The crash risk of disqualified/suspended and other unlicensed drivers

Unlicensed driving remains a serious problem in many countries, despite ongoing\ud improvements in traffic law enforcement practices and technology. In the USA, over 10%\ud of the drivers involved in fatal crashes do not hold a valid licence, while approximately\ud 20% of all fatal crashes involve at least one of these drivers (1, 2). In Australia, unlicensed\ud drivers represent over 5% of the drivers involved in fatal crashes. The crashes involving\ud unlicensed drivers and riders account for almost 10% of the national road toll (3).\ud Unlicensed driving represents a major problem for road safety in two respects. Firstly, it\ud undermines the effectiveness of driver licensing systems by preventing the allocation of\ud demerit points and reducing the impact of licence loss, which has otherwise been\ud demonstrated to be a very effective deterrent to illegal behaviour (4, 5). Secondly, there is\ud a growing body of evidence linking unlicensed driving, particularly disqualified driving, to a\ud cluster of high-risk behaviours including drink driving, speeding, failure to wear seat belts\ud and motorcycle use (1, 6, 7). Consistent with this, the crashes involving unlicensed drivers\ud tend to be more severe than those involving licensed drivers, resulting in higher rates of\ud fatality and serious injury (7).\ud While the above data suggest that unlicensed drivers engage in more risky behaviour than\ud licensed drivers, it does not necessarily confirm that they have a higher crash risk. This is\ud because the crash data does not take into account possible differences in the exposure\ud patterns of unlicensed drivers. Indeed, there is a common assumption in the literature that\ud unlicensed drivers drive in a more cautious manner, or at least restrict their driving, to\ud avoid detection. For example, Hurst (8) proposed the existence of a disqualified driver\ud effect, whereby disqualified and other unlicensed drivers are rewarded for driving safely\ud and inconspicuously because it reduces the threat of detection. Similarly, Scopatz et al (2)\ud suggest that the lower rates of drink driving reoffences among drivers who fail to have their\ud licence reinstated (compared with those who do become reinstated) may not necessarily\ud be due to these people giving up driving altogether. Rather, it may be the product of\ud continued driving which is less frequent and more cautious. However, others have\ud questioned the assumption that unlicensed drivers drive in a more cautious manner.\ud Warren (9) argued that the behaviour of unlicensed drivers is primarily motivated by the\ud desire to avoid apprehension, rather than to drive safely.\ud In an attempt to clarify the crash risks associated with unlicensed driving, DeYoung, Peck\ud and Helander (10) examined the fatal crash involvement of different drivers in California.\ud Given the difficulty in obtaining accurate exposure data, they used a quasi-induced\ud exposure procedure to estimate the exposure and subsequent fatal crash rates of licensed\ud and unlicensed drivers. This procedure involves dividing the proportion of at-fault drivers\ud in a particular group by the proportion of innocent drivers to calculate an estimated crash\ud rate. It is based on the assumption that the proportion of crash-involved innocent drivers\ud should be indicative of their overall representation in the driving population. Based on the\ud method, DeYoung et al (10) estimated that suspended/revoked drivers and other unlicensed drivers were over-involved in fatal crashes by a factor of 3.7:1 and 4.9:1,\ud respectively, compared to licensed drivers.\ud The objectives of the current study were two-fold. The first was to replicate the quasiinduced\ud exposure method used by DeYoung et al (10) to estimate the crash risks\ud associated with unlicensed driving in the Australian state of Queensland. The second\ud objective was to confirm whether the crashes involving unlicensed drivers tend to be more\ud severe than those involving licensed drivers, after controlling for driver-related\ud characteristics such as age, gender and vehicle driven