Analyzing securitized products typically entails two steps: collateral analysis, and structure analysis. In either case it is necessary to access very detailed data in order to perform any analysis. Of note, European products are less advanced than US products from this standpoint, as loan-level data is almost systematically available in the US, but more scarce in Europe.
The ideal approach for collateral analysis includes creating a loan-level database that tracks the performance of each and every loan in the securitized product universe. Such a database needs to store over a billion lines (each loan, at each point in time). Specialized vendors offer the data in raw form, and it generally takes a few months to put in place a efficient system to exploit them.
Leveraging a loan-level database, it is possible to then build econometric models as well as an infrastructure allowing to design and run particular scenarios. These scenarios are used to create loan-specific projections.
Structures are usually represented through a structuring language that is generic enough to account for the features of most existing securitization deals. Specialized vendors offer analytics solutions that allow users to programatically obtain bonds cash flows, given the specific assumptions relevant for each underlying loan.
The pricing and risk infrastructure manages the interface between loan-level information, econometric models, and the cash flow engine. Based on these results, an analytics routine can extract the necessary data (present value, average life, ...).
These types of models then allow detailed risk analysis. Securitized products risks can be grouped in three categories: fundamental risk, driving future collateral cash flows, structural risk, driving potential changes in the way collateral cash flows are distributed within a deal, and valuation risk, driving the magnitude of the discount applied to these products by the market as a function of perceived risk.
- Fundamental risk can be analyzed with econometric models, linking certain macro or micro factors to collateral performance. For example, residential real estate prices, interest rates, or refinancing costs are important inputs in such a model.
- Structural risk is normally not modeled in an econometric fashion but rather represented with specific scenarios. For example, although many deals allow for a maximum 5% of loans to be modified Government programs call for a much higher effective rate, which results in an unpredictable number of loans being modified. In this case, structural risk can be measured by particular stress scenarios driving the importance of loan modifications.
- Valuation risk measures are based on historical data, such as historical prices for relevant indices, and stress or correlation scenarios.