% ==========================================================================
\section{Conclusion}
\label{sec:conclusion}
% ==========================================================================

This paper has estimated hedonic price functions for a province-wide
cross-section of Airbnb listings in Quebec from a scrape of 5,000 listing
cards. Its first finding concerns the data: the displayed price field
mixes nightly rates with the total prices of stays of about three weeks,
for roughly 28\% of priced listings, concentrated among small Montreal
rental units with few reviews. A two-regime mixture of hedonic
regressions separates the two populations sharply and leaves 2,480
listings with nightly prices. Every estimate in the first version of this
paper was computed on the mixture; none survives, and the magnitudes
reported there (a bedroom premium above 200\%, an 11\% premium per tenth
of a rating point, negative ``amenity index'' effects) were artefacts of
the contamination and of regressors that should not have been
constructed. We have documented these corrections in full so that the
lesson is usable by others who work with scraped platform prices: the
bimodality was visible in the first summary table, and a ratio of
medians by bedroom count is enough to diagnose it.

On the nightly-price sample the hedonic model behaves as the literature
leads one to expect. Capacity, bathrooms and bedrooms, the shared-bathroom
indicator that stands in for the unobserved room type, the number of
photos and the displayed rating are the priced attributes; title
categories matter once location is held fixed, with chalets and houses
priced about 8 to 10\% above rental units of the same capacity; local
density is positively priced within cities. The Superhost and Guest
Favourite badges carry no premium conditional on capacity, type and
location, and the number of reviews is unrelated to price on average.
The specification with city fixed effects explains half of the variance
of log prices, with a root mean squared error of 0.475 log points, and
its coefficients are stable to the choice of fixed-effect unit, to
clustering, to the regime cut-off and to the exclusion of commercial
lodging or of Montreal.

The three complementary estimators each add something the linear model
does not show. Quantile regression reveals that cheap listings are priced
per guest and certified by reviews and the Superhost badge, whereas
expensive listings are priced per room and per bathroom, closer to the
centre, and are cheaper when heavily reviewed; the inter-quantile
differences for bedrooms, bathrooms and review counts are statistically
significant. The spatial models find moderate spatial dependence
($\hat\rho \approx 0.20$--$0.26$, $\hat\lambda \approx 0.28$) and a spatial
multiplier of about 1.25, but the direct impacts are within a few
percent of the OLS coefficients, so the hedonic estimates are not
materially biased by the omitted spatial lag. The machine-learning
benchmarks improve the out-of-sample $R^2$ from 0.47 to 0.56 under random
cross-validation and from 0.41 to 0.48 when whole areas are held out; the
seven-point loss under spatial blocking applies to every model, and the
SHAP decomposition of the best learner ranks capacity first and the
badges last, as the hedonic coefficients do, and attributes to the
review count a weight that the linear model cannot, because its
association with price changes sign along the distribution. The flexible models refine the hedonic surface
rather than contradict it, and a median pricing error of about one
quarter of the nightly price is the practical limit of valuation from
card-level information.

\textit{Limitations.} The price is an asking price displayed on a single
day and excludes the cleaning fee; the data are a cross-section, so
seasonality and within-listing dynamics are unobserved and all
coefficients are conditional associations rather than marginal
willingness to pay. The room type, the host identifier and the amenity
list were not scraped, so professionalisation and specific amenities
cannot be priced, and the stay-regime listings cannot be converted to
nightly rates. The scrape reflects the platform's search results on the
collection dates and is denser in urban and resort markets.

\textit{Future work.} A repeated scrape that records the search dates,
the stay length behind each displayed price, the room type and the host
identifier would remove the main measurement problems and allow panel
estimation of within-listing price responses to reviews, badges and
seasons. Linking listings to realised bookings or to the calendar would
replace asking prices with revenue. Finally, the Quebec market offers a
natural setting to study how Montreal's registration and
primary-residence rules shift the composition of listings and their
prices, a question that the cross-sectional stay-price pattern documented
here already hints at.