Economic uncertainty measures, experts and ChatGPT

The paper proposes a test for comparing the validity of different measures of economic uncertainty. The test verifies the randomness hypothesis for matching jumps in an uncertainty index with the dates of uncertainty-generating events named by independent experts or ChatGPT. It can also be applied to test whether ChatGPT provides a reliable selection of such events. The test has been used to compare four uncertainty indices constructed for Poland, two of which were developed for this paper. A panel of highly qualified experts was created to name and date uncertainty-generating events for Poland for the period from January 2021 to March 2021. The results confirm the validity of measures of uncertainty constructed using internet searches if the searches are done in the local language, which is Polish in this case. They also show that in cases where panels of experts are not available, ChatGPT provides a reasonable alternative. An example for the US confirms our results.