The following section summarizes the evidence available from causal studies in urban transport. The chapter is organized by looking at different types of urban transport interventions, which have been implemented in different countries according to their context. In several cases, the outcomes of interest and methodologies of evaluation are similar across interventions, for example, the exploration of employment and land value effects and the use of difference-in-differences (DID) models are quite common in BRT and metro or light rail studies. In other cases, the nature of the intervention drives the outcome of interest and the methodological approach, such as the focus on environmental outcomes (e.g., pollution) arising from traffic restriction policies and the use of regression discontinuity designs (RDD) given the sharp policy changes.
Evidence on bus rapid transit systems
While designs can vary widely, BRT systems generally are bus-based systems that operate in dedicated lanes, with rapid service through the implementation of several operational features such as off-board payment, at-level bus boarding, intersection signal priority, passing lanes, and frequent service.Footnote 3 The introduction of these systems has also generally been accompanied by government reforms of the bus sector that employ various public–private contracting schemes, or a mix of centralized planning and private investment and service operation. As these systems are considered a cost-effective and flexible approach to providing high-capacity and more environmentally sustainable transportation, they have grown rapidly around the world.
There is not a straightforward theoretical explanation of the impact that BRT systems can have on land values and land-use changes. Increases in property values may depend upon the quality of public transport systems and its subsequent ability to reduce travel times (Cervero and Kang 2011; Bocarejo et al. 2013). As BRT system designs and their associated travel time savings can vary widely, the effects on property values can also vary. As posited by Medda (2012), travel time reductions may also be valued differently by customers depending on the locations to which these new systems provide access. Moreover, the extent to which accessibility benefits translate into land values will depend on the sensitivity of users to improvements in access (Rodríguez and Mojica 2009). BRT systems and other mass transit investments that operate at surface or above ground can also have negative effects on property values and land use near the system due to nuisance effects such as noise, air pollution, and crowds of passengers. Finally, some studies hypothesize that the potential of BRT systems to increase property values may be lower compared to heavy and light rail systems due to their flexibility and perceived lower level of permanence and rigidity as an infrastructure service (Rodriguez and Targa 2004; Vuchic 2002).
Studies that examine the effects of BRT systems on property values and real estate development yield mixed results,Footnote 4 but only a handful of papers use empirical strategies seeking to determine attribution. Perdomo (2011) uses propensity score matching to evaluate the impact on property values of TransMilenio, Bogota’s BRT system, finding a positive impact on areas in the vicinity of the system. The absence of time-varying information, however, does not allow for controlling for unobservable characteristics that might affect the results. Rodríguez and Mojica (2009) evaluate the impact of the extension of TransMilenio on property asking prices. They exploit a DID estimation and find increases in prices in areas that were already served by the TransMilenio but that benefited from an extension, and detect no impact in areas that gain new access to the system. Their empirical approach lacks an analysis of how similar treatment and control groups were at baseline and a discussion about whether the parallel trends assumptions, required for DID to be valid, holds in this case.
The increased value of property is also theorized to stimulate land-use change by increasing the attractiveness of development or redevelopment of parcels near the stations (Rodríguez and Mojica 2009). Previously, vacant parcels may become more attractive to real estate investors, and those that are built up may become targets for more intensive development or infill. Timing of effects might be important: while land-price effects can be instantaneous, land-use changes tend to occur more slowly, partly due to institutional lags (e.g., securing building permits and zoning amendments) (Perez et al. 2003). The literature looking at the urban development effects or land-used changes from BRT investments is still scarce and also shows mixed results (Stokenberga 2014). Again, most of this literature is based on before-and-after comparisons, or cross-sectional analysis exploiting distance to the system, without explicit consideration of the counterfactual scenarios.Footnote 5 The only study identified here that offers a DID estimation is that of Bocarejo et al. (2013), which shows that areas served by Bogota’s TransMilenio have higher population growth than areas without access to the system, particularly feeder areas, but that there are no significant changes in land use.Footnote 6
By reducing transport costs and improving accessibility, BRT investments may also have effects on facilitating access to markets and services. Along these lines, multiple studies, mostly conducted in developed countries, have analyzed the impact of urban transit investments on employment outcomes, but the majority are non-causal.Footnote 7 Among recent studies seeking to address causality, Scholl et al. (2018) rely on a combination of individual-level DID and area-level propensity score-based overlap analyses to evaluate the Metropolitano BRT of Lima, Peru. They find that several years after the introduction of the system, there are positive effects on employment outcomes (employment, formal employment, hours worked, and monthly labor income) for individuals living close to the BRT stations, but not for those who live close to the feeder lines. Tsivanidis (2018) looks at the aggregate and distributional effects of TransMilenio. Based on work featuring gravity equations for commute flows (Ahlfeldt et al. 2015), he proposes a new reduced-form methodology derived from general equilibrium theory based on “commuter market access,” arguing that distance-based approaches might be misleading in capturing the intensity of treatment. To address the non-random route placement, he uses instrumental variables estimation exploiting historic data on the tram system and engineering estimates of the cost to build BRT systems on different types of land. The author finds that while the system caused increases in welfare and output larger than its cost, gains accrued slightly more to high-skilled workers. The analysis of mechanisms suggests a potential increase in residential segregation by skills.
Another recent strand of the BRT literature has analyzed the effects of such systems on pollution. Bel and Holst (2018) study the effect of Mexico City’s Metrobus on air pollution emissions. Using DID and quantile regression techniques, they estimate the atmospheric concentration of pollutants in Mexico City between 2003 and 2007 to assess the impact of the introduction of the Metrobus. They conclude that the BRT system constitutes an effective environmental policy, reducing emissions of CO, NOX, PM2.5, and PM10.
Limited access to safe transportation is one of the greatest challenges to labor force participation faced by women in developing countries, reducing their participation probability by 15.5% points (ILO 2017). Two recent studies looking at employment outcomes of bus services show that the effects are especially strong for women. Martinez et al. (forthcoming) look at both BRT and metro systems in Lima and explore the differential effects on employment by gender. Using DID regressions looking at comparable areas (selected through overlap in propensity score at the area level) that are closer and farther away from the systems, they show increases in the probability of being employed among women living closer to the systems and no significant changes for men. Changes are driven by women not previously in the labor market and, although there is an increase in earnings per hour, no improvements are observed in job quality. Abu-Qarn and Lichtman-Sadot (2019) provide evidence of a trade-off between investment in education and time allocated to work by women after the introduction of bus services in Arab towns in Israel. They support their identification strategy based on the argument that bus line introductions and schedule changes were random due to the long bureaucratic approval processes.
Evidence on light rail and subway systems
Urban light rail and subway systems are expensive enough that these projects generally require large subsidies. To justify these subsidies, proponents often assert the ability of these systems to have a transformative effect on the city and to encourage employment growth (Gonzalez-Navarro and Turner 2018). However, the causal evidence is limited regarding such transformative effects. This section focuses on the small set of papers that attempt to solve the causality problem caused by the non-random assignment of these systems and their stations.
Baum-Snow and Kahn (2000) study the impact of new rail transit on usage and housing values. They exploit variation in transit access changes among census tracks within five major cities in the USA that upgraded their rail transit systems in the 1980s, using distance as a proxy for transit access. The authors find that rail transit improvements lead to increased mass transit use for commuting, but to a small capitalization of transit infrastructure into housing prices and rents. In related research, Gibbons and Machin (2004), for the case of the London Underground and Docklands Light Railway in South East London, and Billings (2011), for the case of the new light rail line in Charlotte, North Carolina, show increases in prices in areas closer to the systems using DID approaches. More recently, Dorna and Ruffo (2017), using a DID approach combined with matching analysis, find that the electrification of a light rail suburban line in Buenos Aires had positive effects on housing prices around the areas of influence of the stations. Using a synthetic control methodology, they also find that the increased reliability of the service had large effects on ridership.
While these studies provide evidence regarding the effects of mass transit on property values, they do not provide information on the relationship between light rail and subway systems and the growth of cities. If these systems affect urban growth, those effects will appear both near and far from the stations and might take more time to appear. Such citywide effects are, by construction, not captured by a DID methodology, as noted by Gonzalez-Navarro and Turner (2018), who studied the relationship between the extent of a city’s subway network, its population, and its spatial configuration in the 632 largest cities in the world. For this, they construct panel data describing the subway systems in these cities, their population, and measures of centralization calculated from night lights data. Their evidence suggests that when big cities build subways, the subways have at most a small effect on urban population growth. However, they find that subways allow the central cores of large cities to spread out and reorganize activity in the cities, suggesting that when transportation costs fall, economic activity can spread out.
Regarding employment, Holzer et al. (2003) exploit the exogenous change in accessibility to employment brought by the expansion of the Bay Area Rapid Transit System (BART) (including heavy rail and subway). Using a DID approach, the authors estimate the impact of the expansion of BART on the propensity of suburban firms to hire minority populations, finding sizable increases in the hiring rates of Latino workers but no increases in the hiring rates of African-Americans. In another study in the United States, using historical data on manufacturing establishments from 1850 to 1870, in DID and instrumental variables models, Atack et al. (2008) find that the introduction of the railroad increased the number of factories and thus employment. Results also show that firms located in counties that gained rail access were more likely to employ women relative to men. More recently, the impact of an exogenous shock from Hurricane Sandy, which shut down a portion of New York’s metro system (the R train) in 2013, was exploited to estimate the effect of the system on access to employment (Tyndall 2017). Findings show that living next to the R train during the shutdown resulted in an overall increase in the probability of being unemployed, and that effects were lower for individuals who had access to a vehicle and much higher for those who were transit dependent. On the gender front, Asahi (2016) exploits fixed effects models to show that increased proximity to the subway network in Santiago, Chile, is associated with higher employment rate and hours of work especially for women.
Another important strand of the literature looks at impacts on air pollution. Chen and Whalley (2012) use a sharp RDD to examine rail transit ridership on the opening day of a new rail transit system in Taipei, China. The assumption behind this design is that in the absence of opening the Taipei Metro, air quality would have changed smoothly for that day (i.e., air pollution levels on the days just before the opening of the Taipei Metro form a valid counterfactual for air pollution levels in Taipei on days just after the opening of the Taipei Metro), conditional on differences in weather, a host of time-specific fixed effects, and a very flexible smooth time trend. The authors found that the opening of the Taipei Metro reduced CO2 air pollution by 5–15%, but they found little evidence that ground-level ozone pollution was affected by the opening of the Metro. Goel and Gupta (2015) use a similar strategy to measure the effects of the Delhi Metro in India on air pollution. The authors exploit the sharp discontinuities in metro ridership resulting from each extension of the rail network and examine whether they coincide with corresponding discontinuities in pollutant measures. They found evidence of large reductions in NO2 and CO2 levels.
Evidence on cable cars
Cable cars are mainly touristic attractions in rich western countries, but in LAC cities they have been implemented as transport systems to connect isolated low-income neighborhoods with the city center. Cable cars offer multiple advantages over subways or light rails systems. They can be built in a shorter amount of time, do not require the displacement of large groups of people, and seem more suited for cities with mountainous geographies (The Economist 2017). However, these systems tend to be heavily subsidized and do not have the same capacity as other massive transport systems. The first cable car designed as a transport system in LAC opened in Medellín, Colombia, in 2004. Since then, Caracas (Venezuela), Cali (Colombia), Mexico City (Mexico), Rio de Janeiro (Brazil), and La Paz (Bolivia) have built similar systems.
Notoriously, all causal evaluations available for cable cars pertain to cities in the LAC region, namely Medellín and La Paz.Footnote 8 Cerdá et al. (2012) examine the effects of the Metrocable in Medellín on violence, based on homicide reports at the neighborhood level and household surveys. The empirical strategy compares neighborhoods that are serviced by the Metrocable versus comparable neighborhoods not serviced by this system (obtained through propensity score-matching techniques) before and after completion of the transit project. Their findings show that the decline in homicide rates was greater in treated neighborhoods and that resident reports of violence also decrease in the proximity of the system. Using more detailed geo-coded information, Canavire-Bacarreza et al. (2016) also found reduced homicide rates in neighborhoods served by the Metrocable.
Bocarejo et al. (2014) also studied the effects of the Metrocable in Medellín, looking at changes in accessibility to jobs, travel time savings and costs, and housing values. The authors use data from origin-and-destination (OD) surveys before and after the project’s implementation. Their results show that the access provided by Metrocable to the main employment centers doubled the number of job opportunities reachable by people in the area of influence of the project. However, they did not observe large changes in reported travel time savings and costs. Moreover, the authors did not find a statistically significant relationship between the Metrocable and housing costs.
It is often assumed that cable cars, as well as other urban transport systems, lead to travel time savings, but few impact evaluation studies quantify these savings. Suárez-Alemán and Serebrisky (2017) conducted a quantitative estimation of travel time savings arising from Mi Teleférico, the cable car system in La Paz. The authors used individual-level, OD surveys and compared travel times between trips, with the same origin–destination pair that were made on Mi Teleférico versus those that were made on alternative transport systems. Their findings suggest that, on average, Mi Teleférico reduced travel times by 22%.
More recently, Martinez et al. (2018) estimated the impact of Mi Teleférico on changes in household-level transport expenditures, individual time allocation decisions, and employment outcomes using cross-sectional data. Given that stations were located in an ad hoc manner, similar to an exogenous shock for nearby households, and that households cannot easily manipulate their location in the short term (particularly property owners), the identification strategy exploits distance to the closest station of the Mi Teleférico system as an instrumental variable to predict the use of the system. The results point toward a transport modal shift, as treated households report larger expenditures on public transportation and lower expenditures on private transportation. In terms of time allocation, there is a significant reduction in transport time and an increase in the time devoted to educational and recreational activities. Finally, there is evidence of increases in self-employment activities and associated increases in labor income.
Evidence on driving restrictions
Currently, capital cities in LAC exceed the recommended annual WHO limits for PM10 and PM25 emissions (WHO 2017) while also ranking poorly on the Waze (2016) driver satisfaction index and the TomTom (2017) traffic index. To deal with both pollution and congestion levels, and to ultimately try to influence the mode of transport choice of the population, several cities around the world have tried policies to curb emissions and/or reduce traffic congestion during peak hours. These policies include large investments in public transport, designated lanes for high-capacity vehicles, and congestion pricing schemes. However, these initiatives are either costly or politically charged for city governments to easily implement. Several cities, primarily in LAC, have opted for a less expensive alternative: vehicle driving restrictions. LAC has actually been the pioneer on this issue (Fig. 1). Following the restrictions imposed in Santiago, Chile in 1986, several LAC cities followed, including Mexico City, São Paulo, La Paz, San José, Quito, and several cities in Colombia. More recently, Delhi, India, and Beijing, China, have implemented similar approaches.
Normally, these restrictions ban the use of private light vehicles during specific times of a weekday in certain areas of a city, based on a given rule like the last digits of the vehicles’ license plates. The restrictions are in place during rush hours in both the morning and the evening. Compliance with the program comes from enforcement either by police officers on the street or traffic cameras, and hefty fines are imposed for violations.Footnote 9
Several studies have used empirically robust methodologies to assess the effectiveness of these types of restrictions, and their findings have been mostly disappointing. Permanent restrictions do not have a lasting effect on reducing pollution or traffic and might even induce households to buy a second highly polluting car to circumvent the restriction completely.Footnote 10 Studies of the “Hoy No Circula” program in Mexico City, such as Eskeland and Feyzioglu (1997) concluded that the driving restrictions increased gasoline use, most likely from a second car in the household. Davis (2008) found an increase of 20% in the car fleet with a decrease in bus ridership and an increase in car sales, while also finding that it had no discernible effect on air quality. Gallego et al. (2013) found that households adjusted their vehicle stock in a little less than a year and thus the benefits of the restrictions disappeared by the second year. Davis (2017) suggests that expanding the restriction to Saturdays did not reduce air pollution, as households relied on other private car trips. Finally, Blackman et al. (2018b) used a contingent valuation methodology for the willingness to pay (WTP) to avoid the traffic restriction. They found an average annual WTP equivalent to US$130 per vehicle, which represents up to 2% of a driver’s annual income.
Methodologies for assessing the impact of these restrictions typically exploit the timing of the intervention. Studies of the impact on pollution of Mexico City’s restrictions use air quality monitoring stations throughout the city and an RDD centered around the date of implementation, while also controlling for environmental covariates and flexible polynomial adjustment terms (Davis 2008, 2017; Gallego et al. 2013). This methodology has been replicated in other studies (see Blackman et al. 2018a, for a review of developing countries studies) and the results are comparable to the one for Mexico City’s program. Studies of the programs in Bogotá, Santiago, São Paulo, and Quito have found short-term gains, but mixed or even negligible long-term results because drivers adopt strategies to circumvent the restriction (Troncoso et al. 2012; Bonilla 2016; Carrillo et al. 2016; Zhang et al. 2017). In sum, the existing literature suggests that driving restrictions can work for short-term pollution emergencies, but that using driving restrictions as long-term fixes for pollution and congestion must take into account available public or non-pollutant substitutes and behavioral responses by drivers.
Evidence on transport network companies
In recent years, transport network companies (TNC, also known as ride-sourcing companies) have received remarkable attention from consumers, the media, and policymakers. These types of companies have emerged as app-based, on-demand ride services and they have generated a debate over their role in urban transport. TNC have become more common over the last decade, with small local or regional services giving way to national and global companies. Examples include Car2go, Zipcar, ReachNow, Via, Cabify, Lyft, and Uber.
The transport sector of many cities (those in LAC included) is now experiencing a high level of disruption with the introduction and evolution of technology and transport services. As these new layers of technology-based transportation options spread, it is important to understand how they affect the transportation systems and society. The literature on this topic appears to be very limited, in part due to their novelty and lack of open data on these services. In addition, there are difficulties in constructing valid counterfactual scenarios, given that in many cases transport network companies are introduced in whole cities or countries at the same time.
Ride sourcing has been mainly compared with taxis. This is primarily because both services involve passengers paying a fee for the travel. However, there are many differences between them, including the use of technology, labor market differences, and government regulations. In the different countries where ride-sourcing companies have tried to enter, there has been resistance by current providers (mostly taxis) and controversy because the new companies disrupt the industry, competing and taking away many customers from taxis. Rayle et al. (2014) compare ride-sourcing and traditional taxis in San Francisco using an intercept survey. Their findings suggest that ride sourcing meets a latent demand for urban travel, appealing to generally younger, well-educated users looking for short wait times and fast point-to-point service, while avoiding the inconveniences of driving like parking, or restrictions on drinking and driving.
One of the main concerns with the rise of ride-sourcing companies has been the effect that they may have on traffic congestion, especially considering that the areas in which they operate are large cities with heavy traffic. Using a DID approach, Li et al. (2017) found that the entry of Uber into the USA market significantly decreased traffic congestion time, congestion costs, and excessive fuel consumption. The authors argue that ride-sharing services have the potential to reduce car ownership, shift the traffic mode from single occupancy to ride sharing, and delay travel plans during peak hours, thus reducing overall traffic congestion in an urban area. This is to the best of our knowledge the only study that attempts to show a causal effect of the entry of a TNC.
Incentives to increase demand of urban transport systems
As urban transport systems consolidate their operations, new evaluation questions emerge related to the operational aspects of the system. From the operators’ perspective (in several cases a private sector actor), some of these questions could relate to what the most appropriate tariffs are to maximize the demand of the system or what type of promotion strategies could be most effective to incentivize its use, among others. From a policy perspective, incentivizing demand could be key to promoting a modal shift to transport systems that are more environmentally friendly. It could also be relevant to understand how transport systems can maximize social inclusion effects through well-designed and targeted subsidies. This section presents evidence around some of these questions.
Willingness-to-pay studies
As important as it is to know the effects of different urban interventions, it is also important to understand how much people are willing to pay for those interventions. Several studies, the majority in developed countries, have looked at the WTP of travelers or consumers for different transport services or attributes. In several cases, this information is elicited through experimental designs (stated preferences), while in others it is based on observations of actual behavior or choices (revealed preferences approach).
Some studies have looked at the WTP to reduce travel time in the context of toll highways, which has been used to guide the design of congestion pricing or time of day pricing programs. The main idea behind these estimations is to obtain the value of time or the amount of money that a respondent would be WTP in tolls for 1-h time savings to keep the respondent’s transport choice unchanged (Brownstone et al. 2003). Calfee and Winston (1998) applied stated preference models to a sample of drivers who regularly drove to work in major metropolitan areas of the USA. They found that the WTP was surprisingly low (between $3.5 and $5 per hour) and insensitive to travel conditions and to how toll revenues were used. Using revealed preference data, Brownstone et al. (2003) estimated that users have a median WTP of $30 to reduce travel time by 1 h on San Diego’s Route I-15, which highlights the fact that stated preference studies generally yield lower values than revealed preference studies (Wardman 2001). For Route 91 in Southern California, Lam and Small (2001) estimated the value of time to be between $19 and $24 per hour depending on model specification. They looked at day pricing, which might explain the variation in results in comparison to Brownstone et al. (2003) who looked at congestion pricing.
WTP studies have also been applied to value improvements in the quality of transport services. Molin and Timmermans (2006) and Khattak et al. (2003) evaluated the value for consumers of different information aspects that can be included in web-enabled or electronic public transport information systems in the Netherlands and the USA, respectively. Their results indicate that travelers are willing to pay for better quality and more interactive information systems. Eboli and Mazzulla (2008) estimated the WTP for improving the quality levels of a bus service among Spanish students. Their results showed that the maximum valuation corresponded to service frequency and the minimum value pertained to information at bus stops. More specifically, users would pay an increase of 44% in weekly and monthly cards for more service frequency. In a similar vein, Worku (2013) examined the willingness to use and pay for improved public transport services in the United Arab Emirates. The results suggest that residents are willing to use and pay higher fees for public buses, provided that the quality of service is improved.
Another way of using WTP results is to characterize the demand of existing or new transport systems. Results from multiple studies highlight the heterogeneity in WTP across individuals, which is key to improving the design and targeting of certain transport interventions. For example, for certain settings, females seem to value more the expressways when compared to men (Senbil and Kitamura 2004). Higher-income persons are willing to pay more to reduce travel time because of their greater opportunity cost (Markose et al. 2007). Individuals who are more aware of environmental issues are more likely to use public transport (Carson 2000; Lee and Cheah 2014), and those with more children and who are older save travel time by using more expensive or shorter routes (Asensio and Matas 2008).
Experiments with subsidies to increase demand and encourage employment search
Subsidies for public urban transport have been adopted, both in developed and developing countries, to make transport more affordable. As these subsidies usually encourage the use or more frequent use of transport systems, they might also facilitate access to services and economic opportunities. Phillips (2014) studied whether transportation costs constrained job searches in urban low-wage labor markets. He provided transit subsidies to randomly selected clients of a non-profit employment agency in Washington, DC. The subsidies generated a large, short-run increase in job search intensity for the transit subsidy group relative to a control group receiving standard job search services but no transit subsidy. In the first 2 weeks, individuals assigned to the transit subsidy group applied and were interviewed for 19% more jobs than those not receiving subsidies. These results provide experimental evidence in support of the theory that search costs over time can depress job search intensity, contributing to persistent urban poverty in neighborhoods far from job opportunities. Similar findings have been obtained by Franklin (2017) looking at the case of young job seekers who live far from the center of Addis Ababa, Ethiopia. The author concludes that search costs impose significant constraints to find employment, as experimentally treated individuals increase job search intensity and are more likely to find good and permanent jobs.