At a meeting last week, the Federal Reserve Board voted to put out for comment an innovative proposal on regulating the risks posed by banks' trading operations. The proposal, drawn up by Fed economists Paul Kupiec and James O'Brien, drew praise from Fed Chairman Alan Greenspan and Vice Chairman Alan Blinder - and surprised bankers and Fed watchers. Here are excerpts:
Techniques for measuring and managing market risk have been progressing rapidly in recent years, and further advances can be expected in the future. It is important that capital requirements provide incentives for such advances and that these requirements remain compatible with best practices as they evolve.
Recognizing the need for further evolution in supervisory approaches to capital adequacy, the board is requesting comment on a novel approach that has been termed the "precommitment" approach.
While in theory this approach might offer significant advantages over existing alternatives, many of the practical details have not yet been worked out.
The precommitment approach draws its inspiration from the economic literature on "incentive-compatible" regulatory schemes.
As in the internal models approach to market risk capital requirements that the board has proposed, the regulatory objective is to require a bank to maintain sufficient capital to cover potential losses in its trading activities from all but the most extreme price movements.
The internal-models approach seeks to ensure compliance with this objective by standardizing the parameters under which a bank would calculate the value at risk of its trading portfolio and then applying a multiplication factor to each bank's calculated (value at risk), in part to cover potential losses over longer horizons.
By contrast, the precommitment approach would seek to induce banks to meet the regulatory objective by providing them with a common set of economic incentives.
Specifically, in the precommitment approach a bank would specify its desired amount of capital for supporting market risks and would commit to manage its trading portfolio so as to limit any cumulative trading losses over some subsequent interval to an amount less than that capital allocation.
The length of the interval would be established by the bank's regulator, based on the regulator's ability to monitor losses from the bank's trading activities and, if necessary, to force reductions in the size of the bank's open positions.
To ensure that the bank committed an amount of capital commensurate with the risks in its trading portfolio and its capacity to manage those risks, the regulator would need to provide appropriate incentives in the form of economic costs or "penalties" for failing to limit losses to less than the capital commitment.
Given these costs, the bank's choice of a capital commitment would be based on a self-assessment of its capabilities to measure and control the risks of its trading activities.
The bank's choice of a capital commitment for market risk could be subject to review by supervisory authorities. Bank management could be expected to explain how cumulative losses would be contained within the amount of the commitment.
This necessarily would require documentation of how internal models are used to measure risks, how limits are applied to the measured risks, how compliance with limits is ensured, and how management would respond to unanticipated losses.
It would be important to emphasize, however, that any supervisory review of the commitment would in no way diminish the bank management's responsibility for setting aside adequate capital to cover its market risks.
An attractive feature of the precommitment approach is that it would underscore the responsibility of bank management for maintaining adequate capital, even if the amount needed exceeds what otherwise might be regulatory minimum requirements.
The key to the effectiveness and applicability of the precommitment approach is the specification of the penalties that would result from a failure to limit trading losses to an amount less than the commitment.
These penalties could take various forms. Fines (monetary penalties) would be especially effective in creating appropriate incentives because of their transparency. (U.S.-insured banks could be required to pay any fines into the Bank Insurance Fund.)
As an alternative to fines, supervisors could impose punitive capital charges.
The severity of fines or capital penalties could be reduced if they were accompanied by supervisory sanctions, such as restrictions on future trading activity.
For the precommitment approach to be credible, banks would need to be reasonably certain that supervisory authorities would impose the specified penalties when losses exceed the commitment.
The certainty of the penalty would strengthen the incentive for the bank to make the initial capital commitment commensurate with the supervisor's desired coverage of potential losses.
Nonetheless, supervisors would need to reserve the right to suspend the penalties in the event of extreme price movements that reflect macroeconomic instability.
This would help ensure that banks could continue to provide liquidity to markets following such stressful episodes.
Market forces might also be utilized to provide banks with incentives to allocate adequate capital.
If the capital commitment were publicly disclosed, the reporting of losses in excess of the commitment not only would imply that supervisory sanctions had been imposed on the bank, but could also cast doubts on the effectiveness of the bank's risk management capabilities. Together, these factors could adversely affect its share price and its funding costs.
For this reason, some banks might actually be tempted to commit more capital than is necessary to meet regulatory objectives.
However, this tendency toward conservatism would be tempered by fears that an excessive capital commitment would cause the public (including stock analysts and rating agencies) to overestimate the riskiness of the bank's trading activities. Thus, market forces could be harnessed to induce banks to make realistic capital commitments.