Transparency, Reproducibility, and the Credibility of Economics Research

Causal Question / Estimand

Meta-scientific: not a treatment effect but the credibility of the body of empirical estimates — how research practices distort the published distribution of findings and what reforms restore trust.

Identification Strategy

Surveys the evidence that current practice produces biased bodies of evidence: publication bias (significant results selectively published), specification searching / p-hacking (researcher degrees of freedom), and underpowered studies (Statistical-Power) that, conditional on significance, exaggerate effects. Reviews the reform toolkit: study registration and [[Pre-Analysis-Plan|pre-analysis plans]], disclosure and reporting standards, replication/reproducibility, and open data and materials. The unifying logic is constraining the analytic flexibility that divorces published estimates from the true effect distribution.

Key Assumptions

n/a — review/meta-science, not an identified causal estimate. Implicitly invokes the Statistical-Power and selection mechanisms that bias literatures.

Threats to Validity

The paper catalogues the threats to the credibility of empirical work: selective reporting, publication bias, low power, non-reproducibility, and the incentives that sustain them.

Setting / Data

n/a — survey of economics (and adjacent social-science) research practices.

Key Claims

  • Publication bias, specification search, and low power jointly make a substantial share of published findings unreliable.
  • Pre-registration and pre-analysis plans, replication, and open data/disclosure are the core tools for restoring credibility.
  • These reforms are the methodological complement to the design-based “credibility revolution” — credible designs still need credible reporting.

Connections

Citation

Christensen, G., & Miguel, E. (2018). Transparency, Reproducibility, and the Credibility of Economics Research. Journal of Economic Literature, 56(3), 920–980.