Index to Measure Traffic Enforcement Effectiveness in Each Police Region of Victoria

Monash University Accident Research Centre – Report #172 - 2000

Authors: K. Diamantopoulou, M. Cameron & M. Shtifelman

Abstract:

The Victoria Police requested MUARC to develop a procedure that would measure traffic enforcement outcome levels against expected levels for each Victoria Police Region. To achieve this objective, relationships were developed that connected monthly casualty crashes in each of the five Police Regions with monthly variations in variables representing exposure, enforcement activity and other factors measured in each Region for the period 1989-1997. These relationships were achieved using structural (‘state-space’) time-series regression modelling techniques.

The models developed revealed the relative contribution of an increase in each enforcement operation to reducing the risk of casualty crashes in each Region after the effects of exposure changes and other factors had been taken into account. The coefficients (obtained from the regression models in most cases) measured the relative contribution of each enforcement operation, specific to each Region, and formed the basis of an index which allowed the overall contribution to traffic safety produced by the mix of enforcement activities in each Region to be assessed and compared between Regions.

An index for each Region was developed for the months January-December 1998 using the monthly average during 1997 as the base period. There was some variation found in the monthly indices both between and within Regions. When considering the overall index (i.e. for all regions combined – based on the five regions’ average), the estimated index has shown that relative to the previous year (1997), the Police performed better than average during the first quarter of 1998, but decreased their performance for the rest of the year.

Although the Police performance, as reflected by the index, was below average for most of 1998, the index should still be considered as a valuable tool for Police. This is because the index specifies which enforcement operations Police should increase their resources in per Region, to reduce the risk of casualty crash in that Region.

The indices developed for each Region during January-December 1998 were tested against actual road safety performance by comparing the observed crash frequencies per month with the expected levels (projected from the estimated models). This testing procedure appeared to work best when Police were performing better than average (as measured by the index), with reductions in casualty crash risk occurring in most cases. This was even more evident when the current month’s index was compared with the next month’s crash risk outcomes.

Executive Summary

The Victoria Police requested the Monash University Accident Research Centre (MUARC) to develop a procedure that would measure traffic enforcement outcome levels against expected levels for each Victoria Police Region. This followed a MUARC review of a Police internally developed index that had been established to measure the effectiveness of Police Districts in meeting their road safety objectives.

The purpose of this study was to develop relationships connecting monthly casualty crashes in each of the five Police Regions with monthly variations in variables representing exposure, enforcement activity and other factors measured in each Region.

These relationships were achieved using ‘state-space’ or structural time-series regression modelling techniques. These models are a more flexible family of models than econometric or traditional ‘fixed-effects’ models in that they allow certain parameters to have stochastic (random) variation from one period of observation to the next.

Form of Index

The relationship between casualty crashes and explanatory factors during time t was assumed to have the following form:

Crashes_t = a*(exposure factor_t)^b
*(enforcement factors_t)^c
*(other factors affecting risk_t)^d.

However, the only terms in this equation that were relevant to the calculation of the index were the enforcement factors. This is because the other variables or factors represent either the total potential for crashes (i.e. exposure such as population or vehicle travel) or represent the effects of other, non-enforcement factors on risk (i.e. unemployment rate, road safety advertising, alcohol sales) that are beyond the control of traffic police management.

Pragmatically, therefore, the index was expressed as:

Index_t = (random breath tests_t)^a*(speed camera traffic infringement notices_t)^b
*(penalty notice offences_t)^c*(mobile radar hours_t )^d* (laser hours_t)^e.

The parameters relating to random breath tests (RBTs), speed camera traffic infringement notices (TINs) and Penalty Notice offences were found by regression methods (i.e. using ‘state-space’ modelling techniques) because there was sufficient historical data to allow for this type of estimation. However, the parameters relating to mobile radar and laser hours of operation were estimated from independent evaluation studies since the data for these types of operations covered a period of one year (for laser operations) and under three years (for mobile radar operations).

There was also a need to make the index independent of scale so that Regions could be compared. This involved re-expressing the index in terms of ratios of Police activity. The enforcement activity (hours or offences) achieved during a particular month was divided by the monthly average of a base period (e.g. the previous year) so that the indices developed for each Region would be on the same scale and therefore be comparable.

Index for January-December 1998

An index for each Region was developed for the months January-December 1998 using the monthly average during 1997 as the base period. There was some variation found in the monthly indices both between and within Regions. For example, during January 1998 each Region performed better than what would have been expected on average during 1997, with Region 2 and Region 5 performing best amongst all Regions. However, during April, June, October and December each Region performed below what was expected on average for 1997. For the other months, some regions performed above average in a particular month whilst others performed below average.

When considering the overall index (i.e. for all regions combined – based on the five regions’ average), the estimated index has shown that relative to the previous year (1997), the Police performed better than average during the first quarter of 1998, but decreased their performance for the rest of the year.

Although the Police performance, as reflected by the index, was below average for most of 1998, the index should still be considered as a valuable tool for Police. This is because each Region’s developed index consists of negative parameters relating to all or some of the following enforcement operations – the number of random breath tests, the number of speed camera TINs issued, the number of penalty notice offences issued, the hours of mobile radar operations and the hours of laser operations.

The parameters or coefficients of these operations measured the relative contribution of an increase in each enforcement operation, specific to each Region, to reducing the casualty crash risk per Region. Because the parameters were, in most circumstances, negative, then this suggests that Police should increase their operations of that type of enforcement in a particular Region to reduce the risk of casualty crashes. Thus, the value of the index to the Police is that it indicates which enforcement operations Police should increase their resources in per Region to reduce the risk of casualty crash in that Region.

Testing the Index Against Crash Outcomes

Even though the indices developed for each Region had face value in that they could be assessed and compared, they needed to be tested against actual road safety performance. This was achieved by comparing the observed crash frequencies per month during January-December 1998 with the expected levels (projected from the estimated models).

This testing procedure appeared to work best when Police were performing better than average (as measured by the index), with reductions in casualty crash risk occurring in most cases. This was even more evident when the current month’s index was compared with the next month’s crash risk outcomes. However, when Police were performing below average, the index did not reflect casualty crash outcomes as well as it did when the performance was above average – both increases and decreases in casualty crash risk occurred.

Sponsoring Organisation: Baseline Research Program - Department of Justice, Transport Accident Commission, Royal Automobile Club of Victoria (RACV) Ltd, VicRoads