Evaluating the Basle Guidelines for Backtesting Banks' Internal Risk Management Models

Article excerpt

The 1996 Amendment to the Basle capital accord to incorporate market risks constitutes a breakthrough in the determination of capital requirements. Rather than dictating these requirements through a uniform supervisory approach, banks are allowed to use their own, internal models for computing the capital required. In order to mitigate moral hazard problems and stimulate banks to use adequate internal models, the models must be subjected to a backtesting procedure. If a model produces too many incorrect predictions, increased capital requirements result. This paper provides an evaluation of the current internal models approach in conjunction with the proposed backtesting procedure. In particular, using a stylized representation of the present supervisory framework, we investigate whether banks are provided with the right incentives to come up with the correct internal model. We find that, under the current regulatory framework, banks are prone to underreporting their true market risk. A much stricter penalty scheme is required in order to align banks' incentives with those of the supervisor. We check the sensitivity of our results to changes in the length of the planning horizon, portfolio risk, time preferences, risk attitudes, and the distribution of financial returns.

CAPITAL REQUIREMENTS play a major role in the banking industry [see, for example, Berger, Herring, and Szego (1995) for an overview]. Following Estrella (1995), we can distinguish between market capital requirements and regulatory capital requirements. Market capital requirements serve to reduce agency problems, balance the benefits of tax evasion versus increases in financial distress costs, and reduce transaction costs if new investment capital is needed. Regulatory capital requirements, by contrast, aim at protecting the government (and ultimately the taxpayers) against financial distress costs and guaranteeing the soundness and stability of the financial system. From this regulatory or supervisory perspective, capital serves as a cushion to absorb part of the effect of adverse economic developments on financial institutions.

A key statistic in formulating capital requirements is the captial ratio, that is, the ratio of capital to total assets. A typical measure of capital is equity, possibly increased by the level of subordinated debt; see Berger, Herring, and Szego (1995). They show that over the past 150 years U.S. capital ratios have gradually declined from around 40 percent in 1850 to about 6-8 percent from 1940 onward. Several sharp drops in capital ratios follow major changes in the regulatory framework. For example, the creation of the Federal Deposit Insurance Company (FDIC) in 1933 resulted in a decrease in capital ratios of about 50 percent over a ten-year period.

It is well known that flat premium deposit insurance schemes like that of the original FDIC can cause perverse incentives for banks, inducing them to engage in more risky activities (see, for example, John, John, and Senbet 1991). Boot and Thakor (1991) show that these effects may be mitigated by certain off-balance-sheet activities. The historically low capital ratios since the creation of the FDIC and the growing awareness of banks' misaligned incentives have spurred the introduction of new, risk-based capital requirements. The prime breakthrough is the Basle accord of 1988. Following this accord and its implementation in 1990, banks are required to hold a capital reserve between 0 percent and 8 percent for each transaction, depending on its riskiness, for example, government (0 percent) versus corporate (8 percent) loans. The accord also postulates capital requirements for certain off-balance-sheet activities. Avery and Berger (1991) and Berger, Herring, and Szego (1995) argue that capital ratios have generally increased due to the new risk-based capital standards. Others comment that the requirement of 8 percent is ad hoc and insufficient to meet capital needs in times of stress (see Bradley, Wambeke, and Whidbee 1991). …