Experimental evidence shows that hierarchical organization is more inefficient than generally realized.
Ernst Fehr and colleagues got subjects to play an authority-delegation game, in which subjects were divided into principals and agents, and then asked to work on selecting projects with varying payoffs. They made two important discoveries.
First, subordinates put in less effort than you'd expect rational income maximizers to; depending on the treatment, up to half put in no effort at all, even though this was almost never the income-maximizing option.
One reason for this, says Fehr, might lie in regret aversion. People have an aversion to being treated unfairly - which is why they reject unfair but wealth-enhancing offers in ultimatum games - and the fear of not getting a fair reward for their effort makes them loath to work.
This suggests that the trade-off between the allocation of control rights and provision on incentives is greater than conventional theory (pdf) predicts.
Secondly, Fehr and colleagues say:
We find a strong behavioral bias among principals to retain authority against their pecuniary interests and often to the disadvantage of both the principal and the agent.
Some two-fifths of principals did not delegate even when income-maximization required it. This suggests that people get a non-pecuniary buzz from being in control, and seek this benefit at the cost of economic payoffs to themselves and others. This is consistent with the findings of other experiments by Fehr and colleagues, which suggests that hierarchy facilitates exploitation rather than pure economic efficiency.
All this represents a challenge to conventional transactions costs economics, as developed by the likes of Oliver Williamson, which predicts that hierarchy is efficient (pdf) in many plausible conditions. And this poses a challenge to defenders of hierarchical capitalism: do real-world mechanisms such as competition between firms strongly select against the adverse effects of hierarchy which we see in experiments? My suspicion is: perhaps not.