Since the mid 1990’s, Value at Risk (VaR) has played a central and ever-expanding role in financial risk management. But throughout this time, a fundamental error in the accepted method of VaR calculation has gone largely unacknowledged:  moments of a nonlinear transformation of the moments of a random variable (in this case sample volatility) do not equal the transformation of its moments.

Tsafack and Cataldo (Empirical Economics 61:1351–1396, 2021) demonstrated that this results in a systematic and material downward bias in VaR estimates.  However, several conceptual and practical challenges to implementation remain unaddressed.  This paper explores key interdependences between the returns process, volatility forecast methods, and the statistical properties of sample volatility to develop an effective, practical protocol for unbiased estimation of VaR.

Bio:
Jim is on the Faculty of the University of Rhode Island School of Business, where he advises DBA candidates on their dissertation research and conducts research seminars.   He  received his PhD in Economics from Columbia University, an MA in Accounting from Suffolk University, and a BA in English and American Literature and Economics from Brandeis University.  His career has included turns as a Federal banking regulator, Director of Treasury Risk Management at the Federal Home Loan Bank of Boston, Structured Products trader at Cantor Fitzgerald, Assistant Professor at Suffolk University’s Sawyer Business School, and Director of Model Risk Management at Citizen’s Bank.  His current research focuses on volatility estimation and financial risk.