Power supply disruptions deter electric vehicle adoption in cities in China

Low-carbon electrification is central to achieving deep decarbonization and limiting global warming to 1.5 °C. Despite growing renewable shares, end-use sectors such as transport and buildings remain heavily reliant on fossil fuels. Electrifying vehicles and buildings alongside a cleaner power grid can reduce fossil fuel use, and many jurisdictions offer incentives for electrification. A core premise of electrification is a stable electricity supply. However, grids face two pressures: challenges from integrating high shares of intermittent renewables and rising electricity demand from electrification, which can elevate outage risks during peaks. China and the U.S. have both experienced large power outages linked to extreme weather and other factors. In China, unprecedented outages in Dec 2020–Jan 2021 and Sep–Oct 2021 were driven by extreme cold, surging demand, coal price spikes, supply disruptions from floods, and rationing. While prior work in developing contexts indicates outages reduce benefits from rural electrification, there is little empirical evidence on how outages affect EV adoption. This study provides nationwide evidence from China (Nov 2019–Sep 2021) on the causal impact of power outages on EV adoption, quantifies associated carbon benefit losses, and informs policy on grid resilience to support electrification. The study’s contributions are: building a nationwide high-frequency point-level outage database; demonstrating empirically that power system failures deter EV adoption (with a placebo showing substitution toward non-EVs); and estimating reduced carbon benefits from delayed EV uptake due to outages.

Existing literature highlights multiple determinants of EV adoption including range anxiety, charging availability, environmental awareness, and income. A recent statistical study reports no causal link between charging station counts and EV purchases. Surveys (e.g., in California) suggest outage exposure lowers stated intent to buy EVs, but such studies rely on intentions rather than revealed purchases. Research in rural electrification shows outages diminish expected benefits from electricity access. Analogies from combustion-engine markets show fuel price increases or petroleum supply disruptions reduce purchases of ICE vehicles, indicating that supply reliability shapes vehicle purchase decisions. The present study fills a gap by providing large-scale empirical evidence, using observed sales data across Chinese cities, on the effect of electricity supply disruptions on EV adoption and downstream carbon benefits.

Study design and data: The authors assemble a monthly panel at the city level spanning Nov 2019–Sep 2021. EV adoption is proxied by the number of vehicles obtaining new automobile insurance each month (from the China Association of Automobile Manufacturers), covering BEVs and PHEVs and, where relevant, other categories for heterogeneity tests. High-frequency, geo-referenced point-level outage data are scraped daily from official city websites, including start/end times, addresses, affected areas, and causes. Outage measures are aggregated to city-by-month averages across districts: (1) average outage counts per district per month; (2) average outage hours per district per month (summing all simultaneously affected locations within a district/day can exceed 24 h). Control variables include per capita GDP (China City Statistical Yearbook) and, in robustness, EV charging stations (China Electric Vehicle Charging Infrastructure Promotion Alliance). The final sample includes 310 cities (140 northern, 170 southern). Identification: Two sources of variation support causal inference: (i) large cross-sectional variation in outage hours across cities (e.g., in 2020 average 162 h/year/district with SD ~188 h); and (ii) sharp temporal spikes during two nationwide outage periods (Dec 2020–Jan 2021; Sep 2021) yielding strong longitudinal variation within cities. Empirical model: Main specification is a semi-log fixed-effects panel regression linking ln(monthly EV sales) to lagged outage measures (1- and 2-month lags), controlling for ln(GDP per capita). Rich fixed effects include city-by-year and city-by-month, capturing unobserved time-varying city characteristics (e.g., policies, population, environmental awareness, infrastructure, seasonal factors, climate, temperature effects on batteries, business cycles). Standard errors are clustered at the city level. The authors also estimate separate models for BEV and PHEV, and conduct heterogeneity analyses by vehicle type (non-commercial vs for-hire), region (north vs south divided by the Huai River), and GDP level (high vs low using 60,000 RMB per capita threshold). Robustness and endogeneity: Placebo regressions test effects on non-EV sales (gasoline/diesel/CNG/ethanol). Additional controls include counts of charging stations and a dummy for mandatory rationing orders. COVID-19 shocks are addressed using lockdown dummies and their lags as proxies for supply chain disruptions; models excluding heavily COVID-impacted cities are also estimated. Potential simultaneity is probed via an instrumental variables approach using monthly counts of extreme temperature days (max > 89.6 °F or min < 32 °F) as an instrument for outages; first-stage relevance and weak instrument diagnostics (F > 10) are reported. Granger-type panel tests indicate EV purchases do not cause outages, while lagged outages do predict EV sales. Additional checks include alternative functional forms (linear, semi-log, double-log, exponential) with link tests favoring the semi-log model; unit-root tests (most variables stationary; non-stationary cumulative charging stations addressed via first differences); panel cointegration tests (variables cointegrated). Per-capita definitions of outage measures yield consistent results. Carbon benefits calculation: Using the average estimated elasticity (0.99% decrease in EV adoption per additional monthly outage per district), average city monthly EV sales (405), number of cities (310), and an estimated carbon benefit per EV of $210/year (based on an average social cost of carbon and per-vehicle CO2 reductions), the study estimates national climate benefit losses from increased outages. A conceptual S-curve framework illustrates decade-long delays in benefits if consumers switch to fossil vehicles during outage spikes.

• Each additional power outage per district in a city in a given month reduces that city’s new monthly EV sales by about 0.99% on average (mean of 1-month and 2-month lag coefficients: 1.1% and 0.88%).
• By technology: BEV sales decline by about 0.92% and PHEV by about 1.4% per additional outage. Given 2021 sales volumes (BEV 2.734 million; PHEV 0.6 million), the implied reduction is larger in absolute units for BEVs (≈25,152) than PHEVs (≈8,400).
• Outage duration effects: A 1-hour increase in monthly average outage hours per district reduces EV sales by 0.024% (1-month lag) and 0.021% (2-month lag).
• Placebo on non-EVs: No negative impact is found; instead, non-EV sales increase with outages (e.g., L1 outage times coefficient ≈ +0.038; L1 outage hours ≈ +0.0007), supporting a substitution effect and reinforcing the causal interpretation for EV impacts.
• Heterogeneity: Non-commercial private vehicles are negatively affected; for-hire vehicles (taxis, rental fleets, government/corporate fleets) show no significant impact, likely due to policy-driven procurement and organizational resilience. Northern and lower-GDP regions experience more negative impacts than southern and higher-GDP regions.
• Magnitude of national climate impact: If outages double for one year nationwide (e.g., from 10 to 20 outages per district per month on average), the depressed EV adoption implies a loss of approximately $31.3 million per year in carbon reduction benefits; assuming a 10-year delay in benefits yields a discounted total of about $254 million (5% discount rate). Considering combined health and climate co-benefits under high-renewables scenarios, an upper-bound annual loss is estimated at about $911 million.
• Contextual outage severity: During Dec 2020–Jan 2021 and Sep–Oct 2021, several provinces experienced extreme outage frequencies and durations (e.g., Zhejiang 642, Jiangsu 619, Anhui 498 outage counts/district in two months; Hunan 4,719, Hubei 3,909 hours/district).

The findings directly address the research question by showing that power supply disruptions causally deter EV adoption in Chinese cities, even after controlling for extensive covariates and applying IV strategies to address endogeneity. This demonstrates that the success of transportation electrification hinges not only on vehicle technology and incentives but critically on the reliability of the power system. The implications are far-reaching: increased outage risks arising from extreme weather, aging infrastructure, rising electrification loads, and integration of variable renewables can slow EV diffusion and postpone environmental and carbon benefits at scales material to climate goals. Policy responses include investing in grid resilience (upgrades, hardening), staged re-optimization of generation capacity and operations to meet evolving load profiles, advanced forecasting and balancing, demand-side management (e.g., time-of-use pricing, critical peak designs), vehicle-to-X integration, and distributed storage to buffer outages. Behavioral interventions that shift or reduce peak demand can further stabilize grids. Strengthening power infrastructure and managing demand can mitigate the deterrent effect of outages, preserving the pace of electrification crucial for emissions reduction targets.

This paper compiles a novel, high-frequency, point-level outage database across Chinese cities and links it with monthly EV insurance data to quantify the causal impact of power outages on EV adoption. The study finds that each additional outage per district per month reduces EV sales by roughly 1%, with meaningful heterogeneity by vehicle type, region, and income levels, and a substitution toward non-EVs. The estimated climate benefit losses from outage-induced adoption delays are economically significant and can persist for up to a decade. Policy should account for the costs of unreliability by reinforcing grid resilience, optimizing capacity and operations, deploying demand-side strategies and storage, and ensuring that electrification policies are complemented by power system reliability. Future research could examine longer time horizons and post-2021 data, micro-level charging availability and reliability, consumer expectations and risk perceptions, interactions with renewable integration and storage, and spillovers to other electrification domains (e.g., buildings and industry).

Key caveats include: extrapolating short-term adoption impacts to decade-long carbon benefits requires assumptions about policy stability, consumer preferences, EV benefits per vehicle, and the social cost of carbon; outage-hour aggregation (summing across locations within districts) can exceed 24 h/day and is not normalized by the number of electricity accounts, although fixed effects mitigate bias; limited city-level supply chain disruption data led to using lagged lockdowns as proxies; V2G capabilities were nascent during the study period and not expected to offset effects; one control (cumulative charging stations) is non-stationary but addressed via differencing; the study period (Nov 2019–Sep 2021) encompasses unusual shocks (pandemic, coal price spikes, extreme weather) and may not capture all longer-run dynamics; functional form choices favor the semi-log model based on tests, but alternative nonlinearities could exist; despite IV and Granger tests, residual unobserved factors cannot be fully ruled out.