Skip to content

result-extraction

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

Curator-private skill — copy text from 100xOS/shared/skills/replication/result-extraction.md.

Skill: Result Extraction

You are cataloging empirical results from an academic paper for replication purposes.

What Counts as a "Result"

  • Every row in every regression table
  • Every panel in every table
  • Summary statistics (Table 1 is almost always summary stats)
  • Figures that show empirical findings (coefficient plots, event study plots)
  • Statistical tests mentioned in text (Hausman, Wald, F-statistics)
  • Robustness checks, even if in appendix

Priority Classification

  • Primary: Main specification that answers the core research question. Usually Table 3 or 4.
  • Secondary: Extensions, heterogeneity analysis, mechanism tests
  • Robustness: Alternative specifications, placebo tests, subsample analysis

Coefficient Extraction

  • Extract exact values: coefficient, standard error, t-statistic, p-value
  • Note significance: * p<0.10, ** p<0.05, *** p<0.01 (or paper's convention)
  • Note if SEs are robust, clustered, bootstrapped, etc.
  • Record N, R-squared, F-statistic for each specification

Model Identification

  • What is the dependent variable?
  • What are the independent variables (treatment, controls)?
  • What fixed effects are included?
  • How are standard errors computed?
  • Is it OLS, IV, DiD, RDD, probit/logit, GMM, etc.?

Replicability Assessment

For each result, assess: - High: Public data + clear methodology + code available - Medium: Public data + clear methodology, no code - Low: Proprietary data or unclear methodology