Cultural diversity of religion across 117 countries

Abstract

Humans show marked diversity in cultural values, beliefs, and preferences. Here, we analyze survey responses from 640,110 individuals in 117 countries (representing 89% of the World’s population), and 2,333 sub-national districts, surveyed in the World Values Survey and the European Values Study. Across multiple sets of comparable cultural traits, we find that religion consistently emerges as the cultural trait that exhibits the largest cross-country divergence (measured by the Cultural Fixation Index or simple variances). For example, personal emphasis on religion varies 5-10 times more across countries than emphasis on family, politics, or work effort, and religion is the most divergent quality valued for raising children. Likewise, differences in religious group membership across countries exceed those in other voluntary affiliations. These cross-country differences in religiosity stem partly from differences between religious traditions, but mostly from individual differences in religious involvement intensity. Our findings underscore the importance of religion in shaping cultural differences across societies.

Societies differ widely in cultural values across the world. Here, the authors show that religious values display the greatest cross-country diversity, exceeding that of comparable values, such as values concerning family, work, or politics.

Introduction

Worldwide, there is considerable population variability in human values, beliefs, and preferences, and this diversity in cultural traits is orders of magnitude larger than diversity in genetic traits^1–3. The nature of this diversity, and its persistence over time, remain a scientific puzzle that continues to occupy the social sciences.

Of the myriad ways in which cultural traditions may sustain global divergence in values and preferences, people’s relationship with religion is particularly important. People affiliated with distinct religious traditions differ in their attitudes and preferences^4–6. For instance, religion galvanizes beliefs, rituals, and sacred commitments as means of creating a moral community^7–9. This key feature of world religions consolidates values, beliefs and preferences within a superordinate identity^6,10–12. In addition, secularization has spread in some parts of the world such as Europe, North America, Australia, and East Asia, altering moral values and other social preferences, further increasing differences in levels of religiosity across the globe^13,14. Based on this, we hypothesize that global diversity in religiosity is high. We aim to estimate the magnitude of this global diversity, relative to the diversity of other cultural traits.

Our focus on religion is motivated by evidence that both the strength of religiosity and affiliation with different religious traditions shape important outcomes, including education, income, crime, well-being, conservatism, and prosociality^4,11,15–19. However, attempts to quantify the relative contribution of religion compared to other cultural traits have been few¹². In this study, we investigate religion as one domain of worldwide cultural diversity, and compare to other domains of cultural life, such as family, work, politics, and leisure. This allows us to assess the relative contributions of each domain to overall cultural diversity. Moreover, we distinguish between the intensity of individual religiosity and affiliation with specific religious traditions, in order to evaluate their respective contributions to cultural variation across the 640,110 individuals from 117 countries in our sample.

Previous research has examined the development of cultural diversity over time. Using data for 76 countries available in up to seven waves of the WVS, Jackson and Medvedev examined 40 cultural values and documented increased divergence in emancipative values (defined as a favorable stance on issues such as homosexuality, euthanasia, divorce, prostitution, suicide, abortion), but convergence in values across countries in the same region¹⁴. They attribute this mainly to the widening gap between high-income Western countries and the rest of the world. Comparing 18 countries in 1990 and 2010, other research finds convergence in cultural traits typical for Western rich individualistic countries²⁰. Research analyzing preferences for 80,000 individuals in 76 countries reaches similar conclusions: preferences—such as patience, willingness to take risk, altruism, trust, positive and negative reciprocity—are becoming more similar across countries²¹. Most relevant to the main question we address here is the study by White et al. (2021), which used three waves of the World Values Survey (WVS) across 88 countries to document that individuals who share the same religious tradition and level of religious commitment also tend to share similar cultural values and preferences both within and between countries¹².

In the current study we build on this research but also extend it in several important respects. First, we restrict the analyses across values based on groups of questions that are directly comparable across cultural domains. This enables us to compare cultural traits for the same individual, thus removing confounds related to individual differences and features of survey questions. Second, by exploiting information on the subnational districts of the interviews (instead of just the country), we are able to conduct the analysis exclusively within countries, thus accounting for a large pool of confounders that vary across countries, such as differences in geography, national culture, or institutions. Third, we are equipped with a much larger sample than previous research, which means that our sample is much more representative of the full global sample and that we can analyze changes over time. This method allows us to assess cultural change in a balanced sample over 60 years. Balancing the sample is crucial, as the inclusion of different countries at different points in time would otherwise distort the computation of diversity. Taken together, these steps increase our confidence that it is religious diversity that we detect, rather than other variables associated with it.

In this research, we exploit answers from surveys of 640,110 individuals from 117 countries surveyed in seven waves of the World Values Survey (WVS) (1981–2022) and five waves of the European Values Study (EVS) (1981–2021). These 117 countries represent 89% of the world’s population, all major religious denominations, and all continents. These surveys have been used in foundational work on the importance of cultural values for various outcomes^13,22. We compute the Cultural Fixation Index (CF_st), which is rapidly becoming a standard in cross-cultural studies. The CF_st measures the variance across groups (here: countries or sub-national districts), relative to the within-group variance in cultural traits^1,23,24. Not all cultural values can be meaningfully compared, but the surveys include sets of questions that can. We pick all such sets that include at least one question on religion for at least 350,000 individuals and restrict analysis to comparisons only within each of the sets separately. This leaves us with four sets of questions: 1) qualities that are important to pass on to one’s children, 2) qualities important to one’s own life, 3) voluntary active and 4) inactive membership in organizations. Since we exclusively compare questions on religion to questions regarding other cultural dimensions answered by the same individual, our method reduces differences in socioeconomic background and other demographics. We also conduct additional robustness checks, and show that results hold after controlling for several relevant socio-demographic variables.

A general problem with analyses based on cross-country comparisons is that differences in various other characteristics across countries may explain results. Our methodology of comparing answers by the same individuals narrows the potential bias caused by this substantially. However, cross-country differences influencing values of the same person differently could bias our results. To overcome this problem, we further exploit information on the sub-national district in which respondents were interviewed. Treating the sub-national district as the population group, we document that individuals across 2333 sub-national districts differ more in their values regarding religion, compared to other cultural values. This latter conservative analysis allows comparing sub-national districts within countries, removing variation across countries altogether.

Results

To identify which cultural values differ most across our sample of 640,110 individuals from 117 countries, we compute the Cultural Fixation Index (CF_st), which is commonly used to measure cultural diversity and differentiation^12,24. It is adapted from the Genetic Fixation Index (F_st), which compares the genetic variability within and between any populations^23,25. The CF_st ranges from 0 to 1, where 0 corresponds to the lowest diversity, indicating that the cultural variations across examined populations are identical and 1 indicates the highest diversity, where the populations do not share any cultural traits. We extend beyond standard usage in three respects: We compute the CF_st for each cultural value separately, allow the population to comprise either countries or sub-national districts, and treat the CF_st as a measure of the total existing between-population variance for each cultural trait.

We compute:

$$C{F}_{st,i}=\frac{va{r}_{b,i}}{va{r}_{b,i}+va{r}_{w,i}}$$ 1

for each of the cultural values i examined. var_b,i is the cross-population variance in cultural value i. The population is either a country or a sub-national district. var_w,i is the within-population variance for cultural value i. The index ranges between 0 and 1 with 0 indicating that countries (or sub-national districts) in the sample are similar in terms of cultural value i and 1 indicating that countries (or sub-national districts) differ completely in terms of value i.

Comparability

We ensure comparability of each measure of cultural values i by selecting subsets of questions that are asked within one group of questions, such as “important child qualities" or “values important in your own life." We choose these groups of questions based on the criteria that they i) include a question related to religion and ii) are available for at least 350,000 respondents.

Figure 1 shows the CF_st for each of the cultural values within each of the four comparable value sets. The left panels show the CF_st for the full sample and the right panels show the development in CF_st over time, where time indicates the birth-cohorts of the respondents.

Fig. 1. CF_st across countries.

The left panels show the diversity measured using the CF_st score across countries for the full sample, which includes 599,313 respondents from 117 countries for Personal values, 640,110 respondents from 117 countries for Child qualities, 344,149 respondents from 100 countries for Membership, and 321,507 respondents from 101 countries for Active membership. The right panels show the diversity measured using the CF_st score across countries for each birth cohort, with 502,666 respondents from 85 countries for Personal values, 535,540 respondents from 87 countries for Child qualities, 260,688 respondents from 64 countries for Membership, and 231,417 respondents from 59 countries for Active membership. Code to reproduce the figure is provided in Harvard Dataverse. Result: Religiosity exhibits greater cross-country diversity compared to other cultural values. This pattern has remained consistent for the past 100 years.

Personal values

The first group of comparable values includes answers to the question: “For each of the following aspects, indicate how important it is in your life." The list of qualities includes family, friends, leisure, politics, work, and religion. These questions are answered by 599,313 individuals from 117 countries in six waves (1989–2022). The upper panel of Fig. 1 reveals that values regarding religion differ much more across the 117 countries, compared to any of the other dimensions. The cultural diversity in personal emphasis on religion is 3.4 times larger than the value with the second-largest diversity–emphasis on leisure–or 9.7 times larger than the value exhibiting least diversity–emphasis on family. This pattern persists across cohorts from the 1930s to the 1990s (as shown in the right panel of Fig. 1). If anything, countries are becoming more different over time in terms of individuals’ personal emphasis on religion (i.e., religiosity). Differences based on the other cultural dimensions have not changed much during the past 100 years. These patterns are not due to great changes in the mean values of these cultural dimensions. For instance, the average person has regarded religion as rather important (score 2) throughout the past 100 years, while the average person has regarded family as very important (score 3) throughout the period (Supplementary Fig. S1 in Supplementary Information). The lack of a change in the average emphasis on religion covers declines in some parts of the world and increases in others. For instance, if we had focused the analysis on Europe, we would have seen a decline in the average emphasis on religion, as expected.

Child qualities

Our second set of values includes answers to the question: “Here is a list of qualities that children can be encouraged to learn at home. Which, if any, do you consider to be especially important? Please choose up to five.” The listed qualities are: Independence, hard work, feeling of responsibility, imagination, tolerance and respect for other people, thrift (saving money and things), determination and perseverance, religious faith, unselfishness, and obedience. These questions are answered by 640,110 individuals from 117 countries in seven waves (1981–2022). These are depicted in the second panel of Fig. 1 (Child qualities). Again, these 640,110 individuals from 117 countries differ more in their emphasis on the importance of religion, compared to any of the other dimensions considered. The variation in the emphasis on the importance of religion is 49% larger than the value exhibiting the second-largest diversity–emphasis on hard work–and 5.2 times larger than the value exhibiting the least diversity–emphasis on determination. The pattern has not changed much over time, except that countries have become more similar in terms of their emphasis on the importance of hard work.

Self-reported values

Surveys are sometimes criticized for their exclusive reliance on self-reported values, which may not always accurately reflect individuals’ true beliefs. We have no a priori reason to believe that the biases inherent in self-reports significantly impact the results. Moreover, studies have shown that values and preferences in surveys spanning multiple countries do predict relevant behaviors in laboratory experiments as well as institutional features of societies^3,26. Nevertheless, we proceed with two groups of questions that—although they are still self-reported—are more likely to reflect the respondents’ actual behaviors, not only their self-reported values.

Voluntary membership

Our third values set includes answers to the question: “Please look carefully at the following list of voluntary organisations and activities and say which, if any, do you belong to?" The list of organization types includes religious or church organisations (we term this group Religious); education, arts, music or cultural activities (Culture), labour unions (Labour); political parties or groups (Political); professional associations (Professional); sports or recreation (Sports). Corroborating our previous results, countries differ more in terms of individuals’ memberships of religious organizations than any other voluntary memberships. The pattern is visible in the aggregate data and in every decade of the past 100 years. The pattern is even more pronounced with regard to whether respondents are active members of these organizations (the lower panel of Fig. 1). Interestingly, there is a slight tendency for convergence in membership in religious organizations, but a divergence in personal values. This signifies that the reason for the divergence has more to do with differences in actual personal religious beliefs, rather than differences in membership of religious organizations across the 100 countries in this sample.

Religiosity vs. affiliation with religious denominations

We test the extent to which religious divergence constitutes divergence in the intensity of religiosity, or affiliation to different religious denominations. To examine this question, we remove variation due to religious denominations from the measures of cultural values before computing the CF_st. Unsurprisingly, cultural diversity declines, particularly for values regarding religion (Supplementary Fig. S10 in Supplementary Information). For instance, diversity declines by 25% for Personal values and 9% for Child qualities. The difference between diversity in religiosity relative to the remaining values declines by 35% for Personal values, 44% for Child qualities, 2% for Membership, and 26% for Active membership. This indicates that around one-third of the observed diversity in religiosity can be attributed to differences between denominations, while the majority of the variation is not due to differences across religious denominations.

To further examine the differential contribution of religiosity relative to religious denominations, we perform the analysis separately within each major religion. Specifically, we divide respondents into four religious affiliation categories: “None," “Christianity," “Islam," and “Other" (which includes Judaism, Hinduism, Buddhism, and Other religion) and compute the CF_st separately within each group. We find the greatest divergence in religiosity within each religious group, with the exception of child-rearing values among Christians, where “Hard Work" shows greater diversity (Supplementary Figs. S11–S14 in Supplementary Information). Notably, even among individuals who do not identify with any religion, their stance towards religion remains the most culturally diverse dimension.

These findings suggest that differences in religiosity take up a larger share of cultural diversity than affiliation with different religious denominations. This aligns with previous research indicating that religious affiliation explains only a minor share of the global variation in religiosity²⁷.

Within vs. between variance

The CF_st,i statistic implicitly accounts for differences in within-country variation. Consequently, for a given level of variance across countries, the index will increase as within-country variance decreases. Computing instead the within and between country variances (var_w,i and var_b,i) separately, we find that the results are driven by a higher variance across countries for religiosity (Supplementary Figs. S2, S3 in Supplementary Information). Indeed, the pattern of between-country variance is very similar to the pattern of the Cultural Fixation Index, whereas the within-country variances do not differ much across the cultural dimensions and do not follow a similar pattern across the four groups considered. This corroborates that religiosity differs more across the 640,110 individuals from 117 countries than any of the other values considered and that this is not a byproduct of more homogenization within countries.

Diversity across sub-national districts

A potential caveat of the above analysis—and any other analysis of cross-country differences—is that countries differ from each other in countless ways. A concern is that these cross-country differences may be the true drivers of the identified associations. By restricting our comparisons to answers by the same person, we limit the potential bias caused by this substantially. Only if these cross-country differences influence, say, values regarding work and religion differently would they matter for our results. In theory, however, such cross-country within-individual influences pose a concern.

To address this, we leverage information provided by the WVS and EVS on the sub-national district where each survey was conducted. This information is available for 467,690 individuals in 2333 sub-national districts in 109 countries for the child qualities and 435,241 individuals in 2252 sub-national districts in 109 countries for the personal values. We then remove cross-country variation in each of the values before computing the CF_st measures in order to investigate cultural diversity across sub-national districts within countries. This accounts for variation in national-level factors such as religious majority, Western influence, national institutions, national culture, and geographic features.

Although this exercise reduces the variation in all values considered and narrows the difference between religion and the remaining dimensions, it does not change the conclusion: religion remains the cultural dimension along which these sub-national districts differ the most for all four groups of cultural dimensions considered (cf. Fig. 2). For an analysis of diversity across sub-national districts without removing cross-country variation, see Supplementary Fig. S4 in Supplementary Information.

Fig. 2. CF_st across sub-national districts, controlling for country fixed effects.

The sample includes 435,241 respondents across 2252 districts for Personal values, 467,690 across 2333 districts for Child qualities, 198,560 across 1601 districts for Membership, and 305,818 across 1809 districts for Active membership. Code to reproduce the figure is provided in Harvard Dataverse. Result: Religious values continue to exhibit the highest diversity when considering variation across sub-national districts within countries.

Accounting for age and gender of respondents

Another potential concern is that the observed diversity over time may reflect demographic changes in the sample, such as age and gender. For instance, respondents born in the 1930s were on average older at the time of interview than those from younger cohorts during the survey period (1981–2022). Age and gender could influence the self-reported cultural values, affecting the observed diversity across countries.

To address this, we first remove variation due to age and gender before calculating CF_st over time. This approach allows us to disentangle demographic effects from broader cultural trends. The patterns in CF_st are virtually unchanged (Supplementary Fig. S5), suggesting that the temporal trends are not driven by demographic differences. Applying this approach to the full sample, without considering changes over time, yields the same result: religion remains the cultural dimension along which countries differ the most (cf. Supplementary Figs. S6, S7).

Alternative balancing

When analyzing changes over time (cf. the right panels in Fig. 1), we restrict the main sample by respondents’ birth year so that each country is represented in each of the 13 half-decades. This balancing is crucial for analyzing variation in cultural diversity over time, since later waves of the WVS and EVS include countries that differ more from the first waves. For instance, developed nations were more represented in the first waves than their actual global representation. If this is not accounted for, measures of cultural diversity will mechanically rise over time due to a more diverse sample of countries over time. We show that conclusions are unaltered if we instead balance the sample by interview decade. In Supplementary Fig. S8, we restrict the sample to countries with respondents interviewed in each of the 1990s, 2000s, and 2010s. This results in samples of 61 countries for Personal values, 61 countries for Child qualities, 35 countries for Membership, and 25 countries for Active membership.

Additional robustness checks

We conduct various additional robustness checks.

Cross-country differences

The examination across sub-national districts within countries rules out that country-level factors explain the results. Nevertheless, we include another measure of cross-country differences as an additional robustness check. A substantial part of differences across countries occurs across continents. If such differences explain results, we should see results change if we account for differences across continents. Removing variation across continents before computing the CF_st does not change results (Supplementary Fig. S9).

Socioeconomic status

To account for socioeconomic differences, we consider the highest educational attainment and income of respondents. We account for this variation by removing variation caused by 8 different education levels and 10 different income levels before computing the CF_st. The results are unchanged (Supplementary Figs. S15, S16).

Excluding invalid responses

As an additional robustness check, we exclude respondents who selected more than five qualities in the Child qualities question group. Although the survey explicitly instructs respondents to mark up to five qualities, some individuals select more. This means dropping 22,658 respondents in the full sample and 18,133 respondents in the cohort-based sample. Dropping these responses has no impact on the results (Supplementary Fig. S17 in Supplementary Information).

We conclude that results are not sensitive to differences across countries or individual-level differences in religious denominations, socioeconomic status or demographics. This includes the fact that we compare values for the same individual. Thus, all factors that influence the individual values similarly are implicitly accounted for.

Discussion

This research examined the question of which cultural traits are the most globally diverse. We found that religious cultural traits are the traits exhibiting the largest diversity. Across countries, people diverge most in their emphasis on religion with regard to their own personal values, the qualities they value for their children, and their choice of memberships in voluntary groups, compared to their emphasis on values such as hard work, politics, or family. These findings align with and extend previous research that shows affiliation with religion and religiosity are both important components of cultural similarity across nations¹². It is also consistent with recent work showing that religion is a key part of the observed divergence in cultural values across countries¹⁴.

Here we extend our understanding of religion’s share of global diversity. Our analyses revealed that one-third of religion’s divergence across countries was attributable to affiliation with different faith traditions (Buddhist, Hindu, Christian, Muslim, and Jewish) with the remaining two-thirds due to variation in religiosity, irrespective of the tradition followed. This result suggests that the secularization process that began almost two centuries ago and has taken root in some parts of the world has become important enough that it contributes more to divergence than affiliation with different faiths. The primacy of religion in cultural diversity emerged in variation across countries, across sub-national regions (removing variation across countries), and over time across several birth cohorts. It was also robust to accounting for socioeconomic and demographic characteristics of individuals, such as education, income, gender, and age. Furthermore, because the same individuals responded to survey items that featured religion along with a set of other cultural traits, this method further reduces other individual differences and features of surveys that could have explained the results.

This research has important limitations. Although religion was compared to a wide array of other cultural traits (personal values such as family, work, politics, etc, or virtues in children such as determination, thrift, and unselfishness), it is possible that the World Values Survey missed some important cultural traits that could have been even more dominant than religion in global diversity. For example, White et al. found that affiliation with country, perhaps not surprisingly, was an even larger contributor to cultural similarity across their sample of countries than religion¹². Other sources of diversity that could matter when measured on a global scale are linguistic diversity²⁸ and genetic diversity²⁹. While national identity is a potentially important cultural trait that was unmeasured, it is expected to correlate with other factors that vary across countries, such as national institutions. It is unlikely that this variable would be a counter-explanation, however, since our results were unchanged when holding constant these potential national differences.

As well, despite its very large sample of countries, the WVS and EVS are just two databases of cultural preferences. It is important for future research to assess how well these results replicate in other databases using different features and methods. Another limitation of this study is that–although we are able to examine a longer time-span than most other studies in the field–it still captures diversity in a relatively brief period of time (birth cohorts 1930–1994). While it is encouraging that the finding largely replicated across thirteen birth cohorts and the identified differences are large and persistent across the entire period, it is possible that cultural diversity in the emphasis on religion changes dynamically in different historical periods. This requires further research that extends the time span of data to centuries rather than decades. Work on the creation of historical databases is in progress and could answer this question in the future^30,31. One interesting possibility is that in longer time-spans the global divergence in religiosity would increase. This might be because both religious diversity and strength of religion worldwide were greater in the past than in the present. However, the further back in time we probe, the less secularization there is. We found that individual differences in religiosity explained 2/3 of the cross-country variance in cultural traits. If this effect is largely driven by secularization in the last two centuries, it is quite possible that global diversity in religion might have been smaller in the distant past.

Our study does not allow us to disentangle the different mechanisms that generate diversity in cultural traits, and how religion is implicated in these processes. In the literature three broad factors have been identified: Geography, institutions, and culture. Differences in geography matter as a result of parasite stress, ecological duress, vulnerability to natural disasters, and subsistence, among other factors^32–35. Institutions also shape values and preferences, and tend to be slow-changing and thus can contribute to persistent cultural differences^36,37. Cultural traditions are transmitted cumulatively across generations, and consequently also contribute to persistent differences³⁸. Of course, these factors typically interact with one another and are potentially mutually reinforcing. For example, societies often respond to geographical circumstances by developing locally-appropriate institutions and cultural norms; these institutions in turn shape cultural traditions and traditions shape institutions³⁹. Moreover, societies within the same geographical niche often diverge due to distinct cultural ancestry that could encourage the development of different norms and traditions^40,41. The high, and potentially rising, diversity in religion could have other explanations that include drift, recognition signaling, and competition between religious groups^11,42,43. Future work could address the question of mechanisms by incorporating historical or phylogenetic approaches to test competing hypotheses of religious change and diversity, potentially based on theories of cultural evolution²⁸ or sociological theories of diversification⁴⁴. As well, within secularized groups there is important diversity in cultural attitudes and preferences; that would be another valuable extension in future studies.

Summary of results

Societies differ across the globe in their norms, psychological outlook, and economic preferences. A puzzle within the social sciences is why such differences tend to persist over time^45–48. For millennia, religious systems have been intertwined with cultural traditions, galvanized social cohesion, and informed what people consider true, sacred, and moral in societies worldwide. It is, therefore, plausible that people’s relationship with religion reflects a significant share of global diversity in cultural values and preferences. We find that different levels of religiosity and affiliation with different religious traditions take up a potent share. This result points to the influential place of religions, as well as secularization, in the cultural evolution of societies worldwide.

Methods

Here we describe our data and outline the main aspects of our methodology. Further details are provided in the Supplementary Information.

Exemption from ethics approval

This study used anonymous and publicly available data. No additional ethical approval was required. This was confirmed by the Research Ethics Committee at the Department of Economics, University of Copenhagen.

Data source

To measure cultural values across as many countries or subnational regions as possible, we use the pooled World Values Survey and European Values Study⁴⁹. We use all available seven waves (1981–2022) of the World Values Survey (WVS) and all available five waves (1981–2021) of the European Values Study (EVS). The surveys are representative of the full population at the country-level, and hold numerous questions on cultural values on trust, individualism, and religiosity, for instance.

Cultural values

To ensure comparability of the different dimensions of culture, we only compare within values that are asked within the same set of questions. We restrict to sets of questions that include at least one question on religion and that are answered by at least 350,000 respondents. The 350,000 limit is chosen to keep the task manageable and to ensure representation across countries (maximum N = 117), birth decades (T = 13), and sub-national districts (maximum N = 2333). We have checked results for questions available for fewer respondents and conclusions are unaltered; religion clearly dominates cultural diversity. We arrive at four sets of values. Two sets include respondents’ stated values, two sets include memberships of voluntary organizations and thus reflect respondents’ revealed values.

Our first set of values–Personal values–includes answers to the question: “For each of the following aspects, indicate how important it is in your life" (Questions A001–A006). The list of qualities are family, friends, leisure, politics, work, and religion. The original variables took values 1 (Very important), 2 (Rather important), 3 (Not very important), and 4 (Not at all important). We recoded the variables to 0 (Not at all important), 1 (Not very important), 2 (Rather important), 3 (Very important).

Our second set of values–Child qualities–consists of answers to the question: “Here is a list of qualities that children can be encouraged to learn at home. Which, if any, do you consider to be especially important? Please choose up to five" (Questions A027–A043). The listed qualities are: Independence, hard work, feeling of responsibility, imagination, tolerance and respect for other people, thrift (saving money and things), determination and perseverance, religious faith, unselfishness, and obedience. The variables take on values 1 if the particular quality was mentioned, 0 if not.

Our third set of values–Member–includes answers to the question: “Please look carefully at the following list of voluntary organisations and activities and say...which, if any, do you belong to?" Questions include A065, A066, A067, A068, A072, and A074. The questions in between are not asked in all waves and thus include answers by fewer than 350,000 respondents. These omitted organization types are “Local community action on issues like poverty, employment, housing, racial equality," “Third world development or human rights," “Conservation, environment, animal rights groups," and “Youth work (e.g. scouts, guides, youth clubs etc.)." The conclusion is unaltered if we include these questions. The list of organization types includes religious or church organisations (we term this group Religious); education, arts, music or cultural activities (Culture), labour unions (Labour); political parties or groups (Political); professional associations (Professional); sports or recreation (Sports). The variables take on values 1 if the particular membership was mentioned, 0 if not.

Our fourth set of values–Active member–includes answers to the question of whether the respondents are active or inactive members of the same organizations. Questions include A098–A105. In addition to the organizations in the third group, this set also includes environmental organizations, as this is answered by more than 350,000 respondents. The variables take on values 0 if the person is not a member, 1 if he/she reports being a member, and 2 if he/she reports being an active member.

Within each group, we remove respondents who do not answer all questions within the particular group.

Time-variation and balanced samples

To analyze the diversity in cultural values over time, we attach a survey response to the respondent’s birth year. We hereby assume that an individual’s cultural values are closely correlated with the predominant values of his or her birth year. This assumption is based on a literature that documents how attitudes, preferences, and beliefs to a large extent are formed in childhood and early adulthood^50–52. Our main sample is restricted to respondents with birth years between 1930 and 1994. This time span is chosen to maximize the length of time studied and the number of countries with information for each half-decade. When analyzing variation across time, we balance the sample so that the same countries are represented (by at least 10 respondents) in each half-decade within the time span. The average country-half-decade has 450 respondents. In Supplementary Fig. S8, we restrict the sample based on interview year instead. In this analysis, we restrict to countries with respondents interviewed in every decade of the 1990s, 2000s, and 2010s. To group respondents by decade, we take the first year of each wave’s time interval and include 2020 in the 2010s¹⁴.

Cultural diversity

We examine cultural diversity using two different measures: the Cultural Fixation Index described in Eq. (1) and simple between-country or between-sub-national district variances.

The CF_st index was originally designed to capture cultural distances between specific populations, such as countries or religious groups^12,24. This dyadic approach focuses on measuring how much two populations differ in their cultural values. In contrast, we compute the variance across individual responses. This variance captures how much individuals within the entire sample differ from each other in their cultural values and can be decomposed into between-group and within-group variances based on countries or sub-national districts. By structuring the data at the individual level and using the xtsum command in Stata, we can calculate the relevant variances for each cultural value across different population definitions without needing to structure the data into specific dyads.

Sub-national analysis

To construct Fig. 2, we remove cross-country variation from each of the cultural dimensions before computing the variance within and across sub-national districts. We do so by the following procedure. We regress each of the cultural values on a full set of country dummies and save the residuals. The variation in these residuals equals the variation in the cultural values, excluding the variation from cross-country differences. Thereafter, we compute the variance across and within sub-national districts for these residuals.

Individual-level controls

We use information on the socioeconomic status of the respondent, which includes self-reported education levels (variable X025) and income levels (variables X047_WVS and X047_EVS), and their demographics, including gender (X001) and age (X003). To account for variation at the individual-level, such as religious denominations, educational attainment, and income, we follow the same procedure as described for removing cross-country variation. For instance, to account for variation in religious denominations, we first regress each of the cultural values on a full set of dummies for religious denominations, save the residuals, and compute the CF_st for the residuals. These estimations include accounting for dummies for 10 religious denominations, 8 levels of educational attainment, 10 income groups, 92 age levels, and 2 genders.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

L.L.S., J.S.B., and A.B.K. designed the first version of the paper based on explanatory power models, and did the first analysis. J.S.B. re-designed, did the analysis, and wrote the draft for the current version of the paper. A.B.K. conducted supplementary analysis and A.N. contributed to the framing of the results and writing of the paper.

Peer review

Peer review information

Nature Communications thanks Joseph Bulbulia and Joshua Jackson for their contribution to the peer review of this work. A peer review file is available.

Data availability

The analysis uses the integrated European Values Study (EVS) and World Values Survey (WVS) dataset (1981-2022), available from the European Values Study website (https://europeanvaluesstudy.eu/methodology-data-documentation/integrated-values-surveys/data-and-documentation/). Access requires free registration at the data portal. Detailed instructions for obtaining and processing the data to replicate our analysis are provided in the replication repository at Harvard Dataverse (10.7910/DVN/OQONVO). No additional restrictions apply to the data used in this study.^53–55

Code availability

All code and documentation required to reproduce the analyses and figures are available in the replication repository at Harvard Dataverse (10.7910/DVN/OQONVO).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-67671-y.