The general belief is that dark-colored vehicles such as black or gray are poorly visible on open roads. This is assumed to increase the risk of darker vehicles being involved in a collision, whereas the opposite is believed to hold for vehicles of more conspicuous lighter colors such as white or yellow. 1–4 However, surprisingly few studies have tested the effect of vehicle color on the risk of collision. 5
Assuming that the effect of vehicle color on the risk of being involved in a collision is dependent only on its relative visibility, dark-colored vehicles might be expected to be overrepresented in a sample of vehicles passively involved in collisions. In other words, dark vehicles may be more likely to be hit simply because the driver of another vehicle did not see them. Since 1993 the Spanish register of traffic crashes (Dirección General de Tráfico [DGT]) has collected information about the color of all vehicles involved in traffic crashes with victims. We designed the present study to estimate the specific effect of vehicle color (conspicuity) on the risk of being passively involved on a collision between vehicles in Spain from 1993 to 1999.
Materials and Methods
The method used in the present study is an extension of that proposed by Perneger and Smith 6 in 1991. These authors studied collisions between two vehicles in which only one of the drivers had committed an infraction; this driver was considered to be the actively involved driver. The other driver was considered to have been passively involved in the collision. Analysis of the pairs of actively and passively involved drivers resembled a paired case-control study.
We too considered collisions between two or more vehicles in which only one of the involved drivers had committed any infraction. In accordance with the hypothesis raised in the Introduction, the effect of color upon conspicuity may specifically affect those vehicles passively involved in the collision. The passively involved vehicles were considered cases in the present analysis, whereas actively involved vehicles were considered the control group. Therefore the approach to data analysis is analogous to that of a retrospective, paired case-control study with a variable number of cases per control.
The data were obtained from the Spanish register of traffic crashes recorded by the DGT for crashes that resulted in personal injury or death between 1993 and 1999. We considered only the 57,472 collisions between two or more vehicles (with four or more wheels) in which only one of the drivers involved had committed any traffic infraction recorded in the DGT database (classified according to 20 categories, excluding administrative and speed infractions, which are recorded in two separate variables; see Appendix) and for which information about all the study variables was available. The 57,472 ticketed (actively involved) drivers constituted the control group; the other 66,154 nonticketed (passively involved) drivers were considered the corresponding case group, paired by collision.
We recorded the color of each vehicle as originally encoded in the DGT register. We also recorded the following driver-related variables: sex, age, psychophysical circumstances (normal, under the influence of alcohol when no breath test was used, under the influence of alcohol as documented by a positive breath test, under the influence of other drugs, sudden illness, sleepiness or drowsiness, worried or other), administrative infractions (none, expired driving license, expired motor vehicle inspection certificate or other), speed-related infractions (none, inappropriate speed for the road or weather conditions, driving above the speed limit, or slow driving that interfered with traffic), physical disabilities (none, vision, hearing, upper limbs, lower limbs or other), years in possession of driving license, type of driver (professional or nonprofessional), and safety belt use.
As vehicle-related variables we recorded type (car, ambulance, van, truck, tanker, bus, articulated vehicle or other) and years since the vehicle was first registered for driving on public roads. The environmental variables were: type of road where the accident occurred (open road or urban area), visibility conditions dependent on environmental light (daytime, twilight or night), and weather conditions (good weather, light fog, thick fog, light rain, heavy rain, hail, snow, strong winds or other).
The data were subjected to conditional logistic regression analysis, 7 using ticketed or nonticketed driver as the dependent variable, to estimate crude odds ratio (OR) for each category of color. Considering passively involved vehicles as cases, and assuming that the frequency of colors among vehicles that were actively involved was representative of all vehicles on the road at the time of the accident, an odds ratio higher than 1.0 for any given color is associated with a higher risk for vehicles of that color being passively involved in the collision. However, as certain driver and vehicle characteristics may be associated with the color of the vehicle, 8 we included all driver- and vehicle-dependent conditions in the model to obtain adjusted odds ratio (aOR) estimates for the effect of color, while controlling for the confounding effect of the other variables. Adjusted OR estimates corroborated this assumption and yielded figures quite different from those obtained in the crude analysis, justifying the need for adjustment. These aOR estimates were stratified according to the environmental variables collected.
Results
White was the most common color (40%), followed by red (17%) and gray (16%) (Table 1). Compared with white, all colors except yellow showed aOR estimates higher than 1.0, with black yielding the highest value (aOR = 1.05; 95% CI = 0.99–1.12). When white was compared with the remaining colors (used as the reference), a protective estimate was obtained (aOR = 0.97; CI = 0.94–1.00). The results were similar for light colors (white plus yellow) compared with all remaining colors (aOR = 0.96; CI = 0.94–0.99).
;)
TABLE 1: Association of Vehicle Color with the Risk of Being Passively Involved in a Collision Between Two or More Vehicles
Table 2 shows the effect of light colors compared with all remaining colors, stratified by environmental factors. The protective effect of light colors was observed for open roads but not in urban areas. Regarding weather conditions, the protective effect of light colors was stronger under conditions other than good weather (aOR = 0.91; CI = 0.86–0.99) than under good weather conditions. The aOR was lower than 1.0 (aOR = 0.93; CI = 0.90–0.96) under daylight conditions, whereas values higher than 1.0 were observed both in twilight and at night.
TABLE 2: Association of Light Colors with the Risk of Being Passively Involved in a Collision Between Two or More Vehicles, Stratified According to Type of Road, Weather Conditions, and Light Conditions
The results of stratification for white compared with other colors (not shown) were similar to those for light colors (white plus yellow). The remaining individual colors generally showed the opposite pattern to that described for light colors (not shown); their aORs were higher than 1.0 in open roads and daylight conditions. Although the estimates for both good and other weather conditions were higher than 1.0, those for other weather conditions were generally higher.
Discussion
In general our results were as expected: white and light colors were slightly less frequent among vehicles passively involved in collisions than in the group of actively involved vehicles. This finding supports the hypothesis that light colors (white and yellow) are associated with better vehicle conspicuity 2,9 and hence with a slightly lower risk of being involved in collisions. 9 The opposite effect was seen for darker colors such as gray, red, black and blue.
We are aware of no studies that have specifically investigated the influence of vehicle color on the risk of being involved in collisions. 5 Most studies of passive involvement have focused on the visibility of pedestrians, motorbikes and bicycles. 3,10,11 Our findings support previous recommendations to use yellow or white colors for certain vehicles such as schoolbuses 12 and patrol, 13 emergency and firefighting vehicles. 9
The influence of color on the risk of being passively involved in a collision varied depending on environmental conditions. These changes are logical, as the effect of color upon conspicuity is not the same under different visibility conditions. In cities, for example, low visibility was influenced by factors other than vehicle color. The influence of color also varies with visibility conditions during various times of day or night; for example, at twilight and at night, drivers turn on the headlights, reducing the effect of color on visibility. The effect of color on the risk of being involved in a collision was largest on open roads and during daytime driving. The protective effect of white and yellow was greatest under weather conditions associated with worsening visibility.
The main drawbacks of the present study are related with the study design and its specific application to test our hypothesis. The main problem with the former lies in the validity of the classification of drivers as actively or passively involved in the collision depending on whether they committed any infraction. This point is addressed in detail in previous papers. 6,14
Some comments are also in order regarding the validity of the hypothesis that underpins the study. Because the specific effect of color depends only on the conspicuity of each color, color may be influential only in vehicles passively involved in the collision. However, this assumption is probably not entirely accurate, as certain driver-related characteristics associated with a higher risk of being actively involved in a collision are also presumably related with certain vehicle colors. 8,15 In fact, the association between color and driver characteristics was manifested indirectly in our study by the different estimates of the effect of color obtained in the crude (not shown) and adjusted analyses. It is therefore essential to adjust the OR estimates for the effect of color considering all potentially driver- or vehicle-dependent confounding factors. Furthermore, because the adjusting variables we considered do not include all possible factors related with driving skills or risk-taking behaviors, residual confounding by these factors cannot entirely be ruled out. Moreover, the information recorded in the DGT register did not specify all the available body colors of vehicles, which makes it impossible to provide a more detailed analysis of the effect of color.
In conclusion, our data suggest that light vehicle color is associated with a slightly lower risk of being passively involved in serious collisions in Spain. Although the effect of vehicle color is small and undoubtedly other vehicle-dependent factors deserve much more attention in efforts to decrease the frequency of traffic crashes, the influence of color may deserve consideration by traffic authorities, car manufacturers and the public in general.
We thank the Dirección General de Tráfico of Spain for providing the raw data from their traffic accident database, and Karen Shashok for translating parts of the manuscript into English.
Appendix
TABLE
Table: Appendix. Distribution of Drivers According to Traffic Infractions Recorded by the Spanish Dirección General de Tráfico
References
1. Alvarez González FJ. Seguridad y medicina de tráfico. Barcelona, Masson, SA:; 1997.
2. Academia de Tráfico de la Guardia Civil. Investigación de accidentes de tráfico. Madrid: Ministerio del Interior, Dirección General de Tráfico, 1991.
3. Hills BL. Vision, visibility, and perception in driving. Perception 1980; 9: 183–216.
4. Allen MJ. Automobile visibility problems. J Am Optom Assoc 1965; 36: 807–810.
5. AAA Foundation for Traffic Safety. Car Color and Safety. White paper. AAA Foundation for Traffic Safety, 2000. Available at:
http://www.aaafoundation.org/resources/index. cfm?button=carcolor. Accessed December 11, 2001.
6. Perneger T, Smith GS. The driver’s role in fatal two-car crashes: a paired “case-control study.” Am J Epidemiol 1991; 134: 1138–1145.
7. Hosmer DW, Lemeshow S. Applied Logistic Regression. New York: John Wiley and Sons; 1989.
8. Salter CA, Salter CD. Automobile color as a predictor of driving behavior. Percept Mot Skills 1982; 55: 383–386.
9. Solomon SS. Lime yellow color as related to reduction of serious fire apparatus accidents: the case for visibility in emergency vehicle accident avoidance. J Am Optom Assoc 1990; 61: 827–831.
10. Olsen P. Problemas de visibilidad en la conducción nocturna. Carreteras 1991; 52: 6–27.
11. Harrell WA. Effects of pedestrians’ visibility and signs on motorists’ yielding. Percept Mot Skills 1994; 78: 555–562.
12. National Highway Traffic Safety Administration.
Pupil Transportation Safety. Uniform Guidelines for State Highway Safety Programs. Highway Safety Program Guideline no. 17. Washington, DC: National Highway Traffic Safety Administration, 2001. Available at:
http://www.nhtsa.dot.gov/nhtsa/whatsup/tea21/GrantMan/HTML/05h_ProgGuidlines. html#17. Accessed October 4, 2001.
13. Advisory Committee on Highway Safety of the Interna-tional Association of Chiefs of Police (IACP).
The Highway Safety Desk Book. Available at:
http://www.nhtsa.dot. gov/ people/injury/enforce/DESKBK.html. Accessed October 2, 2001.
14. Lardelli P, Luna JD, Jiménez JJ, Bueno A, García M, Gálvez R. Age and sex differences in the risk of causing vehicle collisions in Spain, 1990–1999.
Accid Anal Prev (in press).
15. Newman MC, Willis FN. Bright cars and speeding tickets. J Appl Soc Psychol 1993; 23: 79–83.
