MLS HPI vs Average Price: Why Canadian Home-Price Headlines Disagree

MLS HPI vs Average Price: Why Canadian Home-Price Headlines Disagree

Short answer: The average price is the average of every home that sold in a month, so it moves when the mix of sales changes. The MLS Home Price Index prices a fixed "benchmark" home, so it moves only when prices for comparable homes change. In September 2026 the two told different stories: CREA's national average sale price was $666,538, down 1.6% from a year earlier, while the national composite MLS HPI was down 3.4%. Teranet's repeat-sales index fell 3.5% and Statistics Canada's new-home index about 2.0%.

Every month a Canadian reader can find a headline saying prices fell sharply and another saying they barely moved. Usually both are accurate. They are measuring different things, and the gap between them is information, not noise. This guide explains the four national measures that show up most often, using one month of published data, so the next headline can be read for what it is.

One month, four answers

The four measures below were all published for September 2026. They cover the same country and the same month.

Measure	Publisher	September 2026, year over year	Published
National average sale price, not seasonally adjusted	Canadian Real Estate Association (CREA)	-1.6% ($666,538)	October 16, 2026
MLS Home Price Index, national composite, not seasonally adjusted	CREA	-3.4%	October 16, 2026
Teranet-National Bank Composite House Price Index (11 CMAs)	Teranet and National Bank of Canada	-3.5%	October 20, 2026
New Housing Price Index, Canada, total (house and land)	Statistics Canada	-2.0% (index 122.7 to 120.2)	Table 18-10-0205-01

Show the data behind this chart

Measure	What it prices	Sep 2026 vs Sep 2025	Latest month
CREA average sale price	Every home sold on MLS that month, all types	-1.6% ($666,538)	Not stated as a monthly change in the release
Statistics Canada NHPI	Builders' prices for unchanged new house models	-2.0% (122.7 to 120.2)	-0.2% (120.4 to 120.2)
CREA MLS HPI composite	Benchmark homes modelled from MLS sales	-3.4%	-0.4%
Teranet-National Bank HPI	Repeat sales of the same properties, 11 CMAs	-3.5%	+0.1% (seasonally adjusted)

The spread is 1.9 percentage points, from a 1.6% decline to a 3.5% decline. The two indexes designed to hold the home constant, the MLS HPI and Teranet, agree closely with each other. The average price, which does not hold the home constant, shows the smallest decline of the four. The new-home index, which measures a different product altogether, sits between them.

On a month-over-month basis the picture is quieter: CREA says the composite MLS HPI "edged down 0.4%" from August to September, Teranet's composite "rose 0.1%" on a seasonally adjusted basis, and the NHPI moved from 120.4 to 120.2, a decline of about 0.2%. Which number leads a story depends on the writer's choice of measure and period, and that choice is rarely explained.

What the average price measures, and what CREA says about it

The average sale price is the simplest number: add up the prices of all homes sold through Canadian MLS Systems in the month and divide by the number sold. CREA's release says the national figure was "$666,538 in September 2026, down 1.6% from the same month last year," and that the statistics in the report include "all housing types."

That makes it a measure of spending, not of value. It tells you what the homes that happened to trade that month cost on average. If more of the month's sales were detached houses in Vancouver and fewer were condos in Windsor, the average rises, even if every individual home is worth less than a year earlier. If the mix shifts the other way, the average falls faster than prices.

CREA is direct about the limit. Its release carries a standing note: "CREA cautions that average price information can be useful in establishing trends over time, but does not indicate actual prices in centres comprised of widely divergent neighbourhoods or account for price differential between geographic areas." In other words, the average is a useful long-run trend line and a weak guide to what a particular kind of home is worth.

A worked example of the mix problem

The arithmetic below is hypothetical, built only to show the mechanism. Suppose a market has two kinds of homes: one selling for $500,000 and one for $1,000,000. In the first year, 30% of sales are the expensive type, so the average is $650,000.

Case	Each home type, year over year	Share of sales that are the $1M type	Average sale price	Change in the average
Starting year	n/a	30%	$650,000	n/a
Prices fall 3%, mix tilts to expensive homes	-3% for both types	33%	$645,050	-0.76%
Prices fall 3%, mix tilts to cheaper homes	-3% for both types	27%	$615,950	-5.24%

In both cases every home is worth exactly 3% less. The average reports a decline of under 1% in one and over 5% in the other. A benchmark index that held the home type fixed would report about 3% in both. That is the whole difference between the two kinds of measure, and it is why a gap such as September's 1.6% against 3.4% should prompt a question about which homes sold, not a conclusion that one figure is wrong.

CREA's release does not decompose September's gap, so we do not claim to know how much of it came from regional mix (fewer sales in Ontario, for example) and how much from property type. What the release does say is that the monthly HPI decline "reflected a fresh step down for prices in Ontario where demand softened in September." A shift in where sales happen is exactly the kind of change that moves an average and a benchmark apart.

There is a practical way to use the gap rather than argue about it. When the average falls less than the benchmark, as in September 2026, the simplest reading is that the homes that sold were, on balance, a pricier mix than a year earlier: more sales in expensive regions or property types, fewer in cheaper ones. When the average falls more than the benchmark, the mix has tilted the other way. One month's gap can be noise, so look at several months in a row. A gap that persists in one direction is a signal about who is still buying, which matters to anyone forecasting sales volumes, commissions or land transfer tax revenue, even when it says little about what any particular home is worth.

What the MLS HPI benchmark measures

The MLS Home Price Index was launched in February 2012. CREA's history page says it was developed with Altus Group for six founding partners: the real estate boards of Calgary, Fraser Valley, Greater Montreal, Greater Vancouver and Greater Toronto, and CREA itself. Its central idea is the benchmark home.

CREA's methodology backgrounder describes the benchmark home as "a virtual representation of what a typical, middle-of-the-road home would look like for a particular property type in a particular neighbourhood." It "does not necessarily represent an actual existing home." Each neighbourhood and property type combination has its own set of models.

The benchmark is built from the middle of the market. For a number such as bedrooms, CREA collects the sales of that property type in that neighbourhood over the five years before the models were built or last reviewed, sorts them, and takes the median. Bathrooms, living area and lot size are set the same way. For qualitative features such as exterior, heating fuel and foundation, it takes "the most common value," the mode.

The model then prices that benchmark. CREA's worked example for a two-storey single-family home lists features such as a finished basement, 1,600 square feet of living area, two bathrooms, one fireplace and seven rooms, each multiplied by a coefficient estimated from sales. The products are added, giving 14.00983845 in the example, and the benchmark price is e raised to that sum. We checked the arithmetic: e to the power 14.00983845 is about $1,214,494, which CREA rounds to $1,214,500. CREA notes that the figures are for demonstration only and that coefficients "vary from one month to the next."

What that buys, and what it costs

Because the benchmark home's features are fixed, a change in the index is a change in what buyers pay for the same bundle of features. That is the "apples to apples" comparison CREA describes on its HPI page, where it also says that average or median prices "can change a lot from one month to the next and paint an inaccurate or even unhelpful picture of price values and trends."

There are trade-offs. The benchmark is a model output, so it inherits the model's choices. It is revised: CREA's annual-review FAQ says the HPI "is reviewed annually," with updates "released in December," and that a benchmark home must have "at least 12 sales in the three most recent years" to keep being tracked. When the review changes historical figures, CREA says the updated data "depicts the most accurate home prices levels and trends, and should be used until the next review." A chart saved last year and a chart pulled today may not match, and neither is an error.

What the Teranet-National Bank index measures

Teranet and National Bank take a third approach: watch the same houses sell twice. The methodology page says the index "is estimated by tracking the observed or registered home prices over time," and that "properties with at least two sales are required." Each such "sales pair" gives a price change for one property between two dates, and the index averages those changes across many pairs.

Teranet's own simplified example shows the logic. House A sold for $300,000 in 2005 and $330,000 in 2006, a 10% rise. That implies house B, bought for $250,000 in 2005, was worth $275,000 in 2006; B then sold for $297,000 in 2007, an 8% rise from 2006. House C went from $280,000 in 2006 to $308,000 in 2007, a 10% rise. The 2006 to 2007 index change is the average of B's and C's, 9%. We reproduced each step.

The method filters out sales that would break the "same property" assumption. The methodology lists four exclusions: non-arm's-length sales, a change of property type "for example after renovations," data errors, and properties that trade very often, twice a year or more.

The September 2026 release covers the composite of 11 census metropolitan areas. It reported a 0.1% monthly rise and a 3.5% decline from a year earlier, with Toronto down 6.0%, Vancouver down 5.6%, Hamilton and Halifax each down 3.9%, and Quebec City up 7.7%, Winnipeg up 3.3%, and Edmonton and Montreal each up 1.4%. A national story built from those cities is a story about large markets: the composite is not a measure of the whole country.

Repeat-sales indexes have their own blind spots. New homes have no previous sale, so they enter only when they resell. Renovated homes that change type are excluded. And to the extent the method works from registered prices, which are recorded when a transfer is registered rather than when an offer is signed, the index reflects deals agreed earlier than the month it is labelled with. None of that makes it worse than the HPI. It makes it a different instrument.

What Statistics Canada's New Housing Price Index measures

The NHPI answers a narrower question: what are builders charging for new houses? Statistics Canada describes it as "a monthly series that measures changes over time in the builders' selling prices of new residential houses, where detailed specifications pertaining to each house remain the same between two consecutive months."

Its scope is specific. The survey "covers the following dwelling types: new single homes, semi-detached homes and townhomes (row or garden homes)." It excludes custom-built houses and is limited to houses in 27 census metropolitan areas that represent every province. Prices exclude GST, QST and HST. Builders are asked for their three top-selling models, and when a model is dropped, Statistics Canada matches a replacement and adjusts for quality differences using hedonic methods. Cities are weighted using sales values from CMHC's Market Absorption Survey.

For Canada as a whole, the total (house and land) index stood at 120.2 in September 2026 (December 2016 = 100), against 122.7 in September 2025 and 120.4 in August 2026. The index was 103.4 in January 2020, so even after the past year's decline new-house prices on this measure remain about 16% above where they started 2020.

Two cautions follow from the design. First, it is not a resale measure, and condominium apartments are not among the dwelling types the survey lists, so a condo-heavy market such as downtown Toronto is not what this index describes. Second, the survey prices models that builders can price repeatedly, so it describes production builders, not the custom-home market. Statistics Canada's documentation also notes that the split between land and structure "can contain some level of respondent bias," because builders find that split hard to estimate.

The city spread hidden inside one national number

A national index is an average of very different local markets, and September 2026 shows how wide the spread can be. Within Teranet's 11-city composite, the year-over-year change ran from a 6.0% decline in Toronto to a 7.7% rise in Quebec City, a gap of 13.7 percentage points. Six of the 11 cities posted declines and the others did not. The composite's 3.5% decline is the weighted result of those moves, not a description of any one of them.

The same is true month to month. Teranet's September release lists four cities with increases (Hamilton up 2.5%, Victoria up 2.2%, Vancouver up 0.6% and Ottawa-Gatineau up 0.3%) and six with declines, led by Halifax at 1.6%, while Calgary was stable. A reader in Hamilton and a reader in Halifax saw opposite monthly moves under a national headline of "up 0.1%."

This is the strongest practical argument for local benchmarks over national ones, and it applies to all four measures. CREA says its benchmark homes "are tracked across Canadian neighbourhoods and different types of houses," and it offers an online MLS HPI tool; Statistics Canada's NHPI table carries metropolitan, provincial and regional series alongside the Canada total; Teranet's release tables every one of its 11 cities. If the decision in front of you is local, the national figure is context, and the local series is the evidence.

How to quote a home-price number without misleading anyone

Much of the confusion about Canadian home prices comes from figures quoted without their labels. Five details make a price figure checkable, and a careful writer, analyst or agent can include all of them in one sentence:

• The series name. "Average sale price," "MLS HPI composite benchmark," "Teranet-National Bank composite," or "New Housing Price Index."
• The geography. National, provincial, metropolitan, or a board area.
• The month the data describe, which is not the month it was published. CREA published September 2026 figures on October 16, 2026; Teranet on October 20.
• The comparison period. Month over month or year over year.
• The adjustment. Seasonally adjusted or not.

Compare two ways of reporting the same release. "Home prices fell 3.4%" invites the reader to apply that to their own home and leaves out which index fell. "CREA's national composite MLS HPI was down 3.4% in September 2026 from a year earlier, not seasonally adjusted, while the national average sale price was down 1.6%" is longer, but every number in it can be checked against the release, and it tells the reader that two measures disagreed. When BubbleWatch quotes a price, this is the standard we try to meet.

The same discipline applies to revisions. Because CREA revises HPI history in its annual December review, a benchmark quoted from an old chart may not match the current series. Say when you pulled the figure.

How the four compare

	CREA average price	CREA MLS HPI	Teranet-National Bank HPI	Statistics Canada NHPI
What is priced	Every MLS sale that month	A modelled benchmark home per type and area	The same properties sold at least twice	Unchanged new house models from builders
Holds the home constant?	No	Yes, by model	Yes, by matching sales	Yes, by specification
Housing covered	All types sold on MLS	Types and areas with enough sales	Properties in 11 CMAs, resale	New singles, semis, townhomes in 27 CMAs
Main strength	Shows what buyers actually spent	Clean comparison of a typical home over time	Uses actual repeat transactions	Separates new-build pricing from resale
Main weakness	Moves with the mix of sales	Model output, revised each December	Large-city coverage; closing-date timing	No resale, no condo apartments in its listed types
September 2026, year over year	-1.6%	-3.4%	-3.5%	-2.0%

Reading a headline: a five-point check

1. Which measure? "Prices fell 3%" and "prices fell 1.6%" can describe the same month. Look for the words average, benchmark, HPI, Teranet or new housing.
2. Which period? Monthly and yearly changes routinely point in different directions. In September 2026, Teranet's monthly change was up 0.1% while its yearly change was down 3.5%.
3. Seasonally adjusted or not? CREA's year-over-year comparisons in the September release are stated as not seasonally adjusted; Teranet's monthly change is seasonally adjusted. A spring-to-summer average price jump in unadjusted data is partly just the season.
4. Which geography? A national composite averages very different markets. Teranet's September composite combined a 6.0% decline in Toronto with a 7.7% rise in Quebec City.
5. Which homes? A national average built from all sales, and a new-home index that excludes condo apartments, describe different slices of the housing stock.

If a story uses an average to claim that prices for a type of home rose or fell, ask whether a benchmark measure for that type agrees. If the two disagree, the change in the mix of sales is part of the story.

Which number to use for which question

• "Are prices for a typical home rising or falling?" Use the MLS HPI benchmark for that area and property type, and check it against the Teranet index if your city is one of its 11.
• "How much did buyers spend this month?" Use the average sale price. It is the right number for total dollar volume and for questions about what was bought.
• "Is new construction getting cheaper?" Use the NHPI, and remember it covers ground-oriented houses, not condo towers.
• "What is my house worth?" None of these. National and even city figures describe markets; your own value comes from recent sales of similar homes near you.
• "Is the market in a bubble?" No single index answers that. Prices need to be read against incomes, rents and borrowing costs. Our Canada housing bubble tracker puts the benchmark, average price, inventory and affordability together, and the national housing market dashboard shows the current readings in one place.

How this affects the numbers on BubbleWatch

BubbleWatch uses more than one of these series, and labels matter. The $666,538 figure CREA published for September 2026 is the national average sale price, not the benchmark; we show it as an average price. Our historical price view tracks average sale prices over decades, which is the use CREA itself says average price information is suited to: establishing trends over time. When we compare cities, as in our property tax comparison or income needed to rent, we state which price or rent series each figure comes from.

For how interest rates feed through to these indexes over time, our analysis of mortgage rates and home prices traces the lag, and our review of CREA's 2026 forecast looks at why a forecast for the national average price hides large regional differences.

Sources and method

• CREA national statistics, September 2026 release (published October 16, 2026): average sale price, MLS HPI changes, the average-price caution and the all-housing-types note.
• CREA, MLS Home Price Index, About MLS HPI, MLS HPI Methodology Backgrounder and MLS HPI Annual Review FAQ (2025).
• Teranet-National Bank House Price Index, release of October 20, 2026 and methodology.
• Statistics Canada, Table 18-10-0205-01, New housing price index, monthly (Canada, total house and land) and New Housing Price Index survey documentation.

The NHPI percentage changes, the benchmark-price check and the Teranet example are our arithmetic from the published figures. The two-segment example is hypothetical. This guide explains statistics; it is not investment, appraisal or real estate advice.

Frequently asked questions

What is the MLS Home Price Index?

It is CREA's measure of the price of a typical home. CREA says the MLS HPI uses more than 15 years of MLS System data and statistical models to define a benchmark home, a virtual middle-of-the-road home for a property type in a neighbourhood, and tracks its price. Because the benchmark's features stay the same, the index changes when prices change, not when the mix of homes sold changes.

Is the benchmark price the same as the average price?

No. In September 2026 CREA's national average sale price was $666,538, the average of homes that actually sold. The benchmark is a modelled price for a typical home with fixed features. The two can move in different directions in the same month: that September the average was down 1.6% from a year earlier while the national composite HPI was down 3.4%.

Why does the average home price go up when prices are falling?

Because an average reflects which homes sold, not only what they sold for. If more expensive homes make up a bigger share of sales, the average can rise even when every type of home is cheaper than a year ago. CREA itself cautions that average price information does not indicate actual prices in centres with widely divergent neighbourhoods or account for price differences between areas.

Which home price index is most accurate in Canada?

Each answers a different question, so none is most accurate for every use. The MLS HPI tracks a typical resale home with fixed features. Teranet-National Bank tracks repeat sales of the same properties in 11 large metropolitan areas. Statistics Canada's NHPI tracks builders' prices for unchanged new house models. The average tells you what buyers actually spent that month. Use the one that matches your question and say which one you used.

Why did my neighbourhood's benchmark price change for past months?

CREA reviews the MLS HPI every year, with updates released in December. Its annual-review FAQ says benchmark homes and coverage are updated, a benchmark home needs at least 12 sales in the three most recent years to stay tracked, and the updated data should be used until the next review. Earlier figures were representative of the information available at the time.

Does the New Housing Price Index include condos?

Statistics Canada says the survey covers new single homes, semi-detached homes and townhomes, and excludes custom-built houses. Condominium apartments are not among the dwelling types it lists, and the index excludes GST, QST and HST. It measures what builders charge for the same house model from one month to the next, so it is a new-construction price signal, not a resale one.

Which number should I use to judge my own home's value?

None of the national figures. They describe markets, not houses. For a specific home, the useful evidence is recent sales of similar homes nearby and, where available, the local benchmark for that property type. A national decline of 3.4% does not mean a given Toronto condo or Calgary detached house fell by that amount.