Whence LASSO? A Rational Interpretation

This paper rationalizes the LASSO algorithm based on uncertain fat-tail priors and max-min robust optimization. Our rationalization excludes heuristic learning or restrictive prior assumptions in the original interpretation of LASSO (Tibshirani (1996)). In our setting, economic agents (arbitrageurs) face ambiguity about fat-tail shocks and in equilibrium, they ignore a reasonable range of ambiguous signals but respond linearly to almost unambiguous signals. With this LASSO equivalent strategy, arbitrageurs can amass extra market power which induces a ``cartel'' to protect their aggregate profit from being competed away. This result shows a new mechanism for limited arbitrage