Constructing indicators of unobservable variables from parallel measurements

The social and economic research often focuses on the construction of composite indicators for unobservable (or latent) variables using data from a questionnaire with Likert-type scales. Within the variety of procedures, we focus on the data analysis technique of Principal Components Analysis, in its Linear and NonLinear versions. This paper shows that when the variables are parallel measurements of the same latent unobservable variable, Linear and NonLinear Principal Components Analyses practically lead to the same composite indicators.