Undergraduate Honors Theses

Thesis Defended

Spring 2015

Document Type


Type of Thesis

Departmental Honors



First Advisor

Carlos Martins-Filho


This paper determines and compares the efficacy of systematic approaches to portfolio construction using hourly assets from the PJM (Pennsylvania-New Jersey-Maryland) Independent System Operator (PJMISO). The two models addressed herein are an unconditional mean-covariance model using ordinary least squares (OLS) estimation and a lagged seemingly unrelated regression (SUR) model using feasible generalized least squares (FGLS) estimation. Despite the abundance of quantitative research on electricity markets, market participants are known for doing anecdotal analysis on the qualitative features of the market, e.g. weather patterns, transmission and generation outages, gas prices, etc. to construct daily portfolios and manage risk. Further, there is a lack of academic research that takes a practical approach to daily, portfolio risk in these markets. This paper bridges the gap between academic and industry analysis by incorporating quantitative analysis into a practical, systematic approach to portfolio construction. To test for and compare the efficacy of optimized daily portfolios, this paper will simulate portfolios using realized returns from the PJMISO market.