Global economic impact of weather variability on the rich and the poor

Climate change already causes considerable and unevenly distributed economic impacts worldwide, with increasing extreme events and changing variability disrupting economic activity and growth. Prior econometric evidence shows large regional heterogeneity of macroeconomic impacts from temperature variability, extremes, and rainfall, and highlights regressive impacts whereby low-income populations are disproportionately affected. Exposure and vulnerability drive these inequalities, with lower-income communities often more exposed to floods, storms, and landslides. Empirical work indicates heat extremes are regressive and climate change has increased macroeconomic inequality between countries; within-country studies often find stronger impacts on low-income groups. Against this background, the study asks how weather-induced production disruptions propagate through global supply chains to affect consumer consumption, and how these risks differ by income both within countries (across income quintiles) and between countries (across World Bank income groups). The purpose is to perform a stress-test under changing climate conditions, holding socioeconomic conditions fixed, to identify risk factors and hotspots that drive higher vulnerability of specific consumer groups.

The paper situates itself within literature documenting unequal climate impacts and the role of variability and extremes on economic outcomes. Studies have shown regressive impacts of heat and rainfall extremes on households and regions, with higher exposure and lower adaptive capacity among low-income groups. Evidence links climate change to increased global economic inequality and highlights mechanisms such as exposure of informal settlements and unequal flood risk. The authors also reference work on long-term growth impacts of weather variability and extremes and complement it by focusing on short-term consumption losses propagated through trade. The COVID-19 pandemic literature underscored the importance of accounting for supply-chain disruptions, reinforcing the need to study indirect, trade-mediated effects beyond local direct impacts.

The study performs a global, short-term risk assessment using the Acclimate dynamic agent-based supply-chain model. Inputs include: (1) subnational econometric estimates (DOSE dataset) of impacts of temperature variability and extremes, and precipitation variability and extremes, mapped to three sectors (agriculture, manufacturing, services); (2) climate forcing from five CMIP6 models (GFDL-ESM4, IPSL-CM6A-LR, UKESM1-0-LL, MPI-ESM1-2-HR, MRI-ESM2-0), bias-adjusted (ISIMIP) and combined across three emissions pathways (RCP2.6, 7.0, 8.5) for three decades: recent past (2011–2020), present (2021–2030), and near future (2031–2040). Because pathway differences are within natural variability over this horizon, scenarios are not distinguished in results. The econometric yearly marginal effects are downscaled to daily damage functions across five independent impact channels: (i) extreme high/low temperature (quadratic threshold), (ii) daily temperature variability (deviation from monthly mean), (iii) extreme precipitation (exceeding 99.9th percentile), (iv) wet-day precipitation (>1 mm), and (v) annual total precipitation deviations. Subnational parameters are estimated via least squares to match yearly damages over 1979–2014, using population weights as a proxy for spatial economic activity, and applied to bias-corrected CMIP6 daily fields to generate grid-level daily production capacity disruptions, aggregated to regions and sectors. To avoid double counting of indirect effects embedded in the econometric estimates, daily forcing magnitudes are conservatively halved. The independent impact channels are combined multiplicatively, then smoothed with a 7-day rolling mean before feeding into Acclimate. Acclimate simulates 6,243 producing agents across 26 sectors and 264 regions (including subnational detail for the United States and China) and 1,320 utility-maximizing consumer agents (five income quintiles per region) who optimize a two-level CES utility over three consumption baskets (necessary, relevant, other), subject to fixed baseline budgets derived from EORA 2015 MRIO tables and World Bank income quintile shares (Engel’s law representation). Firms are profit-maximizing with myopic adjustments; the model captures propagation of local production shocks along supply chains, price adjustments from disequilibria, and substitution within consumption baskets. Risk is quantified as the 90th percentile of relative consumption loss (baseline deviations) for each consumer group and time period, holding socioeconomic conditions (trade structure, production capacity, income shares) fixed and assuming no adaptation. Robustness checks include alternative smoothing windows (e.g., 14-day), with qualitatively similar patterns though lower absolute risks due to stronger smoothing of extremes.

• Within-country inequality: Lower-income quintiles face higher consumption loss risks across all country income levels and study periods. Inequality is larger in UMICs and HICs, where the lowest quintile’s risk is about twice that of the highest quintile; in LICs the additional risk gap is smaller (about 30%). The mechanism is lower substitutability for necessity-heavy consumption of low-income groups and potential price competition effects where high-income consumers crowd out low-income consumers for necessities.
• Between-country heterogeneity: UMICs and LMICs face about double the consumption risks of LICs or HICs. Three key risk factors explain differences: (1) seasonal compounding of local impacts (e.g., Northern Hemisphere summer heat extremes coinciding with monsoonal rainfall extremes in subtropics), (2) trade dependence patterns—most countries source consumption largely from same-income-level countries, except LICs which import ~65% from HICs and have low self-dependency (~12.5%), thus dampening exposure, and (3) transmission from production to consumption—HICs exhibit stronger dampening (lower production–consumption correlation), reflecting larger capacities and central network positions.
• Exemplary cases: Among LICs, North Korea has the highest consumption risk (~2.1%), linked to high self-dependency and strong production–consumption correlation (≈0.82), while Syria shows lower risk (~1.31%). In LMICs, the Philippines’ higher within-country inequality coincides with consumption risk roughly double its production risk, unlike Ukraine and Uzbekistan where consumption risk is ~25% lower than production risk. In HICs, Germany’s larger economy and centrality help mitigate consumption risk despite larger direct impacts, while New Zealand and Japan show weaker dampening; U.S. states almost avoid transmission from production to consumption.
• Risk amplification with warming: From 2011–2020 to 2021–2030, median consumption risks increase across all groups. Examples: LICs +15% (6–26%) for all quintiles; LMICs show the smallest amplification, +11% (6–15%) lowest quintile to +15% (10–17%) highest; UMICs +14% (6–19%) lowest to +20% (12–25%) highest; in HICs, the highest quintile increases +27% (14–35%) vs +17% (8–24%) for the lowest. By 2031–2040, amplifications continue with greater uncertainty; in HICs the highest quintile faces a median +51% (28–80%) vs +28% (12–41%) for the lowest. Despite higher relative increases among high-income consumers, absolute risks remain highest for low-income quintiles. The United States experiences the strongest relative increases due to low initial risk levels.
• Aggregation by consumption shares indicates macroeconomic risks are dominated by higher-income quintiles because they account for a large share (often >40%) of baseline consumption.

The findings show that weather-induced production shocks propagate through global supply chains to generate unequal consumption risks. The study addresses the research question by demonstrating that low-income consumers are more vulnerable due to necessity-heavy, low-substitutability consumption, and that country-level risk depends on seasonal exposure, trade dependencies, and the efficiency of shock dampening from production to consumption. While HICs currently exhibit lower risk levels and better shock dampening, they also face the largest relative increases under warming, potentially eroding resilience advantages and increasing macroeconomic risks given their larger consumption shares. Incorporating trade-mediated effects extends prior econometric studies focused on direct local impacts, revealing how global supply chains can either buffer or amplify risks. The results underscore the need for policies that enhance resilience not just locally but also along supply chains, tailored to specific country profiles and inequality structures.

This study integrates subnational, sectoral climate impact estimates with a dynamic global supply-chain model to quantify short-term consumption risk inequalities within and between countries under changing climate conditions. It identifies key risk drivers—necessity-focused consumption among low-income groups, dependence on domestic production, inequality-driven market competition for necessities, seasonal compounding of extremes, and trade dependencies—which together shape heterogeneous risk profiles. Policy-relevant insights include: (1) climate-related consumer risks are already widespread and likely to intensify; (2) national adaptation plans should include resilience to supply-chain disruptions and diversification of trade dependencies; and (3) poverty alleviation and inequality reduction can mitigate consumer risks and yield macroeconomic co-benefits. Future work could incorporate adaptive behavior, dynamic socioeconomic evolution, savings and credit mechanisms, and livelihoods outside global trade networks, and refine daily impact downscaling to better capture high-frequency extremes and compound events.

• Behavioral and budget assumptions: Consumers are modeled as myopic utility maximizers without savings; budgets are static and unaffected by short-term income shocks. This may underestimate inequality effects if higher-income households use savings to buffer shocks.
• No adaptation or socioeconomic evolution: Trade structures, production capacities, and income distributions are fixed; adaptation (local or via trade reconfiguration) is not modeled, limiting long-term applicability.
• Scope of economic system: Results apply to agents embedded in global trade networks; subsistence livelihoods are not represented and may experience different direct impact dynamics.
• Impact downscaling and data uncertainties: Yearly econometric effects are downscaled to daily functions using bias-corrected CMIP6 data and population weighting; daily estimates are approximate and smoothed (7-day window). To avoid double counting of indirect effects, daily forcings are halved—an assumption adding uncertainty. High-frequency and threshold processes (e.g., flood triggers) are not fully captured.
• Model abstraction: The production–consumption correlation as a resilience proxy and the chosen CES structure and substitution elasticities entail simplifications that may influence quantitative outcomes, though qualitative patterns are robust in sensitivity checks.