VEC) models, Granger causality tests and impulse response
functions. Because of power and size problems associated with these methods in small
samples, single-equation autoregressive distributed lag (ARDL) models were also employed
for each energy variable. The validity of inferences made with such models has been reestablished
in the late 1990s thanks to the work of Pesaran-Shin-Smith.
Results from cointegration tests and VEC models show that a long-run equilibrium relationship
between energy, income and prices exists for most energy uses. The long-term impact of
income and prices on energy use is significant, with elasticities similar to those reported for
other countries (above unity for income, less than 0.5 for prices in absolute terms). Weather
fluctuations seem to be the most significant cause of short-term variation in electricity use
(albeit with small elasticity values). Granger causality tests indicate that energy prices can be
treated as purely exogenous, income and prices often Granger-cause energy use, and there is
bidirectional causality between most energy forms and income or economic activity.
ARDL test results have to be interpreted in conjunction with those of the cointegration tests. In
the cases of residential and commercial electricity consumption, elasticities are found to be
similar with those of the VEC model, whereas results of the two methods are different for the
long-term elasticities of gasoline and industrial electricity consumption.
Despite the quite small sample size, which poses limitations on the analysis, the evidence from
both the VEC and ARDL models shows that results are meaningful and robust for residential,
commercial and industrial electricity as well as gasoline consumption. This finding is important
as it allows the corresponding income, price and weather elasticities to be used for forecasting
purposes and policy analyses.