1 The Causal Question
Vehicle emissions are a leading source of localised air pollution in urban and peri-urban areas. A large epidemiological literature has established associations between residential proximity to high-traffic roadways and adverse infant health outcomes including low birth weight, preterm birth, and infant mortality [Deschenes et al., 2017]. Yet translating these associations into causal estimates is difficult: families do not locate randomly near roads. Those who live closest to congested highways may be systematically poorer, less educated, or more exposed to other hazards—factors that independently worsen infant health.
Currie and Walker [2011] exploit a natural experiment to solve this identification problem: the staggered rollout of electronic toll collection (E-ZPass) at highway toll plazas in New Jersey and Pennsylvania. By replacing cash toll collection—which requires trucks and cars to stop, idle, and accelerate—with electronic transponders that allow vehicles to pass through at highway speed, E-ZPass dramatically reduced the traffic pollution in the immediate vicinity of toll plazas. This created an exogenous, localized reduction in vehicle emissions at a specific set of geographic points, with a timing that varied across toll plazas and was unrelated to the health of nearby pregnant women.
2 The Identification Strategy
2.1 E-ZPass as a Natural Experiment
The key insight of Currie and Walker [2011] is that E-ZPass adoption was a policy change driven by traffic management and revenue-collection considerations, not by the health or characteristics of residents near toll plazas. The timing of E-ZPass installation varied across individual toll plazas, providing within-plaza before-after variation that can be exploited in a difference-in-differences design.
The treatment is E-ZPass adoption at a specific toll plaza. The comparison group is births to mothers who lived somewhat farther from the plaza (2 km versus 0.2 km, for example). The key assumption is that conditional on plaza-level fixed effects and time trends, the timing of E-ZPass adoption is uncorrelated with trends in health outcomes of mothers living near the plaza.
This is a geographic difference-in-differences design: the treated group consists of births to mothers living within 0.2 km of a toll plaza (in the "treatment corridor"), while the comparison group consists of births to mothers living 0.2-2 km from the same plaza (in a farther ring). Both groups experience the same time trends in all macro shocks; only the treatment group is near enough to benefit from the reduction in idling-related emissions.
2.2 The Biological Mechanism
The primary mechanism linking toll plaza emissions to infant health is short-range particulate matter and carbon monoxide generated by vehicles idling in queues at cash toll booths. Trucks, in particular, produce disproportionate quantities of diesel particulates while idling. Pre-natal exposure to fine particulate matter (PM₂.₅) and carbon monoxide has been linked to intrauterine growth restriction, preterm delivery, and placental abnormalities [Currie and Walker, 2011]. The effect is expected to be highly localised: ambient pollution from idling dissipates rapidly with distance, which is why only mothers living within 0.2 km of a plaza should show meaningful health improvements from E-ZPass.
The localised nature of the treatment also helps validate the design: if E-ZPass improves infant health only within the 0.2 km radius and not at 0.5 km or 1 km, this is consistent with a pollution mechanism rather than with confounding trends that would affect both near and far births. This spatial gradient is a key falsification test.
3 Data and Setting
The paper uses New Jersey and Pennsylvania birth certificate records from the early 1990s through the early 2000s, covering the period before and after E-ZPass adoption at individual toll plazas. The birth records include the mother's residential address (geocoded to within-census-block precision), birth weight, gestational age, and maternal characteristics (age, race, education, smoking).
By geocoding birth records to specific residential locations and matching them to the nearest toll plaza, Currie and Walker [2011] can assign each birth to a treatment corridor (within 0.2 km) or a comparison ring (0.2-2 km). The staggered rollout of E-ZPass across plazas provides temporal variation in treatment timing within each plaza-ring pair.
The DiD estimating equation is:
where Yigt is the health outcome (e.g., low birth weight indicator or log birth weight) for birth i near plaza g in period t; EZPassgt is an indicator for whether plaza g has adopted E-ZPass by period t; Nearig is an indicator for whether mother i lives within 0.2 km of plaza g; γg and δt are plaza and period fixed effects; and Xigt is a vector of maternal characteristics. The coefficient β is the DiD estimate: the change in health outcomes for near births relative to far births, before versus after E-ZPass adoption.
4 Key Findings
4.1 Main Results
Currie and Walker [2011] find that E-ZPass adoption significantly improved infant health outcomes for mothers living within 0.2 km of a treated toll plaza:
- The probability of low birth weight (below 2,500 grams) fell by approximately 10.8% among births within 0.2 km of adopting plazas, relative to the comparison group. Given a baseline low birth weight rate of around 6%, this corresponds to a meaningful absolute risk reduction.
- Preterm birth (below 37 weeks gestation) fell by approximately 11.3% in the treatment corridor following E-ZPass adoption.
- These effects were not detectable for mothers living 0.2-2 km from the plaza, consistent with the rapid spatial decay of idling-related pollution.
The gradient across distance rings is a powerful validation of the pollution mechanism. If E-ZPass adoption correlated with some broader trend in the health of residents near toll plazas (e.g., neighbourhood improvement associated with reduced congestion), we would expect health improvements at all distances—not just within 0.2 km.
4.2 Heterogeneity
Effects were larger for lower-education mothers and for mothers of black children, groups who are disproportionately exposed to environmental risks and may have less ability to compensate for pollution shocks through other margins of health investment. This heterogeneity is consistent with a biological mechanism and with the broader literature on environmental justice.
4.3 Mechanisms
The authors also show direct evidence of the pollution channel: monitor data indicate that carbon monoxide and particulate concentrations near toll plazas fell significantly after E-ZPass adoption. The reduction in pollution was largest very close to the plazas and dissipated rapidly with distance, precisely mirroring the spatial pattern of health effects.
5 Validity Tests
5.1 Pre-Trend Analysis
Event-study estimates in the paper show no significant differences in low birth weight rates between near and far births in the years preceding E-ZPass adoption at each plaza. This parallel pre-trend is reassuring: it suggests that health outcomes in the treatment and comparison groups were on similar trajectories before the natural experiment and that the post-adoption improvement is not driven by pre-existing trends.
5.2 Placebo Tests
The authors check whether E-ZPass at a distant plaza (unrelated to the birth's residential location) also predicts health improvements for births near the focal plaza—a placebo with no biological plausibility. These placebo estimates are small and statistically insignificant, supporting the identification strategy.
5.3 Robustness to Comparison Ring Definition
The results are robust to varying the definition of the near and far rings (e.g., using 0.1 km and 1 km as cutoffs instead of 0.2 km and 2 km) and to restricting the sample to births within specific distance bands.
6 Limitations
Like all natural experiment studies, this paper has limitations worth noting:
- Local effects: The LATE is specific to mothers who lived within 0.2 km of a toll plaza during pregnancy. The effect may not generalise to pollution reduction in other contexts (e.g., highway expansion, urban traffic management).
- Selection into residence: Mothers who lived very close to a toll plaza during the study period may have characteristics that differ systematically from those farther away. While plaza fixed effects absorb persistent differences and the DiD design controls for time trends, the possibility of differential time-varying sorting cannot be entirely ruled out.
- Measurement: Birth certificates capture birth weight and gestational age but not other potentially important health dimensions such as neurological development or longer-run outcomes.
7 Broader Contributions
Beyond its specific findings, the Currie and Walker [2011] paper has been enormously influential methodologically. It exemplifies the use of a policy-induced geographic discontinuity in treatment intensity as a source of causal identification—a design template that has since been applied to highways and infant health [Chay and Greenstone, 2003], E-ZPass and pollution [Deschenes et al., 2017], and other settings.
The paper also contributed to the environmental economics literature by providing some of the first credibly causal estimates of the infant health consequences of localised traffic pollution, complementing earlier work by Chay and Greenstone [2003] who used changes in total suspended particulates (TSP) induced by the Clean Air Act as a quasi-experiment to estimate the effect of pollution on infant mortality.
References
- Chay, K. Y. and Greenstone, M. (2003). The impact of air pollution on infant mortality: Evidence from geographic variation in pollution shocks induced by a recession. Quarterly Journal of Economics, 118(3), 1121-1167.
- Currie, J. and Walker, R. (2011). Traffic congestion and infant health: Evidence from E-ZPass. American Economic Journal: Applied Economics, 3(1), 65-90.
- Deschenes, O., Greenstone, M., and Shapiro, J. S. (2017). Defensive investments and the demand for air quality: Evidence from the NOx budget program. American Economic Review, 107(10), 2958-2989.
- Isen, A., Rossin-Slater, M., and Walker, R. (2017). Every breath you take— every dollar you'll make: The long-term consequences of the Clean Air Act of 1970. Journal of Political Economy, 125(3), 848-902.