Latent variable scores: An option for customer satisfaction index of Brazilian Electricity Regulatory Agency

Structural Equation Models with unobservable variables and measurement error have been used in the production of customers satisfaction indexes to evaluate products and services quality as well as the economic performance of companies, sectors and nations. Previous studies pointed out the PLS (Partial Least Square) better than Maximum Likelihood method to estimate parameters in marketing applications, mainly because of data marketing feature. However, no study has shown the effects of estimation methods in the latent variable scores, especially in the satisfaction scores. The objective of this study is to analyze the effects of estimation methods in the satisfaction scores based on the IASC model (Customer Satisfaction Index of Brazilian Electricity Regulatory Agency). The data and true scores will be generated by Monte Carlo Simulation and the estimated scores will be compared to the true ones through the following information measures: linear correlation, mutual information and an empirical information measure.