Low-carbon consumer choices reflect attitudes, context and demographics and can be nudged either proactively or through follow-up policies, but spillover effects are unpredictable; the literature needs better causal measurement — including AI-enabled, high-frequency data and experimentation — to design targeted, equitable interventions.

Main Finding

Residents' low-carbon consumption behavior results from the interaction of multiple factors—individual cognitive factors, external situational factors, and demographic factors—and can be influenced by two broad classes of policy tools (proactive vs. follow-up). Spillover effects from private low-carbon actions exist but are ambiguous (can be positive or negative). The literature has conceptual and methodological gaps; the review synthesizes existing work and proposes an integrated theoretical framework for antecedents, policy guidance, and spillovers to support urban/regional low-carbon governance.

Key Points

• Antecedents
  • Individual cognitive factors: attitudes, environmental knowledge, values, personal norms, perceived behavioral control.
  • External situational factors: economic incentives (prices/subsidies), infrastructure/availability (public transit, recycling), social norms and peer effects, institutional/market context.
  • Demographic factors: age, income, education, household composition—these modify how cognitive and situational factors operate.
• Policy guidance (timing-based typology)
  • Proactive guiding policies: interventions applied before or at the point of decision (information provision, incentives, nudges, default options); strengths include shaping initial choices and preventing high-emission defaults; limitations include information overload, limited persistence without structural change.
  • Follow-up guiding policies: interventions after an action (feedback, rewards for sustained behavior, enforcement, follow-through infrastructure); strengths include consolidation and reinforcement of behavior; limitations include monitoring costs and potential intrusiveness.
• Spillover effects
  • Behavioral spillovers from low-carbon consumption can be positive (one environmentally friendly choice encourages more) or negative (moral licensing/rebound effects).
  • Direction and magnitude depend on factors such as identity, framing, perceived impact, and policy design.
• Research limitations and controversies
  • Heterogeneous theoretical frameworks and inconsistent operationalizations of “low-carbon behavior.”
  • Methodological weaknesses in the literature: heavy reliance on self-report surveys and cross-sectional designs, limited longitudinal and experimental evidence, small or non-representative samples, and under-explored contextual heterogeneity.
• Contributions of the review
  • Integrates antecedents, policy timing, and spillovers into a coherent theoretical framework aimed at informing urban/regional low-carbon governance and policy design.

Data & Methods

• Study type: Narrative literature review synthesizing existing empirical and theoretical research on residents’ low-carbon consumption behavior, policy tools, and spillover effects.
• Evidence base (as reported/covered in the reviewed literature): cross-sectional surveys, behavioral experiments (lab and field), case studies, natural/quasi-experiments, policy analyses, and some longitudinal studies—though longitudinal and causal designs are relatively scarce.
• Common methodological shortcomings identified: predominance of self-reported measures, limited causal identification, contextual narrowness (few cross-cultural or multi-city comparative studies), and inconsistent definitions/measures of spillover.
• Result: a conceptual/theoretical framework constructed from cross-study synthesis rather than primary data analysis.

Implications for AI Economics

• Measurement & evaluation
  • AI and big-data methods can improve measurement of actual low-carbon actions (e.g., smart-meter, mobility, and transaction data) and reduce reliance on self-report. Economists can combine administrative, sensor, and platform data to better estimate behavior and spillovers.
  • Causal ML and heterogenous treatment-effect methods can help identify which policies work for which subpopulations and when spillovers occur.
• Policy design & targeting
  • Reinforcement-learning and personalization algorithms can optimize timing and content of proactive policies (nudges, defaults, targeted subsidies) for different demographic and cognitive segments to increase cost-effectiveness.
  • AI-driven segmentation can reveal behavioral archetypes (based on observed actions and inferred preferences) enabling more efficient allocation of limited policy budgets.
• Simulating dynamics & spillovers
  • Agent-based models augmented with ML-calibrated behavioral rules can simulate diffusion and potential positive/negative spillovers across communities and markets.
  • Structural and reduced-form economic models can be combined with AI to forecast macro effects (consumption patterns, rebound effects, market demand shifts) of large-scale low-carbon programs.
• Experimentation & causal inference
  • Automated experimentation platforms (A/B tests, multi-armed bandits) can be used in the field to compare proactive vs. follow-up interventions and to detect spillover patterns in real time.
  • Synthetic control and difference-in-differences augmented with ML covariate adjustment can strengthen causal claims about policy impacts.
• Welfare, distributional and market implications
  • AI-enabled cost-effectiveness analysis can incorporate heterogeneity to assess who benefits or loses (distributional consequences) and identify regressive policy effects.
  • Predictions of behavioral change at scale can inform market responses (e.g., demand shifts, firm strategy), regulatory needs, and potential for rebound effects that negate emissions gains.
• Practical cautions & ethics
  • Privacy, fairness, and consent are critical when using personal data to design targeted interventions; AI-driven targeting must be evaluated for equity impacts.
  • Over-reliance on opaque algorithms risks reducing public trust; transparent, interpretable models are preferable for policy settings.
• Research opportunities for AI economists
  • Design and evaluate AI-powered, privacy-preserving nudging systems with rigorous randomized trials.
  • Use causal ML to map heterogenous spillover effects and uncover mechanisms (identity, moral licensing, social norms).
  • Integrate high-frequency behavioral data into dynamic policy optimization models that balance proactive and follow-up interventions.
  • Develop frameworks to quantify macroeconomic and distributional consequences of widespread low-carbon behavior change driven by digital platforms.

Suggested next steps for researchers/policymakers: adopt richer behavioral data sources, prioritize experimental and longitudinal designs, use AI/ML for heterogeneity and spillover estimation while embedding strong privacy and equity safeguards, and pilot combined proactive-plus-follow-up intervention strategies to test persistence and net emissions outcomes.