TESTING THE LEAP-FROG HYPOTHESIS: The impact of existing infrastructure and telecommunications policy on the global digital divide

Abstract

This paper tests the ‘leap-frog’ hypothesis by modeling the impact of existing telecommunications infrastructure, controlling for economic, political and demographic factors, on changes in information communication technology (ICT) access for over 200 countries between 1995 and 2005. This study has significantly greater coverage than previous research, in terms of both time frame and country cases. First, the analysis demonstrates that in the first decade of the information society successful leap-frog countries are few and far between. Second, the relative distribution of personal computers, internet hosts and secure servers among the nations of the world has barely improved over the last decade. Third, contrary to received wisdom, most of the countries that might qualify as successful leap-frog countries are actually among the wealthiest in the world. Finally, while policy reform in the telecommunications sector can sometimes speed the diffusion of digital communication tools, the record of market reforms is mixed, and the overall effect of economic wealth is still paramount. In sum, a few poor countries have leapt ahead in the development of a few aspects of ICT infrastructure and use, but these relatively rare successes are more likely to be due to economic productivity than to privatization, regulatory separation and depoliticization, or market liberalization in the telecommunications sector.

Keywords:

• Global digital divide
• Gini coefficient
• leap-frog
• telecommunications policy
• privatization

Acknowledgements

For their helpful feedback, I am grateful to Sandra Ball-Rokeach, Gina Neff, Christian Sandvig, Eric Schulenberger, Crispin Thurlow and the anonymous reviewers of iCS. For research assistance, I am grateful to Tamara Barnett and the World Information Access Project.

Notes

1 In addition, it is assumed that the data obey a first-order Markov process in that the values of a given year depend only on the values during the preceding year, but not beyond. These OLS tests make assumptions about homoscedasticity and normality.

2 Additional variables, including literacy rates, were tested but were dropped from the models if the effects were negligible, statistically insignificant, contributed little to explained variation, and the literature reviewed above did not recommend their use.