When it Pays to Think about the Competition, and When it Doesn’t: Exploring Overconfidence Bias in Dynamic Games

Despite robust behavioral research that shows a widespread bias towards overconfidence in competitive scenarios, e.g., underestimating the competitor's skill level, there is little research on the long term costs associated with this bias. We develop a theoretical framework that allows us to explore systematic long-term ramifications of opponent skill estimation bias across different competitive contexts relevant to managers. We capture these contexts with dynamic branching games that are parametrized by four features. We use Monte Carlo estimation methods to test how the expected game outcomes compare under different types biases. The results suggest that bias in evaluating an opponent's skill level is less harmful when the opponent is more skilled, and when there is greater first-mover advantage. Furthermore, they suggest that if there is any effort cost associated with making a decision, then a bias towards overestimating the opponent's skill is never advantageous, while a bias towards underestimating can be advantageous in many contexts.