jole-data-analysis

Data Analysis (jole-data-analysis)

When to trigger

• You are building the analysis sample from CPS/ACS/IPUMS, administrative, or register data
• You are running wage decompositions (Oaxaca / RIF) or AKM firm–worker models
• Standard errors, weighting, or robustness need to meet labor-referee expectations
• You want to make sure the empirical work will be replicable before you write it up

Labor empirical norms at JOLE

JOLE publishes empirical / simulation / experimental labor papers only if the data are documented and available for replication, so build the analysis so it can be deposited later (data + programs + documentation) to the JOLE Dataverse (see jole-replication-and-data-policy). Beyond reproducibility, labor referees expect disciplined data work:

• Sample construction is part of identification. Document the universe, age/labor-force restrictions, top-coding handling, and how you treat zeros/imputed earnings (CPS allocation flags, ACS PUMS edits). Report sample sizes at each restriction.
• Weights and design. Use survey weights appropriately (CPS/ACS) and account for complex sampling; for registers, be explicit about coverage and linkage rules.
• Earnings measures. Be precise: hourly vs. weekly vs. annual; nominal vs. real (state the deflator); winsorizing/top-coding decisions and their sensitivity.
• Standard errors. Cluster at the level of the variation (often state or firm); use heteroskedasticity-robust SEs by default; wild-cluster bootstrap with few clusters; randomization inference for experiments.

Common labor estimations (and their pitfalls)

• Wage decompositions: Blinder–Oaxaca for mean gaps; RIF / unconditional-quantile (rifreg) for distributional gaps. State the reference group and the index-number problem; do not over-interpret the "unexplained" component as discrimination without argument.
• Two-way (AKM) firm–worker FE: estimate on the connected set; correct limited-mobility bias (leave-out / KSS) before decomposing wage variance; report the share of movers.
• Labor-supply elasticities: be explicit about extensive vs. intensive margin, and about which elasticity (Marshallian/Hicksian/Frisch) is identified.
• Returns to schooling/training: distinguish OLS from IV/RDD estimates; report both and reconcile.
• Event studies / DID: use modern estimators on staggered timing (see jole-identification-strategy) and plot leads.

Robustness a labor referee will ask for

• Alternative samples (age bands, full-time/part-time, with/without imputed earnings)
• Alternative SE clustering and few-cluster corrections
• Specification curve / leave-one-out on key controls or sub-populations
• Placebo outcomes and placebo timing/cutoffs
• Heterogeneity by the labor-relevant dimensions (gender, education, age, sector) where theory predicts it

Execution bridge (StatsPAI / Stata MCP)

Run the battery, don't just enumerate it. Full map:

[execution-with-mcp](../../../shared-resources/empirical-methods/execution-with-mcp.md). JOLE is labor economics — the home of clean identification; DiD/IV/RDD and selection corrections are the binding constraint.

• Many outcomes / specifications: romano_wolf (step-down FWER) or benjamini_hochberg.
• OVB sensitivity: oster_delta / sensemakr.
• Inference: wild_cluster_bootstrap (few clusters), twoway_cluster / conley.
• Re-fit off one handle: audit_result(result_id) lists missing checks + the exact

suggest_function for each.

• Exhibits: etable / did_summary_to_latex from the handle — no retyped numbers.

Decisive checks in the body, exhaustive battery in the appendix.

[JF execution walkthrough](../../../Journal-of-Finance-Skills/resources/worked-examples/02-execution-walkthrough.md).

Checklist

• [ ] Sample restrictions documented with counts at each step
• [ ] Earnings measure and deflator stated; top-coding/winsorizing sensitivity shown
• [ ] Survey weights / register coverage handled correctly
• [ ] SEs clustered at the variation level; few-cluster issues addressed
• [ ] Decompositions report reference group; AKM corrects limited-mobility bias
• [ ] Robustness covers samples, SEs, placebos, and theory-motivated heterogeneity
• [ ] Every table/figure regenerable from a master script (replicability built in)

Anti-patterns

• Undocumented sample cuts that drive the result
• Ignoring CPS/ACS allocation flags and imputed-earnings issues
• Default i.i.d. SEs when variation is at the state/firm level
• Interpreting the Oaxaca "unexplained" gap as discrimination with no further argument
• Reporting AKM firm-effect dispersion without limited-mobility-bias correction
• Leaving reproducibility to the end instead of scripting it as you go

Output format

【Data】source(s) + sample universe + restrictions (with counts):
【Earnings measure】hourly/weekly/annual, real/nominal, deflator:
【Estimator】OLS / Oaxaca / RIF / AKM / IV / DID:
【SEs】clustering level + few-cluster handling:
【Robustness done】[samples, SEs, placebos, heterogeneity]:
【Replicability】master script regenerates all exhibits? [Y/N]
【Next step】jole-contribution-framing or jole-tables-figures