It`s about time: how, why, and when networks change

SOCIAL
NETWORKS
ELSEVIER
Social Networks 19 (1997) 1-7
It's about time: how, why, and when networks
change
J. Jill Suitor a.*, Barry Wellman b, David L. Morgan c
a Department of Sociology, Louisiana State University, Baton Rouge, LA 70803, USA
b Centre for Urban and Community Studies, University of Toronto, Toronto, Ont. M5S 2G8, Canada
¢ Institute on Aging, Portland State University, Portland, OR 97207, USA
1. Introduction
During the past 20 years scholars have provided an increasingly complex image of
individuals' social networks, using a variety of theoretical and methodological approaches to show that such networks are major providers of sociability, support,
information, and a sense of belonging. One feature common to almost all of this work
on 'personal communities' is that it has provided a picture of the ties that exist at only
one time. Although the limitations of such a single snapshot approach have been
recognized, financial and organizational constraints have limited efforts to study the
inevitable changes that occur in networks over time.
This special issue of Social N e t w o r k s brings together a groundbreaking set of studies
that provides theoretical, methodological, and substantive contributions to the study of
change in social networks. To this end, the authors use a variety of approaches to
analyze different longitudinal data sets. Taken together, this set of papers provides us
with the first concerted effort to understand: (a) the extent to which personal community
networks change over time; and (b) the processes that underlie such changes.
2. The papers
The first three papers in this issue are similar in their approach to analyzing changes
in personal community networks. By collecting data at multiple measurement points
across a year, Morgan et al. investigate the stability of the networks themselves. Next,
* Correspondingauthor.
0378-8733/96/$15.00 Copyright © 1996 ElsevierScienceB.V. All rights reserved.
PII S0378-8733(96)00287-0
2
J,J. Suitor et al./Social
Networks 19 (1997) 1 - 7
the paper by Wellman et al. examines changes in personal communities across a decade,
including an investigation of how changes in the Egos' social statuses affect their
networks. Suitor and Keeton's paper has a similar time perspective, reporting on the
effects of changes in homophily on social support networks of midlife women across a
10-year period following a return to school.
The second group of three papers addresses a different set of theoretical and
methodological issues involving network change. Leik and Chalkley's paper presents a
theoretical framework for analyzing sources of instability in networks across time. Ruan
et al.'s paper uses data collected in two separate surveys more than 5 years apart, to
show differences in the networks reported by urban Chinese. Last, Feld's paper uses
data from friendship ties within a bounded population to examine how the initial
strength of ties, especially in terms of their embeddedness in shared social networks,
affects the continuity of ties across time.
3. Methodological approaches
Most of the prior work in this area has focused on assessing the reliability of repeated
survey measurement of ego-centered networks. (See the paper by Morgan et al. for a
partial review of this literature.) Typically, such studies collected data at two times,
separated by a relatively brief interval. By contrast, many of the papers in this special
issue depart from this tradition by presenting data collected over a much more lengthy
period, or at multiple data points. The time-frames of these papers range from six
months in Leik and Chalkley's and Feld's studies to 1 year in Morgan et al.'s project, 7
years in Ruan et al.'s investigation, and a full decade in both Suitor and Keeton's, and
Wellman and colleagues' studies.
All of the authors elicited information by repeatedly interviewing respondents about
the members of their personal communities. There is significant variation in the
frequency with which information has been gathered from respondents. Where WeUman
et al., Leik and Chalkley, and Feld each interviewed respondents twice, at the beginning
and end of their time-frames, Suitor and Keeton interviewed respondents three times,
while Morgan et al. interviewed his respondents seven times in a year. All of these
scholars obtained their information by comparing what the same respondents reported
about their networks at different times. By contrast, Ruan et al. are the only authors who
did not repeat interviews with the same respondents. They obtained their information
about change in Chinese networks by comparing distributions in two separate random
samples, with similar sampling criteria and stimulus questions.
All of the articles in this issue present data from field studies of real populations, and
not experimental laboratory situations. Most of the articles in this special issue use data
collected to study particular populations within the United States, including university
students (Feld), low SES parents (Leik and Chalkley), widows (Morgan et al.), and
women who returned to school in midlife (Suitor and Keeton). Two studies use data
from outside of the United States - the residents of the East York section of Toronto,
Canada (Wellman et al.), and residents of Tianjin, China (Ruan et al.).
./.J. Suitor et al. / Social Networks 19 (1997) 1-7
3
All of the authors except Feld examine ego-centered networks. Although Feld studies
a 'whole network' (all the classmates in a university cohort), he analyzes the data in
ways similar to the other authors' ego-centered analyses, inquiring into such matters as
the percentage of ties retained.
Several authors adopt a broad definition of network membership, which includes
relatives, friends, neighbors, and co-workers (Morgan et al.; Ruan et al.; Suitor and
Keeton; Wellman et al.). Because people may have more than 1000 ties in their
ego-centered networks, these studies restrict their view to only ties that were deemed
'strong' in some respect, typically resulting in networks of less than 20 members. By
contrast, Feld examines only relations with other university classmates, while Leik and
Chalkley examine only relations within family households. These authors' definitions
allow them to take into account all the ties within these networks, not just the strong
ones.
All of the papers rely primarily on statistical analysis. However, both Suitor and
Keeton and Wellman et al. complement their quantitative analysis with qualitative
accounts by respondents of how and why their personal community networks changed.
4. Questions addressed
The common approaches taken in this set of papers allow them to address collectively
a set of key questions.
4.1. To w h a t extent do ties persist?
Just knowing the extent to which ties persist is an important piece of information.
Remarkably, there is a great deal of similarity in persistence of ties across specific
time-frames among some of these studies. For example, both Morgan and colleagues
and Suitor and Keeton find that about two-thirds of the associates named at T1 were
named again 1 year later. Further, both Suitor and Keeton and Wellman and colleagues
report that between about one-quarter and one-third of the ties from T1 continue to be
mentioned again a full decade later.
These two sets of findings on persistence are especially interesting when taken in
combination. Clearly these networks are not losing a third of their membership every
year; if this were the case, there would be virtually no associates remaining over a
10-year span. Instead, this pattern fits the claim by Morgan and colleagues that networks
are generally comprised of both a 'persistent core' of ties that is maintained across time,
and a 'periphery' of ties that is regularly replaced.
It is interesting to note how differently various authors interpret the persistence they
found. For example, Suitor and Keeton interpret their degree of persistence as remarkable, considering the magnitude of changes in all of the women's lives across the
decade, while Wellman and colleagues note how small a percentage persisted. Is the cup
one-quarter full or three-quarters empty?
As ties come and go, a person's cumulative inventory of relationships grows over
4
J.J. Suitor et al./Social Networks 19 (1997) 1 - 7
time, as Morgan and colleagues point out. The accumulation of ties can facilitate the
building of bridges among social circles when individuals maintain ties from one focus
of activity (Feld, 1981) to another. But to what extent do once-strong ties remain
connected? Both Wellman and colleagues' and Suitor and Keeton's findings suggest that
most former intimates drop out of the network over the course of a decade.
4.2. Why do some ties persist more than others?
Several authors find that strong ties are more likely to persist than weaker ones. (We
should note that the authors are actually talking about gradations of strength within a
small subset of ties that are all relatively strong in comparison with the other thousand
or so members of a personal community network.) Yet different studies emphasize
various dimensions of tie strength: Feld, Suitor and Keeton, and Wellman et al. each
show that supportive ties are the most likely to persist. Feld and Wellman et al.
demonstrate that frequent contact between network members is also associated with the
persistence of relationships.
Other relational criteria are also associated with persistence: Feld, Morgan et al., Leik
and Chalkley, Ruan et al., and Wellman et al. all find that kinship ties are more
persistent than ties between friends, neighbors or workmates. Distinguishing among
different types of networks, Suitor and Keeton find that kin tend to persist in emotionally supportive relationships but not as people with whom to socialize. Suitor and
Keeton are also the only authors that identify a social characteristic of network members
as contributing to the persistence of a tie: The greater the educational homophily
between the associate and the respondent, the more likely the associate was to remain in
the network.
Two authors differ regarding the extent to which network structure fosters persistence: Where Feld finds that 'structural embeddedness' (interacting as part of a set of a
ties that is more than a two-person relationship) is associated with the persistence of ties
among university studenLs, Wellman et al. do not find that a network member's
centrality or the social density of the overall network is associated with the persistence
of ties among Toronto residents.
4.3. Which social characteristics of egos affect turnover in networks?
As personal community networks in the Western world are primarily oriented toward
providing sociability and support for domestic situations, it is not surprising that
Wellman et al. identify changes in marital status (particularly the move into marriage) to
be responsible for major turnover in networks. They also find that changes in Egos'
neighborhood or employment status are not associated with changes in their personal
community networks. Leik and Chalkley also point to the importance of domestic
situations in understanding the extent of turnover in the networks of low-income parents.
In contrast, Suitor and Keeton do not find that change in marital status is related to
change in social networks; as they note, this is probably due to the fact that only women
who remained in the same geographical area were included in the sample. Last, Ruan et
J.J. Suitor et al. // Social Networks 19 (1997) 1-7
5
al. concentrate on the demographically central concerns of age and gender: As China has
changed, young adults have had fewer ties to workmates than their age-mates reported 7
years earlier, while women have experienced an increase in their number of friends.
4.4. What causes o f change are analyzed?
Are there inherently different types of changes in networks? Leik and Chalkley's
paper focuses on teasing out four distinct potential causes o f change: the unreliability of
data from time to time (background noise); inherent instability in systems (such as
psychological ups and downs in intimacy); change due to normal system dynamics (such
as childbearing); and external factors (such as economic depression in society or, on a
more microscopic level).
Other papers in this issue take somewhat different approaches. For example, Morgan
et al. use their seven waves of interviews within a year to analyze the unreliability of
data and the instability of systems. Feld has a similar focus in his whole network study.
Suitor and Keeton and Wellman et al. analyze changes due to dyadic relations between
Egos and network members. Wellman et al. focus on changes in the Egos' aging and
marital situations; while Suitor and Keeton focus on changes in educational homophily
between the respondents and members of their social networks. In contrast to the rather
unchanging external dynamics of North American societies, Ruan et al. are able to
analyze the effects of the massive external socioeconomic changes in China during the 7
years between interviews.
4.5. How do changes in personal communities intersect with changes in society?
Ruan et al.'s paper analyzing changing social networks in China is the only study
conducted in a society that is rapidly changing in fundamental ways. The finding of a
decrease in the number of co-workers in these networks is associated with the shift in
China from a state bureaucratic to a market society. As part of this shift, work
organizations are no longer claiming total control over people's lives. Concomitantly,
the evidence from this paper suggests that workmates no longer play as important a role
in people's social lives. They find more of their sociability and support with friends, as
their society becomes socially liberated.
Two other papers, while not addressing such large-scale issues, do attempt to
consider how the nature of people's personal communities may change in response to
social changes. Morgan et al. find a core-periphery structure following the transition to
widowhood: a stable core populated principally by kin, plus a more transitory set of
friendships. Morgan and colleagues' findings are consistent with Wellman et al.'s view
of personal communities. They too, see a stable core of supportive friends and
immediate kin, but they note that there is some change in this core over a decade.
Wellman et al. suggest that changes in personal community networks have two
dynamics. Both the core and periphery of the networks are usually slowly changing sets
of ties. They adapt to external circumstances and they help the Egos at their centers to
adapt, just as Ruan et al. find. Yet Wellman et al. point to the almost complete turnover
6
J.J. Suitor et al./Social Networks 19 (1997) 1-7
of relationships associated with change in marital status to propose that a second process
was at work: rapid 'catastrophic' change that is associated with fundamental change in
the social circumstances around which these networks are organized. In the case of the
Torontonians, this was marital change, but in other societies it might well be political or
economic change.
4.6. The future of studying change in ego-centered networks
We believe that future work will broaden the study of changes in networks along
three dimensions: (1) the type of longitudinal data; (2) the type of change; and (3) the
type of network.
With regard to the type of data, longitudinal research designs have historically
distinguished between time-series data, panel data and trend data. Time series studies
collect many measurements at regular intervals, and typically use curve-fitting procedures to track the course of key indicators. Given the cost and complexity of network
studies, time-series analysis will probably be the rarest form of study of network change,
although the paper by Morgan et al. shows some of the possibilities for such work. Panel
studies follow a given sample across a discrete number of time points, and typically
concentrate on changes between pairs of time points. The papers by Wellman et al.,
Suitor and Keeton, Leik and Chalkley, and Feld all fit this pattern - just as this is the
common design among the studies presented here, it is also likely to be the dominant
direction for future work on network change. Trend studies represent a third alternative,
using separate samples to track key variables across a limited number of data collections, and typically concentrating on changes between pairs of variables at subsequent
time points. Ruan et al.'s work on contemporary China demonstrates how comparisons
of trend data on networks can give empirical definition to the nature of social change.
The type of change that one investigates provides a further source of research designs
for studies of network change. In particular, there is a distinction between changes that
follow some well-defined focal event, versus broader studies of social change. Most of
the studies in this special' issue use specific events as their starting points, including
widowhood (Morgan et al.), returning to school (Suitor and Keeton), participation in an
intervention program (Leik and Chalkley), and entering college (Feld). Although these
studies follow individuals' networks, similar designs could apply to other units of
analysis, such as examining networks at the community level following a major
transformation such as a natural disaster; or at an organizational level following a major
economic or regulatory change, etc. An example of this is provided by Ruan et al., who
use their data on China to illustrate how studies at the community or national level can
track network variables across time as key social indicators.
A final way to think about future directions for the study of network change
distinguishes between types of networks, especially bounded versus unbounded networks. With the partial exception of Feld's paper, all of the articles in this special issue
deal with unbounded networks. This emphasis on change in unbounded networks differs
from the majority of the work published in this journal which has dealt with bounded
networks. Further, work on change in bounded networks has a relatively long history,
J.J. Suitor et aL /Social Networks 19 (1997) 1 - 7
7
going back at least to the original sociometric analyses by Moreno in the 1930s
(Moreno, 1934). It is our expectation that work on changes in bounded networks will
continue to be a direction for future research. We hope that the papers presented in this
special issue show how work on unbounded networks can also be fruitful.
In sum, it is clear that there is a wide-ranging research agenda associated with
changes in social networks. We hope that the articles in this special issue provoke
additional interest in research that increases our understanding of social networks
through investigations of changes in their structure and function.
References
Feld, S.L., 1981, The focusedorganizationof social ties, AmericanJournal of Sociology86, 1015-1035.
Moreno,J.L., 1934, Who shall survive? Foundations of sociometry, group psychotherapy,and sociodrama.
Nervous and mental disease monograph58 (Washington, DC).