Do we know why the number of traffic fatalities is declining? If not, can we find out?

Elvik, Rune; Høye, Alena · 2022 · Crossref

DOI: 10.55329/oyhu8693

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper investigates the difficulty of identifying the causal factors behind the 80–90% decline in traffic fatalities observed in many highly motorized countries since their peaks in the 1960s and 1970s. The authors argue that establishing a rigorous scientific explanation for this long-term trend is nearly impossible because historical data precludes randomized controlled trials or quasi-experimental designs, which are necessary to establish counterfactuals. Consequently, the study evaluates two suboptimal methods for attribution: multivariate statistical analysis and historical reconstruction. The authors first critique multivariate statistical analyses, noting they are inherently flawed due to multicollinearity and omitted variable bias. To demonstrate this, they developed a negative binomial regression model for Norway (1997–2018) using 11 independent variables, such as infrastructure length, seat belt usage, and vehicle safety features. Although the model appeared to fit the data closely, it was over-fitted and produced implausible results, such as suggesting that 2+1 roads reduced fatalities by 90% or that electronic stability control increased fatalities by a factor of nearly 300,000. These errors stemmed from high correlations among variables that all increased over time and the inability to account for missing data on other influential factors, such as driver distraction or mean speed. As an alternative, the paper examines historical reconstruction, a method that estimates the contribution of specific factors by reconstructing hypothetical counterfactuals based on available data and evaluation studies. The authors review six studies from Norway, Great Britain, Sweden, and the Netherlands. These studies consistently identify safer road user behavior as a major contributor to fatality reductions, followed by infrastructure improvements and vehicle safety features. For instance, a recent Norwegian study estimated that known factors explained 59% of the decline in killed or seriously injured road users from 2000 to 2019. However, the authors emphasize that these estimates are not causal proofs but rather educated guesses. The relative importance of factors is heavily dependent on data availability; factors with abundant data appear more significant than those with missing information. Furthermore, these reconstructions cannot be empirically tested because history cannot be rerun. The study concludes that no ideal method exists to scientifically explain the decline in traffic fatalities. Both statistical models and historical reconstructions suffer from incomplete data and the inability to establish true causality. While historical reconstructions provide a plausible impression of relative contributions—highlighting the roles of behavior, infrastructure, and vehicles—they remain limited by their reliance on available information. The authors assert that any quantification of factor contributions carries a high risk of being misleading due to omitted variables, and thus, definitive scientific explanations for the decline remain out of reach.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• fatality injury trends
• comparative international

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: crash risk outcomes