Drinking Driver and Traffic Safety Project. Volume 2, Probabilities for Drinking Drivers

McEachern, Alexander · 1973 · ROSA P / United States. National Highway Traffic Safety Administration

archive: archived pipeline: cataloged verified

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

Summary

This report, titled *Drinking Driver and Traffic Safety Project, Vol. II, Probabilities for Drinking Drivers*, presents the findings of a four-year study (1968–1972) conducted by the University of Southern California’s Public Systems Research Institute for the National Highway Traffic Safety Administration. The primary objective was to develop a statistical prediction model to estimate the likelihood that an individual is a drinking driver or a recidivist drinking driver based on specific demographic and behavioral characteristics. The study aimed to provide administrative and judicial decision-makers with data-driven tools to identify high-risk individuals, thereby aiding in the allocation of resources for traffic safety and the formulation of legal dispositions. The prediction model was developed using a dataset comprising over 4,000 cases, which included convicted drinking drivers, recidivist drinking drivers, and driver’s license applicants with no prior drunk driving convictions. The model utilizes five specific variables to generate a five-digit index number for each individual: education level, total number of minor traffic violations in the past three years, age, number of traffic accidents in the past three years, and total number of non-traffic arrests in their lifetime. Each variable is assigned a numerical code based on specific categories (e.g., education ranges from grades 1–6 to graduate work; age ranges from 21 or under to over 50). The report provides comprehensive tables listing probability estimates for every possible combination of these five characteristics. The results consist of two distinct probability estimates for each index number. The first estimate indicates the probability that an individual belongs to the class of those convicted of drunk driving, as opposed to those who have not. The second estimate indicates the probability that a convicted drunk driver will reoffend (recidivism), as opposed to committing only a single offense. The report demonstrates that these probabilities correspond closely to actual outcomes in the sample data; for instance, among individuals with a calculated probability of 0.80 for being a drunk driver, 80% were indeed drunk drivers. Similarly, for recidivism, 50% of those with a 0.50 probability were recidivists. The text emphasizes that these probabilities are numerical expressions of uncertainty derived from historical data and do not explain the causes of drunk driving. The significance of this work lies in its application as an aid for administrative and judicial decisions. The authors suggest that these probability tables can help officials identify high-risk intersections or individuals, similar to how traffic engineers use accident data to prioritize infrastructure improvements. However, the report cautions that these probabilities should not be used as sole determinants for punishment or treatment, as they inevitably involve a margin of error. For example, treating everyone with an 0.80 recidivism probability as a recidivist would result in errors in 20% of cases. The final interpretation and application of these probabilities are left to the judgment of the decision-maker, who must weigh the statistical likelihood against the potential consequences of error.

Key finding

The prediction model accurately estimates the probability of an individual being a drunk driver or a recidivist drunk driver based on five demographic and behavioral variables, with predicted probabilities closely matching actual observed frequencies in the sample.

Methodology

dataset

Sample size: 4000

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. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

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

• incidence prevalence

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, observational prevalence