Revealing the Safety of the Road Environment  from Driver Responses: Investigation of Driver Behaviour under Specific Road  and Traffic Situations

Baojin Wang

This thesis develops an empirical approach to investigate a driver's  perception of safety and behavioural response when faced with specific road and  traffic situations. The roundabout was selected as a context for empirical  inquiry. We reviewed the literature on driver's perception of safety, driving  behaviour and the safety, driving behaviour and the safety properties of the  roundabout. A conceptual framework was proposed within which the effects of  attributes describing road and traffic situations can be empirically measured.  The stated-preference technique was used to investigate how attributes of  roundabouts and characteristics of drivers influence their perception of safety  and/or behavioural response. A statistical design was used to reduce the number  of combinations of attribute levels to a practical size and to ensure the main  effects of attributes can be observed and the effects of correlation among  attributes can be minimised. The experimentally designed road and traffic  scenarios were visualised using a video image system and developed in a  computerised survey instrument. A faced-to-face survey of a sample of Sydney  drivers provides the data used in model estimation.

The computerised survey instrument automatically recorded the time  that respondents allocated on each evaluated scenario and how they made use of  detailed information provided in interactive windows. These allowed us to  investigate how respondents assigned time and attention in a survey. We  identified three distinctive stages in the response process. At the beginning  of the survey, respondents learnt the task and spent a longer time on each  evaluation situation. After becoming familiarised with the survey task and  developing a response strategy, they allocated a reduced but relatively  constant amount of time on each evaluation situation. In the last stage, it  appeared that respondents became fatigued or somewhat lost interest in the  survey, thus a further reduced response time on each evaluation situation was  observed.

The thesis has two major contributions. The first is to investigate  preference equality and response consistency with the design and implementation  of a stated preference equipment. Two formats of the survey instrument, a Picture and Word format and a Picture Only format, were implemented in  a two-wave survey. An important aspect of survey design is the extent that the  medium used to present information (eg picture or word descriptions) acts as a  source of response bias, and the likelihood of response consistency over time  (eg in two surveys). We found that data evaluated with the Picture and Word format were statistically equal to the data  evaluated with the Picture Only format,  suggest that bias caused by the medium used for presenting information is not  significant for this study. Data obtained at the first wave of the survey are  statistically equal to data obtained at the second wave of the survey,  suggesting that the respondent's preferences are relatively stable over time.  The behavioural response variance in data obtained in the first wave of the survey  was consistently larger than that in the second wave of the survey, suggesting  that response consistency improves in a repeated survey. These findings not  only support the appropriateness of using stated-preference data for eliciting  driver's perception of safety and behavioural response in this study, but also  add to our knowledge of the appeal of the stated-preference technique in  general.

The second major contribution of the thesis is to investigate  drivers' perception of safety and behavioural response at specific road and  traffic situations. Ordered probit (logit) models were estimated to investigate  how attributes of roundabouts and characteristics of drivers influence the  perception of safety. An important output is an indicator of perceived safety (IPS). We found that attributes  describing a roundabout and its traffic situation in addition to the  characteristics of drivers have a significant influence on the perception of  safety. The IPS is very sensitive to  the change in the levels of attributes such as the size of a roundabout, the  number of circulating lanes, visibility to other traffic, size of a potentially  conflicting vehicle, general traffic level at a roundabout, presence of a  pedestrian and the speed of vehicles, suggesting that these attributes are  important determinants of a driver's perception of safety. Given a road and  traffic situation, the IPS varies  among different drivers, suggesting that the heterogeneity property of the  perception of safety between drivers with different socio-economic  characteristics and driving experience.

To investigate behavioural response at specific roundabouts under  specific traffic situations, we estimated multinomial logit and mixed logit  models. The mixed logit model permits us to account of heterogeneity in  preference parameters and to examine choice set correlation and correlation  between alternatives. We found that correlation between some pairs of  attributes was statistically significant. However, once individual  heterogeneity in means was taken into account, the correlation was negligible,  suggesting that correlation could be spurious due to a failure to account for  unobserved heterogeneity. Estimation results suggest that drivers tend to  select a less cautious behavioural response when facing a perceptually safer  driving environment. The simulated probabilities based on estimated models  suggest that attributes describing a roundabout and its traffic situation have  a significant influence on a driver's behavioural response. Obstructed  visibility, relatively fast speed of a potentially conflicting vehicle, presence  of a potentially conflicting pedestrian, a large-sized potentially conflicting  vehicle, busy traffic at a roundabout and multi-circulating lanes contribute to  a driver's choice of stopping or slowing down before a roundabout. On the  other hand, light traffic at a roundabout, a small-sized potentially  conflicting vehicle, relatively slow speed of a potentially conflicting  vehicle, a small-sized roundabout and a driver in a hurry contribute to his or  her choice of not slowing down response before a roundabout. A driver's socio-economic characteristics and  driving experience also has a significant influence on their behavioural  response. Relatively inexperienced drivers (less than 5 year driving  experience) and more experienced drivers (more than 15 year driving experience)  are less likely to stop or slow down before a roundabout. Drivers  involved in an accident in the last two years are less likely to stop or slow down when approaching a roundabout. Low-income drivers (less  than $30,000) are more likely to stop or slow down before a roundabout.  Commuter drivers are less likely to stop or slow down when approaching a  roundabout. Young drivers (25 years or younger) are more likely to stop or slow down when approaching a roundabout, but male young drivers are  less likely to stop or slow down before a roundabout.