# Week 10 research memo and data-access/validation appendix

Suggested submission: four pages of main text plus a concise appendix. This is a proposal, not a claim that a publishable paper or approved data access already exists.

## Page 1 — Economic decision, contribution, and claim

**Working title:**

**Decision-maker and choice:** What can the organization actually change?

**Labor outcome and economic question:** Do not start with a dataset or an estimator.

**Two competing mechanisms:** Write a distinct observable implication for each.

**Literature position:** One personnel application, one relevant method, and one data/resource reference. Name the unresolved distinction, not only a new country or dataset.

**Strongest feasible claim in one sentence:**

**Tempting but unsupported claim in one sentence:**

## Page 2 — Population, timing, records, and measurement

**Initial eligibility / observation unit / assignment unit:**

**Treatment and control regimes:** Specify components, access, timing, versions, and possible crossover. Keep assignment and usage separate.

**Primary outcome and independent outcome family:** Include units, denominator, horizon, data owner, creation time, and revisions.

**Selection and missingness:** Which candidates, workers, sites, or outcomes are not observed? How are exits handled?

**Event ledger:** Attach a mock schema with IDs and timestamps. Use synthetic records only unless real access is authorized.

## Page 3 — Comparison, inference, and falsification

**Source of identifying variation:** What comparisons are permitted, and why?

**Estimand:** Write the population, treatment/exposure regime, weighting rule, and horizon.

**Assumptions and largest remaining threat:** Separate assignment, measurement, interference, and attrition assumptions.

**Inference:** Number of independent assignment units; cluster/stratification logic; randomization test null or other uncertainty method; power/design simulation.

**Discriminating evidence:** What outcome pattern would weaken the favored mechanism? Which proposed “control” is actually post-treatment?

**Sensitivity:** Pre-specified plausible alternatives; no menu selected by significance.

## Page 4 — Portability, incidence, and feasible next step

**Source and target:** Whose effect is learned, and whose decision is being informed?

**Effect modifiers and support:** Which pre-existing worker/organization attributes plausibly matter? Does the target require an unobserved context or treatment version?

**Welfare/accounting:** Firm, initially eligible workers, entrants and external parties; keep leavers nested correctly; distinguish transfers and resource costs.

**Deliverable in the next month:** One bounded evidence product, not a promise of a completed paper.

**Access fallback:** What can proceed with public documentation, a released archive, or synthetic testing? What substantive claim must wait?

## Appendix A — Minimal data-access map

| Required object | Owner/source | Fields and keys | Timing/coverage | Sensitive content | Status: checked / pending / absent | Fallback |
|---|---|---|---|---|---|---|
| Eligibility and openings | | | | | | |
| Assignment or rollout | | | | | | |
| Production and audit | | | | | | |
| Pay and conditions | | | | | | |
| Exits and follow-up | | | | | | |
| Target context | | | | | | |

No data-provider contact, restricted application, employee recruitment, or data transfer is authorized by filling out this form.

## Appendix B — Validation and reproducibility plan

Specify one check for each: joins/IDs; time ordering; risk sets; recording changes; independent labels/audits; survey response; missing destinations; permissions; versioned input snapshots; source-to-output scripts; restricted-file documentation; and disclosure review.

State whether the proposed reproduction is exact, reduced, a published-statistic reconstruction, or synthetic. Keep these labels separate in every report.

## Appendix C — Source-to-target comparison

| Dimension | Study/source | Target | Why it may modify effects | Observed? | Consequence for claim |
|---|---|---|---|---|---|
| Workers/skills | | | | | |
| Tasks and review capacity | | | | | |
| Pay/authority | | | | | |
| Outside options | | | | | |
| Implementation/version | | | | | |
| Scale/peer exposure | | | | | |
| Horizon/outcome definition | | | | | |

## Response to peer review

One justified revision, one unresolved limitation, and the smallest extra measurement or comparison that would help. Do not fabricate access or convert uncertain assumptions into facts to improve the proposal's appearance.