Monday, May 23, 2016

Dealing with residuals and working on narratives in finance

The first law of pharmacology - every drug has two effects the one you know and the other one.

The simple phrasing of this "law" show a deep understanding of measurement and uncertainty. The same law could be applied to any finance model. There is the amount explained or known and everything else; the effects we know and have measured from the past and the rest. One of the failures of quant modeling is that the effects you know can only describe a small portion of the variance. The other effects are very big. All of the action and all of the risk is with the other effect, the portion that cannot be explained.

Most investors implicitly know this. This is one of the reason why there is a high degree of skepticism about relying only on models.  This is why telling stories about what is going in the markets at any particular time is so important. The models will get you close to an answer and provide guardrails on what is possible in the market but there still need to be judgement concerning a particular situation.

Now, the type of manager is a function of how they want to deal with these two effects, the explained from a model and the unexplained. The quant will say that even though there is a high degree of unexplained variation, the model is more dependable than any ad hoc decision process. The quant concludes that his ability to provide value on the unexplained is low and may actually take away from the signal generated from the explained.

The discretionary manger will say that the model explains too little of the variation and developing a current narrative is more reliable. Models may provide some guidance but context is most important for making a decision. The discretionary manager will develop a story for "explaining" the current market situation and suggest a reason for opportunity. There is nothing inherently wrong with the approach and it can be based on data and facts.

The key money management question is how to develop effective narratives for the unexplained portions of asset returns. How should these narratives be constructed in a manner that is repeatable? How can decisions be framed through using narratives? If money management decisions are to be consistent, there has to be a process for generating the unexplained narrative.

No comments: