Understanding the Impact of Automated Trade Algorithms
The economic crisis that began in 2007 triggered a sense of financial insecurity among big and small institutions as well as individual investors. The housing bubble, subprime lending and deregulation were all thought to contribute to market instability. Ru Jun Han ’14 believes another, more technological, component was also responsible.
Interested in economic regulation, Han is discovering how automated trading impacts financial markets. In particular, he is looking at high frequency trading (HFT), a complex system of computer hardware and programs that use small transactions to accumulate wealth over time. A Levitt Summer Research Fellowship makes his research about HFT possible. Automated trading, which has existed for decades but is new in practice, uses a computer algorithm to monitor stocks and trade assets without human assistance.
Han said the process has become “unfathomably quick,” and transactions can occur within milliseconds. HFT is commonplace in hedge fund markets, and Han observed that the programs could not function at such incredible speeds without modern technology.
Professor of Economics Chris Georges, who has developed a financial market model for previous research, is advising Han. Han “is mostly interested in using programming and mathematical tools to solve real problems,” and is writing additional source code to increase the complexity of the model. Everything in their model can be manipulated, from agents’ decisions to stock values, providing the sort of flexibility “that is not attainable working with historical data.” The program generates unique stock price trends, dictates agent’s behaviors and sets trade frequency. It can generate scenarios that closely mirror current markets, or it can generate different values for more experimental purposes.
Han “definitely wants to modify the current model and make it more sophisticated,” and his project involves “adding heterogeneity to the artificial markets.” He is using agent-based modeling, a method where programming tools simulate real financial markets, to examine HFT.
HFT has been heavily criticized since the crisis, and many analysts suggest that this kind of trading introduces volatility to financial markets. Han said, “Automated trades happen so fast; even a minor fault in the algorithm can lead to huge mistakes.” Computers are programmed to capitalize on price fluctuations and may not recognize errors in securities listings.
According to Han, “in order to regulate these new markets, we need to first understand how computer trading works.” Financial markets are entering a new digital era, and humans cannot compete with computers’ speeds. A computer’s longer memory allows it to remember past stock trends and make more informed forecasting decisions.
Previous models involved both uniform markets and computer agents. Han is enhancing the model by introducing more realistic agents who are risk averse, risk neutral or risk seeking. Agent characterization leads to more financial instability, so it is essential that we understand the impact of their varied decisions.
Han is taking this work even further; he will “make agents change their attitude toward risk depending on prices.” The algorithm will allow agents to become more risk-seeking when prices rise or more risk-averse when prices begin to fall.
Han hopes his work develops a more realistic model for economists and government officials to study. Once researchers create a model that mimics real word markets, policymakers will better understand how to monitor computer decision-making and when human intervention might be necessary.
Han is a graduate of the Manhattan Comprehensive Night and Day School (N.Y.)