- Email: [email protected]
Development of a short form of the driving anger expression inventory
Accident Analysis and Prevention 72 (2014) 169–176
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Development of a short form of the driving anger expression inventory Amanda N. Stephens a,∗ , Mark J.M. Sullman b a b
Monash University Accident Research Centre, Melbourne, Australia Driving Research Group, Cranfield University, UK
a r t i c l e
i n f o
Article history: Received 17 April 2014 Received in revised form 21 June 2014 Accepted 23 June 2014 Keywords: Driving anger Aggressive driving DAX DAS Anger expression
a b s t r a c t The present study developed a revised version of the driving anger expression inventory (25-items) and a short (15-item) version using data from 551 drivers. Split half factor analyses on both versions confirmed the original four factors; personal physical aggressive expression, use of a vehicle to express anger, verbal aggressive expression and adaptive/constructive expression. The two DAX versions were strongly correlated, demonstrating the suitability of both forms of the scale and the aggressive forms of expression were higher for drivers who reported initiating road rage interactions. Total aggressive expression was also higher for drivers who reported recent crash-related conditions, such as: loss of concentration, losing control of their vehicle, moving violations, near-misses and major crashes. The revised DAX and DAX-short provide shorter versions of the 49-item DAX that can more easily be combined with other questionnaires and require smaller sample sizes to analyse. Further research is required to validate these tools among different samples and populations. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction Driving aggressively increases a driver’s chances of becoming involved in a motor vehicle crash and also their chances of a serious crash. A survey of Irish drivers found that 40% shout or become angry behind the wheel on a weekly basis and 13% had left their car to confront another driver (Automobile Association Ireland, 2011). Moreover, in 2011 aggressive driving was a factor in 3% of all reported traffic crashes; 3% of minor crashes; 4% of major crashes and 7% of all fatalities in the UK (DfT, 2011). It is also likely that the actual contribution of aggression to traffic crashes is considerably higher, as crash statistics rely on a police officer being called to the scene and specifically noting aggression as a contributory factor. Moreover, anger can result in behaviours that are also independently considered to contribute to motor vehicle crashes, such as: careless, reckless or hurried driving, losing control of one’s vehicle, sudden braking, excessive speed and close following (DfT, 2011). These are all behaviours that may be aggressive expressions of anger and are consistently found by researchers to result from driving anger (Deffenbacher et al., 2002; Stephens and Ohtsuka, 2014; Sullman et al., 2013). Indeed in studies not measuring
∗ Corresponding author at: Monash University Accident Research Centre, Monash University, Clayton, Melbourne, Australia. Tel.: +61 39905 1191. E-mail addresses: [email protected] (A.N. Stephens), m.sullman@cranfield.ac.uk (M.J.M. Sullman). http://dx.doi.org/10.1016/j.aap.2014.06.021 0001-4575/© 2014 Elsevier Ltd. All rights reserved.
aggression directly, driving anger has been found to relate to selfreported crashes (Deffenbacher et al., 1994; Sullman and Stephens, 2013) and near crashes (Underwood et al., 1999) and to increase the likelihood of crashing in driving simulator scenarios (Deffenbacher et al., 2003; Stephens, Trawley, Madigan & Groeger, 2013; Stephens and Groeger, 2011). The majority of self-report studies on the expression of driving anger have used the Driving Anger Expression Inventory (DAX; Deffenbacher et al., 2002). The most commonly used version of the DAX comprises 49-items and four factors, which are: verbal aggressive expression (VAE) – verbally expressing anger (e.g. yelling at the other driver); personal physical aggressive expression (PPAE) – using themselves to express anger (e.g. make hostile gestures); use of the vehicle to express anger (UoV) (e.g. drive a little faster); and adaptive/constructive expression (A/C) – constructive behaviours the driver can engage in (e.g. try to ignore it). These four factors have been found to have good internal reliability (˛ = 0.80–0.90; Deffenbacher et al., 2002). A total aggressive expression (TAE; summed scores for VAE, PPAE and UoV) is also used. The validity of the DAX has also been shown through correlations with self-reported driver aggression (see Deffenbacher et al., 2007). The DAX has been administered to drivers from France (Villieux and Delhomme, 2010), Malaysia (Sullman et al., submitted for publication), Romania (Sarbescu, 2012), Serbia (Jovanovic´ et al., 2011), Spain (Herrero-Fernández, 2011), Turkey (Esiyok et al., 2007) and the USA (Deffenbacher et al., 2002). Surprising, the
170
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
DAX has not been administered in either of the two countries in the British Isles1 (Ireland or the UK). Recently, Özkan et al. (2011) reported a cross-cultural study comparing aggressive driving behaviour among British, Dutch, Finnish and Turkish drivers. Aggressive expression was measured using a bespoke index, which did not include adaptive mechanisms to reduce anger. This omission is important as research using the DAX shows adaptive mechanisms are the most commonly reported response to driver anger. Furthermore, it is equally important to understand the trait or state factors influencing the use of positive coping strategies. There have been a number of different structures used with the DAX. Initial development of the DAX produced a 53-item, five factor model, which included an additional four item factor labelled displaced aggression. However, low reliability of the factor led the authors to advise against use of this factor in future research (Deffenbacher et al., 2002). Recently, using Confirmatory Factor Analysis (CFA), Herrero-Fernández, 2011 found support for the longer five-factor model, which also included the displaced aggression factor. To achieve acceptable model fit, three items were dropped from the analysis. However, a 49 or 53 item scale is lengthy to administer, especially if used in conjunction with other scales. Further, given statistical assumptions regarding ratios of questionnaire items to number of participants, a large sample is required to appropriately analyse the longer form of the DAX. Other researchers have reduced the number of factors in the model. For instance, Sarbescu (2012) reported a 30-item threefactor model, which combined verbal and physical aggression into one factor. However, these results are contentious as the researchers performed Principal Components Analysis and CFA on the same data using a relatively small sample. In contrast, using data from French drivers, Villieux and Delhomme (2010) reported a three-factor solution using only 11 of the original 49items. These researchers removed all personal physical aggression expression items, but found a three factor solution that included: adaptive/constructive, Verbal aggressive expression and use of a vehicle to express anger. Nonetheless, some research has supported the four-factor model, but only obtained acceptable fit after deleting one or two items and allowing a large number of errorpairs to co-vary (Sullman et al., 2013; Sullman et al., submitted for publication), which suggests some degree of redundancy across the 49-items. Although the DAX has been validated across a variety of different driving populations there is some inconsistency regarding the driver characteristics that relate to the propensity to become aggressive while driving. Some researchers have found that females report more adaptive/constructive means of reducing anger (Esiyok et al., 2007; Jovanovic´ et al., 2011). Males tend to engage more often in personal physical aggressive expression (Dahlen and Ragan, 2004; Esiyok et al., 2007). However, others have found no sex differences (Villieux and Delhomme, 2010). It is unclear whether this represents underlying differences in driving populations, sampling techniques or is an artefact of the differing factor structures used. Age has consistently been found to be related to driving aggression tendencies. Jovanovic´ et al. (2011) found a significant negative relationship between age and the total of all the aggressive expression items. Furthermore, age was found to be negatively related to personal physical aggressive expression, use of a vehicle and verbal aggressive expression, while adaptive/constructive expression was positively related to age (Deffenbacher et al., 2007). Total aggressive expression and verbal aggressive expression were also both negatively related to age amongst Turkish taxi drivers (Sullman et al., 2013). In fact, with the exception of Moore and Dahlen (2008), most
1 The British Isles is the geographic term for the group of islands off the north western coast of continental Europe, which includes the UK and Republic of Ireland.
Table 1 Descriptive variables. Variable
Mean (SD)
Range
Annual mileage (miles) Mean age (years) Average length of licence (years) Minor crashesa Major crashesa Near-missesa Summonsa Lost concentrationa Lost controla Experienced road ragea Engaged in road ragea
9543 (7081) 37.85 (12.93) 17.15 (12.63) 0.18 (0.67) 0.03 (0.21) 1.53 (3.38) 0.17 (1.15) 6.22 (20.07) 0.94 (2.37) 1.81 (2.40) 2.10 (2.59)
0 –200,000 18 –75 0 –55 0–10 0–2 0–50 0–25 0–365 0–30 0–20 0–20
a
Past twelve months.
previous research has found some relationships between age and the DAX subscales (Esiyok et al., 2007; Herrero-Fernández, 2011; Villieux and Delhomme, 2010). Aside from the inconsistencies in item inclusion mentioned above, the length of the 49-item DAX makes it unwieldy, reduces the ability to combine it with other measures and necessitates large sample sizes. Therefore, the main aims of the present research were to revise and reduce the DAX and to validate these scales using a sample of drivers from the British Isles. A further aim was to examine age and gender differences across the sample to provide additional evidence about the contribution of these to driver aggression. 2. Method 2.1. Participants A total of 551 drivers from the British Isles (males = 49%) completed the online questionnaire. Of these, 52% were from the UK and 48% from the Republic of Ireland. Participants’ had an average age of 38 (SD = 13) years, had held a driving licence for 17 (SD = 13) years and had an annual mileage of approximately 9543 (SD = 7081) miles (Table 1). 2.2. Materials The 49-item four-factor DAX was used to measure how drivers express their anger whilst driving (Deffenbacher et al., 2002). This version was chosen as it is the most commonly used form of the DAX and the developers of the DAX had advised against the longer 53-item version. For each of the 49-items, respondents rate, on a four-point scale (1 = almost never, 4 = almost always), the frequency of each potential reaction to feeling angry while driving. The 14-item Driving Anger Scale (DAS) provided an overall measure of trait driving anger (Deffenbacher et al., 1994). The DAS and the DAX have demonstrated moderate correlations (Deffenbacher et al., 2002) and as such the measure was included to support validation of the new forms of the DAX. Participants rate the level of anger elicited by each of the 14 potentially anger inducing situations on a five-point scale (1 = not at all, 5 = very much). Higher scores on the DAS indicate stronger tendencies towards anger while driving. The DAS short-form has demonstrated good internal consistency (˛ = 0.80; Deffenbacher et al., 1994). The validity of the DAS has been shown through correlations with Spielberger’s (1988) Trait Anger Scale (Deffenbacher et al., 1994; Sullman and Stephens, 2013). Six items from the Driving Survey (Deffenbacher et al., 2002), were used to measure how many times, in the last year, drivers had: been fined or prosecuted for a driving offence (excluding parking tickets), lost concentration, lost control of their vehicle, experienced a near-miss, had a minor crash or a major crash.
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
171
Fig. 1. CFA model of the DAX revised (25-items) and the DAX short (15-items). Dashed lines represent items in 25 item model that are removed for the 15 item model. The factor loadings are presented in Table 1.
Drivers also reported how often in the past year they had experienced (four items) or initiated road rage incidents (four items). These measures were included to validate aggressive tendencies against self-reported detrimental and aggressive behaviours. Previous research has shown positive relationships between scores on the DAX and crash-related conditions (Deffenbacher et al., 2002; Sullman et al., 2013; Sullman et al., submitted for publication) and self-reported road rage (Sullman et al., submitted for publication). 2.3. Procedure A link to the online questionnaire was distributed to drivers via social media sites, on-line advertising websites and emails sent to staff and students at Cranfield University and University College Cork. Drivers were informed that the study was voluntary, responses were anonymous and that anyone who held a current driving licence and had driven at least once in the past six months could complete the survey. The ethics committees of both universities approved the study. 2.4. Statistical analysis EQS for windows (version 6.1) was used to conduct Confirmatory Factor Analysis (CFA) on the DAX scales. The robust method of maximum likelihood (ML) was utilised due to the violation of multivariate normality across questionnaire responses and as Mardia’s normalised estimate was >5. Goodness-of-fit indices included the Santorra-Bentler Scales Chi-Squared (S-B2 ), S-B2 /df, adjusted Comparative Fit Index (CFI), Root Mean Square Error of Approximation (RMSEA). Acceptable model fit was indicated by an S-B2 /df < 5; adjusted CFI of 0.90 or greater and an RMSEA of 0.06 or less (Hu and Bentler, 1999). The confidence interval (C.I.) reporting a 90% interval surrounding the RMSEA was also examined. 3. Results 3.1. Descriptive variables Table 1 shows the mean responses on the driver behaviour survey questions. The most common of the six behaviours was losing concentration while driving, with some drivers reporting this to
be an almost daily occurrence. Overall crash involvement was low, averaging less than one minor or major crash per year. 3.2. Driving anger expression Table 2 shows the means and standard deviations of the DAX items2 . Items from the adaptive/constructive factor received the three highest scores. In order, these were: “I just try to accept that there are bad drivers on the road”; “I decide not to stoop to their level”; and “I just try to accept that there are frustrating situations”. The three lowest scored items all belonged to the personal physical aggressive expression factor. These were: “I try to get out of the car and have a physical fight with the other driver”; “I bump the other driver’s bumper with my own”; and “I try to get out of the car and tell the other driver off”. Overall, drivers reported relatively low aggressive responses to anger, scoring an average of 1.50 (SD = 0.37) out of a possible 4. Adaptive/constructive expressions were the most commonly reported, followed by Verbal aggressive expression, and then use of a vehicle to express anger. personal physical aggressive expressions were the least commonly reported. The order of factor means was the same as that reported in previous research (Sullman et al., 2013; Sullman et al., submitted for publication; Villieux and Delhomme, 2010). 3.3. Reduction of the DAX items Split half validation was used to confirm shorter versions of the DAX. A random sample of 200 cases was selected from the full data set and subjected to Principle Components Analysis (PCA). The Kaiser–Meyer–Olkin measure of sampling adequacy was 0.87 and Bartlett’s test of sphericity was significant (2 (300) = 2894, p < 0.0001), indicating the data were suitable for factor analysis. Initially, all 49-items were analysed using the direct Oblimin method of rotation, resulting in thirteen factors with eigenvalues >1. The direct Oblimin method was used as the DAX factors
2 The total DAX scores did not differ between drivers when compared across UK and Ireland.
172
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
Table 2 DAX item means. DAX items (N = 551) M (SD) Adaptive/constructive expression (˛ = 0.89) Accept there are bad drivers on the road 42 30 Not stoop to their level 45 Accept there are frustrating situations 36 Tell myself it’s not worth getting involved 49 Pay closer attention to others’ driving to avoid accidents 23 Pay closer attention to being a safe driver 48 Tell myself to ignore it 25 Think things through before I respond Tell myself it’s not worth getting mad at 29 Think of positive things to do 35 Think of positive solutions to deal with the 26 situation 32 Turn on the radio/music to calm down Think about things that distract me from 47 the frustration 24 Think about things that distract me from thinking about Take deep breaths to calm down 44
2.30 (0.58) 2.71 (0.93) 2.70 (0.97) 2.69 (0.91) 2.66 (0.98) 2.64 (0.94)
Verbal aggressive expression (˛ = 0.87) Shake my head at the other driver 37 Make negative comments about the driver 6 aloud Call the other driver names aloud 5 Give the other driver dirty looks 40 Glare at the other driver 11 Make negative comments about the driver 39 under my breath 28 Swear at the other driver aloud 31 Swear at the other driver under my breath Call the other driver names under my 14 breath Think things like “where did you get your 43 licence?” 38 Yell at the other driver Yell questions like “where did you get your 9 licence?”
2.02 (0.58) 2.40 (0.88) 2.34 (0.89)
2.61 (0.91) 2.54 (0.94) 2.51 (0.91) 2.48 (0.93) 1.88 (0.87) 1.95 (0.87) 1.78 (0.89) 1.76 (0.82) 1.76 (0.86) 1.76 (0.89)
2.20 (0.93) 2.13 (0.92) 2.12 (0.87) 2.11 (0.84) 2.07 (0.99) 2.04 (0.92) 2.00 (0.87) 1.82 (0.95) 1.60 (0.86) 1.39 (0.76)
Use of vehicle to express anger (˛ = 0.87) Flash my lights at the driver 33 Drive a little faster 3 46 Slow down to frustrate other driver Block the other driver from doing what 16 he/she wants to do Speed up to frustrate other driver 15 Drive right up on the other driver’s bumper 2 Drive a lot faster 27 Do to drivers what they did to me 22 Leave my lights on full in their mirrors 19 Follow right behind for a long time 7 Try to cut in front 4
1.33 (0.40) 1.75 (0.83) 1.60 (0.69) 1.47 (0.76) 1.36 (0.61)
Personal physical aggressive expression (˛ = 0.90) Give the finger 1 Make other hostile gestures 34 Go crazy behind the wheel 18 Shake my fist 12 Roll down the window to communicate my 10 anger 21 Try to scare the driver Try to force the driver to the side of the 20 road 13 Stick my tongue out 8 Try to get out of the car and tell the other driver off 17 Bump the driver’s bumper with my own Try to get out and have a physical fight 41
1.16 (0.37) 1.33 (0.69) 1.30 (0.65) 1.21 (0.60) 1.20 (0.59) 1.17 (0.56)
Total
1.50 (0.37)
1.27 (0.62) 1.26 (0.55) 1.24 (0.57) 1.23 (0.55) 1.18 (0.55) 1.15 (0.50) 1.12 (0.42)
1.14 (0.51) 1.11 (0.47) 1.10 (0.44) 1.08 (0.37) 1.06 (0.35) 1.05 (0.33)
have been shown to be correlated (Sullman et al., 2013; Sullman et al., submitted for publication). A parallel analysis, which outlines minimum eigenvalues for factor retention, suggested a four-factor solution that accounted for 60% of the variance. Several iterations of the PCA were performed to inform item elimination. Items were excluded if they did not conceptually fit within a factor, if their factor loading was less than 0.40 or if they cross-loaded onto another factor (>0.40). This process resulted in a 25-item scale that retained the four dimensions: Adaptive/constructive; verbal aggressive expression; use of a vehicle; and personal physical aggressive expression (see Table 3). Adaptive/constructive expression was reduced from 15 to 10 items, with items, 24, 25, 32, 44 and 47 excluded. These all involve active distractions such as thinking about non-traffic related situations, turning on the radio or focusing on calm breathing. Personal physical aggressive expression was reduced from 11 to 5 items (excluding items 1, 12, 13, 17, 18 and 34). Item 18 “I go crazy behind the wheel” has previously been identified as problematic and removed from the DAX analysis (Esiyok et al., 2007; Sullman et al., 2013; Villieux and Delholmme, 2010). The Verbal aggressive expression factor was reduced from 12 to 5 items by excluding items (removing items 9, 11, 31, 37, 39, 40, 43) involving non-verbal communications (i.e., giving a dirty look) and conceptually similar items. For example, Item 9 “I yell questions like where did you get your licence” is similar to Item 38 “I yell at the other driver”. The use of a vehicle to express anger factor was reduced from 11 items to 5 (excluding 3, 4, 19, 33) by removing the lower correlated items that had a similar meaning to another item (e.g., 3 and 27). For example, Item 3 “I drive a little faster” and Item 27 “I drive a lot faster”. Items related to flashing or leaving on the headlights and cutting in front of the driver were also removed. All factors demonstrated good reliability (˛ ranging from 0.75 to 0.88). 3.4. CFA on the revised DAX CFA was conducted on the remaining 351 cases not selected for PCA. The model showed less than adequate fit to the data (S-B2 (269) = 815.94, p < 0.001, S-B2 /df = 3.03; CFI = 0.61; RMSEA = 0.08; 90% C.I. = 0.07:0.08). Lagrange Multiplier Tests revealed four error-pairs to be covaried. As can be seen in Table 3, it made conceptual sense to co-vary each of the error-pairs. Three of these were from the adaptive/constructive factor. Items 26: “Think of positive solutions to deal with the situation” and Item 35 “Think of positive things to do”; Item 23 “Pay closer attention to being a safe driver” and Item 49 “Pay closer attention to other’s driving to avoid accidents”; and, Item 30 “Not stoop to their level” and 36 “Tell myself it’s not worth getting involved”. The other error-pair was from the verbal aggressive expression factor and both involved making negative comments aloud: Item 5 “Call the other driver names aloud” and Item 6 “Make negative comments about the driver aloud”. When the model was respecified, allowing the errors for these conceptually similar items to co-vary, it produced an acceptable fit (S-B2 (265) = 363.23, p < 0.001, S-B2 /df = 1.37; CFI = 0.92; RMSEA = 0.04; 90% C.I. = 0.03:0.05), with acceptable internal consistency (Rho = 0.81) and composite reliability scores >0.80 for all factors. Within method convergent validity was also demonstrated by statistically significant regression coefficients and 96% of the loadings being >0.40 (Fig. 1). 3.5. Developing the DAX-short As a 25-item scale may still be too long in some studies, the revised DAX was subjected to further item reduction to obtain the shortest possible measure. The same 200 cases were used that were used to develop the 25-item measure. Items were removed that were conceptually similar, retaining the item with the largest
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
173
Table 3 Factor loadings for the DAX revised and DAX short (PCA, N = 200) and (CFA, N = 351). Revised (25-item) PCA Factor 1: adaptive/constructive Item 26e1 Think of positive solutions to deal with the situation 48 Tell myself to ignore it e4 36 Tell myself it’s not worth getting involved e1 Think of positive things to do 35 Pay closer attention to others’ driving to avoid accidents 49e3 Accept there are frustrating situations 45 Tell myself it’s not worth getting mad at 29 42 Accept there are bad drivers on the road 23e3 Pay closer attention to being a safe driver e4 30 Not stoop to their level Factor 2: Personal physical aggressive expression Items Try to get out and have a physical fight 41 Try to get out of the car and tell the other driver off 8 Bump the driver’s bumper with my own 17 Try to scare the driver 21 Roll down the window to communicate my anger 10 Factor 3: Verbal aggressive expression Items 5e2 Call the other driver names aloud Make negative comments about the driver aloud 6e2 Swear at the other driver aloud 28 Yell at the other driver 38 Swear at the other driver under my breath 31 Factor 4: Use of vehicle to express anger 7 22 15 2 27
Follow right behind for a long time Do to drivers what they did to me Speed up to frustrate other driver Drive right up on the other driver’s bumper Drive a lot faster
33% variance ˛ = 0.88 −0.75 −0.75 −0.74 −0.74 −0.73 −0.71 −0.70 −0.69 −0.67 −0.67 12% variance ˛ = 0.75 0.93 0.87 0.86 0.59 0.49 9% variance ˛ = 0.84 0.83 0.80 0.75 0.72 0.62 6% variance ˛ = 0.74 −0.65 −0.55 −0.44 −0.40 −0.36
Short (15-item) CFA ω = 0.87 0.45 0.89 0.78 0.44 0.34 0.82 0.73 0.69 0.34 0.69 = 0.91 0.95 0.89 0.77 0.67 0.79 = 0.83 0.72 0.71 0.85 0.73 0.43 ω = .82 0.83 0.63 0.77 0.69 0.63
PCA
CFA
39% variance ˛ = 0.87 −0.76 −0.83 −0.82
ω = 0.84 0.44 0.88 0.72
−0.82 −0.78
0.80 0.71
15% variance ˛ = 0.75 0.93 0.93
= 0.89 0.88 0.89
0.62 0.63 9% variance ˛ = 0.81
0.67 0.80
0.83 0.78 0.74
0.70 0.87 0.72
7% variance ˛ = 0.78
ω = 0.73
0.88
0.75
0.73 0.75
0.60 0.70
= 0.81
e1, e2, e3, e4: error pairs covaried during the CFA; ˛ = Cronbach’s alpha; ω = composite reliability.
contribution. This process resulted in a 15-item scale (see Table 3), which demonstrated good reliability (˛ ranged from 0.75 to 0.86). CFA confirmed the 15-item four-factor model adequately fitted the data (S-B2 (84) = 131.32, p < 0.001, S-B2 /df = 1.56; CFI = 0.91; RMSEA = 0.04; 90% C.I. = 0.03:0.05). Internal consistency was satisfactory (Rho = 0.82) and composite reliability scores were >0.70 for all factors. Within method convergent validity was also demonstrated by statistically significant regression coefficients and all of the factors loadings were >0.40. The factors derived from the DAX revised (25-item) and DAXshort (15-items) were strongly correlated (adaptive/constructive: r (549) = 0.95, p < 0.001; verbal expressions of anger: r (549) = 0.97, p < 0.001; use of a vehicle to display anger: r (549) = 0.88, p < 0.001). The total aggressive expression scores were also strongly correlated (r (549) = 0.97, p < 0.001). As both scales (15-item and 25-item) produced similar findings, to avoid extensive repetition only the results of the most parsimonious version (i.e., 15-item) are presented below. 3.6. Intercorrelations among variables Adaptive/constructive aggression, verbal aggressive expression and trait driving anger scores (DAS scores) were all normally distributed. However, use of a vehicle, personal physical aggressive expression and consequently, total aggressive expression were positively skewed and were thus analysed using non-parametric tests. Table 4 shows that the DAX factors shared moderate correlations (rs 0.23–0.38) indicating related but separate constructs. Verbal aggressive expression scores were strongly related to the total aggressive expression score, suggesting that a large
proportion of the variance in the total aggression score is from this factor. Thus, using the DAX as a four-factor measure (rather than a positive/negative structure) may be more appropriate in understanding a driver’s tendency for different types of aggressive expression. DAS scores were moderately, positively related to verbal aggressive expression, use of a vehicle to express anger and personal physical aggressive expression and negatively related to adaptive/constructive means of dealing with anger, demonstrating the relationship between driving anger and aggressive expressions of anger. Likewise, scores on verbal aggressive expression, use of a vehicle and personal physical aggressive expression decreased with age, whereas adaptive/constructive factor scores increased with age. 3.7. DAX by sex The DAX scores were compared across sex using independent samples t-tests on parametric variables and Mann–Whitney U on non-normally distributed variables. Table 5 shows there were no differences in average scores between males and females for the adaptive/constructive means of dealing with anger, nor for verbal aggressive expression or total aggressive expression scores. Males scored higher for use of a vehicle to express anger. 3.8. DAX by crashes Mann–Whitney U tests were conducted to compare scores on adaptive/constructive means of dealing with anger, verbal aggressive expression, use of a vehicle to express anger and total aggressive expression between drivers who, during the last 12 months, had and those who had not: received a traffic ticket
174
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
Table 4 Intercorrelations between variables (N = 551).
1. Age 2. Years licensed 3. Annual mileage 4. DAS 5. A/C 6. VAE 7. UoV 8. PPAE 9. TAE Means (SD)
1
2
3
– 0.92*** 0.03 −0.12** 0.15*** −0.10* −0.13* −0.04 −0.13** See Table 1
– 0.17*** −0.12** 0.13** −0.07 −0.10* −0.04 −0.10*
– 0.06 0.06 0.04 0.10* −0.01 0.03
4
5
6
7
8
9
– −0.26*** 0.41*** 0.31*** 0.19*** 0.40*** 2.80 (0.66)
– −0.38*** −0.33*** −0.28*** −0.43*** 2.47 (0.73)
– 0.31*** 0.23*** 0.91*** 2.00 (0.78)
– 0.37*** 0.58*** 1.25 (0.45)
– 0.43*** 1.11 (0.37)
– 1.45 (0.42)
A/C, adaptive constructive; PPAE, personal physical; VAE, verbal; UoV, use of vehicle; TAE, total aggressive expression; DAS, driving anger scale. Upper panel: Pearson’s correlation coefficient; lower panel Spearman’s Rho. * p < .05. ** p < .01. *** p < .001.
(excluding parking tickets); lost concentration while driving; lost control of the vehicle; experienced a near-miss; had a minor crash; or a major crash. One driver had been licensed for less than a year and was thus excluded. Yes or no membership was determined by metric responses on each crash related condition (0 = no; >0 = yes). Several differences emerged between drivers who reported crashes or crash-related conditions over the past year and those who did not (Table 6). Verbal aggressive expression scores were reliably higher for drivers who had also reported losses of concentration (N = 398) near-misses (N = 326), and a major crash (N = 16), when compared to those who had not (N = 152; N = 224; N = 534). Use of the vehicle to express anger and personal physical aggressive expression scores were also higher for drivers involved in a major crash. Drivers who reported more road-rage type behaviours (N = 340) scored higher on verbal aggressive expression, use of a vehicle and personal physical aggressive expression than those who had not (N = 210). In contrast, adaptive/constructive means of dealing with anger was higher for those who had not initiated road rage interactions.
4. Discussion The present research developed and validated a refined and short version of the DAX. Split half validation confirmed both the revised 25-item and short (15-item) scales, which both retained the original four factors of: adaptive/constructive means of dealing with anger, verbal aggressive expression, use of vehicle to express anger and personal physical aggressive expression (Deffenbacher et al., 2002). As both scales (25-item and 15-item) produced similar findings, to avoid extensive repetition only the results of the 15-item scale are presented. As the goal in reduction analysis is to find the most parsimonious model, the 15-item DAX provides a good balance between fitting the data well while using a few parameters. However, the 25-item model, produced adequate fit and allows researchers to have a reduced scale with additional items should they desire to use them. Prior to reduction analysis, the pattern of item means were checked for consistency across previously published research. The
means for the 49-item scale were similar to those found using drivers from Malaysia (Sullman et al., submitted for publication), France (Villieux and Delhomme, 2010) and Turkey (Sullman et al., 2013). For example, the nine most commonly reported responses were all from the adaptive/constructive factor (i.e., try to accept there are bad drivers on the road; “Decide not to stoop to their level”; and “Try to accept that there are frustrating situations on the road”) and were among the most commonly reported previously (Sullman et al., submitted for publication; Sullman et al., 2013; Villieux and Delhomme, 2010). Furthermore, the items with the lowest means were all from the personal physical aggressive expression factor (i.e., “I try to get out of the car and have a physical fight with the other driver”; “I bump the other driver’s bumper with my own”; and, “I try to get out of the car and tell the other driver off”), which was also consistent with drivers from Malaysia, France and Turkey. This suggests that globally, drivers are most likely to use constructive mechanisms to deal with anger and rarely resort to extreme displays of anger involving personal physical aggression. This might represent the moderate levels of driving anger tendencies reported by the participants. It would be interesting to assess whether drivers reporting a problem with driving anger display similar patterns expression, or whether more extreme behaviours are more common in that sample. The intercorrelations between the DAX factors were all moderate, showing similar patterns to those from other countries using the longer scale (Dahlen and Ragan, 2004; Deffenbacher et al., 2002; Herrero-Fernández, 2011). For example, verbal aggressive expression, use of a vehicle to express anger and personal physical aggressive expression were all positively correlated and the adaptive/constructive factor shared negative relationships with the other three aggressive factors. The moderate nature of these relationships indicates that these factors are related but measure separate components of anger expression. Also, in accord with previous research, the DAS was positively correlated with the three aggressive factors and negatively correlated with adaptive/constructive behaviour (Dahlen and Ragan, 2004; Herrero-Fernández, 2011). Previous research has produced mixed findings regarding sex differences on the DAX factors, but when these have emerged, they
Table 5 Gender differences in the DAX short factors.
Adaptive/constructive Verbal aggressive expression Personal physical aggressive expression Use of vehicle to express anger Total aggressive expression
Males (N = 269) M (SD)
Females (N = 282) M (SD)
2.45 (0.77) 2.01 (0.78) 1.14 (.041) 1.29 (0.53) 1.48 (0.47)
2.49 (0.69) 2.00 (0.77) 1.07 (0.32) 1.19 (0.53) 1.42 (0.35)
t < 1, ns t < 1, ns U, ns U = 34,747, z = −2.01, p < 0.05 U, ns
0.12 2.14 (1.11) 1.43 (0.36) <0.01 0.09 1.88 (1.34) 1.09 (0.26) <05 0.14 0.09 0.03 2.62 (0.81) 2.46 (0.73) NS
2.50 (1.04) 1.99 (0.76) <0.05
2.04 (1.13) 1.22 (0.40) <0.001
0.06 1.61 (0.69) 1.43 (0.35) NS 0.09 1.30 (0.74) 1.08 (0.25) <0.05 0.07 0.05 0.05 2.36 (0.68) 2.48 (0.74) NS
2.12 (0.87) 1.99 (0.76) NS
1.41 (0.73) 1.22 (0.38) NS
0.18 1.51 (0.43) 1.37 (0.37) <0.001 0.09 <0.05 1.12 (0.37) 1.09 0.09 <0.05 0.17 0.03 2.44 (0.71) 2.50 (0.76) NS
2.12 (0.82) 1.83 (0.69) <0.001
1.28 (0.48) 1.20
0.10 1.51 (0.48) 1.41 (0.37) <0.05 0.09 <0.05 1.15 (0.45) 1.08 0.10 0.07 0.07 2.40 (0.67) 2.51 (0.76) NS
2.09 (0.83) 1.95 (0.74) NS
1.30 (0.51) 1.21 (0.41) <0.05
0.14 1.48 (0.43) 1.37 (0.36) <0.01 0.03 1.11 (0.39) 1.10 (0.29) NS 0.08 0.14 0.07 2.44 (0.70) 2.54 (0.81) NS
2.07 (0.80) 1.84 (0.72) <0.01
1.26 (0.47) 1.19 (0.37) NS
0.11 1.71 (0.74) 1.42 (0.35) <0.01 0.16 1.38 (0.82) 1.08 (0.25) <0.001 0.09 0.11 0.08 2.32 (0.81) 2.48 (0.71) NS
2.28 (0.88) 1.97 (0.76) <0.01
1.46 (0.78) 1.22 (0.40) <0.05
0.49 1.59 (0.45) 1.23 (0.21) <0.001 0.18 1.15 (0.44) 1.04 (0.17) <0.001 0.26 0.45 0.31 2.29 (0.70) 2.76 (0.68) <0.001
2.28 (0.80) 1.55 (0.47) <0.001
1.33 (0.52) 1.10 (0.23) <0.001
0.20 1.49 (0.44) 1.34 (0.33) <0.001 0.07 1.12 (0.41) 1.05 (0.19) NS 0.11 1.27 (0.47) 1.17 (0.37) <0.05 0.17 2.09 (0.79) 1.79 (0.69) <0.001 0.04
No M (SD) r p No M (SD)
2.45 (0.72) 2.51 (0.75) NS
Road rage (experienced) YES: N = 394 NO: N = 156 Road rage (expressed) YES: N = 340 NO: N = 210 Traffic Tickets YES: N = 56 NO: N = 494 Loss of concentration YES: N = 398 NO: N = 152 Loss of control YES: N = 206 (0.30) NO: N = 344 Near-miss YES: N = 326 (0.40) (0.35) NO: N = 224 Minor Crash YES: N = 75 NO: N = 475 Major Crash YES: N = 16 NO: N = 534
p No M (SD) Yes M (SD) r p No M (SD) Yes M (SD) p No M (SD) Yes M (SD) Yes M (SD) Yes M (SD)
p
Verbal aggressive expression Adaptive/constructive
Table 6 DAX short factor means across self-reported involvement in crash-related conditions.
r
Use of vehicle to express anger
r
Personal physical aggressive expression
Total aggressive expression
r
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
175
have supported the finding here that males reported higher levels of use of a vehicle to express anger (Esiyok et al., 2007; Deffenbacher et al., 2002). Males tend to exhibit more direct aggression than females (Archer, 2004) and it seems this remains consistent in the driving context. The Use of a Vehicle to express anger factor consisted of items such as tailgating and increasing speed to express anger as well as a more ambiguous item for reciprocating the annoying behaviour of other drivers. It would be interesting to assess how much each of these behaviours creates anger in the first place and the extent to which anger and aggression on the road are exacerbated by this reciprocal relationship. Further, the lack of sex differences for the other factors of the DAX is worthy of more investigation. Given that males have a tendency to score higher on violent aggression measures (Hennessy and Wiesenthal, 2001), a similar finding for these data would have been expected. However, whether this might be an artefact of the low scores for this factor or the gender role anonymity afforded in a motor vehicle (Ellison-Potter et al., 2001) is unclear. A number of relationships were found between DAX factors and self-reported crash-related conditions. Drivers who reported more verbal aggressive expression, use of a vehicle to express anger or personal physical aggressive expression also reported initiating more incidents of road-rage. Total aggressive expression differed significantly across all but one of the crash-related conditions. Drivers who had recently received a traffic fine, lost concentration, lost control of the vehicle, had a near-miss or a major crash also reported more aggressive anger expression while driving. These are broadly similar to the findings made by Sullman et al. (2013) and Sullman et al. (submitted for publication), although the latter did not find a significant relationship between total aggressive expression and traffic fines. Drivers who had received a traffic fine, lost concentration while driving or had a near-miss in the past year had higher verbal aggressive expression scores when compared to drivers who had not. Drivers who reported recent loss of concentration or control while driving and near-misses also had higher means on the use of a vehicle to express anger subscale. The shorter scale was therefore also able to differentiate between those who self-reported crashes and more frequent engagement in crashrelated conditions from those who had not crashed and reported less frequent engagement in crash-related conditions. 4.1. Limitations The study reported here suffers from the usual perceived weaknesses of self-report, such as social desirability bias. This is particularly the case for questions about illegal, dangerous or socially undesirable behaviours, which includes several of the variables measured in this study. However, as the participants’ names were not recorded and they were all assured of total confidentiality and anonymity, the impact of social desirability bias is likely to have been minimal, as has been found by research investigating the topic (e.g. Sullman and Taylor, 2010). Given that the questionnaire was made available on online, the issue of self-selection bias cannot be overlooked. However, self-selection bias may suggest people who perceive themselves as good drivers are likely to opt into the research. Therefore, stronger relationships between aggressive driving tendencies and self-reported crash history may have emerged with a more randomised sampling method. 4.2. Summary and practical implications The most commonly used form of the DAX contains 49 items, which makes it lengthy to complete and difficult to use in conjunction with other scales. Further, the 49-item DAX often requires respecification during Confirmatory Factor Analysis (Sullman et al., 2013; Sullman et al., submitted for publication) meaning a large
176
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
sample size is required to undertake the appropriate analyses. These limitations can make the longer version impractical for researchers to use. In the research reported in this paper, the 49item scale was reduced to a 25-item scale as well as a 15-item DAXshort, and both the reliability and validity of these shortened scales confirmed. The integrity of the structure proposed by Deffenbacher et al. (2002) was retained, resulting in four factors related to anger expression: Adaptive/constructive means of dealing with anger, verbal aggressive expression, use of a vehicle to display anger and personal physical aggressive expression. The findings support previous research that has shown positive relationships between the aggressive forms of anger expression and crash-related conditions, such as traffic tickets, losing control of the vehicle, loss of concentration and minor and major crashes. Further, the present study has demonstrated relationships between self-reported aggressive tendencies and actual recent road rage behaviour. The 25-item and 15-item versions of the scale could also be used to identify whether an intervention has been successful in reducing dysfunctional expressions of anger while driving (i.e., PPAE, UoV and VAE) or increasing the use of adaptive/constructive methods of dealing with anger. This could be undertaken at a group level (e.g., has this media campaign reduced dysfunctional expressions of driving anger) or on a more individual basis (e.g., has this treatment improved the way this individual deals with potentially anger inducing situations. Although they were not using the DAX, the latter benefit has been used previously by Deffenbacher et al. (2000), where they evaluated two methods for treating angry drivers. Furthermore, as this research has shown there are age differences and sex differences for the use of a vehicle, the present research can also be used to target interventions at these groups aiming to promote engagement in adaptive/constructive forms of dealing with driving anger. In summary, the practical applications of the current study are a revised 25-item scale and a short 15-item version of the DAX, which can be readily incorporated into questionnaires to further understand the causes and consequences of aggressive driving behaviour. Research is currently underway to validate the use of the short DAX with other scales to compare responses on the shorter versions with those of the 49-item version. Moreover, both scales are valid representations of the original DAX and are able to differentiate between those who self-report crashes and crash related conditions. Acknowledgements Dr Stephens would like to acknowledge the support of the School of Applied Psychology, University College Cork, Ireland. This study was conducted while she was based there. References Archer, J., 2004. Sex differences in aggression in real-world settings: a meta-analytic review. Rev. Gen. Psychol. 8 (4), 291.
Automobile Association Ireland, 2011. Bouts of road rage more common among http://blog.aaireland.ie/bouts-of-road-rage-more-common-amongmales. males/ (accessed 08.04.14). Dft, 2011. Reported Road Casualties in Great Britain: 2011: Annual Report, Retrieved 05.13 from https://www.gov.uk/government/uploads/system/ uploads/attachment data/file/9277/rrcgb2011-04.pdf Dahlen, E.R., Ragan, K.M., 2004. Validation of the propensity for angry driving scale. J. Saf. Res. 35 (5), 557–563. Deffenbacher, J.L., Deffenbacher, D.M., Lynch, R.S., Richards, T.L., 2003. Anger, aggression, and risky behavior: a comparison of high and low anger drivers. Behav. Res. Ther. 41 (6), 701–718. Deffenbacher, J.L., Kemper, C.C., Richards, T.L., 2007. The driving anger expression inventory: a validity study with community college student drivers. J. Psychopathol. Behav. Assess. 29, 220–230. Deffenbacher, J.L., Huff, M.E., Lynch, R.S., Oetting, E.R., Salvatore, N.F., 2000. Characteristics and treatment of high-anger drivers. J. Counsel. Psychol. 47 (1), 5–17. Deffenbacher, J.L., Lynch, R.S., Oetting, E.R., Swaim, R.C., 2002. The Driving Anger Expression Inventory: a measure of how people express their anger on the road. Behav. Res. Ther. 40, 717–737. Deffenbacher, J.L., Oetting, E.R., Lynch, R.S., 1994. Development of a driving anger scale. Psychological reports 74 (1), 83–91. Ellison-Potter, P., Bell, P., Deffenbacher, J., 2001. The effects of trait driving anger, anonymity, and aggressive stimuli on aggressive driving behavior. J. Appl. Soc. Psychol. 31, 431–443. Esiyok, B., Yasak, Y., Korkusuz, I., 2007. Anger expression on the road: validity and reliability of the driving anger expression inventory. Turk. Psikiyatri. Derg. 18, 231–243. Hennessy, D.A., Wiesenthal, D.L., 2001. Gender, driver aggression, and driver violence: an applied evaluation. Sex Roles 44, 661–676. Herrero-Fernández, D., 2011. Psychometric adaptation of the Driving Anger Expression Inventory in a Spanish sample: differences by age and gender. Transport. Res. Part F 14, 324–329. ´ D., Lipovac, K., Stanojevic, ´ P., Stanojevic, ´ D., 2011. The effects of personJovanovic, ality traits on driving-related anger and aggressive behaviour in traffic among Serbian drivers. Transport. Res. Part F 14, 43–53. Moore, M., Dahlen, E.R., 2008. Forgiveness and consideration of future consequences in aggressive driving. Accident Analysis & Prevention 40 (5), 1661–1666, http://dx.doi.org/10.1016/j.aap.2008.05.007. Özkan, T., Lajunen, T., Parker, D., Sümer, N., Summala, H., 2011. Aggressive driving among British, Dutch, Finnish and Turkish drivers. International Journal of Crashworthiness 16 (3), 233–238, http://dx.doi.org/10.1080/13588265.2010. 536687. Sarbescu, P.S., 2012. Aggressive driving in Romania: psychometric properties of the Driving Anger Expression Inventory. Transport. Res. Part F 15, 556–564. Spielberger, C.D., 1988. State-Trait Anger Expression Inventory, 1st ed. Psychological Assessment Resources, Odessa, FL. Stephens, A.N., Groeger, J.A., 2011. Anger-congruent behaviour transfers across driving situations. Cognition & emotion 25 (8), 1423–1438. Stephens, A.N., Trawley, S.L., Madigan, R., Groeger, J.A., 2013. Drivers Display AngerCongruent Attention to Potential Traffic Hazards. Applied cognitive psychology 27 (2), 178–189. Stephens, A.N., Ohtsuka, K., 2014. Cognitive biases in aggressive drivers: does illusion of control drive us off the road? Pers. Indiv. Differ. 68, 124–129. Sullman, M.J.M., Stephens, A.N., 2013. A comparison of the Driving Anger Scale and the Propensity for Angry Driving Scale. Accident Analysis & Prevention 58, 88–96, http://dx.doi.org/10.1016/j.aap.2013.05.002. Sullman, M.J.M., Stephens, A.N., Kuzu, D., 2013. The expression of anger amongst Turkish taxi drivers. Accid. Anal. Prev. 56, 42–50. Sullman, M.J.M., Stephens, A.N., Yong, M. (submitted for publication). Anger, aggression and road rage behaviour in Malaysian drivers. Sullman, M.J.M., Taylor, J.E., 2010. Social desirability and self-reported driving behaviours: should we be worried? Transport. Res. Part F 13, 215–221. Underwood, G., Chapman, P., Wright, S., Crundall, D., 1999. Anger while driving. Transport. Res. Part F: Traffic Psychol. Behav. 2 (1), 55–68. Villieux, A., Delhomme, P., 2010. Driving anger and its expressions: further evidence of validity and reliability for the Driving Anger Expression Inventory French adaptation. J. Safety Res. 41, 417–422.
Contents lists available at ScienceDirect
Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap
Development of a short form of the driving anger expression inventory Amanda N. Stephens a,∗ , Mark J.M. Sullman b a b
Monash University Accident Research Centre, Melbourne, Australia Driving Research Group, Cranfield University, UK
a r t i c l e
i n f o
Article history: Received 17 April 2014 Received in revised form 21 June 2014 Accepted 23 June 2014 Keywords: Driving anger Aggressive driving DAX DAS Anger expression
a b s t r a c t The present study developed a revised version of the driving anger expression inventory (25-items) and a short (15-item) version using data from 551 drivers. Split half factor analyses on both versions confirmed the original four factors; personal physical aggressive expression, use of a vehicle to express anger, verbal aggressive expression and adaptive/constructive expression. The two DAX versions were strongly correlated, demonstrating the suitability of both forms of the scale and the aggressive forms of expression were higher for drivers who reported initiating road rage interactions. Total aggressive expression was also higher for drivers who reported recent crash-related conditions, such as: loss of concentration, losing control of their vehicle, moving violations, near-misses and major crashes. The revised DAX and DAX-short provide shorter versions of the 49-item DAX that can more easily be combined with other questionnaires and require smaller sample sizes to analyse. Further research is required to validate these tools among different samples and populations. © 2014 Elsevier Ltd. All rights reserved.
1. Introduction Driving aggressively increases a driver’s chances of becoming involved in a motor vehicle crash and also their chances of a serious crash. A survey of Irish drivers found that 40% shout or become angry behind the wheel on a weekly basis and 13% had left their car to confront another driver (Automobile Association Ireland, 2011). Moreover, in 2011 aggressive driving was a factor in 3% of all reported traffic crashes; 3% of minor crashes; 4% of major crashes and 7% of all fatalities in the UK (DfT, 2011). It is also likely that the actual contribution of aggression to traffic crashes is considerably higher, as crash statistics rely on a police officer being called to the scene and specifically noting aggression as a contributory factor. Moreover, anger can result in behaviours that are also independently considered to contribute to motor vehicle crashes, such as: careless, reckless or hurried driving, losing control of one’s vehicle, sudden braking, excessive speed and close following (DfT, 2011). These are all behaviours that may be aggressive expressions of anger and are consistently found by researchers to result from driving anger (Deffenbacher et al., 2002; Stephens and Ohtsuka, 2014; Sullman et al., 2013). Indeed in studies not measuring
∗ Corresponding author at: Monash University Accident Research Centre, Monash University, Clayton, Melbourne, Australia. Tel.: +61 39905 1191. E-mail addresses: [email protected] (A.N. Stephens), m.sullman@cranfield.ac.uk (M.J.M. Sullman). http://dx.doi.org/10.1016/j.aap.2014.06.021 0001-4575/© 2014 Elsevier Ltd. All rights reserved.
aggression directly, driving anger has been found to relate to selfreported crashes (Deffenbacher et al., 1994; Sullman and Stephens, 2013) and near crashes (Underwood et al., 1999) and to increase the likelihood of crashing in driving simulator scenarios (Deffenbacher et al., 2003; Stephens, Trawley, Madigan & Groeger, 2013; Stephens and Groeger, 2011). The majority of self-report studies on the expression of driving anger have used the Driving Anger Expression Inventory (DAX; Deffenbacher et al., 2002). The most commonly used version of the DAX comprises 49-items and four factors, which are: verbal aggressive expression (VAE) – verbally expressing anger (e.g. yelling at the other driver); personal physical aggressive expression (PPAE) – using themselves to express anger (e.g. make hostile gestures); use of the vehicle to express anger (UoV) (e.g. drive a little faster); and adaptive/constructive expression (A/C) – constructive behaviours the driver can engage in (e.g. try to ignore it). These four factors have been found to have good internal reliability (˛ = 0.80–0.90; Deffenbacher et al., 2002). A total aggressive expression (TAE; summed scores for VAE, PPAE and UoV) is also used. The validity of the DAX has also been shown through correlations with self-reported driver aggression (see Deffenbacher et al., 2007). The DAX has been administered to drivers from France (Villieux and Delhomme, 2010), Malaysia (Sullman et al., submitted for publication), Romania (Sarbescu, 2012), Serbia (Jovanovic´ et al., 2011), Spain (Herrero-Fernández, 2011), Turkey (Esiyok et al., 2007) and the USA (Deffenbacher et al., 2002). Surprising, the
170
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
DAX has not been administered in either of the two countries in the British Isles1 (Ireland or the UK). Recently, Özkan et al. (2011) reported a cross-cultural study comparing aggressive driving behaviour among British, Dutch, Finnish and Turkish drivers. Aggressive expression was measured using a bespoke index, which did not include adaptive mechanisms to reduce anger. This omission is important as research using the DAX shows adaptive mechanisms are the most commonly reported response to driver anger. Furthermore, it is equally important to understand the trait or state factors influencing the use of positive coping strategies. There have been a number of different structures used with the DAX. Initial development of the DAX produced a 53-item, five factor model, which included an additional four item factor labelled displaced aggression. However, low reliability of the factor led the authors to advise against use of this factor in future research (Deffenbacher et al., 2002). Recently, using Confirmatory Factor Analysis (CFA), Herrero-Fernández, 2011 found support for the longer five-factor model, which also included the displaced aggression factor. To achieve acceptable model fit, three items were dropped from the analysis. However, a 49 or 53 item scale is lengthy to administer, especially if used in conjunction with other scales. Further, given statistical assumptions regarding ratios of questionnaire items to number of participants, a large sample is required to appropriately analyse the longer form of the DAX. Other researchers have reduced the number of factors in the model. For instance, Sarbescu (2012) reported a 30-item threefactor model, which combined verbal and physical aggression into one factor. However, these results are contentious as the researchers performed Principal Components Analysis and CFA on the same data using a relatively small sample. In contrast, using data from French drivers, Villieux and Delhomme (2010) reported a three-factor solution using only 11 of the original 49items. These researchers removed all personal physical aggression expression items, but found a three factor solution that included: adaptive/constructive, Verbal aggressive expression and use of a vehicle to express anger. Nonetheless, some research has supported the four-factor model, but only obtained acceptable fit after deleting one or two items and allowing a large number of errorpairs to co-vary (Sullman et al., 2013; Sullman et al., submitted for publication), which suggests some degree of redundancy across the 49-items. Although the DAX has been validated across a variety of different driving populations there is some inconsistency regarding the driver characteristics that relate to the propensity to become aggressive while driving. Some researchers have found that females report more adaptive/constructive means of reducing anger (Esiyok et al., 2007; Jovanovic´ et al., 2011). Males tend to engage more often in personal physical aggressive expression (Dahlen and Ragan, 2004; Esiyok et al., 2007). However, others have found no sex differences (Villieux and Delhomme, 2010). It is unclear whether this represents underlying differences in driving populations, sampling techniques or is an artefact of the differing factor structures used. Age has consistently been found to be related to driving aggression tendencies. Jovanovic´ et al. (2011) found a significant negative relationship between age and the total of all the aggressive expression items. Furthermore, age was found to be negatively related to personal physical aggressive expression, use of a vehicle and verbal aggressive expression, while adaptive/constructive expression was positively related to age (Deffenbacher et al., 2007). Total aggressive expression and verbal aggressive expression were also both negatively related to age amongst Turkish taxi drivers (Sullman et al., 2013). In fact, with the exception of Moore and Dahlen (2008), most
1 The British Isles is the geographic term for the group of islands off the north western coast of continental Europe, which includes the UK and Republic of Ireland.
Table 1 Descriptive variables. Variable
Mean (SD)
Range
Annual mileage (miles) Mean age (years) Average length of licence (years) Minor crashesa Major crashesa Near-missesa Summonsa Lost concentrationa Lost controla Experienced road ragea Engaged in road ragea
9543 (7081) 37.85 (12.93) 17.15 (12.63) 0.18 (0.67) 0.03 (0.21) 1.53 (3.38) 0.17 (1.15) 6.22 (20.07) 0.94 (2.37) 1.81 (2.40) 2.10 (2.59)
0 –200,000 18 –75 0 –55 0–10 0–2 0–50 0–25 0–365 0–30 0–20 0–20
a
Past twelve months.
previous research has found some relationships between age and the DAX subscales (Esiyok et al., 2007; Herrero-Fernández, 2011; Villieux and Delhomme, 2010). Aside from the inconsistencies in item inclusion mentioned above, the length of the 49-item DAX makes it unwieldy, reduces the ability to combine it with other measures and necessitates large sample sizes. Therefore, the main aims of the present research were to revise and reduce the DAX and to validate these scales using a sample of drivers from the British Isles. A further aim was to examine age and gender differences across the sample to provide additional evidence about the contribution of these to driver aggression. 2. Method 2.1. Participants A total of 551 drivers from the British Isles (males = 49%) completed the online questionnaire. Of these, 52% were from the UK and 48% from the Republic of Ireland. Participants’ had an average age of 38 (SD = 13) years, had held a driving licence for 17 (SD = 13) years and had an annual mileage of approximately 9543 (SD = 7081) miles (Table 1). 2.2. Materials The 49-item four-factor DAX was used to measure how drivers express their anger whilst driving (Deffenbacher et al., 2002). This version was chosen as it is the most commonly used form of the DAX and the developers of the DAX had advised against the longer 53-item version. For each of the 49-items, respondents rate, on a four-point scale (1 = almost never, 4 = almost always), the frequency of each potential reaction to feeling angry while driving. The 14-item Driving Anger Scale (DAS) provided an overall measure of trait driving anger (Deffenbacher et al., 1994). The DAS and the DAX have demonstrated moderate correlations (Deffenbacher et al., 2002) and as such the measure was included to support validation of the new forms of the DAX. Participants rate the level of anger elicited by each of the 14 potentially anger inducing situations on a five-point scale (1 = not at all, 5 = very much). Higher scores on the DAS indicate stronger tendencies towards anger while driving. The DAS short-form has demonstrated good internal consistency (˛ = 0.80; Deffenbacher et al., 1994). The validity of the DAS has been shown through correlations with Spielberger’s (1988) Trait Anger Scale (Deffenbacher et al., 1994; Sullman and Stephens, 2013). Six items from the Driving Survey (Deffenbacher et al., 2002), were used to measure how many times, in the last year, drivers had: been fined or prosecuted for a driving offence (excluding parking tickets), lost concentration, lost control of their vehicle, experienced a near-miss, had a minor crash or a major crash.
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
171
Fig. 1. CFA model of the DAX revised (25-items) and the DAX short (15-items). Dashed lines represent items in 25 item model that are removed for the 15 item model. The factor loadings are presented in Table 1.
Drivers also reported how often in the past year they had experienced (four items) or initiated road rage incidents (four items). These measures were included to validate aggressive tendencies against self-reported detrimental and aggressive behaviours. Previous research has shown positive relationships between scores on the DAX and crash-related conditions (Deffenbacher et al., 2002; Sullman et al., 2013; Sullman et al., submitted for publication) and self-reported road rage (Sullman et al., submitted for publication). 2.3. Procedure A link to the online questionnaire was distributed to drivers via social media sites, on-line advertising websites and emails sent to staff and students at Cranfield University and University College Cork. Drivers were informed that the study was voluntary, responses were anonymous and that anyone who held a current driving licence and had driven at least once in the past six months could complete the survey. The ethics committees of both universities approved the study. 2.4. Statistical analysis EQS for windows (version 6.1) was used to conduct Confirmatory Factor Analysis (CFA) on the DAX scales. The robust method of maximum likelihood (ML) was utilised due to the violation of multivariate normality across questionnaire responses and as Mardia’s normalised estimate was >5. Goodness-of-fit indices included the Santorra-Bentler Scales Chi-Squared (S-B2 ), S-B2 /df, adjusted Comparative Fit Index (CFI), Root Mean Square Error of Approximation (RMSEA). Acceptable model fit was indicated by an S-B2 /df < 5; adjusted CFI of 0.90 or greater and an RMSEA of 0.06 or less (Hu and Bentler, 1999). The confidence interval (C.I.) reporting a 90% interval surrounding the RMSEA was also examined. 3. Results 3.1. Descriptive variables Table 1 shows the mean responses on the driver behaviour survey questions. The most common of the six behaviours was losing concentration while driving, with some drivers reporting this to
be an almost daily occurrence. Overall crash involvement was low, averaging less than one minor or major crash per year. 3.2. Driving anger expression Table 2 shows the means and standard deviations of the DAX items2 . Items from the adaptive/constructive factor received the three highest scores. In order, these were: “I just try to accept that there are bad drivers on the road”; “I decide not to stoop to their level”; and “I just try to accept that there are frustrating situations”. The three lowest scored items all belonged to the personal physical aggressive expression factor. These were: “I try to get out of the car and have a physical fight with the other driver”; “I bump the other driver’s bumper with my own”; and “I try to get out of the car and tell the other driver off”. Overall, drivers reported relatively low aggressive responses to anger, scoring an average of 1.50 (SD = 0.37) out of a possible 4. Adaptive/constructive expressions were the most commonly reported, followed by Verbal aggressive expression, and then use of a vehicle to express anger. personal physical aggressive expressions were the least commonly reported. The order of factor means was the same as that reported in previous research (Sullman et al., 2013; Sullman et al., submitted for publication; Villieux and Delhomme, 2010). 3.3. Reduction of the DAX items Split half validation was used to confirm shorter versions of the DAX. A random sample of 200 cases was selected from the full data set and subjected to Principle Components Analysis (PCA). The Kaiser–Meyer–Olkin measure of sampling adequacy was 0.87 and Bartlett’s test of sphericity was significant (2 (300) = 2894, p < 0.0001), indicating the data were suitable for factor analysis. Initially, all 49-items were analysed using the direct Oblimin method of rotation, resulting in thirteen factors with eigenvalues >1. The direct Oblimin method was used as the DAX factors
2 The total DAX scores did not differ between drivers when compared across UK and Ireland.
172
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
Table 2 DAX item means. DAX items (N = 551) M (SD) Adaptive/constructive expression (˛ = 0.89) Accept there are bad drivers on the road 42 30 Not stoop to their level 45 Accept there are frustrating situations 36 Tell myself it’s not worth getting involved 49 Pay closer attention to others’ driving to avoid accidents 23 Pay closer attention to being a safe driver 48 Tell myself to ignore it 25 Think things through before I respond Tell myself it’s not worth getting mad at 29 Think of positive things to do 35 Think of positive solutions to deal with the 26 situation 32 Turn on the radio/music to calm down Think about things that distract me from 47 the frustration 24 Think about things that distract me from thinking about Take deep breaths to calm down 44
2.30 (0.58) 2.71 (0.93) 2.70 (0.97) 2.69 (0.91) 2.66 (0.98) 2.64 (0.94)
Verbal aggressive expression (˛ = 0.87) Shake my head at the other driver 37 Make negative comments about the driver 6 aloud Call the other driver names aloud 5 Give the other driver dirty looks 40 Glare at the other driver 11 Make negative comments about the driver 39 under my breath 28 Swear at the other driver aloud 31 Swear at the other driver under my breath Call the other driver names under my 14 breath Think things like “where did you get your 43 licence?” 38 Yell at the other driver Yell questions like “where did you get your 9 licence?”
2.02 (0.58) 2.40 (0.88) 2.34 (0.89)
2.61 (0.91) 2.54 (0.94) 2.51 (0.91) 2.48 (0.93) 1.88 (0.87) 1.95 (0.87) 1.78 (0.89) 1.76 (0.82) 1.76 (0.86) 1.76 (0.89)
2.20 (0.93) 2.13 (0.92) 2.12 (0.87) 2.11 (0.84) 2.07 (0.99) 2.04 (0.92) 2.00 (0.87) 1.82 (0.95) 1.60 (0.86) 1.39 (0.76)
Use of vehicle to express anger (˛ = 0.87) Flash my lights at the driver 33 Drive a little faster 3 46 Slow down to frustrate other driver Block the other driver from doing what 16 he/she wants to do Speed up to frustrate other driver 15 Drive right up on the other driver’s bumper 2 Drive a lot faster 27 Do to drivers what they did to me 22 Leave my lights on full in their mirrors 19 Follow right behind for a long time 7 Try to cut in front 4
1.33 (0.40) 1.75 (0.83) 1.60 (0.69) 1.47 (0.76) 1.36 (0.61)
Personal physical aggressive expression (˛ = 0.90) Give the finger 1 Make other hostile gestures 34 Go crazy behind the wheel 18 Shake my fist 12 Roll down the window to communicate my 10 anger 21 Try to scare the driver Try to force the driver to the side of the 20 road 13 Stick my tongue out 8 Try to get out of the car and tell the other driver off 17 Bump the driver’s bumper with my own Try to get out and have a physical fight 41
1.16 (0.37) 1.33 (0.69) 1.30 (0.65) 1.21 (0.60) 1.20 (0.59) 1.17 (0.56)
Total
1.50 (0.37)
1.27 (0.62) 1.26 (0.55) 1.24 (0.57) 1.23 (0.55) 1.18 (0.55) 1.15 (0.50) 1.12 (0.42)
1.14 (0.51) 1.11 (0.47) 1.10 (0.44) 1.08 (0.37) 1.06 (0.35) 1.05 (0.33)
have been shown to be correlated (Sullman et al., 2013; Sullman et al., submitted for publication). A parallel analysis, which outlines minimum eigenvalues for factor retention, suggested a four-factor solution that accounted for 60% of the variance. Several iterations of the PCA were performed to inform item elimination. Items were excluded if they did not conceptually fit within a factor, if their factor loading was less than 0.40 or if they cross-loaded onto another factor (>0.40). This process resulted in a 25-item scale that retained the four dimensions: Adaptive/constructive; verbal aggressive expression; use of a vehicle; and personal physical aggressive expression (see Table 3). Adaptive/constructive expression was reduced from 15 to 10 items, with items, 24, 25, 32, 44 and 47 excluded. These all involve active distractions such as thinking about non-traffic related situations, turning on the radio or focusing on calm breathing. Personal physical aggressive expression was reduced from 11 to 5 items (excluding items 1, 12, 13, 17, 18 and 34). Item 18 “I go crazy behind the wheel” has previously been identified as problematic and removed from the DAX analysis (Esiyok et al., 2007; Sullman et al., 2013; Villieux and Delholmme, 2010). The Verbal aggressive expression factor was reduced from 12 to 5 items by excluding items (removing items 9, 11, 31, 37, 39, 40, 43) involving non-verbal communications (i.e., giving a dirty look) and conceptually similar items. For example, Item 9 “I yell questions like where did you get your licence” is similar to Item 38 “I yell at the other driver”. The use of a vehicle to express anger factor was reduced from 11 items to 5 (excluding 3, 4, 19, 33) by removing the lower correlated items that had a similar meaning to another item (e.g., 3 and 27). For example, Item 3 “I drive a little faster” and Item 27 “I drive a lot faster”. Items related to flashing or leaving on the headlights and cutting in front of the driver were also removed. All factors demonstrated good reliability (˛ ranging from 0.75 to 0.88). 3.4. CFA on the revised DAX CFA was conducted on the remaining 351 cases not selected for PCA. The model showed less than adequate fit to the data (S-B2 (269) = 815.94, p < 0.001, S-B2 /df = 3.03; CFI = 0.61; RMSEA = 0.08; 90% C.I. = 0.07:0.08). Lagrange Multiplier Tests revealed four error-pairs to be covaried. As can be seen in Table 3, it made conceptual sense to co-vary each of the error-pairs. Three of these were from the adaptive/constructive factor. Items 26: “Think of positive solutions to deal with the situation” and Item 35 “Think of positive things to do”; Item 23 “Pay closer attention to being a safe driver” and Item 49 “Pay closer attention to other’s driving to avoid accidents”; and, Item 30 “Not stoop to their level” and 36 “Tell myself it’s not worth getting involved”. The other error-pair was from the verbal aggressive expression factor and both involved making negative comments aloud: Item 5 “Call the other driver names aloud” and Item 6 “Make negative comments about the driver aloud”. When the model was respecified, allowing the errors for these conceptually similar items to co-vary, it produced an acceptable fit (S-B2 (265) = 363.23, p < 0.001, S-B2 /df = 1.37; CFI = 0.92; RMSEA = 0.04; 90% C.I. = 0.03:0.05), with acceptable internal consistency (Rho = 0.81) and composite reliability scores >0.80 for all factors. Within method convergent validity was also demonstrated by statistically significant regression coefficients and 96% of the loadings being >0.40 (Fig. 1). 3.5. Developing the DAX-short As a 25-item scale may still be too long in some studies, the revised DAX was subjected to further item reduction to obtain the shortest possible measure. The same 200 cases were used that were used to develop the 25-item measure. Items were removed that were conceptually similar, retaining the item with the largest
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
173
Table 3 Factor loadings for the DAX revised and DAX short (PCA, N = 200) and (CFA, N = 351). Revised (25-item) PCA Factor 1: adaptive/constructive Item 26e1 Think of positive solutions to deal with the situation 48 Tell myself to ignore it e4 36 Tell myself it’s not worth getting involved e1 Think of positive things to do 35 Pay closer attention to others’ driving to avoid accidents 49e3 Accept there are frustrating situations 45 Tell myself it’s not worth getting mad at 29 42 Accept there are bad drivers on the road 23e3 Pay closer attention to being a safe driver e4 30 Not stoop to their level Factor 2: Personal physical aggressive expression Items Try to get out and have a physical fight 41 Try to get out of the car and tell the other driver off 8 Bump the driver’s bumper with my own 17 Try to scare the driver 21 Roll down the window to communicate my anger 10 Factor 3: Verbal aggressive expression Items 5e2 Call the other driver names aloud Make negative comments about the driver aloud 6e2 Swear at the other driver aloud 28 Yell at the other driver 38 Swear at the other driver under my breath 31 Factor 4: Use of vehicle to express anger 7 22 15 2 27
Follow right behind for a long time Do to drivers what they did to me Speed up to frustrate other driver Drive right up on the other driver’s bumper Drive a lot faster
33% variance ˛ = 0.88 −0.75 −0.75 −0.74 −0.74 −0.73 −0.71 −0.70 −0.69 −0.67 −0.67 12% variance ˛ = 0.75 0.93 0.87 0.86 0.59 0.49 9% variance ˛ = 0.84 0.83 0.80 0.75 0.72 0.62 6% variance ˛ = 0.74 −0.65 −0.55 −0.44 −0.40 −0.36
Short (15-item) CFA ω = 0.87 0.45 0.89 0.78 0.44 0.34 0.82 0.73 0.69 0.34 0.69 = 0.91 0.95 0.89 0.77 0.67 0.79 = 0.83 0.72 0.71 0.85 0.73 0.43 ω = .82 0.83 0.63 0.77 0.69 0.63
PCA
CFA
39% variance ˛ = 0.87 −0.76 −0.83 −0.82
ω = 0.84 0.44 0.88 0.72
−0.82 −0.78
0.80 0.71
15% variance ˛ = 0.75 0.93 0.93
= 0.89 0.88 0.89
0.62 0.63 9% variance ˛ = 0.81
0.67 0.80
0.83 0.78 0.74
0.70 0.87 0.72
7% variance ˛ = 0.78
ω = 0.73
0.88
0.75
0.73 0.75
0.60 0.70
= 0.81
e1, e2, e3, e4: error pairs covaried during the CFA; ˛ = Cronbach’s alpha; ω = composite reliability.
contribution. This process resulted in a 15-item scale (see Table 3), which demonstrated good reliability (˛ ranged from 0.75 to 0.86). CFA confirmed the 15-item four-factor model adequately fitted the data (S-B2 (84) = 131.32, p < 0.001, S-B2 /df = 1.56; CFI = 0.91; RMSEA = 0.04; 90% C.I. = 0.03:0.05). Internal consistency was satisfactory (Rho = 0.82) and composite reliability scores were >0.70 for all factors. Within method convergent validity was also demonstrated by statistically significant regression coefficients and all of the factors loadings were >0.40. The factors derived from the DAX revised (25-item) and DAXshort (15-items) were strongly correlated (adaptive/constructive: r (549) = 0.95, p < 0.001; verbal expressions of anger: r (549) = 0.97, p < 0.001; use of a vehicle to display anger: r (549) = 0.88, p < 0.001). The total aggressive expression scores were also strongly correlated (r (549) = 0.97, p < 0.001). As both scales (15-item and 25-item) produced similar findings, to avoid extensive repetition only the results of the most parsimonious version (i.e., 15-item) are presented below. 3.6. Intercorrelations among variables Adaptive/constructive aggression, verbal aggressive expression and trait driving anger scores (DAS scores) were all normally distributed. However, use of a vehicle, personal physical aggressive expression and consequently, total aggressive expression were positively skewed and were thus analysed using non-parametric tests. Table 4 shows that the DAX factors shared moderate correlations (rs 0.23–0.38) indicating related but separate constructs. Verbal aggressive expression scores were strongly related to the total aggressive expression score, suggesting that a large
proportion of the variance in the total aggression score is from this factor. Thus, using the DAX as a four-factor measure (rather than a positive/negative structure) may be more appropriate in understanding a driver’s tendency for different types of aggressive expression. DAS scores were moderately, positively related to verbal aggressive expression, use of a vehicle to express anger and personal physical aggressive expression and negatively related to adaptive/constructive means of dealing with anger, demonstrating the relationship between driving anger and aggressive expressions of anger. Likewise, scores on verbal aggressive expression, use of a vehicle and personal physical aggressive expression decreased with age, whereas adaptive/constructive factor scores increased with age. 3.7. DAX by sex The DAX scores were compared across sex using independent samples t-tests on parametric variables and Mann–Whitney U on non-normally distributed variables. Table 5 shows there were no differences in average scores between males and females for the adaptive/constructive means of dealing with anger, nor for verbal aggressive expression or total aggressive expression scores. Males scored higher for use of a vehicle to express anger. 3.8. DAX by crashes Mann–Whitney U tests were conducted to compare scores on adaptive/constructive means of dealing with anger, verbal aggressive expression, use of a vehicle to express anger and total aggressive expression between drivers who, during the last 12 months, had and those who had not: received a traffic ticket
174
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
Table 4 Intercorrelations between variables (N = 551).
1. Age 2. Years licensed 3. Annual mileage 4. DAS 5. A/C 6. VAE 7. UoV 8. PPAE 9. TAE Means (SD)
1
2
3
– 0.92*** 0.03 −0.12** 0.15*** −0.10* −0.13* −0.04 −0.13** See Table 1
– 0.17*** −0.12** 0.13** −0.07 −0.10* −0.04 −0.10*
– 0.06 0.06 0.04 0.10* −0.01 0.03
4
5
6
7
8
9
– −0.26*** 0.41*** 0.31*** 0.19*** 0.40*** 2.80 (0.66)
– −0.38*** −0.33*** −0.28*** −0.43*** 2.47 (0.73)
– 0.31*** 0.23*** 0.91*** 2.00 (0.78)
– 0.37*** 0.58*** 1.25 (0.45)
– 0.43*** 1.11 (0.37)
– 1.45 (0.42)
A/C, adaptive constructive; PPAE, personal physical; VAE, verbal; UoV, use of vehicle; TAE, total aggressive expression; DAS, driving anger scale. Upper panel: Pearson’s correlation coefficient; lower panel Spearman’s Rho. * p < .05. ** p < .01. *** p < .001.
(excluding parking tickets); lost concentration while driving; lost control of the vehicle; experienced a near-miss; had a minor crash; or a major crash. One driver had been licensed for less than a year and was thus excluded. Yes or no membership was determined by metric responses on each crash related condition (0 = no; >0 = yes). Several differences emerged between drivers who reported crashes or crash-related conditions over the past year and those who did not (Table 6). Verbal aggressive expression scores were reliably higher for drivers who had also reported losses of concentration (N = 398) near-misses (N = 326), and a major crash (N = 16), when compared to those who had not (N = 152; N = 224; N = 534). Use of the vehicle to express anger and personal physical aggressive expression scores were also higher for drivers involved in a major crash. Drivers who reported more road-rage type behaviours (N = 340) scored higher on verbal aggressive expression, use of a vehicle and personal physical aggressive expression than those who had not (N = 210). In contrast, adaptive/constructive means of dealing with anger was higher for those who had not initiated road rage interactions.
4. Discussion The present research developed and validated a refined and short version of the DAX. Split half validation confirmed both the revised 25-item and short (15-item) scales, which both retained the original four factors of: adaptive/constructive means of dealing with anger, verbal aggressive expression, use of vehicle to express anger and personal physical aggressive expression (Deffenbacher et al., 2002). As both scales (25-item and 15-item) produced similar findings, to avoid extensive repetition only the results of the 15-item scale are presented. As the goal in reduction analysis is to find the most parsimonious model, the 15-item DAX provides a good balance between fitting the data well while using a few parameters. However, the 25-item model, produced adequate fit and allows researchers to have a reduced scale with additional items should they desire to use them. Prior to reduction analysis, the pattern of item means were checked for consistency across previously published research. The
means for the 49-item scale were similar to those found using drivers from Malaysia (Sullman et al., submitted for publication), France (Villieux and Delhomme, 2010) and Turkey (Sullman et al., 2013). For example, the nine most commonly reported responses were all from the adaptive/constructive factor (i.e., try to accept there are bad drivers on the road; “Decide not to stoop to their level”; and “Try to accept that there are frustrating situations on the road”) and were among the most commonly reported previously (Sullman et al., submitted for publication; Sullman et al., 2013; Villieux and Delhomme, 2010). Furthermore, the items with the lowest means were all from the personal physical aggressive expression factor (i.e., “I try to get out of the car and have a physical fight with the other driver”; “I bump the other driver’s bumper with my own”; and, “I try to get out of the car and tell the other driver off”), which was also consistent with drivers from Malaysia, France and Turkey. This suggests that globally, drivers are most likely to use constructive mechanisms to deal with anger and rarely resort to extreme displays of anger involving personal physical aggression. This might represent the moderate levels of driving anger tendencies reported by the participants. It would be interesting to assess whether drivers reporting a problem with driving anger display similar patterns expression, or whether more extreme behaviours are more common in that sample. The intercorrelations between the DAX factors were all moderate, showing similar patterns to those from other countries using the longer scale (Dahlen and Ragan, 2004; Deffenbacher et al., 2002; Herrero-Fernández, 2011). For example, verbal aggressive expression, use of a vehicle to express anger and personal physical aggressive expression were all positively correlated and the adaptive/constructive factor shared negative relationships with the other three aggressive factors. The moderate nature of these relationships indicates that these factors are related but measure separate components of anger expression. Also, in accord with previous research, the DAS was positively correlated with the three aggressive factors and negatively correlated with adaptive/constructive behaviour (Dahlen and Ragan, 2004; Herrero-Fernández, 2011). Previous research has produced mixed findings regarding sex differences on the DAX factors, but when these have emerged, they
Table 5 Gender differences in the DAX short factors.
Adaptive/constructive Verbal aggressive expression Personal physical aggressive expression Use of vehicle to express anger Total aggressive expression
Males (N = 269) M (SD)
Females (N = 282) M (SD)
2.45 (0.77) 2.01 (0.78) 1.14 (.041) 1.29 (0.53) 1.48 (0.47)
2.49 (0.69) 2.00 (0.77) 1.07 (0.32) 1.19 (0.53) 1.42 (0.35)
t < 1, ns t < 1, ns U, ns U = 34,747, z = −2.01, p < 0.05 U, ns
0.12 2.14 (1.11) 1.43 (0.36) <0.01 0.09 1.88 (1.34) 1.09 (0.26) <05 0.14 0.09 0.03 2.62 (0.81) 2.46 (0.73) NS
2.50 (1.04) 1.99 (0.76) <0.05
2.04 (1.13) 1.22 (0.40) <0.001
0.06 1.61 (0.69) 1.43 (0.35) NS 0.09 1.30 (0.74) 1.08 (0.25) <0.05 0.07 0.05 0.05 2.36 (0.68) 2.48 (0.74) NS
2.12 (0.87) 1.99 (0.76) NS
1.41 (0.73) 1.22 (0.38) NS
0.18 1.51 (0.43) 1.37 (0.37) <0.001 0.09 <0.05 1.12 (0.37) 1.09 0.09 <0.05 0.17 0.03 2.44 (0.71) 2.50 (0.76) NS
2.12 (0.82) 1.83 (0.69) <0.001
1.28 (0.48) 1.20
0.10 1.51 (0.48) 1.41 (0.37) <0.05 0.09 <0.05 1.15 (0.45) 1.08 0.10 0.07 0.07 2.40 (0.67) 2.51 (0.76) NS
2.09 (0.83) 1.95 (0.74) NS
1.30 (0.51) 1.21 (0.41) <0.05
0.14 1.48 (0.43) 1.37 (0.36) <0.01 0.03 1.11 (0.39) 1.10 (0.29) NS 0.08 0.14 0.07 2.44 (0.70) 2.54 (0.81) NS
2.07 (0.80) 1.84 (0.72) <0.01
1.26 (0.47) 1.19 (0.37) NS
0.11 1.71 (0.74) 1.42 (0.35) <0.01 0.16 1.38 (0.82) 1.08 (0.25) <0.001 0.09 0.11 0.08 2.32 (0.81) 2.48 (0.71) NS
2.28 (0.88) 1.97 (0.76) <0.01
1.46 (0.78) 1.22 (0.40) <0.05
0.49 1.59 (0.45) 1.23 (0.21) <0.001 0.18 1.15 (0.44) 1.04 (0.17) <0.001 0.26 0.45 0.31 2.29 (0.70) 2.76 (0.68) <0.001
2.28 (0.80) 1.55 (0.47) <0.001
1.33 (0.52) 1.10 (0.23) <0.001
0.20 1.49 (0.44) 1.34 (0.33) <0.001 0.07 1.12 (0.41) 1.05 (0.19) NS 0.11 1.27 (0.47) 1.17 (0.37) <0.05 0.17 2.09 (0.79) 1.79 (0.69) <0.001 0.04
No M (SD) r p No M (SD)
2.45 (0.72) 2.51 (0.75) NS
Road rage (experienced) YES: N = 394 NO: N = 156 Road rage (expressed) YES: N = 340 NO: N = 210 Traffic Tickets YES: N = 56 NO: N = 494 Loss of concentration YES: N = 398 NO: N = 152 Loss of control YES: N = 206 (0.30) NO: N = 344 Near-miss YES: N = 326 (0.40) (0.35) NO: N = 224 Minor Crash YES: N = 75 NO: N = 475 Major Crash YES: N = 16 NO: N = 534
p No M (SD) Yes M (SD) r p No M (SD) Yes M (SD) p No M (SD) Yes M (SD) Yes M (SD) Yes M (SD)
p
Verbal aggressive expression Adaptive/constructive
Table 6 DAX short factor means across self-reported involvement in crash-related conditions.
r
Use of vehicle to express anger
r
Personal physical aggressive expression
Total aggressive expression
r
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
175
have supported the finding here that males reported higher levels of use of a vehicle to express anger (Esiyok et al., 2007; Deffenbacher et al., 2002). Males tend to exhibit more direct aggression than females (Archer, 2004) and it seems this remains consistent in the driving context. The Use of a Vehicle to express anger factor consisted of items such as tailgating and increasing speed to express anger as well as a more ambiguous item for reciprocating the annoying behaviour of other drivers. It would be interesting to assess how much each of these behaviours creates anger in the first place and the extent to which anger and aggression on the road are exacerbated by this reciprocal relationship. Further, the lack of sex differences for the other factors of the DAX is worthy of more investigation. Given that males have a tendency to score higher on violent aggression measures (Hennessy and Wiesenthal, 2001), a similar finding for these data would have been expected. However, whether this might be an artefact of the low scores for this factor or the gender role anonymity afforded in a motor vehicle (Ellison-Potter et al., 2001) is unclear. A number of relationships were found between DAX factors and self-reported crash-related conditions. Drivers who reported more verbal aggressive expression, use of a vehicle to express anger or personal physical aggressive expression also reported initiating more incidents of road-rage. Total aggressive expression differed significantly across all but one of the crash-related conditions. Drivers who had recently received a traffic fine, lost concentration, lost control of the vehicle, had a near-miss or a major crash also reported more aggressive anger expression while driving. These are broadly similar to the findings made by Sullman et al. (2013) and Sullman et al. (submitted for publication), although the latter did not find a significant relationship between total aggressive expression and traffic fines. Drivers who had received a traffic fine, lost concentration while driving or had a near-miss in the past year had higher verbal aggressive expression scores when compared to drivers who had not. Drivers who reported recent loss of concentration or control while driving and near-misses also had higher means on the use of a vehicle to express anger subscale. The shorter scale was therefore also able to differentiate between those who self-reported crashes and more frequent engagement in crashrelated conditions from those who had not crashed and reported less frequent engagement in crash-related conditions. 4.1. Limitations The study reported here suffers from the usual perceived weaknesses of self-report, such as social desirability bias. This is particularly the case for questions about illegal, dangerous or socially undesirable behaviours, which includes several of the variables measured in this study. However, as the participants’ names were not recorded and they were all assured of total confidentiality and anonymity, the impact of social desirability bias is likely to have been minimal, as has been found by research investigating the topic (e.g. Sullman and Taylor, 2010). Given that the questionnaire was made available on online, the issue of self-selection bias cannot be overlooked. However, self-selection bias may suggest people who perceive themselves as good drivers are likely to opt into the research. Therefore, stronger relationships between aggressive driving tendencies and self-reported crash history may have emerged with a more randomised sampling method. 4.2. Summary and practical implications The most commonly used form of the DAX contains 49 items, which makes it lengthy to complete and difficult to use in conjunction with other scales. Further, the 49-item DAX often requires respecification during Confirmatory Factor Analysis (Sullman et al., 2013; Sullman et al., submitted for publication) meaning a large
176
A.N. Stephens, M.J.M. Sullman / Accident Analysis and Prevention 72 (2014) 169–176
sample size is required to undertake the appropriate analyses. These limitations can make the longer version impractical for researchers to use. In the research reported in this paper, the 49item scale was reduced to a 25-item scale as well as a 15-item DAXshort, and both the reliability and validity of these shortened scales confirmed. The integrity of the structure proposed by Deffenbacher et al. (2002) was retained, resulting in four factors related to anger expression: Adaptive/constructive means of dealing with anger, verbal aggressive expression, use of a vehicle to display anger and personal physical aggressive expression. The findings support previous research that has shown positive relationships between the aggressive forms of anger expression and crash-related conditions, such as traffic tickets, losing control of the vehicle, loss of concentration and minor and major crashes. Further, the present study has demonstrated relationships between self-reported aggressive tendencies and actual recent road rage behaviour. The 25-item and 15-item versions of the scale could also be used to identify whether an intervention has been successful in reducing dysfunctional expressions of anger while driving (i.e., PPAE, UoV and VAE) or increasing the use of adaptive/constructive methods of dealing with anger. This could be undertaken at a group level (e.g., has this media campaign reduced dysfunctional expressions of driving anger) or on a more individual basis (e.g., has this treatment improved the way this individual deals with potentially anger inducing situations. Although they were not using the DAX, the latter benefit has been used previously by Deffenbacher et al. (2000), where they evaluated two methods for treating angry drivers. Furthermore, as this research has shown there are age differences and sex differences for the use of a vehicle, the present research can also be used to target interventions at these groups aiming to promote engagement in adaptive/constructive forms of dealing with driving anger. In summary, the practical applications of the current study are a revised 25-item scale and a short 15-item version of the DAX, which can be readily incorporated into questionnaires to further understand the causes and consequences of aggressive driving behaviour. Research is currently underway to validate the use of the short DAX with other scales to compare responses on the shorter versions with those of the 49-item version. Moreover, both scales are valid representations of the original DAX and are able to differentiate between those who self-report crashes and crash related conditions. Acknowledgements Dr Stephens would like to acknowledge the support of the School of Applied Psychology, University College Cork, Ireland. This study was conducted while she was based there. References Archer, J., 2004. Sex differences in aggression in real-world settings: a meta-analytic review. Rev. Gen. Psychol. 8 (4), 291.
Automobile Association Ireland, 2011. Bouts of road rage more common among http://blog.aaireland.ie/bouts-of-road-rage-more-common-amongmales. males/ (accessed 08.04.14). Dft, 2011. Reported Road Casualties in Great Britain: 2011: Annual Report, Retrieved 05.13 from https://www.gov.uk/government/uploads/system/ uploads/attachment data/file/9277/rrcgb2011-04.pdf Dahlen, E.R., Ragan, K.M., 2004. Validation of the propensity for angry driving scale. J. Saf. Res. 35 (5), 557–563. Deffenbacher, J.L., Deffenbacher, D.M., Lynch, R.S., Richards, T.L., 2003. Anger, aggression, and risky behavior: a comparison of high and low anger drivers. Behav. Res. Ther. 41 (6), 701–718. Deffenbacher, J.L., Kemper, C.C., Richards, T.L., 2007. The driving anger expression inventory: a validity study with community college student drivers. J. Psychopathol. Behav. Assess. 29, 220–230. Deffenbacher, J.L., Huff, M.E., Lynch, R.S., Oetting, E.R., Salvatore, N.F., 2000. Characteristics and treatment of high-anger drivers. J. Counsel. Psychol. 47 (1), 5–17. Deffenbacher, J.L., Lynch, R.S., Oetting, E.R., Swaim, R.C., 2002. The Driving Anger Expression Inventory: a measure of how people express their anger on the road. Behav. Res. Ther. 40, 717–737. Deffenbacher, J.L., Oetting, E.R., Lynch, R.S., 1994. Development of a driving anger scale. Psychological reports 74 (1), 83–91. Ellison-Potter, P., Bell, P., Deffenbacher, J., 2001. The effects of trait driving anger, anonymity, and aggressive stimuli on aggressive driving behavior. J. Appl. Soc. Psychol. 31, 431–443. Esiyok, B., Yasak, Y., Korkusuz, I., 2007. Anger expression on the road: validity and reliability of the driving anger expression inventory. Turk. Psikiyatri. Derg. 18, 231–243. Hennessy, D.A., Wiesenthal, D.L., 2001. Gender, driver aggression, and driver violence: an applied evaluation. Sex Roles 44, 661–676. Herrero-Fernández, D., 2011. Psychometric adaptation of the Driving Anger Expression Inventory in a Spanish sample: differences by age and gender. Transport. Res. Part F 14, 324–329. ´ D., Lipovac, K., Stanojevic, ´ P., Stanojevic, ´ D., 2011. The effects of personJovanovic, ality traits on driving-related anger and aggressive behaviour in traffic among Serbian drivers. Transport. Res. Part F 14, 43–53. Moore, M., Dahlen, E.R., 2008. Forgiveness and consideration of future consequences in aggressive driving. Accident Analysis & Prevention 40 (5), 1661–1666, http://dx.doi.org/10.1016/j.aap.2008.05.007. Özkan, T., Lajunen, T., Parker, D., Sümer, N., Summala, H., 2011. Aggressive driving among British, Dutch, Finnish and Turkish drivers. International Journal of Crashworthiness 16 (3), 233–238, http://dx.doi.org/10.1080/13588265.2010. 536687. Sarbescu, P.S., 2012. Aggressive driving in Romania: psychometric properties of the Driving Anger Expression Inventory. Transport. Res. Part F 15, 556–564. Spielberger, C.D., 1988. State-Trait Anger Expression Inventory, 1st ed. Psychological Assessment Resources, Odessa, FL. Stephens, A.N., Groeger, J.A., 2011. Anger-congruent behaviour transfers across driving situations. Cognition & emotion 25 (8), 1423–1438. Stephens, A.N., Trawley, S.L., Madigan, R., Groeger, J.A., 2013. Drivers Display AngerCongruent Attention to Potential Traffic Hazards. Applied cognitive psychology 27 (2), 178–189. Stephens, A.N., Ohtsuka, K., 2014. Cognitive biases in aggressive drivers: does illusion of control drive us off the road? Pers. Indiv. Differ. 68, 124–129. Sullman, M.J.M., Stephens, A.N., 2013. A comparison of the Driving Anger Scale and the Propensity for Angry Driving Scale. Accident Analysis & Prevention 58, 88–96, http://dx.doi.org/10.1016/j.aap.2013.05.002. Sullman, M.J.M., Stephens, A.N., Kuzu, D., 2013. The expression of anger amongst Turkish taxi drivers. Accid. Anal. Prev. 56, 42–50. Sullman, M.J.M., Stephens, A.N., Yong, M. (submitted for publication). Anger, aggression and road rage behaviour in Malaysian drivers. Sullman, M.J.M., Taylor, J.E., 2010. Social desirability and self-reported driving behaviours: should we be worried? Transport. Res. Part F 13, 215–221. Underwood, G., Chapman, P., Wright, S., Crundall, D., 1999. Anger while driving. Transport. Res. Part F: Traffic Psychol. Behav. 2 (1), 55–68. Villieux, A., Delhomme, P., 2010. Driving anger and its expressions: further evidence of validity and reliability for the Driving Anger Expression Inventory French adaptation. J. Safety Res. 41, 417–422.