THE COMPONENTS OF STRONG TIES AMONG HOMELESS

Social Networks
North-Holland
9 (1987) 37-47
37
THE COMPONENTS OF STRONG TIES
AMONG HOMELESS WOMEN
J. Clyde MITCHELL
NujfieId College, Oxjord *
The sociological imagination seems to have been captivated by the
engaging paradox of “the strength of weak ties” almost to the exclusion of ~nsideration of what is nevertheless surely the most significant
aspect of social relationships - the role of strong ties in everyday life.
Yet as Marsden and Campbell (1984) point out, when Granovetter
(1973, 1982) was developing the idea of “weak ties” he was reasonably
clear about what the characteristics of strong ties were, so it was not
necessary for him to pursue the matter in any detail. Recently Marsden
and Campbell (1984) using data previously used by Laumann (%73),
Marsden and Laumann (1978) and by Verbrugge (1977) have pursued a
sophisticated piece of analysis from which they conclude that: a
measure of “closeness” or the emotional intensity of a relationship, is
on balance the best indicator of the concept of tie strength among those
available to us (1984: 498). Conflation
of this point is available from
very different material using different methods of analysis.
The data
The data in question were derived from a study of a small sample of
women who were the focus of attention by the Homeless Families
Social Work Team in Manchester ‘. Under the British Housing (Homeless Persons) Act of 1977 Local Authorities are obliged to provide
housing from the stocks of public housing managed by their Housing
* 25 Staunton Road, Headington, Oxford, United Kingdom.
1 Data collection was made possible by a small grant from the Nuffield Foundation, London. The
research could only have been conducted with the cooperation of the social work team attached
the the Homeless Families Unit in the city in which we worked. I am grateful to Dr Elinor Kelly
who was my close associate and colleague during this research.
0378-8733/87/%3.50
@ 1987, Elsexier Science Publishers B.V. (North-Holland)
38
J. C. Mitchell
/ Strong tm among homeless women
Departments to families who are deemed to have become homeless
through no fault of their own. A Social Work Team in another
department had the responsibility of providing other services for homeless families. This team over the years had become aware that rehousing was sometimes unsuccessful because people were rehoused in areas
remote from their kin and other sources of social support. The prima
facie significance of social relationships in these families was indicated
by the fact that some 85 percent of the families who presented
themselves for housing help to the Housing Department had declared
that the ostensible reason for their having become homeless was a
dispute with a spouse or kinsperson. The social workers approached Dr
Elinor Kelly of the Extra-mural Studies Department, Manchester University, for help in researching the social links of homeless families: Dr
Kelly had previously conducted research, in which network techniques
had been used, among a number of homeless young people in the same
city.
In the event we were able to secure a small grant from the Nuffield
Foundation to enable us to collect data from a small sample of families
who could be contacted through the Homeless Families Social Workers.
Homeless families can approach a number of agencies to obtain
housing and other services but the large majority are catered for by the
Local Authority. Women who have been physically assaulted can
secure help in a number of Women’s Aid refuges in the city. Others in
less urgent need were able to obtain help directly through housing
associations. The women in our sample were drawn from all three of
these sources who cooperated with us in giving us access, on conditions
of strict confidentiality, to families on their housing waiting lists.
The research reported here is based on interviews we carried out
with 10 women who had been homeless and who were rehoused with
their families during our fieldwork. The sample of women can in no
way be taken to be random or even representative of any wider
population because so many factors beyond our control affected our
access to the families among whom we wanted to work. However,
among the 10 women several different reasons for homelessness and
also different household structures were represented. The data we
present here are drawn from our second round of interviews conducted
after the women had been rehoused.
The focal point of our interest was the extent to which people
undergoing the crisis of homelessness were able to secure personal
J.C. Mitchell / Strong ties among homeless women
39
support from others to whom they were connected. The significance of
interpersonal relationships in homelessness is indicated by the fact that
such an enormous proportion of those applying for aid from the
Housing Department had instanced “disputes with spouses or relatives”
as the reason why they were homeless at the time that they applied for
help. We should not assume, of course, that the reason given to an
administrative officer at the time of interview is a reliable indicator of
the “cause” of homelessness. Issues such as unemployment, poverty
and similar structural circumstances were usually crucial elements in
the situation. Disputes with a spouse or kinsperson were rather the
precipitating, and no doubt the most immediately influential factor
from the point of view of the person applying for aid.
We decided that we could best approach the issue of the extent to
which people currently homeless were involved in’ supportive relationships, or otherwise, by systematically recording what they considered
their own networks to be at the time when they became homeless. To
this end we decided to record details on five different aspects of social
relationships. These were:
(1) Frequency of contact: i.e., whether the respondent saw the person
daily, weekly, monthly, intermittently, rarely or never.
(2) Self-defined “closeness”: i.e., which of the persons in the network
the respondent classified as “most close”, “least close” or in
between the two extreme categories.
(3) Sociability or conviviality: i.e., whether the respondent met the
persons for sociable or convivial purposes, e.g. for entertainment,
for a drink etc.
(4) Emotional support: i.e., the extent to which the respondent could
expect to get emotional support from or could expect to give
emotional support to the persons in question.
(5) Practical aid: i.e., the extent to which the respondent could expect
to give practical aid to or receive practical aid from the person, as
for example help with shopping, looking after the children etc.
The procedure we adopted was to note carefully the identities of all
people mentioned by the women in their first interview, when they were
questioned about their family, housing histories, employment histories
and the factors which had led to their becoming homeless. We checked
this list of persons in the second interview and added to it the kin who
were identified in questions about the women’s genealogies. For purely
40
J.C. Mitchell
/ Strong ties among homeless women
practical reasons we arbitrarily limited the number of persons about
whom we enquired to 20. On occasions the women were so isolated that
they did not mention even 20 persons as significant in their lives
whereas with other respondents, our interviewer was forced to make
rapid judgements about where to concentrate because there were so
many family and friends to handle. As it was, even this limit of 20
proved to be onerous and tedious for the respondent and occasionally
embarrassingly demanding for the interviewer to complete. The respondent was asked to estimate the impact of five aspects of social relationships on all possible pairs in the set of twenty persons. Hence for
“frequency” she was asked to estimate the frequency of contact each
person had with her, i.e. “daily contact”, “weekly contact” etc. When
this was completed for all persons in relation to herself, she was asked
about the frequency of contact between each person and all others.
Thus ultimately the procedure was carried out for all five aspects of
relationship in respect of all members in her network. For a person
with the maximally sized network of twenty, it follows that the respondent was asked to make 20 * 19 * 5 = 1900 judgements.
The data, of course, relate to the respondent’s own construction of
the characteristics of the relationships amongst the people in her
network, and should not be taken to represent behavioural data. The
completed matrices form the material for extensive analysis of the data
in terms of block-modelhng and other procedures.
Multiplexity profiles
A simple computer program allowed us to extract the multiplexity
profiles of each relationship from the five basic input matrices. I mean
by multiplexity profile the actual combination of frequency, “closeness”,
sociability, emotional support and practical aid. The multiplexity profile can be represented by a simple vector of length 5 defined as
follows:
1
1
1
1
1
if
if
if
if
if
frequency is monthly or more frequently, otherwise 2
self-defined “closeness” is highest, otherwise 2
they meet for sociable purposes, otherwise 2
emotional support is provided, otherwise 2
practical aid given, otherwise 2
J.C. Mitckefl/ Strong ties
41
amonghomekss women
Table 1
Multiplexity profiles for ten homeless families.
FCSEP
FCSEP
lllll
11112
11121
11122
11211
11212
11221
11222
12lll
12112
12121
12122
12211
12212
12221
12222
231
48
35
128
14
2
4
10
6
9
18
147
4
3
21
20
21111
2lll2
21121
21122
21211
21212
21221
21222
22111
22112
22121
22122
22211
22212
22221
22222
241
29
31
89
0
1.
3
17
8
2
13
355
1
1
10
1495
Total
2895
F, fre4uency; C, “close.ness”; S, sociability; E, emotional
support;
P, practical aid.
Clearly a m~tiple~ty profile of (11111) implies that the relationship
involved reflects all the characteristics of a strong relationship while a
profile of (22222) reflects a weak relationship. The frequency of all
profiles in all 10 families studied is set out in Table 1.
By far the greatest proportion of multiplexity profiles was the maxim&y weak one of (22222) (51.6 percent), which is a salutary finding
when we consider that we are working with personal networks made up
very largely of the kin of the respondents, Nevertheless the issue before
us is whether the multiplexity profiles could be thought of as being
subsumed by a relatively small number of latent classes which reflect
particular patterns of interaction. Accordingly several possibilities were
explored using Clogg’s Maximum Likelihood Latent Structure Analysis
Program which performs a modified version of Goodman’s maximum
likelihood latent structure analysis (Clogg 1977).
42
J. C. Mitchell / Strong ties among homeless women
After fitting four unrestricted latent classes to the data the maximum
likelihood &i-square dropped from 5134.9 to 27.95 which with 18
degrees of freedom suggested a passable fit (p = 0.06). The Pearson
&i-square for the independent model was 24006.9 but after the model
was fitted the &i-square dropped to 26.34 representing an improvement of some 0.9989. The index of dissimilarity was 0.0122. A three
class model reflected a maximum likelihood &i-square of 77.45 with 11
degrees of freedom. The improvement of 51.11 for the loss of 7 degrees
of freedom suggests that the four-class model is an appreciably better
fit.
The characteristics of the latent classes may be shown by comparing
the distribution of positive counts on each of the different aspects of
the relationship in the profile compared with the positive count for
each aspect as a whole. An effective way of doing this is to calculate the
odds ratios that a positive count will exist for some attribute given that
the relationship is in some latent class, as against occurring if the
relationship is in some other class. For example the odds that individuals in Latent Class 4 will associate frequently with others, as against
infrequently, is 336/210 = 1.6 whereas the odds that they will interact
frequently if they are not in Latent Class 4 is 364/1986 = 0.1833. The
odds ratio of 1.6/0.1833 = 8.73 reflects the extent to which frequent
association is characteristic of relationships in Latent Class 4. In terms
of the reasoning behind this measure, if the ratio of positive to negative
links within the latent class were exactly the same as outside the latent
class the ratio of the two ratios would be 1.0, indicating no association
between the given latent class and frequency of association. The size of
Table 2
Odds ratios ’ for aspects of relationships with latent classes
Latent
Class
I
II
III
IV
Components of relationships
F
0.00
11.67
16.18
8.73
C
S
E
P
0.003
0.309
4.17
296.03
0.024
O.OCQ
73.30
64.97
0.00
0.739
0.110
847.27
0.00
8.41
0.31
179.16
F, frequency; C, “closeness”; S, sociability; E, emotional support; P, practical aid.
a Odds ratios calculated from:
J.C. Mitchell / Strong ties among homeless women
43
the odds ratios therefore reflects the extent to which the aspect of
relationship in question is associated with the latent class in question.
The odds ratios for the different latent classes derived from this
analysis are shown in Table 2.
The latent classes
These classes may be interpreted in the following way;
Class I. Overall weak relationship. The percentage of relationships
classified in this class was 64.5.
Class II. Relationships between people which involve frequent contacts and the performance of practical tasks for one another. This was a
rare class with only 2.3 percent of all relationships in it.
Class III. Relationships between people involving frequent association and sociable activities in which some closeness is involved. These
appear to be typical of what normal long-standing sociable relationships might entail. There were 14.3 percent of all relationships in this
class.
Cluss IV. These are overall strong relationships. There were 18.9
percent of all relationships in this class.
These findings, derived from women who were in some distress,
confirm those of Marsden and Campbell (1984) in that the principal
components of strong relationships seems to be those of either emotional support or alternatively their own definition of “closeness”,
which the women themselves often linked with emotional support. This
is shown very clearly by the large odds-ratios for “emotional aid” of
847.16, the odds-ratio of 296.03 for self-defined closeness, the odds-ratio
of 179.16 for practical aid. These findings also confirm those of
Marsden and Campbell that mere frequency is not a very reliable
indicator of strong relationships, nor is being involved in simple
convivial relationships. The odds-ratio for “frequency of contact” for
our data was 8.73, considerably lower than that for other elements in
the multiplexity profile, and we found an odds-ratio of 64.97 for simple
convivial relationships, once again considerably lower than the other
odds-ratios for the strong relationships.
The four different patterns of relationship cannot be arranged in a
simple order of strength to weakness without further theoretical argu-
44
J.C. Mitchell
/ Strong ties among homeless women
ment. While there is little doubt about the “strong” and “weak” ties,
whether Class II should be considered to be “stronger” than Class III
would turn on the extent to which we would consider the odds ratio of
8.41 in respect of practical aid in Class II as indicating a “stronger” tie
than the odds ratio of 73.3 for sociability in Class III. Much depends
on how one appreciates the significance of practical aid as against
convivial relationships in “strong” ties.
The procedure used, of course, is a classical data reduction technique
and its utility in the analysis of data with which I am concerned is:
(a) To enable me to compare the number of weak and strong ties in
families with different histories of homelessness and so deepen our
understanding of the role of interpersonal relationships in the lives of
the families concerned;
(b) To enable me to reduce five sets of individual links to a single
matrix of ties of different types and strength and so simplify the
interpretation of our findings.
To illustrate these uses of latent classes we may apply the analysis of
two cases drawn from our homeless family data. For example Figure 1
shows the strength of ties linking the various members of one particular
network together.
The case refers to a woman in a woman’s refuge who had been
severely assaulted by the man with whom she had been living after her
marriage had broken up. After the marriage had broken up relationships with her own family had become estranged. The arrangement of
network members around the circle was determined on the basis of a
CONCOR analysis involving all five of the types of relationship we had
enquired into. What emerges very clearly from this analysis is the
extent to which the strong ties which the respondent sees herself
involved in were entirely confined to some other women in the refuge
and to the social workers with whom she had established ties. The other
women in the refuge, of course, were in the same predicament as she
was so that she had common ground with them. We might contrast the
types of tie Mrs Wyatt had with the members of her network with those
that a second woman in our sample - Mrs Maxwell - experienced.
These were as follows:
Mrs Maxwell had moved to Manchester with her husband from
Ireland. He was looking for work in Manchester but found it difficult
to get. The couple and their children squatted in unoccupied premises
4.5
J.C. Mitchell / Strong ties among homeless women
Figure 1. Ties of different sorts in Mrs Wyatt’s network
Key: (1) Respondent; (2) Co-habitee; (3) Ex-husband; (4) Mother; (5) Sister A; (6) Sister A’s
husband; (7) Sister B; (8) Co_habite& mother; (9) Co-habitee’s father; (10) Co-habitee’s brother;
(11) Co-habit&s sister A; (12) Co-habit&s sister B: (13) Friend A; (14) social worker A; (15)
Social worker B; (16) Friend B; (17) Friend C, (18) Friend D.
iVole: Thickest I&s indicate “strong” ties, slightly thinner Class III ties, thin Class II ties. Weak
ties not shown.
Mrs MaxweII
Mrs Wyatt
No
%
No
%
Strong ties
Predominantly frequent ties with practical aid
Predominantly frequent with convivial Iinks
weak ties
24
21
24
237
1.84
6.86
1.84
77.45
134
11
23
212
35.26
2.89
6.05
55.79
Totals
306
100.00
380
100.00
for some time until one of her friends suggested that they approach the
Homeless Families Unit for help. Mrs Maxwell’s network was composed entirely of her own and her husband’s kin, most of whom were
still in Ireland. Her kinship ties were still very active and two of her
children were being cared for by kin in Ireland. These circumstances
are well illustrated in the distribution of ties of different kinds as
determined by latent structure analysis. No less than 35.26 percent of
Mrs Maxwell’s links could be classified as strong in terms of the latent
class categories. The somewhat impoverished network of Mrs Wyatt,
however, stands out in marked contrast. Only 7.84 percent of the links
in her network could be classified as “strong”. These differences could
be related back immediately to the histories of the two families before
they came under the care of the social workers attached to the Homeless Families Unit.
A detailed analysis of the distribution of ties in Mrs Wyatt’s network
shows that she herself had strong ties only with one of the social
workers attached to the refuge and two of her friends in the refuge,
although she saw strong ties linking her friends and the social workers
to one another. She also saw strong ties linking her own kin to one
another and her co-habitee’s kin to one another but not between these
two categories of kin. Mrs Wyatt saw her relations~ps with her own
kin as confined entirely to frequent ties combined with a practical aid
component, but lacking any “closeness” or emotional content.
Mrs Maxwell’s links with people in her network were entirely different. While she saw the same proportion of frequent and convivial (i.e.
Class II) links among her network members as Mrs Wyatt there were
considerably fewer frequent links with some form of practical aid
involved (i.e. Class III). This is not due to the fact that most of Mrs
Maxwell’s kin did not live in Manchester but in Ireland, since the
questions relating the relationships amongst network members were
asked about those in Ireland as weIl as about Mrs Maxwell’s immediate
family in Manchester. The main difference between Mrs Wyatt and
Mrs Maxwell reldtes instead to the proportion of “strong” ties that
they saw in their networks. While Mrs Wyatt saw just under 8 percent
of all links as “strong” Mrs Maxwell saw over 35 percent - a factor of
some 4.5 times as much. Mrs Maxwell saw these strong ties as occurring between close kin-either on her own side or on her husband’s side.
There was also a fair sprinkling of strong ties linking her own kin with
her husband’s kin. The picture that emerges is one of an active kinship
J.C. Mitchell / Strong ties among homeless women
41
system which Mrs Maxwell still sees as operative in spite of the
geographical distance separating herself from her kin.
The allocation of each multiplexity profile to a latent class of
network relationships has enabled us to separate out the main component of strong and weak ties and has enabled us to deepen our
understanding of the pattern of relationships in which two of the
families in our data set are involved. The same procedures can of
course be applied equally well to the other families in the set.
References
Clogg, Clifford C.
1917 Unrestricted and Restricted Maximum Likelihood Latent Structure Analysis: A Mom& for
Users. Working Paper no 1977-79 Pennsylvania. Population Issues Research Office.
Granovetter, M.S.
1973 “The strength of weak ties”. American Journal of Sociology 78: 1360-1380.
1982 The strength of weak ties: A network theory Revisited”, in: V. Peter Marsden and Nan
Lin (eds.) Social Structure and Nehvork Analysis (pp. 105-130) London: Sage Publications.
Laumann, E.O.
1973 The Bon& of Pluralism: The Form and Subsfonce of Urban Socrol Networks. New York:
Wiley.
Marsden, P.V. and K.E. Campbell
1984 “Measuring tie strength”. Social Forces 63: 482-501.
Marsden, P.V. and E.O. Laumamr
1978 “The social structure of religious groups: A replication and methodological critique”, in:
S. Shye (ed.) Theory Construction and Dotu Analysis in the Behoviorol Sciences. San
Francisco: Jossey-Bass.
Verbrugge, L.M.
1977 “The structure of adult friendship choices”. Social Forces 56: 576-97.