I agree with Harold Feld that the premise of Michael Palage's work is flawed. Michael is talking about the "representativeness" of particular groups. But it doesn't really make sense to try to make "representativeness" a qualification for the constituencies. The world of "businesses" and "users" is a big one, and there's no chance that anyone (other than the sets of registries/registrars themselves) could claim to be numerically "representative" of their group.
This is yet another argument for the appropriateness of the consensus idea -- ICANN's board can't claim to have the power to make rules itself (it isn't the FCC), and ICANN can't be a one-person-one-vote "democracy." So the consensus ideal offers a way to find out how stakeholders feel about proposed policies: listen to whoever shows up. If there's an affected group that's troubled about a proposed policy, it can "show up" and engage in logrolling and discussion, and the consensus notion says that group has to be listened to because it's likely that it is expressing widely-held views. Without a consensus system in place, this incentive to "show up" doesn't exist, because whoever shows up will be dismissed as not being representative enough.
Consensus reflects democratic values (like making sure people's interests are taken into account in decision-making, and respecting local rules) without claiming to be statistically "representative." Consensus is a way to bring these values to bear in a context (a) in which we don't have national boundaries or censuses to define an electorate and (b) in which which there is no "sovereign" that has been given power to make rules.
Susan Crawford (speaking only on behalf of myself)