Computing Economic Capital via Portfolio Simulation
Joseph L. Breeden, Ph.D.; Signals, Spring 2002
The Basel II Accord offers benefits to financial services institutions that develop internal ratings-based (IRB) models for setting economic capital. Those institutions that do not may face rather substantial penalties in the form of increased capital requirements in the future.
Although Basel II creates a strong incentive for IRB models, it offers scant advice on how these should be created. Retail portfolios have historically received much less attention than commercial portfolios for modeling total risk in part due to the inherent complexities of modeling retail portfolios. Portfolio-level forecasts of unexpected losses are unable to capture changes in portfolio mix, economic climate, originations quality, portfolio maturation, and more. Because of these shortcomings, it is clear that few if any existing models will meet the stringent requirements for IRB models under Basel II.
Another commonly discussed approach is to use account-level bureau scores to assess portfolio risk. Although these scores have excellent performance characteristics as rank-order models, precisely calibrating them to future loss rates carries the same problems described above. In other words, scores accurately reflect risk, but not necessarily the future magnitude of such risk.
Dual-time Dynamics (DtD) was developed by Strategic Analytics to explicitly address these issues. DtD quantifies the account maturation process, changes in originations quality, management impacts, seasonality, and response to the economic environment for retail portfolios. By replacing the past macroeconomic environment with a scenario for the future environment, this approach leads naturally to robust, scenario-based forecasts. DtD runs without need of account-level information, operating instead on segment-level data.
With DtD, we know that the portfolio maturation process, seasonality, and marketing plans (originations quality and management impacts) are statistically predictable using the proper nonlinear algorithms. The great unknown is the future economic environment. Because our process has quantified the past outside (or "exogenous") environment for the portfolio, a natural solution is to use Monte Carlo Simulation to generate a very large number of future exogenous curves representing the portfolio's response to future macroeconomic changes. Importantly, our version of Monte Carlo Simulation is conducted preserving the dominant structures in the exogenous curve: auto-correlation (trend-length), distribution of changes, and seasonality (see inset). To properly capture these effects, SA utilizes a more sophisticated version of Monte Carlo Simulation than typically encountered.
After generating many possible future exogenous curves, forecasts of future revenue and losses are generated via the simulation engine in LookAhead™. These forecasts are collected to create probability distributions of revenue and losses. From these distributions, levels of expected and unexpected losses can be set easily, and economic capital computed. Dual-time Dynamics provides a high-precision approach to predicting future losses while capturing the spirit of economic capital—preparing for the unexpected. |
Top of page