Where Canadian Manufacturing Output Actually Went
Every mining page the front page indexes is built from one annual survey, and Statistics Canada files that survey under the subject Manufacturing. That is not an interpretation of the table. It is written in the metadata file that ships inside the same archive as the data, immediately below the block naming the survey, under a heading giving the subject code and the subject name.
Read the product list and the filing stops looking like a clerical accident. Lime is a member of that table, with quicklime and hydrated lime beneath it. Clay products sit beside clay. Titanium dioxide has its own line, and so does gold and silver bullion, crude and doré. A kiln makes the first of those and a refinery the last, and none comes out of the ground in the state the table names. The survey follows the material a short way past the pit and stops, and the classification follows what the survey does rather than what its title advertises.
None of which puts an output figure on this section. Nothing in the cache behind these pages establishes the output of a single Canadian factory in a single year, and that is still the position the section page states. What the cache does hold is three federal documents that each draw a boundary around manufacturing, and they draw it in three different places. Where the boundary sits turns out to be more of the answer to the question in the title than any quantity would be.
The department that annexes three manufacturing subsectors
Natural Resources Canada defines its own subject in a list of industry codes. The minerals and metals sector, on that page, means mining and quarrying other than oil and gas, two support activity codes covering contract drilling and exploration services, and then three more: non-metallic mineral product manufacturing, primary metal manufacturing, and fabricated metal product manufacturing.
Those last three name themselves. Fabricated metal products are cutlery, stampings, wire and pipe, made in factories by people who would not describe themselves as working in mining. A department whose remit is minerals has drawn its sector boundary around them, openly, in a definition anybody can read.
That is a legitimate thing for a department to do. An account of the minerals economy that stopped at the mine gate would be missing most of it. But a reader who opens a departmental minerals page for context on industrial output is reading a document that has already moved three manufacturing subsectors onto the mining side of the line.
The same page draws the line a second time
Further down, the page explains where its economic indicators come from, and the explanation quietly replaces the definition above it. The gross domestic product and employment values rest on the Natural Resources Satellite Account, which the page describes as going beyond an industry perspective in order to identify natural resource activities and products within the economy. A second list follows, for those indicators only.
The second list is not built of industry codes at all. It is built of activities. Extraction, and services for mining and exploration. Primary metallic products such as aluminium ingots, iron and steel basic shapes and refined metals. Primary non-metallic products such as glass and cement. Then a downstream group: secondary and tertiary metal products such as pipe, cutlery and forged and stamped goods, miscellaneous metal products including motor vehicle metal stamping, and services such as coating, engraving and heat treating.
Two definitions, one page, and not the same shape. One is a set of industries. The other is a set of activities that can occur inside industries the first list never mentions. The page is not careless about it and says which indicators sit on which. But the consequence for anybody quoting it is unforgiving: two figures from the same document can measure two subjects, and nothing in the figures will announce the change.
That is the first thing worth distrusting in a confident account of where output went. Rarely the arithmetic. Usually the subject, changed silently between one paragraph and the next.
Ten headings, and not one of them is a factory
The federal list of project types that pull a proposal into an impact assessment is organised under headings, and the headings are quicker to read than the items. National parks and protected areas. Defence. Mines and metal mills. Nuclear facilities. Oil, gas and other fossil fuels. Electrical transmission lines and pipelines. Renewable energy. Transport. Hazardous waste. Water projects.
No heading names manufacturing, and the nearest item to it is measured from the wrong end. A new metal mill other than a uranium mill is caught at a stated ore input capacity. The expansion item that follows is written on total ore input capacity after the work. Both ask what the building consumes. Neither asks, anywhere, what it produces.
The mill turns up a third time in a different character. Section 1 defines the area of mining operations as the ground occupied by the open pit or underground workings, the mill complex, and the storage areas for overburden, waste rock, tailings and ore. So one regulation treats a processing plant as part of a mine's footprint when it sizes an expansion, and as an item standing alone when it decides whether a project enters the system. The permitting page states the capacities at which those items bite. The plant holds three positions in one instrument and not one of them is about output.
There is exactly one place in the whole Schedule where an ordinary factory is caught, and it is a place rather than a size. Inside a wildlife area, a migratory bird sanctuary or a protected marine area, the construction, operation, decommissioning and abandonment of a new industrial facility is a designated project. No capacity qualifies it. The phrase industrial facility occurs once in the Schedule, carries no threshold, and is not among the terms section 1 defines. Everywhere else in the country the same building is invisible to the instrument.
A project registry is therefore not an industrial census. Anyone reaching for it as the only federal enumeration of large facilities comes away with a map of protected areas, corridors and mines with a few mills in it.
The frame under the one table that does exist moved
The mineral table has a documented discontinuity in it, and it is in the notes rather than in any cell. An establishment extracting sand, gravel or stone as a secondary activity is dropped from the survey when its production value falls below a stated amount, and that amount was raised starting with reference year 2024. Quebec applied no exclusion at all in the earlier years, and from the same reference year excludes establishments with one employee or fewer instead.
Two further changes carry a date in the same month. Region level data was discontinued as of March 2024. The regional rows carry nothing in any year of the extract these pages were built from. As of that month, dollar values for every reference period are presented in thousands, a restatement applied backwards across the whole series.
A fourth caution in the same file covers the newest end of the series. Respondents supply the survey year together with forecast values for the year after it, and the note warns that revisions to preliminary estimates can move further than revisions to historical data.
None of this is a scandal. Agencies revise frames, and this one wrote down what it changed and when. The point is about reading. A series with a frame change inside it shows a movement at the change whether or not anything in the world moved, and in the published numbers that movement is indistinguishable from a real one. Two adjacent years on either side of it are two collection rules.
Where a national number is actually collected
The table is national and is not collected by one office. A note records that the Institut de la Statistique du Québec collects and processes the mineral production data from Quebec respondents. So the Quebec line that the Quebec page works through arrives through a provincial agency, on an exclusion rule that is not the one applied elsewhere.
A second note runs the other way. The quantities for secondary aggregates in Newfoundland and Labrador, being stone, sand and gravel, are supplied to Statistics Canada by a provincial department, and the note names that department under a title the province's own current page no longer uses. The Newfoundland and Labrador page works through why a name has to be read off the organisation's own document. What is worth noting is where this one sits, in a live footnote on a current federal table, which is close to the last place anybody thinks to check a name.
What a claim about a shift has to carry
Three questions, the same three whatever the industry. Which collection produced the numbers. Which definition of the sector was in force when each was quoted. And whether the frame held constant across the years compared. An account that answers none of them is a difference between two figures whose subjects were never shown to be the same, and a difference inherits every discontinuity in both of its terms.
The section page already names the records it would take before a number could stand here. This piece adds one condition on the first of them, and it is the one easiest to skip. Whatever table arrives has to arrive with the metadata file beside it. Everything load bearing above came out of a metadata file rather than a data cell: the subject classification, the frame change, the second collecting agency, the provincial supplier. None of it shows in a chart or a downloaded column of values.
So the position is not that the numbers are missing. It is that three federal documents here each cut a boundary around manufacturing, one of them cuts two, and every cut is right for the job it was made for. A survey frame, a departmental account and a trigger list are under no obligation to agree. What none of them licenses is a sentence carrying a quantity across one of those cuts without saying so, which is what to look for in the next confident account of where Canadian output went: not whether the number is big, but whether the writer noticed there was more than one edge in the way.