The problem with quant models

I have a bias and it is with the use quantitative investing strategies. I like the idea that there is no commentary such as "I feel the market is going higher". I like the idea that back-tests can be done for any model and there is clear way to measure the success of a model. However, quant modeling has its own set of biases for which special care should be taken.

Here are three. Similar comments in the context of numbers based models have been made by Aswath Damodaran, the NYU professor, in his blog.

The illusion of precision - There is a precision with using a model and focusing on the numbers but as anyone who has worked with numbers knows, there is error with any model and with the inputs used for a model. Economic data is subject to revisions. Good models of risk premia can only explain a small portion of the variation in the data. There is a lot of imprecision with models and it takes skill to appreciate and understand those limits.

The illusion of objectivity - Some will say that you just look at the cold hard facts, but no model is truly objectivity. A model incorporates the biases of the model-builder. This cannot be helped. Every choice made by the model-builder includes his view on how markets operate.

The illusion of control - Markets change. There are structural adjustments and changes in the sensitivity between the market and variable inputs. Models are inherently backward-looking so they are never in control of the future. 

Quantitative modeling may avoid some behavioral biases, but that does not mean they are without there own set of biases.