RISK MANAGEMENT
An overview of the industry-standard market risk management framework
Disclaimer: the sole purpose of this article is to educate the readers by expressing the author’s personal views. The content does not constitute any financial advices nor reflect any views of the author’s affiliated organizations.
Why should we care about market risk?
Market risk is the possibility of financial losses in positions arising from changes in values of risk factors.
During 2008 global financial crisis, S&P 500 index fell more than 50% and countless companies such as Lehman Brothers and Bear Stearns collapsed. Banks were lending aggressively so that they can sell packages of junk mortgages — underestimating the risk of the housing market crash.
The purpose of this article is to provide a big sketch of the market risk management process used by large financial institutions today. Hence, we are not getting lost in the details of fancy quantitative models such as Black-Scholes and Hull-White, despite peoples’ predilections for them.
Identify risk factors and obtain market data
Market risk management simply begins with knowing what we own. Risk managers spend countless hours building a comprehensive inventory of risk factors, which are categorized into 5 risk classes. While this is a piece of cake for individual investors with U.S. stocks or ETFs, it is rather perplexing for large institutions with complex derivatives all over the world.
Based on the inventory, risk managers can acquire relevant market data either internally from their trade records or externally from vendors such as Bloomberg or Thomson Reuters. Market data are not only used to value the portfolio on a particular date but also used to generate scenarios to forecast how our portfolio will look like tomorrow if there is a black swan event.
Generate scenario and compute profit & loss
To generate these scenarios, most banks use historical simulation, which is, in a nutshell, computing day-over-day changes of historical time series data. It gained popularity due to its conceptual simplicity, computational efficiency, and non-parametric method (i.e. does not assume any distributions). On the other hand, alternatives like Monte Carlo simulation are relatively complex, computationally expensive, and parametric methods that assume data follow normal or log-normal distributions when they usually do not. As an illustration, a historical scenario distribution of DAX index is presented below.
With 500 most recent historical scenarios, banks can simulate the changes in their portfolio value (i.e. P&Ls) 500 times. Then, they can take the fifth-worst or sixth-worst value as a risk measure and call it tomorrow’s 99% Value-at-Risk (VaR), a loss threshold not expected to breach 99% of the times.
Compute VaR, stressed VaR and stress test P&Ls
The main caveat of using VaR is that it can understate the risk if the market has been calm recently. To complement VaR, banks select the most volatile period based on their portfolios such as the 2008 global financial crisis and repeat the same procedure as VaR to calculate stressed VaR (sVaR).
Even with VaR and sVaR, banks are still far away from being safe. They are still unsure if they can survive a historical event like 9/11 or a hypothetical event like escalating China-United States trade war, which will lead to panic selling in stock markets. To bridge such information gaps, banks create stress tests scenarios classified into 3 types: historical, hypothetical and single risk factor stress scenario.
Historical stress scenarios are drawn from the past, while hypothetical stress scenarios are anticipated events based on the current economic outlook. On the contrary, single risk factor scenarios are not defined as events and simply defined by shifting the values of key risk factors by large amounts. For instance, the U.S. treasury yield curve shifting upward by 1% is a commonly used single risk factor scenario.
Compile a risk report for risk management decisions
The portfolio is fully revalued under each stress test scenario to compile a list of stress P&Ls. Combined with VaR and sVaR in a daily risk report, this will be presented to senior management, who will compare the today’s measures with predefined limits.
For instance, if an equity trading desk has a VaR of $3MM but has a VaR limit of $2MM, senior management will find the equity trading business too risky and reduce risks by selling some stocks. If a bond trading desk does not breach VaR and sVaR limit while it performs poorly under an unanticipated inflation scenario, senior management may decide to substitute some bonds with inflation-linked bonds to hedge inflation risk.
This article is just the tip of the iceberg. Upcoming posts will deep dive into risk methodologies and quantitative concepts that are integral in computing the market risk measures presented in this post. Stay tuned!
Sources:
