What is your methodology?
Academic basis of our approach: The methodology is broadly similar to the approach outlined in the following academic paper in Nature.
Summary of the approach: We use parameterized Lorenz curves (LC) that are time invariant, however we use one LC per demographic group (age and gender) and predict growth rates based on the national accounts total household expenditure rather than the survey means directly.
Primary elements of the approach: The main drivers of the method are the distributions and mean spending forecasts. The latter is based on medium term forecasts by the IMF. The IMF uses a “bottom-up approach” where the forecasts for each country are made by country teams, so methodology can and does vary from country to country. So do the input variables. The long-term medium spending forecasts (after 2023) come from the Shared Socioeconomic pathways, whose main drivers are age and educational structure and changes of countries.
What are your data sources?
Main input sources include:
- World Development indicators (World Bank)
- Poverty Equity Database (World Bank)
- World Economic Outlook (IMF)
- Shared Socioeconomic Pathways (IIASA, OECD)
- United Nations population projections
- PovcalNet (World Bank)
- UNU- WIDER (UN)
- National Statistic offices
How do you compensate for gaps in the various data sources?
For each type of “gap” we apply a unique and tailored approach. Any informational gaps in distributional data (sources from Povcal, UNU-WIDER, and Poverty Equity) is imputed by finding countries that are similar along the dimensions of GDP, age structure, sectoral employment and education structure.
How often do you update your forecasts?
Twice per year, once in November and once in May. These updates are aligned with the biannual release of the World Economic Outlook (WEO) by the International Monetary Fund (IMF). Throughout the year we update parts of the data as updates become available.
Does the data change that much from update to update?
It varies from update to update, and country to country.
What countries do you have data for?