Paul Krugman asks:
What I’d like to have is a Guttman scale of positions on political matters, such that almost everyone who gave the “liberal” answer to question 7 also gave liberal answers to questions 1-6, while almost everyone who gave the conservative answer to question 7 also gave conservative answers to questions 8-13. And we’d want population shares associated with each point on the scale. So we could then take known positions of public figures and place them on the scale: say, we might find that only 19 percent of Americans are to the right of Michelle Malkin, while 23 percent are to the left of Michael Moore.The problem is, it's been known since Converse's paper that no such scale can be constructed for the general public. It can only be constructed in parts of the citizenry who are politically aware, and Converse found that that was a minority, 11.5% of subjects, 15.5% of voters. The largest plurality of people, 42% of subjects, 45% of voters, in Converse's five-ranked classification system took positions based on understandings of group allegiances. So, for instance, Converse found a "socialist" who supported privatization of utilities! For the rest, people would answer poll questions, but their answers didn't correlate, so you'd get, for instance, people who (post-Converse example) support Medicare but are opposed to government-financed health care. The classic and horrible example, of course, is what we are seeing writ large in California, where people have voted for government programs and against the taxes that fund them. Many votes and positions are apparently made and taken at random, and change when the voters are consulted some time later.
There's more to be said about this: the intellectual response, the implications for democratic governance, and so on. I continue to be astonished that these nearly 50-year old results have drawn almost no attention outside of political science.