Factor structure and measurement invariance of the multidimensional driving style inventory across gender and age: An ESEM approach
Introduction
Despite worldwide efforts over the last decade to increase road safety, motor vehicle crashes remain one of the leading causes of death and injury across all ages (WHO, 2018). Among driving-related factors underlying to crash risk, driving style has been recognized as one of the most important factors (Eboli et al., 2017). Driving style refers to the typical behavioral pattern of drivers, including driving speed, headway, compliance with traffic rules and habitual levels of attentiveness; it is expected to be influenced by drivers’ attitudes and beliefs regarding driving as well as more general needs and values (Elander, West, & French, 1993).
The interest in driving style has generated a number of self-report measures to evaluate different aspects of driver behavior, such as the Driver Behavior Questionnaire (Reason, Manstead, Stradling, Baxter, & Campbell, 1990), the Driving Style Questionnaire (French, West, Elander, & Wilding, 1993), the Driving Behavior Inventory (Gulian, Matthews, Glendon, Davies, & Debney, 1989), the Attitude Towards Traffic Safety Scale (Iversen, 2004), and the Driving Anger Scale (Deffenbacher, Oetting, & Lynch, 1994), to name a few. Based on an integrative review of the literature and existing measures, Taubman-Ben-Ari, Mikulincer, and Gillath (2004) proposed that driving style can be classified into four broad domains: (a) reckless and careless (b) angry and hostile (c) anxious and (d) patient and careful. The reckless and careless driving style refers to seeking sensation and thrill while driving and deliberate transgression of road safety norms. The angry and hostile driving style refers to exhibit feelings of irritation and rage, hostile attitudes while driving and a general tendency to act aggressively towards road users. The anxious driving style reflects feelings of anxiety, tension and alerteness, as well as difficulty engaging in relaxing activities whilst driving. The patient and careful driving style reflects a well-adjusted driving style characterized by attention, calmness, politeness and respect for traffic rules.
On the basis of the above conceptualization, Taubman-Ben-Ari et al. (2004) constructed the Multidimensional Driving Style Inventory (MDSI), a 44-item self-report measure designed to assess various driving styles simultaneously. Results from exploratory factor analysis revealed eight factors: reckless driving, high-velocity driving, angry driving, anxious driving, dissociative driving, distress-reduction driving, patient driving and careful driving. Although the number of factors differed from those hypothesized, the authors concluded that the eight driving styles are internally coherent with the four driving style domains originally proposed, while making finer distinctions within each of the domains. In particular, risky driving domain is represented by the reckless and high-velocity driving styles, while the anxious domain is represented by the anxious, dissociative and distress-reduction driving styles. All the MDSI factors showed good internal consistency (Cronbach’s alpha coefficients ranged from 0.72 to 0.86), correlated in the expected direction with measures of personality and sociodemographic characteristics, and significantly predicted self-report crashes and driving offenses, thus supporting the validity of the scale. Subsequent studies also demonstrated the robustness of the MDSI against social desirability bias (Taubman-Ben-Ari, 2006), providing further support for the validity of the scale.
Since its publication in 2004, the MDSI has been traslated into several languages and validated in different countries, including Brasil (Silva, 2004), Argentina (Poó, Taubman-Ben-Ari, Ledesma, & Díaz-Lázaro, 2013), Spain (Padilla, Castro, Doncel, & Taubman-Ben-Ari, 2020), the Netherlands (van Huysduynen, Terken, Martens, & Eggen, 2015), China (Wang, Qu, Ge, Sun, & Zhang, 2018), and Romania (Holman & Havârneanu, 2015). These studies have strongly supported the reliability and validity of the MDSI; however, the factor structure encountered across the studies has been inconsistent. Specifically, Poó et al. (2013) failed to replicate the original MDSI eight-factor structure through confirmatory factor analysis (CFA), and obtained six factors using exploratory factor analysis (EFA): risky (which includes items from the risky and high-velocity styles of the original MDSI), careful and patient (which includes items from the careful and patient styles of the original MDSI), dissociative, anxious, distress-reduction, and angry driving styles. Similarly, Holman and Havârneanu (2015) found no support for the original MDSI factor structure using CFA, and a subsequent EFA revealed seven factors, six of which were comparable to those of Poó et al. (2013) and one, termed violation of rules contextually perceived as irrational, which was specific to the Romanian driving context and comprised new items targeting behaviors not tapped in the original scale. On the other hand, van Huysduynen et al. (2015) identified five underlying factors to MDSI items that correspond to angry, anxious, dissociative, distress reduction, and a factor that combines items of the careful and risky driving styles from the original MDSI, which was labelled careful driving (as the items of risky driving were negative in this factor). Finally, Wang et al. (2018) found a four-factor structure composed of risky style, angry-high-velocity style, careful style and anxious style in a sample of Chinese drivers, once again after rejecting the original MDSI factor structure using CFA. Overall, the factor structure of the MDSI found across different studies matches the four broad domains of driving styles originally proposed by Taubman-Ben-Ari et al. (2004).Yet, certain differences in the number and configuration of driving styles also appeared which calls for further research.
A number of possible reasons could account for the inconsistencies in previous MDSI factor-analytic findings. On the one hand, one might think of cultural differences since that driving-related behaviors are influenced by cultural and contextual factors varying across cultures (e.g., Özkan and Lajunen, 2011, Şimşekoğlu et al., 2012). Consequently, the original MDSI factor structure may not be universal. On the other hand, there are methodological issues that might account for the findings as well. As already mentioned, most of the studies failing to replicate the MDSI factor structure used the CFA approach as a first strategy. Under this approach, items are allowed to load only on the factor that it was designed to measure and are constrained to zero on the other factors (Assis Gomes et al., 2017). This simple structure imposed by CFA may be too restrictive and problematic, particularly with complex multidimensional constructs (Morin, Marsh, & Nagengast, 2013), such as driving style. Indeed, characterizing driver behavior in terms of a single driving style is too simplistic (van Huysduynen et al., 2015) since drivers often exhibit more complex patterns of driving characterized by behaviors belonging to different styles. In other words, although a driver’s typical behavior may correspond to a single driving style, he/she may also engage, to a certain extent, in behaviors theoretically associated with other driving styles (cf. Chung & Wong, 2010). Moreover, certain driving behaviors may be influenced by several different styles (Holman & Havârneanu, 2015). Accordingly, imposing a simple structure on MDSI data would appear inadequate in view of the complex structure of driving style and would lead to model misspecifications and misfit (Browne, 2001).
Exploratory structural equation modeling (ESEM) has been proposed as an alternative approach to the traditional CFA (Asparouhov & Muthén, 2009). In the ESEM model, items are allowed to cross-load on non-intended factors, thus imposing fewer restrictions on the measurement model (e.g., allows for complex structure) than common CFA while at the same time allows for the calculation of standard SEM statistics (e.g., goodness-of-fit indices, standard errors). Based on these considerations, ESEM provides a particularly useful framework to test the latent structure of MDSI.
Furthermore, the MDSI scores have often been used to examine gender and age differences. Results have generally shown that females score higher than males on the anxious, dissociative and careful driving styles, whereas males score higher than females on the angry and risky driving styles (e.g., Holman and Havârneanu, 2015, Poó et al., 2013, Taubman-Ben-Ari and Skivirsky, 2016). Regarding age, most studies indicate that anxious, dissociative, risky and angry driving styles decrease with age, and patient and careful styles increase (e.g., Navon and Taubman-Ben-Ari, 2019, Taubman-Ben-Ari and Yehiel, 2012, Trógolo et al., 2014). Nonetheless, none of these studies have tested the measurement invariance of MDSI in these groups. This is worrisome since valid comparison between groups requires that the construct being measured by the scale be equivalent across the groups (Dimitrov, 2010). A lack of invariance could lead to serious consequences, including erroneous driving course referrals for certain groups of drivers considered to be of higher risk based on MDSI profile scores, when driving styles are in fact assessed differently in these groups.
In line with the above, the objective of the present study was to analyze the factor structure of the Argentinian version of the MDSI and test the measurement invariance of the retained model across gender and age groups of drivers using an ESEM approach. We therefore sought to provide new evidence of validity that support using MDSI scores to compare different groups.
Section snippets
Participants
An incidental sample of 1282 drivers from Argentina was recruited for the study. Five respondents had missing values and were deleted from the dataset. The final sample consisted of 675 males and 602 females, aged 18 to 80 (M = 35.73; SD = 13.80). The majority of the participants drove regularly (83.2%) and held a university degree (41.3%).
Measures
The Argentinian version of the Multidimensional Driving Style Inventory (Argentinian MDSI-S; Poó et al., 2013) was used. It consists of 40 items assessing
Preliminary analysis
Five participants (0.4%) showed missing values in one or more items; since they represented so few cases, we decided to delete them and keep only cases with complete responses. Item mean (and standard deviation) values ranged from 1.54 to 5.19 (0.94–1.73), and various items had skeweness and kurtosis above 1 in absolute values, thus indicating a significant deviation from univariate normality. In addition, Mardia’s kurtosis multivariate test showed a significant departure from multivariate
Discussion
The purpose of this study was to provide further evidence the factor structure of Argentinian MDSI-S and its measurement invariance across gender and age, using an ESEM approach. Findings showed a six-factor model represented by risky style, angry style, dissociative style, anxious style, careful and patient style, and distress-reduction style to be the best model based on the joint consideration of statistical fit indices and interpretability. The findings were similar across two separate
Conclusion
Despite these limitations, to the best of our knowledge this study is the first to test the measurement invariance of the Argentinian MDSI-S across gender and age. Providing empirical support for the measurement equivalence across frequently compared groups is essential to judge the validity of findings. From a practical standpoint, our paper indicates that the Argentinian MDSI is an adequate tool for examining gender- and age-related differences in driving styles and, consequently, to identify
Financial disclosure
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (36)
- et al.
How to define the accident risk level of car drivers by combining objective and subjective measures of driving style
Transport. Res. Part F: Traffic Psychol. Behav.
(2017) - et al.
The Romanian version of the multidimensional driving style inventory: Psychometric properties and cultural specificities
Transport. Res. Part F: Traffic Psychol. Behav.
(2015) Risk-taking attitudes and risky driving behaviour
Transport. Res. Part F: Traffic Psychol. Behav.
(2004)- et al.
Driven by emotions: The association between emotion regulation, forgivingness, and driving styles
Transport. Res. Part F: Traffic Psychol. Behav.
(2019) - et al.
Person and environment: Traffic culture
- et al.
Reliability and validity of a Spanish-language version of the multidimensional driving style inventory
Transport. Res. Part F: Traffic Psychol. Behav.
(2013) Couple similarity for driving style
Transport. Res. Part F: Traffic Psychol. Behav.
(2006)- et al.
The multidimensional driving style inventory a decade later: Reviewof the literature and re-evaluation of the scale
Accident Anal. Prevent.
(2016) - et al.
Driving styles and their associations with personality and motivation
Accident Anal. Prevent.
(2012) - et al.
The multidimensional driving style inventory—scale construct and validation
Accident Anal. Prevent.
(2004)