I've seen, in the past several months, a number of proposals that essentially would destroy American federalism if implemented. There seems to be an increasing boldness in insisting that everyone needs to have the same laws under one and only one government. This is, of course, a step toward totalitarianism, in the sense of removing one of the major structural impediments to it; precisely one of the consistent advantages of federalism is that it makes it very difficult to engage in total capture of the forces of the state. If the federal government gets too pushy, there is some leverage in the states; if the states get too pushy, there is some leverage in the federal government. And in practice I think even people making such proposals in fact regularly take advantage of this when they can. But in politics people are often in the grip of an abstract theory, and this leads to being attracted to speculations about what would happen under ideal circumstances rather than about what will actually deal with the real problems that must be faced in government. And under ideal circumstances, of course, many things that are wholly dangerous can seem quite nice; totalitarianism itself can seem initially attractive if you attribute to it the qualities it would have if it were implemented by virtuous angels with extraordinary skill. It is foolish; but it is perhaps unsurprising that people are occasionally dazzled by the brightness of their own abstract schemes.
However, here I would not like to rehearse the ordinary arguments for federalism -- that it is the organic structure of the United States, that it is part of the tradition carried forward from the Founders, that it is a bulwark against fascism, that it creates the sort of perpetual political laboratory required for genuine progress in the long run, etc. -- and suggest one that I have not seen people give but which I think is in fact even more important than the others. And this argument might be summarized in the following way: No single method of representation is adequate to representing people; to do so with perfect adequacy would indeed take an unimaginable variety of methods of representation, too many perhaps for us to handle in practice; but in practice, being as adequate as we can will still always require multiple methods of representation. Call this feature 'superrepresentationality', exceeding any given scheme for representing them, and we can express this argument even more succinctly by saying, In any society in which the state is good enough at representing people, the People will be recognized as superrepresentational. This is not a sufficient condition, of course, but it is a necessary one.
When we have people, they only become a People in politics by political representation in one form or another. A People represented only one way -- say, by a national popular vote -- is a flat and one-dimensional People, and participating in it can only be valuable for the good of the people contributing to it in a flat and one-dimensional way. Multiple methods of representation, however, add new dimensions to what it is to be a People, and thus become more useful for contributing to the good of the people participating in that People. How much more useful, of course, will depend on how well the interests of the people are actually represented. But representing the interests of the people well will require representing the people in multiple ways.
In the American scheme, of course, we have a primary kind of representation and a secondary kind; the secondary kind would be things like city councils or county sheriffs, which serve a genuine representational function, but are also entirely subordinate to some other kind of representation. Primary kinds of representation in the United States are state legislatures and governors and at the federal level, Congress, and the President. Because of this, 'We the People' are not governed wholly from D.C., but have distinguishable modes by which our interests show up in representation: as individuals forming the People of a state, as individuals forming the People of the United States, as Peoples of the states forming the People of the United States. These are not reducible to each other; our interests as Peoples of states, for instance, don't show up at all if you only considering us as individuals. Simply considered as individuals, people are largely interchangeable; as Peoples of states, they are very obviously not, and woe to Montanans if they are governed according to what benefits Californians.
Of course, nothing about this tells us how this superrepresentationality is best structured. The American system, of course, has undergone several changes, and will no doubt undergo more for as long as it lasts. Perhaps we have not yet found the most workable form of federalism, and perhaps there are ways of having superrepresentationality compared to which federalism is quite crude and primitive. But any government that does not incorporate it is inevitably a government inadequate for governing real people in the real world, and federalism is the way we currently know how to do.