Fuel choice, fuel switching and improved cook stoves in Vietnamese households: Analysis, models and proposals for new solutions

Doctoral thesis, 2015

A majority of rural households in the developing world use solid biomass fuels for cooking. This use has severe negative health effects, is often either expensive or time consuming, and contributes to global warming. Options for policy interventions include the promotion of improved cook stoves (ICS) and enabling households to switch to more modern fuels, like liquefied petroleum gas. The main aims of this thesis is 1) to explore whether rural households’ fuel use can be modeled in new ways that focus on prediction, 2) to investigate whether area level differences in fuel use may have impacts for ICS programs, and 3) to address new solutions for ICS programs in areas where the current fuel use is mainly collected biomass. Methods used to model fuel use are ordinary linear regression and a machine learning algorithm called Random Forest. Further models are developed in order to evaluate possible implications and proposed solutions for ICS programs based on variations in current fuel use. The papers use data from two different surveys. The first data set is from a survey, carried out in the Vĩnh Phúc province of northern Vietnam in 2010. The second survey is representative of most of rural Vietnam and was collected in 2002, 2005, and 2008 as part of the Vietnam Rural Electrification program. The results from the regression and Random Forest analysis include new ways to model fuel use, enabling easy and accurate prediction. The results also provide possible alternative explanations for some previous modeling results. The modeling of stove interventions reveals large potential differences between communes, as well as a possible non-linear relationship between stove efficiencies and benefits, but also large uncertainties in estimations depending assumptions of fuel choice. Lastly, a model for a new type of ICS program that offers possibilities to overcome some of the barriers to adoption and sustained use reported by previous studies is evaluated. Combining the conclusions from the respective papers, a possibility of modeling variations in possible outcomes for stove programs, and the effects of such programs based on area descriptions, is found. However, further research is needed in order to make more robust estimations.