“ A Two-Stage Model of Innovation Adoption with Partial Observability : Model Development and Application . ”

We show how one can specify micro-level hazard models of adoption that separately represent the awareness and evaluation/choice stage of the adoption decision process even when one observes only the final outcome, i.e. adoption or not. These models can be derived from random utility theory. Using a small simulation study, we show that the two-stage partial observability specification retrieves weak advertising or contagion effects better than a more traditional one-stage model does when people really follow a two-stage decision process. We illustrate the value of the modeling strategy in a re-examination of Medical Innovation, a classic study of the diffusion of an antibiotic, traditionally credited for establishing the role of social contagion in innovation diffusion. Using newly collected data on advertising and applying both one-stage and two-stage models, we conclude that the traditional interpretation is based on confounding contagion and marketing effort. We also explain why other re-analyses have failed to document strong contagion effects.