Investigate bias from mis-specified lag structures between policy announcements and epidemic outcomes

Investigate whether mis-specifying the lag length between national policy announcements and observed epidemic outcomes in time-series regression models biases estimated policy effects, and characterize the nature and direction of any such bias under plausible epidemic dynamics.

Background

The paper argues that 2-week and 4-week lags used between policy announcements and outcomes (cases or deaths) are implausibly short given biological and reporting delays. The author raises the question of whether lag mis-specification can bias results, potentially even flipping the apparent direction of policy effects.

This uncertainty is explicitly acknowledged as unclear, motivating a need to formally analyze the consequences of lag mis-specification in epidemic policy-effect regressions.

References

But the question is whether mis-specifying the time-structure of the relationship in the regression can bias outcomes? It is unclear but consider the example in the figure below.

— Analysis of Potential Biases and Validity of Studies Using Multiverse Approaches to Assess the Impacts of Government Responses to Epidemics  (2409.06930 - Goldhaber-Fiebert, 2024) in Approach and Findings – Bias Due to Specification?