AI Cable News-based EPU Indexes

Download Data                                   Annotated Chart

We extract economic policy uncertainty from the major US cable news networks (CNN, Fox News, and MSNBC) using the Stanford Cable TV News Analyzer. We begin with the keywords used in Baker, Bloom and Davis (2016) (BBD) and compute a daily measure (from 4pm on day t-1 until 4pm on day t) of minutes that economic policy uncertainty (EPU) is discussed across the three cable news networks. In our search, we set the threshold for nearness to 30 seconds such that the words from the different strings must be mentioned within a 30-second interval. BBD creates their index by searching newspapers for the triple text string containing terms to identify the economy, policy, and uncertainty. This trio is formed by E AND P AND U where 1) [E] "economic" or "economy"; 2) [P] one or more of "Congress," "deficit," "Federal Reserve," "legislation," "regulation" or "White House; and 3) [U] "uncertain" or "uncertainty."

We provide 5 indices. We construct all 4 combinations of BBD_EPU: E AND P (TV_EP), P AND U (TV_PU), E AND U (TV_EU) and E AND P AND U (TV_EPU). We also construct an index with an expanded set of BBD terms. Specifically, we expand E and U with the incremental terms used in Baker, Bloom, Davis and Renault (2021) and include 'risk' and 'risky' in the U set (TV_EPU_EXP).

If using these indices please cite both papers below:

• Bergbrant and Bradley, 2023. Did they just say that? Using AI to extract economic policy uncertainty from cable news networks
• 2) Hong, Crichton, Zhang, Fu, Ritchie, Barenholtz, Hannel, Yao, Murray, Moriba, Agrawala, and Fatahalian, 2021. 'Analysis of Faces in a Decade of US Cable TV News,' ACM Conference on Knowledge Discovery and Data Mining (KDD)

The data can be downloaded above. We plan to update these measures quarterly.