Finance Internal Seminar: Gyuri Venter
Title: Regularized Learning from Prices
Info about event
Time
Location
Universitetsbyen 51, 8000 Aarhus C, Building 1816, Room 613
Organizer
Presenter: Gyuri Venter, AU
Title: Regularized Learning from Prices
Abstract: Investors forecast with shrinkage estimators such as ridge regression, Ledoit–Wolf covariance shrinkage, and betas shrunk toward one. We embed these estimators in a market with dispersed information, in which investors learn from prices. Because the estimators shrink the most precise signals most, traders can underuse the price precisely when it is informative. Ridge generates trading volume that grows with market size, momentum, and too little price volatility; Ledoit–Wolf generates the same volume with reversal and excess volatility. When traders also have hedging needs, regularization improves risk sharing; because more data means less regularization, better data can lower welfare.
Organizers: Stefan Hirth and Mads Markvart Kjær