Traffic accidents involving pedestrians are a relatively frequent cause of death and injury in Spain. According to official data, in 2016 there were 14147 pedestrians who were involved in traffic accidents. Of them, 389 were killed and 1989 were injured. During 2016, pedestrians were the 21% of the total amount of people death in traffic accidents in Spain. Moreover, about 44% of the total of pedestrians that were run over outside the cities had committed a traffic infraction, while it was about 22% of the cases of pedestrians crashed into the cities committing and infraction (Dirección General de Tráfico, DGT, 2017). Therefore, it seems that pedestrians´ human factor is crucial in traffic accidents involving pedestrians as in other road accidents (Evans, Reference Evans1991).
Within human factor, risky behaviors have been proposed as one of the main predictors of road accidents among pedestrians (Barić, Pilko, & Starcevic, Reference Barić, Pilko and Starcevic2018). Following the well-supported Driver Behavior Questionnaire theoretical model, it has been proposed that there are three types of risky behaviors that are applicable to pedestrians as well as to drivers: violations, errors and lapses (Reason, Manstead, Stradling, & Baxter, Reference Reason, Manstead, Stradling and Baxter1990). Violations are those behaviors in which drivers contravene a traffic rule deliberately, having a motivational basis. They are the most dangerous type of risky behavior, being related to anger and aggressive behavior (Berdoulat, Vavassori, & Muñoz Sastre, Reference Berdoulat, Vavassori and Muñoz Sastre2013). Otherwise, errors are characterized for violating a traffic rule by mistake (with no intention about it), so it is due to misbehavior because of a bad calculation (the intention is not appropriate). Finally, lapses are non-deliberated risky behaviors which are usually due to distractions. Of these tree types of risky behavior, violations are the most related to traffic accidents (Zhang, Yau, & Chen, Reference Zhang, Yau and Chen2013). This theoretical approach has been applied to the study of both drivers´ behavior (Berdoulat et al., Reference Berdoulat, Vavassori and Muñoz Sastre2013; Reason et al., Reference Reason, Manstead, Stradling and Baxter1990; Zhang et al., Reference Zhang, Yau and Chen2013) and pedestrians´ behavior (Elliott & Baughan, Reference Elliott and Baughan2004; Granié, Pannetier, & Guého, Reference Granié, Pannetier and Guého2013). Finally, it has been proposed that risky behavior is a result of the perception of risk and the threshold of risk. Therefore, pedestrians who have a low threshold of risk, and perceive the risk in a situation to be high might adopt safer behaviors (Herrero-Fernández, Macía-Guerrero, Silvano-Chaparro, Merino, & Jenchura, Reference Herrero-Fernández, Macía-Guerrero, Silvano-Chaparro, Merino and Jenchura2016).
Then, predictors of risky behavior should be analyzed in order to prevent this kind of behavior, and ultimately road accidents. In this sense, one of the variables that has been closely related to risky behavior is anger, both in a general context (Ferrer, Maclay, Litvak, & Lerner, Reference Ferrer, Maclay, Litvak and Lerner2017) and in the specific field of driving (Herrero-Fernández & Fonseca-Baeza, Reference Herrero-Fernández and Fonseca-Baeza2017). Anger has also shown strong relationships with aggressive behavior, both in a general context (Spielberger, Krasner, & Solomon, Reference Spielberger, Krasner, Solomon and Janisse1988) and in driving (Deffenbacher, Deffenbacher, Lynch, & Richards, Reference Deffenbacher, Deffenbacher, Lynch and Richards2003). Therefore, anger, aggression and risky behavior are closely related each other. On one hand, it has been proposed that anger and risky behavior could reinforce each other in a kind of feedback process (Berkowitz & Harmon-Jones, Reference Berkowitz and Harmon-Jones2004). On the other hand, in spite of the strong relationship and the apparent theoretical overlap between aggression and risk, it is important to differentiate them. Whereas an aggressive behavior is characterized by the intention of causing harm to other persons (e.g. other drivers or pedestrians), risky behavior does not have this intention, being characterized by making dangerous behaviors which increase crash-risk, such as running red lights (Suhr & Dula, Reference Suhr and Dula2017). Then, the same behavior could be risky or aggressive, depending on the intention of the person who engages in it.
According to the Berkowitz´s theory, anger results from a combination of feelings, cognitions and physiological reactions, which are associated with the intention to punish whoever caused the individual’s anger (Berkowitz & Harmon-Jones, Reference Berkowitz and Harmon-Jones2004). Besides this, it has been proposed that anger is due to both situational variables, such as a provocative situation, and personal variables, such as trait anger (Deffenbacher, Oetting, & Lynch, Reference Deffenbacher, Oetting and Lynch1994). Therefore, situational variables have been explored with the aim of identifying the events which are more likely to generate anger in people. One of the better empirically supported approaches is the original Deffenbacher´s trait driving anger theoretical model, empirically conceptualized in the Driving Anger Scale (DAS). It proposed that there were six types of situations where drivers with higher anger trait were more likely to become angered: Hostile gestures, illegal driving, police presence, slow driving, discourtesy and traffic obstructions (Deffenbacher et al., Reference Deffenbacher, Oetting and Lynch1994). However, this questionnaire has been analyzed in several countries and cultures, obtaining different structures as situations generators of anger. All in all, almost all of the versions share three common factors: Reckless driving, referred to anger provoked as a result of a risky behavior from other road user; hostile gestures, referred to anger provoked because of direct aggressions from other road users such as obscene gestures, and progress impeded, referred to anger provoked as a result of the slowdown of the progress. It has been observed in versions from such different countries as Spain (Herrero-Fernández, Reference Herrero-Fernández2011), Argentina (Escanes & Poo, Reference Escanes and Poo2018) and United Kingdom (Lajunen, Parker, & Stradling, Reference Lajunen, Parker and Stradling1998). These three situations could be applicable to pedestrians, as they as referred to general aspects of traffic. Whereas there are important differences among drivers and pedestrians (for example the anonymity), both of them share other characteristics: risk of suffering an accident (related to anger for reckless driving of other drivers), movement or desire of going from one place to another one (related to anger because of progress impeded), and the possibility of being the target of the aggressive behavior from others (related to hostile gestures). Besides, the interaction of both situational variables and trait anger is on the basis of the Driving Anger Scale model and the general trait anger model (Spielberger et al., Reference Spielberger, Krasner, Solomon and Janisse1988).
On the other hand, age and gender have been explored as demographic variables in their relationship with anger. Regarding age, there is an agreement about the negative relationship between age and anger (Deffenbacher, Lynch, Oetting, & Swaim, Reference Deffenbacher, Lynch, Oetting and Swaim2002; Herrero-Fernández, Reference Herrero-Fernández2011). Regarding gender, there is no agreement about it, so whereas some studies propose that men are more prone to getting angry behind the wheel than women (Vanlaar, Simpson, Mayhew, & Robertson, Reference Vanlaar, Simpson, Mayhew and Robertson2008), some others have found the opposite results (Sullman, Stephens, & Yong, Reference Sullman, Stephens and Yong2014), and some others have not found significant relationship between gender and anger propensity (Deffenbacher et al., Reference Deffenbacher, Oetting and Lynch1994; Herrero-Fernández, Reference Herrero-Fernández2011).
Therefore, it is crucial to develop instruments which allow assessing both risky behavior and anger in pedestrians. In the case of risky behaviors, three instruments have been developed, taking the Drivers´ Behavior Questionnaire theoretical approach (Reason et al., Reference Reason, Manstead, Stradling and Baxter1990). The first one is the Pedestrian Behavior Scale (Moyano, Reference Moyano1997; Torquato & Bianchi, Reference Torquato and Bianchi2010), which assesses violations, errors and lapses. The second one is the Adolescent Road User Behavior Questionnaire (Elliott & Baughan, Reference Elliott and Baughan2004), which assesses both unsafe road crossing behavior and protective planed behavior. Finally, the most recent one is the Pedestrian Behavior Questionnaire (Granié et al., Reference Granié, Pannetier and Guého2013), which is composed of four factors: Transgression, lapses, aggressive behavior and positive behavior. Regarding aggression, to our knowledge this last questionnaire is the only which includes a subscale to measure it. In this case, it is composed of four items based on scales of aggressive driver behaviors (Lawton, Parker, Manstead, & Stradling, Reference Lawton, Parker, Manstead and Stradling1997), and they referred to getting anger and expressing aggression by different ways. Then, there is no instrument which measures specifically anger as emotion in pedestrians, so it would be relevant to develop an instrument properly focused on exploring situations that generate this emotion in pedestrians.
The main aim of the current research was the development of a new psychometric tool to measure the frequency of anger in different situations in pedestrians. This scale will be analyzed in terms of factorial structure, reliability (internal consistency) and validity (convergent, divergent and incremental). Afterwards, the effect of age and gender will be analyzed.
Method
Participants
Initially the self-report questionnaires set was sent by Internet or by paper-and-pencil (see Procedure section) and 600 people from Spanish general population (convenience sampling) answered them. However, in the case of 50 participants, there were some missing data, so they were erased from the data set and the final sample consisted of 550 participants. Of them, 224 (40.73%) were male, 326 (59.27%) were female. The age ranged from 14 to 65 years (M = 27.91, SD = 13.21, Mdn = 22.50). Four groups were made based on the participants´ age for the analysis of the relationship between age and anger: 14–18 years, 19–30 years, 31–45 years, and > 45 years. Finally, 295 participants (53.64%) had a car license. All participants volunteered and signed informed consents guaranteeing the confidentiality of the collected data.
Instruments
Pedestrian Anger Scale (PAS)
Following the theoretical framework of the trait driving anger and the most used questionnaire to measure it (the DAS) (Deffenbacher et al., Reference Deffenbacher, Oetting and Lynch1994), the Pedestrian Anger Scale was developed as a self-report questionnaire to measure the trait pedestrian anger. More concretely, three sources of drivers´ anger were taken into account and applied to pedestrians´ experience: First, impeded progress by others, which implies a slowdown in the movement (in this case, in the walking) regarding the preferred speed. In this case 8 items were designed, representing situations in which pedestrians are forced to reduce the speed of their walking. Second, reckless driving, which is referred to as anger provoked by risky situations generated by other road users. In this case 4 items were designed representing risky situations that could put them into risk of suffering an accident. Third, direct hostility, which is referred to as anger generated as a consequence of offences received from other road users. In this case 3 items were designed representing annoying situations (e.g. being honked at, a loud motor noise nearby). Then, a pool of 15 items was built. In the drafting process of the items it was taken into account the items of the aggressive behaviors factor from the Pedestrian behavior Scale (Granié et al., Reference Granié, Pannetier and Guého2013). The instructions given to the participants were as follows: “Rate from 0 (nothing at all) to 4 (very much) the intensity of anger that the following situations provoke to you”.
Risky Pedestrian Behavior (RPB)
It is a 8-item self-report which measures the frequency of risky behaviors in pedestrians (Herrero-Fernández, Reference Herrero-Fernández2015b). The participants rate the frequency in which they behave in the described ways (e.g. “I cross the road by non-permitted places”) in a four-point Likert scale (0 = never / 4 = always). The internal consistency of the scale was high with the data of the current research (α = .85).
State-Trait Anger Expression Inventory (STAXI–2)
The trait anger scale of the Spanish version of the STAXI–2 (Miguel-Tobal, Casado, Cano-Vindel, & Spielberger, Reference Miguel-Tobal, Casado, Cano-Vindel and Spielberger2001) was applied, in order to measure the level of anger experienced in a general context. This scale is composed of 10 items rated in a four-point Likert scale (1 = Almost Never to 4 = Almost Always), so participants indicate the amount of anger that every situation provokes him / her. Finally, this scale showed a good internal consistency with the current research data (α = .81).
Risk Perception
Risk perception was measured through a set of ten videos. Each one lasted between 12 and 17 seconds. In three of them it appeared a pedestrian behaving in a risky way as a traffic rule was being violated (e.g. crossing the road in a non-permitted place or when the traffic light is red). In the other two it appeared a pedestrian behaving in a safe way according to the traffic rules (e.g. crossing with traffic light in green). In the other three it appeared vehicles circulating in a risky way (e.g. overtaking with continuous line). Finally, in the other two it appeared vehicles circulating in a safe way. In each of the four types of videos, a total score was calculated by adding up the scores of each video, so it resulted in four scores: One for risk perception in situations with pedestrians behaving riskily, another one for risk perception in situations with pedestrians behaving safely, another one for risk perception in situations with vehicles ridding riskily, and another one for risk perception in situations with vehicles ridding safely. The order of presentation of the videos was the same for all the participants, having been randomized previously. The instructions given to the participants were as follows: “Rate the amount of risk that you perceive in the following situations, from 0 (no risk) to 5 (much risk)”. Then, this task assesses the perception of how risky the situations are. This procedure have been used in previous research (Herrero-Fernández, Reference Herrero-Fernández2015b; Herrero-Fernández et al., Reference Herrero-Fernández, Macía-Guerrero, Silvano-Chaparro, Merino and Jenchura2016).
Procedure
The questionnaires were sent to the participants in paper-and-pencil or by Internet (Facebook), so a snowball technique was used to collect the data. This procedure is justified in some research that have shown the equivalence in terms of validity and scores of both paper-and-pencil and Internet methods to gather self-reported data in traffic-related behavior variables (Herrero-Fernández, Reference Herrero-Fernández2015a). Given that the main variable (anger in pedestrians) can be applied to everyone, there were no exclusion criteria to take part in the research.
Data Analysis
Data analyses consisted of four stages. First, a parallel analysis (PA) was carried out in order to explore the factorial structure. In this case, Factor 10.4.01 software was used (Lorenzo-Seva & Ferrando, Reference Lorenzo-Seva and Ferrando2013). The procedure selected for determining the number of dimensions was the optimal implementation of PA (Timmerman & Lorenzo-Seva, Reference Timmerman and Lorenzo-Seva2011), and the method for factor extraction and parameters estimation was the Unweighted Least Squares (ULS). This is the better method when the variables are non-continuous, like in this case as items were responded in a Likert scale. The rotation method to achieve factor simplicity was the Weighted Oblimin, because it was expected that the factors were correlated with each other. The quality of the global fit of the model was assessed through several robust goodness of fit statistics: the χ2 / degrees of freedom (df) ratio, which should be lower than 3 (Carmines & McIver, Reference Carmines, McIver, Bornstedt and Borgatta1981), and the Non-Normed Fit Index (NNFI), the Comparative Fit Index (CFI), the Goodness of Fit Index (GFI) and the Adjusted Goodness of Fit Index (AGFI), whose values should be above .95 (Hu & Bentler, Reference Hu and Bentler1999). The residuals of the model were assessed through the Root Mean Square of Residuals (RMSR), whose value should be lower than Kelley´s criterion (Kelley, Reference Kelley1935). Finally, the simplicity of the structure was assessed through the Bentler´s S simplicity index, with values close to 1 indicating a simple structure (Bentler, Reference Bentler1977).
In a second step, descriptive statistics of both items and factors, reliability analysis (internal consistency through α) were conducted. In the third step, both convergent and divergent validity of the PAS was analyzed by correlating it with the criteria through Pearson´s r. Based on this correlational analysis, a hierarchical multiple regression was conducted in order to check the amount of variance that the anger and risk measures explained of the pedestrians anger by controlling the common variance among them. This was made as an incremental validity test.
In the fourth step age and gender differences in the PAS scores were analyzed through a 2 (gender) x 4 (age) MANOVA. Age was analyzed by creating four groups: 14–18 years, 19–30 years, 31–45 years, and older than 45 years. In this case, both interaction and individual effects of age and gender were tested. Significance and effects sizes (η2) were calculated. In this last case, effect sizes were interpreted according to the Cohen´s criterion, so values under .04 were considered small; between .04 and .14 medium, and above .14 large (Cohen, Reference Cohen1988). After the multivariate contrast, when the univariate tests attained significance, Hochberg´s GT2 Post Hoc test was applied. This is the better choice when sample sizes are quite different among the groups.
Results
First, a PA was carried out to explore the factorial structure of the PAS. The Kaiser-Meyer-Olkin test (KMO = .84) and the Bartlett´s statistic, χ2(105) = 2,308.6, p < .001, showed the adequacy of the sample. The results suggested the existence of four factors, with a good fit of the model: χ2 / df = 2.65; NNFI = .98; CFI = .99; GFI = .99; AGFI = .98. The analysis of residuals supported also the existence of this four factor structure, as the value of the RMSR was lower than the Kelly´s test: RMSR = .049; Kelley´s test = .052. Besides, the factorial solution showed a very simplicity structure, S = .94. Then, it was analyzed the fit of the model by forcing a three-factor solution (according to the original theoretical framework explained before) and a single-factor solution, in order to verify the most parsimonious model. However, both one-factor and three factor solution showed worse fit than the four factor solution: In the case of the three-factor solution: χ2 / df = 4.07; NNFI = .96; CFI = .98; GFI = .98; AGFI = .97; RMSR = .068, Kelley´s criterion = .052. In the case of a single-factor solution: χ2 / df = 12.55; NNFI = .86; CFI = .88; GFI = .87; AGFI = .85; RMSR = .142, Kelley´s criterion = .052. Then, four factor solution was accepted. Considering the items of each factor, the first one was labeled as “Anger because of obstructions or slowdowns caused by other pedestrians” (4 items, 30.76% of the variance, α = .79), the second one was labeled as “anger because of hostility from other drivers” (3 items, 10.80% of the variance, α = .64), the third one was labeled as “anger because of bad conditions of the infrastructure”, (4 items, 9.96% of the variance, α = .62), and the fourth one was labeled as “anger because of dangerous situations caused by vehicles” (4 items, 7.51% of the variance, α = .71). On the whole, the four factors accounted for the 59.00% of the total variance, and the fifteen items of the PAD showed a high internal consistence (α = .83).
Descriptive statistics of both items and factors are detailed in the Table 1. As it can be observed, the mean score of the items ranged from 1.24 (Item 8) to 2.91 (Item 7). Besides, all of the variables (both items and factors) were symmetrically distributed (skewness ranged from 0.00 to |0.90|). On the other hand, factor loadings were from 0.32 to 1.08, so every item has a significant contribution to the validity of its factor. Finally, correlations item-total ranged from .35 (Item 7) to .66 (Item 14), showing a relevant contribution of every items to the reliability of its factor.
Table 1. Descriptive Statistics (Mean, Standard Deviation, Skewness, Factor Loading, and Correlation between Item and Total Scale) of Items and Factors of the PAD
Note. M = Mean, SD = Standard Deviation; Sk = Skewness. R without item = Reliability without item.
In the next step, both convergent and divergent validity of the PAS were analyzed. The results of the bivariate correlations are represented in Table 2. As it can be observed, the PAS factors were significantly related to each other but with medium effect sizes. Besides, the total PAS score was positively and significantly correlated with all the variables except for the low risky situations involving vehicles, although the relationship with general trait anger was higher than with both risky behavior and risk perception.
Table 2. Descriptive Statistics (Mean and Standard Deviations) and Correlation Coefficients (Pearson’s r) among the Variables
Note. OSP = Obstruction / Slowdown caused by other pedestrian; HFD = Hostility from other drivers; BCI = Bad conditions of infrastructure; DSCV = Dangerous situations caused by vehicles; High R. Ped. = Risk perception = high risk situations for pedestrians; Low R. Ped. = Risk perception = low risk situations for pedestrians; High R. Veh. = Risk perception = high risk situations for vehicles; Low R. Veh. =: Risk perception = low risk situations for vehicles; Risky Behav. = Risky behaviour in pedestrians.
*p < .05. *p < .01. ***p < .001.
Based on previous correlations, the incremental validity of the PAS was analyzed through a hierarchical multiple regression. The results are showed in Table 3. As it can be observed, general trait anger was the only predictor which stayed significant in the last step. It supports the incremental validity of the PAS, as the previous significant correlations of risk perception and risky behavior with it were due to common variance shared with general trait anger.
Table 3. Hierarchical Multiple Regression Predicting the Total Score of the PAS (Trait Pedestrian Anger)
Note. Trait pedestrian anger ΔR 2 = .15 (p = .203) in Step 1; ΔR 2 = .15 (p = .024) in Step 2.
*p < .05. ** p < .01.
Finally, differences in the PAS scores by age and gender were analyzed. The first 2 (gender) x 4 (age) MANOVA showed a significant multivariate interaction effect, F(12, 1,488) = 2.53, p = .003, η2 = .02. More specifically, it was a multivariate significant effect by age, F(12, 1488) = 4.60, p < .001, η2 = .04, but not by gender, F(4, 494) = 1.47, p = .209, η2 = .01. Then, univariate effects by age were explored. Significant interaction effects with low effects sizes were obtained for the “Anger because of bad conditions of the infrastructure” factor, F(3, 505) = 3.91, p = .009, η2 = .02, for the “Anger because of obstructions or slowdowns caused by vehicles” factor, F(3, 505) = 4.59, p = .004, η2 = .03, and for the total score in the PAS, F(3, 505) = 3.59, p = .014, η2 = .02. In general, 19–30 year men and older than 45 women were the less anger people, while 19–30 year women were the most anger ones. Regarding the univariate differences by age, the results are showed in Table 4. As it can be observed, there were significant effects in all of the factors except for the second one (“Anger because of hostility from other drivers”), with low to medium effect sizes. In general, the 19–30 years group was the more prone to getting anger as pedestrians, while the older group (> 45 years) were the less one.
Table 4. Univariate Differences by Age Group in Each Way of Trait Pedestrian Anger
Note. Differences by age, being < 23 (a), 23–30 (b), 31–42 (c), 43–53 (d) and > 53 (e), according to the Hochberg´s GT2 Post Hoc test. OSP = Obstruction / Slowdown caused by other pedestrian; HFD = Hostility from other drivers; BCI = Bad conditions of infrastructure; DSCV = Dangerous situations caused by vehicles.
*p < .05. **p < .01. p < .001.
Discussion
Anger experience in pedestrians could be one of the main predictors of risky behaviors in pedestrians, and these last are one of the main predictors of road accidents. Therefore, it is relevant to get psychometric tools which allow measuring both risky behavior and anger in this specific context and activity. Whereas there are some well-validated questionnaires which measure risky behaviors, there is a gap in the options to measure anger. The aim of the current research was to propose a new tool to assess this variable, analyzing its factorial structure and its reliability and validity properties. Furthermore, the effect of age and gender on the amount of anger experienced in pedestrians were analyzed.
The results showed a good fit of the model in a four factor solution according to the PA. These factors showed in general good values of internal consistency (the factors Anger because of hostility from other drivers and Anger because of bad conditions of the infrastructure had slightly lower values of internal consistency than the common criterion, .70, which could be due to the number of items composing them) and they were easily interpretable according to the Driving Anger Scale theoretical approach (Deffenbacher et al., Reference Deffenbacher, Oetting and Lynch1994), especially regarding the three most common situations that have been detected throughout the different cultures as anger generators. Then, the factors were labeled Obstruction / slowdown caused by other pedestrians, Hostility from other drivers, Dangerous situations caused by vehicles, and Bad conditions of infrastructure. The first one was equivalent to progress impeded factor from the DAS, as both of them are referred to the slowdown in the movement because of other road or pavement users. The second one was equivalent to the hostile gestures from the DAS, as both of them are referred to annoying situations which imply a frustration in the road user. The third one was equivalent to the reckless driving from the DAS, as both of them are referred to dangerous situations because of a misbehavior of a driver. More interesting could be the fourth factor. Despite it seems not to have equivalence in the DAS model, it would be related to a dangerous situation because of a bad road or pavement design, which at some point makes pedestrians to behave riskily. Moreover, it has been recently proposed an Argentinean version of the DAS (Escanes & Poo, Reference Escanes and Poo2018), and it includes a factor labelled Bad conditions of the road network, showing a clear similitude. Whereas these authors suggest that this factor could be applicable to underdeveloped cities and countries, the current results suggest that it probably could be applicable to any country. Furthermore, it highlights the importance of designing safe roads and traffic-related environments, as they could have an important role in transportation safety as it is usually explored as one of the three big road safety factors: vehicle, infrastructure and human factor (Evans, Reference Evans1991).
The Pedestrian Anger Scale has also shown good values of convergent and divergent validity, as bivariate correlations evidenced significant associations with both risk measures (perception and behavior) and general trait anger. In this case, the coefficient was higher in the case of the anger measure than in the case of risky behavior and risk perception. The significant but lower relationship with risk measures is congruent with previous driving related research, which show that risk and anger are both theoretically and empirically related, as a same behavior could be risky or aggressive depending on the intention of the person who makes it (Herrero-Fernández & Fonseca-Baeza, Reference Herrero-Fernández and Fonseca-Baeza2017). Finally, when both risk and anger measures were analyzed together as predictors in the hierarchical regression, risk measures lost their significance, showing the incremental validity of the PAS.
In the next part the effect of age and gender in the experience of anger in pedestrians were analyzed. On one hand, the results showed that there were no significant gender differences. It is in the same line with the previous research with pedestrians thorough PBQ, in which no gender differences were observed in aggressive behavior subscale (Granié et al., Reference Granié, Pannetier and Guého2013). Otherwise, in this same study it was observed that men tended to behave more riskily than women. It supports the general traffic-related studies, which shows in general gender differences in risky perception and behavior (Herrero-Fernández, Reference Herrero-Fernández2015b; Herrero-Fernández et al., Reference Herrero-Fernández, Macía-Guerrero, Silvano-Chaparro, Merino and Jenchura2016), but not in aggressive behavior (Deffenbacher et al., Reference Deffenbacher, Oetting and Lynch1994; Herrero-Fernández, Reference Herrero-Fernández2011). This result could be also an additional support of the validity of the PAS regarding the real measurement of anger rather than risk as an associate concept. On the other hand, significant differences have been observed for age, as in general middle-age people (19–30 year) are more prone to getting anger than both youngest (14–18 year) and eldest ones (> 45 year). Previous studies have found negative associations between age and anger, concluding that young people are more prone to getting anger than elder (Deffenbacher et al., Reference Deffenbacher, Lynch, Oetting and Swaim2002; Herrero-Fernández, Reference Herrero-Fernández2011). However, in the current research the youngest group was 14–18 years, while in the rest of the studies normally the youngest group is > 18 years. Therefore, for future research anger in young people and teenagers should be also taken into account in order to analyze and compare them with older participant groups.
These results have several practical implications, for both clinical and road safety fields. Regarding the first one, it has been proposed a new reliable and valid tool to measure anger in pedestrians. Taking into account that anger in very closely related to aggression (Bogdan-Ganea & Herrero-Fernández, Reference Bogdan-Ganea and Herrero-Fernández2018), an anger prone person could be likely engaged in aggressive behaviors against other road users. Then, an efficient assessment and intervention could be crucial for the person. Regarding the road safety viewpoint, it has already been commented the relationship between anger, aggression, risky behavior and crash-related events (Herrero-Fernández & Fonseca-Baeza, Reference Herrero-Fernández and Fonseca-Baeza2017).
Future research should follow the current findings. First, the psychometric properties of the PAS should be verified with other samples. Confirmatory factor analyses should be conducted to verify the factorial structure that has been found with the current sample. Second, the anger propensities of pedestrians with characteristics associated with anger should be analyzed, such as in certain mental disorders or in people with antecedents of aggressive behaviors. Third, knowing the kind of situations that generate anger in pedestrians, specific treatments and interventions could be designed in order to desensitize them. It would contribute to both a more peaceful coexistence among pedestrians and a reduction of accidents in this collective. Fourth, anger should be analyzed in other road users, as bicyclists (Marín Puchades et al., Reference Marín Puchades, Prati, Rondinella, De Angelis, Fassina, Fraboni and Pietrantoni2017). It would allow analyzing the interaction among every kind of road users regarding anger experience. Finally, some important variables such as rumination or displaced aggression, which have been related to both general anger (Denson, Pedersen, & Miller, Reference Denson, Pedersen and Miller2006) and anger in specific fields as driving (Herrero-Fernández, Reference Herrero-Fernández2013) should be explored in relation to trait pedestrian anger. Hypothetically, both rumination and displaced aggression could explain why some people could behave aggressively in other context than a pedestrian as a result of having being angered as pedestrians.
Finally, the current research has some limitations that should be taken into account for future research. First, this is the first study regarding the trait pedestrians anger and just a pilot study of the PAS. Further research is needed to confirm the factorial structure. Given that the data have been analyzed only through a PA, confirmatory factor analyses should be conducted with different samples to verify the internal structure and other aspects such as the invariance of the instrument by different variables (e.g. gender, age groups, etc.) and some other more sophisticated statistics, such as the construct reliability (as an internal consistency measure) and the average variance extracted (as a validity measure). Besides, considering the relatively low internal consistency values of two of the factors (Anger because of hostility from drivers and Anger because of bad conditions of the infrastructure), due probably to the low number of items, and relatively low factor loadings of some items (two items were under .40) psychometric properties of the scale must be replicated. However, the global scale showed a high value, which suggests that at least the instrument could be used to measure trait pedestrian anger as a whole. Second, the study was based on self-report instruments. Despite this methodology is very extended in social sciences and have demonstrated to be trust-worthy in Traffic Psychology (Lajunen & Summala, Reference Lajunen and Summala2003), the results should be completed with research using other kind of measures, such as behavior observation or psychophysiological studies related to anger generating situations. Third, risk perception has been measured though an instrument (set of videos) that are not properly validated. Despite it has been used in previous similar research (Herrero-Fernández, Reference Herrero-Fernández2015b; Herrero-Fernández et al., Reference Herrero-Fernández, Macía-Guerrero, Silvano-Chaparro, Merino and Jenchura2016), it should be validated in future research to ensure its reliability and validity.
