In order to better understand the various contributors, the Ministry of Transport commissioned an econometric analysis of the 2011 road toll. The objective was to assess the degree to which the drop in 2011 could be attributed to any particular causes, rather than to random variation.

Download the full Analysis of potential factors behind the 2011 reduction in New Zealand road fatalities (PDF, 841kb)

The econometric analysis was extended in 2013 to include 2012 road toll data

Download the further analysis of potential factors behind the 2011 reduction in New Zealand road fatalities [PDF, 281 KB]

An explanation of the methodology for econometric modelling

The method adopted was two stage. Firstly a long-term model was used to capture the downward trend in fatalities, and secondly a short-term model was developed to explain the remaining variation.

Long-term trend

The underlying trend from 2000 to December 2010 was established to be 10.4 fewer deaths each year. Ongoing safety interventions and technology improvements have contributed to this ongoing improvement.

The graph below compares the actual road toll with that predicted by the long term model. Quarter three is typically lower than the other quarters, so the model includes a seasonal effect for that quarter. That is why the predicted quarter three numbers are lower than the trend for the other quarters.

Graph showing road toll and long-term from 1999 Q1 to 2010 Q2

 

The long-term trend model predicts a road toll of 355 for 2011, a decrease of 20 on 2010. The actual toll was 284, which was a decrease of 91 on 2010. That means the long-term trend explains 20 of the decrease of 91.

Short-term model

The short-term model was developed to explain the reduction in 2011, over and above the underlying trend. It was formulated using the data from 2000 to 2010, and provided the best predictor of the variation from the long term model that was seen in those years.

Among the factors considered were travel, advertising, enforcement, short-term changes in vehicle quality, weather, real GDP, terms of trade (possibly affecting rural travel), light vehicle prices, interest rates, alcohol consumption, Waitangi day and ANZAC day falling on weekends, unemployment, real wages, fuel prices, motorcycle registrations, real private consumption and vehicle speeds.

Many combinations of those data items were tested, and the best explanatory variables were determined to be petrol prices, real wages and motorcycle registrations. The short-term model was then applied to 2011.

There was a drop of 91 road deaths from 2010 to 2011. The short-term model (petrol prices, wage levels and motorcycle registrations) explained 21 of these deaths, the long-run trend explained 20 deaths. The remaining 50 deaths cannot yet be explained by trends and models, and could be random variation, but with the passage of time and further work, contributing factors may become more apparent.

In the graph below what the model predicts is shown by the red line and the actual road toll is shown by the blue line. The model predicts the toll road well for the period it was fitted to (2000-2010) but does not predict all the drop experienced in 2011.
Graph showing road toll, long term model and short-term model  from 2000 Q1 to 2011 Q2