The Herd Behavior Index and predicting market fear.
The past has learned that stock prices tend to move together. Moreover, at moments of high market fear, this co-movement is stronger and stock prices move predominantly downwards. In such a market situation, diversification benefits dry up and stock picking does not help to protect an investment portfolio.
In this project we implement the Herd Behavior Index, also called HIX, which was invented in Dhaene et al. (2012). The HIX is an option-implied measure taking values between 0 and 1, which can be calculated daily. If at a certain day, the HIX is close to 1, this is a sign that stock prices are likely to move together in the near future.
The aim of this project is to determine values of the HIX for the S&P 100 between 1996 and 2017. A next step is to study the time series of the HIX and investigate the behavior of the HIX in times of market stress (i.e. during the dot-com bubble, 9/11 attacks, European debt crisis, Lehman brother, Brexit, etc.). Moreover, we compare the HIX with the VIX (the volatility index). The HIX and the VIX are very similar, but whereas the VIX measures volatility, the HIX measures co-movement.
Students: Owen Adhikaputra, Supasin Chalermpoonsup, Zhihan Hui, Da Xu, Yi Yuan
Supervisor: Daniël Linders
Optimal Investment with Forward Preferences and uncertain parameters under binomial market model.
Given a financial market environment, an agent aims to solve her optimal investment strategy. This project is a continuation of “forward and backward preferences” proposed in IGL and IRL last year. In the previous project, under the binomial market model, comparing to the classical backward approach, we showed substantial improvement in both computation time and cumulative earning when the forward approach, introduced by Musiela and Zariphopoulou (2008), is numerically implemented to the historical data of S&P500. Although the agent could update the real-time market information at the end of each period due to the forward nature, she inevitably encounters model uncertainty in each period. Inspired by Chong and Liang (2019) in continuous-time setting, this project studies the optimal investment problem with forward preferences and uncertain parameters under the binomial market model.
Students: Norman Dewantoro, Rhyxian Lim, Suyoung Park, Himanshi Sharma
Supervisor: Alfred Chong
European-type basket option pricing: independence and comonotonicity approximations.
This project solves the European-type basket option pricing problem. Finding analytical solutions or stable numerical schemes for the corresponding high-dimensional PDE is still an open problem. Hanbali and Linders (2019) propose an approximation of the problem using the element of comonotonicity. Their theoretical results have been further strengthened by Ling (2019) using a modern machine learning approach with tremendous improvement in terms of computation time without deteriorating numerical solutions much. In view of these, this project aims to relax the full comonotonicity approximation in Hanbali and Linders (2019), to reduce the pricing error arising from dependence approximation, while to implement modern machine learning approaches, to rectify the expenses in computation time due to a more realistic approximated dependence structure.
Students: Ruizheng Bai, Kara Wong
Supervisor: Alfred Chong, Daniël Linders
Graduate Supervisor: Biwen Ling
Cyber risk profile construction via individual cyber losses aggregation.
Cyber risk refers to the potential losses that a firm might suffer due to a failure of its information system. The exponential increase in the use and the complexity of information systems has made cyber risk one of the most important and vulnerable operational risks for a company. In order to determine the total cyber risk exposure of a company, one first has to determine potential losses due to different sources of cyber risk, such as DDoS, ransomware, etc; one then has to aggregate individual cyber losses of the company.
In this project, we investigate how to determine the individual cyber loss distributions using historical data. We then investigate different methodologies to aggregate the individual loss distributions to construct the aggregated cyber loss distribution of a company. Lastly, we investigate different models to quantify the aggregated cyber loss using risk measures.
Students: Nargiz Alekberova, Joshua Immanuel, Linshan Jiang, Evelyn Lai Jia Yi, Hanqing Wang