The Aggregate

When your own experience contradicts the official statistic, notice which one gets the authority. The number is “the data.” Your life is “anecdote.” That asymmetry is the case. The aggregate is not a lie — it is built from real work — but it is built from definitional choices and coverage gaps, and it is weakest exactly where life is hardest to measure: precarious, informal, churning work. This Edition is not the claim that statistics are fake, or that one person’s story beats a dataset. It is narrower and harder: the authority of the number is laundered — the precision it projects stands in for a certainty it does not have, and that borrowed certainty is then used to overrule the very people it was meant to describe.

§01 — The case for the number

Grant the defense first, in full, because it is strong. Official statistics are a genuine public good. You cannot govern, fund a hospital, draw an electoral map, or hold power to account without a shared, independent count — and the alternative to public statistics is not no numbers but worse ones: figures supplied by whoever has an interest in the answer. A full enumeration, like the census head-count, is genuinely reliable; you cannot hide a population. None of what follows is an argument against measuring. It is an argument against a specific move that measuring makes possible — and the move only works because the underlying enterprise is trusted.

§02 — Measured worst where it matters most

The sectors hardest to count are precisely the precarious ones. Temporary and staffing-agency work — is the employer the agency or the client? — gig and platform work, and the cash-and-informal economy all sit at the edge of what a survey can see. Statistics Canada estimates an underground economy of $72.4 billion in 2023, about 2.5% of GDP — and publishes it explicitly as an upper-bound estimate, because the thing by definition resists direct observation. That is not a scandal; it is honest. But read the consequence: the people whose working lives are most precarious are the people the official numbers describe worst. The aggregate’s precision is highest where the stakes are lowest, and thinnest where they are highest — a bias baked not into anyone’s intent but into what is easy to count.

Counter: ask what a number leaves out before you ask what it says. The gap is not random; it falls on the same people twice.

§03 — The definitions decide the answer

Before a single figure is published, definitions have already chosen what it will say — the definitional dodge at the level of a methodology note. In the Labour Force Survey, you are counted as “employed” if you did one hour of paid work in the reference week — one hour, regardless of how far your hours have fallen. And you are counted as “unemployed” only if you actively looked for work; a person who has stopped looking because they believe no suitable work exists — a “discouraged searcher” — drops out of the labour force entirely and out of the unemployment rate. None of this is hidden, and each choice is defensible. But the choices are not neutral: they make the headline number smaller and calmer than the experience underneath it. Statistics Canada knows this, which is why it also publishes R8, the broad measure that counts the underemployed and the discouraged — the rate the headline isn’t.

Counter: a definition is a decision. Ask who is one notch outside it — the person at thirty hours lost, the one who stopped looking — and why the headline is drawn to leave them out.

§04 — The number depends on how you ask

The clearest proof that a statistic is a construction and not a simple fact is to measure the same thing three ways and watch it move. Take informal and gig work in Canada:

Same country, same phenomenon, three official-grade numbers spanning two orders of magnitude. The “fact” is a function of the instrument that produced it. There is no neutral vantage from which one of these is simply the answer — which is exactly the thing the single published headline is built to obscure.

Counter: before citing a rate, name the instrument. A number without its method is a verdict without its evidence.

§05 — The lived, called anecdote

Here is where the measurement becomes a power move. The aggregate is true on average and false for the person — both at once — but only one of those two truths is allowed to call itself “the data.” When a worker who has lived an industry says the official figure does not match the floor they stand on — as a staffing insider once noted of revenue numbers that did not square with the labour market he worked in — the reply writes itself: that’s anecdotal; the data says otherwise. The headline’s borrowed certainty is spent to convert a person’s direct knowledge into noise. But the lived experience is not the error term. It is the territory the map was drawn to represent — and when the map and the territory disagree, the honest question is which one was simplified, not which one is allowed to speak.

Counter: “the data says” is a claim about a method, not a refutation of a life. Ask what the method could not see.

§06 — The integration stamp, in a lab coat

Strip it to the move and it is the fourth one in the grammar. We measured it → therefore we know → therefore your experience is mistaken. The running of a procedure — sampling, weighting, publishing — offered as the verdict on the procedure’s own subject. It is the integration stamp Case 16 documents, wearing the most trusted uniform there is: the lab coat, the official seal, the decimal place. And it completes the arc the last three cases trace — the census sold as help, the data sold as protected, and here the figure sold as true, each promise resting on an authority larger than the thing it actually delivers. The measurement is real. The certainty stamped onto it is the wash.

Counter: the procedure ran; that is not the same as the answer being settled. Separate “we measured” from “we know,” and the authority lets go of its grip.

§07 — What this does and does not claim

It does not claim official statistics are fabricated, worthless, or produced in bad faith — §01 makes the opposite case, and Statistics Canada’s own R8 and upper-bound estimates are evidence of an agency naming its limits in public. It does not claim a single person’s experience always outranks a dataset; one life is a sample of one, and the aggregate often is the better guide.

It does claim this: every published figure embeds definitional choices and coverage gaps; those gaps fall hardest on the precarious, who are therefore measured worst; the same phenomenon yields wildly different numbers depending on the instrument; and the authority of “the official number” is routinely laundered into a certainty it has not earned and then used to overrule the lived. The number is a tool. The laundering is the sentence that begins “the data says,” and ends by telling a person that what they are living is not happening.

§08 — Where this sits in the volume

This closes a quartet on measurement. Vol. I · Case 10 · The Index named measurement as a laundering layer; Case 22 read the census as a compulsion sold as help; Case 23 read the data as a protection sold past its own re-identification; and this case reads the figure as a construction sold as a fact. The thread through all four is a single inversion: the apparatus built to see the citizen becomes the authority that overrules them. Read through the sentence-level grammar in The Grammar of the Con; the de-nominalization counter here is three questions — what was not counted, who chose the definition, and whose experience was filed as noise.

The aggregate is true on average and false for you. Only one of those gets called “the data.”