Relaxing Identifying Assumptions

Résumé du livre

Every empirical approach to causal inference relies on untestable identifying assumptions. Yet, just 5% of recent empirical studies in top management journals test for sensitivity to the violation of these assumptions. To address this gap, we highlight recent methods that examine how sensitive estimates are to violations of identifying assumptions for three identification strategies: controlling for confounders, difference-in-differences (DD), and instrumental variables. We apply these tools by examining how an inventor's first patent affects future mobility. In each case, we find that first patents are associated with increased mobility, though the results are sensitive to moderate violations of the identifying assumptions. Our DD analyses also reveal that inventors are on a downward-mobility trend in the years before they patent, which we call the “Inventor's Dip,” but are more mobile after receiving a patent. These results suggest that a first patent likely increases mobility, while the pre-patenting innovation process reduces it.