We analyse job grading within the UK National Health Service nursing profession, using 1994 survey data. We start from the ordered probit model, for which we develop and apply appropriate specification tests. Threshold constancy and covariate exogeneity are rejected, with important consequences for estimates of the influence of gender, ethnicity, training and career interruptions. We find little evidence of disadvantage for females relative to males, but significant differences in speed of promotion between ethnic groups, implying non-negligible differences in lifetime earnings.