Ontario auto insurance rates have turned negative
Every auto insurer in Ontario needs regulatory approval to change its rates, and every approval is published with a date and a percentage. Weight those approvals by market share and you get something unusual: a price series that runs ahead of the prices. Right now it is pointing down for the first time since the pandemic.
The signal
Ontario regulates auto insurance rates. An insurer that wants to charge more has to file with the Financial Services Regulatory Authority of Ontario and wait for approval, and the approved figure is published along with the date it takes effect. That makes auto insurance one of the few consumer prices where you can watch the decision before you feel the consequence.
We pulled every published approval — — filings from 2018 through — — and weighted each insurer's approved change by its share of the Ontario market, so a large insurer's decision counts for more than a small one's. Aggregated by quarter, the result is the series below.
Approved rate changes, weighted by market share
Quarterly effect on Ontario auto premiums, by the quarter the change takes effect
View as table
The shape is a market that pushed hard through 2023 and 2024 — compounding to — and — respectively — then decelerated through 2025 and crossed into negative territory in 2026.
Approvals carry a future effective date. The most recent approval in this dataset was granted on —, but the changes already approved run as far forward as —. Those rate changes are decided. They have not reached anyone's renewal notice yet.
That is the practical value of this series. Consumer price indexes tell you what happened. This tells you what has already been agreed and is waiting in the queue.
Does it match what drivers actually pay?
An approved rate change is not a price. It moves an insurer's base rates; what a driver actually pays also depends on their vehicle, their record, where they live, and — critically — whether they shop around. So the honest test of this series is whether it tracks a genuinely independent measure of realized prices. Statistics Canada publishes one: the Ontario consumer price index for passenger vehicle insurance premiums.
Approved changes vs. prices actually paid
Both indexed to 100 at 2018 Q1
View as table
They track each other closely, which is the first thing worth saying: a series built from regulatory filings is a fair proxy for the direction of Ontario auto prices. In 2025 the two agree almost exactly — — approved against — realized.
Over the full period they do not. Since 2018, approved changes compound to — while realized prices rose —. Approved increases have run roughly — ahead of what drivers ended up paying.
The gap is not an error in either measure. It is what consumers do in response. Filed increases apply to an insurer's book; drivers facing them switch insurers, raise deductibles, drop optional coverage, or requalify on a different rating variable. The approvals show the pressure applied. The price index shows the pressure that survived contact with customers who can leave.
The clearest illustration is 2020–2021. Approved changes stayed mildly positive, then roughly flat. Realized prices fell — — in 2021 alone. Insurers were returning premium through pandemic rebates and reduced-driving credits, which never appear as filed rate decreases. Divergence between the two lines is itself the finding: when they part company, something is moving premiums that is not a filed rate change.
Home insurance is the bigger story
Home insurance in Ontario is not rate-regulated the way auto is. There is no approval feed, no published filings, and no equivalent series to build. What does exist is the same Statistics Canada index methodology applied to homeowners' insurance — which turns out to be enough to show that the two lines have decoupled.
Auto vs. home insurance prices in Ontario
Both indexed to 100 at 2018 Q1
View as table
Since 2018, home insurance has risen — against auto's —. The divergence opened in 2021 and 2022, when home rose — and — while auto prices were falling and then flat.
That is the period auto insurers were handing back pandemic rebates and property insurers were absorbing a run of severe weather losses. Two lines of business, same province, same customers, opposite direction. Anyone reasoning about "insurance costs" as a single number was wrong by about — over this period.
What this cannot tell you
Four limits worth stating plainly, because they bound every claim above.
- This is a market trend, not your premium. An insurer's approved change does not match any individual policyholder's change. Personal risk factors move premiums independently, and often by more than the filed rate.
- The most recent quarters are incomplete. Filings continue to arrive for effective dates already in the series. Quarters shaded as provisional will move, and they will generally move up as more filings land.
- Weights are fixed at one year. Every filing is weighted by its insurer's 2024 market share, including filings from 2018. This keeps the series internally comparable, but it misattributes weight to insurers whose share has changed materially over nine years.
- Filing type is not published. FSRA does not expose whether an approval was a major filing or a minor one, so all approvals are treated identically. A regulatory change effective March 2026 also replaced the previous filing categories, which may make later data behave differently from earlier data.
Method
Everything here is reproducible from public sources with no special access. The pull and the derivation are a single script; the numbers on this page are generated from its output rather than typed in, so a refresh updates the prose as well as the charts.
-
Collect the approvals
FSRA publishes every approved auto rate change through a public endpoint on its rate approvals application. A request with a wide date window returns the full history in one call — currently — records, earliest approval 2017-08-11. Each record carries six fields: insurer, insurer group, approved percentage change, approval date, renewal effective date, and a market share figure.
Departure from the original planWe intended to capture filing type as a fourth field. It is not published — the categories exist in regulation but never appear in the data. Also note that a request without a date filter silently returns only the latest filing per insurer (about 60 rows), not an error.
-
Get market-share weights
One insurer's +6% should not count the same as a much smaller insurer's. The same application exposes a per-insurer endpoint returning four years of market share. We take 2024, the latest complete year, which resolves for — of — insurers and covers — of the market. The remainder are merged or wound-down entities.
Departure from the original planWeights were to come from FSRA's annual review documents. The per-insurer endpoint is better — machine-readable and self-consistent with the filings. Note that the market share field on the filings endpoint is unusable: many rows are zero and the snapshot sums to little over half the market. Use the per-insurer endpoint only. 2023 is also unusable as a weight year — one large insurer is recorded as zero.
-
Aggregate into a quarterly series
Bucket each approval by the quarter of its effective date, not its approval date — the effective date is when the change reaches policyholders, and it is what makes the series leading. For each quarter, compute the sum of approved change times market share. Quarterly is the right granularity: monthly is too noisy, annual hides the turning points.
-
Compound, never average
Annual figures compound the quarterly impacts. This is the step most likely to be got wrong, and getting it wrong is not a rounding matter.
A mistake worth flaggingAn earlier pass at this analysis averaged the four quarterly impacts to get an annual figure. That produces roughly 1.3% for 2025 instead of —, and it made the approvals look like they were running three to six times below realized prices — a dramatic finding that was entirely an artifact of the arithmetic. Quarterly rate changes stack on top of each other; they do not take turns.
-
Cross-check against an independent measure
Approvals tell you what changes are coming. They do not tell you where prices landed. Statistics Canada table 18-10-0004 gives an Ontario consumer price index for vehicle insurance, monthly, through —, with a documented and stable API. Plot both. Divergence is itself a signal.
Departure from the original planThe intended cross-checks were FSRA's own average premium figure and its quarterly approval announcements. The quarterly announcements were discontinued in 2019. The average premium page publishes only a rolling four-column window, so it destroys its own history on each update — if you use it, archive every pull. Statistics Canada is the durable cross-check.
-
Flag the fluctuations
Compute a trailing mean and standard deviation over the quarterly series and mark quarters that fall outside the band. We use a trailing 8 quarters at 2 standard deviations, which flags — of — quarters.
Departure from the original planThe plan specified a trailing four quarters at one standard deviation. Run against the real series, that flags 20 of 36 quarters — more than half — because four observations produce a small, noisy standard deviation that almost any quarter clears. It detects nothing. The wider window and higher threshold were chosen by testing against the data, and isolate the events that have meaning: the 2020 rebates, the 2022–23 hard market, and the 2026 turn.
-
Handle home insurance separately
There is no approvals feed for home insurance, so there is no equivalent series to build. We use the Statistics Canada homeowners' insurance index for Ontario and keep it on its own figure. Auto and home are never merged into a single index: one is built from regulatory filings and one is a price index, and averaging them would launder the weaker signal into the stronger one.
Departure from the original planInsurance Bureau of Canada publications were the intended source. They are not usable for this: the most recent public Facts Book is 2023, the figures are national rather than provincial, and it publishes no average premium. Fixed-profile quote tracking — re-quoting one frozen risk profile on a schedule — remains the rigorous DIY option, but it produces nothing publishable until several quarters have accumulated.
Refresh cadence
Quarterly: re-run the pull and regenerate the series. Twice a year: check whether the regulatory benchmark reports have moved. Annually: advance the market-share weight year. Every pull is written to a dated snapshot before anything is derived from it, so the series can be audited against what the source actually said at the time.
The Statistics Canada index carries a published methodology and a stability guarantee. The FSRA endpoint does not — it is a public application programming interface with no documentation, versioning, or commitment to keep working. That is the main fragility in this pipeline, and the reason each pull is archived rather than fetched live by this page.
Sources
- Auto insurance rate approvals and insurer market share Financial Services Regulatory Authority of Ontario — autorateapprovals.fsrao.ca. Retrieved —; — approval records.
- Consumer price index, Ontario — vehicle and homeowners' insurance Statistics Canada, table 18-10-0004-01, vectors v41691997 and v41691958. Monthly through —.
This page publishes derived aggregates only — quarterly and annual weighted figures and the index series plotted above. The underlying filing-level records are not reproduced here.
Contains information licensed under the terms of the King's Printer for Ontario. © King's Printer for Ontario, 2026. Adapted from Statistics Canada, table 18-10-0004-01; this does not constitute an endorsement by Statistics Canada of this product.
Nothing here is financial, insurance, or legal advice. It is a description of a market trend derived from public data, and it does not predict what any individual will pay.