When Is There Enough Data to Create a Global Statistic ?

To monitor progress toward global goals such as the Sustainable Development Goals, globalstatistics are needed. Yet cross-country data sets are rarely truly global, creating a trade-off for producers ofglobal statistics: the lower is the data coverage threshold for disseminating global statistics, the more statistics canbe made available, but the lower is the accuracy of these statistics. This paper quantifies the availability-accuracytrade-off by running more than 10 million simulations on the World Development Indicators. It shows that if the fractionof the world’s population for which data are lacking is x, then the global value will on expectation be off by 0.37*xstandard deviation, and it could be off by as much as x standard deviations. The paper shows the robustness of thisresult to various assumptions and provides recommendations on when there is enough data to create global statistics.Although the decision will be context specific, in a baseline scenario, it is suggested not to create globalstatistics when there are data for less than half of the world’s population.