Econometric teaching is not focusing on new tools for financial data analysis

I was recently asked to review a new book on financial econometrics for a investment journal. I was disappointed with the direction of the book. It is well-written, detailed, and covers all the important finance topics researched over the last three decades. Unfortunately, it is missing many of the advances in data analytics that are changing how current research is conducted. The old or traditional thinking about econometrics is not serving the investment world well. It has not helped practitioners of finance.

This issue of not focusing on cutting edge techniques and research was further reinforced after reading the notes “The 7 reasons most econometric investments fail” by Marcos Lopez de Prado. 

The greatest advances in data science are occurring in fields outside of finance. Machine learning is being used in finance; however, it is not being taught as foundational for use in econometrics or finance. In more simple terms, there are broad techniques in statistical work that are largely ignored by finance in an attempt to maintain the focus on well-defined linear models. Theory is critical to good science, but given the large amount of data, there also needs to be a broadening of fundamental work in analysis. 

Some of the issues and topics that should be addressed in a new financial econometrics book:

• the high number of false positives with any analysis and the over-reliance on p-values and statistical significance, this has caused the factor zoo we currently face;
• the overuse of the multivariate linear regression model;
• the avoidance of clustering and classification methods;
• an effective introduction to machine learning techniques and neural networks; 
• an increased emphasis on forecasting and model overfitting and specification bias;
• a discussion of unstructured data, hierarchical relations, and sparse datasets;
• a focus on conditional correlation, outliers, and nonlinearities. 

These topics are practical and useful for many finance problems today, yet there is limited time spent on looking outside the traditional econometric toolbox. Of course, there have been significant advancements and greater sophistication of testing today relative to two decades ago. However, my point is increase the focus on using the right tools for the right problem and a linear stable world is not what investors face.