Does Context Matter? Factorial And Measurement (Non-) Invariance of McGee et al.’s Entrepreneurial Self-Efficacy Scale - In age and Culturally Diverse Cohorts

Understanding, predicting and developing entrepreneurial behaviour in demographically diverse cohorts is important to society as entrepreneurship in its many shapes and forms, builds and sustains economies. Educators and car er counsellors, working with diverse groups, face challenges in how best to measure change in learners’ understandings, capabilities and confidence (self-efficacy) in entrepreneurial activities. The McGee et al (2009) nineteen item entrepreneurial self-efficacy scale offers a multidimensional instrument for measuring ESE based on task-domains, but in the current ad-hoc approach using confirmatory factor analysis, it is appearing in shortened forms in the literature. The primary aim of the study was to compare and to test for factorial structure and invariance of the ESE scale as a measuring instrument in a populations of diverse ages and cultures. In doing so, outcomes were realised that were and weren’t expected. We show that when using the ESEM model estimation and the MLR estimator, that the five dimensions of the ESE scale (McGee et al, 2009) are robust constructs without the need for co-varied error terms. The five-factor ESEM model of the ESE scale (McGee et al, 2009) was the model of best fit though the three factor ESEM model, ICM-CFA five and three-factor models also had well-defined and reliable factors. In our final model, five dimensions of ESE emerged aligned with the original structure of the scale (1) searching, (2) planning, (3) marshalling, (4) implementing-people and (5) implementing-finance. Additionally, we demonstrate application of the alignment method to identify non-invariant items and to avoid the step-wise and often cumbersome traditional approach. We find eleven fully invariant items for our study population and these exhibit more than 50% overlap with items in the shortened scales (McGee and Peterson, 2017; Caines, 2017; Douglas and Prentice, 2019) (see Table 4). Of particular note, in the study of Douglas & Prentice (2019), no items were retained from the ‘marshalling’ dimension. We wonder if less restrictive confirmatory factor analysis approaches were used such as ESEM, more items would have been retained in their models and the scale retain its ability to act multi-dimensionally. We wonder if use of ICM-CFA may have contributed to the shortened versions of the ESE scale (McGee et al, 2009) used in recent studies (McGee and Peterson, 2017; Caines, 2017; Douglas and Prentice, 2019). Overall, construct validity and partial scalar invariance was established for the ESE scale (McGee et al, 2009) in our age and culture diverse study population. Additionally, our finding of partial scalar invariance results strengthens and extends the validation of the scale in the context of non-nascent entrepreneurs, older people and the PNG context. Scalar MI indicates very similar understanding, conceptualisation and use of the scale across groups. The items with full scalar MI also supported face-validity of the main dimensions (constructs). These particular invariant items could be used as anchor items in further multi-group CFA.