Demystifying the Measurement of Complex Social Constructs: Assessing Social Capital Across Sectors WILLIAM STORY and TIM FRANKENBERGER

1. Will Story, PhD, MPH
Department of Community and
Behavioral Health
University of Iowa College of Public Health
Tim Frankenberger, PhD
TANGO International
CORE Group Global Health Practitioner Conference
Portland, OR
May 18, 2016

2.  Reviewed the approaches to creating valid and
reliable measures of complex social constructs.
 Examined the design of social capital
assessment tools and discovered how the tools
have been used in development programs.
 Compared various assessment tools for
measuring social capital and discussed how to
adapt them for wider use in population-based
surveys.

3.  A social mechanism, phenomenon, or
category created and developed by society
 A perception of an individual, group, or idea
that is 'constructed' through cultural or social
practice
 A proposed attribute of a person (or group of
people) that often cannot be measured
directly, but can be assessed using a number
of indicators (i.e., latent variable)

4.  The construct must be carefully defined and
delineated before it can be measured
 Single measure
 With or without a scale
 Multiple measures
 Additive score
 Factor score
 Continuous, ordinal, dichotomous

5.  Validity:The accuracy of a measurement
 The degree to which the measures in your project
reflect the theoretical constructs on which those
measures are based.
 Reliability:The consistency of a measurement
 The degree to which a measure would give you the
same result over and over again, assuming the
underlying phenomenon is not changing.

6. (Trochim and Donnelly, 2008)

7.  Translation validity
 Focus on whether the operationalization is a good
reflection of the construct
 Criterion-related validity
 Examine whether the operationalization behaves
the way it should given your theory of the
construct
(Trochim and Donnelly, 2008)

8.  Expert review: Assess whether the survey questions
are appropriate for measuring social capital
 Suggestions from peer-reviewed literature (De Silva et al.,
2006); independent review by social capital expert; and
final review by our research team from ICDDR,B
 FGDs and cognitive interviews: Explore what people
know about social capital and examine how
respondents formulate their answers
 We conducted four FGDs and 32 cognitive interviews
consisting of 18 structured survey questions followed by
scripted verbal probes
(Story et al., 2015)

10.  Internal consistency
 Homogeneity of multiple items within an
instrument or specific domain
 Inter-rater reliability
 Extent that two or more observers agree with one
another
 Test-retest reliability
 Extent that the same results are obtained on
repeated administrations of the same instrument
(Trochim and Donnelly, 2008)

11.  Five dimensions of community capacity were
used to develop a survey instrument:
leadership, resources, problem assessment,
connections and networks, and community
attitudes
 The initial scale with 41 items had a
Cronbach's alpha coefficient of 0.89
indicating good internal consistency and this
was evident over all four communities
(Lovell et al., 2014)

12.  Trochim,W. M. K., & Donnelly, J. P. (2008).The research methods
knowledge base. Mason,OH: Cengage Learning or
http://www.socialresearchmethods.net/kb/
 Story,W.T.,Taleb, F.,Ahasan, S.M., & Ali, N.A. (2015).Validating the
measurement of social capital in Bangladesh:A cognitive approach.
Qualitative Health Research, 25(6), 806-819.
 De Silva, M. J., Harpham,T.,Tuan,T., Bartolini, R., Penny, M. E., & Huttly,
S. R. (2006). Psychometric and cognitive validation of a social capital
measurement tool in Peru andVietnam. Social Science & Medicine, 62,
941–953.
 Lovell, S.A., Gray, A.R., Boucher, S.E. (2014). Developing and validating a
measure of community capacity:Why volunteers make the best
neighbors. Social Science and Medicine, 133, 261-268.

13.  What complex social constructs are you
currently trying to measure?
 What have you done (or can you do) to assess
the validity and reliability of these
constructs?

14.  What is social capital?
 Household social capital and socioeconomic
inequalities in child undernutrition:
Evidence from rural India
 The effects of social capital on resilience:
Evidence from Ethiopia

15.  A broad term including social relationships,
networks, and values that facilitate collective
action for mutual benefit
 Theories of social capital tend to be
dichotomized into two forms: structural and
cognitive
 Structural:What people do as measured by
individuals’ actions and behaviors
 Cognitive: How people feel as measured by
individuals’ attitudes and perceptions
(Woolcock, 1998; Harpham et al., 2002; Krishna & Shrader, 2000)

16.  A survey-based approach is commonly used to
measure specific components of social capital
 Structural social capital
 Group membership/involvement
 Informal social relationships
 Collective action
 Cognitive social capital
 Trust (generalized or interpersonal)
 Social cohesion
(Harpham et al., 2002; Narayan & Cassidy, 2001)

17.  Bonding: social ties that connect similar people (e.g., a
religious or ethnic group), build in-group loyalty, and
provide resources
 Bridging: social ties that connect dissimilar people (e.g.,
a social movement, different demographic groups) and
are more ideal for accessing external assets and
information
 Linking: social ties to people of power in a society (e.g.,
health care providers, school teachers, and government
officials), who can help leverage resources, ideas, and
information
(Szreter &Woolcock, 2004)

18. William T. Story, PhD, MPH
Department of Community and
Behavioral Health
University of Iowa College of Public Health
Richard M. Carpiano, PhD, MA, MPH
Department of Sociology
School of Population and Public Health
University of British Columbia
CORE Group Global Health Practitioner Conference
Portland, OR
May 18, 2016

19.  3.1 million annual deaths among
children under the age of five due
to undernutrition
 Strong association between child
nutrition status and household
wealth persists in the developing
world
 Social capital can potentially be
used by the poor as a primary
means of protection against risk
and vulnerability
(Black et al. 2013; Boyle et al. 2006; Carroll 2001; Story 2013)

20.  India is home to the largest number of under-
nourished children globally
 42.5 % of children below the age of 5 years are
underweight for age (highest in South Asia)
 Two-thirds of the population is rural, which leads
to greater disparities in access to resources
 Prior studies have shown unequal access to
social capital and differential associations
between social capital and health
(Ramachandran 2014; Story 2014;Vikram et al. 2012; De Silva et al. 2007)

21. 1) How does household SES
shape access to bonding,
bridging, and linking social
capital?
2) If these social capital forms
are accessed, how are their
influences on child
underweight shaped by
household SES?

22.  SES and Possession of Social Capital
 H1a – Opportunity Hoarding Explanation:
Higher household SES is associated with higher
odds of bonding, bridging, and linking social
capital
 H1b – Coping Explanation: Lower household SES
is associated with higher odds of membership in
bonding organizations
(Wright 2009; Cattell 2001; Altschuler, Somkin, andAdler 2004)

23.  SES as a Moderator of Social Capital and Child
Underweight
 H2a – Resource Substitution Explanation: Social
capital demonstrates a stronger reduction in the odds
of child underweight in lower (versus higher) SES
households.
 H2b – Fundamental Cause Explanation: Social
capital demonstrates a stronger reduction in the odds
of child underweight in higher (versus lower) SES
households.
(Uphoff et al. 2013; Carpiano 2006; Story 2014; Carpiano, Link, and Phelan 2008)

24.  Data source
 2005 India Human Development Survey (IHDS)
 Analytic sample
 9,008 children under the age of 5
 6,753 households
 1,347 rural villages
 Unit of analysis
 Children within households within villages
(Desai et al. 2005)

25.  Child underweight
 Below -2 standard
deviations from the median
weight-for-age of the
international reference
population
 Binary variable where 1 =
Underweight
(Fishmann et al., 2004)

26.  Household SES
 Index of 30 dichotomous items reflecting asset
ownership and housing quality
 Bonding ties (e.g., religious or caste groups)
 Bridging ties (e.g., women’s groups, saving and
credit groups, youth groups)
 Linking ties
 Doctor or nurse
 Teacher or principal
 Government officer

27.  Exploratory factor analysis was used to
validate the components of social capital in
the IHDS
 6 factors were identified and explained 82.6%
of the total variance in the 15 social capital
questions
 3 of those factors (bonding, bridging, and
linking social capital) were used in this study

28.  Multilevel logistic regression analysis was used
to estimate:
 The overall association between SES, social capital
and child underweight adjusting for individual,
household, and village characteristics
 The variation in child underweight between
households and communities
 Interactions between each social capital variable
and household SES were examined using the
margins command in Stata 13.0
(Ai and Norton 2003)

29. Adjusted odds ratios (95% confidence intervals) of possessing each form of household social
capital as predicted by household assets after controlling for other household and village
characteristics.
Organizational Ties LinkingTies
Bonding Bridging Medical School Government
Household
assets
1.07***
(1.04-1.10)
1.07***
(1.05-1.09)
1.10***
(1.07-1.12)
1.10***
(1.07-1.12)
1.16***
(1.12-1.20)
* p<0.05; ** p<0.01; *** p<0.001
Note: n=7,568 households within 1,379 villages; Each cell represents a separate model. Each
model controls for all household-, and village-level independent variables.

30.  H1a – Opportunity Hoarding Explanation:
Higher household SES is significantly
associated with greater odds of bonding,
bridging, and linking social capital.
 H1b – Coping Explanation: Our findings do
not support the hypothesis that lower
household SES is associated with greater odds
of membership in a bonding organization.

32.  H2a – Resource Substitution Explanation:
Poorer households with ties to bridging
organizations and school personnel are less
likely to have an underweight child compared
to poorer households without such ties.
 H2b – Fundamental Cause Explanation:
Wealthier households with a medical tie are
less likely to have an underweight child
compared to wealthier households without a
medical tie.

33.  One cannot assume that a social capital
strengthening initiative will be equally
beneficial to all members of society
 Socially excluded households must be given
special attention
 Future research should assess equity in the
delivery and utilization of social capital
strengthening interventions

34.  Interventions should continue to promote
social ties to bridging organizations,
especially among poorer households
 Need to better understand the pathways
through which bridging ties lead to better
child nutrition by exploring organizational
strategies related to better child feeding and
treatment seeking practices

35.  Social connections to medical professionals
are insufficient to reduce disparities in child
nutrition
 Access to a doctor or nurse must be
combined with sufficient household
resources to rehabilitate malnourished
children

36.  Social ties to teachers and principals may have
an impact beyond the effect of formal education
 It is important to identify what aspects of the
education system are helping to eliminate
socioeconomic disparities in child nutrition
 Are children and their parents establishing new
social connections to help improve the overall
health and well-being of their household?

37. The effects of social capital on
resilience: Evidence from Ethiopia
Tim Frankenberger
May 18th, 2016
Core Group Global Health Practitioner
Conference

38. Importance of Social Capital
• Previous research demonstrates that the extent
and application of social capital strongly
influences community level resilience.
• Disasters may sometimes enhance social capital
because they activate or give rise to
neighborhood associations and collective
organizations that can be used to disseminate
vital information, provide community members
with a voice, and afford leverage to assist in
taking control of rebuilding efforts.

39. *Aldrich 2012; Wilson 2012; Magis 2010; Elliott et al. 2010
Figure Source: Reproduced with permission from Aldrich (2012, p. 34) in Frankenberger, T., Mueller M., Spangler
T., and Alexander S. October 2013. Community Resilience: Conceptual Framework and Measurement Feed the
Future Learning Agenda. Rockville, MD: Westat.

40. Hypotheses to be Tested
• Households with greater levels of social capital (bonding,
bridging, and linking) achieve greater levels of food security
than those with less social capital, all else equal.
• Households with greater levels of social capital (bonding,
bridging, and linking) are able to recover better than those
with less social capital, all else equal
• For a given level of exposure to shocks, households with
more social capital report fewer negative impacts of shocks
than households with less social capital, all else equal.

41. Empirical Evidence
• This presentation will examine empirical
evidence from several studies focused on
measuring resilience
• Pastoralist Areas Resilience Improvement and
Market Expansion (PRIME) program in
Ethiopia

42. Studies: PRIME
• Pastoralist Areas Resilience
Improvement through Market
Expansion
– USAID Ethiopia Feed the Future
– Project goals:
– increase household incomes
– enhance resilience
– Improve climate change adaptive capacity
• Program beneficiaries
– pastoralists, non-pastoralists, and other
– Geographic location
– 2 areas in Ethiopia (Borena and Jijiga)
– Data
– Baseline (2013)
– Interim monitoring data (2014 – 2015, 6
months)

43. Samples from Project areas
Project area
# of
households
# of
communities
PRIME
Jijiga 1398 32
Borena 1744 41

44. Methodology
• In order to measure the impact social capital has on
resilience, indices were created for bonding, bridging and
linking social capital.
• The bonding social capital index is based on eight yes/no
questions about whether the household would be able to
give or receive help from relatives or non-relatives in their
community.
• The bridging social capital index is based on eight similar
yes/no questions, but about giving and receiving help from
relatives or non-relatives living outside their community.

45. Methodology
• The linking social capital index measures the
amount of information received from
government agents (i.e., rural development
agents and government/political officials).
• The index also measures households’ access
to services generally provided by the
government and the quality of those services

46. Methodology
• All three forms of social capital are composite measures
based on multiple other measures.
• In many of these cases Principal Components Analysis
(PCA) or polychoric factor analysis were used to validate
each form of social capital.
• Simply put, these techniques reduce a set of “input”
variables that are hypothesized to be related to one
another.
• Three indices were constructed using this technique

47. Methodology
• The dependent variable Household food
security is the inverse of an experiential
indicator of food insecurity, the Household
Food Insecurity Access Scale (HFIAS).
• The HFIAS is an index constructed from the
responses to nine questions regarding
people’s experiences of food insecurity in the
previous four weeks

48. Methodology
• The dependent variable, recovery takes into
account households’ ability to recover from
climatic, conflict, and/or economic shocks using a
4-point likert scale.
• The dependent variable HH shock impact is an
index which takes into account if a household
experienced a shock within the last 12 months,
how many times they experienced a shock with in
the last 12 months, and how severe the impact of
the shock was on income and food consumption.

49. Multivariate regression analysis

50. Multivariate regression analysis

51. Food Security Results for PRIME
Table 1. Relationship between social capital and household food security for PRIME baseline
Indicators
Jijiga Borena
Coefficient Elasticity Coefficient Elasticityn n
Social capital
Bonding
social capital
0.005 0.030 1236 0.072 *** 0.732 1566
Bridging
social capital
0.015 0.057 1253 0.054 *** 0.402 1624
Linking social
capital
0.025 0.105 1253 -0.005 -0.029 1624
NOTES: Stars represent statistical significance at the 0.01 (***), 0.05 (**) and 0.1 (*) levels.

52. Food Security Summary
• Bonding and bridging social capital are
significantly associated with increased food
security in Borena but not in Jijiga

53. Recovery Results for PRIME
Table 2. Relationship between social capital and recovery for PRIME baseline
Indicators
Jijiga Borena
Coefficient Elasticity Coefficient Elasticityn n
Social capital
Bonding
social capital
0.009 *** 0.212 1127 0.005 *** 0.152 1430
Bridging
social capital
0.007 *** 0.110 1146 -0.002 * -0.041 1476
Linking social
capital
0.043 *** 0.757 1146 0.004 0.073 1476
NOTES: Stars represent statistical significance at the 0.01 (***), 0.05 (**) and 0.1 (*) levels.

54. Recovery Summary
• In both Jijiga and Borena bonding and bridging
social capital enabled households to recover
• Linking social capital was important for
recovery in Jijiga but not Borena

55. Shock Impact Results for PRIME
Table 3. Relationship between social capital, asset index, and number of shocks on shock exposure for
PRIME baseline
Jijiga Borena
Indicators
Only
bonding
social
capital
Only
bridging
social
capital
Only
linking
social
capital
Only
bonding
social
capital
Only
bridging
social
capital
Only
linking
social
capital
(A) (B) (C) (A) (B) (C)
Social capital
Bonding
social capital -0.011
***
-0.008
***
Bridging
social capital -0.011
***
0.012
***
Linking
social capital 0.000 0.004
Asset index 0.002 -0.002 -0.006 -0.015
*
-0.032
***
-0.020
***
Number of
shocks 3.564
***
3.558
***
3.563
*
*
* 3.611
***
3.599
***
3.592
***
Number of
observations
1324 1351
1352
1618 1618
1618
NOTES: Stars represent statistical significance at the 0.01 (***), 0.05 (**) and 0.1 (*) levels.

56. Shock Impact Summary
• Bonding and bridging social capital help
mitigate the effect of shocks in Borena and
Jijiga
• Linking social capital does not have an effect
in either Jijiga or Borena

57. Conclusions
• Social capital appears to have a positive effect on food
security, helps households recover and mitigates the
effect of shocks across the different data sets
• Thus social capital appears to be critical to resilience
• Wealthier households appear to receive the benefits of
social capital more than poorer households
• Social capital can be used up in the early phases of a
prolonged covariate shock and its downstream effects

58.  Review the three social capital assessment
tools individually.
 In your groups discuss the following:
 How might you adapt one or more of these tools
for use in your own programs?
 What barriers might you face when trying to
measure social capital?
 How will you overcome those barriers?

59. Will Story
william-story@uiowa.edu
Tim Frankenberger
tim@tangointernational.com