The 2008 financial crisis triggered a series of interventions by regulators across the world and has driven the continuous evolution of stress testing requirements and practices since. With changing regulations in mind, it’s important for institutions to have effective Stress Testing programmes in place. Our risk specialists explore more in our latest stress testing blog, with a particular focus on the consumer credit sector.
Stress Testing exercises are designed to assess the resilience of a financial institution to a macroeconomic shock or a contraction in the financial markets, quantifying the implications of plausible but extreme events on the institution’s profits, capital, value or liquidity.
There are many different components to a successful stress testing programme and some will be specific to consumer credit products, especially regarding the choice of the appropriate modelling of macroeconomic effects.
1. Data management and infrastructure creation
Having a strong data infrastructure is the first key step to any successful stress testing programme. Good data is the backbone of any successful analytical exercise. It’s a fundamental part of the modelling process but is often poorly collected and organised, especially relatively young organisations who have not had time to invest in their data infrastructure or more established organisations that have not fully recognised its true importance.
Once the data is available, its quality needs to be validated before moving to the modelling phase.
Important validation steps include:
- Assessment of data suitability
- Analysis of missing values, outliers or ‘abnormal’ periods
- Analysis of basis descriptive statistics and distributions
- Reconciliation with existing MI packs if possible
- Documentation
2. Modelling methodology
In order to quantify the relevant drivers of losses, the choice of the appropriate modelling methodology is fundamental. The measure of interest to be used as target variable depends on the product, but in the stress testing context it can be expressed in general terms as the Expected Loss deriving from the exposure. Different approaches can be applied to quantify the values of losses in case of default. Typically, the choice depends on the asset type, the nature of the exposure, the guarantees or collateral in place and the different macroeconomic variables that are likely to be significant drivers of loss.
A fundamental factor influencing the model choice is the amount of data available. Portfolios with large amounts of data available can benefit from more sophisticated modelling techniques while younger portfolios with less data available will need to be modelled applying more pragmatic approaches typically using proxy series or comparable data.
3. Incorporate macroeconomic effects and scenario effects
Incorporating macroeconomic effects is a fundamental part of stress testing modelling as the main objective is to quantify the impact of the economy on portfolio losses. Economic effects are measured historically incorporating macroeconomic variables in PD, LGD and EAD models and the different economic scenarios are then applied using the estimates relationships to produce based and stressed forecasts.
Typically, economic effects are embedded in the estimation of PD models and in measuring losses within LGD models, but in some circumstances, particularly on revolving exposures, economic impacts can be identified also in the estimation of exposure at default.
Despite variations in macroeconomic drivers, based on analysis of historical statistical relationships, it is possible to identify which drivers play the biggest role in explaining the losses of an asset class. However, the future may be different from the past. The magnitude of the impact could vary as the economic landscape evolves and the relationship between macroeconomic variables could change considerably.
For this reason, it is important to always use expert judgement on portfolio trends forecasts or other specific knowledge about the portfolio or wider business plan and future strategies.
4. Model validation and oversight
Model validation is employed to validate the model by deploying it on a sample outside the build, be that either a ‘hold out sample’ or on a different timeframe. Different measures of performance are available to judge the results, depending on the modelling context and the objectives.
In a time series model, which is typically the framework where stress testing models are built, it is standard practice to leave part of the data available on the side (if possible) during the estimation phase to test the ability of the model to predict the target variable on new data ‘unknown’ to the estimation sample.
5. Model execution
Interaction between business units
Stress testing involves all risk areas in one institution, it does not concern only credit risk, but liquidity risk, market risk and operational risk. It requires the coordination of different areas, some of which have traditionally operated separately such as risk, treasury, corporate and financial planning.
The importance of planning
Having an efficient programme management in place is fundamental for a successful stress testing execution.
Regulators have highlighted on several occasions how specific resources should be dictated to stress testing and have reiterated this point in relation to the additional analytic effort required to assess the risks emerging in the consumer credit sector.
6. Management and Governance
It is imperative that a firm’s senior leadership team influences every aspect of the stress testing process. The results are an essential element of any ICAAP and are important in identifying, assessing and managing key risk drivers in any balance sheet. The influence and control exerted by senior leaders in relation to stress testing and wider capital management is placed under heavy scrutiny by regulators.
Stress testing modelling is still largely perceived as a one-off exercise. The regulators intend the process and approach to become a credit risk tool embedded in a firm’s usual business activity. The economic and financial landscapes are continuously evolving but there is no reason to assume that stress testing and ICAAP models should be rebuilt every year if a portfolio has not changed substantially.
A formal monitoring approach can assess the performance of each model on an ongoing basis. Firms that use the IRB approach are already required to have formal monitoring of rating systems and models as a condition of IRB permission. An established framework can be adapted to incorporate stress testing.