16
Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271. © 2018 ATS All rights reserved 256 Public Acceptance of Designated Driver Services for Alcohol-Drinking Drivers in Hanoi Cong Duc PHAN a* , Makoto CHIKARAISHI b , Akimasa FUJIWARA c , Junyi ZHANG d a Transport Development and Strategy Institute, Hanoi, Vietnam; E-mail: [email protected] b Graduate School for International Development and Cooperation, Hiroshima University, Higashi-Hiroshima, 739-8529, Japan; E-mail: [email protected] c Same as the second author; E-mail: [email protected] d Same as the second author; E-mail: [email protected] Abstract: To reduce the impact of drunk-driving behavior on accident risks, this study made an initial attempt to evaluate the public acceptance of designated driver services (DDS) in the Vietnamese context. A stated preference (SP) survey was implemented to capture more than 300 car and motorcycle drivers’ preferences for various hypothetical DDS in Hanoi. The collected SP data were analyzed by employing a panel mixed logit model with both social interactions and heterogenous tastes. Positive social interactions leading the choice behavior to an equilibrium were found. This new service seems more effective to meet the needs of middle-aged people and office workers, who have the highest risk of drunk driving among all population groups. The modeling results further suggest that to reduce the prevalence of drunk-driving behavior and encourage people to use the new DDS, the government needs to tighten law enforcement, raise drunk-driving penalties, maintain the low cost of DDS, and educate people to internally mitigate their negative attitudes about drunk-driving behavior. Keywords: Designated Driver Services, Drunk-Driving Behavior, Panel Mixed Logit Model, Social Interaction, Hanoi City 1. INTRODUCTION In 2015, around 5.3 million motorcycles and 0.6 million cars were registered in Hanoi, which is the central city for business, commerce, and tourism in Vietnam (Tuan, 2015). The average ownership of motorcycles is nearly 0.7 motorcycles per person. On the one hand, the growth rates for motorcycle and car use are 11% and 17%, respectively, which is 1.5 to 2 times higher than the growth rate of Gross Domestic Product (GDP). On the other hand, the city operated a fleet of only 1,145 buses with 80 routes on main roads in 2014. In addition, the buses were operated less frequently; i.e., only 10% of the routes had a peak period headway being shorter than 10 minutes and the remaining 90% had a headway of 10–15 minutes or even longer (Bray et al., 2016). In comparison to other Asian megacities, the share of cars and motorcycles in Hanoi reached 88%, which is 8 times higher than the 11% share of buses (Tuan, 2015). The combination of lower transit share and higher motorcycle ownership makes Hanoi a highly motorcycle-dependent city. Consequently, traffic congestion and accidents occur on main streets daily in addition to serious air pollution. Among these issues, traffic * Corresponding author.

Public Acceptance of Designated Driver Services for

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Public Acceptance of Designated Driver Services for

Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271. © 2018 ATS All rights reserved

256

Public Acceptance of Designated Driver Services for Alcohol-Drinking Drivers in Hanoi Cong Duc PHAN a*, Makoto CHIKARAISHI b, Akimasa FUJIWARA c, Junyi ZHANG d a Transport Development and Strategy Institute, Hanoi, Vietnam;

E-mail: [email protected] b Graduate School for International Development and Cooperation, Hiroshima

University, Higashi-Hiroshima, 739-8529, Japan; E-mail: [email protected]

c Same as the second author; E-mail: [email protected] d Same as the second author; E-mail: [email protected] Abstract: To reduce the impact of drunk-driving behavior on accident risks, this study made an initial attempt to evaluate the public acceptance of designated driver services (DDS) in the Vietnamese context. A stated preference (SP) survey was implemented to capture more than 300 car and motorcycle drivers’ preferences for various hypothetical DDS in Hanoi. The collected SP data were analyzed by employing a panel mixed logit model with both social interactions and heterogenous tastes. Positive social interactions leading the choice behavior to an equilibrium were found. This new service seems more effective to meet the needs of middle-aged people and office workers, who have the highest risk of drunk driving among all population groups. The modeling results further suggest that to reduce the prevalence of drunk-driving behavior and encourage people to use the new DDS, the government needs to tighten law enforcement, raise drunk-driving penalties, maintain the low cost of DDS, and educate people to internally mitigate their negative attitudes about drunk-driving behavior. Keywords: Designated Driver Services, Drunk-Driving Behavior, Panel Mixed Logit Model,

Social Interaction, Hanoi City 1. INTRODUCTION In 2015, around 5.3 million motorcycles and 0.6 million cars were registered in Hanoi, which is the central city for business, commerce, and tourism in Vietnam (Tuan, 2015). The average ownership of motorcycles is nearly 0.7 motorcycles per person. On the one hand, the growth rates for motorcycle and car use are 11% and 17%, respectively, which is 1.5 to 2 times higher than the growth rate of Gross Domestic Product (GDP). On the other hand, the city operated a fleet of only 1,145 buses with 80 routes on main roads in 2014. In addition, the buses were operated less frequently; i.e., only 10% of the routes had a peak period headway being shorter than 10 minutes and the remaining 90% had a headway of 10–15 minutes or even longer (Bray et al., 2016). In comparison to other Asian megacities, the share of cars and motorcycles in Hanoi reached 88%, which is 8 times higher than the 11% share of buses (Tuan, 2015). The combination of lower transit share and higher motorcycle ownership makes Hanoi a highly motorcycle-dependent city. Consequently, traffic congestion and accidents occur on main streets daily in addition to serious air pollution. Among these issues, traffic * Corresponding author.

Page 2: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

257

accidents are a very urgent problem that has attracted much attention from both policy makers and the public. In 2002, there were more than 30,000 accidents with around 27,000 injuries. Since then, a downward trend has been observed because of the various policies for traffic safety (Government of the Socialist Republic of Vietnam, 2002). The numbers were reduced by a third after 10 years to about 10,000 accidents with 8,000 injuries. However, the number of deaths has remained stable during the previous decade at around 10,000 deaths per year. Since 2005, the number of deaths has surpassed the number of injuries, indicating that despite the reduction in the number of accidents and injuries, the number of deaths has remained at a higher level.

There are many reasons for such a higher rate of deaths. The World Health Organization (2009) reported that road traffic deaths involving drunk driving accounted for 34% of all cases. Therefore, examining alcohol usage issues may reveal a good solution to reduce the severity of accidents.

In 2014, the total amount of alcoholic beverages production in Vietnam was 3.89 billion liters of beer, which was a 234% increase compared to that in 2004 (Kirin Holdings Company, Ltd., 2015a). Over 10 years, Vietnam grew from the 27th biggest beer producer worldwide to become the 10th biggest and is ranked 3rd biggest in Asia after China and Japan. Thus, beer production is a very large business that brings several benefits to beer manufacturers in Vietnam. In addition, there is a rapidly increasing trend in the demand for alcoholic beverages in Vietnam, with 3.64 billion liters of beer consumed in 2014 (Kirin Holdings Company, Ltd., 2015b).

Alcohol has been consistently shown to be a major risk factor for road traffic crashes resulting in injuries and death. Studies have examined that alcohol acts through multiple pathways in increasing the risks for road traffic crashes and injuries. Alcohol can impair judgment and increase crash risk even at relatively low blood alcohol concentration (BAC) levels (Lang, 1992). However, the effects become progressively worse as the BAC increases. Apart from its direct impact on crash outcomes, alcohol is believed to affect other aspects of driver safety such as seat-belt wearing, helmet use, and speed choice. Drunk driving has therefore been linked with both an increased likelihood of a crash and increased severity.

In October 2012, the Vietnamese Government issued Decree No. 71, which enacted higher fines stipulating a maximum penalty for motorcyclists up to 3 million VND (approximately US$150), loss of license for 60 days and vehicle impoundment for 10 days. For all other vehicle drivers, the refusal of a BAC or breath alcohol concentration (BrAC) test attracts a fine of up to 15 million VND (approximately US$750), loss of license for 60 days, and vehicle impoundment for 10 days. The law states that if a road crash fatality was a result of drunk driving, then the driver is subject to a maximum jail time of 10 years. However, prosecution is rare and without greater policy enforcement and monitoring, the prevalence of drunk driving is likely to remain high.

As discussed above, with a very high frequency of using private vehicles (mostly motorcycles) and high consumption of alcoholic beverages, drinkers in Vietnam, in general, and those in Hanoi, in particular, prefer to ride their own vehicles during frequent drinking trips. Moreover, due to the lack of parking capacity at the restaurant for private vehicles after drinking, drinkers usually must take their own vehicles back home. Therefore, drunk driving is very common among drinkers in Vietnam.

Reducing the impact of drunk-driving behavior on accident risk when constrained by low parking capacity should be balanced with preserving the development of the alcoholic beverage industry; therefore, an alternative transportation mode should be introduced to help drinkers return home in their vehicles after drinking but without having to drive their own vehicles.

Page 3: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

258

One of the potential alternative transportation modes is a type of taxi service, which was originally called the SafeRide (SR) program in the US (Caudill et al., 2000). The SR service is operated almost the same as a normal taxi service. However, the difference is this taxi service usually comes with 2 drivers, 1 driver will drive the drinkers’ vehicle (with the drinker in it) back to their house and the other driver will follow in the taxi. Therefore, this type of service can accommodate both the alcohol-drinking drivers and their vehicles. In the scope of this study, this type of taxi service is called designated driver services (DDS) based on the name of a real service being operated in the US and Canada. In other countries, the names of such taxi service vary, such as Daiko taxi in Japan and Dae-ri Un-jeon (i.e., the replacement of drivers) in Korea. Unfortunately, no services like DDS or relevant studies can be found in Vietnam.

By introducing a new type of DDS in Hanoi, this study aims to evaluate the public acceptance of this service by alcohol-drinking drivers based on a stated preference (SP) survey. The final goal of this research is to propose proper policies to reduce the prevalence of drunk driving and promote the use of DDS.

The remainder of this paper is organized as follows. A brief review of literature on SR programs and social interaction follows in the next section. The third section explains the methodology, i.e., estimating the panel mixed logit model with the heterogeneity of social interaction. The SP survey and data-collection process in Hanoi are described in the fourth section followed by an overview of the data with aggregation analysis. The simulation results are presented in the fifth section, considering equilibrium points found at a place where the DDS choice (model output) is equal to the expected DDS market share (model input). Furthermore, some simulations were implemented to examine the effect of different policies. Finally, conclusions, limitations, and future research issues are presented in the last section. 2. LITERATURE REVIEW 2.1 SafeRide Program Two measures have been proposed to decrease alcohol-impaired driving: i.e., designated driver (DD) and SR programs. DD programs require a group of individuals to plan who will drive whom after drinking, so that the DD ideally does not consume any alcohol. SR programs provide an alternative means of transportation such as taxis, vans, and limousines, or provide a driver to drive the drinking individual’s car home (Caudill et al., 2000). Therefore, the DDS used in this study is not a DD program but like an SR program.

The earliest start date for the SR program can be traced back to 1979. The program spread across the US in the following years (Harding et al., 1988). In the early years, SR programs were used less extensively compared with DD programs (Lavoie et al., 1999). However, many studies found that SR programs could reduce the number of intoxicated drivers on roads (Rivara et al., 2007; Caudill et al., 2000; Apsler, 1988). A published report on SR services was a “bar room” survey conducted by Caudill (1990) in which at-risk drinkers indicated they would be likely to use an SR program if it was introduced in their communities. Of the 1,522 bar room patrons surveyed by the same author in two California communities, 79% said they “might” to “definitely would” use SR programs if they were available.

Many previous studies examine the SR program’s potential users to obtain a deeper understanding the program. Molof et al. (1995) studied two SR programs and described SR users as being younger in age (most were 26 to 35 years old), male, and individuals who drink

Page 4: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

259

more frequently and consume more alcohol when they do drink. Sarkar et al. (2005) also identified this group as being more likely to engage in drunk driving. Drivers aged 21–24 years showed the greatest level of alcohol-related fatalities (33%), followed by the 25–34 (28%) and 35–44 (25%) age groups. Male drivers were also more likely than female drivers to be involved in a fatal crash with a BAC of 0.08 g/dl or higher. The results of Rivara et al.’s (2007) research confirm the earlier surveys on frequent hazardous drinking by young adults aged 21–34 years. Forty percent of people in this age group had binged on alcohol at least once in the last 12 months and many of them drank more than once a week (usually at some place other than their own home); therefore, it is not surprising that 1 in 5 drove after drinking too much in the last month. Moreover, the quantity and frequency of drinking were the factors most strongly associated with drunk driving. Other factors such as age, education, and income were much less important predictors.

Most SR passengers studied by Molof et al. (1995) were judged by taxi drivers as being “moderately to severely impaired.” In situ breath testing of 17 SR passengers in a holiday-based program in Minneapolis revealed an average BAC of 0.096; however, only five of the passengers tested (29%) had driven the night they were tested. In other research (Caudill et al., 2000), in situ breath-test data showed that SR users’ BAC ranged from 0.031 to 0.248, with a mean of 0.146. Caudill et al. (2000) found that users of SR programs reported higher levels of intoxication, were more likely to drink outside the home, and had been arrested in the past for driving while intoxicated (DWI).

Shore and Sanchez (1993) reported the assessment of 30 SR users by SR service drivers who described their passengers as typically male, most often alone, and moderately to severely intoxicated. Shore and Sanchez (1993) noted that this population, namely the “moderately drunk lone male in his twenties”, is overrepresented in DWI-related incidents and that the SR services may reach a population of drinkers that other prevention efforts have been missing. However, concerns have been raised that the SR programs users increase the amount of alcohol consumed (Harding et al., 2001). 2.2 Social Interaction Over the past few decades, a considerable number of studies in the social sciences have examined the role of social interaction in human behavior (Fukuda and Morichi, 2007; Asch, 1951). In general, the term “social interaction” refers to the utility or payoff that an individual derives from his/her action, which is dependent on the actions of other members in his/her reference group (Brock and Durlauf, 2001).

When the utility of an action is higher for an individual if others behave similarly, there is a positive social interaction among individuals. This phenomenon is known as “conformity and peer effects” in the field of social psychology (Asch, 1951), “bandwagon effects” in economic literature (Leibenstein, 1950), and “strategic complementarity” in game theory (Cooper, 1999). Regarding the collective behavior of individuals, positive social interaction induces a tendency to conform to the majority, which sometimes results in undesirable social conditions. Smoking, crime, spreading rumor or disease, opinion formation, school dropout, and voting behavior are typical examples of the kinds of social behaviors thus created (Granovetter, 1978).

Previous studies showed social interactions in the transportation field, such as illegal bicycle parking (Fukuda and Morichi, 2007), illegal car parking (Muromachi et al., 2004), and affirmative rejection of the implementation of road pricing (Jakobsson et al., 2000). Therefore, the current study investigates the effect of social interactions on drunk-driving behavior in Vietnam.

Page 5: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

260

Brock and Durlauf (2001) described the general framework of a binary choice model with social interactions. Their model has been proposed as a tool for understanding the collective behavior that emerges from interdependencies among individuals. They also included a brief overview of the empirical studies of social interactions that range from educational achievement to crime. Brock and Durlauf (2001) emphasized that empirical studies should be elaborated continuously. This approach can be used to analyze travel behavior, which is a field where the study of the role of social interaction has been relatively superficial to date. 2.3 Framework Analysis Based on the literature review, this study presents a framework for analyzing individual drivers’ DDS choices (Figure 1). Factors affecting these DDS choices were classified into external and internal factors. External factors consist of drivers’ sociodemographic characteristics, their current trip attributes, habitual drinking behavior, and the level of services for the nonexistent DDS. In addition, the effects of social interaction may not be ignorable because the expected DDS market share (i.e., other anonymous drivers’ choices) may also influence drivers’ choice of DDS. Thus, many drivers’ DDS choices will change the expected market share for DDS and this loop may continue until it reaches an equilibrium point where the DDS choices equal the expected DDS market share. However, the internal factors are the negative and risk-adverse attitudes about DWI. These attitudes were indicated by subjective answers. The internal and external factors are simultaneously associated with the individual’s choice preference and intention to use a DDS.

Figure 1. Framework of drivers’ DDS choices in this study

3. METHODOLOGIES: ESTIMATING DDS CHOICES In the practice of SP surveys, each respondent is requested to answer several SP cards repetitively, which allows us to identify the heterogeneity between individuals by using a panel modeling approach, i.e., a panel mixed logit model. Here, we consider heterogeneous

Page 6: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

261

responses with respect to the social interaction terms between individuals. Discrete choice models with social interactions are based on the recent developments in

microeconometrics, which were summarized as interaction-based models (Brock and Durlauf, 2001). Recently, an increasing number of empirical studies included an assessment of social interactions. This study formulates a binary choice model between the drivers’ current transportation mode and the newly introduced DDS with social interactions, which allows the endogenous relationship among individuals’ choices to be analyzed. In addition, the collective behavior of the driver’s reference group is formulated mathematically.

A binary action of an individual on the t-th SP card is coded into the binary variable , which takes a value equal to 1 if an individual chooses DDS, and −1 if he/she does not

choose DDS (i.e., the individual chooses a private car or motorcycle). The total indirect utility of an individual ’s action on the t-th SP card, coded by , is assumed to be partitioned into three components: individual-specific utility , social utility , and unobserved individual-specific utility . The existence of social utility enables us to incorporate the choice results of other agents into an individual choice. The total indirect utility is given in Eq. (1):

(1) where , measured in terms of social utility, denotes the subjective expected value of choices in the population. The support is between −1 and 1, in accordance with the definition of . The expected proportion of the population choosing DDS, which is an expectation of the DDS share in the model input (range from 0 to 1), is established by . The random utility term is assumed to be independently and identically distributed (IID) across samples.

The individual-specific utility is defined as , where is a random variable assumed to be normally distributed with the mean and variance , is a parameter vector, and is a vector of individual-specific explanatory variables. The random variable is introduced to capture the unobserved interindividual heterogeneity, which may correspond to the correlated effects stated by Manski (1993).

We also consider the unobserved heterogeneity across individuals for the social utility term as follows:

(2) where is a random variable assumed to be normally distributed with the mean and variance , and represents the heterogeneous response to the average choice in the population. The properties of are explained in Brock and Durlauf (2001) under the assumption of the homogeneous response. When the heterogeneous responses are assumed, it may be difficult to calculate the equilibrium points, although the basic features of the model are similar with that in Brock and Durlauf (2001) in the sense that a multiplicative interaction between the individual and expected average choice in the market allows us to measure the degree of social interaction.

Assuming is IID Gumbel distributed, a binary logit model of individual choice probability incorporating the expected average choice of the driver’s reference group can be derived as:

(3)

Page 7: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

262

where is the scale parameter of the random utility component .

The expectations of the average choices are then determined. The assumption that all individuals possess rational expectations leads to a situation where an individual’s subjective belief concerning the average choice is equal to its mathematical expectation, while the response to the average choice could vary between individuals.

As a preliminary calculation, the individual-specific utility for each alternative was replaced as and ), respectively, where and are transformation parameters. Using this linearization, the mathematical expectation of each individual choice, given his/her subjective beliefs regarding the average choice of others, was described as:

(4)

This study uses a standard maximum likelihood estimator (with Gaussian quadrature approximation for random terms). Note that some studies use structural estimation methods such as the nested pseudo likelihood method (Aguirregabiria, 2004), which may not be suitable for SP survey data, since structural estimation methods assume the observed data is in the equilibrium state: the SP data is obviously not the case.

After the model estimation, the next step is to evaluate the impacts of social interactions: when the social interaction exists, policy measures promoting DDS will be amplified to some extent until reaching a certain equilibrium point. To identify the equilibrium point, this study uses Fukuda and Morichi’s (2007) approximation method, which can be written as follows:

(5) where the hat indicates the estimated parameters and associated explanatory variables. Solving this equation with respect to , the expected average choice at the equilibrium point can be calculated. Note that, for random variables, we use the so-called shrinkage estimates to obtain the individual specific parameters (Pinheiro and Bates, 1995). 4. SURVEY ON PUBLIC ACCEPTANCE OF DESIGNATED DRIVER SERVICES IN

HANOI 4.1 SP Survey To collect drivers’ behavioral intentions for a nonexistent DDS in Hanoi, this study employed an SP survey method, which measured the reactions of car and motorcycle drivers to various hypothetical DDS scenarios.

Based on the research framework in Figure 1, the questionnaire survey was designed in four sequential parts. In the first part, after defining DDS to help the respondents have a clearer imagination, they produced a diary for their activity on the last day that they drove a motorcycle or car after outside drinking within 1 month. The questions also included weather,

Page 8: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

263

trip purpose, time, place, distance, mode choice, and cost. The second part collected the alcohol-related details for 1 chosen trip from the diary, including gender and weight (to calculate the BAC), amount of alcohol consumed, recognition of penalty for drunk driving, and the respondent’s drinking habits. The third part is an SP experiment in which eight SP cards combining the 12 different attributes with 2 or 3 levels (Table 1) are presented to each respondent based on their reported travel modes. The questions examine the respondents’ intentions to choose their current travel modes or shift to DDS under the condition of their chosen trip attributes and the hypothetical attributes of the nonexistent DDS. The current distance of the chosen trip was used to calculate the cost of DDS.

Table 1. Attributes and their levels in the SP experiment Attributes Levels Expected share in the market (Private vehicle/DDS) 10%/90% 50%/50% 90%/10%

Current transportation mode (motorcycle) Average penalty 500,000 VND 1,500,000 VND 3,000,000 VND Likelihood of punishment by police 1 per 100 times 1 per 20 times 1 per 5 times License suspension No 2 months 6 months Report to employers No Yes — Current transportation mode (car) Expected penalty 2,000,000 VND 7,500,000 VND 15,000,000 VND Likelihood of punishment by police 1 per 100 times 1 per 20 times 1 per 5 times License suspension No 2 months 6 months Report to employers No Yes — DDS Waiting time 5 mins 15 mins 30 mins Fare 5,000 VND/km 15,000 VND/km 25,000 VND/km Vehicle type Motorcycle (1 seat) Car (4 seats) —

Following the above three parts, the final part gathers drivers’ attitudes on drunk-driving behavior, including individual attitudes about DWI, subjective norms, and perceived behavioral control, and their sociodemographic information. 4.2 Hypothetical Attributes for SP Design The SP questions about DDS were designed based on an orthogonal fractional factorial design method for reducing the number of combinations but ensuring that the attributes represented to respondents varied independently of each other. The SP techniques are characterized using experimental designs to set up the hypothetical alternatives and attributes presented to respondents. This avoids multicollinearity between attributes, which is a common problem with revealed preference data.

The level of average penalty and license suspension attributes were calculated based on the current drunk-driving law in Vietnam (Government of the Socialist Republic of Vietnam, 2013). Monetary values of fare and waiting time were introduced by considering standard taxi services. Meanwhile, the other attributes (expected DDS market share, report to employers, likelihood of punishment by police, and vehicle type) were presented with hypothetical levels (Table 1).

Page 9: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

264

4.3 Survey Data Profiles In October 2015, the survey was conducted in 12 Hanoi urban districts with the support of students from the University of Transport and Communication. People who drove their private vehicles after an outside drinking trip within 1 month to the time of the survey were identified and then interviewed in person. The collected data profile is outlined in Table 2. The sample of 303 respondents were 92% and 8% motorcycle and car drivers, respectively. This percentage ratio reflects the real-world situation of a higher motorcycle ownership rate than for cars in Hanoi.

There are two groups of exogenous variables in the panel mixed logit model applied in this study: i.e., external and internal factors. Seventy percent of total trips are because of one of the DDS attributes (external factor); i.e., the main purpose of the drinking trip is to hang out with friends. The trip distance varied within 3 categories (£2 km, from >2 km to 5 km, and from >5 km to 10 km) roughly the same at 30%, while the longer trips only account for 5%. Interestingly, from these survey results, we found that no respondents went out to drink alone. Thus, the average number of people accompanying the drinking respondent was around 3.89 persons.

Based on the alcohol consumption, drinking duration, sex, age, and weight data derived from the questionnaire, we applied the Widmark formula to approximate the BAC index of respondents (Widmark, 1981). More than half of the respondents (55%) had a high level of BAC (i.e., 80 mg/100 mL), which sharply increases the accident risk when drunk driving. Collating the BAC index calculated above and the drunk-driving law, we found that only one fourth of respondents did not commit an illegal act. In contrast, the other 74% did violate the drunk-driving law on many different levels. However, only few reported that the police had caught them drunk driving.

The internal factors employed in this study are two latent attitudes: i.e., “negative attitude about DWI” and “risk-averse attitude to DWI.” These factors were derived by confirmatory factor analysis as described in the following subsection 4.4.

After considering the scenario in SP cards carefully, respondents decided between choosing their current transportation mode or the new DDS. The results are the exogenous variables in the panel mixed logit models. As shown at the bottom of Table 2, for motorcycle drivers, the number of respondents who chose the current transportation mode was slightly larger (52.7%) than those who shifted to DDS (47.3%). About 24% more car drivers decided to change to DDS than those who chose their current transportation mode (38.1%). 4.4 Internal Factors of Driver Attitudes about DWI Individual attitudes toward drunk-driving behavior were measured with 5 rating scores each in 15 subjective DWI-related questions. For example, a little more than 90% of drivers strongly disagreed or disagreed with Q48 (“DWI feels fun”). Thus, most drivers have negative attitudes about DWI. Moreover, because almost 90% of drivers strongly agree or agree with Q50 (“DWI increases the risk of severe accident”), many drivers recognize the risk to their safety and are risk-averse. However, the responses to Q46 (“Self-driving after drinking helps you to arrive at your destination more quickly”) show that although the disagreement rate was higher than the agreement rate, both agree and disagree responses lacked a majority. Attitudes about DWI were found to be heterogeneous between different drivers, but somehow inconsistent within an individual.

Page 10: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

265

Table 2. Survey data profiles

Endogenous Variables of Panel Mixed Logit Models Motorcycle drivers Car drivers

Mean SD Mean SD

External factors

Alternative attributes set in SP questions

Expected DDS market share (%) 39.7 32.9 42.3 34.1

Likelihood of punishment by police (%) 6.7 7.8 6.7 7.9

Penalty (mil VND) 1.382 1.03 6.42 5.041

License suspension (months) 1.957 2.43 2.193 2.536

Report to employers (dummy), 1: yes (%) 50.5 51.7

DDS fare (thousand VND) 12.397 8.28 13.182 8.148

DDS waiting time (mins) 13.597 10.2 14.943 10.378

DDS vehicle type (dummy), 1: car (%) 50.5 46.6

Trip attributes Drinking purpose (dummy), 1: hang out with friends (%)

81.8 43.2

Distance (km) 4.76 3.7 6.8 8.592

Number of companions (persons) 3.94 4.26 2.881 1.664 High BAC (dummy), 1: higher than 80 mg/100 mL (%)

56.4 47.7

Sociodemographic characteristics

Gender (dummy), 1: male (%) 81.9 84.1

Age (dummy), 1: 20 to 50 years old (%) 86.7

Age (dummy), 1:50 to 70 years old (%) 9.1

Marital status (dummy), 1: married (%) 40.4 90.9

Job (dummy), 1: office worker (%) 31.7 Educational status (dummy), 1: bachelor, master, or PhD

69.8 65.3

Individual income (million VND) 6.7 5.895 18.665 13.531 Drunk-driving habits (dummy), 1: high frequency (%)

53 61.4

Internal factors Attitudes

Negative attitude about DWI 11.7 2.65 16.842 4.138 Risk-averse attitude to DWI 5.79 2.84 1.067 5.073

Endogenous Variables of Panel Mixed Logit Models Current DDS Stated intention to use DDS by motorcycle drivers

(%) 52.7 47.3

Stated intention to use DDS by car drivers (%) 38.1 61.9

To overcome the above problem, we performed confirmatory factor analysis to identify the latent factors behind the attitudes observed in the responses. The estimation results of the

Page 11: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

266

factor analysis are shown in Table 3 with high goodness-of-fit indicators in both motorcycle and car drivers. In the case of motorcycle drivers, their negative attitudes about DWI are logically and significantly indicated by three representative answers, but their risk-averse attitudes are indicated by two other factors. Those two attitudes will be used as explanatory variables of the DDS choice model estimated in the following section.

Table 3. Confirmatory factor analysis of drivers’ attitudes about DWI

Direct effects Standardized coefficient Motorcycle Car

Q48 DWI is fun <– Negative attitude about DWI

0.693 * 0.781 *

Q47 DWI makes you feel independent and free <– Negative attitude about DWI

0.867 *** 0.641 ***

Q46 Self-driving after drinking helps to arrive at my destination more quickly <– Negative attitude about DWI

0.584 *** 0.972 ***

Q50 DWI increases the risk of severe accidents <– Risk-averse attitude about DWI

0.370 *** 0.270 ***

Q45 You can bribe the police to escape <– Risk-averse attitude about DWI

−0.266 *** −0.880 ***

Q49 DWI is stressful <– Risk-averse attitude about DWI

0.800 *** 0.753 ***

Adjusted goodness of fit index 0.922 0.797 Root mean square error of approximation 0.092 0.000 5. MODEL ESTIMATION RESULTS The estimation results for the panel mixed logit models are shown in Table 4. The results show that the likelihood of punishment by the police and drunk-driving penalties have a significantly negative impact on DDS choices by both motorcycle and car drivers. Thus, people may avoid drunk driving if the government proposes stronger law enforcement activities such as increasing the frequency of police checks and higher penalties. This study proposes a new policy towards drunk driving; i.e., our research results show that the number of drunk motorcycle drivers may decline if they are reported to their employer after being caught by the police. However, the significantly negative value for the fare variable indicates that motorcycle users are sensitive to the new taxi service fares. License suspension, DDS waiting time, and the DDS vehicle type show no statistical significance in both models.

In terms of trip attributes, the dummy for drinking purpose (hang out with friends) and distance are both negatively significant in the motorcycle model, while nonsignificant in the car model. The results show that motorcycle users who go out drinking with friends in a distant place do not prefer to use DDS for their return trip. One of the reasons for this finding may be because of the increasing cost of DDS when the distance is increased. The significance of the parameter “number of companions” indicates that people may prefer to use DDS when going out drinking in a large group. Unfortunately, the dummy for high BAC in these results shows a low confidence level only for the car model.

Page 12: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

267

Table 4. Panel mixed logit model with social interaction considering heterogeneity across individuals

Explanatory Variables Motorcycle drivers Car drivers Parameter z value Parameter z value

Intercept −4.630 ** −3.743 2.670 0.447 Alternative attributes Expected DDS market share (%) 1.184 ** 4.068 2.776 + 1.700 Likelihood of punishment by the police (%) 34.220 ** 15.536 0.413 ** 3.475

Penalty (million VND) 0.133 ** 13.112 0.456 ** 3.854 License suspension (months) −0.001 −0.027 −0.022 −0.135 Report to employers (dummy), 1: yes 0.405 * 2.141 0.122 0.148 DDS fare (thousand VND) −0.116 ** −9.211 −0.153 * −2.370 DDS waiting time (mins) −0.002 −0.253 −0.022 −0.579 DDS vehicle type (dummy), 1: car 0.326 + 1.668 1.146 1.186 Trip attributes Drinking purpose (dummy), 1: hang out with friends −1.033 * −2.545 −4.099 + −1.826

Distance (km) −0.213 ** −4.583 −0.276 −1.597 Number of companions (persons) 0.033 0.818 High BAC (dummy), 1: higher than 80 mg/100 mL 0.177 0.499 −6.977 * −2.305

Sociodemographic attributes Gender (dummy), 1: male −0.185 −0.396 −4.841 −1.116 Age (dummy), 1:20 to 50 years old 0.995 * 1.991 Age (dummy), 1:50 to 70 years old 12.421 ** 2.779 Marital status (dummy), 1: married 0.105 0.280 2.130 0.671 Job (dummy), 1: office worker 0.728 + 1.748 Educational status (dummy), 1: bachelor, master, or PhD −0.607 −1.365 5.035 + 1.796

Individual income (million VND) −0.058 + −1.855 −0.361 ** −2.703 Drunk-driving habits (dummy), 1: high frequency −1.030 ** −2.743 7.307 * 2.255

Attitudes Negative attitude about DWI 0.334 ** 3.622 0.110 0.368 Risk-averse attitude about DWI −0.023 −0.158 −0.081 −0.426 Random effects within an individual Var. SD Corr. Var. SD Corr. Intercept 6.015 2.453 1.848 1.359 Expected DDS market share (%) 4.34 2.083 −0.38 22.434 4.736 1.00 Log-likelihood at convergence −795.5 −46.7 Rho-squared 0.47 0.62 Akaike information criterion 1641 139.4 Number of observations 2167 176 Note: (**) Significant at 1% level; (*) Significant at 5% level; (+) Significant at 10% level

The 20–50 years age group in the motorcycle model and 50–70 years age group in the car model were both positively significant, which suggests that they may be the potential DDS user group. Car drivers with higher individual income prefer to use a car after drinking because this parameter is negatively significant. The reason for this finding is that the high-income groups appear to be less sensitive to drunk-driving penalties. However, motorcycle and car drivers who habitually drunk drive with high frequency also tend to continue to act illegally.

Page 13: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

268

In terms of external factors, the variable of risk-adverse attitude shows no significance in both models. In a thought-provoking result, however, drivers with a negative attitude about drunk driving chose the motorcycle DDS model. Therefore, this finding implies that some sort of education or communication policies to foster these negative attitudes about DWI may encourage people to use the DDS and by extension prevent traffic accidents in Hanoi.

In terms of the social interaction effects, the expected DDS market share (%) was confirmed to positively affect both motorcycle and car drivers’ use of DDS. Simultaneously, nonnegligible unobserved variations were found between individuals, which indicates that there would be variation in individuals’ responses to DDS in the market. 6. SIMULATION ANALYSIS OF THE DDS MARKET SHARE UNDER DIFFERENT

POLICY SCENARIOS As mentioned, social interaction causes the DDS market share to vary nonlinearly. For example, as the DDS market share increases, individual choice probability will nonlinearly increase, which is often called amplified effects. In particular, the panel mixed logit models developed in the previous section can promise different equilibrium points under policy scenarios. In this section, we attempted to calculate such equilibrium states considering social interactions in the panel mixed logit model.

Figure 2 initially illustrates the relationship between the expected DDS market share and the DDS market share outlined by the motorcycle and car driver choice probabilities, respectively. The intersections between the two lines show the equilibrium points. In this case, the equilibrium points were around 40% and 65% for motorcycle and car drivers, respectively, at which point the drivers choose to use DDS.

(a) Motorcycle (b) Car

Figure 2. Equilibrium points for motorcycle and car drivers

Figure 3 compares the effects of a policy that prescribes a monetary penalty for DWI between two cases of “equilibrium” and “nonequilibrium,” where equilibrium refers to the average choice probability after solving eq. (5) with respect to , while nonequilibrium refers to the average choice probability without solving eq. (5), i.e., the expected DDS market share is considered as an exogenous variable. The solid and dotted lines in the figure represent the equilibrium and nonequilibrium cases, respectively.

Page 14: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

269

(a) Motorcycle (b) Car

Figure 3. Impact of drunk-driving penalties on DDS choices by motorcycle and car drivers

Figure 3 shows that the effect of drunk-driving penalties on DDS choices differs between the equilibrium and nonequilibrium cases. In general, in the equilibrium cases for motorcycle and car drivers, the effect of the drunk-driving policies on DDS choice on the solid line becomes lower in the beginning compared with the dotted line. However, due to the positive feedback effects of social interactions, the increasing speed of DDS choices in the equilibrium case are higher and exceeds the nonequilibrium cases at the equilibrium points of penalties of around 2 million VND and 5 million VND, respectively. 7. CONCLUSIONS This study is the first to deal with a new type of DDS and social interactions in the Vietnamese context. Analyses were based on SP data and a panel mixed logit model. The results suggest that to reduce the prevalence of drunk-driving behavior and encourage people to use the new DDS, the Vietnamese government needs to strengthen law enforcement activities, such as increasing the number of random alcohol checkpoints, introducing a higher penalty for drunk-driving behavior and a warning system that reports the actions of drunk drivers to their employers, and educating people about the dangers of drunk driving. Moreover, the cost of DDS may need to be kept at a reasonably low level to attract more users. The use of DDS appears to be attractive to middle-aged and office worker groups, who usually show the riskiest drunk-driving behavior. Furthermore, by considering positive social interactions that lead the choice behavior to an equilibrium, less resources may be required to achieve the same result.

This research successfully clarified the awareness of Hanoi citizens about drunk-driving behavior in addition to estimating the demand for the new DDS. However, some important points still may need to be improved, which suggests some directions for further study. 1) Our research results show that there is only one equilibrium point for social interactions.

However, stronger social interactions may lead to multiple equilibrium points, in which the saddle point between low and high equilibrium can be seen. To find such stronger social interactions, a deeper understanding of individuals’ preferences across their reference groups based on a latent-class model may be a promising solution.

2) In this study, the number of car users was quite small (23 respondents provided 184 SP responses). Therefore, a large-scale study on drunk driving in general and on car drivers

Page 15: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

270

in particular may help to derive more sound conclusions. This factor is now increasingly becoming important because of the rapid increase in car ownership in Vietnam and other developing countries.

3) Drunk driving is a sensitive topic. The SP methodology attempts to mitigate the impacts of respondents’ unfamiliarity and inexperience with new services such as DDS. In reality, the responses may differ from those in the SP experiment. Therefore, the reliability of the SP data should be checked carefully.

4) To understand the effects of social interactions on drunk-driving behavior more clearly, a join-decision or negotiation game between individuals should be developed in a SP context.

REFERENCES Aguirregabiria, V. (2004). Pseudo maximum likelihood estimation of structural models

involving fixed-point problems. Economics Letters, 84(3), 335–340. Apsler, R. (1988). Transportation alternatives for drinkers. In: Surgeon General’s Workshop

on Drunk Driving: Background Papers, 157-168, Office of the Surgeon General, Washington D.C.

Asch, S.E. (1951). Effects of group pressure upon the modification and distortion of judgments. In: Guetzkow, H.S. (ed.), Groups, Leadership and Men: Research in Human Relations, 177–190, Carnegie Press, Oxford, England.

Bray, D.J., Patella, D., Holyoak, N., Tuan, V.A., Tran, M.H. (2016). Motorcycle use and mode choice in Hanoi, Vietnam. A paper presented at the Transportation Research Board 94th Annual Meeting, Washington D.C., January 11-15.

Brock, W.A., Durlauf, S.N. (2001). Discrete choice with social interactions. The Review of Economics Studies, 68(2), 235–260.

Caudill, B.D. (1990). Driving while intoxicated: increased deterrence or alternative transportation for the drunk driver. Journal of Substance Abuse, 2(1), 51–67.

Caudill, B.D., Harding, W.M., Moore, B.A. (2000). At-risk drinkers use safe ride services to avoid drinking and driving. Journal of Substance Abuse, 11(2), 149–159.

Cooper, R. (1999). Coordination Games: Complementarities and Macroeconomics. Cambridge University Press, Cambridge, MA.

Fukuda, D., Morichi, S. (2007). Incorporating aggregate behavior in an individual’s discrete choice: an application to analyzing illegal bicycle parking behavior. Transportation Research Part A: Policy and Practice, 41(4), 313–325.

Government of the Socialist Republic of Vietnam (2002). Solutions for mitigating the increase and then gradually reducing traffic accident and congestion. In: Government of the Socialist Republic of Vietnam (ed.), Resolution No.13/2002/NQ-CP.

Government of the Socialist Republic of Vietnam (2013). Regulation On Sanctioning Administrative Violations In The Road And Railway Traffic, In: Government of the Socialist Republic of Vietnam (ed.), Decree No.171/2013/NĐ-CP.

Granovetter, M. (1978). Threshold models of collective behavior. The American Journal of Sociology, 83(6), 1420–1443.

Harding, W.M., Apsler, R., Goldfein, J. (1988). The Assessment of Ride Service Programs as an Alcohol Countermeasure. Final Technical Report. United States Department of Transportation, Washington D.C.

Harding, W.M., Caudill, B.D., Moore, B.A., Frissell, K.C. (2001). Do drivers drink more when they use a safe ride? Journal of Substance Abuse, 13(3), 283–290.

Page 16: Public Acceptance of Designated Driver Services for

Phan, C.D., et al. / Asian Transport Studies, Volume 5, Issue 2 (2018), 256-271.

271

Jakobsson, C., Fujii, S., Gärling, T. (2000). Determinants of private car users’ acceptance of road pricing. Transport Policy, 7(2), 153–158.

Kirin Holdings Company, Ltd. (2015a). Global Beer Consumption by Country in 2014 (http://www.kirinholdings.co.jp/english/news/2015/1224_01.html; Accessed May 31, 2018).

Kirin Holdings Company, Ltd. (2015b). Global Beer Production by Country in 2014 (http://www.kirinholdings.co.jp/english/news/2015/0810_01.html; Accessed May 31, 2018).

Lang, A.R. (1992). Alcohol, Teenage Drinking. The Encyclopedia of Psychoactive Drugs. Chelsea House Publishers, Langhorne, PA.

Lavoie, M., Godin, G., Valois, P. (1999). Understanding the use of a community-based drive-home service after alcohol consumption among young adults. Journal of Community Health, 24(3), 171–186.

Leibenstein, H. (1950). Bandwagon, snob, and Veblen effects in the theory of consumers’ demand. The Quarterly Journal of Economics, 64(2), 183–207.

Manski, C.F. (1993). Identification of endogenous social effects: the reflection problem. The Review of Economic Studies, 60(3), 531–542.

Molof, M.J., Dresser, J., Ungerleider, S., Kimball, C., Schaefer, J. (1995). Assessment of Year-Round and Holiday Ride Service Programs: Final Report. National Highway Traffic Safety Administration, Washington D.C.

Muromachi, Y., Takeuchi, D., Harata, N., Ohta, K. (2004). Illegal on-street parking problems as social dilemma. 10th World Conference on Transport Research, Istanbul, July 4-8.

Pinheiro, J., Bates, D.M. (1995). Approximations to the log-likelihood function in the nonlinear mixed-effects model. Journal of Computational and Graphical Statistics, 4(1), 12–35.

Rivara, F.P., Relyea-Chew, A., Wang, J., Riley, S., Boisvert, D., Gomez, T. (2007). Drinking behaviors in young adults: the potential role of designated driver and safe ride home programs. Injury Prevention, 13(3), 168–172.

Sarkar, S., Andreas, M., Faria, F.D. (2005). Who uses safe ride programs: an examination of the dynamics of individuals who use a safe ride program instead of driving home while drunk. The American Journal of Drug and Alcohol Abuse, 31(2), 305–325.

Shore, E.R., Sanchez, S. (1993). Characteristics of users of a drunk-driving prevention ride service program. Accident Analysis & Prevention, 25(1), 112–114.

Tuan, V.A. (2015). Development of urban transport demand model for assessing impacts of traffic management strategies in Hanoi City. REMON Final Conference, Hanoi, July 2-3.

Widmark, E.M.P. (1981). Principles and Applications of Medicolegal Alcohol Determination. Biomedical Publications, Davis California.

World Health Organization. (2009). Violence and injury prevention. Road safety in Viet Nam. Website, World Health Organization (http://www.who.int/violence_injury_prevention/road_traffic/countrywork/vietnam/en/; Accessed May 31, 2018).