The Arrival of Fast Internet and Skilled Job Creation in Africa∗

This paper provides evidence on how the expansion of fast Internet along its current frontier in developing countries affects job creation. We exploit the gradual arrival in African coastal cities of submarine Internet cables from Europe and maps of the continent’s terrestrial cable network that connects those cities with users. Robust difference-in-differences estimates from four datasets covering 14 countries show large positive effects on employment rates. A decrease in unskilled jobs is offset by a bigger increase in jobs in higher-skill occupations, and outside of South Africa job inequality between more and less educated workers falls. We use detailed firm level data available for some countries to investigate how higher average speeds for existing users and higher take-up enable greater (skilled) job creation. We find an increase in (i) workers and skilled positions per firm, firm level productivity, and the productivity of workers in skilled positions in existing Ethiopian manufacturing firms, (ii) firm entry in South Africa in the sectors that hire workers in the occupations that see growth when fast Internet becomes available, and (iii) workers in skilled positions, exports, and use of websites among firms in seven countries. Finally, we show that the impact on job creation is accompanied by a rise in average incomes. Our findings shed light on how modern information and communications technology affects total job creation, structural change, job inequality, and firm growth in the poorest region of the world. JEL codes: D2, J2, O1, O3 ∗[email protected], [email protected]. We thank Sebastian Axbard, Niklas Bengtsson, Esther Duflo, Ray Fisman, Francois Gerard, Amit Khandelwal, Mushfiq Mobarak, Kalle Moene, Eva Mörk, Anders Olofsgård, Michel Serafinelli, Kjetil Storesletten, Eric Verhoogen, Frank Windmeijer, Chris Woodruff, and seminar participants at Barcelona Summer Forum, Columbia, Harvard, IGC Growth Week, NBER Summer Institute, Oslo, Oxford, and Uppsala for comments and suggestions, and Patrick Kennedy and Roxanne Rahnama for great research assistance. We are especially grateful to Henrik Sigstad and Matthieu Teachout for many helpful conversations, and to Akamai and the World Bank for data access. Hjort thanks the Center for Development Economics and Policy at Columbia University for financial support.