3.1. Descriptive Analysis
provides the descriptive statistics based on age and gender groups, including the frequency and percentage of corresponding population. The results of Chi-Square tests were also presented in the Table 2
. Regarding daily mobile phone use, 75 male (34.2%) and 58 female (34.5%) participants responded they spend about 2–4 h per day on mobile phone use which were groups with the largest proportions in each gender population, followed by 4–6 h (32.4% and 32.7%, respectively). According to the Chi-Square test, these two factors were independent of each other. That is to say the difference in daily mobile phone use between males and females was not statistically significant (p
> 0.05). Indeed, the proportions for males and females were quite comparable within each pairs. In contrast to gender, the daily mobile phone use among age groups were somewhat different (p
< 0.05): compared to the younger participants, the older ones spend much less time on device use. In other words, the result showed that young adults were the population of the most frequent users of mobile phones. Thus, to some extent, distracted crossing was more likely to happen among young adults.
Reported main usage of mobile phone on a daily basis was similar between males and females (p > 0.05). Participants spent the most time on social applications. Indeed, a variety of social apps are prevalent in China due to their entertainment and communication functions, such as Wechat, QQ and Sina Microblog. Among the age groups, the percentage for younger participants on social apps option was greater than the older, 41.4% of 41–60 year olds chose “others”, and none of these participants chose “play games” category.
As for whether they had experience of using mobile phones while crossing in past two weeks, over half of males (50.7%) and females (56.5%) reported “Yes.” Although the proportion for females to do so was somewhat higher, but no statistically significant difference was found (p > 0.05). In contrast, differences emerged across the age groups on previous experience (p < 0.05): the younger participants were more likely to engage in the distracted behavior, while 21 (72.4%) of over 41 year olds did not have such experience.
In addition, this study also considered the main usage of device during street crossing among pedestrians in China, the most frequently chosen option for male (49.5%) and female (54.7%) participants was phone calls, followed by social apps (23.4% and 31.6%). One previous study conducted by Byington and Schwebel [24
] found that the most common reason for pedestrians using mobile internet while crossing was related to the use of social apps (23.9% of 92 participants). Just from this point of view (i.e., mobile phone usage involving internet), the present study was consistent with the previous one. Furthermore, females were less likely to be distracted by playing games than males. The result of Chi-Square test showed that the difference across age groups was statistically significant (p
< 0.05). Specifically, over half of participants in group 31–40 years and over 41years reported that they used the device for voice calls (63.2% and 75.0%) while crossing street, followed by social apps related usage (26.3% and 12.5%). It seems that most respondents who using mobile phones while crossing due to communication-related reasons rather than entertainment-related requirements.
3.2. Differences in Constructs
Before comparing the differences in constructs and using a binary logistic regression model to predict the past behavior (i.e., the behavior of using a mobile phone while crossing the street before), factor extraction was conducted. First, an exploratory factor analysis (EFA) was performed to identify the construct validity of 7 components in present study (KMO = 0.895, Bartlett’s test p
< 0.001). A principal component analysis with fixed number of factors was applied to reduce dimensions of all 19 items, then 7 identified factors which were consistent with the factors previously defined within TPB framework accounted for a total of 75.18% of the variance. In addition, the varimax rotation method was used to display factor score coefficient matrix and used these factors as 7 independents (i.e., 4 standard and 3 extended constructs) for logistic regression analysis. Table 3
presents the details of factor loadings.
All aggregate measures were compared between males and females, respectively. Due to the violation of homogeneity of variance, two methods were used here, MANOVA (including attitudes, intention, MPI and situation) and t
-test (including SN, PBC and safety awareness). The results of MANOVA analysis indicated that no significant multivariate main effect was found for gender, with F
(4, 382) = 0.868, p
> 0.05, Wilk’s Lambda = 0.991. On the other hand, by using t
-test, significant difference in SN and safety awareness was found, but not in construct PBC. The result suggested that males had higher scores than females on both constructs (i.e., 1.566 vs. 1.417, 2.160 vs. 1.854). Consequently, it could be concluded that males perceived a slightly greater favorable norm compared to their counterparts, at the same time, they were also more sensitive to traffic risk results from distracted crossing behavior. Such phenomenon may be attributed to the role of males and females in mundane life. The differences in constructs between participant groups were detailed in Table 4
Similar to gender, difference in constructs among daily mobile phone use groups was examined. In terms of MANOVA results (including SN, PBC, intention and MPI), a significant multivariate main effect was found, F
(12, 1006) = 4.709, p
< 0.05, Wilk’s Lambda = 0.865. Moreover, significant univariate main effects were found for intention, F
(3, 383) = 8.415, p
< 0.001, η2
= 0.062, and MPI, F
(3, 383) = 14.288, p
< 0.001, η2
= 0.101. Subsequently, pairwise comparisons were conducted for the two constructs respectively, and the results were similar: those participants who spend more hours on mobile phone use per day had higher scores than the participants who spend less time on device use, namely participants in the first group had the smallest score, while participants in the last group had the highest score (Intention: 1.780 vs. 2.806; MPI: 2.240 vs. 3.204). Table 4
presents the details of comparisons as well as mean values for each group. And it was as expected that the more time spent on device use, the more addicted the person was, finally a higher responded score for MPI would be obtained. Consequently, it could be reasonably inferred that for the individual who spent more time on device use, she/he was also more likely to engage in such behavior while crossing street. In other words, the intention to use mobile phone while crossing was influenced by their daily use habits, these two variables were positively correlated. There were no significant differences in PBC/SN between the groups (p
The other three constructs were investigated by Kruskal–Wallis analyses, and the results indicated significant differences among groups within situation and attitude (p < 0.05), but not for safety awareness (p > 0.05). In order to further study the mean difference between groups, post-hoc tests with a Bonferroni correction (i.e., ANOVA) were performed. For pairwise comparisons on attitudes, the only difference between the first and last group was significant (p < 0.05). Means, medians and mean ranks were shown below: T ≤ 2 (mean = 1.390, median = 1.00, mean rank = 128.02), 6 < T (mean = 1.862, median = 1.75, mean rank = 211.32). Results of pairwise comparisons for construct situation were similar to attitudes, participants in the first group had a significant lower score than those in the last group: T ≤ 2 (mean = 1.500, median = 1.00, mean rank = 133.18), 6 < T (mean = 2.078, median = 2.00, mean rank = 204.47). As a whole, it was found that the individuals who spent more hours on mobile phone-related activities daily also had a more positive attitudes toward the using behavior when crossing street, and their behavior of using the device were also more likely to be influenced by others and/or surrounding.
For the variable of previous exposure to pedestrian injury/critical event, 35.66% (138) of participants responded that they had either bumped into other pedestrians or stationary objects (e.g., trash cans/lamp posts), and even had some more dangerous experiences (e.g., a close call with a vehicle) that directly caused by distraction of using mobile phone while walking. The remaining 249 participants (64.34%) responded no exposure. The difference between the two groups was examined by a series of t
-tests. The results revealed that there were significant differences in all constructs between the two groups (p
< 0.05), the details were presented in Table 4
. Meanwhile, it should be noted that the pattern for each construct was similar: the group in which participants reported such an exposure had a higher score than the group that reported no exposure. Specifically, compared with those not reporting exposure to pedestrian injury/critical event, participants who reported such an exposure perceived more support from significant others (1.652 vs. 1.418), valued the behavior in a more positive way (1.908 vs. 1.605), and showed more confidence in the ability of behavioral control (1.877 vs. 1.605) and their behavioral intention to use a mobile phone in a street crossing scenario was significantly higher than their counterparts (2.757 vs. 2.361). In addition, it was also found that the participants in exposure group were more addicted to the device use than those in non-exposure group (3.044 vs. 2.819), and their safety awareness was relatively weak (2.196 vs. 1.934). As a result, it was not surprising that they were more likely to be involved in pedestrian injury/critical event. It seems that there may be potential relationships between the previous exposure and these psychological constructs.
A series of t-tests were carried out to examine the difference in constructs between participants who reported using a mobile phone while crossing and not using the device. The results demonstrated that there were significant differences between the two groups for all the constructs. Compared with the participants who did not have an experience of using mobile phone while crossing the street in the past, those distracted pedestrians had significant higher scores on constructs (see in Table 4
), indicating that they had more positive attitudes towards such distracting behavior (2.027 vs. 1.356), perceived higher control over the behavior (1.975 vs. 1.392), had weaker safety awareness (2.180 vs. 1.854), were also more addicted to mobile phone use in their daily life (3.071 VS 2.704) and had higher intention to perform the behavior (2.932 vs. 2.014).Moreover, the participants reporting a past experience of distracted crossing were significantly more susceptible to the situation factor than those not reporting such an experience (2.296 vs. 1.588), and the former group also scored higher on construct SN (1.653 vs. 1.329). Consequently, they were indeed more likely to use a mobile phone during street crossing, and they also did so. To some extent, the results indicated that there may be potential associations between these constructs and behavior.
In order to examine the difference in constructs between age groups, MANOVA were performed for intention, MPI and situation, a significant multivariate main effect was found, F
(12, 1006) = 4.165, p
< 0.05, Wilk’s Lambda = 0.880. Furthermore, the results showed that there were univariate main effects for intention F
(4, 382) = 5.157, p
< 0.001, η2
= 0.051, and MPI, F
(4, 382) = 6.186, p
< 0.001, η2
= 0.061, the age difference in construct situation was not significant (p
> 0.05). Additional pairwise comparisons were conducted for intention and MPI (please see Table 4
for details). With regards to intention, it was found that the youngest participants reported a higher level of intention to perform the distracted behavior than the oldest group (p
< 0.05, 2.917 vs. 1.931), indicating that the younger participants were more likely to be distracted. Moreover, significant age differences were also found between 26–30 year olds and 41–60 year olds (p
< 0.01, 2.649 vs. 1.931). The results of other pairwise comparisons were not statistically significant. With regards to MPI, participants aged 19–25, 26–30 and 31–40 years were significantly more addicted to mobile phone use than those over 41 years old (p
< 0.05, 2.794/2.996/3.195 vs. 2.264). This phenomenon may be not surprising given that younger people are more susceptible to the new technologies and electronic devices than the older ones in general.
The age differences in SN, PBC, SA and attitudes were examined using Kruskal–Wallis analyses due to the violations of homogeneity of variance. The significant differences between age groups were found for PBC and attitudes (p < 0.01), while the differences in SN and SA were not statistically significant (p > 0.05). With respect to PBC, it was found that the younger participants had significantly higher scores than older participants (p < 0.01). The results of post-hoc tests with Bonferroni correction found that 17–18-year-olds had significantly higher scores than 31–40 as well as 41–60 year olds (p < 0.01, 2.167 vs. 1.512/1.336), while the remaining pairwise comparisons were not significant (p > 0.05). These results indicated that the older individuals’ perception of their ability to perform the distracted crossing involving mobile phone use was lower than the younger individuals’, this may be largely due to the difference in physical function between ages, such as the hearing of older people may be less good, and they respond slowly to a sudden event in general. As for attitudes, the results indicated that younger participants had higher scores than the older participants (p < 0.01), indicating that younger participants had a more positive attitude towards the behavior. Furthermore, according to the results of post-hoc tests with Bonferroni correction, the differences between age groups 17–18 and 41–60 (p < 0.01, 2.097 vs. 1.345), and group 19–25 and 41–60 (p < 0.05, 1.788 vs. 1.345) were found, the remaining pairwise comparisons were not statistically significant (p > 0.05).