DP17392 Microfoundations of Low-Frequency High-Impact Decisions
We show that grounding low-frequency high-impact strategic decisions under uncertainty on the development and test of theories is a foundational strategic choice of firms, and an important determinant of performance. Our normative Bayesian model shows that decision-makers benefit from comparing alternative theories and when theories are reliable (less uncertain) they should not experiment and they should focus on their most plausible theory (highest prior). Otherwise, they ought to experiment with their less reliable theories because they learn more. We also show that the variability of the optimal experiment matches the variability of the prior. In particular, large scale experiments may be overprecise and penalize radical new theories with more variable priors. Overall, our framework explains how structured exploration improves strategic decisions by reducing model mis-specification and why performance heterogeneity and competitive advantage is created by exploring theories with higher variance
