Q5/T1 pilot, BC side — Surrey passes the vacant-lot test too
2026-07-19 · wave 34 · ACADEMIC lane · the second half of the T1 pilot
(analysis/q5-t1-sherbrooke.md ran the Quebec side). Question: does BC
Assessment over-allocate value to land on built parcels — the other way the
BC(~¾)–Quebec(~⅓) gap could be bookkeeping?
Method (mirrors the Sherbrooke pilot, adapted to BC data)
Surrey's open Property Detail List carries per-parcel lot size (parseable from its text field), coordinates, and BC Assessment gross land/improvement values. 101,488 standard lots with parseable areas were pulled in full. Surrey's dataset has no usage codes, so vacant = improvements assessed at exactly zero (6,820 parcels), and micro-markets are ~500 m coordinate grid cells (587 cells with ≥3 vacant and ≥10 improved lots) rather than assessor neighbourhood units.
Results
| Comparison (improved-implied ÷ vacant-benchmark land $/m², per-cell medians) | Median | IQR | Cells |
|---|---|---|---|
| All qualifying cells | 1.97 | [1.17, 4.44] | 587 |
| Size-matched (vacant within 0.5–2× the cell's median improved lot area) | 1.11 | [0.96, 1.45] | 301 |
The size confound is bigger than Sherbrooke's (vacant lots median 2,682 m² vs 681 m² improved — Surrey's vacant stock skews to acreage and ALR-edge land), which is why the raw ratio is inflated and the size-matched number is the meaningful one. Size-matched, the typical cell's built parcels carry land allocations within ~11% of what same-sized vacant lots are assessed at, 63% of cells within [0.8, 1.5].
The two-sided conclusion
| Pilot | Regime | Land share | Size-matched T1 ratio |
|---|---|---|---|
| Sherbrooke (w33) | Quebec / MAMH | 27–30% | 1.06 |
| Surrey (w34) | BC Assessment | 78% | 1.11 |
Two assessment regimes, one at each end of a 40-point land-share spectrum, both allocating built-parcel land within ~10% of their own observable vacant-land benchmarks. The suppression and inflation versions of H2 both fail their cheapest test. What remains of the BC–Quebec gap after this: substantially real scarcity (H1), minus a few points of Quebec vintage lag (H3, measured at 2.5 years in w33). The measured spectrum — and everything downstream of it: incidence, the container arithmetic, the two-lots story — now rests on allocations that have been checked against market-adjacent benchmarks on both sides, not taken on faith.
Caveats
- Vacant-by-zero-improvements is a proxy (BC's dataset has no usage codes); it may include derelict-structure parcels assessed as land-only — which are themselves market evidence of land value, so the contamination is benign.
- Grid cells are coarser than assessor neighbourhoods; the wider right tail (p90 ≈ 7) is concentrated where cells straddle the ALR boundary or redevelopment zoning, where BC's highest-and-best-use standard prices vacant land at development potential. Restricting to mid-city cells narrows the tail; the median is insensitive.
- One city per regime; the Q5 note's remaining tests (teardowns, Ottawa–Gatineau, IAAO ratio studies) stay open for academic hands, and the method is reproducible from the committed dataset and this file's recipe.
Sources
- City of Surrey Open Data, "Property Detail List" (fields LOT_SIZE, GROSS_LAND, GROSS_IMPROVEMENTS, LATITUDE/LONGITUDE, LOT_TYPE), full paged pull of standard lots, retrieved 2026-07-19. https://services5.arcgis.com/YRpe0VKTJytZSSIB/arcgis/rest/services/property%20detail%20list/FeatureServer
- Quebec-side pilot and test design:
analysis/q5-t1-sherbrooke.md(w33),analysis/q5-allocation-note.md(w32). - Aggregate context:
data/land-shares.csv;analysis/surrey-land-share.md(w28).