Literature map and novelty statement — UQO WP 26
Can Autonomous AI Agents Discover the Fundamental Value of Real Estate? Evidence from Quebec Transactions and Multi-Agent Deliberation
Phase 1 deliverable (literature review and novelty map). Compiled 2026-10-27. Every citation key below exists in paper/references.bib and was verified against Crossref (DOI), the arXiv API (identifier) or a live URL on 2026-10-27; the per-key log is in the final section and the machine-readable record is docs/_bib_verification.json. Entries that were verified but held in reserve to respect the 90–140 budget are listed separately; they can be restored by flipping included in the JSON and re-running the generator.
1. Hedonic pricing and automated valuation
The hedonic framework (rosen1974hedonic) remains the workhorse of residential valuation. Applied reviews (malpezzi2003hedonic, sirmans2005composition) show that specification choices — functional form, attribute sets, treatment of location and time — vary enormously across studies, and hill2013hedonic provides the standard taxonomy of hedonic index methods (time-dummy, imputation, characteristics). Repeat-sales indexes (bailey1963regression, case1987prices) are the main alternative for index construction and are directly relevant to our arm B and to the repeat-transaction triangulation of fundamental value.
Spatial dependence is first-order in housing data. pace1998spatial and pace1998spatiotemporal introduced spatial and spatiotemporal autoregressive structures to real estate; lesage2009introduction is the textbook treatment. For Quebec, the Laval group (Des Rosiers, Thériault, Dubé) supplies the closest hedonic antecedents: access and neighbourhood factors in Quebec City (desrosiers2000sorting), spatial difference-in-differences around Montreal transit investments (dube2014spatial), joint modelling of price and time-on-market (dube2016spatiotemporal), and residential reconversions (dube2023residential). None of these uses LLMs; all use assessment-roll or MLS-type transaction data similar in nature to Vrai-Prix.
On the AVM side, mullainathan2017machine frames prediction as a distinct econometric problem; kok2017big compare manual appraisal and automated valuation; mayer2019estimation test estimation and updating methods for hedonic valuation (Swiss data, including gradient boosting and mixed-effects models); perezrave2019machine propose a big-data ML approach for inferential and predictive purposes; bogin2020appraisal show that AVM and appraisal accuracy both degrade in rural (thin) markets — an explicit motivation for RQ3 and RQ5. Tree ensembles (chen2016xgboost, prokhorenkova2018catboost) remain the strongest tabular baselines (grinsztajn2022why), which is why arm C/D uses them. Text improves AVMs: nowak2017textual, shen2021information, zhang2024describe. These papers set the bar that any LLM arm must clear and define what "privileged access to data" means in RQ2 (listing text is information, not method).
2. Appraisal accuracy and bias
Appraisal is the institutional benchmark. quan1991price model appraisal as signal extraction with optimal shrinkage toward comparables; northcraft1987experts show anchoring on list price among professionals. cho1996empirical and nakamura2010how connect appraisal to redlining and to house-price uncertainty. The modern purchase-appraisal literature documents contract-price confirmation bias: calem2021appraising, eriksen2019influence, eriksen2020contract, and fout2022when (low appraisals pull contract prices down). agarwal2015collateral link collateral valuation to borrower constraints; agarwal2020regulation test whether regulation de-biases appraisers; shui2018appraisal (FHFA WP 18-01) compare AMC and non-AMC appraisals. Racial valuation gaps are documented by howell2018neighborhoods, howell2021increasing, perry2018devaluation, freddiemac2021racial, and algorithmic credit disparities by fuster2022predictably.
Implication for the paper: appraisals are anchored, asymmetric and sometimes discriminatory; transaction prices embed appraisal influence through the mortgage channel. This supports treating the sale price as a noisy proxy, the planned audit of severe over/undervaluation rates (RQ6), and the fairness stratification.
3. Information aggregation and deliberation
3.1 Classical results
- Averaging:
galton1907vox,surowiecki2004wisdom; forecast combinationbates1969combination,clemen1989combining,timmermann2006forecast; robust poolsranjan2010combining,satopaa2014combining(extremising/logit pooling),jose2014trimmed(trimmed opinion pools). - Correlated signals:
clemen1985limits(information value of dependent sources is bounded — the formal root of our common-model-risk propositions). Diversity vs ability:hong2004groups,lamberson2012optimal(optimal forecasting groups),davisstober2014when. - Social learning and cascades:
degroot1974reaching,golub2010naive(naive learning is wise only without dominant agents),acemoglu2011bayesian,bikhchandani1992theory,morris2002social(public signals crowd out private information). - Experiments:
lorenz2011how(social influence narrows diversity and undermines crowd wisdom),becker2017network(decentralised networks can improve accuracy),navajas2018aggregated(small deliberating groups beat large crowds),prelec2017solution(surprisingly-popular meta-knowledge).
These give the testable predictions behind RQ2–RQ3: deliberation helps when signals are heterogeneous and influence is decentralised; it hurts when agents share a model, when a public anchor (list price, assessment) dominates, or when updating is uniform.
3.2 LLM multi-agent debate and ensembles
- Foundational debate papers:
du2023improving,liang2024encouraging,chan2023chateval; safety-via-debateirving2018ai,khan2024debating,michael2023debate. - Frameworks:
li2023camel,wu2023autogen,hong2023metagpt,park2023generative; surveyguo2024large. - Sceptical/compute-matched evidence:
wang2022selfconsistency(majority voting over samples),smit2024should,wang2024rethinking,kaesberg2025voting,wu2025can(controlled study: reasoning strength and diversity dominate, majority pressure suppresses correction),estornell2024multi(theory: similar models converge to majority, including shared misconceptions; interventions = diversity pruning, quality pruning, misconception refutation),zhu2026demystifying(homogeneous agents + uniform updates cannot improve expected correctness). - Inference-compute scaling:
chen2024more(more calls can be non-monotone),li2024more,brown2024large,snell2024scaling. - Calibration:
kadavath2022language,tian2023just,xiong2024can. Faithfulness:turpin2023language,lanham2023measuring, with the definitional standard ofjacovi2020towards.
Implication: the LLM literature has largely converged on the same conclusion as the human literature — debate is not free information, it is a function of diversity and correlation. Our treatments F–J (independent agents, debate, blinded-skeptic debate, placebo roles, cost-matched single-agent and ensemble) are designed to separate these mechanisms, which no prior real-estate study has done.
4. LLMs as forecasters and valuers
General forecasting: halawi2024approaching, schoenegger2024wisdom (silicon crowd of 12 LLMs rivals human crowd), schoenegger2023large (single model lags). Real-estate-specific prior work is listed in the table below.
Closest prior work
| Paper (key) | Data | Method | What they test | Gap vs. ours |
|---|---|---|---|---|
Geerts, Reusens, Baesens, vanden Broucke, De Weerdt 2025, ECML-PKDD (geerts2025performance) |
Four international housing datasets (listing/transaction tabular data) | Single LLMs (GPT-4o-mini and others) with zero-shot, few-shot, market-report and hybrid prompts; LightGBM benchmark | Whether in-context examples chosen by hedonic and geographic similarity make LLM price estimates competitive; interval calibration | Single-agent only; no deliberation, no cost-matched ensembles, no leakage-audited as-of-date comparables, no Quebec/Canadian data; reports overconfident intervals, which we treat as a hypothesis to test under conformal and debate treatments |
Faxvaag, Aarsnes et al. 2025, Journal of Real Estate Research (faxvaag2025incorporating) |
9,842 Oslo transactions, 2022–23, with listing text | GPT models extract features from text → XGBoost AVM; hedonic OLS benchmark | Whether LLM-derived text features improve a tabular AVM | LLM used as feature extractor, not valuer; no agents, no uncertainty, no explanation audit |
Vedula, Dhyani et al. 2025, Findings of ACL (vedula2025quantile) |
Price-prediction tasks incl. housing | Fine-tuned LLM quantile regression | Whether LLMs can output calibrated quantiles for prices | Single fine-tuned model; no deliberation, no economic baselines or fundamental-value construct |
Chen & Si 2024, arXiv (chen2024predicting) |
Shanghai lane-house rentals | ChatGPT few-shot vs. ML regressors | Rental price prediction accuracy | Rents not sales; single model; no uncertainty or leakage controls |
Bae, Jung, Nam et al. 2025, Springer chapter (bae2025housing) |
Korean housing transactions | LLM-based appraisal with reasoning vs. hedonic model | Predictive power and explainability of a single LLM | Single agent; explanation quality assessed by plausibility, not faithfulness |
Li 2026, ERES conference (li2026large) |
Understanding Society (UK household panel), self-reported valuations 2010–24 | GPT-4o prompted with user-cost and hedonic logic; implied WTP vs. observed valuations | Whether a single LLM internalises user-cost theory | Closest in spirit to our fundamental-value question but uses owner self-valuations, one model, no transactions, no deliberation |
Schoenegger et al. 2024, Science Advances (schoenegger2024wisdom) |
Forecasting-tournament binary questions | Ensemble of 12 LLMs, no communication | Whether a silicon crowd matches a human crowd | Not housing; no debate treatment; motivates our arm F |
Estornell & Liu 2024, NeurIPS (estornell2024multi) |
Four reasoning benchmarks | Theory + interventions for multi-LLM debate | When debate converges to majority misconceptions | Not valuation; we import their diversity/misconception logic into a continuous-valued, economically grounded task |
Searches run (2026-10-27): Crossref, OpenAlex (works?search=), Semantic Scholar graph API, arXiv API and web search for "LLM house price", "GPT real estate valuation", "multi-agent real estate", "LLM-based AVM", "LLM hedonic", "LLM agents housing market". We found no paper that deploys role-specialised multi-agent LLM deliberation for residential valuation on transaction data, and no LLM valuation study on Quebec or Canadian data. Adjacent items found but not included: a 2026 HSE Business Informatics paper on commercial real estate with LLM-derived features (Moscow), the REAL benchmark of housing-transaction Q&A (arXiv 2507.03477), the HouseTS spatiotemporal dataset (arXiv 2506.00765), Karanikolas et al. 2025 (Information, multimodal AVM), and Cheung 2023 (JPIF, ChatGPT in valuation reports); these are verified and held in reserve.
5. Uncertainty, evaluation and model comparison
Scoring rules matheson1976scoring (CRPS), gneiting2007strictly, gneiting2014probabilistic; quantile loss koenker1978regression. Conformal prediction: vovk2005algorithmic, lei2018distribution, romano2019conformalized (CQR), angelopoulos2023gentle. Forecast comparison: diebold1995comparing, harvey1997testing, giacomini2006tests, hansen2011model (MCS). Dependence: kunsch1989jackknife, politis1994stationary, cameron2008bootstrap (cluster bootstrap). Multiplicity: holm1979simple. These support the pre-registered endpoints (log-error, MdAPE, tail error, coverage at 50/80/90/95 %, pinball/CRPS) and the clustered-bootstrap / DM / MCS inference plan.
6. Fundamental value versus transaction price
User cost and price-to-rent: poterba1984tax, himmelberg2005assessing, campbell2009what. Efficiency and expectations: case1989efficiency, glaeser2017extrapolative. Search, liquidity and bargaining: wheaton1990vacancy, han2015microstructure, genesove2001loss. Credit and collateral: stein1995prices, kiyotaki1997credit, iacoviello2005house, mian2011house. Institutional sources: statcan2024chsp (Canadian Housing Statistics Program), mamh2024manuel (Manuel d'évaluation foncière du Québec, the assessment-roll methodology that generates the municipal assessment feature). Together these justify three operational value definitions in the paper: (i) transaction-price prediction target; (ii) repeat-sale-adjusted value; (iii) user-cost/rent-implied structural value.
7. Common-model risk and fairness
kleinberg2021algorithmic (monoculture can lower welfare despite higher individual accuracy), bommasani2022picking (outcome homogenisation across deployers of a shared model), plus the appraisal-fairness items in §2. These frame the welfare section and the provider-balanced design (all-Claude, all-OpenAI, mixed committees).
7b. Mapping the literature to the research questions
| RQ | What the literature already establishes | What remains open and is tested here |
|---|---|---|
| RQ1 (agents vs. baselines) | Tree ensembles beat deep and linear models on tabular housing data (grinsztajn2022why, mayer2019estimation); single LLMs are competitive only with carefully chosen in-context comparables and remain behind LightGBM on most datasets (geerts2025performance) |
Whether a role-specialised committee with as-of-date comparables closes or reverses that gap on unseen Quebec transactions; no prior estimate exists |
| RQ2 (deliberation vs. compute, data, ensembling) | Majority voting captures much of the "debate" gain (wang2022selfconsistency); more calls can be non-monotone (chen2024more); similar models converge to shared errors (estornell2024multi, zhu2026demystifying) |
The decomposition of any gain into communication, complementary information, compute and ensembling in a continuous-valued economic task with cost-matched arms (treatments F, G, H, I, J) |
| RQ3 (role diversity, calibration, tail errors, thin markets) | Diversity beats ability under correlated errors (hong2004groups, clemen1985limits, davisstober2014when); social influence narrows estimate dispersion (lorenz2011how); LLM intervals are overconfident (geerts2025performance, xiong2024can); AVMs fail in rural markets (bogin2020appraisal) |
Whether blinded-skeptic debate and placebo-role controls change coverage at 50/80/90/95 % and P90/P95 errors, especially under spatial holdouts and distribution shift |
| RQ4 (faithful explanations) | Chain-of-thought explanations can be unfaithful (turpin2023language, lanham2023measuring); the faithfulness standard is defined by jacovi2020towards |
Provenance precision/recall against evidence IDs and numerical consistency of stated adjustments with the point estimate in a valuation dossier |
| RQ5 (heterogeneity by region, tier, type, cycle, liquidity) | Quebec hedonic evidence is concentrated in Quebec City and Montreal (desrosiers2000sorting, dube2014spatial, dube2023residential); time-on-market and price are jointly determined (dube2016spatiotemporal) |
Stratified performance across Montreal, Quebec City, Outaouais and other regions, condominium vs. detached vs. plex, price deciles and transaction density |
| RQ6 (economic usefulness, fairness, cost) | Appraisal bias has collateral consequences (calem2021appraising, fout2022when); valuation gaps by neighbourhood race are documented (howell2021increasing, freddiemac2021racial); monoculture can lower welfare (kleinberg2021algorithmic) |
Net value of agentic valuation after inference cost, latency and human review, with severe over/undervaluation rates and subgroup conditional coverage reported rather than assumed |
Two institutional notes shape the Quebec setting. Municipal assessment rolls follow the Manuel d'évaluation foncière du Québec (mamh2024manuel), so the assessment feature is a mass-appraisal estimate produced on a triennial cycle with a fixed reference date, not a market value at the sale date; it is therefore a legitimate as-of-date covariate but a biased proxy for value. Canadian housing statistics (statcan2024chsp) provide region-level context but no transaction-level ground truth, which is why the study depends on authorised Vrai-Prix transaction snapshots and must document their provenance and usage rights before any evaluation.
8. Novelty statement (candid)
What is not new:
- Using an LLM to estimate house prices (Geerts et al. 2025; Bae et al. 2025; Chen & Si 2024; Vedula et al. 2025) or to extract features for an AVM (Faxvaag et al. 2025).
- Multi-agent debate as a method, its theoretical critique (Estornell & Liu 2024; Zhu et al. 2026) and the finding that debate often fails to beat voting at matched cost (Smit et al. 2024; Wang et al. 2024; Kaesberg et al. 2025; Wu et al. 2025).
- The classical theory of aggregation under correlated signals (Clemen & Winkler 1985; Hong & Page 2004; Lorenz et al. 2011).
- Hedonic and spatial evidence on Quebec housing (Des Rosiers, Thériault, Dubé).
- Asking whether an LLM embodies user-cost logic (Li 2026, ERES).
What is new, conditional on execution:
- A role-specialised 12-agent (5–20) deliberation architecture for residential valuation evaluated on arm's-length Quebec transactions with as-of-date comparable retrieval and a leakage-audit suite.
- A pre-registered, information-matched, compute-matched and provider-balanced comparison against hedonic, repeat-sales, gradient-boosting, calibrated AVM, single-LLM, no-communication ensemble, placebo-role and cost-matched single-agent arms — the design needed to attribute any gain to communication rather than compute, correlation or data.
- Joint evaluation of point accuracy, calibrated intervals (including conformal post-processing), tail errors relevant to collateral risk, and explanation faithfulness (evidence provenance and numerical consistency of stated adjustments), rather than any one of these.
- A formal link between the correlated-signal aggregation literature and LLM debate, yielding falsifiable propositions about diversity, correlation, communication topology and aggregation weights in a continuous-valued economic task.
- Separation of transaction-price prediction from fundamental-value identification, with repeat-sale, rent and user-cost triangulation on Canadian data.
Priority claims we explicitly do not make: first LLM valuation study; first multi-agent debate study; first Quebec hedonic study; superiority of agents over humans (the human arm is pending authorisation and is not assumed).
9. Could not verify / not included
| Item requested | Outcome |
|---|---|
| "Kahn/Ohls" (appraisal accuracy) | No paper by these authors on appraisal accuracy or bias could be located via Crossref, OpenAlex or web search. Omitted. If the intended reference is known (title or year), it can be added. |
| Calem, Lambie-Hanson & Nakamura "appraisal bias" | Resolved to the published version calem2021appraising (Calem, Kenney, Lambie-Hanson & Nakamura, Real Estate Economics 49(S1), 2021). |
| Fout/Mota/Rosenblatt "FHFA appraisal gaps" | The paper is a Fannie Mae working paper published as fout2022when (JREFE 65(3), 2022), not an FHFA paper. The FHFA appraisal-gap paper is Grodzicki et al. (2024, WP 24-06), verified and held in reserve as grodzicki2024home. |
| Shui & Murthy 2019 (AMC appraisals) | Exists only as FHFA Staff Working Paper 18-01 (2018); no journal version found. Included as shui2018appraisal (URL-verified). |
| Chen et al. 2024/2025 "compute-matched ensembles" | Interpreted as Chen, Davis, Hanin, Bailis, Stoica, Zaharia & Zou 2024 (chen2024more). No 2025 follow-up with a distinct title was found; the OpenReview version is retitled "Towards the Scaling Properties of Compound AI Systems". |
| Perry et al. (Brookings) 2021 update | Only the 2018 report could be URL-verified (perry2018devaluation). |
| Case & Shiller 1989 (AER) | Not in Crossref; verified via the JSTOR stable URL (article metadata entered manually). |
| Holm 1979 | No DOI exists; verified via JSTOR stable URL. |
| Surowiecki 2004, Cochrane 2005 (books) | No DOI; verified via publisher/Google Books pages. |
| Semantic Scholar bulk sweep | The API rate-limited the second batch of queries; coverage was completed with OpenAlex, arXiv and web search instead. |
Template entries (fama1993common, bollerslev1986generalized, newey1987simple, white1980heteroskedasticity, cochrane2005asset, petersen2009estimating) |
Carried over from the working-paper template because the placeholder introduction/methodology sections still cite them; all six were verified. Remove them once those sections are rewritten. |
10. Verified entries held in reserve (not in the .bib)
Seventy-four further entries were verified (see JSON, included: false) and can be restored if needed. Highlights by theme: hedonic identification (lancaster1966new, ekeland2004identification, kuminoff2010which), spatial/mass-appraisal (fik2003modeling, bourassa2010predicting, hill2018geospatial, schulz2014automated, mccluskey2013prediction, eljaouhari2024mapping, geerts2023survey), images (poursaeed2018vision, law2019take), appraisal (lacourlittle2003appraisal, diaz1997how, tzioumis2017appraisers, mayer2022appraisal, eriksen2024attribute, shi2015appraisal, ding2016impact, grodzicki2024home), crowds (banerjee1992simple, winkler1981combining, larrick2006intuitions, mannes2014wisdom, budescu2015identifying, koriat2012when, bahrami2010optimally, mellers2014psychological, ottaviani2006strategy, stock2004combination, smith2009simple, kerr2004group, hastie2005robust), LLM agents (tran2025multi, cemri2025why, zhang2024exploring, huang2024large, sharma2023towards, lin2022teaching, wei2022chain, zheng2023judging, gruver2023large, hegselmann2023tabllm, kenton2024scalable), LLMs in economics (lopezlira2023can, kim2024financial, horton2023large, ludwig2024machine, korinek2023generative), real-estate LLM adjacent (karanikolas2025artificial, cheung2023unleashing, zhu2025real, wang2025housets), uncertainty (barber2023conformal, tibshirani2019conformal, diebold2015comparing, benjamini1995controlling, kuleshov2018accurate, guo2017calibration), housing (glaeser2015housing, case1990forecasting, mian2009consequences, desrosiers2001size, theriault2003modelling, dube2013spatiotemporal, hossain2009determinants, cmhc2024outlook), fairness/monoculture (toups2023ecosystem, creel2022algorithmic, rudin2019stop, bartlett2022consumer), ML (breiman2001random, wolpert1992stacked, athey2019machine).
11. Verification log (included entries)
| Key | DOI / arXiv / URL | Verification method | Date |
|---|---|---|---|
acemoglu2011bayesian |
doi:10.1093/restud/rdr004 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
agarwal2015collateral |
doi:10.1287/mnsc.2014.2002 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
agarwal2020regulation |
doi:10.1016/j.jfi.2019.04.003 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
angelopoulos2023gentle |
doi:10.1561/2200000101 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
bae2025housing |
doi:10.1007/978-981-96-5833-6_2 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
bailey1963regression |
doi:10.1080/01621459.1963.10480679 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
bates1969combination |
doi:10.1057/jors.1969.103 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
becker2017network |
doi:10.1073/pnas.1615978114 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
bikhchandani1992theory |
doi:10.1086/261849 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
bogin2020appraisal |
doi:10.1007/s11146-019-09712-0 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
bollerslev1986generalized |
doi:10.1016/0304-4076(86)90063-1 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
bommasani2022picking |
arXiv:2211.13972 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
brown2024large |
arXiv:2407.21787 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
calem2021appraising |
doi:10.1111/1540-6229.12326 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
cameron2008bootstrap |
doi:10.1162/rest.90.3.414 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
campbell2009what |
doi:10.1016/j.jue.2009.06.002 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
case1987prices |
https://www.bostonfed.org/publications/new-england-economic-review/1987-issues/issue-september-october-1987/prices-of-single-family-homes-since-1970-new-indexes-for-four-cities.aspx | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
case1989efficiency |
https://www.jstor.org/stable/1804778 | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
chan2023chateval |
arXiv:2308.07201 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
chen2016xgboost |
doi:10.1145/2939672.2939785 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
chen2024more |
arXiv:2403.02419 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
chen2024predicting |
arXiv:2405.17505 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
cho1996empirical |
doi:10.1007/BF00174550 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
clemen1985limits |
doi:10.1287/opre.33.2.427 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
clemen1989combining |
doi:10.1016/0169-2070(89)90012-5 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
cochrane2005asset |
https://books.google.com/books?vid=ISBN9780691121376 | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
davisstober2014when |
doi:10.1037/dec0000004 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
degroot1974reaching |
doi:10.1080/01621459.1974.10480137 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
desrosiers2000sorting |
doi:10.1108/14635780010338245 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record (located via Crossref bibliographic query, top hit checked) |
2026-10-27 |
diebold1995comparing |
doi:10.1080/07350015.1995.10524599 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
du2023improving |
arXiv:2305.14325 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
dube2014spatial |
doi:10.1016/j.trb.2014.02.007 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
dube2016spatiotemporal |
doi:10.1111/1540-6229.12121 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
dube2023residential |
doi:10.1007/s10901-023-10041-1 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
eriksen2019influence |
doi:10.1016/j.jue.2019.04.007 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
eriksen2020contract |
doi:10.1007/s11146-019-09716-w | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record (located via Crossref bibliographic query, top hit checked) |
2026-10-27 |
estornell2024multi |
https://proceedings.neurips.cc/paper_files/paper/2024/hash/32e07a110c6c6acf1afbf2bf82b614ad-Abstract-Conference.html | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
fama1993common |
doi:10.1016/0304-405X(93)90023-5 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
faxvaag2025incorporating |
doi:10.1080/08965803.2025.2587313 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
fout2022when |
doi:10.1007/s11146-020-09800-6 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
freddiemac2021racial |
https://www.freddiemac.com/research/insight/20210920-home-appraisals | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
fuster2022predictably |
doi:10.1111/jofi.13090 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
galton1907vox |
doi:10.1038/075450a0 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
geerts2025performance |
arXiv:2506.11812 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
genesove2001loss |
doi:10.1162/003355301753265561 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
giacomini2006tests |
doi:10.1111/j.1468-0262.2006.00718.x | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
glaeser2017extrapolative |
doi:10.1016/j.jfineco.2017.06.012 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
gneiting2007strictly |
doi:10.1198/016214506000001437 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
gneiting2014probabilistic |
doi:10.1146/annurev-statistics-062713-085831 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
golub2010naive |
doi:10.1257/mic.2.1.112 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
grinsztajn2022why |
arXiv:2207.08815 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
guo2024large |
doi:10.24963/ijcai.2024/890 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
halawi2024approaching |
arXiv:2402.18563 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
han2015microstructure |
doi:10.1016/B978-0-444-59531-7.00013-2 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
hansen2011model |
doi:10.3982/ECTA5771 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
harvey1997testing |
doi:10.1016/S0169-2070(96)00719-4 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
hill2013hedonic |
doi:10.1111/j.1467-6419.2012.00731.x | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
himmelberg2005assessing |
doi:10.1257/089533005775196769 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
holm1979simple |
https://www.jstor.org/stable/4615733 | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
hong2004groups |
doi:10.1073/pnas.0403723101 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
hong2023metagpt |
arXiv:2308.00352 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
howell2018neighborhoods |
doi:10.1177/2332649218755178 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
howell2021increasing |
doi:10.1093/socpro/spaa033 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
iacoviello2005house |
doi:10.1257/0002828054201477 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
irving2018ai |
arXiv:1805.00899 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
jacovi2020towards |
doi:10.18653/v1/2020.acl-main.386 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
jose2014trimmed |
doi:10.1287/mnsc.2013.1781 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
kadavath2022language |
arXiv:2207.05221 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
kaesberg2025voting |
arXiv:2502.19130 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
khan2024debating |
arXiv:2402.06782 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
kiyotaki1997credit |
doi:10.1086/262072 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
kleinberg2021algorithmic |
doi:10.1073/pnas.2018340118 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
koenker1978regression |
doi:10.2307/1913643 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
kok2017big |
doi:10.3905/jpm.2017.43.6.202 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
kunsch1989jackknife |
doi:10.1214/aos/1176347265 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
lamberson2012optimal |
doi:10.1287/mnsc.1110.1441 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
lanham2023measuring |
arXiv:2307.13702 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
lei2018distribution |
doi:10.1080/01621459.2017.1307116 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
lesage2009introduction |
doi:10.1201/9781420064254 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record (located via Crossref bibliographic query, top hit checked) |
2026-10-27 |
li2023camel |
arXiv:2303.17760 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
li2024more |
arXiv:2402.05120 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
li2026large |
doi:10.15396/eres2026_013 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
liang2024encouraging |
doi:10.18653/v1/2024.emnlp-main.992 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
lorenz2011how |
doi:10.1073/pnas.1008636108 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
malpezzi2003hedonic |
doi:10.1002/9780470690680.ch5 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
mamh2024manuel |
https://www.mamh.gouv.qc.ca/evaluation-fonciere/manuel-devaluation-fonciere-du-quebec/ | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
matheson1976scoring |
doi:10.1287/mnsc.22.10.1087 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
mayer2019estimation |
doi:10.1108/JERER-08-2018-0035 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
mian2011house |
doi:10.1257/aer.101.5.2132 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
michael2023debate |
arXiv:2311.08702 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
morris2002social |
doi:10.1257/000282802762024610 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
mullainathan2017machine |
doi:10.1257/jep.31.2.87 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
nakamura2010how |
https://www.philadelphiafed.org/-/media/frbp/assets/economy/articles/business-review/2010/q3/brq310_how-much-is-that-home-really-worth.pdf | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
navajas2018aggregated |
doi:10.1038/s41562-017-0273-4 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
newey1987simple |
doi:10.2307/1913610 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
northcraft1987experts |
doi:10.1016/0749-5978(87)90046-X | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
nowak2017textual |
doi:10.1002/jae.2550 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
pace1998spatial |
doi:10.1023/A:1007783811760 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
pace1998spatiotemporal |
doi:10.1023/A:1007799028599 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
park2023generative |
doi:10.1145/3586183.3606763 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
perezrave2019machine |
doi:10.1080/09599916.2019.1587489 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
perry2018devaluation |
https://www.brookings.edu/articles/devaluation-of-assets-in-black-neighborhoods/ | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
petersen2009estimating |
doi:10.1093/rfs/hhn053 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
politis1994stationary |
doi:10.1080/01621459.1994.10476870 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
poterba1984tax |
doi:10.2307/1883123 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
prelec2017solution |
doi:10.1038/nature21054 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
prokhorenkova2018catboost |
arXiv:1706.09516 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
quan1991price |
doi:10.1007/bf00173120 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record (located via Crossref bibliographic query, top hit checked) |
2026-10-27 |
ranjan2010combining |
doi:10.1111/j.1467-9868.2009.00726.x | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
romano2019conformalized |
arXiv:1905.03222 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
rosen1974hedonic |
doi:10.1086/260169 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
satopaa2014combining |
doi:10.1016/j.ijforecast.2013.09.009 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
schoenegger2023large |
arXiv:2310.13014 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
schoenegger2024wisdom |
doi:10.1126/sciadv.adp1528 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
shen2021information |
doi:10.1016/j.jue.2020.103299 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
shui2018appraisal |
https://www.fhfa.gov/research/papers/wp1801 | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
sirmans2005composition |
doi:10.1080/10835547.2005.12090154 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record (located via Crossref bibliographic query, top hit checked) |
2026-10-27 |
smit2024should |
arXiv:2311.17371 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
snell2024scaling |
arXiv:2408.03314 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
statcan2024chsp |
https://www.statcan.gc.ca/en/subjects-start/housing | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
stein1995prices |
doi:10.2307/2118444 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
surowiecki2004wisdom |
https://www.penguinrandomhouse.com/books/175380/the-wisdom-of-crowds-by-james-surowiecki/ | HTTP GET returned 200; bibliographic fields entered manually | 2026-10-27 |
tian2023just |
doi:10.18653/v1/2023.emnlp-main.330 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
timmermann2006forecast |
doi:10.1016/S1574-0706(05)01004-9 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
turpin2023language |
arXiv:2305.04388 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
vedula2025quantile |
arXiv:2506.06657 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
vovk2005algorithmic |
doi:10.1007/b106715 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
wang2022selfconsistency |
arXiv:2203.11171 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
wang2024rethinking |
arXiv:2402.18272 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
wheaton1990vacancy |
doi:10.1086/261734 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
white1980heteroskedasticity |
doi:10.2307/1912934 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
wu2023autogen |
arXiv:2308.08155 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
wu2025can |
arXiv:2511.07784 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
xiong2024can |
arXiv:2306.13063 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |
zhang2024describe |
doi:10.1017/S1351324923000360 | Crossref API works/<doi> resolved; title/authors/venue/year taken from the record |
2026-10-27 |
zhu2026demystifying |
arXiv:2601.19921 | arXiv API query?id_list= resolved; title/authors/year taken from the record |
2026-10-27 |