Congestion charge

Since the 1970s, congestion charges have been adopted in Asian and European cities [e.g., refs. 1,2,3,4]. Congestion charges serve not only as traffic management tools but also as a form of climate policy. By imposing direct fees on road use5, these policies aim to manage local congestion while simultaneously raising revenues for public transportation6, and also yielding indirect benefits, such as improved air quality and reduced greenhouse gas emissions.

Despite decades of use abroad7, the initiative for a NYC congestion charge, as originally proposed by then-Mayor Michael Bloomberg back in 20078, was implemented on January 5, 2025. Tolls were introduced for drivers entering lower Manhattan below 60th Street. Since the implementation, congestion has eased, vehicle volumes have decreased, and public transportation use has increased9.

Previous research indicates that public support for congestion charges tends to be strengthened after implementation5,10. In this study, we attend to whether and how such a policy implementation is related to shifts in attitudes, behaviors, and social norms.

Current research

To better understand these dynamics, we recruited New Yorkers across three waves, ranging from three weeks before implementation to six months after implementation. Participants (n = 1522) were recruited via Prolific Academic. First, we screened participants living in the State of New York via the platform. Second, we excluded participants who lived outside the New York City metropolitan area. The final sample included 1417 at wave 1 (three weeks before implementation), 1031 at wave 2 (during the first week after implementation), and 603 at wave 3 (six months after implementation).

Attrition and weighting

To analyze attrition, we used first-wave data comparing participants who completed all waves to non-completers (i.e., participants only participating in one or two waves). Attrition was not significantly related to attitude strength, education, income, or ideology (all pHolm > 0.05). The attrition analysis showed that non-completers held stronger support for the congestion charge (60.3 vs 51.6%, pHolm = 0.01), stronger car-restriction norms within the tolls (d = 0.37, pHolm < 0.001), and expressed stronger behavioral intentions (d = 0.40, pHolm < 0.001) than completers (see supplementary material). To mitigate attrition bias, we used mixed-effect models with stabilized inverse-probability weights (IPW), trimmed at the 99th percentile, for observations in each wave. The weights diagnostics indicated minimal variance inflation (see supplementary materials).

Support

Did support for the NYC congestion charge change over six months? Three weeks before the implementation, 57.1% of the participants (n = 1417) supported it. A logistic mixed-effect model with IPW weights did not reject the null hypothesis between waves 1 and 2 (β = −0.23, SE = 0.15, p = 0.12) or waves 1 and 3 (β = 0.16, SE = 0.18, p = 0.40). In other words, we found no evidence of change in support for the NYC congestion charge across the first six months after implementation.

Attitude strength

In addition to measuring participants’ support (yes vs. no), we assessed strength of attitude by asking to what extent they supported or opposed the congestion charge. A longitudinal mixed-effects model with IPW weights showed that attitudes were strengthened for both supporters and opponents between the first and second waves (β = 0.24, SE = 0.06, p < 0.001). This implies that supporters expressed stronger supportive attitudes, while opponents expressed stronger oppositional attitudes. Between the second and third waves, supporters’ attitudes were further strengthened (β = 0.28, SE = 0.08, pHolm < 0.001), while opponents showed a non-significant descriptive decline (β = −0.14, SE = 0.08, pHolm = 0.11, see Fig. 1). Pairwise comparisons showed no statistically significant difference between supporters and opponents in either wave 1 or 3 (pHolm > 0.05). During the implementation week, however, opponents held stronger attitudes than supporters (d = 0.22, pHolm = 0.02). These results are in line with valence asymmetry, meaning that people tend to weigh negatively valenced evaluations more heavily than positive ones [e.g., ref. 11]. Applied to the NYC congestion charge, opponents expressed stronger “negative” attitudes than supporters expressed “positive” attitudes during the implementation week. Taken together, although support remained stable throughout the six months, supporters’ attitudes became gradually stronger. Opponents, on the other hand, expressed stronger attitudes than supporters during the implementation week.

Fig. 1: Attitude strength for opponents and supporters across three waves (before, during, and six months after implementation) of the NYC congestion charge.
figure 1

For comparability, values for opponents have been negated to display oppositional attitude strength on the same visual scale as supporters’ attitudes. Thus, larger absolute values indicate stronger attitudes in both groups. Opponents expressed the strongest opposition during the implementation week, while supporters’ attitudes gradually strengthened over time. Points represent model-based means (lmer with IPW weighting) and shaded ribbons represent 95% confidence intervals.

Full size image

Intention and behaviors

Did intention to support or oppose the congestion charge translate into behaviors? A longitudinal mixed-effects model with IPW weights showed an intention–behavioral gap in supporting or opposing the NYC congestion charge. Self-reported behaviors were weaker than stated intentions for both supporters (Wave 1 vs. 2: β = −1.46, SE = 0.05, pHolm < 0.001; Wave 1 vs. 3: β = −0.24, SE = 0.07, pHolm = 0.0005) and opponents (Wave 1 vs. 2: β = −1.39, SE = 0.06, pHolm < 0.001; Wave 1 vs. 3: β = −0.21, SE = 0.07, pHolm = 0.003). Across all waves, supporters reported engaging in more behaviors than opponents (drange = 0.60–0.67, all pHolm < 0.001), and there was no evidence that the intention-behavior gap differed by group (wave × group interaction: F(2, 1750.5) = 0.39, p = 0.68). Taken together, participants’ self-reported behaviors to support or oppose the NYC congestion charge were less extreme than their intentions. Furthermore, opponents engaged in fewer opposing behaviors than supporters engaged in supportive ones.

Social car-restriction norms

Did the congestion charge undermine or foster social car-restriction norms? A longitudinal mixed-effects model with IPW weights first showed that social car-restriction norms were stronger within than outside the tolls (β = 0.29, SE = 0.04, p < 0.001), and that supporters held stronger social car-restriction norms than opponents (β = 0.82, SE = 0.06, p < 0.001). Of particular interest, simple slopes showed that car-restriction norms increased within the tolls for both supporters (β = 0.07, SE = 0.03, pHolm = 0.007), and opponents (β = 0.11, SE = 0.03, pHolm < 0.001). Outside the tolls, however, social car-restriction norms were weakened among supporters (β = -0.08, SE = 0.03, pHolm = 0.003), and unchanged among opponents (β = 0.02, SE = 0.03, pHolm = 0.45, see Fig. 2). Taken together, we observed strengthened social car-restriction norms within the Manhattan toll area, while negative or no statistically significant change outside the tolls.

Fig. 2: Social car-restriction norms were strengthened within tolls (below 60th street in Manhattan), while negative or no statistically significant change in social car-restriction norms was observed outside the tolls.
figure 2

The points and lines represent model-based means, and ribbons represent 95% confidence intervals.

Full size image

In brief

To understand psychological responses to a real-world implementation of environmental policy, we conducted a six-month survey of public responses to the NYC congestion charge. Building on research suggesting both misalignments between anticipated and experienced consequences of a congestion charge4, and that congestion charges may spill over to environmental behaviors12. We tracked how New Yorkers voting intentions, attitudes, behaviors, and car-restriction norms evolved. Support for the NYC congestion charge remained stable over time, while attitudes became gradually stronger for supporters. The latter aligns with past research on congestion charges [e.g., refs. 4,5]. Opponents’ attitudes peaked at the implementation week. Supporters and opponents thus seem to differ in attitude strength13; supporters expressed more durable and gradually strengthened attitudes, while opponents expressed more immediately intense attitudes. Opponents’ attitude intensity may stem from a negativity bias, forming attitudes from perceived unfairness or tax aversion14,15. Drawing on how public opinion evolved during implementation of the Stockholm congestion charge4, New Yorkers’ experiences with the congestion charge may have been more positive than anticipated.

Public responses

Both supporters and opponents expressed stronger behavioral intentions than self-reported behaviors. Implying an intention–behavior gap in both policy support and opposition. Given that opponents held stronger attitudes than supporters during the implementation week, it would be reasonable to assume that opponents would also engage in more behaviors. However, our data showed the opposite pattern: Opponents engaged in fewer opposing behaviors than supporters engaged in supportive ones. Behavioral engagement, such as posting content or using hashtags on social media, attending public events or rallies, and supporting crowdfunding campaigns to support or oppose the congestion charge, were more common among supporters than opponents. We encourage future research to develop instruments to assess changes in public opposition or support for planned or newly implemented environmental policies16.

Car-restriction norms grew stronger within, but not outside, the tolls. This advances past research by suggesting not only that social norms may increase policy support17, but also that policy implementation may strengthen social norms, potentially creating a positive feedback loop18,19. The latter is especially important given the central role of social norms in motivating climate mitigation behaviors20.

Limitations

Five main limitations warrant mention. First, the sample was more educated than the population of New York City. Second, attrition analysis showed that participants who completed all three waves differed from those who only took part in one or two waves. To assess robustness, we ran all analyses using the sample of completers without IPW. Results confirmed the general patterns of attitudes, behavioral engagement, and social norms. Support, however, was not stable; it dropped during the implementation week (-9.6 percentage points) compared to before being implemented (see supplementary materials). Third, supportive and oppositional engagement was measured by 21 behaviors, while New Yorkers might have engaged in additional behaviors. Fourth, we only provide data from one congestion charge, generalizability remains uncertain. Fifth, any causal claims are unwarranted, and long-term effects beyond this period remain unknown. Overall, it should also be mentioned that effects of attitude and norm changes are modest.

Practical recommendations

For policymakers, we want to underscore that 1) patterns of attitude strength differ between supporters and opponents; 2) intended opposition and support were stronger than consequent behaviors, and 3) the NYC congestion charge seems to serve an expressive function, fostering social car-restriction norms21. These results suggest that well-designed policies not only regulate behavior but may also reshape attitudes, public norms, and political engagement in the face of local environments and climate change.

Design

This study employed a three-wave longitudinal survey design, starting on December 18, 2024, and ending on June 17, 2025. 1522 participants, located in the state of New York, were recruited via Prolific Academic on the first wave. Data was cleaned to exclude participants not living in the New York City metropolitan area (including New York City, Long Island, Westchester, Hudson Valley, Newburgh, Yonkers, Rockland County, and Middletown). Eligible participants were reinvited to the second and third waves to answer a survey that was accessible for one week and took approximately 10 minutes to complete (Mwave1 = 11.48, Mwave2 = 8.07, Mwave3 = 4.47 min). Participants were compensated with a gradually increasing hourly payment of £6.10 on the first wave, £14.78 on the second wave, and £33.87 on the third wave. The main items of the survey assessed participants’ voting intention and strength of support (i.e., attitude) toward the congestion charge, intentions and self-reported supportive or oppositional behavioral engagement (i.e., demonstrating and posting on social media), and social car-restriction norms, defined as others (dis)approval (i.e., perceived social acceptance, expectations of social sanctions, and perceived social pressure). This data collection is part of a larger project with six waves (see supplementary material on OSF).

Sample

Demographic statistics from the first wave showed that 45.0% voted for the Democratic party, 35.6% for the Republican party, 12.6% did not vote, 2.5% for another candidate, and 4.4% preferred not to say. Implying that, among those who voted, 55.9% voted for the Democratic party and 44.1% for the Republican party. The sample is thus fairly representative of general voting patterns in the NYC metropolitan area (i.e., Bronx, Brooklyn, Dutchess County, Manhattan, Nassau County, Orange County, Putnam County, Queens, Rockland County, Staten Island, Suffolk County, Ulster County, Westchester County). Furthermore, the full sample was highly educated (70.6% having a bachelor’s degree or higher), and the most frequent annual household income was 50,000 to 79,999. In terms of gender, the full sample was somewhat balanced (50.0% women and 47.4% men, 0.6% consent revoked), and the mean age was 36.2 years (SD = 13.79).

Survey

The data collection was part of a larger project assessing multiple aspects of the congestion charge. Relevant items include information, attitudes, behavior, social norms, and demographics (full survey is available on OSF). First, participants were informed that the survey was part of a six-wave project, provided with ethical declarations, and asked to give their consent. Second, participants were provided with materials describing the implementation, purpose, extent, and cost of the NYC congestion charge. In assessing voting intentions, participants were asked “If given the opportunity to vote in a referendum about the implementation of the congestion charge in Manhattan (below 60th St), as described on the previous page. How would you vote?1) vote for the congestion charge, 2) vote against the congestion charge”. In measuring attitude strength, supporters and opponents received a matched item: To what extent do you support/oppose the congestion charge? Ranging from 1 (not at all supportive/ opposed) to 7 (Extremely supportive / opposed). By combining both voting intention and attitude, we aim to provide a complementary measure of support, rather than measure one of them alone [see ref. 14]. Intentions and behaviors were measured by 21 behaviors matched to express support or opposition (see supplementary material on OSF). At the first wave, we measured participants’ intentions by asking “Would you do any of the following to support/oppose the congestion charge?” Ranging from 1 (definitely would not) to 7 (Definitely would). During the following waves we assessed self-reported behavior by asking “Have you done any of the following to support/oppose the congestion charge?” Ranging from 1 (Never) to 7 (Always, whenever possible). Before measuring social car-restriction norms, participants were provided with a text aiming to clarify the following questions: “We’re interested in your opinion on whether driving in Manhattan (below 60th St.) is socially acceptable or unacceptable. Please note, we’re not asking about the legality of driving, but how you perceive it socially—what you consider to be socially acceptable or unacceptable behavior. For example, downloading copyrighted material is illegal but socially acceptable to some, while talking loudly on the phone in public is legal but often seen as socially unacceptable to some”. In assessing injunctive social car-restriction norms, defined as others (dis)approval, we draw on past research22,23 and used three items aimed to measure perceived social acceptance, expectations of social sanctions, and perceived social pressure: 1) “How socially acceptable do you think it is to drive a car in Manhattan (below 60th St) presently?” measured on a scale from 1 (“Totally socially unacceptable”) to 7 (“Totally socially acceptable”, 2) “If someone drives a car in Manhattan (below 60th St) presently, how likely is it that others in the area would disapprove (e.g., eye-rolling, verbal remarks, or gestures)?” measured on a scale from 1 (“Not at all likely”) to 7 (“Absolutely likely”), 3) “Is there “social pressure” to avoid driving a car in Manhattan (below 60th St) presently?”, measured on a scale from 1 (“Absolutely no social pressure”) to 7 (“Absolutely strong social pressure”). To be able to control for a general change in social norms within the whole NYC, participants also answered the three items above about “other parts of New York City” (e.g., “Is there “social pressure” to avoid driving a car in other parts of New York City presently?”). In the first wave, we measured demographic variables including political affiliation, education, where in New York participants live, where in New York participants live and work, income, and car ownership. In addition, via the Prolific academic platform, we accessed participants’ age, sex, race, employment status, nationality, country of birth, and country of residence.