Should congested cities reduce their speed limits? Evidence from São Paulo, Brazil

https://doi.org/10.1016/j.jpubeco.2020.104155Get rights and content

Highlights

  • The São Paulo speed limit program reduced accidents by 21.7% and resulted in 1889 averted accidents within 18 months.

  • Analysis of policy effects using Google transit API

  • Social benefits from speed limit reductions are at least 1.32 times larger than the social costs.

  • The benefits of accident reductions accrue largely to lower income pedestrians and motorcyclists.

Abstract

Road accidents are the leading cause of unnatural deaths worldwide. Cities are experimenting with more stringent speed limits in an effort to reduce them. The impacts of these policies are unclear in many developing country cities, where a disproportionate share of accident damages occur but also where speed regulations could exacerbate already high levels of congestion. We evaluate a speed limit reduction program in São Paulo, Brazil using a dynamic event study design and measurements of 125 thousand traffic accidents, 38 million traffic tickets issued by monitoring cameras, and 1.4 million repeat observations of real-time trip durations before and after a regulatory change. We find that the program resulted in 1889 averted accidents within the first 18 months and reduced accidents by 21.7% on treated roads, with larger effects on roads with camera-based enforcement. The program also affected travel times on treated roads (5.5%), though the social benefits from reduced accidents are at least 1.32 times larger than the social costs of longer trip times. The benefits of accident reductions accrue largely to lower income pedestrians and motorcyclists, indicating that speed limit reductions may have important impacts on low income residents in developing country cities.

Introduction

Road injuries are the leading cause of unnatural deaths worldwide, representing 27.1% of all human unnatural deaths and approximately one third of the external costs related to private transportation (Parry and Small, 2005, Parry et al., 2007, WHO, 2015). Road safety is particularly problematic in the developing world. While low- and middle-income countries account for only half of world's vehicles, 90% of road fatalities occur in those countries. In recent years, cities around the world have responded by reducing speed limits and improving speed limit enforcement on urban roads.1 There is evidence that well-enforced speed limits tend to reduce vehicle speeds and damages from road accidents (Archer et al., 2008, Musicant et al., 2016, Wilmot and Khanal, 1999). However, reducing speed flows may increase travel costs as commuters face longer journeys. The empirical literature is less clear on the magnitude and significance of these effects (Archer et al., 2008). As a result, the net benefit of increasing the stringency of speed regulations is still largely an open question, particularly in developing country cities where they could have the greatest impact on both accident reductions and increased drive times. Not surprisingly, many highly congested cities have been reluctant to adopt speed limit reductions.

This paper brings together a number of novel data sources to provide estimates of the effects of a major speed limit reduction program in one of the most congested cities of the developing world.2 In 2015, the government of São Paulo, Brazil reduced speed limits on the main highways in the city from 90 km/h to 70 km/h. However, the policy became highly contentious and in early 2017, the speed limits on urban highways were reverted to pre-reduction levels. Using a semi-dynamic event study design (Abraham and Sun, 2018, Borusyak and Jaravel, 2016), we exploit exogenous variation in the timing of speed limit reductions adopted on each road in São Paulo to identify the effect of changing the speed limit on traffic accidents and travel time. Speed limit reductions could alter the travel and routing decisions of residents, potentially reducing travel on treated roads or displacing accidents from treated roads to alternate routes. We examine changes in the volume of individual traffic violations that are unrelated to speeding to test for substitution or route switching in response to the policy and analyze changes in individual speeding tickets to learn how drivers adjust to the regulatory change.

Our results indicate that the 2015 speed limit program reduced road accidents by 21.7% on treated road segments, resulting in approximately 1889 averted road accidents and 104 averted fatalities within the first 18 months of adoption. These impacts are considerably larger than policies aimed at addressing violent crime, which is the other primary cause of unnatural deaths in the city.3 Roads where speed limit reductions occur alongside contemporaneous onset of camera-based enforcement experienced an additional 11.5–11.8 percentage point reduction in accidents. We compare the social cost of increased trip duration from reduced speeds with the benefits of reduced accidents and find that the benefits associated with reduced accident damages are at least 1.32 times larger than the costs of longer drive times.4

This paper builds upon prior studies that have evaluated the positive and negative economic impacts of speed limit changes. Two notable examples are van Benthem, 2015, Ashenfelter and Greenstone, 2004, which study speed limit increases on American highways in the 1980s and 1990s. Our paper extends this literature in a number of ways. First, while van Benthem (2015) finds that the social costs of increasing speed limits on interstate highways in the United States are 2–7 times larger than the benefits, the current paper provides estimates of the impacts of speed limit changes on the key transportation corridors of one of the most congested cities in the world. Our findings suggest that speed limit reductions are benefit-enhancing in São Paulo and provide evidence on enforcement and distributional concerns that may be common in many developing country cities. As is true in the case of most road safety policies, urban highways and arterial roads were not randomly selected for speed limit reductions in the São Paulo program. We estimate cohort-specific average treatment effects (CATT) to support the internal validity of our event study results (Abraham and Sun, 2018).6

This study also builds on recent work that uses real-time web routing services to study the impact of urban transportation policies and congestion in developing country cities (Akbar and Duranton, 2017, Akbar et al., 2018, Hanna et al., 2017).7 We combine a representative origin-destination travel survey with repeated observations of real-time trip durations before and after the speed limit change to attribute policy impacts to individuals who vary in the types of trips that they are taking, in their value of time, and in their accident risk. In our setting, we find that failing to account for (otherwise unobservable) heterogeneity in the population of travelers who use treated roads would attenuate our cost estimates by more than 35%.

This paper makes an additional contribution by examining the extent to which speed limit reductions have progressive/regressive impacts, which is an important and unresolved issue in the transportation literature (van Benthem, 2015). By pairing the travel survey with real-time congestion data, we find that the costs associated with increased travel time (slower trips) accrue disproportionately to wealthier individuals that have higher rates of private vehicle ownership and tend to commute on treated roads. This is expected. However, we also find evidence of a striking difference in the distribution of the benefits from accident reductions across income groups: 86% of the benefits from reduced accident damages accrue to low-income residents, who bear a disproportionate share of the accident risk as pedestrians and motorcyclists. Our analysis reveals that speed limit reductions likely have strongly progressive impacts in São Paulo and could have important effects on reducing unnatural deaths in other urbanizing regions where low-income residents rely upon motorcycles and other high-risk modes of transport.

The remainder of the paper is structured as follows: Section 2 describes the speed limit changes investigated in our study, Section 3 details the data sets used for our analysis, and Sections 4 and 5 presents our main empirical strategies and reduced-form estimates. In Section 6, we compare the costs and benefits of the policies, including our distributional analysis. Section 7 concludes.

Section snippets

Speed limit reductions of 2015 and 2017 reversal

With more than 20 million residents, São Paulo is one of the largest and most heavily congested metropolitan areas in the world. Over 30 million motorized trips are made on an average weekday in São Paulo (METRO, 2013). While the length of the average trip is 8 km, the average duration is 51 min. Travel-related injuries are also a major problem in the city. In an average year during the 2005–2014 period, there were 12.5 deaths per 100,000 residents in the Metropolitan Area of São Paulo, which

Data

This section describes the data that we use to evaluate the impacts of speed limit changes on accidents, commuting time, traffic volume and traffic violations.14

Did São Paulo's program reduce accidents ?

We estimate the effects of speed limit changes on road accidents using a semi-dynamic event study design that makes use of exogenous variation in the timing of speed limit reductions. Our empirical setting includes road segments that were treated at different points in time during 2015 and a data set of monthly accidents between 2012 and ending before the reversal of the policy in January of 2017.26

Did São Paulo's program affect travel times ?

This section examines the effect of the January 2017 speed limit increase on the travel times of drivers in São Paulo. Using representative trips simulated on the Google Directions API at the time of the reversal, we compare the estimated duration of trips that utilize the Marginais Highways to those that do not, before and after the policy change. A trip is defined as a pair of origin and destination coordinates queried in real-time at a specific time of day.42

Cost–benefit analysis

In this section, we extend our reduced-form estimates to analyze the social costs and benefits of the speed limit changes that were implemented in São Paulo.46 We estimate the monetary value of road accidents and travel time using standard parameters from the literature (Viscusi and Masterman, 2017). We acknowledge the limitations of this procedure given substantial parameter uncertainty in this

Conclusion

This paper evaluates the effect of policies that altered traffic speed limits in one of the most highly congested and dangerous cities for drivers in the world: São Paulo, Brazil. We demonstrate that a series of speed limit reductions in 2015 resulted in a substantial (21.7%) reduction in road accidents on treated road segments, resulting in 1889 averted road accidents and 104 averted fatalities within the first 18 months of adoption. Our findings provide evidence that camera-based enforcement

References (43)

  • Ang et al.

    this paper

    (2020)
  • J. Archer et al.

    The impact of lowered speed limits in urban and metropolitan areas

  • O. Ashenfelter et al.

    Using mandated speed limits to measure the value of a statistical life

    J. Polit. Econ.

    (2004)
  • K. Borusyak et al.

    Revisiting Event Study Designs

  • M.V.d.F. Cabral

    Avaliação do impacto do INFOCRIM sobre as taxas de homicídios dos municípios paulistas: uma aplicação do método de diferenças em diferenças espacial

    (2016)
  • CET

    Benefícios Imediatos da Redução das Velocidades Máximas Permitidas. O Caso das Marginais Tietê e Pinheiros

  • CET

    Acidentes de Trânsito – Relatório Anual - 2016

  • P. Christensen et al.

    Behavioral responses to automated enforcement: evidence from Sao Paulo speed cameras

  • DATASUS, 2018. Informações de saúde (tabnet). Online. URL...
  • M.J. dos Santos et al.

    A cointegration analysis of crime, economic activity, and police performance in São Paulo City

    J. Appl. Stat.

    (2013)
  • I.K. Folgero et al.

    Going Fast or Going Green? Evidence from Environmental Speed Limits in Norway

    (2017)
  • Cited by (14)

    • Highways and pedestrian deaths in US neighborhoods

      2023, Regional Science and Urban Economics
    • Investigating the impacts of urban speed limit reduction through microscopic traffic simulation

      2021, Communications in Transportation Research
      Citation Excerpt :

      The results, in line with other studies (Archer et al., 2008; Pei et al., 2012; Aarts and Van Schagen, 2006), have shown that there is a strong correlation between the value of speed reduction and the risk of accidents as well as the severity of accidents. This correlation is supported by a Brazilian study (Ang et al., 2020) conducted in 2015. The speed limits were reduced in São Paulo, from 90 to 70 ​km/h on inner city motorway segments, and from 60 to 50 ​km/h on urban main roads.

    • Pedestrian deaths and large vehicles

      2021, Economics of Transportation
      Citation Excerpt :

      The relationship between vehicle characteristics and pedestrian fatalities is one element of overall road safety. Significant economic research has been undertaken to investigate other causes of traffic fatalities such as vehicle speed (Ang et al., 2020; Van Benthem, 2015), road congestion (Green et al., 2016) alcohol consumption (Baughman et al., 2001; Green et al., 2014; Hansen, 2015; Jackson and Owens, 2011; Levitt and Porter, 2001; Ruhm, 1996), public policy and regulation (Basili and Belloc, 2020; Borsati et al., 2019; Bourgeon and Picard, 2007; Carpenter and Stehr, 2008; KaracaMandic and Ridgeway, 2010; Peltzman, 1975), electronic distractions (Blattenberger et al., 2013; Oviedo-Trespalacios, 2018) and the driver's state of mind (Giulietti et al., 2020). The current study is focused specifically on the effect of vehicle characteristics on road fatalities, with a particular focus on pedestrian effects.

    • Uber and traffic safety: Evidence from Brazilian cities

      2021, Journal of Urban Economics
      Citation Excerpt :

      As argued in the previous sections, also note that one possible reason for the results to be larger for Brazilian cities is that, in the context of poor urban infrastructure and low traffic safety as it is the case of Brazil, perennial and straightforward interventions can generate a significant effect on traffic accidents. Ang et al. (2020), for instance, found that reducing speed limits in the Brazilian city of So Paulo lead to 21.7% reduction of traffic accidents on treated roads and that this effect grows to 35.5% percent after 1.5 years of the change in speed limits. This evidence is in line with the results of the study by Lichtman-Sadot (2019) about the effect of the introduction of night buses in the Israeli cities.

    • Why are highway speed limits really justified? An equilibrium speed choice analysis

      2020, Transportation Research Part B: Methodological
      Citation Excerpt :

      In order to curb the external costs associated with highway speeding and to achieve/enhance efficiency, regulatory measures such as (maximum) speed limits are regularly proposed (Elvik, 2010). There is a large body of scientific research showing that measures taken to curb speeding behavior lead to fewer and less severe accidents as well as a better environmental performance in terms of lower local and global air pollutants (Abdel-Aty et al., 2006; Ang et al., 2020; Castillo-Manzano et al., 2014; Elvik, 2013a; Liu et al., 2015; Madireddy et al., 2011; Vadeby and Forsman, 2018; Yang et al., 2012; among others).2 As long as the number/severity of road crashes and/or the extent of air pollutants are the main arguments in the underlying objective function measures like maximum speed limits can be regarded as effective instruments (Elvik, 2019).

    View all citing articles on Scopus

    We are grateful for helpful comments from Andre Chagas, Sandy Dall’erba, Edward Glaeser, Eduardo Haddad, Marieke Kleemans, Gabriel Kreindler, Mary Arends-Kuenning, Adam Osman, Julian Reif, Nick Tsivanidis, and Matthew Turner. We thank participants at the Federal University of São Carlos Economics Seminar, the University of São Paulo Economics Seminar, the University of Illinois pERE Seminar, the Urban Economics Association Annual Meeting (2018), and the Cities and Development Workshop (SAIS) for helpful comments. We thank computer scientists in the UIUC Big Data in Environmental Economics and Policy Research Team for excellent programming work. In particular, we acknowledge Sloane Sullivan, Elena Lee, Nate McCord, Surya Tadigadapa, and Abhishek Banerjee. This work was made possible by generous support from the Lemann Institute for Brazilian Studies and the National Center for Supercomputing Applications. All data and replication files will be made available in the following data repository: https://github.com/uiuc-bdeep/speed_change (DOI: 10.5281/zenodo.2000257). All errors are our own.

    View full text