Transportation

, Volume 39, Issue 6, pp 1079–1103

Mobility behavior of the elderly: an attitude-based segmentation approach for a heterogeneous target group

Authors

    • ILS—Research Institute for Regional and Urban Development
Article

DOI: 10.1007/s11116-011-9380-7

Cite this article as:
Haustein, S. Transportation (2012) 39: 1079. doi:10.1007/s11116-011-9380-7

Abstract

The western population is ageing. Based on the assumption that the elderly are a quite heterogeneous population group with an increasing impact on the transport system, mobility types of the elderly were identified. By means of 1,500 standardized telephone interviews, mobility behavior and possible determinantes including infrastructural, sociodemographic and attitudinal variables, were assessed. The most important factors, identified by five regression analyses, served as type-constituent variables in a series of cluster analyses. The final cluster solution resulted in four segments of the elderly named Captive Car Users, Affluent Mobiles, Self-Determined Mobiles, and Captive Public Transport Users. The groups showed distinct mobility patterns as well as significant differences in infrastructural, sociodemographic and attitudinal variables. The study provides a more comprehensive understanding of the diverse lifestyles, attitudes, travel behavior and needs of the elderly. Furthermore, it identifies starting points for the reduction of car use.

Keywords

Elderly travelMobility behaviorSegmentationTarget-groupAttitudesAgeing population

Introduction

The western population is ageing. While at present 20% of the European population is 60 years or older, by 2030 almost every third European and about every fourth North American and Australian will belong to this age-group (United Nations 2007). This development will affect almost every aspect of our lives, last but not least the transport sector. Older people are recognized as a heterogeneous population group that will have a much greater influence on the transport system in the future than it has today.

Against this background, a mobility-related segmentation approach for the elderly that allows for the identification of target-group specific measures is presented in this paper. Based on a short overview of seniors’ mobility behavior today and the expectations for the future, it is argued in the section below why such an approach is needed. As the segmentation relies on the most important determinants of elderly mobility behavior, the subsequent section summarizes which variables turned out to have a significant impact on elderly mobility behavior in previous research. Finally, existing segmentation approaches of elderly people in the mobility sector are presented, as well as the concept of the present study.

Why do we need a segmentation approach for the elderly?

There are developments in the group of the elderly that strongly affect the transport system. While the elderly nowadays have a higher share in walking and public transport use than younger adults (e.g. BMVBW 2003; Cao et al. 2007) in future more elderly people, especially women, will hold licenses and have access to a private car. This will most probably result in a higher share of elderly drivers and a decrease in captive public transport users.

Although there is a negative association between increasing age and trip-making frequency today (e.g. Páez et al. 2007), due to lifestyle changes and better health older people are also becoming more mobile (Rosenbloom 2001), especially in the social/leisure category (e.g. Arentze et al. 2008; Hjorthol et al. 2010) and with regard to car trips (e.g. Newbold et al. 2005; OECD 2001).

The expected growth rate of elderly drivers poses both safety-related and environmental consequences. Road safety issues associated with elderly people is already a well established research topic. Programs designed to enhance driving skills and to raise awareness of the special challenges elderly people face while driving are being developed, as well as cars that better address older people’s needs. However, little effort is taken to address environmental issues and to reduce senior citizen’s car use.

Measures are required that on the one hand improve the mobility options of older people and on the other hand offer more environmentally friendly alternatives to the private car. Besides a promotion of walking and cycling, including improvements of the respective infrastructure, public transport has to be enhanced to meet the needs and wishes of the elderly. The respective requirements of the elderly are well researched in several national and European studies and a multitude of recommendations for policy and practice have already been derived (e.g. Engeln and Schlag 2001; Haindl and Risser 2006; Kaiser and Kraus 2007; Rudinger et al. 2004; Mollenkopf and Flaschenträger 2001; Holz-Rau et al. 2004). The level of implementation of necessary improvements differs within Europe and the whole western world. However, in contrast to ‘hard measures’ like barrier-free accessibility, user-friendly ticket machines, speed limits, etc., soft policy measures (with the exception of concessionary fares) are often overlooked. Better services, however, will not lead to higher use if elderly people are not aware they exist.

A necessary condition to successfully and efficiently implement hard measures as well as information campaigns or new mobility services is that measures are developed that address specific target groups rather than all potential user groups at once. So far, most measures targeted to the elderly address mainly the sub-group of mobility-impaired people. It is often overlooked that higher age does not necessarily mean restricted mobility. According to quantitative survey results from eight European countries, most of the elderly in Europe are highly mobile and rarely report any mobility constraints: 85% of people aged 65–74 and 70% of people aged 75 and above feel negligible limitations in terms of their out-of-home mobility (SIZE, Kaiser and Kraus 2007). The elderly of today have different lifestyles than past generations and also higher expectations (Kirchmair 2005). Only a small percentage can be described as solitary and mobility-impaired. The more active subgroups of seniors have requirements and needs that go beyond barrier-free access and should also be considered when aiming at a decrease in car use.

Determinants of elderly mobility behavior

While research on elderly mobility behavior has formerly often been restricted to the aspect of mobility impairment, the research focus has recently widened such that other factors influencing mobility behavior of the elderly have also been taken into account. Results of the study MOBILATE (Mollenkopf et al. 2005) for example found that people with greater mobility are younger and typically live in urban areas. In accordance with other studies (e.g. Jansen et al. 2001; Scheiner 2004; Smith and Sylvestre 2001), a significant impact of physical health and the size of the social network on elderly mobility was shown. Additionally, an important factor is the accessibility to either public or private vehicular transportation (e.g. Páez et al. 2007; Scheiner 2004; Smith and Sylvestre 2001) as well as neighborhood accessibility in general, which may reduce the level of car use (Cao et al. 2007).

Mobility-related attitudinal factors

Mobility-related attitudes are known to be correlated with mobility behavior. However, the direction of influence is debatable. Within the framework of the Theory of Cognitive Dissonance (Festinger 1957) it is argued that after a choice, stated attitudes will be aligned to this choice. Evidence that the relationship from mobility behavior to attitudes may be stronger than the reverse can be found in Dobson et al. (1978) or Golob (2001). However, most studies that consider attitude-behavior-relationships in the transport sector refer to the Theory of Planned Behavior (TPB), Ajzen (1991). According to the TPB, the most important determinant of a person’s behavior is the intention to perform a behavior. Behavioral intention is defined as a combination of attitude, subjective norm (SN), and perceived behavioral control (PBC). Attitude toward a behavior is the degree to which the performance of the behavior is positively or negatively valued, SN is defined as perceived social pressure to engage or not to engage in a behavior, and PBC refers to people’s perceptions of their ability to perform a behavior. PBC is assumed to be a direct predictor of both intention and behavior. The TPB has been applied successfully to predict the use of public transportation and car use (e.g., Bamberg and Schmidt 2001, 2003; Haustein and Hunecke 2007; Heath and Gifford 2002). In some transport studies TPB has been extended by the construct personal norm (PN) derived from the Norm-Activation Model (Schwartz 1977) (e.g. Bamberg et al. 2007; Harland et al. 1999). PN is defined as the intrinsic feeling of moral obligation to behave in accordance with the person’s individual value system.

When applying the TPB in the context of travel mode choice, attitude is often measured as a simple shaping of ‘good’ or ‘bad’, or ‘pleasant’ or ‘unpleasant’. Other, more specific mobility-related attitude dimensions result from symbolic-affective evaluations of travel modes (e.g. Anable 2005; Anable and Gatersleben 2005; Hunecke 2000, Hunecke and Haustein 2007; Steg 2005; Steg et al. 2001). Steg et al. (2001) demonstrated that symbolic-affective functions like excitement and prestige as well as instrumental-reasoned functions like financial costs and driving conditions are important dimensions underlying the attractiveness of car use. Hunecke (2000) differentiated four basic symbolic dimensions of mobility: autonomy, excitement, status, and privacy. On the basis of these dimensions, behavior-relevant attitudes concerning different travel modes can be operationalized (Haustein and Hunecke 2007).

However, only a few studies dealing with mobility behavior of elderly people considered attitudes as determinants of mobility behavior. In a study by Haustein et al. (2008a) mobility-related attitudes were included and played a significant role in the prediction of travel mode choice of the elderly. An important factor turned out to be the perceived ability to use public transportation (public transport control), which influenced both car use (negatively) and public transportation use (positively). Similarly, Cao et al. (2007) found the related construct of car dependency to have a significant impact on vehicle miles driven. In both studies travel attitudes were of higher relevance for younger adults compared to the elderly. However, both studies examined the general population, therefore, some potentially influential attitudinal factors specific to the elderly were not measured.

Perceived danger

Perceived danger is often considered as a relevant factor that restricts mobility behavior of older people. Results of several studies indicate that older people perceive a greater danger in public space than younger people and attach greater importance to safety from crime (e.g. Flade 2002; Scheiner and Holz-Rau 2002). Perceived danger has a negative impact on the experience of public transport and security is regarded as a relevant aspect for the attractiveness of mobility services by the elderly (Engeln and Schlag 2001; Megel 2002). However, an influence of perceived danger on the number of leisure time activities of the elderly could not be shown. There is only a weak correlation between travel mode choice and perceived danger, which may indicate that trips in situations being deemed dangerous (e.g. in darkness) are either shifted to some other point of time or made when accompanied (Haustein and Kemming 2008).

Existing segmentation approaches and the present study

Studies that examine mobility behavior, needs and limitations of the elderly often conclude that they deal with a quite heterogeneous group (e.g. Alsnih and Hensher 2003; BASt and DVR 2000; OECD 2001; Siren and Hakamies-Blomqvist 2004). Yet recommendations, strategies or policies are typically devised for the entire group of elderly people. In order to effectively develop and apply new mobility services and improve the existing system, a differentiation of relevant subgroups of older people, which may serve as target groups, is required.

Segmentation approaches in the transport sector can be differentiated according to the factors on which the segmentation is based, i.e., mobility behavior, sociodemographic variables, and psychographic factors. Psychographic approaches are mainly based on attitudes and values. Here one can distinguish between general lifestyles approaches (e.g. Sinus Sociovision 2006) and the more specific mobility styles (e.g. Götz et al. 1998; Lanzendorf 2001) or mobility types (e.g. Hunecke et al. 2010). For each of these segmentation approaches there is also at least one example for the elderly population.

A segmentation approach on the basis of mobility behavior was applied by Rudinger and Käser (2007). Based on activities’ variety and frequency they differentiated four groups of older people, which primarily fulfill a descriptive function.

Based on sociodemographic variables (e.g., age, gender, driving license) Hildebrand (2003) identified six distinct clusters, “Workers”, “Mobile Widows”, “Granny Flats”, “Mobility-Impaired”, “Affluent Males”, and “Disabled Drivers”. The segments were found to have significant differences in mobility behavior and activity engagement patterns.

Jansen et al. (2001) classified seniors by lifestyle variables (leisure time activities, preferred TV programs, furnishing). The five resulting lifestyle clusters differed in mobility-related personality traits and some aspects of mobility behavior.

Finally, Haustein et al. (2008b) used mobility specific attitudes as well as car accessibility and age to create six distinct segments of older people, which showed strong differences in mobility behavior, i.e., modal split, distances travelled as well as frequency of activities.

The different types of segmentation approaches have specific pros and cons. Which one is most appropriate, highly depends on the field of application they are used for (Hunecke et al. 2010).

An advantage of the segmentation approach that includes mobility-related attitudes as constituent variables is that it shows greater differences in mobility behavior than segmentations based on sociodemographic variables (Hunecke et al. 2010) or life-style variables (Hunecke and Schweer 2006). Moreover, it provides better starting points for interventions compared to behavior-based and sociodemographic segmentations. However, the attitude-based approach by Haustein et al. (2008b) has some limitations. First, it was based on a survey involving the whole population rather than targeting the elderly population specifically. Thus, potential factors that are especially relevant for mobility of the elderly were not included or not measured as differentiatedly as desirable. Moreover, the scope of activities considered was tailored to the whole group and thus disregarded such activities that are more relevant for the elderly (e.g., health care, church service). Finally, the sample was restricted to inhabitants of three large cities.

These limitations shall be overcome in the segmentation approach presented in this paper.

Crucial to a segmentation approach for older citizens are in my view aspects of personal living conditions, such as the level of impairment, financial resources, and social networks. Together with infrastructural conditions these aspects determine the individual’s objective scope of action on the one hand. On the other hand, attitudes and perceived constraints, which underlie individual mobility behavior, determine the subjective scope of action (Tanner 1998, 1999). Based on the most relevant subjective and objective factors, subgroups of the elderly will be identified in this study on the basis of an extended version of the TPB, Ajzen (1991).

Method

Procedure and participants

Data for this study was collected in April and May 2009 by means of standardized computer-assisted telephone interviews (CATI) that lasted 22 min on average. A random sample was drawn based on a procedure that also included non-registered telephone numbers (for details see Rösch 2004). The target population was individuals aged 60 years and above living in the German federal state of North Rhine-Westphalia (NRW). In terms of population, NRW is the largest Federal State of Germany. It has over 18 million inhabitants and covers a land area of 34,080 km². The selection of this region was mainly based on two reasons. First, in Germany public transport is administrated by federal states, and thus it was decided to concentrate on one state so that the respondents’ basic conditions would be comparable. Second, NRW comprises a high variety of urban areas ranging from larger cities with more than 500,000 inhabitants to rural districts in generally urbanized regions. At the same time it excludes pure rural areas which are not in the focus of this study as a minimum of public transport availability was required for answering several questions related to public transportation.

Altogether, 1,500 interviews were conducted. The overall response rate was 10% (63% refusals, 26% not reached, 1% not completed). The low response rate can be explained with the interview duration, which was quite long for a telephone interview. Nevertheless, the sample was representative for the elderly population of NRW with regard to sociodemographic data, such as gender, and structure of urban development of the municipalities the individuals belonged to. The sample consisted of 893 women (59.5%) and 607 men (40.5%) with a mean age of 71. Education level was above average (26.9% with higher education) due to a higher willingness to participate of well-educated people, which is usually found in scientific surveys.

Measures

In the following section the content of the standardized interviews is described.

Infrastructural variables: spatial characteristics and accessibility to transport systems

Spatial aspects were operationalized through objective spatial indicators based on the municipality where the people lived, such as the level of centrality derived from the ‘accessible daytime population’ (BBR 2009). In addition individuals were asked if they lived either in/near the city center or at the city border/in a suburb.

Accessibility to transport systems was measured by individual ratings of access to public transport (closest bus stop, tram and rail station within walking distance or not) and car accessibility (five-point frequency scale). Additionally, the participants were asked about the number of cars per household, possession of a driving license, a season ticket for public transport, free access to public transport and possession of a ‘‘Bahncard’’—a discount card for frequent users of German Rail. Moreover, it was asked which of the following facilities could be reached by foot: grocery, bakery, pharmacy, general practitioner, and post office. The number of accessible facilities was used as an indicator of general accessibility. Finally, the overall satisfaction with personal mobility options (“getting everywhere I want to get”) was rated on a five-point rating scale.

Sociodemographics, social network, and health

Sociodemographic data including gender, age, employment-status, educational background, household income, and household size were recorded, and participants were asked if they owned a mobile phone and had access to the internet.

Participation in social networks was assessed with two questions: “How many relatives do you see or speak to at least once a month?” and “How many friends do you see or speak to at least once a month?” The mean of both variables was used as an indicator for the size of individuals’ social networks.

Respondents were further asked if they owned a dog.1

Participants’ health status was measured by asking if they were limited in their mobility on account of a disability and if they used a walker or wheelchair. Additionally, they were asked to assess their health status and level of individual mobility (ability to move, walk around etc.) on a five-point rating scale ranging from “very good” (1) to “very poor” (5).

Perceived danger

Two aspects related to perceived danger were included for each of the following transport modes: foot, bicycle, public transport, and car. Firstly, it was asked how safe individuals felt with regard to the danger of being involved in a traffic accident. Secondly, they were asked how safe they felt alone in the dark with regard to harassments and attacks.

The responses were provided on a five-point scale from “not safe at all” (1) to “totally safe” (5).

Mobility-related attitudinal variables

Attitude, social norm, and PBC were derived from the TPB (Ajzen 1991) and PN from the Norm-Activation Model (Schwartz 1977). Attitude was measured as an evaluation of the symbolic dimensions of different modes of transportation (Hunecke et al. 2007). In addition some newly developed health-related items concerning non-motorized modes as well as perceived car stress were measured. They were supplemented by perceived mobility necessities (Haustein and Hunecke 2007).

All responses were provided on a five-point agreement-scale from “not agree at all” (1) to “agree totally” (5). The statements measuring the psychological constructs were presented in random order. All psychological items are listed in Appendix Table 4.

A principal component analysis with varimax rotation was carried out in order to reduce the number of psychological variables to their underlying dimensions. Retaining only factors with eigenvalues greater than one, an easily interpretable ten factors solution was obtained, which explained 65.1% of the variance.

In Appendix Table 5 the loadings of the single items on the ten factors are presented. Mean scales were constructed following the resulting factor solution with one exception: Although, items referring to excitement of public transport and social norm loaded on the same factor, two separate scales were constructed for two reasons: First, these two dimensions have a high relevance for the design and promotion of public transport services. Secondly, they loaded on separate factors in preceding studies (e.g. Hunecke et al. 2007, 2010).

Table 1 displays means, standard deviations, and internal consistencies (Cronbach’s alpha) for the calculated multi-item scales, which are briefly described below.
Table 1

Description of psychological scales

Scale

Construct (number of items)

n

M

SD

Cronbach’s α

Pt control

Pt autonomy (1)

1481

3.22

1.27

.70

Car autonomy (1)

Perceived behavioral control (1)

Pt excitement

Pt excitement (4)

1411

2.41

1.11

.72

Pt privacy

Pt privacy (2)

1365

2.02

1.21

.66

Car attitude

Car excitement (3)

1442

3.14

1.02

.71

Car privacy (1)

Car autonomy (1)

Car stress

Car stress (2)

1159

1.67

1.11

.72

Cycling attitude

Cycling excitement (2)

1034

3.47

1.38

.84

Cycling autonomy (2)

Weather resistance

Weather resistance (2)

1009

2.39

1.35

.56

Walking attitude

Walking excitement (3)

1488

3.88

1.16

.85

Walking/cycling health (2)

Social norm

Social norm (2)

1354

2.04

1.26

.68

Personal norm

Personal norm (2)

1413

3.16

1.34

.64

Perceived mobility necessities

Perceived mobility necessities (2)

1486

3.20

1.37

.64

Pt public transportation

Public transport control measures how easy or difficult individuals perceive the use of public transport and whether they feel restricted or not in their autonomy when using public transport (instead of the private car).

Public transport excitement assesses in how far positive aspects are associated with public transport use, such as relaxation or social communication.

The privacy dimension measures if the use of public transport is perceived as a restriction of privacy, e.g., because other passengers come too close.

Car attitude summarizes aspects concerning excitement, privacy and autonomy with regard to driving a car. As in previous studies, for the car the different dimensions cannot be separated but load on the same factors and are thus integrated in one scale—meaning that the car is either evaluated positively or negatively, without differentiation between sub-dimensions. However, the newly constructed items referring to car stress loaded on a separate factor. This variable describes in how far driving a car is perceived as increasingly difficult or exhausting.

Cycling attitude comprises cycling autonomy and excitement.

Weather resistance measures the willingness to use a bicycle in all weather conditions. In other studies this variable turned out to be an important positive predictor of cycling as well as a significant negative predictor of car use (Hunecke et al. 2007, 2010). It can be used to divide people into two groups: Those who use the bike only for leisure purposes and those who use the bike as an everyday mode of transport (Haustein et al. 2007).

Walking attitude comprises two aspects: general walking excitement and health-related motives for walking.

Social norm is defined as the perceived social pressure or support to engage or not to engage in a behavior (here: public transport use). In contrast, personal norm measures the intrinsic moral obligation to behave in a morally correct way, i.e., to use environmentally friendly means of transport.

Finally, perceived mobility necessities are defined as people’s perceptions of mobility-related consequences of their personal living circumstances, e.g., if people perceive that they have to be mobile all the time to meet their obligations.

Except for weather resistance all scales have acceptable internal consistencies. As Cronbach’s alpha for weather resistance was much higher in former studies, the scale is used despite its rather bad performance in this study.

Mobility behavior

The mobility behavior was measured by participants’ specifications about their daily activities. Participants were asked to state, how often they performed each of 16 activities (if ever). Appendix Table 6 provides an overview on the activities, which were allocated to three categories: leisure time activities, work (including unpaid work) as well as shopping and private errands.

For one selected trip per category additional characteristics were requested, i.e., for the most frequent leisure time trip, the trip to work and for shopping trips. If no work trips/shopping trips were performed, the next activity of the allocated category was chosen. For the three selected trips it was asked how often the trip was performed by car (driver or passenger), public transport, by foot, by bicycle or by motorcycle. On this basis the modal split (percentage of trips by travel mode) was calculated for each category separately as well as the mean modal split over all three trips.

Besides the activity-based request, it was also assessed how often people generally used public transport and the bicycle (days per week/month). For the car the distance travelled per year as a driver was recorded.

Results

Identification of mobility types of the elderly

We identified target groups or “mobility types” in two successive steps: First, we analyzed the personal determinants of mobility behavior in order to identify the factors with the highest predictive power. On the basis of these factors, which were used as constituent variables, we identified target groups by means of cluster analysis.

Multivariate analyses of mobility behavior

Five regression analyses were conducted in order to identify the most important personal determinants of mobility behavior. Three of the regression analyses refer to the used travel mode, predicting the use of private motorized modes (modal split), public transportation (categorized frequency of use per week/month), and bicycle (categorized frequency of use per week/month), respectively. In addition, two regression analyses predicting the frequencies of leisure time and other activities were calculated. Subject to the dependent variable, either linear or ordinal regression analyses were conducted. As independent variables infrastructural, sociodemographic, and attitudinal variables were entered into the regression models. Table 2 summarizes the results of all regression analyses.
Table 2

Review of the results of regression analyses predicting different mobility behavior variables

Independent variables

Share of private motorized modes

Public transport use

Bicycle use

Leisure time activities/year

Other activities/year

 

β

Estimate

Estimate

β

β

Infrastructure/accessibility

     

 Number of cars/household

.05

−.40

.23

.01

−.01

 Car availability [from “never” (1) to “always” (5)]+

.27***

−.38**

.15

.10*

.04

 Season ticket

−.08**

 

.48

−.02

−.01

 Free access to public transportation

−.05

−1.57*

.85

−.03

.01

 Bahncard

−.01

−.52

.38

.00

.00

 Rail station within walking distance

.01

.58**

.23

.07

−.02

 Tram station within walking distance

.01

−.71**

.26

−.09

−.02

 Centrality (accessible population at daytime)

.00

.00***

.00

.01

.00

 City centre

−.09**

.31

.26

.03

.06

 Number of facilities within walking distance+

−.11***

−.13

.07

−.01

.12**

Socio-demographics

  

 

 

 

 Age

−.03

.03

.02

−.04

−.10*

 Gender [1 = male, 2 = female]

.02

−.16

.24

.03

.02

 Education level

.01

−.17

.25

.00

.04

 Employed

.00

−.23

.33

−.04

.29***

 Income+

.08*

.16

.07

.04

.13**

 Number of persons/household

.02

−.19

.16

−.12**

−.01

 Living apart together relationship

.01

.19

.40

.04

.01

Social network/health

     

 Size of social network+

 

 

 

.13***

.07*

 Dog owner

.02

.36

.34

.09**

−.02

 Mobility disabled

.03

−.29

.43

.03

.03

 Walking frame/wheelchair

−.03

.67

.88

.02

−.04

 Motivity [from “very good” (1) to “very poor” (5)]

.06

−.18

.18

.00

.00

 Health status [from “very good” (1) to “very poor” (5)]

−.07

.13

.19

.00

.05

Perceived danger [from “not save at all” (1) to “totally save” (5)]

 

  

 

 

 Perceived danger (accident) car

.00

  

 

 

 Perceived danger (accident) pt

 

.20

 

 

 

 Perceived danger (accident) bicycle

 

 

.11**

 

 

 Perceived danger (harassments) car

−.03

  

 

 

 Perceived danger (harassments) pt

 

.17

 

 

 

 Perceived danger (harassments) bicycle

 

 

.10

 

 

Mobility-related attitudes [from “not agree at all” (1) to “agree totally” (5)]

 

  

 

 

 Pt control+

−.14***

.28**

.12

.06

.01

 Pt excitement

−.01

.41***

.12

.01

.03

 Pt privacy

−.02

.00

.09

.06

−.01

 Car attitude

.10**

−.06

.13

.00

.03

 Car stress

.05

−.07

.13

.00

.02

 Cycling attitude+

−.06

−.09

.14***

−.09*

−.03

 Weather resistance+

−.15***

−.07

.09***

.03

.03

 Walking attitude+

−.16***

.14

.15

.35***

.01

 Social norm

−.04

.13

.11

.07

.03

 Personal norm

−.05

.05

.10*

.00

−.02

 Perceived mobility necessities+

.08***

−.13

.09

.00

.18***

 R2

.453

  

.187

.243

Adjusted R2

.425

  

.147

.206

Nagelkerke’s R2

 

.387

.480

  

McFadden’s R2

 

.166

.229

  

For the dependent variables “private modes”, “leisure time activities/year” and “other activities/year” linear regression analyses were conducted; for “public transport use” and “bicycle use” ordinal regression analyses (link function: logit) were conducted. Predictors were checked for multicollinearity: variance inflation factors (VIF) of all variables were <3. Pairwise deletion of missing data was used. Missing coefficients indicate that variables were not included in the respective analyses

Because of the encoding, the sign of a coefficient corresponding to a categorial predictor has a different meaning in the linear and ordinal regressions. In the linear regressions a positive coefficient indicates a positive relationship between predictor and independent variable, whereas in the ordinal regressions a positive coefficient indicates a negative relationship. In the case of a continuous variable, a positive coefficient always represents a positive relationship

+Selected for cluster analyses

*p < .05, **p < .01, ***p < .001

Unsurprisingly, the percentage of car use is highest correlated with car availability. The more facilities are within walking distance, the fewer trips are conducted by car. In addition, several attitudinal variables, especially walking attitude, weather resistance, and public transport control were identified as significant negative predictors of car use behavior. The most important predictors of public transport use are centrality of the municipality as well as public transport excitement. Bicycle use is highest correlated with cycling attitude and weather resistance (p < 001). Additionally, perceived danger bears some importance (p < .01). With regard to the number of leisure time activities it is interesting that the most relevant variable is walking attitude, i.e., individuals who like walking are generally more active in their leisure time than individuals to whom walking appears exhausting. Here the extent of individuals’ social networks also has some influence as well as the number of persons per household: the smaller the household the more dependent individuals are on social contacts and activities outside their homes. Finally, as predictors of the number of other activities, variables related to work and other duties are of great importance, i.e., employment-status, income, and perceived mobility necessities. This is not surprising as the variable “other activities” includes trips to work (paid and unpaid).

Across all analyses the most important variables, having a significant impact (p < .05) on at least two mobility behavior variables, turned out to be: car availability, number of facilities within walking distance, income, size of social network, public transport control, cycling attitude, weather resistance, walking attitude, and perceived mobility necessities (see Table 2). They were thus selected as constituent variables for the cluster analysis.

Cluster analysis

Cluster analyses in general do not offer a test to calculate the optimal number of clusters. Thus, cluster analyses using the k-means algorithm were conducted for two to eight cluster solutions. The solutions were compared according to the criteria of predictive power, stability, and interpretability. Appendix Table 7 presents the results of ANOVAs for the different cluster solutions with the five dependent behavioral variables. The F-values as well as the η2-values do not provide an explicitly best solution. As an additional criterion the stability of the cluster solutions was considered. For this purpose, the sample was randomly split into halves of the same size and cluster analyses restricted to each subsample were conducted. The results of the subsamples were compared with each other and with the corresponding results of the whole sample in order to find the most stable solution. All considered cluster solutions turned out to be stable, i.e. the clusters of the whole sample could also be identified in the sub-samples. As a last and most important step, the interpretability of the different solutions was compared, which finally lead to the decision in favor of the four cluster solution.

Description of mobility types of the elderly

Profiles of mobility types of the elderly

The four resulting mobility types of the elderly were labeled “Captive Car Users”, “Affluent Mobiles”, “Self-Determined Mobiles” and “Captive Public Transport Users”. Figure 1 displays the cluster centers of the four types which result in distinct profiles (z-standardized scores).
https://static-content.springer.com/image/art%3A10.1007%2Fs11116-011-9380-7/MediaObjects/11116_2011_9380_Fig1_HTML.gif
Fig. 1

Cluster profiles

Captive Car Users have good (but not the best) access to a car. The number of facilities they can reach by foot is below average. In addition, they perceive low public transport control and neither like walking nor cycling. Affluent Mobiles have a high car availability and by far the highest income and the largest social network. They also perceive low public transport control but evaluate all other modes rather positively. They can further be characterized by high perceived mobility necessities. In contrast, Self-Determined Mobiles do feel no pressure to be mobile all the time. They have good access to both car and public transport, can reach many facilities by foot and show the most positive attitudes with regard to cycling and walking. Finally, Captive Public Transport Users have by far the lowest car access but perceive high public transport control. Their attitude towards the bike is negative, whereas walking is evaluated more or less on average. Finally, they have the lowest incomes.

Differences of mobility types in mobility behavior

Differences in attitudes, car access, facilities within walking distance, income and social network are reflected in the mobility behavior of the four mobility types. Figure 2 present the percentage of used transport modes. Captive Car Users show the highest percentage of trips by private motorized modes (78%), followed by Affluent Mobiles (66%). In contrast, the smallest share of trips by private motorized modes is conducted by Captive Public Transport Users, who have the highest share of public transport use (19%). Public transport use of all other types is below 5%. The bicycle is used most often by Self-Determined Mobiles (22%), followed by Affluent Mobiles (14%), whereas for Captive Car and Captive Public Transport Users the bike is of no relevance as mode of transport. Finally, Captive Public Transport Users do half of the considered trips by foot. For Captive Car Users and Affluent Mobiles walking is of much lower importance.
https://static-content.springer.com/image/art%3A10.1007%2Fs11116-011-9380-7/MediaObjects/11116_2011_9380_Fig2_HTML.gif
Fig. 2

Modal split

With regard to distances travelled by car (Fig. 3) we also find significant differences between the four types: While three quarter of Captive Public Transport Users do not drive a car at all, more than half of the Affluent Mobiles drive 10,000–15,000 km or more per year. The distances travelled of the other types lie in-between.
https://static-content.springer.com/image/art%3A10.1007%2Fs11116-011-9380-7/MediaObjects/11116_2011_9380_Fig3_HTML.gif
Fig. 3

Distances travelled per year by car as driver

Considering how many activities are conducted per year (see Fig. 4), Self-Determined Mobiles and Affluent Mobiles are most active with regard to leisure time activities, whereas Captive Car Users are least active. Most trips for (paid or unpaid) work are conducted by Affluent Mobiles, and least by Captive Public Transport Users. With regard to shopping trips and trips for private errands the differences are not that pronounced. The only significant difference is between Captive Car Users and Affluent Mobiles, who do the most shopping trips (p < .001).
https://static-content.springer.com/image/art%3A10.1007%2Fs11116-011-9380-7/MediaObjects/11116_2011_9380_Fig4_HTML.gif
Fig. 4

Number of activities per year

Further characteristics of the four mobility types

In the following section the four mobility types are described by sociodemographic and infrastructural data to get a broader picture of their living circumstances.

Table 3 lists descriptive results for sociodemographic, infrastructural and health variables, whereby it becomes apparent that mobility types show also great differences with regard to variables that were not included in the cluster analyses. Compared to the other groups, Captive Car Users as well as Captive Public Transport Users are older, more often disabled, and live more often alone. In this respect both groups can be considered to be disadvantaged compared to Affluent Mobiles and Self-Determined Mobiles. Besides these similarities, they strongly differ with regard to car and public transport access. Captive Public Transport Users on the one hand have the best access to public transport, on the other hand they depend on it as they often have no access to a private car (mostly women, who do not possess a driver’s license), while Captive Car Users depend more strongly on the car, as they live less central or find it more difficult to use public transport for other reasons.
Table 3

Description of the four elderly mobility types

 

Captive car users (23.9%)

Affluent mobiles (22.9%)

Self-determined mobiles (30.4%)

Captive public transport users (22.9%)

Total

Socio-demographics

     

 Percent female***

57.8%

45.5%

53.9%

82.8%

57.8%

 Average age***

73.1

68.8

69.7

74.5

71.4

 Percent high school level completed***

22.5%

47.8%

23.0%

17.2%

27.3%

 Percent income < 1000 EUR***

21.3%

2.5%

12.8%

33.9%

17.0%

 Percent income > 3000 EUR***

4.3%

51.3%

5.0%

.4%

14.6%

 Percent employed***

7.7%

16.7%

9.1%

2.1%

8.8%

 Average household size***

1.7

2.0

1.8

1.5

1.8

 Percent single-person household***

42.6%

16.7%

32.7%

60.0%

37.6%

 Percent couple household***

51.4%

78.4%

63.1%

34.1%

57.2%

Infrastructure/accessibility

     

 Percent driving license***

90.5%

97.1%

87.1%

48.7%

81.4%

 Percent car availability “never” ***

13.4%

4.1%

18.6%

74.1%

26.7%

 Percent car availability “often/always” ***

84.9%

94.8%

78.3%

22.2%

70.8%

 Percent rail station within walking distance***

31.1%

40.4%

54.3%

50.9%

44.8%

 Percent tram station within walking distance***

21.5%

24.5%

28.7%

41.2%

28.9%

 Percent season ticket***

3.4%

4.7%

9.0%

20.7%

9.3%

 Percent free access to public transportation***

5.6%

1.5%

1.5%

14.0%

5.3%

 Percent living in/near city centre vs. suburb***

19.5%

25.7%

34.9%

36.3%

29.5%

 Average number of facilities within walking distance***

2.2

3.8

4.4

4.1

3.7

 Satisfaction with mobility options [from “not satisfied at all” (1) to “totally satisfied” (5)]***

4.1

4.5

4.6

4.3

4.4

Social network/health

     

  Average size of social network***

4.8

8.6

5.1

5.4

5.9

 Percent mobility disabled***

41.5%

7.9%

9.2%

36.8%

22.9%

 Motivity*** [from “very good” (1) to “very poor” (5)]

2.7

1.8

1.9

2.5

2.2

 Health status*** [from “very good” (1) to “very poor” (5)]

2.6

2.1

2.0

2.6

2.3

Access mobile phone/internet

     

 Percent possession of mobile phone***

70.9%

91.0%

79.9%

57.9%

75.3%

 Percent internet access***

40.1%

74.0%

48.8%

24.0%

46.8%

***p < .001 [χ2-test (frequencies) or ANOVAs (means)]

In contrast, the two other segments are more self-determined and better off. They most often live together with their partner, are mostly not mobility-impaired and in good shape. Affluent Mobiles are well-educated, and 17% are still working. Both types are more satisfied with their mobility options than the two Captive Types and differ significantly from Captive Car Users, who exhibit the lowest satisfaction (p < .001, Scheffé).

With regard to information and communication technology, the greatest gap is between Affluent Mobiles of which 91% have a mobile phone and 74% access to the internet, compared to Captive Public Transport Users, for which it is 58 and 24%, respectively.

Finally we have a look at those attitudinal aspects of travel mode choice which were not included in the cluster analyses. Captive Public Transport Users differ significantly from all other types with regard to higher public transport excitement (p < .001, Scheffé), lower car attitude (p < .05), and higher car stress (p < .001). Compared to all other types, they also feel higher social pressure to use public transport (SN, p < .01), whereas Self-Determined Mobiles feel stronger personally obliged to use environmentally friendly modes than Affluent Mobiles and Captive Car Users (PN, p < .01).

Discussion and conclusions

In the following section the results are discussed in the context of findings of related studies. After that, target-group specific measures to reduce personal car use are derived from the mobility type profiles. Finally, future research directions are discussed.

Results in the context of other research findings

In this study based on data from 1,500 individuals aged 60 years and above, determinants of different aspects of elderly people’s mobility behavior have been analyzed. Studies that considered a similar range of variables as predictors of travel mode choice—but were not restricted to an elderly sample—largely showed comparable results (Hunecke et al. 2007, 2010). For car use, car availability, weather resistance, and public transport control are important factors, for public transport use it is public transport attitudes, aspects of centrality and car availability and finally for bicycle use cycling attitude and weather resistance.

However, former studies that also considered determinants of mobility behavior of the elderly in multivariate analyses showed some differences compared to the results in the present study. Scheiner (2005) found health and motivity (ability to move, walk around etc.), social networks and living with a partner to be significant predictors of the frequency of leisure time activities for older people. While we also detected an influence of social networks and household size, health and motivity surprisingly showed no significant impact in our analysis. However, in a study by Haustein et al. (2008), both age and disabilities also were significant predictors. A possible explanation for the insignificance of health in the present study may be that mobility-related health aspects are covered by walking attitude. Such attitudes were not included in the other two studies.

Cao et al. (2007) who predicted transit trip frequency of the elderly also showed significant impacts of car access and public transport attitudes. In their analysis attitudes towards walking/cycling also positively influence public transit use of the elderly, whereas this is not the case in our study.

Based on the most important determinants of elderly mobility behavior, four distinctive groups with specific mobility patterns were identified. The four mobility types can be differentiated according to perceived and actual restrictions of their ability to use different modes of transport. These restrictions concern health, social status, infrastructural conditions, and personal access to transport systems. The highest restrictions exist for Captive Car Users and Captive Public Transport Users. In contrast to the other two types, they can be described as socially disadvantaged. As for the mobility behavior of these two types, objective constraints seem to be of higher relevance than personal attitudes. In the case of Captive Car Users limited access to public transport and a generally poor infrastructure leads to car dependency and is correlated with low satisfaction with mobility options. In contrast, the mostly female Captive Public Transport Users have very limited access to a private car, but as they have much better infrastructural conditions, they can reach their important destinations more easily, either by walking or public transport, which is reflected in their modal split as well as in a higher amount of activities, which indicates that favorable infrastructural conditions can help to compensate mobility restrictions of the elderly.

Besides these two rather restricted types, we identified two types which appear more self-determined with regard to their mobility options. For both types, men are slightly overrepresented. Affluent Mobiles face also restrictions with regard to public transport use, but they evaluate walking and cycling more positively and reach close destinations better by these modes, due to better health conditions compared to the two older segments. But even if they were more restricted, their large social network and good access to the internet would probably partly compensate arising mobility restrictions. Self-Determined Mobiles have good access to both, the car and public transportation, and also evaluate walking and cycling positively. Their modal split shows preferences for all individual modes, i.e. driving, cycling, and walking. Their travel mode choice appears to be most strongly driven by individual choices and preferences, maybe supported by the low perceived mobility necessities. It does not surprise that Affluent Mobiles and Self-Determined Mobiles are most satisfied with their mobility options. At the same time, they exercise the highest amount of leisure time activities, which is generally associated with well-being (cf. Jansen et al. 2001; Scheiner 2004).

The resulting mobility types show some similarities with segments of the elderly that have been identified in former studies. If we compare the present types with the sociodemographic clusters created by Hildebrand (2003), three of his seven clusters resemble the ones presented here. Affluent Mobiles are reflected in Hildebrand’s Affluent Males, all of whom possess driver’s license, have high car availability, are rather young and well off. Our Captive Car Users show some similarities with the so-called Disabled Drivers. Finally, our Captive Public Transport Users can be compared to Hildebrand’s Mobility-Impaired, who are not licensed, often handicapped, older and mostly female.

Compared to the six clusters presented by Haustein et al. (2008) based on similar variables as used in the present study, we can identify a related segment for each of our clusters: Our Captive Car Users can be associated with the Restricted Mobiles, the Affluent Mobiles with the Mobile Car-Oriented, Self-Determined Mobiles exists in both studies and finally our Captive Public Transport Users show much similarity to Eco-friendly Public Transport Users. These similarities across different studies indicate that our classification represents reliable core segments that can be found in different populations of the elderly.

Conclusions and recommendations

On the one hand car access is associated with better health and well-being (Ellaway et al. 2003; Macintyre et al. 2001). It enables older people with physical limitations to still live independently and participate in normal daily activities, and as such the car can act as a compensation tool for functional limitations (Siren and Hakamies-Blomqvist 2009). This specific affordance of cars can also reduce the collective financial pressure that care for the elderly places on the public budget. On the other hand car use by the elderly is associated with both negative safety and environmental consequences.

The segmentation approach presented here aims to improve the mobility options for older people by identifying the most suitable mode options and interventions to facilitate their take up, for each segment group, while simultaneously reducing the demand for car use. Here, the four segments can serve as target groups, as their profiles provide the information necessary to identify starting points for intervention strategies in the mobility sector.

The negative view of all modes but the private car as well as health restrictions and the rather peripheral location of Captive Car Users make it very difficult to achieve a switch from the car to other modes of transport in this segment. Here, compensating mobility services, such as delivery services or escort services for trips by public transport—if public transport access is available—are suitable. However, for the future it seems to be more important to prevent that individuals become Captive Car Users as this type appears to be the most disadvantaged of all types. Once Captive Car Users are no longer able to drive by themselves, they will be highly dependent on others to fulfill their mobility needs. This dependency is also recognized by adult children of older drivers and often regarded as a burden (Rosenbloom 2010). As preventative measures, awareness campaigns that advertise the relation between walking, cycling and health could be useful (Cairns et al. 2008). Moreover, measures supporting older people to move to more central locations or improvements of neighborhood design seem appropriate. Enhancing accessibility can be regarded as a promising strategy to promote walking trips, especially for the elderly (Cao et al. 2007). Ideally, these measures should be directed to people when they are at a younger age, and still likely to have relatively high mobility capabilities and options.

As Affluent Mobiles evaluate all transport modes quite positively, a shift from the car to other modes seems much more realistic. However, their high mobility necessities enhance the use of the private car, especially due to time-pressure (cf. Haustein and Hunecke 2007). For short trips, technically developed, high quality bikes could be advertised to this affluent target group. For longer trips, especially to the city center or neighboring cities, public transport could be an alternative to the car, if the usage was more flexible, e.g. by means of electronic ticketing. The high percentage of Affluent Mobiles with internet access indicates that they are quite receptive in terms of dealing with technology.

Showing a positive view of walking and cycling, high public transport control, and low perceived mobility necessities, Self-Determined Mobiles deliver the best preconditions for a voluntary change from car to alternative modes. However, as most of them have a high car availability, they simply use the car. As they approve the health effect of walking and have strong pro-environmental norms, awareness campaigns that emphasize the health- and environment-related advantages of non-motorized modes and public transport could be effective in this group. Moreover, Self-Determined Mobiles appear as an adequate target group for car-sharing. Environmental and financial reasons could convince people of this group to dispose of their private car and become a car-sharing member.

Finally, most of the Captive Public Transport Users have no choice but to use environmentally friendly modes. As most of them live rather central, this does not seem to be difficult, as positive ratings for public transport and comparable negative ratings for the car suggest. Compared to Captive Car Users they show a higher level of leisure time mobility, although both groups are at a comparable age. This cannot only be explained with infrastructural differences—the regression analysis only shows a small impact of infrastructure on the number of activities—but especially with different attitudes towards walking and cycling, which are significant predictors of the number of leisure time activities.

However, it can be assumed that the group of Captive Public Transport Users is decreasing as in the future a growing number of older people (especially women) will possess a driver’s license and thus will not depend on public transport any longer—at least as long as they can afford a car, which is another restricting factor in this group. Thus, public transport suppliers have to be prepared that their older customers become more ambitious than today.

The results of the present study demonstrate that elderly are a heterogeneous group, that can be divided into four basic, distinct segments. A further differentiation seems not appropriate as it does not increase explained variance in mobility behavior significantly.

The used approach that also includes attitudes as type-constituent variables provides important information for different aspects of mobility behavior, especially travel mode choice. Besides attitudes, especially accessibility is a key variable for the elderly to stay mobile, which is a precondition for high quality of life (e.g. Jansen et al. 2001; Scheiner 2004). While in districts of high accessibility restricted car access can be compensated by good infrastructural conditions, for elderly living in the suburbs improvements of accessibility are necessary to ease car dependency. The results of this study indicate that car dependency is associated with lower satisfaction with mobility options, even if the question of cause-and-effect cannot be answered based on correlational data. In the future, longitudinal research could give insight into the question how stable mobility types of the elderly are and in how far mobility attitudes follow from specific mobility patterns (such as using the car only) or vice versa.

Footnotes
1

Dog ownership was recorded as people who own a dog go more often for a walk (Brown and Rhodes 2006). In addition, the dog might be an obstacle when travelling with German rail as you have to pay the same fare as for children between 6 and 14 for a larger dog (cf. www.bahn.de).

 

Acknowledgments

The author would like to thank the reviewers for their constructive advice, her colleagues at ILS for fruitful discussions and Michael Carreno for very helpful comments on the manuscript.

Copyright information

© Springer Science+Business Media, LLC. 2011