How Widespread and Predictable is Stock Broker Misconduct?

Abstract

In this paper we reconcile widely diverging recent estimates of broker misconduct. Qureshi and Sokobin report that 1.3% of current and past brokers are associated with awards or settlements in excess of a threshold amount. Egan, Matvos, and Seru find that 7.8% of current and former brokers have financial misconduct disclosures including customer complaints, awards, and settlements. Qureshi and Sokobin arrive at their low estimate by excluding 85% of all brokers, including those brokers most likely to have engaged in misconduct. Applying Qureshi and Sokobins restrictive definition of potential misconduct to all brokers, we find that misconduct is much more widespread.

Qureshi and Sokobin find that the top quintile of brokers sorted by estimated likelihood of being subject of a customer complaint contains 55% of the brokers subsequently subject to complaints. We demonstrate that more sophisticated data mining techniques can sort brokers so that the highest risk quintile contains 75% of the brokers subsequently subject to complaints.

We also evaluate Qureshi and Sokobin's claim that FINRA's BrokerCheck website provides helpful information to investors seeking to avoid bad brokers and answer the question posed by Egan, Matvos, and Seru: If BrokerCheck data can predict broker misconduct, why don't investors use that data to protect themselves? We find that BrokerCheck is worthless in its current hobbled form, but could easily be modified so that investors could protect themselves and market forces would substantially reduce broker misconduct.

Keywords: BrokerCheck, Investor harm, Disclosures, CRD, FINRA, Predictive modeling

JEL Classification: D18, K20, K22, G2, G24, G28,C10