Safe Routes to School: A Pilot Evidence Review
Safe Routes to School: A Pilot Evidence Review
What the two-axis framework reveals when you actually try to fill in the table
Suzanne Childress | Unicairn | April 2026
Safe Routes to School is a good place to test an evidence framework because it looks like a success story until you try to be precise about what the evidence actually supports. Congress allocated $612 million for the national SRTS program in 2005. The program has broad bipartisan appeal, intuitive logic, and a substantial body of research behind it. It is also a case where the evidence, examined honestly, tells a more complicated story than either advocates or skeptics tend to acknowledge.
This is a pilot application of a two-axis evidence framework I am developing. The framework assesses transportation interventions on two dimensions: parametric confidence (how precisely have we estimated the effect, given the available studies?) and structural uncertainty (how confident are we that the causal model is correctly specified for the decision at hand?). Standard evidence reviews assess the first. Almost none assess the second. The distinction matters most for interventions where the evidence is generated within a model boundary that excludes the dominant variable, and SRTS turns out to be a clear case.
What SRTS is
SRTS programs combine infrastructure improvements (sidewalks, crosswalks, traffic calming) with education, encouragement, and enforcement activities to make it safer and easier for children to walk and bike to school. The federal program funds both engineering projects and non-infrastructure programs. Most implementations combine several of these elements, which makes isolating the effect of any single component difficult.
Safety: the evidence is real
The strongest evidence for SRTS is on pedestrian and cyclist injury reduction, and it is reasonably good.
DiMaggio and Li (2013) analyzed ten years of geocoded crash data in New York City and found that the annual rate of school-age pedestrian injury during school-travel hours decreased 44% in census tracts with SRTS interventions, compared to no change in tracts without them. They used difference-in-differences methods with adult injury rates as the comparison group. A follow-up Bayesian changepoint analysis confirmed that the timing of the injury decline matched the timing of SRTS implementation.
DiMaggio et al. (2016) extended the analysis to 18 states covering 55% of the nation’s school-age children over a 16-year period. SRTS was associated with a 23% reduction in pedestrian and bicyclist injury risk and a 20% reduction in fatality risk, using multilevel negative binomial models with adult injury rates as the comparison. A Texas-specific analysis found a 14% reduction.
These are the best-designed studies in the SRTS literature. They use quasi-experimental methods, they have comparison groups, and the effect sizes are consistent across different contexts. But the authors themselves note important caveats: pedestrian injuries were already declining nationally during the study period, the study windows were too short for interrupted time series analysis, and concurrent safety interventions (New York City was pursuing aggressive pedestrian safety efforts during the same period) cannot be fully separated from the SRTS effect. The difference-in-differences framework helps, but the authors are careful to describe the findings as associations rather than causal estimates.
Parametric confidence: Moderate. Multiple quasi-experimental studies with consistent findings across contexts. Cannot fully isolate SRTS from concurrent safety interventions or secular trends.
Structural uncertainty: Low to moderate. The causal mechanism (safer crossings, lower speeds, better sidewalks reduce pedestrian crashes) is direct and well-specified. The intervention targets the proximate cause of the outcome. System-level feedbacks are unlikely to undermine the finding.
Mode shift: the evidence falls apart
The mode shift evidence is where the honest assessment diverges from the standard narrative.
The Community Preventive Services Task Force systematic review found a median increase of 5.9 percentage points in active school travel across 26 studies. A subset of 12 studies evaluating U.S. SRTS programs specifically found a median increase of 6.5 percentage points. The most comprehensive evaluation (McDonald et al. 2014, 801 treatment and control schools across four jurisdictions) found a 31% relative increase in the proportion of students walking and biking after five years of SRTS participation.
These numbers sound positive. But the evidence quality is poor. A 2019 systematic review of 23 interventions focused on active commuting to school rated 21 of 23 studies as methodologically “weak.” Only three were randomized controlled trials. Most used pre-post designs without control groups, relied on parent-reported travel mode, and did not control for confounders. The one well-designed cluster RCT (Mendoza et al., Houston walking school bus) showed positive results, but the walking school bus was researcher-led, making it unclear whether results would hold for a school-run program.
The 31% relative change from McDonald et al. also needs context. A 31% relative increase from a baseline of 13% means going to about 17%. That is a real change at the school level. It is not a transformation of how children get to school in America.
And at the national level, the trend line is damning. The proportion of children who walked or biked to school dropped from 48% in 1969 to about 13% by 2009. That decline continued through the period of federal SRTS investment. SRTS did not reverse the national trend. It may have slowed the decline in participating schools. But the aggregate pattern is one of continued loss of active school travel despite hundreds of millions of dollars in federal investment.
Parametric confidence: Low. Large number of studies but predominantly weak designs. Most lack control groups. Parent-reported outcomes. Small absolute effect sizes. The few stronger studies show modest effects from small baselines.
Structural uncertainty: High. This is the critical finding.
The structural problem: school siting
The mode shift evidence is generated within a model boundary that excludes the dominant variable.
Over the past fifty years, American schools have gotten bigger and been sited farther from the families they serve. School consolidation reduced the number of U.S. schools from 200,000 in 1940 to 62,000 by 1990, despite a 70% population increase. New schools are routinely built on large sites along arterial roads, in locations selected for land cost and traffic access rather than walkability. A study of new schools built in six high-growth California counties from 2003 to 2011 found that most were sited in areas that fell short of walkability thresholds. Only 21% of children lived within a mile of their school by 2001, down from 34% in 1969.
If a school sits on a 30-acre campus off a four-lane arterial two miles from the nearest neighborhood, no amount of sidewalk improvement or walking school bus programming will produce meaningful mode shift. The SRTS intervention operates within a land use and school siting constraint that the intervention cannot change. This is not a methodological limitation of the studies. It is the reason the intervention underperforms at the population level.
This is a textbook case of the structural uncertainty problem the two-axis framework is designed to surface. The parametric evidence on SRTS and mode shift is weak but real: at the school level, in contexts where walking is physically feasible, SRTS programs produce small positive shifts. But the structural uncertainty is high because the dominant variable (whether walking is physically possible given school location and surrounding land use) is outside the model boundary. The evidence tells you about the effect of SRTS conditional on walkable school siting. It tells you almost nothing about the effect of SRTS in the American land use context where most children go to school.
The rest of the table: what is not there
Physical activity. The Community Guide review found mixed results on whether active school travel translates to increases in overall daily physical activity. Most studies cannot isolate the SRTS contribution. The causal chain from “child walks 0.5 miles to school” to “child is meaningfully more physically active” is plausible but not well-established. Confidence: Very low.
Equity and distribution. Almost no rigorous evidence. Some indication that low-income schools benefit more from infrastructure improvements, which makes sense because they are more likely to lack basic pedestrian infrastructure to begin with. But the equity evidence is largely anecdotal. The interaction between SRTS and school siting is especially important here: low-income children face greater pedestrian injury risk and are more likely to attend schools in environments with poor walkability. The intervention that would help them most (walkable school siting) is the one SRTS does not address. Confidence: Very low. Structural uncertainty: High.
VMT and environmental outcomes. Not directly studied in the SRTS literature. The implied logic is that children walking instead of being driven reduces VMT, but the effect sizes on mode shift are small enough that the VMT impact would be negligible at any meaningful scale. Confidence: Very low.
Wellbeing and experience. No studies found that assess the effect of SRTS on children’s subjective experience of their commute, sense of independence, or quality of life. This is a gap, not a null finding. Not assessed.
The table
| Outcome | Parametric confidence | Range of effects | Structural uncertainty |
|---|---|---|---|
| Safety (ped/cyclist injury) | Moderate | 14-33% injury reduction (DiD studies) | Low-moderate |
| Mode shift (walking/biking) | Low | 0-6.5 pp increase; weak study designs | High |
| Physical activity | Very low | Mixed; can’t isolate SRTS | Moderate |
| Equity/distribution | Very low | Almost no rigorous evidence | High |
| VMT/environmental | Very low | Not directly studied | High |
| Wellbeing/experience | Not assessed | No studies found | N/A |
What this means
SRTS is reasonably effective at making existing school trips safer for children who are already walking. It has not demonstrated the ability to meaningfully shift mode share in the context of American land use patterns, and the reason is structural rather than programmatic.
The honest version of the SRTS story is not “it doesn’t work.” It is: the intervention addresses a proximate problem (dangerous walking conditions near schools) while leaving the dominant constraint (school siting and land use) untouched. The safety evidence is real. The mode shift evidence is weak. And the gap between them is not a failure of program design. It is a reflection of the fact that you cannot solve a land use problem with a sidewalk.
This is the kind of finding that a standard evidence review would miss, because a standard review would rate the mode shift evidence as “low” and move on. The two-axis framework adds the structural uncertainty dimension, which says: even if the mode shift evidence were stronger, the intervention would still underperform at the population level because the causal model excludes the variable that matters most. That is a different and more useful finding for practitioners and policymakers.
A note on method
This is a pilot review, not a systematic review. I have not conducted a comprehensive literature search, applied formal inclusion/exclusion criteria, or assessed every relevant study. The purpose is to test whether the two-axis framework produces useful findings when applied to a real intervention, and whether the table structure surfaces information that standard reviews miss. On both counts, I think it does. A full review would require a team, a protocol, and substantially more time. What this pilot demonstrates is what the full review would look like and why it is worth doing.
The evidence assessment framework, including operational definitions for the parametric confidence and structural uncertainty levels, is described in a companion document: Drawing Lines of Evidence: From What We Can See to What We Can Prove.
Suzanne Childress | childressssuzanne@gmail.com | unicairn.com
This post is part of an ongoing project to build systematic evidence infrastructure for transportation planning. If you are interested in collaborating, I would like to hear from you.