Wednesdays @Hariri

Simulated Likelihood Estimators for Discretely Observed Jump-Diffusions

Gustavo Schwenkler

Junior Faculty Fellow, Hariri Institute for Computing
Assistant Professor of Finance, Questrom School of Business
Boston University

With an introduction by Marcel Rindisbacher, Associate Professor of
Finance and Chairman, Questrom School of Business

Abstract: This paper develops an unbiased Monte Carlo approximation to the transition density of a jump-diffusion process with state-dependent drift, volatility, jump intensity, and jump magnitude. The approximation is used to construct a likelihood estimator of the parameters of a jump-diffusion observed at fixed time intervals that need not be short. The estimator is asymptotically unbiased for any sample size. It has the same large-sample asymptotic properties as the true but uncomputable likelihood estimator. Numerical results illustrate its computational advantages.

Bio: Gustavo is an assistant professor of finance at the Boston University Questrom School of Business. His research focuses on the development of computational and statistical tools for the measurement of financial risks. Gustavo has presented his work at major international conferences, including the American Finance Association Meeting, the Annual Meeting of the Econometric Society, and the Joint Mathematics Meeting. He received his PhD in management science and engineering in 2013 from Stanford University and his diploma in applied mathematics and economics from the University of Cologne. Prior to pursuing his doctorate, Gustavo worked at Goldman Sachs and Deutsche Bank.