Lamentable Oversimplification

To a trained economist, the basic Ricardian model seems almost trivial. Two goods, two countries, one productive factor, perfect competition: what could be simpler? — Paul Krugman, from here

It's simple, all right, but the simplicity is also an error.

Uncertain social status makes primates panic. People are primates. The effort, including the hugely "no, that's not really in your best interest" effort, that goes into making social status stable is simple, obvious, and wrong from an economic point of view. So there's no such thing as perfect competition because humans are ground-dwelling plains apes that gang up in problems and apportion resources on the basis of social status and will do so, even when they are fully aware that economically, this is insane.

To a very large (I would argue completely dominant) extent, government exists to implement a choice about what kinds of economic competition will be permitted.

So we have a sort of general case where you aren't allowed to compete by the use of military force; that took a long time to establish (post-Rome, in Europe, to the banning of mercenaries in the Treaty of Vienna, is something like 1400 years. The Dark Ages can be defined as the period where "go conquer some place and extract taxes/tribute from it" was your best economic bet in purely rational terms, and so on.)

We also have a widespread general case that you're not allowed to unseat an incumbent industry, no matter how inefficient, corrupt, polluting, or just surpassed that incumbent might be, because the value of the incumbent to the existing social hierarchy is great and the value of the innovative efforts that might unseat that incumbent are negative. This is the disease of empires; trade away economically rational behaviour (or, rather, laws and customs that encourage pseudo-rational conduct in most people, most of the time) for a predictable—NOT stable; one of the predictable things about it is that it will eventually collapse—social order.

So we get things like the US bank bail out, where members of a ruling class agree that other members of the ruling class belong in that ruling class and will have the social status, irrespective of how much diversion of public funds to maintaining the personal wealth that guarantees that status is required; the various government bailouts of auto workers in the US and Canada, where some simulation of a good job–with its attached social status–is being preserved despite terrible price/benefit, no obvious public value to (and considerable obvious harm from...) the business, and obvious areas where government funding in transportation would have much, much greater long term economic and social value; or Atrios' cogent complaint that most of the US, through various mostly municipal laws, forbids high density residential development, so that there are entire classes of economic activity that just don't happen because there's been a political decision that they shall not.

So, sure; if everything was being run by a bunch of Culture Minds as a bizarre bet-settling experiment, Ricardian model comparative advantage would be a fine thing.

Everything is not run by a bunch of Culture Minds; everything is run by people who, to a first approximation, insist that they never, ever, lose any social status. By no means can they always pull that off, but they're always, always trying.

Because of that insistence, we do not and will not get smooth movement from local maximum to local maximum (as new and better local maxima come into reach due to accumulating capital, new technology, etc.); we get jerky, painful, abrupt transitions as existing local maxima get crushed out of existence, and the people who used to do that actual work are thrown out of work in an environment that makes it really, really tough to get equivalently good work again as an inherent, structural part of a "the economy is for maintaining and displaying social status" view. (Which is worse now than it used to be, at least in Anglo NorAm; insecurity drives out charity, hope, and risk-taking.)

As information management ability increases with technological improvement, this tendency, which is mostly driven by illusions of control, trends worse, not better.

A really good government would be set up to arrange smooth transition from local maxima to higher local maxima; that would require forbidding economically dominant industries, making money effectively useless to politics, and probably a couple of other things that would take those Culture Minds to actually implement. But we haven't got that. What we've got a diversity of ruling classes more or less determined to reintroduce peonage in one form or another because that it what makes their percieved position most secure, even if this makes no economic sense whatsoever.