Revisiting the Trade Policy Uncertainty Index

Abstract

Newspaper coverage-based uncertainty measures are made popular by Baker et al. (2016), who created the Trade Policy Uncertainty (TPU) index by analyzing appearances of uncertainty-, economy-, policy- and trade-related search terms. This paper shows their set of search terms leads to systematic inaccuracies such as misclassification and omission of articles. I then construct an improved U.S. TPU index by expanding and modifying the set of search terms, and restricting attention to national newspapers. The new set of search terms uncovers 31 times more articles, and the individual newspaper-level series that aggregate up to the new TPU index are more highly correlated, hence agreeing more with each other about the movements of U.S. trade policy uncertainty. I also provide a detailed mapping between major U.S. trade policy events and the new U.S. TPU index, and shows that Baker et al. (2016)'s U.S. TPU index sometimes mistakes financial market and political uncertainty for trade policy uncertainty.