Search

Effects of income on travel behavior have been widely examined in the literature; however, a majority of existing studies focused mainly on the direct effects of income and neglected the indirect effects. This focus is especially true in the context of developing countries in which income per capita is increasing rapidly. Effects of income on travel behavior may be observed through its impacts on other life choices. Such indirect effects are called “mediation effects” in this study. To fill the foregoing research gap, this study focused on travel mode choice in the Hanoi metropolitan area of Vietnam and developed a path model with multiple discrete choices. In the path model, mediation effects of income on travel mode choice were captured by modeling residential location choice as a discrete mediator that generated nonlinear indirect effects of income on mode choice (the proposed model is called the “mediation model”). As a comparison, a joint model of residential location and travel mode choices was built with only direct effects of income on both types of choice. Based on data collected at three new urban areas in Hanoi in 2015, model estimation results confirm the existence of both direct effects and mediation effects of income on mode choice. Simulation analyses further show that as income increases, the share of motorcycles sharply decreases up to a certain level of income; however, it rises after that level.

Effects of income on travel behavior have been widely examined in the literature; however, a majority of existing studies focused mainly on the direct effects of income and neglected the indirect effects. This focus is especially true in the context of developing countries in which income per capita is increasing rapidly. Effects of income on travel behavior may be observed through its impacts on other life choices. Such indirect effects are called “mediation effects” in this study. To fill the foregoing research gap, this study focused on travel mode choice in the Hanoi metropolitan area of Vietnam and developed a path model with multiple discrete choices. In the path model, mediation effects of income on travel mode choice were captured by modeling residential location choice as a discrete mediator that generated nonlinear indirect effects of income on mode choice (the proposed model is called the “mediation model”). As a comparison, a joint model of residential location and travel mode choices was built with only direct effects of income on both types of choice. Based on data collected at three new urban areas in Hanoi in 2015, model estimation results confirm the existence of both direct effects and mediation effects of income on mode choice. Simulation analyses further show that as income increases, the share of motorcycles sharply decreases up to a certain level of income; however, it rises after that level.

Finding ways to encourage people to participate in domestic tourism has been a central political issue in Japan for many years. One issue is the unbalanced regional trend in domestic tourist destinations. Various transport policies have been proposed. Under such policy considerations, jointly representing tourism participation, destination choice, and travel mode choice becomes important. However, tourism demand shows monthly variations. To date, the above three aspects have not been satisfactorily analyzed in a dynamic fashion. This study seeks to gain a better understanding of interrelated tourist behavior and provide a scientific tool to support tourism policy decisions by jointly analyzing these choice aspects to build a dynamic nested logit model while taking the influence of state dependence into account. Empirical analysis confirmed the effectiveness of the model with the use of retrospective panel survey data collected from 1,253 respondents in Japan in 2010. To the authors' best knowledge, this was the first panel survey in Japan to investigate tourists' behavior over the course of 1 year on a monthly basis. Model estimation results confirmed significantly negative effects of state dependence with respect to tourism participation behavior (implying a decreasing tendency of continuous tourism participation at the monthly level), but positive effects with respect to destination and travel mode choice behaviors. The positive value of state dependence for the travel mode choice behavior suggests that people prefer to use the same type of travel modes when they travel. The influence of travel mode choice on destination choice shows largely regional variations. Results also clarified effects of tourism motivation, individual characteristics, destination-specific attributes, and travel mode-specific attributes on the three choice aspects. Implications of model estimations for policy decisions are discussed.

Finding ways to encourage people to participate in domestic tourism has been a central political issue in Japan for many years. One issue is the unbalanced regional trend in domestic tourist destinations. Various transport policies have been proposed. Under such policy considerations, jointly representing tourism participation, destination choice, and travel mode choice becomes important. However, tourism demand shows monthly variations. To date, the above three aspects have not been satisfactorily analyzed in a dynamic fashion. This study seeks to gain a better understanding of interrelated tourist behavior and provide a scientific tool to support tourism policy decisions by jointly analyzing these choice aspects to build a dynamic nested logit model while taking the influence of state dependence into account. Empirical analysis confirmed the effectiveness of the model with the use of retrospective panel survey data collected from 1,253 respondents in Japan in 2010. To the authors’ best knowledge, this was the first panel survey in Japan to investigate tourists’ behavior over the course of 1 year on a monthly basis. Model estimation results confirmed significantly negative effects of state dependence with respect to tourism participation behavior (implying a decreasing tendency of continuous tourism participation at the monthly level), but positive effects with respect to destination and travel mode choice behaviors. The positive value of state dependence for the travel mode choice behavior suggests that people prefer to use the same type of travel modes when they travel. The influence of travel mode choice on destination choice shows largely regional variations. Results also clarified effects of tourism motivation, individual characteristics, destination-specific attributes, and travel mode–specific attributes on the three choice aspects. Implications of model estimations for policy decisions are discussed.

A dynamic discrete choice model for multiple social interactions was developed on the basis of retrospective panel data in the context of household vehicle-type choice. "Social interactions" refer to the influence of reference groups on household choice behavior. However, most studies in transportation have considered a household as a decision maker independent from the society to which it belongs. Social interactions have therefore not been well represented, especially if there were two or more types of reference groups, which may further vary over time. The model was built within the dynamic generalized extreme value framework, which included a set of dynamic elements, such as initial conditions, state dependence, and future expectation. This study defined three types of social interactions: diffusion rates of a vehicle type at the national level and at the neighborhood level and diffusion rate of households with the same income level. A survey was conducted in local Japanese cities in 2006. The survey focused on household vehicle ownership behavior over the 10 years from 1997 to 2006. Model estimation results confirmed that social interactions from reference groups of neighborhoods and households with the same income level were especially influential in decisions about household vehicle-type choice. On the contrary, social interaction from the whole society did not significantly influence the household choice. It was further found that effects of dynamic elements on the household choice were statistically significant.

To improve methods of evaluating the benefits from investments in the transport infrastructure, this paper attempts to develop a method for capturing changes in variations in travel time expenditure and shows empirical results from data from the German Mobility Panel, a multiday and multiperiod panel survey. The proposed methodology can simultaneously and explicitly deal with changes and variations. The empirical analysis indicates that over time, travel time expenditures become more dependent on situational attributes than on individual or household attributes; such dependence implies that travel time expenditures change toward diversification and that longer periods of behavioral observation to describe people's travel time expenditures become more important. The results also imply that understanding of the dynamic aspects of behavior is still important and remains a challenging issue to be revisited, not only for improving forecasting models but also for reconsidering the benefits from investment in transportation infrastructure.

To improve methods of evaluating the benefits from investments in the transport infrastructure, this paper attempts to develop a method for capturing changes in variations in travel time expenditure and shows empirical results from data from the German Mobility Panel, a multiday and multiperiod panel survey. The proposed methodology can simultaneously and explicitly deal with changes and variations. The empirical analysis indicates that over time, travel time expenditures become more dependent on situational attributes than on individual or household attributes; such dependence implies that travel time expenditures change toward diversification and that longer periods of behavioral observation to describe people's travel time expenditures become more important. The results also imply that understanding of the dynamic aspects of behavior is still important and remains a challenging issue to be revisited, not only for improving forecasting models but also for reconsidering the benefits from investment in transportation infrastructure.

A dynamic discrete choice model for multiple social interactions was developed on the basis of retrospective panel data in the context of household vehicle-type choice. “Social interactions” refer to the influence of reference groups on household choice behavior. However, most studies in transportation have considered a household as a decision maker independent from the society to which it belongs. Social interactions have therefore not been well represented, especially if there were two or more types of reference groups, which may further vary over time. The model was built within the dynamic generalized extreme value framework, which included a set of dynamic elements, such as initial conditions, state dependence, and future expectation. This study defined three types of social interactions: diffusion rates of a vehicle type at the national level and at the neighborhood level and diffusion rate of households with the same income level. A survey was conducted in local Japanese cities in 2006. The survey focused on household vehicle ownership behavior over the 10 years from 1997 to 2006. Model estimation results confirmed that social interactions from reference groups of neighborhoods and households with the same income level were especially influential in decisions about household vehicle-type choice. On the contrary, social interaction from the whole society did not significantly influence the household choice. It was further found that effects of dynamic elements on the household choice were statistically significant.

To realize environmentally sustainable transport (EST), it becomes more and more important to reduce the environmental load from the transport sector while maintaining the level of mobility. Because mobility policies often conflict with environmental ones, policy decision makers need to find a way to solve the exclusiveness between these two policies. This study attempts to apply the concept of ecoefficiency, originally proposed by the World Business Council for Sustainable Development, to tackle the dilemma between urban mobility and environmental load in cities in both developed and developing countries. This study proposes a new model, the environmental efficiency (EE) model, which expands the concept of the data envelopment analysis cost-efficiency model. This new model aims to measure the efficiency of energy consumption at a given level of mobility in each transport system by incorporating some feasible conditions. In addition, the model can find multiple sets of frontier cities that are the most efficient among homogeneously developed cities. Consequently, the proposed method contributes to providing each city with a feasible transport energy consumption goal. Moreover, a panel analysis was carried out to examine temporal changes in environmental efficiency. Finally, the effects of some EST policies implemented to raise efficiency are simulated on the basis of the EE model, including modal shifts in a city, technological innovations to improve the intensity of energy consumption in cities, and the emissions trading scheme between cities in developing and developed countries. The method is confirmed to be a useful tool for the establishment of more plausible targets for transport energy-saving policies.

This study attempts to examine the variation properties of time use behavior on the basis of a multilevel multiple discrete-continuous extreme value model, which describes both activity participation and time allocation behavior by incorporating various variance components. Five major variation components are dealt with, including interindividual, interhousehold, temporal, spatial, and intraindividual variations. The MobiDrive data, continuous 6-week daily travel data, allow identification of these variations simultaneously. Two types of models are empirically examined: the model without consideration of the influence of explanatory variables (null model) and the model that introduces explanatory variables (full model). Estimation results of the null model confirm that intraindividual variation still accounts for more than 50% of the total variation (except for mandatory activities) even after incorporation of the aforementioned four other types of variations jointly. In contrast, the results from the full model reveal that most types of unobserved variations (especially the intraindividual variation) still dominate in the total variation even after the relevant observed information is introduced. These findings provide useful insights into both model development and data collection methods as well as to the understanding of the mechanisms of time use decisions.

This study attempts to examine spatiotemporal changes of traffic demand elasticities regarding gasoline prices and economic conditions and focusing on the substantial fluctuations that occurred throughout 2008 in Japan. This research is aimed at a better understanding of the spatial and temporal instability of the elasticities and possible reasons for changes in the elasticities. First, a random coefficient model is built to represent spatial heterogeneity in the elasticities and is estimated by using a hierarchical Bayesian method. Next, a sequential Bayesian updating method is applied to examine monthly changes in the elasticities. The data used in the empirical analysis were monthly traffic volume data collected on 53 expressways. The results showed that although the monthly changes in the average elasticities over all routes were mainly observed before August 2008, different directions of change across routes were observed after September 2008, when gasoline prices began to fall. The results also indicated that responses to gasoline price changes might be dependent on the causes of price changes. Furthermore, on urban expressways, it was found that once a reduction in traffic demand was attained because of rising gasoline prices, the demand did not fully recover even after the actual prices fell again to the original level.

This study attempts to examine spatiotemporal changes of traffic demand elasticities regarding gasoline prices and economic conditions and focusing on the substantial fluctuations that occurred throughout 2008 in Japan. This research is aimed at a better understanding of the spatial and temporal instability of the elasticities and possible reasons for changes in the elasticities. First, a random coefficient model is built to represent spatial heterogeneity in the elasticities and is estimated by using a hierarchical Bayesian method. Next, a sequential Bayesian updating method is applied to examine monthly changes in the elasticities. The data used in the empirical analysis were monthly traffic volume data collected on 53 expressways. The results showed that although the monthly changes in the average elasticities over all routes were mainly observed before August 2008, different directions of change across routes were observed after September 2008, when gasoline prices began to fall. The results also indicated that responses to gasoline price changes might be dependent on the causes of price changes. Furthermore, on urban expressways, it was found that once a reduction in traffic demand was attained because of rising gasoline prices, the demand did not fully recover even after the actual prices fell again to the original level.

This study attempts to examine the variation properties of time use behavior on the basis of a multilevel multiple discrete–continuous extreme value model, which describes both activity participation and time allocation behavior by incorporating various variance components. Five major variation components are dealt with, including interindividual, interhousehold, temporal, spatial, and intraindividual variations. The MobiDrive data, continuous 6-week daily travel data, allow identification of these variations simultaneously. Two types of models are empirically examined: the model without consideration of the influence of explanatory variables (null model) and the model that introduces explanatory variables (full model). Estimation results of the null model confirm that intraindividual variation still accounts for more than 50% of the total variation (except for mandatory activities) even after incorporation of the aforementioned four other types of variations jointly. In contrast, the results from the full model reveal that most types of unobserved variations (especially the intraindividual variation) still dominate in the total variation even after the relevant observed information is introduced. These findings provide useful insights into both model development and data collection methods as well as to the understanding of the mechanisms of time use decisions.

To realize environmentally sustainable transport (EST), it becomes more and more important to reduce the environmental load from the transport sector while maintaining the level of mobility. Because mobility policies often conflict with environmental ones, policy decision makers need to find a way to solve the exclusiveness between these two policies. This study attempts to apply the concept of ecoefficiency, originally proposed by the World Business Council for Sustainable Development, to tackle the dilemma between urban mobility and environmental load in cities in both developed and developing countries. This study proposes a new model, the environmental efficiency (EE) model, which expands the concept of the data envelopment analysis cost-efficiency model. This new model aims to measure the efficiency of energy consumption at a given level of mobility in each transport system by incorporating some feasible conditions. In addition, the model can find multiple sets of frontier cities that are the most efficient among homogeneously developed cities. Consequently, the proposed method contributes to providing each city with a feasible transport energy consumption goal. Moreover, a panel analysis was carried out to examine temporal changes in environmental efficiency. Finally, the effects of some EST policies implemented to raise efficiency are simulated on the basis of the EE model, including modal shifts in a city, technological innovations to improve the intensity of energy consumption in cities, and the emissions trading scheme between cities in developing and developed countries. The method is confirmed to be a useful tool for the establishment of more plausible targets for transport energy-saving policies.

A study applied a web-based stated preference (SP) survey to investigate household preferences for transit-oriented residential choices, incorporating intrahousehold interaction. The SP survey included five alternatives: one current residential location and four combinations of two residential areas and two commuting modes (a car and a transit system). The assumed residential areas were all located near stations of a transit system: one was close to and another was far from the city center. Commuting modes were only for household heads. Households with only couples were recruited, and the survey was conducted with the help of an Internet survey company. In the SP survey, the couple was first asked to answer the SP questions separately, and then the two members jointly answered the questions. It was found that members in 40% of the households changed their stated choices after joint decisions. This switch suggests the necessity to incorporate intrahousehold interaction into any household residential choice model. To this end, a group discrete choice model with multilinear utility was established; members' relative influences and intrahousehold interaction were simultaneously represented in the model. Model estimation results first confirmed the effectiveness of the established model. It was then revealed that the existence of intrahousehold interaction reduced total household utility and that husbands wielded relatively greater influence than wives. It was also shown that with increases in age, members' relative influence grew, and that dominant decision makers concerning household asset management had greater influence power on household residential choice behavior.

The variation properties of departure time choice behavior are examined by activity type with the use of a continuous 6-week travel survey collected in the cities of Karlsruhe and Halle, Germany, in 1999. The total variation in departure time choice is divided into five variation components: spatial variation, temporal variation at the aggregate level, interhousehold variation, interindividual variation, and intraindividual variation. These variations are first quantitatively analyzed by using a multilevel modeling approach without considering the influences of explanatory variables. Then, on the basis of the clarified variations patterns, how much of the variations can be captured by explanatory variables is examined. It is confirmed that different activities show quite different variations, and of the variations the intraindividual variation accounts for the largest percentage of the total variation. Model estimation results further underscore the need for simultaneously dealing with unobserved macrolevel variations, especially interindividual, interhousehold, and spatial variations, as well as the intraindividual variation, even when the explanatory variables are included in the model. Such quantitative assessment of various types of variations based on a multilevel modeling approach provides a useful guide for the specification of behavior models and deepens the knowledge of behavioral variability by going beyond the traditional paradigm, which focuses on a limited set of variations.

The variation properties of departure time choice behavior are examined by activity type with the use of a continuous 6-week travel survey collected in the cities of Karlsruhe and Halle, Germany, in 1999. The total variation in departure time choice is divided into five variation components: spatial variation, temporal variation at the aggregate level, interhousehold variation, interindividual variation, and intraindividual variation. These variations are first quantitatively analyzed by using a multilevel modeling approach without considering the influences of explanatory variables. Then, on the basis of the clarified variations patterns, how much of the variations can be captured by explanatory variables is examined. It is confirmed that different activities show quite different variations, and of the variations the intraindividual variation accounts for the largest percentage of the total variation. Model estimation results further underscore the need for simultaneously dealing with unobserved macrolevel variations, especially interindividual, interhousehold, and spatial variations, as well as the intraindividual variation, even when the explanatory variables are included in the model. Such quantitative assessment of various types of variations based on a multilevel modeling approach provides a useful guide for the specification of behavior models and deepens the knowledge of behavioral variability by going beyond the traditional paradigm, which focuses on a limited set of variations.

A study applied a web-based stated preference (SP) survey to investigate household preferences for transit-oriented residential choices, incorporating intrahousehold interaction. The SP survey included five alternatives: one current residential location and four combinations of two residential areas and two commuting modes (a car and a transit system). The assumed residential areas were all located near stations of a transit system: one was close to and another was far from the city center. Commuting modes were only for household heads. Households with only couples were recruited, and the survey was conducted with the help of an Internet survey company. In the SP survey, the couple was first asked to answer the SP questions separately, and then the two members jointly answered the questions. It was found that members in 40% of the households changed their stated choices after joint decisions. This switch suggests the necessity to incorporate intrahousehold interaction into any household residential choice model. To this end, a group discrete choice model with multilinear utility was established; members’ relative influences and intrahousehold interaction were simultaneously represented in the model. Model estimation results first confirmed the effectiveness of the established model. It was then revealed that the existence of intra-household interaction reduced total household utility and that husbands wielded relatively greater influence than wives. It was also shown that with increases in age, members’ relative influence grew, and that dominant decision makers concerning household asset management had greater influence power on household residential choice behavior.

Traffic accidents at signalized intersections are influenced by many environmental factors, including site layout, traffic volume, and signal operation. However, most traffic accidents are attributable to driver behavior in response to the environment, because responses to the environment vary with drivers' individual characteristics, as observed in the concept of a dilemma zone (DZ) at intersection approaches. For DZ protection, the installation of traffic signals closer to the stop line (i.e., before the intersection) has been proposed in South Korea. Although such proposals have been implemented experimentally, the actual effect of this measure on driver behavior has not yet been investigated. A discrete choice model is used to identify the factors that influence drivers' crossing and stopping behaviors at signalized intersections, under the assumption that traffic safety is at risk if the influence of the signal location is ignored. Two analyses are presented to assess the impact of signal location on driver behavior at intersections. Violations of stop line crossing behavior and remaining time to stop line are examined, and two types of driver behavior models are presented that address the effects of signal location and other factors. The results consistently reveal that traffic signal location affects drivers' stopping behavior, suggesting that traffic signal location can jeopardize road safety and therefore should be considered in intersection design.

Traffic accidents at signalized intersections are influenced by many environmental factors, including site layout, traffic volume, and signal operation. However, most traffic accidents are attributable to driver behavior in response to the environment, because responses to the environment vary with drivers’ individual characteristics, as observed in the concept of a dilemma zone (DZ) at intersection approaches. For DZ protection, the installation of traffic signals closer to the stop line (i.e., before the intersection) has been proposed in South Korea. Although such proposals have been implemented experimentally, the actual effect of this measure on driver behavior has not yet been investigated. A discrete choice model is used to identify the factors that influence drivers’ crossing and stopping behaviors at signalized intersections, under the assumption that traffic safety is at risk if the influence of the signal location is ignored. Two analyses are presented to assess the impact of signal location on driver behavior at intersections. Violations of stop line crossing behavior and remaining time to stop line are examined, and two types of driver behavior models are presented that address the effects of signal location and other factors. The results consistently reveal that traffic signal location affects drivers’ stopping behavior, suggesting that traffic signal location can jeopardize road safety and therefore should be considered in intersection design.

Focusing on car tourists' 1-day tours, a new scheduling model combines a nested paired combinatorial logit (NPCL) type of destination and route choice model and a time allocation (TA) model. The NPCL model, developed previously from the generalized extreme value family of discrete choice models to represent the similarity between pairs of alternatives in the same choice nest as well as the influence of inclusive value, indicates destination choice in the bottom level and route choice in the top level. The TA model applies Becker's theory to determine the time allocated to each touring site. Utility of destination choice is influenced by the time spent at each site. Different route choices result in a level of service for the road network that varies hourly, varying available time used in the TA model. The TA model endogenously incorporates the influence of hourly variance in level of service at the site of interest, which is affected by the allocated time. An iteration estimation procedure is proposed to estimate the parameters consistently in both models. Finally, revealed preference tourist travel survey data collected in a tourist attraction region near the Sea of Japan indicate that the proposed scheduling model is effective in representing car tourists' scheduling behavior for 1-day tours.

Focusing on car tourists’ 1-day tours, a new scheduling model combines a nested paired combinatorial logit (NPCL) type of destination and route choice model and a time allocation (TA) model. The NPCL model, developed previously from the generalized extreme value family of discrete choice models to represent the similarity between pairs of alternatives in the same choice nest as well as the influence of inclusive value, indicates destination choice in the bottom level and route choice in the top level. The TA model applies Becker's theory to determine the time allocated to each touring site. Utility of destination choice is influenced by the time spent at each site. Different route choices result in a level of service for the road network that varies hourly, varying available time used in the TA model. The TA model endogenously incorporates the influence of hourly variance in level of service at the site of interest, which is affected by the allocated time. An iteration estimation procedure is proposed to estimate the parameters consistently in both models. Finally, revealed preference tourist travel survey data collected in a tourist attraction region near the Sea of Japan indicate that the proposed scheduling model is effective in representing car tourists’ scheduling behavior for 1-day tours.