jleo-identification

Identification Strategy (jleo-identification)

When to trigger

• The causal claim rests on OLS + controls, or TWFE on staggered institutional reforms
• The unit chooses its own governance form (make-or-buy, integration, contract type) and the estimate confounds selection with effect
• The "treatment" is a court/committee/judge/case assignment whose as-good-as-random status is asserted but not shown
• A referee asks "what is the source of variation that identifies this institutional effect?"
• The design borrows institutional variation across countries/states/firms without ruling out the obvious confounds

The JLEO identification bar

JLEO empirics live or die on whether the institutional or organizational variation that does the identifying is credible — because the object of interest is usually itself chosen by the agents (firms pick governance forms; legislatures pick rules; litigants select into courts). The credible-identification standards are the same as any top empirical economics journal, but the threats are institution-specific: endogenous boundaries, endogenous reform timing, selection into contract form, and strategic case selection. Name the variation, defend it, and never let the institutional story substitute for the design.

Design branches and their institution-specific threats

Reforms / institutional change (DID, event study)

• With staggered adoption move beyond TWFE: Callaway–Sant'Anna, Sun–Abraham, de Chaisemartin–D'Haultfœuille; show clean event-study leads; report a Goodman–Bacon decomposition.
• The first-order JLEO threat is endogenous reform timing: institutions reform when conditions change. Defend timing exogeneity (or instrument it), and test for anticipation.

Governance-form / make-or-buy choice (the classic TCE empirics)

• The boundary of the firm is chosen. A raw comparison of integrated vs. arm's-length transactions confounds the governance effect with the selection that produced it.
• Use within-transaction variation, an IV for the governance choice, a selection model with a defensible exclusion, or matching on asset-specificity/uncertainty proxies — and say which, and why it breaks the selection.

Court / committee / judge / case assignment

• The cleanest JLEO-style natural experiment: random or rotation-based assignment of cases to judges/panels, or bills to committees.
• Show the assignment mechanism is actually as-good-as-random (balance on case characteristics); address case settlement/selection that re-introduces endogeneity downstream.

Cross-institutional variation (states, countries, firms)

• Defend the variation against the obvious omitted institutional confounds; prefer within-jurisdiction-over-time variation or a sharp discontinuity to a cross-section of institutions.

IV / RDD where applicable

• IV: strong first stage; weak-IV-robust inference (Anderson–Rubin) if weak; defend exclusion in the institutional context. RDD: density test (McCrary / Cattaneo–Jansson–Ma), bandwidth robustness, covariate smoothness, bias-corrected CIs.
• Cluster inference at the level of the institutional treatment; address few-cluster problems (wild-cluster bootstrap).

Execution bridge (StatsPAI / Stata MCP)

Estimate and audit the design, don't only describe it. Full map:

[execution-with-mcp](../../../shared-resources/empirical-methods/execution-with-mcp.md). JLEO is law-and-economics/organizations; institutional designs with endogeneity — foreground identification.

• detect_design → recommend → fit with as_handle=true → audit_result.
• Observational causal claims: staggered DiD (callaway_santanna / sun_abraham +

bacon_decomposition + honest_did_from_result); IV (effective_f_test +

anderson_rubin_ci); RDD (rdrobust + mccrary_test).

• Experiments: randomization-based inference + romano_wolf for many-outcome control.
• Sensitivity: oster_delta / sensemakr for observational claims.

Report the magnitude in interpretable units; route the full battery to the appendix. A

run end-to-end (synthetic data, real returns) is in the [JF execution walkthrough](../../../Journal-of-Finance-Skills/resources/worked-examples/02-execution-walkthrough.md).

Checklist

• [ ] The identifying variation is named in one sentence and tied to the institutional setting
• [ ] Selection into the governance form / institution is explicitly addressed, not assumed away
• [ ] Reform timing exogeneity or assignment-as-random is defended and tested (anticipation, balance)
• [ ] Staggered designs use a modern estimator; event-study leads shown flat
• [ ] Inference clustered at the institutional treatment level; few-cluster issues handled
• [ ] The causal claim never exceeds what the institutional variation supports

Anti-patterns

• Comparing integrated vs. non-integrated transactions as if governance form were randomly assigned
• TWFE on staggered institutional reforms with no heterogeneity-bias discussion
• Asserting judge/case/committee assignment is random without showing balance
• Cross-country institutional regressions with controls standing in for identification
• An "institutional story" doing the work the research design should do

Worked vignette (illustrative)

A paper estimates the effect of vertical integration on hold-up costs using firm transaction data. The naive comparison is biased: firms integrate precisely the transactions most prone to hold-up. A JLEO-credible move uses a shock to asset specificity (e.g., a technology change raising relationship-specific investment for a subset of input categories) as the source of variation, with an event study around the shock and a placebo on unaffected categories. Suppose integrated-transaction disputes fall 18% (s.e. 6, illustrative) post-shock for treated categories only — that pattern, not the cross-section, identifies the governance effect.

Output format

【Design】reform-DID / governance-choice / case-assignment / cross-institution / IV / RDD
【Identifying variation】one sentence, tied to the institution
【Selection/endogeneity threat + fix】[...]
【Diagnostics】event-study leads / balance / first-stage / density / placebo
【Inference】clustering level; few-cluster handling
【What it does NOT identify】[...]
【Next skill】jleo-theory-model