|
|
 |
Impact of the graduated driver licensing program in Nova Scotia
Report from the Traffic Injury Research Foundation
Method: Control for Other Factors
The validity of the evaluation can be compromised by the impact of other
uncontrolled factors or initiatives. This section considers what they might
be and describes the procedures used to control them.
- Effects of other safety initiatives. The only obvious initiative
introduced in Nova Scotia during the study period that may have influenced
collisions was a change in the reporting definition for a property-damage
collision – it was increased from $500.00 to $1,000.00. This change
took effect in June 1995, nine months after the implementation of the
graduated licensing program. Given that most collisions involve only
property damage, this change raises concerns about using the total number
of collisions as a criterion to measure the effectiveness of the program.
An option would be to exclude property damage collisions and use only
those involving a casualty but this would significantly reduce the number
of cases and drastically reduce the sensitivity of the design. It is
more desirable to retain all collisions as a dependent measure, so it
was necessary to determine if the reporting threshold had any independent
effect on the number of collisions in the database.
The findings are shown in Figure
1, which plots the monthly number of property damage crashes involving
drivers age 16 and over, from 1990 through 1997. The figure also displays
the 12-month moving average to adjust for seasonal variations. The date
of the change in reporting of property damage crashes is shown by the vertical
line. As can be seen, the number of drivers involved in property damage
crashes has been steadily declining in the province since about 1992 and
there is no noticeable departure from this trend following the change in
reporting definition. This visual inspection is supported by the analysis.
Statistical modeling of the series, with the inclusion of an intervention
parameter representing the new definition for a reportable collision, did
not reveal a significant change in the series after June 1995.
However, it is still possible that a change in the reporting threshold had
a differential impact on young, novice drivers that is obscured in the aggregate
data. A differential effect would occur if, for example, young drivers were
more likely than older drivers to report minor collisions involving $500.00
to $1,000.00 in property damage. If this happened, then an observed decrease
in the number of collisions among young drivers over the study period could
be partially, or totally, attributable to the change in reporting threshold
and not the implementation of the graduated licensing program. It is unlikely
that this was the case, for several reasons.
Relative to older drivers, young driver crashes are typically associated
with higher, not lower costs. For example, research by the Highway Loss
Data Institute (1991) showed that cars insured for teenagers had a 90% higher
collision claim frequency than those insured for adults and a 16% higher
average loss payment per collision claim. Research has also shown that young,
novice drivers are reluctant to report property damage crashes in the first
place. For example, in a study of learner and novice drivers in the United
Kingdom, Forsyth et al. (1995) report that only 35% of all their accidents,
most of which would be exclusively property damage, were reported to the
police. By contrast, 90% of injury accidents were reported to the police.
This is not surprising since there is probably an economic disincentive
for young drivers to report minor crashes in the first place – i.e., a significant
increase in their insurance premiums.
It is, therefore, unlikely that the reporting of crashes by young, novice
drivers was differentially influenced by the change in the reporting threshold
from $500.00 to $1,000.00, so total collisions was retained as a dependent
measure. The evaluation also focuses on casualty collisions, which would
be unaffected by the new reporting threshold.
- Controlling for other unknown factors. As mentioned previously,
various control groups were used to reduce threats to the validity of
the evaluation. The internal control group involved drivers age 25 and
over, who would be largely unaffected by the graduated licensing program.
They would have received their initial license before the program and
would already have accumulated several years driving experience.
The external control groups came from three other jurisdictions – New Brunswick,
Saskatchewan and Maine – that did not introduce graduated licensing programs
during the study period. Although New Brunswick did introduce a version
of graduated licensing, it was not implemented until January 1996, 15 months
after the Nova Scotia program came into effect. Saskatchewan also introduced
a Probationary Driver’s program but it was not implemented until August
1996, 21 months after the Nova Scotia program was operational.
The selection of the control jurisdictions was based on proximity and similarities
to Nova Scotia as well as data availability, completeness and quality. In
this regard, New Brunswick is adjacent to Nova Scotia and Maine is a proximate
jurisdiction. And, as shown in Table
1, both these jurisdictions are similar to Nova Scotia in terms of geographic
size, overall population, population per square kilometer, length of the
public road system, number of people per kilometer of roadway, number of
cities with a population of over 50,000, number of persons aged 16, and
the percent of the population aged 16 to 19.
Although, Saskatchewan differs from Nova Scotia, in terms of such things
as geographic size and location, similarities do exist in terms of overall
population and the number of 16 year olds. In addition, any general or age-specific
trends affecting collisions in Canada as a whole should be apparent in Saskatchewan.
- Controlling for exposure. Other factors that could potentially
influence the frequency of collisions are changes in the number of licensed
drivers. This could occur as a direct result of the program itself or
independently of it.
The graduated licensing program itself may discourage some individuals from
becoming licensed, which could result in a decrease in collisions among
the target group. Although this would certainly be viewed as a positive
impact, the benefits would be largely independent of the primary intent
of the program – gaining driving experience under conditions of low risk.
And, there is some evidence in the literature that the introduction of the
program does affect licensing rates. For example, in the evaluation of the
California provisional licensing program, Hagge and Marsh (1988) note that
the "reductions found in the total accident rate (per capita) for 15-17
year olds may be due, at least in part, to a decrease in the percentage
of licensed drivers in that age group" (p. 39). They report that the percent
of licensed drivers age 15-17 decreased substantially following program
implementation but the percent of licensed people over the age of 24 increased
over this period. According to Hagge and Marsh, "the provisional licensing
program itself may have been the mechanism reducing the teen licensing rate,
and thereby accident rate, through postponement of licensure" (p. 71). Similar
conclusions were reached by Langley et al. (1996) regarding the positive
effect on injury collisions of the New Zealand program.
In contrast, McKnight et al. (1983) reported that the implementation of
the provisional licensing program in Maryland did not appear to have any
effect upon the licensing rate of 16-year old drivers. More recently, Ulmer
et al. (1999) in the evaluation of the Florida program, also reported that
"the introduction of graduated licensing has not diminished the rate at
which young teenagers apply for licenses and that the noted declines in
crash involvements were not due to fewer people holding licenses" (p. 8).
Nonetheless, the authors were sensitive to the potential impact of the graduated
licensing program on licensing rates and the confounding effects this can
have on the results.
Changes in the number of licensed drivers could also be a result of changes
in factors unrelated to the program, such as demographic changes in the
population of young people, which could affect the number who become licensed
and, thereby, the number of collisions.
The present investigation was sensitive to these problems and the analyses
take into account changes in population in Nova Scotia and the control sites.
In addition, pre-post comparisons control for changes in the number of novice
drivers in Nova Scotia by using collision rates. Similar licensing data
for calculating rates were not available from the control jurisdictions.
|
|
Table of Contents Acknowledgements
|
|
