What does it cost to live on a Trust island?

Read this before the numbers

We want the Islands Trust abolished, so you should assume we will publish whatever helps us. The hypothesis above is the publisher's own, written down before the data was assembled, with the result that would refute it fixed in advance. 3 claims have already been withdrawn and are listed below, 2 of them squarely against our case — including the strong form of this page's own hypothesis.

The instrument is not ours either. The Province paid Statistics Canada to cut the census to the Trust's own jurisdiction boundaries, on the indicator list that provincial housing regulation forces on local trust committees and regional districts alike. We did not choose these measures, which is what makes them usable against us.

The twelve islands, on the Province's own instrument

Method. Read from the Province's custom census tabulation, one sheet per local trust area (IST_590002 Denman through IST_590013 South Pender), 215 variables, four censuses, percentage change precomputed. Extracted by scripts/extract-hnr-census.mjs, which locates each column by its header label so a layout change fails loudly instead of shifting every number by one column. The 13th IST sheet, IST_590243 'South Wellington part B', is suppressed throughout and carries no data. Source. Statistics Canada custom tabulation for the Islands Trust local trust areas, 2021 Census, commissioned by the Province of British Columbia (BC Data Catalogue, Housing Needs Reports group). Licence 'Access Only': derived figures published with the source named; the file itself is not rehosted. (measured 2026-08-22). Geography. Islands Trust local trust area (StatCan type IST) — the jurisdiction the Trust actually regulates, NOT a designated place and NOT a census subdivision.

Why shelter cost is not an affordability measure. Average monthly shelter cost is a paid cost, not a measure of affordability: an island of mortgage-free owners and unserviced cabins reports a low one. Publish it as a ratio to income, or beside core housing need, never alone. Hornby is the case in point: the lowest household income of any island with published figures, and one of the lowest shelter costs. That describes who is still able to live there, not a bargain.

The spread is the finding, and it breaks the simple version of our own claim

Median household income runs from $34,800 on Lasqueti to $78,500 on Salt Spring — and the island with the highest income also has the lowest core housing need in the set at 6.0%. So “Trust islands are poorer and worse housed” is not true as a general statement about the Trust area, and we do not publish it as one. The islands where high housing need sits alongside low income are Denman, Hornby, Lasqueti — and even there, this page does not claim the Trust is the cause.

The one pattern that holds everywhere

Every local trust area has a median age above 42.8, which is British Columbia's. Not one is within thirteen years of the province, and South Pender is 24.7 years above it. It cuts two ways and both belong here: it is what a place looks like when working-age families cannot afford to live there, and it is also retirement in-migration, which is a preference people are entitled to. Saying only the first would be a slogan.

One electoral area, one director, two land-use authorities

The sharpest comparison available needs no control island. Comox Valley Electoral Area A contains Denman, Hornby and the Baynes Sound mainland: one elected director, one set of regional services, one assessment authority — and two different bodies deciding what may be built.

Method. Computed two independent ways that agree exactly: a direct 8-dissemination-area sum, and Comox Valley A CSD minus Denman DPL minus Hornby DPL. One electoral area, one director, one set of regional services, one assessment authority, two land-use authorities. This is a DIFFERENCE across the boundary, not an EFFECT of it: island status is confounded with Trust status here, and seasonal share is simultaneously an outcome and our amenity-pressure covariate. Source. Statistics Canada, 2021 Census of Population; docs/trust-comparison-study-protocol.md §2. (measured 2026-08-22). Geography. Designated place (DPL) and dissemination area (DA). NOTE: Hornby's core housing need is 13.0% here and 13.9% on the IST cut in TRUST_ISLANDS. Both are correct for their own geography. Never mix them.

Inside one electoral area, the share of homes with nobody living in them is 6.7% on the mainland against 21.1% and 41.5% on the two Trust-planned islands. Core housing need runs the same way.

Read that carefully before using it. It is a difference across the boundary, not an effect of the boundary. The line is also a water line, so island status and Trust status are tangled together here, and the seasonal-home share is both a result we are measuring and the thing we use to measure development pressure. It is the strongest single number this project has, and it is not a causal claim.

What running your own island actually costs

Bowen Island incorporated in 1999 and runs its own land use while staying inside the Trust area. It is the only place in British Columbia that has made the change this campaign asks for, which makes its books the only real answer to “what would it cost us?” — 19 audited years of them.

Method. Schedules 201 (population and dwellings), 401 (consolidated revenue including own-purpose taxation) and 701 (assessed values by class), 2005 through 2023, nineteen consecutive years, with per-year distribution percentiles across all BC municipalities. The full series and eleven comparator municipalities are in docs/bowen/lgs-fiscal.json; only the endpoints and the 2005 distribution are restated here, so nothing is transcribed twice. Source. Province of British Columbia, Local Government Statistics. Fetched through the Wayback Machine because the provincial host refuses the TLS connection for these assets; the bytes are the Province's own. (measured 2026-08-30). Geography. Bowen Island, island municipality (census subdivision) — a municipality, which is the ONLY reason this data exists at island level.

Tax per resident went from $839.7 to $2,359.73 — 2.81 times — while the population grew by 634. In 2005 that put Bowen at the 80th percentile of all 155 British Columbia municipalities, against a median of $662.51. This is the finding on this page that most obviously cuts against the campaign, and it leads the section rather than closing it: self-government on an island is not cheap. The honest case for it is representation, not savings.

And it cannot be compared with what you pay now. Nothing in the Bowen series can be compared with an unincorporated island's cost of government, because that figure does not exist in any source. Provincial Local Government Statistics carries one row per regional district with no electoral-area breakout. Never present a regional-district figure as an island's.

Why the comparison islands are harder than they look

The obvious controls are the ferry-dependent islands outside the Trust — Cortes, Quadra, Texada. Two of them can be measured as islands. One cannot, and that is worth knowing before anyone quotes a number about it.

Method. Each unit resolved to the statistical geography that actually covers it, and recorded as island-only or not. Established by checking every sheet in the Province's workbook (13 IST + 29 whole regional districts + BC + two metadata sheets = 45) and by identifying each control's electoral area. Cortes, Quadra and Texada appear in no file under data/hnr/: the Province's instrument covers the treated side only. Source. Statistics Canada census geography; BC Data Catalogue Housing Needs Reports group; docs/trust-comparison-study-protocol.md §4a for the Quadra dissemination-area resolution. (measured 2026-09-11). Geography. Mixed by design — that is what this table records.

The bias in this comparison also runs toward us, and we would rather say it than be caught with it. The Trust area was drawn in 1974 around islands already under development pressure from Vancouver and Victoria; the obvious comparison islands were left out partly because they faced less of it. So finding lower incomes inside the Trust area would prove nothing on its own. That is the methodological problem this whole study is built around.

Where the numbers stop

3 islands have figures Statistics Canada withholds, by rule rather than by oversight. The rules are theirs, not ours.

Method. Read from Statistics Canada's own dissemination and confidentiality documentation rather than inferred from which cells happen to be blank. Source. Statistics Canada, Census of Population Guide 2021, Chapter 10 (Dissemination); confidentiality and random rounding rules. (measured 2026-09-11). Geography. Applies to every geography used here.

• Saturna (425 people). Core housing need is not usable for Saturna. The Province's published summary records it as suppressed, while the extracted workbook cell reads 0. Those are different statements and we do not know which is right, so neither is published. Resolve against the source workbook before this cell is ever filled.

• South Pender (235 people). Income and core housing need withheld by Statistics Canada: under 250 people in private households.

• Gambier (220 people). Income and core housing need withheld by Statistics Canada: under 250 people in private households. Gambier's 2021 census response was also the weakest in the set (38.2% short-form and 40.0% long-form non-response), so even its published cells carry that caveat.

What we got wrong

Dated, in full, at the same prominence as everything above. 2 of these make our case weaker and they are left standing. If this list ever stops growing while the findings keep coming, stop believing this page.

• 2026-09-11 · against our case

  Withdrawn: “Trust islands are uniformly worse off economically than they would be under ordinary local government — higher costs and lower incomes across the board.”

  It is not uniform, and the Province's own tabulation says so before we published anything. Salt Spring, the largest Trust island, has the HIGHEST median household income in the set at $78,500 and the LOWEST core housing need at 6.0% — better on both than many off-island BC communities. Thetis ($68,500, 8.8%) and North Pender ($67,000, 7.2%) also run against the pattern. The real pattern in the data is a spread, from Lasqueti at $34,800 and 18.2% core need to Salt Spring at $78,500 and 6.0%, and any honest version of this study reports that spread rather than an average. The hypothesis survives only in the specific, weaker form that SOME Trust islands — Denman, Hornby, Lasqueti — show high core housing need alongside low income, and even there the cause is not established.

• 2026-09-11 · against our case

  Withdrawn: “A fifty-year economic history of the Trust islands, measuring what changed after the Trust was created in 1974.”

  Cannot be built, and the reason is structural rather than a matter of effort. Statistics Canada introduced the designated place in 1996, twenty-two years after the Islands Trust Act. There is no island-level census geography before then, so there is no pre-treatment series, no pre-trend to fit, and no difference-in-differences on the founding of the Trust. Pre-1996 material exists only as historical census geography products reachable through a university data-library affiliation or a direct request to StatCan, and we hold none of it. The study is therefore a documented series from 2001/2006 onward, and the pages must not imply a before-and-after that the data cannot support. This removes the strongest claim the campaign could have made from this work.

• 2026-09-11 · our own method

  Withdrawn: “The control-island comparison can be built from the Province's custom census tabulation by pairing each local trust area against its regional district's electoral areas in the same file — 'same file, same variables, same censuses'.”

  Impossible as written, and it was our own written instruction in docs/hnr-custom-census.md, so anyone following it would have built the comparison out of sheets that do not exist. Every sheet in the workbook was listed and typed: 13 island trust areas, 29 WHOLE regional districts, British Columbia, and two metadata sheets. There are no electoral-area sheets. Strathcona appears once at 47,280 people and qathet once at 21,065; Cortes, Quadra and Texada appear in no file we downloaded. Pairing Denman's 1,340 against Strathcona's 47,280 would have compared an island to Campbell River, which is the exact error that discredited the only prior Trust comparison study. The instruction is retired in place and the control side must be assembled from island-only census subdivisions, designated places, or dissemination-area aggregations, each labelled.

What we cannot compute yet

Listed so nobody has to ask, and so a gap in the evidence cannot be mistaken for a finding.

• Assessed values and land-value trends for ten of thirteen local trust areas. Held for Denman, Hornby and the Strathcona-sourced controls only. The gap is BC Assessment's terms of use, not unfinished work. An open lead: BC Assessment publishes 'Roll Totals by Jurisdiction' as a spreadsheet, and if its jurisdiction field resolves to a local trust area it would bypass census suppression entirely — unverified, and not to be relied on until checked.

• Rent and vacancy rates for any island in this study. CMHC's Rental Market Survey samples urban areas of 10,000 or more, and its rural survey 2,500 to 10,000. Denman (1,340) and Hornby (1,105) are below even the rural threshold. There is no rent series for them, and no proxy is substituted.

• Per-capita cost of government for any unincorporated island. Provincial Local Government Statistics has one row per regional district and no electoral-area breakout. The figure does not exist for any Trust island or for Cortes, Quadra or Texada. Bowen is the sole exception because it is a municipality.

• Island-level agricultural economics. The Census of Agriculture publishes nothing below the census consolidated subdivision, which is coarser than the electoral areas. Island-specific farm data would require a custom StatCan tabulation.

• Business counts for Denman and Hornby. Canadian Business Counts is published for census subdivisions only. Cortes, Texada and Salt Spring have island-only CSDs and can be measured; Denman and Hornby have none, so only the coarse combined electoral area is available.

• Retail food and fuel price comparison. No standing series exists at this geography. Either a fixed basket is collected primarily at named stores with dates, or nothing is published.

• BC Ferries traffic by individual route. The annual report to the Ferries Commissioner publishes by route GROUP (the Minor routes as one). Whether individual small-route counts appear anywhere is unresolved and needs a full appendix read before any method depends on it.

Method. Each blocker recorded with the specific rule, licence or absence that causes it, so a reader can tell a limit of the data from a limit of our effort. Source. docs/economic-history-study.md §2.4, where each item carries its source. (measured 2026-09-11). Geography. Various — each blocker names the geography it fails at.

What this study will not do

• We will not score ferry cost against the Islands Trust. Ferry dependency is a cost shared by Trust and non-Trust islands alike — Cortes and Texada pay it too. It belongs in the study because an urban comparison would silently omit a real cost of island life, not because the Trust causes it.

• We will not publish the Islands Trust tax levy without the matching regional-district planning requisition beside it. Every rural BC property pays for land-use planning somewhere; showing the Trust levy against zero would be dishonest.

• We will not pool Cortes, Quadra and Texada into one 'non-Trust' bloc. They differ from each other more than some differ from the Trust islands, and pooling them is the error that discredited the only prior study of this question.

• We will not put a figure from one statistical geography in the same sentence as a figure from another without naming both.

• We will not fill a suppressed cell with an estimate. The line is omitted or marked withheld, with the rule that withheld it.

• We will not describe any difference between islands as an effect of the Trust unless a design in the protocol licenses that claim. Most of what this study can produce is description.

Status

Designed and pre-registered 2026-09-11. The treated side is measured from the Province's own census tabulation; the matched control comparison is not yet built, and the protocol permits publishing that it could not be. The full design, its decision rule and its open verification items are in docs/economic-history-study.md in the public repository, fixed before the results existed.