Bikson 1999). A few academic studies or market surveys have produced different results, but
Inequities based on gender have diminished over the years (US Department of Commerce
2002), and some predictions have been made that racial and ethnic gaps are currently
insignificant or will soon disappear of their own accord. The “strong version” of this scenario is
that all differences between groups, including those based on income and education, are being
erased by the rapid diffusion of the Internet and computers throughout society (Compaine 2001;
US Department of Commerce 2002). To understand why this is not likely requires a closer
examination of what digital inequalities entail, at least among some disadvantaged groups.
Policy debate and research have often shared an overly-narrow definition of the divide as
an issue of access alone. Access, however, is insufficient if individuals lack the skills needed to
use technology. Technical skill, or the ability to use computer hardware and software, is only
one dimension of the skills needed to use computers. With the advent of the Internet, technology
use requires reading comprehension and the ability to search for, use, and evaluate information.
Evidence indicates that this is a more challenging threshold for technology use. Twenty percent
of Americans report needing help using a mouse or keyboard, but 37 percent say they need help
navigating the Internet (Mossberger, Tolbert, and Stansbury 2003, 45). Segments of the
population that have limited basic literacy and little education will not likely develop the more
sophisticated skills required for effective use of the Internet. According to the National Adult
Literacy Survey conducted in 1992, between 21 and 23 percent of the population operates at the
lowest level of literacy, unable to perform more than the most rudimentary tasks (Kaestle et al.
1
Some market research has found that Latinos have higher rates of access than whites (Walsh 2001). This market
survey has been quoted by academic sources (see Compaine 2001, Chapter 14), but it was based on a mail survey,
for which the response rate was not disclosed. Nie and Erbring (2000) and Wilhelm (2000) dismiss the influence of
race, but Nie and Erbring do not use multivariate statistical controls, and Wilhelm’s findings on race and ethnicity
are suspect because of the way in which he analyzed the statistical data. Wilhelm included two dummy variables for
whites in his analysis, one variable for race, and one for ethnicity. This created a situation of near perfect
multicollinearity. He also used the residual category “other race” as the left-out group in his analysis, again
fostering multicollinearity, because of the small number of individuals in that category. As a result, his analysis
obscures the real impact of race and ethnicity.
6
Convention