Estimating the Effects of a New Technology using a Duration Model for Staggered Adoption
Joint with Aureo de Paula
Measuring the impact of new technologies on productivity is a classic question in economics. However little work has focused on the modern office place for knowledge workers. In this paper, we study the diffusion of Microsoft SharePoint, a new technology aimed at bringing seamless collaboration for Microsoft Office documents, serving as a direct substitute for traditional email attachments. The classical parallel trends assumption is violated in this case because unobserved heterogeneity may generate correlation between adoption timing and productivity shocks. We propose a new type of staggered difference-in-difference estimator using the adoption timing of direct neighbours in the collaboration network as instruments to measure the effect of Share-Point adoption. Through a parametric model of the adoption diffusion process, the proposed estimator allows us to account for latent variables that may moderate inter-temporal shocks before adoption. We show that our estimator works well in simulations and our empirical results show that the traditional email attachments can be reduced by 25 pieces per user after 10 months of adoption of SharePoint. This estimation protocol is applicable more broadly in staggered adoption settings where the parallel trends assumption is likely to be violated.