Financial Risk Management

CHAPTER 20Understanding the Root Causes of Poor Conduct

In this final chapter, I will draw on my consulting experience with numerous banks to describe what they need to do to detect misconduct early enough to do something about it. The dataset used in what follows was designed and built jointly with an investment bank I advise (as a contractor within a team of consultants), to redesign its process for managing conduct risk.¹

The dataset records 157 conduct risk indices of traders and sales. As shown in Figure 36, most risk‐takers lie in the range from 56 to 75, and the lowest conduct performers only represent 4% of the distribution. At first sight, this bank reports quite a sound conduct risk situation. But what about the 25% of those employees who do not behave as expected, or are just above the 50% borderline? What about the impact on the bank's conduct performance? It is difficult to say that it is positive. This is where a more detailed analysis is necessary to better understand the root causes of poor conduct.

CLUSTERING RISK‐TAKERS

To get to the bottom of this non‐satisfactory conduct performance, we used a cluster methodology to uncover potential hidden realities among employees. Clustering is a method of unsupervised machine learning. Its goal is to partition the observations in a dataset (here, we have 157 risk‐takers) into groups or clusters so that the pairwise dissimilarities between those assigned to the same cluster tend to be smaller than those in different ...