Changes in social norms during the early stages of the COVID-19 pandemic across 43 countries

The study investigates whether a global collective threat—the COVID-19 pandemic—altered social norms across societies. Prior research on cultural tightness–looseness posits that societies exposed historically to ecological and social threats (e.g., disease, warfare, environmental catastrophes) evolve tighter cultures with stronger norms and more sanctioning, while less-threatened societies evolve looser norms. Although computational models and cross-sectional evidence suggest threat-tightening associations, cultural evolution may be slow, raising the question of whether an acute, global pandemic triggers short-term tightening or leaves cultural patterns largely stable. The authors aim to test changes in: (i) cultural tightness–looseness, (ii) situation-specific social norms across various domains, (iii) metanorms (approval of punishments for norm violations), (iv) frequency of punishing norm-breakers, and (v) hand hygiene norms. They also examine mechanisms potentially associated with norm change, including perceived disease prevalence, fear of COVID-19, and government policy stringency. This work informs theories of cultural change and practical understanding of societal responses to collective risks.

The paper builds on tightness–looseness theory and a body of work linking historical threat exposure to stronger social norms and stricter socio-political institutions. Prior studies include cross-national surveys, ethnographic data, longitudinal experiments, and analyses of regional tightness, suggesting threat correlates with tighter norms and differing innovation and governance patterns. Recent work linked looser cultures to poorer early COVID-19 outcomes. Computational models indicate that sudden threat increases can induce tightening. Conversely, theories of cultural evolution emphasize slow change, implying resilience of norms to short-term shocks. Experimental findings suggest that specific threats may strengthen domain-relevant norms (e.g., cooperation under collective loss), implying threat-specific, mosaic-like adjustments rather than uniform tightening. The literature thus motivates testing both general tightening and domain-specific changes (e.g., hygiene norms) in response to the pandemic.

Design and sample: Two-wave international survey with pre-registration (OSF: initial March 23, 2020; detailed plan Oct 22, 2020). Wave 1 collected April–December 2019 (pre-pandemic); Wave 2 collected March–July 2020 (early pandemic). The dataset includes 30,431 valid respondents from 55 cities in 43 countries, sampling both students (~200 per country target) and non-students; translations in 30 languages with forward–back translation. Attention checks applied; under-18s excluded. For two locations (Israel, Poland), a subset of participants were matched across waves; six additional locations had identifiable repeat participation without one-to-one matching. Ethical approvals obtained across multiple institutions. Measures: - Cultural tightness–looseness (TL): Standard six-item scale (1–7), standardized for response styles. Country-level stability checks linked to Wave 0 (2000–2003) data show high rank-order stability prior to COVID-19. - Situation-specific social norms: Appropriateness ratings (0–5) for norm-violating behaviors across multiple domains (mean ≈ 1.13, SD 0.6). - Metanorms: Appropriateness (0–5) of responses to norm violations (verbal/physical confrontation, ostracism, gossip, non-action) across five scenarios (25 items total; mean ≈ 2.22, SD 1.25). - Punishing frequency: Self-reported frequency (1–5) of confronting, gossiping, ostracizing inappropriate behavior; analyzed collectively and separately (mean ≈ 2.98, SD 0.59). - Hand washing norms: Endorsement of hand washing in six situations (before/after meals, after defecation/urination, when coming home, after shaking hands); both count of endorsed situations and item-level probabilities analyzed. Indonesia excluded due to translation error. - Mechanisms: Fear of COVID-19 (three items; 6-point scales; Cronbach’s α = 0.84 overall; country α mostly ≥ 0.60), perceived local COVID-19 prevalence (0–100%), and government Stringency Index (0–100; Oxford tracker). External daily-matched cases/deaths per million from Our World in Data used for robustness. Analytic strategy: - Stage 1 (between-wave changes): Multilevel models with random intercepts (country, city, and individual/observation levels as appropriate) estimating Wave 2 effect on each outcome, controlling for age, gender, and student status; varying-slope models to assess cross-country heterogeneity; robustness checks controlling for cases/deaths. Equivalence tests (TOST) with pre-registered SESOIs assess practical significance. - Stage 2 (mechanisms): Country-level OLS regressions of change scores (Wave 2 – Wave 1) on fear, perceived prevalence, and policy stringency with heteroskedasticity-robust SEs; akin to a multi-level exposure design (no untreated control), with analyses repeated for overall hand washing and COVID-relevant items (after shaking hands, coming home) and for punishment sub-items. Additional checks examined whether geography or sample composition drove results. Power: Simulations indicated high power (>95%) to detect small effects (Cohen’s d ≈ 0.10) for TL change; sensitivity analyses suggested detectable medium-to-large effects for mechanism models with ≈41 countries.

- Cultural tightness–looseness: Small but statistically significant decrease after COVID-19 emergence (Cohen’s d = 0.11; b = -0.028, 95% CI [-0.047, -0.009], p = 0.003). Heterogeneous across countries (varying slope b = -0.037, 95% CI [-0.073, -0.001], p = 0.042; τ² = 0.01). Most country-specific changes not significant (35/43). Higher fear of COVID-19 associated with greater TL decreases (b = -0.081, 95% CI [-0.157, -0.005], p = 0.037). Perceived prevalence and stringency not significantly associated. - Situation-specific norms: Slight decrease (Cohen’s d = 0.04; b = -0.017, 95% CI [-0.028, -0.006], p = 0.006). - Metanorms: No significant overall change (Cohen’s d = 0.03; b = 0.006, 95% CI [-0.001, 0.031], p = 0.120). Heterogeneous by punishment type: approval of ostracism slightly increased (b = 0.028, 95% CI [0.015, 0.042], p < 0.001); approval of passing (non-action) slightly decreased (b = -0.024, 95% CI [-0.035, -0.013], p < 0.001). No significant change for verbal or physical confrontation. - Punishing frequency: Significant decrease (Cohen’s d = -0.07; b = -0.034, 95% CI [-0.047, -0.022], p < 0.001). Robust in varying slopes models. Decreases observed for confrontation (β = -0.221, p = 0.035) and ostracism (β = -0.091, p < 0.001); gossip showed no significant overall change (β = -0.011, p = 0.335). Countries with higher COVID-19 levels tended to have greater decreases in gossip frequency (β = -0.139, 95% CI [-0.261, -0.016], p = 0.028). Mechanism associations for the overall index were not robust. - Hand washing norms: Substantial increase across nearly all countries, especially for items most relevant to COVID-19 (e.g., after shaking hands, when coming home). Overall effect sizable (e.g., β ≈ 0.420, 95% CI [0.390, 0.450], p < 0.001; Cohen’s d ≈ -0.32 reported descriptively, indicating a marked change in endorsement). Effects stronger for COVID-relevant items. Fear of COVID-19 associated with increased hand washing norms; perceived prevalence and stringency showed limited or inconsistent predictive value. - Equivalence tests: For tightness–looseness, situation-specific norms, metanorms, and punishing frequency, observed changes were statistically equivalent to SESOIs (i.e., small in practical terms). For hand washing norms, changes exceeded upper equivalence bounds, indicating a practically meaningful increase. - Stability checks: TL rank-order stability from 2000–2003 to 2019 and to 2020 suggests high pre-pandemic stability; little evidence of pre-existing TL time trends.

The results indicate that, in the short term, the COVID-19 pandemic did not broadly tighten social norms. Instead, cultures appeared largely stable, with small decreases in overall tightness and punishment frequency, minimal changes in situation-specific norms and metanorms, and a pronounced strengthening of hand hygiene norms—those most directly relevant to disease mitigation. This pattern supports the notion of domain-specific norm adaptation to salient threats rather than uniform tightening across all domains. Fear of COVID-19 emerged as a modest correlate of TL decreases and hand washing increases, whereas perceived prevalence and policy stringency had limited explanatory power. Together, these findings suggest that immediate responses to collective threats may selectively reinforce norms that directly mitigate the threat, while broader cultural structures change slowly. The results refine tightness–looseness theory by highlighting temporal dynamics and domain specificity, pointing to a mosaic of norm adjustments shaped by the nature of threats.

The study provides cross-national evidence that early in the COVID-19 pandemic, social norms largely persisted with minimal change, except for a significant strengthening of hand hygiene norms. Cultural tightness and punishment frequency slightly decreased, and most other norms and metanorms remained stable. These findings suggest that acute global threats may induce targeted, domain-relevant norm shifts rather than broad-based tightening, at least in the short term. Future research should examine longer-term trajectories, differential effects of various threat types (e.g., pandemics vs. natural disasters), individual and socio-demographic moderators, and causal pathways linking perceived threat, policy signals, behavior, and norm internalization.

- Sampling and generalizability: Convenience samples (students and non-students) with over-representation of younger participants may limit generalizability, particularly to older or high-risk populations. - Causality: The design mitigates some endogeneity by comparing pre/post waves and modeling exposure heterogeneity, but cannot conclusively identify causal effects; unobserved time trends may remain. - Measurement and translation: Although standard translation procedures were used, one country (Indonesia) had a translation error for hand washing items; fear scale reliability was lower in Kenya (analyses robust to excluding Kenya). - Power for mechanisms: Country-level mechanism analyses may lack power to detect small effects; detectable effects were medium-to-large in size. - Heterogeneity and model assumptions: Effects varied across countries; although robustness checks were conducted (including cases/deaths and varying slopes), multiple testing adjustments were not applied. - Short-term window: Wave 2 captured early pandemic responses; longer-term norm changes may differ and were not observed.