Page 117 - Culture Technology Communication

102 Carleen F. Maitland and Johannes M. Bauer

INDEPENDENT VARIABLES
To explain and predict the diffusion of the Internet the independent
variables suggested above will be used. Each of these variables is rep-
resented by one or more measures. It is nearly impossible to find na-
tional-level indicators for all countries in the world. The result is that
one measure may be available for 85 countries while another is avail-
able for 110. It may also be the case that those 85 are not all included
in the 110 of the other variable. In general, the year of the measure
was chosen based on the number of countries reporting the data in
the 1990–1995 time period. To maximize the number of variables
available for the analysis it was necessary to take a cross-sectional
measure for each variable and these cross-sectional data were taken
from different years. For example, the income measure is from 1994
while the newspaper variable is from 1992. Although not ideal, the
impact on the analysis should be minimal as the measure reflects the
rank of country vis-à-vis the other countries and these rankings, in
national level macroeconomic and demographic variables over a short
12
time period, are fairly stable. Also, use of multiple measures for one
variable does not imply that scales will be formed. It merely adds
flexibility in choosing a set of independent variables that are (1) not
highly correlated and (2) represent the largest number of countries
possible.
The variables were gathered from a variety of sources. The
source, year of the data, number of observations, and expected cor-
relation with the dependent variable are presented in Table 2.

Multivariate Analyses

In the first part of this analysis the independent variables dis-
cussed above will be regressed onto the dependent variable
(START). The start variable was constructed by the number of time
periods a country has had an Internet connection between 1991 and
1997. For example, a country having adopted in January 1996 has
a START value of four, compared to twelve for a country having
adopted in January 1992. The data are available for 185 countries.
Thus, this is an analysis of the impact of independent variables,
which are assumed to be stable over this time, on the adoption tim-
ing of countries.
To begin the analysis, the correlation matrix of all independent
variables was examined. Special attention was paid to the number of
countries reporting data for each bivariate case. The correlations