Technology Adoption in Input-Output Networks

We study how input-output networks affect the speed of technology adoption. In particular, we model the decision to adopt the programming language Python 3 by software packages. Python 3 provides advanced features but is not backward compatible with Python 2, which implies it comes with adoption costs. Moreover, packages are dependent on other packages, meaning one package’s adoption decision is affected by the adoption decisions of other packages because many packages are linked to each other.

We build a dynamic model of technology adoption that incorporates an input-output network and estimate it using a complete dataset of Python packages. We are among the first to link the literature of dynamic discrete choice models to network analysis. We also contribute to the literature on technology adoption by showing the adverse effects that input-output networks can have on how technology is adopted in a dynamic setting.

We show that a package’s adoption decision is significantly affected by the adoption decisions of its dependency packages. We conduct counterfactual analyses of cost subsidies that target a community level and show that network structure is crucial to determining an optimal policy of cost subsidy.