Social and Behavioral Sciences Commons^™

Converged Subset Portfolios: An Extension To Subset Optimization, Coleman Cornell

CMC Senior Theses

The limited span of useful data, coupled with increasingly expansive asset universes, cripples the traditional mean-variance problem. When optimizing in these environments, the pronounced effect of estimation error yields extremely unstable portfolios when evaluated out-of-sample. As a proposed solution to the "curse of dimensionality," Gillen (2016) presents subset optimization as a technique to reduce the impact of estimation error. Instead of optimizing jointly over the entire asset universe, subset optimization na\"ively aggregates over many "subset portfolios" that each optimize over a much smaller random sample of assets. Given the inefficiencies when using naive aggregation, converged subset optimization is presented as …