Perceptions of carbon dioxide emission reductions and future warming among climate experts

The IPCC uses emission scenarios to help policymakers understand potential future climate outcomes, ranging from strong mitigation pathways to futures with very high warming. The IPCC does not assign likelihoods to scenarios to avoid underrepresenting uncertainty, leaving policymakers to select scenarios without guidance on relative probabilities. Recent discourse within the climate community indicates that worst-case scenarios such as RCP8.5/SSP5-8.5 are increasingly viewed as implausible, while the feasibility of achieving best-case outcomes declines with the shrinking carbon budget. Given the influential role of IPCC authors in research, modeling, and public communication, understanding their beliefs about the likelihood of key climate outcomes is important. Moreover, scientists' perceptions may be shaped by second-order beliefs (what they think their peers believe), potentially introducing biases such as the false consensus effect. This study investigates IPCC authors' personal and perceived peer beliefs regarding temperature outcomes, net zero timing, and carbon dioxide removal (CDR).

Background literature highlights the IPCC’s longstanding practice of presenting scenario families without assigning probabilities, due to deep uncertainties in future socioeconomics and climate response. Recent assessments and commentary suggest worst-case high-forcing scenarios (e.g., SSP5-8.5) have become less plausible, while progress toward stringent targets is constrained by dwindling carbon budgets. Research in psychology documents the false consensus effect, where individuals project their own views onto others, potentially shaping second-order beliefs among scientists. Prior surveys of IPCC authors indicated substantial expected warming by 2100 (many expecting ≥2.5–3 °C), and scholarship discusses how scenario selection and communication can influence policy planning. This study builds on that literature by quantifying both first- and second-order beliefs among IPCC authors across working groups and relating them to scenario expectations.

Ethics approval was obtained from Concordia University (Certification Number: 3001621). The sampling frame included IPCC AR6-cycle authors (from SR1.5 onward). Email addresses were identified for 924 authors; four were contacts/collaborators, 11 had autoresponders indicating they would not receive the invitation, and five addresses could not be obtained, yielding 909 invited authors. The online survey ran from October 17, 2022 to December 15, 2022. Participants provided estimates for four outcomes and corresponding peer (second-order) estimates: (1) maximum global warming by or before 2100, (2) likelihood that warming reaches/exceeds 3 °C by or before 2100, (3) year global human CO₂ emissions reach net zero, and (4) rate of carbon dioxide removal (CDR) in 2050. Respondents were told that the most accurate average estimates would direct a $100 USD donation to a charity of their choosing (with a small guaranteed donation per response). Due to non-normal distributions, medians are reported. Statistical analysis used Spearman rank correlations for associations between first- and second-order beliefs; Wilcoxon signed-rank tests to compare first- vs second-order medians; Kruskal–Wallis tests to test differences across working groups (WG1–WG3) based on self-identified AR6 chapter of central expertise; Bonferroni-adjusted pairwise Wilcoxon tests for post hoc comparisons; and Fisher Z tests for comparing correlations. Analyses were conducted in R (Version 4.2.0). To address representativeness, proportional iterative weighting (raking) with the survey package was applied using gender, continent of citizenship, and working group; weighted means were compared to unweighted results. Data cleaning and outlier handling: for net zero year, responses ≥2400 and entries like “9999” (interpreted as NA/never) were treated as missing; five first-order and one second-order net zero responses were removed. For CDR in 2050, values ≥15 GtCO₂/yr were treated as unit-misinterpretations and removed (six first-order, seven second-order). For maximum warming by 2100, three values were removed (one NA; two misinterpreted the question as the highest conceivable warming). Robustness checks re-ran hypothesis tests including outliers, yielding no substantial changes in statistical significance.

• IPCC authors were generally skeptical that warming will be limited to the Paris temperature targets. The median estimate for maximum warming by or before 2100 was 2.7 °C. A majority expected end-of-century warming to exceed 2 °C, and 58% estimated at least a 50% chance of reaching/exceeding 3 °C by or before 2100 (median likelihood = 50%). In the discussion, 86% of respondents were noted as believing warming would exceed 2 °C this century. - Net zero and CDR estimates: Participants expected net zero CO₂ to occur later in the century (e.g., personal median reported as 2075 in later sections), with a median CDR rate of 5 GtCO₂/yr in 2050. - Strong alignment between first-order (personal) and second-order (peer) beliefs supported H1, with Spearman correlations: maximum warming rs = 0.62 (p < 0.001), likelihood of ≥3 °C rs = 0.72 (p < 0.001), net zero year rs = 0.55 (p < 0.001), CDR in 2050 rs = 0.78 (p < 0.001). - Differences across working groups were limited. Only likelihood of reaching 3 °C showed a significant difference (Kruskal–Wallis H = 10.4, p = 0.006), with medians WG1 = 50%, WG2 = 60%, WG3 = 40%; pairwise WG2 vs WG3: p = 0.02. No significant WG differences for predicted maximum warming (H = 3.03, p = 0.222), net zero year (H = 1.46, p = 0.482), or CDR in 2050 (H = 0.31, p = 0.857). - Weighting the sample to reflect IPCC demographics (gender, continent, working group) produced similar mean responses, suggesting limited impact of sample composition on central estimates. - A notable internal inconsistency emerged: many participants projected net zero CO₂ timelines consistent with lower warming (<2 °C) but simultaneously expected higher end-of-century warming (median 2.7 °C), indicating potential disconnects between perceived emissions pathways and modeled climate responses.

The findings address the study’s aims by revealing how IPCC authors perceive future climate outcomes and how those perceptions relate to their beliefs about peers. Strong correlations between first- and second-order beliefs indicate a robust false consensus pattern: individuals at either end of the distribution tend to think their view is close to the community average. This cognitive bias may influence expert communication and scenario selection in research and policy contexts. Differences across working groups were small, occurring only for the likelihood of ≥3 °C, suggesting that perceived scenario likelihoods are broadly shared across domains (physical science, impacts/adaptation, mitigation). The observed inconsistency between expected temperatures and net zero timing suggests either differing assumptions about climate sensitivity, non-CO₂ forcing (e.g., aerosols), or unfamiliarity with the modeled relationship between net zero CO₂ and temperature outcomes. These patterns underscore the importance of clarifying how scenario pathways relate to warming outcomes and of explicitly discussing likelihoods and uncertainties to support more consistent expert judgments and communication to policymakers.

IPCC authors hold influential roles in assessing and communicating climate science. This survey indicates that most authors believe Paris-aligned temperature limits are unlikely and that higher warming is more probable this century, a view consistent across working groups. There is a strong alignment between individuals’ own expectations and what they think peers believe, reflecting a pronounced false consensus effect. These insights can help the climate community better understand internal beliefs shaping research and communication, and encourage open dialogue about perceived scenario likelihoods. The study highlights the need for clearer connections between emissions pathways, net zero timelines, and expected warming, and suggests future work should further probe the reasoning behind expert estimates and improve the clarity and interpretation of scenario information for decision-making.

• Sample representativeness: Although weighting by gender, continent, and working group was applied, the respondent sample may not fully represent the IPCC author population. - Self-reported, rapid estimates: Responses reflect subjective judgments that may rely on heuristics (e.g., availability, anchoring), not detailed analyses. - Question interpretation: Some respondents misinterpreted questions (e.g., maximum conceivable vs best-estimate warming); outlier handling and NA treatments were necessary and could influence results. - Inconsistencies in reported medians across sections (e.g., net zero year) suggest either response variability or reporting/interpretation challenges. - The study did not directly elicit reasoning behind responses, limiting insight into the causes of discrepancies between temperature and net zero expectations. - Scenario likelihoods remain unassigned externally by the IPCC, which may contribute to varied internal beliefs and complicate generalization of findings to formal assessments.