Aggregate Stability and Individual-Level Flux in Mass Belief Systems: The Level of Analysis Paradox

Converse's findings of low constraint and stability among mass attitudes are only one side of the story; mass attitudes often manifest much more coherent and stable patterns at the aggregate level than would seem possible if one took the results of panel survey analysis at face value. Items designed by Rokeach and Inglehart to tap basic value priorities, show modest individual-level stability, together with remarkably high aggregate stability structured in ways that could not occur if random answering were the prevailing pattern. Materialist/postmaterialist values show large differences between birth cohorts that not only persisted throughout 1970-1984 but seem to reflect distinctive formative experiences that occurred as much as 50 years ago. These aggregate results are much too skewed to result from equiprobable random answering and cannot be attributed to methods effects. They reflect underlying attitudinal predispositions in the respondents themselves. While random response to given items does play an important role, it is much less widespread than Converse's Black and White model implies and does not generally reflect an absence of relevant preferences. Structural equation analysis of multiple indicators demonstrates much stabler, broader orientations underlying the response to given items that account for the high aggregate-level stability observed here. Because it usually measures such orientations imperfectly, individual-level survey data tend systematically to underestimate constraint and stability in mass attitudes.