Absolute levels and direction of pre-treatment self-censorship

Determine whether scholars self-censored their research before the January–February 2025 U.S. policy shock and, if so, identify the direction of that pre-treatment self-censorship, which the paper’s relative-reaction design cannot establish.

Background

The empirical strategy compares researchers at universities with different levels of federal-funding dependence before and after the U.S. government’s 2025 scrutiny of research containing flagged words. This difference-in-differences design identifies changes associated with the policy shock relative to a comparison group, rather than the absolute prevalence of self-censorship.

Consequently, the authors explicitly leave unresolved whether researchers had already suppressed or altered their research content before the treatment period and whether any such pre-existing response increased or decreased the use of politically sensitive topics. Resolving this issue would require a design capable of identifying the baseline level and direction of self-censorship.

References

Our setting allows us to detect to what extent academics react to political pressure, but we are not able to make general statements about the absolute levels of self-censorship, i.e., our data would not allow us to know if, during the pre-treatment period, scholars have self-censored themselves, and if yes, in what direction.

— Science under Threat? A Natural Experiment in Economics  (2608.14164 - Rohner et al., 14 Aug 2026) in Section 2, “Empirical Framework” (within Section 3.2, Methods)