Do German drivers use their smartphones safely?—Not really!

Accident Analysis and Prevention 96 (2016) 29–38

Contents lists available at ScienceDirect

Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap

Do German drivers use their smartphones safely?—Not really! Mark Vollrath a,∗ , Anja Katharina Huemer a , Carolin Teller a , Anastasia Likhacheva b , Jana Fricke a a b

Technische Universität Braunschweig, Institut für Psychologie, Ingenieur- und Verkehrspsychologie, Gaußstraße 23, 38106 Braunschweig, Germany Humboldt-Universität zu Berlin, Institut für Psychologie, Ingenieurpsychologie & Kognitive Ergonomie, Rudower Chaussee 18, 12489 Berlin, Germany

a r t i c l e

i n f o

Article history: Received 10 March 2016 Received in revised form 3 June 2016 Accepted 6 June 2016 Keywords: Driver distraction Handheld cell phone Texting while driving Eating Drinking Smoking Observational study Epidemiology Prevalence Safety Situational influences Driver characteristics

a b s t r a c t Research in the laboratory as well as in naturalistic driving studies has shown that texting while driving seems to be the most dangerous driver distraction. However, there is still some discussion about the extent to which drivers adapt their behavior to the traffic situation. Accordingly, they might use their phones only in easy driving situations but refrain from doing so when driving becomes more demanding. For Germany, no reliable data on these topics could be found although overall smartphone use has also increased exponentially in this country. As observational studies have proven to be an effective means to gather these data, such a study was done observing 11,837 drivers in three big German cities (Braunschweig, Hannover, Berlin) during daytime. An alarmingly high rate of texting while driving was found (4.5%) as compared to other international studies. This was even more frequent than the use of handheld (2.2%) and hands-free (1.7%) phones combined. Thus, there seems to be a special problem in Germany with texting which should be further examined as this activity is highly distracting. Finally, there was some indication that drivers adapt their secondary task activities to the requirements of the driving task (e.g. somewhat less texting when moving than when stationary at a red traffic light). However, these adaptations were not very strong. Thus, drivers seem to underestimate the dangers due to distraction. This could be a starting point for countermeasures which increase this awareness of danger. © 2016 Elsevier Ltd. All rights reserved.

1. Background Distraction seems to become one of the major causes of accidents. For example, Singh (2015) reported for the US that recognition errors (driver’s inattention, internal and external distractions, and inadequate surveillance) were responsible for 41% of the crashes examined. In Austria, driver distraction is recorded as a causing factor in vehicle crashes by the police. Accordingly, in Austria in 2014 distraction was the most common cause for traffic crashes, being responsible for 38% of all crashes (VVO, 2015). One of the major sources of distraction seems to be the use of smartphones while driving. While only about 30% of the German households had at least one smartphone in 2000, the percentage has risen to 90% in 2012 (Statistisches Bundesamt, 2013). This increase may also result in an increased use while driving. For example, in the Naturalistic Driving Study SHRP2, Victor et al. (2014) found drivers to be clearly attentive to the road only in about 30% of their analyzed driving episodes, but clearly distracted

∗ Corresponding author. E-mail address: [email protected] (M. Vollrath). http://dx.doi.org/10.1016/j.aap.2016.06.003 0001-4575/© 2016 Elsevier Ltd. All rights reserved.

in more than 50% of these episodes. The most common, identified types of distraction were passenger related (about 10%), portable electronics (about 10%) and texting (about 8%). There is still a discussion about which kinds of distractions are really dangerous and increase the risk of an accident. While the first case-control studies seemed to indicate that talking on the phone increases accident risk (Redelmeier and Tibshirani, 1997), more recent studies even find a decrease in accident risk while phoning (Young, 2015). Methodological errors in the early risk studies may have led to these first estimations of an increased risk while phoning (Young, 2015). The first analyses of rear-end accidents in the SHRP2 study also showed that the risk of a rear-end accident while phoning is only 0.1 of that when driving without distraction (Victor et al., 2014). However, in a newer analysis from the same database (the SHRP2 data), Dingus et al. (2016) compared the relative contribution of driver distractions, driver errors and other driver impairments in all crashes recorded during the study with episodes in that drivers were clearly unimpaired (so-called model driving). In this comparison, they found that talking on the cell phone increased the accident risk by 2.2. The major cause for these differences in results may be that in the car following scenario drivers on the phone may have been

30

M. Vollrath et al. / Accident Analysis and Prevention 96 (2016) 29–38

driving slower with larger headways. Moreover, drivers who phone tend to focus on the road ahead while phoning (Victor et al., 2014) and do not look so much around as they do otherwise. Thus, they are very well able to rapidly detect if the car in front stops. In contrast, the scenarios investigated by Dingus et al. (2016) were more complex and far more diverse. Additionally, the use of model driving as the baseline comparison may have influenced the estimation of the accident risk. Overall, these results show that a secondary task like phoning increases accident risk for some situations but may even decrease the risk in other situations if drivers adapt their behavior and drive more cautiously. Thus, the situations in which drivers engage in secondary tasks should be examined as well as the effect of the secondary tasks themselves. This can be done relatively easily by means of traffic observations. In one of the first studies by Johnson et al. (2004) digital photographs were used. When examining about 40,000 photographs of drivers passing through a New Jersey turnpike, the most frequent activities were using the cell phone (1.39%), interactions with passengers (0.56%) and smoking (0.39%). Drinking accounted for 0.22%, eating for 0.08%. However, as photographs only capture a very short moment in time, other studies have used human observers at the side of the road (e.g., in the US, g. McCartt et al., 2010; in the UK, Broughton and Buckle, 2007; in New Zealand, Dury et al., 2012; in Australia Young et al., 2010; in Spain, Prat et al., 2014). The results from these and other observation studies show that this is quite an effective method to gather a large number of observations and to estimate the frequency of different, observable distractions. Moreover, quite a large number of possible influencing factors can be observed regarding the driver and the circumstances of the trip. This can be used to better understand which drivers engage in secondary tasks under which circumstances. Finally, as only very basic characteristics of the driver like age and sex are recorded, the data are anonymous and cannot be related to any specific driver. For example, a large number of studies found higher phone use in young drivers as compared to middle aged drivers (e.g. Dury et al., 2012; Horberry et al., 2001; Young et al., 2010; Gras et al., 2012). The presence of passengers seems to reduce phone use (e.g., Astrain et al., 2003; Glasbrenner, 2005; Johnson et al., 2004; NHTSA, 2009; Wenners et al., 2013). The time of day implying different kinds of trips may also influence phone use. For example, in Europe, Sullman (2012) found more eating and cell phone use in the morning, but more texting in the afternoon. This information gathered by traffic observations is also quite useful to develop countermeasures for those distractions that are most frequent. The information about different groups of drivers and the circumstances under which secondary tasks occur is also very interesting in order to define relevant target groups and situations. However, for Germany there is still hardly any information about the frequency of distraction while driving, especially caused by smartphones. In Germany, phoning is only allowed using a hands-free device. It is forbidden to have a phone in your hand and to operate it in any way (§23 StVO; German Traffic Law). Since 2005, the number of drivers caught using the phone while driving has increased from about 290,000 to about 450,000 (KBA, 2012). However, it is impossible to say in which manner this corresponds to the percentage of the drivers using the phone. Compared to about 50,000,000 cars in Germany, this figure seems very small. In 2009, a pilot study was conducted interviewing 289 drivers at parking lots at highways and in town, including cars and trucks (Huemer and Vollrath, 2011). It was found that 34% of them had tuned their radio during the last half hour of the trip, about 20% had made a phone call, 3% had made an input with their navigation system and 1% had written a text message.

More recent data on secondary tasks while driving were gathered by some surveys in German speaking countries. In these surveys, the frequency of secondary tasks while driving was estimated by asking participants how often they did specific tasks while driving. Answers could be given on a five point scale from “never” to “always”. Accordingly, the data give only a very rough estimation about the willingness to engage in different types of tasks. For example, Kubitzki (2011) questioned 600 drivers from Germany, Austria and Swiss about 40 specific tasks done while driving and split data for drivers who had a crash reported in the insurance companies’ files and those who had no crash. He found the most frequently reported activity to be listening to the radio (91% of drivers who had an accident, 82% of those who had not), followed by talking on the phone hands-free after having been called (88%/76%). In all tasks examined, those drivers who have had crashes in the previous three years resulted in higher proportions reporting the task (Kubitzki, 2011; p. 17). Similar estimates come from Goodyear Dunlop Germany (2013), AachenMünchener Versicherung AG (2013) and Ford-Werke GmbH (2014). All these studies from Germany show that a large percentage of drivers are willing to engage in secondary tasks at least at some occasions. However, these percentages do not tell how many trips are done with secondary tasks and in which situations. Accordingly, the results cannot be compared to those gathered by driver observations. Thus, although the data from other countries indicate that distraction especially by smartphones may be an increasing problem in traffic, respective observational data are missing for Germany. The first aim of the study presented here was to gather this information. Different methods to this aim were examined beforehand. Naturalistic driving studies require an enormous effort but usually give only information about a limited number of drivers. Conducting interviews with drivers gives only quite imprecise information (e.g., “I do this sometimes”). Even interviews directly after the trip as in Huemer and Vollrath (2011) may suffer from the inability to precisely recollect the involvement in secondary task and especially their duration. Moreover, some drivers may not want to tell their activities truthfully. Thus, an observation study was conducted in three large cities in Germany (Braunschweig, Hannover and the capital Berlin). Following the discussion above, one should not only measure how many drivers engage in smartphone use, but also in which situations they do so. This is the reason for the question in the title of this paper: Do German drivers use their smartphones safely? In order to address this question, the observations were done at situations where using a smartphone might be less dangerous (e.g., when waiting at a red traffic light) and others where drivers might refrain from using their phone because of the complexity of the driving situation (e.g., in dense traffic on larger roads in town). To summarize, the study has two aims: (1) To estimate the frequency of different distracting activities while driving in Germany. (2) To examine whether drivers adapt their behavior to the traffic situation and to what extent the behavior depends on driver characteristics. 2. Method 2.1. Locations and timing As the observations require that the observer stands quite close to the passing cars and has sufficient time to observe what the driver is doing, it was decided to focus the study on traffic within towns. As the observations are more effective when more

M. Vollrath et al. / Accident Analysis and Prevention 96 (2016) 29–38

traffic is present and as more traffic leads to a larger coverage of the overall trips, three larger cities in the North of Germany were selected. Braunschweig is the smallest with about 250,000 inhabitants followed by Hannover (capital of Lower Saxony, about 500,000 inhabitants) and, finally, Berlin as the capital of Germany (about 3.5 Mio inhabitants). Observations were made at six locations in the city of Braunschweig between March and April 2015 and at six locations in Hannover in May 2015. In Berlin data was gathered in two municipalities, in Charlottenburg in the west part of the town and Friedrichshain in the east of the city center. In both municipalities, 9 locations were chosen and observed between July and November, 2015. Overall, the observations took about 130 h which corresponds to about 90 observations per hour. The main criterion for the selection of the locations in all three towns was to find large arteries with a speed limit of 50 km/h (which is the usual speed limit in towns in Germany) and with at least a medium traffic density. This was done in order to capture a sample of the traffic coming in and out of town while ensuring sufficient traffic flow to enable an effective observation. While these selections above were done in order to gather a large sample of the traffic on major roads in the three towns, additional characteristics of the locations were selected in order to examine the influence of the traffic situation on the engagement in secondary tasks. The most important influencing factor was supposed to be the movement of the car. Thus, one part of the observation in each town was done for cars stationary at red traffic lights, a second one about 150 m after the traffic light in moving traffic and a third part even further away on the free road. It was expected that there would be a higher frequency of secondary tasks when stationary than when moving. As a second factor, the number of lanes going in one direction was varied by selecting roads with only one lane and others with two lanes. As there are more cars to observe by the driver on two-lane roads it was supposed that drivers would do less secondary tasks on two-lane roads as compared to roads with only one lane. At each of the three towns, a trained individual observer studied the drivers of the cars. The observers were three of the authors who did the studies as their Bachelor theses. Thus, each of them had been extensively introduced to the background of the study. They had developed and tested the observation sheet and had practiced the observation before the real study extensively. As the detection of the secondary tasks was limited to well-observable tasks it was decided to use only one observer at each location. Moreover, for two observers it would have been more difficult to be unobstrusive and not be noticed by the drivers which might have changed their behavior. Finally, two observers might have impaired each other in getting a really good view into the car. In each town, observations were done at three different times, as some international studies had found differences: In the morning between 8 and 9 am, in the early afternoon between 1 and 2 pm and in the later afternoon between 5 and 6 p.m. These times were distributed on the locations so that for each time period observations were done at each of the locations. In all three cities, the observations were done during weekdays and at different days of the week. As the observations in Braunschweig and Hannover had shown to be able to capture a large amount of drivers, and the Berlin study was conducted somewhat later, it was decided to observe additional drivers in Berlin during the weekend. 2.2. Procedure and measures In Braunschweig and Hannover, the observer stood on the right side of the traffic between parked cars or on the sidewalk, in Berlin additionally on the middle-island of the two-lane roads to observe the left lane. First the number of passing cars in one minute was

31

counted to get an indication of the traffic density at the different locations. After that, the observation of single drivers started. These were selected from the oncoming traffic as the observer decided per random which of the next few cars to observe. In order to describe the sampling fraction that was achieved, we used the count of traffic density to estimate how many cars went by during the observation time and related that to the number of observations. This estimations shows that about 10% of the cars passing by were observed. Following the procedure described, these 10% should be a random sample of the overall traffic passing by. The observers rated the age (young: about 18–24 years old, middle aged about 25–64, and old drivers from 65 years onwards) and the sex of the driver. As the observers noticed in the first two cities that the presence of passengers seemed to influence the frequency of distracting activities, and the Berlin study was conducted somewhat later, in the Berlin part of the study the presence of passengers was also recorded. The following distractions could be observed and were recorded: • Handheld phoning: Drivers were talking on the phone which they held in their hand pressed against their ears. • Hands-free phoning: Drivers were talking and their phone was placed somewhere in the car. • Using the smartphone: Drivers had their smartphone in their hands and were operating it, looking away from the road and on the screen. • Eating/drinking/smoking: These activities were recorded separately. For the Berlin study, the presence of passengers was also recorded. Moreover, for two-lane roads it was recorded whether the driver was on the left or right lane of these. Only very few drivers were doing more than one activity (if any). Thus, it did not make sense to analyze them separately. Instead, each single activity was analyzed separately and combined (“at least one secondary task”). Additionally, for control purposes, weather and light conditions were recorded. However, the observations were done during the day with good viewing conditions and mostly good weather. Thus, a further analysis of these factors was not necessary. 2.3. Analyses In order to create a predictor “traffic density”, the frequency distribution of the number of cars counted at each observation site was separately computed (four frequency distributions) for observations of stationary and moving traffic and for roads with one and two lanes. This was done because the values differed quite a lot. For each of the four sub-groups, a median split was done to create “low” vs. “high” traffic density. For Braunschweig, the four values were 12, 7, 19 and 15 cars in a minute of observation (one lane, moving and stationary; two lanes, moving and stationary, respectively). In Hannover, the values used were 13, 11, 9 and 7. In Berlin, it was 10, 8, 28 and 26. These values show that the traffic density was somewhat different between the cities and also between the different roads observed. Thus, this factor is useful for comparisons within one city, but not between the cities. The time of day was recoded as morning (8.00 h–9.00 h) and afternoon (13.00 h–14.00 h and 17.00 h–18.00 h) as preliminary analyses showed no differences between the two time-points in the afternoon. In the first step, separate logistic regressions were computed for each of the three sites for each of the single distracting activities and additionally for being engaged in any secondary task. All relevant predictors (see below) were included in each of these analyses. Due

32

M. Vollrath et al. / Accident Analysis and Prevention 96 (2016) 29–38

Table 1 Results of the logistic regression for any distracting activity for the Braunschweig site (n = 2116). Significant effects (p < 0.05) are indicated by “*”. Any secondary Task

Confidence Interval

Predictor

Wald

p

OR

lower

upper

Female vs. Male Age Medium vs. Young Old vs. Young 2 Lanes vs. 1 Lane Stationary vs. Moving High vs. Low Density Afternoon vs. Morning

0.0 18.5 0.3 16.0 0.7 5.4 0.7 4.1

0.877 0.000* 0.582 0.000* 0.391 0.020* 0.407 0.043*

1.0

0.7

1.3

0.9 0.3 1.1 1.4 0.9 0.7

0.6 0.2 0.9 1.1 0.7 0.6

1.3 0.5 1.5 1.9 1.2 1.0

Braunschweig Any Activity Overall

11.1

Young Medium Old

13.3 12.3 4.4

Stationary Moving

14.0 10.2

Morning Afternoon

13.4 9.9 5

0

to the relatively large number of predictors, the number of cases was too small for analyzing interactions between the predictors.

10 15 Number of Drivers [%]

20

Fig. 1. Percentage of drivers at the Braunschweig site who did any kind of activity, overall and for the significant factors due to the logistic regression.

3. Results

3.1. Braunschweig In Braunschweig, 2116 drivers were observed. Table 1 gives the results of the logistic regression with the criterion of doing any secondary task. Age, stationary vs. moving and afternoon vs. morning were found as significant predictors. For age, older drivers had an OR of 0.3 as compared to younger drivers. The OR when stationary as compared to moving was 1.4. The OR in the afternoon was 0.7 as compared to the morning. In order to describe the effects, Fig. 1 shows the percentages for the significant predictors. Overall, 11.1% of the observed drivers in Braunschweig were engaged in some kind of secondary task. But only 4.4% of older drivers as compared to 12.3% middle aged and 13.3% younger drivers did this. When stationary at a traffic light 14% were distracted, as compared to 10.2% when moving. Finally, more drivers engaged in secondary task in the morning (13.4%) than in the afternoon (9.9%).

Drink

Smoke

7.7 4.8

Medium

1.2

1.6 Old

3.0 0.3

Young

Stationary

1.5

Moving

3.0

Two lanes

4.1

Old

4.2

One lane

As the observations in the three study towns differed with regard to these predictors, we first examined for each town separately whether these predictors influence the frequency of the different secondary tasks (using logistic regressions). In order to combine the data sets, the observations from the three sites were then weighted using the predictors that influenced the different secondary tasks and which were different between the sites due to the study design. The rationale of this weighting procedure is described in more detail below. Thus, the analyses using these predictors are presented in the first section of the results for the three study towns. Additionally, as control variables, age and sex of the driver were also included besides the four predictors described above. Afterwards, the weighting procedure is described. Finally, the estimation of overall secondary task involvement is presented using the weighted data. The final section presents the investigation of situational influences on secondary task activity for the total data set.

Smartphone

Medium

Stationary at a red light vs. moving in traffic One-lane road vs. two-lane road High vs. low traffic density Morning vs. afternoon

20 18 16 14 12 10 8 6 4 2 0

Young

1) 2) 3) 4)

Number of Drivers [%]

Braunschweig Single Activities Besides describing the frequency of different distracting activities, the second aim of the study was to examine to what extent drivers adapt their behavior to the driving situation. To this aim, the following situational predictors were examined:

Fig. 2. Significant predictors for using the smartphone, drinking and smoking. The bars show the percentage of drivers engaging in these activities with regard to the different predictors.

The same pattern of logistic regressions was then done for all single activities. Due to the smaller frequencies, fewer effects were found (see Table 2). None of the predictors was significant for phoning (neither handheld nor hands-free nor combined) and eating. For typing on the smartphone, there was a significant effect of age and a trend for the number of lanes. In order to describe the effects, Fig. 2 gives the percentages of the significant predictors. Young drivers use the smartphone much more frequently than middle aged and older drivers (7.7% as compared to 4.2% and 4.1%, respectively). Against expectations, the smartphone is used somewhat more frequently on roads with two lanes (4.8% as compared to 3.0%). Drinking is done more frequently when stationary (1.5%) as compared to moving (0.3%). The age effect for smoking may reflect the frequency of smoking in the population of drivers, as middle aged drivers were found to smoke more frequently (3.0%) than young (1.2%) and older (1.6%) drivers. When comparing the effects for all activities combined to these single effects, it becomes clear that the main effect of age (young and medium aged drivers have more activities than older drivers) is the result of two somewhat different effects: younger drivers use the smartphone more frequently than older and middle aged drivers. Middle aged drivers smoke more frequently than both young and older drivers. The difference between stationary and moving is only significant for drinking. However, for all single activities besides hands-free phoning, the percentages of doing it while stationary is somewhat larger than when moving. There seems to be a trend to refrain a bit from secondary tasks when moving as compared to being station-

M. Vollrath et al. / Accident Analysis and Prevention 96 (2016) 29–38

33

Table 2 Results of the logistic regressions for the different activities (columns) and the predictors examined (rows) for the Braunschweig site. Significant effects (p < 0.05) and trends (p < 0.10) are indicated by “*” and “+”, respectively. Predictor

Female vs. Male Age Medium vs. Young Old vs. Young 2 Lanes vs. 1 Lane Stationary vs. Moving High vs. Low Density Afternoon vs. Morning

Phoning Handheld

Phoning Handsfree

Using Smartphone

Eating

Wald

p

Wald

p

Wald

p

Wald

p

Wald

p

Wald

p

1.5 2.1 0.1 1.2 0.2 0.8 0.3 0.1

0.219 0.353 0.737 0.277 0.688 0.373 0.589 0.782

1.4 2.7 2.0 0.3 0.5 0.0 0.5 0.5

0.234 0.257 0.154 0.614 0.499 0.962 0.488 0.463

0.2 15.1 6.6 11.0 3.2 2.3 0.8 0.2

0.635 0.001* 0.010+ 0.001* 0.072+ 0.133 0.382 0.639

0.9 2.7 0.2 2.7 0.0 1.5 0.2 0.2

0.345 0.259 0.645 0.101 0.915 0.221 0.619 0.665

0.1 0.8 0.1 0.3 0.4 8.1 0.6 0.6

0.768 0.677 0.743 0.603 0.512 0.005* 0.429 0.454

0.6 4.9 3.1 0.1 0.1 1.4 2.6 1.4

0.449 0.085+ 0.079+ 0.701 0.777 0.230 0.105 0.243

Table 3 Results of the logistic regression for any distracting activity for the Hannover site (n = 3473). Significant effects (p < 0.05) are shown with “*”. Any secondary Task

Confidence Interval

Predictor

Wald

p

OR

lower

upper

Female vs. Male Age Medium vs. Young Old vs. Young 2 Lanes vs. 1 Lane Stationary vs. Moving High vs. Low Density Afternoon vs. Morning

1.3 30.0 0.0 23.4 6.8 4.7 0.2 2.4

0.249 0.000* 0.887 0.000* 0.009* 0.030* 0.632 0.120

1.1

0.9

1.4

1.0 0.1 1.3 1.3 0.9 0.8

0.7 0.1 1.1 1.0 0.8 0.7

1.4 0.3 1.6 1.6 1.2 1.0

Hannover Any Activity Overall

14.5

Young Medium Old

15.5 15.7 2.8

Stationary Moving

16.9

Drinking

Smoking

drivers were more often distracted on two-lane roads (16.0%) than on one-lane roads (13.1%). When looking at the single activities, there are overall similar patterns as in Braunschweig with few exceptions (see Table 4). For handheld phoning there was an effect of the number of lanes and stationary vs. moving. Fig. 4 (right) shows these effects. Drivers were phoning more frequently on two-lane roads (3.1% as compared to 1.6%). They were also phoning more often when moving (2.6%) then when being stationary (1.3%). There were no effects for hands-free phoning, but effects of age and number of lanes for using the smartphone (see Fig. 4 left). As in Braunschweig, young drivers used their smartphone more frequently (9.2%) than older drivers (0.6%). However, middle aged drivers in Hannover used their smartphone similarly often as young drivers (9.6% and 9.2%, respectively). There was no effect for eating, but as in Braunschweig for drinking. Drivers drank more often when stationary (1.2%) than when moving (0.1%), and also more often in the morning (0.7%) than in the afternoon (0.1%). This may be due to drinking a coffee in the morning in the car on the way to work. Finally, different from Braunschweig, there was no age effect for smoking, but people were smoking more often when stationary (4.6%) than when moving (1.0%).

13.9

3.3. Berlin One Lane Two Lanes

13.1 16.0 0

5

10 15 Number of Drivers [%]

20

Fig. 3. Percentage of drivers at the Hannover site who did any kind of activity, overall and for the significant predictors due to the logistic regression.

ary, but this is not really a very large effect as it did not become significant for any single activity besides drinking. 3.2. Hannover In Hannover, 3473 drivers were observed. The logistic regression for doing any kind of secondary task found age, the number of lanes and stationary vs. moving as significant effects (see Table 3). There was an OR of 0.1 for comparing old with young drivers, and OR of 1.3 for comparing stationary vs. moving and two lanes vs. one lane. In contrary to the results from Braunschweig, there was no effect of the time of day, but an effect of the number of lanes which could not be found in Braunschweig. For the description of these results, Fig. 3 shows the percentages with regard to the significant predictors. Overall, 14.5% of the drivers engaged in any kind of activity. Just like Braunschweig, younger and middle aged drivers were much more active (15.5% and 15.7%, respectively) than the older drivers (2.8%). There were more secondary tasks when being stationary (16.9%) as compared to moving (13.9%). Finally, as in Braunschweig,

In Berlin, 6248 drivers were observed. As described above, in Berlin there were also observations on the weekends and the presence of passengers was recorded. Thus, two additional predictors were included in the logistic regression, namely the day of the week (weekday vs. weekend) and the presence of passengers (with vs. without passengers). Table 5 shows the results. Sex, age, the number of lanes, traffic density and the presence of passengers were significant predictors. Females had an OR of 0.8 as compared to males. Middle aged and older drivers also had ORs smaller than one when compared to young drivers (0.8 and 0.4, respectively). The OR of two-lane roads was 0.8 as compared to one-lane roads. With higher traffic density, the OR was 0.9. Finally, with passengers the OR was 0.4 as compared to driving alone. For the description of the effect, Fig. 5 shows the respective percentages. Overall, 12.9% of all Berlin drivers engaged in at least one secondary task. Male drivers were more active than female drivers (13.3% vs. 12.1%). The age effect is very similar to the other two cities with 15.3% of young drivers, 12.7% of middle aged drivers and 7.2% of old drivers engaging in a secondary task. Drivers were more active on one-lane roads (14.0%) as on twolane roads (12.4%). This is opposite to the effect in Hannover. The presence of passengers has a very strong effect. Only 6.9% drivers were engaged in a secondary task with passengers but 15.1% when driving alone. Finally, with high traffic density the frequency of secondary tasks is somewhat reduced (12.1% as compared to 13.8%). When looking at the logistic regressions for the single activities (see Table 6) there are quite a number of significant effects, which

34

M. Vollrath et al. / Accident Analysis and Prevention 96 (2016) 29–38

Table 4 Results of the logistic regressions for the different activities (columns) and the predictors examined (rows) for the Hannover site. Significant effects (p < 0.05) and trends (p < 0.10) are shown with “*” and “+”, respectively. Phoning Handsfree

Using Smartphone

Eating

Wald

p

Wald

p

Wald

p

Wald

p

Wald

p

Wald

p

1.6 0.6 0.6 0.0 8.2 3.0 1.7 2.6

0.210 0.733 0.431 0.994 0.004* 0.081+ 0.192 0.109

0.7 1.8 1.6 1.3 0.5 0.5 1.5 0.2

0.402 0.406 0.210 0.249 0.463 0.490 0.216 0.639

0.0 15.6 0.1 13.1 5.8 0.1 0.0 0.1

0.827 0.000* 0.735 0.000* 0.016* 0.720 0.861 0.758

1.1 0.1 0.1 0.0 1.7 1.9 1.4 1.3

0.305 0.957 0.766 0.994 0.194 0.173 0.243 0.250

0.3 0.0 0.0 0.0 0.0 8.1 2.7 5.3

0.580 1.000 0.995 1.000 0.849 0.004* 0.103 0.021*

0.7 0.2 0.0 0.1 0.1 34.9 0.1 0.0

0.411 0.920 0.935 0.752 0.799 0.000* 0.813 0.870

4.6

Standing

Afternoon

Morning

Driving

1.0

0.1 0.7 0.1 Moving

Two lanes

One lane

Old

Medium

Stationary

1.2

0.6

1.6

3.1

2.6

1.3

Moving

7.6

20 18 16 14 12 10 8 6 4 2 0

One lane

Smoke

Number of Drivers [%]

Drink 10.0

9.2 9.6

Young

Number of Drivers [%]

Smartphone

Smoking

Hannover Handheld Phoning

Hannover Single Activities 20 18 16 14 12 10 8 6 4 2 0

Drinking

Stationary

Female vs. Male Age Medium vs. Young Old vs. Young 2 Lanes vs. 1 Lane Stationary vs. Moving High vs. Low Density Afternoon vs. Morning

Phoning Handheld

Two lanes

Predictor

Fig. 4. Significant predictors for using the smartphone, drinking and smoking (left) and for the handheld phoning. The bars show the percentage of drivers engaging in these activities by the different predictors.

Table 5 Results of the logistic regression for any distracting activity for the Berlin site (n = 6248). Significant effects (p < 0.05) are indicated with “*”.

Berlin Any Activity Overall

12.9

Male Female

13.3 12.1

Young Medium Old

12.7

7.2

One Lane Two Lanes

Any secondary Task

15.3

14.0 12.4

Alone Passengers

15.1

6.9

Low Density High Density

13.8 12.1 0

5 10 15 Number of Drivers [%]

20

Confidence Interval

Predictor

Wald

p

OR

lower

upper

Female vs. Male Age Medium vs. Young Old vs. Young 2 Lanes vs. 1 Lane Stationary vs. Moving High vs. Low Density Afternoon vs. Morning Weekend vs. Workday Passengers vs. Alone

8.3 27.1 8.8 24.8 4.1 0.7 4.2 1.5 1.7 76.2

0.004* 0.000* 0.003* 0.000* 0.043* 0.405 0.041* 0.214 0.189 0.000*

0.8

0.7

0.9

0.8 0.4 0.8 0.9 0.9 1.1 1.1 0.4

0.7 0.3 0.7 0.7 0.7 0.9 0.9 0.3

0.9 0.6 1.0 1.1 1.0 1.3 1.3 0.5

Fig. 5. Percentage of drivers at the Berlin site who did any kind of activity, overall and for the significant predictors due to the logistic regression.

also differ between the activities. Due to the large number of effects, not all are given as figures. Instead, for the description of the effects, Table 7 shows percentages for all significant effects. For handheld phoning, female drivers did this less than male drivers (1.9% vs. 3.0%). They also smoked less frequently (2.6% as compared to 4.1%), but ate more often (2.0% vs. 0.9%). There were also age effects. Younger drivers were phoning as often as middle aged drivers with a handheld phone (2.7%), but more frequently with a hands free phone than middle aged drivers (4.1% vs. 1.8%). In comparison to the middle aged drivers, younger drivers used their smartphone more often (3.1% vs. 1.8%) and ate more frequently (2.2% vs. 0.7%), but smoked less often (2.5% vs. 3.8%). Older drivers did the least phoning handheld (1.1%) as well as hands free (0.4%), using the smartphone (0.4%) and eating (0.7%). Old Berlin drivers were the most frequent smokers (4.4%).

There was an effect of the number of lanes for handheld phoning (more on two lanes) and smoking (less on two lanes). Interestingly enough, drivers phoned handheld more when moving than when being stationary (2.8 vs. 1.2%) and ate more when moving (1.5% vs. 0.5%), but used their smartphone more often when stationary than when moving (2.8% vs. 1.9%). High vs. low traffic density was significant for drinking only (more in low density). Drivers used their handheld and hands free phone more often in the afternoon than in the morning (2.9% vs. 2.0% and 2.5% vs. 1.9%) and ate more in the afternoon (1.6% vs. 1.0%). Finally, phoning (hands free and handheld), using the smartphone and eating was done less frequently when a passenger was present. The frequency of none of the activities was significantly different between workday and weekends (no significant main effect for any of the activities). Thus, for the overall estimation of the frequency of distraction it was decided to include these data from the weekends in Berlin in order to increase the number of observations.

M. Vollrath et al. / Accident Analysis and Prevention 96 (2016) 29–38

35

Table 6 Results of the logistic regressions for the different activities (columns) and the predictors examined (rows) for the Berlin sites. Significant effects (p < 0.05) and trends (p < 0.10) are given as “*” and “ + ”, respectively. Predictor

Female vs. Male Age Medium vs. Young Old vs. Young 2 Lanes vs. 1 Lane Stationary vs. Moving High vs. Low Density Afternoon vs. Morning Weekend vs. Workday Passenger vs. Alone

Phoning Handheld

Phoning Handsfree

Using Smartphone

Eating

Wald

Wald

Wald

Wald

10.0 5.5 0.1 5.4 5.2 7.5 0.1 4.3 1.2 32.8

p *

0.002 0.064+ 0.725 0.020* 0.023* 0.006* 0.801 0.039* 0.264 0.000*

0.1 34.7 27.0 11.5 1.3 0.0 0.3 4.8 0.7 27.8

p 0.704 0.000* 0.000* 0.001* 0.248 0.913 0.592 0.029* 0.408 0.000*

0.9 16.3 9.6 9.1 0.2 3.8 2.4 0.0 1.4 19.5

p 0.354 0.000* 0.002* 0.002* 0.675 0.052+ 0.119 0.942 0.233 0.000*

7.3 12.4 11.2 3.3 2.2 4.4 5.8 1.5 0.3 6.4

Drinking p *

0.007 0.002* 0.001* 0.069+ 0.139 0.035* 0.016* 0.213 0.582 0.011*

Smoking

Wald

p

Wald

p

0.0 3.9 2.1 0.5 1.4 0.8 12.8 0.7 2.3 0.6

0.904 0.142 0.150 0.461 0.232 0.364 0.000* 0.392 0.127 0.441

9.0 6.9 6.5 3.4 7.7 1.0 1.6 0.0 1.1 0.1

0.003* 0.032* 0.011* 0.064+ 0.006* 0.315 0.204 0.887 0.287 0.768

Table 7 Percentages of drivers engaging in different activities (columns) depending on the predictors with a significant effect in the logistic regression. Predictor (1) vs. (0)

Female vs. Male Medium vs. Young Old vs. Young 2 Lanes vs. 1 Lane Stationary vs. Moving High vs. Low Density Afternoon vs. Morning Weekend vs. Workday Passenger vs. Alone

Phoning Handheld

Phoning Handsfree

Using Smartphone

Eating

(1)

(0)

(1)

(0)

(1)

(0)

(1)

(0)

1.9 2.7 1.1 2.9 1.2

3.0 2.7 2.7 1.7 2.8

1.8 0.4

4.1 4.1

1.8 0.4

3.1 3.1

2.0 1.0 0.7

0.9 2.2. 2.2

2.8

1.9

0.5

1.5

2.9

2.0

2.5

1.9

1.6

1.0

0.5

3.4

0.2

3.1

0.6

1.6

0.7

Time of Day

# Lanes

Motion

Braunschweig

Hannover

Berlin

Overall

Morning

1

Moving Stationary Moving Stationary Moving Stationary Moving Stationary

11.0 3.1 13.9 5.5 21.3 6.0 29.3 10.1

12.1 6.9 12.6 2.7 27.8 4.7 27.6 5.7

9.8 2.0 17.5 3.4 14.7 2.3 42.6 7.7

10.93 3.98 14.67 3.85 21.25 4.34 33.16 7.81

Afternoon

1 2

(1)

0.8

Table 8 Distribution of the samples in percent in the three cities and overall with respect to time of day, number of lanes and moving vs. stationary.

2

Drinking

3.4. Estimating the overall frequency of distraction in three large cities in Germany Three predictors were actively varied in all three observations: (1) stationary vs. moving, (2) one vs. two-lane roads, (3) morning vs. afternoon. The observations in all three cities were done during the week, but additional observations were conducted in Berlin at the weekend. As there was no effect of workday vs. weekend in Berlin, both observations were used in the pooled data set. As the first three predictors influenced at least partly (in some of the three towns as described above) the frequency of secondary tasks, it was examined to which extent the distribution of the observations with regard to the three predictors differed between the three cities. Table 8 shows the three-dimensional distribution of the observations for each of the three cities. Overall, the distributions of observations in the three cities are quite similar. The largest differences are less observations in Berlin in the afternoon with people driving on onelane roads (14.7% as compared to 21.3% and 27.8%), but more on two-lane roads (42.6% in comparison to 29.3% and 27.6%). As in Berlin drivers were more active on two-lane roads than on onelane roads, more observations with secondary tasks would enter the pooled data set if the observations from the three cities were just combined. Thus, it would seem that drivers were more active in Berlin as compared to the other two cities. However, this would

2.6

Smoking (0)

(1)

(0)

2.6 3.8 4.4 3.1

4.1 2.5 2.5 4.5

1.9

only be due to a larger sample of afternoons in Berlin as compared to the other towns. In order to avoid this kind of sampling bias, it was decided to weight the data in such a manner that the resulting threedimensional distributions for the three cities after the weighting were identical. To this aim, the combined distribution of the three cities was used as the target distribution (right column of Table 8). Then, each observation in each town received a weighting factor which was computed in such a manner that observations that were too frequent in one cell of the distribution of a town received a weight smaller than one. Thus, after the weighting, this observation did not count as one observation any more but as more than one. Observations in cells that were too infrequent received a weight larger than one. The computation of the weights will be described for an example. For example, overall (target distribution) 10.93% of all observations were done in the morning on one-lane roads with people moving. In Berlin, only 9.8% of all observations were done in this time-period. Thus, every observation from Berlin when driving on one-lane roads in the morning was given a weight of 10.93/9.8 = 1.12. As described above, there were too few observations in Berlin. Thus the weight was larger than one. Similarly, in Hannover the weight was 10.93/12.1 = 0.90. As there were too many observations in Hannover for this cell of the distribution, the resulting weight is less than one. When applying the weights to the observations and re-computing the three-dimensional distributions for each town, these were identical and corresponded to the overall distribution combining the three cities. If differences between the three cities are then found afterwards, these cannot be due to differences in the time of day, number of lanes, or movement of the cars in the observations of the three cities. Using these weighted data, the frequency of the different activities will be described in the following section. The combined data set comprises 11,837 observations. When looking at the distribution of observations with regard to age and sex, there were about 62% males and 38% females, 19%

Smoking

Berlin

Fig. 7. Comparison of the frequencies (percentage of drivers) in the three cities. The whiskers correspond to the 95% confidence interval.

young drivers, 70% middle aged drivers and 11% older drivers. This corresponds quite well to the distribution from the representative study about driving in Germany from 2008 (MID2008, distribution of trips between 8 a.m. and 6 p.m. for car drivers; Infas and DLR, 2010). In this study, also about 62% of all trips in Germany were done by males and 38% by females. Young drivers did 9% of all trips, middle aged drivers 76% and old driver 15%. Thus, the observations are very similar to trips in general in Germany. Fig. 6 gives the percentages as well as the 95% confidence intervals, which are quite small due to the large sample size. Overall, in 13.2% of all observations drivers were engaged in any of the observed activities. The most frequent one was using the smartphone with 4.5%, followed by 2.9% smoking. Phoning was found in 3.9% of the observations with 2.2% with a handheld phone (which is not allowed in Germany) and 1.7% with a hands-free phone (which is allowed in Germany). Using the smartphone, either for phoning or for other activities, sums up to 8.4% of all observations. Eating and drinking were done with about 1% (1.1% and 0.9%, respectively). As Fig. 7 shows, there are some differences in the frequencies between the three cities even after weighting. The overall percentage of any distracting activity is lowest in Braunschweig (10.9%) and largest in Hannover (15.0%) with Berlin in the middle (13.0%). For the single activities, the largest difference is found for the use of smartphones with 4.0% in Braunschweig, 9.1% in Hannover and 2.1% in Berlin. Hands-free phoning, which is allowed in Germany, ranges from 0.8% in Hannover to 1.3% in Braunschweig and 2.3% in Berlin. However, nearly the same percentages are found for handheld phoning, which is forbidden by German law. This is done by 0.9% of drivers in Braunschweig, 2.4% in Berlin and 2.5% in Hannover.

Smoking

Drinking

Moving

Eating

Using Smartphon e

Phoning Handsfree

Drinking

Hannover

Drinking

Using Smartph…

Phoning Handsfree

Phoning Handheld

Braunschweig

Eating

20 18 16 14 12 10 8 6 4 2 0 Any

Number of Drivers [%]

Fig. 6. Percentage of different activities in the observation study. The whiskers correspond to the 95% confidence interval (n = 11837).

Stationary

Phoning Handheld

0.9 Smoking

1.1 Eating

Phoning Handsfree

2.9

Using Smartphon e

1.7

Phoning Handheld

4.5 2.2

20 18 16 14 12 10 8 6 4 2 0 Any

13.2

Any

20 18 16 14 12 10 8 6 4 2 0

Number of Drivers [%]

M. Vollrath et al. / Accident Analysis and Prevention 96 (2016) 29–38

Number of Drivers [%]

36

Fig. 8. Percentages of the different activities when stationary (white bars) and when moving (grey bars). The whiskers correspond to the 95% confidence interval (n = 2366 stationary and 9471 moving).

The second question of this study was to which extent drivers adapt their behavior to the traffic situations. Our hypothesis was that engaging in distracting activities would be more frequent when stationary than when moving. Fig. 8 shows the comparison. Overall, the frequency is a bit larger when stationary (14.4%) than when moving (12.9%), but the 95% confidence intervals overlap. The pattern even reverses for phoning (handheld and hands-free) and eating. Handling the smartphone, drinking and smoking is somewhat more frequent when stationary then when moving. Thus, there is some adaptation for some activities, but that is not large. 4. Discussion Overall, the results show an alarming frequency of using the smartphone to type and read in all three cities. This was least the case in Berlin with 2.1%, but largest in Hannover with 9.1%. Although all three observers had been trained in an identical manner and used the same method, it cannot fully be excluded that in some part this may be due to differences in observation style between the three observers, e.g. looking for a longer time inside the car or from a different perspective. However, it possibly also reflects differences in the driver population, the purposes of the trips, and the structure of roads and traffic. Phoning with a handheld phone (or smartphone) is also quite frequent although forbidden in Germany. It is even somewhat more frequent than using hands-free phones. Combining these different instances of illegal smartphone use, this seem to pose a serious problem for traffic safety in Germany. Using apps and typing on the smartphone seems to have become more frequent then just phoning. Reviews of experimental studies on distraction (e.g., Vollrath et al., 2014) as well as simulator studies (e.g. Schömig et al., 2015) show that typing on the smartphone leads to much larger distraction than phoning. This does not seem to be clear to drivers. Moreover, conversations on the phone may take minutes and thus lead to long time periods of distracted driving. But also using the smartphone with shorter messages, but with a rapid exchange between different persons, may be a more or less continuous activity. This will then, of course, further increase the probability that something may happen while is driver is thus distracted. This longer lasting distraction might not be so dangerous, if drivers are well able to estimate the difficulty or risk of the current traffic situation and adapt their usage of electronic devices while driving accordingly, like some studies suggest (e.g., Metz et al., 2011; Schömig et al., 2011). Indeed, the frequencies found in this observational study were partly larger when stationary than when moving. However, looking at the overall frequency of distraction of any kind, this adaptation was only seen in Braunschweig and Hannover, but not in Berlin. And the frequency was only reduced by

M. Vollrath et al. / Accident Analysis and Prevention 96 (2016) 29–38

about 3%–4% from a level of 14% or 17% in Braunschweig and Hannover, respectively. Furthermore, at the level of single activities, an adaptation of the frequency of use was only found in Braunschweig for drinking and in Hannover for handheld phoning, drinking and smoking. In Berlin, for handheld phoning and eating there was even an inverted effect. For the other factors like the number of lanes and traffic density, the effects were quite mixed and also not large. Thus, overall we have to conclude that drivers do not really much adapt their engagement in secondary tasks to the suspected requirements of the traffic situation in real driving situations. When looking at the frequency of secondary tasks in Germany in our study, overall our data fit well to the international studies. At the ground of single studies, our recent German data is most similar to the Austrian data of Sullman and Metzger (2012), being in a German speaking country and observing secondary tasks in a large city (Salzburg). They also found a rather high use of cell phones with 4.3% of people talking on the phone compared to 3.9% in our study. Texting was found to be less frequent with only 0.5% in 2010. The 4.5% in our study may very well be due to an increase in texting behavior in general, as SMS and other text messages now cost almost nothing compared to high prices per text message back then. The frequencies of eating, drinking and smoking are also very similar. In city observations in Spain, Gras et al. (2012) and Prat et al. (2014) found about 1.3% cell phone use. Recent data from the UK (Scoons, 2013; Sullman, 2012; Sullman et al., 2015) combined found somewhat larger frequency for mobile phone use (phoning and texting) with about 4.9%. However, this is still much smaller than the 8.4% in our study. They found also a somewhat smaller percentage of eating and drinking (combined) with 1.1% as compared to 2.0%. Smoking is similar with 2.2% in the UK and 2.9% in Germany. These differences may be due to different populations of drivers observed, as the UK studies data collection was not exclusively concentrated in cities. The differences for eating/drinking may also point to cultural differences. Finally, it may also be that phoning behavior has increased during the last few years. Only in the US studies, a higher amount of handheld phoning (about 7%, and very stable in the last five years; e.g. NHTSA, 2015) is found than in our study. But even in the US and in Australia where texting also seems to be high, in the most recent available data (observational US: from 2013, NHTSA, 2015; NDS data from Dingus et al., 2016; Australian: data from 2009, Young et al., 2010) texting activities are much less with about 2.5% than in our German population with 4.5% of driver texting. This large amount of activity in one of the most dangerous secondary tasks is alarming and shows that German drivers do not use their smartphone safely while driving at all. While the results of this study give a first approximation of secondary task prevalence in German inner-city traffic, several limitations need to be mentioned. As we found some differences in the frequencies between the three towns, additional data from other cities in Germany should be gathered in order to be really representative for Germany as a whole. Observations were only made in good weather conditions and mostly in daylight, to ensure data quality. Secondary task occupation may be very different in more adverse conditions such as heavy rain or in worse visual conditions like during the night. Observations are in general also limited to secondary tasks that actually can be observed. Thus, some tasks reported by drivers in surveys (like listening to music or daydreaming) are not accounted for in our data. Therefore, overall secondary task occupation may very likely be underestimated. Despite these limitations, observational studies are resource efficient (in time and costs per data point) methods to gather cross-sectional information on the prevalence of drivers’ activities. The high number of drivers observed to be texting clearly needs to be followed up upon in future observational studies. As observations now have only been made at very few locations in city-centers,

37

for a representational picture of German drivers’ secondary task activities, observational sites need to be stratified in cities and observations for rural roads and large highways (the German Autobahn) should be added. For large highways with their high speeds, observers cannot be standing at the side of the roadway. Moreover, at the high speeds of the cars secondary tasks are difficult to observe. Thus, observers need to be moving with the observed traffic as it has been done by Westat (cited after Sharpiro et al., 2001) for the NHTSA. For rural traffic on one-lane roads, speed is also a problem, but one cannot move with the traffic. Video or photo observations may be the only way to obtain these data. As data form the US (e.g. Utter, 2001; Glasbrenner, 2005) and the UK (e.g., Broughton and Buckle, 2007; Broughton and Hill, 2005) show, repeated observations are a feasible measures to obtain data for changing behavioral patterns in driver activities. These repeated observations should also be done for Germany, as it seems to have a rather special driving culture in terms of secondary task occupation. With regard to countermeasures, the frequencies show that using the smartphone (for phoning, texting, apps) is the major source of distraction in German drivers and should be the major topic for prevention. One especially relevant target group are the young drivers (18–24 years in our study) who already have a larger accident risk than middle aged drivers (for Germany, e.g. Krüger and Vollrath, 2004). In this group, smartphone use was more frequent than in the other age groups at least in Braunschweig and Berlin. Thus, it would be especially important to address this group of drivers as successful countermeasures would have the largest impact here. This has already been noted by prevention organizations like, for example, the Deutsche Verkehrssicherheitsrat with their campaign “Abgelenkt” (distracted; see http://www. abgelenkt.info/). A northern German radio station for young people (NJOY) has launched a major awareness campaign with hourly spots especially directed against smartphone use while driving by young drivers. From the legal side, a revision of the Traffic Law to better address smartphone use is planned. Here, it could make sense to discuss special rules for younger drivers similar to the Zero-BAC limit which already exists for German novice drivers. Observational studies are a very efficient way to examine the effects of such countermeasures which can also be done locally (e.g. in a city before and after specific campaigns have been introduced). In order to obtain better estimations of the accident risk introduced by these different activities, these observational studies should be complemented by analyses of traffic accidents. Facing the large frequencies found in our study, respective accident studies in Germany are dearly needed.

References AachenMünchener Versicherung AG, 2013. Studie Deutschland deine Autofahrer, Retrieved February 22, 2016 from https://www.amv.de/studie-ablenkung980978/. Astrain, I., Bernaus, J., Claverol, J., Escobar, A., Godoy, P., 2003. Prevalencia del uso de teléfonos móviles durante la conducción de vehículos (Prevalence of mobile phone use while driving vehicles). Gac. Sanit. 17, 66–69. Broughton, J., Buckle, G., 2007. Mobile Phone and Seat Belt Usage Rates in London, March 2006. TRL Limited, Published Project Report PPR232, Retrieved January 28, 2016 from http://www.trl.co.uk/reports-publications/trl-reports/report/ ?reportid=6187. Broughton, J., Hill, J., 2005. Mobile Phone Use by Drivers, 2000–03. TRL Limited, TRL Report TRL634, Retrieved January 28, 2016 from www.trl.co.uk/reportspublications/trl-reports/report/?reportid=2767. Dingus, T.A., Guo, F., Lee, S., Antin, J.F., Perez, M., Buchanan-King, M., Hankey, J., 2016. Driver crash risk factors and prevalence evaluation using naturalistic driving data. PNAS, http://dx.doi.org/10.1073/pnas.1513271113, published ahead of print February 22, 2016. Dury, C., Abussaud, Z., Allison, G., Bhindi, E., Bustard, J., Chamberlain, J., et al., 2012. Mobile Phone Use While Driving After a New Law: Observational Study, Retrieved January 28, 2016 from http://www.otago.ac.nz/wellington/ otago041209.pdf.

38

M. Vollrath et al. / Accident Analysis and Prevention 96 (2016) 29–38

Ford-Werke GmbH, 2014. Europaweite Ford-Studie: Jeder dritte deutsche junge Autofahrer hat bereits “Selfies während der Fahrt geknipst, Retrieved February 22, 2016 from http://www.presseportal.de/pm/6955/2801110. Glasbrenner, D., 2005. Driver Cell Phone Use in 2005—Overall Results. Report No. DOT HS 809 967. National Highway Traffic Administration, Washington, D.C, Retrieved January 28, 2016 from www-nrd.nhtsa.dot.gov/Pubs/809967.pdf. Goodyear Dunlop Germany, 2013. Junge Autofahrer: Ein Drittel telefoniert während der Fahrt, Retrieved February 22, 2016 from http://www.goodyear. eu/de de/media-and-promotions/news-items/articles/2013/6/jungeautofahrer-ein-drittel-telefoniert-waehrend-der-fahrt 145825. Gras, M.E., Planes, M., Font-Mayolas, S., Sullman, M.J.M., Jiménez, M., Prat, F., 2012. Driving distractions in Spain. In: Dorn, L. (Ed.), Driver Behaviour and Training, Vol. V. Ashgate, Farnham, pp. 299–305. Hill, J.P., 2005. A Survey of Mobile Phone Use by Drivers, April 2004. TRL Limited, TRL Report TRL635, Retrieved January 28, 2016 from http://www.trl.co.uk/ reports-publications/trl-reports/report/?reportid=2768. Horberry, T., Bubnich, C., Hartley, L., Lamble, D., 2001. Drivers’ use of hand-held mobile phones in Western Australia. Transp. Res. Part F Traffic Psychol. Behav. 4, 213–218. Huemer, A.K., Vollrath, M., 2011. Driver secondary tasks in Germany: using interviews to estimate prevalence. Accid. Anal. Prev. 43, 1703–1712. Infas & DLR, 2010. Mobilität in Deutschland 2008. Ergebnisbericht. Bundesministerium für Verkehr, Bau und Stadtentwicklung, Berlin. Johnson, M.B., Voas, R.B., Lacey, J.H., McKnight, A.S., Lange, J.E., 2004. Living dangerously: driver distraction at high speed. Traffic Inj. Prev. 5 (1), 1–7. KBA, 2012. Anzahl der registrierten Verstöße gegen das Handyverbot am Steuer in Deutschland von 2005 bis 2011 nach Altersgruppe. In Statista – Das Statistik-Portal, Zugriff am 10. Mai 2016, von http://de.statista.com/statistik/ daten/studie/224047/umfrage/registrierte-verstoesse-gegen-dashandyverbot-am-steuer-nach-alter-seit-2005/. Krüger, H.-P., Vollrath, M., 2004. The alcohol-related accident risk in Germany: procedure, methods and results. Accid. Anal. Prev. 36 (1), 125–133. Kubitzki, J., 2011. Ablenkung im Straßenverkehr: Die unterschätzte Gefahr. Allianz Deutschland AG, München. McCartt, A.T., Hellinga, L.A., Strouse, L.M., Farmer, C.M., 2010. Long-term effects of handheld cell phone laws on driver handheld cell phone use. Traffic Inj. Prev. 11, 133–141. Metz, B., Schömig, N., Krüger, H.-P., 2011. Attention during visual secondary tasks in driving: adaptation to the demands of the driving task. Transp. Res. Part F Traffic Psychol. Behav. 14 (5), 369–380, http://dx.doi.org/10.1016/j.trf.2011.04. 004. NHTSA, 2009. Driver Electronic Device Use in 2008—Overall Results. Report No. DOT HS 811 184. National Highway Traffic Administration, Washington, D.C, Retrieved January 28, 2016 from www-nrd.nhtsa.dot.gov/Pubs/811184.pdf. NHTSA, 2015. Driver Electronic Device Use in 2013—Overall Results. Report No. DOT HS 812 114. National Highway Traffic Administration, Washington, D.C, Retrieved January 28, 2016 from www-nrd.nhtsa.dot.gov/Pubs/812114.pdf. Prat, F., Planes, M., Gras, M.E., Sullman, M.J.M., 2014. An observational study of driving distractions on urban roads in Spain. Accid. Anal. Prev. 74, 8–16. Redelmeier, D., Tibshirani, R., 1997. Association between cellular-telephone calls and motor vehicle collisions. New Engl. J. Med. 336 (7), 453–458.

Schömig, N., Metz, B., Krüger, H.-P., 2011. Anticipatory and control processes in the interaction with secondary tasks while driving. Transp. Res. Part F Traffic Psychol. Behav. 14 (6), 525–538, http://dx.doi.org/10.1016/j.trf.2011.06.006. Schömig, N., Schoch, S., Neukum, A., 2015. Simulatorstudien zur Ablenkungswirkung fahrfremder Tätigkeiten BASt-Bericht M 253. Bundesanstalt für Straßenwesen, Bergisch Gladbach. Scoons, J., 2013. Mobile Phone and Seat Belt Usage Rates in Surrey 2012. TRL Limited, Published Project Report PPR642, Retrieved January 28, 2016 from www.trl.co.uk/reports-publications/trl-reports/report/?reportid=6799. Sharpiro, G., Utter, D., Mix, W., Croos, J., Marker, D., Bondy, N., 2001. Sample design issues for surveys involving the observation of drivers. In: Proceedings of the Annual Meeting of the American Statistical Association, August 5–9, 2001, Retrieved January 28, 2016 from https://www.amstat.org/sections/srms/ Proceedings/y2001/Proceed/00505.pdf. Singh, S., 2015. Critical Reasons for Crashes Investigated in the National Motor Vehicle Crash Causation Survey. (Traffic Safety Facts Crash Stats. Report No. DOT HS 812 115). National Highway Traffic Safety Administration, Washington, D.C. Statistisches Bundesamt, 2013. Pressemitteilung: Zahl der Woche vom 18. Juni 2013, In the Internet: https://www.destatis.de/DE/PresseService/Presse/ Pressemitteilungen/zdw/2013/PD13 025 p002.html. Sullman, M.J.M., Metzger, M., 2012. A roadside survey of driving distractions in Austria. In: Sullman, M., Dorn, L. (Eds.), Advances in Traffic Psychology. Ashgate, London, pp. 53–62. Sullman, M.J.M., Prat, F., Tasci, D.K., 2015. A roadside study of observable driver distractions. Traffic Inj. Prev. 16, 552–557, http://dx.doi.org/10.1080/ 15389588.2014.989319. Sullman, M., 2012. An observational study of driver distraction in England. Transp. Res. Part F Traffic Psychol. Behav. 15, 272–278. Utter, D., 2001. Passenger Vehicle Driver Cell Phone Use—Results from the Fall 2000 National Occupant Protection Use Survey. National Highway Safety Administration, Report no. DOT HS 809293. Retrieved January 28, 2016 from http://www-nrd.nhtsa.dot.gov/Pubs/809-293.pdf. VVO, 2015. Unfallursache Ablenkung: Herausforderung der Zukunft! Verband der Versicherungsunternehmen Österreichs, Retrieved February 22, 2016 from www.kvf.at/uploads/tx news/8. PI Ablenkung.pdf. Victor, T., Bärman, J., Boda, C.N., Dozza, M., Engström, J., Flannagan, C., Lee, J.D., Markkula, G., 2014. Analysis of Naturalistic Driving Study Data: Safer Glances, Driver Inattention, and Crash Risk. SHRP 2 Safety Project S08A, Retrieved January 28, 2016 from http://onlinepubs.trb.org/onlinepubs/shrp2/SHRP2 prepupS08Areport.pdf. Vollrath, M., Huemer, A.K., Nowak, P., Pion, O., 2014. Ablenkung durch Informations- und Kommunikationssysteme. Unfallforschung der Versicherer, Berlin. Wenners, K.E., Knodler, M.A., Kennedy, J.R., Fitzpatrick, C.D., 2013. Large-Scale observational study of drivers’ cell phone use. Transp. Res. Rec. J. Transp. Res. Board 2365, 49–57, http://dx.doi.org/10.3141/2365/07. Young, K.L., Rudin-Brown, C.M., Lenné, M.G., 2010. Look who’s talking! A roadside survey of drivers’ cell phone use. Traffic Inj. Prev. 11, 555–560. Young, R., 2015. Driver compensation: impairment or improvement? Hum. Factors 57, 1334–1338, http://dx.doi.org/10.1177/0018720815585053.