The risk-return tradeoff is an important feature of the financial system because riskier assets usually command higher returns, as a compensation for the higher risk. This compensation for risk is referred to as the "risk premium". Recognizing the sources of risk that are worth the premium and determining the appropriate amount of this premium is far from a straightforward task. Here is where financial models come in. Based on common knowledge and some axiomatic principles, financial models attempt to identify the priced risks and also attempt to provide suitable pricing of those risks.

Beyond pricing risk and pricing risky assets, financial models are used to determine credit quality of financial instruments (e.g., loans and other obligations), individuals, companies, institutions, nations. In this application, an attempt is made to quantify the ability to pay, the probability of default and the loss given default. This is important for informing various regulatory and risk-management decisions. Yet other financial models are used for forecasting and assessing most likely, mean, worst-case, best-case future cash flows, costs, and other factors that may affect today's decision making.

The more sophisticated the financial system, the greater the need for financial modeling. This, however, exposes the financial system to yet another level of risk, known as "model risk". Because financial models are so essential to decision making at all levels, and because the models are often far from simple, any inappropriate or unreasonable feature of a model may cause incorrect decisions to be made, and severe material consequences and large financial losses may result. This is particularly true because many basic assumptions that lay the foundation of any model are often subjective to the model builder. This potentially opens the door to model manipulation, where the underlying principles of a model may be adjusted to produce the results desired by the decision makers that use the models to justify their decisions. Some instances of this have occurred in the recent past.

The US financial crisis in the first decade of the XXI Century is case in point. It was characterized by an environment in which much lending was done at interest rates that were too low to compensate for the risks underlying the loans. Such low interest rates were often dictated by the underwriters', lenders', and securitizers' desires to increase the volume of business and to earn the market share; so, underpricing the loans was one way to achieve this goal. Additionally, the benefits of risk sharing and securitization of loans were overstated, and the risks of highly correlated defaults across different loan classes were considerably understated. Sometimes, financial models were "calibrated" to produce results that could justify such managerial decisions. This, clearly, reduced the objectivity of the resulting financial models.

All these concerns have lead to the rapid growth of the sub-industry of model validation, especially subsequent to the financial crisis. Many financial institutions now have independent department devoted to Model Validation and Model Audit. Regulatory agencies have their model validation teams, as well. This is the area of the financial industry that experiences rapid job growth, but this is also the area that requires highly educated, skilled, and experienced talent, knowledgeable in financial and economic theories, statistical and econometric techniques, as well as having common sense and a level of integrity. Such talent is less than easily available and is actively sought after by financial institutions.

Our model validation approach consists of several phases. First, we strive to understand the modeling environment at our client, the purposes of the models and their uses. Then, we review in detail the assumptions and other fundamental principles that make the model work, followed by the review of data sources and the treatment and preparation of the datasets. If the client desires, we can also review the mathematics and computer codes associated with the model. The review of the quality and reliability of model outputs, as well as any additional testing of the model is performed before the final validation report is drafted. Optionally, model process and governance can be reviewed as well.

We have helped our clients with model validation and audit in the areas of probability of default, loss severity, loss forecasting modeling, automatic valuation modeling, counterparty risk, risk ratings, and some other models. We also look forward to performing model validation in other areas, as requested by our prospective clients.