Social clustering in high school transport choices 

Get Complete Project Material File(s) Now! »

Chapter 3:  Social clustering in high school transport choices

Publication reference

Please note that Chapter 3 has been published as the following article: Long, J., Harré, N., Atkinson, Q. D. (2015). Social clustering in high school transport choices. Journal of Environmental Psychology, 41, 155-165. The article is included in this thesis with permission from Elsevier. The text is identical to its published format with the exception of adjustments to the numbering of title, figure and table headings.


Active transport offers opportunities to reduce the environmental impacts of car travel and improve health. During adolescence, friends and parents may influence transport mode to school. Using a social network survey of 934 high school students we investigated whether students’ walking, cycling, bus and car travel to school were predicted by their friends’ transport behaviour, accounting for parent encouragement, ride availability, distance to school, gender, school unit and age. In addition, we examined whether descriptive norms, friend encouragement or co-travel requests mediated the effect of friends’ active transport behaviour. We found that friends’ transport behaviour predicted ego behaviour, particularly for cycling. Descriptive norms and co-travel requests, but not friend encouragement, approached significance as mediators of friends’ active transport similarities. Parent encouragement for active transport was a particularly strong predictor of transport mode. Implications for future research and interventions are discussed.


Transport generates a substantial portion of greenhouse gas emissions, comprising nearly 23% of the world’s energy related emissions (International Energy Agency, 2009). Private car use produces substantially more greenhouse gases per passenger kilometre than public transport in most countries, whilst walking and cycling are virtually emission free (IPCC, 2007), replacing car journeys with alternative forms of transport also reduces traffic congestion and improves the overall safety of pedestrians, passengers and other road users. Active transport such as walking or cycling also provides an opportunity to increase regular physical activity (Wanner, Götschi, Martin-Diener, Kahlmeier, & Martin, 2012) which can in turn contribute to physical and psychological health (Garrard, Rissel, & Bauman, 2012). Local car trips that could be walked or cycled are an important and feasible target for change (Maibach, Steg, & Anable, 2009).
Adolescence may be a particularly important time for shaping adult transport patterns (e.g. Line, Chatterjee, & Lyons, 2012; Simons, Clarys, De Bourdeaudhuij, de Geus, Vandelanotte, & Deforche, 2013) and adult health outcomes (Lawlor & Chaturvedi, 2006). Peers are salient during adolescence and have been found to be influential for a range of behaviours (Brechwald & Prinstein, 2011; Brown et al., 2008). Social interventions, including those involving peers, may increase participation in active transport (Orsini & O’Brien, 2006; Panter, Jones, van Sluijs, & Griffin, 2010) but little is known about the role of peers in adolescents’ transport choices to and from school.

Clustering of behaviour within social networks

Social networks describe relationships between individuals in a given setting or community. Social network methods generally represent individuals as nodes in a network and social relations (e.g. friendships, interactions, associations) as the links between nodes. Social clustering (also known as network autocorrelation) describes a situation in which linked individuals in a network are more similar on a given attribute than would be expected due to chance. To establish similarities in friends’ attributes, each individual’s behaviour is measured independently and mapped onto the network. People often assume others’ behaviour is more similar to their own than it actually is (McPherson, Smith-Lovin, & Cook, 2001; Prinstein & Wang, 2005). Therefore using independent reports collated on a social network avoids a similarity bias or “false consensus effect” that can arise if individuals are asked to estimate the behaviour of their friends.
Social clustering can arise from a combination of processes that can be broadly categorized as social contagion, homophily or secondary homophily. Social contagion captures processes whereby an individuals’ behaviour is influenced by the behaviour of their peers. In contemporary work, the term social contagion is used synonymously with socialisation, friend influence and peer effects (e.g. Brechwald & Prinstein, 2011; Christakis Fowler, 2008; Dishion & Tipsord, 2011; Eisenberg, Golberstein, Whitlock, & Downs, 2013; Shalizi & Thomas, 2011). The term has historically been used to describe a myriad of sub-types of influence, particularly subtypes of imitation (see Levy & Nail, 1993; Wheeler, 1966). In this paper we use social contagion as it is most commonly used in current literature, to describe processes in which friends influence the ego (focal individual) to behave in ways that are consistent with their own behaviour.
Homophily, refers to the predisposition to select people with similar traits as friends. Homophilic selection of friends may be based on the behaviour of interest (manifest homophily), which in our case would be transport choices (Shalizi & Thomas, 2011). Friendship selection may also relate to a trait that is associated with the behaviour of interest (secondary homophily when the trait is measured, latent homophily if the trait is unmeasured) (Shalizi & Thomas, 2011). For transport behaviour, secondary or latent homophily could include selecting friends on the basis of gender or distance from school, or other traits likely to influence transport choices. For example, adolescents are more likely to select friends who live close by (Preciado, Snijders, Burk, Stattin, & Kerr, 2012) and who are the same age and gender (McPherson et al., 2001) which are all factors that have been linked to transport choices (Sirard & Slater, 2008). Features of the home environment such as parent encouragement and ride availability may also play a role here. For example, parent encouragement is known to correlate with transport choices (Panter, Jones, & van Sluijs, 2008) and may give rise to secondary homophily if students tend to form friendships with those whose parents have similar attitudes toward particular transport choices.
It can be difficult, if not impossible to conclusively differentiate between these three classes of explanation in social network surveys. Social contagion, homophily and secondary homophily are not mutually-exclusive processes (Brechwald & Prinstein, 2011; de la Haye et al., 2011b) and if homophily exists on the variable of interest this can contaminate estimation of social contagion unless very strong assumptions are made (Shalizi & Thomas, 2011). Nevertheless, simple tests for clustering of behaviour on a network can identify whether at least one of the three processes is likely to be present. Further, including potential secondary homophily variables in the analysis makes it possible to quantify their relative importance and may allow contagion effects to be ruled out. That is, if there is no clustering in transport behaviour after controlling for secondary homophily variables, this makes a contagion explanation unlikely. Conversely, incorporating variables linked to possible social contagion mechanisms into the analysis makes it possible to test the plausibility of these causal pathways and potentially provides indirect support for the role of contagion.

READ  Best Practice in Teaching English as Foreign Language

Mechanisms of contagion within social networks

When behaviours cluster, and we suspect there is some degree of social contagion present, we can ask what interpersonal mechanisms are likely driving this. Empirical work on the mechanisms underlying social contagion has been a gap in the literature on social contagion although more attention has been paid to these mechanisms in recent years (Brechwald & Prinstein, 2011).
One potential factor driving contagion effects is individuals’ perception that a behaviour is common among their friends. According to normative focus theory (Cialdini et al., 1990) information about common behaviour (descriptive norms) may provide a short-cut to decision making, leading people to adopt the common behaviour in a particular context (Cialdini et al., 1990). People may also consciously adopt the common behaviour because they assume that these behaviours are likely to be rewarded by their friendship group (Brown et al., 2008) or, drawing on social categorisation theory, because the common behaviour may become part of their identity as group members (Turner et al., 1987). If people are consciously adopting the common behaviour then descriptive norms around what behaviours are most common should mediate similarities between the individual’s travel mode choice and that of their friends. A Dutch study found that adult car use was related to perceptions of how often important others travel by car (Steg, 2005) and descriptive norms appear to be consistently related to physical activity (Maturo & Cunningham, 2013). No research to our knowledge has assessed whether descriptive norms predict adolescent transport behaviour, nor whether they underpin contagion processes if these are present.
A second mechanism potentially driving contagion effects involves reward and encouragement from peers. People may promote behaviours that match their own through verbal influence, requests and teasing (Brown et al., 2008) because conformity to in-group relevant norms increases positive emotions for the perceiver (Christensen, Rothgerber,Wood, & Matz, 2004) and affirms the influencers’ own behaviour. The term “encouragement” is often used to capture verbal influence, particularly within the health promotion literature.
Encouragement has consistently been found to predict physical activity in adolescence (Maturo & Cunningham, 2013) and friend encouragement of physical activity is also related to adolescent active transport (Deforche, Van Dyck, Verloigne, & De Bourdeaudhuij, 2010; Hohepa, Scragg, Schofield, Kolt, & Schaat, 2007). One study of UK children found that friend encouragement for active transport related to whether students cycled to school, but only for students living close to school (Panter et al., 2010). No research to our knowledge has explored whether friend encouragement of active transport is related to transport choices for adolescents. If encouragement is important for active transport and people tend to encourage this behaviour when they do it themselves then encouragement may generate similarities in friends’ behaviour.
Social contagion may also arise from opportunities to travel with friends. Transport with friends is likely to be more enjoyable than travelling alone or with parents and the time spent travelling together may also contribute to a sense of belonging, which Baumeister and Leary (1995) propose is a fundamental human motivation. Pre-adolescents in Scotland and New Zealand have reported that travelling to school can be a fun opportunity to socialise with friends and suggested that travelling with friends might boost active transport participation (Orsini & O’Brien, 2006; Panter, Jones, van Sluijs, & Griffin, 2010). Older adolescents also appear interested in co-travel: Belgian youth reported that opportunities to travel with friends altered their choice of transport mode or the distance they were willing to cycle for leisure journeys (Simons, Clarys, De Bourdeaudhuij, de Geus, Vandelanotte, & Deforche, 2013).

The present study

This study investigates similarities in friends’ transport behaviour in a New Zealand high school social network. In particular it examines students’ walking, cycling, car travel and bus travel choices at a single time point. First, we aim to identify whether adolescent transport behaviour to school shows social clustering. Second, we aim to test whether and to what extent this clustering holds when controlling for a range of demographic and context variables. These additional variables are primarily of interest because they may contribute to friend selection and could therefore explain any social clustering in terms of secondary homophily effects. Third, we explore indirect evidence for social contagion via social mechanisms that mediate similarities in friends’ behaviour.
We test the following hypotheses:
Transport behaviour will cluster socially, such that pairs of friends will be more likely than randomly selected pairs to share the same transport behaviour and friends’ average behaviour will predict the likelihood that the ego will typically use that mode. This will be true for walking (H1a), cycling (H1b), car travel (H1c) and bus travel (H1d).The predictive power of friends’ transport behaviour will hold even after controlling for demographic and context variables that may give rise to secondary homophily (H2a-d).Descriptive norms (i.e. perceptions of friends’ behaviour) will mediate relationships between ego and friends’ behaviour (H3).Friend encouragement of active transport will mediate relationships between ego and friends’ behaviour (H4).Active transport co-travel requests will mediate relationships between ego and friends’ behaviour (H5a). As shared transport is likely to be a key mechanism of social contagion or homophily, we also expect that the behaviour of friends who travel separately will not predict ego transport choices (H5b–e).


Participants were students of a large relatively affluent co-educational public high school in Auckland, New Zealand. Students who live within the school “zone” are prioritised for enrolment at this school, however around 15% attend the indigenous language unit that permits students from outside the school zone.
All students present in their weekly administration class were invited to take part in the survey. Thirty seven consenting students were excluded as they did not provide friend nominations or an ID number required to link their data with that of their friends. The final social network dataset which was used for all analyses included 934 students (71% of enrolled school population). Slightly more males (54.5%) than females took part, consistent with school demographics (57.9% male). The average age was 14.7 years and ages ranged from 12 to 19 years. Chi-square tests indicated no evidence of a difference in the gender or school year between participants and students enrolled at the school. Fewer NZ European individuals took part in the survey than would be expected based on the ethnic makeup of the school population.
Participants provided informed consent prior to the questionnaire and parents of students under 16 years old were given the opportunity to opt their child out of the study. Friend nominations were separated from the rest of the questionnaire and de-identified with ID codes before being sent to the researchers. To encourage participation, students were offered the chance to win one of five $20 (NZD) gift vouchers. The project was approved by the University of Auckland Human Participants’ Ethics committee.

READ  The flipped classroom model


The questionnaire was part of a larger research project on sustainability. Only the items of interest to the current study will be presented here. Friends were identified by asking students to “Please name [School name] students who are your close friends. You do not have to fill in all the spaces but we are providing space just in case anyone has this many close friends”.
We used an aided recall free recall approach allowing seven spaces for friend names. Similar friend nomination approaches have been used previously (e.g. Poulin et al., 2011). We also asked students to tick a box beside each name indicating whether they “often travelled to or from school with that person”.
Transport to and from school was measured by asking participants to tick their major transport mode for each journey they made to or from school last week (a total of 10 journeys). Possible response boxes for each journey were; Walk, Cycle, Motor scooter, Skate/scooter, Bus, Driven to school, I drive myself in a car, Other (please specify). Street name, suburb name and weeknights spent in each house were collected so that approximate distance to school could be calculated. No house numbers were collected to maintain student privacy. Students were also asked to report their age and gender and home number which was used to assess if they attended the language unit.
The remaining items were rated on Likert scales ranging from 1 (strongly disagree) to 7 (strongly agree). Verbal anchors were given for values 1 through 7 as this can increase the reliability of likert scale ratings (Weng, 2004). Likert scale items included active transport descriptive norms (“My friends from [school name] often walk, cycle or skate to and from school”) and active transport co-travel requests from friends (“One or more of my friends sometimes asks me to walk, cycle or skate to or from school with them”).
We also assessed active transport encouragement from friends (e.g. “My friends from [school name] encourage me to walk, cycle or skate to and from school”) and from a male or female caregiver (e.g. “My Mum (or a caregiver who acts like a mum) encourages me to walk, cycle or skate to and from school”). Perceived encouragement from mothers and fathers was later combined to form a parent encouragement variable (α = .85). Availability of rides from family members was assessed with two questions, one about the morning (“A member of my family is able to drop me at school by car”) and the other about the afternoon (“A member of my family is able to pick me up from school by car”). The ride availability questions were combined to form a single family ride availability variable (α = .76).
Finally, we asked students about their age, gender and collected their home room number in order to assess whether they attended the main school or the indigenous language unit.

Table of contents
Thesis abstract
List of Tables 
List of Figures 
Chapter 1: Introduction
1.1 Background
1.2 Social networks
1.3 Potential mechanisms of contagion
1.1 Diffusion of innovations theory
1.2 Conversations and the construction of appropriate behaviour
1.3 Individual differences and pro-environmental behaviour
Chapter 2: Methodological background
2.1 Overview of the research approach
2.2 The research context
2.3 Social network analysis
2.4 Analytical approaches
2.5 Qualitative approach
2.6 Ethics
2.7 Data collection stages
Chapter 3: Social clustering in high school transport choices 
3.2 Abstract
3.3 Introduction
3.4 Method
3.5 Results
3.6 Discussion
Chapter 4: Do high school friends influence each other’s transport to and from school? Investigating change processes over one year
4.1 Abstract
4.2 Introduction
4.3 Method
4.4 Results
4.5 Discussion
Chapter 5: Understanding change in recycling and littering behavior across a school social network 
5.2 Abstract
5.3 Introduction
5.4 Method
5.5 Results
5.6 Discussion
Chapter 6: Personality and adolescent pro-environmental behaviour 
6.1 Abstract
6.2 Introduction
6.3 Method
6.4 Results
6.5 Discussion
Chapter 7: ‘You can’t dedicate your whole life to such a thing’: Stereotypes, justifications and the discursive marginalisation of pro-environmental behaviour
7.1 Abstract
7.2 Introduction
7.3 Method
7.4 Results and analysis
7.5 Conclusion
Chapter 8: Thesis summary and conclusion 
8.1 Social network analysis opportunities and pitfalls
8.2 Social clustering and local norms
8.3 Individual differences in behaviour
8.4 Pro-environmental behaviour maybe “contagious”
8.5 Wider social contextual factors that may impact on social influence in pro-environmental behaviour
8.6 Limitations, generalisability and future research
8.7 Closing remarks
Pro-environmental behaviours in a high school social network

Related Posts