U.S. Topical Economic Sentiment

Monthly mean across the 44 topics, one series per lexicon. The lexicons do not share a scale — compare shapes, not levels.

What it measures

Daily sentiment for a range of topics relevant to the U.S. economy, computed from a large news corpus using several lexicon-based methods. The indices are the empirical output of Questioning the news about economic growth (International Journal of Forecasting), which proposes generating thousands of candidate sentiment indices and letting a sparse regression select the ones that forecast — the methodology that became the sentometrics R package.

Winner of the International Journal of Forecasting best paper award 2018–2019.

Coverage

Frequency Daily
Period 2 January 1994 – 31 December 2017
Region United States
Series One value per date, lexicon and topic — long format, 1.37 million rows
Format .csv

Dates missing from the file mean no news articles were available for that date.

How to cite

Ardia, D., Bluteau, K., Boudt, K. (2019). Questioning the news about economic growth: Sparse forecasting using thousands of news-based sentiment values. International Journal of Forecasting, 35, 1370–1386. doi:10.1016/j.ijforecast.2018.10.010

BibTeX is on the How to cite page.

Terms

Free for academic research. By downloading you agree to cite the reference above, to place https://sentometrics-research.com in a footnote so others can find the data, and to assume all risk associated with its use.

Download

Gzipped CSV, one header row, ISO 8601 dates. No decompression step needed:

read.csv(gzfile(url("https://sentometrics-research.com/data/us-econ/us-topical-economic-sentiment.csv.gz")))

See the folder README for every column, and what differs from the archived original.

Original as published — the file this is derived from, byte for byte: Sentometrics_US_Topical_Economic_Sentiment.csv. It will move to a Zenodo deposit with a citable DOI.

Full licence