TeX 55.1%
Python 44.2%
1# Novelty scan: close predecessors of an adaptive, hierarchical, state-space house price index23Scope: what each close predecessor does, with the citation as verified on 2026-10-21 (keys refer to4`paper/references.bib`; metadata in `docs/literature/references_verified.json`), and what an adaptive5hierarchical state-space index (the SHARP-HPI candidate in `CLAUDE.md` §6–7) would have to add to be6distinct. Descriptions are based on the published abstracts retrieved from Crossref / Semantic Scholar /7RePEc; where an abstract could not be retrieved this is stated and only title-level information is used.8No claim of novelty is made here; the purpose is to delimit what already exists.910---1112## 1. Rolling-time-dummy (RTD) hedonic indexes1314**Citations.** Hill, Scholz, Shimizu & Steurer (2022), *Journal of Official Statistics* 38(1), 127–151,15DOI 10.2478/jos-2022-0007 (`hill2022rtd`); Hill, Scholz, Shimizu & Steurer (2018), *Economie et16Statistique* 500-501-502, 221–238, DOI 10.24187/ecostat.2018.500t.1953 (`hill2018eval`); the method is17also described in the Eurostat et al. (2013) handbook (`eurostat2013handbook`).1819**What it does.** A hedonic regression with time dummies is re-estimated on a rolling window of the20most recent periods; the index movement for the latest period is taken from the newest window and21chained (linked) onto the published series, so published values are never revised. The 2022 JOS paper22(abstract) addresses three issues: a method for choosing the window length, variants of the linking23rule and how to pick among them, and three modifications to make RTD robust to periods of low24transaction volume. RTD "requires less data and is more adaptable than other hedonic methods", which25is why several European statistical offices use it for official HPIs (Hill et al. 2018 evaluate the26hedonic methods used by European NSIs).2728**What an adaptive hierarchical state-space index would need to add.** RTD pools information over29time only through a fixed (or one-off optimised) window and treats each geography separately; the30hedonic coefficients are not shrunk across areas, there is no explicit model of the latent price level,31and uncertainty is the regression standard error of the time dummies. To differ, the new estimator must32(i) make the degree of temporal and cross-sectional pooling cell-specific and data-driven (estimated33variance components rather than a chosen window), (ii) share hedonic and movement information across34a spatial hierarchy with estimated shrinkage weights, (iii) provide filtered (real-time) and smoothed35(retrospective) estimates with model-based intervals, and (iv) treat revision explicitly rather than by36construction (RTD avoids revision by never updating). Any claim must be benchmarked against RTD with37the 2022 window-length and low-volume modifications, not against a naive fixed window.3839## 2. HMTS (hedonic multilateral time-series re-estimation with splicing)4041**Citation.** Ishaak, Ouwehand & Remøy (2024), "Constructing Limited-Revisable and Stable CPPIs for42Small Domains", *Journal of Official Statistics* 40(3), 380–408, DOI 10.1177/0282423X24124661743(`ishaak2024hmts`). This is the paper the official Statistics Netherlands R package `REPS` cites as the44source of HMTS; an earlier four-author (with de Haan, 2023) version is mentioned in the package manual45but could not be located bibliographically (see `unverified.md`).4647**What it does.** From the abstract: for commercial property indices in small domains, where common48methods give volatile series and attempts to reduce volatility cause frequent revisions of the whole49series, the authors introduce "a combination of hedonic imputation, multilateral calculations, time50series analysis, and window splicing", producing "stable and limited-revisable indices with the ability51to detect turning points in an early stage", tested on Dutch commercial transactions. The `REPS`52implementation (CRAN manual) makes the steps explicit: average imputed prices are computed for every53period against every possible base period; each base-period series is re-estimated with a state-space54(Kalman-filter) model; the series are converted to indices; a window of index figures is spliced; and55the geometric average over the remaining figures gives the final index.5657**What an adaptive hierarchical state-space index would need to add.** HMTS already contains a58state-space smoothing step and already targets small domains, so the overlap is substantial. The59differences available are: (i) in HMTS the Kalman filter is applied series-by-series to imputed60averages after a hedonic model estimated separately, whereas a joint model would estimate the hedonic61quality function, the latent movements and their variances in one likelihood with the small-domain62movements shrunk toward higher-level movements; (ii) HMTS has no cross-domain hierarchy (each domain is63smoothed on its own) and no robust (heavy-tailed) observation model; (iii) HMTS's stability comes from64averaging over base periods and splicing a fixed window, not from an estimated signal-to-noise ratio;65(iv) HMTS's uncertainty is not model-based. Because HMTS is already in official use (and the mandated66benchmark G in `CLAUDE.md` §5), the empirical comparison must use the `REPS`-style implementation with67its re-estimation and splicing options switched on.6869## 3. Structural time-series / hierarchical repeat-sales for thin markets (Francke and co-authors)7071**Citations.** Francke (2010), "Repeat Sales Index for Thin Markets", *JREFE* 41(1), 24–52, DOI7210.1007/s11146-009-9203-1 (`francke2010thin`); Francke & de Vos (2000), "Efficient Computation of73Hierarchical Trends", *JBES* 18(1), 51–57 (`franckedevos2000`); Francke & van de Minne (2017), "The74Hierarchical Repeat Sales Model for Granular Commercial Real Estate and Residential Price Indices",75*JREFE* 55(4), 511–532, DOI 10.1007/s11146-017-9632-1 (`francke2017hierarchical`); Francke, Rolheiser &76van de Minne (2025), "Estimating Census Tract House Price Indexes: A New Spatial Dynamic Factor77Approach", *JREFE* 70(3), 483–514, DOI 10.1007/s11146-023-09957-w (`francke2025censustract`); van de78Minne, Francke, Geltner & White (2020), *JREFE* 60(4), 514–553 (`vandeminne2020revisions`); Schwann79(1998), *JREFE* 16(3), 269–287 (`schwann1998thin`) and Goetzmann (1992), *JREFE* 5(1), 5–5380(`goetzmann1992`) are the earlier thin-market / random-walk antecedents.8182**What they do.** Francke (2010), abstract: replaces the time dummies of the Bailey–Muth–Nourse83repeat-sales model (and Goetzmann's random walk with drift) by a *local linear trend* structural84time-series model in which both level and drift vary over time, adds the reciprocal of the time between85sales as a regressor, and obtains index estimates even for periods with no observations; it is86explicitly designed for thin markets. Francke & van de Minne (2017), abstract: a repeat-sales model87with "multiple stochastic log price trends having a hierarchical additive structure: one common log88price trend and cluster specific log price trends in deviation from the common trend", with89*t*-distributed errors to handle outliers; the estimated degrees of freedom are small, rejecting90normality; applied to Philadelphia/Baltimore commercial property and part of Amsterdam, giving91"reliable indices on a very detailed level based on a small number of observations". Francke, Rolheiser92& van de Minne (2025), abstract: augments the repeat-sales model with a spatial dynamic factor model in93which loadings on latent trends follow a spatial random walk, producing quarterly census-tract indexes94that are less noisy than non-spatial ones and that replicate standard repeat-sales indexes where pairs95are plentiful. van de Minne et al. (2020) use revisions as a quality criterion for repeat-sales indices.9697**What an adaptive hierarchical state-space index would need to add.** This strand already has (a)98state-space latent price levels, (b) an additive common-plus-deviation hierarchy, (c) heavy-tailed99errors, (d) spatial sharing of latent trends, and (e) revision-based evaluation. It is the closest100predecessor and must be cited as such. The remaining space is: (i) the observation equation is101repeat-sales (pairs), which in Quebec municipalities with few resales discards most single-sale102transactions and inherits repeat-sales selection issues (Gatzlaff & Haurin 1997, 1998; Clapp,103Giaccotto & Tirtiroglu 1991); a hedonic (single-sale) observation equation with the same hierarchical104trend structure would be a different estimator, not a new idea in itself (hybrid models go back to Case105& Quigley 1991 and Quigley 1995); (ii) the pooling strength in these models is governed by estimated106variance hyper-parameters that are constant over time and across clusters — an *adaptive* scheme in107which the local-versus-pooled weight responds to a cell-level reliability measure (effective sample108size, covariate overlap, residual spatial dependence) would need to be shown to improve on constant109hyper-parameters, in simulation and out of time; (iii) a multi-level (province / region / MRC /110municipality) hierarchy rather than one common trend plus one deviation level; (iv) coherence of111aggregates (see §7). Without (ii)–(iv) demonstrated empirically, the proposal is a hedonic variant of112Francke & van de Minne (2017).113114## 4. Timeliness from listings (Anenberg & Laufer)115116**Citation.** Anenberg & Laufer (2017), "A More Timely House Price Index", *Review of Economics and117Statistics* 99(4), 722–734, DOI 10.1162/rest_a_00634 (`anenberg2017timely`). Related: Hill, Pfeifer,118Steurer & Trojanek (2024), *RIW* 70(2), 320–344 (`hill2024warning`) on stale new-build transaction119prices and the use of preliminary agreements; Sayag, Ben-hur & Pfeffermann (2022), *IJF* 38(1), 253–266120(`sayag2022nowcasts`).121122**What it does.** From the RePEc abstract: using listings data, the authors build a repeat-sales index123that dates house values at the *contract* date (when the price is determined) rather than at the124closing date, which helps explain short-run serial correlation and weak correlation with stock prices125and news; a version using only recent listings reveals trends several months before existing indexes126such as Case–Shiller are published. Sayag et al. (2022), abstract: provisional HPIs are revised heavily127because late-reported transactions differ from those reported on time; they nowcast, at sub-district128level, the characteristics, average price and number of late-reported transactions and feed them into129an extended hedonic model, cutting revision magnitude by more than 50% (Israel).130131**What an adaptive hierarchical state-space index would need to add.** These contributions are about132*data timing* (contract vs. deed date; late registrations), not about the estimator's pooling133structure. A Quebec index built on registry deeds faces exactly the registration-lag problem; an134explicit real-time vintage design (`CLAUDE.md` §11) and a late-registration model are therefore135required to compare with Sayag et al., and the filtered state-space estimate is a nowcast that should136be evaluated against a nowcast-augmented hedonic index, not only against the retrospective series.137Nothing in a state-space hierarchy by itself addresses the contract-vs-closing timing issue.138139## 5. High-frequency indices with time-series modelling (Bollerslev, Patton & Wang)140141**Citation.** Bollerslev, Patton & Wang (2016), "Daily House Price Indices: Construction, Modeling, and142Longer-run Predictions", *Journal of Applied Econometrics* 31(6), 1005–1025, DOI 10.1002/jae.2471143(`bollerslev2016daily`). Related: Bokhari & Geltner (2012), *JREFE* 45(2), 522–543144(`bokharigeltner2012`); Francke & van de Minne (2022), *REE* 50(5), 1257–1281 (`francke2022daily`).145146**What it does.** From the abstract: daily repeat-sales indices for 10 US metropolitan areas, built147with a method that "closely mimics" the monthly Case–Shiller methodology; the daily returns show mild148autocorrelation and strong conditional heteroskedasticity; a multivariate time-series model for the149daily returns "explicitly allowing for commonalities across cities and GARCH effects" forecasts150longer-run monthly changes better than lower-frequency alternatives.151152**What an adaptive hierarchical state-space index would need to add.** Bollerslev et al. model the153*index returns after construction* (a two-step approach) with cross-city commonality and GARCH; the154index itself remains a standard repeat-sales estimate on large metropolitan samples. A joint estimation155of the index and its dynamics in sparse municipalities is a different setting; the relevant lesson to156carry over is that cross-area commonality in returns is empirically strong, and that forecasting157performance must be tested with proper out-of-time designs. Daily frequency is not at issue for158Quebec (monthly target).159160## 6. Fay–Herriot and small-area estimation (SAE)161162**Citations.** Fay & Herriot (1979), *JASA* 74(366), 269–277, DOI 10.1080/01621459.1979.10482505163(`fayherriot1979`); Datta & Lahiri (2000), *Statistica Sinica* 10(2), 613–627 (`dattalahiri2000`);164Datta, Rao & Smith (2005), *Biometrika* 92(1), 183–196 (`datta2005variability`); Pfeffermann (2013),165*Statistical Science* 28(1), 40–68 (`pfeffermann2013sae`); Rao & Molina (2015), *Small Area Estimation*,1662nd ed., Wiley (`raomolina2015`). Empirical-Bayes shrinkage roots: Efron & Morris (1973, 1975)167(`efronmorris1973`, `efronmorris1975`); hierarchical modelling: Gelman & Hill (2007), Gelman (2006).168Housing applications at small spatial scale: Ren, Fox & Bruce (2017), *Annals of Applied Statistics*16911(2), 808–839, DOI 10.1214/17-AOAS1019 (`renfoxbruce2017`); Lopez & Hewings (2018), *RSUE* 70, 57–71170(`lopezhewings2018`); Bogin, Doerner & Larson (2019), *REE* 47(2), 365–398 (`bogin2019local`); Contat &171Larson (2024), *REE* 52(6), 1551–1583 (`contatlarson2024`); Ahlfeldt, Heblich & Seidel (2023), *RSUE*17298, 103836 (`ahlfeldt2023micro`; abstract not retrieved, title-level only); Guntermann, Liu & Nowak173(2016), *JRER* 38(1), 93–128 (`guntermann2016thin`; abstract not retrieved).174175**What they do.** The Fay–Herriot area-level model combines a direct (design-based) estimate for each176small area with a regression synthetic estimate, shrinking the direct estimate toward the synthetic one177with a weight that depends on the ratio of sampling variance to model variance; Datta & Lahiri (2000)178derive a second-order MSE approximation and estimator for EBLUPs under REML/ML, and Datta, Rao & Smith179(2005) treat MSE under the basic area-level model; Pfeffermann (2013) reviews design-based, frequentist180and Bayesian SAE developments including time-series and spatial extensions. In housing: Ren, Fox &181Bruce (2017), abstract: a Bayesian nonparametric latent-factor model that clusters census tracts with182correlated valuation dynamics to overcome sparse sales, yielding monthly tract-level indices for183Seattle; Lopez & Hewings (2018), abstract: median-based HPIs are misleading for small spatial units, and184they propose a matching-based Fisher HPI that controls for characteristics and is not restricted to185repeat sales (Illinois); Bogin, Doerner & Larson (2019), abstract: the first public set of constant-186quality (repeat-sales) indices for US counties, ZIP codes and census tracts, annual, 40 years; Contat &187Larson (2024), abstract: an algorithm making granular repeat-sales indices feasible with low transaction188counts, a balanced panel of 63,084 tract-level annual indices, and city indices built from tract189aggregates that are robust to heterogeneous submarket appreciation and non-random sampling.190191**What an adaptive hierarchical state-space index would need to add.** The SAE literature supplies the192formal notion of *borrowing strength with an estimated shrinkage weight and a valid MSE*, which is193exactly the "adaptive pooling" idea in `CLAUDE.md` §7; so adaptive shrinkage per se is not new. What is194not standard is a *dynamic, multi-level* Fay–Herriot structure for index movements in which the195direct estimate is a cell-level hedonic movement, the synthetic estimate is the parent-level movement,196and the weights evolve over time through a state-space model with spatial neighbours — i.e. the197combination of SAE (time-constant, area-level) with structural time-series (dynamic, single level) and198a geographic hierarchy. Ren et al. (2017) already do data-driven clustering of tracts with latent199factors in a Bayesian model, and Francke et al. (2025) already do spatially varying loadings; the new200estimator would have to show, with MSE and interval-coverage evidence as SAE practice requires, that201hierarchy-based pooling with an observable reliability score outperforms (or is at least comparable202to) these latent-factor approaches in Quebec-like sparsity. Reporting direct vs. model-assisted cells,203as in SAE practice, is a requirement not a contribution.204205## 7. Forecast reconciliation / coherent hierarchies206207**Citations.** Hyndman, Ahmed, Athanasopoulos & Shang (2011), *CSDA* 55(9), 2579–2589208(`hyndman2011hierarchical`); Wickramasuriya, Athanasopoulos & Hyndman (2019), *JASA* 114(526), 804–819,209DOI 10.1080/01621459.2018.1448825 (`wickramasuriya2019mint`); Athanasopoulos, Hyndman, Kourentzes &210Panagiotelis (2024), "Forecast reconciliation: A review", *IJF* 40(2), 430–456211(`athanasopoulos2024reconciliation`).212213**What they do.** From the abstracts: hierarchical time series are collections that aggregate across214products or geographies; "coherent" forecasts respect the aggregation constraints. Hyndman et al.215(2011) forecast all series at all levels independently and combine them by a regression (GLS)216projection that is unbiased and minimum-variance among combinations under stated assumptions.217Wickramasuriya et al. (2019) show the coherency-error covariance of the 2011 approach is not218identifiable, and propose MinT, which uses the full forecast-error covariance to minimise the MSE of219the coherent forecasts under unbiasedness, with a closed form and a scalable implementation. The 2024220review covers cross-sectional, temporal and general linear-constraint hierarchies, and the Bayesian,221probabilistic and machine-learning perspectives.222223**What an adaptive hierarchical state-space index would need to add.** Reconciliation operates on224*forecasts* of series that aggregate additively. Price indices do not aggregate additively: a provincial225index is a weighted (stock- or transaction-weighted) combination of municipal indices in logs or226levels, and the weights are themselves estimands (`CLAUDE.md` §15). Applying MinT-type projections to227index *estimates* (not forecasts) therefore requires (i) specifying the linear constraint that defines228coherence for the chosen aggregation (log-linear with fixed weights is the tractable case), (ii) an229estimate of the covariance of the estimation errors across cells — available from the state-space230posterior/filter, which is the natural link to §3 and §6 — and (iii) evidence that reconciliation231reduces error rather than merely smoothing local series toward the aggregate (the review notes232reconciliation "can also improve forecast accuracy", which is an empirical property, not a guarantee).233If the hierarchical state-space model already imposes the hierarchy through its prior, reconciliation234may be redundant; this has to be tested by ablation, and `CLAUDE.md` §7 requires reporting when235reconciliation introduces bias or excessive smoothing.236237---238239## 8. Other verified work that constrains novelty claims (brief)240241- **Hedonic imputation vs. time dummy.** Diewert, Heravi & Silver (2009); Silver & Heravi (2007), *JBES*242 25(2); de Haan (2010), *Jahrbücher* 230(6); Hill & Melser (2008), *Economic Inquiry* 46(4); Hill (2013),243 *JES* 27(5); Hill & Scholz (2018), *RIW* 64(4) (geospatial splines in hedonic imputation); Hill & Steurer244 (2026), *RIW* 72(1) (post-GFC survey of methods, data and policy). Any claim about quality adjustment245 must be positioned against hedonic imputation with flexible location controls (Hill & Scholz 2018).246- **Multilateral methods and non-revisable indexes.** Ivancic, Diewert & Fox (2011), *J. Econometrics*247 161(1) (rolling-window GEKS); de Haan & Krsinich (2014), *JBES* 32(3); Diewert & Fox (2022), *JBES*248 40(1); de Haan, Hendriks & Scholz (2021), *RIW* 67(2). "Stable multilateral temporal linking"249 (`CLAUDE.md` §7) is a well-developed topic; window splicing options are catalogued there and in HMTS.250- **Robustness.** Bourassa, Cantoni & Hoesli (2013), *REE* 41(3) and (2016), *IJHMA* 9(1) (robust251 repeat-sales and robust hedonic indexes); Francke & van de Minne (2017) (*t* errors); Hill, Pfeifer,252 Steurer & Trojanek (2024) (detrimental transaction prices). Robust likelihoods in an HPI are not new.253- **Unobserved heterogeneity / location.** Francke & van de Minne (2021), *REE* 49(4): property and254 spatial random effects in hedonic models improve out-of-sample prediction, most for properties255 without prior sales — directly relevant to the `alpha_location(i)` term in `CLAUDE.md` §6.256- **Revisions.** Clapp & Giaccotto (1999), *REE* 27(1); Clapham, Englund, Quigley & Redfearn (2006),257 *REE* 34(2); van de Minne et al. (2020); Croushore (2011), *JEL* 49(1) (real-time data analysis).258- **Quality adjustment with text / ML.** Nowak & Smith (2020), *AER: Insights* 2(3): the constant-quality259 assumption of repeat-sales HPIs is violated in a time-varying way; textual features from agent260 remarks reduce the bias (up to 7% downward during the crisis, 20% upward after). Relevant to any use261 of listing-text or image features (`CLAUDE.md` §8).262- **Matching estimators.** McMillen (2012), *REE* 40(4); Deng, McMillen & Sing (2012), *RSUE* 42(3);263 Lopez & Hewings (2018).264- **Canadian official/industry indices.** Teranet–National Bank (repeat-sales, registry-based, smoothed265 composite; `teranet2017methodology`); CREA MLS HPI (hybrid repeat-sales/hedonic benchmark-price model;266 `crea2023hpi`); Statistics Canada RPPI (resale component = arithmetic repeat-sales, Shiller 1991;267 `statcan2019rppi`) and NHPI technical guide (`statcan2022nhpi`). No Bank of Canada or CMHC268 methodology document was found (see `unverified.md`).269270## 9. Summary of the gap as it stands after this scan271272Every individual ingredient listed in `CLAUDE.md` §6–7 exists in the verified literature: structural273time-series index levels (Francke 2010), hierarchical common-plus-deviation trends with *t* errors274(Francke & van de Minne 2017), spatially varying latent factors for tracts (Francke et al. 2025; Ren et275al. 2017), estimated shrinkage with MSE (Fay–Herriot / SAE), state-space smoothing with splicing for276small domains in official statistics (HMTS, Ishaak et al. 2024), rolling-window hedonics with low-volume277safeguards (RTD, Hill et al. 2022), nowcast-based revision reduction (Sayag et al. 2022), and coherent278hierarchical combination (MinT). What has **not** been found in a single published estimator is the279combination of (a) a hedonic single-sale observation equation, (b) a multi-level administrative280hierarchy (province / region / MRC / municipality × type), (c) *time-varying, reliability-driven*281pooling weights rather than constant variance hyper-parameters, (d) real-time vintage evaluation with282an explicit registration-lag model, and (e) coherence-aware aggregation evaluated by ablation. The283contribution, if any, is this combination and its documented performance in Quebec-like sparse markets;284it should be described as an integration and extension of Francke & van de Minne (2017) and Ishaak et285al. (2024), not as a new class of model.286