Global internet participation

The question

Which conditions are most strongly associated with internet use?

Internet access is often discussed as an infrastructure problem: build networks, reduce costs, and participation should follow. I tested a broader proposition: that actual use depends on the conditions under which people encounter access. The analysis compared educational, economic, demographic, technological, and institutional explanations for cross-national differences in internet use.

Why it matters

Coverage and subscription counts tell us about infrastructure and connection; they do not tell us who actually uses it. If participation depends on more than infrastructure, policies that stop at provision may miss the educational and institutional conditions that help turn connection into something usable.

The approach

I assembled a cross-national dataset covering 104 countries and evaluated ten indicators across equality, education, economics, health, and governance. The outcome was the percentage of individuals using the internet, a closer measure of participation than infrastructure availability alone.

I estimated simple and multiple regression models, then used AIC and BIC model selection to compare competing explanations. I also checked model diagnostics and multicollinearity before interpreting the optimized model.

What I found

The education model explained roughly three quarters of the observed variation in internet use (adjusted R² = .747), while the economic model explained closer to half (adjusted R² = .5197). Mean years of schooling was the strongest individual predictor.

The optimized model included urban population, secondary education, mean years of schooling, ICT affordability, and political stability. Together, these variables explained more than 80 percent of the observed variation in internet use across the countries analyzed (adjusted R² = .8222).

Interpretation

The analysis does not show that education causes internet use. It does show that an account built around affordability and national income is incomplete. Education and political context carry substantial explanatory weight alongside infrastructure and cost.

Limits

The data are cross-sectional, so the results identify associations rather than causal effects or their direction. The indicators are proxies for more complex conditions, data collection varies across countries, and national averages conceal within-country inequality. The models are best read as evidence about structural relationships that require more targeted and longitudinal investigation.

Related work

• Published paper: Modeling Internet Use in the Global Development Context: Preliminary Findings and Future Directions
• Dissertation research program: Innovation as Freedom