Validity of the Satisfaction with Life Scale in Colombia and factorial equivalence with Spanish data

Full reference: Ruiz, F. J., Suárez-Falcón, J. C., Flórez, C. L., Odriozola-González, P., Tovar, D., López-González, S., & Baeza-Martín, R. (2019). Validity of the Satisfaction with Life Scale in Colombia and factorial equivalence with Spanish data. Revista Latinoamericana de Psicología, 51(2), 58-65. https://doi.org/10.14349/rlp.2019.v51.n2.1

Study type: Cross-cultural validation and measurement invariance analysis

Background and objectives

The Satisfaction with Life Scale (SWLS; Diener et al., 1985) is a widely used instrument for assessing life satisfaction, conceptualized as a global subjective evaluation of quality of life based on the comparison between an individual's current state and their standard of what is desirable. Although the SWLS has been translated into Spanish and used in various Colombian studies, it had not been formally validated in the Colombian context, which limited the ability to conduct research with solid scientific guarantees. Furthermore, although the SWLS is frequently used to compare life satisfaction across countries, there is limited evidence for factorial equivalence between Spanish-speaking cultures, particularly between Colombia and Spain.

The main objective of this study was twofold: (1) to explore the psychometric properties of the SWLS in Colombia using three distinct samples (undergraduate students, general population, and clinical sample), and (2) to analyze the factorial equivalence of the instrument between Colombian and Spanish samples. The authors hypothesized that the SWLS would show adequate internal consistency, unidimensional structure, measurement invariance by gender in Colombian samples, and positive correlations with valued living measures as well as negative correlations with emotional symptoms, negative thoughts, experiential avoidance, and cognitive fusion.

Method

Participants

The study included a total of 1,587 Colombian participants distributed across three samples:

• Sample 1 (Colombian undergraduate students): 762 undergraduates (age range 18-63, M = 21.16, SD = 3.76) from seven universities in Bogotá. Forty-six percent were studying psychology; others were enrolled in law, engineering, philosophy, communication, business, medicine, and theology. Sixty-two percent were women. Twenty-six percent reported having received psychological or psychiatric treatment at some point, but only 4.3% were currently in treatment. 2.9% were taking psychotropic medication.

• Sample 2 (Colombian general population online): 724 participants (74.4% women, age range 18-88, M = 26.11, SD = 8.93). 17.8% had primary or secondary education, 63.8% were undergraduate or college graduates, and 18.4% were pursuing or had completed postgraduate studies. They responded to an anonymous online survey distributed through social media. Forty-five percent reported having received psychological or psychiatric treatment at some point, but only 8.4% were currently in treatment. 5.4% reported using psychotropic medication.

• Sample 3 (Colombian clinical sample): 101 patients (52 women, age range 18-67, M = 32.22, SD = 12.09) being treated at a private psychological consultation center in Bogotá. 67.3% presented with an emotional disorder and 32.7% with a sexual disorder. Only 5% reported taking psychotropic medication.

The three Spanish samples totaling 1,057 participants were:

• Sample 4 (Spanish undergraduate students): 235 students (62.6% women, age range 18-61, M = 27.95, SD = 10.55) from three Spanish universities. Twenty-seven percent were studying psychology. 17.4% had received psychological or psychiatric treatment at some point, 3.4% were currently in treatment, and 3.8% were taking psychotropic medication.

• Sample 5 (Spanish general population online): 731 participants (78.1% women, age range 18-89, M = 35.09, SD = 11.72). 34.7% had primary or secondary education, 42% were undergraduate or college graduates, and 23.3% were pursuing or had completed postgraduate studies. They responded to an anonymous online survey distributed through social media. Forty-four percent had received psychological or psychiatric treatment at some point, 13.1% were currently in treatment, and 12.2% reported using psychotropic medication.

• Sample 6 (Spanish clinical sample): 91 patients (54 women, age range 18-63, M = 39.45, SD = 13.04) being treated at a private psychological consultation center in Spain. 76.9% presented with an emotional disorder. Thirty-six percent reported taking psychotropic medication.

Instrument(s) under study

Satisfaction with Life Scale (SWLS; Diener et al., 1985; Spanish translation by Atienza et al., 2000): A 5-item, 7-point Likert-type scale (7 = strongly agree; 1 = strongly disagree) that measures self-perceived well-being. Items include statements such as "I am satisfied with my life" and "In most ways, my life is close to my ideal." Prior to administration, a pilot study with 10 Colombian undergraduate students confirmed the comprehensibility of the items.

Other outcome measures

• Valuing Questionnaire (VQ; Smout et al., 2014; Spanish translation by Ruiz et al., 2019): 10 items, 7-point Likert scale. Assesses valued living with two subscales: Progress (value enactment) and Obstruction (disruption of valued living). Alphas reported in the study: Progress .81-.88; Obstruction .78-.86.

• Depression, Anxiety and Stress Scales-21 (DASS-21; Antony et al., 1998): 21 items, 4-point Likert scale. Measures negative emotional states with three subscales: Depression, Anxiety, and Stress. Alphas reported: Depression .86-.91; Anxiety .80-.85; Stress .80-.88.

• General Health Questionnaire-12 (GHQ-12; Goldberg & Williams, 1988): 12 items, 4-point Likert scale. Screening tool for psychological distress. Alphas reported: .88-.91.

• Automatic Thoughts Questionnaire-8 (ATQ-8; Netemeyer et al., 2002): 8 items, 5-point Likert scale. Measures frequency of negative automatic thoughts. Alphas reported: .85-.91.

• Acceptance and Action Questionnaire-II (AAQ-II; Bond et al., 2011): 7 items, 7-point Likert scale. Measures experiential avoidance and psychological inflexibility. Alphas reported: .88-.91.

• Cognitive Fusion Questionnaire (CFQ; Gillanders et al., 2014): 7 items, 7-point Likert scale. Measures cognitive fusion. Alphas reported: .89-.93.

Data analysis

LISREL 8.71 was used for missing data imputation using matching response pattern (only one value was missing) and confirmatory factor analysis (CFA). Because the SWLS uses an ordinal scale, a weighted least squares (WLS) method with polychoric correlations was employed, recommended for large samples with fewer than 20 items. Fit indices evaluated included RMSEA (root mean square error of approximation), CFI (comparative fit index), and NNFI (non-normed fit index). RMSEA < .05 was considered very good fit, < .08 acceptable; CFI and NNFI > .95 good fit and > .90 acceptable.

To analyze metric invariance (equal factor loadings) and scalar invariance (equal factor loadings and item intercepts), three nested models were compared: multiple-group baseline model (configural invariance), metric invariance model, and scalar invariance model. Criteria from Cheung and Rensvold (2002) and Chen (2007) were applied: ΔRMSEA < .01, ΔCFI and ΔNNFI ≥ -.01.

Cronbach's alpha and McDonald's omega were calculated with 95% confidence intervals using percentile bootstrap in the MBESS package in R. Corrected item-total correlations, descriptive analyses, independent samples t-tests (with Cohen's d for significant effects), and Pearson correlations for convergent validity were obtained.

Results

Factor structure

The fit of the unidimensional model was very good in the combined Colombian samples: χ²(5) = 26.18, p < .01; RMSEA = .052, 90% CI [.034, .073]; CFI = .99; NNFI = .99. In separate samples, Sample 1 (students) showed χ²(5) = 18.42, p < .01; RMSEA = .059, 90% CI [.032, .090]; CFI = .98; NNFI = .99; Sample 2 (general population) showed χ²(5) = 14.94, p < .01; RMSEA = .053, 90% CI [.040, .070]; CFI = .99; NNFI = .99. Sample 3 (clinical) showed a higher RMSEA (χ²(5) = 9.50, p < .01; RMSEA = .095, 90% CI [.00, .18]; CFI = .97; NNFI = .99) attributable to the small sample size. Standardized factor loadings in the overall Colombian sample were: Item 1 = .79; Item 2 = .82; Item 3 = .91; Item 4 = .73; Item 5 = .78.

Reliability

Internal consistency was excellent. Cronbach's alpha ranged from .85 (Samples 1 and 3) to .89 (Sample 2), with an overall alpha for the total Colombian sample of .89 [95% CI .88, .90]. McDonald's omega was very similar: Sample 1 = .85 [95% CI .83, .87]; Sample 2 = .89 [95% CI .88, .91]; Sample 3 = .86 [95% CI .79, .90]; overall = .87 [95% CI .86, .88]. Corrected item-total correlations were good across all samples (range: .59 to .83).

Validity evidence based on relationships with other variables

The SWLS showed correlations in theoretically coherent directions with all assessed constructs. Positive correlations with Valued Living-Progress: r = .49 (Sample 1), .64 (Sample 2), .56 (Sample 3), all p < .001. Negative correlations with Valued Living-Obstruction: r = -.27 (Sample 1), -.50 (Sample 2), -.47 (Sample 3), all p < .001.

Negative correlations with DASS-21 Depression: r = -.41, -.57, -.52 for Samples 1-3 respectively. With DASS-21 Anxiety: r = -.27, -.40, -.45. With DASS-21 Stress: r = -.26, -.44, -.49. With GHQ-12: r = -.43 (Sample 1), -.37 (Sample 3). With ATQ-8 (negative thoughts): r = -.44, -.60, -.54. With AAQ-II (experiential avoidance): r = -.42, -.57, -.57. With CFQ (cognitive fusion): r = -.36, -.52, -.51. All correlations were significant at p < .001.

Regarding criterion validity, Sample 3 (clinical) showed significantly lower scores than nonclinical samples (Samples 1 and 2). Comparison with Sample 1: t(855) = 5.77, p < .001, d = .66 [95% CI .45, .87]. Comparison with Sample 2: t(823) = 3.24, p = .001, d = .36 [95% CI .15, .56]. Mean scores were: Sample 1 (M = 24.87, SD = 6.93); Sample 2 (M = 22.76, SD = 6.95); Sample 3 (M = 20.26, SD = 7.62).

No significant gender differences were found in Sample 1 [t(752) = .82, p = .41] or Sample 3 [t(99) = 1.11, p = .27]. In Sample 2, a small difference favoring women emerged [men: M = 21.84, SD = 6.85; women: M = 23.09, SD = 6.93; t(708) = -2.07, p = .039, d = 0.18].

Measurement invariance

Metric and scalar invariance analyses were successful at all comparison levels. Across the three Colombian samples, scalar invariance was demonstrated: ΔRMSEA = -.002, ΔCFI = -.003, ΔNNFI = -.001, all < .01. Between men and women in Colombia, scalar invariance was also confirmed: ΔRMSEA = .0052, ΔCFI = .000, ΔNNFI = .002. Between Colombia and Spain, scalar invariance was observed: ΔRMSEA = .0045, ΔCFI = -.001, ΔNNFI = .001. These results indicate that the SWLS measures the same construct in the same manner across all these groups.

Discussion and conclusions

The SWLS demonstrated excellent psychometric properties in Colombia. Results replicated the original unidimensional model, showed excellent internal consistency, and evidenced solid convergent and criterion validity. The most important finding was the factorial equivalence demonstrated between Colombian and Spanish samples at the scalar invariance level, meaning that scores can be compared between these countries without measurement bias. This result is particularly noteworthy because, according to the authors, previous research has found scalar invariance in only one of eleven published studies on SWLS invariance across cultures.

The authors acknowledge several limitations. First, no systematic diagnostic information was collected for clinical participants. Second, aspects of validity such as sensitivity to therapeutic change were not analyzed. Third, the proportion of women was significantly higher than men across all samples, although this limitation was reduced through gender invariance analysis. Fourth, the mean age of Sample 2 was relatively low, possibly due to higher social media use among young people in Colombia.

Discussion and conclusions

The Spanish version of the SWLS translated by Atienza et al. (2000) showed robust psychometric properties in Colombia and factorial equivalence with Spanish samples. The most important finding was the demonstrated factorial equivalence between Colombian and Spanish samples at the level of scalar invariance, meaning that scores can be compared between these countries without measurement bias. The authors acknowledge several limitations: no systematic diagnostic information was collected for clinical participants; aspects of validity such as sensitivity to therapeutic change were not analyzed; the proportion of women was significantly higher than men; and the mean age of one sample was relatively low.

Significance and contribution

This study contributes to cross-cultural validation of the SWLS by demonstrating robust psychometric properties in Colombia and establishing transnational factorial equivalence with Spanish samples. By demonstrating scalar invariance across countries, the work facilitates valid comparison of life satisfaction across Spanish-speaking contexts and provides solid psychometric guarantees for international comparative studies. Expected correlations with measures of psychological inflexibility confirm the theoretical relevance of the instrument for assessing well-being and clinical outcomes.