Last week, we shared a post recapping some key themes and a general framework discussed by Dr. Dan Rosen, Managing Director of Risk Analytics at S&P Capital IQ, during a recent webinar entitled “Re-Thinking Scenarios: Integrating Economic Scenarios with Advanced Scenario Analytics to Manage Investment Portfolios.” Let’s go a little deeper into some of Dr. Rosen’s key strategies for building out a joint simulation model that combines macroeconomic factors with risk factors. This is a core component of being able to blend traditional risk analysis with economic risk factors.
Before creating a joint simulation model, the risk factors and economic factors to be used in simulation must be selected. This process requires three key steps:
- Understanding The Portfolio – This critical, often-overlooked step aids in ensuring the outcome of the scenario analysis is usable. Effective scenario analysis begins with choosing the risk factors that can drive substantial changes in the portfolio. This can be achieved by analyzing the portfolio exposures.
- Defining The Basic Risks To Be Considered – Looking at the value at risk, marginal value at risk, and net market value for each position by sector, currency, and other characteristics will lend further insight into the factors that are most likely to impact the portfolio most significantly.
- Distilling An Economic Forecast Into Relevant Risk Factors – Creating a joint distribution of all the factors to use in building out the simulation model will represents how the joint economic and risk factors will evolve together in the future.
Constructing The Joint Model
A joint simulation model contains two parts: Economic and Risk Factor Residuals and a Co-Dependent Structure. In building out the one-year simulation in the presentation, Dr. Rosen used 20 years of quarterly data for all of the economic and market risk factors. Then, he used the time series data to create a statistical model by first converting the data to logs (or log differences depending on factor type), extracting the auto correlations, and finally extracting the residuals.
Next, we model the co-dependent structure using the historical residuals. The co-dependence will be described by a matrix of residuals for each quarter by first determining the correlations for all the market risk factors affecting the portfolio. With this information, correlations for both market risk and economic factors can be constructed. Once the model is prepared for each quarter, the joint simulation can be used to create scenarios and stress testing analyses on any of the factors defined.
With this model created, we can use it to apply economic scenarios to specific portfolios. Our next post will highlight how to do this. Access Dr. Rosen’s webinar presentation to learn how to apply a joint simulation model such as this to a multi-asset class portfolio.