Skip to content
SVERIGEFAKTA.com

What happened to Sweden?A data-driven storyChapter 6 of 6Method chapter

Methods, sources and uncertainty

How do we know?

The methods behind the numbers — how public statistics are produced, what they can show and what they cannot.

read time
11 min read
charts
3
primary sources
12

After this chapter you will understand

  • Which authorities produce Sweden's official statistics and what each of them covers.
  • Why definitions often explain more of the debate's disagreement than the numbers themselves.
  • The difference between correlation, causation and confounders — with simple examples.
  • The four evidence levels Sverigefakta uses and how they are applied.
  • Where knowledge still runs out — and how Sverigefakta discloses uncertainty.

Section 1

Where does the data come from?

Almost all Swedish social statistics come from a small number of agencies and international bodies. Each is responsible for its domain, its methodology and its definitions. Sverigefakta reads from these primary sources — not from second-hand summaries.

  • SCB

    Sweden's official statistics — population, work, income, education.

  • Brå

    Crime statistics, reported crimes, cleared crimes and special studies.

  • Polismyndigheten

    Shootings, bombings, lethal violence, situation reports on vulnerable areas.

  • Skolverket

    School results, grades, eligibility, Swedish PISA follow-up.

  • Socialstyrelsen

    Health, social assistance, causes of death, care registries.

  • Migrationsverket

    Applications, granted residence permits, enforcement.

  • Eurostat

    Comparable EU statistics on migration, work, income and demographics.

  • OECD

    International comparisons — PISA, Migration Outlook, economy.

  • Konjunkturinstitutet

    Economic outlooks, fiscal policy analysis, long-term scenarios.

  • ESO

    The Expert Group for Public Economics — reports on welfare and migration.

  • V-Dem

    Varieties of Democracy — international democracy indices, University of Gothenburg.

Statistics are built in four steps: authority → data → processing → analysis. Sverigefakta sits at the end of the chain, as a presentation layer over what is already public.

  1. Step 1

    Authority

    SCB, Brå, Police, School Agency …

  2. Step 2

    Data

    Registers, surveys, measurements

  3. Step 3

    Processing

    Classification, quality, publication

  4. Step 4

    Analysis

    Researchers, agencies, ESO, OECD

  5. Step 5

    Sverigefakta

    Source, definition, update date

Observation

Sweden's official statistics are governed by the Official Statistics Act (SFS 2001:99). Publication is independent of political priorities.

Section 2

Definitions matter

The same word can mean different things depending on who is measuring. The difference is rarely rhetorical — it changes the number. Below are six concepts where the definition often causes confusion in the public debate.

Foreign background
SCB: a person who is either foreign-born or born in Sweden to two foreign-born parents. A narrower definition (foreign-born only) yields a lower figure.
Unemployed
SCB LFS uses the ILO definition: actively seeking, available, without work. The Public Employment Service counts registered jobseekers, giving a different figure.
Employed
At least one hour of work during the reference week. A low threshold — a person can be employed without being self-sufficient.
Fatal shooting
Police definition since 2017: death caused by a firearm shot in a conflict or criminal setting. No comparable official series exists before 2017.
Bombing
Brå counts completed detonations. The Police also report attempts and preparation — these series differ.
Self-sufficiency
No single official definition. ESO uses labour income above a set base level; others use absence of social assistance. The figures differ markedly.
Research

Meta-analyses of migration research show that a significant share of the differences between studies is driven by definitional choices — not by disagreement about reality. Source: OECD Migration Outlook.

Section 3

Statistics show patterns

Statistics describe groups and trends — not individual people. A 3 percent risk means three in a hundred are affected, not that a specific person will be. Statistics also say nothing about motive, intent or the future.

  • Individual ≠ group. Group averages hide large within-group variation.
  • Trend ≠ level. A falling trend can still start from a high level.
  • History ≠ future. Trends can break — through politics, technology or crises.
  • Motive is not in registries. Register data measures what, not why.

Section 4 — the most important part

Correlation is not causation

Two things moving together does not mean one causes the other. Ice-cream sales and drownings correlate — but ice cream does not drown anyone. The underlying factor is summer. In research it is called a confounder.

Correlation

Two variables move together statistically. Could be chance, could be causal, could reflect a third variable.

Causation

One variable causes the other. Requires experiments, natural experiments or very careful control of alternative explanations.

Confounder

A third variable that influences both observed variables and produces a spurious correlation between them.

Research

Rigorous research uses instrumental variables, difference-in-differences and natural experiments to separate correlation from causation. Source: IFAU methodology manuals.

Section 5

Why do researchers sometimes reach different conclusions?

Studies sometimes reach different conclusions. That does not automatically mean someone is wrong. Five common reasons:

  • Different samples. A study of big cities gives different numbers than one of the whole country.
  • Different time periods. 2010–2015 and 2018–2024 are not the same Sweden.
  • Different models. Linear regression or causal model — same data, different answers.
  • Different definitions. See Section 2 — the content of concepts drives the numbers.
  • Different statistical methods. Controls for age, sex and education shrink or amplify associations.
Observation

When results diverge, the divergence itself is information: it says the phenomenon is sensitive to methodological choices — and that caution is required in interpretation.

Section 6

How strong is the knowledge?

Sverigefakta uses four evidence levels throughout the series. They reflect how consistent the research is — not how confident any single researcher happens to be.

  • Very strong evidence

    Several independent studies, different methods and countries point the same way. Robust over time.

    Examples: Education affects income. Age structure affects the dependency ratio.

  • Strong evidence

    Established research with a clear direction. Occasional divergent results but the main pattern holds.

    Examples: First job within five years predicts long-run establishment. Refugee cohorts have lower employment in years 1–8.

  • Moderate evidence

    Studies exist but results vary by method, sample and time period. Direction likely, magnitude uncertain.

    Examples: Segregation's effect on school outcomes after socioeconomic controls. Lifetime fiscal net effect.

  • Uncertain evidence

    Limited evidence, conflicting results or very long time horizons. Draw conclusions cautiously.

    Examples: Effects across three generations. The future labour market after AI. Long-run political stability.

Section 7

What do we still not know?

An honest method chapter must also state what research cannot answer. Five questions where evidence today is weak or the time horizon is too long.

  • What will the third generation look like?

    For several large immigrant groups the second generation is still young. Adult outcomes will only be visible in 15–25 years.

  • How will AI affect the labour market?

    Both low- and high-skilled jobs are affected. Which occupations disappear or emerge is still open.

  • How will Sweden's demographics evolve 2035–2050?

    Projections rest on assumptions about migration, fertility and life expectancy — small changes produce large long-run differences.

  • What long-run fiscal effect will today's immigration have?

    Studies span decades and depend on retirement age, employment and public spending. Magnitude and sign vary.

  • How stable will institutional capacity remain?

    Courts, schools, healthcare and social services face demographic and fiscal pressure. The outcome is not predetermined.

Political interpretation

Where research is uncertain, an interpretive space opens up. Sverigefakta tries to fill that space with open disclosure — not with over-confident conclusions.

Section 8

How Sverigefakta works

Sverigefakta is a documentation platform. We do not push a political line. We work by six open principles:

  1. Principle 1

    Public primary sources

    We work from SCB, Brå, the Police, the National Agency for Education, the Board of Health, the Migration Agency, Eurostat and OECD. Not from summaries of summaries.

  2. Principle 2

    Source citation on every figure

    No figure without a link to the primary source. Missing sources are clearly flagged.

  3. Principle 3

    Method descriptions

    We describe how calculations were made, which definitions were used and which price-level adjustments were applied.

  4. Principle 4

    Version history

    Every page shows a visible update date. Corrections are documented separately in /changelog.

  5. Principle 5

    Corrections are the norm

    When we find errors we correct them openly, with date and description. Acknowledging errors is part of transparency.

  6. Principle 6

    Uncertainty is disclosed

    Where evidence is weak, methods differ or data is missing — we say so.

The combination of open sources, version history and public corrections means anyone can trace how a figure came to be — and when it was last changed.

End of the series

Where the story ends and the site begins

Six chapters have shown what happened, how it developed, who came, what the consequences were, what drove the change and how we know it. From here the reader picks the road.

Knowledge is not static. As new research and new statistics appear, our understanding also changes. That is why Sverigefakta is a living knowledge platform.

Primary sources

Sources in this chapter

Updated 2026-07-21.

Summary

Key takeaways

  • All statistics rest on definitions, measurements and methodological choices.
  • Almost all Swedish social statistics come from a few primary sources: SCB, Brå, the Police, the School Agency, the Board of Health, the Migration Agency, Eurostat and OECD.
  • The same concept can have different definitions — and that changes the figure.
  • Statistics describe groups and patterns, not individuals or motives.
  • Correlation is not causation; a third confounding factor can explain both observations.
  • Different results across studies often reflect samples, time periods, definitions, models or methods — not error.
  • Sverigefakta uses four evidence levels: very strong, strong, moderate and uncertain.
  • Some questions remain open — long-run demographics, AI effects, the third generation.
  • Sverigefakta works with primary sources, citations, method descriptions, version history and public corrections.
  • Knowledge is not static. New data shifts understanding — that is why the platform is living.

Further reading