In an attempt to lower the level of road trauma in New Zealand, the Land Transport Safety Authority introduced the Supplementary Road Safety Package (SRSP) in October 1995. The package consists of targeted speed and alcohol enforcement, and features graphic television advertising highlighting the consequences of unsafe driving. Over the first four years the campaign was allocated a budget of NZ$50.06 million and charged with reducing 80 fatalities, 450 serious injuries and 1600 minor injuries. Although a requirement of the package's approval was that it be thoroughly evaluated, no consistent conclusion has been drawn. Recognising the discrete and strictly positive nature of road trauma measures, this dissertation adds to the body of literature by adopting statistical modelling techniques specifically designed for the analysis of such count variables: The Poisson and Negative Binomial regression models. While the Poisson model finds a significant level effect on the number of serious injuries from the SRSP's introduction, no statistically significant effect is found using the more appropriate Negative Binomial model. (Author/publisher)
Abstract