15

number of computer per 100 people has the most significant impact on the output

of internet population. Model 2 shows that adding of the number of schools in the

model increase the explanatory power. In the third model, predictors from all the

three factors are presented, and the Adjusted R square reaches 90.5%.

__________________________

Table 6 about here

__________________________

**3. Determinants of penetration rate **

Running factor analysis on all variables that have significant correlation

with Internet penetration rate (Jan., 2001), only one factor can be retrieved

(Table 7), which

means all these variables are highly correlated to each other.

In order to obtain a precise regression model and avoid multicollinearity

(Tabachnick & Fidell, 2001), statistical regression is used again is used again.

----------------------------------------------------------

Table 7 about here

---------------------------------------------------------

In four regression models with significant predictors, rate of population

with higher education has the highest explanatory power, with largest Beta value

in all models. Rate of home computer ownership, percentage of urban population

and income of urban residents also contribute to explain the variance in adoption