révision scientifique complète (editorial board, 2026-10-17)

v1.2 → v1.3 : passage au pipeline reproductible (N 140 931 → 142 121), province dérivée
de la RTA (11,7 % d'annonces mal étiquetées corrigées), décomposition Shapley–Owen,
inférence alternative + bootstrap wild-cluster par province, spline, holdout spatial par
RTA, sensibilité de Moran, effets type/tenure (Kennedy), 91 références vérifiées (Crossref),
annexe (construction des variables, réconciliation v1.2/v1.3), README, docs/REVIEW-REPORT.md
et docs/CHANGELOG-REVISION.md.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

Simon-Pierre Boucher committed 22 days ago (Oct 18, 2026) parent 8eb9bf6 Browse files

78 changed files +4,381 −2,079

modified README.md +154 −134

		`@@ -1,23 +1,26 @@`
1	1	# 🏘️ A Grand Hedonic Model of the Canadian Housing Market
2	2
3	3	UQO Working Paper No. 9 — *Decomposing the Value of Structure and Location across
4		−140,931 MLS Listings with High-Dimensional Neighbourhood Fixed Effects*
	4	+142,121 MLS Listings with High-Dimensional Neighbourhood Fixed Effects*
5	5
6		−[![Paper](https://img.shields.io/badge/paper-PDF%20(34%20p.)-b31b1b?logo=latex&logoColor=white)](paper/main.pdf)
	6	+[![Paper](https://img.shields.io/badge/paper-PDF%20(52%20p.)-b31b1b?logo=latex&logoColor=white)](paper/main.pdf)
7	7	[![Python](https://img.shields.io/badge/python-3.11%2B-3776AB?logo=python&logoColor=white)](requirements.txt)
8	8	[![DuckDB](https://img.shields.io/badge/data-DuckDB%20·%20172%2C019%20listings-FFF000?logo=duckdb&logoColor=black)](data/raw/README.md)
9	9	[![Estimation](https://img.shields.io/badge/estimation-statsmodels%20·%20linearmodels%20(AbsorbingLS)-4051b5)](src/wp9/models.py)
10	10	[![Reproducible](https://img.shields.io/badge/reproducible-end--to--end%20pipeline-2e7d32)](scripts/)
11		−[![Tables verified](https://img.shields.io/badge/tables-identical%20to%20published-2e7d32)](docs/AUDIT.md)
	11	+[![Revised](https://img.shields.io/badge/revised-editorial%20board%202026--10--17-2e7d32)](docs/REVIEW-REPORT.md)
12	12	[![Institution](https://img.shields.io/badge/UQO-D%C3%A9pt.%20des%20sciences%20administratives-16365c)](https://uqo.ca)
13	13	[![Author](https://img.shields.io/badge/author-Simon--Pierre%20Boucher-16365c)](mailto:simon-pierre.boucher@uqo.ca)
14	14	[![Contact](https://img.shields.io/badge/contact-simon--pierre.boucher%40uqo.ca-a02020?logo=maildotru&logoColor=white)](mailto:simon-pierre.boucher@uqo.ca)
15	15
16		−> TL;DR — On 140,931 Canadian MLS listings, absorbing 1,153 neighbourhood (FSA) fixed
17		−> effects lifts explained log-price variance from 46% → 77%. Location alone is worth
18		−> ~30 percentage points of R² — more than every structural attribute combined. The model
19		−> values held-out homes with a median absolute error of 15.8% (OOS R² = 0.764),
20		−> competitive with commercial AVMs, while staying fully transparent.
	16	+> TL;DR — On 142,121 Canadian MLS listings (asking prices), absorbing 1,155
	17	+> neighbourhood (FSA) fixed effects lifts explained log-price variance from 44% → 78%.
	18	+> A Shapley–Owen decomposition assigns ≈34 pp to structure and ≈36 pp to location: the
	19	+> two are of comparable importance, and location is the single largest block once province
	20	+> and neighbourhood are grouped (the sequential ladder would say 44 vs 32). Living-area
	21	+> elasticity 0.53, +12% per full bathroom, bedrooms ≈ 0. Out of sample: MdAPE 15.6%
	22	+> (R² 0.77) in neighbourhoods seen in training, 17.3% (R² 0.74) in neighbourhoods never
	23	+> seen when the FSA effect is interpolated from its neighbours.
21	24
22	25	---
23	26
		`@@ -25,10 +28,10 @@`
25	28
26	29	\| Item \| Value \|
27	30	\|---\|---\|
28		−\| Version \| 1.2 (May 2026) \|
29		−\| SSRN \| [7626678](https://ssrn.com/abstract=7626678) — submitted 2026-10-17 (preliminary upload) \|
	31	+\| Version \| 1.3 (October 2026) — revised 2026-10-17 by the editorial-board review ([`docs/REVIEW-REPORT.md`](docs/REVIEW-REPORT.md), [`docs/CHANGELOG-REVISION.md`](docs/CHANGELOG-REVISION.md)); v1.2 (May 2026) archived as `docs/main_pre-revision-20261017.pdf` and git tag `pre-revision-20261017` \|
	32	+\| SSRN \| [7626678](https://ssrn.com/abstract=7626678) — v1.2 uploaded 2026-10-17; v1.3 to be uploaded (title N changed 140,931 → 142,121) \|
30	33	\| Journal \| not yet submitted \|
31		−\| Code & data \| reproducible from `scripts/` (see Quick start); raw DuckDB snapshot not redistributed \|
	34	+\| Code & data \| fully reproducible from `scripts/` on the committed `data/processed/analysis.parquet`; raw DuckDB snapshot not redistributed \|
32	35
33	36	## 📖 Table of contents
34	37
		`@@ -40,12 +43,11 @@`
40	43	5. [The pipeline, step by step](#-the-pipeline-step-by-step)
41	44	6. [Data](#-data)
42	45	7. [Methodology](#-methodology)
43		−8. [The two results tiers (reference vs. reproduced)](#-the-two-results-tiers-reference-vs-reproduced)
	46	+8. [The two results tiers (reproduced vs. reference)](#-the-two-results-tiers-reproduced-vs-reference)
44	47	9. [Figures & tables inventory](#-figures--tables-inventory)
45		−10. [Reproduction verification](#-reproduction-verification)
46		−11. [Limitations](#-limitations)
47		−12. [Citation](#-citation)
48		−13. [Author & contact](#-author--contact)
	48	+10. [Limitations](#-limitations)
	49	+11. [Citation](#-citation)
	50	+12. [Author & contact](#-author--contact)
49	51
50	52	---
51	53
		`@@ -59,50 +61,60 @@ estimates a semi-logarithmic hedonic price equation at national scale for Ca`
59	61	ln P_i = α + β·ln(Area_i) + x_i'γ + d_i'δ + μ_f(i) + ε_i
60	62	```
61	63
62		−where `μ_f(i)` is a fixed effect for the Forward Sortation Area (FSA — the first three
63		−characters of the postal code) of listing i. With 1,153 absorbed FSA intercepts, every
	64	+where `P_i` is the asking price of listing i and `μ_f(i)` is a fixed effect for its
	65	+Forward Sortation Area (FSA — the first three characters of the postal code). With 1,155
	66	+absorbed FSA intercepts (1,146 FSAs + 9 province-level pools of small FSAs), every
64	67	neighbourhood gets an arbitrary price level that soaks up schools, transit, coastline,
65	68	employment density — observed or not — and the structural implicit prices (β, γ) are
66	69	identified purely from within-neighbourhood variation. Estimation is by absorbing
67		−least squares (numerically identical to full-dummy OLS); inference is clustered by FSA.
68		−
69		−Beyond the headline decomposition the paper delivers: out-of-sample valuation accuracy
70		−(80/20 split, Duan-smeared retransformation), robustness across six sample cuts, quantile
71		−hedonic regressions, a quadratic test of diminishing returns to floor space, an urban
72		−price gradient in distance to the nine major metros, Moran's I spatial diagnostics,
73		−leave-one-province-out transferability, and a ranked map of Canada's most and least
74		−expensive neighbourhoods net of structure.
	70	+least squares (numerically identical to full-dummy OLS); baseline inference is clustered by FSA.
	71	+
	72	+Beyond the headline decomposition the paper delivers: a Shapley–Owen decomposition of
	73	+R² across regressor blocks (order-free alternative to the sequential ladder); inference
	74	+under HC1 / FSA-cluster / province-cluster / two-way SEs and a **province-level wild-cluster
	75	+bootstrap*; out-of-sample valuation accuracy (repeated random 80/20 splits and* a
	76	+spatial holdout of entire FSAs with interpolated effects, Duan-smeared retransformation);
	77	+robustness across eight sample cuts; quantile hedonic regressions; quadratic and spline
	78	+tests of diminishing returns to floor space; an urban price gradient in distance to nine
	79	+metros (pooled, metro intercepts, quadratic); Moran's I diagnostics with k-sensitivity;
	80	+leave-one-province-out transferability; dwelling-type/ownership effects with the Kennedy
	81	+correction; and a ranked map of Canada's most and least expensive neighbourhoods net of
	82	+structure.
75	83
76	84	## 🏆 Headline results
77	85
78	86	\| Quantity \| Value \|
79	87	\|---\|---\|
80		−\| Estimation sample \| 140,931 listings — 82,334 houses, 57,857 condos, 9 provinces \|
81		−\| Absorbed neighbourhood effects \| 1,153 FSAs \|
82		−\| R²: structural only (M1) \| 0.464 \|
83		−\| R²: + dwelling type & ownership (M2) \| 0.469 \|
84		−\| R²: + province FE (M3) \| 0.567 \|
85		−\| R²: grand model, FSA FE (M5) \| 0.767 \|
86		−\| Living-area elasticity (M5) \| 0.547 (cluster SE 0.009) \|
87		−\| Full-bathroom premium \| 0.109 log points ≈ +11% per bathroom \|
88		−\| Bedrooms, conditional on area \| ≈ 0 (the textbook hedonic result) \|
89		−\| Out-of-sample R² (log price) \| 0.764 \|
90		−\| Median absolute valuation error \| 15.8% (59% of homes priced within ±20%) \|
91		−\| Moran's I of residuals, structural → grand \| 0.46 → 0.08 (−82%) \|
92		−\| Urban gradient \| −8.5% location premium per doubling of distance to metro \|
93		−\| Neighbourhood premia span \| ×9 between the most and least expensive FSAs \|
94		−
95		−The most expensive neighbourhoods net of structure are all in the City of Vancouver
96		−(V6S, V8E, V6T: +150–200% vs. the national median); the cheapest are in rural
97		−Saskatchewan, Manitoba and Newfoundland (−60 to −67%).
	88	+\| Estimation sample \| 142,121 listings — 83,524 houses, 57,853 condos, 744 other; 9 provinces (province from FSA letter) \|
	89	+\| Absorbed neighbourhood effects \| 1,155 (1,146 FSAs ≥ 25 listings + 9 pooled residual categories) \|
	90	+\| R² houses: M1 structural / M2 +type / M3 +province / M4 +FSA \| 0.467 / 0.470 / 0.605 / 0.769 \|
	91	+\| R²: grand model, all residential, FSA FE (M5) \| 0.776 (structure-only on the same sample: 0.436) \|
	92	+\| Variance decomposition (pp of R², all residential) \| sequential: structure 43.6, type 1.6, province 14.3, FSA 18.1 · Shapley: 33.0 / 7.5 / 8.9 / 28.2 · Owen: structure 34.0, type 7.3, location 36.3 \|
	93	+\| Living-area elasticity (M5) \| 0.529 (FSA-cluster SE 0.009; province-cluster 0.037; wild-cluster p = 0.004) \|
	94	+\| Full-bathroom premium \| 0.112 log points ≈ +11.9% per bathroom (0.148 ≈ +16% on houses with province FE) \|
	95	+\| Bedrooms, half baths (conditional on area) \| −0.006, −0.034 — not significant under province-level inference \|
	96	+\| Dwelling type (vs apartment, same size & FSA) \| house +46%, triplex +43%, fourplex +39%, duplex +16%, townhouse ≈ 0 (Kennedy-corrected) \|
	97	+\| Out-of-sample, random 80/20 (within support) \| R² 0.772, MdAPE 15.6%, 60% within ±20% (5 seeds: 0.769 ± 0.010, 15.5 ± 0.2%) \|
	98	+\| Out-of-sample, held-out FSAs (20% of neighbourhoods) \| province-mean effect: R² 0.52, MdAPE 25% · IDW of 5 nearest FSAs: R² 0.74, MdAPE 17.3% \|
	99	+\| Moran's I of residuals, structural → grand (k = 10) \| 0.50 → 0.06 (−87%); k = 5: 0.52 → 0.11; k = 20: 0.47 → 0.03 \|
	100	+\| Floor space \| quadratic: elasticity 0.63 at 60 m² → 0.41 at 350 m²; spline: 0.57 / 0.65 / 0.58 / 0.41 / 0.48 by quintile; ΔR² = 0.001 \|
	101	+\| Urban gradient \| −0.082 per log unit of distance ⇒ −5.5% per doubling; +21% within 5 km, −38% beyond 320 km \|
	102	+\| LOPO transfer (within-province R²) \| mean 0.34; MB/AB/NB/ON/SK/NS 0.46–0.52; QC 0.19; BC −0.49 \|
	103	+\| Neighbourhood premia span \| ×10.5 between extremes (V6S +226% vs A0B −69%), ×6.7 between P99 and P1 \|
	104	+
	105	+The most expensive neighbourhoods net of structure are on Vancouver's west side (V6S
	106	+Dunbar–Point Grey, V6T UBC), in Whistler (V8E) and West Vancouver (V7T); the first
	107	+Ontario FSA (M5R, the Annex) ranks 14th. The cheapest are in rural Newfoundland (A0B, A0G),
	108	+Saskatchewan (S0N, S0C, S0H, S0A), Manitoba and New Brunswick.
98	109
99	110	## 📁 Repository layout
100	111
101	112	```
102	113	wp9_uqo/
103	114	├── README.md ← you are here
104		−├── docs/ AUDIT.md (forensic audit + verification), CHANGES.md,
105		−│ PAPER_REVIEW.md, UPGRADE_REPORT.md
	115	+├── docs/ REVIEW-REPORT.md + CHANGELOG-REVISION.md (editorial-board
	116	+│ revision, 2026-10-17), main_pre-revision-20261017.pdf (v1.2),
	117	+│ AUDIT.md, CHANGES.md, PAPER_REVIEW.md, UPGRADE_REPORT.md (history)
106	118	├── requirements.txt pinned Python dependencies
107	119	├── .gitignore
108	120	├── data/
		`@@ -121,19 +133,22 @@ wp9_uqo/`
121	133	│ ├── 01_build_sample.py
122	134	│ ├── 02_estimate_core.py
123	135	│ ├── 03_estimate_extended.py
124		−│ ├── 04_make_figures.py (--results reference\|reproduced)
125		−│ └── 05_make_tables.py (--results reference\|reproduced)
	136	+│ ├── 04_make_figures.py (--results reproduced\|reference; default reproduced)
	137	+│ ├── 05_make_tables.py (--results reproduced\|reference; default reproduced)
	138	+│ └── 06_revision_analyses.py Shapley/Owen, SE comparison + wild bootstrap, spline,
	139	+│ spatial holdout, Moran sensitivity, metro gradient
126	140	├── figures/ all 17 paper figures (PNG, generated)
127	141	├── results/
128		−│ ├── reference/ ORIGINAL published outputs — canonical paper numbers
129		−│ ├── reproduced/ outputs regenerated end-to-end by this pipeline
130		−│ └── tables/ the 6 LaTeX tables consumed by the paper
	142	+│ ├── reference/ archived outputs of the May 2026 version (v1.2)
	143	+│ ├── reproduced/ outputs of the current pipeline — THE PAPER'S NUMBERS (v1.3)
	144	+│ └── tables/ the 13 LaTeX tables consumed by the paper
131	145	└── paper/
132	146	├── main.tex preamble + metadata; \input's the sections
133		− ├── main.pdf compiled paper (34 pages)
	147	+ ├── main.pdf compiled paper (52 pages)
134	148	├── sections/ titlepage, introduction, literature, data,
135		− │ methodology, results, robustness, conclusion
136		− ├── references.bib 32 entries, natbib author-year, aer style
	149	+ │ methodology, results, robustness, discussion, conclusion
	150	+ ├── appendix/appendix.tex variable construction, Moran sensitivity, v1.2 ↔ v1.3 reconciliation
	151	+ ├── references.bib 91 entries (all Crossref-verified, DOIs), natbib author-year, aer style
137	152	├── Makefile / .latexmkrc build config
138	153	└── uq_logo.jpg
139	154	```
		`@@ -149,15 +164,17 @@ python3 -m pip install -r requirements.txt`
149	164	python3 scripts/01_build_sample.py # duckdb → data/processed/analysis.parquet
150	165	python3 scripts/02_estimate_core.py # M1–M5 ladder → results/reproduced/
151	166	python3 scripts/03_estimate_extended.py # OOS, robustness, quantile, Moran, LOPO…
	167	+python3 scripts/06_revision_analyses.py # Shapley/Owen, SE comparison, spline, spatial holdout…
152	168	python3 scripts/04_make_figures.py # figures/ (17 PNG)
153		−python3 scripts/05_make_tables.py # results/tables/ (6 .tex)
	169	+python3 scripts/05_make_tables.py # results/tables/ (13 .tex)
154	170
155	171	# Paper:
156	172	cd paper && latexmk -pdf main.tex # or `make`
157	173	```
158	174
159		−Without the raw DuckDB you can still run steps 02→05: the committed
	175	+Without the raw DuckDB you can still run steps 02→06: the committed
160	176	`data/processed/analysis.parquet` (6 MB) is the estimation sample produced by step 01.
	177	+The whole pipeline (01→06) runs in about 2 minutes on an Apple M5 Max.
161	178
162	179	## 🔬 The pipeline, step by step
163	180
		`@@ -165,9 +182,10 @@ Without the raw DuckDB you can still run steps 02→05: the committed`
165	182	\|---\|---\|---\|---\|---\|
166	183	\| 01 \| `01_build_sample.py` \| raw DuckDB (172,019 rows) \| `analysis.parquet` (142k rows, 21 cols) \| ~40 s \|
167	184	\| 02 \| `02_estimate_core.py` \| parquet \| `fit.json`, `coef_M{1,2,3,5}.csv`, `grand_model.parquet`, `fsa_premia.csv` \| ~1 min \|
168		−\| 03 \| `03_estimate_extended.py` \| parquet + step 02 \| `oos.json`, `robustness.csv`, `heterogeneity.csv`, `quantile.csv`, `lopo.csv`, `ext2.json`, Moran arrays, gradient/nonlinearity bands \| ~6 min \|
	185	+\| 03 \| `03_estimate_extended.py` \| parquet + step 02 \| `oos.json`, `robustness.csv` (8 cuts), `heterogeneity.csv`, `quantile.csv`, `lopo.csv`, `ext2.json`, Moran arrays, gradient/nonlinearity bands \| ~10 s \|
	186	+\| 06 \| `06_revision_analyses.py` \| parquet + step 02 \| `shapley.csv`, `se_comparison.csv`, `spline.csv`, `spatial_holdout.json`, `moran_sensitivity.csv`, `gradient_metro.json` \| ~35 s \|
169	187	\| 04 \| `04_make_figures.py` \| results tier + parquet \| 17 PNG figures \| ~30 s \|
170		−\| 05 \| `05_make_tables.py` \| results tier \| 6 LaTeX tables \| ~5 s \|
	188	+\| 05 \| `05_make_tables.py` \| results tier \| 13 LaTeX tables \| ~5 s \|
171	189
172	190	\* Apple Silicon, single process.
173	191
		`@@ -181,9 +199,13 @@ Without the raw DuckDB you can still run steps 02→05: the committed`
181	199	`"50 x 120"` frontage×depth) with a missingness indicator; the top 1% of positive
182	200	parsed lots (multi-acre rural acreage strings) treated as "no usable lot info";
183	201	- dwelling type consolidated to 8 groups, ownership to 6 groups;
184		−- FSA from the postal code; FSAs with < 25 listings pooled into `<PROV>_other`;
	202	+- FSA from the postal code; province from the first letter of the FSA (Canada Post
	203	+ convention; the scraper's `_province` seed mislabelled 11.7% of listings and is kept as
	204	+ `prov_seed`); territories / invalid letters (27 records) dropped; FSAs with < 25
	205	+ listings pooled into `<PROV>_other`;
185	206	- filters: strictly positive price, non-missing living area / bedrooms / bathrooms,
186		− valid FSA; then the extreme 1% tails of price and living area are trimmed.
	207	+ valid FSA (147,236); then the extreme 1% tails of price and living area are trimmed
	208	+ (142,121).
187	209
188	210	## 📊 Data
189	211
		`@@ -197,89 +219,79 @@ file limit, plus its content is scraped listing data). It lives on the author's`
197	219	place it at `data/raw/realtor_mls_unique.duckdb` to run step 01. All downstream artifacts
198	220	— including the committed estimation sample — derive from it programmatically.
199	221
200		−Estimation-sample snapshot (Table 1 of the paper): median list price ≈ $639,888;
201		−median living area ≈ 135 m²; 3 bedrooms; 2 full bathrooms. Both price and price/m²
202		−are strongly right-skewed, motivating the log transformation. Ontario, Quebec, BC and
203		−Alberta dominate; PEI and the territories have no listings in this snapshot.
	222	+Estimation-sample snapshot (Table 1 of the paper): median asking price $639,900;
	223	+median living area 135 m²; 3 bedrooms; 2 full bathrooms. Both price and price/m² are
	224	+strongly right-skewed, motivating the log transformation. ON 55,897 · BC 32,994 · QC 19,949
	225	+· AB 19,463 · SK 3,458 · MB 3,062 · NS 3,044 · NB 2,980 · NL 1,274; PEI and the territories
	226	+have no listings. Prices are asking prices, not transactions — every coefficient is an
	227	+implicit list price.
204	228
205	229	## 📐 Methodology
206	230
207	231	- Specification ladder — M1 structural only → M2 + type/ownership → M3 + province FE
208	232	(house subsample) → M4 houses + FSA FE → M5 grand model (all residential, FSA FE).
209		− The M3→M5 gap ≈ 20 pp of R² is the value of resolving location at neighbourhood scale.
210		−- Absorbing least squares (`linearmodels.AbsorbingLS`) sweeps out the 1,153 FSA
211		− intercepts without materialising dummies; slopes are numerically identical to
212		− full-dummy OLS. SEs clustered by FSA throughout.
213		−- Retransformation — predictions in levels use Duan's (1983) smearing estimator (no
214		− log-normality assumption).
215		−- Validation — random 80/20 split, evaluation restricted to FSAs observed in
216		− training (out-of-support FE are not identified); leave-one-province-out CV with
217		− province-specific intercepts; Moran's I with row-standardised k-NN weights (k = 10,
218		− 15,000-listing sample, 199 permutations).
219		−
220		−## 🧭 The two results tiers (reference vs. reproduced)
221		−
222		−The original upstream cleaning code (`RE_DB_QC/hedonic`) was lost before this
223		−repository was assembled; `src/wp9/sample.py` is a careful reconstruction from the
224		−paper's own data section, calibrated against every stored output (see `docs/AUDIT.md`). It
225		−reproduces the published sample to +0.86% and all headline estimates closely — but
226		−not to the last digit. To keep the published record intact:
227		−
228		−- `results/reference/` — the original outputs (verbatim files, plus values
229		− transcribed from the published tables). Default source for figures/tables, so the
230		− paper always shows exactly the published numbers.
231		−- `results/reproduced/` — regenerated end-to-end by scripts 02–03 on the
232		− reconstructed sample.
233		−
234		−```bash
235		−python3 scripts/04_make_figures.py --results reproduced # opt in to regenerated numbers
236		−python3 scripts/05_make_tables.py --results reproduced
237		−```
	233	+- Variance decomposition — sequential increments and Shapley values over the four
	234	+ blocks (16 regressions) plus Owen values grouping province + FSA into one location block
	235	+ (Israeli 2007; Huettner & Sunder 2012). The sequential ladder credits the ~10 pp shared by
	236	+ structure and location to whichever enters first.
	237	+- Absorbing least squares (`linearmodels.AbsorbingLS`) sweeps out the 1,155 FSA
	238	+ intercepts; slopes identical to full-dummy OLS. Baseline SEs clustered by FSA; also HC1,
	239	+ province-cluster, two-way, and a restricted wild-cluster bootstrap-t at the province level
	240	+ (Rademacher, full enumeration of 2⁹ = 512 draws).
	241	+- Percentage effects — e^γ − 1 for counts; Kennedy (1981) correction for indicators.
	242	+- Functional form — quadratic in ln(area) and a linear spline with quintile knots.
	243	+- Retransformation — Duan (1983) smearing.
	244	+- Validation — (i) random 80/20 splits × 5 seeds, evaluation on FSAs seen in training;
	245	+ (ii) spatial holdout: 20% of FSAs held out entirely × 5 reps, FSA effect imputed as the
	246	+ province mean or as the inverse-distance mean of the 5 nearest training FSAs; (iii)
	247	+ leave-one-province-out with province intercepts.
	248	+- Spatial diagnostics — Moran's I, row-standardised k-NN weights (k = 5, 10, 20), two
	249	+ 15,000-listing subsamples, 199 permutations.
	250	+- Urban gradient — location residual on ln(distance to nearest of 9 metros): pooled,
	251	+ with nearest-metro intercepts, quadratic; FSA-clustered SEs.
	252	+
	253	+## 🧭 The two results tiers (reproduced vs. reference)
	254	+
	255	+- `results/reproduced/` — the paper's numbers since v1.3. Regenerated end-to-end by
	256	+ scripts 01–06 on the committed sample, with the province derived from the FSA.
	257	+- `results/reference/` — the archived outputs of the May 2026 version (v1.2), whose
	258	+ upstream cleaning code was lost (see `docs/AUDIT.md`). Kept for the reconciliation table
	259	+ in Appendix C of the paper. `python3 scripts/0{4,5}_*.py --results reference` rebuilds
	260	+ the v1.2 figures/tables (the six original tables only).
238	261
239	262	## 🖼️ Figures & tables inventory
240	263
241		−\| # \| Figure \| Shows \| Numbers from \|
242		−\|---\|---\|---\|---\|
243		−\| 1 \| `fig_price_dist` \| raw vs. log price distribution \| micro sample \|
244		−\| 2 \| `fig_province_ppm2` \| median $/m² by province \| micro sample \|
245		−\| 3 \| `fig_r2` \| R² across the M1–M5 ladder \| reference \|
246		−\| 4 \| `fig_forest` \| structural implicit prices, 95% CI (M3) \| reference \|
247		−\| 5 \| `fig_size_gradient` \| median price by size bin × dwelling type \| micro sample \|
248		−\| 6a/6b \| `fig_map` / `fig_fsa_map` \| listing map & FSA medians, coloured by ln $/m² \| micro sample \|
249		−\| 7 \| `fig_premia` \| top/bottom 12 FSA premia vs. national median \| reference \|
250		−\| 8 \| `fig_decomp` \| variance decomposition bar \| reference \|
251		−\| 9 \| `fig_nonlinear` \| marginal elasticity vs. size (quadratic model) \| reproduced band + reference anchor \|
252		−\| 10 \| `fig_gradient` \| location premium vs. distance to metro \| reproduced \|
253		−\| 11 \| `fig_quantile` \| quantile coefficients τ = 0.1…0.9 \| reference \|
254		−\| 12 \| `fig_moran` \| Moran scatterplots, structural vs. grand \| reproduced clouds, reference I \|
255		−\| 13 \| `fig_heterogeneity` \| size elasticity by province, 95% CI \| reference \|
256		−\| 14 \| `fig_oos` \| OOS error buckets (±10/±20%) \| reference \|
257		−\| 15/16 \| `fig_fit` / `fig_resid` \| predicted vs. actual; residual diagnostics \| reproduced \|
258		−
259		−(The leave-one-province-out results appear as a table only.)
260		−
261		−Tables (`results/tables/`): `summary_stats`, `regression` (M1/M2/M3/M5),
262		−`robustness` (6 sample cuts), `quantile`, `lopo`, `oos` — all six verified
263		−numerically identical to the originally published tables.
264		−
265		−## ✅ Reproduction verification
266		−
267		−Full side-by-side in [`docs/AUDIT.md`](docs/AUDIT.md) §5. Highlights (original → reproduced):
268		−
269		−- Sample: 140,931 → 142,146 (+0.86%); condos match to 3 listings.
270		−- R² ladder: 0.464/0.469/0.567/0.762/0.767 → 0.467/0.470/0.569/0.766/0.770.
271		−- OOS: R² 0.764 → 0.770; median APE 15.8% → 15.6%.
272		−- Moran's I: 0.459→0.082 vs. 0.494→0.072 (same −82~85% conclusion).
273		−- Known gaps (flagged, not hidden): M5 living elasticity 0.547 vs. 0.530; lot-field
274		− coefficients differ (the original free-text lot parser could not be fully recovered);
275		− LOPO mean 0.362 vs. 0.315 with identical province ranking.
	264	+\| # \| Figure \| Shows \|
	265	+\|---\|---\|---\|
	266	+\| 1 \| `fig_price_dist` \| raw vs. log asking price \|
	267	+\| 2 \| `fig_province_ppm2` \| median $/m² by province (FSA-derived) \|
	268	+\| 3 \| `fig_r2` \| R² across the M1–M5 ladder \|
	269	+\| 4 \| `fig_decomp` \| sequential vs. Shapley variance decomposition \|
	270	+\| 5 \| `fig_forest` \| structural implicit prices, 95% CI (M3) \|
	271	+\| 6 \| `fig_size_gradient` \| median price by size bin × dwelling type \|
	272	+\| 7a/7b \| `fig_map` / `fig_fsa_map` \| listing map & FSA medians, ln $/m² \|
	273	+\| 8 \| `fig_premia` \| top/bottom 12 FSA premia vs. median FSA \|
	274	+\| 9 \| `fig_nonlinear` \| marginal elasticity vs. size: quadratic + spline \|
	275	+\| 10 \| `fig_gradient` \| location premium by distance band to metro \|
	276	+\| 11 \| `fig_quantile` \| quantile coefficients τ = 0.1…0.9 \|
	277	+\| 12 \| `fig_moran` \| Moran scatterplots, structural vs. grand \|
	278	+\| 13 \| `fig_heterogeneity` \| size elasticity by province, 95% CI \|
	279	+\| 14 \| `fig_oos` \| OOS error buckets (±10/±20%) \|
	280	+\| 15 \| `fig_fit` / `fig_resid` \| predicted vs. actual; residual diagnostics \|
	281	+
	282	+Tables (`results/tables/`, 13): `summary_stats`, `shapley`, `regression` (M1/M2/M3/M5),
	283	+`type_effects`, `spline`, `gradient`, `robustness` (8 cuts), `se_comparison`, `quantile`,
	284	+`lopo`, `oos`, `spatial_holdout`, `moran_sensitivity` (appendix).
276	285
277	286	## ⚠️ Limitations
278	287
279		−List prices, not transactions; no construction year / renovation status / interior
280		−quality (absorbed into FSA effects and the residual); lot information sparse and noisy;
281		−a single cross-section — levels, not dynamics. See the paper's conclusion for the
282		−research agenda these imply.
	288	+Asking prices, not transactions (implicit list prices; accuracy is an upper bound for
	289	+transaction-price models); no construction year / renovation / interior quality (within-FSA
	290	+sorting remains); noisy fields (lot parsed from free text, storeys/parking with implausible
	291	+extremes and zero when unreported, province reconstructed from the postal code); a single
	292	+cross-section of one platform; FSAs are postal, not planning, units and 9 pooled residual
	293	+categories are not neighbourhoods; structural prices do not transfer to BC (LOPO R² < 0).
	294	+See §7 of the paper.
283	295
284	296	## 📚 Citation
285	297
		`@@ -288,14 +300,15 @@ research agenda these imply.`
288	300	author = {Boucher, Simon-Pierre},
289	301	title = {A Grand Hedonic Model of the Canadian Housing Market:
290	302	Decomposing the Value of Structure and Location across
291		− 140,931 MLS Listings with High-Dimensional Neighbourhood
	303	+ 142,121 MLS Listings with High-Dimensional Neighbourhood
292	304	Fixed Effects},
293	305	institution = {Universit\'e du Qu\'ebec en Outaouais,
294	306	D\'epartement des sciences administratives},
295	307	type = {Working Paper},
296	308	number = {9},
297	309	year = {2026},
298		− month = {May}
	310	+ month = {October},
	311	+ note = {Version 1.3}
299	312	}
300	313	```
301	314
		`@@ -317,3 +330,10 @@ redistributed; code and paper are shared for research reproducibility.`
317	330	- Numbering: UQO Working Paper No. 9 since the first version (May 2026).
318	331	- The original flat May 2026 project (`immo-wp9-spb-20260529`) is archived in `Recherche-UQO/_archives/versions-mai-2026/grand-hedonic-model-canada--immo-wp9-20260529`; `docs/AUDIT.md` and `docs/CHANGES.md` describe how this repository was rebuilt from it.
319	332	- Cover page, metadata and layout follow `Recherche-UQO/HARMONISATION.md` (2026-10-17).
	333	+- Revision 2026-10-17 (v1.2 → v1.3): full scientific revision by the editorial-board
	334	+ protocol (`Recherche-UQO/REVISION-MASTER-PROMPT.md`). Report: `docs/REVIEW-REPORT.md`;
	335	+ audit trail of every change (old → new numbers, reasons, verification sources):
	336	+ `docs/CHANGELOG-REVISION.md`. The paper now reports the fully reproducible pipeline
	337	+ (`results/reproduced/`) with FSA-derived provinces; the N in the subtitle changed from
	338	+ 140,931 to 142,121 accordingly. The pre-revision PDF is `docs/main_pre-revision-20261017.pdf`.
	339	+ Deviation from the standard: none (appendix added per the template).

modified data/processed/analysis.parquet +0 −0

Binary file not shown.

added docs/CHANGELOG-REVISION.md +91 −0

		`@@ -0,0 +1,91 @@`
	1	+# CHANGELOG-REVISION — Audit trail of the 2026-10-17 editorial-board revision (v1.2 → v1.3)
	2	+
	3	+All original files are preserved under git tag `pre-revision-20261017`; the v1.2 PDF is
	4	+`docs/main_pre-revision-20261017.pdf`; the v1.2 numerical outputs remain in
	5	+`results/reference/` (untouched). Every number in the v1.3 paper comes from
	6	+`results/reproduced/` produced by `scripts/01…06`.
	7	+
	8	+## 1. Decision to switch the paper to the reproduced tier
	9	+
	10	+Why. (a) The v1.2 numbers came from an upstream pipeline that no longer exists
	11	+(`docs/AUDIT.md`); the shipped code reproduced them only approximately (+0.86 % sample).
	12	+(b) The province variable used by v1.2 (`_province`, the scraper's regional seed) mislabels
	13	+16,558 of 142,121 listings (11.7 %): e.g. 13,600 listings with A/B/E/K/P FSAs (NL, NS, NB,
	14	+Ottawa-area and northern ON) labelled "QC"; 4,278 V-FSAs (BC) labelled "AB"; 1,326 T-FSAs
	15	+(AB) labelled "BC". This affected M3 (province FE), LOPO, provincial heterogeneity, the
	16	+ON/QC/BC cut, the `<PROV>_other` pooling, and the province labels in the premia figure
	17	+(v1.2 showed "E2A (QC)", "R0M (SK)"). (c) Fixing (b) requires re-running everything, which
	18	+is only possible on the reconstructed sample. Evidence: cross-tab FSA letter × seed
	19	+province, 2026-10-17 (reproduced in `scripts/01_build_sample.py` logic; see §2).
	20	+
	21	+Consequence. N changes 140,931 → 142,121 (title, abstract, README, BibTeX). All
	22	+headline magnitudes preserved (Appendix C of the paper, Table A2).
	23	+
	24	+## 2. Code changes
	25	+
	26	+\| File \| Change \|
	27	+\|---\|---\|
	28	+\| `src/wp9/config.py` \| `FSA_PROVINCE` map (Canada Post first letter → province) \|
	29	+\| `src/wp9/sample.py` \| `prov` derived from FSA letter; seed kept as `prov_seed`; FSAs with letters O/W/X/Y → dropped (27 records) \|
	30	+\| `scripts/02_estimate_core.py` \| records `n_fe` in `fit.json` \|
	31	+\| `scripts/03_estimate_extended.py` \| 2 robustness rows added (storeys/parking capped at P99; excl. pooled residual FSAs); quantile vcov made explicit (`robust`, Epanechnikov, Hall–Sheather) \|
	32	+\| `scripts/06_revision_analyses.py` \| new: Shapley/Owen decomposition; SE comparison (HC1, FSA, province, two-way) + restricted wild-cluster bootstrap-t at province level (Rademacher, 512 enumerated draws); spline elasticities; spatial holdout (20 % of FSAs × 5 reps, province-mean vs IDW-5 imputation) + 5 random seeds; Moran sensitivity (k = 5/10/20 × 2 subsamples); gradient with metro intercepts and quadratic \|
	33	+\| `scripts/04_make_figures.py` \| default tier `reproduced`; `fig_decomp` = sequential vs Shapley; `fig_nonlinear` adds spline segments \|
	34	+\| `scripts/05_make_tables.py` \| default tier `reproduced`; 7 new tables (`shapley`, `se_comparison`, `spline`, `spatial_holdout`, `moran_sensitivity`, `type_effects`, `gradient`); notes corrected (quantile SEs, regression N of FE, robustness) ; `[t]` → `[htbp]` \|
	35	+\| `paper/main.tex` \| v1.3, October 2026, subtitle N, abstract, keywords (8), `multirow`, `appendix`, float parameters, `\clearpage` before bibliography, appendix input \|
	36	+\| `paper/sections/*.tex` \| all eight sections rewritten (see §3) \|
	37	+\| `paper/appendix/appendix.tex` \| new (variable construction; Moran sensitivity; reconciliation v1.2 ↔ v1.3) \|
	38	+\| `paper/references.bib` \| 64 → 91 entries (§4) \|
	39	+\| `README.md` \| Status, TL;DR, headline table, pipeline, tiers, inventory, limitations, citation, notes \|
	40	+
	41	+## 3. Major corrections (where · before · after · why · evidence)
	42	+
	43	+\| # \| Where \| Before (v1.2) \| After (v1.3) \| Why / evidence \|
	44	+\|---\|---\|---\|---\|---\|
	45	+\| 1 \| Title, abstract, data \| N = 140,931; 1,153 FSAs \| N = 142,121; 1,155 effects = 1,146 FSAs + 9 pooled \| §1; `01_build_sample.py` log \|
	46	+\| 2 \| Data §3.2 \| province = scraper seed \| province = FSA letter; 27 territory/invalid records dropped \| §1 \|
	47	+\| 3 \| Data §3.3 text \| median "$565,000 / 125 m²" \| $639,900 / 135.3 m² \| text contradicted Table 1 even in v1.2 ($639,888 / 134.6) \|
	48	+\| 4 \| Data \| "nine provinces" with NB absent from tables \| NB 2,980 listings (was hidden under "QC") \| cross-tab \|
	49	+\| 5 \| Results §5.1 \| R² M1/M2/M3/M4/M5 = 0.464/0.469/0.567/0.762/0.767 \| 0.467/0.470/0.605/0.769/0.776 \| re-estimation; M3 rises because provinces are now correct \|
	50	+\| 6 \| Results §5.2 (new) \| "location alone ≈ 30 pp, more than all structure" (sequential only) \| sequential 43.6/1.6/14.3/18.1; Shapley 33.0/7.5/8.9/28.2; Owen structure 34.0, type 7.3, location 36.3 \| `shapley.csv`; order-dependence of sequential R² (Israeli 2007; Huettner & Sunder 2012) \|
	51	+\| 7 \| Results §5.3 \| elasticity 0.547 (0.009); bath 0.109 (+11 %) \| 0.529 (0.009); bath 0.112 (+11.9 %); M3 0.148 (+16 %) \| `coef_M5.csv`, `coef_M3.csv` \|
	52	+\| 8 \| Results §5.3 \| half-baths "proxy for older floor plans"; bedrooms "textbook ≈ 0" \| half-baths −0.034 and bedrooms −0.006 not significant under province-level inference (wild p = 0.52 / 0.58); described as negligible \| `se_comparison.csv` \|
	53	+\| 9 \| Results §5.3 \| has_lot −0.196, ln lot 0.030 read separately \| −0.459 / 0.067 read jointly (≈ −0.02 at median lot); indicator ≈ dwelling-type marker (84 % of condos, 7 % of houses lack lot) \| `coef_M5.csv`; has_lot by category \|
	54	+\| 10 \| Results §5.4 (new) \| type/ownership effects never shown \| Table 4 with Kennedy %: house +45.7 %, triplex +43.0 %, fourplex +38.7 %, duplex +15.8 %, townhouse +0.1 %; freehold +10.7 %, condo +10.2 %, leasehold −12.2 % vs co-op \| `coef_M5.csv` \|
	55	+\| 11 \| Results §5.5 \| top FSAs "all in the City of Vancouver" (+150–200 %); intro: "Vancouver and GTA" \| V6S +226 %, V6T +201 %, V8E Whistler +197 %, V6G +185 %, V7T West Vancouver +179 %; first ON (M5R) 14th at +142 %; ratio ×10.5 extremes, ×6.7 P99/P1 \| `fsa_premia.csv` \|
	56	+\| 12 \| Results §5.5 \| bottom FSAs "rural SK, MB, NL"; labels E2A (QC), R0M (SK) \| A0B −69 %, A0G −64 %, S0N/S0C/S0H/S0A ≈ −62 %, R2W Winnipeg −62 %, E3Y/E2A (NB), V0C \| `fsa_premia.csv` \|
	57	+\| 13 \| Results §5.6 \| quadratic b1 1.06, b2 −0.053; elasticity 0.65 → 0.45 \| b1 1.147, b2 −0.063; 0.63 at 60 m² → 0.41 at 350 m²; spline 0.57/0.65/0.58/0.41/0.48; ΔR² 0.001 \| `ext2.json`, `spline.csv` \|
	58	+\| 14 \| Results §5.7 \| "a doubling of distance → 8.5 % lower premium" \| slope −0.082 per log unit ⇒ −5.5 % per doubling; metro-FE slope −0.079; quadratic steepening; bands +21 % (<5 km) → −38 % (>320 km); non-monotone 10–40 km \| 2^λ − 1; `gradient_metro.json`, `gradient_bins.csv` \|
	59	+\| 15 \| Robustness §6.1 \| 6 cuts; elasticity 0.51–0.62 \| 8 cuts; 0.49 (houses) – 0.62 (condos); capped counts 0.513; excl. pooled 0.522 \| `robustness.csv` \|
	60	+\| 16 \| Robustness §6.2 (new) \| FSA-cluster SEs only \| HC1 0.0037 / FSA 0.0094 / province 0.037 / two-way 0.0375 for ln area; wild p 0.004 for area, bath, lot \| `se_comparison.csv` \|
	61	+\| 17 \| Robustness §6.3 \| "elasticity rises 0.56 → 0.60; bath stable 0.10–0.13" (table showed 0.18 → 0.07) \| elasticity flat 0.57–0.59; bath declines 0.19 → 0.10; lot rises 0.006 → 0.053 \| `quantile.csv` \|
	62	+\| 18 \| Robustness §6.4 \| Moran 0.46 → 0.08 (−82 %), z 145/23, "p < 0.01" \| 0.50 → 0.06 (−87 %), z 141/16, p = 0.005 (minimum with 199 perms); k = 5: 0.52 → 0.11; k = 20: 0.47 → 0.03 \| `ext2.json`, `moran_sensitivity.csv` \|
	63	+\| 19 \| Robustness §6.5 \| heterogeneity BC 0.49 … MB 0.66; NL 0.44 \| NL 0.41, ON 0.47, BC 0.49, NS 0.50, QC 0.57, AB 0.59, NB 0.61, MB 0.66, SK 0.73 \| `heterogeneity.csv` \|
	64	+\| 20 \| Robustness §6.6 \| LOPO mean 0.36; "above 0.40 for large central and western markets"; AB −0.10 \| mean 0.34; MB/AB/NB/ON/SK/NS 0.46–0.52; NL 0.40; QC 0.19; BC −0.49 \| `lopo.csv` \|
	65	+\| 21 \| Robustness §6.7 \| OOS R² 0.764, MdAPE 15.8 %, 59.5 % within ±20 % \| 0.772, 15.6 %, 60.1 %; 5 seeds 0.769 ± 0.010, 15.5 ± 0.2 %; FSA holdout: province-mean R² 0.52 / MdAPE 25 %; IDW-5 R² 0.74 / 17.3 % \| `oos.json`, `spatial_holdout.json` \|
	66	+\| 22 \| Abstract/intro/conclusion \| "elasticity 0.51–0.62", "11–15 % per bathroom", "×9" \| 0.53 (0.49–0.62), ≈12 %, ×10 \| consistency with tables \|
	67	+\| 23 \| Throughout \| "list price" / "price" interchangeably \| "asking price" / "implicit list price" consistently; limitation (i) expanded with mechanisms \| master prompt priority \|
	68	+
	69	+## 4. References added (29) — verification source
	70	+
	71	+Crossref REST API `works?query.bibliographic=` on 2026-10-17 unless noted; DOI recorded in `.bib`.
	72	+
	73	+allen2009canadian 10.1111/j.1540-5982.2009.01541.x · sheppard1999hedonic 10.1016/S1574-0080(99)80010-8 · bishop2020best 10.1093/reep/reaa001 · bishop2019estimating 10.1016/j.jue.2018.11.006 · sirmans2006value 10.1007/s11146-006-9983-5 · kennedy1981estimation (AER 71(4):801; RePEc `aea:aecrev:v:71:y:1981:i:4:p:801`) · abadie2023when 10.1093/qje/qjac038 · mackinnon2017wild 10.1002/jae.2508 · israeli2007shapley 10.1007/s10888-006-9036-6 · gromping2007estimators 10.1198/000313007X188252 · huettner2012axiomatic 10.1214/12-EJS710 · kelejian1998generalized 10.1023/A:1007707430416 · kelejian2001asymptotic 10.1016/S0304-4076(01)00064-1 · hill2018geospatial 10.1111/roiw.12303 · bogin2019local 10.1111/1540-6229.12233 · roberts2017cross 10.1111/ecog.02881 · combes2019costs 10.1093/restud/rdy063 · davis2008price 10.1016/j.jue.2007.02.003 · albouy2018metropolitan 10.1162/rest_a_00710 · diewert2015hedonic 10.1080/07474938.2014.944791 · horowitz1992role 10.1002/jae.3950070202 · merlo2004bargaining 10.1016/j.jue.2004.05.004 (DOI record fetched) · anglin2003tradeoff 10.1023/A:1021526332732 · han2016what 10.1016/j.jue.2016.03.008 · genesove2012search 10.1016/j.jue.2012.01.002 · hossain2009determinants 10.1080/00036840701522861 · dube2013commuter 10.1016/j.tra.2013.07.015 · theriault2003modelling 10.1108/02637470310464472 · statcan2022fsa (Statistics Canada catalogue 92-179-G2021001, release 2022-09-21, catalogue page fetched).
	74	+
	75	+Removed: cmhc2018canadian (unverifiable), gabaix2016power (uncited).
	76	+Corrected: sirmans2005composition pages 3–43 → 1–44 (+DOI); Court 1939, Griliches 1961, Combes & Gobillon 2015 → `@incollection`; Correia 2017 → `@unpublished`; DOIs added to rosen1974, lancaster1966, malpezzi2003, can1992, bourassa2007, case2004, kuminoff2010, glaeser2005, duan1983, cameron2015, mullainathan2017, ekeland2004, bajari2005, black1999, chay2005, linden2008, gibbons2012, moran1950, koenker1978, zietz2008, mcmillen2010, pace1998.
	77	+
	78	+## 5. Re-analyses executed
	79	+
	80	+See `REVIEW-REPORT.md` §5. Total wall-clock ≈ 2.5 min (Apple M5 Max). Seeds: `SEED_OOS=12345`, `SEED_MORAN=3`, `SEED_EXT2=7`, `SEED_REV=2026`; random-split repeats use seeds 12345, 1, 2, 3, 4.
	81	+
	82	+## 6. Items for the author's attention
	83	+
	84	+1. Title N on SSRN (7626678): v1.2 says 140,931; v1.3 says 142,121 — upload v1.3 and update the title there.
	85	+2. The repository's `.gitignore` excludes `data/processed/*.parquet` but the file is tracked (6 MB) — intentional for reproducibility; left as is.
	86	+3. The scraper seed is retained as `prov_seed` in the parquet for audit only; it is not used anywhere.
	87	+4. Historical docs (`AUDIT.md`, `CHANGES.md`, `PAPER_REVIEW.md`, `UPGRADE_REPORT.md`) describe the August 2026 rebuild and v1.2 upgrade; they are kept unchanged as history and some of their statements (e.g. "paper uses results/reference") no longer hold.
	88	+
	89	+## 7. Not modified for lack of data
	90	+
	91	+Transaction prices, time on market, age/renovation/quality fields, finer-than-FSA geography, amenity data for premium decomposition, additional snapshots for dynamics.

added docs/REVIEW-REPORT.md +103 −0

		`@@ -0,0 +1,103 @@`
	1	+# REVIEW-REPORT — Editorial-board revision of UQO WP 9 (2026-10-17)
	2	+
	3	+Paper: *A Grand Hedonic Model of the Canadian Housing Market: Decomposing the Value of
	4	+Structure and Location across 142,121 MLS Listings with High-Dimensional Neighbourhood
	5	+Fixed Effects* — v1.2 (May 2026, 34 pp., 64 refs) → v1.3 (October 2026, 52 pp., 91 refs).
	6	+Protocol: `Recherche-UQO/REVISION-MASTER-PROMPT.md` + `REVISION-PROTOCOLE.md`. Pre-revision
	7	+state: git tag `pre-revision-20261017`, `docs/main_pre-revision-20261017.pdf`.
	8	+Companion audit trail: `docs/CHANGELOG-REVISION.md`.
	9	+
	10	+## 0. Summary verdict
	11	+
	12	+A well-executed national measurement paper whose v1.2 had three structural problems that a
	13	+referee would have caught: (1) the headline "location ≈ 30 pp, more than all structure"
	14	+rested on a sequential R² ladder whose attribution is order-dependent; (2) ~12 % of listings
	15	+carried the wrong province (the scraper seed, not the postal code), visible on the published
	16	+premia figure ("E2A (QC)", "R0M (SK)"); (3) the paper's numbers came from a lost pipeline
	17	+that the shipped code reproduced only approximately. All three are fixed in v1.3: the paper
	18	+now reports the fully reproducible pipeline with FSA-derived provinces, complements the
	19	+ladder with a Shapley–Owen decomposition, and adds inference, functional-form and
	20	+spatial-validation analyses. The contribution survives in a more defensible form:
	21	+**structure and location are of comparable importance (≈34 vs ≈36 pp of R²), and location
	22	+is the largest block only once province and FSA are grouped.**
	23	+
	24	+## 1. Diagnosis by section (v1.2 → action in v1.3)
	25	+
	26	+\| Section \| Strengths \| Weaknesses / errors found \| Action \|
	27	+\|---\|---\|---\|---\|
	28	+\| Abstract \| Clear, quantified \| "location alone ≈ 30 pp … dominant" unqualified; elasticity range 0.51–0.62 mixes samples; list-price nature not stated \| Rewritten with attribution rule, inference caveat, within/out-of-support accuracy, explicit asking-price statement \|
	29	+\| Introduction \| Good motivation, 3 contributions \| Novelty claim unsupported by Canadian lit; "GTA among top premia" false (all top-12 are BC); no statement of list-price caveat \| Rewritten; Canadian lit cited (Allen et al. 2009; Hossain & Latif 2009; Dubé et al. 2013; Thériault et al. 2003); design caveats stated upfront \|
	30	+\| Literature \| 7 thematic subsections \| Missing: Sheppard 1999, Sirmans et al. 2006 meta-analysis, Bishop et al. 2020, Bishop–Timmins 2019, Kennedy 1981, Kelejian–Prucha 1998/2001, Abadie et al. 2023, MacKinnon–Webb 2017, Shapley/Owen refs, Davis–Palumbo 2008, Albouy–Ehrlich–Shin 2018, Combes–Duranton–Gobillon 2019, Diewert et al. 2015, Hill–Scholz 2018, Bogin et al. 2019, list-price literature (Horowitz 1992, Merlo & Ortalo-Magné 2004, Anglin et al. 2003, Han–Strange 2016, Genesove–Han 2012), spatial CV (Roberts et al. 2017); one unverifiable reference (Mok & Chan 2018) \| 27 verified refs added, 2 removed, 23 DOIs added; new subsection "List prices and the Canadian evidence" \|
	31	+\| Data \| Transparent parsing \| Median price "$565,000 / 125 m²" in text vs $639,888 / 134.6 in table; province = scraper seed (11.7 % mislabelled); NB present but reported as absent; territories listed as absent though 22 records existed; storeys/parking zero-fill and extreme values undocumented; "1,153 FSAs" counted 9 pooled province categories as FSAs \| Province rebuilt from FSA letter; territories/invalid letters dropped; all counts recomputed and reported (147,236 → 142,121; 1,146 FSAs + 9 pools); StatCan FSA definition cited \|
	32	+\| Methodology \| Correct equation, HDFE well explained \| No equation for decomposition, inference, quantile, Moran, gradient, Duan; "ladder" as bullet list; Halvorsen–Palmquist cited but Kennedy correction absent; clustering level not discussed \| Six numbered equations added (Kennedy, Shapley, quantile, Moran, Duan, gradient reading); inference subsection (FSA vs province, wild-cluster bootstrap); validation designs defined \|
	33	+\| Results \| Clear figures \| Order-dependent decomposition presented as fact; "doubling of distance → −8.5 %" wrong (slope −0.085 per log unit = −5.7 % per doubling); half-bath/bedroom effects over-interpreted; has_lot coefficient read in isolation; type/ownership effects never reported though invoked; gradient non-monotonicity unmentioned; Whistler (V8E) called "City of Vancouver" \| Shapley/Owen table + figure; gradient table (bands, metro FE, quadratic) and correct per-doubling reading; type/ownership table with Kennedy %; joint reading of lot variables; spline table; geography paragraph corrected \|
	34	+\| Robustness \| Six cuts, quantiles, Moran, LOPO, OOS \| Quantile text claimed bathroom premium "stable 0.10–0.13" while table showed 0.18 → 0.07; LOPO text claimed ">0.40 for large central and western markets" while BC 0.32, QC 0.30, AB −0.10; Moran p "<0.01" with 199 permutations (min 0.005); OOS random split only; "analytical SEs" for quantiles undefined \| Text rewritten to match recomputed tables; SE-comparison table (HC1/FSA/province/two-way + wild bootstrap); spatial holdout table (FSA-level, province-mean vs IDW); Moran sensitivity (k = 5/10/20, two subsamples); two robustness rows added (capped counts; excl. pooled FSAs) \|
	35	+\| Discussion \| Good engagement with literature \| Repeated v1.2 overstatements; "upper end of published estimates" for bathrooms unsupported \| Rewritten; new "Inference" paragraph; six limitations incl. list-to-sale wedge mechanics; methodological message on sequential ladders \|
	36	+\| Conclusion \| Modest tone \| Numbers inconsistent with new results \| Rewritten (four conclusions) \|
	37	+\| Appendix \| none \| — \| Added: variable construction, Moran sensitivity, v1.2 ↔ v1.3 reconciliation table \|
	38	+
	39	+## 2. Methodology — verdict on each point of §3 of the master prompt
	40	+
	41	+- Specification justified: yes; semi-log with log area; constant elasticity tested by quadratic and spline (ΔR² ≤ 0.001; curvature concentrated above 150 m²).
	42	+- Estimation assumptions explicit: now yes (conditional mean zero given attributes and FSA; within-FSA sorting of unobserved quality acknowledged as the residual threat).
	43	+- Variables correctly defined: yes after revision (zero-fill of storeys/parking, lot indicator + ln(1+lot), province from FSA all documented, Appendix A).
	44	+- Fixed effects appropriate: yes. One-way FSA FE, exact demeaning. Note: 9 "pooled residual" categories are not neighbourhoods; dropping them changes nothing (Table 7).
	45	+- Spatial dependence defensible: FE remove 87 % of Moran's I; residual I = 0.03–0.11 significant → clustered inference required; result robust to k and subsample.
	46	+- Standard errors: baseline FSA-cluster (G = 1,155) appropriate; province-cluster SEs ×4 larger; restricted wild-cluster bootstrap at province level (G = 9, 512 enumerated draws) → area, bathrooms, lot robust (p = 0.004 = minimum); half-baths and bedrooms not (p = 0.52, 0.58). Text revised accordingly.
	47	+- Sample construction transparent: yes (raw 172,019 → core 147,236 → trimmed 142,121; Appendix A).
	48	+- Robustness addresses genuine threats: 8 cuts incl. capped counts and no pooled FSAs; quantiles; functional form; alternative SEs; spatial holdout; LOPO.
	49	+- Coefficient interpretation: e^γ−1 for counts (11.9 % per bathroom), Kennedy (1981) for indicators (house +45.7 % vs apartment), gradient per doubling 2^λ−1 (−5.5 %). The v1.2 "−8.5 % per doubling" was a mathematical error.
	50	+- Statistical vs economic significance: distinguished (bedrooms −0.6 %: statistically significant under FSA clustering only, economically nil).
	51	+- Causal claims: none; FE described as measurement device; premia as composite capitalisation.
	52	+
	53	+## 3. Literature
	54	+
	55	+- Existing references verified: all 62 retained entries checked against Crossref (query + DOI) or publisher/RePEc records; DOIs added to 23 entries that lacked one; Sirmans et al. (2005) pages corrected to 1–44 (Crossref); Court (1939), Griliches (1961), Combes & Gobillon (2015) retyped as `@incollection`; Correia (2017) as `@unpublished`.
	56	+- Removed (2): `cmhc2018canadian` (Mok & Chan, "House Prices and the Geography of the Canadian Housing Market", CJUR 2018 — no Crossref/Scholar record; treated as unverifiable) → replaced by Allen, Amano, Byrne & Gregory (2009, CJE); `gabaix2016power` (orphan, never cited).
	57	+- Added (29, all Crossref-verified with DOI except Kennedy 1981 verified via RePEc/AER and the StatCan catalogue record): Sheppard 1999; Bishop et al. 2020; Bishop & Timmins 2019; Sirmans et al. 2006; Kennedy 1981; Abadie et al. 2023; MacKinnon & Webb 2017; Israeli 2007; Grömping 2007; Huettner & Sunder 2012; Kelejian & Prucha 1998, 2001; Hill & Scholz 2018; Bogin, Doerner & Larson 2019; Roberts et al. 2017; Combes, Duranton & Gobillon 2019; Davis & Palumbo 2008; Albouy, Ehrlich & Shin 2018; Diewert, de Haan & Hendriks 2015; Horowitz 1992; Merlo & Ortalo-Magné 2004; Anglin, Rutherford & Springer 2003; Han & Strange 2016; Genesove & Han 2012; Allen et al. 2009; Hossain & Latif 2009; Dubé, Thériault & Des Rosiers 2013; Thériault et al. 2003; Statistics Canada 2022 (FSA reference guide). Full list with DOIs in `CHANGELOG-REVISION.md` §4.
	58	+- Count: 64 → 91 entries, 91/91 cited, 0 unverified.
	59	+
	60	+## 4. Results — inconsistencies found and resolved
	61	+
	62	+1. Text vs table medians (price, area) — fixed by regeneration from the sample.
	63	+2. Quantile bathroom premium "stable" vs declining 0.18 → 0.07 — text corrected (now 0.19 → 0.10).
	64	+3. LOPO narrative vs table — corrected (new ranking: BC −0.49, QC 0.19).
	65	+4. Gradient per doubling −8.5 % → −5.5 % (correct transformation).
	66	+5. "All top FSAs in the City of Vancouver" — V8E is Whistler, V7T West Vancouver; corrected.
	67	+6. "GTA among the highest" — first ON FSA ranks 14th; corrected.
	68	+7. Premia labelled with wrong provinces (E2A "QC", R0M "SK") — fixed at source.
	69	+8. Moran p "<0.01" — reported as 0.005 (minimum attainable).
	70	+9. "1,153 FSAs" — 1,146 FSAs + 9 pooled categories; stated.
	71	+10. Storeys coefficient 0.001*** driven by a 593-storey record — explained; capped-row robustness.
	72	+11. Mean APE in repeated splits dominated by one listing with 3,620 parking spaces — trimmed mean reported with note.
	73	+
	74	+## 5. Re-analyses executed (all on this Mac, total < 3 min; scripts versioned)
	75	+
	76	+\| Analysis \| Script \| Output \| Runtime \|
	77	+\|---\|---\|---\|---\|
	78	+\| Sample rebuild with FSA-derived province \| `01_build_sample.py` \| `data/processed/analysis.parquet` \| 40 s \|
	79	+\| Ladder M1–M5, premia \| `02_estimate_core.py` \| `results/reproduced/{fit.json,coef_M*.csv,fsa_premia.csv,grand_model.parquet}` \| 5 s \|
	80	+\| OOS, 8 robustness cuts, heterogeneity, quantiles, quadratic, gradient, Moran, LOPO \| `03_estimate_extended.py` \| `results/reproduced/*` \| 8 s \|
	81	+\| Shapley/Owen (16 regressions ×2 samples), SE comparison + wild-cluster bootstrap (512 draws × 5 coefs), spline, spatial holdout (5 FSA reps + 5 random seeds), Moran sensitivity (12 configs), metro gradient \| `06_revision_analyses.py` (new) \| `shapley.csv`, `se_comparison.csv`, `spline.csv`, `spatial_holdout.json`, `moran_sensitivity.csv`, `gradient_metro.json` \| 33 s \|
	82	+\| 13 tables, 17 figures \| `05_make_tables.py`, `04_make_figures.py` \| `results/tables/`, `figures/` \| 35 s \|
	83	+
	84	+## 6. Recommended analyses NOT executed (data unavailable or out of scope)
	85	+
	86	+- Linking listings to transaction prices / time on market (no sales data in the snapshot).
	87	+- Age, renovation and interior-quality controls (fields absent).
	88	+- Explicit spatial error model (Kelejian–Prucha GS2SLS) on the within-FSA residuals — feasible but adds complexity without changing conclusions; left for future work.
	89	+- Fully interacted submarket-specific coefficient systems (province × attribute slopes) — partially covered by province-by-province estimation.
	90	+- Finer geography (6-character postal codes or dissemination areas) — requires a boundary file / geocoding of ~140k points to census units; not attempted.
	91	+- Panel of snapshots for dynamics — single cross-section only.
	92	+- Decomposition of FSA premia into amenities (schools, transit) — requires external amenity data.
	93	+- ML benchmark (XGBoost/LightGBM) against the hedonic model — dead code in the original project was dropped in Aug 2026; not revived (would need `xgboost`, `lightgbm`).
	94	+
	95	+## 7. Remaining limitations (stated in §7.4 of the paper)
	96	+
	97	+Asking prices; unobserved quality and within-FSA sorting; noisy lot/storey/parking fields and reconstructed province; single cross-section of one platform; postal (not planning) geography and 9 pooled pseudo-neighbourhoods; structural prices do not transfer to BC.
	98	+
	99	+## 8. Publication maturity
	100	+
	101	+- Target journals: Journal of Housing Economics, Journal of Real Estate Finance and Economics, Regional Science and Urban Economics (empirical note), Canadian Journal of Economics (Canadian scope), Journal of Property Research.
	102	+- Readiness: suitable for submission as a measurement/benchmark paper after the author checks the items flagged in `CHANGELOG-REVISION.md` §6 (title N change on SSRN; wording of the novelty claim).
	103	+- Main rejection risks: (i) asking prices rather than transactions — the central vulnerability; now framed honestly but a referee at a top field journal may still ask for a list-to-sale validation on a subsample; (ii) descriptive rather than causal contribution — framed as measurement; (iii) FSA geography as "neighbourhood"; (iv) single snapshot, unknown platform coverage; (v) the two lot coefficients are hard to interpret and depend on a free-text parser.

modified figures/fig_decomp.png +0 −0

Binary file not shown.

modified figures/fig_fit.png +0 −0

Binary file not shown.

modified figures/fig_forest.png +0 −0

Binary file not shown.

modified figures/fig_fsa_map.png +0 −0

Binary file not shown.

modified figures/fig_gradient.png +0 −0

Binary file not shown.

modified figures/fig_heterogeneity.png +0 −0

Binary file not shown.

modified figures/fig_map.png +0 −0

Binary file not shown.

modified figures/fig_moran.png +0 −0

Binary file not shown.

modified figures/fig_nonlinear.png +0 −0

Binary file not shown.

modified figures/fig_oos.png +0 −0

Binary file not shown.

modified figures/fig_premia.png +0 −0

Binary file not shown.

modified figures/fig_price_dist.png +0 −0

Binary file not shown.

modified figures/fig_province_ppm2.png +0 −0

Binary file not shown.

modified figures/fig_quantile.png +0 −0

Binary file not shown.

modified figures/fig_r2.png +0 −0

Binary file not shown.

modified figures/fig_resid.png +0 −0

Binary file not shown.

modified figures/fig_size_gradient.png +0 −0

Binary file not shown.

added paper/appendix/appendix.tex +125 −0

		`@@ -0,0 +1,125 @@`
	1	+% Author: Simon-Pierre Boucher — contact@spboucher.ai
	2	+% ============================================================================
	3	+% Appendix
	4	+% ============================================================================
	5	+\newpage
	6	+\begin{appendices}
	7	+
	8	+\renewcommand{\thetable}{A\arabic{table}}
	9	+\renewcommand{\thefigure}{A\arabic{figure}}
	10	+\setcounter{table}{0}
	11	+\setcounter{figure}{0}
	12	+
	13	+\section{Variable construction}
	14	+\label{app:variables}
	15	+
	16	+The raw \texttt{listings} table stores every attribute as text. The estimation sample is
	17	+built by the first script of the pipeline (\texttt{01\_build\_sample.py}) from the rules
	18	+below, which were reconstructed from the description of the original (May 2026) pipeline and validated
	19	+against its stored outputs; the original cleaning code was not preserved.
	20	+
	21	+\paragraph{Bedrooms.} The integer components of the raw string are summed, so that
	22	+``3~+~1'' (three bedrooms on the main level and one below grade) is coded as four.
	23	+
	24	+\paragraph{Living area.} The explicit floor-area measurement is preferred. Exact entries
	25	+are used as reported; banded entries such as ``1100--1500 sqft'' are mapped to the upper
	26	+bound of the band, the convention that reproduces the published living-area distribution.
	27	+When the field is absent the interior-size field is used. Square feet are converted at
	28	+$1\text{ ft}^2 = 0.0929\text{ m}^2$; entries labelled m$^2$ are used as such.
	29	+
	30	+\paragraph{Lot area.} The free-text lot field is parsed when it contains a number with a
	31	+recognisable unit (square feet, square metres, acres or hectares), a phrase of the form
	32	+``under $x$ acre'', an acreage range (mapped to its upper bound), or a frontage-by-depth
	33	+pair in feet. Parsed lots above the 99th percentile of positive values are treated as
	34	+unparsed, because they are dominated by multi-acre acreage strings whose relationship to
	35	+the dwelling is unclear. The variable entering the regressions is $\ln(1+\text{lot area})$
	36	+together with an indicator equal to one when a positive lot area was parsed.
	37	+
	38	+\paragraph{Counts.} Full bathrooms, half bathrooms, parking spaces and storeys are taken as
	39	+reported; half bathrooms, parking and storeys are set to zero when absent. No cap is
	40	+applied in the baseline; Table~\ref{tab:robustness} caps storeys and parking at their 99th
	41	+percentiles.
	42	+
	43	+\paragraph{Dwelling type and ownership.} The raw building type is consolidated into eight
	44	+groups (house; apartment; row or townhouse; duplex; triplex; fourplex; manufactured home,
	45	+which pools manufactured, mobile and park-model homes; other), and the ownership type into
	46	+six (freehold; condominium or strata; leasehold; co-operative or co-ownership; other; not
	47	+reported). The broad listing category (house, condominium, plex, townhouse, apartment) is
	48	+also retained as a control.
	49	+
	50	+\paragraph{Neighbourhood and province.} The FSA is the first three characters of the postal
	51	+code, retained when it matches the pattern letter--digit--letter. The province is the
	52	+Canada Post region of the first letter: A, Newfoundland and Labrador; B, Nova Scotia; C,
	53	+Prince Edward Island; E, New Brunswick; G, H and J, Quebec; K, L, M, N and P, Ontario; R,
	54	+Manitoba; S, Saskatchewan; T, Alberta; V, British Columbia \citep{statcan2022fsa}. The
	55	+letters X and Y (territories) and the unassigned letters O and W (27 records in total) are
	56	+dropped. The scraper's own province field, which records the regional seed of the
	57	+extraction, is retained for reference only; it disagrees with the FSA-based province for
	58	+11.7\% of the estimation sample.
	59	+
	60	+\paragraph{Filters.} Listings are kept when the asking price is strictly positive, living
	61	+area, bedrooms and bathrooms are non-missing and the FSA is valid (147{,}236 of 172{,}019
	62	+raw records); the 1st and 99th percentiles of price and living area are then applied as
	63	+inclusive bounds, leaving 142{,}121 listings. FSAs with fewer than 25 listings are pooled
	64	+into one residual category per province.
	65	+
	66	+\section{Additional results}
	67	+\label{app:tables}
	68	+
	69	+\input{../results/tables/moran_sensitivity}
	70	+
	71	+\section{Reconciliation with the May 2026 version}
	72	+\label{app:reconciliation}
	73	+
	74	+The May 2026 version of this paper (v1.2) reported estimates from an upstream pipeline that
	75	+was not preserved. The present version re-estimates every result from the reconstructed
	76	+pipeline described in Appendix~\ref{app:variables}, with two substantive differences: the
	77	+province of each listing is derived from its FSA rather than from the scraper's regional
	78	+seed, and the reconstructed trimming rule retains 142{,}121 rather than 140{,}931 listings
	79	+(a difference of 0.8\%, concentrated at the upper living-area bound). All headline
	80	+magnitudes are preserved (Table~\ref{tab:reconciliation}); the estimates that changed
	81	+materially are those that use the province variable---the province-effects $R^2$, the
	82	+leave-one-province-out ranking and the provincial heterogeneity---and the two lot
	83	+coefficients, whose free-text parser could not be recovered exactly. The variance
	84	+decomposition, the inference comparison, the spline, the spatial holdout and the Moran
	85	+sensitivity analysis are new to this version.
	86	+
	87	+\begin{table}[htbp]\centering
	88	+\caption{Headline estimates: May 2026 version (v1.2) and present version (v1.3)}
	89	+\label{tab:reconciliation}
	90	+\begin{threeparttable}
	91	+\begin{tabular}{lcc}
	92	+\toprule
	93	+Quantity & v1.2 (May 2026) & v1.3 (this version) \\
	94	+\midrule
	95	+Estimation sample & 140{,}931 & 142{,}121 \\
	96	+Absorbed location effects & 1{,}153 & 1{,}155 \\
	97	+$R^2$: M1 / M2 / M3 (houses) & 0.464 / 0.469 / 0.567 & 0.467 / 0.470 / 0.605 \\
	98	+$R^2$: M4 (houses, FSA) / M5 (grand) & 0.762 / 0.767 & 0.769 / 0.776 \\
	99	+Living-area elasticity, M5 (SE) & 0.547 (0.009) & 0.529 (0.009) \\
	100	+Full-bathroom coefficient, M5 (SE) & 0.109 (0.004) & 0.112 (0.004) \\
	101	+Bedrooms, M5 & $-0.002$ & $-0.006$ \\
	102	+$\ln$ lot / lot indicator, M5 & 0.030 / $-0.196$ & 0.067 / $-0.459$ \\
	103	+Out-of-sample $R^2$ / MdAPE & 0.764 / 15.8\% & 0.772 / 15.6\% \\
	104	+Moran's $I$, structure $\to$ grand ($k=10$) & 0.46 $\to$ 0.08 & 0.50 $\to$ 0.06 \\
	105	+Gradient slope on $\ln$ distance & $-0.085$ & $-0.082$ \\
	106	+Quadratic in $\ln$ area: $\beta_1$ / $\beta_2$ & 1.065 / $-0.053$ & 1.147 / $-0.063$ \\
	107	+LOPO mean within-province $R^2$ & 0.362 (AB negative) & 0.341 (BC negative) \\
	108	+Size elasticity range across provinces & 0.44 (NL)--0.66 (MB) & 0.41 (NL)--0.73 (SK) \\
	109	+Premium ratio, most / least expensive FSA & $\approx 9$ & $\approx 10$ \\
	110	+\bottomrule
	111	+\end{tabular}
	112	+\begin{tablenotes}\footnotesize
	113	+\item Notes: v1.2 values are the published numbers of the May 2026 version, archived in
	114	+\texttt{results/reference/}; v1.3 values are produced by the current pipeline
	115	+(\texttt{results/reproduced/}). The v1.2 province variable was the scraper's regional seed,
	116	+which mislabelled about 12\% of listings; the v1.2 statement that the lowest-premium FSAs
	117	+included ``E2A (QC)'' and ``R0M (SK)'' reflected this error (E is New Brunswick and R is
	118	+Manitoba). In v1.2 a doubling of distance was described as lowering the location premium by
	119	+8.5\%; the correct reading of a slope of $-0.085$ on $\ln$ distance is $-5.7\%$ per
	120	+doubling.
	121	+\end{tablenotes}
	122	+\end{threeparttable}
	123	+\end{table}
	124	+
	125	+\end{appendices}

modified paper/main.pdf +0 −0

Binary file not shown.

modified paper/main.tex +42 −23

		`@@ -4,6 +4,7 @@`
4	4	% A Grand Hedonic Model of the Canadian Housing Market
5	5	%
6	6	% Build: latexmk -pdf main.tex (or `make` in this directory)
	7	+% Revised by the editorial-board review of 2026-10-17 (v1.3): see docs/REVIEW-REPORT.md.
7	8	% Figures are read from ../figures/, tables from ../results/tables/.
8	9	% ============================================================================
9	10	\documentclass[12pt,letterpaper]{article}
		`@@ -29,11 +30,18 @@`
29	30	\usepackage{booktabs}
30	31	\usepackage{threeparttable}
31	32	\usepackage{makecell}
	33	+\usepackage{multirow}
32	34
33	35	% ---------------------------------------------------------------- figures
34	36	\usepackage{graphicx}
35	37	\usepackage{subcaption}
36	38	\graphicspath{{../figures/}{./}}
	39	+\renewcommand{\topfraction}{0.9}
	40	+\renewcommand{\bottomfraction}{0.6}
	41	+\renewcommand{\textfraction}{0.07}
	42	+\renewcommand{\floatpagefraction}{0.75}
	43	+\setcounter{topnumber}{3}
	44	+\setcounter{totalnumber}{4}
37	45
38	46	% ---------------------------------------------------------------- captions
39	47	\usepackage[font=small,labelfont=bf,labelsep=period,justification=justified,singlelinecheck=false]{caption}
		`@@ -58,16 +66,19 @@`
58	66	\usepackage{enumitem}
59	67	\setlist{nosep,leftmargin=*}
60	68
	69	+% ---------------------------------------------------------------- appendix
	70	+\usepackage[toc,page]{appendix}
	71	+
61	72	% ============================================================================
62	73	% METADATA
63	74	% ============================================================================
64	75	\newcommand{\WPnumber}{9}
65	76	\newcommand{\WPtitle}{A Grand Hedonic Model of the Canadian Housing Market}
66	77	\newcommand{\WPsubtitle}{Decomposing the Value of Structure and Location across
67		− 140{,}931 MLS Listings with High-Dimensional Neighbourhood Fixed Effects}
68		−\newcommand{\WPdate}{May 2026}
69		−\newcommand{\WPversion}{1.2}
70		−\newcommand{\WPkeywords}{hedonic pricing, housing markets, Canada, fixed effects, spatial heterogeneity, automated valuation}
	78	+ 142{,}121 MLS Listings with High-Dimensional Neighbourhood Fixed Effects}
	79	+\newcommand{\WPdate}{October 2026}
	80	+\newcommand{\WPversion}{1.3}
	81	+\newcommand{\WPkeywords}{hedonic pricing, housing markets, Canada, high-dimensional fixed effects, neighbourhood effects, spatial autocorrelation, automated valuation model, list prices}
71	82	\newcommand{\WPjel}{R31, R21, C21, C55}
72	83
73	84	\newcommand{\WPauthor}{Simon-Pierre Boucher}
		`@@ -78,24 +89,30 @@`
78	89	283, boulevard Alexandre-Tach\'e\\ Gatineau, Qu\'ebec, Canada J9A 1L8}
79	90
80	91	\newcommand{\WPabstract}{%
81		−We estimate a large-scale hedonic price model for the Canadian residential real-estate
82		−market using a cross-section of 140{,}931 active MLS listings drawn from the nine provinces
83		−present in the data. A semi-logarithmic specification decomposes dwelling prices into
84		−structural attributes, dwelling type and ownership form, and 1{,}153 neighbourhood
85		−(Forward Sortation Area) fixed effects estimated by absorbing least squares. Moving from a
86		−purely structural model to the full specification raises the explained share of price
87		−variation from 46\% to 77\%, establishing that location is the dominant price determinant
88		−in Canada: neighbourhood effects alone account for roughly thirty percentage points of
89		−explanatory power. The living-area elasticity is 0.51--0.62, each full bathroom commands a
90		−premium of about 11--15\%, and---conditional on floor space---bedroom counts are
91		−economically negligible. The implicit prices are stable across alternative samples and
92		−trimming rules, and the model delivers strong out-of-sample valuation accuracy (held-out
93		−$R^2=0.76$; median absolute error of 16\%). The neighbourhood effects absorb the bulk of
94		−the spatial dependence in prices---Moran's~$I$ falls from 0.46 to 0.08---and we further
95		−document diminishing returns to floor space, an urban price gradient that decays with
96		−distance to major metros, and quantile and cross-province heterogeneity in the implicit
97		−prices. We map the resulting neighbourhood premia, which span a factor of roughly nine
98		−between the most and least expensive areas.}
	92	+We estimate a national hedonic price function for the Canadian residential market on a
	93	+cross-section of 142{,}121 active MLS listings from nine provinces. A semi-logarithmic
	94	+specification relates the log list price to structural attributes, dwelling type and
	95	+ownership form, and 1{,}155 absorbed neighbourhood (Forward Sortation Area) fixed effects
	96	+estimated by absorbing least squares with neighbourhood-clustered inference. Moving from a
	97	+structural model to the full specification raises the explained share of log-price
	98	+variation from 44\% to 78\%. Because structure and location are correlated, the attribution
	99	+of this gain depends on the order in which blocks enter: a Shapley--Owen decomposition
	100	+assigns about 34 percentage points to structure and 36 to location, so that the two are of
	101	+comparable importance and the neighbourhood is the single largest block once province and
	102	+neighbourhood are grouped. Conditional on location, the living-area elasticity is 0.53
	103	+(0.49--0.62 across samples), each full bathroom adds about 12\%, and bedroom counts are
	104	+economically negligible. The implicit prices of area and bathrooms survive province-level
	105	+wild-cluster bootstrap inference, whereas the small half-bathroom and bedroom effects do not.
	106	+Neighbourhood effects remove 87\% of the spatial autocorrelation in residuals (Moran's~$I$
	107	+0.50 to 0.06). The model values held-out listings in neighbourhoods seen in training with
	108	+an out-of-sample $R^2$ of 0.77 and a median absolute percentage error of 15.6\%; for
	109	+neighbourhoods never seen in training, spatial interpolation of the neighbourhood effect
	110	+still yields $R^2=0.74$ and a median error of 17\%. Floor space displays mild diminishing
	111	+returns, the location premium falls by about 5.5\% per doubling of distance to the nearest
	112	+metropolitan centre, and implicit prices vary across the price distribution and across
	113	+provinces. Neighbourhood premia span a factor of about ten between the most and least
	114	+expensive areas. Because the dependent variable is the asking price rather than the
	115	+transaction price, all estimates describe the pricing of listings, not of sales.}
99	116
100	117	\hypersetup{pdftitle={A Grand Hedonic Model of the Canadian Housing Market}, pdfauthor={Simon-Pierre Boucher}}
101	118
		`@@ -113,7 +130,9 @@ between the most and least expensive areas.}`
113	130	\input{sections/discussion}
114	131	\input{sections/conclusion}
115	132
116		−\newpage
	133	+\clearpage
117	134	\bibliography{references}
118	135
	136	+\input{appendix/appendix}
	137	+
119	138	\end{document}

modified paper/references.bib +382 −63

		`@@ -9,7 +9,8 @@`
9	9	volume = {82},
10	10	number = {1},
11	11	pages = {34--55},
12		− year = {1974}
	12	+ year = {1974},
	13	+ doi = {10.1086/260169}
13	14	}
14	15
15	16	@article{lancaster1966new,
		`@@ -19,23 +20,28 @@`
19	20	volume = {74},
20	21	number = {2},
21	22	pages = {132--157},
22		− year = {1966}
	23	+ year = {1966},
	24	+ doi = {10.1086/259131}
23	25	}
24	26
25		−@article{court1939hedonic,
26		− author = {Court, Andrew T.},
27		− title = {Hedonic Price Indexes with Automotive Examples},
28		− journal = {The Dynamics of Automobile Demand, General Motors},
29		− pages = {99--117},
30		− year = {1939}
	27	+@incollection{court1939hedonic,
	28	+ author = {Court, Andrew T.},
	29	+ title = {Hedonic Price Indexes with Automotive Examples},
	30	+ booktitle = {The Dynamics of Automobile Demand},
	31	+ publisher = {General Motors Corporation},
	32	+ address = {New York},
	33	+ pages = {99--117},
	34	+ year = {1939}
31	35	}
32	36
33		−@article{griliches1961hedonic,
34		− author = {Griliches, Zvi},
35		− title = {Hedonic Price Indexes for Automobiles: An Econometric Analysis of Quality Change},
36		− journal = {The Price Statistics of the Federal Government, NBER},
37		− pages = {173--196},
38		− year = {1961}
	37	+@incollection{griliches1961hedonic,
	38	+ author = {Griliches, Zvi},
	39	+ title = {Hedonic Price Indexes for Automobiles: An Econometric Analysis of Quality Change},
	40	+ booktitle = {The Price Statistics of the Federal Government},
	41	+ publisher = {National Bureau of Economic Research},
	42	+ address = {New York},
	43	+ pages = {173--196},
	44	+ year = {1961}
39	45	}
40	46
41	47	@article{halvorsen1980interpretation,
		`@@ -54,8 +60,9 @@`
54	60	journal = {Journal of Real Estate Literature},
55	61	volume = {13},
56	62	number = {1},
57		− pages = {3--43},
58		− year = {2005}
	63	+ pages = {1--44},
	64	+ year = {2005},
	65	+ doi = {10.1080/10835547.2005.12090154}
59	66	}
60	67
61	68	@incollection{malpezzi2003hedonic,
		`@@ -65,7 +72,8 @@`
65	72	editor = {O'Sullivan, Tony and Gibb, Kenneth},
66	73	publisher = {Blackwell},
67	74	pages = {67--89},
68		− year = {2003}
	75	+ year = {2003},
	76	+ doi = {10.1002/9780470690680.ch5}
69	77	}
70	78
71	79	@article{can1992specification,
		`@@ -75,7 +83,8 @@`
75	83	volume = {22},
76	84	number = {3},
77	85	pages = {453--474},
78		− year = {1992}
	86	+ year = {1992},
	87	+ doi = {10.1016/0166-0462(92)90039-4}
79	88	}
80	89
81	90	@book{anselin1988spatial,
		`@@ -92,7 +101,8 @@`
92	101	volume = {35},
93	102	number = {2},
94	103	pages = {143--160},
95		− year = {2007}
	104	+ year = {2007},
	105	+ doi = {10.1007/s11146-007-9036-8}
96	106	}
97	107
98	108	@article{case2004modeling,
		`@@ -102,7 +112,8 @@`
102	112	volume = {29},
103	113	number = {2},
104	114	pages = {167--191},
105		− year = {2004}
	115	+ year = {2004},
	116	+ doi = {10.1023/B:REAL.0000035309.60607.53}
106	117	}
107	118
108	119	@article{kuminoff2010which,
		`@@ -112,7 +123,8 @@`
112	123	volume = {60},
113	124	number = {3},
114	125	pages = {145--160},
115		− year = {2010}
	126	+ year = {2010},
	127	+ doi = {10.1016/j.jeem.2010.06.001}
116	128	}
117	129
118	130	@article{glaeser2005why,
		`@@ -122,7 +134,8 @@`
122	134	volume = {95},
123	135	number = {2},
124	136	pages = {329--333},
125		− year = {2005}
	137	+ year = {2005},
	138	+ doi = {10.1257/000282805774669961}
126	139	}
127	140
128	141	@article{duan1983smearing,
		`@@ -132,7 +145,8 @@`
132	145	volume = {78},
133	146	number = {383},
134	147	pages = {605--610},
135		− year = {1983}
	148	+ year = {1983},
	149	+ doi = {10.1080/01621459.1983.10478017}
136	150	}
137	151
138	152	@article{cameron2015practitioner,
		`@@ -142,25 +156,17 @@`
142	156	volume = {50},
143	157	number = {2},
144	158	pages = {317--372},
145		− year = {2015}
	159	+ year = {2015},
	160	+ doi = {10.3368/jhr.50.2.317}
146	161	}
147	162
148		−@article{correia2017reghdfe,
	163	+@unpublished{correia2017reghdfe,
149	164	author = {Correia, Sergio},
150	165	title = {Linear Models with High-Dimensional Fixed Effects: An Efficient and Feasible Estimator},
151		− journal = {Working Paper},
	166	+ note = {Working paper, Duke University},
152	167	year = {2017}
153	168	}
154	169
155		−@article{gabaix2016power,
156		− author = {Gabaix, Xavier},
157		− title = {Power Laws in Economics: An Introduction},
158		− journal = {Journal of Economic Perspectives},
159		− volume = {30},
160		− number = {1},
161		− pages = {185--206},
162		− year = {2016}
163		−}
164	170
165	171	@article{mullainathan2017machine,
166	172	author = {Mullainathan, Sendhil and Spiess, Jann},
		`@@ -169,18 +175,10 @@`
169	175	volume = {31},
170	176	number = {2},
171	177	pages = {87--106},
172		− year = {2017}
	178	+ year = {2017},
	179	+ doi = {10.1257/jep.31.2.87}
173	180	}
174	181
175		−@article{cmhc2018canadian,
176		− author = {Mok, Diana and Chan, Sheryl},
177		− title = {House Prices and the Geography of the Canadian Housing Market},
178		− journal = {Canadian Journal of Urban Research},
179		− volume = {27},
180		− number = {1},
181		− pages = {1--18},
182		− year = {2018}
183		−}
184	182
185	183	@article{ekeland2004identification,
186	184	author = {Ekeland, Ivar and Heckman, James J. and Nesheim, Lars},
		`@@ -189,7 +187,8 @@`
189	187	volume = {112},
190	188	number = {S1},
191	189	pages = {S60--S109},
192		− year = {2004}
	190	+ year = {2004},
	191	+ doi = {10.1086/379947}
193	192	}
194	193
195	194	@article{bajari2005hedonic,
		`@@ -199,7 +198,8 @@`
199	198	volume = {113},
200	199	number = {6},
201	200	pages = {1239--1276},
202		− year = {2005}
	201	+ year = {2005},
	202	+ doi = {10.1086/498586}
203	203	}
204	204
205	205	@article{black1999better,
		`@@ -209,7 +209,8 @@`
209	209	volume = {114},
210	210	number = {2},
211	211	pages = {577--599},
212		− year = {1999}
	212	+ year = {1999},
	213	+ doi = {10.1162/003355399556070}
213	214	}
214	215
215	216	@article{chay2005does,
		`@@ -219,7 +220,8 @@`
219	220	volume = {113},
220	221	number = {2},
221	222	pages = {376--424},
222		− year = {2005}
	223	+ year = {2005},
	224	+ doi = {10.1086/427462}
223	225	}
224	226
225	227	@article{linden2008estimates,
		`@@ -229,7 +231,8 @@`
229	231	volume = {98},
230	232	number = {3},
231	233	pages = {1103--1127},
232		− year = {2008}
	234	+ year = {2008},
	235	+ doi = {10.1257/aer.98.3.1103}
233	236	}
234	237
235	238	@book{lesage2009introduction,
		`@@ -246,16 +249,20 @@`
246	249	volume = {52},
247	250	number = {2},
248	251	pages = {172--191},
249		− year = {2012}
	252	+ year = {2012},
	253	+ doi = {10.1111/j.1467-9787.2012.00760.x}
250	254	}
251	255
252		−@article{combes2015empirics,
253		− author = {Combes, Pierre-Philippe and Gobillon, Laurent},
254		− title = {The Empirics of Agglomeration Economies},
255		− journal = {Handbook of Regional and Urban Economics},
256		− volume = {5},
257		− pages = {247--348},
258		− year = {2015}
	256	+@incollection{combes2015empirics,
	257	+ author = {Combes, Pierre-Philippe and Gobillon, Laurent},
	258	+ title = {The Empirics of Agglomeration Economies},
	259	+ booktitle = {Handbook of Regional and Urban Economics},
	260	+ editor = {Duranton, Gilles and Henderson, J. Vernon and Strange, William C.},
	261	+ volume = {5},
	262	+ publisher = {Elsevier},
	263	+ pages = {247--348},
	264	+ year = {2015},
	265	+ doi = {10.1016/B978-0-444-59517-1.00005-2}
259	266	}
260	267
261	268	@article{moran1950notes,
		`@@ -265,7 +272,8 @@`
265	272	volume = {37},
266	273	number = {1/2},
267	274	pages = {17--23},
268		− year = {1950}
	275	+ year = {1950},
	276	+ doi = {10.1093/biomet/37.1-2.17}
269	277	}
270	278
271	279	@article{koenker1978regression,
		`@@ -275,7 +283,8 @@`
275	283	volume = {46},
276	284	number = {1},
277	285	pages = {33--50},
278		− year = {1978}
	286	+ year = {1978},
	287	+ doi = {10.2307/1913643}
279	288	}
280	289
281	290	@article{zietz2008determinants,
		`@@ -285,7 +294,8 @@`
285	294	volume = {37},
286	295	number = {4},
287	296	pages = {317--333},
288		− year = {2008}
	297	+ year = {2008},
	298	+ doi = {10.1007/s11146-007-9053-7}
289	299	}
290	300
291	301	@article{mcmillen2010issues,
		`@@ -295,7 +305,8 @@`
295	305	volume = {50},
296	306	number = {3},
297	307	pages = {712--733},
298		− year = {2010}
	308	+ year = {2010},
	309	+ doi = {10.1111/j.1467-9787.2010.00664.x}
299	310	}
300	311
301	312	@article{pace1998spatiotemporal,
		`@@ -305,7 +316,8 @@`
305	316	volume = {17},
306	317	number = {1},
307	318	pages = {15--33},
308		− year = {1998}
	319	+ year = {1998},
	320	+ doi = {10.1023/A:1007799028599}
309	321	}
310	322
311	323	% ============================================================================
		`@@ -648,3 +660,310 @@`
648	660	pages = {628--649},
649	661	doi = {10.1177/1536867X1001000406}
650	662	}
	663	+
	664	+% ============================================================================
	665	+% Added in the October 2026 editorial-board revision (v1.3).
	666	+% Every entry verified against Crossref (query + DOI) on 2026-10-17; see
	667	+% docs/CHANGELOG-REVISION.md for the verification record.
	668	+% ============================================================================
	669	+
	670	+% ---- Hedonic surveys and best practice ----
	671	+@incollection{sheppard1999hedonic,
	672	+ author = {Sheppard, Stephen},
	673	+ title = {Hedonic Analysis of Housing Markets},
	674	+ booktitle = {Handbook of Regional and Urban Economics},
	675	+ editor = {Cheshire, Paul and Mills, Edwin S.},
	676	+ volume = {3},
	677	+ publisher = {Elsevier},
	678	+ pages = {1595--1635},
	679	+ year = {1999},
	680	+ doi = {10.1016/S1574-0080(99)80010-8}
	681	+}
	682	+@article{bishop2020best,
	683	+ author = {Bishop, Kelly C. and Kuminoff, Nicolai V. and Banzhaf, H. Spencer and Boyle, Kevin J. and von Gravenitz, Kathrine and Pope, Jaren C. and Smith, V. Kerry and Timmins, Christopher D.},
	684	+ title = {Best Practices for Using Hedonic Property Value Models to Measure Willingness to Pay for Environmental Quality},
	685	+ journal = {Review of Environmental Economics and Policy},
	686	+ volume = {14},
	687	+ number = {2},
	688	+ pages = {260--281},
	689	+ year = {2020},
	690	+ doi = {10.1093/reep/reaa001}
	691	+}
	692	+@article{bishop2019estimating,
	693	+ author = {Bishop, Kelly C. and Timmins, Christopher},
	694	+ title = {Estimating the Marginal Willingness to Pay Function Without Instrumental Variables},
	695	+ journal = {Journal of Urban Economics},
	696	+ volume = {109},
	697	+ pages = {66--83},
	698	+ year = {2019},
	699	+ doi = {10.1016/j.jue.2018.11.006}
	700	+}
	701	+@article{sirmans2006value,
	702	+ author = {Sirmans, G. Stacy and MacDonald, Lynn and Macpherson, David A. and Zietz, Emily Norman},
	703	+ title = {The Value of Housing Characteristics: A Meta Analysis},
	704	+ journal = {Journal of Real Estate Finance and Economics},
	705	+ volume = {33},
	706	+ number = {3},
	707	+ pages = {215--240},
	708	+ year = {2006},
	709	+ doi = {10.1007/s11146-006-9983-5}
	710	+}
	711	+
	712	+% ---- Interpretation of semilogarithmic coefficients ----
	713	+@article{kennedy1981estimation,
	714	+ author = {Kennedy, Peter E.},
	715	+ title = {Estimation with Correctly Interpreted Dummy Variables in Semilogarithmic Equations},
	716	+ journal = {American Economic Review},
	717	+ volume = {71},
	718	+ number = {4},
	719	+ pages = {801},
	720	+ year = {1981}
	721	+}
	722	+
	723	+% ---- Inference with clustered data ----
	724	+@article{abadie2023when,
	725	+ author = {Abadie, Alberto and Athey, Susan and Imbens, Guido W. and Wooldridge, Jeffrey M.},
	726	+ title = {When Should You Adjust Standard Errors for Clustering?},
	727	+ journal = {Quarterly Journal of Economics},
	728	+ volume = {138},
	729	+ number = {1},
	730	+ pages = {1--35},
	731	+ year = {2023},
	732	+ doi = {10.1093/qje/qjac038}
	733	+}
	734	+@article{mackinnon2017wild,
	735	+ author = {MacKinnon, James G. and Webb, Matthew D.},
	736	+ title = {Wild Bootstrap Inference for Wildly Different Cluster Sizes},
	737	+ journal = {Journal of Applied Econometrics},
	738	+ volume = {32},
	739	+ number = {2},
	740	+ pages = {233--254},
	741	+ year = {2017},
	742	+ doi = {10.1002/jae.2508}
	743	+}
	744	+
	745	+% ---- Variance decomposition (relative importance) ----
	746	+@article{israeli2007shapley,
	747	+ author = {Israeli, Osnat},
	748	+ title = {A {S}hapley-Based Decomposition of the {R}-Square of a Linear Regression},
	749	+ journal = {Journal of Economic Inequality},
	750	+ volume = {5},
	751	+ number = {2},
	752	+ pages = {199--212},
	753	+ year = {2007},
	754	+ doi = {10.1007/s10888-006-9036-6}
	755	+}
	756	+@article{gromping2007estimators,
	757	+ author = {Gr{\"o}mping, Ulrike},
	758	+ title = {Estimators of Relative Importance in Linear Regression Based on Variance Decomposition},
	759	+ journal = {The American Statistician},
	760	+ volume = {61},
	761	+ number = {2},
	762	+ pages = {139--147},
	763	+ year = {2007},
	764	+ doi = {10.1198/000313007X188252}
	765	+}
	766	+@article{huettner2012axiomatic,
	767	+ author = {Huettner, Frank and Sunder, Marco},
	768	+ title = {Axiomatic Arguments for Decomposing Goodness of Fit According to {S}hapley and {O}wen Values},
	769	+ journal = {Electronic Journal of Statistics},
	770	+ volume = {6},
	771	+ pages = {1239--1250},
	772	+ year = {2012},
	773	+ doi = {10.1214/12-EJS710}
	774	+}
	775	+
	776	+% ---- Spatial econometrics ----
	777	+@article{kelejian1998generalized,
	778	+ author = {Kelejian, Harry H. and Prucha, Ingmar R.},
	779	+ title = {A Generalized Spatial Two-Stage Least Squares Procedure for Estimating a Spatial Autoregressive Model with Autoregressive Disturbances},
	780	+ journal = {Journal of Real Estate Finance and Economics},
	781	+ volume = {17},
	782	+ number = {1},
	783	+ pages = {99--121},
	784	+ year = {1998},
	785	+ doi = {10.1023/A:1007707430416}
	786	+}
	787	+@article{kelejian2001asymptotic,
	788	+ author = {Kelejian, Harry H. and Prucha, Ingmar R.},
	789	+ title = {On the Asymptotic Distribution of the {M}oran {I} Test Statistic with Applications},
	790	+ journal = {Journal of Econometrics},
	791	+ volume = {104},
	792	+ number = {2},
	793	+ pages = {219--257},
	794	+ year = {2001},
	795	+ doi = {10.1016/S0304-4076(01)00064-1}
	796	+}
	797	+@article{hill2018geospatial,
	798	+ author = {Hill, Robert J. and Scholz, Michael},
	799	+ title = {Can Geospatial Data Improve House Price Indexes? {A} Hedonic Imputation Approach with Splines},
	800	+ journal = {Review of Income and Wealth},
	801	+ volume = {64},
	802	+ number = {4},
	803	+ pages = {737--756},
	804	+ year = {2018},
	805	+ doi = {10.1111/roiw.12303}
	806	+}
	807	+@article{bogin2019local,
	808	+ author = {Bogin, Alexander and Doerner, William and Larson, William},
	809	+ title = {Local House Price Dynamics: New Indices and Stylized Facts},
	810	+ journal = {Real Estate Economics},
	811	+ volume = {47},
	812	+ number = {2},
	813	+ pages = {365--398},
	814	+ year = {2019},
	815	+ doi = {10.1111/1540-6229.12233}
	816	+}
	817	+@article{roberts2017cross,
	818	+ author = {Roberts, David R. and Bahn, Volker and Ciuti, Simone and Boyce, Mark S. and Elith, Jane and Guillera-Arroita, Gurutzeta and Hauenstein, Severin and Lahoz-Monfort, Jos{\'e} J. and Schr{\"o}der, Boris and Thuiller, Wilfried and Warton, David I. and Wintle, Brendan A. and Hartig, Florian and Dormann, Carsten F.},
	819	+ title = {Cross-Validation Strategies for Data with Temporal, Spatial, Hierarchical, or Phylogenetic Structure},
	820	+ journal = {Ecography},
	821	+ volume = {40},
	822	+ number = {8},
	823	+ pages = {913--929},
	824	+ year = {2017},
	825	+ doi = {10.1111/ecog.02881}
	826	+}
	827	+
	828	+% ---- Land, location and agglomeration ----
	829	+@article{combes2019costs,
	830	+ author = {Combes, Pierre-Philippe and Duranton, Gilles and Gobillon, Laurent},
	831	+ title = {The Costs of Agglomeration: House and Land Prices in {F}rench Cities},
	832	+ journal = {Review of Economic Studies},
	833	+ volume = {86},
	834	+ number = {4},
	835	+ pages = {1556--1589},
	836	+ year = {2019},
	837	+ doi = {10.1093/restud/rdy063}
	838	+}
	839	+@article{davis2008price,
	840	+ author = {Davis, Morris A. and Palumbo, Michael G.},
	841	+ title = {The Price of Residential Land in Large {US} Cities},
	842	+ journal = {Journal of Urban Economics},
	843	+ volume = {63},
	844	+ number = {1},
	845	+ pages = {352--384},
	846	+ year = {2008},
	847	+ doi = {10.1016/j.jue.2007.02.003}
	848	+}
	849	+@article{albouy2018metropolitan,
	850	+ author = {Albouy, David and Ehrlich, Gabriel and Shin, Minchul},
	851	+ title = {Metropolitan Land Values},
	852	+ journal = {Review of Economics and Statistics},
	853	+ volume = {100},
	854	+ number = {3},
	855	+ pages = {454--466},
	856	+ year = {2018},
	857	+ doi = {10.1162/rest_a_00710}
	858	+}
	859	+@article{diewert2015hedonic,
	860	+ author = {Diewert, W. Erwin and de Haan, Jan and Hendriks, Rens},
	861	+ title = {Hedonic Regressions and the Decomposition of a House Price Index into Land and Structure Components},
	862	+ journal = {Econometric Reviews},
	863	+ volume = {34},
	864	+ number = {1--2},
	865	+ pages = {106--126},
	866	+ year = {2015},
	867	+ doi = {10.1080/07474938.2014.944791}
	868	+}
	869	+
	870	+% ---- List prices, search and bargaining ----
	871	+@article{horowitz1992role,
	872	+ author = {Horowitz, Joel L.},
	873	+ title = {The Role of the List Price in Housing Markets: Theory and an Econometric Model},
	874	+ journal = {Journal of Applied Econometrics},
	875	+ volume = {7},
	876	+ number = {2},
	877	+ pages = {115--129},
	878	+ year = {1992},
	879	+ doi = {10.1002/jae.3950070202}
	880	+}
	881	+@article{merlo2004bargaining,
	882	+ author = {Merlo, Antonio and Ortalo-Magn{\'e}, Fran{\c c}ois},
	883	+ title = {Bargaining over Residential Real Estate: Evidence from {E}ngland},
	884	+ journal = {Journal of Urban Economics},
	885	+ volume = {56},
	886	+ number = {2},
	887	+ pages = {192--216},
	888	+ year = {2004},
	889	+ doi = {10.1016/j.jue.2004.05.004}
	890	+}
	891	+@article{anglin2003tradeoff,
	892	+ author = {Anglin, Paul M. and Rutherford, Ronald and Springer, Thomas M.},
	893	+ title = {The Trade-off Between the Selling Price of Residential Properties and Time-on-the-Market: The Impact of Price Setting},
	894	+ journal = {Journal of Real Estate Finance and Economics},
	895	+ volume = {26},
	896	+ number = {1},
	897	+ pages = {95--111},
	898	+ year = {2003},
	899	+ doi = {10.1023/A:1021526332732}
	900	+}
	901	+@article{han2016what,
	902	+ author = {Han, Lu and Strange, William C.},
	903	+ title = {What Is the Role of the Asking Price for a House?},
	904	+ journal = {Journal of Urban Economics},
	905	+ volume = {93},
	906	+ pages = {115--130},
	907	+ year = {2016},
	908	+ doi = {10.1016/j.jue.2016.03.008}
	909	+}
	910	+@article{genesove2012search,
	911	+ author = {Genesove, David and Han, Lu},
	912	+ title = {Search and Matching in the Housing Market},
	913	+ journal = {Journal of Urban Economics},
	914	+ volume = {72},
	915	+ number = {1},
	916	+ pages = {31--45},
	917	+ year = {2012},
	918	+ doi = {10.1016/j.jue.2012.01.002}
	919	+}
	920	+
	921	+% ---- Canadian evidence ----
	922	+@article{allen2009canadian,
	923	+ author = {Allen, Jason and Amano, Robert and Byrne, David P. and Gregory, Allan W.},
	924	+ title = {Canadian City Housing Prices and Urban Market Segmentation},
	925	+ journal = {Canadian Journal of Economics},
	926	+ volume = {42},
	927	+ number = {3},
	928	+ pages = {1132--1149},
	929	+ year = {2009},
	930	+ doi = {10.1111/j.1540-5982.2009.01541.x}
	931	+}
	932	+@article{hossain2009determinants,
	933	+ author = {Hossain, Belayet and Latif, Ehsan},
	934	+ title = {Determinants of Housing Price Volatility in {C}anada: A Dynamic Analysis},
	935	+ journal = {Applied Economics},
	936	+ volume = {41},
	937	+ number = {27},
	938	+ pages = {3521--3531},
	939	+ year = {2009},
	940	+ doi = {10.1080/00036840701522861}
	941	+}
	942	+@article{dube2013commuter,
	943	+ author = {Dub{\'e}, Jean and Th{\'e}riault, Marius and Des Rosiers, Fran{\c c}ois},
	944	+ title = {Commuter Rail Accessibility and House Values: The Case of the {M}ontreal {S}outh {S}hore, {C}anada, 1992--2009},
	945	+ journal = {Transportation Research Part A: Policy and Practice},
	946	+ volume = {54},
	947	+ pages = {49--66},
	948	+ year = {2013},
	949	+ doi = {10.1016/j.tra.2013.07.015}
	950	+}
	951	+@article{theriault2003modelling,
	952	+ author = {Th{\'e}riault, Marius and Des Rosiers, Fran{\c c}ois and Villeneuve, Paul and Kestens, Yan},
	953	+ title = {Modelling Interactions of Location with Specific Value of Housing Attributes},
	954	+ journal = {Property Management},
	955	+ volume = {21},
	956	+ number = {1},
	957	+ pages = {25--62},
	958	+ year = {2003},
	959	+ doi = {10.1108/02637470310464472}
	960	+}
	961	+@techreport{statcan2022fsa,
	962	+ author = {{Statistics Canada}},
	963	+ title = {Census Forward Sortation Area Boundary File, Reference Guide, Census Year 2021},
	964	+ institution = {Statistics Canada},
	965	+ type = {Catalogue no.},
	966	+ number = {92-179-G2021001},
	967	+ address = {Ottawa},
	968	+ year = {2022}
	969	+}

modified paper/sections/conclusion.tex +46 −37

		`@@ -4,44 +4,53 @@`
4	4	\label{sec:conclusion}
5	5	% ============================================================================
6	6
7		−Using a nationwide cross-section of 140{,}931 MLS listings and 1{,}153 absorbed
8		−neighbourhood fixed effects, we estimate a grand hedonic model of the Canadian
9		−residential housing market. The exercise yields three robust conclusions. First,
10		−location dominates: resolving geography from the provincial to the neighbourhood scale
11		−raises explained price variation from 57\% to 77\%, and location as a whole accounts for
12		−roughly thirty percentage points of $R^2$---more than every structural attribute
13		−combined, and a listing-level counterpart to the large land shares documented in
14		−aggregate data \citep{davis2007price,knoll2017no}. Second, the structural implicit
15		−prices behave as hedonic theory predicts and are strikingly stable across samples:
16		−living area carries an elasticity near 0.55, each full bathroom adds about 11--15\%, and
17		−bedroom counts are economically negligible once floor space is held
18		−fixed---magnitudes squarely within the range catalogued by decades of hedonic work
19		−\citep{sirmans2005composition}. Third, the neighbourhood premia are vast---a factor of
20		−roughly nine separates the most and least expensive FSAs---and the model translates into
21		−a credible valuation tool, predicting held-out prices with an out-of-sample $R^2$ of
22		−0.76 and a median error of 16\%. These conclusions survive an extensive battery of
23		−checks: the implicit prices are stable across subsamples and across the price
24		−distribution, the neighbourhood effects absorb 82\% of the spatial autocorrelation in
25		−residuals, floor space displays diminishing returns, value decays with distance to major
26		−metros as the monocentric tradition predicts, and the structural prices transfer across
27		−provinces in leave-one-province-out cross-validation.
	7	+Using a nationwide cross-section of 142{,}121 MLS listings and 1{,}155 absorbed
	8	+neighbourhood fixed effects, we estimate a hedonic price function for the Canadian
	9	+residential market and decompose the variance of log asking prices into structure and
	10	+location. The exercise yields four conclusions. First, location and structure are of
	11	+comparable importance: resolving geography from the provincial to the neighbourhood scale
	12	+raises explained variation from 60\% to 77\%, and a Shapley--Owen decomposition assigns
	13	+about 34 percentage points of $R^2$ to structure and 36 to location, so that the
	14	+neighbourhood is the single largest block once province and neighbourhood are grouped,
	15	+while the sequential ladder that most hedonic papers report would credit structure with 44
	16	+points and location with 32. The answer to ``how much is location worth?'' is therefore
	17	+inseparable from the attribution rule, and we have made that rule explicit. Second, the
	18	+structural implicit list prices behave as hedonic theory predicts and are stable across
	19	+eight samples: living area carries an elasticity of 0.53, each full bathroom adds about
	20	+12\%, and bedroom counts are economically negligible once floor space is held fixed. The
	21	+area and bathroom effects survive province-level wild-cluster bootstrap inference; the
	22	+small half-bathroom and bedroom effects do not. Third, the neighbourhood premia are
	23	+vast---a factor of about ten separates Vancouver's west side from rural Newfoundland for an
	24	+identical dwelling---and the neighbourhood effects remove 87\% of the spatial
	25	+autocorrelation in residuals. Fourth, the model is a credible valuation tool whose accuracy
	26	+depends on spatial support: it values held-out listings in known neighbourhoods with an
	27	+out-of-sample $R^2$ of 0.77 and a median absolute error of 15.6\%, and listings in
	28	+neighbourhoods never observed with an $R^2$ of 0.74 and a median error of 17\% when the
	29	+neighbourhood effect is interpolated from adjacent FSAs. Floor space displays mild
	30	+diminishing returns concentrated above 150~m$^2$, the location premium declines by about
	31	+5.5\% per doubling of distance to the nearest metropolitan centre, implicit prices vary
	32	+sensibly across the price distribution and across provinces, and the structural prices
	33	+transfer across most provinces but not to British Columbia.
28	34
29		−The reading we propose is deliberately modest and, we believe, well supported: in
30		−Canada, the single largest priced component of a home is its neighbourhood. That fact
31		−is simultaneously a measurement result (for assessment, taxation and index
32		−construction), a validation result (a transparent fixed-effects design removes the
33		−spatial dependence that motivates far more elaborate machinery), and a policy result
34		−(the premium dispersion that supply constraints and agglomeration generate is the
35		−dominant term in the decomposition). Section~\ref{sec:discussion} details the
36		−limitations---list prices, unobserved quality, noisy lot data, a single
37		−cross-section---and the research agenda they imply. Even within those bounds, the model
38		−provides a transparent, reproducible benchmark for automated valuation, market
39		−monitoring and the welfare analysis of local amenities across the Canadian residential
40		−market.
	35	+The reading we propose is deliberately modest. In Canada, the neighbourhood is at least as
	36	+large a component of a dwelling's asking price as the dwelling itself, and it is the
	37	+component whose dispersion is greatest. That fact is simultaneously a measurement result
	38	+(for assessment, taxation and index construction), a validation result (a transparent
	39	+fixed-effects design removes most of the spatial dependence that motivates far more
	40	+elaborate machinery, and its fixed effects are smooth enough to be interpolated), and a
	41	+policy result (the premium dispersion that supply constraints and agglomeration generate is
	42	+largest precisely in the most expensive markets). Section~\ref{sec:discussion} details the
	43	+limitations---asking rather than transaction prices, unobserved quality, noisy fields, a
	44	+single cross-section, postal rather than planning geography---and the research agenda they
	45	+imply. Within those bounds, the model provides a transparent and reproducible benchmark for
	46	+automated valuation, market monitoring and the analysis of location value across the
	47	+Canadian residential market.
41	48
42	49	\vspace{0.6em}
43	50	\noindent\textbf{Reproducibility.} Data engineering was performed in \texttt{DuckDB} and
44		−\texttt{pandas}; estimation used \texttt{statsmodels} (cluster-robust OLS) and
45		−\texttt{linearmodels} (absorbing least squares); figures were produced in
46		−\texttt{matplotlib}. All tables and figures are generated programmatically from the
47		−source database by the numbered scripts in the accompanying repository.
	51	+\texttt{pandas}; estimation used \texttt{statsmodels} for cluster-robust OLS and quantile
	52	+regression and \texttt{linearmodels} for absorbing least squares; figures were produced
	53	+in \texttt{matplotlib}. All tables and figures are generated programmatically from the source
	54	+database by the numbered scripts in the accompanying repository, and
	55	+Appendix~\ref{app:reconciliation} reconciles the present estimates with those of the May
	56	+2026 version of this paper.

modified paper/sections/data.tex +85 −47

Large diff (157 lines) is not loaded by default to keep the page fast.Load diff

modified paper/sections/discussion.tex +159 −114

Large diff (290 lines) is not loaded by default to keep the page fast.Load diff

modified paper/sections/introduction.tex +99 −80

Large diff (189 lines) is not loaded by default to keep the page fast.Load diff

modified paper/sections/literature.tex +161 −136

Large diff (316 lines) is not loaded by default to keep the page fast.Load diff

modified paper/sections/methodology.tex +167 −52

Large diff (246 lines) is not loaded by default to keep the page fast.Load diff

modified paper/sections/results.tex +220 −123

Large diff (398 lines) is not loaded by default to keep the page fast.Load diff

modified paper/sections/robustness.tex +159 −93

Large diff (307 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/coef_M1.csv +9 −9

Large diff (19 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/coef_M2.csv +14 −14

Large diff (29 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/coef_M3.csv +22 −22

Large diff (45 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/coef_M5.csv +25 −25

Large diff (51 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/ext2.json +11 −11

Large diff (33 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/fit.json +12 −10

Large diff (35 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/fsa_premia.csv +870 −869

Large diff (1,740 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/gradient_bins.csv +8 −8

Large diff (17 lines) is not loaded by default to keep the page fast.Load diff

added results/reproduced/gradient_metro.json +59 −0

Large diff (60 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/grand_model.parquet +0 −0

Binary file not shown.

modified results/reproduced/heterogeneity.csv +9 −8

Large diff (18 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/lopo.csv +9 −8

Large diff (18 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/moran_coords.npy +0 −0

Binary file not shown.

modified results/reproduced/moran_resid_grand.npy +0 −0

Binary file not shown.

modified results/reproduced/moran_resid_struct.npy +0 −0

Binary file not shown.

added results/reproduced/moran_sensitivity.csv +13 −0

Large diff (13 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/nonlinear_band.csv +60 −60

Large diff (121 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/oos.json +9 −9

Large diff (21 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/quantile.csv +6 −6

Large diff (13 lines) is not loaded by default to keep the page fast.Load diff

modified results/reproduced/robustness.csv +8 −6

Large diff (15 lines) is not loaded by default to keep the page fast.Load diff

added results/reproduced/se_comparison.csv +6 −0

Large diff (6 lines) is not loaded by default to keep the page fast.Load diff

added results/reproduced/shapley.csv +9 −0

Large diff (9 lines) is not loaded by default to keep the page fast.Load diff

added results/reproduced/spatial_holdout.json +236 −0

Large diff (237 lines) is not loaded by default to keep the page fast.Load diff

added results/reproduced/spline.csv +6 −0

Large diff (6 lines) is not loaded by default to keep the page fast.Load diff

added results/reproduced/summary_stats.csv +10 −0

Large diff (10 lines) is not loaded by default to keep the page fast.Load diff

added results/tables/gradient.tex +31 −0

Large diff (31 lines) is not loaded by default to keep the page fast.Load diff

modified results/tables/lopo.tex +11 −10

Large diff (27 lines) is not loaded by default to keep the page fast.Load diff

added results/tables/moran_sensitivity.tex +19 −0

Large diff (19 lines) is not loaded by default to keep the page fast.Load diff

modified results/tables/oos.tex +10 −10

Large diff (31 lines) is not loaded by default to keep the page fast.Load diff

modified results/tables/quantile.tex +8 −8

Large diff (21 lines) is not loaded by default to keep the page fast.Load diff

modified results/tables/regression.tex +11 −11

Large diff (39 lines) is not loaded by default to keep the page fast.Load diff

modified results/tables/robustness.tex +10 −8

Large diff (29 lines) is not loaded by default to keep the page fast.Load diff

added results/tables/se_comparison.tex +21 −0

Large diff (21 lines) is not loaded by default to keep the page fast.Load diff

added results/tables/shapley.tex +27 −0

Large diff (27 lines) is not loaded by default to keep the page fast.Load diff

added results/tables/spatial_holdout.tex +22 −0

Large diff (22 lines) is not loaded by default to keep the page fast.Load diff

added results/tables/spline.tex +21 −0

Large diff (21 lines) is not loaded by default to keep the page fast.Load diff

modified results/tables/summary_stats.tex +11 −11

Large diff (33 lines) is not loaded by default to keep the page fast.Load diff

added results/tables/type_effects.tex +29 −0

Large diff (29 lines) is not loaded by default to keep the page fast.Load diff

modified scripts/02_estimate_core.py +2 −0

Large diff (8 lines) is not loaded by default to keep the page fast.Load diff

modified scripts/03_estimate_extended.py +13 −1

Large diff (26 lines) is not loaded by default to keep the page fast.Load diff

modified scripts/04_make_figures.py +49 −26

Large diff (113 lines) is not loaded by default to keep the page fast.Load diff

modified scripts/05_make_tables.py +270 −23

Large diff (385 lines) is not loaded by default to keep the page fast.Load diff

added scripts/06_revision_analyses.py +393 −0

Large diff (393 lines) is not loaded by default to keep the page fast.Load diff

modified src/wp9/config.py +9 −0

Large diff (15 lines) is not loaded by default to keep the page fast.Load diff

modified src/wp9/sample.py +10 −4

Large diff (39 lines) is not loaded by default to keep the page fast.Load diff

spb/wp9_uqo

révision scientifique complète (editorial board, 2026-10-17)

Theme