A confirmatory test of the factor validity of scores on the Spiritual Well-Being Scale in a community sample of African Americans.

This study examined a five-factor model of the Spiritual Well-Being Scale (SWBS; Ellison, 1983) proposed by Miller, Fleming, and Brown-Anderson (1998). A confirmatory factor analytic procedure was conducted to determine whether the Miller et al. model fit the data for the current sample of African Americans. For comparative purposes, several alternative SWBS models were included in the study's design. The findings indicated that none of the SWBS models provided an adequate fit to the data for the current sample. Implications for future use of the SWBS with African American populations are discussed.

**********

The Spiritual Well-Being Scale (SWBS; Ellison, 1983) is a popular and widely used measure of subjective well-being and overall life satisfaction. Since its development the SWBS has been used for research purposes in a variety of settings (e.g., university, community, clinical, religious), and with diverse participant groups (e.g., psychiatric, community, college students, persons with cancer, and persons with HIV/AIDS). A review of the health, subjective well-being, and spirituality literature revealed that the instrument is increasingly popular with researchers studying quality-of-life issues among African Americans. Despite this increasing popularity, there have been few efforts aimed at evaluating the instrument's psychometric integrity with African American populations. One study, conducted by Miller, Fleming, & Brown-Anderson (1998), compared the underlying factor structure of the SWBS in a sample of Caucasian American and African American participants. Their findings indicated that a three-factor SWBS model best fit the data for the Caucasian American sample while a five-factor model emerged for the African American sample. The current study was a confirmatory test of the Miller et al (1998) five-factor solution and several alternative SWBS models with a community sample of African Americans.

The SWBS is a 20-item paper and pencil self-report measure of well-being that consists of the following three dimensions: (a) religious well-being (RWB), (b) existential well-being (EWB), and (c) overall spiritual well-being (SWB; Paloutzian & Ellison, 1982). Building on the conceptual work of Moberg (1971), Ellison and his colleagues (Ellison, 1983; Paloutzian & Ellison, 1979, 1983) developed the SWBS to capture both the vertical and horizontal nature of spiritual well-being. The vertical component assesses spiritual well-being in relation to an individual's relationship with God. A sample item for this domain is, "My relation with God contributes to my sense of well-being." The horizontal component, however, assesses one's sense of purpose and satisfaction with life, absent any religious dimension. A sample item from this domain includes, "I believe there is some real purpose for my life." Items on the SWBS are rated on a 6-point Likert-type scale ranging from 1 to 6 (1 = strongly agree and 6 = strongly disagree), with higher scores indicating greater levels of well-being. The odd-numbered items comprise the RWB subscale while the even-numbered items comprise the EWB subscale. To derive a SWB score, the RWB and EWB scores are totaled (Ellison & Paloutzian, 1991).

Psychometrically, there is considerable evidence to support the SWBS as a reliable and valid measure of subjective well-being and quality of life. For example, Paloutzian and Ellison (1982) reported test-retest coefficients with a one-week interval ranging as follows: .96 for the RWB, .86 for the EWB, and .93 for the SWB. Bufford, Paloutzian, & Ellison (1991) reported test-retest coefficients ranging from .88 to .99 for the three scales at 4, 6, and 10 weeks. With regard to internal consistency, researchers have reported Cronbach's alpha coefficients ranging from .87 to .91 for the RWB, .78 to .91 for the EWB, and .76 to .93 for the SWB (Ellison, 1983; Fernander, Wilson, Staton, & Leukefeld, 2004; Genia, 2001). Evidence of the instrument's convergent validity was demonstrated in a study where the RWB, EWB, and SWB were all found to be negatively correlated with a measure of loneliness and positively correlated with measures of life purpose, intrinsic religious orientation, and self-esteem (Ellison, 1983). In more recent studies the SWB was found to correlate with emotional well-being and life satisfaction (Kim, Heinemann, Bode, Sliwa, & King, 2000), self-esteem and depression (Genia, 2001), and ethnicity (Fernander et al., 2004).

The Spiritual Well-Being Scale was developed using a sample of 206 Caucasian American students from three religious-oriented colleges (Ellison, 1983). An exploratory factor analysis indicated that the SWBS was a multidimensional measure consisting of 2 subscales--Religious Well-Being and Existential Well-Being, and an overall Spiritual Well-Being score. This finding failed to hold up in a number of subsequent factor analytic studies (Fernander et al., 2004; Genia, 2001; Ledbetter, Smith, Fischer, Vosler-Hunter, & Chew, 1991; Miller et al., 1998). To evaluate the psychometric integrity of the SWBS across racial and ethnic populations, Miller et al. conducted a comparison study with Caucasian American (n = 119) and African American (n = 97) samples. The study's participants were solicited from undergraduate and graduate classes at colleges and universities in the southwestern United States. Findings from this study produced a three-factor model (accounting for 64% of the total variance) for the Caucasian American sample and a five-factor model (likewise, 64% of the variance was accounted for by this model) for the African American sample. The five-factor African American model included the following subscales: (a) Connection with God, (b) Satisfaction with God and day-to-day living, (c) Future/Life Contentment, (d) Personal Relationship with God, and (e) Meaningfulness. While the factor analytic work of Miller et al. (1998) contributes to our understanding of the underlying structure of spiritual well-being among African Americans, it is limited by a number of critical flaws related to the study's methodology and data analyses.

The study conducted by Miller et al. (1998) was flawed for several reasons. First, the sample size was inadequate for what is typically required for conducting exploratory factor analysis. Most rules of thumb require at a minimum 150-300 cases (Gorsuch, 1983). Even the most lenient subject-to-variable ratios require at minimum 5 participants for every variable (Bryant & Yarnold, 1995). Second, two of the five factors extracted in the Miller et al. study have as few as three items thereby making them difficult to interpret. Because the authors failed to report alpha coefficients for the SWBS the viability of the three-item subscales cannot be further evaluated and therefore remains questionable. Third, the authors do not provide adequate detail in reporting the findings from their factor analysis. For example, we do not know what criteria was used for retaining items to factors, nor is there any indication of whether some items loaded on more than one factor. Last, given that the study's participants were exclusively college students the degree to which the study's findings can be generalized to other populations of African Americans is limited.

Hence, the primary purpose of the current study was to conduct a confirmatory test of the Miller et al. (1998) SWBS five-factor model in a sample of African American adults from the community. In addition, for comparative purposes, we evaluated the original SWBS two-factor model proposed by Ellison (1983), a one-factor model proposed by Ledbetter et al. (1991), and a three-factor model proposed by Miller et al. (1998). Both varimax and oblimin solutions were tested for the two-, three, and five-factor models. Moreover, we tested hierarchical models in the confirmatory procedure for the two-, three-, and five-factor SWBS models. We anticipate that the current study's finding will contribute to further our understanding of spiritual well-being among African American populations.

METHOD

Participants

A total of 291 African Americans were solicited to participate in the current study. Individuals were recruited from community-sponsored events, an adult learning center, and a residential job-training program all located in the northeastern United States. Of the 291 participants in the study, 238 were female (54.2%) and 189 were male (45.6%). There was one missing value for gender (.3%). The participants ranged in age from 18 to 69 years, with a mean age of 25.82 and a standard deviation of 10.28. With regard to education, participants varied such that 69 (23.7%) did not complete high school, 79 (27.1%) completed high school, 92 (31.6%) completed some college, 34 (6.2%) had bachelor's degrees, and 13 (5.2%) had graduate degrees. There were four (1.3%) missing values for education.

Instrument

The Spiritual Well-Being Scale (SWBS; Ellison, 1983) is a 20-item pencil and paper self-report measure of religious and existential well being. The SWBS requires 10-15 minutes to complete. Ten items comprise the Religious Well-Being (RWB) and Existential Well-Being (EWB) subscales; the two subscales combine to form the SWBS. The RWB measures a respondent's perception of God and the perceived involvement of God in their life. The EWB refers largely to one's sense of meaning in life and satisfaction with life's present direction.

Procedure

Participants were administered a packet of survey questionnaires that included the SWBS and a demographic data sheet. Instructions for completing the questionnaires were given verbally and participants were made aware of their right to withdraw from the survey without consequences. The questionnaires were counter-balanced to control for instrumentation effects. The surveys were collected and the participants were debriefed individually and as a group. All participants were compensated $10 for their time.

RESULTS

Confirmatory Factor Analysis of the SWBS

A confirmatory factor analytic (CFA) procedure using the maximum likelihood method was used to examine how well the SWBS five-factor model proposed by Miller et al. (1998) fit the data for the current sample of African Americans. In addition, we evaluated the fit of several alternative models for comparison to the Miller et al. model. Specifically, a null SWBS model, and one-, two-, and three-factor SWBS models were evaluated for their fit to the data in the current sample. For the two-, three-, and five-factor models, we evaluated the fit of both varimax and oblimin solutions. Moreover, we examined whether including a hierarchical component to the two-, three-, and five-factor models improved their fit to the data. Means, standard deviations and Cronbach's alpha coefficients were calculated for each model and are reported in Table 1.

Following the established procedures for structural equation modeling (SEM) with AMOS 4.0 (Byrne, 2001), several fit indexes were calculated for all of the competing SWBS models. First, we computed the chi-square ([chi square]) statistic for each of the competing models and tested its significance. The likelihood ratio test statistic ([chi square]) represents the discrepancy between the unrestricted sample covariance matrix and the restricted covariance matrix (Byrne, 2001). Results from the SEM procedure indicated that Miller et al (1998) five-factor solution and its hierarchical extension was inadmissible. Furthermore, all of the models produced a statistically significant ([chi square]) (see Table 2), indicating that the observed and specified models differed. As noted, this is not uncommon for large samples and does not necessarily reflect a poor fit to the data (Bollen, 1989; Fassinger, 1987).

The AMOS program provides additional measures of fit that are less sensitive to sample size and considered to be better estimates of a model's fit to the data. One such index is the root mean square error of approximation (RMSEA). The RMSEA is an estimate of the lack of fit in a given model compared to a perfect model and is used to compare the fit of two different models to the same data set (Browne & Cudeck, 1993). RMSEA values of .05 or less are said to reflect a model with a good fit to the data, while values less than .08 suggest an adequate fit to the data (Browne & Cudeck, 1993). Based on the fit indexes reported in Table 2, the original two-factor model (oblimin solution) proposed by Ellison et al. (1983) and the five-factor model (oblimin solution) proposed by Miller et al. (1998) had RMSEA coefficients of .08 or less, suggesting an adequate fit to the data. Note that the two-factor and five-factor hierarchical SWBS models also produced an adequate fit (i.e., RMSEA [less than or equal to] .08) to the data for the current sample of African Americans.

In addition to [chi square] and RMSEA, we calculated the standardized root mean square residual (SRMR) for each of the competing SWBS models. The SRMR represents the average discrepancy between the observed and hypothesized correlation matrix, with values ranging from 0 to 1 (Hu & Bentler, 1995). Similar to RMSEA, SRMR is a measure of global fit and is used to compare the fit of models for different data. Standardized RMR values of .05 or less suggests a good fitting model. Based on the findings reported in Table 2, only the five-factor oblimin SWBS model proposed by Miller et al. (1998), with a Standardized RMR of .06, approximated a reasonable fit to the data for the current sample of African Americans. Although the five-factor model had the best SRMR among the competing models, the findings, including the solution's inadmissibility, are indicative of a failure to confirm its fit to the data.

DISCUSSION

The SWBS is the product of a movement to identify quality-of-life indicators among various clinical samples. The original measure consisted of three factors representing the following constructs: religious well-being (RWB), existential well-being (EWB), and overall spiritual well-being (SWB). Miller et al. (1998) proposed a five-factor model as best representing the underlying factor structure of the SWBS in an African American sample. Their model consisted of the following five factors: connection with God, satisfaction with God and day-to-day living, future/life contentment, personal relationship with God, and meaningfulness. The purpose of this study was to examine how well the five-factor model proposed by Miller et al. (1998), as well as several competing SWBS models, fit the data with the current sample of African Americans.

AMOS 4.0 (Arbuckle, 1999), a software program for structural equation modeling was used to evaluate how well each of the competing SWBS models fit the data for the current sample of African Americans. To evaluate how well the Miller et al. (1998) five-factor model fit the data in comparison to the competing SWBS models, we calculated several goodness-of-fit indexes. Goodness-of-fit indexes were also calculated for a two-, three-, and five-factor hierarchical model. Our findings indicated that none of the models demonstrated an adequate fit to the data for the current sample of African Americans. This assessment is based on statistically significant [chi square] for all competing models, RMSEAs > .05, and RMSRs > .05. Taken together, these results suggest the absence of an interpretable underlying factor structure for the SWBS in the current African American sample.

These findings are not inconsistent with the body of prior research on the factor validity of scores on the SWBS. Several researchers, working independently, have evaluated the underlying factor structure of scores on the SWBS with different results. For example, Ledbetter et al. (1991), using archival data on the SWBS conducted a confirmatory test of a general factor model and the original two-factor model proposed by Ellison (1983). His findings indicated that both models produced a poor fit to the data. In contrast, Genia (2001) conducted a Principal Axis Factor analysis with the SWBS. Her findings supported the original two-factor model proposed by Ellison. That the current study also failed to confirm an adequate fitting SWBS model for the current sample brings into question both the validity of spiritual well-being as a psychological construct and/or our ability to measure this construct. Psychological tests such as the SWBS are said to possess construct validity to the degree there is evidence that it measures what it purports to measure. As such, the SWBS is without any consistent evidence of its construct validity as a measure of individuals' subjective wellbeing in relation to the spiritual domain of human functioning.

As regards to the validity of the underlying factor structure of the SWBS with African American populations, several factors are worth noting. First, although spirituality as a construct is not unique to African Americans, its expression is culturally determined. For example, anthropologists have found that African Americans, heavily influenced by their West African cultural origins, tend to view themselves as essentially religious and spiritual beings (Ani, 1990; Mbiti, 1989). According to Stewart (1999), African Americans view spirituality as a force that is active, transformative, and purposeful in day-to-day life. For African Americans, living in spiritual harmony, which includes fulfilling one's life purpose, is viewed as a necessary condition for protection against life's adversities (Newlin, Knafl, & Melkus, 2002). Another factor to consider regarding the validity of the SWBS with African American populations is the absence of any substantial African American representation in the early development of the measure. Determining whether the SWBS is a valid measure of spiritual well-being with African American populations was a methodological afterthought. Note that fifteen years had passed since the SWBS' initial development (Ellison, 1983) before the measure's validity in cross-cultural research was evaluated (i.e., Miller et al., 1998). Given that the SWBS was not initially intended for use with African American populations, perhaps it is psychometrically advisable not to force a fit.

Although the current study provides some useful information as to the validity, or lack thereof, of the SWBS' underlying factor structure with African American populations, several limitations are worth noting. First, the generalizability of the study's findings is limited in that the participants came from urban communities in the northeastern region of the United States. Therefore, we might anticipate a different result with African Americans from other geographical locations within the United States. Moreover, the use of a self-report measure to assess participants' spiritual well-being is problematic. Self-report measures are prone to increased measurement error due to retrospective recall bias, problems with accuracy of reporting, and a potential for respondents to be confused by the directions. A final limitation is the use of a cross-sectional, single method design to assess factor validity of the SWBS with an African American sample. A multi-trait/multi-method design might provide more reliable information regarding the validity of the SWBS with African American populations. Hence, caution is warranted in interpreting the current study's findings.

Future research may include qualitative studies to determine, contextually, the role and practice of spirituality within African American communities. This information might inform the development of new and more appropriate measures of spiritual well-being in African American populations. Although extensive work has been done to identify appropriateness of the SWBS with differing clinical populations, similar rigor should capture diverse spiritual definitions and practices with African American samples.

REFERENCES

Ani, M. (1990). Let the circle be unbroken: The implications of African spirituality in the Diaspora. New York, NY: Nkonimfo Publications.

Bollen, K. (1989), Structural equations with latent variables. New York: John Wiley.

Byrne, B. M. (2001). Structural equation modeling with AMOS: Basic concepts, applications, and programming. Mahwah, NJ: Lawrence Erlbaum.

Bryant, F. B., & Yarnold, P. R. (1995). Principal-components analysis and exploratory and confirmatory component analysis. In L. G. Grimm & P.R. Yarnold (Eds.), Reading and understanding multivariate statistics (pp. 99-136). Washington, DC: American Psychological Association.

Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 445-455). Newbury Park, CA: Sage.

Bufford, R., Paloutzian, R., & Ellison, C. (1991). Norms for the Spiritual Well-Being Scale. Journal of Psychology and Theology, 19, 56-70.

Ellison, C. (1983). Spiritual Well-Being: Conceptualization and measurement. Journal of Psychology and Theology, 11, 330-340.

Ellison, R., & Paloutzian, C. (1991). Manual for the Spiritual Well-Being Scale. Nyack, NY: Life Advance, Inc.

Fassinger, R. (1987). Use of structural equation modeling in counseling psychology research. Journal of Counseling Psychology, 34, 425-436.

Fernander, A., Wilson, J. F., Staton, M., & Leukefeld, C. (2004). An exploratory examination of the Spiritual Well-Being Scale among incarcerated Black and White male drug users. International Journal of Offender Therapy and Comparative Criminology, 48, 403-413.

Genia, V. (2001). Evaluation of the Spiritual Well-Being Scale in a Sample of College Students. International Journal for the Psychology of Religion, 11, 25-33.

Gorsuch, R. L. (1983). Factor analysis (2nd ed.). Hillside, NJ: Erlbaum.

Hu, L-T., & Bentler, P. M. (1995). Evaluating model fit. In R. H. Hoyle (Ed.), Structural equation modeling: Concepts, issues, and applications (pp. 76-99). Thousand Oaks, CA: Sage.

Kambon, K. K. (1998). African/Black psychology in the American context. Tallahassee, FL: Nubian Nation Publications.

Kim, J., Heinemann, A. W., Bode, R. K., Sliwa, J., & King, R. B. (2000). Spirituality, quality of life, and functional recovery after medical rehabilitation. Rehabilitation Psychology, 45, 365-385.

Ledbetter, M.F., Smith, L. A., Fischer, J. D., Vosler-Hunter, W. L., & Chew, G. P. (1991). An evaluation of the construct validity of the Spiritual Well-Being Scale: A confirmatory factor analytic approach. Journal of Psychology and Theology, 19, 94-102.

Mbiti, J. S. (1989). African religions and philosophy (2nd ed.). Oxford: Heinemann.

Miller, G., Fleming, W,. & Brown-Anderson, F. (1998). Spiritual Well-Being Scale Ethnic Differences Between Caucasians and African Americans. Journal of Psychology and Theology, 26, 358-364.

Moberg, D. (1971). Spiritual well-being: Background and issues. Washington, DC: WhiteHouse Conference on Aging.

Newlin, K., Knafl, K., & Melkus, G. D. (2002). African-american spirituality: A concept analysis. Advances in Nursing Science, 25(2), 57-70.

Paloutzian, R. F., & Ellison, C. W. (1982). Loneliness, spiritual well-being, and quality of life. In L. A. Peplau, & D. Perlman (Eds.), Loneliness: A sourcebook for current theory, research, and therapy. New York: Wiley Interscience.

Paloutzian, R. F., & Ellison, C. W. (1979, August). Developing a measure of spiritual well-being. Paper presented at the American Psychological Association convention, New York, NY.

Stewart, C. F. (1999). Black spirituality and Black consciousness: Soul force, culture, and freedom in the African American experience. Trenton, NJ: African World Press.

AUTHORS

UTSEY, SHAWN O.: Address: Department of Psychology, Virginia Commonwealth University, 800 W. Franklin Street, Richmond, VA 23284. Title: Associate Professor of Counseling Psychology. Degrees: BA, N.C. A & T State University; MA, New York University; Ph.D., Fordham University. Specializations: African American culture and resilience, African American coping, psychological effects of racism on African Americans.

LEE, ANGELA. Address: Howard University, Washington, D.C. Title: Graduate student.

BOLDEN, MARK. Address: Howard University, Washington, D.C. Title: Graduate student.

LANIER, YZETTE. Address: Howard University, Washington, D.C. Title: Graduate student.

SHAWN O. UTSEY

Virginia Commonwealth University

ANGELA LEE, MARK A. BOLDEN, and YZETTE LANIER

Howard University

Correspondence concerning this article may be sent to Shawn O. Utsey, Ph.D., Department of Psychology, Virginia Commonwealth University, 800 West Franklin Street, Richmond, VA 23284. Email: [e-mail]

TABLE 1 Means, Standard Deviations, and Cronbach's Alpha Coefficients
for the One-, Two-, Three-, and Five-Factor SWBS

Model                      M          SD         [chi]

One-Factor SWBS
  Factor 1               96.40       15.55        .89
Two-Factor SWBS
  Factor 1               47.21        8.07        .79
  Factor 2               49.04        9.36        .87
Three-Factor SWBS
  Factor 1               49.04        9.36        .87
  Factor 2               28.70        5.09        .71
  Factor 3               18.38        3.89        .62
Five-Factor SWBS
  Factor 1               19.30        4.40        .82
  Factor 2               28.14        5.77        .73
  Factor 4               15.55        2.81        .54
  Factor 5               15.32        2.66        .49

TABLE 2 Confirmatory Factor Analysis of the Original Five-Factor
Spiritual Well-Being Scale Proposed by Miller et al. (1998).

Model                              [chi square]    p       df

Ledbetter et al. (1991)
  One-factor                          670.69      .0001    170
Ellison et al. (1983)
  Two-factor
    Oblimin                           490.59      .0001    167
    Varimax                           615.34      .0001    168
  Two-factor hierarchical             480.39      .0001    166
Miller et al. (1998)
  Three-factor (Euro American)
    Oblimin                           497.70      .0001    163
    Varimax                           729.40      .0001    166
  Three-factor hierarchical           482.13      .0001    162
  Five-factor (African American)
    Oblimin                           421.32      .0001    160
    Varimax                           982.08      .0001    170
  Four-factor hierarchical            443.81      .0001    165

Model                              [chi square]/df    RMSEA   RMSR

Ledbetter et al. (1991)
  One-factor                            3.94           .10    .08
Ellison et al. (1983)
  Two-factor
    Oblimin                             2.93           .08    .07
    Varimax                             3.66           .10    .19
  Two-factor hierarchical               2.89           .08    .07
Miller et al. (1998)
  Three-factor (Euro American)
    Oblimin                             3.05           .09    .08
    Varimax                             4.39           .11    .20
  Three-factor hierarchical             2.97           .09    .08
  Five-factor (African American)
    Oblimin                             2.63           .08    .06
    Varimax                             5.77           .13    .25
  Four-factor hierarchical              2.69           .08    .07s