This report profiles trends in access and usage of U.S. telephones, computers, and the Internet. NTIA's first report, Falling Through the Net: A Survey of the "Have Nots" in Rural and Urban America (July 1995), was the first survey of its kind regarding household computer and modem ownership by degree of urbanization. NTIA's second report, Falling Through the Net II: New Data on the Digital Divide (July 1998), presented updated data regarding household telephone and computer ownership, but focused on household on-line access instead of modems. This third survey, Falling Through The Net: Defining the Digital Divide, further defines the digital divide, and provides new information on Internet access and usage.
As in our previous reports, we utilize data from the U.S. Department of Commerce Census Bureau. NTIA contracted with the Census Bureau to add questions to its December 1998 "Current Population Survey" ("CPS") on household penetration, specifically to formulate a Computer and Internet Use Supplement survey. This survey asked additional questions regarding points of Internet access, methods of access, types of use, and reasons for discontinuing use, among other topics. All respondents were at least fifteen years old, knowledgeable about the Internet or computers, and gave proxy responses for other members of the household.
The Census Bureau obtained data on these surveys by interviewing 48,000 sample households. The CPS and Computer and Internet Use Supplement samples were selected from the 1990 Decennial Census files with coverage in all fifty states and the District of Columbia. The sample is continually updated to account for new residential construction. The Census Bureau divided the United States into 2,007 geographic areas, each typically comprised of a county or several contiguous counties. It selected a total of 754 geographic areas for the 1998 CPS survey.
As in 1994 and 1997, the Census Bureau cross-tabulated the information gathered according to specific variables, such as income, race, education level, household type, and age as well as by geographic categories, such as rural, urban, and central city, plus state and region. NTIA adopted these categorizations in presenting information in the attached charts. The Census Bureau determined that some of the data were statistically insignificant for any meaningful analysis because they were derived from small samples. We have noted this, where appropriate, in the charts.
All statistics are subject to sampling error, as well as non-sampling error such as survey design flaws, respondent classification and reporting errors, data processing mistakes and undercoverage. The Census Bureau has taken steps to minimize errors in the form of quality control and edit procedures to reduce errors made by respondents, coders, and interviewers. Ratio estimation to independent age-race-sex-Hispanic population controls partially corrects for bias attributable to survey undercoverage. However, biases exist in the estimates when missed people have characteristics different from those of interviewed people in the same age-race-sex-Hispanic group.
NTIA used Census data to create its own cross-tabulation references throughout the report. We also conducted a logistic regression to evaluate how race, income, degree of urbanization, education, and access to a computer at home, influence Internet usage through a library or community center.
The variables are binary and defined as follows: PUBACCESS is Internet usage at a public library or community center; LOWINCOME is a household with annual income less than $20,000; BLACK is Black non-Hispanic; HISPANIC is Hispanic, any race; MINORITIES is minority, non-Black, non-Hispanic; NOSUBURB is central city or non-metropolitan (rural); NOCOMPUTER is no computer in the household; NOCOLLGRAD is no four-year college degree. The logistic regression analyzes how changes in the above variables affect the probability of a person utilizing a library or community center for Internet access. Because of the binary nature of PUBACCESS, (i.e., either individuals get access to the Internet from a public library or community center or they do not), the logistic regression technique is well suited for this study of the Census data.
|<-- Executive Summary|