Measuring Swiss Uncertainty from Historical Text

Historical text archives are attractive for measuring uncertainty, but long-run archives change in size, composition, and language over time. Raw word counts therefore mix uncertainty with corpus growth and source turnover. I show how to recover a usable broad uncertainty measure from such a changing archive by handling comparability at the source level before aggregation. I implement that design for Switzerland and construct a quarterly series for 1848Q1–2022Q1 from newspapers, business publications, association material, and official reports. The paper also builds on a substantial archival effort that combines existing digitized sources with newly assembled and digitized firm and association material. The resulting series lines up with major historical stress episodes and is negatively related to modern Swiss activity benchmarks. Local projections and a benchmark recursive VAR show that the index also behaves sensibly in standard macro applications. JEL Codes: C38, D81, E32, N10