Canada missed two self-set launch dates for a national flood insurance backstop — 1.5 million highest-risk households remain outside any federal program

What is happening?

Canada has no national flood insurance program. The federal government has said since 2019 that it would build one, and as of 2026-08-20 it has not.

The Task Force on Flood Insurance and Relocation, convened by the federal government, published its report titled Adapting to Rising Flood Risk on 2022-08-30. It found that about 90 percent of the flood risk in Canada sits on about 10 percent of the housing stock, roughly 1.5 million households, and that for that group the premium a private insurer would have to charge cannot be reconciled with the payouts expected, so coverage is often not offered at all. The Public Safety Canada page that owns the file still describes the government as reviewing that report to determine next steps, and that page carries a last-updated date of 2026-07-28.

Two launch dates have come and gone. Budget 2023 set aside 31.7 million Canadian dollars over three years for design and consultation. Budget 2024 allocated 15 million Canadian dollars to the federal housing and mortgage Crown corporation and said the program would be operating by the 2025 flood season, which placed the target at about April 2025. In January 2025 a federal, provincial and territorial working group began meeting to finish the design, and the target was reset to April 2026. April 2026 passed with no program.

In reporting dated 2026-06-25 the federal minister responsible for emergency management said that near-term delivery could not be promised, and the national property and casualty insurance industry association said that no schedule had been shared with it. The provincial minister responsible for emergency management in British Columbia said the program appeared to have stalled. No source opened here gives a replacement target date.

Whose problem is this?

The body that would have to deliver the program is the same body that set both of the dates it did not meet, and no external clock runs against it.

Where does this problem end?

The measure taken here is the distance between the dates the federal government set for itself and what exists on those dates. The value question next door, whether people should be insured to stay where they are or helped to leave, is a choice among goals rather than a gap between a stated goal and a result, so it sits outside this frame.

What is the state now, and what should it be?

Now

What the government itself said should hold

No source opened here names a target for how many of the 1.5 million households should hold a policy, or a date by which the residual gap closes. Every target above is a calendar date for the arrival of an instrument, not a stated level of coverage that the instrument is meant to reach.

How big is it?

About 1.5 million households. That is the figure the federal task force published in 2022, describing roughly 10 percent of the housing stock that carries roughly 90 percent of the flood risk in the country. The same figure was restated at the Budget 2024 announcement, where the industry side described more than 1.5 million households as being in scope, and it appears again in 2026 reporting without revision.

This count measures households standing outside a federal program that does not exist, not households flooded in any single year. The annual harm is counted separately and by a different instrument: insured flood losses exceeded 4 billion Canadian dollars in 2024 and 1 billion Canadian dollars in 2025, and federal post-flood disaster assistance averaged about 230 million Canadian dollars a year from 2016 to 2025.

The two quantities answer different questions. The first says how many households sit in a market that will not price them, and the second says what the country paid in the years while nothing was built. A separate figure of 9.4 billion Canadian dollars for total insured catastrophe losses in 2024 covers all perils rather than floods alone, so it is not substituted for the flood figure and not added to it.

Under what conditions does it arise?

1. The risk is concentrated, and concentration is what breaks the private market. When 90 percent of the exposure sits on 10 percent of the homes, the premium that would cover the expected payout for that tenth approaches the payout itself. The task force described exactly this in 2022, and the market outcome it described is the one still observed in 2026 — coverage refused, or offered at a price reported to reach 15,000 Canadian dollars a year.

2. The instrument requires two orders of government to agree before it can exist. A backstop that shares losses has to name who pays which share. The absence of a federal, provincial and territorial cost sharing agreement was flagged at the Budget 2024 announcement and was still described as unresolved in 2026 reporting.

3. The information layer the program would price against is itself incomplete. The audit published in May 2026 by the Commissioner of the Environment and Sustainable Development recorded 131 flood hazard mapping projects in progress at 2025-12-31, of which 26 were complete and 11 had been posted publicly, and about half of the 200 high-risk priority areas designated in 2022 had been mapped.

4. The cost of not acting is paid through a channel that requires no program. Post-flood disaster assistance arrives after each event without anyone having to design, fund or agree on anything in advance, and it averaged about 230 million Canadian dollars a year from 2016 to 2025.

A program that has not launched cannot be judged by how well it performs, only by the dates it was supposed to arrive on. That is why the measurable object here is a calendar rather than a coverage rate.

What has been tried?

Two tracks were run at once — build the insurance backstop, and build the flood risk information the backstop would need. Neither has reached the public, and the second has now been audited while the first has not been delivered.

What was found?

Why is it still unsolved?

Institutional gap — the instrument that would close it has been designed, funded in part and announced twice, and nothing in the system obliges anyone to finish it by a date.

The first part of the gap is that the deadline and the deadline-setter are the same party. A statutory deadline can be enforced, a court order can be enforced, and a condition attached to an appropriation can be enforced. A target announced in a budget document or agreed in an intergovernmental working group has none of those properties. Both of the dates that passed were dates the federal government chose, which means missing them changes nothing except the calendar.

The second part is that the design has a veto structure built into it. A loss-sharing backstop cannot be declared by one order of government alone, because the shares have to be agreed. The absence of that agreement was already named at the 2024 announcement, and two years later the opened sources still describe it as the unresolved item. Where agreement is the precondition, the number of parties who can hold the outcome equals the number of parties at the table, and no source opened here records what any of them are asking for.

The third part is that the delay is invisible as an event. Nobody publishes a notice that a program has slipped. The department page stays up, the working group keeps meeting, the portal gets a completion announcement, the disaster assistance keeps flowing after each flood season, and the file looks the same from outside on the day the target passes as it did the day before. An institution that has stopped and an institution that is still working look identical from outside when neither publishes a date.

What observation would mean it is solved?

Candidates — (a) a national flood insurance program begins operating, with a published effective date and published eligibility rules (b) the share of the 1.5 million highest-risk households able to buy flood coverage at a stated price rises (c) federal post-flood disaster assistance falls as the insured share of flood losses rises.

(a) alone is weaker than it looks. A program can launch with an eligibility threshold high enough, a deductible large enough or a participation requirement narrow enough that the households the 2022 task force identified are still outside it. A launch also does not say whether provinces signed the cost sharing agreement or whether coverage reaches the concentrated tenth of the housing stock that is the whole reason the program was proposed.

(b) is the observation that matches the problem, and it is the one nobody currently counts. No opened source reports how many of the 1.5 million are refused outright and how many are quoted a price they cannot pay. Without that split there is no baseline, and a program could launch into a measurement vacuum where improvement and no improvement look the same.

(c) alone can move for the wrong reason. Disaster assistance falls in years with mild flooding and rises in years with severe flooding, independently of any program. It also falls if the assistance rules are tightened, which shifts cost onto households rather than removing it. The three have to be read together, and (c) has to be read against the flood seasons it came from.

What is it connected to?

Fills with researchmunicipal and provincial floodplain zoning, managed retreat and relocation policy, the mortgage and property valuation consequences of uninsurable homes, federal disaster financial assistance arrangements as a standing instrument, and comparable national flood insurance backstops in other countries. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• Why the 2026-04 target was missed. The opened sources describe the discussions as very complex and stop there. One 2024 report lists the kind of issues in general terms, such as differing definitions and shares, but no opened source names the specific items on which the parties are apart, and no opened source carries an on-the-record explanation from any of the governments involved.
• Whether the 450 million Canadian dollars pledged in the 2025 campaign entered an enacted budget. No opened source states either that the pledge was appropriated or that it was not, and none of them cites a budget document that would settle it. The pledge is therefore recorded here as a commitment and not as funding in place.
• Whether a replacement target date now exists. The most recent opened source is dated 2026-06-25 and states that no schedule had been shared with the industry side. No opened source carries a later announcement, so the position as of 2026-08-20 is recorded as the position of that date and not as a confirmed present state.
• How the 1.5 million households split between refusal and unaffordable pricing. The opened sources describe the outcome qualitatively, saying most of the group is affected, and give a single upper premium figure. None of them gives the proportion refused coverage outright, the proportion quoted a price above some threshold, or the distribution of prices in between.
• What the audit says at the source. The audit of flood hazard mapping published by the Commissioner of the Environment and Sustainable Development in May 2026 could not be opened, because the federal domain returned HTTP 403 to automated retrieval. Every figure attributed to that audit here is taken from one secondary report, and none of it was checked against the audit text.
• What the flood risk portal will do once provinces opt in. The opened source records that the tool depends on each province and territory supplying data and that it does not account for climate change. Nothing opened here states how many jurisdictions have agreed to supply data, or on what schedule.
• What the highest-risk households do instead. No opened source describes what uninsured households in that group actually rely on after a flood, how much of the loss disaster assistance covers, or how many of them carry a mortgage on a property that cannot be insured.

See the evidence

No primary federal document was read at the source. The Public Safety Canada page was opened directly and is the only government-owned surface here; it carries the position of the department in the words of the department, that the 2022 report is under review to determine next steps. The already-public position of the government is otherwise present in this document at three points: that departmental page, the statement of 2026-06-25 by the federal minister responsible for emergency management, and the announcement of 2026-04-27 that the portal technology was complete. Everything else comes from professional, industry and news publications that cite the budget documents, the task force report, the working group and the audit. The audit could not be opened, so its figures ride on a single secondary carrier and are graded low for that reason, below values that come from sources actually opened here. Where the opened sources overlap they agree: the figure of about 1.5 million highest-risk households appears in three of them across four years without revision, the 2026-04 target appears in two, and the absence of a federal, provincial and territorial cost sharing agreement appears in two separated by two years. Where they are silent the silence is recorded in the block above rather than filled. This is a Path A output, a research-based definition, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 1,500,000–1,500,000 As of 2022, restated in 2024 and 2026

Sensitivity The interval has zero width and is not a confidence interval. It is a single published estimate, rounded to two significant figures, restated across four years by three opened sources without revision and without any of them attaching a margin. Both bounds sit at a floor rather than at a midpoint, because the 2024 restatement describes more than 1.5 million households as being in scope, so the true value is at or above the number used here. The unit is households and not persons, and the conversion is not derivable from any source opened in this round. Two corrections would move the count in opposite directions and neither could be computed from the material available. Downward, some households inside the highest-risk group do obtain flood coverage, at premiums reported to reach 15,000 Canadian dollars a year, and the opened sources say only that most of the group is affected without giving the split between outright refusal and unaffordable pricing, so the subset with no coverage at all is smaller than 1.5 million by an unstated amount. Upward, the count is defined by concentration of risk rather than by absence of a coverage path, so it excludes households at moderate flood risk who may be underinsured, and it is anchored to a 2022 measurement against a hazard mapping layer that the audit published in May 2026 recorded as roughly half complete for the 200 high-risk priority areas designated in 2022, which means the denominator the count was drawn from was itself incomplete when the count was made. The count also measures exposure to a missing instrument rather than annual harm. The harm counted separately was more than 4 billion Canadian dollars in insured flood losses in 2024 and more than 1 billion Canadian dollars in 2025, alongside about 230 million Canadian dollars a year in federal post-flood disaster assistance averaged over 2016 to 2025, and those are payment quantities rather than household counts, so they answer a different question and are not combined with this figure.

Regional breakdown No source opened in this round gives a province-by-province or territory-by-territory count of highest-risk households, and no opened source breaks the national figure down by watershed, municipality or floodplain. The only regional material found is the audit finding that about half of the 200 high-risk priority areas designated in 2022 have been mapped, which counts mapping projects rather than households and does not name the areas. Splitting the national figure by provincial population would be proportional allocation, and it would be wrong in both directions at once, because flood exposure concentrates by watershed and floodplain rather than distributing with population, so provinces with large populations away from mapped flood hazard would be overstated and smaller jurisdictions with concentrated riverine or coastal exposure would be understated. The artifact that would permit a regional split is the same one the audit records as incomplete, namely a finished national flood hazard mapping layer.