Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
ROSE
MICHAEL E. WOOLLEY
GARY L. BOWEN
University of Maryland
545
546
Family Relations
547
supports: the organization, various combinations, and singularity of these supports determine how effectively they can be processed
into investments that can be utilized to achieve
objectives (Brunie, 2009; Hayami, 2009). In the
most typical approach to measurement, however,
these factors are overlooked in favor of a variable reduction strategy: social capital spanning
multiple dimensions is typically quantified by
summing items reported by youth respondents,
across or within microsystem domains, using a
simple index or a measurement method such as
factor analysis (e.g., Woolley & Bowen, 2007).
These counting approaches lead to a quantification of social capital as a continuous or count
outcome, rather than a discrete measure of the
typology of different combinations of social capital (Croninger & Lee, 2001; Morrow, 1999;
Reimer, Lyons, Ferguson, & Polanco, 2008).
The result is a measure that contains fewer
dimensions but does not differentiate according to the source or nature of the social capital
(Croninger & Lee, 2001; Reimer et al., 2008),
or differentiates between domains but treats
these domain-specific sources as inherently distinct supports (Wenz-Gross, Siperman, Untch,
& Widaman, 1997; Woolley & Grogan-Kaylor,
2006). However, two youths having similar levels of social capital as represented by a count
may in fact have very distinct portfolios that constitute that capital. The literature indicates that
these differences matter. A portfolio approach
to measurement retains important information
about not only the level of social capital obtained
through a count, but also information about the
source, combination, and pattern of social capital, which is lost in the process of counting.
We suggest that operationalizing and measuring social capital as Coleman (1988) conceived
it requires measures of resources from multiple
sources as well as a method that captures how
those varied sources come together to influence
youth development. A discrete latent construct
of social capital preserves information about the
sources from which the portfolio is constructed;
defining typologies within the student population
represent groups of students that have common
patterns.
Accordingly, the purpose of this study was to
investigate a LPA strategy for measuring social
capital as a capital investment consisting of a
portfolio of multisystem social supports. We
now turn to discuss LPA, a method that has
shown some promise in this regard.
548
Family Relations
METHOD
549
Minimum
Maximum
Mean as Proportion
of Maximum
SD
Reliability
0
0
0
0
0
12
11
5
7
6
8.1
8.7
3.2
4.8
3.8
0.68
0.79
0.64
0.69
0.63
3.1
2.8
1.8
2.5
2.3
0.81
0.86
0.86
0.91
0.92
0
0
238
232
8
3
303
286
6.6
1.2
265.9
258.5
0.83
0.40
1.7
1.2
10.1
8.9
0.79
0.78
550
these dimensions ranged from .81 for neighbor
support to .92 for parent support. The range of
each scale varied by the number of items; the
smallest (friend support) had a range of 0 to 5
and the largest had a range of 0 to 12 (neighbor
support). We normed the scales on a 0-to-1
metric such that a 0 indicated having none of
that type of social capital, while a 1 indicated
all of that type. Renorming the scales facilitated
an intuitive interpretation of the results for each
portfolios means on the manifest indicators:
these means now constituted proportions of the
available resources within each social capital
dimension. This is reported in Table 1 in the
column Mean as Proportion. We ran the LPA
on these renormed indicators.
Model Fit and Class Labels. Following Nylund,
Asparouhov, and Muthen (2007), we determined
the most appropriate number of classes in the
LPA model by starting with the minimum
number of classes required (N = 2), then adding
one class at a time while evaluating each
model for improvement in fit, a process that
continued until the addition of an additional
class did not improve model fit. The LoMendell-Rubin (LMR) likelihood ratio test
(Lo, Mendell, & Rubin, 2001) was used to
compare model fit. Other criteria such as
the Akaike information criterion (AIC) and
Bayesian information criterion (BIC), including
a sample adjusted BIC, were considered as well.
Entropy was considered to evaluate the overall
fit of the model to the data. Qualitative factors
were also important, such as the face validity and
conceptual coherence of the classes that emerged
(Nylund et al., 2007). Mplus (Muthen & Muthen,
2004) was used for estimating the models. The
expectation-maximization algorithm was used
to fit the LPA solution.
Predictive validity. To determine the predictive
value of the portfolios identified through LPA,
we conducted a predictive validity analysis
incorporating the portfolios into multilevel
models of student self-reported outcomes and
end-of-grade scores for the same academic
year (2004 2005), with adjusted standard
errors for school nesting using multilevel
modeling (Raudenbush & Bryk, 2002). All
models were also conditioned on students
race/ethnicity, gender, receipt of free/reduced
price lunch, and past grade retention. We
hypothesized that classes revealed to have
Family Relations
more robust portfolios consisting of higher
proportions of more types of social capital
would have better outcomes. Two school
outcome dimensions from the SSPschool
engagement and trouble avoidanceare selfreports by youth that describe their involvement
and behavior in school. School engagement was
measured using a three-item scale and trouble
avoidance using an eight-item scale. End-ofgrade standardized test performances in math
and reading for school year 2004 2005 derive
from administrative data collected by the North
Carolina Education Data Research Center at
Duke University. Univariate statistics of selfreport and end-of-grade outcomes are reported
in Table 1. To incorporate each students
portfolio, portfolios being probabilistic not
deterministic, we assumed that each student was
best characterized by the portfolio on which he
or she had the highest probability of membership
(a classify-analyze type of post hoc analysis).
Missing data. The percentage of missing data
for each dimension ranged from 3 (peer group
acceptance) to 10 (social support use). Of
the 6,976 students surveyed, 6,914 students
in sixth through eighth grades were retained
for the present analysis; 62 students (less than
1%) did not have valid responses on enough
of the constituent items used in the analysis
and their surveys were consequently discarded.
Mplus implements full information maximum
likelihood (FIML) estimation for missing values.
Although FIML requires an assumption of
missing at random (Schafer 1997), it uses all
available data for each participant, resulting in
no additional loss of cases. Variables used in
the predictive analysis had higher percentages
of missing data. Missing scale score dependent
variables (850 for math and 864 for reading)
were also treated as FIML (Schafer, 1997).
Among the remaining records, 125 students were
missing the free/reduced price lunch indicator,
15 students were missing race/ethnicity, 21
students were missing the retention indicator,
and 3 students were missing gender; a total of
161 (2.3% of sample) records were not used in
the predictive validity analysis as a result.
RESULTS
Number of Portfolios
Starting with two portfolios and working
upwards, we determined that the best-fitting
551
43
12
9
6
9
4
6
13
Neighbor Support Teacher Support Friend Support Family Togetherness Parent Support
0.78
0.69
0.56
0.60
0.62
0.56
0.66
0.49
0.90
0.82
0.71
0.77
0.78
0.28
0.80
0.61
0.77
0.63
0.51
0.56
0.53
0.56
0.63
0.45
0.95
0.82
0.33
0.44
0.35
0.91
0.91
0.08
0.96
0.54
0.39
0.10
0.84
0.94
0.06
0.05
552
Family Relations
553
High
HO
H
NP
MHO
HO
H
AA
Medium
MHO
EFA
AA
T
EFA
NP
LH
LH
Parent Support
Family Togetherness
Low
HO
MHO
NP
AA
EFA
T
HO
MHO
NP
HO
MHO
NP
AA
EFA
T
H
LH
EFA H
AA
T
LH
H
LH
Neighborhood Support
Friend Support
Teacher Support
Teacher = T
Extra-Familial Adults = EFA
All Adults = AA
Home = H
Non-parental = NP
Low Home = LH
School Engagement
High overall
Medium-high overall
Teacher
Extrafamilial adults
All adults
Home
Nonparental
Low home
Trouble Avoidance
Math
Reading
Reference condition
0.34
0.68
0.50
0.50
0.77
0.16
0.82
0.18
0.52
0.64
0.34
0.90
0.47
0.68
0.37
0.00
1.03
0.22
1.43
2.55
0.39
0.00
0.01
1.53
0.75
0.44
2.27
0.16
Note: All models conditioned on race/ethnicity, gender, free/reduced lunch, and retention.
p < .05. p < .01. p < .001.
554
Family Relations
These findings endorse the inclusion of multiple sources of social capital spanning the four
central microsystems of youth. Within this multimicrosystem approach, the portfolios displayed
some unique tendencies that challenged and confirmed our hypotheses. First, in confirmation of
our assumptions, in the high overall class, students reported feeling high levels of support and
satisfaction from their interactions with neighborhood adults, teachers, peers, and at home.
Further, these adolescents tended to outperform
their peers in school engagement and trouble
avoidance, as well as in selected portfolios in
academic performance.
Second, though one portfolio was characterized by very low social capital at home, the
students in this portfolio demonstrated greater
reading and math performance than some of the
moderate portfolios. This may be due to the level
of teacher support in this low home portfolio;
prior research including multiple microsystem
measures of social capital has revealed teacher
support as the most powerful predictor of school
performance (e.g., Woolley & Grogan-Kaylor,
2006). It may also be because of the otherwise
relative uniformity of the levels of support across
microsystems, whereas in the moderate portfolios, there were often significant differences by
setting. Thus, the effect of diversification and
consistency may be key facets to examine.
Third, and confirming our assumptions, we
found that students with portfolios with multiple
sources of adult social capital, such as all adults
or nonparental, had higher school engagement
than those with only one source of adult social
capital. On the other hand, the critical sources
of social capital differed for trouble avoidance;
any set of multiple sources were better than any
single source, even if one of those sources was
from peers.
Fourth, the five moderate portfolios that
emerged exhibit the variability in form and level
we hypothesized LPA would identify, and we
suggest that this is an advantage over simple
counting methods. These portfolios indicate that
most of the variability across moderate portfolios
lies in the relative levels of the two home
dimensions. Aside from high and medium-high,
the portfolio with the highest levels of home
supports, nonparental, was among those most
predictive of school outcomes; the difference
between this moderate portfolio and the high
overall and medium-high overall portfolios was
not determined by the level of parent support.
555
an excellent job of measuring the resources in
youths social networks, new measures developed from the social capital conceptualization
advocated above would differ, which presents
an important and ambitious agenda for social
capital research going forward.
Second, because the levels of social capital
cannot be randomly assigned, the evidence of
portfolio pattern effects in the predictive validity
analysis should not be inferred as causal. Other
correlates that are unmeasured may be responsible for the observed relationships, and we cannot
be certain of who is doing the investing in
the portfolios; for example, high teacher support
could derive from the agency of the student or of
the teacher. In future studies examining the relative benefit of these portfolios, some statistical
adjustments such as propensity score matching
of children with the high social capital may help
build causal inferences (Guo & Fraser, 2009).
Third, the predictive validity assessment
uses probabilistic constructs in a deterministic
way, which we remark on further in the
Implications. Fourth, although this approach
can quantify from whom, in what settings,
and how much or often in aggregate, these
dynamics still may not capture the quality of
the social capital experienced, a problem that
plagues all measurement strategies, even those
using putatively qualitative methods such as the
discrete latency of LPA. Sixth, although the
supports can be seen as representing a form
of investment, it is not clear that two children
with the same portfolio will benefit similarly.
Part of this uncertainty is due to the reported
portfolios consisting of the means of the support
levels in each dimension (such that there is
variability within each portfolio) and partly due
to differences in the other forms of capital
(financial, cultural, and human, primarily) that
children can access.
In addition, although we characterize LPA
as an empirically derived approach to classification, the approach is not completely free of
investigator bias. The selection of dimensions
that constitute the measurement of social capital
can be a source of bias. For example, in an
earlier investigation of this approach, a second
dimension of peer support, peer group acceptance, which measured adolescents perceptions
of relative standing in their peer groups and
their ability to retain their individuality although
resisting peer pressure, was included. Unlike
friend support, the items used to measure peer
556
Family Relations
REFERENCES
Bourdieu, P. (1986). The forms of capital. In J. E.
Richardson (Ed.), Handbook of theory for research
in the sociology of education (pp. 46 58).
Westport, CT: Greenwood Press.
Bowen, G. L., Rose, R. A., & Bowen, N. K. (2005).
The reliability and validity of the school success
profile. Philadelphia: Xlibris Press.
557
Bowen, N. K., Bowen, G. L. & Ware, W. B. (2002).
Neighborhood social disorganization, families, and
the educational behavior of adolescents. Journal
of Adolescent Research, 17, 468 490.
Bronfenbrenner, U. (Ed.). (2005). Making human
beings human: Bioecological perspectives on
human development. Thousand Oaks, CA: Sage.
Brunie, A. (2009). Meaningful distinctions within a
concept: Relational, collective, and generalized
social capital. Social Science Research, 38,
251 265.
Coleman, J. C. (1988). Social capital in the creation
of human capital. American Journal of Sociology,
94(Suppl), S95 S120.
Croninger, R. G., & Lee, V. E. (2001). Social capital
and dropping out of high school: Benefits to atrisk students of teachers support and guidance.
Teachers College Record, 103, 548 581.
Crosnoe, R. ( 2004). Social capital and the interplay
of families and schools. Journal of Marriage and
Family, 66, 267 280.
Dika, S. L., & Singh, K. (2002). Applications of
social capital in educational literature: A critical
synthesis. Review of Educational Research, 72,
31 60.
Fraser, M. W., & Galinsky, M. J. (2010). Steps in
intervention research: Designing and developing
social programs. Research on Social Work
Practice, 20, 459 466.
Furstenburg, F. F., & Hughes, M. E. (1995). Social
capital and successful development among at-risk
youth. Journal of Marriage and the Family, 57,
580 592.
Gerard, J. M., & Buehler, C. (2004). Cumulative
environmental risk and youth maladjustment: The
role of youth attributes. Child Development, 75,
1832 1849.
Glanville, J. L., & Bienenstock, E. J. (2009).
A typology for understanding the connections
among different forms of social capital. American
Behavioral Scientist, 52, 1507 1530.
Gofen, A. (2009). Family capital: How firstgeneration higher educational students break the
intergenerational cycle. Family Relations, 58,
104 120.
Guo, S., & Fraser, M. W. (2009). Propensity score
analysis: Statistical methods and applications.
Thousand Oaks, CA: Sage.
Hagenaars, J. A., & Halman, L. C. (1988). Searching
for ideal types: The potentialities of latent
class analysis. European Sociological Review, 5,
81 96.
Hayami, Y. (2009). Social capital, human capital
and the community mechanism: Toward a
conceptual framework for economists. Journal of
Developmental Studies, 45, 96 123.
Hoffman, J. P., & Dufur, M. J. (2008). Family and
school capital effects on delinquency: Substitutes
558
or complements? Sociological Perspectives, 51,
29 62.
Lillibacka, R. (2006). Measuring social capital:
Assessing construct stability of various operationalizations of social capital in a Finnish sample.
Acta Sociologica, 49, 201 220.
Lo, Y., Mendell, N. R., & Rubin, D. B. (2001). Testing
the number of components in a normal mixture.
Biometrika, 88, 767 778.
Loken, E. (2004). Using latent class analysis to
model temperament types. Multivariate Behavioral Research, 39, 625 652.
Morrow, V. (1999). Conceptualising social capital in
relation to the well-being of children and young
people: A critical review. Sociological Review, 47,
744 764.
Muthen, B. O., & Muthen, L. (2004). Mplus users
guide. Los Angeles: Author.
National Council on Family Relations. (2007).
CFLE practice analysis report. Minneapolis, MN:
Author.
Nylund, K. L., Asparouhov, T., & Muthen, B. O.
(2007). Deciding on the number of classes in latent
class analysis and growth mixture modeling: A
Monte Carlo simulation study. Structural Equation
Modeling, 14, 535 569.
Parcel, T. L., Dufur, M. M., & Zito, R. C. (2010).
Capital at home and at a school: A review and
synthesis. Journal of Marriage and Family, 72,
828 846.
Portes, A. (1998). Social capital: Its origins and
applications in modern sociology. Annual Review
of Sociology, 24, 1 24.
Raudenbush, S. W. & Bryk, A. S. (2002). Hierarchical
linear models: Applications and data analysis
methods (2nd ed.). Thousand Oaks, CA: Sage.
Reimer, B., Lyons, T., Ferguson, N., & Polanco, G.
(2008). Social capital as social relations: The
Family Relations
contribution of normative structures. Sociological
Review, 56, 256 274.
Rosenfeld, L. B., Richman, J. M., & Bowen, G.
L. (2000). Social support networks and school
outcomes: The centrality of the teacher. Child &
Adolescent Social Work Journal, 17, 205 226.
Schafer, J. L. (1997). Analysis of incomplete
multivariate data. Boca Raton, FL: Chapman
Hall/CRC.
Van Lier, P. A. C., Verhulst, F. C., & Crijnene, A.
A. M. (2003). Screening for disruptive behavior
syndromes in children: The application of latent
class analysis and implications for prevention
programs. Journal of Consulting and Clinical
Psychology, 71, 353 363.
Vaquera, E., & Kao, G. (2008). Do you like me as
much as I like you? Friendship reciprocity and
its effects on school outcomes among adolescents.
Social Science Research, 37, 55 72.
Wenz-Gross, M., Siperman, G. N., Untch, A. S., &
Widaman, K. F. (1997). Stress, social support,
and adjustment of adolescents in middle school.
Journal of Adolescence, 17, 129 151.
Woolley, M. E., & Bowen, G. L. (2007). In the
context of risk: Supportive adults and the school
engagement of middle school students. Family
Relations, 56, 92 104.
Woolley, M. E., & Grogan-Kaylor, A. (2006). Protective family factors in the context of neighborhood:
Promoting positive school outcomes. Family Relations, 55, 93 104.
Woolley, M. E., Kol, K., & Bowen, G. L. (2009). The
social context of school success for Latino middle
school students: Direct and indirect influences of
teachers, family, and friends. Journal of Early
Adolescence, 29, 43 70.