révision scientifique complète (editorial board, 2026-10-17) — citations complétées (94/94 rendues)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
3 changed files +3 −3
modified
paper/main.pdf
+0 −0
Binary file not shown.
modified
paper/sections/conclusion.tex
+1 −1
@@ -10,4 +10,4 @@ The semantic block raises the adjusted $R^2$ of a structural hedonic model from | ||
| 10 | 10 | |
| 11 | 11 | The methodological conclusion is twofold. Reference-based projection does what it promises at the level of the block and of a few well-separated dimensions: it turns an opaque embedding into named, auditable features whose joint information content is robust, and it does so at negligible cost. It does not deliver twenty stable implicit prices, because twenty projections of the same vector are too collinear for their individual coefficients to be invariant to wording and representation choices that have no economic content. The practical recommendations that follow---report and control for the language composition of the corpus or use a language-aligned encoder, design references with disjoint vocabularies, treat block- and domain-level tests as the primary inferential objects, report the sensitivity of individual coefficients to reference phrasing, and validate out of sample with spatially grouped folds---apply to any application of continuous dictionaries in the text-as-data tradition \citep{gentzkow2019text, ash2023text}, in housing or elsewhere: product descriptions in online marketplaces, job postings \citep{marinescu2020opening}, patent texts \citep{kelly2021measuring}, corporate disclosures \citep{loughran2011liability, garcia2023colour}. |
| 12 | 12 | |
| 13 | −Several extensions would strengthen the evidence. Transaction prices would replace the pricing behaviour of sellers by the valuation of buyers and allow the list-to-sale gap to be modelled as a function of text. Successive listings of the same property, or platform experiments, would offer within-property variation for causal statements. A principled design of the references---disjoint vocabularies, orthogonalisation in embedding space, or data-driven anchors validated against expert labels---would address the attribution problem directly. A French-specific or domain-adapted encoder \citep{martin2020camembert} and listing photographs \citep{glaeser2018computer, poursaeed2018vision} would complete the measurement of quality from unstructured listing content. The replication pipeline accompanying the paper, which reproduces every number reported here from the raw database, is intended to make these extensions straightforward. | |
| 13 | +Several extensions would strengthen the evidence. Transaction prices would replace the pricing behaviour of sellers by the valuation of buyers and allow the list-to-sale gap to be modelled as a function of text. Successive listings of the same property, or platform experiments, would offer within-property variation for causal statements, and buyer-side data on search behaviour would show whether the qualitative segments that brokers write to are the segments buyers search in \citep{piazzesi2020segmented}. A principled design of the references---disjoint vocabularies, orthogonalisation in embedding space, or data-driven anchors validated against expert labels---would address the attribution problem directly. A French-specific or domain-adapted encoder \citep{martin2020camembert} and listing photographs \citep{glaeser2018computer, poursaeed2018vision} would complete the measurement of quality from unstructured listing content. The replication pipeline accompanying the paper, which reproduces every number reported here from the raw database, is intended to make these extensions straightforward. | |
modified
paper/sections/discussion.tex
+2 −2
@@ -8,7 +8,7 @@ | ||
| 8 | 8 | |
| 9 | 9 | Three findings survive every specification in this paper. First, broker-written descriptions contain price-relevant information beyond the structural fields. The 20 semantic similarities are jointly significant in the baseline, within language subsamples, within forward sortation areas and municipalities, with clustered standard errors, after correcting the lot-size measure, on the floor-area subsample, under every perturbation of the reference texts and with both encoders; and the information generalises out of sample, to random held-out listings and to held-out areas alike. Second, the magnitude of this information is smaller than the first version of the paper suggested and depends on what one controls for. The headline gain of six percentage points of $R^2$ over a structural model with no location variables shrinks to about two points once the language of the listing and area fixed effects are accounted for with the baseline encoder, and to about four points with a multilingual encoder; location explains far more of the variation in asking prices than text does. Third, the individual implicit price gradients are robust for a core of five dimensions---Luxury, Land \& Nature, Family-Friendly, Quiet \& Peaceful, Needs Renovation---and not for the others. |
| 10 | 10 | |
| 11 | −The interpretable dimensions tell a coherent story. Listings that use luxury vocabulary are priced higher conditional on structure, in every specification and with both encoders, with a gradient of roughly 4 to 14\% per standard deviation depending on the controls; the quantile regressions show the gradient rising with the price level, consistent with \citet{zietz2007determinants} and \citet{mcmillen2008changes}. Listings that describe a natural setting are priced higher; the gradient is large in the baseline (+13\%) and falls to +4\% once the lot area is measured correctly, which indicates that the ``large wooded lot'' reference was in part a proxy for lot size and in part a genuine amenity signal---a decomposition that structured data alone could not have provided. Using Quebec City transactions and site inspections, \citet{desrosiers2002landscaping} estimate landscaping premiums of roughly 4 to 12\%; the text-based gradient, net of lot area, lies at the lower end of that range. Listings that need renovation are priced lower, with a gradient of 4 to 8\%, the sign expected from the capitalisation of deferred maintenance \citep{nowak2017quality, goodwin2014broker}. Listings that market a family-friendly or quiet setting are priced lower, in every specification including those with area fixed effects, which means that these words mark lower-priced segments within areas as well as between them; the equilibrium-sorting logic of \citet{kuminoff2013new} and the evidence of \citet{delmelle2021language} that advertisement text predicts neighbourhood composition both point the same way: listing language is as much about where a property is as about what it is. | |
| 11 | +The interpretable dimensions tell a coherent story. Listings that use luxury vocabulary are priced higher conditional on structure, in every specification and with both encoders, with a gradient of roughly 4 to 14\% per standard deviation depending on the controls; the quantile regressions show the gradient rising with the price level, consistent with \citet{zietz2007determinants} and \citet{mcmillen2008changes}, and with the view of luxury finishes as positional goods whose marginal valuation rises with the quality of the surrounding bundle \citep{frank2007falling}. Listings that describe a natural setting are priced higher; the gradient is large in the baseline (+13\%) and falls to +4\% once the lot area is measured correctly, which indicates that the ``large wooded lot'' reference was in part a proxy for lot size and in part a genuine amenity signal---a decomposition that structured data alone could not have provided. Using Quebec City transactions and site inspections, \citet{desrosiers2002landscaping} estimate landscaping premiums of roughly 4 to 12\%; the text-based gradient, net of lot area, lies at the lower end of that range. Listings that need renovation are priced lower, with a gradient of 4 to 8\%, the sign expected from the capitalisation of deferred maintenance \citep{nowak2017quality, goodwin2014broker}. Listings that market a family-friendly or quiet setting are priced lower, in every specification including those with area fixed effects, which means that these words mark lower-priced segments within areas as well as between them; the equilibrium-sorting logic of \citet{kuminoff2013new} and the evidence of \citet{delmelle2021language} that advertisement text predicts neighbourhood composition both point the same way: listing language is as much about where a property is as about what it is. | |
| 12 | 12 | |
| 13 | 13 | \subsection{What the Evidence Does Not Establish} |
| 14 | 14 | |
@@ -36,7 +36,7 @@ For automated valuation, the results are encouraging but measured. Text features | ||
| 36 | 36 | |
| 37 | 37 | \textbf{Asking prices.} The data are asking prices from a single snapshot of active listings. List and sale prices differ systematically, and the gap varies with market conditions and marketing time \citep{haurin2010list}; if the gap varies with the semantic content of the description---plausibly larger for listings that disclose urgency---the coefficients on sale prices would differ from those reported here. The estimates describe the pricing behaviour of sellers and brokers conditional on the text they wrote. |
| 38 | 38 | |
| 39 | −\textbf{Identification.} The cross-sectional design cannot separate the information, persuasion and omitted-location channels of Section~\ref{sec:identification}, and the fixed effects absorb only between-area location. Text is written after the asking price is set and by the same agent who sets it, so the estimates are associations in a jointly determined pair of outcomes. Within-property variation across successive listings, or platform experiments on listing text, would be required for causal statements; \citet{alfano2022word} point toward designs that exploit variation in listing composition. | |
| 39 | +\textbf{Identification.} The cross-sectional design cannot separate the information, persuasion and omitted-location channels of Section~\ref{sec:identification}, and the fixed effects absorb only between-area location; a spatial autoregressive or Durbin specification \citep{lesage2009introduction, dube2014spatial} would be needed to model the within-area dependence that the clustered standard errors only accommodate. Text is written after the asking price is set and by the same agent who sets it, so the estimates are associations in a jointly determined pair of outcomes. Within-property variation across successive listings, or platform experiments on listing text, would be required for causal statements; \citet{alfano2022word} point toward designs that exploit variation in listing composition. | |
| 40 | 40 | |
| 41 | 41 | \textbf{Missing structural information.} The baseline lacks floor area and measures lot size with mixed units. The revision corrects the second and addresses the first on a 36\% subsample; on that subsample the semantic block's increment is about one percentage point of $R^2$, and a full-sample estimate with floor area would likely be of that order. |
| 42 | 42 | |
| 43 | 43 | |