Trust over time — two series that diverge
Sweden has topped World Values Survey rankings of interpersonal trust for decades. The SOM Institute's series shows that the share of Swedes reporting high trust in other people held around 60–65 % from 1996 into the mid-2000s, then slowly slid to about 57 % by 2025. Trust is lower and falling faster among socially vulnerable groups.
Confidence in the police is measured by Brå's National Safety Survey (NTU), which began in 2006 — not the 1990s. That series moves the opposite way: the share reporting fairly or very high confidence in the police rose from around 60 % in 2006 to approximately 72 % in 2023. The interesting story is that interpersonal trust and institutional trust in the police do not move together.
High interpersonal trust (national)(SOM)
- Then
- ≈ 62 % (1996)
- Now
- ≈ 57 % (2025)
High trust in socially vulnerable groups(Holmberg & Rothstein 2022)
- Then
- lower
- Now
- lower and falling
Fairly / very high confidence in the police(Brå NTU 2024)
- Then
- ≈ 60 % (2006)
- Now
- ≈ 72 % (2023)
| Mått | Then | Now |
|---|---|---|
| High interpersonal trust (national)(SOM) | ≈ 62 % (1996) | ≈ 57 % (2025) |
| High trust in socially vulnerable groups(Holmberg & Rothstein 2022) | lower | lower and falling |
| Fairly / very high confidence in the police(Brå NTU 2024) | ≈ 60 % (2006) | ≈ 72 % (2023) |
Why trust matters for the welfare state
Robert Putnam's early claim — that ethnic diversity itself lowers trust — remains contested. Meta-studies (Dinesen, Schaeffer & Sønderskov, Annual Review of Political Science 2020) find a small, context-dependent negative effect, with European replications pulling in different directions. What is more robust is the link between perceived shared norms in the neighbourhood and trust: when people no longer feel they share basic norms with their neighbours — around child-rearing, gender roles or law — willingness to contribute to shared solutions declines. The mechanisms behind this are still being debated.
Swedish trust can be described as society's most important invisible infrastructure — possible to damage in a generation, hard to rebuild.
Consequences for the welfare state
When interpersonal trust falls, willingness to pay for shared systems, compliance with rules, and willingness to help strangers fall with it. The picture is not uniform: confidence in the police as an institution is at historic highs in Brå's NTU and has risen during the 2010s, while witness cooperation in vulnerable areas is documented as low and local interpersonal trust there is falling. These parallel-but-opposite movements — high institutional trust, low local trust — are what make the Swedish welfare model vulnerable in a new way.
The question this page answers
Is Sweden still a high-trust society — and what does 'trust' actually mean?
At a glance
Key figures — the most important numbers on this page
1996
≈ 62 %
SOM — high general interpersonal trust
2023
≈ 57 %
SOM — high general interpersonal trust
ESS 2020
≈ 6.3
Scale 0–10, 'most people can be trusted'
2006
≈ 60 %
Brå NTU — high confidence in the police
2023
≈ 72 %
Brå NTU — high confidence in the police
Series
1986–
SOM Institute — annual since 1986
What does the data show?
Objective observations — not interpretations
- Sweden has for decades ranked at or near the top of the World Values Survey and European Social Survey for general interpersonal trust.
- The SOM Institute's series shows a slow, steady decline: the share of Swedes reporting high trust in other people has fallen from around 62 % (mid-1990s) to about 57 % (2020s).
- Institutional trust moves at a different pace. Brå's National Safety Survey shows that confidence in the police rose during the 2010s — from about 60 % (2006) to about 72 % (2023) — while interpersonal trust was falling.
- The national average hides large differences between groups and areas. Trust is lower among young people, lower-educated, low-income earners and residents of socially vulnerable areas — and it is falling faster there.
- No single indicator captures 'trust'. SOM, ESS, WVS and OECD ask different questions with different scales, samples and periodicity — the series are not interchangeable and must not be combined in a single chart.
- Survey data show correlation, not causation. Statistical differences between groups or areas cannot be read as one trait 'causing' high or low trust.
Definitions
How the numbers are counted — and what they do not cover
- General (interpersonal) trust
- An individual's baseline expectation that people in general can be trusted. Usually measured via the SOM/ESS standard item 'Generally speaking, would you say that most people can be trusted…' on a 0–10 scale, or the WVS yes/no variant 'Most people can be trusted'.Source: SOM Institute codebook; European Social Survey Rounds 1–11; World Values Survey Wave 7.
- Particularised trust
- Trust in a defined circle — family, kin, clan, colleagues, neighbours. High in almost all societies and therefore varies little between countries. Not to be confused with general trust, which is the variable that sets Sweden apart internationally.Source: Rothstein & Uslaner (2005); Trägårdh & Wallman Lundåsen (SNS 2013).
- Institutional trust / confidence in public authorities
- Share reporting high or fairly high confidence in a single institution (police, courts, health care, parliament, municipality). Measured separately for each institution in SOM and NTU and cannot be added together into a single 'institutional trust' score.Source: SOM — Society Opinion Media; Brå NTU 2024.
- Perceived safety
- Self-reported feeling of being able to move around outdoors without fear of crime, measured mainly in Brå's NTU for Sweden. Safety is not the same as trust — a person can hold high general trust but still feel unsafe in their own neighbourhood.Source: Brå — National Safety Survey (NTU).
- Social capital
- Resources produced by networks, norms of reciprocity and trust between people. Robert Putnam's foundational concept (Bowling Alone, 2000). Contains both relationships (bonding/bridging) and norms and cannot be reduced to a single trust score.Source: Putnam (2000); OECD — How's Life? chapter on social capital.
- Social cohesion
- Broader concept describing the degree to which members of a society share basic norms, trust each other and cooperate across group lines. Usually measured as a composite index in OECD and EU work, rarely as a single question.Source: OECD — Perspectives on Global Development, Trust chapter.
- Self-reported measure
- A measure where the respondent reports their own perception or experience. Affected by non-response (who answers), cultural interpretation of the wording, and the general mood of the moment — and is not the same as an observed behaviour.Source: SOM — method reports; Brå NTU — technical report.
- Area vs individual effect
- When people with different characteristics live in the same area, it is difficult to tell whether differences in trust are caused by the individuals or by the local environment. Ecological correlations at area level cannot be transferred to individual persons.Source: Dinesen, Schaeffer & Sønderskov, Annual Review of Political Science 2020.
How does it look in your municipality?
Trust and safety are reported by municipality in SOM's regional volumes and in local NTU rounds. Open the municipality tool to see trust and safety indicators for your municipality.
Primary sources
Agencies and research institutions behind this page
4
agencies/institutions
6
reports & studies
4
primary datasets
Frequently asked questions
Short answers to what is most often discussed
Related topics
Logical next steps if you want to understand the background
The question this page answers
Is Sweden still a high-trust society — and what does 'trust' actually mean?
At a glance
Key figures — the most important numbers on this page
1996
≈ 62 %
SOM — high general interpersonal trust
2023
≈ 57 %
SOM — high general interpersonal trust
ESS 2020
≈ 6.3
Scale 0–10, 'most people can be trusted'
2006
≈ 60 %
Brå NTU — high confidence in the police
2023
≈ 72 %
Brå NTU — high confidence in the police
Series
1986–
SOM Institute — annual since 1986
What does the data show?
Objective observations — not interpretations
- Sweden has for decades ranked at or near the top of the World Values Survey and European Social Survey for general interpersonal trust.
- The SOM Institute's series shows a slow, steady decline: the share of Swedes reporting high trust in other people has fallen from around 62 % (mid-1990s) to about 57 % (2020s).
- Institutional trust moves at a different pace. Brå's National Safety Survey shows that confidence in the police rose during the 2010s — from about 60 % (2006) to about 72 % (2023) — while interpersonal trust was falling.
- The national average hides large differences between groups and areas. Trust is lower among young people, lower-educated, low-income earners and residents of socially vulnerable areas — and it is falling faster there.
- No single indicator captures 'trust'. SOM, ESS, WVS and OECD ask different questions with different scales, samples and periodicity — the series are not interchangeable and must not be combined in a single chart.
- Survey data show correlation, not causation. Statistical differences between groups or areas cannot be read as one trait 'causing' high or low trust.
Method note
Interpersonal trust is measured by the SOM Institute using the same core question since 1996 (0–10 scale, share answering 7–10). Institutional trust in the Police is measured by Brå's NTU since 2006 (share expressing ‘fairly/very great confidence’). International comparisons draw on the World Values Survey and European Social Survey; these measure related but not identical concepts.
Definitions
How the numbers are counted — and what they do not cover
- General (interpersonal) trust
- An individual's baseline expectation that people in general can be trusted. Usually measured via the SOM/ESS standard item 'Generally speaking, would you say that most people can be trusted…' on a 0–10 scale, or the WVS yes/no variant 'Most people can be trusted'.Source: SOM Institute codebook; European Social Survey Rounds 1–11; World Values Survey Wave 7.
- Particularised trust
- Trust in a defined circle — family, kin, clan, colleagues, neighbours. High in almost all societies and therefore varies little between countries. Not to be confused with general trust, which is the variable that sets Sweden apart internationally.Source: Rothstein & Uslaner (2005); Trägårdh & Wallman Lundåsen (SNS 2013).
- Institutional trust / confidence in public authorities
- Share reporting high or fairly high confidence in a single institution (police, courts, health care, parliament, municipality). Measured separately for each institution in SOM and NTU and cannot be added together into a single 'institutional trust' score.Source: SOM — Society Opinion Media; Brå NTU 2024.
- Perceived safety
- Self-reported feeling of being able to move around outdoors without fear of crime, measured mainly in Brå's NTU for Sweden. Safety is not the same as trust — a person can hold high general trust but still feel unsafe in their own neighbourhood.Source: Brå — National Safety Survey (NTU).
- Social capital
- Resources produced by networks, norms of reciprocity and trust between people. Robert Putnam's foundational concept (Bowling Alone, 2000). Contains both relationships (bonding/bridging) and norms and cannot be reduced to a single trust score.Source: Putnam (2000); OECD — How's Life? chapter on social capital.
- Social cohesion
- Broader concept describing the degree to which members of a society share basic norms, trust each other and cooperate across group lines. Usually measured as a composite index in OECD and EU work, rarely as a single question.Source: OECD — Perspectives on Global Development, Trust chapter.
- Self-reported measure
- A measure where the respondent reports their own perception or experience. Affected by non-response (who answers), cultural interpretation of the wording, and the general mood of the moment — and is not the same as an observed behaviour.Source: SOM — method reports; Brå NTU — technical report.
- Area vs individual effect
- When people with different characteristics live in the same area, it is difficult to tell whether differences in trust are caused by the individuals or by the local environment. Ecological correlations at area level cannot be transferred to individual persons.Source: Dinesen, Schaeffer & Sønderskov, Annual Review of Political Science 2020.
Primary sources
Agencies and research institutions behind this page
4
agencies/institutions
6
reports & studies
4
primary datasets
Frequently asked questions
Short answers to what is most often discussed
Related topics
Logical next steps if you want to understand the background
Sources for this page
The data comes from official primary sources.
Brå · Annually
Nationella trygghetsundersökningen (NTU)NTU — utsatthet och otrygghet
Latest data point · Source published
Sources
- SOM Institute — Swedish trends 1986–2025
- Brå — National Safety Survey (NTU)
- Holmberg & Rothstein — Swedish trust remains high — but falls in vulnerable groups (SOM 2022)
- Trägårdh & Wallman Lundåsen — The Cool Swedish Trust (SNS 2013)
- European Social Survey Round 10 (2020)
- Dinesen, Schaeffer & Sønderskov (ARPS 2020) — Ethnic diversity and social trust: a narrative and meta-analytical review
Read the full investigation of how Sweden has changed.
Back to the overview →