# Week 4 transfer memo: an incentive reform

**Length:** two pages plus a compact data/contract dictionary.  
**Name / setting / date:** …  
**Status:** research proposal; any illustrative numbers must be labeled synthetic.

## 1. Rule, population and decision

State the employer decision and exact contract term changed. Define the baseline eligible population, who knows the terms before work, announcement/work/payment dates, frozen targets, exceptions and outside options. Distinguish formal eligibility from realized receipt. State which technology, demand or staffing conditions are held fixed and which may change.

## 2. Competing explanations

| Mechanism | Distinct observable prediction | Counterprediction or possible falsifier | Additional observation needed |
|---|---|---|---|
| Mechanism 1: … | … | … | … |
| Mechanism 2: … | … | … | … |

Explain why the same headline performance gain could arise under both mechanisms. Examples include incumbent effort versus selection, task substitution versus productive learning, or payment credibility versus reciprocity. Avoid interpreting ratings, hours or pupil outcomes as direct observations of latent effort.

## 3. Variation and estimand

Describe randomization or a specific quasi-experiment, its assignment unit and permissible comparisons. Name the outcome target, weights, eligible risk set and analysis period. Define how people who leave, decline or move jobs remain represented in the participation ledger. Explain how variation separates recruitment, incumbent response and retention; state what remains unidentified. Do not condition the causal comparison on bonus receipt or remaining employed.

## 4. Outcomes and inference

| Outcome family | Measure and unit | Recorder and independence from pay rule | Timing / missingness |
|---|---|---|---|
| Prespecified primary | … | … | … |
| Independent quality | … | … | … |
| Participation / exit | … | … | … |
| Worker welfare | … | … | … |

Choose inference at the actual assignment unit, define the confirmatory family and multiplicity strategy, and explain interference or spillovers. Address delayed quality, returns, discounting or shifted work dates. State whether outcomes remain observed after exit; missing output is not automatically zero. A large p-value alone cannot establish equivalence.

## 5. Interpretation and access

State the proposed decision rule and evidence that would change your recommendation. Distinguish gross sales, contribution, full profit and welfare. Identify missing conditions behind an optimal contract or credible relational arrangement; an employer payment inequality alone is insufficient. List feasible data access and confidentiality arrangements without treating an archive listing as completed replication. Do not propose creating harmful breaches to test credibility.

## Compact data and contract dictionary

Add one row per requested field. Cover unit identifiers and merges; baseline eligibility; assignment; awareness; contract version; fixed pay/rate/threshold; work timing; payout timing; realized receipt; primary and quality outcomes; exits; welfare and relevant costs.

| Field | Unit / key | Producer and availability date | Known before work? | Definition / legal values | Risk set / missingness | Access status |
|---|---|---|---|---|---|---|
| … | … | … | … | … | … | … |

**Reproducibility note:** give the lab command, seed, modified files and exact mode. Distinguish computations already executed from proposed future data collection or empirical analysis.