Improving Portfolio Construction through Side Information and Quantified UncertaintydissertationsShah, Anish RasiklalcreatorHarrison, MatthewAdvisorUpfal, EliReaderKolm, PetterReaderBrown University. Department of Applied Mathematicssponsor202311, 149 p.born digitalThesis (Ph. D.)--Brown University, 2023thesesMotivated largely by the problems of estimation error in investment portfolio optimization, this dissertation develops techniques to improve covariance forecasting and portfolio construction. Covariance forecasting is improved by incorporating contemporaneous side information and by propagating uncertainty of the parameters in factor-modeled covariance to uncertainty of the covariance matrix itself. The latter makes it possible to approach several investment problems explicitly modeling uncertainty. Markowitz mean-variance portfolio optimization becomes robust optimization with an uncertainty set on portfolio utility. Uncertain risk parity considers deviation from the risk budget. Risk under uncertainty and price movement measures the stability of hedges under forecasting uncertainty and shifts in risk exposure caused by dynamic prices.RiskUncertainty--Mathematical modelsUncertainty QuantificationForecastingCovariancePortfolio management--Mathematical modelsRobust optimizationRisk ParitySide InformationUncertaintyPortfolio optimizationHedging (Finance)EnglishRPB20230602