Conferences

Nonspecific cognitive capabilities of respondents who participate in mass surveys are commonly considered to influence their opinions and attitudes, separately from other variables such as socioeconomic position, or class or educational attainment, yet are difficult to measure.

Following a brief review of previous approaches to such measurement, employed in survey studies of psychosocial functioning in
relation to respondent’s location in social structure by Melvin Kohn and his associates, a new method of measurement will be outlined, based on selected Raven’s items related to progressive matrices, with emphasis on analysis of the resulting data from a national-level high quality Polish survey (PolPan) conducted in 2003.

The emphasis will be on predicting smartness in a flexible manner as a latent factor that
explains joint distribution of indicator responses via nonlinear dependencies fitted via kernel smoothing.

This modelling framework reveals the structure of dependencies of the responses to Raven’s items, enabling adjustements that arise in connection with possible nondistinguishability of categories, their nonmetrical character and differential functioning of the items. Robust prediction of smartness
will also be addressed, leading to categorical measurement on ordinal scale.
Efficaciousness of the proposed analysis of smartness based on Raven’s items will be evaluated by presenting selected results
concerning distribution of smartness across educational and income strata, or association of smartness with political opinions.

Keywords: Reven’s progressive matrices, smartness, psychosocial functioning, nonparametric latent factor model, mass surveys