A rating is a model output, and the model rests on a correlation assumption

The idea

A credit rating on a structured product is not a judgement about each loan inside it. It is the output of a model, and the model's most important input is an assumption about how the loans will default together. If mortgages across different cities were assumed to fail independently, a pool of them looked safe, and the senior slices could be rated as if they were government bonds. When house prices fell everywhere at once the loans defaulted together, the assumption broke, and the ratings broke with it. The lesson is not that ratings are useless. It is that any rating or score is only as good as the correlation it assumes, and that the assumption is the thing to ask about.

Ask what a rating assumes about defaults happening together before trusting what it says about the pool.

How to use it

Asked what got you interested in markets, name the correlation assumption and what you read next rather than the film. In a credit interview, when shown a tranche structure, ask how losses in the pool are assumed to be distributed before commenting on the rating.