Algorithmic Finance

Deep neural networks (DNNs) are powerful types of artificial neural networks (ANNs) that use several hidden layers. They have recently gained considerable attention in the speech transcription and image recognition community (Krizhevsky et al., 2012) for their superior predictive properties including robustness to overfitting. However their application to algorithmic trading has not been previously researched, partly because of their computational complexity. This paper describes the application of DNNs to predicting financial market movement directions. In particular we describe the configuration and training approach and then demonstrate their application to backtesting a simple trading strategy over 43 different Commodity and FX future mid-prices at 5-minute intervals. All results in this paper are generated using a C++ implementation on the Intel Xeon Phi co-processor which is 11.4x faster than the serial version and a Python strategy backtesting environment both of which are available as open source code written by the authors.

This study examines the network dynamics of fourteen Asian Stock Markets (ASMs) in three phases (pre, during, and post) of financial crisis of 2008. Based on network statistics, I find that ASMs network is more interconnected during the crisis period than pre-and post-crisis period. Furthermore, using the Minimum Spanning Tree (MST) diagram, I find that the stock markets of Hong Kong, Japan, Korea, and India play a significant role in these networks and any shock to these markets can lead to contagion. The trade and the interest rate differential are the major driving forces behind these linkages. This work has practical implications as it provides insight on portfolio diversification during the crisis period and can also be used in anticipating the route of crisis.

This paper documents the Russian exchange-traded funds (ETF) puzzle. Since 2014-2015, most Russian fund-of-funds have been investing only in one preselected Western ETF. During that period, these funds have raised more money than any other fund category in Russia. However, if an investor buys shares in an ETF via a mutual fund (MF) rather than doing it directly, she overpays up to 36% of the invested capital over a 10-year horizon. Additionally, the paper provides a brief overview of possible reasons for this anomalous, suboptimal index investing behavior, and formulates hypotheses for further research.

In a Semi-Strong Form (SSF) Efficient Market, asset prices should respond quickly and completely to the public release of new information. In the period from his election on 11/8/16 to his swearing in ceremony on 1/20/17, President-elect Trump posted numerous statements (‘tweets’) on his Twitter messaging service account that identified ten publicly traded firms. In the absence of new information, the Efficient Market Hypothesis (EMH) predicts that these announcements should have little or no price impact on the common stocks of these firms. Using standard event study methods, we find that positive (negative) content tweets elicited positive (negative) abnormal returns on the event date and virtually all of this effect is from the opening stock price to the close. Within five trading days, the CARs are no longer statistically significant. President-elect Trump’s tweets were associated with increases in trading volume and Google Search activity. Taken as a whole, the price and trading volume response, combined with Google Search activity is consistent with hypothesis that it was small/noise traders who were acting on President-elect Trump’s tweets and that their impacts were transitory.