ONE of the greatest frustrations for survey analysts with a historical bent is the fact that reliable surveys date back only to the mid-1930s. Our inability to survey a random sample of Americans from, say, 1776, 1861, or even 1932 deprives us of fascinating data about historical events and has led scholars to devise ingenious ways to try to reconstruct attitudes from a bygone era. One such approach is Kristi Andersen's attempt to create a data base of surveys that can tell us how the Republican electoral majority of the 1920s became the New Deal majority coalition of the 1930s (Andersen, 1976:74-95). Andersen's method is as simple as it is clever: she analyzed the University of Michigan's Survey Research Center surveys from 1952 on, concentrating on responses to questions about one's past partisanship. Each Democrat or Republican is asked by the SRC whether, and if so when, he or she identified with the other party. Each independent is asked whether, and if so when, he or she identified with one of the major parties. Andersen took those respondents old enough to vote in, say, 1920, used the responses to these questions to ascertain their party identification at that time, and by adjusting for the size of age groups tried to reconstruct the 1920 electorate's partisan composition. From her research she concluded that there was little conversion of Republicans to Democrats, and that