Spring 2019 Conference Agenda

Geopolitical events may impact volatilities of all asset classes, sectors and countries. It is shown that innovations to volatilities are correlated across assets and asset classes and therefore can be used to forecast or hedge geopolitical risks. The paper develops a statistical model of a Geopolitical Volatility Factor or GPVF, based on the standardized residuals from a factor model with GARCH style residuals. A test for GPVF is developed with estimation algorithms. These are applied to 9 ETFs of sectors of the US and to 45 MSCI country ETFs. The model has asset pricing implications for hedging geopolitical risks.

Standard sequential generation methods assume a pre-specified generation order, such as text generation methods which generate words from left to right. In this work, we propose a framework for training models of text generation that operate in non-monotonic orders; the model directly learns good orders, without any additional annotation. Our framework operates by generating a word at an arbitrary position, and then recursively generating words to its left and then words to its right, yielding a binary tree. Learning is framed as imitation learning, including a coaching method which moves from imitating an oracle to reinforcing the policy's own preferences. Experimental results demonstrate that using the proposed method, it is possible to learn policies which generate text without pre-specifying a generation order, while achieving competitive performance with conventional left-to-right generation.

There’s a tremendous amount of interest in the use of machine learning in modern day financial decision making. Much of this interest is fueled by increasing amounts of available data and the general success of machine learning in other domains such as perception. I start by assessing the opportunities and key challenges for machine learning in exchange traded and OTC markets, and how finance problems are uniquely challenging due to their low inherent predictability. I present results from experiments that highlight how the well-known “variance problem” problem in machine learning manifests itself in building models from data. The results have implications for when should trust autonomous machine learning systems with decision making.

We investigate how one can use variables built on implied and historic realized variances to predict future realized variance in the frame of machine learning. By using daily price data on SPX, options written on it and VIX, We compare the predictability and stability of each model, and discuss model selection in a few different scenarios. Then we apply mean decrease accuracy type of feature importance analysis to derive the usefulness of each variable in realized variance forecast. The tradability of certain models is also discussed.

The most successful financial innovations – including indexing – have deep roots in academic research. Aging populations and lack of savings imply a huge potential opportunity for disruptive innovation in retirement solutions driven by new research. This talk focuses on how technology and data science intersect to drive new innovation in the retirement space including factor glide paths, smart decumulation using analytics, and overcoming behavioral biases. An important focus for the discussion is robo-advising, which lies at the intersection of asset management, retirement solutions, and fintech. We will also emphasize the difference between the private and social benefits from financial innovation.