Study Brings Light to Heated Discussion over Rising Electricity Bills
Article: Ryan Wiser et al., “Factors influencing recent trends in retail electricity prices in the United States,” The Electricity Journal 38:4 (December 2025). https://doi.org/10.1016/j.tej.2025.107516.
Consumers and policymakers have noticed the recent increase in electricity bills, sparking considerable debate over potential causes. After years of relative stability, rates surged with a sharp rise in natural gas prices in 2022 and went up again in the first half of 2025. A lot of attention has been paid to the role of growing electricity use by data centers; however, most commentary has proceeded without much in the way of thoughtful analysis. A study on drivers of state-level trends in U.S. retail electricity prices from the Lawrence Berkeley National Lab begins to fill the gap.
The Berkeley team relied on data collected by the U.S. Energy Information Administration, supplemented with information on state policies and other factors of interest. Their analysis examines how average state-level retail prices of electricity appear to correlate with several potential causes over the period of nominal price growth (2019-2024). The authors used inflation-adjusted prices to focus more clearly on electric industry-related forces. Among the potential factors of interest were changes in electric consumption, utility-scale wind and solar power production, state renewable energy mandates, distributed solar power, and natural gas prices. The Berkeley team also used regression models to link changes in potential factors to changes in average state-level electricity prices.
Despite concerns over rising costs, national average electricity prices mostly tracked inflation from 2019 to 2024. In other words, real electricity prices did not increase despite nominally higher rates. Yet movements in the national average conceal significant differences among states. The authors found that real prices fell in 31 states and rose in 17. Additionally, while overall prices were stable, rates for residential customers were somewhat higher after adjusting for inflation while commercial and industrial rates trended lower.
The most striking correlation identified in the analysis is that between load growth and retail prices. Although many analysts have attributed increased electricity prices to growing demand for power by data centers, the study found that growing consumption actually correlates with falling rates. Each 10 percent increase in energy consumption in a state over the five-year period yielded a reduction of about 0.6 cents per kilowatt-hour. Earlier work from the Berkeley Lab identified spending on distribution and transmission infrastructure as the fastest growing component of electricity bills. Spreading such costs over a broader base reduces the per-unit share of the cost burden.
The analysis produced several other key findings. No simple link was found between renewable energy and retail rates. States with more aggressive renewable energy policies and weaker renewable resources tended to see rates increase, as did states with substantial growth in distributed energy resources like rooftop solar. However, states with significant growth in renewable energy production did not see associated customer rate increases. In several states, these increases were driven by costs of recovery from wildfires, hurricanes, and other natural hazards and by spending to mitigate future exposure. In 2022 and 2023, states relying on natural gas generation were exposed to sharply increased fuel costs resulting from the Russian invasion of Ukraine.
The Berkeley study does have limitations, as the authors candidly admit. Statistical power is limited, and findings reflect broad averages rather than utility specifics. Prices are also shaped by varied cost‑recovery rules and lags across states and utilities. State-level average price data does not reveal these differences. Several variables move together (e.g., renewal portfolio standards growth and rooftop solar), making it hard to cleanly separate their effects. Results further depend on the chosen five-year window, and five‑year snapshots can miss shorter- or longer-term effects. The authors note these issues and provide robustness checks and additional analysis in supplemental information accompanying the publication.
Beyond these limits, the regression analysis itself, while useful, is limited in establishing causal links between variables. More recent “causal inference” methods aim to address this by introducing quasi-experimental approaches to better identify causal effects. Techniques like difference-in-differences, synthetic control, and regression discontinuity designs can isolate causality by leveraging external factors or comparing cases around treatment thresholds. While careful analysis and judgment are critical, new tools have opened promising avenues for identifying causal effects in real-world data.
Future research could build on this study by applying statistical methods that better capture both short- and long-term dynamics in electricity prices. For short-term impacts—such as shocks from natural gas price spikes, wildfires, or storm recovery costs—polynomial distributed lag models or autoregressive distributed lag frameworks can be used. These approaches allow the influence of a variable to unfold gradually over several months or years rather than all at once, offering a clearer view of timing and persistence. To explore long-term effects, dynamic panel regression models like the generalized method of moments estimator could be used to account for persistence in prices and effect between variables like load growth, renewable deployment, and regulatory changes. Combining these approaches with causal inference tools like difference-in-differences would help identify whether policy shifts, such as new renewable standards or rate design reforms, truly cause observed changes in retail prices.
Overall, the paper is careful, clear, and candid about what the study design can and cannot do. The associations found between policies and prices are of great interest. A preliminary step in understanding the many factors driving rates up in some parts of the country is to accurately describe trends and explore potential relationships. However, regression methods are limited in their ability to reveal causal links among data sets. Deeper study using more-sophisticated methods and focusing on particular states and regions is the necessary next step. Finally, the study does not provide direct guidance to policymakers—for example, while it addresses the apparent cost impact of certain policies, it does not attempt to measure their benefits. Complete policy analysis requires both.