Print
E-MailColleen Honigsberg is an Associate Professor of Law at Stanford Law School and Matthew Jacob is a PhD candidate in the department of economics at Harvard University. This post is based on their recent paper, forthcoming in the Journal of Financial Economics.
BrokerCheck, a public-facing website maintained by financial regulators, provides employment and disciplinary history for all US-registered securities brokers in an easy-to-search format. There are many indications that the website is well utilized and provides important information that can be used to predict broker misconduct (Qureshi and Sokobin, 2015; Egan, Matvos, and Seru, 2019). However, BrokerCheck has become subject to increasing scrutiny for the controversial practice, known as “expungement,” which allows brokers to remove select allegations of misconduct through an arbitration process.
Our paper Deleting Misconduct: The Expungement of BrokerCheck Records, forthcoming in the Journal of Financial Economics, provides the first academic analysis of BrokerCheck’s expungement process. From 2007 to 2016, we identify 6,660 expungement requests, suggesting that brokers request to expunge 12% of the allegations of misconduct made by customers and firms. Of the expungement requests that are adjudicated on the merits, over 80% are successful. We show that expungements significantly predict future misconduct; brokers with prior expungements are 3.3 times as likely to engage in new misconduct as the average broker. Further, using an instrumental variable based on the random assignment of arbitrators, we present evidence that brokers who receive expungement are more likely to reoffend than brokers who are denied expungement. We also show that successful expungements improve long-term career prospects.
Our paper thus provides important evidence on a longstanding policy debate over expungement (Grassley & Reed, 2013; Lipner, 2013; Edwards, 2017a; Edwards, 2017b; Warren, 2019). On the one hand, FINRA states that expungement is an extraordinary remedy that should be granted only if the allegations against the broker are erroneous, the broker was not involved in the misconduct, or the claim is false. If the process functions as intended—meaning that the expunged information is inaccurate or otherwise does not reflect the broker’s conduct—removing the information has many benefits. Doing so should (1) improve the accuracy of the BrokerCheck database, (2) incentivize brokers to maintain a clean record by sharpening the signal between “clean” and “misconduct” brokers, and (3) allow regulators, firms, and consumers to perform more effective monitoring, as they can better predict the brokers likely to commit misconduct.
On the other hand, if brokers are abusing the expungement process, as many state regulators have alleged, removing misconduct from BrokerCheck will reduce the utility of BrokerCheck and hamper the effectiveness of FINRA’s disciplinary regime. Moreover, if the expungement process is abused, behavioral literature suggests that it could lead to an increase in socially undesirable behavior, as studies have found that a higher incidence of unethical behavior is likely to occur if prior unethical decision making is rewarded (e.g., Hegarty and Sims, 1978).
Therefore, a key issue in understanding the impact of expungement is the relation between expungement and broker recidivism. At a descriptive level, as noted above, successful expungements predict future misconduct. However, this simple ordinary least squares (OLS) regression does not address whether expungement affects recidivism, as many of the characteristics associated with successful expungements are also likely to be associated with a lower likelihood of recidivism. To answer the causal question of whether expungement affects recidivism, we use an instrumental variable (IV) analysis where our instrument is based on the randomized list of arbitrators assigned to adjudicate the expungement request. The arbitrators on this list are chosen by an algorithm, and FINRA states explicitly—and has undergone an audit to confirm—that the algorithm selects the initial list of arbitrators randomly (subject only to geographic limitations). Conceptually, our instrument is the relative leniency of the average arbitrator on this randomly generated list.
Our analysis provides evidence that successful expungements increase recidivism. The two-stage least squares (2SLS) results, which exploit plausibly exogenous variation in expungement from the random assignment of FINRA’s arbitrator list, show that expunged brokers are more likely to reoffend. Notably, this result appears to be driven by repeat expungements—in other words, successful expungements cause an increase in future (successful) expungements. With full controls, the 2SLS results show that the marginal expunged broker has 0.16 to 0.20 more years with successful expungements than brokers who are denied expungement. Additional robustness tests provide evidence that the increase in successful expungements is jointly driven by an increase in expungement requests and a greater likelihood of success.
Our paper further examines the career consequences of expungement. Our descriptive analysis suggests that brokers who receive a successful expungement are more likely to remain with their firm and, conditional on leaving the firm, to be rehired by another brokerage firm. Similarly, the results from our IV analysis using the leniency of the mean arbitrator show that, relative to those denied expungement, marginal expunged brokers are 21 percentage points less likely to separate from their firm (with full controls). Marginal expunged brokers also remain in the BrokerCheck database for a greater number of years (meaning they remain employed as registered brokers). Therefore, there is evidence that expungement improves career outcomes, and further tests show that this effect is strongest for brokers who remove the weakest claims of misconduct (those expungements classified as “erroneous” under FINRA Rule 2080).
FINRA has recently proposed updated rules to govern the expungement process, and our analysis suggests several avenues for reform. Further, in addition to our policy contribution, we provide two main academic contributions. First, our work contributes to literature on the use of personal brands and reputation as a regulatory tool. Regulators commonly require financial advisors to disclose substantial personal information to regulators, much of which is then made available online in accordance with a market-based theory of deterrence: public disclosure will allow markets to weed out the “bad actors.” Regulators frequently post allegations with relatively limited verification, such as customer complaints, because these allegations have predictive power. But, although these allegations have predictive power, on average, there are false positives as well. Thus, one tension with this approach is the degree to which regulators should verify disciplinary information ex ante versus allowing individuals to remove information ex post through expungement. Our study suggests that, at least at present, regulators should rely more heavily on ex-ante verification rather than ex-post expungement. Second, our work provides other academics with increased visibility into the expungement process. We show that the selection of brokers who request, and are granted, expungement is not random, and we suggest other academics should consider whether the non-random removal of misconduct data has implications for their research.
The complete paper is available here.
