# Week 5 proposal: manager effects or authority reform

**Length:** two pages plus a compact data dictionary.  
**Name / setting / date:** …  
**Status:** proposed research; label all illustrative synthetic numbers.

## 1. Decision and estimand

Define one organizational decision. Is the target assignment to a person, a manager–team match, a managerial practice, a selection rule or a change in decision rights? State the population and outcome weights. Explain which tempting alternative interpretation this design does not identify.

## 2. Chronology, comparison and remaining threat

List vacancy, appointment decision, announcement, arrival, resource/task changes and outcome dates. Define baseline eligibility, treatment histories, comparison group, overlapping-event rules and repeat exposure. For L→H versus L→L or H→L versus H→H comparisons, define H/L using information independent of future focal outcomes. State the largest remaining matching or selection threat and a concrete observation that could reveal it.

## 3. Ledgers, outcomes and composition

Describe reporting links and decision rights separately: who initiates, recommends, approves, vetoes and appeals? Retain withdrawn and abandoned decisions as well as completions. Specify primary operational output, independently recorded quality, worker welfare, resources, headcount and exits. Distinguish a baseline cohort from all-current-worker outcomes, and average productivity from total output. Explain whether a post-treatment task control would remove part of the total effect.

## 4. Two mechanisms and a discriminating prediction

| Mechanism | Observable prediction | Counterprediction / possible falsifier | Required observation |
|---|---|---|---|
| Mechanism 1: … | … | … | … |
| Mechanism 2: … | … | … | … |

Explain why persistence or timing alone would not uniquely identify either mechanism. A manager fixed effect is not automatically a portable skill or a practice effect.

## 5. Inference, score construction and limits

Name the assignment/sampling unit, dependence structure, actual permitted assignments and clustering/resampling rule. Distinguish graph connectedness from exogeneity. If constructing scores, fix training dates, focal exclusions and independent folds at a defensible dependence unit. State the confirmatory outcome family and treatment of uncertainty; do not equate a large p-value with equivalence. Give one justified conclusion and one explicitly unjustified welfare, mechanism or external-validity claim.

## Compact data dictionary and access plan

| Field | Unit / join key | Recorder and timestamp | Availability before exposure | Definition / allowed values | Risk set / missingness | Access status |
|---|---|---|---|---|---|---|
| … | … | … | … | … | … | … |

Cover worker/team/site/manager spells, reporting changes, acting/concurrent roles, score provenance, decision rights, tasks, resources, outcomes and exits. State how joins avoid duplicating people or decisions. Classify sources as public, restricted, requiring a partnership or unavailable; a public archive pointer does not mean its underlying microdata are public.

**Reproducibility note:** record the executed lab command, seed and modified files. Separate completed synthetic calculations from proposed empirical work. No provider contact or data application is required.