Overinference from Weak Signals and Underinference from Strong Signals

We use experimental and empirical data to show that whether people overreact or underreact to news depends on how informative the news is. Numerous experiments have studied how people respond to highly informative news; however, in many important settings, people typically receive only weakly informative news. In an experiment, we show that people systematically overinfer when signals are weak despite underinferring when signals are strong. Misperceptions are more severe for subjects who are less experienced and less cognitively sophisticated. These results align with a model of cognitive imprecision we adapt to study misperceptions about signal informativeness. We then use complementary empirical data from betting and financial markets to show how these patterns play out in high-stakes real-world settings. We develop an empirical method to test over- vs. underreaction in observed prices across differing signal-strength environments, using time to resolution as an observable correlate of signal strength. When signals are weak, prices exhibit excess movement, indicating overreaction; the pattern is reversed with stronger signals, in accord with our theory and experiment