Network effects and strategic effects: essays on multilevel technology adoption
This dissertation analyzes the network effects and strategic effects in technology and certification adoption. The first chapter analyzes the externalities of the Electronic Medical Records (EMR) technology. EMR is a multi-level technology, which is characterized by both network effects and strategic effects. Differentiated adoption levels might yield different external effects to neighboring potential adopters. By using a panel of U.S. hospitals' EMR adoption and applying a set of flexible reduced-form regressions, I find the presence of a complementary effect for the first-level EMR adoption, and a competitive effect for higher-level adoption. Chapter two studies the network effects and strategic effects of EMR technology adoption among hospitals using a dynamic structural model. I estimate this dynamic structural model by applying the methods by Bajari, Benkard and Levin (2007) (BBL), and Pakes, Ostrovsky and Berry (2007) (POB), which provide a two-step algorithm for estimating dynamic games. The primary result is the presence of both competition and complementation in this adoption. Furthermore, I perform some counterfactual experiments. The first experiment is to compare hospitals' adoption behaviors in monopoly and duopoly markets, and I find that network effects and strategic interactions are important in the duopoly market. I then perform a policy experiment to examine the effect of the government's incentive programs for EMR adoption, and I find it would greatly stimulate adoption. The third chapter studies the adoption of LEED (Leadership in Energy & Environmental Design) certification, which is an internationally recognized building certification system that evaluates environmental impact. Competition may affect not only certification adoption but also the adoption of quality standards. LEED offers four levels of certifications to indicate the different standards of environmental impact. By using a detailed dataset of LEED registrations in the U.S. and applying the Multinomial Test of Agglomeration and Dispersion (MTAD) developed by Rysman & Greenstein (2005), we find the allocation of certification levels is more agglomerated than independent random choice would predict.