Karen Mossberger et al's book on the digital divide in the United States is an excellent starting point. In a discussion the other day one of my students raised the issue of the racial divide, pointing out that, after controlling for other variables, there were few good hypotheses for why the divide might exist. One of Mossberger et al's findings was that even after controlling for educational attainment and income there was still a small but significant digital divide based on race. They found that African-Americans and Latinos were less likely overall to have technology skills, even though they had a much more optimistic outlook on the difference the Internet could make to their social and political influence.

The authors revisited this conundrum in a useful paper (pdf) published by the Center for Digital Government (formerly at Harvard, now at the University of Massachussetts, Amherst). The paper argues that environmental and neighborhood constraints go some way toward explaining the divide.