Does adoption of broadband internet in firms enhance labor productivity and increase wages? And is this technological change skill biased or factor neutral? We exploit rich Norwegian data with firm-level information on value added, factor inputs and broadband adoption to answer these questions. We estimate production functions where firms can change their technology by adopting broadband internet. A public program with limited funding rolled out broadband access points, and provides plausibly exogenous variation of broadband adoption in firms. This enables us to address endogeneity of broadband adoption and examine how it shifts the production technology and changes the productivity and labor outcomes of different types of workers. We find that broadband adoption favors skilled labor by increasing its relative productivity. The increase in productivity of skilled labor is especially large for college graduates in fields such as science, technology, engineering and business. By comparison, broadband internet is a substitute for workers without high school diploma, lowering their marginal productivity. Consistent with the estimated changes in labor productivity, wage regressions show the expansion of broadband internet improves (worsens) the labor outcomes of skilled (unskilled) workers. We explore several possible explanations for the skill bias of broadband internet. We find suggestive evidence that broadband internet complements skilled workers in executing nonroutine abstract tasks, and substitutes for unskilled workers in performing routine tasks. When we use our production function estimates to construct measures of firm level productivity, we find that broadband internet accounts for a few percent of the standard deviation in total factor productivity across firms. Taken together, our findings have important implications for the ongoing policy debate over government investment in broadband infrastructure to encourage productivity and wage growth.
broadband internet labor productivity tasks technological change skill bias