Making Data Measurement Errors Transparent: The Case of the IMF
Peter A.G. van Bergeijk
Published: September 2017
In 1950 Morgenstern pointed out that absolute precision and certainty are impossible in economic observations, but estimates are often hampered by a substantial degree of measurement error. Unlike the natural sciences, economists in general do not report measurement errors for the key concepts such as prices, value or production that it seeks to define, measure and explain. For most macroeconomic concepts two approaches are available: the Implicit Minimal Measurement Error and the Maximum Ratio. Studying different vintages of the IMF World Economic Outlook data base it was found that the estimates on average have an implicit minimal measurement error of 4.3% and maximum ratio of 17.9%. An agenda is proposed for removing disincentives (creating incentives) for stakeholders (academics, data collectors and producers) since reporting measurement error will result in better research, better policy and ultimately better data.