1996-10-11

Promotie: ir. M.E. Kraan

Time to travel: A model for the allocation of time and money
Promotor:
prof.dr.ir. M.F.A.M. van Maarseveen 

1. 1 Research goal and background

The goal of this dissertation was to examine time expenditures on activity patterns. With these time expenditures the thesis tried to find whether limited time or money budgets impose limits to mobility growth. In the past, travel has been increasing continuously. Together with an increase in the distances, also travel time has increased, averaged over the total population. Total mobility is expected to continue to grow in the future, raising a lot of problems, in particular in a small] and densely populated country as the Netherlands. Existing network capacity is insufficient to cope with alt traffic, so congestion increases. Moreover, increasing mobility leads to negative environmental impacts.

In the Netherlands it was decided to diminish the increase of travel. Therefore, policy strategies were developed and examined. The second Traffic and Transport planning scheme ("Tweed Vervoersschema Verkeer en Vervoer", SVV2, Tweede Kamer der Staten Generaal, 1990) considers various policy measures to reduce travel growth. Travel is assumed to increase by 70% from 1986 to 2010, if no additional measures are taken. The Dutch Ministry of Transportation wants to reduce this growth to an increase of only 35%, by taking policy measures.

Various policy strategies have to be analyzed to determine their impact on travel. Due to the long term of these explorations, they have a strategy character. This means that direction and order of magnitude of future developments are important, not the precise values of developments. This thesis examined whether a large increase in travel growth is inherently impossible due to limited time and money budgets of individuals. The main research question was formulated as:

Can limited time or money budgets lead to limits in travel growth?

In order to answer this question the total activity pattern has to be considered. All activities, including travel take time. All duration together count up to 24 hours a day. So if travel requires more time, another activity (for example work) should take less time' Or inversely, when working week reduces, more time is available for other activities, including travel. Similar observations can be made for money: the more spent on travel the less money is available for other activities or goods. This was formulated in a research question, preliminary to the main question:

1 How do people allocate their time and money to their total activity pattern?

1.2 Research approach

To answer these questions, various concepts from the literature for modeling travel expenditures were compared and studied for their applicability to describe the total activity pattern. Then a theory was formulated on activity behavior, consisting of assumptions on the allocation of time and money to the activity pattern. This theory was compared with empirical data. Both the theory and the empirical data were considered for different population groups defined by their time and money budgets. The assumptions were translated in a strategy, mathematical model. In a theoretical analysis of the model its performance was examined, in particular whether impacts of various changes in exogenous variables were logical.

Estimating the model parameters then tested the theory. Due to practical considerations it was not possible to estimate the combination of the time and money expenditures. Therefore, the model was restricted to time allocation and the parameters were estimated on time budget data. With the estimated model, future scenarios could be explored. Changes in demographics, time budgets, and individual behavior have various impacts on the activity pattern. The analysis concentrated on travel times. Increases in total travel time, for the whole population were then translated into developments of distances, for various scenarios on travel speed. The explorations are based on the assumption that the transport systems in the Netherlands is assumed not to change drastically, during the period considered. Comparison of the results with results of other studies, in particular, with the SVV2, showed that

An increase of car mobility growth by 70% over a 25 year period is not likely to occur, under the given assumptions of limited time budgets and hardly any changes in the transport system in the Netherlands.

1.3 Research questions and answers

This research approach can be formulated by research questions ordered by the chapters where the questions are answered. The answers and main conclusions are given immediately.

Chapter 2 answered the next question by comparing various concepts from the literature.

Which techniques flor modelling travel expenditure can be found in the literature and which are applicable flor our goal?

The classicfour-stage model was found to be inappropriate to model travel expenditures, because no account is taken of individual budget constraints. Time and money expenditures are calculated in terms of travel time and costs, representing the network performance. These expenditures are then " back into assignment, modal split, and distribution modules to act as explanatory variables. This is iterated until demand and supply is in equilibrium. However, there is no feedback into the trip generation module. This means that the number of trips is independent of network performance.

Furthermore, classic four-stage models are trip based. Travel is considered independently of the total activity pattern. As travel is a derived demand, understanding travel behavior requires the understanding of activity behavior. New versions of four-stage models are closely related to activity approaches, because the generation module determines activity chains, or trip chains instead of single trips.

But still, no interaction between activity duration and travel time expenditure is incorporated.

The travel budget approach regards time and money budgets explicitly, hut is also found to be inapplicable to search for limits. By assuming that the amount of time and money spent on travel is constant, this approach allocates those expenditures to distances for various modes. Time or money saved on travel is thus spent on more or other travel, leading to counter-intuitive results if travel speed or costs change.

This fixed budget approach considers time and money expenditures on travel as exogenous, fixed input. However, this thesis assumes them to depend on the demand for travel (the activity pattern) and the supply characteristics (the network). Therefore, travel expenditures should be considered as endogenous variables, as output of activity behavior.

An improvement was made in the flexible budget approach, which allows exchange between travel time expenditure and leisure time or between travel costs and consumption goods. However, both budget approaches consider travel not in the context of the activity pattern, because they maximize people's spatial opportunities (distances) under budget constraints irrespective of the duration and frequency of activities for which one’ as to travel. Again, the conclusion was stated that the total activity pattern should be considered.

In the time allocation approach the inclusion of the total activity patterns possible. This approach considers two characteristics of the activities: time duration and costs in terms of goods consumed. The combination of goods purchased and time spent on activities yields an amount of satisfaction. This is represented by a utility function, which is maximized under a time and money budget constraint. Implementations of the theory mostly consider households allocating time and money to leisure activities and consumption goods. Travel related applications are hardly developed, nor are individuate based implementations.

This time allocation approach can be extended and applied to explore limits to nobility growth. But first, more knowledge on activity patterns is required. Chapter 3 answered therefore the questions:

·

What do activity patterns look like?

·

Theoretically: what components does an activity pattern consist of? Empirically: how much time do people spend on various activities?

People's activity behavior is based on various choices and consents. Long-term, medium, and short term choices can be distinguished. Constraints can he exogenous or endogenous, imposed by choices on a higher level. The individual time and money budgets are such endogenous constraints, imposed by the choices of work. The activity pattern is the outcome of the combination of both choices and constraints.

An activity pattern consists of activities in- and out-of-home, with a duration and frequency.-This thesis distinguishes tbree types of activities: obligatory (work or study: diserctionary in the long term, but obligatory in the short term), maintenance (housework, child care, shopping, etc.), and leisure (social and cultural, active and passive recreation). Personal needs (sleep, eating, washing) are considered as fixed activities on all time horizons. They are left out of the analyses. For activities out-of- home a certain distance bas to be travelled. Activities and travel not only take time, but also cost an amount of money. Except work, which yields an amount of money. There exists a trade-off between time and money through the working hours.

Activity patterns differ over the population. People with similar time and money budgets are assumed to have a similar activity pattern. The budgets are the outcome of various long-terrn life style choices, in particular, the employment status. So activity patterns are considered for population groups, classified by employment status.

Survey data describing these total activity patterns combining time and money are not available in the Netherlands. The combination of various survey data, on the other hand, is not reliable for studying the interaction between time and money. Therefore, the theory was restricted to time expenditure. The Time Budget Survey of the Netherlands ("TijdsBestedings Onderzoek", TBO) was found appropriate to study activity patterns.

The TBO data showed that nearly half the time is spent on physiological needs (sleep and personal care) and a quarter on leisure. The other quarter is spent on labor, education, maintenance and travel. Most variation is in this last quarter, due to people's task (work, study, household). The activity patterns were further analyzed for population groups. Significant differences were found in the distribution between in-home and out-of-home activity duration’s, frequencies and travel.

With respect to the activity patterns, the population groups could be categorized into three main groups. People combining tasks have a busy activity pattern. These people combine work or study with maintenance activities. Single workers, students living on their own, and part-timers belong to this category. A lot of time is spent on work or study out-of-home. A large proportion of the small amount of leisure time is spent out-of-home. The number of trips and travel time expenditure are large.

Opposite this group, people with a calm activity pattern are distinguished. These are people with no obligatory activities. Pensioners, unemployed, and housewives belong to this group. They spend a lot of time at home, mostly on leisure. Housewives divide their time at home equally over leisure and maintenance. The number of trips and travel time expenditure are small.

In between, people with one task (full-timers and pupils not living on their own) are considered. They also spend a lot of time out-of-home on obligatory activities. Little time is left for leisure and hardly any time is spent on maintenance. Travel time expenditure is also large.

Concerning travel modes another categorization can be made. Working people, full- timers in the extreme, travel mostly by car. Pupils and students (including those on their own) use slow modes and public transport a lot. People with no obligatory activities are grouped in between these extremes.

With these differences in activity patterns the assumption to distinguish between people based on their employment status is justified.

Next, the theory on activity behavior had to be translated into a mathematical model. Therefore, chapter 4 answered the following research questions:

·

How do people allocate their time and money to activities and goods?

·

How is this formulated by a mathematical strategy model?

·

And how does the model perform theoretically?

People allocate the available time and money, including the income gained from work, to various activities, in-home and out-of-home, travel and goods. The allocation is based on the utility maximization approach. Participation in activities and consumption of goods both yield an amount of utility, which is, maximized under time and money budget constraints. The utility function depends on activity duration, out- of-home and in-home, frequency, distance, and goods. Utility is concave in all its arguments, in order to impose the principle of "more is better". The larger the activity duration, frequency, etc. the higher the utility. Marginal utility diminishes.

Activities, travel and goods cost money; activities and travel also take time. The sum of all time and money expenditures should add up to the time and money budgets. The time budget is given by 24 hours a day minus the time spent on physiological needs and work. The money budget is given by unearned income plus income from labour: the hours spent working multiplied by the wage rate. There is a relation between time and money budgets through the working hours.

With T the activity duration out-of-home, d the distance travelled, fl the frequency of out-of-home activities, T, the duration at home, G the amount of money spent on goods and savings, T, the working hours, T"" the time budget (24 hours a day minus time for personal needs), Y unearned income, w the wage rate, v travel speed, c, the costs per hour for out-of-home activities, of the costs per occurrence of out-of-home activities, and c, the costs per kilometre travelled. 0, y, i5, p, X are parameters in the utility function, such that P + y + l@ + p + X = 1.

Demand functions were derived, and theoretical analysis was done by means of elasticity’s, value of time, and substitution effects. These all describe the short-term impacts of (small) changes in time or money budgets, wage rate, costs, and travel speed. Contrary to a lot of applications of time allocation models, the model is defined for individuals, instead of households, because variation between individuals is much larger than between households.

Theoretically the model performs quite well. Empirical testing of the model is the next step. Therefore, the following questions for chapter 5 is posed:

·

How can the model be estimated?

·

Which data are required and what procedure can be used?

·

How does the model perform empirically?

Due to the lack of combined data the model was estimated on time use data, after restricting the model to time expenditure. It was already explained why the Time Budget Survey was found applicable to test the model. The estimation procedure chosen was the ordinary least squares approach.

The model was first estimated for all activities together, distinguishing activity duration, out-of-home versus in-home, and travel time. Secondly, discretionary activities were considered, that is maintenance, leisure, and all travel. The time budget was reduced with the time spent on obligatory activities. The parameter values were in most cases significantly different from zero, which means that the model could be estimated in agreement with the theory.

Comparison of the parameter values, elasticity’s, and marginal rates of substitution was made. Parameter values indicate variations in time duration’s for variations in the time budget. Elasticity’s indicate how additional time, by larger budget is allocated to activities. The marginal rate of substitution indicates how travel time is substituted for time at home or out-of-home, to obtain equal utility. It is a proxy for the different values of time.

The results of the estimation show large differences between the population groups. The results of all these analyses for discretionary activities distinguish three typical, extreme, groups. Part-timers, combining works with maintenance-, pupils and students, with only one task: education-, and pensioners, with no obligatory activities. Part- timers show large variations in activity duration at home. They will allocate additional time to activities, and not to travel. They have a small value of travel time. Pupils show large variations in out-of-home activities and travel. Additional time will be allocated more than proportional (to the actual allocation) to out-of-home activities, less than proportional to in-home activities. Their value of time at home is very small, even travel time gives them more utility than time at home. Pensioners show large variations in time durations at home. They will allocate additional time proportional to all durations. Their value of time is equal for all duration (in-home, out-of-home, and travel).

The results are very encouraging. The three typical groups correspond with the classification given for the activity patterns. These groups are the people combining tasks with busy activity patterns (part-timers), people with only one task (pupils), and people with no tasks and a calm activity pattern (pensioners).

A comparison of estimation results over more years shows very stable parameter values. It can he concluded that the model performs quite well, empirically. The estimated model for the allocation of time was then applied for future forecasts. The research questions for chapter 6 are given by:

·

What are the impacts on travel time of changes in population size and composition,in time budgets, and in model parameters?

·

What are the implications flor total travel?Which conclusions can be drawn flor Mure travel developments?

·

Is travel growth limited?

Future scenarios for 2015 on demographics (population size and composition), changing working weeks, and time dependency of model parameters (autonomous changing behaviour) were developed. Long term scenarios for the Netherlands developed by the Central Planning Office ("Centraal PlanBureau", CPB) were found to have little variation in demographics. These were considered together in one global scenario. This scenario was called Moderate, because population composition hardly differs from 1990. Therefore, the impacts for this scenario were moderate. In search for limits, two extreme scenarios were developed. The Labour scenario assumes a large proportion of people working. The Aged People scenario assumes a large proportion of aged people. In order to have consistency with total labour force, next, the individual working week in the Labour scenario was assumed to reduce, while in the Aged People scenario it was assumed to enlarge. Comparison of parameter values over more years gave some indications for future values. Upper and lower hounds for travel time parameters were defined.

For the scenarios future travel times were calculated. The Moderate scenario hardly shows a change in individual travel times, averaged over the total population, explained by the similar population composition. Due to an increase in population size the total travel time for the whole population, is expected to increase. The Labour scenario shows an increase in both individual and total travel time. The Aged People scenario shows a reduction in individual travel time, but due to an increase in population size a growth in total travel time.

Changes in the working week amplify the effects of population composition. A reduction in working week induces larger travel times, an enlarging working week induces smaller travel times.

The changes in parameter values also reinforce the effects. Total travel time is expected to increase at least by 9%, in case of the Aged People scenario with enlarging working week and for minimum parameter values. Total travel time might increase by 34% at the most, in the case of the Labour scenario with reduced working week and maximum parameter values.

Comparison of these extreme potential developments with other studies requires scenarios on average travel speed, because most studies give forecasts on travel distances. Considering the reconstructed history development of travel speed, this thesis assumes that average travel speed increases by 10% at the most over a period of 25 years. Also scenarios for speed to remain constant or even decrease were analysed. The most extreme total scenario implies total travel distance to increase by 45% (over the period 1990-2015) at maximum. Translating this into car distances showed that car distances are expected to increase by 60% at the most.

These extreme scenarios show a difference with a forecast given by the SVV2 of an increase of car mobility by 70% for unchanged policy. The extreme scenarios are based on extreme population composition, limited time budgets, extreme changes in behaviour, and hardly any change in the transport system in the Netherlands. The difference can be explained by the different approaches, including the different forecasting period (1985-2010 in SVV2 and 1990-2015 in this thesis). Furthermore, due to the lack of money expenditure en prizes, the model in this thesis is still incomplete. Chapter 7 shows how the approach of this thesis and the model can be improved by incorporating money expenditure, income, and prizes. But also by distinguishing the population by car ownership, personal characteristics (age, gender, education), and travel speed. It is recommended to carry out the scenario analysis per population group. And thus determining travel distances per mode for each population group. Another important explanation of the difference are the changes in the transport system. In the SVV2 the transport system is assumed to change such that travel speed increases more than assumed in this thesis.

The model is not yet complete, but a first step has been taken in the direction of modelling the allocation of time and money. The next step is to improve the model in order to search for limits in individual budgets and limits in mobility growth.

2 Conclusions

This dissertation analysed activity patterns and developed a model for the allocation of time and money. This model was intended to analyse whether the limited individual time and money budgets lead to limits in mobility growth. Due to the lack of combined data for time and money expenditures, the model was estimated, for different population groups, on time budget data only. With the model extreme future scenarios on population composition, changes in working week, and changes in behaviour, were examined. These calculations resulted in travel growth less than the number found in the Second Traffic and Transport planning scheme ("Tweede Vervoersschema Verkeer en Vervoer", SVV2), for unchanged policy. The difference in growth is explained by the different approaches of both models used. Mainly because the model used in this thesis is still incomplete. It is recommended to improve the model by including money expenditure, income, and prizes. Therefore, combined data on time and money expenditures have to be gathered.

On the other hand, the Dutch National Model System (Landelijk Model Systeem, LMS), used for the calculations for the SVV2, does not account explicitly for individual time budget and time expenditure on the total activity pattern. Although this thesis does not prove the existence of limits to mobility growth, it suggests that the total activity pattern should he considered, in combination with individual time and money budgets and expenditures to account for individual limitations.

Another important difference between both methods is given by the changes in the Dutch transport system. This thesis assumes hardly any increase in the average travel speed, while in the SVV2 the transport system is assumed to improve such that larger increases in average speed are considered. Based on these findings, it is argued that travel speed might be an interesting instrument for transport policy measures. Enlargement of travel speed will increase total travel, while reduction in speed will lead to a reduction of travel growth. As long as speed will not increase, mobility growth is limited.

Voor meer informatie: a.m.klijnstra@utwente.nl