---

mode: empirical
discipline: economics
title: "<Working title: an empirical study>"
---

# Research brief — <working title> (empirical)

## 1. Research objective and scope

- **Question.** <e.g., Does X affect Y in population P over period T?>
- **Hypotheses.** H1: <…>. H0 is a real possibility: a well-identified null result is a valid outcome.
- **Scope and unit of analysis.** <firms/countries/households…; frequency; sample window>.

## 2. Domain, mode and methodology

- Discipline: <economics/finance/statistics/…>. Primary mode: empirical.
- Methodology: <OLS/panel FE/DiD/IV/event study/GARCH/ML forecasting…>, chosen because <identification argument>.
- Inference: <clustered/robust SE, bootstrap, multiple-testing policy>.

## 3. Contribution and novelty

- <Expected contribution; the 3–5 closest papers to position against.>

## 4. Literature and source rules

- Indexes: OpenAlex + Crossref (+ Semantic Scholar). Target ≥ 15 verified sources; must-read: <DOIs>.
- Map every background claim to a verified key.

## 5. Data

- **Sources:** <exact dataset/API/series ids, e.g., FRED series XYZ; World Bank indicator ABC; SEC EDGAR filings; public Kaggle/Zenodo DOI>. Only legally usable, documented data; record license and access date.
- **Construction:** <variable definitions, cleaning rules, sample selection, treatment of missing values and outliers>.
- **Size limits:** keep raw + derived data under 200 MB in the workspace (subsample and document if larger).
- No synthetic data unless explicitly allowed here: <allowed/not allowed>.

## 6. Procedure (pre-registered)

1. Data acquisition and manifest (`data/manifest.json`, SHA-256, license).
2. Descriptive statistics table and at least one figure of the raw series/distribution.
3. Main specification(s): <exact equations/models, controls, fixed effects, sample>.
4. Secondary/exploratory analyses: <…> (label as exploratory).
5. Write `code/reproduce.sh` regenerating all tables/figures/metrics from raw inputs; log every run.

## 7. Baselines, robustness and falsification

- Robustness: <alternative controls, windows, estimators, placebo/pre-trend tests, leave-one-out, winsorization>.
- Report robustness results that weaken the finding. State what would falsify H1.

## 8. Reproducibility and artifacts

- Pinned environment (uv), seeds, `results/experiment-log.jsonl`, `results/metrics.json` with every reported value registered.

## 9. Manuscript

- AutoScholar class; tables with units/notes/standard errors; figures with legible axes; 10–20 pages.

## 10. Review and revision

- Reviewer focus: identification threats, standard-error validity, data leakage, p-hacking signals, claim-to-evidence gaps.

## 11. Limitations, ethics, legal

- <Measurement error, external validity, data license constraints; no personal data.>

## 12. Acceptance criteria

- Main specification estimated on the full sample with CIs; ≥ 3 robustness checks; reproduction script regenerates `metrics.json`; ≥ 15 verified citations; limitations stated.