4.3. Estimation approachTo examine the latent variables within their causal structure, we applied structural equation modeling (SEM) using EQS 6.1 software (Bentler 1995) and the Maximum Likelihood (ML) procedure.We chose an SEM approach as it is a powerful generalization of earlier statistical approaches with the key benefit being that each explanatory and dependent variable is associated with measurement error in contrast to OLS regression, for example, that is based on the assumption that variables are measured perfectly (Bollen 1989). In addition, SEM allows for multiple indicators of latentvariables which are a more realistic representation of the variables under study.