Transportation Systems Planning Methods and Applications 15

Transportation Systems Planning Methods and Applications 15 Transportation engineering and transportation planning are two sides of the same coin aiming at the design of an efficient infrastructure and service to meet the growing needs for accessibility and mobility. Many well-designed transport systems that meet these needs are based on a solid understanding of human behavior. Since transportation systems are the backbone connecting the vital parts of a city, in-depth understanding of human nature is essential to the planning, design, and operational analysis of transportation systems. With contributions by transportation experts from around the world, Transportation Systems Planning: Methods and Applications compiles engineering data and methods for solving problems in the planning, design, construction, and operation of various transportation modes into one source. It is the first methodological transportation planning reference that illustrates analytical simulation methods that depict human behavior in a realistic way, and many of its chapters emphasize newly developed and previously unpublished simulation methods. The handbook demonstrates how urban and regional planning, geography, demography, economics, sociology, ecology, psychology, business, operations management, and engineering come together to help us plan for better futures that are human-centered.

Trang 1

15 Assessing the Effects of

Constrained and Unconstrained Policy

Scenarios on Activity–Travel Patterns Using a Learning-Based

CONTENTS

15.1 Introduction 15.2 Albatross 15.3 Data 15.4 Model Performance 15.5 The Scenario Agent 15.6 Scenario 1: Decrease in Two-Adult Households

Scenario Definition • Results • Interpretation of Results

15.7 Scenario 2: Change of Work Start Times

15.8 Scenario 3: Increase in Part-Time Workers

15.9 Scenario 4: Friday Afternoon Off

15.10 Conclusions References

1 Paper presented at the WCTR Conference, Seoul 2001.

Theo Arentze

Eindhoven University of Technology

Frank Hofman

Ministry of Transportation, Public

Works and Water Management

Henk van Mourik

Ministry of Transportation, Public

Works and Water Management

Harry Timmermans

Eindhoven University of Technology

Trang 2

15.1 Introduction

Over the last decade, it has been realized that the reduction of car use and mobility requires a multitude

of policy instruments Capacity-oriented measures, transport demand management, land use planning initiatives, and institutional approaches in principle all influence people’s activity and hence travel patterns Unfortunately, such policies are not always formulated and implemented with this goal in mind Moreover, some policies may turn out to be counterproductive because not all factors and implications were taken into consideration For example, the policy of stimulating to leave the car at home for the work trip has often ended up increasing travel, as other members of the household had an opportunity

to use the car for other activities Likewise, the stimulation of teleworking may have adverse affects, as time is now used for becoming involved in social and recreational activities (Harvey et al., 2000) To reduce the risk of stimulating the wrong set of policies, a rather advanced tool that incorporates the mechanisms described above may be helpful

It was against this background that the Dutch Ministry of Transportation, Public Works and Water Management commissioned European Insitute of Retailing and Services Studies (EIRASS)/Urban Plan-ning Group to develop a prototype of a rule-based model of activity behavior The system developed for this purpose received the acronym Albatross The system, which has been described in detail elsewhere (Arentze et al., 2000), predicts which activities are conducted where, when, for how long, with whom, and the travel involved Household interaction and scheduling are modeled The predictive ability of the system for a sample of activity diaries, plus a validation set, turned out to be relatively good

This chapter reports the results of an application of the system to a set of scenarios These simulations were conducted primarily to assess the face validity of the system, but may also have some substantive significance Four scenarios were developed: (1) decrease in two-adult households, (2) change of work start times, (3) increase in part-time workers, and (4) Friday afternoon off Note that some of these policies are unconstrained and relate to demographic change, while other scenarios constrain activity scheduling options All these scenarios are related to ongoing trends in The Netherlands

These policies affect the conditions and properties of activity programs The policies were translated into the relevant attributes and conditions of the Albatross system The model was then used to simulate the impact

of the policy scenarios on activity–travel patterns The impact of the scenarios was assessed in terms of such performance indicators as total travel distance, travel distance by car, number of tours, trip–tour ratio, car driver–total distance ratio, car passenger–total distance ratio, and public transport–total distance ratio This chapter will report the main results of the effects of these policy scenarios on activity–travel patterns It will first briefly summarize Albatross Next, the data collection process will be summarized This will be followed by a description of the various policy scenarios, and the outcomes of the application

of the model system to each of these scenarios Interpretations will be given The chapter will be completed with a summary of the main findings

15.2 Albatross

Albatross is the latest, most comprehensive, and only operational computational process model of transportation demand (Arentze et al., 2000) It can be considered a multiagent rule-based system that predicts activity patterns Essential to the system is that choice heuristics are used to simulate behavior The system consists of a series of agents that together handle the (consistency of the) data, the derivation

of choice heuristics from activity diary data, the simulation or prediction of activity patterns, the assess-ment and reporting of model performance, the calculation of various system performance indicators, and the evaluation of alternative model scenarios

A series of rules have been derived from empirical activity diary data to identify the mechanisms underlying the organization of activities in time and space It is assumed that individuals and households will try to meet particular criteria, subject to personal, household, temporal, spatial, institutional, and space–time constraints The simulation process involves the application of the choice rules, subject to this wide range of constraints

Trang 3

Central to the system is the scheduling agent It controls the scheduling processes in terms of a sequence

of steps In each step, this agent identifies the condition information required for making principal scheduling decisions, sends appropriate calls to agents for the required analyses, passes the obtained information to the rule-based system, and translates returned decisions into appropriate operations on the current schedule

The simulation starts with an activity program This defines the set of activities that need to be completed on any given day This program consists of mandatory activities (e.g., work, school) and discretionary activities The mandatory activities constitute the skeleton of the activity program Scheduling then involves selecting the discretionary activities, adding these to the skeleton, and determining the schedule position and profile of each added discretionary activity A series of decision tables, used to represent the choice heuristics, are consulted to simulate this process in a series of steps

The first step simulates the selection of discretionary activities, travel party, and the duration of the discretionary activities The selection decision is modeled as a yes or no decision for each of a predefined set of optional activities If a particular activity is selected, the model system simulates whether another instance of the same activity should be added to the activity program If this is not the case, the selection of possible other activities is further modeled The order in which activities are evaluated is predefined based on assumptions about activity priorities Temporal, household, and institutional constraints (facility opening times) are systematically considered to assess the feasibility

of selection and duration decisions

The next two steps involve the identification of the scheduling position of each selected activity The set of possible positions is first reduced to positions that comply with a specific episode of the day (e.g., early morning, around noon, etc.) by selecting a start time interval Next, trip chaining decisions simulate whether the activity is to be linked to a previous or future activity The outcome of this step simulates whether the conduct of that activity involves a single-stop tour from home or a more complex pattern The combination of start time and trip chaining decisions often determines the schedule position If more options still prevail, the system selects the position that leaves maximum freedom of choice for subsequent scheduling steps Temporal constraints, available locations, and normative travel times deter-mine the feasibility of choice options in this step

The next steps in the scheduling process involve the choice of transport mode and location Mode decisions are made at the level of tours Possible interactions between mode and location choices are explicitly taken into account by using location information as conditions of mode selection rules The location of fixed activities is assumed given For nonfixed activities, the system dynamically defines the location choice set while accounting for available locations, mode-specific travel times, opening times of available facilities, and the time window and minimum duration for the activity The temporal constraints follow from the earlier duration, schedule position, start time, trip chain-ing, and mode decisions A location decision involves the choice of a heuristic that uniquely identifies

a location from the choice set in terms of a noninferior combination of attractiveness and travel time The option “other” is included to cover alternative heuristics If this option is chosen, the system selects a location at random from a drive time band, if the choice set includes more than one location in that band

15.3 Data

The simulations reported in this chapter are based on an activity diary survey that was administered

in two municipalities in the Rotterdam region in The Netherlands in 1997 These diaries were also used to derive the rules underlying the Albatross model system The diary involved two consecutive days Days were designated for households such that the sample was balanced across the days of the week Respondents were invited to report the sequence of activities that they conducted during these days, and to detail the start and end times, the location where the activity took place, the transport mode (chain), the travel time per mode, and the travel party A precoded scheme was used for activity

Trang 4

reporting Several modes of administration were used Response rates ranged between 64 and 82%, dependent on mode of administration for those households that indicated during a prescreening that they would be willing to participate in the survey This resulted in a total of 2198 household days that were used for the analysis The diaries were cleaned using the special-purpose program Sylvia (Arentze

et al., 1999)

15.4 Model Performance

The activity diary data described above were used to derive a decision table for each decision step in the scheduling model To allow validity testing, the data set was split into two parts The first part includes 75% of the cases and was used to induce the decision tables, whereas the other cases were used to validate the derived decision rules A CHAID-based tree induction algorithm was used to derive the decision rules

It goes beyond the scope of the present chapter to discuss the induced decision tables and model performance in any detail Readers are referred to Arentze and Timmermans (2000) for such details Suffice to say that, although performance varied by step in the scheduling model, the overall performance was good compared to other model types (Arentze et al., 2000)

15.5 The Scenario Agent

One of the agents in the Albatross system is the scenario builder agent, called Agnes This agent allows users to change the attributes of the land use and transportation environment, household character-istics, and the schedule skeleton In addition, users can change the composition of the sample or population The agent can be activated in three modes: population composition, behavioral change, and impact assessment

In the population composition mode, population scenarios can be implemented by selecting one or

more segments and specifying a multiplication factor The factor determines the number of times or, if the factor is smaller than 1, the probability that each case of that segment will enter the prediction cycle

In the behavioral change mode, users can simulate the effects of policy measures that are not

incor-porated as explanatory variables in the model In that case, policy effects can only be predicted if the user can change directly some or all of the behavioral exogenous variables For example, the current system does not have an adjustable road pricing variable Yet, one may assume that variable road pricing may induce some people to leave home early Users can modify the skeleton of the input schedules In the current version, change to the structure–response pattern is not possible, unless one changes the decision rule base outside the system

Scenarios about such behavioral change can be built as follows First, a three-dimensional cross-tabulation can be created to select the target segment This three-dimensional functionality allows users

to vary either anticipated behavioral change by zone and two sociodemographic variables simultaneously

or three sociodemographic variables A drop-down list allows the user to select any combination of activity program facets considered relevant Next, users are requested to indicate (1) the degree of adaptation penetration (i.e., the percentage of the selected segment or each category that will adapt), and (2) the size and direction of behavioral change The degree of adaptation penetration may be set for all cases or a target segment

In case of continuous variables, users can specify the size of changes in terms of an exact value, a minimum, a maximum, a proportion, a constant change, or a range For nominal variables, the user should specify percentage change for all categories minus 1, by specifying proportions in a matrix of given and new categories of the dimension Furthermore, users can define a standard deviation dependent

on the assumed heterogeneity under change A truncated normal distribution is assumed to avoid predicted change in the wrong direction Based on these specifications, Albatross uses Monte Carlo simulation to identify the specific respondents that will experience the change of interest

Trang 5

In the impact assessment mode, users can change the exogenous variables of the system The same

principles apply It should be noted that in addition to these individual scenarios, accumulative scenarios can be built

In the following sections, we will report the development of the scenarios and the changes in activ-ity–travel patterns that the system predicts should these scenarios be implemented The formulation of these scenarios used the scenario-building agent

15.6 Scenario 1: Decrease in Two-Adult Households

15.6.1 Scenario Definition

The reduction of household size has been a constant trend over the last decades in The Netherlands as well as other postindustrial countries Demographic forecasts suggest that this trend will continue to be influential in the foreseeable future To explore the impact of such demographic change on activity–travel patterns, this first scenario assumes that the number of two-adult households will be reduced by 10% and that a proportional increase in the number of single-adult households will occur

This scenario was implemented by decreasing the number of two-adult-households and increasing the number of single-adult households in the sample This was done by setting the multiplication factor in the scenario builder to 0.90 for the two-adult cases and 1.58 for the single-adult cases Due to these settings, each two-adult case has a chance of 0.90 of being selected in the new sample and each single-adult case has

a chance of 1 of being selected once and a chance of 0.58 of being selected twice in the new sample The multiplier for the single-adult cases was determined such that the number of person days keeps constant Thus, the scenario has no impact on population size, but only on the composition of the sample

15.6.2 Results

The number of prediction runs is set to N = 20 for both the scenario and the null scenario As it appears, this number results in sufficient power to be able to measure all effects of interest Table 15.1 represents

a selection of the most relevant indicators

In the after case, the total distance traveled equals 83,054 km This means a decrease in 1255 km (–1.5%) compared to the before case The distance traveled by car is estimated by a weighted summation

of kilometers traveled as car passenger and kilometers traveled as car driver, to prevent double counting

of kilometers that are traveled by partners on the same trip (as car passenger and car driver) That is, a relative weight of 0.5 is used for car passenger, assuming that in 50% of the car passenger cases the same trip is already represented in the partner’s schedule Measured this way, the total distance traveled by car decreases approximately proportionally with the decrease in total travel distance (–1.3%) The shares of other modes in total kilometers traveled remain the same The decrease in total travel distance is not caused by a decrease in the number of tours (the decrease is not significant) This implies that the effect

is caused by a decrease in average distance traveled per tour The increase in ratio between trips and tours indicates that activities are more often combined on the same tour

Several other observations are relevant First, although the total number of out-of-home activities has stayed constant, the distribution across activities has changed An increase in frequencies of service activities (+3.8%) and “other” activities (+2.8%) compensates for a decrease in the frequency of bring and get activities (–4.0%).2 The distribution of out-of-home activities across days of the week has stayed the same except for Tuesday On Tuesday, the frequencies of nondaily shopping, social, and leisure activities have increased The distribution across times of the day has undergone a change too The share

of activities starting before 10 A.M has decreased, and the proportion of activities initiated during 4 to

6 P.M has increased The relative frequency of single-stop trips is smaller in the after case This is consistent

2 The “other” category includes medical visits, personal business and activities labeled as “other” in the question-naire

Trang 6

with the earlier observation that the trip–tour ratio has increased The choice of transport mode shows

an increase in car passenger share in out-of-home activities

15.6.3 Interpretation of Results

The simulation results suggest that this scenario leads to a reduction of total travel distance caused by a decrease in the average distance traveled per tour There are at least three possible explanations for this decrease First, the choice of more local destinations may be related to the change in distribution across activity types, which is occurring simultaneously However, given the nature of the redistribution — fewer bring and get activities and more service and “other” activities — this is not very likely to be the case Second, the preference for nearest-location choice heuristics has increased or location choice sets have been narrowed down This explanation is not very plausible either, considering the location selection decision tables (DTs) Third, individuals from single-adult households tend to live closer to the place where they work, because in choosing the residence location they do not have to take the workplace of a partner into account Thus, according to this last explanation, the average tour length has decreased due to shorter commuting trips None of the above explanations take into account that other population attributes have changed simul-taneously with the manipulation of household size In the after case, older-age households, no-child house-holds, and females are more strongly represented in the sample The decrease in average trip length may

be related to differences in behavior (i.e., lifestyles) of these groups Such differences may be reflected in input schedule skeletons as well as the DTs With respect to the latter, the age index, child index, and gender variable indeed recur in many decision rules Given the fact that other sample characteristics co-vary with the manipulation, a critical question is whether the whole package of changes is realistic in the context of the scenario If the scenario assumes that age index, child index, and gender co-vary with the decrease in household size as they do in the present sample, then the outcomes are still interpretable as consequences

of the scenario Nevertheless, the present method does not allow users to control the correlation structure

15.7 Scenario 2: Change of Work Start Times

To simulate the effects of changing work start times, the second scenario assumes that 15% of the trips from home to work that currently take place during morning rush hour start earlier or later to avoid rush hour This behavioral change is implemented as a change in the schedule skeleton using the scenario builder Trips are considered to take place during morning rush hour if the work start time falls in the time interval of 7:00

to 9:30 A.M In a random selection of 15% of these work activities, the work start time is set to 7:00 or 9.30

A.M., whichever manipulation implies the smallest change End times are changed simultaneously in such a way that the duration of the work activity stays the same The scenario further assumes that every sleep activity and every other work activity (e.g., the afternoon work episode) simultaneously shift on the timescale such that the intervals between the activities remain constant A change in start and end times of an activity

TABLE 15.1 Comparison of Means between Scenario 1 and the Null Scenario on a Selection of Indicators

Indicator Mean s ∆m % ∆m t-value df Total travel distance 83054 1316 –1255 a –1.5107 –3.187 41 Car distance (driver + passenger) 60176 1083 –808 a –1.3434 –2.567 39 Number of tours 5306 75 –39 –0.7398 –1.990 33 Trips–tours ratio 2.420 0.007 0.0077 a 0.3199 3.301 42 Car driver–total distance ratio 0.677 0.011 0.0032 0.4797 0.968 41 Car passenger–total distance ratio 0.095 0.007 –0.0041 –4.3711 –1.870 43 Public–total distance ratio 0.128 0.012 0.0011 0.8182 0.320 37 Slow–total distance ratio 0.100 0.004 –0.0002 –0.1678 –0.135 39

a Significant at the α = 0.05 level in a two-sided test.

Trang 7

may give rise to conflicts with the timing of other activities in the schedule skeleton Conflicts were consistently solved by means of moving activities on the timescale while keeping the duration of activities constant

15.7.2 Results

The prediction of schedules under scenario conditions was repeated 20 times Table 15.2 shows differences

in means between after and before runs for a selection of travel demand indicators There are no significant differences in terms of the total distance traveled, the total car distance traveled, or the other indicators There are also no significant differences in distributions of relevant characteristics of schedules, tours, or activities Thus, the model predicts that this scenario does not lead to a rescheduling of activities

to an extent that has measurable impacts on output variables

The scenario involves a change in the timing of a small proportion (15%) of activities of the input schedule skeletons The amount of time spent on the activities and the travel times do not change The change may have an impact on at least three types of conditions that play a role in the DTs First, it may lead to more time available in preferred time periods of the day for specific optional activities (e.g., more time in the afternoon for shopping) More time generally means higher probabilities of selecting the activity and choosing a long duration for the activity Second, shifts in work start times may impact relationships with the partner’s schedule In particular, the characteristics of tours of the partner over-lapping in time play a role in decision rules related to mode and location selection Obviously, the extent

to which tours of the partner overlap in time may increase or decrease if the work start time changes Third, a changed start time of the work activities may impact opportunities for travel party and trip chaining choices Theoretically, therefore, the scenario may lead to changes in activity schedules However, the results indicate that if it does, the effects are not measurable at the aggregate level

Finally, we should note that the system is not sensitive to all theoretically possible impacts of the scenario A 15% reduction of work trips during the morning rush hour probably leads to a meaningful decrease in congestion Faster travel times leave more room in schedules for other activities The present system uses predicted travel times under free-floating conditions only and, therefore, is not sensitive to such changes Furthermore, coupling constraints may lead to rescheduling of the activities of the partner Such decision rules are only partly represented in the current model

15.8 Scenario 3: Increase in Part-Time Workers

Shortening the work hours of full-time workers is a national policy in The Netherlands, aimed at reducing the unemployment rate and stimulating flexible work hours The third scenario assumes a 10% reduction

of full-time workers and a proportional increase in part-time workers

TABLE 15.2 Comparison of Means between Scenario 2 and Null Scenario on a Selection of Indicators

Total travel distance 84545 1423 237 0.280 0.574 39 Car distance (driver + passenger) 61420 802 436 0.709 1.616 43 Number of tours 5359 39 14 0.258 1.015 43 Trips–tours ratio 2.409 0.006 –0.003 –0.128 –1.466 42 Car driver–total distance ratio 0.678 0.009 0.004 0.577 1.274 43 Car passenger–total distance ratio 0.097 0.007 –0.002 –1.572 –0.697 43 Public–total distance ratio 0.127 0.010 0.000 –0.345 –0.150 41 Slow–total distance ratio 0.098 0.003 –0.002 –1.990 –1.793 43

Trang 8

This scenario could be implemented by editing the schedule skeletons and the job status of individuals and households However, schedule skeletons may change in more ways than just work hours For example, fixed activities may be reallocated across household members Therefore, the scenario was implemented by increasing the share of part-time workers in the current sample and decreasing the share

of full-time workers This was done such that the number of person days remains constant Full-time and part-time workers were identified based on weekly work hours of individuals More than 32 h of work in a week was identified as full-time and 17 to 32 h as part time.3

The change was applied to individuals belonging to a dual-earner household with the same probability

as to those belonging to a single-earner household (although one may argue that the first group is more prone to change to a part-time contract than the second group) Table 15.3 shows the multipliers that were used to implement the scenario for the different groups

15.8.2 Results

As in the previous cases, the impacts of the scenario were analyzed based on 20 prediction runs Table 15.4 shows the results for a selection of travel demand indicators This scenario leads to a reduction of total distance traveled of 3.0% The percentage reduction of total distance traveled by car is of the same magnitude (i.e., 2.8%) The share of distance traveled by slow mode has increased (3.7%), while the shares for the other modes have remained the same This suggests that the extra share of slow mode is drawn from no specific fast mode in particular

The number of tours has remained the same, implying that the decrease in distance traveled is exclusively caused by a decrease in average distance by tour (as was the case in scenario 1) Furthermore, the absence of a change in the trip–tour ratio suggests that the average number of activities per tour has stayed the same We conclude therefore that the reduction of travel demand is caused by a decrease in travel required per activity

The analysis of activity patterns confirms and details this picture Although the total number of out-of-home activities has remained the same, the distributions across activity type, day of week, and time

TABLE 15.3 Multipliers Used to Change the Sample Composition According to Scenario 3

Household Type Weekly Work Hours No of Cases Multiplier

Single, 1 worker > 32 220 0.9

Double, 1 worker > 32 632 0.9

Double, 2 worker > 32 496 0.9

3 For dual-earner households the same cut-off points of 17 and 32 hours work at the household level were used

As it appears, these cut-off points result in the desired size of the shift from full-time workers to part-time workers.

Total travel distance 81862 1682 –2446 a –2.988 –5.340 35 Car distance (driver + passenger) 59350 1012 –1634 a –2.753 –5.397 41 Number of tours 5334 41 –11 –0.210 –0.802 43 Trips–tours ratio 2.415 0.008 0.003 0.132 1.266 40 Car driver–total distance ratio 0.674 0.009 0.000 0.031 0.068 43 Car passenger–total distance ratio 0.102 0.011 0.003 2.939 1.017 33 Public–total distance ratio 0.120 0.013 –0.007 –5.843 –2.002 35 Slow–total distance ratio 0.104 0.004 0.004 a 3.691 3.007 39

a Significant at the α = 05 level in a two-sided test.

Trang 9

of day show changes We see a decrease in the number of work activities and increases in both bring–get and voluntary work activities In addition, we see a decrease in out-of-home leisure activities and an increase in service activities

As for day of the week, there are more activities on the early days (Monday and Tuesday) and fewer

on Saturdays in the after case Finally, with respect to time of day, the shares of activities starting before

10 A.M and after 6 P.M decrease, and the share of activities starting between 10 and 12 A.M increase Furthermore, the composition of tours has undergone changes The percentage of tours consisting of only one activity has decreased, whereas the share of two-activity tours has increased.4 Finally, we mention that the share of car mode in tours has decreased and the percentage of slow modes has increased The scenario involved a redistribution of cases across part-time and full-time workers The third section

of the report allows us to assess the extent to which other dimensions of the sample composition have changed simultaneously Redistributions have occurred on almost all other dimensions Only for age group, child index, and gender was the size of the differences meaningful With regard to age, a shift has occurred from the 25 to 45 years old group (–1.7%) to the older than 45 years old group (+2.4%) As for child index, the percentage of households without children has increased (+1.4%), while households with young children decreased (–2.0%) With regard to gender, the after case comprises fewer males (–3.3%) and more females (+2.4%)

As in the case of scenario 1, the increase in part-time work leads to a reduction of total travel distance The reduction is less than one would expect based on the reduction of work activities implied by the scenario The decrease in work activities is compensated by an increase in other out-of-home mandatory activities (bring–get, voluntary work) There is a slight increase in the average number of activities conducted on tours, but this cannot explain the reduction in the total distance traveled Instead, the reduction of travel demand is caused by a decrease in the average distance traveled per trip This is probably related to the shift of activities from work to other mandatory activities Locations of work tend

to be more distant from the home location than destinations of other activities Therefore, replacing work activities by other mandatory activities entails that more distant locations are replaced by destina-tions closer to home, resulting in less travel demand per trip Although activities have a more local character in the after case, approximately the same proportion of the distance is traveled by car as before Besides the reduction of total travel distance, the scenario has impacts on the spread of activities across days of the week and times of the day In the after case, there are more activities during the early days

of the week (Monday and Tuesday) and fewer activities on Saturday At the same time, a smaller proportion of activities is initiated during morning rush hour (as well as in the evening) These changes probably lead to a slightly more even distribution of traffic on a daily and weekly basis

Just as in the case of scenario 1, it is not possible to attribute the impacts to the variable of interest alone Changing the sample composition with respect to work status of households has led to changes

on other dimensions correlated with work status Specifically, the shares of households belonging to older age groups, households without children, and females have increased in the after case The present analysis does not allow us to unravel the separate impacts of the work status variable and these covariates

15.9 Scenario 4: Friday Afternoon Off

This last scenario assumes a 50% reduction of work hours on Friday The scenario is implemented by editing the schedule skeleton (as in scenario 2) rather than changing the sample composition (as in scenarios 1 and 3) Only Friday schedule skeletons involving more than 360 min work time were selected

4 The decrease of single-stop tours has not led to a significant increase of the trip/tour ratio.

Trang 10

In each selected skeleton, the duration of each work episode was reduced 50% by changing the end time while keeping the start time constant In most cases, this means canceling the afternoon work part Shortening the workweek in this way is a realistic scenario in the Dutch context

15.9.2 Results

As in the foregoing scenarios, predictions were repeated 20 times Sample composition is the same in the before and after cases, because the scenario involves changes in schedule skeletons only A selection

of most relevant travel demand indicators is represented in Table 15.5 As it appears, there is no significant difference in total travel distance between the before and after cases However, the distance traveled by car and the number of tours have increased 1.3 and 1.6%, respectively The increase in car kilometers is accompanied by a decrease in the public transport share in the total distance traveled, whereas the share

of slow mode has not changed Finally, we note that the average number of trips per tour has remained the same

The increase in number of tours is caused by an increase in flexible activities conducted on Friday (+9.5%) These concern shopping, service, and social activities The number of out-of-home leisure activities remains constant, but the duration of these activities tends to be longer in the after case Furthermore, we see a shift in the start time of activities The share of activities initiated between 10

A.M and 4 P.M increases at the expense of the before 10 A.M and after 6 P.M periods of the day A bigger proportion of activities is conducted with other members of the same household and on multiple-stop tours

In a substantial number of cases, part of the time that has become available because of the 50% work time reduction is used for conducting flexible activities such as shopping, service, and social activities

In other cases, the extra time leads to decisions to substitute short-duration leisure activities with longer-duration leisure activities These effects could be expected considering the underlying activity selection and activity duration DTs Available time in preferred time slots for the concerned activity is an important condition variable in both DTs (dependent on household and individual characteristics) The effect is not limited to activity selection and duration decisions Start time decisions also tend to change The observed shift in start time of activities can be interpreted as a shift toward the more preferred times of day for conducting activities such as shopping and social activities In summary, all these effects can be understood as consequences of relaxing temporal constraints on activity schedules

The increase in the number of out-of-home activities leads to an increase in the total number of trips and tours undertaken by the individuals It is noteworthy that the total distance traveled does not increase substantially The distance traveled by car does increase significantly, accompanied by a decrease in kilometers traveled by public transport Probably, there is a substitution between car and public transport directly related to work activity duration (on Friday) As the DTs for mode choice suggest, the preference for public transport decreases and the preference for car increases when the duration of work activities

Total travel distance 84613 997 305 0.3602 0.887 43 Car distance (driver + passenger) 61765 1028 781 a 1.2645 2.557 40 Number of tours 5429 45 84 a 1.5537 5.811 43 Trips–tours ratio 2.414 0.007 0.002 0.0973 1.011 43 Car driver–total distance ratio 0.677 0.010 0.003 0.4020 0.847 43 Car passenger–total distance ratio 0.106 0.007 0.008 a 7.2188 3.364 42 Public–total distance ratio 0.117 0.006 –0.010 a –8.8994 –4.339 41 Slow–total distance ratio 0.100 0.004 0.000 0.0040 0.003 39

a Significant at the α = 0.05 level in a two-sided test.

Định dạng
Số trang	11
Dung lượng	318,92 KB