Date of Award


Document Type


Degree Name

Bachelor of Science


Mathematical Economics

First Advisor

Dr. Saif Mehkari

Second Advisor

Dr. Paul Kvam


Since the 2008 financial crisis, interest rates and bond yields have been low all through the recovery and expansion that followed, and they are still low. As a result, more investors have been attracted to US equities, a space of possibly higher returns. However, these returns come with a potential downside: risk of loss. One of the methods to assess this potential downside is value-at-risk (VaR), which gained momentum in the late 1990s. At the time, the market risk amendment to the 1988 Basle Capital Accord required commercial banks with significant trading activities to put aside capital to cover market risk exposure to their trading accounts. VaR was used to determine the amount to be set aside (Lopez, 1998).

Formally, VaR is the maximum expected loss in a portfolio with a (1-θ)% confidence. This measure is developed to forecast the θ quantile of the profit-loss (P&L) distribution for a time period ahead (day, month, year). Much of the research done in this area has concerned the theoretical implementation of this method, and accuracy comparisons among calculation variations. At the same time, given the purpose of VaR’s wide-spread use in the 1990s, the model has been mainly implemented to assess portfolios with short time-horizons and US-only exposure. For this reason, much of the little practical research in the topic has focused on comparing different VaR calculations in commercial banks against their daily P&L realizations. My aim is to test differently constructed VaR models using the holdings of the University of Richmond student-led ETF investment fund. The latter has exposure to non-US equities, making it a non-conventional case to test the accuracy of the models.