A multiple indicator model for panel data: An application to ICT area-level variation

• We present a multiple indicators dynamic panel data model of ICT activity.
• The model is estimated using sub-regional data from a Spanish ICT survey.
• We find that the level of ICT activity in each area is highly stable over time.
• Less dynamic and more rural areas lag behind in terms of ICT development.

Consider the case in which we have data from repeated surveys covering several geographic areas, and our goal is to characterize these areas on a latent trait that underlies multiple indicators. This characterization occurs, for example, in surveys of information and communication technologies (ICT) conducted by statistical agencies, the objective of which is to assess the level of ICT in each area and its variation over time. It is often of interest to evaluate the impact of area-specific covariates on the ICT level of the area. This paper develops a methodology based on structural equations models (SEMs) that allows not only the ability to estimate the level of the latent trait in each of the areas (building an ICT index) but also to assess the variation of this index in time, as well as its association with the area-specific covariates. The methodology is illustrated using the ICT annual survey data collected in the Spanish region of Catalonia for the years 2008–2011.