Breath alcohol measurements and other data collected at randomly selected roadside sites were combined with data on fatally injured drivers in crashes occurring on the same weekdays and times (Friday and Saturday nights) at locations matched by the size of the nearest town. A logistic model was fitted to these data for the years 1995–2000 to estimate the effects of alcohol, driver’s age and the influence of passengers carried on the risk of driver fatal injury in New Zealand. The estimated risks increased steeply with increasing blood alcohol concentration (BAC), closely following an exponential curve at levels below about 200 mg/dl (i.e. 0.2%) and increasing less than exponentially thereon. The model fitted to data for drivers under 200 mg/dl showed that risks at all BAC levels were statistically significantly higher for drivers aged under 20 (over five times) and for drivers aged 20–29 (three times) than for drivers aged 30 and over. Further, controlling for age and BAC level, driving with a single passenger was associated with approximately half the night-time risk of driver fatal injury relative to driving either solo or with two or more passengers. According to a recent travel survey, the types of passengers carried at the times of night and days of week studied appear to differ significantly from the types of passengers carried generally, which may lead to different passenger effects on driver behaviour. The high relative risk of teenage drivers means that they reach high risk levels commonly regarded as unacceptable in the field of road safety even at their current legal limit of 30 mg/dl, particularly when more than one passenger is carried in the car.