"My general inclination is to think that "applying" general social theories to specific social circumstances is not a valid way of creating new knowledge or understanding. This is because I believe that social ensembles reflect an enormous degree of plasticity and contingency; so general theories only "fit" them in the most impressionistic and non-explanatory way. We may have a pure structural theory of feudalism; but it is only the beginning of a genuinely knowledge-producing analysis of fourteenth-century French politics and economy or the Japanese samurai polity. At best the theory highlights certain issues as being salient -- the conditions of bonded labor, the nature of military dependency between lord and vassal. But the theory of feudalism does not permit us to "derive" particular features or institutions of French or Japanese society. "Feudalism" is an ideal type, a heuristic beginning for social analysis, rather than a general deductive and comprehensive theory of all feudal societies. And we certainly shouldn't expect that a general social theory will provide the template for understanding all of the empirical characteristics of a given instance of that theorized object...
allow Marxism, or Weber or Durkheim or Tilly, to function as a suggestive program of research for empirical investigation. Let it be a source of hypotheses, hunches, and avenues of inquiry. But be prepared as well for the discovery of surprising outcomes, and don't look at the theory as a prescription for the unfolding of the social reality. Most importantly, don't look to theory as a deductive basis for explaining and predicting social phenomena."
Daniel Little's post should be a must read for all social scientists who often elevate theories to a pedestal and engineer social explanations and predictions to fit the underlying theories. No where is this more ubiquituous than in the field of economics.
Social science theories should function as enabling methods of research rather than act as all encompassing and comprehensive theories. They should be employed to help find explanations for past events and help predict future ones, instead of becoming blanket explanations and predictions themselves.