ERIC ED608897: Paths 2 the Future: Evidence for the Efficacy of a Career Development Intervention for Young Women with Disabilities

2020·0.57 MB·English

by  ERIC

924851
research-article2020 ECXXXX10.1177/0014402920924851Exceptional ChildrenLindstrom et al.
Original Research
Exceptional Children
Paths 2 the Future: Evidence 2020, Vol. 87(1) 54 –73
© The Author(s) 2020
for the Efficacy of a Career hDttOpsI:/:/ d1o0i..o1r1g/7170./1010717/40401042490229029029424885511
journals.sagepub.com/home/ecx
Development Intervention for
Young Women With Disabilities
Lauren Lindstrom1, David DeGarmo2, Atika Khurana2,
Kara Hirano3, and Leslie Leve2
Abstract
Young women with disabilities face unique barriers in the transition from school to adulthood,
yet very few studies have examined the effectiveness of gender-specific career interventions.
Using an intent-to-treat analysis, this study tested the efficacy of the Paths 2 the Future (P2F)
career development curriculum to produce beneficial impacts as compared to business-as-usual
career and transition services within a clustered, school-randomized trial. The sample included
366 young women with disabilities enrolled in 26 high schools. Controlling for student and school
characteristics, multilevel growth models showed that young women in the P2F intervention
schools grew in career development skills at a greater rate relative to those in the control
schools. Findings suggest that P2F was effective in promoting individual student differences in
career development skills for young women with disabilities and support the need for further
research examining effectiveness of gender-specific career interventions in special education.
Although the Individuals With Disabilities women with disabilities are more likely to
Education Act (2006) states that the funda- experience poverty in adulthood (Fins, 2019).
mental purpose of a free appropriate public The process of career development for young
education is to prepare all students with dis- women with disabilities is impacted by both indi-
abilities for “further education, employment, vidual and structural barriers. Individual attributes,
and independent living” (20 U.S.C. § 1400[d] such as low self-esteem, limited self-efficacy, and
[1][A]), youth with disabilities consistently a lack of self-advocacy skills can restrict the
experience lower rates of employment and are ability to fully explore a wide range of career
less likely to enroll in or complete postsecond- options. In addition, young women with disabili-
ary education than their nondisabled peers ties often have limited opportunities for career
(Butterworth & Migliore, 2015; U.S. Bureau exploration activities during high school (Ferri &
of Labor Statistics, 2020). Young women with Conner, 2010; Lindstrom et al., 2012), are less
disabilities, in particular, face unique chal- likely than their male peers to enroll in vocational
lenges in accessing postsecondary education, courses or participate in community work experi-
gaining employment, living independently, ences, and may face discrimination related to both
and fully participating in their communities
(Hogansen et al., 2008; Lindsay et al., 2017).
1University of California, Davis
After leaving high school, females with dis-
2University of Oregon
abilities are more likely than males to be 3Illinois State University
employed part-time and earn lower wages and
Corresponding Author:
are less likely to work in high-skill or high-
Lauren Lindstrom, School of Education, University of
wage jobs (Doren et al., 2011). The persistence
California, Davis, One Shields Ave., Davis, CA 95616.
of these inequities over time suggests that Email: [email protected]
Lindstrom et al. 55
gender and disability identities (Lindsey et al., der awareness. Initial field test data indicated
2017; Noonan et al., 2004). This restricted set of improved employment rates and enrollment in
opportunities and experiences translates to a nar- postsecondary education for the young women
row range of career interests and aspirations, (N = 54) who completed the Girls at Work pro-
ultimately resulting in poor long-term educa- gram (Parent & Tanis, 2011).
tional and employment outcomes (Doren et al., More recently, the Post-School Achievement
2011; Lindsey et al., 2017). Through Higher Skills (PATHS) curriculum was
Despite the complex combination of barri- developed to promote social cognitive career
ers faced by young women with disabilities as and self-determination outcomes and address
they transition from school to adult roles, rela- the unique learning needs of young women with
tively few career development interventions disabilities (Lindstrom et al., 2013). Lesson top-
have been developed to specifically focus on ics, derived from special education and career
ameliorating these gender gaps. One of the few development literature, include self-determina-
exceptions is Career Connections, a 24-lesson tion and strengths, gender identity, disability
curriculum designed to improve prosocial and knowledge, and career exploration. Participants
career development outcomes for young in PATHS were young women identified by
women with disabilities. This weekly program school personnel as either eligible for special
aims to improve participants’ knowledge about education services or at risk for dropping out of
self-advocacy, gender equity, leadership, and school (e.g. chronic absenteeism, credit defi-
workforce issues and provides opportunities to cient, unstable family situations). A pretest-post-
investigate a variety of career options. A pre- test control group design with 110 young women
test-posttest study showed that immediately found that that participation in this curriculum
after participating in Career Connections, resulted in significant gains in autonomy and
female students with disabilities who partici- disability and gender-related knowledge (Doren
pated (N = 102) had a range of improved out- et al., 2013). The Paths 2 the Future (P2F) cur-
comes, including (a) more positive relationships riculum used in the current study is an extension
with adults, (b) increased ability to advocate for of the PATHS curriculum, with updated content
themselves in difficult situations, (c) increased and language to reflect current career- and col-
participation in school and community activi- lege-readiness standards.
ties, and (d) more proactive communication Overall, the reviewed literature highlights
skills (Lindstrom et al., 2008). the potential benefits of gender-specific career
and transition interventions; however, to date,
Despite the complex combination of there have been no randomized controlled
studies targeting improved career outcomes
barriers faced by young women
for young women with disabilities receiving
with disabilities as they transition
special education services. The current study
from school to adult roles,
builds on previous research to test the efficacy
relatively few career development of a fully developed gender-specific career
interventions have been developed development intervention (i.e., P2F) using a
to specifically focus on randomized controlled trial.
ameliorating these gender gaps.
Social Cognitive Career
Another such intervention, Girls at Work,
Theory (SCCT)
developed by Wehmeyer and colleagues (2009),
utilized a self-determined career development The P2F intervention utilizes SCCT as a frame-
model to improve career outcomes for young work for understanding the career development
women with intellectual and developmental processes for young women with disabilities.
disabilities. Girls at Work is a web-based, stu- SCCT is an integrated career development
dent-driven, transition-oriented curriculum that model that describes how people, their behavior,
includes a set of eight action steps designed to and environments influence each other to shape
promote self-determination and encourage gen- occupational and academic interests, choices,
56 Exceptional Children 87(1)
and attainment of career goals (Lent, 2005). cial impact on career development outcomes for
SCCT takes into account the influence of indi- young women with disabilities relative to a
vidual and contextual variables, such as race- counterfactual (e.g., typical high school career
ethnicity, gender, disability, and socioeconomic and transition services ). P2F is a fully devel-
status, as well as self-efficacy and expectations oped and pilot-tested curriculum designed to
for the future while at the same time accounting address the unique career development and tran-
for the importance of significant learning experi- sition needs of young women with disabilities
ences in shaping career behaviors and educa- (Lindstrom et al., 2019). Developed and field-
tional outcomes. A key tenet of SCCT is that tested initially as PATHS (Lindstrom et al.,
self-efficacy (beliefs about a person’s capabilities 2013), the P2F curriculum includes 75 lessons
to achieve their goals) and outcome expecta- divided into four modules: (a) self-awareness, (b)
tions (imagined consequences of actions) have disability knowledge, (c) gender identity, and (d)
important effects on the formation of career career and college readiness. The modules are
interests, choices, and aspirations. Aspirations designed to build individual skills and attributes
for the future, in turn, are a significant predictor and expose young women to a wide range of
of employment in adulthood (Ashby & Schoon,
career options. Lessons are taught in sequence
2010). Self-efficacy and outcome expectations
within a group or classroom setting in the course
predict career interests and work outcomes for
of the school day. (For more details on the P2F
students with academic learning challenges and
curriculum, see Lindstrom et al., 2019.)
intellectual disabilities (Nota et al., 2010; Ochs
Guided by the SCCT framework and prior
& Roessler, 2001), and previous interventions
findings, we hypothesized that young women
designed to enhance self-efficacy beliefs have
in the P2F intervention schools would exhibit
been demonstrated to be successful for adoles-
greater rates of increased career development
cents with disabilities (Lindstrom et al., 2019;
skills relative to those in the control schools.
Sheftel et al., 2014 ). Thus, we posit that self-
Hypotheses were tested using an intent-to-
efficacy and outcome expectations are crucial
treat (ITT) approach. That is, effectiveness of
concepts for understanding and promoting
a treatment relies on comparison of assigned
career development outcomes among young
groups regardless of dropout, changes in
women with disabilities.
school protocol, variation in implementation
or noncompliance of teachers, or any other
A key tenet of SCCT is that self-
unobserved factors that occur after random
efficacy (beliefs about a person’s
assignment (DeGarmo & Gewirtz, 2019).
capabilities to achieve their goals)
Thus, the randomization and ITT evaluation
and outcome expectations provide causal inference and unbiased esti-
(imagined consequences of actions) mates of intervention effects due to group
have important effects on the assignment within real-world conditions.
formation of career interests,
Method
choices, and aspirations.
Study Procedures
Thus, we posit that self-efficacy and
outcome expectations are crucial After receiving institutional review board
approval, students were recruited from 26 pub-
concepts for understanding and
lic high schools in the northwestern region of
promoting career development
the United States to participate in an efficacy
outcomes among young women with
trial of an intervention designed to improve
disabilities. career outcomes for young women with dis-
abilities. Special education teachers and school
counselors at each school were asked to iden-
Study Purpose
tify a sample of young women in their schools
The purpose of this study was to determine to participate in the study based on the follow-
whether the P2F curriculum produces a benefi- ing inclusion criteria: (a) identified as female;
Lindstrom et al. 57
(b) currently enrolled in Grades 9 through 12 in were trained in the data collection instruments
a participating high school; (c) determined eli- and procedures to be used during the study.
gible for special education services with a Students in both control and intervention
high-incidence disability, including learning groups completed a web-based survey at four
disability, other health impairment (including time points: (a) preintervention (T1), (b) mid-
attention deficit hyperactivity disorder), speech way through the academic year (for schools
or language disability, and emotional disabil- implementing a full-year schedule; T2), (c)
ity; and (d) demonstrated fourth- to fifth-grade postintervention (T3), and (d) 6-month follow-
reading, writing, and language skills. Students up (T4). The P2F survey was administered
with more significant cognitive or intellectual individually or in groups. Each student was
disabilities were not included. provided with a unique login code. Students
Schools were randomly assigned to inter- took between 15 and 60 min to complete the
vention and control conditions using propen- survey. Research team members were avail-
sity score matching (PSM) procedures. PSM is able for the entire duration to answer any ques-
recommended for group randomized trials of tions the students had about survey items or
20 or fewer or correlations of .30 and above on complex vocabulary. Teachers separately pro-
matching factors (Diehr et al., 1995; Luellen vided information about each participating
et al., 2005; Murray, 1998). Over half of the student’s diagnosis qualifying them to receive
matching factor correlations were above .30. special education services as well as informa-
Schools were matched on data extracted from tion on academic, family, health, work, and
the Department of Education on school size other barriers that often limit educational
(number enrolled), percentage special educa- attainment and may influence postschool out-
tion, percentage Black, percentage White, per- comes.
centage Hispanic, and percentage receiving
free and reduced lunch. The propensity score P2F intervention. Prior to the start of the inter-
(PS) was defined as the conditional probability vention and data collection activities in each
of a subject being assigned to the treatment school, all P2F instructors received one full
group given the observed covariates and where day of professional development from the pro-
the log odds of assignment to treatment was gram developers and research team. This
included an explanation of the P2F logic
 PS  model, an overview of the content of the four
ln1−PS=β0+β1SchoolSize program modules, and modeling of procedures
, for implementing the standard daily lesson
+β %SPED+β %EA
2 3 plans. Once appropriate consent and assent
+β %AA+β %HA+β %FRL
4 5 66 processes were complete, intervention teach-
ers began implementing the lessons and activi-
where covariates were the size of the school and
ties as outlined in the P2F curriculum guide
the percentage of students enrolled who were
(Lindstrom & Post, 2015). Lessons included
receiving special education, were white, were
small- and large-group discussions, role-plays
Black, were Hispanic, and were eligible for free
to practice key skills, guest speakers, and field
or reduced lunch. The matching tolerance was
trips to explore a wide range of college and
set to .2, and paired scores were randomly allo- career options. Young women attended the
cated to either treatment or control. Participants class daily (or on alternate days for those on
in the control condition received “business-as- block schedules) and received high school
usual” career and transition services within their credit upon completing the 75 lessons.
high school. A total of 156 participants from 13 The research team used multiple measures to
schools were randomized to the intervention observe and document fidelity of implementa-
condition, and 230 participants from 13 schools tion for the P2F study. First, we collected infor-
were randomized to the control condition. mation on student dosage through tracking
Parent consent and student assent were student attendance and lesson completion at
obtained prior to the study onset. Teachers each school. Curriculum dosage was calculated
58 Exceptional Children 87(1)
at a school level as the number of lessons deliv- Table 1. Paths 2 the Future Student
ered divided by the total number of curricular Demographics.
lessons (75). Second, we conducted a minimum
Demographic n %
of eight classroom observations in each of the
intervention sites and completed a five-item Grade
teacher adherence observation protocol and six- 9 53 14.5
item quality-of-instruction observation proto- 10 124 33.9
col. Both the adherence and quality observation 11 110 30.1
12 79 21.6
rubrics were completed by P2F research per-
Missing 20 5.2
sonnel during the same classroom observation.
Hispanic or Latina
Across the 13 intervention schools, lesson dos-
Yes 71 19.4
age ranged from 49% to 100% (mean dosage =
Missing 20 5.2
87.8%), with average observer-rated quality at
Race-ethnicity
79.7% and overall adherence to lesson compo-
White 225 61.5
nents measured at 81.7%.
Black or African 17 4.46
American
Sample. The current study sample comprised American Indian or 16 4.4
366 participants (mean baseline age = 16.54 ± Alaskan Native
1.12 years) who had baseline data available Asian or Pacific Islander 8 2.2
on all outcome variables. The study sample Other 44 12.0
(N = 366) included 61.5% non-Hispanic More than one race 49 13.4
White and 19.4% Latina youth, enrolled in Missing 7 1.9
Grades 9 (14.5%), 10 (33.9%), 11 (30.1%), IDEA classification
and 12 (21.6%). Most participants were receiv- Specific learning disability 202 55.2
ing special education services under the catego- Other health impairment 54 14.8
ries specific learning disability (55%) or other Emotional disturbance 22 6.0
health impairment (14.8%). (See additional dis- Intellectual disability 22 6.0
No disability specified 21 5.7
ability demographics in Table 1.) Teachers were
Speech or language 14 3.8
asked to indicate whether each student was
impairment
experiencing any additional risk factors or bar-
Other 25 6.8
riers in five areas, including academics, family
Missing 6 1.6
or living, work, at-risk behaviors, and health
challenges. Most common teacher-reported Note. N = 366. IDEA = Individuals With Disabilities
barriers included difficult family circumstances Education Act.
(43.7%), mental health issues (42.6%), chronic
absences (27.6%), no prior work or volunteer questions (14 items), the survey contained
experience (25.1%), or being behind in com- questions on future goals (four items), career
pleting credits toward graduation (16.4%). and technical education course enrollment (one
According to teacher reports, 83 participants to three items), work experience (one to 20
(22.7%) did not experience any barriers, but the items), and 87 scale items. Eight previously
majority of participants (53.6%) experienced validated scale measures were used to compute
more than one barrier. the focal career development score.
Two scales were from the Arc’s Self-Deter-
Measures. We developed a web-based survey mination instrument (Wehmeyer & Kelchner,
that included a compilation of validated mea- 1995). The Autonomy scale included 14 items
sures. The survey was originally developed and rated from 1 (not even if I had a chance) to 4
validated through a pilot study (Doren et al., (every time I have the chance). Sample items
2013) and was augmented for the current assess- were “I work on schoolwork that will improve
ment. The survey included between 107 and 128 career chances,” “I keep my appointments and
questions, depending on responses to items meetings,” and “I volunteer for things I am
with branching. In addition to demographic interested in.” Cronbach’s alpha for internal
Lindstrom et al. 59
consistency was .74, .85, .82, and .83 for T1, way I am,” “Students at my school are there
T2, T3, and T4, respectively. The Self-Real- for me when I need them,” and “Students
ization scale included 15 items rated from 1 here respect what I have to say” (α = .90, .91,
(never agree) to 4 (always agree). Sample .93, and .92 for T1 through T4, respectively).
items were “It is better to be yourself than Finally, Disability and Gender Awareness
popular,” “I feel free to be angry at people I (Doren et al., 2013) was a six-item scale rated
care for,” and “I don’t accept my own limita- from 1 (no confidence at all) to 5 (complete
tions” (reversed) (α = .78, .78, .80, and .79 confidence). Sample items were “Identify dif-
for T1 through T4, respectively). ferent types of disabilities,” “Describe how
The Self-Advocacy scale was from the being female can impact career choices,” and
College Students With Disabilities Campus “Identify key characteristics of women lead-
Climate Survey (Lombardi et al., 2011) and ers” (α = .79, .87, .89, and .91 for T1 through
included five items rated from 1 (strongly dis- T4, respectively).
agree) to 5 (strongly agree). Sample items After factor analyses (see next section), an
were “I know about my rights and responsi- overall career development construct was
bilities as a student with a disability” and “I computed as a mean total score of the eight
feel comfortable advocating for myself and subscales drawn from above. Each compo-
my needs at this school” (α = .79, .82, .84, nent score was rescaled 1 to 5 and reflected to
and .80 for T1 through T4, respectively). assess career skills and then averaged to com-
Two scales employed in the pilot study pute the overall career development score
were adapted from McWhirter et al. (2000), (Cronbach’s alpha was .82, .82, .86, and .87
the Vocational Skills Self-Efficacy scale and for T1 through T4, respectively).
the Vocational Outcome Expectancy scale.
Vocational Skills Self-Efficacy included 29
Analytic Strategy
items measuring confidence in job prepara-
tion skills, time management, and goal set- Analyses were conducted in two stages. The
ting rated from 1 (no confidence at all) to 5 first stage focused on the measurement of the
(complete confidence). Sample items were criterion outcome construct of career develop-
“Know what to wear for a job interview,” ment. In this preliminary stage, we conducted
“State my general career interests,” and “State factor analysis, more specifically, principal
my educational goals” (α = .97, .97, .97, and components analysis (PCA) to evaluate the
.98 for T1 through T4, respectively). The configural structure of the career development
Vocational Outcome Expectancy scale construct over time as measured by the con-
included six items rated from 1 (strongly dis- stituent scale scores. The second phase of the
agree) to 4 (strongly agree). Sample items analyses focused on evaluation of the P2F
were “My career planning will lead to a sat- efficacy hypothesis, which involved multi-
isfying career for me” and “My talents and level latent growth curve analyses to address
skills will be used in my career/occupation” the repeated-measures structure of the data
(α = .88., .92, .91, and .89 for T1 through T4, and to address the nonindependence of data
respectively). observations due to the clustering of the data
Two scales were from the Student Engage- within students and schools.
ment Instrument (Appleton et al., 2006).
Future Aspirations and Goals included 5 Career development construct factor analy-
items rated from 1 (strongly disagree) to 4 sis. We used several related approaches to
(strongly agree). Sample items were “Going build the career development construct score.
to school after high school is important” and First, we conducted exploratory factor analy-
“I plan to continue my education following sis (EFA) of the scale score indicators. Spe-
high school” (α = .88, .91, .91, and .91 for cifically, we used PCA across T1 through T4
T1 through T4, respectively). Peer Support using the following recommended criteria for
for Learning included six items. Sample determining the factor structure of the focal
items were “Other students here like me the construct outcome score (Henson & Roberts,
60 Exceptional Children 87(1)
2006; Thompson & Daniel, 1996): (a) PCA LGM provide advantages for testing
was conducted with oblimin rotation allowing developmental changes in student outcomes
exploratory factors to be potentially corre- within randomized trials (Brown et al., 2008)
lated, (b) eigenvalues of unrotated factors because they more reliably estimate within-
were greater than 1, and (c) each factor and between-student differences in change
accounted for greater than 5% of the total compared to repeated-measures ANOVA. In
variance. LGM, a two-level growth model is a special
Second, to best determine the optimal case of an SEM measurement model, with
number of factors we conducted both parallel repeated-measures observations specified as
analysis (PA) and Velicer’s minimum average the measurement indicators of a latent vari-
partial correlation analysis (MAP) (O’Connor, able intercept and latent growth rate or slope
2000). The PA directly compares eigenvalues factor. The first step of LGM is to estimate a
for the ranked ordered PCA loadings with series of sequential unconditional models to
eigenvalues generated from randomly ordered determine the optimal pattern of growth.
data (O’Connor, 2000). Criteria for meaning- Using fixed chronometric time weights, we
ful factors are those with higher eigenvalues compared three models: a random intercept-
from those randomly generated (i.e., eigne- only model (i.e., no mean growth but indi-
values greater than the random parallel 95% vidual differences or variance in intercepts), a
quantile are retained), and thus meaningful linear growth model (i.e., mean increases or
variance is obtained above chance. The ratio- decreases and variance in individual trajecto-
nale underlying MAP is that the average of the ries), and finally, accelerated or quadratic
squared partial correlations should be mini- growth to capture curvature in the mean tra-
mum when an adequate number of factors are jectories. The intercept-only model is speci-
partialed or extracted. fied by fixing repeated waves at 1, 1, 1, and 1
Finally, for tests of measurement invari- for T1 through T4. The linear growth slope
ance, we conducted a replicability test recom- model adds a latent factor with loadings fixed
mended by Osborne and Fitzpatrick (2012) in at 0, 1, 2, and 3, representing variation in lin-
which the standardized structure item load- ear trajectories. Finally, a quadratic model
ings from wave to wave are compared using a adds a latent factor allowing for acceleration
longitudinal panel design. Squared differ- in growth in trajectories, or nonlinear curva-
ences less than .04 represent replication; dif- ture, by fixing loadings at 0, 1, 4, and 9.
ferences greater in magnitude do not replicate. As in SEM, the MLSEM model fit was
evaluated using recommended fit indices
Multilevel P2F efficacy analyses. This study (Byrne, 2011; McDonald & Ho, 2002) of a
used a clustered randomized control trial to chi-square minimization p value greater than
test the efficacy of the intervention (Murray, .05, a comparative fit index (CFI) greater than
1998; Raudenbush, 1977). The clustering, or .95, a chi-square ratio (χ2/df) less than 2.0,
nesting of data, involved longitudinal data and a root mean square error of approximation
nested within students and students nested (RMSEA) less than .08. Data were also mod-
within schools. The P2F efficacy hypotheses eled using full-information maximum likeli-
were tested with structural equation modeling hood (FIML), which uses all available
(SEM) specified as latent growth models information from the observed data in han-
(LGM) using Mplus 8.2 (Muthén & Muthén, dling missing data. FIML estimates were
1998–2018). We employed multilevel SEM computed by maximizing the likelihood of a
(MLSEM) to estimate the two-level growth missing value based on observed values in the
model at the within and between school levels data. Any individual who had baseline data
(Heck & Thomas, 2015). The MLSEM growth only and no follow-up data contributed noth-
model statistically addresses the nonindepen- ing to the likelihood of estimates and was
dence of the clustered data and partitions vari- effectively excluded from change analyses
ance in participants’ growth trajectories across (Brown et al., 2008). Compared to mean-
student- and school-level effects. imputation, listwise, or pairwise models,
Lindstrom et al. 61
FIML provides more statistically reliable family socioeconomic status, respectively, as
standard errors. predictors of growth in career development.
The residual errors for initial status, linear and
Within-school two-level growth Level 1. The quadratic growth (ζ , ζ , and ζ ) are each
0ij 1ij 2ij
within-school growth model, or student-level assumed to be normally distributed.
growth model, for career development was
specified as the following: Between-school model Level 3. The between-
school growth model estimates the proportion
Level1DC =η +η Time
ijt 0ij 1ij tij of variation in latent intercepts and latent
+η Time2 +ε , growth factors that is attributable to school-
2ij tij tij
level variables. The latent growth model is
where the Level 1 career development out- identical to the specification of time weights
come variable is a function of the repeated- at the within-school student level, and the
measures random intercepts (η ) for student growth indicators are latent estimates based
0
i in school j and is a function of time as linear on the summary of the estimated observed
growth (η ) and time squared as accelerated data for students in the within model. All
1
or quadratic growth (η ) for student i in school school-level covariates predicting growth are
2
j, plus the time-specific errors of measure- entered in the between-school model. More
ment nested within individuals and schools importantly, because randomization occurred
(ε ). Thus Level 1 includes the time-varying at the school level, the ITT P2F effect is esti-
tij
repeated measures of the outcome score with mated in the between model as
linear time weighted 0, 1, 2, and 3 for the
Level3CDInterceptη =α +ζ
respective outcome scores and time squared 0j 0j 0j
weighted 0, 1, 4, 9. Level3CDLinearGrowthη =α +β P2F
1j 1j 1jj j
+β SchoolSize +β %FRL +ζ
Within-school growth Level 2. The estimated 2j ij 3j ij 1j
time-varying random intercepts and growth Level3CDAcceleratedGrowwthη =α
2j 2j
trajectories for each student are then summa- +β P2F +β SchoolSize
4j j 5j ij
rized for students representing the within-
+β %FRL +ζ
school but “between-student” effects specified 6j ij 2j
as the following:
where β is the ITT effect of the P2F inter-
1j
vention on growth in girls’ career develop-
Level2CDRandom
ment scores, controlling for the effects of
Interceptη =α +ζ
0ij 0j 0ij school-level variables, such as school size
Level2CDLinear (β ) and percentage of free and reduced
2j
Growthη1iij =α1j +γ1ijGradeij lunches served (β3j). The P2F efficacy
hypothesis in the form of an MLSEM growth
+γ Ethnicity +γ SES +ζ
2ij ij 3ij ij 1ij model is summarized graphically as a two-
Level2CDAAcceleratedGrowthη = level within-school portion and a two-level
2ij
between-school portion of variance, shown
α +γ Grade +γ Ethnicity
2j 4ij ij 5ij ij in Figure 1. For clarity’s sake, model specifi-
+γ SES +ζ
6ijj ij 2ij cation is illustrated with linear growth
parameters only. The observed data or mani-
where η is the summary of individual differ- fest variables are represented as squares. All
0ij
ences in student intercepts at Wave 1 for the unobserved latent variables are represented
full sample, η is the predicted growth rate by circles, such as Level 2 and Level 3
1ij
for the sample, α is the mean adjusted growth growth factors and the between-school latent
1j
rate, and γ through γ are the student-level estimates of career development repeated-
1ij 3ij
covariate effects of girls’ grade, ethnicity, and measurement indicators.
62 Exceptional Children 87(1)
Figure 1. Multilevel structural equation modeling analytic specification for test of the Paths 2 the
Future efficacy hypothesis predicting growth in young women’s career development outcomes.
Results of factor dimensions. Results of the PCA, PA,
and MAP each indicated that a single-factor
Career Development Construct
solution was the optimal configural structure
The means and standard deviations for the for the career development construct. Load-
career development indicator scale scores are ings for the obliquely rotated PCA solution
presented in Table 2 by group condition are shown in Table 3 for T1 through T4.
across time. In the first step of the analyses, For T1, T3, and T4, the PCA obtained a
we subjected each of the candidate scale single-factor solution and no second factor
score outcomes to a PCA, followed by the PA was extracted meeting criteria. At T2, the
and the MAP focusing on the optimal number assessment wave with partial data, the PCA
e 4 102) 0.49)0.4800.74)0.79)0.56)0.91)0.50)
Timn( = 2.78 (2.76 (3.55 (3.70 (3.23 (3.49 (3.06 (
n
o
venti e 3 136) 0.47)0.44)0.79)0.80)0.53)0.89)0.52)
e inter Timn( = 2.78 (2.79 (3.52 (3.69 (3.29 (3.55 (3.10 (
r
u
t
u
e F 2 6) 54)39)72)84)56)86)46)
s 2 th Time n 4= 86 (0.73 (0.34 (0.61 (0.17 (0.28 (0.09 (0.
th ( 2.2.3.3.3.3.3.
a
P
e 1 153) 0.46)0.48)0.76)0.85)0.59)0.85)0.50)
Timn( = 2.66 (2.68 (3.38 (3.37 (3.12 (2.95 (3.04 (
n.
nditio e 4 165) 0.50)0.43)0.67)0.83)0.54)0.87)0.50)
up Co Timn( = 2.72 (2.74 (3.50 (3.39 (3.10 (2.99 (3.07 (
o
r
G
ment Indicators by trol condition 2 Time 3 n1)( 192)= 51)2.73 (0.47)45)2.72 (0.46)77)3.51 (0.72)80)3.36 (0.80)60)3.14 (0.51)74)2.94 (0.85)56)3.09 (0.54)
velop Con Time n 4= 66 (0.72 (0.51 (0.49 (0.10 (0.83 (0.15 (0.
De ( 2.2.3.3.3.2.3.
r
e
e
ons for Car Time 1 n( 213)= 2.66 (0.41)2.73 (0.42)3.53 (0.72)3.39 (0.80)3.17 (0.48)2.88 (0.74)3.14 (0.47)
ti
a
vi
e
Means and Standard DTable 2. Scale AutonomySelf-RealizationSelf-AdvocacyVocational Skills Self-EfficacyVocational Outcome ExpectancyDisability and Gender AwarenessStudent Engagement
63