An Exploratory and Confirmatory Factor Analysis of the Commitment to School Scale for Middle School-Aged Youth in Ghana
Youth’s commitment to their academic work may be fundamental to their academic success. Jenkins (1995) finds that students’ committed to school are less likely to engage in school crime, misconduct and nonattendance. The extent to which school commitment impacts performance is not well known partly because of lack of standardized measures of student commitment. This study assesses the validity and reliability of a US-based Commitment to School Scale (CSS) adapted to the Ghanaian context. The CSS is a 10 Likert-type items developed for the Rochester Youth Development Study to measure youth’s school commitment (Dahlberg, Toal, Swahn, and Behrens, 2005). The scale was adapted to the Ghanaian context after rounds of expert review and rigorous pilot testing. Changes made to the original CSS to ensure good fit and adherence to measurement best-practices included: expansion of the response scale from 4- to 11-point scale; reduction of scale items from 10 to nine; restructuring of items from second-person to first-person statements, and reverse-wording of some items.
Methods:
Data for this study came from a random sample of 6252 sixth to eighth graders in Ghana. These youth are part of an ongoing experimental savings project in Ghana called YouthSave. LISREL software (version 8.80) and the Maximum Likelihood estimation method were used for all analyses. First, an exploratory factor analysis (EFA) was conducted to examine the underlying factor structure of the scale. The EFA yielded a 9-item one-factor solution. Confirmatory factor analysis (CFA) was also used to validate the one-factor model. Internal consistency of the scale was examined and a multi-sample analysis was conducted to assess group invariance by gender and geographical region.
Results:
The initial model fit was poor (χ2=1596, p<.0001, 27 df; RMSEA=.096 [90% CI .092 – .01]; CFI=.93). The model was refitted by dropping two indicators that had low factor loadings and reliability scores. Error covariances were added among five indicators. Although the χ2 of the refitted 7-item model was statistically significant, it had a reasonably good fit: (χ2=98.08, p<.0001, df=10; RMSEA=.038 [90% CI .031 – .044]; CFI=.99, Cronbach’s alpha=.754). The model produced statistically significant (p<.001) moderate factor loadings. Imposition of equality constraints did not result in significant difference between genders or among the eight geographic regions, and thus there is measurement invariance by gender and geographic region. Most item-to-total correlation coefficients met the recommended criteria (r = 0.30 to r = 0.70). Items retained are: I like school; I try hard at school; I finish my homework; I do extra work to improve my grades; getting good grades is important to me; homework is a waste of time, and I belong at school.
Conclusion:
Given its short length and good fit, the adapted CSS might be useful for research on school commitment among Ghanaian youth. Also, similarity of the model fit across the diverse geographic regions of Ghana suggests that the scale has potential uses in education research in diverse culture. Further development of the CSS is needed to ensure accurate measurement and better understanding of youth’s school commitment