It really depends on what model of rationality you use. I suspect (but correct me if I am wrong) that most of the poli-sci and econ folk assumes some variant of the rat-choice or "means optimization" model, which itself makes certain dubious imo assumptions about the nature of information being used to optimize the means. That assumption is "omniscient rationality" i.e. that people have all or at least sufficient information that is needed to make the optimal means/ends choice.
That assumption has been questioned by the bounded rationality crowd (Cyert, March and fellow travelers) who, I understand, gained some limited traction in sociology and political science but have not entered the mainstream or, for that matter, the conventional leftist critique of the mainstream, that still relies on the rat choice model.
What the bounded rationality crowd claims is that people usually do not have
sufficient information to optimize means toward the end, and therefore rely
on socially constructed "shortcuts" or "crutches" i.e. conventionally
accepted beliefs, expectations, and rules of behavior aka institutions.
>From that perspective, people are rational in terms of their means-ends
orientation, the irrationality comes from imperfect information and
institutional inertia.
My own approach (self-promotion: see Ch. 6 of my book _Civil Society and the Professions in Eastern Europe_) to the issue involves adding a "meta" level to the decision making process - the "base" level involves selection the best means to a specific end - just like in the conventional rat-choice model - whereas the "meta" level involves selection of information deemed most relevant for making the decision on the "base" level (i.e. what is the optimal means to a given end). The decision on the "meta" level depend on cognitive framing - which in cognitive science is a device that determines the relevance of any information - which in turn is affected by social/institutional context. Several different frames can be used at the meta level at the same time, i.e. one frame to evaluate potential costs, another frame to evaluate potential benefits, etc. Depends on what information is allowed or filtered out by these different cognitive frames - costs can be cognitively minimized while benefits maximized, or vice versa. The text introduces a heuristic device to classify the effects of these cognitive frames along two dimensions: social scope and social time i.e. whose interests are relevant and in what time span). The frames, in turn, are adopted by individuals the way institutionalists claim - by rules, norms, and expectations embedded in social institutions.
This model, I believe, reconciles the apparent contradiction between the assumption that every human actor is essentially a rational actor (save for cases of mental illness) and the seemingly irrational outcomes of their actions (e.g. voting for a candidate whose policies are detrimental to the voter's economic interests). The apparent contradiction results from the fact that the observer uses a different set of cognitive frames to evaluate the voter's actions than the voter does herself.
Take for example the discussion about the "middle class" on this list. Some observers adopt a cognitive frame that amplifies the salience of certain economic interests (such as ownership, or take-home income) while minimize the salience of other factors (such as culture, ethnicity, life-style, education, etc.). The voters, otoh, tend to adopt a cognitive frame that amplifies the salience of the latter while minimizing that of the former. The crux of the matter is that BOTH views are rational in the conventional sense (i.e. optimization of means to achieve an end), but they operate under different rationality "systems" or cognitive frames that regulate information flow.
George Lakoff (_Moral Politics_, _Don't think of an elephant_) argues along similar lines to explain, inter alia, apparent contradictions in political positions (e.g. opposition to abortion and support for death penalty).
Interestingly, this cognitive bent attracts little attention among both mainstream pundits and its critics from the left. For the mainstream pundits it comes as no surprise, for their goal is to present the stats quo as the best of all possible words, and any perspective that undermines the notion "best of all possible worlds" simply destroys their business. But the left critics?
Wojtek