spb/wp4_uqo

1% ==========================================================================2\section{Introduction}3\label{sec:introduction}4% ==========================================================================56A short-term rental listing is a bundle of characteristics priced by an7individual host rather than by a centralised market maker. The theory of8hedonic prices \citep{Rosen1974}, which builds on the characteristics view9of demand of \citet{Lancaster1966}, treats the observed price of such a10differentiated good as the sum of the implicit prices of its attributes,11and the hedonic price function has become the standard empirical tool for12housing and property markets \citep{Malpezzi2003, Sirmans2005}. Short-term13rental platforms offer an unusually complete view of this function: every14listing displays its capacity, its type, its amenities, its host's status15and its reputation, together with a price that the host chose. Since16\citet{WangNicolau2017} and \citet{Gibbs2018}, a sizeable literature has17estimated hedonic price equations for Airbnb listings in individual cities18and across cities \citep{ChenXie2017, Teubner2017, PerezSanchez2018,19Cai2019, Lorde2019, Gyodi2021, TongGunter2022}, with the recurrent finding20that capacity, room type, location and reputation signals explain most of21the cross-sectional variation in displayed prices, and with a secondary22literature on the value of trust signals such as Superhost status, review23scores and host photos \citep{Ert2016, Zervas2021}.2425Quebec is an interesting and little-studied case. The province combines a26large metropolitan market (Montreal), a heritage tourism market (Quebec27City) and an extensive resort and cottage market in the Laurentians,28Charlevoix and the Eastern Townships, so that a provincial sample mixes29urban apartments with rural chalets in a way that single-city studies30cannot. Canadian evidence on short-term rentals has concentrated on their31housing-market consequences \citep{WachsmuthWeisler2018, Combs2020}; the32only hedonic study we are aware of, \citet{Deboosere2019}, estimates a33multilevel model of prices and revenue for Montreal and finds that34accessibility and professionalisation dominate. No provincial hedonic35analysis exists, and none uses the newer platform signals (the Guest36Favourite badge, introduced in 2023) or compares the hedonic model with37quantile, spatial and machine-learning estimators on the same data.3839This paper estimates hedonic price functions for a province-wide40cross-section of Airbnb listings in Quebec scraped from the public search41pages of the platform in early 2026. The data are a snapshot of 5,00042listing cards, of which 3,456 display a price. They are not an Inside43Airbnb extract: the record is the listing card, which carries the title,44the displayed price, the star rating, the review count, capacity, bedroom45and bathroom counts, an amenity count, the Superhost and Guest Favourite46badges, the pet policy and the coordinates, but neither the room type, the47host identifier, the amenity list, the cleaning fee nor the minimum stay.48The limited field set is a constraint that shapes the paper, and we49describe exactly what each variable measures rather than relabelling it.5051The first task of the paper is a measurement problem that any user of52scraped Airbnb prices must confront. The displayed price field is53bimodal on the logarithmic scale: one mode lies near \$150 and a second54near \$2,500, and 26.5\% of priced listings carry a ``nightly'' price55above \$1,000. The second mode is populated by ordinary one- and56two-bedroom rental units in Montreal, its listings have prices that are57about twenty times those of otherwise similar listings in the first mode,58and the pattern is best explained by the platform displaying the total59price of a multi-night stay for part of the search results. We document60this with a two-component mixture model and a regime regression, and we61estimate the hedonic models on the subsample priced at or below \$1,000,62with the threshold varied in robustness checks. The companion paper on63Airbnb and rents (UQO Working Paper No.~5) uses the same scrape and64reaches the same diagnosis; here we treat it as the point of departure.6566The second task is the hedonic analysis itself. We estimate a sequence of67nested ordinary least squares specifications that add amenity, host and68reputation, location and finally city fixed effects to a structural69baseline, with heteroskedasticity-robust and city-clustered standard70errors; quantile regressions at five quantiles of the conditional71log-price distribution, with bootstrap tests of inter-quantile72differences; spatial lag, spatial error and spatial Durbin models73estimated by maximum likelihood on a nearest-neighbour weights matrix,74with direct and indirect impacts computed as in \citet{LeSagePace2009};75and five supervised learning models evaluated under both random and76spatially blocked cross-validation \citep{Roberts2017}, with the best77model interpreted through SHAP values \citep{Lundberg2017, Lundberg2020}.78Throughout, the estimates are read as conditional associations in a79cross-section of asking prices, not as structural marginal willingness to80pay \citep{Kuminoff2010, Bishop2020}.8182Three results stand out. First, once the stay-price regime is removed,83the hedonic model behaves as the literature leads one to expect: capacity84variables, title category and location explain the bulk of price85variation, with an $R^2$ that rises from about one-third in the structural86baseline to about one-half with city fixed effects, compared with87$R^2$ values of 0.05 to 0.27 obtained on the contaminated full sample in88the first version of this paper. Second, reputation signals carry89precisely estimated but modest premia once capacity and location are90held fixed: Superhost status and the Guest Favourite badge are each worth91a few percent of the nightly price, the review score matters mainly at92the lower quantiles, and the number of reviews is negatively associated93with price, consistent with high-turnover listings being cheaper.94Third, flexible estimators improve on the fixed-effects hedonic model95by a meaningful but bounded margin: gradient-boosting models raise the96out-of-sample $R^2$ from 0.47 to 0.56 under random cross-validation and97from 0.41 to 0.48 when whole areas are held out, every model loses about98seven points of $R^2$ in the second design, and the SHAP decomposition of99the best model ranks capacity first and the badges last, as the hedonic100coefficients do, while exposing one sign-switching association (the101review count) that a constant coefficient averages away. The non-linearities that the learners capture102refine the hedonic surface rather than overturn it.103104The contribution is therefore twofold. Substantively, the paper provides105the first province-wide hedonic estimates for Quebec's short-term rental106market and documents how implicit prices differ between the urban and the107resort segments and across the price distribution. Methodologically, it108shows how a measurement defect of scraped platform prices can be detected109and handled, and it benchmarks, on one data set, how much the choice of110estimator matters relative to the choice of sample.111112The remainder of the paper is organised as follows.113Section~\ref{sec:data} describes the data, the price-field diagnosis and114the analysis sample. Section~\ref{sec:methodology} sets out the hedonic,115quantile, spatial and machine-learning specifications and the inference116used. Section~\ref{sec:results} reports the results,117Section~\ref{sec:robustness} the robustness checks, and118Section~\ref{sec:conclusion} concludes with limitations and directions119for future work.120