The Effectiveness and Limits of Time-of-Use Pricing in Public EV Charging Networks

Abstract: Time-of-use pricing is promoted to manage demand at public EV charging stations, yet its effectiveness depends on short run flexibility and local constraints. Using station by day by hour data from Shenzhen and Amsterdam, we estimate intraday price responsiveness on two margins, whether charging occurs in a station hour and, conditional on charging, delivered energy and occupancy time. High dimensional fixed effects absorb station by day demand shocks and hour of week patterns, so identification relies on within station, within day price variation under scheduled tariffs. Responses differ across cities. Shenzhen adjusts mainly through conditional intensity, whereas Amsterdam adjusts mainly through participation. Weather shifts responsiveness in opposite directions, with heat weakening responses in Shenzhen and rainfall strengthening participation responses in Amsterdam. Power upgrades typically outperform network densification except in transit-oriented areas.

• The paper quantifies the intraday EV charging response to TOU pricing using high-frequency station data from two major cities.
• The study employs a two-part model to separate extensive participation effects from intensive usage responses under diverse weather and spatial conditions.
• The findings highlight that power upgrades consistently boost throughput, while ultra-local densification yields variable effects based on local context.

The Effectiveness and Limits of Time-of-Use Pricing in Public EV Charging Networks

Introduction

This paper provides a systematic empirical analysis of intraday price responsiveness in public EV charging networks under Time-of-Use (TOU) pricing in two distinct cities: Shenzhen and Amsterdam. Utilizing high-frequency station-by-day-by-hour datasets, the study quantifies the short-run effects of posted price regimes on charging incidence and usage intensity, and decomposes these responses across the extensive (participation) and intensive (conditional usage) margins. The identification exploits within-station, within-day price variation, leveraging rich fixed effects to address unobserved heterogeneity at the station and temporal level. Beyond average elasticities, the authors pursue heterogeneity across weather states and urban/siting contexts, and elucidate operational implications for infrastructure upgrades versus localized network densification.

Empirical Framework and Methodology

The study adopts a two-part model to distinguish the behavioral response to TOU pricing along two axes:

1. Extensive (Participation) Margin: Whether a charging event occurs within a station-hour.
2. Intensive (Conditional Usage) Margin: The amount of energy delivered and occupancy time, conditional on charging taking place.

High-dimensional fixed effects (station × day, hour-of-week) absorb local daily demand shocks and periodic activity, isolating intraday responsiveness to exogenous, pre-scheduled tariff changes. This design yields semi-elasticities interpretable as short-run operational responses to posted TOU prices rather than long-run demand shifts.

To capture context dependence, the model includes interactions between price and weather (temperature, precipitation), and classifies stations by proximate land use via dominant POIs within 500 m. For spatial and infrastructural heterogeneity, charger-level throughput is regressed on equipment power, station size, nearest-neighbor spacing, ultra-local competition, and POI context.

Core Results

Price Elasticities and Behavioral Margins

Shenzhen: Price responsiveness is dominated by the intensive margin. Intraday price increases primarily manifest as reduced conditional duration of charging sessions; the probability of initiating a charging session is not significantly price-elastic after absorbing fixed effects. The response magnitude is relatively modest; e.g., a 10% price increase yields a precisely measured but small reduction in occupancy conditional on charging but no significant effect on participation.

Amsterdam: The extensive margin is dominant: participation—the decision to initiate charging in a given station-hour—is highly price-elastic. Conditional usage intensity is less sensitive to price once a session is underway, given control for station and temporal effects. Notably, a 10% price increase yields a nontrivial probability reduction (~0.9 pp) in extensive participation, emphasizing that in this context, user flexibility is more expressed as shifting whether and when to start charging rather than how long or how much to charge.

Implication: Aggregate elasticities may mask fundamentally different operational mechanisms across systems. For network operators and planning, it is critical to disambiguate whether TOU tariffs are retiming session start times (Amsterdam) or only altering intra-session energy intake (Shenzhen), as these affect peak management strategies differently.

Weather-Dependent Heterogeneity

Shenzhen: Higher temperatures attenuate the negative price response, especially on the participation and conditional duration margins. Rainfall has a more ambiguous or weaker effect. This attenuation is consistent with increased necessity or reduced flexibility for EV users under hot conditions, weakening the effectiveness of TOU signals during those states.

Amsterdam: In contrast, inclement weather, particularly rainfall, significantly strengthens price responsiveness on the participation margin and, to a lesser degree, on conditional intensity. Warmer hours slightly steepen the price response as well, but the main amplification is found during rainy intervals.

Conclusion: The operational leverage of TOU pricing varies both in sign and magnitude with weather, and the direction of weather-moderation is city/environment dependent.

Spatial Heterogeneity and Infrastructure Effects

Analysis of charger-level throughput reveals:

• Power Upgrades: Higher charger-rated power is uniformly associated with higher energy throughput per charger. Estimated power elasticities exceed unity in both cities (Shenzhen: 1.31, Amsterdam: 1.69).
• Ultra-Local Competition: Densification (adding nearby stations within 0.5 km) has heterogeneous effects. In Shenzhen, ultra-local densification increases utilization in transit-dominated areas (positive competition elasticity), but sharply decreases it in education-dominated areas (negative elasticity). In Amsterdam, densification is consistently associated with reduced per-charger utilization, with a stronger negative effect in parking-dominated neighborhoods.
• Spatial Context: Differences in land use background (commercial, education, transit, parking, other) modulate both baseline throughput and the effects of competition and upgrades.

Policy Counterfactuals

Static scenario simulations (applying estimated cross-sectional elasticities) provide a ranked assessment of operational levers:

• Uniform Power Upgrades (+10%): Yield robust system-wide increases in delivered energy (Shenzhen: +13.1%, Amsterdam: +16.8%).
• Targeted Upgrades (+50% for bottom-quartile chargers): Substantial incremental increases even with focus on low-power devices (Shenzhen: +2.5%, Amsterdam: +3.6%).
• Densification: Context-dependent. Positive in Shenzhen-transit scenarios, neutral/negative in most settings and especially in Amsterdam (parking: −7.4% throughput).
• Combined Interventions: Upgrades combined with targeted densification amplify positive effects only in select spatial contexts; elsewhere, the gains from densification are negated by demand dilution.

Strong Empirical Result: Upgrading existing charger power is uniformly more effective than dense siting of additional stations at sub-kilometer distances, except in spatial and land use micro-environments where empirical competition elasticities are positive.

Robustness and Limitations

Extensive robustness checks (alternative fixed effects, functional forms, sample definitions, weather codings, and falsification tests via permutations and pre-trend checks) confirm the internal consistency of the main findings. Variance decompositions demonstrate sufficiency of within-cell price variability for identification.

Limitations center on the reduced-form nature of the outcome models: the data do not disclose user type, charging intent, real-time congestion, or wait times, leaving open the precise mechanisms underlying observed participation and intensity effects. Queueing behaviors, equity considerations, and long-run dynamics (e.g., EV adoption, home charging substitution) are not directly estimated.

Practical and Theoretical Implications

• Operationally, TOU pricing can induce measurable short-run load reallocation in public EV networks but is subject to fundamental limits stemming from user flexibility and local constraints.
• The behavioral margin of adjustment is context-dependent: for participation-dominated systems, TOU can efficiently manage entry to peak hours, but not necessarily the total load; for intensity-dominated systems, session retiming is limited.
• Weather states must be incorporated into tariff design and operational planning; environmental stress can either mute or amplify price responsiveness depending on the local context.
• Infrastructure policy should prioritize upgrades to charger power as a first-line intervention, reserving ultra-local densification for corridor-like environments with observable complementary demand.
• Overly aggressive local densification, especially in substitution-heavy locations, risks underutilization and does not increase system throughput.
• Integrated design of pricing, infrastructure, and real-time information provision will be required for resilient and context-adapted public EV charging networks.

Directions for Future Research

Major open questions include causal identification of specific behavioral mechanisms, linking charging outcomes to user-specific data (trip purpose, access to alternatives), integrating queueing models and real-time congestion, and explicit welfare and equity analyses. Additionally, dynamic modeling of tariff-induced investment and demand feedbacks will be essential as EV penetration and public charging systems continue to evolve.

Conclusion

This study rigorously characterizes the operational effectiveness and boundaries of TOU pricing for public EV charging, revealing its channel-specific limits, weather interaction, and dependence on spatial/infrastructural context. The findings substantiate the need for nuanced, multifactorial pricing and investment strategies tailored to the behavioral and physical structure of urban EV charging networks.