A cross-national analysis of demographic variation in perceived social trust across 22 countries

Abstract

Although social trust is most often studied as an individual disposition to examine the extent to which an individual trusts other people, less attention has been paid to how much one perceives other people in one’s society to trust one another. This study focuses on the latter, “perceived social trust,” as an indicator of societal well-being. Using nationally representative samples from 22 countries across six continents (N = 202,898), we examine cross-national variation in perceived social trust and its demographic correlates. Although some aspects of national rankings based on this new measure align with those based on conventional trust metrics, notable deviations emerged: Indonesia and the Philippines—previously found to be low-trust societies—ranked among the highest in perceived trust, whereas the United States—often ranked more highly—was at the bottom of perceived trust. These findings suggest that cultural norms and sociopolitical contexts may shape individuals’ perceptions of social trust. A random-effects meta-analysis further reveals that some demographic predictors, such as age, gender, and employment status, follow established patterns, while others, including education and religious service attendance, diverge. These findings suggest that perceived social trust may serve as a descriptive indicator of perceived social cohesion and provide information about how people perceive trust within their societies.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-56453-1.

Introduction

Trust among members of a society is a foundational dimension of societal well-being, reflecting the quality of social relationships that enable societies to flourish¹. Societies with a high concentration of those who trust other people tend to thrive, as widespread social trust underpins many forms of collective well-being^2,3. When people generally trust one another, they are more willing to cooperate, even with strangers, reducing the need for costly monitoring, enforcement, and regulation⁴. This enhances efficiency by lowering transaction costs⁵ and strengthens democratic governance by encouraging civic engagement⁶. Additionally, empirical studies have linked higher levels of social trust with a range of positive societal outcomes, including lower crime rates^7,8, better public health⁹, higher life satisfaction⁹, and greater political stability¹⁰. In contrast, societies with low trust often experience increased polarization, weakened institutions, and social fragmentation¹¹.

Thus, trust is not merely an individual trait but a communal asset that both reflects and fosters the overall health of society. Within the human flourishing framework, community well-being encompasses six key domains. One of these—“good relationships”—captures the extent to which individuals perceive that members of their society have close relationships, respect one another, trust one another, and contribute to one another’s well-being^1,12. Trust, as a measure of these communal well-being, plays a central role in shaping cooperative norms, promoting collective flourishing, and supporting societal functioning.

This study situates trust within the broader context of community well-being, highlighting its significance not only for individuals but for society as a whole. Specifically, we focus on “perceived social trust”—an individual’s perception of how much people in one’s society trust one another—as a measure of communal well-being. By contrast, the conventional concept of social trust is often studied as an individual disposition: a belief in the honesty, integrity, and reliability of others. In other words, it reflects a stable, enduring personal characteristic and captures generalized expectations about others’ behavior. While prior research has largely emphasized one’s trust in others, one’s perceived trust among others offers a complementary lens through which to understand social trust. This distinction deepens our understanding of the relationship between social trust and communal well-being, as personal trust may not necessarily align with perceptions of societal trust. For example, someone with a secure upbringing and a naturally trusting disposition may live in a country marked by corruption or social fragmentation. Such an individual may personally trust others but still perceive low levels of trust among fellow citizens. In other words, perceived trust reflects a broader impression of society, shaped more by shared cultural narratives and collective experiences than by individual interactions alone.

While conceptually distinct, personal trust and perceived trust may be related. Individuals who are generally trusting might be more likely to perceive others as trustworthy, but perceptions of societal trust can also be influenced by broader cultural norms, media portrayals, and collective historical experiences. In this sense, personal trust and perceived trust capture related yet distinct aspects of the social trust concept.

Using a measure of perceived trust, we examine cross-national differences in social trust from a community well-being perspective. Studies based on international surveys such as the European Social Survey and the World Values Survey typically find that social trust is highest in Scandinavian countries (e.g., Sweden), followed by other economically advanced Western societies, including Australia, the United Kingdom, and Germany In contrast, lower levels of trust are typically found in less economically developed countries, such as Brazil, the Philippines, and Indonesia¹³. We expect our alternative measure of social trust to produce national rankings that broadly mirror those of conventional measures, given their anticipated positive correlation; however, some discrepancies are to be expected due to broader societal and political dynamics. For instance, in countries with legacies of slavery or apartheid (e.g., the United States and South Africa), or those experiencing political upheaval or polarization (e.g., Brazil and the United States), individuals might personally trust others yet still perceive that people in society more broadly do not trust one another.

Given that perceptions of trust among others are embedded in wider social contexts, it becomes important to explore whether they also differ according to demographic characteristics. Just as individuals vary in personal trust, they may differ in perceived trust, with some demographic groups viewing people as more trusting one another than other groups. Given the paucity of research on demographic variation in perceived trust, we draw on the broader literature examining demographic differences in personal trust to inform our expectations.

Although social trust often varies across demographic groups, these patterns are not always consistent, and most existing research has focused primarily on the United States and European countries. For example, previous research shows that older individuals tend to report higher levels of trust in the U.S.¹⁴, whereas no significant age difference has been found in European countries¹⁵. Women generally show lower levels of trust than men in both the U.S.¹⁴ and European countries¹⁵. For marital status, married individuals do not show significant differences in social trust compared to other groups in the U.S.¹⁴. However, divorced individuals report lower levels of trust than non-divorced individuals in both the U.S.¹⁴ and Sweden¹⁶. In Germany, being in a domestic partnership is negatively related to social trust, whereas parenthood is positively related¹⁷. Experiencing unemployment is associated with reduced social trust in the U.S.¹⁸. Immigrants also report lower trust levels in both the U.S. and European countries¹⁵.

While education is generally associated with higher social trust in the U.S.¹⁴, Italy¹⁹, and the EU-15 countries¹⁵, a study in Bangladesh presents a contrasting finding—less educated individuals are more likely to trust others than those with higher education²⁰. Similarly, findings regarding religious service attendance and social trust are mixed. A study using European Values Survey data found a positive relationship between the two²¹, a pattern also observed in the U.S.²² and in the U.K.²³. However, in the Middle East and North Africa, this association is statistically not significant²⁴. A Nigerian study even found that higher frequency of religious activity predicted lower social trust²⁵. These conflicting findings call for closer scrutiny of each demographic factor across diverse cultural contexts.

Our dataset includes several non-WEIRD (Western, Educated, Industrialized, Rich, and Democratic) countries²⁶, such as those in North Africa (e.g., Egypt), Sub-Saharan Africa (e.g., Kenya, Nigeria, South Africa, Tanzania), East Asia (e.g., Japan), and South Asia (e.g., India, Indonesia, the Philippines). Using a diverse sample of 202,898 individuals from 22 countries across different continents, we examine key demographic characteristics—age, gender, marital status, employment, education, religious service attendance, and immigration status—and explore the following research questions:

Research Question 1

The distribution and descriptive statistics of key demographic variables will display diverse patterns across the international sample (Table 1).

Table 1.

Frequency distributions of demographic variables and GFS-participating countries (N = 202,898). Note. Country-specific frequency distributions are available in the online supplement.

Variable	Frequency	Percent
Age
18–24	27,007	13.3%
25–29	20,700	10.2%
30–39	40,256	19.8%
40–49	34,464	17.0%
50–59	31,793	15.7%
60–69	27,763	13.7%
70–79	16,776	8.3%
80 or Older	4119	2.0%
(Missing)	(20)	(0.0%)
Gender
Male	98,411	48.5%
Female	103,488	51.0%
Other	602	0.3%
(Missing)	(397)	(0.2%)
Marital Status
Single/Never Been Married	52,115	25.7%
Married	107,354	52.9%
Separated	5195	2.6%
Divorced	11,654	5.7%
Widowed	9823	4.8%
Domestic Partner	14,931	7.4%
(Missing)	(1826)	(0.9%)
Employment
Employed for an Employer	78,815	38.8%
Self-Employed	36,362	17.9%
Retired	29,303	14.4%
Student	10,726	5.3%
Homemaker	21,677	10.7%
Unemployed and Looking for a Job	16,790	8.3%
None of These/Other	8431	4.2%
(Missing)	(793)	(0.4%)
Education
Up to 8 Years	45,078	22.2%
9–15 Years	115,097	56.7%
16 + Years	42,578	21.0%
(Missing)	(146)	(0.1%)
Service Attendance
>1/Week	26,537	13.1%
1/Week	39,157	19.3%
1–3/Month	19,749	9.7%
A Few Times a Year	41,436	20.4%
Never	75,297	37.1%
(Missing)	(722)	(0.4%)
Immigration Status
Born in This Country	190,998	94.1%
Born in Another Country	9791	4.8%
(Missing)	(2110)	(1.0%)
Country
Argentina	6724	3.3%
Australia	3844	1.9%
Brazil	13,204	6.5%
Egypt	4729	2.3%
Germany	9506	4.7%
Hong Kong	3012	1.5%
India	12,765	6.3%
Indonesia	6992	3.4%
Israel	3669	1.8%
Japan	20,543	10.1%
Kenya	11,389	5.6%
Mexico	5776	2.8%
Nigeria	6827	3.4%
Philippines	5292	2.6%
Poland	10,389	5.1%
South Africa	2651	1.3%
Spain	6290	3.1%
Sweden	15,068	7.4%
Tanzania	9075	4.5%
Turkey	1473	0.7%
United Kingdom	5368	2.6%
United States	38,312	18.9%

Research Question 2

Mean levels of social trust will differ significantly across countries (Table 2).

Table 2.

Ordered proportions/percentages of social trust.

Global Flourishing Study (N = 202,898)					Integrated Values Surveys (N = 156,658)a
Country	Proportion	95% CI	SE		Country	Percent
1. Indonesia	0.50	(0.48, 0.51)	0.50		1. Sweden	62.8
2. Philippines	0.49	(0.47, 0.50)	0.50		2. Australia	48.5
3. Hong Kong	0.48	(0.45, 0.50)	0.50		3. United Kingdom	43.3
4. Sweden	0.39	(0.38, 0.40)	0.49		4. Germany	41.6
4. Australia	0.39	(0.37, 0.41)	0.49		5. Spain	41.0
6. India	0.35	(0.33, 0.36)	0.48		6. United States	37.0
7. Japan	0.30	(0.29, 0.31)	0.46		7. Hong Kong	36.4
8. Germany	0.28	(0.27, 0.29)	0.45		8. Japan	33.7
9. Egypt	0.26	(0.25, 0.28)	0.44		9. Poland	24.1
10. Israel	0.24	(0.21, 0.27)	0.43		10. South Africa (2014)b	23.3
11. Poland	0.23	(0.20, 0.25)	0.42		11. Israel (2004)b	22.9
12. United Kingdom	0.21	(0.19, 0.22)	0.40		12. Argentina	19.2
13. Mexico	0.20	(0.18, 0.21)	0.40		13. India	16.9
14. Spain	0.18	(0.16, 0.19)	0.38		14. Turkey	14.0
14. Nigeria	0.18	(0.16, 0.19)	0.38		15. Nigeria	13.0
16. Turkey	0.15	(0.13, 0.17)	0.36		16. Mexico	10.5
17. Tanzania	0.14	(0.12, 0.15)	0.34		17. Kenya	9.5
17. Argentina	0.14	(0.12, 0.15)	0.34		18. Tanzania (2004)b	7.7
19. Kenya	0.12	(0.11, 0.13)	0.33		19. Egypt	7.3
20. South Africa	0.11	(0.09, 0.12)	0.31		20. Brazil	6.5
21. Brazil	0.10	(0.09, 0.10)	0.29		21. Philippines	5.3
22. United States	0.10	(0.09, 0.10)	0.29		22. Indonesia	4.6

Note. Overall proportion = 0.25 (95% CI = 0.20, 0.31; SE = 0.03).

^a Integrated Values Surveys (IVS) is a cooperation in joint data collection between the European Value Study (EVS) and the World Value Survey (WVS), two large-scale, cross-national, and repeated cross-sectional survey research programs. The IVS includes a WVS item asking, “Generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?” Percentages presented above, which came from the fifth version of the Joint EVS/WVS 2017–2022 dataset, are the share of respondents who answered “most people can be trusted” to this question.

^b This country was not included in the Joint EVS/WVS 2017–2022 sample (which consisted of 92 countries and territories), so the only (Israel and Tanzania) or most recent data (South Africa) available from its previous versions are presented.

Research Question 3

Social trust will vary across demographic categories, and these variations will differ by country (Table 3).

Table 3.

Random effects meta-analysis of proportion of social trust by demographic category.

					Prediction Interval
Variable	Category	Proportion	95% CI of Proportion	SE Analogue (CI Width/4)	LL	UL	Heterogeneity (τ)	I2	Global p-value
Age group									< 0.001**
	18–24	0.22	(0.16,0.28)	0.03	0.05	0.57	0.14	95.8
	25–29	0.21	(0.17,0.26)	0.02	0.07	0.49	0.11	93.9
	30–39	0.21	(0.16,0.26)	0.02	0.07	0.47	0.11	94.2
	40–49	0.22	(0.17,0.27)	0.02	0.10	0.49	0.12	94.1
	50–59	0.23	(0.18,0.29)	0.03	0.09	0.57	0.13	95.3
	60–69	0.25	(0.19,0.31)	0.03	0.07	0.52	0.14	95.4
	70–79	0.26	(0.20,0.33)	0.03	0.09	0.51	0.15	95.5
	80 or older	0.18	(0.08,0.34)	0.07	0.00	0.61	0.30	99.4
Gender									< 0.001**
	Male	0.25	(0.20,0.30)	0.03	0.12	0.50	0.12	93.9
	Female	0.21	(0.16,0.27)	0.03	0.08	0.48	0.13	95.2
	Other	0.01	(0.00,0.04)	0.01	0.00	0.74	0.03	99.6
Marital status									< 0.001**
	Married	0.24	(0.19,0.30)	0.03	0.10	0.52	0.13	95.0
	Separated	0.24	(0.19,0.31)	0.03	0.10	0.65	0.14	95.5
	Divorced	0.20	(0.15,0.26)	0.03	0.07	0.49	0.12	95.3
	Widowed	0.24	(0.18,0.31)	0.03	0.08	0.49	0.15	96.0
	Domestic partner	0.14	(0.06,0.27)	0.05	0.00	0.72	0.23	99.2
	Single, never married	0.21	(0.17,0.26)	0.02	0.07	0.48	0.11	93.7
Employment status									< 0.001**
	Employed for an employer	0.23	(0.18,0.28)	0.03	0.08	0.48	0.12	94.4
	Self-employed	0.23	(0.18,0.30)	0.03	0.09	0.60	0.13	95.1
	Retired	0.26	(0.21,0.32)	0.03	0.10	0.49	0.13	94.4
	Student	0.21	(0.16,0.28)	0.03	0.06	0.53	0.13	95.5
	Homemaker	0.21	(0.16,0.27)	0.03	0.07	0.50	0.13	95.2
	Unemployed and looking for a job	0.18	(0.14,0.23)	0.02	0.06	0.52	0.10	93.9
	None of these/other	0.20	(0.15,0.27)	0.03	0.03	0.54	0.13	95.9
Education									< 0.001**
	Up to 8 years	0.22	(0.17,0.28)	0.03	0.04	0.49	0.13	95.3
	9–15 years	0.22	(0.17,0.28)	0.03	0.08	0.50	0.12	94.9
	16 + years	0.22	(0.17,0.28)	0.03	0.09	0.49	0.13	95.0
Religious service attendance									< 0.001**
	> 1/week	0.31	(0.23,0.39)	0.04	0.11	0.86	0.19	96.8
	1/week	0.28	(0.22,0.34)	0.03	0.11	0.58	0.15	95.3
	1–3/month	0.25	(0.20,0.31)	0.03	0.09	0.58	0.13	94.8
	A few times a year	0.21	(0.16,0.26)	0.02	0.07	0.43	0.11	94.3
	Never	0.19	(0.14,0.25)	0.03	0.04	0.42	0.12	95.7
Immigration status									< 0.001**
	Born in this country	0.23	(0.18,0.29)	0.03	0.09	0.49	0.12	94.6
	Born in another country	0.24	(0.20,0.29)	0.02	0.12	0.58	0.11	92.5

* p < .05; ** p < .007 (Bonferroni corrected threshold).

This descriptive, exploratory study examines how perceived social trust varies across countries and demographic groups and provides a basis for future research on factors shaping trust in different contexts.

Methods

The following description of our methods has been adapted from VanderWeele et al.²⁷. Further methodological details are available elsewhere^28–33.

Data

The Global Flourishing Study (GFS) is a longitudinal study of over 200,000 adults (age 18 or older) from 22 geographically and culturally diverse countries, with nationally representative sampling within each country, concerning the distribution of determinants of well-being. Data for Wave 1 were collected principally during 2023, while some countries began data collection in 2022³⁰. Four additional waves of panel data are planned to be collected annually from 2024 to 2027. A total of 202,898 individuals participated in Wave 1 survey in Argentina, Australia, Brazil, Egypt, Germany, Hong Kong, India, Indonesia, Israel, Japan, Kenya, Mexico, Nigeria, the Philippines, Poland, South Africa, Spain, Sweden, Tanzania, Turkey, the United Kingdom, and the United States. These countries were selected to (1) maximize coverage of the world’s population, (2) ensure geographic, cultural, and religious diversity, and (3) prioritize feasibility and existing data collection infrastructure.

Data collection was carried out by Gallup, Inc., and the precise sampling design to ensure nationally representative samples varied by country^30,33. While there are concerns with the sampling design varying by country, the issue boils down to how representative the samples are to the country’s population. To ensure the representativeness of each country’s sample, weights were constructed and post-stratified and adjusted for nonresponse to align the observed samples with the target populations. Survey items included various aspects of well-being such as happiness and life satisfaction, physical and mental health, meaning and purpose, character and virtue, close social relationships, and financial and material stability¹², along with other demographic, social, economic, political, religious, personality, childhood, community, health, and well-being variables^34,35. The data that support the findings of this article are openly available on the Open Science Framework. The specific dataset used was Wave 1 non-sensitive Global data (https://osf.io/sm4cd/) available February 2024 – March 2026 via preregistration and publicly from then onwards⁶. During the translation process, Gallup adhered to TRAPD model (translation, review, adjudication, pretesting, and documentation) for cross-cultural survey research (https://ccsg.isr.umich.edu/chapters/translation/overview/)³⁶.

Measures

Perceived trust. Perceived social trust was assessed using a single-item question: “How many people in this country trust one another?” Responses were coded 1 = all, 2 = most, 3 = some, 4 = not very many, and 5 = none. We dichotomized the item (0 = none/not very many/some, 1 = most/all) based on the default definition of generalized trust, which is the belief that “most people” can be trusted.

Demographic variables. Continuous age was classified as 18–24, 25–29, 30–39, 40–49, 50–59, 60–69, 70–79, and 80 or older. Gender was assessed as male, female, and other. Marital status was assessed as single/never married, married, separated, divorced, widowed, and domestic partner. Employment was assessed as employed, self-employed, retired, student, homemaker, unemployed and searching, and other. Education was assessed as up to 8 years, 9–15 years, and 16 + years. Religious service attendance was assessed as more than once/week, once/week, one-to-three times/month, a few times/year, or never. Immigration status was dichotomously assessed as foreign- or native-born. Religious tradition/affiliation had categories of Christianity, Islam, Hinduism, Buddhism, Judaism, Sikhism, Baha’i, Jainism, Shinto, Taoism, Confucianism, Primal/Animist/Folk religion, Spiritism, African-Derived, some other religion, or no religion/atheist/agnostic with precise response categories varying by country³⁷. Racial/ethnic identity was assessed in some, but not all, countries with response categories varying by country. For additional details on the assessments, see the GFS codebook (https://osf.io/7uj6y/) or Crabtree et al.²⁸.

Analysis

Descriptive statistics for the full sample, weighted to be nationally representative within each country, were estimated for each demographic variable. Nationally representative proportions of social trust (i.e., study participants who reported that most or all people in their country trust one another) were estimated separately for each country and ordered from highest to lowest along with 95% confidence intervals (CIs) and standard errors for the proportions. Variation in proportions of social trust across demographic categories was estimated, with all analyses initially conducted by country (online supplement). Primary results consisted of random effects meta-analyses of country-specific proportions of social trust in each specific demographic category^38,39 along with 95% CIs, standard errors, lower and upper limits of a 95% prediction interval across countries, heterogeneity (τ), and I² for evidence concerning variation within a particular demographic variable across countries⁴⁰. Forest plots of estimates are available in the online supplement. All meta-analyses were conducted in R⁴¹ using the metafor package⁴².

Within each country, a global test of variation of social trust across levels of each demographic variable was conducted, and a pooled p-value⁴³ across countries reported concerning evidence for variation within any country. Bonferroni-corrected p-value threshold is provided based on the number of demographic variables^44,45. Religious tradition/affiliation and race/ethnicity were used as additional demographic characteristics for within-country analyses, but the proportions of social trust within these demographic categories were not meta-analyzed across countries since the availability of these response categories varied by country. As a supplementary analysis, population weighted meta-analyses were also conducted. All analyses were pre-registered with the Center for Open Science (COS) prior to data access (https://osf.io/546q9); all code needed to reproduce analyses are available in an online repository⁴⁶.

Missing data

Missing data on all variables were imputed using multivariate imputation by chained equations, and five imputed datasets were created and analyzed^47–50. To account for variation in the assessment of certain variables across countries (e.g., religious tradition/affiliation and race/ethnicity), the imputation process was conducted separately in each country. This within-country imputation approach ensured that the imputation models accurately reflected country-specific contexts and assessment methods. Sampling weights were included in the imputation model to account for specific-variable missingness that may have been related to probability of inclusion in the study.

Accounting for complex sampling design

The GFS used different sampling schemes across countries based on the availability of existing panels and recruitment needs³⁰. All analyses account for the complex survey design components by including weights, primary sampling units, and strata. Additional methodological detail, including accounting for the complex sampling design is provided elsewhere^29,33.

Results

Table 1 reports the frequency distributions of demographic variables for the total sample, combining data from all 22 countries (N = 202,898). Country-specific sample sizes range from 1,473 respondents in Turkey (0.7%) to 38,312 in the United States (18.9%). The table shows diverse demographic patterns (Research Question 1), with the modal categories being ages 30–39 (19.8%), female (51.0%), married (52.9%), employed by an employer (38.8%), having 9–15 years of education (56.7%), never attending religious services (37.1%), and native-born (94.1%) in the total sample.

Table 2 presents the proportions of perceived trust across GFS countries in descending order. The highest proportion [95% CI] was found in Indonesia, where a half (0.50 [0.48, 0.51]) of respondents said that “most” or “all” of the people in their country trusted one another. This was closely followed by the Philippines (0.49 [0.47, 0.50]) and Hong Kong (0.48 [0.45, 0.50]), all approximately twice as high as the overall average (0.25 [0.20, 0.31]). The lowest levels of perceived trust were observed in the United States (0.10 [0.09, 0.10]), Brazil (0.10 [0.09, 0.10]), and South Africa (0.11 [0.09, 0.12]), where only one in ten respondents expressed the belief that most or all people in their country trust one another.

To explore potential group-level differences, we categorized the 22 countries into three approximately equal groups: lower-trust (0.15 or lower; 7 countries, from the United States to Turkey), middle-trust (greater than 0.15 but less than 0.30; 8 countries, from Nigeria to Germany), and higher-trust (0.30 or higher; 7 countries, from Japan to Indonesia). Then, we examined whether these groupings aligned with established predictors of social trust, including ethnic heterogeneity, national wealth, income inequality, levels of democracy, and dominant religious traditions^13,51–55. Testing whether these factors explain differences in group membership (or proportions among GFS countries) may be beyond the scope of the present study given the limited number of countries; however, no noticeable differences were observed among the groups, seemingly because each group comprised countries that are geographically, economically, religiously, and socio-culturally diverse.

When examining these ordered proportions, it is important to recognize that differences across countries may partly reflect factors other than actual variations in social trust. These factors include translation and interpretation issues (e.g., differences in the understanding of the survey’s trust target) and response biases (e.g., central tendency bias observed among some Asian respondents who tend to select midpoint responses due to traditional Confucian modesty)^56,57. Thus, comparisons across demographic groups or within individual countries may provide more reliable analyses for drawing conclusions.

To situate our results within existing research, we compared our ordered proportions of social trust with the most recent data on social trust reported by the Integrated Values Surveys (IVS)—a joint data collection initiative between the European Values Study (EVS) and the World Values Survey (WVS), two large-scale, cross-national, and repeated cross-sectional survey programs. The IVS includes a WVS item (the conventional measure of social trust) asking, “Generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?” The percentages presented in the last column of Table 2 represent the share of respondents who answered “most people can be trusted” to this question, based on data from the fifth version of the Joint EVS/WVS 2017–2022 dataset (N = 156,658) covering 92 countries and territories. For most countries, the data reflect responses collected in 2022; however, for three countries not included in the 2017–2022 wave, we used the most recent available data: 2004 for Israel and Tanzania, and 2014 for South Africa⁵⁸.

To compare the ordered proportions from the GFS with the percentages from the IVS, we calculated the rank-order correlation coefficient (Spearman’s ρ), which was 0.113 and not significant (p = .616). This indicates that the two rank orders were not especially consistent. However, when using trichotomous typology (low-, middle-, and high-trust), we found that 9 out of 22 countries (40.9%) remained in the same group when classified by IVS percentages instead of GFS proportions. In contrast, 13 countries (59.1%) shifted in group membership—7 countries (31.8%) moved up, and 6 (27.3%) moved down.

On average, a country’s rank changed by approximately six positions (M = 6.05, SD = 6.02, min = 1, max = 21), with three countries shifting by more than two standard deviations. Notably, the U.S., which ranked at the bottom (low-trust) in the GFS data, moved up 16 positions to rank 6th (high-trust) in the IVS data. Conversely, Indonesia and the Philippines—ranked as the top two in the GFS on perceived trust—fell to the bottom two positions in the IVS, becoming the lowest-ranked countries on generalized trust. When these three outliers were excluded, the correlation between the remaining 19 countries’ ranks was positive and significant (ρ = 0.601, p = .007), indicating general consistency between the two datasets for the majority of countries.

While we cannot easily reconcile the drastic differences in rank observed for the U.S., the Philippines, and Indonesia, it is worth noting that the Philippines’ high proportion of social trust (0.49) and the U.S.’s low proportion (0.10) in the GFS data may be anomalies. According to the IVS 1981–2022 dataset, the Philippines has consistently reported relatively low percentages of social trust—ranging from 3.2% in 2014 to 8.3% in 2004—while the U.S. has shown mid-to-high levels, ranging from 34.8% in 2014 to 50.0% in 1993. It is even more difficult to interpret Indonesia’s top-ranked GFS proportion (0.50), collected in 2022–2023, given that its IVS percentage of social trust for the same year was only 4.6%—a dramatic decline from 45.7% in 2004 and 37.5% in 2010.

The difference between the GFS and IVS items may partially account for these discrepancies. However, further research should examine whether other factors may be at play given such dramatic shifts in rankings. Regardless of these anomalies, the proportion of perceived trust clearly varied across countries (Research Question 2). We now turn to comparisons across demographic groups, which may provide a more reliable basis for interpretation than overemphasizing cross-national differences.

Table 3 shows results from random effects meta-analyses of country-specific proportions of perceived trust across demographic category. Consistent with prior research^14,51, older age groupsࣧ60-69 (0.25 [0.19, 0.31]) and 70–79 (0.26 [0.20, 0.33])ࣧhad higher overall proportions of perceived trust than their younger counterparts: 18–24 (0.22 [0.16, 0.28]), 25–29 (0.21 [0.17, 0.26]), 30–39 (0.21 [0.16, 0.26]), 40–49 (0.22 [0.17, 0.27]), and 50–59 (0.23 [0.18, 0.29])ࣧexcept the oldest group, 80 or older, which had the lowest proportion (0.18 [0.08, 0.34]). However, this overall estimate for the oldest age group should be interpreted with caution, as their proportions varied more across countries than those of any other age group. Specifically, the prediction interval for this groupࣧconstructed using the “calibrated effect size”²⁹ rather than the raw effect size, and based on an estimate of heterogeneity (τ = 0.30), which represents the standard deviation of the distribution of proportions across those countriesࣧindicated that approximately 95% of the proportions of 22 countries for this age group fall between 0.00 and 0.61 (see Figure S8 of the online supplement for a forest plot visualizing the cross-country variation). Furthermore, unlike what the meta-analytic overall proportions showed, a review of country-specific results revealed that the oldest age group had higher proportion of perceived trust than younger groups in more than half (13) of 22 countries: Brazil, Germany, India, Indonesia, Japan, Kenya, South Africa, Spain, Sweden, Tanzania, Turkey, the U.K., and the U.S. (Tables S1b to S22b of the online supplement). Also, the I² estimate for the 80 + group was 99.4, slightly higher than those for other age groups (which ranged from 93.9 to 95.8), suggesting that the variability in this group’s proportion was more strongly driven by between-country heterogeneity than by sampling variability. The global p-value for the age group variable was statistically significant (p < .001) and remained below the Bonferroni-corrected threshold (p < .007), indicating that age was significantly associated with perceived trust in at least one of the 22 countries. In fact, the age group variable was significant (p < .001) in 10 of 22 countries: Argentina, Australia, Germany, Hong Kong, India, Japan, Mexico, Poland, Sweden, and the U.S.

As anticipated based on prior literature^14,15, the overall proportion of perceived trust was higher among males (0.25 [0.20, 0.30]) than females (0.21 [0.16, 0.27]), while it was close to zero among participants who identified as “other” gender (0.01 [0.00, 0.04]). However, this estimate should be interpreted with caution due to considerable variation across countries; in 95% of countries, the proportion for this group ranged from 0.00 to 0.74, with the sole exception being Indonesia, where the proportion was substantially higher (0.75 [0.67, 0.81]) (see Figure S11 in the online supplement).

Respondents who were married (0.24 [0.19, 0.30]) exhibited the same overall proportion of social trust as those who were separated (0.24 [0.19, 0.31]) or widowed (0.24 [0.18, 0.31]). These proportions were higher than those observed among divorced individuals (0.20 [0.15, 0.26]), those living with a domestic partner (0.14 [0.06, 0.27]), and those who were single or never married (0.21 [0.17, 0.26]).

Consistent with prior research, employed respondents—whether “employed for an employer” (0.23 [0.18, 0.28]) or “self-employed” (0.23 [0.18, 0.30])—reported higher overall proportions of perceived trust than those in other employment categories: “students” (0.21 [0.16, 0.28]), “homemakers” (0.21 [0.16, 0.27]), “unemployed and looking for a job” (0.18 [0.14, 0.23]), and those in the “none of these/other” category (0.20 [0.15, 0.27]). Notably, the highest overall proportion of social trust was reported by “retired” respondents (0.26 [0.21, 0.32]), a finding that may reflect the higher average age of individuals in this group relative to other employment categories.

In contrast to previous studies that have documented a positive relationship between education and social trust^14,15,19, we found that the overall proportion of perceived trust was identical across all three education levels: “up to 8 years,” “9–15 years,” and “16 + years” (0.22 [0.17, 0.28] for each group). However, country-specific analyses revealed that the anticipated positive association between education and social trust was more likely to appear in Western or economically developed countries. Specifically, as education level increased from “up to 8 years” to “16 + years,” the proportion of perceived trust rose in eight countries: Australia, Germany, Israel, Japan, Spain, Sweden, the U.K., and the U.S. In contrast, this proportion decreased in nine countries: Egypt, Indonesia, Kenya, Mexico, Nigeria, the Philippines, Poland, Tanzania, and Turkey. The remaining five—Argentina, Brazil, Hong Kong, India, and South Africa—displayed a non-linear pattern.

Next, we found that the overall proportion of social trust was positively associated with religious service attendance. Specifically, the proportion increased progressively across frequency categories: from “never” (0.19 [0.14, 0.25]), to “a few times a year” (0.21 [0.16, 0.26]), “one to three times a month” (0.25 [0.20, 0.31]), “once a week” (0.28 [0.22, 0.34]), and “more than once a week” (0.31 [0.23, 0.39]). Finally, we found that immigrants—defined as foreign-born respondents—reported a slightly higher overall proportion of perceived trust (0.24 [0.20, 0.29]) than their native-born counterparts (0.23 [0.18, 0.29]), which contrasts with findings from prior studies²⁴. A review of country-specific results showed significant differences between the two groups in only six countries. In half of these (Brazil, Poland, and the United States), foreign-born respondents reported higher proportions of social trust (see Tables S3b, S15b, and S22b), aligning with the meta-analytic trend. In the other half (Hong Kong, India, and Sweden), native-born respondents reported higher proportions (see Tables S6b, S7b, and S18b).

In sum, the overall proportion of perceived trust varied across demographic categories, and measures of dispersion indicated that these variations were not uniform across countries (Research Question 3).

Discussion

In this study, social trust was assessed as an individual’s perception of how much people in society trust one another, perceived trust, rather than as one’s personal trust in others. This study’s descriptive, exploratory approach allowed us to examine patterns in societal trust and identify findings for further investigation.

First, regarding national rankings, we found that while most countries aligned closely with results from conventional measures of social trust, a few stood out as anomalies. That is, countries such as Indonesia and the Philippines—traditionally considered “low-trust” societies—ranked in the top using the GFS measure, with about half of respondents reporting that most or all people in their country trust one another. In contrast, the U.S., previously found to be a “high-trust” society, ranked bottom in perceived social trust: only one in ten respondents indicated that most or all people in the U.S. trust one another. This striking result merits further investigation.

A recent Pew Research Center report, based on General Social Survey data, shows a 26.1% decline in the proportion of U.S. adults who agree that “most people can be trusted” between 1972 and 2018⁵⁹. Given that individuals’ perceptions of social trust may be shaped by exposure to broader narratives—such as media reports on declining trust—rather than solely by their personal experiences or dispositions, it is plausible that Americans perceive a sharper erosion of trust at the societal level than they experience in their own interpersonal relationships⁶⁰. In addition to increased public awareness of declining trust, the American public has been exposed to narratives of political polarization in recent years, which may have led individuals to view American society as deeply distrustful, even if their personal trust in others remains moderate⁶⁰. Another possible explanation for this pattern may lie in the historical legacy of slavery and racial segregation⁶¹. Countries with historical legacies of social division, such as South Africa, also exhibit notable gaps between the two measures of social trust, ranking low in perceived social trust (#20) in our data despite being classified as a middle-trust society in IVS (#10). While multiple factors may contribute to this discrepancy, the pronounced nature of the U.S. case indicates that further research is necessary to elucidate the mechanisms underlying such exceptionally low levels of perceived trust and the gaps between how much individuals trust others personally and how much they believe others trust each other.

The cases of Indonesia and the Philippines challenge the traditional classification of these countries as low-trust societies. One possible explanation is that both nations are characterized by strong kinship ties and tight-knit communities. In such collectivist contexts, respondents may interpret the question of how many “people in this country trust one another” through the lens of localized or relational trust (i.e., particularized trust)—observing cooperation within families, neighborhoods, or ethnic and religious groups—even if they personally do not extend trust to unfamiliar others¹³. Cultural norms that emphasize social harmony and optimism may also contribute to more favorable perceptions of social trust⁶². Taken together, these factors suggest that respondents may interpret survey questions in a normative rather than empirical way, resulting in higher levels of perceived social trust.

Regarding demographic predictors, our random-effects meta-analysis indicates that some variables—such as age, gender, and employment status—align consistently with patterns found in existing measures of social trust. In contrast, others, such as education and religious service attendance, show divergent results. Notably, for education, we found that the overall proportion of individuals expressing trust remained consistent across different education levels. However, country-specific analyses revealed patterns in line with previous research, suggesting that in economically developed countries, trust levels tend to increase with higher levels of education^14,15.

As a supplement to the random-effects meta-analysis, we conducted a population-weighted meta-analysis in which each country’s results were weighted by its 2023 population size. This approach treats each individual equally across the 22 countries, rather than giving equal weight to each country. Consequently, India—the largest country, comprising 39.6% of the combined GFS country populations—received the most weight, followed by the United States, which accounted for 10.7% of the total. Supplementary meta-analytic results were generally consistent with the random-effects findings for gender, religious service attendance, and, to a lesser extent, marital status, but differed for other variables (see Table S23 of the online supplement). These differences partly reflect the greater influence of larger countries in the population-weighted model. For example, the overall proportion of social trust was higher among native-born respondents (0.21 [0.18, 0.23]) than among foreign-born respondents (0.19 [0.16, 0.21])—a reversal of the pattern found in the random-effects meta-analysis but consistent with results from India (see Table S7b of the online supplement).

For other variables, the population-weighted meta-analysis produced top categories that differed from those in the random-effects model and were not fully aligned with results from India or the U.S. For instance, the 50–59 age group indicated the highest overall proportion of perceived trust (0.21 [0.18, 0.24]), compared to the 60–69 (0.20 [0.18, 0.23]) and 70–79 (0.20 [0.17, 0.22]) groups, whereas the random-effects model showed the older age groups with higher proportions than the middle-aged group. Similarly, the top employment status category was “self-employed” (0.23 [0.20, 0.26]), not “retired” (0.21 [0.18, 0.23]), as in the random-effects model. Additionally, while the three education categories showed similar overall proportions in the random-effects model, the population-weighted results revealed differences: “9–15 years” reported the highest trust (0.20 [0.18, 0.23]), followed by “16 + years” (0.19 [0.17, 0.22]), and “up to 8 years” (0.17 [0.15, 0.19]).

This study has several key limitations. First, perceived trust was measured with a single item. To cover a broad range of well-being constructs using a limited space in the survey, the GFS had to assess most constructs—including social trust—with a single indicator. So, we could not employ multiple items to reliably measure perceived trust. As a result, we had no luxury of including a conventional measure of social trust in the survey to control for personal trust in the analysis of perceived trust and compare results between the two measures of social struct. Second, although our measure of social trust captures individuals’ perceptions of how much others in their country trust one another, it remains unclear whom respondents have in mind when answering this question. The concept of the “radius of trust” has been widely debated in the literature, and our data do not permit us to resolve this ambiguity^63,64. Future research would benefit from incorporating follow-up questions to clarify whom respondents are referencing, thereby capturing the nuances of perceived social trust more precisely. Finally, as illustrated by the cases of Indonesia and the Philippines, interpretations of “how many people” may vary across cultural contexts. Although cognitive and pretest interviews were conducted to enhance the cross-cultural validity of survey items²⁸, the possibility of divergent interpretations cannot be fully ruled out. For these reasons, the findings presented here should be interpreted with caution, particularly in light of potential cultural variation in response patterns.

Despite these limitations, this study examines social trust through individuals’ perceptions of how much others in their society trust one another, rather than their own personal trust. This approach provides a descriptive view of national patterns in perceived trust and highlights how societal trust perceptions may be shaped by collective narratives and socio-political contexts beyond individual experience. By framing perceived trust as an indicator of societal well-being, the present study offers a foundation for further exploration of social cohesion and the factors influencing trust across countries and demographic groups.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Y.K. conducted the literature review, performed the data analysis, and drafted the initial manuscript. S.J.J. contributed to data analysis, interpretation of the results, and manuscript drafting. T.J.V. and B.R.J. acquired funding, coordinated data collection, participated in survey design, oversaw the creation of analysis code, and provided critical feedback and revisions to the manuscript. All authors contributed to the revision of the manuscript, and all read and approved the final version.

Funding

The GFS was supported by funding from the John Templeton Foundation (grant #61665), Templeton Religion Trust (#1308), Templeton World Charity Foundation (#0605), Well-Being for Planet Earth Foundation, Fetzer Institute (#4354), Well Being Trust, Paul L. Foster Family Foundation, and the David and Carol Myers Foundation.

Data availability

Data for Wave 1 of the Global Flourishing Study is available through the Center for Open Science upon submission of a pre-registration ( [https://doi.org/10.17605/OSF.IO/3JTZ8](https:/doi.org/10.17605/OSF.IO/3JTZ8) ) and will be openly available without pre-registration beginning Spring 2026. Please see [https://www.cos.io/gfs-access-data](https:/www.cos.io/gfs-access-data) for more information about data access.

Code availability

The analyses were pre-registered with Center for Open Science prior to data access (10.17605/OSF.IO/KX369 and 10.17605/OSF.IO/35XFV). All code to reproduce analyses are openly available in the online OSF repository (10.17605/osf.io/vbype).

Declarations

Competing interests

The authors declare no competing interests. The authors declare the following competing interests: T.J.V. reports consultingfees from Gloo Inc., along with shared revenue received by Harvard University inits license agreement with Gloo according to the University IP policy. All otherauthors declare no competing interests.