Seat belt use and accident involvement: A comparison of driving behavior before and after a seat belt law

Seat belt use and accident involvement: A comparison of driving behavior before and after a seat belt law

Actid. Anal. and Pew. Printed in the U.S.A. Vol. 25, No. 6. pp. 757-763, 1993 OIXI-4575/93 $6.00 + .OO C 1993 Pergamon Press Ltd. BRIEF COMMUNICAT...

640KB Sizes 49 Downloads 115 Views

Actid. Anal. and Pew. Printed in the U.S.A.

Vol. 25, No. 6. pp. 757-763,

1993

OIXI-4575/93 $6.00 + .OO C 1993 Pergamon Press Ltd.

BRIEF COMMUNICATIONS

AND RESEARCH

SEAT BELT USE AND ACCIDENT A COMPARISON

OF DRIVING

NOTES

INVOLVEMENT:

BEHAVIOR

BEFORE

AND AFTER A SEAT BELT LAW J. RICHARD University

STEWART

of North Carolina Highway Safety Research Center, 134 l/2 E. Franklin Chapel Hill, NC 27599-3430, U.S.A. (Received

27 April

1992; in revised form

18 December

Street, CB #3430,

1992)

Abstract-In an earlier study, researchers at the University of North Carolina Highway Safety Research Center found drivers classified as seat belt nonusers on the basis of direct observation and self-reported belt use to be overrepresented in prior accidents and violations. This study represents a follow-up and extension of the earlier study where accident and violation rates over a 2.5-year interval following the classification by seat belt use status are compared. Seat belt nonusers were again found to be overrepresented in both accidents and violations. In other analyses of these data, changes in seat belt use status were found not to be associated with changes in accident or violation rates, and seat belt use rates reported by police in accidents following the mandatory seat belt law greatly exceeded both the observed and self-reported use rates. This was especially pronounced for drivers who responded that they rarely or never used seat belts.

BACKGROUND

belt use laws. Evans and Wasielewski (1983) using data collected in Detroit and Toronto, found nonuse of seat belts to be associated with riskier driving behavior in the form of reduced vehicle headways and with a greater likelihood of prior accidents. Evans (1987) using the Michigan data cited above together with data from the Fatal Accident Reporting System (FARS), estimated accident involvement rates for unbelted drivers to be 53% higher than those for belted drivers. Results from the North Carolina study were presented by Hunter et al. (1993). Among the findings was that drivers who did not use seat belts had higher prior accident and violation involvement rates than those who did. This was found to hold both in terms of observed belt use at the time of the survey, and in terms of the respondents’ selfreported belt use, where higher reported use rates corresponded to lower accident and violation rates. Moreover, these results could not be attributed solely to differences in the driver characteristics of age, gender, and reported annual mileage, though seat belt users and nonusers did, in fact, differ with respect to each of these characteristics. The current study represents a prospective examination of the driver records of these same drivers covering the time period from January 1987 through June 1990. A primary goal in the analysis was to

For a number of years, beginning just prior to implementation of North Carolina’s mandatory seat belt law, researchers from the University of North Carolina Highway Safety Research Center (HSRC) have estimated statewide seat belt use rates from field observations made at 72 locations across the state. During the months of June and July 1987, while making the field observations the same researchers also distributed some 10,000 mailback seat belt use questionnaires. The questionnaires were color-coded to distinguish between drivers who were using seat belts at the time of the survey from those who were not. A total of 5,074 questionnaires were returned, and of these, 4,505 questionnaires could be linked via name and address to a corresponding driver history record. Thus, the seat belt observation, the questionnaire responses, and the driving record over the 1983-1986 time period formed the basis of a study file for investigating relationships between seat belt use and prior crash involvement. In earlier work the idea that automobile drivers most likely to be involved in crashes tend also to be those least likely to comply with seat belt laws was suggested by Robertson and Williams (1978) and later by McCarthy, Taylor, Sanford, and Lang (1984) in their respective evaluations of mandatory 757

Brief Communications

158

Table Reported

1. Reported

belt use

belt use by observed Observed Belted

Never (%) Rarely (%I Sometimes (%I Most of the time (%I Always (%)

belt use

belt use Not belted 223 (94) 336 (93) 350 (83) 496 (52) 327 (13)

13 (6) 24 (7) c::, 463 (48) 2178 (87)

Total 236 360

959 2,505

RESULTS Basics A total of 4,499 of the original 4,505 subjects were located in the 1990 driver history file. Information concerning accidents and violations involving these subjects during the 3.5-year period January 1987-June 1990 was extracted and added to the study file. Interestingly, over the four-year 1983-1986 period, the 4,505 drivers were involved in 1,038 accidents and 1,848 violations, while in the 31/2-year 1987-June 1990 period the 4,499 drivers were involved in 1,068 accidents and 1,869 violations. Thus, average accidents per driver per year increased from .0576 to .0678, and average violations per driver per year increased from .1026 to .1187 from the earlier to the later period. As will be dis-

2. Follow-up

cussed later, these increases seem, for the most part, to be due to increased exposure among drivers in the youngest age category. Since only six subjects were lost from the study group, descriptions of the subjects in terms of demographics, attitudes, etc. are the same as presented in Hunter et al. (1993).

accidents

Overrepresentation In addition to the color coding that indicated the subject’s seat belt status at the time the questionnaires were distributed, the survey instrument also asked questions concerning the subjects’ current seat belt use. Five response options were listed: never, rarely, sometimes, most of the time, always. Table 1 shows a cross-tabulation of reported vs observed seat belt use. In terms of observed belt use, the reported belt use categories Never and Rarely were, virtually identical, and since their sample sizes were relatively small, it seemed quite logical to combine these two categories. It was not clear if any further collapsing was warranted, and, if so, which other categories should be combined. Nonetheless, in order to have cell sizes sufficiently large to permit cross-classification by other variables, in many of the statistical analyses presented by Hunter et al. (1993) the Sometimes and Most-of-the-time categories were also combined to yield a three-level characterization of reported belt use. With respect to investigating accident and violation involvement in the 1987-1990 data, Tables 2 and 3 show percentages of drivers having specified numbers of accidents (violations) from 0 to 3 or more, classified by observed belt use, the five-level reported belt use, a three-level variation of reported belt use based on combining response categories,

for seat belt users Number

Belt use group I. Observed Not belted Belted II. Reported Never Rarely Sometimes Most of the time Always III. Reported Never or Rarely Sometimes or Mostly Always IV. Combined Never, Rarely & Not Belted Alwavs & Belted

Notes

423

examine the extent to which the overrepresentation previously reported, based on prior accidents and violations, persisted over a time period following the seat belt observations and the survey.

Table

and Research

and nonusers

of accidents

0

1

2

3+

AvgiDriver

76.1% 82.3%

19.4% 15.4%

3.6% 2.0%

0.9% 0.3%

,293

76.3% 75.3% 77.1% 78.5% 82.0%

19.5% 18.9% 19.4% 18.4% 15.5%

3.4% 4.7% 2.8% 2.5% 2.2%

0.9% 1.1% 0.7% 0.6% 0.3%

,288

75.7% 78.1% 82.0%

19.1% 18.7% 15.5%

4.2% 2.6% 2.2%

1.0% 0.7% 0.3%’

,309 .258 .209

75.1% 83.0%

19.5% 14.7%

4.3% 2.1%

1.1% 0.3%

,317 ,197

,202

,322 ,272 ,252 ,209

Brief Communications

759

and Research Notes

Table 3. Follow-up violations for seatbelt users and nonusers Number of violations Belt use group I. Observed Not belted Belted II. Reported Never Rarely Sometimes Most of the time Always III. Reported Never or Rarely Sometimes or Mostly Always IV. Combined Never, Rarely & Not Belted Always & Belted

0

I

2

3+

Avg/Driver

68.7% 78.4%

19.1% 15.3%

7.1% 4.0%

5.1% 2.4%

t.553 ,328

61.0% 63.6% 69.0% 75.2% 78.4%

23.7% 21.9% 19.2% 15.5% 15.3%

8.9% 8.1% 7.8% 4.9% 4.1%

6.4% 6.4% 4.0% 4.4% 2.3%

,724 ,664 ,511 ,440 ,325

62.6% 73.3% 78.4%

22.7% 16.6% 15.3%

8.4% 5.8% 4.1%

6.4% 4.3% 2.3%

,688 ,462 ,325

61.5% 79.0%

23.1% 15.3%

8.8% 3.6%

6.6% 2.1%

,710 .310

and, finally, a classification based on both reported and observed belt use. This last classification contains two categories, one consisting of those who said they always use seat belts and who were also observed wearing belts (n = 2178), and the other consisting of those who said they rarely or never use belts and who were, in fact, not belted when observed (n = 559). Also shown is the average number of accidents (violations) per driver for each belt use classification. Continued overrepresentation in accidents and violations by seat belt nonusers is quite clear from the tables. Based on average numbers per driver, observed belt nonusers had 45% higher accident rates and 69% higher violation rates than did the belt users. This compares with 35% and 69%, respectively, from the earlier data. From the combined observed/reported classification, the Never users had accident and violation rates that were 61% and 129%, respectively, higher than the rates of the Always users. Hunter, et al. (1993) further investigated the question of overrepresentation by noting that seat belt users and nonusers differed with respect to several demographic and other factors (e.g. age, gender, and reported annual mileage) and developed statistical models to take into account certain of these factors. Seat belt use or nonuse was found to have a significant effect on accident and violation rates even after adjustments were made for these other factors. Similar analyses were carried out using the follow-up data. As in the earlier work, the variables included in the analyses along with belt use were age (at the time of the 1987 survey), gender, and reported annual miles driven. Five age categories (16-20,21-25,26-35,36-55,56 and over) and eight

mileage categories as provided on the questionnaire (<5,000; 5,000-9,999; 10,000-14,999; 15,000-19,999; 20,000-29,999; 30,000-39,999; 40,000-49,999; 50,000 and over) were used in the analyses. Logistic categorical regression models were fit to the accident and violation rates. Thus, the dependent variables Proportion of drivers one or more Proportion of drivers with no accidents

Proportion of drivers one or more Proportion of drivers with no violations \

/

were fit as linear functions of seat belt use, driver gender, the age categories, and mileage categories. The age and mileage variables were treated in some models as purely categorical variables and in others as continuous variables with values l-5 and 1-8, respectively. Table 4 shows the analysis of variance table and parameter estimates for a model fit to risk of accident involvement. While accident involvement tended to increase significantly with increasing annual mileage, this variable was not statistically significant when included in the model with the other variables age, gender, and belt use. For this reason annual mileage was not included in the above model, nor in any of the models for accident risk. It was, however, a significant factor in models for violation risk.

Brief Communications

760

Table 4. Analysis Analysis Source

of variance

I I 1

1 I6

X’

7.47 75.53 8.50 15.85 20.66

5. Overrepresentation

Seat belt use status

Notes

for accident

involvement

Parameter Prob. .0063 .oooo .0035 .ooo 1 .I919

Based on the estimated coefficients from the model of Table 4, the relative risk an unbelted driver would be involved in a crash was 35.7% higher than that of a belted driver. Other model effects showed males to have a 25.3% higher risk of crash than females, and that relative crash risk decreased by a factor of .744 with each increasing age category. Other similar models using both observed and reported seat belt use were fit to the data on accident involvement and violation involvement. In each instance statistically significant, (p < .Ol), effects were estimated for the seat belt use variables. Overrepresentation in terms of average numbers of accidents and violations per driver was also investigated by fitting linear models to the data. These models, essentially analyses of covariance, provide estimates of average accidents (violations) per driver within belt use categories after making adjustments for differences in the covariates, age, gender, and annual mileage (violations only). Results of these analyses are given in Table 5. It should be noted that the results of Table 5 were taken from six different models. In each case, the belt use factor was statistically significant (p < .002). Comparison of the values of Table 5 with the raw averages of Tables 2 and 3 show that the adjustments tend to shift, slightly, the extreme values toward more central values. The over-representation of the nonbelt users both in terms of accidents and violations, however, remains quite clear.

Table

estimates

of variance

DF

Intercept Age Gender Belt use Lack of fit

and parameter

and Research

of average Adjusted

Effect Intercept Age Gender Belt use

estimates

Estimate

Standard

,455 ,296 ,226 ~ ,306

error

.I66 ,035 .077 ,077

As in the earlier study, it was also of interest to investigate the extent to which certain characteristics of the accidents involving non-belt users and belt users differ. It was found for the 1983-1986 accidents that the nonuser group differed in having a higher proportion of single vehicle accidents, a higher proportion of rollover accidents, and a higher proportion of accidents in which the driver was charged with a violation. The 1987-1990 accidents of seat belt users and nonusers did not differ significantly with respect to any of these variables. In addition to these variables, no significant differences were likewise found with respect to the variables-accident severity, region of impact, vehicle deformation, or vehicle drivability. Significant differences were found, however, with respect to light condition, accident speed, and driver injury. The distributions of these variables by observed and reported belt use are shown in Table 6. RELATIONSHIPS IN SEAT BELT CHANGES IN VIOLATION

BETWEEN CHANGES USE STATUS AND ACCIDENT AND INVOLVEMENT

A general discussion of the concept of risk compensation in the field of traffic safety is presented by Evans (1991). An opportunity to investigate the question of risk compensation associated with North Carolina’s mandatory seat belt law was made possi-

accidents accidents/driver

and violations

per driver

Adjusted

violations/driver

Observed Belted Not belted

,208 ,282

.353 ,506

Reported Never or Rarely Sometimes or Mostly Always

,288 ,258 ,213

.576 ,450 ,352

Combined Always Never,

,202 ,295

,338 .592

and Belted Rarely, and Not belted.

model

Brief Communications

Table

6. Differences

in accident

characteristics

and Research

between

Observed belt use (%) Characteristics 1. Light cond. Daylight Not daylight 2. Driver injury O-None C-Minor B-Moderate A or K-Serious 3. Accident speed O-29 30-49 50 and over

or fatal (mph)

Notes

seat belt users

Reported

values

Never*

73.6 26.4

79.6 20.4

70.7 29.4

76.2 23.8

7’9.2 20.8

74.9 12.8 7.7 4.7

76.6 15.4 5.3 2.7

76.5 10.9 7.1 5.5

75.6 11.8 9.0 3.6

75.6 16.9 4.5 3.1

35.3 50.6 14.1

45.4 43.8 10.8

30.6 53.0 16.4

43.7 44.3 12.0

42.0 46.9 11.1

of ratios of proportions 1987-1990 compared P(acc. P (act.

belt use (%)

Yes

ble by the fact that the survey questionnaire contained questions concerning seat belt use both before and following the implementation of seat belt law. Using the responses to these questions, three driver groups were identified: one group consisted of those who responded that they never or rarely used seat belts both before the law went into effect and currently and who were observed not wearing seat belts; a second group who responded that they always used seat belts both before the law and currently and who were observed wearing seat belts; and a third group who responded that before the law they rarely or never used seat belts and that currently they always used seat belts and who were observed wearing seat belts. These groups, referred to as no/no, yes/yes, and no/yes, contained, respectively, 518, 605, and 566 drivers. Changes in driving behavior were characterized in terms of differences in the proportions of drivers having accidents and violations in 1987-1990 as compared to 1983-1986, and in changes in mean

7. Predicted

and nonusers

No

Sometimest

* Never indicates the combined Never and Rarely groups. I’ Sometimes indicates the combined Sometimes and Most-of-the-Time

Table

761

in 1987-1990) in 1983-1986)

Always

groups.

numbers of accidents and violations per driver between these two time periods. Driver age, gender, and average annual mileage (from 1987 survey) were again considered as potential covariates. Preliminary analyses showed that among these covariates, the only significant relationship with changing driving behavior was that increased (1983-1986 to 1987-1990) accident and violation rates were found for the youngest (aged 16-20) drivers versus all other age groups. No significant effects were found between other age groups, nor for driver gender, nor by reported annual mileage. To test for significant changes in driving behavior associated with changes in seat belt usage, loglinear categorical models were fit to the ratios of proportions of drivers having accidents (violations) in 1987-1990 relative to proportions having accidents (violations) in 1983-1986. Independent variables in the models were driver age (16-20 vs. older), and the three seat belt before/after categories. Predicted values of these ratios and their standard errors

of drivers having to 1983-1986

accidents

and

P (viol. in 1987-1990) P (viol. in 1983-1986)

violations

in

Age

Belt use

16-20

No/no No/yes Yes/yes

1.58 1.68 1.52

(.3l) (.38) (.33)

1.58 1.65 1.63

(24) (.29) t.29)

21+

No/no No/yes Yes/yes

0.98 1.00 0.94

(.ll) (.l3) (.ll)

0.94 0.97 0.96

(.07) (.09) (.lO)

All

No/no No/yes Yes/yes

1.02 1.05 .98

(.l2) (. 14) (.l3)

.98 1.02 1.01

(.08) (.lO) (.ll)

(s.e.)

(s.e.)

762

Brief Communications

and Research Notes

Table 8. Seat belt use reported in accidents Reported in accidents 1983-1986 Survey data 1987

1987-1990

Belted

Not belted

Belted

238 (43.4) 129 (28.6)

310 (56.6) 322 (71.4)

519 (96.3) 419 (85.7)

242 (47.5)

267 (52.5) 240 (73.2) 123 (78.8)

487 (96.8) 306 (88.7) 140 (80.0)

229 (53.5) 119 (78.3)

403 (97.6) 135 (80.4)

Observed use Belted W) Not belted (%) Reported use Always use (%) Most of the time, Sometimes (W Rarely, Never (%) Observed/reported Belted/always (%) Not belted/Rarely or never (%)

(iE.8) (G.2) 199 (46.5) (::.7,

are shown in Table 7. The effect of the young age group is quite clear from the predicted values, which indicate that this group has substantially higher accident and violation rates in the 1987-1990 period. This would seem to be a reflection of this group’s increased driving exposure in the second time interval. While the ratios for the No/yes belt use group are slightly higher than for the other belt use groups, the overall effects are quite small and not statistically significant (p = .856 for accident ratios and p = .930 for violation ratios). Similar results were obtained from analyses of changes in average numbers of accidents (violations) from the first time interval to the second. Specifically, in analysis of variance models for changes in average numbers of accidents and violations, the significance levels of the estimated effects due to seat belt (before law/after law) use groups were p = .557 and p = .808, respectively. Thus, we find no evidence that the drivers who became seat belt users as a result of the seat belt law had significantly increased accident or violation rates relative to drivers who continued to use or not use seat belts. OBSERVEDANDREPORTEDSEAT BELT USE VERSUS POLICE-REPORTED SEAT BELT USE IN ACCIDENTS A final item of interest was to see how policereported belt use in accidents differed between the 1983-1986 accidents and the 1987-1990 accidents for various user groups based on the 1987 questionnaire and observational data. Table 8 shows frequencies and percentages of drivers reported as belted

Not belted 20 (3.7) (Z.3) 16 (3.2) (E.3) (35 (20.0) 10 (2.4) $6)

or not belted in accidents by time period and by belt use status based on survey responses and observations. Of particular interest is the last line of the table, which pertains to drivers who in the summer of 1987 were observed not using seat belts and who stated that they rarely or never did so. In their prelaw (1983-1986) accidents, 21.7% of these drivers were reported as belted. In their postlaw (1987-1990) crashes, however, over 80% were reported as belted. These findings lend further support to the general feeling that accident reported restraint use can no longer be considered a valid indicator of true restraint use in crashes. SUMMARY Drivers classsified as seat belt nonusers based upon a 1987 survey and field observations, were found to be overrepresented in accidents and violations during the 3 l/2-year period of January 1987-June 1990. The extent of this overrepresentation was very consistent with the overrepresentation of these same nonusers over the four-year period of January 1983-December 1986, as reported by Hunter et al. (1993). In neither of these periods could the overrepresentation of seat belt nonusers in accidents and violations be attributed solely to differences in driver age, gender, or average annual mileage driven. In each of the two time periods (1983-1986, 1987-1990), the accidents of the seat belt nonusers differed in certain characteristics from the accidents involving seat belt users. While the two sets of differentiating characteristics did not overlap, both tended

Brief Communications

to indicate more serious accidents for the nonuser group (i.e. more single vehicle and rollover accidents in 1983-1986, more high speed accidents and more accidents in which the driver was seriously injured or killed in 1987-1990). Using self-reported information concerning seat belt use prior to the implementation of the North Carolina seat belt law and seat belt use during the summer of 1987, analyses were carried out to see if the accident and violation rates of drivers who changed from nonusers prior to the law to belt users after the law, increased more than the comparable rates of drivers who remained nonusers or who were already seat belt users. No statistically significant differences in rate changes were found. Finally, the rates at which drivers were reported by the police as being belted in accidents during the two time periods were examined for different seat belt user categories. The police-reported belt use rate in accidents for the most unlikely user group, increased

from 22% before the law to over 80% after the law.

and Research Notes

763

REFERENCES Evans, L.; Wasielewski, P. Risky driving related to driver and vehicle characteristics. Accid. Anal. Prev. 15:121-136; 1983. Evans, L. Belted and unbelted driver accident involvement rates compared. J. Safety Res. 18:57&t; 1987. Evans, L. Traffic safety and the driver. New York: Van Nostrand Reinhold; 1991. Hunter, W. W., Stewart, J. R., Stutts, J. C., and Rodgman, E. A. Observed and self-reported seat belt wearing as related to prior traffic accidents and convictions. Accid. Anal. Prev. 25~545-554; 1993 McCarthy, R. L. ; Taylor, R. K.; Sanford, S. B. ; Lange, R. C. Seat belts: Effectiveness of mandatory use requirements. In: Advances in belt restraint systems: Design, performance and usage. Warrendale, PA: Society of Automotive Engineers, Inc.; 1984: pp. 161-171. Robertson, L. S.; Williams, A. F. Some international comparisons of the effects of motor vehicle seat belt use and child restraint laws. (Paper presented at the Child Passenger Safety Conference, Nashville, Tennessee, May 1978.) Washington, DC: Insurance Institute for Highway Safety; 1978.