Q5, specified: is the BC–Quebec land-share gap real, or an assessment convention?

2026-07-19 · wave 32 · ACADEMIC lane · turns the question flagged in w21 into a pre-registerable research design, with our nine-roll dataset as its raw material.

The measured fact that demands explanation

Nine full city rolls (data/land-shares.csv, 1.4M parcels) show two tight, far-apart clusters in the land share of assessed property value:

The in-province clustering is remarkably tight (BC within 9 points, Quebec within 8) while the between-province gap is ~40 points. Everything downstream — incidence, the container arithmetic, the two-lots story — treats this as an economic fact. Q5 asks: how much of it is?

Three hypotheses (not mutually exclusive)

Four tests, all feasible with public data

  1. T1 — Vacant-parcel benchmark (within-city, within-regime). Vacant parcels' assessed values are nearly pure land (our parse self-checks: 97–99% in every roll). For neighbourhoods with both vacant and improved parcels, compare improved parcels' implied land value per square metre against adjacent vacant parcels' assessed value per square metre. If improved-parcel land allocations track vacant benchmarks, the convention is anchored to observable land prices (H2 weak); systematic gaps measure the allocation bias directly. The Quebec rolls carry lot areas and the BC city datasets carry lot sizes — this is computable from data we already have.
  2. T2 — Teardown residuals. Properties sold and demolished shortly after sale trade at pure land value regardless of assessment convention. Compare sale prices of identified teardowns against their assessed land values in both regimes (building-permit + sales linkages; Vancouver and Montreal both publish permit data). The urban-economics literature uses exactly this design for land-value measurement.
  3. T3 — The Ottawa–Gatineau natural experiment. One labour market, one river, two assessment regimes. Ontario publishes no split, but sales of vacant land clear on both banks at market prices. A vacant-land sales comparison (price per m², controlling for zoning/servicing) tests whether the economic land-value gradient is continuous across the river while the assessed land shares jump — the cleanest possible H2 detector.
  4. T4 — Assessment-to-sale ratio study on vacant sales. The standard assessment-quality method: for vacant-land sales in each regime, compute assessed-value/sale-price ratios. Ratios near 1 with low dispersion in both regimes ⇒ land values are well-measured where directly observable, pushing any H2 bias into the improved-parcel allocation specifically (where T1 catches it).

What each outcome would mean for Groundshare

Honest prior: H1 is doing most of the work (the supply-constraint pattern reproduces within BC at every distance from downtown, and the US echo point is regime-independent), H3 a few points, H2 unknown and worth measuring — which is why this note exists.

Pilots executed, both sides (w33–w34): T1 ran on Sherbrooke (analysis/q5-t1-sherbrooke.md) and Surrey (analysis/q5-t1-surrey.md). Size-matched results: Sherbrooke 1.06 (improved-parcel land allocations within ~6% of same-neighbourhood vacant benchmarks), Surrey 1.11 (~11%, coordinate-grid micro-markets). Both directions of H2 — Quebec suppression and BC inflation — fail their cheapest test. The vintage lag is measured (2.5 years, w33). Working conclusion pending the heavier tests: the gap is substantially H1 (real scarcity) with a few points of H3; the spectrum's allocations are checked against market-adjacent benchmarks at both ends.

Why this is publishable regardless of Groundshare

A harmonized, open, parcel-level comparison of land-value allocation across two Canadian assessment regimes does not exist in the literature. It matters for property-tax design, for the split-rate debates, for national balance-sheet land estimates (StatCan's own land series leans on assessment data), and for any future LVT discussion in Canada. The dataset is already assembled and reproducible (data/land-shares.csv, tools/parse_quebec_roll.py); T1 and T4 are computable by one graduate student from open rolls.

Sources

  1. Measured rolls and parser: data/land-shares.csv, tools/parse_quebec_roll.py (w9–w29); city sources listed per-row.
  2. Supply-constraint reading: analysis/surrey-land-share.md (w28), analysis/montreal-land-share.md (w21).
  3. Ottawa–Gatineau transparency asymmetry: analysis/assessment-readiness.md (w14/w22/w27), analysis/quebec-rolls.md (w24).
  4. Teardown-based land valuation in the literature: e.g. Dye & McMillen, "Teardowns and land values in the Chicago metropolitan area," Journal of Urban Economics 61(1), 2007 — the design T2 replicates in Canadian data.
  5. Assessment-ratio methodology: IAAO, Standard on Ratio Studies (the T4 template).

Groundshare — a proposal in open development. Every number traces to a cited public source with its retrieval date; corrections are published, not erased. Rebuilt 2026-07-19 from the repo's research files.