Applying quantile regression for modeling equivalent property damage only crashes to identify accident blackspots
DOI: 10.1016/j.aap.2014.01.007
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 addresses the challenge of identifying accident blackspots (hot spots) in transportation networks, a process critical for efficient allocation of safety improvement funds. Traditional methods, such as the Empirical Bayes (EB) approach using negative binomial regression, often fail to adequately account for crash severity, suffer from the underreporting of minor injury and property damage only (PDO) crashes, and struggle with crash data that is heavily skewed by a preponderance of zeros. To resolve these issues, the authors propose a novel methodology combining a PDO equivalency calculation with non-parametric quantile regression. This approach aims to identify sites with disproportionately high risk by reflecting true societal safety costs rather than relying on arbitrary severity weights or mean-based models that are distorted by zero-inflated data. The study utilizes rural road segment data from South Korea, comprising 2,916 highway segments with crash records from 2005 to 2007 and roadway inventory data from 2008. The authors first calculate Equivalent Property Damage Only (EPDO) crashes by weighting fatal, major injury, and minor injury crashes against PDO crashes using societal cost estimates from Blincoe et al. (2002). For instance, a fatal crash is weighted as 1,330 PDO crashes. Because EPDO data do not follow standard statistical distributions, the authors apply quantile regression to model the 90th, 95th, and 97th percentiles of crash risk. This technique avoids distributional assumptions and focuses on the upper tails of the distribution, which is more relevant for identifying high-risk outliers than modeling the population mean. The proposed method is compared against the traditional EB method using negative binomial regression. The results demonstrate that quantile regression identifies a different set of significant predictors compared to traditional mean-based models. For example, while the negative binomial model identified horizontal curves and heavy truck volume as key factors for crash frequency, the quantile regression models for EPDOs highlighted terrain type, median presence, shoulder type, and specific speed limits (e.g., 40 kph, 50 kph, 60 kph) as significant factors at the 90th, 95th, and 97th percentiles. The study found that significant predictors change considerably as increasingly smaller subsets of high-risk outliers are examined, revealing heterogeneity across quantiles. The proposed method successfully identified high-risk sites by calculating the excess of observed EPDOs over predicted EPDOs at specific percentiles. The significance of this work lies in its ability to provide a more accurate and defensible identification of accident blackspots. By incorporating societal crash costs, the method reduces errors associated with the underreporting of minor crashes and accounts for severity without arbitrary weighting. Furthermore, quantile regression overcomes the limitations of traditional models in handling skewed, zero-heavy datasets. This approach allows safety engineers to tailor hotspot identification to specific resource constraints (e.g., inspecting the top 5% vs. 10% of sites) by targeting specific quantiles, thereby improving the efficiency of risk management and public fund allocation.
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | OpenAlex-citations | — | — | 1 | 2026-06-25 |
| archive | success | semantic_scholar | — | — | 6 | 2026-06-26 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-26 |
| chunk | success | chunk | — | — | 1 | 2026-06-26 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-26 |
| enrich | success | semantic_scholar | — | — | 1 | 2026-06-26 |
| promote | success | — | — | — | 1 | 2026-06-25 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-26 |
| verify | success | — | — | — | 1 | 2026-06-26 |
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.
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