Sources of Uncertainty

Definition

Sources of Uncertainty in the context of modelling a Credit Network is the fundamental set of random variables that captures the nature of uncertainties (risks) that are modelled within the framework

Classification

Outcome Variables

Variables capturing the uncertainty of credit system outcomes are of fundamental importance as they help focus on a manageable set of relevant information. Such variables consist of:

Credit Default / Migration Indicators for individual assets (or Credit Default Rates for pools of credit assets)
Uncertain Credit Recovery Rates for individual assets / pools
Uncertain Credit Exposures
Competing Risks such as prepayments

Factor Variables

Risk Factor variables (risk drivers), which in turn are classified as

Idiosyncratic Risk factors or
Macro (systemic) Risk Factors e.g., uncertain Interest Rates

Notation

We denote (abstractly) all information that is used for the formation of expectations as $F_k$ . This is an evolving (growing) set. The realization of uncertain events as time progresses adds information that is used to update future expectations.

The range of credit instruments and their sensitivities to dynamic factors $M_t$ dictates the precise set of variables to include. Given the impact of model assumptions on the outcomes, models should be as simple and transparent as possible so that there is ownership of parameters involved. This suggests a small set of core macro variables $F_t$ as a correlated system, together with the required set of satellite models for producing auxiliary factor variables $A_t$ . Hence we split the set $M_t$ into $(F_t, A_t)$ .

The following is a breakdown of how different types of information enter into the formation of expectations. In the broadest case the information set includes both idiosyncratic information $I^i_t$ specific to an asset / obligor and macro information $M_t$ that is relevant to a broader set of assets. $(F^i_k = \{I^i_t, M_t\})$ .

Issues and Challenges

In the presence of Feedback Effects the separation of information sets may not be very clear