A standard incomplete-information war of attrition is extended to incorporate experimentation and private learning. We obtain a characterization of all equilibria in this extended setup and use this setup to illuminate a tradeoff between short-run and long-run gains of experimentation. The extension yields qualitative impacts on the strategic nature of the problem. The option value of experimentation serves as a credible commitment device to stay in the game, which is instrumental in inducing the other player to concede earlier. As a direct consequence, there may be an equilibrium in which the strictly less efficient player can get the better end of the deal, implying that slow learning can be a blessing in this type of competition. Our analysis gives insight into why an apparently inferior technology often survives in many standards competitions and more broadly offers implications for technology adoption and industry dynamics. We also show that there is a non-degenerate set of parameters that can support the Pareto-efficient allocation as an equilibrium outcome whereas it is never possible in the standard setup.
war of attrition bandit problem experimentation standards competitions private learning