Can transnational municipal networks mitigate the carbon pollution of the world's power plants?: an empirical analysis

National governments’ climate commitments are currently insufficient to limit warming below 2 °C, prompting attention to multilevel, polycentric governance and the role of non-state and subnational actors. Transnational municipal networks (TMNs) are membership-based city networks that provide policy knowledge and resources to help cities reduce greenhouse gas emissions, potentially bypassing national policy constraints. A key open question is whether TMNs can help decarbonize power plants—the largest concentrated sources of CO₂—and whether such effects differ between more and less developed countries, as well as for the worst polluting plants. This study leverages a novel global dataset of plant-level CO₂ emissions (2009–2018) to test whether power plants in cities that are members of TMNs emit less CO₂ over time, whether any effect is stronger in less developed countries, and whether TMNs influence the top emitters.

City networks in climate governance have grown since early efforts like ICLEI’s Cities for Climate Protection, expanding in prominence as gaps persist between current policy trajectories and Paris Agreement goals. TMNs such as ICLEI, C40 Cities, and the Global Covenant of Mayors provide benefits including access to resources, learning, lobbying, and goal setting. Prior research documents that TMN participation correlates with greater city-level GHG reductions globally, suggesting bottom-up diffusion of policy norms can complement or bypass top-down national processes. However, critics argue TMNs may promote one-size-fits-all solutions biased toward wealthier cities, inadequately include high-emitting regions, and lack robust target-setting and monitoring. Evidence of TMN impact on upstream producers like the electricity sector is limited, despite its large and growing share of global CO₂ emissions and strategic importance for broader urban decarbonization (e.g., vehicle electrification). TMNs such as C40 advocate focused acceleration strategies prioritizing grid decarbonization, but rigorous plant-level assessments have been lacking. This study addresses these gaps by examining TMN associations with plant-level emissions, potential heterogeneity across development status, and effects on the most egregious emitters.

Unit of analysis: individual power plants worldwide. Dependent variable: plant CO₂ emissions in 2018 (kg CO₂), log-transformed; a lagged endogenous term captures 2009 emissions (log) to assess change over time. Emissions data derive from CARMA (Carbon Monitoring for Action): 2018 edition combines plant-level reports (US, EU, Australia, Canada, India), Platt’s World Electric Power Plants Database, and IEA country production data; non-reporting plants’ emissions are statistically estimated using plant engineering specifications. Key independent variable: TMN membership, a binary indicator coded 1 if a plant is located in a city that, in the prior year (2017), belonged to a TMN with a program focused on the electricity sector. TMN data source: Carbon Disclosure Project Full Cities dataset. Ten TMNs were considered, including C40 Cities Climate Leadership Group, ICLEI programs, Global Covenant of Mayors, Carbon Neutral Cities Alliance, Urban-LEDS, and others. Alternative operationalizations (years of TMN experience overall and longest membership) were tested but not significant. Controls: plant-level (coal primary fuel; nameplate capacity; capacity factor; plant age; government utility ownership), national- and subnational-level (GDP per capita, inflation, unemployment, population change, national fossil fuel power capacity share, climate risk, national energy consumption, NGOs density, liberal democracy), and an index of national energy-climate policies (count of electricity-related policy instruments from the IEA Policies database, including economic instruments, information/education, policy support, regulatory, RD&D, and voluntary approaches). A less developed country indicator captures development status based on income categories. All predictors are lagged at least one year. Modeling strategy: hierarchical linear mixed-effects regression of 2018 log emissions with three random intercepts to account for cross-nesting in countries, sub-national areas (first-level administrative divisions), and parent companies; unbalanced design accommodates varying plant counts per company. Sample and structure reflected in Table 1 random-effects entries: countries (N=147), sub-national areas (N=1683), parent companies (N=10,360), total observations N=13,985.

• TMN membership is associated with significantly lower plant-level CO₂ emissions over time, net of plant structural characteristics and national policy context. In Model 1, the TMN coefficient is −0.171 (SE 0.053), p<0.001. National energy-climate policies show no significant association (−0.009, SE 0.026).
• The TMN effect is especially strong in less developed countries: interaction term Less Developed Country × TMN = −0.476 (SE 0.129), p<0.001 (Model 2), indicating larger emission reductions where cities may face greater resource constraints.
• TMNs are not significantly related to changes among the highest-emitting plants: interactions with Top 10% polluters (0.136, SE 0.236) and Top 25% polluters (0.132, SE 0.137) are non-significant (Models 3–4), despite these groups’ substantially higher emission levels (Top 10% main effect 0.437, p<0.001; Top 25% main effect 1.420, p<0.001).
• Control variables behave as expected: coal use (+0.240, p<0.001), higher capacity and capacity factor (both positive, p<0.001), government utility ownership (+0.122, p<0.001), and greater national fossil-fuel power capacity share (positive) are associated with higher emissions; older plants emit less (−0.006 per year, p<0.001). Prior emissions (2009) strongly predict 2018 emissions (coefficients 0.687–0.805, p<0.001).
• Robustness notes: Alternative measures of TMN experience duration were not significant. Only a small share of plants are in TMN-member cities (~5%).

The findings address three core questions. First, plants located in TMN-member cities emit less CO₂ over time, even after accounting for plant characteristics and national policies, suggesting that TMNs can provide actionable policy knowledge and capacity that translates to measurable performance at the facility level. Second, the effect is more pronounced in less developed countries, consistent with the notion that TMNs bridge resource and expertise gaps where local access to technical services is limited. Qualitative illustrations from Jakarta (ICLEI and C40), Manila and Hanoi (ICLEI’s Ambitious City Promises), and São Paulo (ICLEI and C40) indicate how TMN-facilitated inventorying, target-setting, energy efficiency, fuel switching, and renewable deployment support power-sector decarbonization. Third, TMNs do not significantly influence the worst polluting plants, implying that additional policy levers, stronger regulations, or targeted interventions may be needed for super-emitters. Overall, results reinforce polycentric governance arguments: subnational networks can drive decarbonization independently of national policies, particularly in developing contexts, yet complementary strategies are required to tackle the highest-emitting facilities.

This study provides the first global, facility-level empirical evidence that TMN participation by cities is associated with reduced CO₂ emissions from power plants, beyond the effects of national energy-climate policies. The decarbonization impact is especially strong in less developed countries, indicating TMNs’ value in addressing capacity gaps and advancing polycentric, bottom-up climate governance. However, TMNs show little effect on the emissions of the highest-emitting plants, highlighting a critical frontier for policy and research. Future work should identify when and how TMNs most effectively support the policymaking process, disentangle TMN contributions from other supports, and determine what targeted measures can complement TMN activities to curb super-polluters. Granular, facility-level analyses can refine city-level assessments and guide strategies to accelerate deep decarbonization of the power sector.

• Emissions data for non-reporting plants are model-based estimates (CARMA), which may introduce uncertainty despite reliance on detailed engineering specifications.
• Only a small fraction of plants are located in TMN-member cities, limiting coverage and potentially affecting generalizability.
• The TMN indicator captures membership in networks with electricity-sector programs in the prior year; alternative exposure measures (e.g., years of experience) were tested but not significant, and the study does not establish which stages of policymaking TMNs most influence.
• While models control for extensive plant- and country-level factors and lag predictors, causal inference remains limited; unobserved factors could influence both TMN participation and plant performance.
• TMNs showed no detectable effect among top-emitting plants, indicating limits to their influence without complementary, possibly more stringent, interventions.