Skip to content

econometrics

Category: review
Field: economics
License: private (curator-owned)
Updated: 2026-05-20
Stages: referee-simulation

Curator-private skill — copy text from 100xOS/shared/skills/review/econometrics.md.

Econometrics Review Checklist

Systematic checklist for evaluating empirical methodology. For detailed method-specific guidance, see the dedicated skill files (did.md, rdd.md, iv-estimation.md, panel-data.md, etc.).

1. Identification Strategy

  • Paper clearly states what it estimates (causal effect, descriptive, structural parameter)
  • Identifying assumption stated explicitly (words AND formally)
  • Most plausible threats discussed with evidence or arguments
  • Estimand well-defined (ATE, ATT, LATE? For whom?)
  • Direction of potential bias from assumption violations discussed

Method-Specific Quick Checks

DiD: Pre-trends shown (event study)? Staggered timing addressed? Anticipation tested? IV: First-stage F reported? Exclusion argued substantively? Reduced form shown? RD: Density test reported? Covariate continuity shown? Bandwidth robustness? RCT: Balance table? Attrition tested? Pre-registration?

2. Standard Errors and Inference

  • Clustering at treatment assignment level (report number of clusters)
  • Few clusters (<40): wild bootstrap or randomization inference
  • Multiple testing correction if many outcomes/subgroups tested
  • Spatial correlation addressed if geographically proximate units

3. Robustness

  • Alternative specifications (controls, functional form, FE)
  • Alternative samples (dropping outliers, subsamples)
  • Alternative variable definitions
  • Placebo treatments (wrong time/place)
  • Placebo outcomes (should-not-be-affected outcomes)
  • Sensitivity analysis (Oster 2019, Conley et al. bounds)
  • Leave-one-out for cross-region/country studies

4. Interpretation and Reporting

  • Economic significance discussed alongside statistical significance
  • Effect sizes in interpretable units (%, SD, benchmark comparison)
  • Sign and magnitude consistent with theory and prior literature
  • Confidence intervals or SEs reported (not just t-stats or p-values)
  • No bright-line p-value treatment (0.049 vs 0.051)
  • Null findings: minimum detectable effect reported

Quick Reference: Common Pitfalls

Pitfall Why it matters Check
Bad controls Post-treatment variables bias estimates Controls are pre-determined?
Wrong clustering Underestimates SEs Clustered at treatment level?
Staggered DiD + TWFE Wrong sign with heterogeneous effects Modern estimator used?
Weak instruments IV biased toward OLS F > 10? LIML reported?
p-hacking False positives Robustness, pre-registration?
Multiple testing False positives Outcomes/subgroups counted?
Log of zero Undefined; log(1+x) distorts Zeros in logged variables?
Survivorship bias Selected sample Entry/exit patterns checked?