|
|
 |
Impact of the graduated driver licensing program in Nova
Scotia
Report from the Traffic Injury Research Foundation
Results: Impact of the Program on Young Novice Drivers
To determine if changes occurred in the collision experience of 16- and
17-year-old drivers after the graduated licensing program was introduced,
a comparison was made between data from 1993 and 1995 (the years before
and after the program was implemented). A similar comparison was made for
the internal control group (drivers over the age of 25) and the external
control groups (the other jurisdictions).
In all cases, to control for changes in population, the comparison data
were based on crash rates – the number of collisions per 10,000 population.
Table
2 presents population estimates, numbers of crash-involved drivers and
per capita crash rates for 16- and 17-year olds as well as persons aged
25 and over for Nova Scotia and two of the control jurisdictions – Saskatchewan
and Maine. Similar information for casualty crashes is shown in Table
3 for Nova Scotia and the three control jurisdictions – Saskatchewan,
Maine and New Brunswick.
To standardize the comparisons, crash ratios were computed using the method
described by Ulmer et al. (1999) in the evaluation of Florida’s graduated
licensing program. To obtain the crash ratio, the crash rate for each target
group is divided by the crash rate for drivers age 25 and over. This standardizes
the crashes of a target group to the population of that group as well as
to the crash rate of the control group -- drivers age 25 and over.
The crash ratios in 1993 (before the program was introduced) were compared
to the crash ratios in 1995 (after the program was introduced). Any changes
observed in these ratios control for changes in the population within that
age group as well as changes in crash rates among older drivers. The changes
in overall crash ratios are summarized in Figures
2; Figure
3 shows changes in casualty crash ratios. The z statistic was used to
determine if the change in crash ratios was statistically significant.
Among 16-year-old drivers in Nova Scotia, there was a 23.8% decrease in
their crash ratio and a 33.9% decrease in their casualty crash ratio between
the pre-program and post-program periods. Both these decreases were statistically
significant. In contrast, no significant declines in collision ratios occurred
in the comparison jurisdictions. In fact, the crash ratio for 16-year-old
drivers in Maine actually increased from 1993 to 1995.
Among 17-year-old drivers in Nova Scotia, no significant decreases were
detected in their crash ratios. None were found in the control sites as
well – the only significant changes were those found in Maine, which were
actually increases in crash ratios.
The failure to find a significant decline among 17-year-old drivers in Nova
Scotia may be due to the fact that many drivers who were 17 years old in
1995 had actually been licensed under the old system and were not in the
graduated licensing program. They received their learner’s permit when they
were 16, which would have been prior to the introduction of graduated licensing
in October 1994. Thus, the impact of graduated licensing on the crashes
of 17-year-old drivers would not be noticed until at least 1996. In fact,
the full impact of the graduated licensing program even on 16-year-old drivers
may not have been realized until 1996 because even some drivers who were
16 in 1995 would have been licensed under the old system.
To test this possibility, the analyses were repeated for 16- and 17-year-old
drivers in Nova Scotia, comparing their crash ratios in 1993 to those in
1995 and 1996. This showed a further decline in the crash ratios for 16-year-old
drivers in 1996, the second year of the program – i.e., from a crash ratio
of 1.62 in 1993 to 1.23 in 1995 and then 1.04 in 1996.
This, represents a 35.8% decline (p<.01) in the overall crash ratio among
16 year olds between 1993 and 1996 and a 48.1% decrease (p<.01) in their
casualty crash ratio from 1993 to 1996.
There was also a significant decline during this timeframe in the crash
ratios among 17-year-old drivers in Nova Scotia – they experienced an 11%
decline (p<.05) in their overall crash ratio and a 17% decrease (p<.10)
in their casualty crash ratio.
Taken together, these findings suggest that the graduated licensing program
in Nova Scotia was associated with a reduction in crashes among 16- and
17-year-old drivers. There was an immediate impact on the crashes of 16-year-old
drivers in 1995 and a further effect in 1996. The impact of the program
on 17-year-old drivers was smaller than observed for 16-year-old drivers
and crash reductions were only noticed in 1996, when all 17 year olds would
have been licensed under the new program.
The previous analyses compared collisions only in the periods immediately
preceding and following implementation of the program – i.e., 1993 compared
to 1995 and 1996. The reductions in collision rates observed in this timeframe
could, however, be influenced by a pre-existing downward trend in collisions
among young drivers. To explore this possibility the monthly numbers of
16-year-old drivers involved in all types of collisions as well as just
casualty collisions from January 1986 to December 1997 were subjected to
time series intervention analysis. The ARIMA procedure (Auto-Regressive
Integrated Moving Average) in the SPSS trends program was used to isolate
the yearly cyclical trends as well as the month-to-month correlation and
to develop a model for the collision series of 16-year-old drivers.
The time series analyses were used to test for the presence of both an abrupt,
permanent change and a gradual, permanent change following the introduction
of the graduated licensing program. The rationale for this arose from the
finding (described above) that collisions among 16-year-old drivers declined
in the year following implementation of graduated licensing (1995) but there
was a further decline in the second year (1996).
The abrupt, permanent change intervention parameter was a dummy variable
with an assigned value of ‘0’ in the pre-GDL period and an assigned value
of ‘1’ in the post-GDL period. The gradual, change intervention parameter
introduced the effect of change in the post-GDL period over a span of twelve
months, with the first month equal to 1/12, the second month equal to 2/12,
and so on, until the twelfth and subsequent months, which equal "1". This
procedure recognized that it may take up to 12 months from the date of program
implementation for all 16-year-old drivers to be licensed under the new
system. The gradual permanent change parameter adjusts for this transitional
period by introducing the intervention with a progressive rather than an
abrupt change.
The same procedures were used with the monthly collision series for drivers
age 25 and over in Nova Scotia as well as for drivers age 16 and those age
25 and over in the three comparison jurisdictions.
Table
4 summarizes the results of the time series analysis for total collisions
as well as for casualty collisions. The analyses revealed significant reductions
in total collisions and casualty collisions among 16-year-old drivers, commensurate
with implementation of the graduated licensing program in Nova Scotia. There
was a significant abrupt, permanent reduction of 6.59 collisions per month
following implementation of the program. And, there was an even larger gradual,
permanent change representing an average decrease of 19.57 collisions per
month and 6.83 casualty collisions per month.
Similar decreases in collisions were not found among drivers age 25 and
over in Nova Scotia. In fact, in contrast to the trend for drivers aged
16, casualty crashes among drivers aged 25 and over actually increased significantly
during the study period.
A similar analysis of the series of monthly collisions of 16-year-old drivers
and those age 25 and over in the control jurisdictions of New Brunswick
and Maine failed to reveal any significant changes. Although there was a
significant gradual permanent reduction in the casualty crashes of 16-year-old
drivers in Saskatchewan, a comparable decrease was also noted for drivers
aged 25 and over in that province. Accordingly, the downward trend in casualty
crashes in Saskatchewan for both young and older age groups appear to be
part of a general and unique trend in that jurisdiction.
The monthly number of collisions involving 16-year-old drivers in Nova Scotia
on which the time series was based is shown in Figure
4. It also displays the 12-month moving average to adjust for seasonal
variations. The vertical line shows the date the program was implemented.
Figure
5 shows comparable information on casualty collisions among 16-year-old
drivers and Figure
6 and Figure
7 show the corresponding data for drivers aged 25 and over. As can be
seen in Figure
4 and Figure
5, the total number of collisions and casualty collisions involving
16-year-old drivers dropped dramatically from the pre- to the post-program
periods. This was not the case for the comparison group. The number of drivers
age 25 and over involved in casualty crashes (Figure
7) actually increased over the study period. Although their involvement
in all types of collisions (Figure
6) did decrease slightly in the post-program period, this downward trend
began in 1992 several years before the implementation of graduated licensing.
However, since collisions involving drivers age 25 and over have generally
declined since about 1992, it is possible that the significant reductions
in collisions observed for drivers age 16, following implementation of the
program, may be at least partially accounted for by an overall downward
trend influencing both young and older age groups. Further analyses were
conducted to explore this possibility. The adult collision series was included
in the time series as a covariant to control for factors or events, such
as a pre-existing downward trend, affecting all drivers. The analysis showed
that the collision series for drivers age 25 and over was not a significant
parameter – i.e., it did not account for the significant decrease in crashes
of drivers age 16 following implementation of graduated licensing.
This suggests that the graduated licensing program and not other factors
that affected both groups equally accounted for the reductions in total
collisions among the primary target group.
The magnitude of the impact of the graduated licensing program can be illustrated
further by comparing the actual number of 16-year-old drivers involved in
collisions to the number that would have been expected in the absence of
the program, based on the pattern of collisions in the pre-program period.
Time series analysis was used to forecast these predicted numbers. In the
39 months since the graduated licensing program was implemented, 980 drivers
age 16 were involved in collisions. This compared to a predicted number
of 1,547 collision-involved drivers age 16. Thus, there were 567 fewer crashes
than would have been expected – this represents an overall reduction in
collisions of 36.7%.
The analysis also showed 153 fewer 16-year-old drivers were involved in
casualty collisions than would have been expected if graduated licensing
was not introduced – this represents a 31.1% reduction (i.e., 492 predicted
collisions compared to 339 actual collisions).
In summary, the time series analysis revealed that the introduction of the
graduated licensing program in Nova Scotia was associated with a significant
decrease in the number of collisions involving 16-year-old drivers and the
magnitude of the reductions are comparable to those found using the simple
pre-post comparisons.
The results of the time series analysis increases confidence that the observed
reductions in collisions as well as casualty collisions were accounted for
by the implementation of the graduated licensing program and not by other
age-specific or general trends.
|
|
|
