The role of fines and rewards in the self-regulation of young drivers

One of the central goals of traffic safety interventions is the modification of driver behavior [9, 26]. This requires several related tasks: understanding the motivational factors leading to risky driving behavior, measuring driving behavior in a reliable and valid manner, developing interventions, and evaluating the effectiveness of these interventions. The current paper reports on an exploratory study into the first two steps, namely, understanding motivations and measuring behaviors, in order to promote and support the latter steps.

Studies of driving behavior usually rely on self-reported assessments, but the validity of this method has been questioned [6, 28], in particular with regard to speeding [5]. In order to assess the validity of self-reported assessments, we obtained both objective and subjective data on driving behaviors, including violations. In addition, traffic safety interventions usually seek to change behavior by relying on the deterrence paradigm [21, 23]. We complemented this approach by exploring the potential role of rewards in motivating behavioral change. Finally, we explored the relationship between objective measures of driver behavior and individual-level attitudinal variables. The current study examined driving behavior among 114 young drivers in Israel, focusing on speeding violations (hereafter speeding). Numerous studies conducted in different countries found that young drivers suffer from a disproportionate risk of crashing [15, 32]. In Israel, a recent report by the Israeli National Road Safety Authority (RSA) found that during 2015, 71% of young drivers (aged 24 and below) who were involved in a fatal accident committed a traffic violation, compared to 58% of drivers aged 25–44 and 54% of those aged 45–64 years old [12]. A common violation exacerbating both the risk of crashing and the potential of its severity is speeding [26]. In Israel, speeding has been identified as the major cause of fatal accidents amongst 28% of young drivers involved in an accident, compared with only 13% amongst drivers aged 25–64 [12].

The results of the current study show a significant correlation between self-reported assessments of speeding and objective measures, suggesting that self-reports can be used to assess trends in driving behaviors. We also found that the magnitude of rewards necessary to motivate self-regulation was significantly smaller than fines. We discuss the implications for future research and policymaking below.

Much of the information on driving behavior is derived from official documentations of traffic violations, mandated for specific population groups rather than individuals. However, this method fails to account for unreported violations. Alternatively, unrecorded violations may be found in other studies that relied on self-reported behavior. This subjective measure, however, may be biased due to cultural norms or experimenter effect (expectancy bias) [17]. Previous studies have found significant differences between self-reported and observed driving behavior [10]. The current study examined both subjective and objective aspects of driving behavior, using a smartphone application especially designed for the study.

Speeding is common and persistent. For example, over the last decades speeding has increased in the US while other risky behaviors, such as driving without a seatbelt, decreased significantly ([26], pp. 273–4). In 2013, just over 10% of reported traffic violations in Israel was due to speeding [3]. This, most likely, is a significant underestimation. Results from the 2010 survey carried out as part of the SARTRE project (Social Attitudes to Road Traffic Risk in Europe), in which speeding was defined as driving at 20 km/h over the limit, suggest that 24% of Israeli driver claim to enjoy driving above the speed limit. Twenty percent believe that speeding does not increase the risk of being involved in an accident, 36% agree that most of their friends are likely to engage in speeding in a residential area, and 63% believe that car drivers speed very often or always on motorways [24]. As a rule, males tend to speed more than females [25, 33], and some studies have found a significant positive correlation between income and speeding [30].

Numerous studies have argued that young males are more likely to speed compared to other groups, citing reasons as peer pressure, lack of driving experience, a greater tendency for thrill-seeking, or inappropriate parental modelling [2, 4, 18, 27, 29]. Excluding inexperience, all these rationals share the same assumption that young drivers expect some benefits from engaging in risky driving behaviors. Accordingly, interventions seeking to modify such behaviors have usually relied on the deterrence paradigm, suggesting that people fear sanctions and change their behaviors in order to avoid them [21, 23]. The question, according to this formulation, is how to change the equation so that the perceived risks of unsafe driving (e.g. injuries or arrests) will outweigh its potential benefits. Attempts to reduce the potential benefits of risky driving include fines of varying magnitudes, revoking the driver’s license, and attempts to modify the social acceptability of such behaviors among the young driver’s peer group [34]. However, the potential of rewards as tools for modifying driving behaviors, including speeding, remains understudied.

The potential benefits of using rewards are clear when considering how individuals decide to engage in speeding. As a rule, a full cost-benefit analysis rarely precedes individual decision making, especially given the immediacy in which decisions are made while driving [31]. Rather, individuals make use of cognitive aids and shortcuts such as heuristics. Studies have shown that particular constellations of rewards can influence decision making beyond their objective value in a cost-benefit analysis [13]. Accordingly, the use of rewards may be particularly beneficial to motivate behavioral change.

The study made use of a unique smartphone application named “Get Location”, which monitors the driver’s location and speed (see Fig. 1). We asked the drivers to turn on the application whenever they began driving and turn it off when arriving at the destination. We then superimposed data from Get Location on a GIS map of Israel, which includes data on the legal speed on all road segments in Israel. This allowed us to identify when and where had a driver exceeded the speed limit. We defined a speeding event as passing the speed limit by at least 20 km/h for up to 10 s. Thus, if a certain driver drove over the speed limit for 27 consecutive seconds, Get Location recorded three speeding events in total.

With relation to the frequency of driving we found a gap between the self-reported average number of trips and the data recorded using Get Location (4.2 vs 3.7). This might suggest that Get Location was not used to record all trips made, but rather approximately 88% of them. It could also indicate students’ tendency to underestimate how often they drive.

We collected subjective data on driving behaviors using an on-line survey. The survey questions covered driving patterns, attitudes about speeding in various contexts, risk perceptions accompanying each context, risk-taking during driving (measured on a scale of 1 to 7, with 1 being “no chance” and 7 “a high chance”), the frequency of near-accident events, and the perceived effectiveness of rewards versus fines. In addition, we collected data on socio-demographic characteristics to assess their association with self-reported driving patterns. We allocated each participant a unique identifier, allowing us to track his or her performance throughout the research.

We conducted a factor analysis, in which 40 attitudinal statements were subjected to principal component analysis with Varimax rotation, and used the results to identify several attitudinal variables. We then examined the influence of these variables and demographic variables on speeding frequency using a logistic regression model.

Table 2 describes the frequency of prominent self-reported traffic violations among the participants over the last 3 years. Speeding, defined as driving at a speed of 20 km/h or more over the limit, is the most common type of traffic violation: about a third (35.2%) of the participants never or rarely drove over the speed limit, and a little less than a third (31.5%) drove over the limit very often. In comparison, other violations were significantly rarer: under 2% reported that they “very often” do not yield the right of way, ignore a stop sign, or run a red light; less than 12% reported using a cellular phone “very often”. This corresponds to the tendencies in Israeli society reported above [24].

Table 4 presents the results of a T-test comparing between tendencies of frequent and non-frequent speeders responding favorably to certain statements (measured on a scale of 1–7). Frequent speeders were significantly less likely to believe that speeding at 20 km/h or 30 km/h over the limit contributes to a higher traffic accident risk factor. We also used a chi-square test to compare the distribution of perceived risk of being apprehended as a factor of speeding frequency (Fig. 2), however significant differences were not found (χ2 = 6.074, df = 4, p = 0.194).

Statement	High frequency (n = 54) (avg. ± s.d.)	Low frequency (n = 55) (avg. ± s.d.)	P value
I believe that speeding at 20 km/h over the limit in an interurban road contributes to the risk of being involved in a traffic accident	2.87 ± 1.78	3.60 ± 1.78	0.035
I believe that speeding at 30 km/h over the limit in an interurban road contributes to the risk of being involved in a traffic accident	3.61 ± 1.93	4.64 ± 1.77	0.005

Next, we examined the association between self-reporting on speeding in the survey and the objective measures acquired using Get Location, calculated for the sample as a whole. Using a Pearson Correlation Test, we found a statistically significant correlation between the reported tendency to engage in violations and the actual number of violations recorded (r = 0.294, p = 0.027). In other words, drivers who committed more speeding violations were also more likely to report more violations. We also found a positive and significant correlation between the magnitude of the sum required for a fine to deter drivers from speeding and the distance driven while speeding (r = 0.415; p = 0.001). In other words, drivers who cover more distance while speeding are more likely to require a higher fine to deter them from speeding. However, we did not find the same correlation between the distance driven while speeding and the magnitude of the reward necessary to deter one from speeding. We found no significant correlations between self-reported frequency of speeding and the distance covered while speeding or between the duration of speeding and the perceived level of risk.

We also found a positive correlation between speeding and the size of the fine required to modify a behavioral change and the more speeding the higher the fine needed. In other words, the more a participant speeds, the higher the fine needs to be to change their travel behavior.

The current study used an innovative method to obtain objective measures for driving violations at the individual level by using a self-application that monitored the participant’s driving patterns in situ. The objective measures were coupled with a subjective self-reported assessment of the frequency of violations. In addition, the study examined the relationship between attitudinal and demographic variables and the frequency of self-reported speeding, and explored the potential use of rewards to motivate behavioral change that will reduce such driving behavior. Finally, we explored the relationship between objective driving behavior and individual-level data on attitudes and demographic characteristics.

We found a significant correlation between self-reporting on the tendency to engage in speeding and the objective measures acquired using Get Location. This suggests that studies can rely on self-reported data to identify trends in driver behaviors, including violations. However, given the well-established tendency of drivers to overestimate their driving capabilities [1, 7, 20], data derived from such self-reporting should be used to evaluate driving trends but not the actual magnitude of these phenomena. At the same time the ease at which participants used the application suggests that existing technological platforms have already created self-monitoring services that are in wide use. Thus, research the same type of methodology could be used to better understand driving behavior at the individual level.

The results of the estimation model demonstrate that none of the demographic variables proved to be efficient in trying to distinguish between frequent and non-frequent speeders. This includes gender, which is in contrast to previous studies [14, 22, 33]. However, this finding should be viewed with caution given the low proportion of females in the sample (19%). We did not examine age due to homogeneity of the sample in that dimension. In contrast to previous studies [30], our results show a negative relationship between income and speeding. This result may derive from the fact that the participants were students, whose income distribution is different from those of other population groups.

Attitudinal and interpersonal variables proved efficient in distinguishing between frequent and infrequent speeders. Drivers who viewed speeding as less risky, were more likely to speed, [11, 25]. In addition, friends’ driving behavior, as perceived by the driver, was shown to have a significant and positive correlation with the risk of being a frequent speeder, and so did risk perception in general; in other words, people who believe that their friends tend to speed, and people who tend to take more risks in general, were more likely to become frequent speeders. We found that a higher level of self-control was correlated with speeding, perhaps demonstrating the tendency to overestimate one’s driving capabilities as discussed above. The relationship between attitudinal and interpersonal variables is of significance when viewing it from the complementary qualities of the objective and subjective measures and tools used. Self-monitoring application on personal devices use a similar model and in away normalize the corresponding behaviors of self-monitoring and self-reporting. The strong reliance on interpersonal considerations suggests the possible advantage of using data gathering techniques that already exist in society at large.

Finally, we found evidence that the use of rewards can influence driving behaviors, including speeding. This is consistent with several previous studies that exhibited their usefulness in reducing speeding tendencies [8, 19]. To stress the matter further, this approach follows the model for human behavior used by the deterrence paradigm [21, 23], but utilizes it in order to underlie the potential benefits of using rewards. The re-organization of insurance schemes can prove to be a fruitful way to achieve this goal, for example by reducing the premiums for safe drivers. Furthermore, this study presents a different model, one that relies on self-monitoring, self-reporting and the behavioral response to rewards and fines. These platforms already exist in various forms throughout the Hi-tech industries, suggesting that future studies could pursue collaboration with existing companies that develop and rely on technologies that gather big data on personal behaviors. This would consequently lend itself to incorporating reward/fine principles that rely on scientific research that could help people modify their driving behavior. The introduction of in-vehicle monitoring devices would also promote such goals. However, the former would circumvent the low social acceptability regarding mandatory monitoring technologies [16].

The study had three main limitations. First, the study did not compare the use of rewards and fines to influence behavior in practice, but only in theory. Future studies should address these issues by trying out different schemes that would improve the understanding of fines, rewards, and various possible combinations, on driving behavior.

Second, the study used a relatively small and homogenous sample. Women were significantly underrepresented in the sample. Ideally, future studies should use larger and more diverse samples, which would be followed for a longer period so the effect of experimenter’s desirability would be reduced. In addition, the participants could decide not to turn on the application before certain trips. However, it should be noted that Get Location recorded approximately 88% of the trips (when compared to the results of the online survey, representing an average week of driving), during which drivers committed a considerable number of violations, even regardless of being aware of the use of the monitoring application. Thus, the bias due to intentional misrepresentation of driving patterns would be minimal.

Third, the participants had to turn the application on whenever they began a trip. We included this feature to increase the participants’ willingness to use the application; however, this allowed participants to avoid recording all their trips. It is possible that participants intentionally chose to avoid using Get Location before particular trips in which the drivers planned to commit violations. However, given the high number of violations recorded overall, we believe that most unrecorded trips were due to forgetfulness.

The main strength of this study was the use of an in-vehicle recording system to obtain objective data on individual level driving behavior, including traffic violations. The study demonstrates the need to use objective measures in order to assess the magnitude of particular driving behaviors in general and traffic violations in particular. Furthermore, with a growing population that is technology savvy and familiar with similar methods, makes the application of this method particularly useful. In addition, the study’s results support future explorations of the use of rewards to modify speeding and driving behavior in general. With the advent of new methodologies for the study of big data, the advantage of cellular-application based methodologies is clear. The impact, however, on risky driving behavior has yet to have been comprehensively studied.