Statistics Canada's own monthly GDP estimates, across eleven years of release vintages
Statistics Canada publishes a real-time table that keeps every vintage of its monthly GDP by industry estimates -- 137 releases between March 2015 and June 2026. Measured across the 130 reference months whose first estimate is observable in that record, no first estimate of 0.20 percentage points or larger has later reversed sign, while roughly a third of first estimates smaller than 0.10 percentage points have. This entry holds the vintage lineage that base rate is measured on, traced through the reference month with the longest observable history: January 2015, whose published growth rate settled over its first twenty months and then changed on just six further dates in nine and a half years -- every one of them an annual or comprehensive revision release, and none in between.
What the updated record shows
Recomputed for the October 4, 2026 review. Source captured 2026-10-04; table metadata released September 14, 2026. The dated study below is retained as a historical reading.
The issuer’s record now contains 139 releases, ending August 28, 2026. It follows 138 first observable monthly estimates from January 2015 through June 2026; 216 earlier months are excluded because their first publication predates the archive.
Outside March–August 2020, the first monthly growth estimate moved by an average of 0.142 percentage points in absolute terms by the latest recorded release. Its sign changed in 18 of 132 months (13.6%). Small initial movements remain the fragile ones.
| Absolute first growth estimate | Months | Sign reversals | Share |
|---|---|---|---|
| Below 0.10 pp | 42 | 15 | 35.7% |
| 0.10 to below 0.20 pp | 23 | 3 | 13.0% |
| 0.20 pp or more | 67 | 0 | 0.0% |
Zero reversals among the larger first estimates is a result in this sample, not a guarantee. Recent months have had less time to revise. The record measures the first estimate in this table; an earlier GDP advance estimate is outside its scope. Growth rates are used because changes in the chained-dollar base make levels unsuitable for this comparison.
| Later releases | Months | Mean absolute revision (pp) | 90th percentile (pp) |
|---|---|---|---|
| 1 | 131 | 0.0479 | 0.0986 |
| 2 | 130 | 0.0732 | 0.1588 |
| 3 | 129 | 0.0780 | 0.1679 |
| 6 | 126 | 0.0921 | 0.1957 |
| 12 | 120 | 0.1172 | 0.2121 |
| 24 | 108 | 0.1328 | 0.2702 |
| 36 | 96 | 0.1380 | 0.2952 |
The signed average revision is slightly positive. Whether it differs statistically from zero depends on the treatment of serial correlation and the maturity of the sample. The full calculation retains those checks, the fixed-cohort horizon comparison and the pandemic-inclusive results; the summary does not establish a dependable upward bias.
Download the complete dated calculation (JSON) · Statistics Canada source table · Statistics Canada Open Licence. Source: Statistics Canada; calculations by Countability. This is a record of past revisions, not a forecast.
A Resolution record follows one institution's figures for a defined quantity. It may contain projections, estimates or measured observations. Each version held here keeps its publication date and source, so a newer figure does not erase the earlier one. What is a Resolution record? →
Maintained record. This page preserves successive versions of an official record. It makes no projection and awaits no forecast verdict. Revisions remain visible, and provisional figures retain that status.
The lineage
| Vintage | Issuer | Published | Position | Source |
|---|---|---|---|---|
| 2015-03-31 | Statistics Canada | 2015-03-31 | superseded | Source → · accessed 2026-08-11 |
| 2016-11-30 | Statistics Canada | 2016-11-30 | superseded | Source → · accessed 2026-08-11 |
| 2017-12-01 | Statistics Canada | 2017-12-01 | superseded | Source → · accessed 2026-08-11 |
| 2018-11-30 | Statistics Canada | 2018-11-30 | superseded | Source → · accessed 2026-08-11 |
| 2019-03-01 | Statistics Canada | 2019-03-01 | superseded | Source → · accessed 2026-08-11 |
| 2019-11-29 | Statistics Canada | 2019-11-29 | superseded | Source → · accessed 2026-08-11 |
| 2023-11-30 | Statistics Canada | 2023-11-30 | superseded | Source → · accessed 2026-08-11 |
| 2025-11-28 | Statistics Canada | 2025-11-28 | superseded | Source → · accessed 2026-08-11 |
| 2026-06-30 | Statistics Canada | 2026-06-30 | current | Source → · accessed 2026-08-11 |
| Vintage | Period | Value | Role | Assurance | Basis |
|---|---|---|---|---|---|
| 2015-03-31 | January 2015 | -0.104 percent (month-over-month change) | preliminary observation | — | — |
| 2016-11-30 | January 2015 | -0.373 percent (month-over-month change) | revised observation | — | — |
| 2017-12-01 | January 2015 | -0.367 percent (month-over-month change) | revised observation | — | — |
| 2018-11-30 | January 2015 | -0.414 percent (month-over-month change) | revised observation | — | — |
| 2019-03-01 | January 2015 | -0.410 percent (month-over-month change) | revised observation | — | — |
| 2019-11-29 | January 2015 | -0.417 percent (month-over-month change) | revised observation | — | — |
| 2023-11-30 | January 2015 | -0.415 percent (month-over-month change) | revised observation | — | — |
| 2025-11-28 | January 2015 | -0.416 percent (month-over-month change) | revised observation | — | — |
| 2026-06-30 | January 2015 | -0.416 percent (month-over-month change) | revised observation | — | — |
Movement between vintages
This record shows what an institution published and how its figures compare across versions or with later outcomes. It does not rank institutions or grade their forecasting ability.
| Comparison | Compared vintages | Movement | Basis |
|---|---|---|---|
| January 2015, release of 2015-03-31 vs release of 2016-11-30 | 2015-03-31 (preliminary observation -0.104 percent (month-over-month change)) vs. 2016-11-30 (revised observation -0.373 percent (month-over-month change)) | -0.269 percent (month-over-month change) | same issuer |
| January 2015, release of 2016-11-30 vs release of 2017-12-01 | 2016-11-30 (revised observation -0.373 percent (month-over-month change)) vs. 2017-12-01 (revised observation -0.367 percent (month-over-month change)) | +0.005 percent (month-over-month change) | same issuer |
| January 2015, release of 2017-12-01 vs release of 2018-11-30 | 2017-12-01 (revised observation -0.367 percent (month-over-month change)) vs. 2018-11-30 (revised observation -0.414 percent (month-over-month change)) | -0.046 percent (month-over-month change) | same issuer |
| January 2015, release of 2018-11-30 vs release of 2019-03-01 | 2018-11-30 (revised observation -0.414 percent (month-over-month change)) vs. 2019-03-01 (revised observation -0.410 percent (month-over-month change)) | +0.004 percent (month-over-month change) | same issuer |
| January 2015, release of 2019-03-01 vs release of 2019-11-29 | 2019-03-01 (revised observation -0.410 percent (month-over-month change)) vs. 2019-11-29 (revised observation -0.417 percent (month-over-month change)) | -0.007 percent (month-over-month change) | same issuer |
| January 2015, release of 2019-11-29 vs release of 2023-11-30 | 2019-11-29 (revised observation -0.417 percent (month-over-month change)) vs. 2023-11-30 (revised observation -0.415 percent (month-over-month change)) | +0.002 percent (month-over-month change) | same issuer |
| January 2015, release of 2023-11-30 vs release of 2025-11-28 | 2023-11-30 (revised observation -0.415 percent (month-over-month change)) vs. 2025-11-28 (revised observation -0.416 percent (month-over-month change)) | -0.001 percent (month-over-month change) | same issuer |
| January 2015, release of 2025-11-28 vs release of 2026-06-30 | 2025-11-28 (revised observation -0.416 percent (month-over-month change)) vs. 2026-06-30 (revised observation -0.416 percent (month-over-month change)) | +0.000 percent (month-over-month change) | same issuer |
| January 2015, first published estimate vs latest release on this axis | 2015-03-31 (preliminary observation -0.104 percent (month-over-month change)) vs. 2026-06-30 (revised observation -0.416 percent (month-over-month change)) | -0.312 percent (month-over-month change) | same issuer |
Evidence qualifications (10)
- This entry compares Statistics Canada against itself: one table, one reference month, the issuer's own successive releases of the same real-time cube. It is a revision record, not a disagreement between institutions, and not a judgment about Statistics Canada. In the issuer's own words about this table: "The revisions in the real-time data tables are not corrections to errors. They represent a normal step in the statistical process, in which statistical agencies produce new vintages of higher quality data as new information becomes available." Nothing here is offered as a measure of the issuer's quality, in either direction.
- "First estimate" means the first estimate for a month recorded in this table's vintage record. It is not the first public number for the month. Statistics Canada publishes an advance estimate of real GDP by industry for the following month inside each monthly Daily release -- roughly a month before that month's official first estimate reaches this table. Verbatim, from the release of 2026-06-30: "Advance information indicates that real GDP by industry increased 0.1% in May." That same release went on to state that, owing to its preliminary nature, the advance estimate would be updated with the official May figure at the end of the following month. This record holds no vintage past 2026-06-30 and does not assert what that later release contained. The advance estimate and its own revision are outside this record and are not measured here.
- The sign-reversal figures in the summary are a base rate over an observed set of 130 reference months, not a forecast and not a probability about any named future month. In full, excluding the six months of the 2020 pandemic window, which are held apart rather than averaged in: of 42 months whose first estimate was smaller than 0.10 percentage points, 15 later read the other way; of 23 between 0.10 and 0.20, three did; of 65 at 0.20 or above, none did. Including the pandemic window, that last figure is none of 71. Restricting instead to months that have since had at least 24 further releases -- the months whose revision is largely done -- it is none of 54 excluding the pandemic window and none of 60 including it. Each of those four counts is on its own base, and the bases are not interchangeable. All eighteen reversals in the record came from first estimates under 0.20 percentage points.
- January 2015 is shown here because it is the reference month with the longest observable vintage history in this table, not because of the size of its revision. That size should be read with the distribution beside it: this month's estimate moved 0.31 percentage points in total, which sits at about the 88th percentile of the 130 months measured, whose mean absolute movement is 0.14 percentage points. It is a long record, not a typical one.
- The lineage below lists the first estimate, the last release of the unbroken run of releases that each changed the figure, every later release at which the figure changed, and the most recent release held. Releases at which the figure did not change are not listed, and there are many: between the releases of 2019-11-29 and 2023-11-30, this month's published growth rate was carried unchanged through 47 successive releases. That pattern is the entry's second finding -- revision to this table arrives on a calendar rather than accumulating smoothly. Across the whole axis, twelve of 136 release-to-release comparisons restate more than a tenth of the history the two releases share; eleven of those twelve fall in late November or early December, and one in March. Two things about the six changes that follow the settling run must be read together. Each is small: in the order they occur they are +0.0055, -0.0463, +0.0041, -0.0073, +0.0020 and -0.0010 percentage points, so every one of them is smaller than 0.05 percentage points and the largest is under a twentieth of a point. And each falls on an annual or comprehensive revision release rather than between them. Neither half is the finding on its own: what remains after a month settles is small AND it arrives on the issuer's revision calendar. The lineage and its movement column below are printed to three decimals for that reason -- at one decimal all six of these changes print as zero and eight of the nine lineage points print as the same number, and the reader is shown a table of movements that looks empty. Those printed figures are the stored values rounded to three decimals, so the first of the six, +0.0055, prints as +0.005.
- This entry is stated in month-over-month percentage change, not in dollar levels, and deliberately. Statistics Canada rebased this series twice inside the vintage record -- the releases of 2018-11-30 and 2023-11-30 moved every historical level at once, by about 9% and 5%. A level-based comparison across those releases would report a rescaling as though it were a restatement of the economy. Month-over-month change survives both, which is why it is the unit used here.
- A companion entry in this register, "Statistics Canada's own December 2025 real GDP estimate, across four release vintages", holds four consecutive vintages of one recent month in dollar levels. The two entries rest on separate archive captures of the same table, taken for different questions, and they do not restate each other: that one preserves a short, recent, level-based path for a single month, while this one traces one month across eleven years in growth terms and reports the distribution measured over 130 months. Where they appear to differ in unit, that is the reason.
- Two further bounds on what is measured. The record begins with the release of 2015-03-03, which already contained every month through December 2014; those 216 earlier months have their true first estimates outside the record and are excluded rather than approximated. And the most recent months cannot yet show their full revision, so they contribute to short spans only.
- One further result, stated operationally and no further. Regressed across the 130 months, the movement a month's estimate goes on to receive is not distinguishable from zero as a function of the first estimate itself (slope -0.08, t -1.46), while it does move with the eventual value (slope +0.28, t +6.04). The practical reading is that there is no basis in this record for adjusting a first estimate using the estimate itself; the honest treatment of revision is to widen the interval around it, not to shift it.
- This entry is filed under the fiscal domain, the closest existing one; it concerns Canada's aggregate monthly GDP, not a fiscal-anchor compliance measure specifically, and no new domain is created to hold it more precisely.
Related on this site
- Big Question: How often do official forecasts come true? →
- Domain: Federal Fiscal Capacity →
- More Resolution lineages →
- Sources & Method →
This record shows what an institution published and how its figures compare across versions or with later outcomes. It does not rank institutions or grade their forecasting ability.
This is the current public record for this lineage. See Work with Countability for a commissioned projection-resolution question.
Work with Countability