Introduction: Differences in crash severity driven by geography and population density are highly relevant not only for traffic authorities, but also for insurance companies. In European countries, populations are concentrated in high‑density areas, while the most severe injuries and fatalities occur in low‑density regions. This study quantifies how much of the severity gap between high‑ and low‑density areas is explained by differences in the distribution of observable risk factors (composition effect) versus differences in how these factors influence outcomes (structure effect).
Methods: Crash severity is measured using a novel monetary approach based on the aggregate Value of a Statistical Life (VSL) for all casualties. The analysis evaluates differences across the full distribution of crash severity. A counterfactual regression method is applied to decompose differences into composition and structure components.
Results: Using Spanish crash data from 2021 to 2023, results show that differences across density regions in road type, collision type, age, sex, and the involvement of two‑wheelers explain a substantial share of the severity gap up to the median. At higher severity levels, differences in the impact of these factors become increasingly important. Summarizing, the composition effect dominates up to the 60th–70th percentiles of the severity distribution, indicating that differences in the distribution of observable characteristics account for most moderate severity gaps. Beyond these percentiles, the structure effect becomes more influential, suggesting a growing role for unobserved factors in the most severe crashes.
Conclusions:
The findings indicate that insurers should integrate population density into premium design to better reflect heterogeneous risk profiles and ongoing demographic trends such as population ageing and geographic relocation in Europe. Counterfactual analysis provides insurers with a valuable tool for identifying the specific sources of the observed severity differences between high‑ and low‑density areas, thereby enabling more accurate risk assessment and more efficient pricing strategies
