Perceiving societal pressure to be happy is linked to poor well-being, especially in happy nations | Scientific Reports

Participants

The present research project was part of a larger cross-national study investigating how individual and cultural values influence emotional well-being and moral attitudes around the world. The initiating sites were based in Australia and Belgium, and the associated researchers contacted potential collaborators via e-mail, in which they outlined the aims and nature of the study, and provided an initial copy of the survey materials (in English). Upon agreeing to participate, all collaborating sites arranged the requisite ethical approval for data collection at their host institution, and translated the questionnaires into their native language (see SI 1 for more information about this process). The original study was approved by the Psychological Sciences Human Ethics Advisory Group in Australia (1647465.2) and the Social-Societal Ethical Committee KU Leuven in Belgium (G-2017 10 954). Each collaborating site was asked to enroll a minimum sample of 100 university students that originated from the nation of testing (e.g., no international or exchange students).

In the end, we collected data from 40 different countries (42 sites), adequately covering all populated continents in the world (i.e., Europe n = 17; Asia n = 10; Africa n = 4; South America n = 4; North America n = 3; Oceania n = 2). A world map with all participating countries can be found in SI 2, together with the final sample size for each site. On average, each country collected 186 participants (SD = 129), with a total sample of 7,443 participants taking part in the study (Mage = 21.81, SDage = 5.60). The balance of gender identification consisted of 32% men, 61.2% women, 0.3% other, and 6.5% unspecified, and the majority of participants (87.6%) were enrolled in a psychology course at the time of the study. All participants provided informed consent.

Procedure and materials

In each country, we adopted a standardized survey battery that was locally translated into participants’ native language (and back-translated by some but not all host institutions; see SI 1 for more information) to evaluate their subjective well-being, alongside their perception of the predominant emotion norms in their country. Participants were only sampled a single time. Next, accessing the public data of the 2019 World Happiness Report, we obtained a global WHI score for each participating country10. Summary statistics and correlations among all measures can be found in Table 1.

Emotional well-being components: natural positive and negative affect

To evaluate the emotional components in subjective well-being, the distinct and global experience of positive (PA) and negative affect (NA67), we compiled a list of four positive (happy, joyful, relaxed, calm) and four negative (sad, depressed, stressed, anxious) emotion items, respectively. The selection of these emotions was based on the circumplex model of affect68 to ensure an adequate representation of different arousal levels. We invited participants to rate their everyday emotional experience both in terms of frequency (How often have you experienced the following emotion during the last month?) and intensity (How intense was your experience of the following emotion?), as both dimensions are known to relate differently to subjective well-being69. For each emotion item, participants provided their response on a 9-point Likert scale that ranged from none of the time (one) to all of the time (nine) for frequency, and from very mild (one) to very intense (nine) for intensity. We averaged same-valenced emotion ratings for each dimension to create a score for PA and NA frequency, and PA and NA intensity.

Cognitive well-being component: satisfaction with life

We assessed life satisfaction with the Satisfaction with Life scale61. This 5-item questionnaire is designed to capture a broad and integrative evaluation of people’s life (e.g., The conditions of my life are excellent.), and concerns the cognitive-judgmental component in subjective well-being13. Participants rated each item on a 7-point Likert scale, ranging from strongly disagree (one) to strongly agree (seven), and we averaged across items to get a global life satisfaction score.

Clinical well-being components: mood complaints

To determine the presence of mood-related symptomatology, experiential factors that usually undermine high subjective well-being52, participants had to complete the Depression Anxiety and Stress Scale70. This 21-item survey is based on the tripartite model of anxiety and depression71, and consists of three 7-item subscales that aim to differentiate between prototypical symptoms of depression (e.g., I felt down-hearted and blue.), anxiety (e.g., I felt scared without any good reason.) and general distress (e.g., I tended to over-react to situations.70). Participants indicated how frequently they experienced each item over the last week on a 4-point scale that ranged from not at all (zero) to most of the time (three), and we averaged responses per subscale to get an indication of each symptom type severity.

Perceived emotion norms

We assessed participants’ perceived societal expectancies to feel positive with the Social Expectancies about Happiness (SEHS51), and not to feel negative with the Social Expectancies about Depression and Anxiety Scale (SEDAS20). The SEHS is a 9-item survey that evaluates people’s global idea about how they think their society expects people to pursue positivity (e.g., I think that society places a great deal of pressure on people to feel happy. or People in my society view people who feel happy as more valuable.; see SI 3 for the full item list). Conversely, the SEDAS is a 13-item instrument that reveals people’s general beliefs about how they think their society disapproves of negative emotional states such as depression or anxiety (e.g., I think society tends to place a lot of pressure on people not to feel depressed or anxious. or I think society accepts people who feel depressed or anxious as normal. [reversed]). For both scales, participants rated each statement on a 9-point Likert scale that ranged from strongly disagree (one) to strongly agree (nine). We averaged across all items (after rescoring the reversed items), so that higher SEHS and SEDAS scores indicated stronger individual beliefs that society pressures people to be happy, and disapproves of negative emotion, respectively. Due to an irreversible coding error the SEDAS scores for Poland are missing.

World happiness index

To get a robust indication of the country-level happiness reported within a particular society, we evaluated countries’ WHI score. A country’s WHI score is based on the average life evaluation of a nationally representative sample10, using the Cantril Ladder54. In this single-item survey, respondents are asked to evaluate the quality of their current life on a 11-rung ladder that ranges from worst possible life (zero) to best possible life (ten). As such, the WHI is more an indication of the average life satisfaction displayed by the inhabitants of a particular country, rather than their global subjective well-being10. Cantril Ladder evaluations and traditional self-report measures for life satisfaction (e.g.,61) are known to correlate very high62.

Because data collection took place in 2019, we adopted the WHI scores for that year (freely accessible online: https://worldhappiness.report/ed/2019/). The countries that took part in our study representatively covered the global ranking (M = 6.11; SD = 0.86), with the Netherlands being the highest ranked country in our sample (7.49; position 5) and Uganda the lowest (4.19; position 136 out of 156). For the participating sites in England, Scotland, Wales and Northern Ireland, we imputed the WHI score of the United Kingdom. In all analyses, we used countries’ actual WHI score, not their corresponding ranking.

Statistical analyses

All analyses in this article were conducted in R (version 4.0.072). To reproduce our results and figures, researchers can consult the data, code and materials at the Open Science Framework (https://osf.io/3aut4/). All methods were carried out in accordance with relevant guidelines and regulations.

Multilevel analysis

To account for the hierarchical structure of the data, we performed our analyses in a multilevel framework, using the lme4 R-package73. Specifically, we ran various two-level models, with persons (n = 7,443) nested within countries (n = 40). In all models, slopes and intercept were allowed to vary randomly across countries to account for possible national differences in the found effects. For an intuitive interpretation of the model parameter estimates, we group-mean centered all person-level predictors. Country-level WHI scores were grand-mean centered. In this way, we effectively separated within- and between-country effects74. All statistical tests were two-sided.

To evaluate how the perception of the societal emotion standard in a country differently relates to subjective well-being as a function of nation’s global happiness level, we ran a series of multilevel models with the various well-being indicators as the outcome of interest (i.e., cognitive, emotional and clinical well-being markers). At the person-level, we either entered participants’ perceived societal pressure to feel positive (SEHS) or not to feel negative (SEDAS) as the focal predictor (separately). At the country-level, we introduced the national WHI scores and evaluated the cross-level interactions with the global intercept and person-level predictor. A generic overview of all model formulae can be found in SI 4.

We emphasize that our multilevel approach inevitably introduces an asymmetry in the specified relation between outcome and predictor75. Because the selection of an outcome and predictor is always somewhat arbitrary with cross-sectional data, we additionally ran all reversed models, together with a third statistical approach in which all variables were within-country standardized (to remove the asymmetry in a multilevel context76). Results can be found in SI 6 and illustrate that this arbitrary decision did not impact our conclusions.

Robustness analysis

With respect to the emotional well-being components, we acknowledge that every item operationalization of a PA and NA composite score is somewhat arbitrary. Because there is little theoretical consensus on how researchers should exactly construct these affective aggregates77, we performed a leave-one-out multiverse analysis for our PA and NA constructs (e.g.,78). For each of the multilevel models that involved PA or NA frequency or intensity as a predictor, we evaluated the robustness of each model parameter under different PA and NA operationalizations. Because we evaluated four specific emotion items for each affective construct, this yielded 15 alternative PA and NA operationalizations, each based on a unique combination of emotion items. We entered each unique affective aggregate as a predictor in the previously outlined models, and evaluated the proportion of models for which the significance test of each estimate (with α = 0.05) yielded identical conclusions as the model in which the PA and NA composites were based on all emotion items (of which the results are presented here). A higher robustness percentage (R%) indicates that the model parameter is less driven by particular PA and NA operationalizations.