Effectiveness of the Ohio Vehicle Action and Administrative License Suspension Laws

C. Results

1. Franklin County

Pooled results for all drivers. Table IV-3 shows that for both first and second DWS drivers pooled together, impounding and/or immobilizing the vehicle marginally reduced the number of DWS offenses but not the number of DUI offenses during the sanction period. This marginal effect may be due to the relatively short time (30 or 60 days) during which a DWS offender’s vehicle was normally impounded or immobilized. The notable feature of the upper portion of this table is that these DWS drivers experienced zero repeat offenses during the time their vehicles were not available to them. Despite this, the number of cases in the sanctioned group was so small, and the amount of exposure time available was so short, that the 100% differences between the vehicle sanctioned and no-vehicle sanctioned groups were, with two exceptions, not statistically significant.

This may be a case where more attention should be given to the effect size to avoid a "type two error." The effect sizes displayed in Tables IV-3 through IV-6 are a measure of the relative difference between the sanctioned group’s rate of reoffense and the comparison group’s rate of reoffense, using the latter group’s rate as the baseline or denominator. Thus, if the comparison group’s rate were 0.20 and the sanctioned group’s rate were 0.12 (i.e., 60% of the baseline rate), the relative difference would be -40%. Obviously, the largest reduction possible is necessarily bounded at 100%, when the sanctioned group’s rate is zero, but the largest increase is not bounded at +100% (which would represent a doubling-in rate).

In the lower section of Table IV-3, which gives the results for the DUI driver groups (where, as above, the total period of vehicle action is evaluated by combining the impoundment and immobilization periods), the relative differences are large: above 50% for both DWS and DUI offenses. DWS and DUI offenses are significantly reduced during the sanction period in both the Cox Regression and Kaplan-Meier analyses. Figure IV-1 shows the cumulative DUI offense hazard (recidivism) rates during sanction using the overall Cox Regression model fit based on the two DWS and two DUI driver groups pooled, which demonstrates a clear separation between those DWS and DUI offenders who actually received a vehicle action (solid line) and comparison drivers who were eligible but avoided the sanction (dashed line).

The upper section of Table IV-4 gives the results of the analysis of the DWS groups up to 2 years following the return of the vehicle to the offender. With the exception of DWS offenses for second DWS drivers, all of the relative differences are in the right direction, but none are significant. It appears that in the case of the DWS drivers, the sanction had no impact on DWS or DUI offenses after the vehicle was returned.

In contrast to the DWS drivers, the DUI drivers whose vehicles were impounded or immobilized for longer periods demonstrated significant reductions in DUI and DWS offenses after the vehicles were returned. The recidivism rate for DUI offenses of second DUI offenders following return of the vehicle was 22% to 38% lower than for offenders who were eligible but did not receive a vehicle sanction. These differences in rates are corrected for variations in age and prior driving record through the entry of these factors as covariates in the Cox Regression analysis (right-hand columns of Table IV-4).

Figure IV-2 shows the pooled cumulative hazard rates for the DUI offense after sanction across both DWS and both DUI driver groups, using the overall Cox Regression model fit. The solid line shows the accumulation of new offenses by the sanctioned group, while the dashed line shows the cumulative rate of DUI offenses for the comparison drivers. The origin of the graph is set at the time when the experimental group had their vehicles returned. The origin for the comparison group is set at the average time after conviction when the sanctioned group had their vehicles returned. The experimental group of drivers who received the vehicle sanction clearly had fewer DUI convictions than similar offenders not sanctioned in the months following return of their vehicles.

Hamilton County

Pooled results for all drivers. In the Hamilton County recidivism analysis, there was no need to consider combining vehicle immobilization and vehicle impoundment since, with few exceptions, vehicles were simply impounded for the length of the sanction period. The two upper sets of rows of Table IV-5 present the Cox Regression and Kaplan-Meier analyses of recidivism rates for DWS drivers during the sanction period. These analyses found significant reductions in DWS offenses, but not DUI offenses, for the combined groups of drivers. The effect sizes for DWS offenses was large, running between 45% and 75%. The reductions in DWS and DUI offense rates for the experimental group of DUI drivers are generally larger than those demonstrated by the DWS drivers whose vehicles were impounded for a shorter time. Pooled effect sizes for both DWS and DUI offenses varied from 40 to 60%. Figure IV-3 shows the pooled cumulative hazard rates for DUI offenses across the two DWS and two DUI driver groups during the sanction period using the overall Cox Regression model fit.

Table IV-6 provides the results of the analysis for DWS and DUI offenses in the postsanction period. The Kaplan-Meier analysis shows some differences in the DWS recidivism for the DWS drivers; however, this was not confirmed in the Cox Regression analysis. In any case, the effects sizes were too small to be of interest. Both the Cox Regression and the Kaplan-Meier analytical procedures detected significant reductions during the postsanction period in both DWS and DUI offenses for the DUI drivers. The Cox regression yielded an effect size of 30% for DUI offenses. Figure IV-4 shows the pooled cumulative hazard rates for DUI offenses across the two DWS and two DUI driver groups after the vehicles were returned to the offenders, using the overall Cox Regression model fit.