How climate policy commitments influence energy systems and the economies of US states

US states have enacted a wide range of policies to mitigate climate change across the past two decades, including electricity market restructuring, renewable energy standards, emissions limits, and incentives for clean energy and efficiency. Despite this policy activity, scholars lack a holistic measure capturing the overall stringency of states’ climate policy regimes, limiting rigorous assessments of policy effectiveness. This study addresses that gap by asking how states’ climate policy regimes have evolved over 2000–2020 and what environmental and economic consequences these regimes have produced. The authors build an aggregate state climate policy stringency index from 25 policy instruments and use it to evaluate impacts on CO₂ emissions, energy production and consumption, and economic outcomes. The purpose is to provide a comprehensive, comparable measure of state climate policy and to assess whether stronger policy regimes reduce emissions without harming state economies.

Prior work has often focused on single policies, especially Renewable Portfolio Standards (RPS). Studies generally find that RPS policies increase renewable generation capacity but provide mixed evidence on the share of renewables in states’ energy mixes. Research on other policies such as public benefit funds, net metering, and green power options has also yielded mixed results. Existing approaches commonly use additive indices or regressions including each policy independently, which can obscure effects due to multicollinearity and fail to account for heterogeneity in policy design and stringency across states. Recent multidisciplinary work suggests that factor-analytic approaches can better capture latent constructs like policy regimes. Building on this, the paper develops a holistic measure that incorporates multiple policies and their design variations to overcome limitations of single-policy and simple additive measures.

Policy measurement: The authors compile a dataset for all 50 states and DC covering 25 climate and energy-related policies from 2000–2020, drawn from advocacy groups, government websites, and academic sources. Policies were selected if they could affect CO₂ emissions, be systematically coded for all states over time, and were eligible for adoption in all states. Each policy is coded using the most granular feasible scheme (binary, ordinal, or continuous), with attention to design heterogeneity across states (e.g., different RPS or net metering designs). The most widely adopted policies include on-site renewable generation incentives (e.g., net metering), solar tax credits, RPS, utility decoupling, and GHG reporting. Fewer states adopted emissions performance standards, GHG caps, or state preemption of local gas bans. Index construction: Using these policy measures, the study estimates a latent state climate policy stringency index via Bayesian factor analysis for mixed ordinal and continuous data (Quinn 2004), implemented with the dbmm package (Stan via CmdStanR). Policies’ intercepts (difficulty) and discrimination (slope) parameters are held constant over time, allowing comparison across years. The model weights each policy according to the information it provides about overall policy stringency and accounts for policy design variation between states. Four chains were run for 1000 iterations (500 warm-up, 500 saved). The resulting index provides annual estimates for each state (standardized mean 0, SD 1). Convergent validity is assessed against ACEEE energy efficiency scorecards, and construct validity is assessed against state policy liberalism and public ideology; correlations are strong and improve over time. Robustness checks show results are insensitive to inclusion/exclusion of any single policy. Outcomes and data: Environmental outcomes are energy-related CO₂ emissions per capita from (a) the electricity sector and (b) the total economy, from EIA. Energy outcomes include total and source-specific production and consumption, and net electricity generation, from EIA; variables are logged, with source-specific measures expressed as shares. Economic outcomes include GDP per capita, jobs per capita, wages per worker (BEA), and average annual electricity prices (EIA). Population (Census) is used to form per-capita measures. Union membership data come from CPS-based databases. Impact estimation: Time-series cross-sectional OLS regressions estimate the association between climate policy stringency and outcomes. Models include state fixed effects to absorb time-invariant unobserved heterogeneity and region-year fixed effects to capture common shocks and regional trends (e.g., COVID-19). Standard errors are two-way clustered by state and region-year. All dependent variables are logged; coefficients are scaled to approximate percent changes associated with a 1-SD increase in the policy index. Because the index is measured with uncertainty, coefficients are corrected for measurement error using the Method of Composition: sampling from the posterior distribution of the index for each state-year, re-estimating regressions across 100 draws, and summarizing the resulting coefficient distributions. Supplementary analyses include additional controls for lagged economic indicators and sensitivity to policy set composition.

- Policy stringency growth: The average state increased its climate policy stringency by 1.76 standard deviations from 2000 to 2020. In 2020, California and New York are among the most stringent; West Virginia is among the least. The 1.76 SD increase is roughly the 2020 gap between California and Arizona, or between Virginia and West Virginia. - Emissions impacts: A 1 SD increase in the climate policy index is associated with approximately a 5% reduction in per-capita CO₂ emissions from the electricity sector (t ≈ -2.3, measurement-error corrected) and about a 2% reduction in economy-wide per-capita CO₂ emissions (t ≈ -2.06, corrected). - Energy system impacts: Stronger climate policy is associated with about 3% reductions in energy and electricity consumption measures (t ≈ -2.3, -2.0, -2.1, corrected). Evidence for reduced electricity production is weaker (t ≈ -1.875). The study does not detect effects on renewable energy production (aggregate) or coal-fired electricity production. - Economic outcomes: No statistically significant effects are found on electricity prices, GDP per capita, jobs per capita, or wages per worker (t ≈ 0.74, -1.4, 1.5, -1.4). Results suggest no detectable economic harms from more stringent climate policy. - Correlations illustrating index validity: Cross-sectional correlation between the index (2019) and electricity-sector CO₂ per-capita emissions (2020) is r = -0.63; correlation between change in the index (2000–2019) and change in emissions (2000–2020) is r = -0.19. The index correlates strongly with simpler policy measures: r = 0.85 with RPS target aggressiveness and r = 0.96 with the count of policies, while capturing additional variation due to policy design and weighting. - Sufficiency relative to goals: At the observed pace (average 1.76 SD increase over two decades), the model implies about a 9.6% reduction in per-capita emissions across two decades—insufficient to meet the Paris Agreement-aligned US goal of a 50% reduction by 2050. States vary widely: e.g., West Virginia increased only ~0.8 SD; California and New York increased ~3 SD.

The study demonstrates that a comprehensive, design-sensitive index of state climate policy captures meaningful variation in policy regimes and clarifies their consequences. Stronger climate policy regimes are associated with lower electricity-sector and economy-wide CO₂ emissions without detectable adverse effects on state-level economic indicators or electricity prices. The mechanism appears to operate primarily through reductions in energy and electricity consumption—consistent with efficiency improvements and with the possibility that subnational heterogeneity and interstate energy trading mitigate in-state shifts in generation portfolios. By integrating multiple policy instruments and their design stringency, the index offers analytical leverage beyond single-policy or simple additive measures, helping disentangle effects otherwise obscured by multicollinearity or coarse coding. Despite measurable benefits, current state efforts, on average, are too modest to achieve Paris-consistent emissions trajectories, underscoring the need for more ambitious and comprehensive policy action.

This paper contributes a validated, annually comparable index of US state climate policy stringency (2000–2020) that accounts for the breadth and design of 25 policy instruments. Using this index, the authors show that stronger policy regimes are associated with meaningful reductions in CO₂ emissions, achieved largely via reduced consumption, and do not impose detectable economic costs in terms of prices, jobs, wages, or GDP. However, current state policies are insufficient to reach Paris-aligned targets. Future research directions include: (1) identifying political and institutional drivers of temporal dynamics (e.g., the 2005–2008 uptick), (2) deeper analysis of the politics and impacts of specific policy instruments, (3) assessing how policy changes shape public opinion on energy and climate, (4) evaluating the distributional impacts of climate policy across communities, and (5) extending the index to monitor evolving policy regimes and their environmental and economic effects over time.

- Policy coverage and scope: The index excludes policies only relevant to a small number of states (due to geography) and cannot capture every action states take on climate, potentially omitting important sector-specific or localized measures. Most policies included focus on the electricity sector, reflecting where most state action has occurred, which may limit generalizability to other sectors (e.g., transportation, industry). - Measurement limitations: Although the Bayesian factor model reduces measurement error and accounts for policy design heterogeneity, latent-variable estimation is subject to remaining uncertainty. The authors correct for this in outcome models, but some measurement error may persist. - Observational design: Estimated effects rely on time-series cross-sectional regressions with fixed effects; while these control for time-invariant state factors and common regional shocks, unobserved time-varying confounders may remain. - Outcome sensitivity: Some targeted outcomes (e.g., renewable generation increases or coal declines) show null effects at the state level, which may reflect interstate energy trading or data aggregation that obscures within-state shifts, limiting interpretation of specific mechanisms.