Accuracy

Two kinds of population data exist. Official statistics know how many people of each kind live in each place. Survey research knows how age, income, education, values and behaviour go together in real people. Neither knows what the other knows. Humanity is one population that reproduces both at once.

Add up any region and you get the figure the national statistical office published. Look inside a person and you find a combination of age, income, education and values drawn from how real people combine them, not averages assigned one at a time. Every country is checked against the figures it was built to reproduce before it is served.

What a number can carry

Counts. The number of people in every served region, by age and sex, matches the official population figures. A country that misses is built and measured, but not served until it meets the bar.

Combinations. Values and attitudes are drawn together, and together with a person's age, income and education, so a high-income, low-education, high-trust person is a real kind of person rather than three averages stacked.

Consistency. Nobody is lost and nobody is counted twice. Households add up to people, spend adds up to budget, and a market sized by households agrees with the same market sized by people.

Place. Every person lives in a specific neighbourhood inside a specific municipality, so a local question gets a local answer rather than a national average worn locally.

Money. Income, spending and emissions follow the region a household lives in, in its own currency. Bucharest earns nearly twice the Romanian average, and its households know it.

Checked against data it never saw

The checks that matter are the ones the model could not have learned. Some published figures are kept out of the build and used only to test it: small-area statistics from national offices, and for attitudes, Eurobarometer. How those tests work is on Validation.

Every region of every country also runs through a battery of automated checks before release, and the results are kept, including the failures. The lineage records what each release could do.

Where we ask for care

Depth. Regional and national figures are precise. One square kilometre with five filters stacked is an estimate, and should be read as one. No count or breakdown covers fewer than 100 people.

Money across borders. Money is kept in local currency. Ranks, quartiles and within-country comparisons are sound. Compare across countries by income quartile or purchasing power, not by converted euro amounts.

Children. Everyone under 15 is in the population, in their household, in every count. Their profile is shorter, because survey evidence on values and behaviour starts at 15.

Attitudes at the edges. Measured attitudinal structure covers most of Europe. Where a country sits outside the surveys' reach it carries a pan-European structure adjusted to its own demographics, and the data flags it.

Time. The reference year is 2024. Projection is built and will ship with a backtest, not before.

Which countries are built, and which are not yet, is on What's inside. Nothing in the population traces back to a real person; What's inside has the short version and the privacy policy the long one.

More detail

How the population is constructed is shared under agreement for procurement, due diligence and audits. Write to hello@hiasynth.co and say what the review needs to cover.