How Much Should We Trust International Statistics? A Study of the World Development Indicators
Economic variables like unemployment, inflation, and GDP growth are not set in stone: they are preliminary estimates that capture the information available at a given point in time and are constantly revised by statistical agencies. Yet political scientists rarely examine whether existing findings are robust to these revisions. Using the case of the World Bank’s World Development Indicators (WDI), I assess the prevalence of revised data in the political science literature, replicate a prominent study to show that there are often substantial differences between revisions, and use tree-based machine learning to understand what predicts variation across data releases between 2005 and 2020. While low-income countries tend to produce less consistent data, even data reported by high-income countries might suffer from systematic bias. These findings reinforce the need to be transparent about the data collection process, as the data source and version we use might affect the conclusions we get.
GDP Growth in 2000 for Selected Countries: Difference in Values Reported by Different WDI Releases