Developing a Bottom-Up Methodology to Project Electricity Demand for Productive Uses in sub-Saharan Africa in 2030 in IMAGE

Abstract

There is a general agreement about the manifold ways in which electricity supply can contribute to sustainable improvement of livelihoods, which is manifested in the SDG 7 of the UN. Many governmental institutions therefore support electrification projects in developing countries with poor access in the context of their development policies. One such institutions is the Dutch Directorate General for International Cooperation, which is aiming to develop a policy strategy to support the electrification of countries in SSA, which currently have the lowest electrification rates in the world. To support this planning, the IMAGE model, an integrated assessment model by the Dutch Environmental Assessment Agency PBL is supposed to project future electricity demand in the region to understand the consequences of different electrification strategies in terms of costs and technologies. Models like IMAGE are increasingly improving their spatial granularity and accuracy by implementing bottom-up modelling approaches to better understand how underlying premises influence different projection scenarios. To this end, the newly developed PrElGen framework is aiming at reflecting electricity demand from different channels, such as residential demand, demand from schools and health facilities and demand for irrigation and productive uses. The present report explains how a methodology for projecting productive uses was developed, i.e. the use for micro-scale income activities on the HH level of crop processing and non-agricultural productive uses. The projection for demand from crop processing is based on a literature review of electricity requirements of different processing technologies for a selection of crops carried out in the context of this research. The final model for the projection for other productive uses, which this study developed, consists of four distinct empirical models developed in the course of this research which are two logistic regression models for the propensity of business activity in a household and the propensity to have an electrical connection and two multivariate linear regression for the performance and consumption of enterprises. The regression models were built based on samples from the World Bank Household and Enterprise Surveys and further variables were added from other sources. Despite a multitude of interactions amongst variables, a variety of variables could be distilled for the different models which showed robust predictive power on the dependent variables. The study could thereby contribute to the existing body of research by confirming significant effects over a broad range of context found in previous literature in rather narrow samples. It also adds some new insights into further predictors and analysed interaction effects of several predictors and outlines potential pathways for future research.