Small Area Estimation for Social Phenomena Using Multidimensional Latent Trait Models

This work addresses the estimation of indicators for monitoring the digital divide, a latent and multidimensional phenomenon. Using 31 binary indicators derived from the Aspects of Daily Life (AVQ – Aspetti della Vita Quotidiana) survey 2021, we adopt a multidimensional latent trait modeling approach to uncover the underlying dimensions of digital divide. To produce reliable estimates for unplanned domains, we integrate Small Area Estimation techniques with the latent trait modeling framework, allowing latent factors to vary with individual-level covariates and area-specific random effects. The proposed methodology enables accurate small area estimates of digital divide latent dimensions, addressing the small sample size limitations that prevent direct survey estimates from being reliable at the sub-regional level.