Interdisciplinary Description of Complex Systems 13(3), 342-353, 2015
*Corresponding author, η: [email protected]; +81 022 277 4760;
*Department of Developmental Biology and Neurosciences, Graduate School of Life Sciences,
*Tohoku University, Katahira 2-1-1, Aoba-Ku, Sendai 980-8578, Japan
*
A THEORY FOR COMPLEX SYSTEM’S
SOCIAL CHANGE: AN APPLICATION OF
A GENERAL ‘CRITICALITY’ MODEL
Ichiro Shimada1, * and Tomio Koyama2
1Department of Developmental Biology and Neurosciences, Graduate School of Life Sciences
1– Tohoku University
1Sendai, Japan
2Institute for Materials Research – Tohoku University
2Sendai, Japan
DOI: 10.7906/indecs.13.3.1
Regular article
Received: 3 June 2015.
Accepted: 23 June 2015.
ABSTRACT
Within the developed nations deterioration in the basis of society, as dramatically demonstrated by the
Lehman collapse, has reached extreme levels, and currently the formation of pro-change agents is
approaching a decisive stage. Here, we will construct a complex systems 'criticality' model, apply it to
social change, and examine its reliability and validity. The model derived a power law distribution of
the output of social change. The validity of the model was verified by examining vote shares of
parties in Japan. Based on the results of this examination, we propose a new quantitative strategy
“information entropy enhancement” for social change.
KEY WORDS
complex systems social change, criticality, power-law distribution, information entropy enhancement
strategy, vote share
CLASSIFICATION
JEL: C65, P16
PACS: 89.75.-k, 89.65.Ef
mailto:[email protected]�
A theory for complex systems social change: an application of a general 'criticality' model
343
INTRODUCTION
Complex societies, such as the developed capitalist country of Japan, are networks of complex
systems that evolve around the focal point of struggle between pro-change and pro-establishment
powers. The pro-change forces are in turn composed of networks of complex systems having
diverse connections (information, needs, movements, cooperation/collaboration etc.) between
various individuals and groups, in order to break free of existing systems and effect a shift to
a new system. Social change manifests in the ‘critical state’ of mutual antagonism and
conflict between the pro-change and pro-establishment powers, while the dimensions of the
change (quantitative and qualitative) are defined by the network pattern of the pro-change
power. Simultaneously, the heretofore dominant pro-establishment power is relegated to a
secondary role. In this article, we will generalize the quantitative theory of adaptive animal
behavior [1-3] into a ‘criticality’ model of complex systems (Appendix), apply it to the
quantitative aspects of social change, construct a model for complex systems social change,
and examine the model’s effectiveness.
A COMPLEX SYSTEMS SOCIAL CHANGE MODEL
The criticality model of complex systems can be applied.
Interdisciplinary Description of Complex Systems 13(3), 342-35.docx
1. Interdisciplinary Description of Complex Systems 13(3), 342-
353, 2015
*Corresponding author, η: [email protected]; +81 022 277 4760;
*Department of Developmental Biology and Neurosciences,
Graduate School of Life Sciences,
*Tohoku University, Katahira 2-1-1, Aoba-Ku, Sendai 980-
8578, Japan
*
A THEORY FOR COMPLEX SYSTEM’S
SOCIAL CHANGE: AN APPLICATION OF
A GENERAL ‘CRITICALITY’ MODEL
Ichiro Shimada1, * and Tomio Koyama2
1Department of Developmental Biology and Neurosciences,
Graduate School of Life Sciences
1– Tohoku University
1Sendai, Japan
2Institute for Materials Research – Tohoku University
2Sendai, Japan
DOI: 10.7906/indecs.13.3.1
Regular article
Received: 3 June 2015.
Accepted: 23 June 2015.
ABSTRACT
2. Within the developed nations deterioration in the basis of
society, as dramatically demonstrated by the
Lehman collapse, has reached extreme levels, and currently the
formation of pro-change agents is
approaching a decisive stage. Here, we will construct a complex
systems 'criticality' model, apply it to
social change, and examine its reliability and validity. The
model derived a power law distribution of
the output of social change. The validity of the model was
verified by examining vote shares of
parties in Japan. Based on the results of this examination, we
propose a new quantitative strategy
“information entropy enhancement” for social change.
KEY WORDS
complex systems social change, criticality, power-law
distribution, information entropy enhancement
strategy, vote share
CLASSIFICATION
JEL: C65, P16
PACS: 89.75.-k, 89.65.Ef
mailto:[email protected]�
A theory for complex systems social change: an application of a
general 'criticality' model
343
INTRODUCTION
Complex societies, such as the developed capitalist country of
3. Japan, are networks of complex
systems that evolve around the focal point of struggle between
pro-change and pro-establishment
powers. The pro-change forces are in turn composed of
networks of complex systems having
diverse connections (information, needs, movements,
cooperation/collaboration etc.) between
various individuals and groups, in order to break free of
existing systems and effect a shift to
a new system. Social change manifests in the ‘critical state’ of
mutual antagonism and
conflict between the pro-change and pro-establishment powers,
while the dimensions of the
change (quantitative and qualitative) are defined by the network
pattern of the pro-change
power. Simultaneously, the heretofore dominant pro-
establishment power is relegated to a
secondary role. In this article, we will generalize the
quantitative theory of adaptive animal
behavior [1-3] into a ‘criticality’ model of complex systems
(Appendix), apply it to the
quantitative aspects of social change, construct a model for
complex systems social change,
and examine the model’s effectiveness.
A COMPLEX SYSTEMS SOCIAL CHANGE MODEL
The criticality model of complex systems can be applied to the
dynamics of complex systems
social change. Here, the pro-change power of the complex
system society is considered to be
the confrontation power, constructed from complex system
networks of diverse connections
between various individuals and groups for the purpose of
breaking free of the existing
systems of the complex system society and effecting a shift to a
new system. We will define
4. the number of diverse and active individuals and organizations
that structure the pro-change
power's network and are related to the output R as N. Following
the criticality model, the
number of network patterns, for example, an information
exchange quantity (information
sorts × frequency) is hypothesized as I(αN) ∼ exp(αN) (see
expression (1) in Appendix1).
It does not touch upon the details of mutual interactions
between the various elements of the
two conflicting powers. Hereafter, figures in parentheses
indicate equations in Appendix. The
existence probability of the number of network patterns of the
pro-change power is given by
P(αN) (2), while the existence (resistance) probability of the
pro-establishment power acting
upon the pro-change network is given by Q(βN) (3). Assuming
the social change output R is
inversely proportional to the resistance probability Q(βN) (4),
we can use it to obtain the
probability density function φ(R) (6) through variable
transformation by R of the existence
probability of the pro-change power, P(αN). Approximating the
information entropy of the
network of a social pro-change power N using HM (9) leads to
αM (12), which incorporates
the information effect in α. The criticality model of complex
systems leads logically to the
power law Φ(R) ∼ R–DM (DM = αM/β) for the social change
output R. It should be noted that
the size (strength) of the pro-change power defined by αM and
NC (largest of N) does not
directly correspond with the individual change outputs of R .
Even if the pro-change power is
constant, the change output R will be unstable and fluctuate
5. greatly, following a power law
distribution. The parameter of existence probability for the pro-
change power is αM; it is
opposed to the pro-establishment power's resistance probability
parameter β in contributions
to the power exponent DM. Consequently, DM reflects
relatively each agonistic probability,
and therefore, real power relationship of both powers. Thus, it
is difficult to grasp the
agonistic structure directly from the output R, which expresses
complex factors, but it can be
captured via the power exponent DM.
In order to increase the probability of true social change, the
agonistic structure expressed by
the power exponent must be greatly altered. To do this, the
power exponent DM should be
I. Shimada and T. Koyama
344
decreased so that the change output should dramatically
increase. In other words,
αM ~ α0 exp(∆H) must be small (decrease ∆H), while β must be
large, and the output R
should be greatly increased. In particular, ∆H is very effective
as it contributes exponentially
to DM. Given ∆H = H0 – HM (10), an effective strategy for
social change should be increasing
information entropy HM in the pro-change power's network
patterns (9) and reducing ∆H (i.e.
amount of memory; quantitative theory of adaptive animal
6. behavior [1-3]). Put another way,
this model suggests that activities for complex systems social
change should focus on
informational and ideological confrontations via peaceful
means, rather than forceful
struggle. In order to test the validity of this model, the question
of what to use to ascertain the
multiple possible social change outputs is extremely important.
Within developed nations, the
foundation of society is a representative democracy composed
of a party government. By way
of addressing one social (change) output, we selected the share
of votes received by political
parties; their rates of change were then aggregated and analyzed
over a number of years. The
rate of change in the vote share is the absolute difference in
vote share between one election
and the next at fixed intervals. In terms of the model, by
examining not the size of the rate of
change, but rather the distribution’s exponents, the agonistic
structure (real power relationship)
can be ascertained. In this paper we focus on the national
elections held in Japan after World
War II. In Japan many political parties were repeatedly
launched, merged and divided during
a period of seventy years after the war and at present eleven
parties exist with seats in the Diet
of Japan which consists of Upper and Lower Houses. Among the
parties only the Japanese
Communist Party, which is the third largest opposition party,
and the Liberal Democratic
Party, which is now in power, have been maintaining the unity
over half a century. The other
parties including the present largest opposition party, the
Democratic party of Japan, have
been founded recently. Hence, one understands that it is
7. appropriate to choose the above two
long-lived parties in the analysis of the agonistic structure in
Japan. Specifically, we used
national (1947-1980) and proportional-representative (1980-
2010) Upper House elections,
which are held regularly every three years and in which all of
Japan is treated as a single
constituency. First we will examine the Japanese Communist
Party (JCP), which has long
operated under the same name and has the largest sample size,
and then the Liberal
Democratic Party of Japan (LDP). Finally, we will look at the
relationships between them.
RESULTS
JCP VOTE SHARE
Figure 1 shows the trend of the JCP’s vote share in Upper
House elections from 1947 to 2010.
Twenty-two elections have been held during this time. The vote
share changes dramatically,
reflecting the complex political, economic, and cultural
conditions of the times. At its peak in
1998, the JCP “surged” when the Hashimoto government
suffered a major defeat after raising
the consumption tax to 5 %, causing economic deterioration. As
a dynamic output of social
change, analysis is performed with respect to the rate of change
R of the vote share2.
Figure 2 shows a double logarithmic plot with the rate of
change R on the horizontal axis, and
the normalized number of elections with a rate of change
greater than R on the vertical axis.
Surprisingly, excluding five points of data in which the rate of
8. change is small, the points fall
along a single line and clearly exhibit a specific power-law
distribution with an exponent of
1,27. As there seems to be nothing special about the times when
the excluded five data points
occurred, they can be regarded as noise from the view point of
social change (since we are
focusing on large R values). This result shows that there is a
clear law governing fluctuations
in vote share determined by complex, diverse factors in each
era; in the case of the JCP’s vote
share, there exists a specific social change equation (power law
distribution with the exponent
A theory for complex systems social change: an application of a
general 'criticality' model
345
Figure 1. The trend of the JCP’s vote share in Upper House
elections from 1947 to 2010.
Figure 2. A double logarithmic plot with the rate of change R of
vote share on the horizontal
axis, and the normalized number of elections with a rate of
change greater than R on the
vertical axis. The power exponent DM obtained in terms of the
least squares method is 1,27.
Year
vo
te
17. th
a
n
R
rate of change R of vote share
I. Shimada and T. Koyama
346
1,27). Large fluctuations (e.g. 1998) conform to this equation
and are not anomalous surges.
That is to say, the election results, including the surges, have
not changed the real power
relationship as expressed by the power exponent 1,27.
LDP VOTE SHARE
Since its formation in 1955 through a conservative alliance, the
LDP has lost twice to
opposition parties, falling out of power for brief periods.
Otherwise, it has occupied a central
position as a consistently pro-establishment power. Figure 3
shows the trend of the LDP’s
vote share in 19 elections from 1956, the year after its
formation, to 2010. Figure 4 shows a
double logarithmic plot with the rate of change R on the
horizontal axis, and the number of
elections with a rate of change greater than R on the vertical
axis. The majority of data points,
18. 10 of 18, diverge from a straight line. Here, the power exponent
is 2,27. Next, a single
logarithmic plot of the same data is shown in Figure 5. All data
falls along a straight line, and
displays exponential distribution. This is in contrast with the
JCP’s power-law distribution.
RELATIONSHIP OF PARTIES’ VOTE SHARE
Figure 6 plots the vote share of the JCP (horizontal axis) and
LDP (vertical axis), showing the
correlation between the two. It indicates a clear negative
correlation. When the JCP vote
share rises, the LDP share falls; the reverse is also true. The
isolated point to the bottom-right
represents data from the 1998 Hashimoto cabinet Upper House
election, but aside from this
point, the correlation coefficient does not fluctuate. This
supports the validity of the
assumption of the complex systems social change model
corresponding to Eq. (4) in Appendix.
Figure 3. The trend of the LDP’s vote share in 19 elections from
1956, the year after its
formation, to 2010.
LDP
Year
vo
te
s
h
23. te
s
h
a
re
(
p
.c
.)
A theory for complex systems social change: an application of a
general 'criticality' model
347
Figure 4. A double logarithmic plot with the rate of change R on
the horizontal axis, and the
number of elections with a margin of change greater than R on
the vertical axis. The power
exponent in terms of the least squares method is 2,27.
Figure 5. A single logarithmic plot of the same data as shown in
Figure 4.
rate of change R of vote share
th
e
28. le
ct
io
n
s
I. Shimada and T. Koyama
348
Figure 6. Plots of the vote share of the JCP (horizontal axis)
and LDP (vertical axis). The
coefficient of the correlation value is –0,48.
DISCUSSION
The JCP displays a definite power law and large changes in its
vote share; within the model,
it exhibits the characteristics of a pro-change power, which
corresponds with its actual role.
In contrast to the JCP, the LDP, which is regarded as the
establishment party, exhibits an
exponential distribution, and has a relatively stable rate of
change with minor fluctuations.
Viewed theoretically from the establishment side, when β is
small, and the pro-change
power’s resistance probability Q(βN) is small, this leads to the
stable continuation of the
exponential distribution (8). However, as indicated by the
fallout from the Lehman collapse
and the severe deflationary economic downturn in Japan, the
buildup of discrepancies in the
29. basis of Japanese society is reaching an extreme level. In order
to break free from this
stagnant era and open up future prospects, new strategy for
social change is needed. Currently,
the country is approaching a dramatic transition from the state
of social change that has
existed up until now (exponent DM = 1,27) to a state of real
criticality, the most probable for
social change (DM = 1). The model suggests that a decrease of
the power exponent DM (1,27)
changes the agonistic structure greatly, leading to an increase in
the change output R.
In order to decrease the power exponent DM = αM/β (slope),
αM ∼ α0 exp(∆H), thus a decrease
in ∆H = H0 – HM will be effective. Consequently, increasing
the information entropy HM (9)
of the pro-change power’s network pattern should be effective
for real social change. In other
words, the model suggests an “information entropy
enhancement” strategy. This raises the
question of what it means to increase information entropy in the
pro-change power’s network
pattern. First it means the expansion of the uncertainty and
variety of information. It denies
the concentration of specialized information, and expands a
united front network composed
of diverse singular issues (single points) – i.e. letting “a
hundred schools of thought contend”.
JCP
L
D
P
30. 5 10 15 20 25 30
0
10
20
30
40
50
60
JCP
L
D
P
A theory for complex systems social change: an application of a
general 'criticality' model
349
Moreover, in the quantitative theory of adaptive animal
behavior, ∆H is the amount of
memory, so reducing it should facilitate new information
creation, sharing and dissemination,
31. and departure from experientialism, the revelation of hidden
social realities, etc. Stirrings
such as the Arab Spring, Occupy Wall Street, and protests at the
Prime Minister's residence
(of Japan) are real examples of a new social change movement
abroad and in Japan.
Therefore, the central issues for social change movements are
the informational and
ideological struggle and the construction of a diverse social
change network.
On the other hand, approval ratings for governments showed
exponential distributions [4].
The results may indicate from our model that the approval
groups examined were not in the
‘critical state’ of mutual antagonism and conflict between
approval and disapproval groups.
The above is a quantitative discussion of social change based on
models. It does not address
the quality (contents) of change, such as what will change or
what kind of society will be
created. The united front strategy based on democratic reform
theory [5] essentially and
qualitatively defines the object of change above all else. The
quantitative “information
entropy enhancement” strategy proposed in this research is not
inconsistent with the united
front strategy; the two strategies can compatibly supplement
one another. A deep analysis of
the relationship between the two will be necessary in the future.
POSTSCRIPT
During the Upper House election on July 21, 2013, the JCP’s
vote share surged from 6,1 %
for the previous election to 9,7 %. This can be seen as the result
32. of an, albeit unintentional,
information entropy enhancement strategy of a diverse united
front of singular issues
(consumption tax, constitutional reform, nuclear power, the
Great East Japan Earthquake,
Trans-Pacific Partnership, American military bases, social
security, etc.). However, upon
analysis via the model for complex systems social change
proposed by this research, there
was almost no change in the power exponent, so this was not
enough to alter the real power
relationship of the agonistic structure. This also indicated that
the JCP’s medium range vote
share goal of 20 % should achieve a significant change in the
power relationship.
CONCLUSIONS
We constructed a complex systems ‘criticality’ model, applied it
to social change, and examined
its reliability and validity. Social change manifests in the
‘critical state’ of mutual antagonism
and conflict between the pro-change and pro-establishment
powers. The model derived a
power law distribution of the output of social change, the
characteristics of a pro-change
power, while a pro-establishment power displayed an
exponential distribution. The power
exponent reflects the agonistic structure (real power
relationship) of both powers. The
validity of the model was verified by examining vote shares of
parties in Japan. Based on the
examination, we propose a new quantitative strategy
“information entropy enhancement” for
effective change of the exponent, and therefore, for social
change.
33. REMARKS
1This hypothesis is analogous to the Boltzmann distribution, but
it is also a solution of
1dI(N)/dN = α I(N) which is often used in various growth
phenomena. Recently, it was shown
1that1the model holds when power law distribution Nα is
hypothesized. Because Nα = exp(α lnN) =
1= exp(α M) when ln N = M, all procedures in the model are
the same for M. This result
1suggests that the assumption of exponential function in
equation (1) is not essential; a model
1can potentially be realized if there exists an increasing
function.
2The rate of change R in the vote share is the absolute
difference in vote share between one
2election and the next.
I. Shimada and T. Koyama
350
APPENDIX: COMPLEX SYSTEMS ‘CRITICALITY’ MODEL
GENERAL THEORY OF THE ‘CRITICALITY’ MODEL IN
COMPLEX SYSTEMS
Criticality refers to “a state in which two powers both act
towards creating a different state,
and in so doing both achieve a state of precise balance”
(Butsurigaku Jiten in Japanese).
However, research [6] indicates that in complex systems
multiple metastable self-organized
critical states are formed. The word ‘criticality’ is used here in
34. quotes to distinguish it from
the precisely balanced criticality of simple systems physics.
Hereafter the term will be used
as-is, without quotes, to indicate criticality in complex systems,
unless otherwise noted.
The criticality model of complex systems is predicated on a
fundamental idea of complex
systems, namely that “complex system dynamics and its output
(quantitative, qualitative) are
defined by a network pattern of diverse mutual interactions
(physical, informational, social)
working among N active elements that are related to output,
within the elements that comprise
the system”. This view of complex system dynamics has come
to be shared in recent years.
However, this model is limited to the quantitative aspects of the
system, and is based on the
assumption that the output is defined by the existence
probability P(N) of the number of
network patterns I(N). In cases where a positive tendency and
opposing anti-tendency are
observed on that network, the existence probability P(N)of the
number of network patterns
I(N)expressing the positive tendency and the existence
probability Q(N) that its anti-tendency
can take, describe the two conflicting tendencies in terms of
probability theory. These do not
touch upon the details of mutual interactions between elements.
As shown in the next section,
the output R of a critical state in a system with two opposing,
mutually-antagonistic tendencies,
logically leads to a power-law distribution. As far as we know,
there are no previous examples
of power-law derivation in complex systems based on the above
assumption, and thus this
35. assumption's validity must be tested by applying it to various
complex phenomena. When
doing so, it must be acknowledged that an output system will
form in which qualitatively
different outputs will exist even for a single complex system
and each of these could in turn
manifest multiple quantitative outputs. For example, in animal
behavior, these could be the
duration of the behavior [1-3], its intensity, scale, etc.; in
earthquakes, the earthquake’s energy
or magnitude [7], duration, fault surface, etc.; in a conflict, the
number of casualties [8],
length of the conflict, etc.; and in social change, these could be
the percentage of votes taken
by the pro-change power in the present paper, participants in
protests, etc. The question of
what to capture as the output is extremely important when
examining the model, but only the
quantitative aspects of the output can be analyzed/examined.
DERIVATION OF THE CRITICALITY MODEL OF COMPLEX
SYSTEMS
Suppose N is the number of active constituent elements that
determine a complex system’s
dynamics. Assume a number of network patterns I(N) of mutual
interactions between those
elements in the following equation (and see remark 1)
( ) e NI N α∝ . (1)
• is a parameter that expresses rate of growth with respect to N
of the number of patterns
I(N). The existence probability P(• N) of the number of patterns
I(N) is expressed as,
36. 0
e
( )
e
C
N
N
N
N
P N
α
α
α
−
−
=
=
∑
, (2)
and so the largest number of active elements is NC. In the
system's critical state, the existence
probability Q(βN) of the number of patterns of anti-tendencies,
37. acting in opposition and
A theory for complex systems social change: an application of a
general 'criticality' model
351
antagonistically to this network of positive tendencies, is
expressed as a similar exponential
function, with the parameter β generally differing from α
0
e
( )
e
C
N
N
N
N
Q N
β
β
β
38. −
−
=
=
∑
(3)
P(α N) and Q(β N) are common simple function types that have
interesting and somewhat
complex characteristics. Both are monotonically decreasing
functions of N, but if one
examines the dependency on • or • , when • is sufficiently
small, P(• N) is also small
(~(NC + 1)
–1), but largely depends on N. That is, when 0 < N < NC/2, a
single extreme value
exists for a given α, and when NC/2 < N, it monotonically
decreases with respect to α
(unpublished).
Next, these two probabilities express two opposing tendencies,
and define the output system
R that is produced from the conflicting critical state; this results
from the following. Initially,
as the output R will probably decrease as anti-tendencies
strengthen, it is assumed to be
inversely proportionate to the probability Q(βN), so R is
defined by βN.
0 0
39. 0
1 1 1 e
( )
e
c
N
N
N
N
Q N
R R R
β
β
β
−
−
=
= =
∑
(4)
Moreover, if R is sufficiently large and integral (and
40. continuous) representation is employed, φ(R)
is the probability density function of R. Through variable
transformation by R of the existence
probability P(αN)dN, the probability ( )R dRφ , which expresses
the structure of the output
system R, is acquired (probability conservation). That is, given
φ(R)dR = P(αN(R))(dN/dR)dR,
d 1
e ,
d
NN
R
β
β
−∝ (5)
which follows from Eq. (4), then
)1/(0
/ )/(e)/1()e(
d
d
)()( −−−− ∝∝= βαββαβ βαφ RR
R
N
NPR NN (6)
41. Integrating this (cumulative distribution)
/0( ) ( )d ( / )R
R R R R R α βφ
∞ −Φ = ∝∫ (7)
yields a power-law distribution expressing the structure of the
complex output system R.
With regards to α and β, opposite tendencies are expressed by
the numerator and
denominator in DM = α/β. However, if β is sufficiently small,
this will become an
approximately exponential distribution:
0( ) exp( / )R R Rα βΦ ∝ − (8)
The complex system is simultaneously a complex material
system and a complex
informational system. In order to quantitatively handle the
complex system, we will introduce
the complex system network's information entropy (mean
information amount) HM to the
model. Shannon’s definition of information entropy HM is
expressed as a discrete equation,
but it can be approximately expressed as the following integral
representation.
0
d lo g( ) ( )[ ]CN M M MN P N P N Hα α− ≡∫ (9)
I. Shimada and T. Koyama
42. 352
Here, αM and P(αM N) are the parameter and existence
probability, incorporating the
information effect. If the change in information entropy of the
complex system network ∆H is
O MH H H∆ = − (10)
(H0 is the maximum value of HM at the time αM changes), then
when NC is large [3]:
HS ≈ 1 – logαS (s = 0 or M). (11)
HS is expressed by αS, therefore:
αM = α0 e∆H. (12)
Consequently
MMD
α
β
= (13)
Thus, DM is shown to change depending on ∆H. However, it
may be necessary to incorporate
the information effect in the probability Q(βN) for specific
complex phenomena.
SIGNIFICANCE OF MODEL
This model illustrates one general expression of the laws of
transformation and criticality,
which comprise the foundation of the
developmental/evolutionary processes of complex
systems. Moreover, it points to the importance of the meaning
of and changes in the power
exponent DM in complex system dynamics that express the
power-law, suggesting that in
43. these changes, the contribution of ∆H is (exponentially) large.
In fact, in the adaptive animal
behavior model [1-3], which is the basis for this model, it has
been clearly demonstrated,
through five independent experiments, that changes of the
power exponent DM depend on
amount of memory (∆H) and gustatory stimulus (β).
Up until now, it seems that many complex systems models are
used to provide feasible, a
posteriori explanations, or to tout new and different
interpretations [9]. Perhaps we are merely
ignorant of their existence, but there seem to be almost no
widely effective and potentially
predictive theories (models). In addition, there have been
almost no examples of research that
directly explore changes in the exponents of power law within
complex systems science. Rather,
the research attempts to evaluate the universality of power
exponents beyond differences of
object. In elucidating the significance of power exponent
dynamics within the same object,
this model may clarify system structural dynamics, which are at
the root of output fluctuations
in complex systems, and suggests the potential of structural
“change” and “control”.
ACKNOWLEDGEMENTS
We thank Drs. Hiroshi Hamada, Satoshi Ihara, Hitoshi Kono,
Yatsuhisa Nagano, Hiroshi Onishi
and Reiji Sugano for useful discussions.
REFERENCES
[1] Shimada, I. and Koganezawa, M.: Animal Behavior as a
Complex Adaptive System.
44. Comparative Physiology and Biochemistry 28, Suppl. (Abs.)
2011,
[2] Shimada, I.; Kawazove, Y. and Hara, H.: A Temporal Model
of Animal Behavior Based
on Fractality in the Feeding of Drosophila Melanogaster.
Biological Cybernetics 68(6), 477-481, 1993,
http://dx.doi.org/10.1007/BF00200806,
[3] Koganezawa, M.; Hara, H.; Hayakawa, Y. and Shimada, I.:
Memory Effects on Scale-
free Dynamics in Foraging Drosophila.
Journal of Theoretical Biology 260(3), 353-358, 2009,
http://dx.doi.org/10.1016/j.jtbi.2009.06.018,
http://dx.doi.org/10.1007/BF00200806�
http://dx.doi.org/10.1016/j.jtbi.2009.06.018�
A theory for complex systems social change: an application of a
general 'criticality' model
353
[4] Honjo, H.; Sano, M.; Miki, H. and Sakaguchi, H.: Statistical
Properties of Approval
Ratings for Governments.
Physica A 428, 266-272, 2015,
http://dx.doi.org/10.1016/j.physa.2015.02.047,
[5] Japanese Communist Party: Program of the Japanese
Communist Party.
http://www.jcp.or.jp/english/23rd-congress/program.html,
[6] Bak, P.: How Nature Works: The Science of Self-organized
45. Criticality.
Springer, New York, 1999,
[7] Gutenberg, B.C. and Richiter, C.F.: Frequency of
Earthquakes in Calfornia.
Bulletin of Seismological Society of America 34, 185-188, 1944
[8] Richardson, L. F.: Variation of the Frequency of Fatal
Quarrels with Magnitude.
Journal of American Statistical Association 43(244), 523-546,
1948,
http://dx.doi.org/10.1080/01621459.1948.10483278,
[9] Watts, D.J.: Everything is Obvious: Once You Know the
Answer.
Crown Business, New York, 2011.
http://dx.doi.org/10.1016/j.physa.2015.02.047�
http://www.jcp.or.jp/english/23rd-congress/program.html�
http://dx.doi.org/10.1080/01621459.1948.10483278�
Copyright of Interdisciplinary Description of Complex Systems
is the property of Croatian
Interdisciplinary Society and its content may not be copied or
emailed to multiple sites or
posted to a listserv without the copyright holder's express
written permission. However, users
may print, download, or email articles for individual use.
Egocentric Social Network Structure, Health, and Pro-
Social Behaviors in a National Panel Study of Americans
A. James O’Malley1*., Samuel Arbesman2*., Darby Miller
46. Steiger3, James H. Fowler4,
Nicholas A. Christakis1,5
1 Department of Health Care Policy, Harvard Medical School,
Boston, Massachusetts, United States of America, 2 Ewing
Marion Kauffman Foundation, Kansas City,
Missouri, United States of America, 3 Gallup Government,
Cleveland, Ohio, United States of America, 4 Departments of
Medical Genetics and Political Science, University of
California San Diego, La Jolla, California, United States of
America, 5 Department of Sociology, Harvard University,
Cambridge, Massachusetts, United States of America
Abstract
Using a population-based, panel survey, we study how
egocentric social networks change over time, and the
relationship
between egocentric network properties and health and pro-social
behaviors. We find that the number of prosocial activities
is strongly positively associated with having more friends, or an
increase in degree, with approximately 0.04 more prosocial
behaviors expected for every friend added. Moreover, having
more friends is associated with an improvement in health,
while being healthy and prosocial is associated with closer
relationships. Specifically, a unit increase in health is
associated
with an expected 0.45 percentage-point increase in average
closeness, while adding a prosocial activity is associated with a
0.46 percentage-point increase in the closeness of one’s
relationships. Furthermore, a tradeoff between degree and
closeness of social contacts was observed. As the number of
close social contacts increases by one, the estimated average
47. closeness of each individual contact decreases by approximately
three percentage-points. The increased awareness of the
importance of spillover effects in health and health care makes
the ascertainment of egocentric social networks a valuable
complement to investigations of the relationship between
socioeconomic factors and health.
Citation: O’Malley AJ, Arbesman S, Steiger DM, Fowler JH,
Christakis NA (2012) Egocentric Social Network Structure,
Health, and Pro-Social Behaviors in a National
Panel Study of Americans. PLoS ONE 7(5): e36250.
doi:10.1371/journal.pone.0036250
Editor: Matjaž Perc, University of Maribor, Slovenia
Received March 15, 2012; Accepted April 3, 2012; Published
May 15, 2012
Copyright: � 2012 O’Malley et al. This is an open-access
article distributed under the terms of the Creative Commons
Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium,
provided the original author and source are credited.
Funding: Research for this paper was supported by National
Institutes of Health (NIH) Grants R01 AG024448 and P01
AG031093 and by Robert Wood Johnson
Foundation Award #58729. The funders had no role in study
design, data collection and analysis, decision to publish, or
preparation of the manuscript.
Competing Interests: The authors have declared that no
competing interests exist.
* E-mail: [email protected] (AJO); [email protected]) (SA)
48. . These authors contributed equally to this work.
Introduction
Although egocentric network studies – wherein a subject is
asked to identify his or her social contacts and their
relationships –
have a long history in sociology, their use in health surveys is
rare.
But increased attention to the role of social networks in
medicine
suggests that a basic understanding of the structure of American
social networks and how they change may be important [1]. We
collected egocentric social network data from a nationally
representative sample of Americans in order to study the
relationship between individuals social networks and their
health
and related social behaviors. The norm in national population
surveys is to sample new individuals each year [2,3,4]. While
this
can allow statistical understanding of population-level,
egocentric
social networks structure at a fixed time, it does not allow us to
49. study the change in individual networks over time. Therefore,
here,
we also sought to characterize how individuals egocentric social
networks change over time, the extent to which close social
contacts are gained or lost, and the extent to which an ego’s
social
contacts (their ‘‘alters’’) come to know each other over time.
Obtaining social network information that is population
representative is challenging. To sample individuals who are
actually connected to each other, non-random methods are
needed, or else a set of null relationships in which no two
individuals know each other is likely [5]. On the other hand, the
necessity of sampling pairs of individuals who are connected is
at
odds with the requirements of population representativeness that
are fundamental to most surveys, for which random or
probability-
weighted random sampling is desirable. Therefore, we imple-
mented a compromise between a population-representative
survey
and a full sociocentric network ascertainment, in order to obtain
50. population representative estimates of quantities related to
networks and various behaviors.
That is, we conducted a national, egocentric study. We fielded a
network survey instrument in a nationally representative sample
in
order to study the relationship between individuals social
networks
and their health and behaviors. The instrument extracts informa-
tion from respondents (egos) on their relationships to the peers
they spend the most time with or discuss important issues with
(alters) and also on the relationships between all pairs of named
peers. Moreover, we collected this information repeatedly
across
time. Thus a respondent’s social network is (partially) revealed
in a
traditional random sample without having to ascertain the full
population network.
Using such a survey instrument, we can examine the change in
the number of close social contacts (‘‘degree’’), tie strength
(‘‘closeness’’), and the number of interconnections between
51. contacts (‘‘transitivity’’), how they are related to each other,
and
how they relate to measures of health and prosocial behavior.
We
focused primarily on prosocial behavior, broadly defined as
PLoS ONE | www.plosone.org 1 May 2012 | Volume 7 | Issue 5 |
e36250
altruism and generosity, or any activity that promotes the
general
welfare of society, e.g., participation in community enrichment,
contributions to charity, and volunteering [6].
Methods
Survey Instrument
We developed a social network survey instrument for deploy-
ment with Gallup’s ongoing, longitudinal, probability-based
panel
of American households. We constructed the survey by
modifying
the social network instrument used previously by the GSS
[7,8,9]
and by the National Health and Social Life Survey [10,11]. The
52. survey consists of items seeking information on ego
characteristics,
alter characteristics, ego-alter relationships, and the relationship
between each pair of alters. Ego health and behavioral traits are
also measured. Surveys were completed online by the egos.
The network component of the survey asked egos to name (up
to) four adults with whom they spent the most free time in the
past
12 months and (up to) four adults with whom they most often
discussed important issues. Alters could be family members,
friends, work/school colleagues, and so forth. Thus, each ego
named up to eight distinct alters yielding a maximum of 28
alter-
alter relationships for which they also provided information.
Individual characteristics included gender, age, race, ethnicity,
education, employment status, income, marital status, religious
preference, political affiliation, health, health behaviors, and
prosocial behavior. Ego-alter variables included the type of
relationship and frequency of contact, and alter-alter variables
53. consisted of type and strength of the relationship. The key alter-
alter variable is the ego’s assessment of the strength of their
relationship.
The health and behavioral traits for the ego consisted of a series
of items that, for analytic purposes, were combined into scales.
Prosocial behavior measured whether the ego donated blood,
donated money, donated clothing, financially supported a
political
candidate, or volunteered to help prepare for a major public
emergency. Health behaviors included smoking status, BMI
(weight/height
2
), whether they wanted to gain or lose weight,
whether they took active steps to improve their health (e.g.,
adhered to a diet, quit smoking, cut back alcohol). The
questions
on health included physical health, mental health, and missed
work due to sickness.
The network measures included the number of alters with
whom the ego named as having a close relationship (degree),
the
54. average strength of these relationships (closeness), and the
connectedness of the named alters to each other according to
the ego (transitivity).
Study Design
The Gallup Panel is a nationally representative, multi-mode
panel recruited through random digit dialing methods. Only the
web-based portion of the Panel was eligible for participation in
this
study. Data was collected on randomly chosen and nationally
representative Americans in June 2009, December 2009, and
July
2010.
A total of 6,000 randomly selected web-based members of the
Gallup panel were sent an email invitation that asked them to
respond to a survey about ‘‘the various people that you spend
your free time with and have important conversations with.’’
Invitations for waves 2 and 3 explained that this was a
continuation of the earlier research in which they had
participated. A reminder email was sent for each wave in order
55. to boost completion rates. The first wave collected data from a
sample of 3,232 respondents (out of 6,000– a 53.9% completion
rate). Of those 3,232 respondents, 2,305 responded to wave 2 in
December 2009 (71.3%) and 2,114 responded to wave 3 in July
2010 (65.4% of wave 1).
The analysis sample consists of all individuals that responded at
wave 1 and, therefore, includes some who have missing data in
waves 2 and 3.
Network Measures
The degree of each ego is the number of alters they name, a
quantity ranging from 0 to the maximum 8. The strength of each
ego-to-alter and alter-to-alter relationship is rated from 0 = no-
relationship to 10 = very strong relationship; 0 strength only
occurs in alter-to-alter relationships. We scaled degree and
strength to obtain quantities that range from 0 to 1; the
closeness
of an ego’s relationships is, therefore, the average strengths of
their reported ties (see Figure 1). We hypothesized that
closeness
would have greater discriminatory power than degree.
56. In egocentric networks, an ego’s transitivity is the average
value of the relationship between all pairs of alters, herein
assumed to be mutual (i.e., the relationship from j to k is the
same as that from k to j). Denote the degree of ego i by ni and
the strength of the relationship between alter j and alter k by
zjk:
When relationships are binary-valued (e.g., corresponding to the
presence of any relationship at all), transitivity is given by
transi~2n
{1
i (ni{1)
{1
X
jvk
I (zjkw0), ð1Þ
and is interpreted as the proportion of pairs of alters for which
some form of relationship exists. If relationships are quantified
in
terms of their strengths, transitivity is given by
transi~2n
{1
i (ni{1)
57. {1
X
jvk
zjk: ð2Þ
Because (2) is our preferred measure, from here on by
‘‘transitivity’’ we mean the weighted form in (2).
Egocentrically,
transitivity is a local measure specific to an ego, not a network
property.
In general, egocentric network measures are more loosely
representative of sociological constructs than their sociocentric
counterparts [12]. For example, egocentric degree does not
account for directionality and therefore cannot characterize
reciprocity (if A is a friend of B then B is more likely to be a
friend of A). Moreover, because the egocentric design used here
does not differentiate between different types of triads, it
confounds transitivity (A and B are more likely to be friends if
C
is a friend of both of them) with three-cycles (A is a friend of B
who
is a friend of C who is a friend of A) and other triadic
58. structures.
Nonetheless, egocentric network measures based on survey data
are still useful for describing how individual’s networks evolve.
Because closeness is undefined if the ego has no alters (i.e.,
degree = 0) and transitivity is undefined if they have one or no
alters (i.e., degree#1), we define them to have value 0 if
no:peeri~I (ni~0)~1 and one:peeri~I (niƒ1)~1 respectively.
We include no:peeri in any model in which closeness is a
predictor
and no:peeri in any model in which transitivity is a predictor.
We
exclude from the analysis of closeness observations with
no:peeri~1 and for the analysis of transitivity we discard
observations with one:peeri ~1:
Egocentric Social Networks and Health Behaviors
PLoS ONE | www.plosone.org 2 May 2012 | Volume 7 | Issue 5 |
e36250
Statistical Models
We fit a series of models where the ego’s health and behaviors
are regressed on the network measures, adjusting for personal
characteristics of the ego such as gender and age. In addition,
we
59. adjusted for the value of the dependent variable at the previous
wave, personal characteristics of the ego (e.g., gender, age), and
survey wave. We also fit similar models in which the network
measures are regressed on the health and behavioral (HB)
measures.
Let yit, wit, and xit be the vectors of HB (BMI, health, health
behavior, prosocial behavior), their network variables (degree,
closeness, transitivity), and exogenous control variables
(gender,
age, survey wave) for individual i at survey wave t (t~1,2,3):
Element h of yit and wit is denoted by yhit and whit
respectively.
When applicable, no:peerit and one:peerit are included in wit:
The model for health and behavioral trait h regressed on the
network measures is
E½yhitjyhit{1,wit,wit{1,xit�~b0zb1yhit{1zb2wit
zb3wit{1zb4xit
ð3Þ
and the model for network measure k regressed on the HB is
E½wkitjwkit{1,yit,yit{1,xit�~h0zh1wkit{1zh2yit
zh3yit{1zh4xit
ð4Þ
60. Both (3) and (4) allow the effects of current and lagged
predictors
to be separated and are jointly useful for detecting
directionality of
effects. For example, the scenario when b2 is not significant
whileb3 and h2 are significant is consistent with network
measures
having a causal effect on HB but not the converse. In general, if
an
element of b3 or h3 is significant, there is stronger evidence of
a
causal relationship than if the corresponding element of b2 or
h2 is
significant.
In the event that the effects of current values of the key
predictors (network measures or HB) are nearly all non-
significant, we consider a reduced model in which they are
excluded. In general, lagged models have the advantage over
cross-sectional models of being more causally defensible. How-
ever, they can lack power.
To investigate the possibility of nonlinearity, we tested whether
transitivity-squared was predictive of ego’s outcomes. However,
no
61. such effects were found and so we omitted transitivity-squared
as a
predictor. We also tested whether the turnover in the ego’s
peers,
defined as
turnoverit~f riendsadditzf riendsloseit{jnit{nit{1j,
where jxj~x if xw0 and 0 otherwise, had an impact beyond
degree and change of degree. In all instances, estimated effects
for
turnoverit were not significant.
To improve scalability of the estimated regression coefficients,
the network measures were scaled to the unit interval when
functioning as predictors (as for models of HB) and to
percentages when functioning as outcomes. In addition, age
was scaled to units of 10 years.
Results
Of the 6,000 subjects in the sampling frame, a total of 3,232
responded to the first survey. Of these, 2,305 subjects
responded
to the second survey, and of these 1,809 responded to all three
surveys. In response to the name generators (‘‘Who do you
62. spend
free time with’’ and ‘‘Who do you discuss important issues
with’’), we found that Americans identify an average of 4.461.8
close social contacts (the average respondent lists 2.2 friends,
0.76
spouses, 0.28 siblings, 0.44 coworkers, and 0.30 neighbors).
This
corresponds to past work with the GSS [9].
Figure 1. Egocentric network involving an ego with N = 8 alters
labeled A, B, …, H. For clarity, the left panel shows only the
ego-alter ties
while the right panel shows only alter-alter ties. Closeness is
computed as the average strength of the ego-alter ties (left
panel) and dividing by 10 to
make the range 0 to 1. Analogously, transitivity is computed as
the average strength of the alter-alter ties (right-panel,
including the 0 strength null
ties that are not depicted) and dividing by 10.
doi:10.1371/journal.pone.0036250.g001
Egocentric Social Networks and Health Behaviors
PLoS ONE | www.plosone.org 3 May 2012 | Volume 7 | Issue 5 |
e36250
Figure 2. Changes and transitions in network measures between
consecutive waves.
doi:10.1371/journal.pone.0036250.g002
63. Egocentric Social Networks and Health Behaviors
PLoS ONE | www.plosone.org 4 May 2012 | Volume 7 | Issue 5 |
e36250
Figure 3. Relationships between changes in network measures.
doi:10.1371/journal.pone.0036250.g003
Egocentric Social Networks and Health Behaviors
PLoS ONE | www.plosone.org 5 May 2012 | Volume 7 | Issue 5 |
e36250
Change in Network Measures
Between the six-month waves, an ego’s degree increased 31% of
the time, stayed the same 27% of the time, and decreased 42%
of
the time. The transition matrix for ego’s degree (not presented)
revealed that unchanged degree occurred most commonly and
that there was a clear tendency towards lower degree, especially
among egos that began with modest degree (2 to 4 alters).
The distributions of closeness and transitivity measures were
very similar between waves with almost equal numbers of
64. increasing and decreasing transitions (Figure 2). Closeness
increased 46%, remained the same 8%, and decreased 46% of
the time, while transitivity increased 50%, stayed the same 5%,
and decreased 45% of the time. All network measures regressed
towards the mean (higher values were more likely to decline
while
lower values were more likely to increase); degree and
transitivity
were most and least affected, respectively.
Relationships Among Network Measures
Change in degree and change in closeness have a distinctly
negative relationship (Figure 3), and people seem to trade of the
number of contacts they have with the closeness of those
contacts.
Furthermore, increased degree is associated with reduced transi-
tivity, while closeness and transitivity are positively correlated.
Thus, as an individual accumulates more alters, the average
closeness of their own relationships and of the relationships
between the alters in their egocentric network decline. These
observations illustrate that correlations often arise in social
65. networks due to natural constraints, even when there is no
mathematical constraint. For example, individuals that name
more alters may on average have ties of lower average strength
due to a limit on the number of very close relationships an
individual can maintain. Likewise, across all pairs of alters,
average
relationship strength is likely to be lower if degree is higher as
the
relative frequency of pairs with a non-close relationship
increases.
Table 1. Regressions of individual outcomes on current and
lagged network measures.
Term BMI Health Health behaviors Pro-social behavior
Estimate 95% CI Estimate 95% CI Estimate 95% CI Estimate
95% CI
Network measures
Degree 0.02 (20.83, 0.88) 0.08 (20.11, 0.27) 20.02 (20.25, 0.21)
0.33 (0.19, 0.48)
Closeness 20.14 (21.37, 1.10) 0.34 (20.03, 0.70) 0.38 (20.09,
0.86) 0.22 (20.09, 0.54)
Transitivity 20.04 (20.86, 0.77) 20.06 (20.24, 0.12) 20.19
(20.38, 0.01) 20.09 (20.23, 0.06)
66. Lag network measures
Degree 0.08 (20.71, 0.88) 0.13 (20.06, 0.32) 0.00 (20.22, 0.22)
20.08 (20.22, 0.07)
Closeness 0.14 (21.24, 1.53) 0.08 (20.32, 0.47) 20.23 (20.69,
0.24) 20.22 (20.51, 0.07)
Transitivity 0.22 (20.44, 0.87) 0.07 (20.12, 0.25) 0.20 (0.01,
0.40) 0.11 (20.04, 0.26)
Other predictors
Wave 0.26 (20.07, 0.60) 0.02 (20.05, 0.09) 0.09 (0.01, 0.17)
20.05 (20.11, 0.00)
Female 20.11 (20.32, 0.10) 0.00 (20.06, 0.05) 0.04 (20.03, 0.10)
0.00 (20.04, 0.04)
Age (10s of years) 0.03 (20.08, 0.15) 0.01 (20.01, 0.03) 0.03
(0.01, 0.05) 0.03 (0.02, 0.05)
Lag dependent variable 0.89 (0.81, 0.96) 0.78 (0.76, 0.81) 0.56
(0.53, 0.60) 0.74 (0.72, 0.77)
doi:10.1371/journal.pone.0036250.t001
Table 2. Regressions of network measures on lagged health and
behavioral traits.
Term Degree (%) Closeness (%) Transitivity (%)
Estimate 95% CI Estimate 95% CI Estimate 95% CI
Lagged health behaviors
67. BMI 0.15 (0.02, 0.27) 0.02 (20.05, 0.08) 0.00 (20.10, 0.10)
Health 0.17 (20.52, 0.85) 0.45 (0.09, 0.81) 0.11 (20.46, 0.68)
Health behavior 0.54 (20.21, 1.28) 20.26 (20.63, 0.11) 0.06
(20.57, 0.68)
Pro-social behavior 0.38 (20.52, 1.27) 0.46 (0.02, 0.91) 0.48
(20.25, 1.20)
Other predictors
Wave 20.88 (22.59, 0.83) 1.65 (0.81, 2.50) 1.83 (0.20, 3.45)
Female 5.98 (4.38, 7.59) 0.86 (0.08, 1.63) 22.18 (23.44, 20.91)
Age (10 s of years) 0.59 (0.02, 1.16) 0.11 (20.16, 0.37) 0.89
(0.42, 1.36)
Lag dependent variable 0.31 (0.26, 0.35) 0.46 (0.41, 0.51) 0.63
(0.59, 0.67)
Note: When the full model in Equation 5 was fit, (current) pro-
social behavior was highly predictive of degree (2.54
percentage-points, 1.39–3.67).
doi:10.1371/journal.pone.0036250.t002
Egocentric Social Networks and Health Behaviors
PLoS ONE | www.plosone.org 6 May 2012 | Volume 7 | Issue 5 |
e36250
Effects between Egocentric Network and Individual
Outcomes
68. Across the four HB we found that current and lagged network
measures were significant predictors in various models (Table
1),
suggesting that the model in Equation (4) was appropriate.
However, in the models of the three network measures
(Equation
5), significant effects were more common for the lagged than
the
current values of the HB (the strong association of prosocial
behavior with degree is the lone exception). Therefore, to
simplify
interpretation, in these models we excluded the current HB
predictor variables and refit them (Table 2).
Table 1 shows that individuals who had more friends (higher
degree) were likely to behave more prosocially (estimate: 0.33;
CI:
(0.19, 0.48)). Because we re-scaled degree to the unit interval to
improve the scalability of estimates, the interpretation of the
coefficient is that 0.33 more prosocial behaviors are expected
for
every 8 friends added (approximately 0.04 more behaviors per
69. friend added). Furthermore, when the roles of degree and
prosocial behavior were reversed, we also found a strong effect
(2.54 percentage-points, 1.39–3.67), the only instance of a
network
measure having a significant effect under the full model
(Equation
5). In both models, the lagged effects are weak, suggesting that
the
change in the predictor is more important than its level. These
results imply that the relationship between degree and prosocial
behavior is bidirectional, fast-acting, and substantial.
Although the association of lagged transitivity with health
behaviors is just significant at the 0.05 level (Table 1), it is of
similar magnitude and opposite in sign to that of transitivity.
Therefore, the overall effect of transitivity is almost entirely
accounted for by change in it, not its level.
In the model of health, the coefficients of the degree terms are
positive (0.08 and 0.13) and the confidence interval of lagged
degree ({0:06, 0.32) overlaps 0 by a small amount (Table 1).
After
70. excluding degree, lagged degree had a significant association
with
health (0.18, 0.03–0.34) suggesting that the collinearity of the
degree terms reduced the level of lagged degree in the original
model. However, the association of lagged health on degree is
not
significant (Table 2) suggesting that the relationship of degree
to
health is directional (i.e., not reciprocated) and perhaps varying
over time.
We discovered several other directional effects of modest
significance (Table 2), including lagged health on closeness
(0.45,
0.09–0.81) and lagged prosocial behavior on closeness (0.46,
0.02–
0.91). Therefore, a unit increase in health is associated with an
expected 0.45 percentage-point increase in average closeness,
while adding a prosocial activity leads to a 0.46 percentage-
point
increase in the closeness of one’s relationships.
The results for the other predictors reveal that age has the
71. strongest association with the outcomes (positive effects on
health
behaviors and prosocial behavior) and network measures
(positive
effects on degree and transitivity). As revealed in the
exploratory
analyses, degree decreased over time while both closeness and
transitivity increased. It is interesting to note that females
named
nearly 6% more friends on average and had stronger
relationships,
but the relationships among their alters were weaker.
Discussion
In addition to providing a quantitative description of the
changes in individuals’ social networks over time in a national
sample, we uncovered interesting relationships among the
egocentric network measures, including the negative
dependence
between degree and either closeness of ties or transitivity
among
alters. Such dependencies would seem to be a consequence of
individuals’ limited capacity to maintain large numbers of
72. close-
ties: as our networks become larger, each tie we have to others
is
expected to weaken. To our knowledge, the phenomena that the
greater the number of ties, the less well ones alters know each
other, has not been reported previously.
Just as there are certain cognitive limits to the number of
individuals one can have as part of one’s social network,
[13,14] it
also appears that there are cognitive and temporal
considerations
for how humans manage their interactions. In particular, we find
that the reported average closeness to all friends decreases as
the
number of one’s friends increases, suggesting an invariant total
expenditure on social interaction. An increase of one in the
number of close social contacts was associated with a decrease
of
0.03 in the average closeness of each individual contact on a
scale
where 0 = do not know and 1 = extremely close. An increase of
73. two close contacts was associated with a decrease in closeness
of
nearly 0.06 (a substantial reduction on this scale). Because, in
prior
research, ties are typically modeled as either present or absent,
with no strength information, these findings are some of the
first of
their kind.
In addition, we evaluated a series of regression models relating
individuals’ egocentric network measures to their health and
behavioral traits. The strongest result we found is the
bidirectional
association between a respondent’s degree and their prosocial
behavior (Figure 4). Moreover, when current network measures
were excluded, degree was associated with health, while the
reverse is not true (i.e., health was not associated with a
subsequent
change in the number of friends); this result is consistent with
prior
work [15]. In addition, being healthier and more prosocial was
associated with the development of closer relationships, but the
74. converse was not observed.
Although our principal finding that degree and prosocial
behavior are highly predictive of each other is a
contemporaneous
effect, and thus open to more scrutiny from a causal standpoint
than a lagged association of the same significance, we
nonetheless
raise the possibility that this provides evidence of some effect.
Whether degree affects prosocial behavior or the converse, or
Figure 4. Flow diagram illustrating the primary effects found
between network measures and HB. The strongest effect is the
contemporaneous bidirectional effect between number of friends
(degree) and prosocial behavior (solid line) while the lagged
directional
effects (dashed lines) were weaker but still statistically
significant.
Relationships within network measures and within HB are not
depicted.
doi:10.1371/journal.pone.0036250.g004
Egocentric Social Networks and Health Behaviors
PLoS ONE | www.plosone.org 7 May 2012 | Volume 7 | Issue 5 |
e36250
whether both occur simultaneously, is a question that may only
be
75. answerable by a study in which subjects are even more closely
followed over time. Across the pairs of individual outcomes and
network measures, there were no other bidirectional
associations
with such a high level of significance.
Considering all results together, the only network measure that
consistently predicted health status and also one’s health and
prosocial behavior is degree, while both health and prosocial
behavior predicted network measures (closeness and degree
respectively). These results are consistent with a circular
evolution
or feedback mechanism among one’s network and one’s health
or
behavior, possibly reflecting a process that for some or all
individuals may tend to an equilibrium position.
Although an egocentric study provides a feasible way of
obtaining network information on a large-scale (e.g., by
augment-
ing an existing national survey), it has several limitations
compared
76. to a full sociocentric study. First, unlike sociocentric studies,
we do
not observe the status of the relationship from the perspective
of
alters. Egocentric data is based purely upon the knowledge,
reflection, and recall of the ego, which may be inaccurate –
especially when describing the relationship between two alters.
Because we do not observe alter’s view of relationships, we are
not
able to validate the ego self-report, nor are we able to ascertain
directionality of relationships [16]. This prevents us from
utilizing
directionality to distinguish the effect of induction from other
phenomena, such as homophilly [17,18,19]. However, the
analysis
is still relevant if we consider the fact that an ego’s perception
of
relationships may be more important than whether or not the
perceived relationship is validated via reciprocation by the
alter.
Second, without some means of validating the relationship
between pairs of alters, relationships elicited via surveys are
77. prone
to confound true relationship status with recall – it is likely that
relationships with whom an individual has recently been in
contact
are the most likely to be reported. Therefore, changes in the size
or
composition of a network may be observed due to changes in
the
set of alters an individual recently interacted with, even if their
true
egocentric network is unchanged [20,21,22]. If measurement
error
is random, estimated effects will be attenuated (biased towards
0).
Thus, we are more likely to not report a true effect (type II
error)
than to claim a significant effect when the true effect is 0 (type
I
error). Such problems with self-reported relationships also
apply to
sociocentric designs if tie-ascertainment is through self-report.
An
alternative to recall is to evaluate relationships by monitoring
78. human interactions [23]. Although monitoring might be a more
precise way of measuring relationships, or at least to generate a
list
of potential alters to use in an interview with the ego, in
practice
such monitoring is likely to involve enormous cost. Clearly, an
important area for future work is to study the properties of
egocentric networks when relationships are not completely
ascertained due to incomplete recollection of ties.
Another limitation is that online surveys have been found to
yield less accurate results than in-person interviews, as
individuals
may answer more mechanically, particularly in completing the
response matrix of the relationships between their alters [24].
However, an advantage of performing the survey online is that
the
expense is much lower and there is no risk of bias from
interviewer
effects.
Egocentric network studies embedded within population surveys
enable population-representative data to be obtained on individ-
79. uals’ relationships. It is clear that, despite the advantages that
sociocentric studies offer, they are not practical on a national
scale
(except, perhaps, if phone data for a whole nation were used).
An
understanding of social network structure and its relationship
with
health and health behaviors can improve understanding of health
phenomena such as collateral effects [25], design of healthcare
interventions, and evaluation of healthcare policy studies [1].
Egocentric studies, such as that described here, can also provide
information about social networks and how they change in ways
relevant to health at the national level.
Acknowledgments
We thank Donald Beck, formerly at Gallup and now at Booz
Allen, for
help with the initial study execution.
Author Contributions
Conceived and designed the experiments: SA NAC JF.
Performed the
80. experiments: DMS. Analyzed the data: AJO. Contributed
reagents/
materials/analysis tools: AJO SA. Wrote the paper: AJO SA
NAC.
References
1. Smith KP, Christakis NA (2008) Social networks and health.
Annual Review of
Sociology 34: 405–429.
2. Smith TW, Marsden PV, Hout M, Kim J (2011) General
social surveys, 1972–
2010: cumulative codebook. Chicago: National Opinion
Research Center:
National Opinion Research Center. 3610 p.
3. Laumann EO, Gagnon JH, Michael RT, Michaels S (2008)
National Health
and Social Life Survey, 1992: [United States]. Inter-university
Consortium for
Political and Social Research (ICPSR) [distributor].
4. Harris KM (2009) The National Longitudinal Study of
Adolescent Health (Add
Health), Waves I & II, 1994–1996; Wave III, 2001–2002; Wave
IV, 2007–2009
[machine-readable data file and documentation]. Chapel Hill,
NC: Carolina
81. Population Center, University of North Carolina at Chapel Hill.
5. O’Malley AJ, Marsden PV (2008) The Analysis of Social
Networks. Health
Services and Outcomes Research Methodology 8: 222–269.
6. Eisenberg N, Mussen PH (1989) The Roots of Prosocial
Behavior in Children.
Cambridge: Cambridge University Press.
7. Burt RS (1984) Network items and the General Social
Survey. Social Networks
6: 293–339.
8. Marsden PV (1987) Core Discussion Networks of Americans.
American
Sociological Review 52: 122–131.
9. McPherson M, Smith-Lovin L, Brashears ME (2006) Social
Isolation in
America: Changes in Core Discussion Networks over Two
Decades. American
Sociological Review 71: 353–375.
10. Laumann EO, Gagnon JH, Michael RT, Michaels S (1994)
The Social
Organization of Sexuality: Sexual Practices in the United
States. Chicago:
82. University of Chicago Press.
11. Laumann EO, Gagnon JH, Michael RT, Michaels S (1995)
National Health
and Social Life Survey Computer file [ICPSR version].
Chicago: University of
Chicago and National Opinion Research Center [Producer]. Ann
Arbor, MI:
Inter-University Consortium for Political and Social Research
[Distributor].
12. Marsden PV (1992) The Reliability of Network Density and
Composition
Measures. 1–29 p.
13. Fowler JH, Dawes CT, Christakis NA (2009) Model of
Genetic Variation in
Human Social Networks. PNAS: Proceedings of the National
Academy of
Sciences 106: 1720–1724.
14. Dunbar R (1992) Neocortex size as a constraint on group
size in primates.
Journal of Human Evolution 22: 469–493.
15. House JS, Landis K, Umberson D (1988) Social
relationships and health.
Science 241: 540–545.
16. Wellman B (2007) Challenges in collecting personal
network data: The nature of
83. personal network analysis. Field Methods 19: 111–115.
17. Christakis NA, Fowler JH (2007) The Spread of Obesity in a
Large Social
Network over 32 Years New England Journal of Medicine 357:
370–379.
18. Christakis NA, Fowler JH (2008) Dynamics of Smoking
Behavior in a Large
Social Network. New England Journal of Medicine 358: 2249–
2258.
19. Fowler JH, Christakis NA (2008) Dynamic spread of
happiness in a large social
network: longitudinal analysis over 20 years in the Framingham
Heart Study.
British Medical Journal. pp doi:10.1136/bmj.a2338.
20. Cooper J, Herring L (2009) Sociologists Debate: Are
American’s Really Isolated?
Available: http://www.asanet.org/press/20090804.cfm. In:
Association AS,
editor. ASA Press Releases: American Sociological Association.
21. Feld SL, Suitor JJ, Gartner Hoegh J (2007) Describing
Changes in Personal
Networks over Time. Field Methods 19: 218–236.
22. Lubbers MJ, Molina JL, Lerner J, Brandes U, Ávila J, et al.
(2010) Longitudinal
84. analysis of personal networks. The case of Argentinean
migrants in Spain. Social
Networks 32: 91–104.
Egocentric Social Networks and Health Behaviors
PLoS ONE | www.plosone.org 8 May 2012 | Volume 7 | Issue 5 |
e36250
23. Marsden PV (2005) Recent Developments in Network
Measurement. In:
Carrington PJ, Scott J, Wasserman S, eds. Models and Methods
in Social
Network Analysis. New York: Cambridge University Press. pp
8–30.
24. Matzat U, Snijders C (2010) The Online Measurement of
Ego Centered Online
Social Networks. Social Networks 32: 105–111.
25. Christakis NA (2004) Social networks and collateral health
effects. British
Medical Journal 329: 184–185.
Egocentric Social Networks and Health Behaviors
PLoS ONE | www.plosone.org 9 May 2012 | Volume 7 | Issue 5 |
e36250
85. Copyright of PLoS ONE is the property of Public Library of
Science and its content may not be copied or
emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission.
However, users may print, download, or email articles for
individual use.
‘Keeping meself to meself’ – How Social Networks Can
Influence Narratives of Stigma and Identity for Long-term
Sickness Benefits Recipients
Kayleigh Garthwaite
Department of Geography, University of Durham, Durham, UK
Abstract
This article focuses upon social networks and their relationship
to stigma and identity for long-term
sickness benefits recipients in the North East of England.
Drawing on empirical qualitative research
with long-term sickness benefits recipients, this article
demonstrates how the co-construction of
stigma is fundamental in shaping how long-term sickness
benefits recipients participate in social
networks with friends, family and the community. The findings
support the idea that the stigma of
receiving benefits can be contrasted with nostalgia for the social
elements of employment. Utilizing
the work of Goffman, the article focuses on how the stigma and
86. shame felt at receiving sickness
benefits for an extended period of time interacts with social
networks and identity. Reluctance to
disclose a claimant identity to friends and family could lead to
social isolation and a perceived need
to ‘keep meself to meself’ which can be linked to a wider
rhetoric surrounding benefits recipients that
characterizes them as ‘scroungers’.
Keywords
Social networks; Sickness benefits; Stigma; Identity;
Neighbourhood; Scrounger
Introduction
Since 2008, the UK has been experiencing a period of welfare
reform and
austerity which has caused increasing stigma, shame and
uncertainty for many
sickness benefits recipients. Briefly, Employment and Support
Allowance
(ESA) was initially introduced by Brown’s Labour Government
in 2008, and
saw the attachment of work-related conditions to the receipt of
sickness benefit
(DWP 2008). The UK coalition Government adopted this
approach, and
under the ESA regime, new claimants must undergo the Work
Capability
Assessment (WCA), a health capacity test to determine their
fitness for work.
From April 2011, those claiming Incapacity Benefit (IB) started
to undertake this
assessment. Ongoing reform has, for example, led to research
that has discussed
88. In the UK, the popular media have contributed significantly to a
hardening
of attitudes to welfare recipients in recent years, characterizing
benefits recipi-
ents as ‘scroungers’, ‘lazy’, ‘workshy’ and ‘fraudsters’. The
accompanying
policy shifts from an emphasis on universalism to one on
conditionality and
selectivity has reaffirmed this (Golding and Middleton 1982;
Garthwaite 2011;
Horton and Gregory 2009; Sefton 2009). Drawing on data
collected during a
qualitative study of long-term sickness benefits recipients in the
North East of
England, this article is particularly interested in how narratives
of those
receiving long-term sickness benefits are influenced and shaped
by social
networks in the form of friends, family, communities and
employment, and
how this relates to stigma and identity.
It can be argued that a stigma is essentially an attribute of the
stigmatized
person. A stigma is a mark of disgrace. The mark may be a
physical one, or
it may be something which attaches to the person, like a stain or
taint.
Goffman (1963) at first refers to stigma as ‘a failing, a
shortcoming, a handicap’
(Goffman 1963: 12); ‘an attribute that is deeply discrediting’
(Goffman 1963:
13); ‘an attribute that makes him different from others . . . and
of a less
desirable kind’ (Goffman 1963: 12); and ‘a shameful
differentness’ (Goffman
89. 1963: 21). Goffman goes on to say that, ‘a stigma . . . is really a
special kind of
relationship between attribute and stereotype’ (Goffman 1963:
14). These
definitions present stigma as a personal flaw – and one which
can be likened
to the rhetoric surrounding benefits recipients as a result of
media and gov-
ernment discourse. Using Goffman’s (1967) notion of stigma
management
including ‘saving face’ and presenting an ‘idealized self’
(Goffman 1959), this
article goes on to illustrate the different arenas within which
stigma is
co-constructed and how people receiving long-term sickness
benefits are
acutely aware of its potential emergence in everyday social
interaction
(Goffman 1963). In response, participants attempted to avoid
stigma at all
costs, by withdrawing from social interactions which might
expose their claim-
ant status or reveal to friends and family the extent of their
health problems,
leading to a compromising of their social networks.
Methods
The research presented here is based on doctoral research which
was
attached to a wider project involving a longitudinal survey of
the health of
SOCIAL POLICY & ADMINISTRATION, VOL. 49, NO. 2,
MARCH 2015
91. Return to Work Credit and enhanced In-Work Support. Initial
contact with
participants was forged following attendance at the Choices
events in venues
such as local colleges, community centres, and leisure centres.
JCP stated
there was no compulsion for people to attend, and as the events
were not
mandatory, non-attendance would not impact upon someone’s
benefits
receipt.
Purposive sampling was used to recruit 25 chronically ill and
disabled
people (15 women and ten men) who were interviewed between
March 2011
and August 2011, with the majority of interviews taking place
in participants’
own homes. Importantly, participants involved in the research
were all long-
term IB recipients and were predominantly yet to undergo the
WCA so
therefore had not been migrated onto ESA or Jobseeker’s
Allowance at the
time of the fieldwork. This should be kept in mind when
references are
made to IB or Disability Living Allowance throughout this
article. Inter-
views lasted between 45 and 120 minutes and were transcribed
verbatim and
fully anonymized before thematic analysis was undertaken. The
age range
of the sample varied from 32 to 63. Only two participants
reported growing
up with health problems which were musculoskeletal in nature.
Diagnoses
93. The importance of social networks – family and family
The importance of friends and family was a common theme
throughout the
narratives. Whilst for some, the support of those in their social
networks was
crucial in terms of their daily coping, for others, friends and
family were shut
out by participants who preferred to keep their health and
illness narratives to
themselves, often due to the stigma of being a benefits
recipient.
Case study: the Wellington Men’s Group
This discussion can be strengthened by looking at a case study
example of the
Wellington Men’s Health Group. Originally set up through
CMP, every
Monday afternoon men with health problems in the Wellington
area meet up
to chat, tend to their allotments, plan what training courses they
would like to
do, arrange day trips and discuss any problems they may be
facing, whether
that may be in terms of health, benefits or other concerns. At
each group,
approximately eight to ten men typically in their 40s and 50s
attend each
week. Of particular importance here is the geographical work of
Gesler (1992,
1993) on the notion of ‘therapeutic landscapes’. Based on an
understanding of
the ways in which environmental, societal and individual factors
can work
94. together to preserve health and well-being, Gesler suggests that
certain envi-
ronments, in this case allotments, promote mental and physical
well-being.
Gesler’s concept suggests that specific landscapes not only
provide an identity
but can also act as the location of social networks, providing
settings for
therapeutic activities. Furthermore, Milligan and colleagues
(2004: 1787)
discuss the importance of allotments and comment that such
communal
activity can have a positive impact upon health and well-being,
but that, ‘the
benefits arising from the social interaction inherent within such
communal
gardening activity also have a powerful potential to address the
UK govern-
ment’s social exclusion agenda’. These explanations fit neatly
into the narra-
tives of the three members of the group who were interviewed –
Shaun, Fred
and Ray – with all of them speaking of the significance the
group has had in
their lives. Fred, 53, had been receiving IB for over eight years.
He used to be
in the Army and had ‘worked all of his life’ until polyarthritis
left him unable to
continue being employed. Fred was referred to the group
through CMP five
years ago. For Fred, the group not only allowed him to enjoy
social activities
such as day trips, but was also a source of information and
support, ‘They may
have experienced something I haven’t like with the benefits
office and they can advise me. I’ve
96. ‘I’ve got the support group and I tried to talk to them and they
said they see me as the
one who sorts problems out. It’s me strength that’s kept me
going all these years and I
just feel like I’m running out of strength. They elected us
chairman and I didn’t even
want to be elected, so I feel I’ve got a responsibility now when
really I can’t face it.’
Fractured relationships
Many participants spoke about how their relationships with
family and friends
had altered following their transition onto sickness benefits,
characterized by
a change in identity. When asked about friends and family,
Mick said:
‘I do miss socialising a lot, I can’t do what I used to do but life
goes on, friends come
to see me as well, we have a chinwag but that friendship is
different. The identity of
the friendship has changed ’cos I can’t do the things I used to
do with them, the daft
things we used to do, play football and we still have the same
laughs and things but
at work that history of all the daft things that happened, that’s
sort of slowly
evaporating, those stored memories. Even though I’ve got
friends the visits aren’t what
they used to be.’
Nostalgia for a past identity was a theme which united the
narratives. Mick
spoke about his feelings of a loss of self and identity in relation
to his friends
98. ‘I think sometimes rightly or wrongly if I’m saying to the
family “Me hands are bad”
I think they must think “Oh she’s off again” and I don’t know
whether they do but I
think they must think I always complain. I dunno I’ve never
actually asked them but
I’m sure they must get sick of us saying can you do this, can
you do that. They shouldn’t
have to be doing it. Like asking Catherine [daughter] to put me
socks on, fasten me bra
or put me knickers on up to here so I can pull them up – it’s
embarrassing. I know she’ll
do it but she shouldn’t have to and that hurts.’
These extracts suggest that suffering chronic illness can serve to
isolate and
separate people from their social networks, which could have a
damaging
effect upon their health; similar sentiments can be found in the
work of
Gallant et al. (2007) on family and friends in relation to chronic
illness
management.
Others such as Sandra chose not to fully share their problems
with family
and friends. Concealing identities and controlling information
meant not only
deciding who can be given information about their illness, but
also how much
and what information they would be given, thereby employing a
form of
stigma management (Goffman 1963). Just as there was an
99. avoidance of accept-
ing the term ‘disabled’, the stigma of receiving sickness
benefits could be so
overwhelming that people refused to admit they were receiving
it (Garthwaite
2013). In some cases, interviewees refused to reveal their
‘claimant identity’ to
close family and friends, and would avoid social situations to
avoid being
asked the question. Sandra, 45, was involved in a car accident
30 years ago
which left her with spinal problems, and has since developed
gastric problems
alongside secondary mental health concerns. Sandra had
received sickness
benefits for 12 years but had not revealed this to anyone other
than her
husband, the relevant authorities and myself. Sandra described
how friends
and family can fail to understand the complexities of sickness
and disability –
something made even more difficult given the fact that Sandra
refused to
disclose her long-term sickness benefits recipient status:
‘I bumped into a friend who I hadn’t seen for 30 years and she
asked if I was working
and when I said no, she was like “Oh I wish I could be a lady of
leisure, I wish I
had nothing to do all day” and I thought you haven’t got a clue.
It’s like my sister she
works full time and I said to her I would love to be earning
£300 a week, getting a
pay packet, earning money – I would love to be in her shoes.
But like I say they don’t
understand why I’m not working, they know I have back
101. Studies have emphasized the continued existence of strong,
local social ties
within disadvantaged neighbourhoods in diverse locations
including the UK,
Ireland and Australia (Gosling 2008; Leonard 2004; Olagnero et
al. 2005;
Warr 2005). These interactions can provide practical help
(Gosling 2008;
Warr 2005) as well as a sense of attachment and belonging to
place (Robertson
et al. 2008). Interestingly, when asked about their local area,
very few partici-
pants reflected upon the history or the importance a place can
have upon
health. Instead, the answer people gave when asked about the
area was the
same time and time again – ‘I keep meself to meself’. This
could be linked to wider
feelings of shame and guilt related to receiving sickness
benefits – as the
findings presented here and elsewhere (Garthwaite 2013)
suggest, people can
be reluctant to reveal a ‘claimant’ identity to friends and
family, so ‘keeping
meself to meself’ can be perceived as an extension of that when
thinking about
place and community. A clear distinction between identifying as
‘deserving’
benefits recipients and those in the area who they perceived as
‘undeserving’
was apparent in the narratives. Angie, 50, had been receiving
sickness benefits
for over seven years following a serious car accident which led
to both physical
and mental health problems. She initially spoke of her
102. perception that many
people were receiving benefits in her neighbourhood, yet when
she reflected
on her comments, she realized that may not be the case:
‘Oh gosh yeah, even if they’re not supposed to be. The girl who
was living next door
she’s gone now but she was working a couple of jobs and then
she was claiming as well
and she got caught but I mean . . . although Amanda next door
has jobs, the house at
the end Stephanie she goes cleaning, Sally works with
handicapped kids, next door they
both work, the next door I think they work so . . . maybes
y’know there’s not that many.
When you sit and think about it, maybe there aren’t many on
benefits here so it might
not be that bad. But like I say I tend to keep meself to meself.’
The importance of community was alluded to by several
participants in the
study, such as Linda and Mick, as shown in these extracts
below. Linda, 54,
had physical health problems which she attributed to working in
factories for
many years, together with mental health problems that
developed following
her exit from the labour market. Linda said, ‘I like getting
outside, getting out in the
back lane when someone’s out. We’ve had some laughs up here
it was all community, a hell
of a community. Like I say we always have little bonfires,
parties . . . its great up here when
it’s like that’.
Welfare and the neighbourhood
104. number of residents
in his study of disadvantaged neighbourhoods articulated a
desire to ‘keep
themselves to themselves’. Crisp (2013) explains that
tendencies to regulate
contact with neighbours was expressed in terms of choice which
can be seen
as fitting into the ideas of ‘community unbound’. This term
refers to broad
changes in the social and economic structure have reduced
reliance on
neighbours and encouraged a ‘privatization of community’
(Blokland 2003)
which includes a growing preference for more intimate networks
of family and
friends.
On occasions, but not often, participants did talk about how the
decline of
the local labour market in County Durham and the North East
had an impact
upon their narratives. For example, Linda, explained how she
felt her job
prospects were being restricted and why, ‘I couldn’t work in a
shop, petrol stations
aren’t the same, I haven’t done anything else. All I’ve ever done
is work in a factory since
leaving school. There is no factories they’re all shut, every one
I’ve worked in has closed down,
every single one’. Joan, 52, reflects upon how the area has
changed since it ceased
to be a working pit village, ‘It’s not as lively an area as it used
to be and there’s clubs
closing down, there’s not a lot of shops open now, the library’s
gone it’s now a car park’.
106. benefits. This transition from paid employment was also
instrumental in
shaping current identities (Garthwaite 2015). Jennifer, 56, and
her husband
were both receiving sickness benefits. Jennifer had arthritis
alongside severe
mental health problems and a host of other physical health
concerns and had
been receiving sickness benefits for 12 years. Jennifer said, ‘I
would love to work,
it’s like you if it happened to you you’d think “I’m stuck what
am I gonna do?” I bet when
you have holidays you get frustrated and want to be back at
work. It’s social, socialising and
we haven’t got that no more’. Talking about the importance of
work to her, Linda
was enthusiastic about how ‘the craic’ or social side of working
in a factory was
appealing to her:
‘It was very important, I loved it. The girls, the craic, we had a
hell of a laugh. Music
on all day, singing, dancing, carrying on . . . it was one big
laugh from start to finish.
There’s nothing like working in a factory I loved it, it was a
blast. As long as you got
your work done it didn’t matter what you were doing, as long as
you kept that line
going. I loved it.’
Angie’s interview revealed a similar sentiment. For her, work
was impor-
tant due to the social aspect that accompanied it:
107. ‘I loved to work. I worked in the doctors we were all friends I
had meals out, things
like that. You know what it’s like, you work. We used to go to
London together, things
like that and [when you come out of work] you lose everything,
you lose your friends,
you lose your job which I loved me job, I love people working
with people and I just
loved it all, I really did.’
Both Jennifer and Angie were keen to stress how as the
researcher, I am
employed and would, like them, miss the social aspect of work
if it was absent.
This again reinforces the stigma they felt at being ‘discredited’
(Goffman 1963)
and having to claim for sickness benefits. Kirsty, 33, a prison
officer for ten
years until an accident at work left her with permanent spinal
problems, spoke
of her concerns over the absence of work within her identity:
‘The first question people always ask you after your name is
“What do you do?” and
it kind of defines you. I usually just say to people “I don’t, I
retired when I was 30”
and they give you a double take and wonder what the heck
you’re going on about but
yeah it does define what you do. People look at you and think
“There’s bugger all wrong
with you”. I’ve had that conversation so many times with people
and you’re having to
justify why you don’t have a job. I would rather be able to turn
around and say
anything really rather than that.’