# UPGRADE_REPORT — Scholarly deepening of the paper (2026-08-05)

Scope: literature, framing, and prose only. No result, number, table, or figure was changed. The paper grew from 43 to 51 pages; references from 24 to 56 (all cited); the PDF builds cleanly (0 unresolved references/citations, 0 BibTeX warnings). The pre-upgrade critical assessment is in PAPER_REVIEW.md.

# 1. New references added (32) — with verification and one-line justification

Every entry was verified against Crossref (title, authors, year, volume, issue, pages, DOI) — the JMLR paper against jmlr.org — before being added. No citation was written from memory alone.

# STR platforms and housing costs (8)

# Canadian evidence (2)

# Sharing-economy economics (3)

# STR pricing hedonics (3)

# Hedonic theory & practice (4)

lancaster1966new (JPE 74(2) — characteristics demand), sirmans2005composition (J. Real Estate Lit. 13(1) — survey of model composition), malpezzi2003hedonic (Blackwell chapter — applied review), kuminoff2010which (JEEM 60(3) — specification guidance that our FE + functional-form strategy follows).

# Spatial econometrics (4)

kelejian1998generalized (JREFE 17(1)) and kelejian1999generalized (IER 40(2)) — the actual sources of the GM_Lag/GM_Error estimators used in Model 4 (previously uncited: a genuine methods-citation hole); can1992specification (RSUE 22(3)) and dubin1988estimation (REStat 70(3)) — spatial dependence in hedonic housing models.

# Quantile & distribution (2)

koenker2001quantile (JEP 15(4) — canonical applied reference), mcmillen2008changes (JUE 64(3) — house-price distribution decomposition).

# ML & software (2)

zou2005regularization (JRSS-B 67(2) — elastic net, used but uncited), pedregosa2011scikit (JMLR 12 — the software that produced every ML number).

# Tourism gentrification & regulation (3)

gotham2005tourism (Urban Studies 42(7) — coined "tourism gentrification"), cocola2016holiday (Sociological Research Online 21(3)), gurran2017when (JAPA 83(1) — planners' response to Airbnb).

# Book (1)

sundararajan2016sharing (MIT Press, 2016 — crowd-based capitalism; verified via publisher metadata/reviews, no DOI as is normal for books).

# Bibliographic fix to an existing entry

glaeser2005urban (Glaeser, Gyourko & Saks, "Urban Growth and Housing Supply", J. Econ. Geography 6(1): 71–89) had year 2005; the article is 2006. Corrected. Note the key name still says 2005 (citations render correctly as 2006).

# 2. Sections expanded and how

• §1 Introduction (+~40%): P2P framing with citations (¶1); the correlational/quasi-experimental evidence base (¶2); the Canadian gap made explicit with Combs et al. and Grisdale (¶3); a new magnitude-preview paragraph comparing our $6–10/listing to Berlin's 7–13 €/listing; the three contributions rewritten with citations and a sharper novelty claim ("first listing-level econometric analysis … in any Canadian market").
• §2 Literature Review (fully rebuilt, ~3× longer): now 10 thematic subsections with signposted structure — STR & housing costs (correlational vs. quasi-experimental waves), Canadian/Quebec evidence, P2P economics, STR pricing, hedonic theory/practice, spatial econometrics, quantile methods, ML in housing, tourism gentrification & regulation, and an explicit Positioning and Contribution subsection stating the three gaps the paper fills. Every strand ends by connecting to this paper rather than summarising in isolation.
• §4 Methodology: every methodological choice now grounded — Rosen/Malpezzi at the hedonic equation, Kuminoff et al. for the FE strategy, Kelejian & Prucha for GM_Lag/GM_Error, Koenker & Hallock for quantiles, Tibshirani / Zou–Hastie / Breiman / Friedman / Pedregosa for the ML stack.
• §5 Results: two literature anchors added — the superhost discount is now explicitly contrasted with the premia in Wang & Nicolau / Ert et al., and the quantile pattern tied to Zietz et al. / McMillen.
• §7 Discussion (+~50%): new subsection 7.1 "Findings in the Context of the Literature" with three quantitative benchmarks (Berlin per-listing, LA ordinance, Boston standardised) and two candid disagreements (superhost discount; null entire-home share); Limitations expanded from 6 to 7 items with citations, adding exposure measurement error (attenuation argument) and buffer/MAUP sensitivity; new "Future Work" subsection with four extensions ordered by data requirements (repeated scrapes → panel; Quebec's 2023 enforcement change as a natural experiment; host-resolved data; lease registry rents).
• §8 Conclusion: strengthened contribution sentence (Canadian evidence base), cross-design corroboration sentence; no overselling added.

# 3. Claims flagged for the author's verification (not resolved by me)

1. Intro ¶3 factual claims — "over 11 million visitors annually", "vacancy rates … historically low": still uncited (kept verbatim from the original). Cite Tourisme Montréal / CMHC Rental Market Survey, or soften.
2. Regulatory description (§7.3 and §2.10): CITQ registration and Montreal zoning are described from the original text; verify against the current state of Quebec's Bill 25 (2023) and Montreal by-laws before submission.
3. "First listing-level econometric analysis in any Canadian market" — worded to be defensible against Combs et al. (descriptive) and Grisdale (qualitative), but a final search for very recent Canadian working papers is advisable before submission.
4. Berlin magnitude comparison (§7.1): 7–13 €/month per listing is taken from Duso et al. (2024)'s headline range; double-check the exact figure in the published version when quoting it in referee responses.

# 4. Literature that potentially contradicts or complicates the findings

Reported in the paper itself (§7.1), not hidden:

• Superhost premium literature (Wang & Nicolau 2017; Ert et al. 2016; Gibbs et al. 2018) finds positive reputation premia; our Model 2 finds a significant superhost discount. The paper now states the disagreement and offers a composition interpretation.
• Entire-home composition: Horn & Merante (2017) find effects concentrated in entire-home density, and Combs et al. (2020) build the Canadian housing-loss argument on frequently-rented entire homes; our Model 1e finds no significant entire-home-share association. Stated as a non-supported prediction with two candidate explanations.
• Cross-market heterogeneity: Ayouba et al. (2020) find no significant rent effect in several French cities — a caution against generalising our positive association across all Quebec markets, reflected in the external- validity limitation.
• Model 2 null (no city-rent premium in nightly prices) runs against the opportunity-cost intuition articulated in, e.g., Barron et al.; the paper frames this as an informative null driven by resort/urban composition.

# 5. Build status

latexmk clean build: 51 pages, 0 unresolved references, 0 unresolved citations, 0 BibTeX warnings, all 56 bibliography entries cited in the text.