The Global Household Energy Database of the World Health Organization (WHO) was used for data related to cooking indicators. The database contained 824 household surveys collected from 161 countries (including high-income countries) between 1970 and 2014. The countries provided for cooking are only those with underlying data, so there are no estimates for Turkey and Libya.
Population data from the World Bank’s World Development Indicators were used for all countries except the Cook Islands and Anguilla (not in that database), so United Nations Population Division data were used. The World Development Indicators database does not have 2013–14 data for Eritrea, 1992–94 data for Kuwait, or 1990–97 data for Sint Maarten, so the 2011, 1991, and 1998 populations were used as proxies.
Estimating missing values
Since household surveys are conducted irregularly, a multilevel nonparametric modeling approach developed by the WHO was adopted to estimate missing values in between surveys for both databases.
For clean cooking fuels, only the model estimates are used due to large variances in survey results.
Multilevel nonparametric modeling takes into account the hierarchical structure of the data: survey points are correlated within countries, which are then clustered within regions. Time is the only explanatory variable; no covariates are used. Regional grouping are based on WHO regions and used for cooking.
Calculating the annual growth rate
In contrast to earlier editions, the 2017 GTF uses a simpler, more intuitive annual increase in the access rate, calculated as the difference between the access rate in year 2 and that in year 1, divided by the number of years to annualize the value:
(Access Rate Year 2 – Access Rate Year 1) / (Year 2—Year 1)
This approach takes population growth into account by working with the final national access rate.