The simulated 20 ft zone is illustrative; the required setback and treatment depend on law, approach speed, street geometry, and agency design.
States began regulating vehicle behavior before a national code existed.
A model framework helped standardize stopping, standing, and parking rules.
Traffic-control devices became increasingly standardized.
Many jurisdictions prohibited parking near crosswalks, signals, hydrants, and intersections.
Guides connect curb clearance explicitly to pedestrian visibility and crossing design.
756 recorded daylighting installations later formed the actual-treatment study.
Required a study and at least 100 new daylighted intersections per year beginning January 1, 2025.
The administration committed to visibility improvements at 1,000 intersections per year within a broader 2,000-intersection program.
314 locations were complete by August; nearly half were hardened. The stated annual goal was 1,000.
NYC DOT announced a barrier-based design for selected high-crash intersections along conventional bike lanes.
Streetsblog reported the Council staff analysis in November 2025. The cited material includes no peer reviewed paper or complete public technical report.
The report strengthens the bus stop exposure and proxy misclassification hypotheses; a complete causal conclusion still requires a documented technical analysis.
Program claims provide context because many interventions occurred simultaneously.
The proxy analysis has a large sample and lacks an installation date, with substantial confounding risk.
The comparison before and after treatment improves the counterfactual while selection, combined treatments, and exposure remain unmeasured.
Directly compare unchanged, sign-only, and hardened treatments with observed exposure and behavior.
Links larger corner radii to higher right-turn speeds and pedestrian-crash frequency.
Parking-setback and curb-extension guidance for uncontrolled crossings.
Street-design guidance on visibility and parking removal near intersections.
Exposure, stop placement, obstructed views, and bus-passenger behavior.
Evidence that roadside constraint and parking can influence operating speed.
Legal definitions, annual minimum, barriers, and reporting requirements.
Secondary reporting on data cleaning and bus-stop versus hydrant associations.
The treatment may be neutral or beneficial, while exposure, selection, measurement, or statistical choices create a harmful association.
Clearing the curb may also change speed, turning geometry, curb use, attention, and behavior in ways that counteract visibility.
Exposure, site selection, treatment records, curb occupancy, and statistical design can change the estimated association.
Bus stops concentrate people who are waiting, boarding, alighting, transferring, and crossing. They also tend to sit on busier streets. A location can have more injuries because many more people use it, even when each crossing is safer.
After controlling for pedestrian crossings, transit boardings, vehicle volume, and turning movements, the adverse coefficient should shrink sharply. Separating bus stops from hydrants should matter.
Install short-duration pedestrian and turning counts at proxy sites; model injuries with pedestrian exposure as an offset and bus ridership as a covariate.
Hydrants and bus stops cluster along major avenues, commercial frontages, schools, subway entrances, truck routes, wide crossings, and areas with high curb activity; these conditions can independently raise injury counts.
The estimate should attenuate after matching on pedestrian and vehicle volume, lane count, street width, transit, land use, control type, truck route, and corner radius.
Use exact or propensity-score matching within the same street class and neighborhood; inspect balance before estimating outcomes.
Actual daylighting may be selected because an intersection already has crashes, visibility complaints, difficult truck turns, or unusual geometry. Selection on risk makes treated sites look dangerous even before treatment.
Treated intersections should show elevated injuries, conflicts, or complaints years before installation. A valid event study should show whether pre-treatment trends are parallel.
Plot annual outcomes for at least three pre-treatment years and use matched high-risk controls chosen before the installation decision.
A location record describes the intended curb rule; observed occupancy determines whether the sightline is open at a particular time.
The estimated effect should vary with the fraction of observed hours when the final curb segment is occupied.
Use repeated curb images or field observations to record vehicle type, dwell time, distance from the crosswalk, and sightline obstruction.
Daylighting acts at one corner and affects specific turns and crosswalks. Intersection-level data can attribute an unrelated crash on the opposite leg to a treatment outside its plausible path of influence.
After assigning crashes to exact corners, approach directions, and movements, any effect should appear only where the sight triangle or turning path intersects the treated space.
Geocode police narratives or diagrams to corner-level movement classes and pre-register the set of causally reachable crashes.
The proxy analysis divided injuries at a study intersection by the average at nearby intersections. When that nearby average is small, a modest numerator produces an extreme ratio. Sparse denominators make averages skewed and noisy.
Poisson, negative-binomial, or hierarchical count models should produce different estimates from an average of ratios, especially after retaining zero-count controls.
Reanalyze the raw counts with exposure offsets, corridor random effects, and robust sensitivity to zero or low-count comparisons.
Nearby intersections share traffic, land use, signal timing, construction, and crash trends. Some may also share comparison intersections. Treating them as independent makes the effective sample size too large. Many subgroup tests create additional false-positive risk.
Corridor-clustered standard errors, spatial random effects, block permutations, and false-discovery correction should widen uncertainty and eliminate some findings.
Estimate intraclass correlation by corridor and rerun inference with spatial block bootstrap and a prespecified family of outcomes.
Excluding protected bike lanes, hardened sites, street improvement projects, and locations without nearby crashes removes many complete designs and retains a selected set of clear zones without physical barriers. Conditioning on inclusion can also create collider bias.
Including all intersections and explicitly modeling co-treatments should change the estimate. Inverse-probability weighting should reveal whether included sites differ systematically.
Model the probability of entering the analytic sample, report covariate balance, and repeat the analysis under alternative inclusion rules.
Transit activity, pedestrian volumes, deliveries, speeding, and commuting changed dramatically during and after COVID-19. A seven-year aggregate can mix different trajectories across neighborhoods and street types.
Monthly or annual panel models with intersection and time fixed effects should differ from the pooled cross-section. Treatment effects may vary before, during, and after the disruption.
Use event-time plots, month-by-year effects, and sensitivity analyses that exclude 2020 to 2021.
Turning geometry, speed, curb operations, attention, and adaptation provide mechanisms that can be observed directly.
Removing a parked vehicle enlarges the asphalt available to a turning driver, while hardening defines the corner edge.
Signs alone should permit a wider path and a higher selected turning speed at susceptible corners.
Trace wheel paths from video; compare approach speed, minimum radius, turning speed, and crosswalk encroachment.
Parked vehicles make the street feel narrower and demand steering attention. Empty curbside asphalt can increase perceived operating width and encourage faster approaches, turns, or departures.
Approach or departure speeds should rise after sign-only daylighting, particularly in uncongested periods, but remain stable or fall where barriers or curb extensions preserve edge friction.
Compare full continuous speed distributions by time of day, and stratify them by hardening.
Daylighting could reduce some low-speed conflicts while increasing the speed of remaining impacts. Total crash frequency may stay flat even as the probability of injury per crash rises.
Total crashes remain approximately stable while injury-to-crash ratios, estimated impact speeds, or injury severity increase.
Analyze all reported crashes and severity jointly, and use near-crash speed or delta-v proxies where available.
Limited visibility forces a driver to creep and pause. A clear view permits an earlier decision and can be used to maintain momentum, accept a shorter gap, or enter the crosswalk before completing the scan.
Complete-stop compliance falls, minimum approach speed rises, accepted gaps shorten, or braking begins later after sign-only daylighting.
Measure stop-line speed profiles and accepted gaps at stop-controlled and uncontrolled sites.
A clearer view lets a turning driver search farther for vehicle gaps. The driver may spend more time looking left at approaching traffic and less time checking the near-side crosswalk or a pedestrian approaching from the opposite direction.
Video or eye-tracking shows longer cross-traffic scans, fewer final pedestrian checks, and conflicts aligned with the neglected approach direction.
Code head orientation and final-check behavior in naturalistic video, or use an instrumented-driving experiment with matched intersection scenes.
An unprotected clear zone remains part of the roadway surface, so a driver can use it during a turn or curb maneuver.
Vehicle incursions should be more frequent at sign only sites than at hardened sites with similar geometry and curb demand.
Mark the intended zone in video; count incursions, maneuvers, dwell time, speed, and proximity to the crosswalk.
A consistently parked car is a predictable obstruction. A bus, delivery van, taxi, or garbage truck arrives and departs, blocks views suddenly, forces merges, and may discharge pedestrians into the roadway.
Conflicts cluster around bus arrivals, deliveries, school pickup, ride-hail stops, and moments of curb-zone entry or exit.
Model risk as a time-varying function of actual curb occupancy and vehicle type; replace the static daylighting label with observed curb conditions.
Curb access removal shifts demand. Vehicles may double park, stop in a bike lane or crosswalk, move to the opposite corner, circle, or make extra U turns. These vehicle movements can create conflicts elsewhere in the block.
The treated corner clears while illegal stopping, merges, U-turns, and conflicts rise on adjacent legs or the next block.
Use a spillover study area of at least one block in every direction and track curb occupancy, circulation, and conflicts.
A clearer corner can change route choice and lower waiting cost. More people may cross there, raising absolute injuries even if each individual crossing becomes safer.
Pedestrian volume increases, injuries per intersection rise or remain flat, but injuries or serious conflicts per 10,000 crossings decline.
Measure route choice and crossing volume at treated and neighboring corners before and after installation.
People who believe drivers can see them may enter sooner, accept smaller gaps, look less, cross outside the marked path, or continue through a developing conflict. Cyclists may make the analogous assumption.
Waiting time falls, accepted gaps shrink, head checks decline, or off-signal entries rise when the clear zone is genuinely unobstructed.
Code pedestrian and cyclist decision behavior and distinguish visibility conditions from the mapped treatment.
Parked vehicles obstruct sight and physically constrain turning paths. A clear zone without barriers improves visibility while remaining open to vehicle encroachment. Hardened daylighting adds a low barrier that preserves visibility.
Sign-only sites show more curbside encroachment and close passes, while bollards, planters, bicycle corrals, granite blocks, or curb extensions eliminate them.
Compare lateral clearance and curb-zone entry across unchanged, sign-only, and hardened corners.
Empty asphalt has ambiguous meaning. A driver may read it as a turn bay or loading space while a pedestrian reads it as protected clear space. Faded markings, darkness, rain, or snow make the ambiguity worse.
Conflicts concentrate at weakly marked sites, in low visibility, or during early adaptation. Standardized hardening should reduce surprise and misuse.
Stratify by marking quality, lighting, weather, treatment age, and driver trajectory; conduct comprehension surveys.
The net effect equals a visibility benefit minus speed, turning, misuse, and exposure costs. Visibility may dominate on narrow, occluded streets; adverse mechanisms may dominate on wide streets, truck routes, or open asphalt without hardening.
Treatment effects interact strongly with width, curb radius, truck and turn volume, control type, pedestrian demand, and hardening. A single citywide average masks these regimes.
Estimate a prespecified heterogeneity model and validate decision rules on held-out intersections.
Bus-stop pedestrian exposure, proxy misclassification, and comparison-method bias. The later reported reanalysis retained the association for bus stops and lost it for hydrants, which points strongly in this direction.
A newly usable turning envelope, higher selected speed, and use of the clear zone as roadway. These mechanisms directly explain why hardening could outperform signs alone.
Dynamic curb obstruction, displaced loading, attention shifts, lost physical buffer, and behavior adaptation. Their importance is likely site-specific.
A general causal claim that installing daylighting raises injuries lies outside the current evidence. The recorded sign-only treatment estimate was essentially zero, while hardened treatment was associated with fewer pedestrian injuries.
Monitor, enforce existing rules, and avoid removing curb access without a clear visibility need.
Sign-only may be testable where speed and incursion risk are demonstrably low, with occupancy monitoring.
Daylighting addresses visibility; turn calming, curb management, or another treatment must address the dominant geometry risk.
Preserve visibility while physically controlling the turning envelope and preventing curb occupation.
No new curb clearance during the first phase.
Parking restriction and markings, with no physical barrier.
Same clear zone plus elements that prevent occupation and constrain the turn.
Watch 20 minutes of one intersection video or conduct a safe field observation.
Write a causal diagram and identify one competing explanation.
Count exposure, speed category, yielding, curb occupation, and one conflict indicator.
State the result that would weaken your hypothesis and the result that would support it.
Recommend status quo, sign-only, hardening, or a different treatment and defend the choice.
The reported +30% result was an observational proxy association. It used a cross sectional comparison without observed installation dates.
The actual-installation study found no detectable pedestrian effect for sign-only treatment and a statistically significant reduction for hardened treatment.
Bus-stop exposure, proxy error, and the comparison method are the strongest explanations for the original adverse association.
A real adverse sign-only mechanism remains plausible: open asphalt can enlarge the turning envelope, raise speed, invite incursion, and shift curb activity.
The engineering target is visibility plus physical control plus curb management. The research target is corner-level, exposure-adjusted, mechanism-aware evaluation.