3. PROGRAM EVALUATION

APPROACH

We used a two-pronged approach in estimating the effect of the Austin DWI enforcement program on alcohol-related crashes. Ultimately, we wanted to know if the number of such crashes was reduced, but first, we wanted to see if there had been any positive changes in APD performance that might plausibly be expected to reduce alcohol-related crashes. Performance measures examined were: DWI arrests, DWI processing time, and DWI conviction rates. These measures were selected as indicative of the kinds of positive effects sought in the program as de-scribed in the prior chapter. If the changes were in the right direction and large enough, then any reduction found in alcohol-related crashes would more likely to be due to the program rather than some other factor.

Our analysis of alcohol-related crashes in Austin used number of fatal-crash involved drivers with a BAC of .01 or more as the measure of effectiveness. We compared changes in such crashes to changes in the rest of Texas as another device for accounting for non-programmatic effects on program effectiveness.

RESULTS

Performance

DWI Arrests. Figure 3 1 shows that the annual number of DWI arrests by APD officers had been trending down from about 3,500 in 1990 to only 2,200 in 1996, but holding fairly steady from 1994 through 1997. With the start of the DWI enforcement task force in 1998, arrests increased to 4,077, an increase of 48% over the 2,747 arrested for DWI in 1997. A further increase to 4,500 occurred in 1999 during the transition from the DWI task force to the full-time DWI unit. The arrest rate remained at a high level in the following two years, with only a slight dip in 2001. These numbers for the latter part of the time period shown are reflected in data from the APD computer aided dispatch system (CAD). Police officers routinely report their involvement in incidents to a dispatcher who enters data for determining response time (time from receipt of a call for service to arrival at the scene of the incident) and time to perform any duties pertinent to the call, including any arrest and post-arrest processing. Incidents initiated for reasons other than a call for service (for example, flowing from a traffic stop) are also entered into the CAD, with a response time of zero.


Figure 3 1: DWI Arrests by the Austin, Texas Police Department, 1985-2001

image - figure 3-1 see "d" for long descriptiond


Figure 3 2 superposes the number of incidents initiated in any way (e.g., call for service or traffic stop) and cleared as DWI, on the above arrest data in monthly form. The data are shown in monthly form for units of all types, including general patrol units and the DWI team. It is seen that the arrest data track very well with the CAD data for the period in which CAD data were available, but indicate slightly more arrests in a given month than incidents reported to the CAD. This suggests that not all DWI incidents resulting in arrests were reported to the CAD, and suggests also that there may have been a systematic bias in reporting DWI incidents to CAD, since the two curves follow the same pattern.

Available data do not permit a breakdown of DWI arrests by type of police unit, but CAD data are available for use as a surrogate. The CAD data indicate that, after an initial high of about 100 – 150 per month, calls answered by the DWI team have settled down to about 50 – 100 per month (Figure 3 3). Calls involving general patrol units have remained fairly constant throughout the period, fluctuating about a mean of roughly 225 per month. This indicates that the increase in calls since the start of the DWI team in September 1998 has been due to the additional calls answered by that team and not to more calls involving general patrol units. Since officers drawn from other elements of the APD have staffed the DWI team, the increase in DWI-related calls suggests an overall increased efficiency with respect to DWI enforcement.

Data are available for DWI arrests by the DWI team for the year 2000 through most of 2002. Figure 3 4 compares the number of arrests made by the team with calls reported by the team for an overlapping period of time. In addition, the figure also shows the number of arrests in which the team participated, including both unassisted arrests and arrests in which


Figure 3 2: Number of DWI Arrests (1985-2001) and Number of Incidents Cleared as DWI Reported to the Austin Police Department Dispatcher (1997-2001)

image - figure 3-2 see "d" for long descriptiond


Figure 3-3: Number of Incidents Cleared as DWI Reported to the Austin Police Department Dispatcher by Type of Police Unit, 1997-2001

image - figure 3-3 see "d" for long descriptiond


assistance was given to general patrol units. Again, arrest patterns track well with call patterns, with the number of calls falling about halfway between the number of unassisted arrests made and the number in which the team participated.

Figure 3-4: Calls, Arrests, and Arrest Participation for the APD DWI Team, 2000-2002

image - figure 3-4 see "d" for long descriptiond


Finally, the CAD data indicate that calls initiated by a traffic stop comprise the largest percentage of DWI calls (Figure 3 5). For example, in December 1998, 251 calls out of 401 (63%) were initiated by a traffic stop, and in December 2000, 242 calls out 359 (67%) were initiated by a traffic stop.

Figure 3-5: Number of Incidents Cleared as DWI Reported to the APD Dispatcher by Type of
Call Initiation, 1997-2001

image - figure 3-5 see "d" for long descriptiond


DWI Processing Time. The CAD data described above provide a means for estimating DWI arrest processing time. Overall, mean processing time has remained in the 2.5 – 3.0 hours range during the 1997 – 2001 period, starting at the lower end of this range and rising slightly toward the higher end in late 1998 (Figure 3 6). The data indicate slightly higher processing times for the DWI team and lower processing times for general patrol units. A possible reason is that general patrol units may be the handing off the more complicated cases to the DWI team, and may in some instances, be reporting back in service after hand-off but before processing. The data also indicate that calls initiated as traffic stops require slightly higher processing time than do calls in general.

Figure 3-6: Mean Processing Time for an Incident Cleared as DWI Reported to the
APD Dispatcher, 1997-2001

image - figure 3-6 see "d" for long descriptiond


DWI Convictions. A majority of DWI arrests in Travis County involve first-time offenders and are initiated by the Austin Police Department. In Texas, the County Courts and the County Court of Law Courts handle first-time DWI offenders as mandated by Sections 49.04 or 49.09 of the Texas Penal Code, unless the arrest involves a felony charge, for example, due to a fatality. Multiple offenders and/or felony cases are handled in District Courts.

During a past NHTSA-sponsored project (Wiliszowski, Jones, and Lacey, 1999), it was calculated that in fiscal year 1996, statewide, County Courts in Texas had a 63% conviction rate in DWI-related cases. Based on published court activity, project staff extracted Travis County convictions and dispositions for that year and calculated a conviction rate for DWI cases of almost 70% (69.78%). In this project, we looked at DWI convictions as a percent of total DWI dispositions from fiscal year 1996 through 2001. Conviction rates for this period are shown in Figure 3 7. Conviction rates for DWI cases adjudicated in County Courts located in Travis County increased from 70% before the start DWI enforcement improvement program to 77% immediately after the start of the program. This 10% increase in conviction rate remained in effect through the year 2001.

Figure 3-7: DWI Conviction Rate in Texas County Courts in Travis County, 1996-2001

image - figure 3-7 see "d" for long descriptiond


Impact

As indicated above, our impact analysis used number of fatal-crash involved drivers at a BAC of .01 or more as the measure of effectiveness. Data from the NHTSA Fatality Analysis Reporting System (FARS) were used in the analysis, with missing BAC data estimated using NHTSA’s new multiple imputation procedure (Subramanian, 2002). An interrupted time series model employing the autoregressive integrated moving average (ARIMA) technique was used in the analysis. Two different time series were analyzed, the first consisting of semi-annual counts of the target drivers in Austin, Texas, and the second consisting of a comparison series of semi-annual counts of the same types of drivers in the remainder of the state. A step-function inter-vent ion was placed at the point corresponding to the start of the DWI enforcement program, taken to be January 1, 1999.

The model for Austin was differenced by one period and had moving average components at lags of one period and two periods. Figure 3 8 shows the results for the model, indicating a statistically significant 25% reduction in drivers at .01+ compared to what would have been expected with no program (p=0.037). By contrast, the comparison model for the rest of the state showed no significant change in drivers at .01+ after the intervention point (p=0.807).

We also examined a time series composed of drivers at a BAC of .10 or more and found a similar positive effect in Austin (Figure 3 9). The effect equated to a 19% reduction in such crashes, but the reduction was not significant at the .05 level (p=0.1097).

Figure 3-8: ARIMA Analysis of Number of Drivers at a BAC of .01+ in Fatal Crashes in
Austin, Texas, 1990-2001

image - figure 3-8 see "d" for long descriptiond


Figure 3-9: ARIMA Analysis of Number of Drivers at a BAC of .10+ in Fatal Crashes in
Austin, Texas, 1990-2001

image - figure 3-9 see "d" for long descriptiond



SUMMARY AND CONCLUSIONS

We examined both the performance and the impact on alcohol-related fatal crashes of the APD DWI enforcement program. We found positive changes in both, lending support to the hypothesis that the program has been an effective alcohol-crash countermeasure. By reassigning some general patrol personnel to the new DWI team and increasing command emphasis of DWI enforcement, DWI arrests nearly doubled. Further, the number of alcohol-impaired drivers in fatal crashes has decreased by an estimated 25% since the start of the program. Several new procedural initiatives designed to increase DWI conviction rates have been placed into operation, and DWI conviction rates have increased by an estimated 10%.

Despite these positive results, some of the hoped-for improvements were not accomplished. Overall, DWI processing times did not decrease, possibly at least in part, due to the physical and procedural changes brought about by a new jail facility operated outside of the control of the APD. Processing times may even have increased slightly for the DWI team, possibly because of an increase in the complexity of cases handled by the team. Also, the DWI-arrests of the general patrol units did not increase, but remained at about the same level as in the year be-fore the program. Nevertheless, the program overall must be viewed as successful and worthy of consideration by other jurisdictions.