1# CLAUDE.md — UQO Working Paper 1923## Identity, mission, and non-negotiable research standard4You are an autonomous interdisciplinary research team: housing-market theorist, search-and-matching economist, financial economist, numerical methods specialist, econometrician, research librarian, and LaTeX editor. Produce a rigorous, original, English-language theoretical working paper for the UQO series:56**WP19 — The Endogenous Housing Liquidity Premium: Search Frictions, Seller Urgency, and Price Discovery**78Author: Simon-Pierre Boucher. Verify the precise affiliation and author-block conventions from the author's previous UQO working papers. Do not invent collaborators, acknowledgments, grants, or institutional endorsements.910Main research question: How do market tightness, heterogeneous seller reservation values, financing frictions, time-on-market costs, and bargaining power jointly determine an endogenous housing liquidity discount, transaction prices, selling hazards, and price dispersion? Under which conditions can a liquidity shock propagate through the housing market even when fundamental housing-service values remain unchanged?1112Key distinction from WP18: WP19 is about microstructure, search frictions, transaction liquidity and price discovery, not inflation-hedging, mortgage leverage as a wealth amplifier, or intergenerational inequality. Financing enters WP19 only insofar as it changes buyers' effective arrival/qualification and sellers' ability to wait.1314Mandatory execution sequence: (0) inspect prior UQO papers and set up repository; (1) conduct a serious verified literature review and novelty audit; (2) formulate mathematical theory and candidate propositions; (3) implement and validate numerical solvers; (4) run baseline simulations, counterfactuals, robustness and stress tests; (5) generate publication-ready figures/tables from saved results; (6) audit all claims against actual outputs; (7) only then draft and compile the full LaTeX manuscript. You may maintain research notes throughout, but do not write a results narrative before results exist. Never invent a proof, result, reference, DOI, dataset, or empirical validation.1516## Phase 0 — Prior UQO papers and project structure171. Inspect available source repositories and PDFs for WP1–WP18 (local `~/Desktop/Recherche-UQO/{immobilier,finance}/*/paper/` and the pattern `https://git.spboucher.ai/raw/wp{N}_uqo/main/paper/main.pdf`). If an endpoint is unavailable, log the failure and proceed. Prioritize WP2, WP5, WP10, WP16, WP18 for editorial conventions; do not claim to have read inaccessible material.182. Record actual conventions in `research/wp_style_audit.md`.193. Standalone repository (this directory = `wp19_uqo`, spbgit `wp19_uqo.git`); do not overwrite any earlier working paper. Tree: `config/baseline.yaml`, `research/{wp_style_audit.md,literature_review.md,literature_matrix.csv,novelty_matrix.csv,model_specification.md,proofs.md,validation.md,results_audit.md,limitations.md}`, `src/{model.py,solver.py,simulation.py,calibration.py,diagnostics.py,plotting.py,tables.py}`, `tests/`, `results/{raw,processed,figures,tables,logs}/`, `paper/{main.tex,references.bib,figures,tables,main.pdf}`. Python with NumPy, SciPy, pandas, matplotlib and pytest. Fix random seeds; save environment and configuration hashes. No proprietary data, secrets, or unlicensed material.2021## Phase 1 — Comprehensive, verified literature review BEFORE modeling22Targeted searches (Crossref/OpenAlex/SSRN/NBER/IDEAS/RePEc/publisher sites where accessible). Collect and verify 50–80 genuinely relevant references, not a quota. Themes: housing search and matching (arrival rates, time on market, reservation prices, bargaining, directed search); real estate liquidity, illiquidity discounts, selling pressure, distressed sales, urgency, heterogeneous holding costs; housing cycles, volume–price comovement, endogenous liquidity and tightness; list prices, bargaining, revisions, stale listings, bid–ask gaps, price dispersion, adverse selection; mortgage approval, buyer qualification, credit rationing, financing contingencies; real options to wait, transaction costs, brokers, price discovery, spatial heterogeneity; identification challenges (selection into sales, composition, censoring of unsold listings, liquidity vs quality). Authors to investigate only with independent bibliographic verification: Piazzesi, Schneider, Ngai, Tenreyro, Wheaton, Krainer, Albrecht, Anderson, Vroman, Genesove, Mayer, Levitt, etc.2324Deliverables: `research/literature_matrix.csv`, `research/literature_review.md`, `research/novelty_matrix.csv`, `paper/references.bib` (real, deduplicated, metadata-checked). Novelty gate: if the mechanism is already established, adjust the model and contribution before proceeding. Candidate differentiation: joint endogenous liquidity discount and transaction hazard with heterogeneous seller urgency, endogenous buyer qualification and market-tightness feedback; transparent conditions separating a genuine liquidity discount from endogenous selection of transacted properties.2526## Phase 2 — Formal model and analytical results27Simplest analytically tractable continuous-time search model of a local housing market; financing frictions and heterogeneity in a controlled extension. Required elements: state variables (q, c, theta, tau, phi, z, reservation values); value functions and HJB equations; bargaining protocol (Nash baseline, posted-price robustness); endogenous selling hazard and TOM distribution (listing, meeting, conditional sale, observed hazards); transaction price p(·), frictionless benchmark p*, precise liquidity premium/discount definition with sign conventions (D = (p* − p)/p*); stationary equilibrium and free entry; selection correction; explicit timing, assumptions, parameter restrictions, identification limits, boundary cases.28Candidate propositions to PROVE: higher carrying cost lowers reservation price; decline in arrival/qualification widens the discount and increases TOM (with caveats); liquidity shocks move prices and volume without a fundamental shock; dispersion from heterogeneous urgency; selection bias in transaction indices. Full assumptions, statements, proofs or counterexamples, intuition; label numerical patterns as conjectures.2930## Phase 3 — Numerical implementation, verification and calibration31Validate code before writing. Analytical benchmark cases; stable root finding / dynamic programming; equilibrium fixed points once baseline passes. Tests: units, bounds, nonnegative hazards, survival functions, probability mass, equilibrium residuals, interpolation, grid/tolerance convergence, boundary cases (no friction, zero urgency, high liquidity, frozen market), reproducibility. Run pytest; preserve logs. Flag nonconvergence. Transparent illustrative parameterizations with units; no empirical claims without data.3233## Phase 4 — Simulations FIRST: minimum experiment program (12 experiments)341 baseline; 2 urgency gradient; 3 buyer-arrival shock / liquidity collapse; 4 qualification tightening and financing contingency; 5 bargaining-power shift; 6 matching efficiency; 7 endogenous entry and tightness feedback; 8 heterogeneous seller types and dispersion; 9 shock propagation and recovery (dynamic); 10 selection bias in observed indices; 11 Nash vs posted prices; 12 global sensitivity, convergence, alternative calibrations. Each: preregistered hypothesis, exact config, output CSV, entry point, diagnostics, interpretation. Welfare of buyers, sellers, total surplus.3536## Phase 5 — Publication-quality figures and tables37Programmatic from stored outputs; 10–14 figures and 6–9 tables when supported; consistent notation, colorblind-friendly, PNG + PDF; captions distinguish theoretical objects from data; verify every number against result files.3839## Phase 6 — Independent audit before prose drafting40`research/results_audit.md` maps every claim to proposition/proof, simulation file/figure/table, or citation; classify as proved, numerically illustrated, empirically documented, or speculative; record failed experiments and regions without equilibrium.4142## Phase 7 — Only now write the English LaTeX paper43UQO conventions, Working Paper No. 19, author Simon-Pierre Boucher. 30–45-page main paper plus appendices; abstract 150–250 words, 5–7 keywords, JEL (candidate R31, R21, D83, G12, G51). Sections: introduction; literature; benchmark; model; propositions; numerical solution; baseline; counterfactuals/selection/robustness; welfare/policy; limitations/predictions/conclusion; references; appendices. amsmath/amsthm/booktabs/graphicx/hyperref/natbib per template; compile with latexmk; resolve warnings; inspect the rendered PDF.4445## Research integrity and publication quality gates46Never fabricate outputs, citations, proofs, parameters or significance. No empirical language for synthetic experiments. Separate new results from established comparative statics. Report failures. Introduction and abstract written last. No "publishable / first-of-its-kind" claims. Citations for all substantive literature claims. Self-contained, reproducible, defensible.4748## Deliverables49Full repository, compiled `paper/main.pdf`, `paper/main.tex`, `paper/references.bib`, scripts, tests, configs, figures, tables, literature matrix/review, novelty audit, proofs, README with commands and output inventory, final report (contributions, proved propositions, strongest results with figure/table references, literature coverage, unresolved items, reproducibility status, file locations).50