When Do Investors Freak Out?: Machine Learning Predictions of Panic Selling
40 Pages Posted: 9 Aug 2021
Date Written: August 4, 2021
Despite standard investment advice to the contrary, individuals often engage in panic selling, liquidating significant portions of their risky assets in response to large losses.Using a novel dataset of 653,455 individual brokerage accounts belonging to 298,556 households, we document the frequency, timing, and duration of panic sales, which wedefine as a decline of 90% of a household account’s equity assets over the course of onemonth, of which 50% or more is due to trades. We find that a disproportionate numberof households make panic sales when there are sharp market downturns, a phenomenonwe call ‘freaking out’. We show that panic selling and freakouts are predictable andfundamentally different from other well-known behavioral patterns such as overtradingor the disposition effect. Investors who are male, or above the age of 45, or married, orhave more dependents, or who self-identify as having excellent investment experience orknowledge tend to freak out with greater frequency. We use a five-layer neural networkmodel to predict freakout events one month in advance, given recent market conditionsand an investor’s demographic attributes and financial history, which exhibited truenegative and positive accuracy rates of 81.5% and 69.5%, respectively, in an out-of-sample test set. We measure the opportunity of cost of panic sales and find that, whilefreaking out does protect investors during a crisis, such investors often wait too longto reinvest, causing them to miss out on significant profits when markets rebound.
Keywords: Panic Selling; Stop-Loss; Tactical Asset Allocation; Freaking Out; Deep Learning; Behavioral Finance
JEL Classification: G11, G01, G02, D14, D91
Suggested Citation: Suggested Citation