1 Formation of social capital for community college students: A Second-order Confirmatory Factor Analysis
Authors Information Yu (April) Chen, Ph.D. Assistant Professor School of Education Louisiana State University 111N Peabody Hall Baton Rouge, LA 70803 Email:
[email protected]
Soko S. Starobin, Ph.D. Email:
[email protected]
2 Abstract Objective. This quantitative study constructed a statistical model to measure two types of social capital (i.e., family social capital and college social capital) among community college students. The authors also examined influences of these two types of social capital constructs on degree aspiration. Methods. This study utilized STEM Student Success Literacy Survey (SSSL) to collect data in all fifteen community college districts in Iowa. With more than 5,000 responses, the authors conducted descriptive analysis, exploratory and confirmatory factor analysis, and structural equation modeling (SEM) analysis. Results. A second-order confirmatory factor analysis confirmed the two-tier structured measurement model for social capital. In particular, college social capital was measured by three latent variables such as interaction with advisors/counselors, interaction with faculty members, and transfer capital. The three latent variables were further measured by 14 survey items. Family social capital was measured by six survey items that described parent-child interaction in high school. The SEM results indicated that college social capital had stronger direct influences on degree aspiration compared to family social capital. The impact of family social capital was delivered through the mediation of college social capital. Contributions. Findings contributed to the literature by emphasizing the important role of institutional agents in promoting educational aspiration among underrepresented and disadvantaged students. Future studies are encouraged to test the two-tier model with a racially diverse student sample. Also, additional social capital measures such as institutional characteristics and peer interaction should be considered for future investigation. Keywords: Social Capital, Community College Students, Confirmatory Factor Analysis
3 Introduction Community colleges in U.S. are enrolling a large number of first generation, low socioeconomic status, and underrepresented minority students (National Center for Science and Engineering Statistics (NCSES), 2013; Hagedorn, 2004; 2008). According to recent national statistics, 36% of the community college students are first generation students; 16% are from single-parent family (American Association of Community College (AACC), 2016). Nearly half of the community college credit students are non-White (AACC, 2016). With an open-door policy, it is the mission of community colleges to serve a diverse student body and help all students achieve their dreams. One of the key factors that may influence success of a diverse student body is social capital. Social capital can provide access to resources and networks that lead to academic success. A plethora of studies in higher education have examined the critical role of social capital. Researchers often demonstrated a significant influence of social capital in fostering educational aspiration as well as desirable educational outcomes (Coleman, 1988; McDonough, 1997; Perna & Titus, 2005). In previous studies focused on educational aspiration, researchers illustrated how social capital played a critical role in students’ college-going aspiration or actual enrollment after high school graduation (Coleman, 1988; Perna & Titus, 2005; Byun, Meece, Irvin & Hutchins, 2012). In community college studies, questions remains regarding students’ subsequent aspiration. In other words, what degree do community college students aspire to obtain in their lifetime after community college enrollment? Do they only aspire to get a certificate or an associate degree? Or, they want to achieve a bachelor degree and above? The aspiration after community college enrollment can greatly influence students’ academic outcome in a long term.
4 Further, although considered different sources of social capital (i.e., from family, school/college, and/or community), previous quantitative studies did not establish a measurement model that considers demographic, academic, and social characteristics of community college students. For community college studies, it is critical to realize that the access to social capital is not equal among different population due to differentiated structural positions and social networks (Lin, 2000). For example, first generation students, low social-economic status (SES) students, and underrepresented minority students might be lack of access to academic networks and resources at home or within the neighborhood. They might primarily rely on resources in community colleges (Moschetti & Hudley, 2015; Museus & Neville, 2012). Thus, it is imperative to develop and test a social capital model that reflect unique characteristics and needs as such for community college students. This study aims at establishing and testing a statistical model that quantitatively measure different types of social capital and examine potential influences of social capital on degree aspiration of community college students. Literature Review Social Capital Theory This study was inspired by social capital theories from two sociologists, Pierre Bourdieu and James S. Coleman. Bourdieu (1986) defined social capital as "the aggregate of the actual or potential resources which are linked to possession of a durable network of more or less institutionalized relationships of mutual acquaintance and recognition." (p. 248) In his view, social capital served as part of an invisible mechanism that allow children from dominant class maintain their inherited advantages in school system. It eventually resulted in social reproduction and inequity (Bourdieu, 1973; 1986). Bourdieu’s theory inspired a series of educational studies that emphasized the role of institutional agents and structural contexts in education system
5 (Stanton-Salazar, 1997; McDough, 1997; Museus & Neville, 2012). These studies highlighted the importance of interaction with institutional agents in promoting social capital acquisition in schools and colleges, which is critical to community college students with disadvantaged background. Coleman (1988), in contrast, emphasized social capital on its function of collectively advancing children’s life chances. The intergenerational closure was one of the most important concept in his theory. It described how networks among parents might positively influence children’s academic achievement. Inspired by Coleman’s theory, parents’ expectation, parentchild interaction, and family structure were often used to measure children’s social capital (Coleman, 1988; 1992; Perna, 2000; Perna & Titus, 2005; Byun, et.al., 2012). In this study, we argue that family influences continue to contribute to community college students’ social capital accumulation. Specifically, students enters community college with diverse family social capital acquisitions. This serves as their starting point of further social capital accumulation in college. Both Bourdieu and Coleman conceptualized social capital in a way that linked family and educational institutions (i.e., K-12, college). For example, in Bourdieu’s theory, family influences can help dominant class students better understand “norms” and “access” in schools and eventually obtain desirable educational credentials to maintain their status. Thus, the reproduction process needs inputs from both family and school. On the other hand, Coleman advocates parental involvement in school activities. He argued that the school and community/family should work together to collectively advance children’s life chances. In this study, we will integrate two types of social capital (i.e., family and college) and examine their different influences. Measuring Social Capital in Higher Education Studies
6 In higher education studies, social capital was quantitatively measured to highlight different sources (family social capital or school/college social capital) and different features (structural and process components) (Byun, et. al., 2012; Israel, Beaulieu, & Hartless, 2001; Perna, 2000; Perna & Titus, 2005). First, the effects of family social capital were elaborated to a structural and a process component in empirical studies. Family structural component was often measured by two-parent/one-parent family, numbers of siblings, and educational outcomes of siblings (e.g., dropouts among siblings) (Coleman, 1998; Israel, et. al., 2001). Common measures for family process component included parental involvement in school-related activities, parental expectation of child attending college, and parental-child interaction at home (Coleman, 1988; Perna, 2000; Perna & Titus, 2005; Byun, et. al, 2012). Both structural and process components of family social capital were found to have significant influences on youth’s educational outcomes such as academic achievement (Israel, et.al., 2001; 2004), high school dropout (McNeal, 1999), college-going aspiration, and college enrollment (Kim & Schneider, 2005; Perna, 2000; Perna & Titus, 2005). Second, school social capital was also quantitatively measured through structural and process components in previous literature. Particularly, researchers utilized measures such as school size, school location (urban/suburban/rural), minority enrollment, and free-lunch percentage to represent the structural component of school social capital (Israel, et. al., 2004; Parcel & Dufur, 2001; Perna, 2000). The process component of school social capital was often measured through teacher’s expectation of students’ academic outcomes, student-teacher interaction, student-advisor interaction, etc. (Israel, et. al., 2004; Smith, et. al., 1995). It should be noted that many qualitative studies also focused on how social capital, either family or school social capital, may influence students’ academic outcomes and college
7 enrollment. In these studies, researchers highlighted that accessing social capital may affect future success for students who are from underrepresented minority racial groups, low SES background, and first-generation status. For example, Kim (2014) described the cultural and structural nuances of how parental involvement influence Chinese and Korean immigrant students’ college-going experiences. Teranishi and Briscoe (2006) utilized social network analysis to demonstrate how social capital from family, school, and community may play a critical role in inequality of accessing higher education for students of color. In addition, Muses and Neville (2012) demonstrated that institutional agents at college could increase the access to social capital for racial minority students by serving as a bridge to the support network. Previous literature inferred that both family and college social capital are important to community college students. Quantitative measures and statistical effects of family and college social capital for community college students, however, are less discussed. The following section summarized studies that may provide directions of measuring and examining social capital in community college context. Forms of social capital for community college students In community college context, researchers often found that college social capital has a more critical role compared to family social capital. For example, Deil-Amen’s (2011) emphasized the “informational benefits” (p. 73) of college social capital. In particular, community college students would obtain beneficial knowledge and skills for success in college through their interactions and relationships with faculty members, counselors, advisors, and other students (Deil-Amen, 2011). Similarly, Wang (2013) indicated that academic integration with institutional agents positively influenced community college students’ baccalaureate expectation. College social capital is especially critical for those who are first generation college students and
8 from underrepresented racial groups. Because they are less likely to obtain related knowledge, skills and information from home (or, through family social capital) and more likely to drop out from college (Deil-Amen, 2011; Sandoval_Lucero, Maes & Klingsmith, 2014). Other empirical studies also had congruent findings. For instance, Wells (2008) found that parental involvement had positive influences on college students’ persistence overall but matter less for community college students. Moschetti and Hudley (2015) argued that family social capital can only serve as an emotional support and/or strong believes towards personal attribute. It may not provide guidance to navigate the college administrative organization and seek institutional support. Further, Tovar (2015) found a small, but significant effect of interactions with institutional agents (instructors and counselors) on Latino/a students’ success in community colleges. Dowd and Colleagues (2013) also highlighted the positive role of administrators in facilitating community college students’ collegiate identity development and successful transfer (Dowd, Pak, & Bensimon, 2013). In addition to the above, transfer capital was identified as a unique type of college social capital that can be obtained from institutional agents at community colleges (Laanan, 2007; Laanan, Starobin, & Eggleston, 2010). Transfer capital refers to the skills and knowledge that community college students acquired for navigating the transfer process from two-year to fouryear institutions (Laanan, 2007). It often can be measured through students’ interaction with faculty members, counselors, advisors regarding transfer process as well as skills necessary for learning at four-year institutions (Laanan, 2004). The scope of transfer capital can be expanded to faculty/staff validation, financial factors, and other related factors (Moser, 2012). Emipiral studies found that transfer capital significantly influenced students’ academic adjustment after the transfer (Laanan, Starobin, & Eggleston, 2010; Starobin, Smith, Laanan, 2016), their degree
9 aspirations (Kruse, Starobin, Chen, Baul, & Laanan, 2015; Myers, Starobin, Chen, Baul, & Laanan, 2015; Chen & Starobin, 2017), and vocational choice (Johnson, Starobin, & Laanan, 2016). Educational aspiration as the outcome variable As mentioned previously, education expectations and aspirations have been examined as a beneficial outcome of social capital acquisition (i.e., Wang, 2013; Kruse, et.al. 2015; Myers, et.al. 2015; Chen & Starobin, 2017). The influences of high educational aspiration on educational attainment has been discussed by various empirical studies. Some studies indicated a robust and positive influence of educational aspiration on educational attainment (Reynolds, Steward, MacDonald, & Sischo, 2006; Adelman, 2006; Domina, Conley, & Farkas, 2011). Yet others warned the negative impact of unrealistically high aspiration (Jencks, Crouse, & Mueser, 1983; Rosenbaum, 2001). In community college context, it is the fact that the actual completion rate remained low despite a high educational aspiration among students (CCSSE, 2016). Nevertheless, we believe that having a high educational aspiration is essential to community college students’ success. A high educational aspiration is the first step towards completion of a postsecondary degree, preferably a bachelor degree through transfer (Wang, 2012). Therefore, in this study, we selected degree aspiration as the outcome variable and hope to discover a positive influences of social capital on degree aspiration. In order to control confounding variables, we also considered other possible predictors of educational aspiration. For instance, family background or social economic status (SES) has been studied as important predictors of educational aspiration. Specifically, the status-attainment theory (Blau & Ducan, 1967) has firstly established the base of discussing how SES may impact youth’s educational and occupational aspirations. In developing Blau and Ducan’s parsimonious
10 model, Sewell and colleagues discussed how social psychological factors such as significant others, self-concept, references groups, and school experiences may mediate the effects of family background and impact youth’s educational aspiration and attainment (Sewell, Haller, & Portes, 1969; Sewell, Haller, & Ohlendorf, 1970). Building upon these classical studies, subsequent empirical studies included family background or SES as significant predictors of educational aspirations. Many of these studies adopted measures such as parents’ occupation, parents’ educational level, and/or family income (Jencks et al, 1983; Wells, 2008; Wang, 2012; Kruse, et.al., 2015). Further, due to the limitation of status-attainment theory as well as the diversity of community college student body (Carter, 2002), we intend to include demographic characteristics as additional predictors of educational aspiration. First, female students might have different aspiration patterns compared to male students. Some researchers found that females might have lower educational aspiration in general (Carter, 2002; Astin, 1977). Others argued that female students have higher educational aspiration regarding postsecondary degrees in particular (Blackhust & Auger, 2008; Byun, et.al., 2012). Females were also more likely to attend community colleges than males (Laanan, 2003). Second, underrepresented racial minority students also have different aspiration patterns. For example, African American students tend to have higher aspiration than White students (Portes & Wilson, 1976; Carter, 1999); Whereas Latino students tend to have lower educational aspirations (Carter, 1999; Kao & Tienda, 1998; Gonzales, Stein & Hug, 2013). Moreover, non-traditiaonl or older students (Carter 1999; Anderson, 2013) and those who have language barriers (Gonzales, et.al., 2013; Wang, Chang & Lew, 2003) may also have lower aspiration. Purpose and Research Questions
11 Although previous studies have examined different forms of social capital and its role in predicting aspiration, there are still gaps to be filled. First, previous studies, especially if quantitatively conducted, often measured the social capital under a high school context (i.e., Coleman, 1988; Kim & Schnedier, 2005; Perna, 2000). For those studies in where community college context was considered, social capital was measured for comparing with four-year students (i.e., Wells, 2008 ) or for a specific group (i.e., Tovar, 2015). Only a few studies highlighted institutional agents’ role in accumulating social capital based on quantitative data (i.e., Tovar, 2015; Wang, 2012). Further and more importantly, rarely a community college research considers family and college social capital simultaneously and examines their influences quantitatively. In this study, we focused on developing a statistical model for examining how family social capital (such as parental involvement) and college social capital (such as interaction with institutional agents) may work together to influence community college students’ success. We started with establishing a measurement model for both family and college social capital. Then, we examined a structural model that emphasized social capital’s influences on degree aspiration. The outcome variable, degree aspiration, was selected based on previous studies regarding its important role in academic success (i.e., Wang, 2012) and its relationship with social capital (i.e., Wang, 2013; Chen & Starobin, 2017). We focused on only community college students who are in the “academic track” versus those are enrolled in Career Technical Education (CTE) programs. This exclusion allowed us to test the model with a population for whom the degree aspiration matters more. The following two research questions guided this study. 1) What is the underlying structure of social capital measures for community college students?
12 2) How do family social capital and college social capital, as two types of social capital, predict community college students’ degree aspiration? Conceptual Framework This study included the theories of Bourdieu (1973; 1986), Coleman (1988), and findings from previous higher education studies into an integrated conceptual framework. Specifically, this study examined the process components of both family and college social capital. In addition, we also incorporated transfer capital to highlight transfer aspiration among community college students who are within the “academic track”. We included various social capital constructs. For example, variables describing parentalchild interaction in high school were adopted to measure family social capital. Parental-child interaction measures were introduced to empirical studies by Coleman (1988) and adopted by subsequent higher education studies (i.e., Perna & Titus, 2005; Byun, et. al, 2012; Kruse, et.al. 2015). It represented the process component of family social capital; and was often found significant in predicting students’ education aspiration. College social capital, on the other hand, was measured by students’ interaction with institutional agents including faculty, academic counselors, and advisors. Previous quantitative studies highlighted the significant impact of these measures on desirable educational outcomes (Israel, et. al., 2004; Smith, et. al., 1995). Further, variables reflecting transfer capital were added to the model as part of the college social capital measures. Previous studies repeatedly utilized transfer capital measures to explain how transferrelated knowledge and skills are important to educational aspiration and performance (Laanan, 2004; 2007; Moser, 2012; Kruse, et.al. 2015). Transfer capital measures involve interactions with institutional agents at both two-year and four-year institutions. Thus, it fit well with other college social capital.
13 It should be noted that we focused on process component (frequency of interaction or activity) rather than structure component (i.e., family structure, institution characteristics) social capital measures. As measures of structure component of college social capital, institution characteristics often contain information such as public/private, geographic location (urban, suburban, rural), size, etc. These measures are at institution level and would be suitable for a hierarchical model (i.e., Perna & Titus, 2005). In this study, we choose to restrict our analysis within the student level and focus on the parsimonious of our model. Similarly, family structure measures such as single parent, siblings’ information, and parents’ educational expectation were not included in the model. These family structure measures were not included mainly due to data availability. The lack of structure component of social capital measures is a limitation of this study; and can be further examined by future studies. According to findings of previous quantitative (i.e., Wells, 2008; Wang, 2013) and qualitative studies (i.e., Moschetti & Hudley, 2015; Deil-Amen, 2011), we hypothesizes that, within our framework, college social capital may have a stronger and more explicit effect than family social capital on degree aspiration. In particular, we assume that influences of family social capital might be strong and significant for college enrollment and college-going aspiration (Kim & Schneider, 2005; Perna, 2000; Perna & Titus, 2005). However, after enrolled in community college, its influences might become indirect. That is, family social capital may significantly influence community college students’ access and possession of college social capital. Students with lower level of family social capital may also have lower level of college social capital (Moschetti & Hudley, 2015). Thus, influences of family social capital are likely to be carried through the effects of college social capital. With the consideration of impact from
14 control variables (i.e., demographics, SES such as parental education level, etc.), a hypothetical model was established. Figure 1 depicted the model graphically. Insert Figure 1 here Methods Instrument The STEM Student Success Literacy (SSSL) Survey was utilized as the instrument in this study. The SSSL survey is an on-line survey disseminated via Qualtrics. It measures community college students’ self-efficacy, social capital, transfer knowledge and demographic characteristics. The survey items were developed based on an extensive review of pre-existing instruments including general self-efficacy scales (Sherer & Maddux, 1982), Campus Life and Learning Survey (Bryant, Spenner, & Martin, 2006), Cooperative Institutional Research Program (CIRP) Freshman Survey (HERI, 2011), Community College Survey of Student Engagement (Center of Community College Student Engagement, 2005), and Laanan Transfer Student Questionnaire (Laanan, 2007). A pilot study was conducted in spring 2012. A number of 565 community college students in the state of Iowa participated in the pilot study. We conducted reliability test and correlational analysis to examine the reliability and validity of measures. For example, based on the results of EFA, we removed survey items with a factor loading lower than 0.6. We then tested reliability (i.e., Cronbach’s alpha) of all emerging constructs with both the entire data and two randomly selected subsets of the pilot data. After the modification, the length of the survey was significantly shortened. The final survey instrument contained 69 questions and 214 items. The data collection was conducted during the fall 2012 semester among all 15 community college districts in the state of Iowa. All potential participants received an email
15 invitation through on line survey software Qualtrics. The survey was active for three weeks. Email reminders were sent to potential participants if they did not respond the initial email within 7 days and 14 days. To obtain a higher response rate, a random lottery drawing for winning one of the five free iPad was conducted. Population and Sample Although the survey instrument highlighted STEM students in community colleges, this study included students from all academic majors. In particular, only students in academic programs, taking classes on campus, have completed at least one semester, and older than 18 were included as potential participants. We did not restrict our sample to first-time students. However, students in Career Technical Education programs, certificate programs, non-credit programs and those with dual enrollment status were excluded. In total, 43,964 potential participants received the e-mail invitation to participate in the survey; and 5,168 students completed the survey, resulted in a response rate of 11.76%. The demographic composition of the sample was slightly different from the characteristics of the population. In particular, during the fall 2012, 55.4% of the enrollment in Iowa community colleges were females. Only 15.2% of the students were minorities (i.e., nonwhite). The average age of all students was 22.96 years old (Iowa Department of Education, 2012). In the survey data, majority of the responses were female (66%), White (78%), and younger adult (18-24 years old, 46.2%). We applied a weighting process to adjust possible bias regarding ethnicity, gender, and age. The demographic information of Iowa community college student population in fall 2012 was used to generate the weight. Variables used in this study
16 The dependent or endogenous variable in this study is degree aspiration. The original scale of this variable ranged from 1 to 7, where 1= do not intend to earn a degree, and 7=doctoral/terminal degree. We recoded it into a three-category variable in where 0= no degree, 1= associate degree, 2=bachelor degree and beyond. The independent variables involved 29 survey items representing students’ demographic characteristics and social capital measures. Specifically, demographic variables included gender, age, race, native language, and parents’ education. Also, race was recoded as a dicthconomous variable in where 1=non-White and 0 = White. This is due to the nature of Iowa demographics. In the fully imputed and weighted sample, we had about 85% White students and only 15% non-White students. Social capital measures included variables of family social capital and college social capital. Family social capital was measured through six survey items. These items related to parental–child interaction during high school. The family social capital items were measured in a five-point Likert scale that indicated the frequencies of each activity, in where 1= never or very rarely, and 5=several times a week. Noticeably, we included parents’ education level as demographic variables rather than family social capital measures. In computing process, it helped us to perform factor analyses with solely Linkert scaled variables that measures activity frequencies. College social capital contained three groups of survey items that measured students’ interaction with advisor/counselors, faculty members, and their transfer related activities. Interactions with advisors/counselors and transfer related activities were measured by sevenpoint Likert scale, in where 1= strongly disagree, and 7=strongly agree. Interactions with faculty members were measured in a five-point Likert scale, in where 1=never of very rarely, and 5=several times a week.
17 Data Analysis This study adopted both descriptive and inferential statistics. Data analysis was conducted in SPSS 22.0 and Mplus 7. Descriptive analysis revealed the demographic and academic characteristics of participants. Then, exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) were adopted to explore and confirm the formation of social capital constructs. We randomly split the sample into halves. We used one-half for EFA (n=2586) and the other half for CFA (n=2582). To reflect the conceptual framework, we proposed a second-order CFA model and fitted it with the data set. The CFA model testing adopted ML (Maximum Likelihood) estimator. Absolute model fit indices such as chi-square, RMSEA (Root Mean Square Error of Approximation), CFI (Comparative Fit Index), TLI (Tucker-Lewis Index), and SRMR (Standardized Root Mean square Residual) were used to examine the model fit of CFA model. Theoretically, a non-significant chi-square indicates a good model fit. However, the value of chi-square is often sensitive to sample size. Among available alternative model fit indices, RMSEA, CFI, TLI and SRMR are mostly used (Hooper, Coughlan & Mullen, 2008; Hu & Bentler, 1999). Each model fit indices has its own criteria. For example, a value smaller than .05 indicates a good model fit in RMSEA and SRMR; a value higher than .9 in CFI and TLI indicates a good model fit (West, Taylor & Wu, 2012). Then, we utilized structural equation modeling (SEM) techniques to test how different types of social capital may influence degree aspiration among community college students. Due to the categorical nature of the dependent variable (i.e., three categories in degree aspiration), we utilized WLSMV as the estimator in SEM (Muthen. 1984; Edwards,Wirth, Houts & Xi, 2012). The WLSMV estimator is a well-known estimator for endogenous categorical variables, especially among Mplus users. It is a limited information approach that considers the likely non-
18 normal distribution of endogenous variables due to their categorical nature (Edwards,Wirth, Houts & Xi, 2012). The term “WLSMV” refers to a diagonally weighted least square estimation with robust standard errors and a mean and variance adjusted test statistics (Muthen, 1984; Muthen & Muthen, 2017). With the WLSMV estimator, several absolute model fit indices (i.e., chi-square, RMSEA, CFI, and TLI) can be generated. Both CFA and SEM analysis requires fully imputed data. The missing data was imputed via EM methods. The EM approach contains an expectation (E) step and a maximization (M) step. The missing observations were imputed using a regression process and combined with the results of maximum likelihood estimation. Additionally, a weighting process was conducted to address any possible bias regarding demographic characteristics. The demographic characteristics of the entire population (i.e., fall 2012 student demographics in Iowa community colleges) were used to provide a bias weight for applying the case-weighting process. It should be noted that the descriptive analysis and EFA were firstly conducted with the raw data and then confirmed with the imputed data. The CFA and SEM analyses were conducted solely with the imputed data. Limitation This study has several limitations to be mentioned. First, because the data collection was conducted within the state of Iowa; this study may provide limited generalization in states other than Iowa or have limited similarity with Midwest states. Second, this study did not exhaust all types of social capital measures. Such measures may include family structure variables (e.g., single family, numbers of siblings, etc.), institutional characteristics (e.g., proportion of underrepresented minority students, proportion of students receiving financial aid, etc.), and variables measuring social capital obtained through the community. Social capital obtained
19 through peer interaction might also be desirable. This limitation is due to the data availability and the structure of statistical model. For example, a hierarchical linear modeling (HLM) technique may be suitable for examining institutional characteristics; a new or revised community college survey might consider adding peer interaction measures. Moreover, just like many other quantitative studies (i.e., Wang, 2012; Tovar, 2015), this study measured the frequencies of related activities rather than the quality. This might call for new topics of future qualitative studies. Lastly, survey data is self-reported. We rely on students’ accuracy on reporting frequencies of different types of interactions, which can be over-estimated or underestimated. Results Descriptive Analysis A descriptive analysis was conducted to summarize the demographic characteristics of participants. A large proportion of the participants were younger adult who were aged from 1824 years old (46.2%), and female (66.2%). As a reflection of the Iowa population, majority of the students were White (78.6%) and more than 80% speak English as their native language. In terms of parental education level, 40.8% mothers and 48.7% fathers did not attend college. Participants showed a high degree aspiration: more than 50% students wanted to transfer to 4year institutions and more than 80% aspired to obtain a Bachelor degree or beyond eventually. Insert Table 1 here Exploratory Factor Analysis The EFA analysis was conducted with half of the sample (n=2586) to explore the formation of social capital measures. Four constructs were emerged: parent-child interaction (family social capital), interaction with advisors and counselors, transfer capital, and interaction with faculty. The KMO and Bartlett’s test results indicated that sample size was sufficient for
20 conducting EFA (KMO=.880; Barlett’s test Chi-square =27094.862***, df=210). We adopted principle component analysis (PCA) and oblique (promax) rotation approaches. PCA was adopted as the extraction method because we want to reduce the number of variables and retain as much as possible the original variances (Tabachnick & Fidell, 2013). Oblique rotation was selected because we assume there are correlations among constructs. All factor loadings were sufficiently high (ranged from 0.649 to 0.912). The alpha levels demonstrated high internal reliability for each construct (ranged from 0.839 to 0.886). Insert Table 2 here Confirmatory Factor Analysis Next, a CFA was conducted with the second half of the sample (n=2582). The CFA model had a good fit (𝜒 2 (124) = 553.682, RMSEA=0.037, CFI=0.981, TLI=0.976, SRMR=0.026). All loadings were high and significant. The model confirmed the proposed twotier structure. While family social capital was a single-tier structure, college social capital involved a second tier. Specifically, college social capital was measured by three constructs: interaction with advisors and counselors, transfer capital, and interaction with faculty. These three constructs were further measured by twelve survey items. Because of this two-tier structure, this analysis was a second-order CFA. It should be noted that two items were deleted from the final CFA model due to cross loading issues emerged within the two-tier structure. Figure 2 displayed the results of the second-order CFA. Insert Table 3 and Figure 2 here Structural Equation Modeling Results
21 The SEM analysis was utilized to examine how social capital may influence community college students’ degree aspiration. The model resulted in a good fit (𝜒 2 (236) = 1145.431, RMSEA=0.029, CFI=0.918, TLI=0.900). The SEM results revealed different patterns of how family and college social capital influence degree aspiration. First, college social capital demonstrated a positive direct effect (𝛽 =.333, 𝑝