PosterPDF Available

A comparative study of contact frequencies among social network members in five countries

Authors:
  • Federal Office for Spatial Development, Bern, Switzerland
  • Swisscom Digital Technology

Abstract

As face-to-face and ICT-mediated social interaction patterns are relevant to explain (social) travel behavior, the objective of this paper is to study comparatively the factors that influence social interaction frequency with different communication modes. The analysis is based on seven recent data collections on personal social networks, from Canada (Toronto), Chile (Concepción), Switzerland (Zurich and nationwide), the Netherlands (Eindhoven), and Japan (Greater Tokyo). A multilevel-multivariate lognormal hurdle model that explicitly accounts for the hierarchical nature of the data and the zero-inflation process is used to jointly analyze contact frequency patterns across all samples. We show the existence of very consistent associations between social interaction frequency and some individual and relational characteristics such as network size, distance and emotional closeness, irrespective of socio-cultural context. For variables such as education level, gender and relationship type, effect patterns are less clear. At the same time, for other characteristics such as gender, income and relationship type, effect patterns are less clear, differences that might be explained by intrinsic contextual characteristics as well as methodological differences among studies.
+41 44 633 3105
info@ivt.baug,ethz.ch
www.ivt.ethz.ch
A comparative study of contact frequencies
among social network members in ve countries
Giancarlos Troncoso Parady1+, Andreas Frei2, Matthias Kowald3, Sergio Guidon4, Michael Wicki5, Pauline van
den Berg6, Theo Arentze6, Harry Timmermans6, Juan-Antonio Carrasco7, Barry Wellman8, Kiyoshi Takami1,
Noboru Harata1, Kay W. Axhausen4
1Department of Urban Engineering, the University of Tokyo; 2Redwood Logistics; 3Faculty of Architecture and Civil Engineering, RheinMain University of Applied Sciences;
4Institute for Transport Planning and Systems (IVT), ETH Zürich; 5Spatial Development and Urban Policy (SPUR), ETH Zürich; 6Department of the Built Environment,
Eindhoven University of Technology; 7Department of Civil Engineering, Universidad de Concepción; 8NetLab Network, Ryerson University.
Background
Social activities are responsible for a substantial
share of trips (Axhausen, 2005). These social trips
are inuenced by new ways of social interaction
made possible by the rise of new information and
communication technologies (ICT), which might af-
fect demand dynamics on transportation systems.
Attempting to tackle this issue, several data collec-
tion eorts have been conducted:
Toronto, Canada: Hogan, Carrasco and Well-
man (2007),
Zurich, Switzerland: Frei and Axhausen (2007),
Guidon et al. (2018),
Switzerland nationwide: Kowald and Axhausen
(2012),
Eindhoven, the Netherlands: van den Berg,
Arentze and Timmermans (2009),
Concepción, Chile: Carrasco and Cid-Aguayo
(2012),
Greater Tokyo Area, Japan: Parady, Takami and
Harata (2019).
In these studies, a personal networks approach has
been used to capture information on individuals
(egos) and their social network members (alters).
Findings from these studies have underscored the
need to incorporate social network characteristics
into social activity-travel behavior research.
Research Objectives
While these studies have independently reported
ndings in their own socio-cultural contexts, there
might be substantial dierences across dierent
contexts in terms of communication patterns, social
activity and travel behavior. As such, in this study,
a comparative analysis of contact frequencies by
mode is conducted across ve countries on three
continents, guided by the following research ques-
tions:
• What are the dierences in social interaction
(contact) patterns between societies and nation-
states?
What role does distance play in these patterns?
How are personal and network characteristics as-
sociated with social interaction patterns?
To what extent are these associations contextual-
or culture-specic, or consistent across societal
and cultural backgrounds?
1 Introduction
This analysis is based on seven recent data collection eorts on personal social networks from Toronto, Cana-
da (TOR), Concepción, Chile (CCP), Zurich and Switzerland nationwide (ZUR1, ZUR2, SWI), Eindhoven, the
Netherlands (EIN), and the Greater Tokyo Area, Japan (TYO).
Face-to-face meeting frequencies show a decreasing trend for all datasets with increasing distance, no clear
pattern observed for telephone and internet-based frequencies.
• Market shares are aected by increases in distance, with similar patterns across datasets.
• Clear temporal dierences observed. For the most recent datasets (TYO and ZUR2) given longer distances,
internet-based communication replaced phone-based communication. For the Greater Tokyo sample, at 10
km the market share is almost 50%.
4 General interaction patterns
3 Data
Fig.1. Yearly average interaction frequency by geographical distance
Fig 2. Mode-specic market shares by geographical distance
To analyze social interaction (contact) frequencies
by mode, a multivariate-multilevel lognormal hurdle
model is estimated using bayesian methods. Such
a model accounts for the hierarchical structure of
the data as well as the zero-ination observed in
the social contact frequency distributions by mode.
2 Methods
Table 1. Survey information and basic network characteristics of respondents
+41 44 633 3105
info@ivt.baug,ethz.ch
www.ivt.ethz.ch
A comparative study of contact frequencies
among social network members in ve countries
Giancarlos Troncoso Parady1+, Andreas Frei2, Matthias Kowald3, Sergio Guidon4, Michael Wicki5, Pauline van
den Berg6, Theo Arentze6, Harry Timmermans6, Juan-Antonio Carrasco7, Barry Wellman8, Kiyoshi Takami1,
Noboru Harata1, Kay W. Axhausen4
1Department of Urban Engineering, the University of Tokyo; 2Redwood Logistics; 3Faculty of Architecture and Civil Engineering, RheinMain University of Applied Sciences;
4Institute for Transport Planning and Systems (IVT), ETH Zürich; 5Spatial Development and Urban Policy (SPUR), ETH Zürich; 6Department of the Built Environment,
Eindhoven University of Technology; 7Department of Civil Engineering, Universidad de Concepción; 8NetLab Network, Ryerson University.
5 Model specication
Prior specications:
• Fixed eects: Normally distributed N(0,5),
• Random eects, lognormal regression variance: half-t(4,0,3),
• Correlation parameters: LKJ-correlation with parameter ζ= 1.
Other specications:
Four independent chains
• 5000 iterations per chain (Burn-in: 1000 iterations; Thinning = 1)
Model convergence assessed via the Gelman and Rubin’s convergence diag-
nostic test (Gelman and Rubin, 1992).
Results showed the existence of very consistent associations between social in-
teraction frequency and some individual and relational characteristics such as
network size, distance and emotional closeness, irrespective of socio-cultural
context. For variables such as education level, gender and relationship type, ef-
fect patterns were less clear. These dierences might be explained not only by
socio-economic and cultural factors, but also by methodological dierences in
terms of sampling, survey medium and question formulation. While the precise
source of these variations remain an open question given the diculties associ-
ated with discriminating between these eects, it is clear that intrinsic contextual
characteristics play an important role and need further understanding.
8 Conclusion
7 Simulation results
6 Key ndings
Hurdle model (Zero-ination):
Age is positively associated with likelihood of phone contact (non-zero frequency)
and negatively associated with likelihood of internet contact.
Education level is positively associated with higher likelihood of ICT-based
contact.
Emotional closeness is positively associated with likelihood of contact irrespective
of mode.
Distance is negatively associated with likelihood of face-to-face contact, and
positively associated with likelihood of ICT-based contact.
Contact frequency:
With few exceptions, no clear patterns were observed for ego-level characteristics.
Age was negatively associated with face-to-face contact frequency for (EIN,
SWI, ZUR1, ZUR2).
Social network size was consistently and negatively associated with contact
frequency irrespective of mode, although signicance diered across samples.
Emotional closeness was consistently associated with higher contact frequencies
irrespective of modes. ZUR1 was the exception exhibiting a negative association.
• Distance showed the most consistent eect, being negatively and signicantly
associated with face to face and phone contact frequencies. In the case of
internet frequency, a negative association was observed although not signicant
in many cases (see Fig 2).
Random eect correlations:
• Zero ination: Face-to-face with phone (0.54); face-to-face with internet (-0.09);
phone with internet (0.09).
• Contact frequency: Face-to-face with phone (0.78); face-to-face with internet
(0.69); phone with internet (0.80).
Fig. 2 Simulated changes in contact frequencies by mode and relationship type given changes in ego-alter distance. Other pa-
rameters set as follows: Education: university; gender: female; age: 30-60; network size: sample averages; emotional close-
ness: very close. Non-signicant distance parameters xed to 0.
... Results showed significant associations between contact frequency by mode and (1) personal and network attributes, (2) relational attributes, and (3) ego-alter spatial distances. The reader is directed to the work of Parady et al. (2020) for a more in-depth comparative analysis of contact frequencies by mode across different socio-cultural contexts using this and similar datasets. ...
Article
Full-text available
This article presents the results of a survey on egocentric social networks in the Greater Tokyo Area. This is, together with our preliminary study, the first study on egocentric social network in Japan that uses an unrestricted name generator to elicit personal networks. It is comparable to previous work conducted in Europe (Switzerland and The Netherlands) and The Americas (Canada and Chile). In addition to a thorough description of the survey design and execution process, basic results regarding network characteristics and social interaction patterns, and estimation results of a multilevel multivariate mixed effect model of social contact frequency by mode are presented and compared against relevant benchmark data. The information provided in this article and the supplementary documents will allow its use as a new benchmark study in the subject of social networks and social interactions research.
Article
Social tie maintenance has always had cognitive and emotional costs and has been leading to uneven distribution of communication volume among interaction partners of individuals. This distribution, known as social signature, is assumed to be stable for each person. Availability of digital traces of human communication allows testing whether this assumption is true and whether it holds in specific channels of computer-mediated communication. In this paper, we investigate private messaging on a popular social networking website on a sample of 39 users and 8063 communication partners of those users over the period of 18 months. We find that this communication channel does not reduce cognitive costs as the overall number of users’ active contacts, on average, is equivalent to the cognitive limit known as Dunbar’s number. Confirming some previous research, we show that the volume of communication is unevenly distributed, related to emotional closeness, and that changes in this distribution (that is, the changes in social signature) over time within an individual are smaller than the distances between social signatures of different individuals. However, as an absolutely novel finding, we demonstrate that the changes within individuals are statistically significant, thus questioning the concept of social signature as a stable phenomenon.
ResearchGate has not been able to resolve any references for this publication.