Notes from the Vault

Larry D. Wall
November 2013

Even bankers overwhelmingly agree that the compensation structure in the financial services industry was one of the causes of the financial crisis, according to a 2009 survey of the members of the Institute of International Finance. Not surprisingly, regulators thought so, too and, in response, the Financial Stability Board and U.S. federal bank supervisors responded with guidelines that would reduce the extent to which compensation structures encourage risk taking.

Now that some time has passed since the initial implementation of the guidelines, a recent conference at the Atlanta Fed, Indices of Riskiness: Management and Regulatory Implications, reviewed what has happened and considered where regulatory guidance should go from here. The conference also included several papers related to the issues of risk sharing and exposure to common shocks, which I discuss in a recent macroblog.

Prior to the crisis, incentive compensation at banks was generally designed and implemented to maximize profit and retain talent rather than to affect risk taking. For example, deferred bonuses were typically withheld only if an employee resigned from the bank. The risk controls in most banks took the form of setting limits on an individual's and groups' independent authority to take risk.

The U.S. bank supervisory incentive compensation guidelines seek to change compensation structures so those structures discourage excessive risk taking. The guidelines focus on those employees who individually or collectively could expose their bank to material risk. The principle set by the guidelines is that if these individuals (or groups) are given incentive compensation based on some measure of their revenue production, the amount of compensation should also take account of the risks employees take to produce revenue. This adjustment for risk could take a variety of forms, including some combination of risk adjusting the revenue performance measures and/or deferring part of the payment of incentive compensation with provisions for reducing payment if the risks turn bad.

The changes to banks' payments of bonuses seem a potentially valuable approach to reducing the risk of a future crisis. However, any change to something as central as the incentive compensation of key people in a banking organization will inevitably have unintended consequences. For example, consider the practice of basing an individual's bonus in part on how the overall bank performs. Intuitively, this practice could be justified as having a variety of beneficial intended causes, including that of incenting bank employees to monitor the riskiness of their bank. Still, three of the papers presented at the conference show that such practices are likely to have a variety of other, unintended consequences—both good and bad.

Paul Kupiec's paper "Incentive Compensation for Risk Managers When Effort Is Unobservable" considers the compensation of two types of risk managers, one responsible for underwriting, and the other for loss mitigation. The risk managers make an unobservable decision on how much effort to expend. The bank then offers the risk managers an incentive contract designed to elicit the value-maximizing amount of effort from each type. Consistent with standard principal-agent models, Kupiec's model finds that risk managers should be paid performance-linked bonuses. A further implication that receives less emphasis in the paper deals with the incentive compensation of those charged with loss mitigation. The importance of properly incenting these managers to exert effort is greatest at the time when loan defaults are at their highest. Yet these are likely to be the times when the bank itself is doing poorly, which results in risk mitigation managers' bonuses being reduced if bonus payments are related to overall bank performance!

The paper "Banker Compensation and Bank Risk Taking: The Organizational Economics View" by Arantxa Jarque and Edward Simpson Prescott focuses on payment of incentive compensation to bank loan officers. One innovative feature of this theoretical model is that rather than extrapolating on the basis of a single agent (single lending officer) to the behavior of all officers, the authors' model explicitly considers the implications of multiple lending officers for the riskiness of the bank. This multi-agent focus allows them to address the vital issue of the correlation of loan returns. If loan returns are uncorrelated, their model produces the result that there is no connection between the form of loan officer compensation and bank risk. Given the large set of loan officers in their model, the idiosyncratic losses from individual loan officer decisions are diversified away. Conversely, if returns are perfectly correlated with some common shock, then the bank's compensation decision is simplified. Loans made by one loan officer who exerted appropriate effort should be the same as other lending officers who made similar effort, and they should receive similar compensation.

The more interesting case, however, arises when the returns on the loans by different lending officers are partially correlated under the control of the lending officers. In this case, the common-sense approach of paying bonuses tied to firm performance induces lending officers to prefer projects that are correlated with the common shock. That way, the lenders are more likely to receive their maximum bonus when their individual loans do well. In contrast, if a lender's loans do well when other loans are doing poorly, the lending officer's bonus is reduced because the bank is doing poorly. As a result, once again basing bonuses on overall bank performance (or merely that of other bank loans) is likely to increase exposure to a common shock and hence make the bank more risky.

These first two compensation papers help emphasize the potential adverse consequences of failing to understand how an individual banker's actions affect overall risk. The third paper brings out an arguably beneficial impact of changes in incentive compensation on a seemingly unrelated prudential issue. My coauthored paper with Timothy W. Koch and Dan Waggoner, "Incentive Compensation, Accounting Discretion, and Bank Capital," relates the effect of changing incentive compensation on the ability of banks to absorb losses in future periods. The underlying idea is that the "true" amount of capital available to absorb losses may be greater than, equal to, or less than its accounting capital, based on whether a bank is following conservative accounting policies (using its discretion to recognize expenses early and revenues later) or being more aggressive (using its discretion to recognize revenues late and expenses early). Any given bank's accounting policies will be determined by its senior management, who in turn will be influenced by the policies guiding a bank's incentive compensation.

My coauthors and I show that the part of the guidelines dealing with accounting-based bonuses will discourage banks from following aggressive policies, especially the part dealing with malus (sometimes called clawbacks). Aggressive accounting procedures have the effect of boosting current income, but at the cost of lower future income. Thus, these procedures tend to increase current bonuses based on overly optimistic measures of performance. Malus is a procedure by which the payment of deferred bonuses is reduced when bank accounting reports low or negative earnings. Malus is intended to take back bonuses that were earned before the true riskiness of the bank's investments was revealed. The reduction of bonuses due to malus has the side effect of punishing aggressive accounting by taking (part of) the bonus back when the firm does poorly. In this respect, reducing the amount of deferred bonuses paid to senior management when a bank does poorly has the unintended consequence of encouraging less aggressive accounting procedures.

A fourth paper, "Bank Pay Caps, Bank Risk, and Macroprudential Regulation," by John Thanassoulis illustrates more ways in which compensation rules could have side effects. Thanassoulis observes that many banks' remuneration to staff is a large fraction of the value of their equity. He argues that this high compensation reflects tough competition in the labor market for more talented bankers. Capping total remuneration would offset the tendency of the market for bankers to obtain high compensation. Moreover, he argues that capping the payments is unlikely to lead to problems recruiting high-quality entrants to the labor market since bankers would continue to earn substantial premiums to comparably educated workers in other industries.

A possible side benefit that receives less attention from Thanassoulis comes from his proposal to scale the cap based on Basel III risk-weighted assets. At present, banks can reduce their capital requirements by using their discretion over risk-weighting estimation techniques to prefer those methods that produce lower risk weights and also to overweight their investment in assets whose risk weighting is below their true risk. These incentives to lowball required capital would be at least partially offset if compensation is capped as an increasing function of risk-weighted assets. In this case, management's own incentives would be brought into closer alignment with that of regulators.

However, as prior papers indicate, Thanassoulis's proposal could easily have other unintended and possibly less benign implications.

The discussion of unintended consequences suggests that supervisors should move with some care in implementing new limits on bank compensation. However, the potential for unintended consequences in the financial sector is not limited to incentive compensation. A large part of contemporary prudential financial regulation is being driven by the safety net's unintended consequence of encouraging financial firms to take more risk. Moreover, in almost all cases, these other prudential regulations create their own set of unintended consequences. Thus, the real issue is how to balance the costs and benefits of these various interventions in the financial system while remaining humble about our ability fully to predict and measure either the costs or the benefits.

The conference was organized by Glenn Harrison (Georgia State University's Center for the Economic Analysis of Risk), Jean-Charles Rochet (University of Zurich), Markus Sticker, Dirk Tasche (Bank of England's Prudential Regulatory Authority), and Larry Wall (Atlanta Fed's Center for Financial Innovation and Stability).

Larry D. Wall is the executive director of the Center for Financial Innovation and Stability at the Atlanta Fed. The author thanks Paula Tkac for helpful comments on the paper. The views expressed here are the author's and not necessarily those of the Federal Reserve Bank of Atlanta or the Federal Reserve System. If you wish to comment on this post, please e-mail