Analysis · Insurance

Ontario hard and soft market indicator — 2026 Q2

Insurance people talk about hard and soft markets the way sailors talk about weather — constantly, confidently, and without a number. This is a number: one phase per quarter per line, derived the same way every time and published with the date it was calculated, so a later reader can check what it said before anyone knew the answer. This edition reads auto as — and residential as —, with both lines below general inflation for the first time since the pandemic rebates.

10x Technologies August 5, 2026 11 min read Data through —
Auto, realized
—
Since —; now reading —
Residential, realized
—
Since —; — against inflation
Already approved
—
Auto, —–—, not yet felt
General inflation
—
Ontario all-items, the bar both lines are measured against

Nominal rate changes hide the cycle

The companion rates paper reports what auto approvals and both price indexes did. Those are readings of level. They are not a reading of the cycle, because a rate change means nothing until you know what it is a change against.

The problem with reading nominal rate changes is that they are quoted against nothing. Residential insurance in Ontario rose — over the year to —. Whether that is a hard market or a soft one depends entirely on a number that never appears in the sentence: general inflation ran — over the same period. A line rising slightly slower than everything else is not extracting more from its customers. It is quietly getting cheaper.

So every line here is read in real terms — its year-over-year change minus Ontario all-items CPI, in percentage points. That is the level. The second reading is momentum: where that real rate sits against the same quarter a year earlier. Level says whether the line is ahead; momentum says whether it is still pulling ahead. Together they give four phases, and the phase is just the quadrant:

The classification

Hard — beating inflation, and by more than it was. Softening — still beating inflation, but by less. Soft — behind inflation, and falling further behind. Hardening — behind inflation, but closing.

Ontario insurance cycle phase, by quarter

Three tracks: what drivers pay, what insurers have been approved to charge, and what homeowners pay

Beating inflation Behind inflation Turning — level and momentum disagree Reading a new phase, not yet confirmed
View as table
Figure 1. Every band is labelled as well as coloured, so the reading survives greyscale and print. Colour carries only the level — warm is ahead of inflation, cool is behind — and the hatch marks the quarters where level and momentum are pulling in opposite directions.

The shape is one full cycle and the start of another. Auto hardened through 2019, collapsed into a deep soft market during the pandemic rebates, bottomed in 2022, hardened hard through 2023–2024, and has been softening since —. Residential ran the same cycle late and shallow: it never went soft at all in 2021, and it did not reach hard until — — against — for auto.

Where both lines are now

Insurance prices measured against general inflation

Percentage points above or below Ontario all-items CPI; zero means insurance is tracking everything else

Auto Residential
View as table
Figure 2. Both series are in the same unit — percentage points against the same deflator — so they share one axis legitimately. The zero line is the whole reading: above it insurance is taking a growing share of household spending, below it a shrinking one.

Both lines are now below zero. Auto sits at — against inflation and residential at —. The last time both were negative together was the pandemic rebate period, and before that, not in this series.

What the indicator will not yet say is that either line is soft. Auto is reading — and the classification still calls it —, because a phase change waits for 2 consecutive quarters before it counts. Property is inside the dead band — at — it is closer to zero than the ±0.5 point margin where the measurement can be trusted to mean anything.

This is the indicator working, not the indicator failing. A rule that called the turn the instant a single quarter crossed a line would have called four turns in this series that never happened — including a one-quarter dip to soft in residential that reversed immediately. The cost of waiting is that the call arrives a quarter late. The benefit is that it arrives once. Residential alone read soft in — and was back above the line the following quarter; the indicator never called it, which is the only reason it does not appear in Figure 1 as a phase.

The part that is already decided

Every track above stops at —, because a real rate needs a deflator and nobody publishes next quarter's CPI. The auto approvals do not stop there — they carry future effective dates. Filings for — are already approved, compounding to —. Those decreases are decided. They have not reached a renewal notice yet, and they cannot be deflated until the CPI catches up to them, so they sit outside the indicator rather than inside it.

Why residential lags, and why that matters more than it looks

Auto rates in Ontario are filed and approved. Residential rates are not — an insurer changes them when it decides to. The intuition is that the unregulated line should turn faster. It does the opposite here, consistently, in both directions: residential reached hard eighteen months after auto and it is leaving hard later too.

The likely reason is what each line is repricing. Auto is repricing a fairly stable book of risks against a claims environment that moves on a policy-year rhythm. Residential is repricing weather, and weather losses arrive in lumps that take several years to work through a reinsurance program. A line that reprices catastrophe exposure does not turn on a quarter's results.

For anyone holding both lines, the practical consequence is that they will not soften together. Auto is already giving rate back. Property, at — against inflation, has only just reached the point where it stops outrunning the cost of everything else.

What this cannot tell you

Method

This runs on the same pull as the auto rate paper — the same script, one additional Statistics Canada vector. Numbers quoted in the prose above are generated from its output rather than typed in, so a refresh updates the sentences as well as the charts.

  1. Add a deflator

    Ontario all-items CPI, Statistics Canada table 18-10-0004, vector v41691919. It joins the two insurance vectors in the same batched request, so the cycle costs one extra series rather than a second pipeline. The pull was widened to 180 months: a real rate is a year-over-year change and momentum compares it to a year earlier, so the first published quarter needs CPI two years before it.

  2. Compute level and momentum

    For each quarter, at the quarter's final month: real is the line's year-over-year change minus all-items, in percentage points. Momentum is that real rate minus its value four quarters earlier.

    Why four quarters and not one

    Quarter-on-quarter momentum re-imports the seasonality that the year-over-year level has already removed, and it makes the classification flip on renewal-cycle timing rather than on the market. The four-quarter comparison is between two measures that have each already absorbed a full year.

  3. Classify, then refuse to trust it twice

    The quadrant gives a raw phase. Two rules sit on top of it. A dead band of ±0.5 points: if either axis is inside it, the axis is not saying anything the measurement error cannot also say, and the quarter holds the standing phase. A confirmation window of 2 consecutive quarters before a change takes effect. Unconfirmed readings are still published — they appear as outlines in Figure 1 and are named in the table.

    The temptation this exists to resist

    Publishing only the confirmed phase would make every past turn look clean and every call look early. The whole point of dating an indicator is that a later reader can see what it said before the answer was known, including the quarters where it was hedging. A series that quietly repaints its own history is worth less than no series.

  4. Back-check against known history

    A classification rule that cannot reproduce events everyone already agrees on is describing something other than the cycle. This one was checked against five: the 2019 hard market, the 2020–21 pandemic rebates, the 2022 trough, the 2023–24 hard market, and the 2025 turn. It reproduces all five, in both lines, at the first set of thresholds tried.

    What would have been wrong to do

    Thresholds were fixed before the back-check and not moved after it. Tuning a band until the history looks tidy produces a rule that fits the past exactly and predicts nothing — the same failure as averaging quarterly rate changes instead of compounding them, documented in the companion paper. Had the check failed, the honest response would have been to publish the recalibration here, not to bury it.

  5. Keep the approvals track separate

    The leading track compounds the trailing four quarters of market-share-weighted approved changes into an annual-equivalent, then deflates it the same way. It is built to be comparable with the realized track, not to be the fastest possible signal — the same construction on both means a difference between them is a difference in the market, not in the arithmetic.

    A limit worth naming

    Making the leading track comparable costs it some of its lead. A trailing four-quarter window cannot use approvals whose effective dates are still in the future, which is exactly where the auto signal is strongest right now. Those quarters are reported separately, in nominal terms, in the callout above — not folded into the indicator, where they would need a deflator that does not exist.

Refresh cadence

Quarterly, with the auto rate paper — they share a script and a generated file, so they cannot drift apart. Each refresh rewrites the phase for every quarter, not just the newest one, which means a revision to old CPI can change an old phase. The git history of the generated file is the record of what was published when.

Sources

  • Consumer price index, Ontario — vehicle insurance, homeowners' insurance, all-items Statistics Canada, table 18-10-0004-01, vectors v41691997, v41691958 and v41691919. Monthly through —.
  • Auto insurance rate approvals and insurer market share Financial Services Regulatory Authority of Ontario — autorateapprovals.fsrao.ca. Retrieved —; — approval records weighted by — market share.
  • Companion paper Ontario auto and residential insurance rates — the approvals series, its derivation, and its limits.

This page publishes derived aggregates only — quarterly real rates, momentum and phase labels. 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.