ARTICLE IN PRESS

Social Science & Medicine 58 (2004) 2267–2277

Mental health and social capital in Cali, Colombia

Trudy Harphama,*, Emma Grantb, Carlos Rodriguezc
a
Urban Policy and Development, South Bank University, London SW8 2JZ, UK
b
ABColombia, London SE1 7RT, UK
c
FUNDAPS, Avenida Cascajal Calle Alferez Casa 04, Ciudad Jardin, Cali, Colombia

Abstract

Mental ill health forms an increasingly significant part of the burden of disease in developing countries. The growing
interest in social risk factors for mental health coincides with the development of social capital research which may
further inform the social model of mental health. The objective of the study reported here was to discover if there is an
independent association between social capital and mental health when taking into account an array of demographic
and violence variables.
A total of 1168 youth (15–25 years) in a low income community in Cali, Colombia were surveyed. Mental health was
measured by a 20 item self-report questionnaire. The instrument used to measure social capital covered structural and
cognitive social capital. Twenty-four per cent of the sample were probable cases of mental ill health. Females had a
prevalence rate three times higher than males. Using a model which considered demographic and social capital
measures as potential risk factors for mental ill health, the significant risk factors emerged as being female, having
limited schooling, working in the informal sector, being a migrant, and having low trust in people. The ‘classic’ poverty
type variables (poor education and employment) were more important than social capital, as was the commonly
dominant risk factor for mental ill health—being a woman. When violence factors were added to the model, the ‘trust’
factor fell out and the most important risk factors became (in descending order of importance): being female; no
schooling/incomplete primary; and being a victim of violence. The dominance of poverty related factors, as opposed to
social capital, prompts renewed attention to the explanatory mechanisms that link income inequality and poor mental
health.
r 2003 Published by Elsevier Ltd.

Keywords: Mental health; Social capital; Youth; Colombia; Violence

Introduction: Linking mental health and social capital
Mental ill health forms an increasingly significant part
of the burden of disease in developing countries (WHO,
2001). Arguably, the social model of mental health (as
opposed to the biological model) offers most hope when
it comes to designing cost-effective interventions to
improve mental health (see, for example, Somasundar-
am, van de Put, Eisebruch, & de Jong, 1999). The main
social risk factors for mental ill health are long-term
difficulties, negative life events and lack of social support
to buffer, or reduce, the effects of such difficulties and
*Corresponding author. Fax: +44-207-815-8392.
E-mail address: t.harpham@sbu.ac.uk (T. Harpham).
events. Examples of long-term difficulties include
poverty, inadequate education, living in a hazardous
physical environment and high levels of insecurity
(violence and accidents). Examples of negative life
events include separation from partner, loss of employ-
ment and forced migration. Changes in social support
can occur with phenomena such as urbanization which
may lead to a reduction of extended families, an increase
in single parent households, reduced fertility, age-
specific rural–urban migration, increased women’s
labour force participation, and under or unemployment
(Harpham, 1994). Patel (2001) eloquently argues the
case for more attention to be paid to mental health in
developing countries. He argues that the social factors
known to be linked to depression and other mental

0277-9536/$ - see front matter r 2003 Published by Elsevier Ltd.

doi:10.1016/j.socscimed.2003.08.013
 ARTICLE IN PRESS
2268 T. Harpham et al. / Social Science & Medicine 58 (2004) 2267–2277
illness are on the increase throughout the developing
world.
The growing interest in social risk factors for mental
health in developing countries coincides with the
development of social capital research which may
further inform the social model of mental health. While
there are various definitions of social capital (and
resulting criticisms, see, for example Forbes & Wain-
wright, 2001), the current study follows Lynch and
Kaplan (1997) who describe social capital as the ‘stock
of investments, resources and networks that produce
social cohesion, trust and a willingness to engage in
community activities’ (p. 307). In other words, social
capital has a structural component (networks, connect-
edness, associational life and civic participation) and a
cognitive component (perceived support, trust, social
cohesion and perceived civic engagement). Several
researchers, including Rose’s (2000) important study of
social capital and health in Russia, limit their definition
of social capital to networks only (typically measured by
membership of organizations) and reject Putnam’s
(1997) idea of social capital being composed of net-
works, norms and trust. Rose (2000) and Coleman
(1990) argue that trust should not be measured as part of
social capital because of the ensuing difficulty of
separating out the effects of behaviour (structural social
capital) and attitudes/perceptions (cognitive social capi-
tal). However, while cognitive social capital might well
be a result of structural social capital, the relative effects
of the two components can be separated if there is
appropriate measurement. For the purposes of this
study, as we shall see below, it is particularly important
that both structural and cognitive social capital be
measured. Moreover, recent work on developing valid
measures of social capital recommend separate measure-
ment of trust, connectedness and reciprocity (Stone &
Hughes, 2002). More on the measurement of social
capital is presented below in ‘Context and Methods’.
Conceptual work that links mental health with social
capital hypothesizes that social capital affects mental
health status through various mechanisms. Kawachi and
Berkman (2001) developed two models to explain this
pathway: one of these is the stress-buffering model
(social support can buffer the effects of life events); the
other is the ‘main effect model’ whereby participation in
social networks results in exposure to social influence
(norms); positive affective states; and beneficial neu-
roendocrine responses. Rose (2000) similarly argues that
‘social capital networks ought to be specially good at
providing emotional support that reduces the likelihood
of emotional depression’ (p. 1426). Harpham, Grant,
and Thomas (2002a) go further and hypothesise that
social capital could reduce the stressors, or risk factors,
for mental ill-health. Social capital could reduce
negative life events (for example, preventing the loss of
a job) and reduce long-term difficulties (for example
promoting good physical health). It is also possible to
hypothesize that different components of social capital
may affect mental health in different ways. The
‘structural’ aspect of social capital (connectedness,
membership of organisations) can provide access to
institutions, formal and informal, which may reduce the
negative impacts of life events by providing additional
support. On the other hand, the ‘cognitive’ aspects of
social capital, such as trust, sense of belonging, and
shared values, may increase feelings of security and self-
esteem and thus improve mental health. This all focuses
on how social capital might affect mental health. As
Cullen and Whiteford (2001) note, no research has
examined the effect of mental health on social capital.
Although the above hypotheses have been generated,
it is generally the case that ‘there has been a lack of
research and theoretical development into social capital
and mental health’ (McKenzie, Whitley, & Weich, 2002,
p. 280). Certainly very little research has examined the
empirical link between social capital and mental health
in a developing country context. An opportunity to
examine this relationship arose when a local, non-
governmental organization and the municipal health
department of Cali, Colombia decided to try to
strengthen the social capital of youths (age 15–25 years)
in order to improve their health and well-being. The
intervention required a baseline study of social capital
and mental health. As the target population has high
levels of violence (perpetrating, witnessing or being a
victim) it was hypothesized that this might also have a
direct relationship with mental health. The objective of
the study reported here is to discover if there is an
independent association between social capital and
mental health, when taking into account an array of
demographic and violence variables.
Context and methods
Cali is one of the major cities in Colombia with a
population of approximately 2 million. Harris (2001)
argues that Cali is in the midst of the worst economic
crisis in its recorded history with: a GDP falling since
1995 by over 2% per year; a decline in mining and
construction; a decline in private and public sector
employment; a move to self-employment; and a third of
the population under the age of 18. On a positive note he
suggests that strengths are the growth of exports in
garments, shoes, cut flowers, asparagus and printing and
publishing. However, these strengths are not unique to
Cali and Harris suggests that the unique culture of the
city/region, particularly music, could provide a distinc-
tive basis for tourism in the city if the problem of
security is ever resolved.
The inequalities are so great in Cali that Urrea and
Ortiz (1999) suggest that it is a city of low paid workers
 ARTICLE IN PRESS
T. Harpham et al. / Social Science & Medicine 58 (2004) 2267–2277
(peones) and professionals (doctores) with no room for
intermediate categories. Such inequality, following
Wilkinson’s (1996) arguments, may cause perceptions
of unfairness, frustration and disrespect which are
associated with increased stress and mental ill health
among the poorest group. Thus, it may be that low-
income youth at ‘the bottom of the heap’ may have
particularly high levels of mental ill health.
Administratively, Cali is divided into 21 urban
boroughs and is very heterogeneous. The poorest area
is in the eastern fringe of the city and is referred to as
Aguablanca (comprised of communities 13, 14,15 and 21
with a total population of around 400,000). Twenty
years ago Aguablanca was a lake. With massive
migration from rural areas, the lake was drained and
slum settlements were established and slowly upgraded.
Poverty has a clear racial connotation in Cali—while the
black population makes up about one-third of the
overall population, they represent more than half the
poor. Aguablanca is heavily populated by black
migrants from the pacific coast. In this area, adolescents
without work have some of the highest levels of income
poverty (Hentschel, 2000).
The city has high homicide rates (for example rising to
134 per 100,000 in one of the poorest boroughs in 1998).
The research reported below was undertaken in bor-
oughs 13 and 14 in Aguablanca. Violence rates here are
amongst the highest in the city and the population is
characterised by poverty, unemployment and lack of
access to basic services such as health and education.
Municipal data indicates that boroughs 13 and 14 are
similar in that they both have an unemployment rate of
approximately 25%, 26–28% live in poverty (defined as
less than US $3.2 per person per day), and around 30%
of the population is between 10–24 years old (Munici-
pality of Cali, 2000a).
Neighbourhood maps that indicate dwellings were
acquired from the city planning department. After field
visits to update the maps, clusters of 20 dwellings each
were marked on the maps. Using ‘Stat-calc’ software,
and a cross-sectional design requiring the ability to
detect a 10% reduction in a key variable of interest in
the future, and assuming a prevalence of a key variable
of 55%, a total sample of 1168 individuals was required.
One hundred and sixteen clusters were randomly
selected. Within each selected cluster approximately 10
dwellings were randomly selected. Each dwelling was
approached to ascertain whether there was one young
person who had lived in the area for at least 1 year. If a
selected dwelling did not contain an eligible respondent,
the dwelling next door was approached (until a youth
within the selected age range was found). In each eligible
dwelling, only one young person was interviewed. In
dwellings where more than one eligible person was
resident the respondent was selected by a random
alphabetical method.
2269
Interviewers were male and female local residents
(from outside the study area) and over 18 years of age.
The 37 interviewers were trained for 4 days and were
monitored by 15 supervisors. Interviews were conducted
in Spanish (the language used by all the respondents)
and averaged 45 min. Fieldwork was undertaken in early
2002.
Mental health was measured by the Self-Reporting
Questionnaire—20 Items (SRQ20) recommended by
WHO (1994) for assessing the prevalence of depression
and anxiety at the community level when diagnoses of
specific illnesses are not required. Its validity and
reliability in developing countries has been widely tested
and its sensitivity is in the range 70–90% and specificity
in the range 62–95%. The tool consists of 20 yes–no
questions and, in low literacy contexts, is administered
by an interviewer. A cut-off point is used to determine
probable cases and non-cases of mental ill health. Lima,
Pai, Santacruz, and Lozano (1991) used a cut-off point
of 7/8 in Armero, Colombia and found a prevalence of
56% of probable cases of mental health among natural
disaster victims. Their results were validated against
DSM-3 (appropriate at that time, although DSM-IV is
now current), an internationally recognised diagnostic
system, and so the same cut-off was used in the current
study.
The instrument used to measure social capital was
developed over the course of 3 years by the research
team and attempts to capture an individual’s percep-
tions and experience of social capital rather than chasing
an elusive ‘community’ level measure of social capital
(see Harpham et al., 2002a). Questions covered group
participation, general, thick and thin trust (see Grano-
vetter, 1973), social cohesion, informal social control,
social support (informational, instrumental and emo-
tional—see Thoits, 1995), and civic participation. All
dimensions consisted of a number of questions which
used a five-point Likert scale, except the measurement of
group participation which asked if respondents were
active members of a range of prompted groups/
organisations. Behavioural questions used a 1-year
recall period. Similar to other measures of social capital,
no validity or reliability measures are available.
The violence measurement tool was adapted from the
ACTIVA violence questionnaire, previously applied in
several cities in Latin America, including Cali (Orpinas,
1998). The instrument covers consequences of violence
in the neighbourhood, consequences of violence in the
family, domestic violence, non-familial acts of violence
(categorised into witness, victim or perpetrator), and
norms relating to violence (justification of violence as
conflict resolution) in the family and neighbourhood. All
attitudinal questions used a five-point Likert scale.
Behavioural questions used a three-point scale asking
if the action had never been undertaken, done once or
more than once in the respective recall period. The recall
 ARTICLE IN PRESS
2270 T. Harpham et al. / Social Science & Medicine 58 (2004) 2267–2277
Demographic characteristics
Social capital dimensions
Participation
Trust (general, thick and thin)
Collective efficacy (social cohesion and informal
social control)
Social support (emotional, instrumental
and informational)
Civic participation
Violence dimensions
Norms
Perpetration
Victim
Witness
Neighbourhood consequences
Note: This framework is intentionally unidirectional as the study’s objective is to test for an
independent association between social capital and mental health, taking into account
demographic and violence dimensions. However, relations between all components
(backwards and forwards) may be hypothesized.
Fig. 1. The conceptual framework of the study.
period was 1 year apart from questions about more
frequent behaviour, for example perpetration of violence
against children, where the recall period was reduced to
1 month. No validity or reliability measures are
available.
Basic demographic data were also collected: age, sex,
length of residence in the neighbourhood, religion,
occupation, schooling, household structure, quality of
housing and degree of overcrowding. All data analysis
were carried out using SPSS.
The conceptual framework of the study is presented in
Fig. 1.
Factor and logistic analyses were used and the
processes are briefly described here. Variables that failed
to divide the sample more evenly than 20%/80% were
not taken forward into the multivariate analysis. When
the variable was measured by a Likert scale, if 80% or
more of the sample fell within the ‘strongly agree’ or
‘agree’ or ‘strongly disagree’ or ‘disagree’ categories, the
variable was excluded. For example, over 80% of
respondents agreed that ‘people around here know each
other’ in the measurement of social cohesion. This was
subsequently excluded from further analysis on the
grounds that it failed to discriminate sufficiently. Table 3
can be used to examine which variables were more
discriminating than others.
Next, correlations were carried out between items
within the same dimension. This allowed an assessment
of whether factor analysis could be usefully carried out.
‘Factor analysis tries to identify statistically the under-
lying dimensions of the set of questions, by locating
clusters of questions that are related to each other. It
does that by correlating every item with every other
item, and identifying the clusters that emerge’ (Onyx &
Bullen, 1997, p. 11). Items were sent to factor analysis by
dimension, in accordance with the theoretical frame-
work used to develop the instrument. As Bajekal and
Purdon (2002) argue in their analysis of social capital in
the English health survey: ‘given that the original
Mental
health module was grouped into separate domains, which were
designed to cover discrete elements of social capital and
exclusion, these domains were retained in the analysis
and separate factor analysis was carried out within each
domain’ (p. 5). Using a varimax rotation, principal
component analysis was carried out on each dimension
to extract the initial factors. Inter-item correlation
(Cronbach’s alpha) was undertaken to provide an
indication of the extent to which each item contributed
to the total social capital scale. Factor analysis and
inter-item reliability were carried out separately for each
dimension, including only discriminating variables.
Table 1 presents a summary of the items that make up
the resulting eight factors of social capital. Group
membership is not included here as the dimension
consists of one item only and is therefore not appro-
priate for factor analysis.
Once the eight factors were extracted, an additive
scale was created for each factor. These scales were
created by calculating the mean score of items within
each single factor. The scales were then treated as
quantitative variables and, together with a group
membership scale, used in logistic regression with
mental health as the dependent (binary) variable.
Results
In order to gain an understanding of the broad picture
of demography, social capital, and mental health of the
respondents, some descriptive statistics are presented
first.
Table 2 presents the demographic profile of the 1060
respondents (there was a 4% refusal rate plus 65
questionnaires were not applied in commune 14 due to
increased insecurity of the time of fieldwork). The profile
illustrates the high poverty level in Aguablanca with
nearly 60% of the youth dependant on their family and
with no jobs, and 54% having left school before
completing secondary level. The fact that nearly three-
quarters of the youth expressed no religious beliefs was
notable in the largely Catholic religious context.
The social capital of the youth is presented in Table 3.
Sixty-five per cent did not participate in any group.
Parches represent a more informal type of group than
the other types of groups named. Qualitative research
with youth in the area found that ‘parche’ was an
informal group of friends with a common interest and
objective: it could be listening to music, chatting or
playing games, or consuming drugs and getting involved
 ARTICLE IN PRESS
T. Harpham et al. / Social Science & Medicine 58 (2004) 2267–2277 2271

Table 1
Social capital factors

Conceptual Factor extracted Item

dimension

Trust in institutions Trust in state institutions
Trust in community institutions
The school provides a service whenever the community needs it
The health centre provides a service whenever the community
needs it
The police provides a service whenever the community needs it
The neighbourhood committee (Junta de Accion ! Comunal)
provides a service whenever the community needs it
The borough committee (Junta Administradora Local) provides a
service whenever the community needs it

Trust in people Thick trust There are neighbours in whom you can confide personal matters
There are neighbours to whom you can lend things without
worrying that they will not return them or will damage them
Thin trust In this neighbourhood the majority of neighbours do not gossip
maliciously about people
In this neighbourhood the majority of neighbours return a lost
wallet if they find it
In this neighbourhood the majority of neighbours do not take
advantage of others even if they have the chance

Social cohesion Solid cohesion and solidarity People around here get on with each other
Solidarity People are close
People often visit one another
There are neighbours who would help you financially if you lost
your income
There are neighbours who would tell you about an interesting
work opportunity

Social control Social control People would get together to improve security in the area
People would get together to demand better health services
People would get together to improve the image of the
neighbourhood
People would get together to prevent formation of waste dumps in
the neighbourhood
People would get together to prevent drug use in public places
People would get together to demand better schools
People would get together to denounce corruption in institutions
in the neighbourhood

Civic participation Neighbour’s civic participation
Personal civic participation
Local residents frequently talk to authorities or local
organizations about local problems
People in this area actively participate in electoral campaigns
People in this area are prepared to participate in citizen marches
and demonstrations
People in this area actively participate in the elections for the
neighbourhood and borough committees
In the last year, have you spoken with authorities or local
organizations about problems in the neighbourhood, areas or
city?
Have you participated in citizen marches or demonstrations?
Have you carried out any voluntary work for the benefit of the
community or neighbourhood?
Have you got together with neighbours to try and solve a problem
affecting the area?
 ARTICLE IN PRESS
2272 T. Harpham et al. / Social Science & Medicine 58 (2004) 2267–2277
Table 2
Demographic profile of the youth respondents
n (1060)
Age
15–17 334
18–20 362
21–23 262
24–24 102
Female 610
No religion 783
Have children 367
No job+depend on family 621
Left school before completing secondary 571
Overcrowded household (more than 3 per 198
sleeping room)
Non-durable (non-brick) housing 111
Female headed household 401
Family from Cali 585
in criminal or delinquent activities. The younger age
range participated more than the older age range and
males participated more than females in cultural/dance/
sports groups and parches.
Although only 16% participated in religious groups,
85% of respondents thought the church provided a good
service to the community.
Thick trust represents close bonds with a few
members of the neighbourhood, whilst thin trust refers
to the general trust in community residents, i.e. is
extended over a greater area to include most residents,
not just a few close neighbours or friends (Granovetter,
1973). In this study, thick trust (such as having someone
who would take care of family or the house) was greater
than thin trust (such as believing that most people would
not take advantage of you). This indicates socially
atomised communities.
Social cohesion and informal social control together
make up the concept of collective efficacy as used by
Sampson, Raudenbush, and Earls (1997) in their
measurement of social capital and violence in the
USA. Overall, collective efficacy was quite high among
adolescents in Cali which provides hope for future
community development interventions.
Perceived social support (sometimes referred to as
solidarity in the social capital literature) available to the
youths varied according to the dimension of social
support. Perceived emotional and informational support
was available but instrumental (mainly financial) sup-
port was substantially lower (37% respondents reported
having someone in the area to provide this type of
support). This appeared to reflect the high levels of
poverty in the area.
The perception of the general civic participation of the
neighbourhood was high with over half the respondents
perceiving that neighbours would participate in elec-
tions, lobbies, marches, etc.
% As sex will emerge as an important factor in the
multivariate analysis we also tested whether there
was a significant association between sex and the eight
32 social capital factors presented in Table 1 plus a
34 group membership index. Significant differences in the
25
mean scores of five of the social capital factors were
10
58
found by sex: trust in community institutions (females
74 higher than males); thick trust; thin trust; social
35 cohesion and solidarity; and group membership
59 (males consistently higher than females in these four
54 factors).
19 Turning to mental health, the percentage of positive
responses to the 20 items in the SRQ 20 are provided in
11 Table 4. A breakdown by sex is also presented. Twenty-
38 four per cent of the sample were probable cases of
55
mental ill health, based on the validated cut-off point
noted in the ‘Methods’ section above. Females had a
prevalence rate three times higher than males, which is
similar to other community-based prevalence surveys in
low income urban areas in developing countries (Harp-
ham & Blue, 1995). Gender differences in support
derived from social networks might partly explain higher
prevalence of mental ill health among women and role
strain associated with obligations to provide social
support through high levels of social connections might
also be a factor. The multivariate analysis below
examines whether social connections (structural social
capital) and/or social support (an element of cognitive
social capital) have an independent effect on mental ill
health. The fact that 13% of the youth admitted
considering suicide in the last month is particularly
disturbing.
Using a model1 which considered the demographic
and social capital measures as potential risk factors for
mental ill health, the main significant risk factors
emerged as: being female which has an odds ratio2 of
3.8 (CI 2.7–5.4; po0:000); having no schooling or
incomplete primary compared to completed secondary
or above with an OR of 2.5 (CI 1.4–4.2; po0:001);
working in the informal sector compared to being
dependent on one’s family (odds ratio=1.5; CI 1.1–
2.2; po0:019); and having low levels of thin trust in
people (odds ratio=0.8; CI 0.7–1.0; po0:044). Thus,
only one of the eight social capital factors as presented
in Table 4 proved to be independently important for
mental health—thin trust in people. (Membership—
structural social capital—did not emerge either). The
1
The selected model has 94% confidence level. A model with
95% confidence excluded all social capital factors. Since the
main focus of interest in this study is social capital, the authors
were willing to accept a slightly lower level of confidence.
2
Odds ratios and confidence intervals have been rounded to
one decimal place.
 ARTICLE IN PRESS
T. Harpham et al. / Social Science & Medicine 58 (2004) 2267–2277 2273

Table 3
Social capital of the youth respondents (items ranked in order of frequency)

Dimension Item n (1060) %

Participation
No participation in any group 689 65
Participate in gang/crew (‘parche’) 212 20
Participate in cultural/sports/dance group 195 18
Participate in religious groups 164 16

General trusta
Trust church 900 85
Trust school 743 70
Trust health centre 684 65
Trust police 367 35
Thick trusta
Can leave family member with neighbours 922 87
Can leave keys with neighbours 896 85
Can lend with confidence 574 54
Can confide in neighbours 422 40
Thin trusta
Most people trust other neighbours 382 36
People do not take advantage of others 299 28
Most would return a lost wallet 193 18
Most people do not gossip or speak badly 161 15
Social cohesiona
People know each other 855 81
People visit one another 484 46
People get on with each other 439 41
People are close 320 30

Informal social control (in this area, people would get together toy)a
Prevent dumps forming 804 76
Improve image of area 768 72
Demand better health services 754 71
Demand better schools 741 70
Improve security 685 65
Denounce corruption in local institutions 619 58
Prevent drug consumption 554 52
Intervene in conflicts 453 43

Social support (there are neighbours who wouldy)a

Support you or counsel you 892 84
Give you information 852 80
Take care of others’ kids 794 75
Be willing to help others 753 71
Tell you about work 713 67
Help you financially 390 37

Civic participation of neighbours (people in this areay)a

Vote in elections 850 80
Participate in electoral campaigns 763 72
Participate in local committee elections 741 70
Talk to authorities about problems 713 67
Participate in marches 619 58

a
Respondents who answered ‘strongly agree’, or ‘agree’ to Likert five-point scale.
 ARTICLE IN PRESS
2274 T. Harpham et al. / Social Science & Medicine 58 (2004) 2267–2277

Table 4
Number and percentage of positive responses to SRQ 20 items (total and by sex) (items ranked in order of frequency according to
‘total’ column

Total n=1060 (%) Males n=450 (%) Females n=610 (n, %)

Feel nervous, tense or worried 53 40 62

Easily frightened 44 22 60
Feel unhappy 44 29 55
Find it difficult to make decisions 41 34 46
Often have headache 40 24 51
Have trouble thinking clearly 35 29 40
Find it difficult to enjoy daily activities 32 28 35
Lost interest in things 28 25 31
Easily tired 25 10 23
Poor appetite 22 14 29
Sleep badly 21 18 24
Have uncomfortable feelings in your stomach 20 14 23
Hands shake 18 15 20
Feel tired all the time 18 10 23
Cry more than usual 17 4 27
Daily work/study suffering 17 18 16
Thought of ending your life been on your mind 13 9 16
Unable to play a useful part in life 12 10 13
Poor digestion 9 17 15
Feel worthless 6 4 7

higher the level of trust across the community (as
opposed to trust in only neighbours/friends—thick
trust) the lower the likelihood of mental ill health. The
‘classic’ poverty type variables (poor education and
employment) were more important than social capital,
as was the commonly dominant risk factor for mental ill
health—being a woman. In fact, when violence factors
were added to the model,3 the ‘trust’ factor fell out and
the most important risk factors became (in descending
order of importance): being female (odds ratio of 4.3; CI
2.9–6.2; po0:000); no schooling/incomplete primary
compared to completed secondary or above (odds
ratio=2.7; CI 1.5–4.7; po0:000); presence of family
violence (odds ratio=1.7; CI 1.3–2.3; po0:000); victim
of violence (odds ratio=1.6; CI 1.2–2.1; po0:002);
perception that violence affects one’s community (odds
ratio=1.5; CI 1.2–1.9; po0.000); and being a perpe-
trator of violence (odds ratio=1.4; CI 1.1–1.9;
po0:021).
Discussion
Are this study’s findings on participation and trust
similar to other studies in Colombia? The low participa-
tion of Cali adolescents in any form of organisation
3
The model used excludes partner violence since this would
restrict the sample to youth with partners (n ¼ 634). However,
in a model that includes partner violence, this variable emerges
as significant.
matches with the only other survey of social capital in
Colombia which was carried out by Sudarsky (1999)
who discovered that 43% of adults did not belong to any
organisation. He found that 53% of respondents
belonged to some sort of religious organisation whereas
the lower participation in religious groups in the current
study (16%) may be explained by the younger age
profile of the current study’s respondents. Sudarsky
(1999) also showed that Colombia had some of the
world’s highest level of interpersonal distrust. In terms
of trust in institutions, he found that people had high
levels of trust in the church, educational systems and
ecological movements; middle levels of trust in the legal
system, police and local government; and least trust
shown in parliament, political parties and guerrilla
movements. The youth in this study have similarly low
trust in the police. It needs to be noted that in terms of
comparison with international literature on social
capital, this study’s respondents were very young. Most
of the ratings (except for financial-related items) were
fairly high. Social capital might mediate differently upon
mental health if an older age group was considered.
This study ultimately aimed at examining the relative
association of structural and cognitive social capital
with mental health. The majority of empirical studies on
social capital are now measuring both structural (usually
membership/activity in informal and formal organiza-
tions and civic engagement of varying nature) and
cognitive (trust, reciprocity, norms) social capital
instead of simply lumping everything together into high,
medium or low social capital. Hjollund and Svendsen
 ARTICLE IN PRESS
T. Harpham et al. / Social Science & Medicine 58 (2004) 2267–2277
(2000) argue: ‘the two dominant (operational) features
describing social capital are ‘trust’ and ‘memberships of
voluntary organizations’ (p. 9). Although it has been
hypothesised that both structural and cognitive social
capital may be associated with mental health, it could be
posited that cognitive (feeling) aspects are more likely to
be significant than structural (behavioural) aspects
because cognition is more closely related to mental
health. Structural social capital was not associated with
mental health in the current study. However, even the
cognitive social capital did not feature very highly in the
explanation of variance in mental health. Cognitive
social capital has always been more complex to measure
than structural social capital. This study used 35
questions (see Table 3) which formed eight factors
(Table 1). It is interesting to note that lack of ‘thin trust’
was the only significant explanatory social capital factor
for mental ill health (before adding in violence factors),
in view of the fact that studies that measure social
capital in a shorter, more superficial manner tend to
focus on this factor. (The classic, now common, question
of whether people in the neighbourhood/community are
likely to take advantage of you is included in the
measurement of this concept of ‘thin trust’.) Therefore
even such shorter measurement tools might identify
associations between social capital and mental health if
they focus on ‘thin trust’.
Are there any other studies which have found similar
results? Another recent study, albeit in the developed
country context of the USA, found no mediating effect
of social capital on the relationship between poverty
(economic and environmental stressors) and mental
health (Mitchell & La Gory, 2002). Also, some years
ago, Furstenberg and Hughes (1995) used secondary
longitudinal data on 252 low income youth in Baltimore,
USA to explore relationships between social capital
(measured in 1984) and several indicators of young adult
success (measured in 1987), including being a graduate,
having a job, avoiding pregnancy before the age of 19,
avoiding crime and having ‘robust mental health’. The
latter was measured by a standard scale measuring
depression. Social capital was measured in terms of:
‘within family’ dimensions, such as support to and from
own mother, father in home, parents’ expectations of
school performance, number of child’s friends mother
knows, etc. (a total of 11 measures); and ‘community-
based’ social capital dimensions, such as religious
involvement, strong help network, neighbourhood as a
place to grow up, etc. (a total of seven measures). In
other words, social capital was measured through a
number of discrete dimensions. They found ‘no evidence
that social capital is implicated in depressive sympto-
matology’ (p. 589).
A single published, quantitative study (McCulloch,
2001) to find an independent, positive relationship
between social capital and mental health used the British
2275
Direct
Increased stressful social comparisons
Frustrations
Income ineq
Poor
mental
health
Indirect
Exclusion/ Lower social
poor services capital
Fig. 2. The direct and indirect association between income
inequality and poor mental health.
Household Survey which asked eight questions about
people’s neighbourhood. Respondents were divided into
low, medium, high and very high levels of social capital.
Mental health was measured by the 12-item General
Health Questionnaire (GHQ). Men and women in the
low social capital category were nearly twice as likely to
be cases of mental ill health than those in the highest
social capital category.
To the authors’ knowledge these are the only other
published, quantitative studies that have examined
whether social capital has an independent effect upon
mental health when considered alongside poverty-
related variables. The fact that three studies found
‘negative’ results suggest that there maybe a ‘need for
some modification of recent claims that social capital is
a critical resource promoting individual well-being’
(Mitchell & La Gory, 2002, p. 199). However, it should
be noted that there are abundant studies which
demonstrate the relationship between social support
(one component of social capital) and mental health
(Thoits, 1995).
What further research do these studies prompt? The
outstanding importance of the poverty-related factors
for mental health in these few studies prompts renewed
consideration of the explanatory mechanisms of the
relationship between poverty and mental health. Fig. 2
suggests the direct and indirect associations between
income inequality and poor mental health. While
income inequality is not the same as absolute poverty
it is perhaps more pertinent for a very unequal society
like Cali. Moreover, as Wilkinson (1999) has argued, at
the point of the epidemiological transition (when the
relative burden of mental ill health grows) the absolute
ownership of material goods becomes less important and
other things, like relative deprivation, become more
important. Cali as a whole is well advanced in the
epidemiological transition with infectious diseases caus-
ing only 8% of deaths between 1988 and 1999 (chronic
and degenerative 36%; accidents and violence 24%;
cancers 15%) (Municipality of Cali, 2000b).
Much of the discussion about the mechanisms to
explain the link between income inequalities and health
(not mental health specifically) has focused on trust.
 ARTICLE IN PRESS
2276 T. Harpham et al. / Social Science & Medicine 58 (2004) 2267–2277
Uslaner (2002) found that societies with lower inequality
have higher levels of trust. Furthermore, Wilkinson’s
(1999) path analyses found that the relationship between
income inequality and mortality went almost wholly
through measure of trust. ‘We now know not only that
there are systematic population differences in psycho-
social characteristics such as [trust], but that these
differences are strongly correlated with structural vari-
ables such as income inequality’ (Wilkinson, 1999, p.
529). However, if studies that measure social capital in a
multidimensional way find that neither trust nor other
dimensions explain variance in health we may need to
re-examine our theoretical models. It is perhaps easier to
theorize the indirect link between income inequality and
poor mental health than the direct link (Fig. 2). If
further studies suggest that there is no independent
effect of social capital on mental health more attention
needs to be given to researching the direct link between
income inequality and mental health. This area includes
exploring feelings of disrespect, being looked down
upon, in other words the cognitive state resulting from
social differentiation. Do these become greater with
increased income inequality, and are they associated
with an increase in poor mental health? We need
longitudinal studies to test hypotheses about causation
(see Harpham, Huttly, Wilson, & de Wet, 2002b). In
policy terms this research area is even more daunting
than social capital. While there is a lively debate on
whether and how social capital can be strengthened
‘from the outside’—increasing individuals’ links with
and trust in, other people and institutions inside and
outside their community—the prospect of interventions
to raise individual’s self-esteem or respect for others
appears particularly challenging.
Acknowledgements
The above research was part of a project funded by
the US National Institutes of Health. The authors are
grateful to Robert Smith and Elizabeth Thomas for their
comments on drafts.
References
Bajekal, M., & Purdon, S. (2002). Social capital and social
exclusion: Development of a condensed module for the health
survey for England. London: National Centre of Social
Research.
Coleman, J. (1990). Foundation of social theory. Cambridge,
MA: Harvard University Press.
Cullen, M., & Whiteford, H. (2001). The interrelations of social
capital with health and mental health. Canberra: The
Commonwealth of Australia.
Forbes, A., & Wainwright, S. (2001). On the methodological,
theoretical and philosophical context of health inequalities
research: A critique. Social Science & Medicine, 53, 801–816.
Furstenberg, F., & Hughes, M. (1995). Social capital and
successful development among at-risk youth. Journal of
Marriage and the Family, 57, 580–592.
Granovetter, M. (1973). The strength of weak ties. American
Journal of Sociology, 78, 1360–1380.
Harpham, T. (1994). Urbanization in mental health in
developing countries: A research role for social scientists,
public health professionals and social psychiatrists. Social
Science & Medicine, 39(2), 233–245.
Harpham, T., & Blue, I. (1995). Urbanization and mental health
in developing countries. Ashgate: Avebury.
Harpham, T., Grant, E., & Thomas, E. (2002a). Measuring
social capital in health surveys: Key issues. Health Policy
and Planning, 17(1), 106–111.
Harpham, T., Huttly, S., Wilson, I., & de Wet, T. (2002b).
Linking public issues with private troubles: Panel studies in
developing countries. Journal of International Development,
15(3), 353–364.
Harris, N. (2001). Economic Reactivation. In Cali city
development strategy. Washington, DC: World Bank.
Hentschel, J. (2000). Poverty in Cali, Colombia. Washington,
DC: World Bank.
Hjollund, L., Svendsen, G. (2000). Social capital: A standard
method of measurement. Scandinavian Working Papers in
Economics. Aarhus: Aarhus School of Business.
Kawachi, I., & Berkman, L. (2001). Social ties and mental
health. Journal of Urban Health, 78(3), 458–467.
Lima, B., Pai, S., Santacruz, H., & Lozano, J. (1991).
Psychiatric disorders among poor victims following a major
disaster-Armero, Colombia. Journal of Nervous and Mental
Diseases, 179(7), 420–427.
Lynch, J., & Kaplan, G. (1997). Understanding how inequality
in the distribution of income affects health. Journal of
Health Psychology, 2, 297–314.
McCulloch, A. (2001). Social environments and health: Cross
sectional national survey. British Medical Journal, 3234,
208–209.
McKenzie, K., Whitley, R., & Weich, S. (2002). Social capital
and mental health. British Journal of Psychiatry, 181, 280–
283.
Mitchell, C., & La Gory, M. (2002). Social capital and mental
distress in an impoverished community. City and Commu-
nity, 1(2), 199–222.
Municipality of Cali (2000a). Cali en Cifras. Cali: Municipal
Department of Cali.
Municipality of Cali (2000b). Diez anos de la salud en Santiago
de Cali 1990–1999. Cali: Municipal Department of Cali.
Onyx, J., & Bullen, B. (1997). Measuring social capital in five
communities in new South Wales: An analysis. Sydney:
Centre for Australian Community Organizations and
Management.
Orpinas, P. (1998). Progress report of the multi-centre project on
violence (Project Activa). Washington, DC: Pan American
Health Organization.
Patel, V. (2001). Poverty, inequality and mental health in
developing countries. In D. Leon, & G. Walt (Eds.),
Poverty, inequality and health. Oxford: Oxford University
Press.
 ARTICLE IN PRESS
T. Harpham et al. / Social Science & Medicine 58 (2004) 2267–2277
Putnam, R. (1997). Democracy in America at century’s end.
In A. Hadenius (Ed.), Democracy’s victory and crisis
(pp. 27–70). New York: Cambridge University Press.
Rose, R. (2000). How much does social capital add to
individual health? Social Science and Medicine, 51(9),
1421–1435.
Sampson, R., Raudenbush, S., & Earls, F. (1997). Neighbour-
hoods and violent crime: A multi-level study of collective
efficacy. Science, 277, 918–924.
Somasundaram, D, Van de Put W, Eisebruch M, de Jong J.
(1999). Starting mental health services in Cambodia. Social
Science & Medicine, 48(8), 1029–1046.
Stone, W., & Hughes, J. (2002). Social capital: Empirical
meaning and measurement validity. Research paper number
27. Melbourne: Australian Institute of Family Studies.
Sudarsky, J. (1999). Colombia’s social capital: The national
measurement with the Barcas. Washington, DC: World
Bank.
2277
Thoits, P. (1995). Stress, and social support processes: Where
are we? What next? Journal of Health and Social Behaviour,
35(Special number), 53–79.
!
Urrea, C., & Ortiz, C. (1999). Patrones Sociodemograficos,
Pobreza y Mercado Laboral en Cali. Documento de trabajo
para el Banco Mundial, unpublished.
Uslaner, E. (2002). The moral foundations of trust. Cambridge:
Cambridge University Press.
WHO (1994). Users guide to the self-reporting questionnaire
(SR20). Geneva: World Health Organisation.
WHO (2001). The world health report: Mental health. Geneva:
World Health Organization.
Wilkinson, R. (1996). Unhealthy societies. London:
Routledge.
Wilkinson, R. (1999). Income inequality, social cohesion and
health. International Journal of Health Services, 29(3),
525–543.