Telfer School Researcher Focuses on Decision Tools For Data Driven Risk Management

Professor Jonathan Li of the Telfer School of Management explores how rich datasets available today can be leveraged to improve business decision-making. Li aims to develop new decision tools that incorporate a holistic view of risk, made possible by the ongoing revolution in data. His research draws on interdisciplinary knowledge from the fields of business analytics (operations research) and financial engineering, and relies heavily on the use of algorithmic tools such as optimization and simulation. Progress in these areas is expected to benefit a wide range of data-intensive business activities that typically have to contend with high degrees of uncertainty, such as financial investment and derivative pricing, supply chain management and revenue management.

Traditional decision models tend to take a narrow view of risk, and often understate the impact of uncertainty on our decisions. The 2008 financial crisis provides only the most famous recent example of how decisions based on naïve risk models can go awry, Li says. Successful implementation of risk management in decision models requires two ingredients: first, access to real-world data, and second, models that can convert the data into risk-relevant information and drive risk-aware decisions. Advances in IT have led to improved data sets, both in terms of quantity and quality, providing new opportunities to better calibrate risk models, Li says. The principal unmet challenge now lies in the design of the risk and decision models that can fully exploit risk-relevant information.

Examples of traditional risk models failing to account for the actual complexity of risk are easy to find. Many models measure risk by calculating simple statistics from the data such as standard deviation. However, lower standard deviation does not necessarily point to less risk. In these models, real-life uncertainties are often poorly approximated by the “well-shaped” normal distribution (or “bell curve”), Li explains. The root problem is the mistaken assumption that uncertainties are “well-structured” when in reality they are not. Li’s research focuses on developing methods that take a broader perspective in modeling uncertainties and hence are better able to project risk more accurately. “Instead of using a single model, one of our strategies is to consider calibrating multiple models simultaneously using large datasets,” Li says. “That will be far more effective in capturing uncertainty in risk measurement.”

Professor Li became assistant professor at the Telfer School after completing his Ph.D. in Operations Research from the Department of Mechanical and Industrial Engineering at the University of Toronto (2012) followed by postdoctoral work at HEC Montreal (2013). He will be teaching and working with colleagues and students at the Telfer School in the area of Business Analytics and Information Systems and master’s students from interdisciplinary master’s programs such as the M.Sc. in Systems Science.

Li continues the work he began during his Ph.D. program, which looked at how data collected directly from users can be used to calibrate the uncertainty of risk measures. This research recognizes that decision makers from different sectors or industries can have different perspectives of risk, depending on the nature of the work undertaken. That represents an exciting development because over the long term, Li’s risk measure could potentially be customized to suit these various risk perspectives. The work will certainly help to bridge the gap between the theory and practice in risk measurement, Li says.

“By relying on models that provide only a very incomplete picture of uncertainty, managers get a false sense of security about their projections, particularly when they don’t recognize the limitations of their models,” he explains. This is a pressing problem for decision making in today’s complex, fast-changing environments. Risk management “is about finding better ways to manage and foresee your worse-case scenarios. Risk measures that are sufficiently robust, that can address multi-faced uncertainty, can put you on the correct path from the outset.”