Abstract
A thorough and careful choice of inputs is the bedrock of all successful energy demand modelling. However, the choice of inputs in most research into fluctuations in electricity demand and its potential determinants is very often made by establishing a simple correlation or by a short descriptive analysis of the specific characteristics of each selected input. As a result, this work proposes a new methodology for input selection based on the interaction between stationarity tests, the Johansen cointegration test, as well as ARDL and VECM modelling, in order to facilitate the highlighting of a real causal link between the inputs brought in the modelling of electricity consumption. Applying this new approach to the Cameroonian case, enables to establish the existence of a unidirectional causal relationship from GDP per capita, CO2 emissions, urbanization and the number of subscribers, respectively, to residential electricity consumption. These inputs are therefore the main factors explaining electricity consumption in this sector.•A new methodology for selecting input variables in modelling power demand is outlined.•The stationarity and Johansen tests as well as the ARDL and VECM models are combined.•The choice of inputs for modelling and forecasting power demand is optimized.