Up to this point we have been dealing in the world of raw preferences, the actual percentage responses reported by survey organizations in response to particularistic queries. But we think politics as practiced is much simpler than that, that people organize sets of beliefs and preferences into far simpler patterns of belief systems. And whatever the well-documented weaknesses and randomness of individual-level political opinions (see Converse 1964 for the still authoritative treatment), we expect to find power and simplicity when we look at the summary preferences of the whole electorate over time. We wish to find the fundamental dimensions of attitudes, track their change over time, and learn what they mean.

THE DIMENSIONS OF OPERATIONAL IDEOLOGY

Measuring policy preferences is messy. We have learned that question wording matters – and often matters a lot. And so careful analysts often describe public opinion narrowly, with respect to the exact words of the question that triggered the response. We write things like “In response to the question “Government should do more to … ” x percent of respondents agreed and y percent disagreed.” When we have written that seven thousand plus times, we will have described our database of measured preferences.

But no voter could comprehend that much specificity. And – important for democratic theory – no politician could respond to it. Because the specificity of measured opinion is overwhelming, when we do political commentary we err in the opposite direction, simplification, and usually just describe people as liberals, moderates, or conservatives.