Public opinion matters in a democracy because it is presumed to influence government. The level at which that influence is realized is the macro level, the polity. Thus work that aspires to connect public opinion to the myriad political activities beyond the public is aggregate. Theories that connect aggregate opinion to politics are macro theories.

My own work (Stimson 1991, 1998) and Erikson, MacKuen, and Stimson (2001) push public opinion to macro theoretic connections, a scholarly trend that can also be seen in works by Page and Shapiro (1992) and Mayer (1992). My own work on domestic policy mood (along with most of Page and Shapiro's efforts) is macro in another sense as well. It begins with fully aggregated data. It attempts universal coverage of the American survey research record of measured policy preferences, an attempt much too ambitious to permit analysis of individual-level data – even though all survey data are collected from individuals and a large proportion of the studies are available as individual-level data. The task of analyzing all those micro data is too daunting. Analyses therefore build on survey marginals, the national percentages choosing one or another survey response option.

Beginning with survey marginals is a wonderfully efficient noise reduction mechanism. The marginals retain the information one wants about movements of opinion between studies and over time but lose the vastly larger variance associated with between-individual differences. The price one pays is twofold. First, one loses the ability to address questions at other levels of aggregation.