Measuring psychological well-being in the Chinese

https://doi.org/10.1016/j.paid.2004.08.013Get rights and content

Abstract

A 4-item-per-construct version of the Scales of Psychological Well-being (Ryff, 1989) was constructed and administered to a sample of adults in Hong Kong (ages 18–86). Three competing models were evaluated using confirmatory factor analysis. Results suggested that a simple 6-factor model without any higher-order factor provided the best fit to the data, though only moderatly fitting. Furthermore, an examination of the factor loadings suggests that individual items might need further refinement or replacement. While the original 3-item scales suffered from poor internal consistencies, the new 4-item scales had relatively more acceptable internal consistency coefficients.

Introduction

Over the past decade and more, Ryff and her colleagues (e.g., Ryff & Singer, 1998) have successfully established a model of psychological well-being, or positive mental health, that consists of six aspects of functioning—autonomy, environmental mastery, personal growth, positive relations with others, purpose in life, and self-acceptance. Her model has been widely studied across the world (e.g., Antonelli & Cucconi, 1998; Casullo & Castro-Solano, 2001; Clarke, Marshall, Ryff, & Wheaton, 2001; Van Dierendonck, 2004). To measure these constructs, Ryff has developed different versions of the Scales of Psychological Well-being, including versions with 20 items per construct, 14 items, 9 items and 3 items (see Ryff, 1989; Ryff, Lee, Essex, & Schmutte, 1994; Ryff & Keyes, 1995). For a review of the theoretical background and the history of scale development, see Van Dierendonck (2004).

The repeated attempts to create shorter versions indicate a need for a psychometrically sound measure, yet short enough to be included in large-scale surveys of multiple variables. However, data available from the 3-item version showed consistently poor internal consistencies (Clarke et al., 2001; Keyes, Shmotkin, & Ryff, 2002; Ryff & Keyes, 1995; Van Dierendonck, 2004).

A begging question of the Ryff scale is its factorial structure. Ryff and Keyes (1995) first tested the structure of the 3-item scales in a large, random sample of American adults aged 25 or older. They drew the conclusion that a second-order factorial model in which a higher-order factor accounting for the intercorrelations of the six aspects of optimal functioning (hereafter called the hierarchical model) provided the best, parsimonious fit to the data because (a) the hierarchical model and the 6-factor model (without the second-order factor) did not result in a significant difference in the Bayesian Information Criteria, and (b) the former is more parsimonious because of the exclusion of first-order factor covariances from the model. The adjusted goodness-of-fit indices of the two models were also the same, being .89 (which is lower than conventional cutoffs). Furthermore, both models evidenced improvements in fit over the 1-factor model and other variants of the model. We want to note, however, that there was a significant chi-square (χ2) difference (p < .001) between the 6-factor and the hierarchical model, and hence the conclusions premissible from the χ2 test and the BICs were inconsistent.

Van Dierendonck (2004) tested a Dutch version of the Ryff scales and found that the 3-item scales suffered from poor Cronbach’s alpha coefficients but had better factorial validity as evidenced from fit indices than the 9-item and the 14-item versions. Their finding concurred with Ryff and Keyes (1995) in that the best fitting model was the hierarchical model, but again the non-normed fit indices (NNFI) and the comparative fit indices (CFI) in a student and an adult sample were all below .90. It needs to be mentioned that Van Dierendonck (2004) did not estimate the factor covariances in the 6-factor model (i.e., assuming the covariances to be zero), as evident in the degrees of freedom for that model. In the hierarchical model, however, a higher-order factor was allowed to explain the interrelationships of the six factors. As it is known that the six factors are interrelated to some extent (e.g., Keyes et al., 2002; Ryff & Keyes, 1995), comparing a 6-factor model without covariances to a hierarchical model incorporating the covariances was unfair for the former. Another analytical issue concerns the fact that factor loadings were not reported in previous studies (cf. Clarke et al., 2001; Ryff & Keyes, 1995; Van Dierendonck, 2004), hence making it difficult to understand clearly what had contributed to the lack of fit.

A more interesting part of Van Dierendonck’s (2004) study was the development of alternative, short versions of the Ryff scales. On the basis of item-total correlations and the lack of cross-loading, 6–8 items per scale were selected from the 14-item versions to form the six Scales of Psychological Well-being. Their recommended versions had Cronbach’s alpha coefficients of .72 to .81 in the student sample. Although the fit indices for this new set of scales (hierarchical model) was still below the acceptable range (NNFI = .84, CFI  = .85, standardized root mean square residuals [SRMR] = .06), they were much better than the original 9- and 14-item versions.

Van Dierendonck’s (2004) study was an attempt to develop psychometrically strong, yet short measures of Ryff’s psychological well-being constructs. The present study was conducted in parallel to that of Van Dierendonck (2004) with a similar aim, but with a different strategy. Our approach differed from Van Dierendonck (2004) in two ways: (a) we aimed at 4 items per scale and started from the 9-item rather than the 14-item version as in Van Dierendonck (2004) and (b) we sought to come up with an equal number of items for each scale as is done in all Ryff versions (Van Dierendonck’s, 2004 version has different number of items for different scales).

An issue that has not been addressed in the literature is whether the Ryff model is equally fitting for different population subgroups. An interest generated from Ryff and colleague’s research is how well-being differs across sociodemographic backgrounds (Keyes et al., 2002; Ryff, 1991; Ryff & Keyes, 1995). The only study that has studied the measurement of the Ryff model in a population subgroup is that of Clarke et al. (2001). Their study, based on a large sample of Canadian elders, focused on the 6-factor model (without the higher-order factor) and found that the CFI was only .77 (without allowing for cross-loadings; Clarke et al., 2001). Their finding suggested that the psychological well-being scales might have problems of measurement in older adults. The present study will investigate whether the proposed measurement model is well-fitting for males and females, and for young- (ages 18–29), middle- (ages 30–59) and old-aged (ages 60+) groups.

To sum up, in view of the low alpha coefficients of the 3-item versions, we sought to develop 4-item versions which would be short enough for large-scale administrations but which would evidence better internal consistencies and factorial validity.

Section snippets

Method

The selection of items for the Chinese version was more difficult than we had thought. Because we were interested in developing a version of the Ryff scales that are applicable to the adult population, we recruited initially a convenience sample of community adults (N = 83, age range = 18–99, M = 47.1, SD = 16.5) for examining item responses. Because we had assumed that the most useful items were retained in each successive shorter versions of the Ryff scales, we began with the 9-item per scale

Method

Twenty hospitals were randomly sampled from the 40 public hospitals all over Hong Kong, and questionnaires were mailed to all those registered as volunteers in these hospitals. Most of these were simply ex-patients who indicated at discharge that they might be available if in the future the hospital needed voluntary assistance with certain services. The psychological well-being items were randomly embedded within a longer questionnaire containing other measures of motivation and volunteer

Discussion

Our initial aim to include Ryff’s 3-item scales was unsuccessful. However, we demonstrate that it was possible to improve the psychometric properties of the scales by marginally increasing their lengths. The new 4-item scales are psychometrically superior to the existing 3-item scales, and yet still suitable for inclusion in lengthy surveys. Our finding that supported the 6-factor model was at odds with claims in the literature (Ryff & Keyes, 1995; Van Dierendonck, 2004). However, as we have

Acknowledgements

The preparation of this manuscript was supported in part by the Hospital Authority, for which we are appreciative. We thank the various hospitals which provided assistance in data collection.

References (25)

  • L. Lu et al.

    Two ways to achieve happiness: When the East meets the West

    Personality and Individual Differences

    (2001)
  • D. Van Dierendonck

    The construct validity of Ryff’s Scales of Psychological Well-being and its extension with spiritual well-being

    Personality and Individual Differences

    (2004)
  • E. Antonelli et al.

    Effects of retirement on psychological well-being, the self-concept, and self-esteem

    Ricerche di Psicologia

    (1998)
  • K.P. Burnham et al.

    Model selection and multimodel inference: A practical information theoretic approach

    (2002)
  • B.M. Byrne et al.

    The issue of measurement invariance revisited

    Journal of Cross-Cultural Psychology

    (2003)
  • M.M. Casullo et al.

    Adolescent students’ perception of psychological well-being meaning

    Revista Iberoamericana de Diagnostico y Evaluacion Psicologica

    (2001)
  • Census and Statistics Department (2002). Hong Kong 2001 population census main report, Vols. I & II. Hong...
  • S.-T. Cheng et al.

    Quality of life in old age: An investigation of well-older persons in Hong Kong

    Journal of Community Psychology

    (2004)
  • F.M. Cheung et al.

    Indigenous Chinese personality constructs: Is the Five-Factor Model complete?

    Journal of Cross-Cultural Psychology

    (2001)
  • P.J. Clarke et al.

    Measuring psychological well-being in the Canadian study of health and aging

    International Psychogeriatrics

    (2001)
  • D.Y.F. Ho et al.

    Indigenization and beyond: Methodological relationalism in the study of personality across cultural traditions

    Journal of Personality

    (2001)
  • L. Hu et al.

    Evaluating model fit

  • Cited by (81)

    • Longitudinal study on affect, psychological well-being, depression, mental and physical health, prior to and during the COVID-19 pandemic in Spain

      2021, Personality and Individual Differences
      Citation Excerpt :

      The scales contain 4 to 6 items. The reliability and validity of the multiple versions of the Ryff's Scales of Psychological Well-Being have been well documented and there have been a number of psychometric studies of the multi-dimensional structure of the SPWB (e.g., Abbott et al., 2010; Cheng & Chan, 2005; Díaz et al., 2006; Lindfors et al., 2006; Ryff, 1989b; Ryff & Keyes, 1995; Springer & Hauser, 2006; Van Dierendonck, 2004). Table 2 presents the results of internal consistency in the population of this study.

    • Psychometric evaluation of the Ryff's Scale of psychological wellbeing in self-identified American entrepreneurs

      2020, Journal of Business Venturing Insights
      Citation Excerpt :

      All in all, the data favored the SPWB-38 parceled CFA Model 1 the best, which aligns with the original theoretical model proposed by Ryff and colleagues (Ryff, 1989; Ryff and Singer, 2006) (see Table 3). The factorial structure of the SPWB scales has been explored across different population samples (adolescents, Chan et al., 2017; Friedman et al., 2017), cultural contexts (e.g., Chinese, Cheng & Chan, 2005; Spanish, Van Dierendonck et al., 2008; Swedish, Lindfors et al., 2006) and versions (e.g., 3, 9, and 14 item versions, Shokri et al., 2008). The original English version has been translated into several other languages (Ryff, 2019).

    • Respect Your Elders: Generativity and Life Satisfaction in Caregiving Grandparents

      2023, International Journal of Aging and Human Development
    View all citing articles on Scopus
    View full text