Joint Stochastics-Math Finance Seminar - Three puzzles in quantitative finance

Other Talks
Wednesday, November 16, 2016 - 1:00pm
1 hour (actually 50 minutes)
Skiles 005
J.P. Morgan
1. One day before the election, the statistics site 538 predicted a 70% chance of a Clinton victory. How do we judge the quality of probabilistic prediction models? Ultimately every quant finance model has a probabilistic prediction model at its core, for instance the geometric Brownian Motion is the core of Black-Scholes.  I will explain the Basel Traffic Ligths Framework and then I'll ask the audience to think how the framework can be extended.     2. Multi-factor local volatility. I will explain Dupire's local volatility model and ask how this model can be extended to a multi-factor framework.     3. Model overfitting. There are objective criteria for statistical model overfitting,  such as AIC. Such criteria don't exist for risk-neutral derivatives pricing models.