# PAPER_REVIEW — Critical assessment before the scholarly upgrade Assessment of `paper/` (UQO WP5, 43-page build of 2026-08-05) and of the analysis code in `scripts/`/`src/`, written **before** the literature expansion. Results, data, and figures are out of scope: only framing, positioning, and scholarship are assessed. ## 1. Core contribution — stated, but under-positioned The paper claims three contributions (intro, ¶6): (i) first granular listing-level analysis of the Airbnb–rent nexus in Quebec; (ii) a multi-method design (hedonic OLS, SAR/SEM, quantile, ML) on a single dataset; (iii) policy relevance for Canadian short-term-rental (STR) regulation. These are credible but under-argued: - The **"first for Quebec"** claim is asserted without engaging the Canadian STR literature at all — notably the McGill group's work on STRs in Canadian cities (Wachsmuth and coauthors), which quantified Montreal's STR market and housing-loss estimates. Without citing it, contribution (i) is exposed. - The **multi-method** point is defensible but currently reads as a list; it should be framed against the identification hierarchy of the field (IV, quasi-experiments, panels) — i.e., what a careful cross-sectional multi-method study adds *given* that stronger designs exist elsewhere. - The buffer-based exposure construction (exact Haversine counts at 250 m–2 km around each rental listing) is arguably the paper's most distinctive methodological feature and deserves to be foregrounded as such. ## 2. Literature review — thin and dated for a 2026 paper The current §2 has ~24 references and six one-paragraph subsections. Major gaps: 1. **Quasi-experimental Airbnb–housing studies are absent.** The review stops at Barron et al., Horn & Merante, Garcia-López et al., Sheppard & Udell. Missing: Los Angeles ordinance evidence (Koster–van Ommeren–Volkhausen), New Orleans regulation (Valentin), Berlin (Duso et al.), Portugal (Franco & Santos), France (Ayouba et al.), London (Shabrina et al.), Amsterdam/hotel margins, and the NYC distributional analysis (Calder-Wang). These are the papers a referee will name first. 2. **No Canadian evidence.** Combs–Kerrigan–Wachsmuth (national STR analysis), Wachsmuth–Weisler is cited but only for the rent-gap concept, and the Toronto displacement work (Grisdale) is absent. For a Quebec paper this is the single most damaging gap. 3. **Sharing-economy economics** (peer-to-peer market design and welfare): Zervas et al. (hotel displacement), Farronato & Fradkin (welfare effects), Einav–Farronato–Levin (P2P markets), Sundararajan (book) — none cited. These ground the supply-diversion mechanism the paper leans on. 4. **Airbnb pricing hedonics** (Model 2 has no dedicated literature): Wang & Nicolau, Gibbs et al., Ert et al. on trust/reputation — the superhost discount found in Table 4 cannot be discussed against prior findings without them. 5. **Hedonic methodology**: only Rosen + two survey chapters. Missing Lancaster (characteristics demand), Sirmans et al. (empirical composition survey), Kuminoff–Parmeter–Pope (specification reliability), Malpezzi (applied review). 6. **Spatial econometrics**: Anselin and LeSage & Pace are cited, but the actual estimators used (GM_Lag / GM_Error) descend from **Kelejian & Prucha (1998, 1999)** — currently uncited, which is a methods-citation hole. Housing applications (Can; Dubin) also absent. 7. **Quantile regression**: only Koenker–Bassett + Zietz et al. Missing the canonical applied reference (Koenker & Hallock) and housing-distribution work (McMillen). 8. **ML methods**: Zou & Hastie (elastic net) uncited although the method is used; scikit-learn (Pedregosa et al.) uncited although it produced every ML number. 9. **Tourism gentrification**: Wachsmuth–Weisler is there, but the strand's antecedents (Gotham's "tourism gentrification"; Cocola-Gant; Gurran & Phibbs on planning responses) are absent, leaving §7's policy discussion free-floating. ## 3. Weak or unsupported argumentation - **Intro ¶3 (Quebec context)**: "over 11 million visitors annually", "vacancy rates … historically low levels" — plausible but uncited; either cite a source that can be verified or soften to qualitative statements. *Flagged for the author rather than silently sourced.* - **§5.2 superhost discount**: the composition-effect interpretation is offered without any literature on reputation pricing (Ert et al.; Gibbs et al. find superhost *premia* — a disagreement worth stating). - **§6.5 ring decomposition** is honest but isolated; connecting it to the spatial-decay findings in Garcia-López et al. and Koster et al. would turn a caveat into a comparison. - **Discussion** interprets magnitudes only against the paper's own numbers; no comparison of the 0.4%-per-listing semi-elasticity with Barron et al.'s elasticity, Horn & Merante's 0.4% per one-SD, Koster et al.'s ~2% price/rent drops from bans, Franco & Santos, or Ayouba et al. The single most useful addition a reader could ask for. - **Limitations** are listed (6 items) but not developed: no discussion of measurement error in scraped exposure (platform coverage), MAUP/buffer-choice issues, or asking-vs-contract rent literature. ## 4. Underdeveloped sections (length audit) | Section | Current state | Verdict | |---|---|---| | 1 Introduction | 7 ¶, good flow | Needs: Canadian/Quebec literature hook, contribution sharpening vs. quasi-experimental strand, magnitude preview | | 2 Literature | ~2.5 pp, 6 mini-sections | **Weakest section** — needs thematic restructuring + ~25 new refs | | 3 Data | Solid | Minor: platform-coverage citation | | 4 Methodology | Solid post-restructuring | Needs method citations (K&P 1998/1999; Zou–Hastie; Koenker–Hallock; Pedregosa) | | 5 Results | Accurate post-correction | Add brief literature anchors when interpreting (superhost, quantile pattern) | | 6 Robustness | Strong after Table 12 | Fine | | 7 Discussion | 5 subsections, ~3 pp | Needs: magnitude comparison table vs. prior studies, expanded limitations, future work | | 8 Conclusion | Fine | Light touch | ## 5. Positioning: current vs. target **Currently reads as:** a competent regional case study with a methods buffet. **Should read as:** the first listing-level, multi-method quantification of the Airbnb–rent association for Quebec — a market with distinctive regulation (CITQ registration, Bill 25) and Canada's second-largest STR market — explicitly placed *between* the correlational early literature and the quasi-experimental strand, with transparent identification limits and magnitudes benchmarked against both. ## 6. Upgrade plan (executed in Phases 2–3) 1. Verify and add ~25–30 references (OpenAlex-verified metadata, DOIs) → target ≈ 50–55 total. 2. Restructure §2 into thematic subsections: (a) STR platforms & housing costs — correlational to quasi-experimental; (b) Canadian & Quebec evidence; (c) sharing-economy market effects; (d) STR pricing hedonics; (e) hedonic theory & practice; (f) spatial econometrics; (g) distributional/quantile; (h) ML in housing; (i) tourism gentrification & regulation. 3. Introduction: +3 paragraphs (positioning, Canadian gap, magnitude preview). 4. Methods: ground every estimator with its source. 5. Discussion: add "Findings in context of the literature" subsection with a magnitude-comparison narrative; expand limitations (measurement error, MAUP, asking rents) and future work. 6. Length target: ≈ +30 % prose (43 → ~55 pages). ## 7. Items the author must verify (not resolved by me) - The uncited factual claims in intro ¶3 (visitor counts, vacancy rates). - Whether "first granular listing-level analysis for Quebec" survives the McGill-group literature now cited (I have worded the claim more precisely: first *listing-level hedonic/econometric* study for Quebec, distinguishing it from the McGill descriptive/market-measurement work). - The regulatory description of CITQ registration and Montreal borough rules (§7.2) is uncited; verify against current provincial law (Bill 25, 2023).