Abstract

My favorite opponent in this debate once made a remarkable concession, not that it interfered with business as usual:No sensible social scientist believes any particular specification, coefficient estimate, or standard error. Social science theories ... imply that specifications and parameters constant over situations do not exist ... One searches for qualitative theory ... not for quantitative specifications Achen (1987, p.149). . With Hooke's law and the like, we are estimating parameters in specifications that are constant across time—at least to a very good degree of approximation But see Cartwright (1983). . What are the social scientists doing when they estimate non-existent parameters, and put standard errors on the output? How can that help them search for qualitative theory? Those are among the first of my questions, and I never get answers