Ten-year panel data confirm generation gap but climate beliefs increase at similar rates across ages

Opinion polls internationally show rising awareness and concern about climate change, yet public discourse often asserts a generation gap in which younger people care more than older generations. Prior studies provide mixed evidence: some report higher concern and climate distress among youth, while others find negligible or inconsistent age effects. Most prior work relies on cross-sectional polling, limiting inference about within-person change and cohort differences over time. This study asks whether climate change beliefs (that climate change is real and caused by humans) have increased over 2009–2018 in New Zealand and whether rates of change differ across 12 five-year birth cohorts (1936–1995). The authors hypothesize that younger cohorts show higher initial levels of belief, but that rates of increase over time are similar across cohorts (i.e., a generation gap in intercepts but not in slopes). Establishing whether belief change reflects ageing versus cohort effects has implications for understanding the malleability of climate beliefs and for intergenerational climate action.

Surveys in the USA and globally show increases in belief that global warming is happening and concern about climate change between 2010 and 2019. In New Zealand, personal importance of climate change rose from 72% to 79% between 2018 and 2019. Pew Research Center data across 23 countries showed increases in viewing climate change as a major threat (56% in 2013 to 67% in 2018), with age differences reliable in only a few countries. Other work indicates youth often report higher concern and climate distress and EU data show greater concern among 15–24-year-olds than those 55+. Conversely, meta-analytic evidence across environmental variables finds mostly negligible relationships with age; older individuals sometimes show higher engagement in conservation behaviors. The mixed evidence suggests age effects may depend on the measure and context, and highlights the need for longitudinal designs to disentangle ageing-related change from cohort differences.

Design and data: Ten annual waves (2009–2018) from the New Zealand Attitudes and Values Study (NZAVS), a longitudinal national probability panel. Participants rated two single-item beliefs on 7-point scales (1=strongly disagree, 7=strongly agree): "Climate change is real" and "Climate change is caused by humans." Sample: 12 five-year birth cohorts spanning 1936–1995; cohort Ns ranged from 495 (1940–1936) to 8819 (1960–1956). Totals reported in Table 1: 56,559 for climate reality and 56,467 for human causation (overall panel referenced as 56,513 New Zealanders). Ethics approval was obtained from the University of Auckland HPEC. Variables: Gender was dummy coded (0=women, 1=men). Analytical approach: Multi-group cohort-sequential latent growth modeling in Mplus to disentangle ageing and cohort effects. Three complementary models were estimated: (1) Age-based trajectory model (single intercept and common linear and quadratic slopes across cohorts; variances/covariances constrained equal across cohorts). (2) Cohort-based (unconstrained) trajectory model (separate intercepts and linear/quadratic slopes for each cohort; variances/covariances constrained equal). (3) Intermediate cohort-based (constrained slopes) model (cohort-specific intercepts with linear and quadratic slopes constrained equal across cohorts). Polynomial growth (linear and quadratic) was used to allow curvature. Fit indices included CFI, RMSEA, SRMR, and comparisons of model fit (including relative chi-square contributions by cohort). The youngest age in each cohort at Time 1 was used to index age across assessment years (e.g., 1990–1986 cohort represented ages 19–28). Gender moderation was tested via multi-group extensions reported in Supplementary Information.

• Mean levels at the sample mean age (~46.17 years): belief in climate change reality M=5.91 (95% CI [5.89, 5.92]); belief in human causation M=5.43 (95% CI [5.42, 5.45]). Belief in reality exceeded belief in human causation across ages. - Trajectories across adulthood were non-linear (U-shaped): quadratic slopes were significant for both beliefs (Table 2). For climate reality, quadratic slope=0.03, p<0.001; linear slope not significant. For human causation, quadratic slope=0.03, p<0.001; linear slope not significant. - Model fit favored cohort-based over age-based trajectories. For climate reality: CFI increased from 0.749 (age-based) to 0.853 (cohort-based), RMSEA decreased from 0.072 to 0.056. For human causation: CFI increased from 0.808 to 0.913, RMSEA decreased from 0.067 to 0.046 (Table 3). - Intermediate model with cohort-varying intercepts but equal slopes fit similarly to the unconstrained cohort-based model, indicating similar rates of change across cohorts with different starting levels. - Cohort-specific change: Climate reality increased longitudinally in 10/12 cohorts (83.3%); the 1955–1951 and 1945–1941 cohorts showed stability. Where change occurred, 8/10 cohorts (80%) showed quadratic (accelerating) increases. Human causation increased in 11/12 cohorts (91.7%); the 1985–1981 cohort showed stability. Where change occurred, 6/11 cohorts (54.5%) showed quadratic increases, with more linear increases relative to climate reality. - Older cohorts started at lower initial belief levels but increased at similar rates as younger cohorts, confirming a generation gap in intercepts but not in slopes. Notably, the strongest quadratic effects were observed in the 1940–1936 cohort (0.52 for climate reality; 0.90 for human causation), suggesting more marked accelerated increases among the oldest participants. - Gender did not moderate findings: women and men exhibited similar cohort patterns—different starting points by cohort but similar rates of increase over time.

The study reconciles seemingly contradictory observations: younger people report higher climate change beliefs, yet belief has risen across the broader public. Longitudinal panel analyses show a generation gap in baseline levels but not in the rate of increase, indicating that all cohorts are increasing in belief at comparable rates. This suggests climate beliefs are malleable across the adult lifespan and offers optimism for broad-based, intergenerational support for climate action. The U-shaped trajectory across ages (higher in early and later adulthood, lower in midlife) may relate to age-linked psychological factors such as openness to change, which tends to decline with age but does not preclude growth in climate beliefs. Although older cohorts begin from lower baselines and may anticipate fewer personal climate impacts, their beliefs increased similarly to younger cohorts, challenging narratives that older generations are unresponsive. The improved fit of cohort-based models underscores meaningful cohort differences in starting points, while the similar slopes imply common societal drivers (e.g., media coverage, scientific consensus visibility, observed climate impacts) that shift beliefs across generations. The lack of gender moderation indicates these patterns are robust across women and men. Implications include potential for coordinated climate communication and policy strategies that can mobilize multiple generations.

This paper demonstrates, using a decade of longitudinal panel data, that New Zealanders across 12 birth cohorts increased in belief that climate change is real and caused by humans from 2009 to 2018. Younger cohorts hold higher initial beliefs than older cohorts, confirming a generation gap in mean levels, but the rate of increase is comparable across ages, suggesting shared societal influences on belief change. Contributions include disentangling ageing versus cohort effects with cohort-sequential latent growth models and providing robust evidence against a generation gap in slopes. Future research should test generalizability in other countries and datasets (e.g., Understanding Society UK), employ richer multi-item measures and behavioral outcomes (e.g., calls for climate action), and experimentally evaluate age-tailored climate communication. Investigating family dynamics and developmental pathways that seed climate beliefs and activism across generations also remains a promising direction.

• Single-country context (New Zealand), which may limit generalizability to other cultural and political environments. - Climate beliefs measured with single items, potentially constraining reliability and construct breadth. - Observational design cannot identify specific causal drivers of belief change; unmeasured factors (e.g., media salience, extreme weather experiences) may differentially influence cohorts. - Age coding used the youngest age within cohorts and the NZAVS samples age 18+, which may affect the youngest cohort’s age range mapping. - Despite large sample size and robust modeling, chi-square statistics are sensitive to minor differences; interpretations rely on comparative fit indices and trajectory coherence. - Data availability is restricted (de-identified dataset available on request), limiting immediate external replication with the exact dataset.