Abstract
A new data set on job related mortality within 347 3 digit occupations is merged with two national probability samples of United States workers to assess which groups are in hazardous and which in safe jobs. Logistic regressions are fitted to the two samples and the following covariates are assumed to explain employment in a hazardous job: age, race, gender, schooling, region, and marital status. Results from both samples are remarkably similar. Odds ratios for non-whites versus whites are roughly 1.2/1; for high school dropouts versus people with some college education roughly 2.2/1; and for men versus women roughly 5.0/1.