It should be happening soon. From Social Science Statistics Blog I get this information about a project aiming to give access to social science research data through virtual data archives (which they call "dataverses"). This meta-repository comes from Harvard University, and IMHO it is a great idea; not just will give access to data before it was only available to very few researchers, or getting it was almost impossible to amateur researchers (like me), but will increase the visibility of the "owners" of the data since the system establishes a quote-like system.
This is the next step to the simple data repositories some academic journals were putting in place from years ago in order to allow for the reproducibility and falsability of the empirical papers they have been publishing, and the broad network of Observatories being created all over the place for generating and distributing data related to the foci of the organisation.
Other concerns that I have not taken a look at yet are about the licensing of the data, Data Protection Act-related mattters, ways of amend errors in the data…
Anyway, another dimension added to the multiverse.