Psychometric Properties of the Multidimensional Scale of Perceived Social Support Within Hispanic College Students

Social Support

Since the mid-1970s, there has been a growing interest in the function of social support as a coping factor and its relationship with mental and physical well-being (Bruwer, Emsley, Kidd, Lochner, & Seedat, 2008; Zimet et al., 1988). Social support has been defined as “an individual’s perception of general support from people in their social network, which enhances their functioning and/or might buffer them from adverse outcomes” (Malecki & Demaray, 2003, p. 232). Social support has been regarded as a complex multifaceted construct that has a wide variety of conceptualizations (Bruwer et al., 2008). Haber, Cohen, Lucas, and Baltes (2007) noted the chief distinction between the enacted support a person obtains and the perceived support one individual considers the adequacy and availability of care as supportive. It is perceived social support that has been strongly linked to physical and mental well-being (Haber et al., 2007).

MSPSS

One instrument addressing subjective measurement of social support adequacy that has spurred much interest in the existing literature is the MSPSS (Zimet et al., 1988). As addressed by Zimet et al. (1988), the MSPSS has a number of characteristics that make this instrument the optimal assessment tool (Bruwer et al., 2008). First of all, the MSPSS was designed to assess the perception of social support adequacy from three specific sources: family, friend, and significant others, whereas most of the measures did not consider these three sources as separate, distinct subscales (Bruwer et al., 2008; Zimet et al., 1988). Second, the MSPSS has been demonstrated to be psychometrically sound, with good reliability, factorial validity, and adequate construct validity. Third, although some other instruments of social support are lengthy and difficult to administer, the MSPSS is a brief, easy-to-use, and time-effective self-report assessment tool, which can be completed in a short time (Bruwer et al., 2008; Rizwan & Aftab, 2009).

Ample reliability and validity evidence has supported the three-factorial structure of the MSPSS, including psychometric data from adults, adolescents, undergraduates, and psychiatric patients among diverse cultures across the world, such as Mexican American adolescents (Edwards, 2004); South African youth (Bruwer et al., 2008); urban, largely African American adolescents; medical students in Thailand (Thai version; Wongpakaran & Wongpakaran, 2012); Turkish college students (Turkish version; Basol, 2008; Duru, 2007); teachers in Malaysia; Arab American adolescents (Ramaswamy, Aroian, & Templin, 2009); patients with myocardial infarction in Iran (Iranian version; Bagherian-Sararoudi et al., 2013); college students in Pakistan (Pakistani version; Rizwan & Aftab, 2009); patients with methadone maintenance treatment in China and women in postpartum period in France (French version; Denis et al., 2014). Only three validation studies have demonstrated a two-dimensional model of social support: a study with elderly (Stanley et al., 1998), a Chinese version of the MSPSS with adolescents (Chou, 2000), and an Urdu version of the scale in Hong Kong (Tonsing et al., 2012). In addition, one study in Pakistani population has highlighted one-factor solution of the social support scale (Akhtar et al., 2010).

Method

Data Analysis

Statistical power analysis

We conducted a power analysis to identify the minimum sample size needed for detecting model fit based on Stevens’s (2009) criteria, n/p ⩾ 10. With this formula in mind, using the 12-item MSPSS would require a sample size of at least 120 participants. Given our sample size of 223, we consider our sample size to be sufficient for making statistical inferences about model fit.

Preliminary analysis

After transferring our data into a Statistical Package for the Social Sciences (SPSS; IBM Corporation, 2013) file, we followed a two-step process to clean the data. First, the data set was examined, and one case was removed due to unanswered items reducing the initial sample to n = 223. Next, we replaced all missing values (125 of 10,035; 1.24%) using the series mean function.

Before conducting factor analyses, we examined the assumption of normality using the AMOS tests for normality and outliers. The results indicated that seven out of 12 were moderately nonnormal according to the critical ratio criterion (e.g., values between 1 and 10; Fouladi, 1998). Mardia’s critical ratio (59.465) also indicated the presence of multivariate kurtosis. In this regard, bootstrapping (set at 500) and Bollen-Stine bootstrap test was used. Results showed that the model marginally fit the data (p = .053). The correlations between all items, as well as item means and standard deviations for the MSPSS scale, are presented in Table 1.

OPEN IN VIEWER

Table 1.

Correlations Between Items, Means (M), and Standard Deviations (SD) of the Multidimensional Scale of Perceived Social Support.

Item	1	2	3	4	5	6	7	8	9	10	11	12
1	1.00
2	.84*	1.00
3	.30*	.30*	1.00
4	.31*	.25*	.84*	1.00
5	.71*	.79*	.33*	.35*	1.00
6	.42*	.47*	.37*	.30*	.41*	1.00
7	.32*	.33*	.36*	.33*	.28*	.66*	1.00
8	.24*	.21*	.69*	.76*	.34*	.39*	.46*	1.00
9	.43*	.46*	.45*	.38*	.36*	.73*	.77*	.41*	1.00
10	.74*	.85*	.33*	.28*	.80*	.46*	.35*	.24*	.46*	1.00
11	.33*	.30*	.79*	.82*	.39*	.38*	.40*	.78*	.45*	.31*	1.00
12	.39*	.39*	.42*	.37*	.36*	.71*	.68*	.49*	.76*	.37*	.51*	1.00
M	5.49	5.68	5.91	5.71	5.68	5.54	5.61	5.32	5.70	5.67	5.73	5.65
SD	1.61	1.64	1.41	1.48	1.57	1.39	1.45	1.72	1.47	1.63	1.47	1.52

Primary analysis

A confirmatory factor analysis (CFA) was conducted for the MSPSS using AMOS version 23. A three-factor model was created based on Zimet et al. (1988). We interpreted the chi square statistic (χ²) and p values, as well as goodness-of-fit index (GFI), comparative fit index (CFI), Tucker-Lewis index (TLI), standardized root mean square residual (SRMR), and the root mean square error of approximation (RMSEA) metrics of model fit. Furthermore, since we had the presence of nonnormality, we reported robusted fit indices using Walker and Smith’s (2017) suggestion and corresponding syntax program (Walker & Smith, 2016). When inspecting these values, we used Dimitrov’s (2012) standards in which an acceptable model fit is represented in values for the χ² (p > .05), GFI > .90, CFI > .90, TLI > .90, SRMR < .06, and RMSEA < .08.

Furthermore, reliability estimates in the normative sample were evaluated using Cronbach’s alpha (α) to assess internal consistency of the MSPSS. In addition, to provide further evidence regarding the internal structure of the MSPSS, an independent sample t-test was performed to examine the gender differences by perceived social support within Hispanic college students.

Results

CFA

To demonstrate evidence of internal structure (American Educational Research Association [AERA], American Psychological Association [APA], & National Council of Measurement in Education [NCME], 2014) and to confirm factor structure of MSPSS, a CFA was conducted. The initial model included three latent variables with 12 items. Results of model fit indices showed that the χ² was significant for the hypothesized model, χ²(51) = 155.88, p < .001. The fit indices indicated an acceptable fit for the data, GFI = .90, CFI = .95, TLI = .94, SRMR = .044, and RMSEA = .09. The results of model indices showed that the RMSEA was not an acceptable level for close model fit (RMSEA < .08). However, these results are biased since they are computed using nonnormal data (Walker & Smith, 2017). Therefore, adjusted fit indices were calculated. After running the analysis, the results indicated that the model with robust indices was a stronger fit than un-adjusted fit indices, Bollen-Stine Adjusted χ² equivalent = 36.415, Bollen-Stine scaling factor = 2.052, GFI = .90, Adjusted CFI = .99, Adjusted TLI = .99, and Adjusted RMSEA = .05. The results indicated that all 12 factor loadings between the observed variables and the latent variable were significant (p < .001; see Figure 1). Additional analyses were not conducted since the results showed that the model was a strong fit for the data.

OPEN IN VIEWER

Figure 1.

The confirmatory factor analysis model tested with the MSPSS.

Discussion

The MSPSS has already been recognized as an internationally well-accepted reliable and valid measurement tool to assess the construct of social support (Denis et al., 2014); however, epidemiological studies have emphasized the careful psychometric evaluation of research instruments before they are adapted for use within a different population and culture (Rizwan & Aftab, 2009). Since the development of the MSPSS, many researchers have examined the psychometric structure of this measure in different cultures, populations, and settings (Denis et al., 2014). However, the psychometric properties of this scale have not been investigated within Hispanic college students even though numerous studies have illustrated the critical role of social support throughout the academic life of this population (Storlie et al., 2014). Therefore, this study aimed to test the psychometric structure of the MSPSS (Zimet et al., 1988) in Hispanic college students.

In order to test the internal structure of the MSPSS and to examine the relationship of this instrument with other measures, the CFA and correlational analysis were addressed, respectively, as a means of construct validity (AERA, APA, & NCME, 2014). In addition, Cronbach’s coefficient alphas were computed for each subscale and for the total scale to test the internal consistency reliability. Furthermore, perceived social support scores of Hispanic college students were examined based on gender differences in order to provide additional evidence relevant to the internal structure of the MSPSS. Confirming the hypothesized three-factor structure (Zimet et al., 1988), the results of this investigation suggest the MSPSS as a valid and reliable tool for assessing perceived social support among Hispanic college students.

Regarding the CFA of the MSPSS within Hispanic college students, adjusted fit indices were calculated due to the presence of nonnormal data; the results indicated a strong model fit. That is to say, the findings demonstrated the three-factor subscale structure (family, friends, and significant other) with 12 items to be the most suitable solution. This factorial structure appears to be consistent with majority of the previous research (Bagherian-Sararoudi et al., 2013) except four validation studies, three of them indicating a two-dimensional model (Chou, 2000; Stanley et al., 1998; Tonsing et al., 2012) and one highlighting one-factor solution of the MSPSS (Akhtar et al., 2010). Specifically, confirming the validity of the MSPSS, the results of the CFA suggest that Hispanic college students make distinctions on the basis of the source of support: family, friends, and significant others.

Furthermore, internal consistency reliability coefficients were computed for each subscale and for the total scale. The results indicate high internal consistency reliability for the total MSPSS (α =.92) and for the subscales of Family Support (α = .93), Support From Friends (α = .91), and Support From a Significant Other (α = .94). Additional evidence regarding the internal structure of the MSPSS was sought by examining the difference between male and female students by perceived social support. Consistent with the previous studies demonstrating no gender difference by perceived social support (Dunn & O’Brien, 2009; Edwards, 2004), the results of the present study also highlighted gender-related similarity across perceived social support. That is to say, these results suggest that male and female Hispanic college students were equally as likely to perceive social support from their families, friends, and significant others.

In conclusion, Hispanic college students encounter myriad academic, social, and emotional stressors during their college lives (Baker & Robnett, 2012; Dunn & O’Brien, 2009). Social support plays a vital role in the lives of these students as a major protective factor that helps them in adaptation with challenging life events and experiences, and serves as a buffer against a number of psychological problems (Baker, 2013; Crockett et al., 2007; Delgado-Guerrero & Gloria, 2013; Solberg & Valdez, 1994). Support from any source has increasingly been recognized as valuable in maintaining good mental health. Advances in social support research depend on the existence of quality instruments to assess or classify the construct (Rizwan & Aftab, 2009). Based on the results of this study, the MSPSS appears to be a psychometrically sound measure with good construct validity and high internal consistency reliability for use with Hispanic college students. In addition, the psychometric characteristics of the MSPSS suggest the potential utility of this instrument for further evaluation of perceived adequacy of social support in this population. As such, researchers investigating perceived social support in Hispanic college students can use this measure for studies about perceived support from family, friends, and significant others, and advance our understanding of social support in this growing population. Furthermore, counselors working with Hispanic college students might well use the MSPSS in the creation of treatment strategies and interventions aimed at improving college adjustment, achievement, mental health, well-being, and acculturation. In essence, counselors and researchers can utilize the MSPSS to measure perceived social support among Hispanic college students for multifarious purposes, as this instrument is psychometrically sound, brief, user-friendly, and requires minimal training.