Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Journal of Transport Geography 19 (2011) 323–331

Contents lists available at ScienceDirect

Journal of Transport Geography


journal homepage: www.elsevier.com/locate/jtrangeo

Estimating social travel demand of senior citizens in the Netherlands


Pauline van den Berg, Theo Arentze, Harry Timmermans *
Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands

a r t i c l e i n f o a b s t r a c t

Keywords: Like many other countries, the Netherlands is experiencing a sharp rise in the ageing population. As age
Travel demand increases, people’s mobility may decrease. However, older people have more leisure time compared to
Social trips their younger (working) counterparts, and potentially spend more time on social activities. Therefore,
Senior citizens this group can possibly increase social travel demand. However, to date, the travel demand for social
activities of senior citizens has received only little attention. This paper studies trip-making for social
purposes, with a special focus on the demographic ageing factors. Using social activity diary data, models
are estimated to predict the number of social trips, the travel distance and mode of transport for social
trips. The results indicate that the elderly of today seem to be as mobile as their younger counterparts
with respect to the number of social trips. High education and involvement in clubs on average result
in more social trips and full time work is found to result in fewer social trips. With regard to trip distance
the results show that the average travel distance does not decrease as people get older. Full time work is
found to result in longer social trips. Shorter trips were found for people in urban as well as rural areas.
Trips for the purpose of visiting or joint activities tend to be longer than average. With regard to transport
mode choice the results indicate that older seniors (75+) are less likely to choose the bicycle, relative to
driving. No other significant age effects were found. Significant effects were found for gender, household
structure, education level, car ownership, having a disability, urban density, distance and the purpose of
the social activity.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction sources (Silvis and Niemeier, 2009), such as instrumental support


(caring if they are sick, help with chores), emotional support or
The population of the Netherlands (and all industrialized coun- companionship, while they in particular rely on their social net-
tries) is getting older. Currently, 15% of the Dutch population is 65 work to obtain these resources (Agneessens et al., 2006). This dif-
or over and it is expected that by 2030 this will be approximately ficulty is reinforced by the fact that social networks are
25% (Statistics Netherlands, 2009). Therefore the quality of life of becoming more spatially dispersed. Therefore people currently rely
this growing group of senior citizens deserves attention. more on fast means of transport and remote means of communica-
Social activities and mobility are important aspects of quality of tion to cover the distance to their social network members. In a
life for the elderly (Tacken, 1998; Banister and Bowling, 2004). survey of the quality of life of the elderly, Banister and Bowling
Increasing age is found to be associated with decreasing mobility: (2004) found that people with access to a vehicle (and people with
older people are found to travel less than younger people in terms access to good local transport) were likely to undertake more social
of trips per day and distance traveled (Tacken, 1998; Silvis and Nie- activities.
meier, 2009). Senior citizens also tend to have a lower propensity Although older people are still found to travel less than younger
to perform social activities than their younger counterparts (Aoki people in terms of trips per day and distance traveled, this is not
et al., 1996; Páez et al., 2006; Kemperman et al., 2006; Carrasco the case for all trip purposes. Moreover, the rapidly growing group
and Miller, 2006; Farber and Paéz, 2009). This might be due to of senior citizens in our society on average is wealthier (but with
the fact that senior citizens have smaller than average social net- substantial within-cohort variability), healthier and more mobile
works (van den Berg et al., 2009). Having a smaller social network than earlier generations (Newbold et al., 2005; Mercado and
might mean that senior citizens have less access to a variety of re- Páez, 2009). Senior citizens nowadays have extended periods of
free time and a desire to travel. Their travel demand and mobility
needs are changing. Their number of trips and kilometers travelled
* Corresponding author. Tel.: +31 40 247 2861; fax: +31 40 243 8488.
E-mail addresses: p.e.w.v.d.berg@tue.nl (P. van den Berg), t.a.arentze@tue.nl
are increasing (Rosenbloom, 2001; Collia et al., 2003) as well as
(T. Arentze), h.j.p.timmermans@tue.nl (H. Timmermans). their levels of car ownership (Schmöcker et al., 2005; Banister

0966-6923/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jtrangeo.2010.03.018
324 P. van den Berg et al. / Journal of Transport Geography 19 (2011) 323–331

and Bowling, 2004; Newbold et al., 2005). Therefore, this growing to compare travel behavior among Canadians in different age co-
group of elderly people can potentially increase the travel demand horts, also found that older people make fewer trips and travel
for social activities. for different reasons than younger generations, but their reliance
Although the studies into the travel behavior of senior citizens upon the private car stays equally high.
in the population have started to increase over the last decade When studying the travel behavior of senior citizens, one
(Collia et al., 2003; Newbold et al., 2005; Páez et al., 2006), the tra- should keep in mind that seniors are not a homogenous group. This
vel demand for social activities of seniors has received only little is pointed out by Alsnih and Hensher (2003) who state that a dis-
attention in academic research. In order to plan future transporta- tinction between younger seniors (65–75) and older seniors (75+)
tion needs for social activities of the ageing population, we need to is useful, as the health of people often starts to decline at an age of
understand the nature of their social activities and the travel in- 75. Hildebrand (2003) developed a simplified activity-based travel
volved in these activities. In this paper we focus on age in relation model, in which six different lifestyle groups, based on socio-
to activity-travel behavior for social purposes. demographic variables, were used. He found diverse mobility and
Analyzing data from a social interaction diary, we examine how travel characteristics among the groups.
age is related to a range of social activity-travel behavior measures, Schmöcker et al. (2005) and Páez et al. (2006) used ordered pro-
such as social trip-making propensity, the travel distance for social bit models to estimate the trip-making frequency of the elderly.
activities and the transport mode choice for traveling to the loca- Schmöcker et al. (2005) estimated a model for the total number
tion of the social activity. Our unique data-set allows us to examine of trips of elderly and disabled people, as well as different models
the social activity-travel behavior of senior citizens in much detail. for different trip purposes (working, shopping, business, and recre-
The remainder of this paper is structured as follows. The next ation). They found that increasing age and walking difficulties re-
section reviews the existing literature on travel behavior and social sult in fewer trips. Household structure, ethnic background,
activities in relation to age. Section 3 describes the data and meth- income, access to a car and taxi or public transport card ownership
ods of analysis used in this study. In Section 4 three models are were found to have an effect on trip-making frequency as well.
presented that are estimated to predict the effect of age on the Páez et al. (2006) also found that trip-making propensity decreases
number of trips: (1) the travel distance (2) and the transport mode with age. Females were found to be more likely to make more trips
choice (3) for social trips. Finally, Section 5 contains the Conclu- than males. License, car ownership, and transport pass ownership
sions and discussion. were positively related to the number of trips as well.
Stern (1993) used a Poisson regression model to estimate the
number of trips for different purposes by elderly and disabled peo-
2. Literature review
ple in rural Virginia. The number of trips was found to be smaller
for females, older seniors and people with walking difficulties.
This section reviews the existing literature on the effects of an
High education and being married resulted in more trips.
ageing population on travel behavior. It focuses on the differences
In order to examine the travel distances of senior citizens,
between younger and older generations, the trip-making frequency
(log-)linear regression models are usually used. Schmöcker et al.
of the elderly, their travel distance and their transport mode
(2005) found that average trip distances decrease with age, how-
choice. This section concludes with a literature review of social
ever, recreational trip distances increased at least until the age of
activities and ageing.
80. Positive effects were found for income, driver’s license and
car ownership. Higher urban density, walking difficulties and a
2.1. Trip-making among senior citizens higher number of trips result in shorter distances.
Mercado and Páez (2009) used multilevel models to study the
Over the last years, the number of studies on different aspects of effect of age on travel distance by different transport modes (car
trip-making of the elderly has increased rapidly. Many studies on as driver, car as passenger and bus). They found a decline in dis-
the travel behavior of seniors are limited to presenting descriptive tance traveled as car driver as age increased. They also found sig-
statistics comparing younger and older people (e.g. Rosenbloom, nificant effects of gender, work status, household size, license
2001; Collia et al., 2003; Newbold et al., 2005). However, studies and car ownership, income and land use.
presenting models to estimate trip frequency, travel distance and Several studies focus on senior citizens’ transport mode choice,
transport mode choice of the elderly are currently appearing (e.g. using multinomial logit (MNL) models (e.g. Stern, 1993; Kim and
Schmöcker et al., 2005, 2008; Páez et al., 2006; Arentze et al., Ulfarsson, 2004; Su et al., 2008; Schmöcker et al., 2008). Stern
2008; Mercado and Páez, 2009), as well as models of trip chaining (1993) used an MNL model to estimate the mode choice by elderly
(e.g. Golob and Hensher, 2007). and disabled people in rural Virginia. The different categories in-
Rosenbloom (2001) used data from the US, Australia, Germany, cluded driving, being driven, bus, taxi, walking, and different local
New Zealand, Norway and the United Kingdom to study the envi- (van) transit systems. Driving was found to be the preferred mode.
ronmental implications of an ageing population and its increased Females were more likely than males to choose one of the other
automobility. The results indicate that senior citizens around the transport modes. People with walking problems were more likely
world, compared to earlier generations, are more likely to have a to use a taxi.
license and to make more trips by car and less by public transport, Examined the modal choice of senior citizens for leisure trips,
which is likely to cause increased environmental pollution. based on data collected in 1996 in the Netherlands. Their results
Collia et al. (2003) compared the basic travel characteristics of indicate that seniors prefer to take the car for leisure trips, espe-
older American adults (65 and over) with those of younger adults cially if they own a vehicle. The authors also found significant ef-
(19–64). They show that although older Americans travel exten- fects for education and the residential environment.
sively, they are less mobile than younger generations in that they Kim and Ulfarsson (2004) presented an MNL model for the
make fewer trips, travel shorter distances, and have shorter travel choice of senior citizens (65 and over) between private car, car
times. This pattern is more pronounced among older women and pool, public transport and walking. Their results indicate that
those with walking disabilities. Older Americans rely on personal the car is the preferred mode choice of senior citizens. The choice
vehicles as heavily as younger generations. 89% of all trips are for a car is negatively influenced by age (younger seniors are
made by car, 9% on foot and 2% by other means. Newbold et al. more likely to choose the car) and positively influenced by vehicle
(2005), who used the 1986, 1992 and 1998 General Social Surveys ownership, high income and living alone. Walking was preferred
P. van den Berg et al. / Journal of Transport Geography 19 (2011) 323–331 325

over public transport. For personal and recreational trips walking studies is not specifically on senior citizens. These studies indicate
was the principal mode. The proportions of trips by public trans- that older people undertake fewer social activities compared to
port increased during the morning and midday, if the distance to younger generations.
a bus stop was shorter, and if the travel distance was longer. Aoki et al. (1996) developed a social activity measure, consisting
The same data-set was used by Su et al. (2008) to estimate MNL of a score on job engagement, social participation or socially-re-
models and nested logit (NL) models to estimate the mode choice lated activities, activities for learning and personal activities. They
of elderly people for shopping trips. They also categorized trans- found that social activities of Japanese seniors decreased as age in-
port mode into car as driver, car as passenger, public transport creased. They found the social activity levels among the Japanese
and walk. They found none of the socio-economic variables to be elderly to be higher for males, higher educated people and senior
significant in the MNL models. Access to different modes, such as citizens who had hobbies and were physically active at middle age.
car ownership and bus stop density, as well as travel time and cost On the other hand, older (retired) people have more free time
were found to have a significant effect on mode choice. The results compared to younger people. Time spent on work is replaced by
of the NL models were similar, but with more significant variables, activities at home, visiting, shopping and recreational activities.
such as income, urban density, having a disability, and age. Driving Tacken (1998) shows that visits and leisure time activities are
was found to be the preferred mode. the most important activities for senior citizens, and that they
Schmöcker et al. (2008) also analyzed mode choice decisions of spend much more time on these activities than on others. The time
elderly and disabled people in London, based on personal and spent on visiting other people increases substantially with age Tac-
home location attributes. They included six mode choices, namely ken (1998).
car as driver, car as passenger, bus/tram, train/tube, taxi and walk- As social activities often involve travel, the growing group of el-
ing. They found a high preference for car use; public transport was derly people can potentially increase the travel demand for social
not preferred in most cases. activities. So far, little is known about the social travel demand
If the transport mode choice of senior citizens is estimated in of senior citizens. This study will therefore aim at increasing our
the Dutch situation, the model should include the bicycle as a understanding of social trip generation of different age cohorts,
transport mode. This is pointed out by Tacken (1998) and Kemper- using a unique data-set showing their performance of social activ-
man and Timmermans (2009). Tacken (1998) analyzed data of the ities and trips for different social purposes.
1994 Dutch National Travel Survey. He found that in the Nether-
lands 19% of all trips by people aged 75 or over are made by bicy-
3. Data and methods
cle. Kemperman and Timmermans (2009) analyzed the 2006
Mobility Research Netherlands (MON) data to study the participa-
3.1. Data collection
tion in physical activities of the elderly and found that almost 23%
of trips by elderly people were biking trips and over 25% were
For this study we use data that were collected between January
walking trips. The car, however, remains the most important mode
and June 2008 in a number of neighborhoods in and around Eind-
among Dutch senior citizens.
hoven. The data-set includes 732 respondents aged 18–89. The
Arentze et al. (2008) used a comprehensive activity-based mod-
data collection instrument consisted of a paper and pencil social
el combined with a trend analysis to predict mobility effects of an
interaction diary, in which the respondents were asked to record
ageing population taking into account cohort effects and activity
all their social interactions during two days. Social interactions
scheduling effects in choice behavior of the elderly. Their results
were defined as all forms of social contact, for instance visiting,
indicate that cohort change leads to a total growth of the mobility
performing a joint activity, having a conversation (face-to-face as
of senior citizens in terms of passenger kilometers (especially by
well as over the phone or online), sending or receiving an e-mail,
car), travel time and trips.
an SMS, a letter or a fax. Interactions at work or school about work
Overall, the findings of the reviewed studies suggest that, in
or school matters were not recorded; nor were interactions at
spite of cultural differences, older people around the world are
home with household members. Interactions that had a more busi-
found to travel less frequently and shorter distances than younger
nesslike nature than a social nature, like interactions with un-
generations. However, their number of trips and travel distances
known shop assistants when paying for merchandise were not
are increasing. Many senior citizens nowadays have access to a
recorded either.
car and the car is their preferred mode of transport. The use of
In the social interaction diary, information was gathered about
alternative transportation, including public transport, is found to
the communication mode used for the interaction, the day of the
be limited.
week and the time of the day, the purpose of the interaction, with
whom and where the interaction took place. If travel was involved,
2.2. Social activities in an ageing population
the travel distance and transport mode were recorded as well. In
addition to the social interactions, a number of personal character-
Older people have different patterns of activities during the day
istics were collected, as well as a number of characteristics of the
compared to younger generations. For the younger age group work
respondents’ residential area. For further details about the data
is an important share of the daily activities. For most senior citi-
collection procedure, see van den Berg et al. (2008).
zens, work is no longer part of their activities. This change that
comes with retirement can affect the social activities of senior
citizens. 3.2. Methods
Social activities have gained only little attention in activity-
based analysis. They have been viewed as a single category or have Using the data from the social interaction diaries and the ques-
even been combined with other leisure activities. However, within tionnaires, the relationship between travel demand for the purpose
the wide range of social activities, different types of activities can of social activities will be estimated. We consider the travel de-
be distinguished, such as visiting, performing a (club or union) mand for social activities to be the result of three decisions:
activity together, or just chatting. The number of studies on the
variety and choice of social activities is limited. Some examples 1. The number of social trips made in 2 days.
are the studies by Kemperman et al. (2006), Carrasco and Miller 2. The distance traveled for social activities.
(2006) and Farber and Paéz (2009). However, the focus of these
326 P. van den Berg et al. / Journal of Transport Geography 19 (2011) 323–331

3. The transport mode used to travel to the location of the social 3.2.3. Mixed logit model
activity. Finally, in the third model, the transport mode choice of senior
citizens for social trip-making is analyzed using a mixed logit mod-
We use a set of regression models to analyze these variables in a el. In the standard logit model, the alternative specific constants
sequential procedure, as a result of a number of personal and and coefficient parameters (b’s), are fixed, which means they are
household characteristics. In addition to the personal and house- the same for everyone. Moreover, standard logit does not take into
hold characteristics, in the second model the dependent variable account any unobserved factors that persist over time for a given
of the first model (number of social trips) is included as an explan- decision maker (with repeated choices). In mixed logit models,
atory variable. In the third model both the number of social trips both random taste variation and correlation in unobserved factors
and social travel distance are included as an explanatory variable over time, can be accounted for. For the mixed logit model, that
in the third model. This way, information overlap between the takes into account heterogeneity and repeated measures, the util-
models is avoided, while the travel demand for social activities of ity for individual i for alternative j on choice occasion t would be:
senior citizens is analyzed exhaustively (Molin et al., 2007). We
use Nlogit (Greene, 2002a) to estimate models 1 and 3, and the U ijt ¼ b0 xijt þ eijt
plm package (Croissant and Millo, 2008) in R for model 2. We where eijt is an unobserved random term that is distributed iid
use a negative binomial model, random effects model and a mixed extreme value, independent of bi and Xijt. Each random parameter
logit model, respectively. bi is defined as the average preference in the population, b, and an
individual deviation, gi which represents the individuals’ prefer-
3.2.1. Negative binomial model ence relative to the average preference for a particular transport
As the number of social trips a respondent made in 2 days is a mode. The utility is:
count variable, Poisson regression or negative binomial regression 0
can be used. We use a negative binomial regression model as this U ijt ¼ b xijt þ g0i xijt þ eijt
allows the variance to differ from the mean. Negative binomial If bi were known, the probability that individual i chooses alter-
regression can be considered a generalization of Poisson regression native j at choice occasion t would be standard logit:
and assumes that the conditional mean ki of yi is not only deter-
0
mined by xi but also a heterogeneity component ei unrelated to Lij ðbi Þ ¼ ebi xijt =Rj ebi xijt
xi. The formulation can be expressed as:
However, since bi is random and not known, the (unconditional)
y
ProbðY  yi Þ ¼ eki ki i =yi !; y ¼ 0; 1; . . . ; choice probability is the integral of this logit formula over the den-
sity of bi. Assuming that the preferences vary in the population
where lnki = b’xi + ei. with density f(b|h), where h are the parameters of this distribution,
In this model exp(ei) has a gamma distribution with mean 1.0 the actual probability is (Greene, 2002a):
and variance a. The formulation of the negative binomial distribu- Z
tion which can be used to model count data with overdispersion is
Pij ðhÞ ¼ Li ðbÞf ðbjhÞdb
then derived as:

Prob½Y ¼ yi  ¼ Cðh þ yi Þ=½CðhÞyi !uhi ð1  ui Þyi 3.3. Explanatory variables

where C is the gamma distribution, ui = h/(h + ki), h = 1/a, a is a dis-


The choice of the explanatory variables for these models is
persion parameter, such that based on a review of the existing studies on trip-making in relation
Var½yi  ¼ E½yi f1 þ aE½yi g to ageing. These studies agree that the explanatory variables
should include personal or household characteristics (age, gender,
ethnicity, number of people in the household, income, work status,
3.2.2. Random effects model level of education) to explain the individual’s propensity to travel.
To analyze the effect of the explanatory variables on the social Others also include characteristics of the neighborhood (popula-
trip distance, a linear regression model can be used. However, tion density, distance to bus stop), different forms of disability
the data on this level has repeated measures (multiple trips per and mobility characteristics (car ownership, license, public trans-
respondent). The unexplained variation across groups (trips per port pass possession, walking difficulties). To explain the propen-
individual) can be captured in simple shifts of the regression func- sity to perform social activities personal and household
tion (i.e., changes in the intercepts) by using an effects model. Let characteristics are usually used, as well as mobility characteristics
yit be the distance of trip t of individual i. The structure is (Greene, and social indicators such as social network size or involvement in
2002b): clubs.
yit ¼ li þ b0 xit þ eit The increasing income and increasing car possession among el-
derly will probably increase future travel demand for social pur-
where eit is a random disturbance term with E[eit] = 0 and Var[eit] = poses. More and more senior citizens in the Netherlands
r2e . nowadays tend to move from urban to rural areas (Jorritsma and
The effects may be fixed or random. In the case of a fixed effects Olde Kalter, 2008). Against this background it is interesting to
model li is a separate constant term for each unit (individual). In a examine the effect of urban density on social trip variables (after
random effects model li is an individual specific disturbance, correcting for socio-economic and socio-demographic variables).
which is assumed to be independent from the regressors. As our The data-set used in this paper was not collected with a special
model contains regressors with no within group variation (per- focus on senior citizens (or people with disabilities). As a result, the
sonal characteristics are constant for the trips of a person), we esti- sample of elderly people and their social trips is rather small.
mated a random effects model of the following form (Greene, Therefore, the number of explanatory variables in the models is re-
2002b): stricted. We therefore included only one variable for residential
location (namely urban density) and one social indicator (namely
yit ¼ a þ b0 xit þ eit þ ui
involvement in clubs). Disability was measured with the question:
where E[ui] = 0, Var[ui] = r2u . ‘‘Do you have a protracted illness or disability which restricts your
P. van den Berg et al. / Journal of Transport Geography 19 (2011) 323–331 327

Table 1 trips. This is in line with results from other studies that indicate
Sample characteristics (N = 735). that the oldest age group makes fewer trips for all purposes (e.g.
N % Stern, 1993; Schmöcker et al., 2005) and that the proportion of so-
Age 18–34 168 23 cially active senior citizens decreases rapidly after the age of 75
35–54 322 43 (Aoki et al., 1996). Although the number of social interactions de-
55–64 137 19 creases with age, the average number of trips made for the purpose
65–69 54 7 of the social interactions does not differ significantly. This means
70–74 29 4
75+ 22 3
that with respect to the number of social trips, senior citizens are
as mobile as their younger counterparts.
Gender Male 283 39
Female 447 61
The social activities (that involve a trip) can be divided by dif-
ferent purposes. We distinguish a joint activity, visiting, talking
Household With partner 579 79
Single 150 21
or chatting and other purposes. The distribution of social trips by
these purposes across the different age groups can be seen in Table
Education Low 162 22
Middle 228 31
3. This Table shows that as people get older, they have smaller
High 342 47 shares of social trips for the purpose of just talking or chatting
Income Low (<€1000 per month) 44 7
and more for ‘other’ purposes. The purpose of the social activity
Medium (€1000–3000 per month) 377 57 will be included in models 2 and 3 as an explanatory variable.
High (>€3000 per month) 245 37 With regard to the travel distance for social activities we found
Work status Full time (>35 h per week) 230 31 significant differences between the age groups. However, the aver-
Part time 267 36 age travel distance does not seem to decrease as people get older.
No work 235 32 This can be seen in Table 4.
# Cars in household Mean 1.29 With respect to transport mode choice we distinguish between
Disability Yes 74 10 driving a car, traveling as a car passenger, cycling and walking.
No 642 90 Other forms of transport, such as public transport were hardly used
Involvement in clubs or 0 Clubs or unions 218 30 (despite the fact that in the Netherlands full time students and
unions 1 Club or union 207 28 people of 65 and over pay reduced fares for public transport) and
2 or more clubs or unions 307 42 therefore not considered in our analyses. Table 5 shows that differ-
Urban density Urban (>1500 addresses per km2) 284 39 ent age groups have somewhat different transport mode prefer-
Suburban (1000–1500 addresses 310 42 ences. For all age groups driving is the most important transport
per km2)
Rural (<1000 addresses per km2) 138 19

Table 3
Social trips by purpose.
ability to travel?” Table 1 shows the characteristics of the sample
Age Joint activity Visit Talk/chat Other
on the explanatory variables that are included in our models.
Using these variables, we can examine the effect of age on social 18–34 115 32% 76 21% 56 15% 116 32%
35–54 236 27% 144 17% 108 13% 376 44%
travel behavior. All the categorical explanatory variables are dum-
55–64 100 30% 69 21% 33 10% 134 40%
my coded, with the second category as the reference category. 65–69 31 22% 19 14% 16 11% 75 53%
Although there is some correlation between the explanatory vari- 70–74 18 20% 18 20% 10 11% 46 50%
ables (e.g. older people are less likely to work full time), this is 75+ 12 31% 6 15% 2 5% 19 49%
not too high to cause a problem of multicollinearity. Total 512 28% 332 18% 225 12% 766 42%

4. Estimating social travel demand of senior citizens

This section presents three models to estimate the social travel Table 4
Mean social trip distance per age group.
demand characteristics of senior citizens: (1) the number of social
trips, and for each social trip (2) the travel distance and (3) the Age N Mean
transport mode choice. Before the estimation results are presented, Social trip distance (km) 18–34 357 16.32
we first compare the different age groups with respect to social 35–54 845 9.01
activities and travel. 55–64 332 12.75
65–69 137 11.05
We find that as people get older, they (on average) have signif- 70–74 89 13.74
icantly fewer social interactions. This goes for face-to-face as well 75+ 37 6.68
as ICT-mediated social interactions. This is shown in Table 2. The Total 1797 11.49
oldest age group (75+) is found to have the lowest number of social

Table 2 Table 5
Mean number of social activities in 2 days per age group. Transport mode choice per age group.

Age N ICT interactions Face-to-face interactions Social trips Age Car driver Car passenger Bicycle Walk
18–34 168 6.85 6.11 2.17 18–34 149 44% 53 16% 92 27% 43 13%
35–54 322 5.10 6.54 2.69 35–54 376 45% 80 10% 225 27% 147 18%
55–64 137 4.44 5.25 2.46 55–64 167 51% 41 13% 82 25% 38 12%
65–69 54 4.02 5.09 2.63 65–69 76 55% 17 12% 25 18% 20 15%
70–74 29 3.62 5.17 3.17 70–74 43 48% 3 3% 26 29% 17 19%
75+ 22 3.59 4.14 1.77 75+ 23 62% 3 8% 3 8% 8 22%
Total 732 5.19 5.97 2.51 Total 834 48% 197 11% 453 26% 273 16%
328 P. van den Berg et al. / Journal of Transport Geography 19 (2011) 323–331

mode for social trips. Senior citizens do not seem to become depen- nior citizens increases in the future, social trip-making might
dent on others to drive them to a social activity location; the share increase as well. The results further indicate that involvement
of trips as car passenger is lowest for the oldest cohorts. The share in clubs or unions is positively related to the number of social
of bicycle trips remains relatively stable with increasing age, but trips. This is a plausible finding, as clubs or unions often imply
drops quickly after the age of 75. The share of walking trips seems (social) meetings and activities. Finally, people who work full
to increase a little with age. time are found to make fewer social trips. This may be ex-
plained by the fact that they have less time to socialize with
4.1. Model 1: number of social trips others.
Gender, household structure, income and car ownership are not
In order to examine the effect of the explanatory variables on found to have a significant effect on the number of trips for social
the number of social trips a negative binomial model is estimated. activities. Strikingly, we did not find a negative effect of having a
The estimation results of model 1 are shown in Table 6. disability either, whereas others (e.g. Schmöcker et al., 2005) found
The results indicate that, when controlling for the other charac- that walking difficulties led to fewer trips. We do acknowledge
teristics, the number of social trips is not significantly influenced though that our data collection did not focus specifically on elderly
by age. This confirms our earlier notion that with respect to the and disabled people, which may cause the data to be less suitable.
number of social trips, senior citizens are as mobile as their youn-
ger counterparts.
Only three variables have a significant effect at the 95% con-
Table 7
fidence level. Higher educated people are found to make more
Estimation results for model 2: social travel distance (N = 1610).
social trips and a negative effect is found for lower educated
people. This finding suggests that if the education level of se- B Sig.
Constant 1.386 0.000
Age 18–34 0.289 0.005
Table 6 35–54 (reference)
Estimation results for model 1: number of social trips in 2 days. 55–64 0.347 0.005
65–69 0.438 0.018
B Sig. 70–74 0.306 0.151
Constant 0.988 0.000 75+ 0.238 0.441

Age 18–34 0.108 0.252 Gender Male 0.117 0.205


35–54 (reference) Female
55–64 0.035 0.700 Household With partner 0.061 0.579
65–69 0.030 0.792 Single
70–74 0.249 0.141
Education Low 0.092 0.445
75+ 0.135 0.447
Middle
Gender Male 0.001 0.700 High 0.052 0.564
Female
Income Low (<1000/month) 0.028 0.862
Household With partner 0.000 0.997 Medium
Single High (>3000/month) 0.046 0.608
Education Low 0.142 0.006 Work status Full time (>35 h per week) 0.315 0.003
Middle Part time
High 0.142 0.006 No work 0.012 0.924
Income Low (<1000/month) 0.082 0.241 # Cars in household 0.095 0.130
Medium
Disability Yes 0.012 0.929
High (>3000/month) 0.082 0.241
No
Work status Full time (>35 h per week) 0.168 0.063
Involvement in clubs or 0 Clubs or unions 0.081 0.444
Part time
unions
No work 0.043 0.613
1 Club or union
# Cars in household 0.000 0.803 2 or more clubs or unions 0.014 0.883
Disability Yes 0.000 0.758 Urban density Urban (>1500 addresses per 0.265 0.002
No km2)
Involvement in clubs or 0 Clubs or unions 0.249 0.008 Suburban (1000–1500 addr.
unions per km2)
1 Club or union Rural (<1000 addresses per 0.289 0.007
2 or more clubs or unions 0.140 0.114 km2)

Urban density Urban (>1500 addresses per 0.068 0.402 Purpose Joint activity 0.203 0.001
km2) Visit 0.406 0.000
Suburban (1000–1500 Talk 0.001 1.000
addresses per km2) Other (reference)
Rural (<1000 addresses per km2) 0.139 0.166 Number of trips 0.005 0.740
Dispersion parameter a 0.419 0.000
Number of observations 735 Number of observations (trips) 1610

Log likelihood function 1487.390 Number of groups (individuals) 522

Restricted log 1598.709 Effects: Idiosyncratic Var [e] 0.765 Std. dev. 0.875 Share 0.65
likelihood Individual Var [u] 0.414 Std. dev. 0.644 Share 0.35

Chi squared 222.637 Total sum of squares 1381.4

McFadden Pseudo R- 0.070 Residual sum of squares 1210.2


squared R-squared 0.124
P. van den Berg et al. / Journal of Transport Geography 19 (2011) 323–331 329

4.2. Model 2: social travel distance controlling for the other characteristics. Positive coefficients are
found for the youngest age group and the 55–64 and 65–69 co-
The second model that is estimated, analyses the travel distance horts. This means that these groups are likely to travel longer dis-
for social activities. As the data has repeated measures (multiple tances for social activities. Our findings show that, when
trips per respondent), a random effects model is estimated to ac- controlling for other personal characteristics, the average travel
count for heterogeneity within and between respondents. For a distance does not decrease as people get older.
better-fitting model the dependent variable distance (in kilome- Beside age, three other variables are found to be significant. The
ters) was transformed into the logarithm of the distance +1. All first one indicates that people who work full time are likely to tra-
explanatory variables are entered in the model. The number of so- vel longer distances for social activities. The results further suggest
cial trips and the purpose of the social activity are included in this that people who live in urban areas are likely to travel shorter dis-
model as explanatory variables as well. tances for social activities. This can be explained by the fact that in
The results are shown in Table 7. They show that age has a sig- urban areas more social activity locations are concentrated. At the
nificant effect on the traveled distance for social activities, when same time we find that social travel distances tend to be shorter

Table 8
Estimation results for model 3: travel mode choice.

Passenger Bicycle Walk


B Sig. B Sig. B Sig.
Random parameters
Constant 4.414 0.000 3.044 0.000 0.000 1.000
Age 18–34 0.132 0.840 0.260 0.563 0.004 1.000
35–54
55–64 1.033 0.166 0.259 0.552 0.569 0.504
65–69 0.528 0.519 1.142 0.271 1.283 0.206
70–74 3.627 0.694 0.437 0.638 1.420 0.191
75+ 0.106 0.964 4.503 0.033 0.907 0.844

Non-random parameters
Gender Male 1.552 0.001 0.138 0.687 0.248 0.561
Female
Household With partner 1.539 0.025 0.707 0.109 0.282 0.655
Single
Education Low 1.322 0.030 0.533 0.302 0.536 0.418
Middle
High 0.202 0.648 0.006 0.987 0.489 0.287
Income High 0.338 0.448 0.352 0.376 0.210 0.648
Middle
Low 0.434 0.622 0.474 0.486 0.916 0.263
Cars # 0.739 0.027 1.787 0.000 1.027 0.005
Disability Yes 1.700 0.006 0.384 0.446 1.401 0.113
No
Number of clubs 0 0.315 0.523 0.580 0.187 0.140 0.790
1
2 or more 0.067 0.886 0.267 0.493 0.145 0.771
Urban density Urban 0.332 0.429 0.997 0.008 0.638 0.149
Suburban
Rural 0.826 0.156 1.049 0.026 0.679 0.248
# Social trips 0.020 0.803 0.039 0.562 0.046 0.555
Ln distance 0.647 0.000 1.303 0.000 4.945 0.000
Purpose Joint activity 0.965 0.007 0.549 0.037 0.736 0.043
Visit 0.966 0.012 0.327 0.325 0.094 0.833
Talk/discuss 1.029 0.056 0.640 0.110 0.363 0.464
Other

Standard deviations
Constant 1.359 0.000 2.206 0.000 1.124 0.102
Age 18–34 3.060 0.001 0.461 0.698 0.001 0.143
55–64 2.005 0.008 0.042 0.972 2.326 0.005
65–69 1.857 0.120 1.428 0.483 0.175 0.951
70–74 3.394 0.717 0.767 0.817 0.530 0.900
75+ 0.328 0.981 0.251 0.996 0.052 0.999
Number of observations 1563
Number of groups 592
Log-likelihood (no coefficients) 2166.748
Log-likelihood 1252.261
R-squared 0.422
R-squared adjusted 0.411
330 P. van den Berg et al. / Journal of Transport Geography 19 (2011) 323–331

than average in rural areas. This finding is in line with Föbker and found that in rural areas the bicycle is more preferred. Their results
Grotz’ (2003) findings of extremely high quota of short leisure trips further indicate that senior citizens in higher urbanized areas make
of elderly people in rural areas and it confirms our earlier notion more walking trips.
that in rural areas stronger local networks exist in which close so- Regarding distance the results show that if the travel distance
cial contacts are kept. increases, people are more likely to travel as a car passenger and
Finally, the results indicate that the purpose of the social activ- less likely to cycle, and especially walk. The purpose of the social
ity affects the travel distance as well. Trips for the purpose of vis- activity is also found to significantly affect travel mode choice.
iting tend to be longer than average. The same goes for trips for the For a joint activity people are more likely to choose to travel as a
purpose of a joint activity. passenger, by bike or on foot, relative to driving. For visiting and
The model has an R-squared of 0.124, which means that 12.4% talking or discussing people are also more likely to ride along with
of the variance can be explained by the independent variables. Ta- someone.
ble 7 shows that 35% of the unexplained variation is associated There is evidence of heterogeneity in the transport mode choice.
with individuals. The unobserved preference heterogeneity terms for passenger and
bicycle are highly significant. This indicates substantial variation
4.3. Model 3: travel mode choice across individuals in the overall preference for riding along and
bicycling relative to driving. We also found preference heterogene-
The last model is used to analyze the travel mode choice for so- ity across decision makers within the age cohorts 18–34 and 55–
cial trips. A mixed logit model is estimated because each respon- 64. An R-square of 0.411 indicates a high goodness of fit of the
dent has several choice situations and we expect preference model.
heterogeneity between the respondents. The choice between car
as driver, car as passenger, bicycle and walking is considered.
The first category, driver, is the largest category and serves as the 5. Conclusions and discussion
reference category in the model. Thus, the coefficients estimated
for passenger, bicycle and walk are interpreted relative to choosing As the population of the Netherlands is getting older, more in-
to drive a car for the social trip. sight into the effect of ageing on travel demand is needed. Older
Again, all explanatory variables are included in the model, people’s travel demand is changing: car ownership and reliance
including the purpose of the social activity and the variables that are increasing, as well as trip distance and trip frequency. Social
were predicted in the previous models (the number of social trips travel is especially important for older people, as their proportion
and the trip distance). Table 8 shows the estimation results of the of travel for social purposes is higher than average. Whereas for
mixed logit model (model 3). For each respondent, 100 Halton ran- other activities, especially work activities, older people make fewer
dom draws were used. trips and travel shorter distances than younger people, we have
The alternative specific constants and the age cohorts were shown that for social activities senior citizens are as mobile as their
introduced as random parameters. The random parameters in this younger counterparts.
study are estimated as normally distributed parameters in order to Using a unique data-set based on social interaction diary data
allow parameters to get both negative and positive values. that was collected in 2008 in the Eindhoven region in the Nether-
The constants in the model represent the estimates for passen- lands, three models were estimated to analyze the relationship be-
ger, bicycle and walking relative to car as driver when all explana- tween age and characteristics of trip-making for social activities.
tory variables in the model are evaluated at zero. The constants First, in order to estimate the number of social trips in 2 days a
indicate that, if all explanatory variables are evaluated at zero, peo- negative binomial model was estimated. We found that high edu-
ple are less likely to travel as a car passenger and more likely to cation and involvement in clubs on average result in more social
choose bicycle to go to the location of the social activity, compared trips, whereas full time work is found to result in fewer social trips.
to driving. The latter may seem unlikely, but we should realize that Even though the focus of other studies is rarely on social trips,
this holds after having corrected for travel distance. In other words, these findings seem to be in line with results from other studies.
this result indicates that a base preference for cycling exists that is However, in contrast with other studies, we did not find significant
offset by distance. If the distance increases, the likelihood of choos- effects for gender and having a disability. No significant age effects
ing bicycle decreases. were found, which shows that senior citizens make as many social
With regard to age the results indicate that older seniors (75+) trips as younger age groups.
are less likely to choose the bicycle, relative to driving. No other The second model is a random effects model which was esti-
significant age effects were found. mated to analyze the travel distance for social activities. Longer
Males are found to be less likely to be car passenger than fe- trip distances were found for people who work full time. Shorter
males. This finding is in line with Golob and Hensher (2007) and trips were found for people in urban as well as rural areas. Visits
Schmöcker et al. (2008) who also found that males, compared to and joint activities were found to be associated with longer dis-
females, were more likely to drive and less likely to be passengers. tances. With regard to trip distance the results showed that the
People who live with a partner are found to be more likely to be a average travel distance for social activities does not decrease as
car passenger. This is a plausible finding, as we can assume that people get older. This again underlines our finding that for social
partners often travel together by car to a social activity location. activities senior citizens are as mobile as younger age groups.
Lower educated people are found to be more likely to travel to a The third model is a random parameters logit model to predict
social activity location as a car passenger. Car ownership is found the travel mode choice for social trips. The choice between the four
to result in a larger likelihood of choosing to travel as a car passen- main transport modes (car as driver, car as passenger, bicycle and
ger, relative to driving. The same goes for people with a disability. walking) was considered. The results indicated a base preference
This is a plausible finding that is in line with other studies. for cycling. However, the tendency to cycle decreases rapidly as
With respect to urban density, the results indicate that people the trip distance increases. Significant effects were found for gen-
who live in urban or in rural areas (as opposed to suburban areas) der: males are less likely to be car passenger. Larger likelihoods
are more likely to cycle. Kemperman and Timmermans (2009), of choosing to ride along were found for people who live with a
who studied the relation between environmental characteristics partner and people with a disability. Significant effects were also
and walking and bicycling patterns of the ageing population, also found for education level, car ownership, urban density, distance
P. van den Berg et al. / Journal of Transport Geography 19 (2011) 323–331 331

and the purpose of the social activity. The model had a relatively Golob, T., Hensher, D., 2007. The trip chaining activity of Sydney residents: a cross-
section assessment by age group with a focus on seniors. Journal of Transport
high R-square.
Geography 15, 298–312.
The data-set used in this paper was not collected with a special Greene, W.H., 2002a. Nlogit Version 3.0 Reference Guide. Econometric Software,
focus on senior citizens. As a result, the sample of elderly people Inc., Plainview, NY.
and their social trips was rather small, which restricted the num- Greene, W.H., 2002b. Nlogit Version 3.0 User’s Manual. Econometric Software, Inc.,
Plainview, NY.
ber of explanatory variables in the models. Hildebrand, E., 2003. Dimensions in elderly travel behavior: a simplified activity-
Nevertheless, the results provide new information on the effect based model using lifestyle clusters. Transportation 30, 285–306.
of age on travel behavior for social purposes in the Dutch context. Jorritsma, P., Olde Kalter, M., 2008. Grijs op reis. Over de mobiliteit van ouderen.
Kennisinstituut voor Mobiliteitsbeleid, Den Haag.
Specific to the Dutch context is the preference for cycling. As in Kemperman, A., Arentze, T., Timmermans, H., 2006. Social commitments and
other countries, the car is the preferred transport mode of the activity-travel scheduling decisions. In: Proceedings of the 85th Annual Meeting
Dutch elderly. However, in the Dutch situation, cycling is the sec- of the Transportation Research Board, Washington, DC.
Kemperman, A., Timmermans, H., 2009. Influences of the built environment on
ond most important transport mode, even for the oldest group. walking and cycling of latent segments of the ageing population. In:
Therefore, future transportation policy in the Netherlands should Proceedings of the 88th Annual Meeting of the Transportation Research
give special attention to the safety of senior citizens traveling by Board, Washington, DC.
Kim, S., Ulfarsson, G., 2004. The travel mode choice of the elderly: effects of
bicycle. personal, household, neighborhood and trip characteristics. In: Proceedings of
Moreover, using data with a specific focus on social activities, the 83th Annual Meeting of the Transportation Research Board, Washington,
this study adds to the understanding of the travel demand of the DC.
Mercado, R., Páez, A., 2009. Determinants of distance traveled with a focus on the
elderly for social purposes, which has received very little attention
elderly: a multilevel analysis in the Hamilton CMA, Canada. Journal of Transport
in (international) academic research. The unique data-set allowed Geography 17, 65–76.
us to study the travel (of senior citizens) for different types of so- Molin E., Arentze T., Timmermans H., 2007. Social activities and travel demands: a
cial activities in more detail compared to the regular travel sur- model-based analysis of social-network data. Paper Presented at Frontiers in
Transportation 2007, Amsterdam.
veys. With regard to the travel behavior for social activities, we Newbold, K., Scott, D., Spinney, J., Kanaroglou, P., Páez, A., 2005. Travel behavior
found that senior citizens are as mobile as their younger counter- within Canada’s older population: a cohort analysis. Journal of Transport
parts. The increasing income level of senior citizens and their Geography 13, 340–351.
Páez, A., Scott, D., Potoglou, D., Kanaroglou, P., Newbold, K., 2006. A mixed ordered
increasing car ownership will probably lead to increasing future probit analysis of elderly trip generation in the Hamilton CMA. In: 11th
travel demand for social purposes. International Conference on Travel Behaviour Research, Kyoto.
Rosenbloom, S., 2001. Sustainability and automobility among the elderly: and
international assessment. Transportation 28, 375–408.
References Schmöcker, J., Quddus, M., Noland, R., Bell, M., 2005. Estimating trip generation of
elderly and disabled people: an analysis of London data. In: Proceedings of the
Agneessens, F., Waege, H., Lievens, J., 2006. Diversity in social support by role 84th Annual Meeting of the Transportation Research Board, Washington, DC.
relations: a typology. Social Networks 28 (4), 427–441. Schmöcker, J., Quddus, M., Noland, R., Bell, M., 2008. Mode choice of older and
Alsnih, R., Hensher, D., 2003. The mobility and accessibility expectations of seniors disabled people: a case study of shopping trips in London. Journal of Transport
in an ageing population. Transportation Research Part A 37 (2003), 903–916. Geography 16 (2008), 257–267.
Aoki, R., Ohno, Y., Tamakoshi, A., Kawakami, N., Nagai, M., Hashimoto, S., Ikari, A., Silvis, J., Niemeier, D., 2009. Social network and dwelling characteristics that
Shimizu, H., Sakata, K., Kawamura, T., Wakai, K., Senda, M., 1996. Lifestyle influence ridesharing behavior of seniors. In: Proceedings of the 88th Annual
determinants for social activity levels among the Japanese elderly. Archives of Meeting of the Transportation Research Board, Washington, DC.
Gerontology and Geriatrics 22 (1996), 271–286. Statistics Netherlands, January 2009. <www.cbs.nl>.
Arentze, T.A., Timmermans, H.J.P., Jorritsma, P., Olde Kalter, M.-J., 2008. More gray Stern, S., 1993. A disaggregate discrete choice model of transportation demand by
hair–but for whom? Scenario-based simulations of elderly activity travel elderly and disabled people in rural Virginia. Transportation Research Part A 27
patterns in 2020. Transportation 35, 613–627. (4), 315–327.
Banister, D., Bowling, A., 2004. Quality of life for the elderly: the transport Su, F., Schmöcker, J., Bell, M., 2008. Mode choice of older people before and after
dimension. Transport Policy 11, 105–115. shopping – a study with London data. In: Proceedings of the 87th Annual
Carrasco, J., Miller, E., 2006. Exploring the propensity to perform social activities: a Meeting of the Transportation Research Board, Washington, DC.
social networks approach. Transportation 33, 463–480. Tacken, M., 1998. Mobility of the elderly in time and space in the Netherlands: an
Collia, D., Sharp, J., Giesbrecht, L., 2003. The 2001 national household travel survey: analysis of the Dutch national travel survey. Transportation 25, 379–393.
a look into the travel patterns of older Americans. Journal of Safety Research 34 van den Berg, P., Arentze, T., Timmermans, H., 2008. Social networks, ICT use and
(2003), 461–470. activity-travel patterns. Data collection and first analyses. In: Paper presented
Croissant, Y., Millo, G., 2008. Panel data econometrics in R: the plm package. Journal at the 9th International Conference on Design & Decision Support Systems in
of Statistical Software 27 (2). Architecture and Urban Planning, 7–10 July, 2008, The Netherlands.
Farber, S., Páez, A., 2009. My car, my friends, and me: an exploratory analysis of van den Berg, P., Arentze, T., Timmermans, H., 2009. Size and composition of ego-
automobility and social activities. Journal of Transport Geography 17 (3), 216– centered social networks and their effect on geographical distance and contact
225. frequency. In: Proceedings of the 88th Annual Meeting of the Transportation
Föbker, S., Grotz, R., 2003. Leisure-related mobility of elderly people: ways to Research Board, Washington, DC.
sustainability. Paper Presented at the 43rd Congress of the European Regional
Science Association (ERSA). Jyväskylä, Finland.

You might also like