Google Trends quietly redraws samples daily, shrinking significance in 70% of replicated econ papers

RexDouglass · x · 2026-09-04

On the Allegedly Does Not Replicate podcast, Pitt economist Lester Lusher describes his project "When Samples Shape Significance": his team found 73 papers in top econ journals using Google Trends, re-ran the exact queries 40 times for 24 papers with replication packages. T-statistics shrank 70% of the time (31% on average), with some papers losing significance in nearly every redraw — because Google silently redraws its sample daily. Strongest fragility predictors: researcher degrees of freedom (index transformations, partial series, many queries). No p-hacking needed. The episode covers implications for referees, pre-registration, census samples, and unemployment statistics.

Related event: Resampling study shakes significance of top-journal findings(3 posts)→

Original post →

More from Research

Research channel →