Academic journal article Economic Perspectives

Early Warning Models for Bank Supervision: Simpler Could Be Better

Academic journal article Economic Perspectives

Early Warning Models for Bank Supervision: Simpler Could Be Better

Article excerpt

Introduction and summary

Capital adequacy has long been central to the regulatory oversight of banking systems around the world. Capital is crucial to bank safety and soundness, because it represents the cushion available to financial institutions to withstand unanticipated losses. By monitoring capital levels, supervisors seek to predict which financial institutions are most likely to be at risk if subjected to an earnings or asset quality shock. In 1991, the passage of the Federal Deposit Insurance Corporation Improvement Act (FDICIA) emphasized capital levels as a key benchmark to use in determining appropriate supervisory interventions to aid ailing financial institutions. Intervention when an institution is beginning to experience problems may allow it to avoid failure.

Over the past two decades, various off-site monitoring systems have been created to identify developing financial problems at banking institutions between on-site examinations. Supervisors use the output from these monitoring or early warning system (EWS) models to determine which organizations need increased supervisory scrutiny, identify specific areas of concern, accelerate on-site examinations of institutions showing financial deterioration, and allocate more experienced or more specialized examiners to institutions with financial problems.

The current models used by bank regulators focus on predicting either CAMELS (1) downgrade or bank failure. (2) In this article, we develop EWS models that focus on identifying banks that will have inadequate capital in the following year. Specifically, our models predict banks with an early stage of capital distress, with a primary capital to assets ratio falling below the 5.5 percent minimum capital adequacy standard (the relevant capital standard for this period). Earlier identification of capital inadequacy would enable supervisors to identify firms at risk and manage timely supervisory interventions.

Based on samples of banks in the late 1980s and early 1990s, we test our EWS models empirically using financial and economic data for individual banks. We chose 1988-90 as the sample period, rather than a more recent period, in order to have a sufficient number of problem banks in the sample. Also, since most troubled banks are those with assets of less than $1 billion, we focus on these banks in our study. We also exclude banks with less than $300 million in assets.

Our objective is to develop a model that predicts one of two states--capital adequate versus capital inadequate--where the latter state represents capital levels that fall below a minimum threshold employed by bank supervisors during this period. Although we use a well-recognized regulatory threshold for adequate capital, our main objective is to examine an early stage of financial distress, rather than regulatory compliance with capital standards. (3)

Our choice of the capital to asset ratio as a plausible proxy for the early onset of financial distress is supported by previous research. Estrella, Park, and Peristiani (2000) examined the relationship between different capital ratios and bank failure and found that the simple capital to assets ratio (leverage ratio) predicts bank failure as well as more complex risk-weighted capital ratios over one-year or two-year horizons. In addition, they recommended using the simple capital ratio as a tool to provide a timely signal of the need for supervisory action.

Our empirical results reveal that banks with impending capital deficiency are much different from other banks in terms of their financial condition. Also, our EWS models are able to detect the early onset of financial distress in commercial banks one year in advance with a reasonable degree of accuracy. Importantly, we compare three EWS models--a simple logit model that includes only the lagged capital ratio and lagged change in capital ratio, a more complete logit model, and a non-parametric trait recognition analysis (TRA) model. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.