spb/wp20_uqo

1% ============================================================================2% UQO Working Paper No. 20 — From Listing Images to Property Valuation3% ============================================================================4\documentclass[12pt,letterpaper]{article}56% --- Encoding & Language ---7\usepackage[utf8]{inputenc}8\usepackage[T1]{fontenc}9\usepackage[english]{babel}1011% --- Page Layout ---12\usepackage[letterpaper,top=1in,bottom=1in,left=1in,right=1in]{geometry}1314% --- Typography ---15\usepackage{mathptmx}16\usepackage{microtype}17\usepackage{setspace}18\onehalfspacing1920% --- Math ---21\usepackage{amsmath,amssymb,amsthm}22\theoremstyle{plain}23\newtheorem{proposition}{Proposition}24\theoremstyle{definition}25\newtheorem{assumption}{Assumption}26\newtheorem{definition}{Definition}27\theoremstyle{remark}28\newtheorem{remark}{Remark}2930% --- Tables ---31\usepackage{booktabs}32\usepackage{tabularx}33\usepackage{multirow}34\usepackage{threeparttable}35\usepackage{siunitx}36\sisetup{input-symbols = {()}, table-format = -1.3, parse-numbers = false}37\usepackage{array}38\usepackage{adjustbox}3940% --- Figures ---41\usepackage{graphicx}42\graphicspath{{figures/}}43\usepackage{subcaption}44\usepackage{float}45\usepackage{tikz}46\usetikzlibrary{arrows.meta,positioning,shapes.geometric,fit}4748% --- Colors & Links ---49\usepackage[dvipsnames]{xcolor}50\usepackage[colorlinks=true,linkcolor=NavyBlue,citecolor=NavyBlue,urlcolor=NavyBlue,breaklinks=true]{hyperref}5152% --- Bibliography ---53\usepackage[authoryear,round,semicolon]{natbib}54\bibliographystyle{aer}5556% --- Captions ---57\usepackage[font=small,labelfont=bf,labelsep=period,justification=justified,singlelinecheck=false]{caption}5859% --- Headers & Footers ---60\usepackage{fancyhdr}61\pagestyle{fancy}62\fancyhf{}63\renewcommand{\headrulewidth}{0pt}64\fancyfoot[C]{\thepage}6566% --- Section Formatting ---67\usepackage{titlesec}68\titleformat{\section}{\large\bfseries}{\thesection.}{0.5em}{}69\titleformat{\subsection}{\normalsize\bfseries}{\thesubsection.}{0.5em}{}70\titleformat{\subsubsection}{\normalsize\itshape}{\thesubsubsection.}{0.5em}{}7172% --- Appendix Support ---73\usepackage[toc,page]{appendix}7475% --- Misc ---76\usepackage{enumitem}77\setlist{nosep,leftmargin=*}78\usepackage{url}7980% --- Generated numbers (every figure quoted in the text) ---81\input{tables/numbers}8283% ============================================================================84% METADATA85% ============================================================================86\newcommand{\WPnumber}{20}87\newcommand{\WPtitle}{From Listing Images to Property Valuation: A Multimodal AI Framework for Automated Building Characterization and the Cost Approach}88\newcommand{\WPsubtitle}{Evidence from Québec single-family listings matched to the assessment roll}89\newcommand{\WPdate}{October 2026}90\newcommand{\WPversion}{1.0}91\newcommand{\WPabstract}{%92The cost approach to value requires a specification of the improvements---materials, finish grade, condition, renovations---that assessment rolls do not record and that appraisers obtain by inspection. We ask whether that specification can be extracted from the photographs and text of a for-sale listing by general-purpose vision--language models, and what a transparent cost indication built on it is worth. We formalise an extraction-to-valuation framework with five evidence classes (administrative, visual, proxy, text-claimed, latent), deterministic cross-image aggregation, text--image--record reconciliation, a fully sourced cost schedule and uncertainty propagation, and we apply it to \nSample{} Québec house listings matched within 15~m to their roll unit (\nImagesSample{} photographs, three extractor tiers, \nRefsVerified{} verified references). Visible exterior attributes are extracted with high agreement against broker fields (roof covering \accroofcovering\%, garage type \accgaragetypeattacheddetached\%, cladding \acccladdingprimarybrokersidings\%) but storeys agree with the roll only loosely ($\kappa=\kappastoreysvisible$), and the most cost-relevant interior evidence is simply not shown: within the first ten photographs, \shownBasement\% of listings show a basement and \shownBathroom\% a bathroom, a missing-not-at-random structure that no extractor can overcome. Relative to a roll-only baseline, image and text attributes raise the median cost indication and move its ratio to the asking price from \priceRatiosonnetBZeroB{} to \priceRatiosonnetBThreeB{} (median absolute deviation \priceMdapesonnetBZeroB\% to \priceMdapesonnetBThreeB\%), mainly through condition-adjusted depreciation; the gain is real but modest, and the level of the indication depends more on the published unit-cost anchor than on anything seen in the images. A calibration against assessor building values on five metropolitan areas transfers to a geographic hold-out and implies a hard cost near \$300 per square foot, consistent with 2026 Québec industry guides and well above a Montréal hard-cost band. We release the ontology, prompts, schedule and code, and we are explicit about what the data cannot establish: no expert labels exist, latent defects are unobservable, and the cost indication is a screening tool, not an appraisal.%93}94\newcommand{\WPkeywords}{cost approach, automated valuation, vision--language models, listing photographs, depreciation, assessment roll, Québec}95\newcommand{\WPjel}{R31, R32, C45, C55, D83}9697% --- Author ---98\newcommand{\WPauthor}{Simon-Pierre Boucher}99\newcommand{\WPaffiliation}{Département des sciences administratives\\ Université du Québec en Outaouais}100\newcommand{\WPemail}{simon-pierre.boucher@uqo.ca}101\newcommand{\WPaddress}{Gatineau -- Pavillon Alexandre-Taché\\ 283, boulevard Alexandre-Taché\\ Gatineau, Québec, Canada J9A 1L8}102103\hypersetup{pdftitle={From Listing Images to Property Valuation: A Multimodal AI Framework for Automated Building Characterization and the Cost Approach},pdfauthor={Simon-Pierre Boucher}}104105% ============================================================================106\begin{document}107108\input{sections/titlepage}109110\input{sections/introduction}111\input{sections/setting}112\input{sections/literature}113\input{sections/framework}114\input{sections/data}115\input{sections/methodology}116\input{sections/results_extraction}117\input{sections/results_cost}118\input{sections/robustness}119\input{sections/discussion}120\input{sections/limitations}121\input{sections/conclusion}122123\newpage124\bibliography{references}125126\input{appendix/appendix}127128\end{document}129