# Progress — WP19 (The Endogenous Housing Liquidity Premium)

## 2026-10-20 — Phase 0 (setup)
- Repository created at `~/Desktop/Recherche-UQO/immobilier/endogenous-housing-liquidity-premium` from `_modele-working-paper/`
  (HARMONISATION.md standard); CLAUDE.md = research brief; local git initialised (`main`).
- Toolchain checked: latexmk/pdflatex/bibtex (TeX Live), Python 3.14.4, numpy 2.4.4, scipy 1.17.1, pandas 3.0.2,
  matplotlib 3.10.9, pyyaml 6.0.3, pytest 9.1.1. Crossref API reachable (test query resolved Krainer 2001 → 10.1006/juec.2000.2180).
- BLOCKER (logged): SSH to the MacLustr cluster (`gitsrv` = M1M32 via ProxyJump m4mn) fails from the laptop at 07:25
  (`kex_exchange_identification: Connection closed`). The spbgit remote `wp19_uqo.git` cannot be created yet; work continues
  in local git, push deferred. OpenAlex MCP server failed to connect (CONNECTION_CLOSED); Crossref used instead.
- Raw endpoint check: `https://git.spboucher.ai/raw/wp2_uqo/main/paper/main.pdf` → 200; `wp18_uqo` → 404 (WP18 has no PDF yet).
  Local copies of all 17 compiled WPs are available in `Recherche-UQO/{immobilier,finance}`; they are the primary style source.

## 2026-10-20 — Phase 1 (literature) : DONE
- Two parallel verified searches (theory 73 refs, evidence 66 refs) merged into `literature_matrix.csv` (132 refs, 129 DOIs, 3 evidence
  files), `paper/references.bib` generated and cross-checked against raw Crossref JSON (`literature_merge_report.md`),
  `literature_review.md`, `novelty_matrix.csv`, `inaccessible_sources.md`. Novelty gate passed with explicit labels (P1 and the
  directions of P2–P4 are benchmarks; new: joint model, Pareto condition, index-selection sign condition, cash-gap decomposition).

## 2026-10-20 — Phase 2 (theory) : DONE — `model_specification.md`, `proofs.md` (3 lemmas, 8 propositions)
- Corrections during Phase 3: Pareto counterexample condition sharpened to $c>\delta w_0+(r+\delta)U$ (the initial statement
  $\beta\eta>(a-1)(r+\delta)$ is equivalent but was wrongly believed reachable for all $c$); Prop. 6(iii) sign condition made
  distribution-dependent after the truncated-normal baseline showed dampening; Prop. 7(ii) made conditional on $v^*_m>v^*_c$.

## 2026-10-20 — Phases 3–5 : DONE
- Solvers, 24 tests passing (`results/logs/pytest_full.log`), Monte Carlo validation (0.39 %), 12 experiments (38 result files),
  14 figures, 21 table fragments + `numbers.tex`. Financing solver rewritten (fixed point diverged → 3-equation root).

## 2026-10-20 — Phase 6 : DONE — `results_audit.md` (24 claims mapped, 6 weakened/removed), `limitations.md`, `validation.md`.