Communicating data uncertainty: experimental evidence for UK GDP

Event Date: 26 February 2020

Speaker:  James Mitchell, University of Warwick

Time:  4.15 pm

Location: Strathclyde Business School, Cathedral Wing CW404b

Many economic statistics are subject to data revisions. But initial estimates of GDP growth are commonly published without any direct quantitative indication of their uncertainty. To assess if and how the public and experts interpret and understand UK GDP data uncertainty, we conduct both a  randomised controlled experiment and a targeted expert survey. The surveys are designed to assess: (1) perceptions of the uncertainty in singlevalued GDP growth numbers; (2) the public's interpretation and understanding of uncertainty information communicated in different formats; and (3) how communicating uncertainty affects trust in the data and the producer of these data. We find that the majority of the public understand that there is uncertainty inherent in GDP numbers, but communicating uncertainty information improves the public’s understanding of why data revisions happen. It encourages them not to take GDP point estimates at face-value but does not decrease trust in the data. We find that it is especially helpful to  ommunicate uncertainty information quantitatively using intervals, density strips and bell curves.

Published: 12 February 2020