Published online by Cambridge University Press: 04 January 2016
In a recent article in Social Science History, Flanigan and Zingale reviewed the old problem of inferring individual relationships from aggregate data, cast doubt on Goodman’s ecological regression technique as a method of estimating such relationships, and contended that the specification analysis approach to the issue was misleading. Instead, they argued that an ad hoc procedure suggested by Shively in a 1974 article was superior to Goodman’s point estimates because it avoids dubious assumptions and forces investigators to make their premises clear. Comparing results from analyses of state-level data on major party voting in the 1968 and 1972 presidential elections with figures from the Michigan surveys, they concluded, to their satisfaction at least, that the assumptions required for ecological regression were badly violated and that it was preferable in this and other cases to use the system that Shively, without making any grandiose claims for it, had originated (Flanigan and Zingale, 1985; Goodman, 1959; Hanushek et al., 1974; Shively, 1974; Langbein and Lichtman, 1978).