Partial Least Squares Path Modelling Approach for Social Composite Indicators Using Different Sources of Data

Many social phenomena are complex and therefore difficult to measure and to evaluate. Complexity implies also multidimensionality and high level of abstraction.

Several social concepts represent multidimensional concepts that are difficult and complex to define. For this reason, researchers in many fields of social sciences, have been focusing on the development and use of a “composite indicators" in order to obtain a global description of the various faces of a complex phenomenon, and to convey a suitable synthesis of information.

The existing literature offers different alternative methods in order to obtain a composite indicator.

Structural Equation Modeling (SEM), and specifically the Partial Least Squares approach to SEM (PLS Path Modeling, PLS-PM) can be used to compute a system of Composite Indicators.

Empirical case on Italian Social Cohesion was analyzed with the aim to research alternative sources to compute the Italian Social Cohesion, using PLS-PM approach.

We apply a theoretical model firstly on data using European Value Study - Italy database (2008), and secondly on mixed data (official data, administrative data an networking data), using some indicators extracted from different sources (I.Stat and SocialCohesion.Stat warehouse). These warehouses have, as reference periods, different years (from 2011 to 2013).

Moreover, for the Latent Variable “Italian Sentiment”, social media data were used, specifically Twitter data.

Finally results are compared. The first important result is the confirmation of the unidimensionality property, in both models, for each latent block. This result shows that the outer model is well specified and that the LVs are well measured by the Manifest Variables, being a good their synthesis.

This result suggests that it is possible to measure the social cohesion using heterogeneous sources of data.