This website is the homepage for the Financial Engineering project for SYS and OR 699 for Fall of 2013.
Many scientists have devoted study to the credit crisis which started in 2007 and have in turn asked, "Could this have been predicted?" Analytically, this would be difficult--dynamics of realistic interactions between a large population of economic agents are far too complicated to compute analytically. However, where traditional economic analysis falls short, it is possible that agent-based modeling (ABM) can provide some insight due to the ability to model interactions between agents and therefore how an economic system changes over time due to these agent-to-agent interactions. Specifically, the purpose of this project is to evaluate whether or not ABM can be used to successfully model a financial system and study the dynamic properties of interactions and their connection to potential financial crisis.
ABM shall be used as the main tool for studying and predicting the emergence of risk events associated with a failed (total collapse and liquidation) hedge fund. The rationale for scoping the study using a failed hedge fund is three-fold. First, modeling the global economy is infeasible due to the size of the global economy (would require potentially millions of specialized agents) and would require in-depth knowledge of mathematics, sociology, and psychology--modeling a hedge fund and its associated entities is achievable given the timeline of the study. Second, as hedge funds have more relaxed regulatory requirements than mutual funds, they can engage in more risky trading behavior, exposing themselves to potentially more chances of making investments which lose value--in turn causing a "financial crisis" for the hedge fund and its associated entities. Third, there are many examples in the history of the financial market of failed hedge funds to calibrate an agent-based model against. One such failed hedge fund is Long Term Capital Management (LTCM).
Although it is uncertain whether the ABM model will produce results similar (i.e. the model may show that the modeled hedge funds' overall portfolio value has decreased similarly to crash levels) to that of the LTCM crash in 1998, the ABM model could have important potential to study general financial failure for hedge funds. Financial failure is defined by extreme portfolio value loss when capital lost exceeds capital required to cover losses. VaR traditionally computes a probability of when a certain capital requirement level is not exceeded within a set number of business days. The complement of that event is the probability of failure that will be compared against the failure rate derived from the ABM model. If the ABM model clearly demonstrates a higher likelihood of extreme events to include heavy portfolio loss for a LTCM-like hedge fund when compared to traditional approaches such as VaR, the proposed ABM model can become a baseline in the future for other hedge funds with their own distinct trading strategies and inherent risks.