Dataset for “Benchmarking Safe System policy performance across Brazilian cities”

Published: 12 September 2026| Version 1 | DOI: 10.17632/mgm8n4gxs5.1
Contributor:
Caio Torres

Description

This dataset supports the study “Benchmarking Safe System policy performance across Brazilian cities: An integrated analytical framework”. It contains city-level data for 110 Brazilian municipalities, using 2019 as the reference year. The dataset includes 27 policy indicators representing five Safe System dimensions—Road Safety Management, Safer Roads and Mobility, Safer Vehicles, Safer Road Users, and Post-crash Response—together with an ordinal road traffic fatality outcome and socioeconomic variables used for contextual peer grouping. The policy indicators were compiled from a structured questionnaire administered to municipal road safety and transport agencies and from official Brazilian administrative data sources. The accompanying data dictionary provides variable definitions, coding schemes, data sources, and analytical roles. The data were used to estimate a city-level structural equation model and subsequently to derive dimension-specific policy performance scores, an overall Policy Performance Index, and socioeconomic peer-group benchmarking measures.

Files

Steps to reproduce

1. Use the city-level dataset provided in the Data worksheet. Variable definitions, coding schemes, sources, and Safe System dimensions are documented in the accompanying Data_dictionary worksheet. 2. Estimate the Safe System structural equation model using the ordered categorical policy indicators and road traffic fatality outcome. The model was estimated in R using the lavaan package with the weighted least squares mean- and variance-adjusted (WLSMV) estimator. 3. Use the standardized factor loadings and standardized structural parameters from the final model to derive the dimension-specific structural relevance coefficients as described in the associated article. 4. Before calculating the Dimension Performance Scores (DPS), rescale each retained ordinal policy indicator to the 0–1 range by dividing its observed value by the maximum value of its corresponding coding scale. 5. Calculate each dimension-specific DPS by combining the rescaled indicators, their standardized factor loadings, and the corresponding structural relevance coefficient. Calculate the Policy Performance Index (PPI) as the sum of the retained dimension-specific DPS values. 6. Standardize municipal population and Human Development Index (HDI) values and apply k-means clustering using Euclidean distance and multiple random initializations. The resulting four clusters are treated as socioeconomic peer groups for benchmarking. 7. Within each peer group, calculate the mean DPS for each retained Safe System dimension. For each city, calculate the relative performance gap as the difference between the peer-group mean DPS and the city DPS divided by the peer-group mean DPS. The dimension with the largest positive relative gap is identified as the priority dimension. In the event of equal maximum relative gaps, select the dimension with the highest structural relevance coefficient. 8. Benchmark cities are defined, separately for each Safe System dimension and peer group, as the city or cities with the highest observed DPS for that dimension. Detailed model specification, equations, indicator selection criteria, and benchmarking procedures are provided in the associated article.

Institutions

Categories

Transportation Engineering, Transport Policy, Public Policy Issue, Road Safety

Funders

Licence