# WP21 — Eligibility criteria (PRISMA 2020 adapted)

# Population / object

Residential property: single-family houses, condominiums/apartments, multifamily buildings, rental units, mixed residential samples (recorded separately). Any country, 1970–2026.

# Inclusion (coefficient-based synthesis)

I1. Empirical study estimating a hedonic-type relationship between a residential price measure (transaction price, listing price, appraised/assessed value, rent) and at least one identifiable property, neighbourhood, environmental, location, market or AI-derived characteristic. I2. Reports at least one coefficient (or marginal effect / implicit price) with a standard error, t-statistic, p-value or confidence interval, or enough statistics to recover one transparently; reports the dependent-variable transformation and the sample size. I3. Peer-reviewed article, working paper, dissertation, book chapter or conference paper with a retrievable full text. I4. Any language (translated screening).

# Exclusion

E1. No residential price/rent outcome (commercial, land-only, agricultural, office). E2. No property-characteristic regressors (pure time-series house-price models, macro studies, index construction only). E3. Purely theoretical, methodological without an empirical application, or simulation-only. E4. Only predictive-performance metrics without interpretable marginal effects → retained for the AI/predictive synthesis (Section 9) but excluded from the coefficient synthesis. E5. No standard errors or recoverable inference → retained for the descriptive/qualitative synthesis only (flag). E6. Duplicate of an included study (same data, same models) → kept as a single study; versions linked. E7. Full text not retrievable → recorded as "not retrieved" with reason.

# Outcome families (never pooled across families)

Transaction price; listing price; appraised/assessed value; rent. Within a family, log and level dependent variables are harmonised into effect-size families (percentage premium, elasticity, semi-elasticity, monetary marginal price, partial correlation) per protocol/preregistration.md §4.

# Screening

Stage 1 (title/abstract): rule-based scoring (keywords in scripts/screening/keywords.yaml) → candidates; LLM-assisted screening by independent agents with a written reason for every decision; disagreements and "unclear" flagged for human review. Stage 2 (full text): eligibility checklist I1–I4 / E1–E7 recorded per study in data/metadata/screening_fulltext.csv.