Methods to Estimate Losses Using Linear Regression Analysis: This Article Outlines How Linear Regression Analysis Can Be Used to Calculate the Allowance for Loan and Lease Losses

Article excerpt

Nature and Purpose of the Allowance for Loan and Lease Losses (ALLL) (1)

The ALLL represents one of the most significant estimates in an institution's financial statements and regulatory reports. Because of its significance, each institution has a responsibility for developing, maintaining, and documenting a comprehensive, systematic, and consistently applied process for determining the amounts of the ALLL and the provision for loan and lease losses (PLLL). To fulfill this responsibility, each institution should ensure controls are in place to consistently determine the ALLL in accordance with GAAP, the institution's stated policies and procedures, management's best judgment, and relevant supervisory guidance.

As of the end of each quarter, or more frequently if warranted, each institution must analyze the collectability of its loans and leases held for investment and maintain an ALLL at a level that is appropriate and determined in accordance with GAAP. An appropriate ALLL covers estimated credit losses on individually evaluated loans that are determined to be impaired as well as estimated credit losses inherent in the remainder of the loan and lease portfolio. The ALLL does not apply, however, to loans carried at fair value, loans held for sale, off-balancesheet credit exposures (e.g., financial instruments such as off-balance-sheet loan commitments, standby letters of credit, and guarantees), or general or unspecified business risks.

The ALLL consists of two components, Accounting Standards Codification (ASC) 450, formerly known as Financial Accounting Statement (FAS) No. 5 (Accounting for Contingencies), and ASC 310, formerly known as FAS No. 114 (Accounting by Creditors for Impairment of a Loan).

The "classified" or "bad" portfolio is analyzed for impairment on a loan level basis in accordance with ASC 310. For loans determined to be impaired, a specific loan loss reserve is calculated. For collateral-dependent loans, the reserve is typically based on the fair market value of the collateral (as-is appraised value less costs to sell); otherwise, the reserve is based on either observable market transactions or a net-present-value discounted cash flow analysis. If it is determined that a loan is impaired but there is no dollar impairment, the loan remains in the ASC 310 portion of the ALLL. If a loan is determined not to be impaired, it is migrated back to ASC 450 and included in the appropriate pool.

The purpose of the ASC 450 calculation is to estimate the dollar amount of potential losses embedded within the "unclassified" portfolio--that is, the "pass" or "good" portion of the portfolio. Loans are aggregated into homogenous pools that exhibit similar risk and performance profiles. Each pool is analyzed separately. Regulators typically require banks to use their actual annualized (pool level) historical loss data covering a two- or three-year period (preferably three) and will consider weighting different periods more than others if there is logic to support it.

To round out the analysis, banks are required to include internal and external metrics. Other than in general terms, regulators do not provide specific guidance on how to derive or apply these metrics. However, regardless of which metrics are used, they must be logical and have a causal influence on future potential losses. Moreover, the end results, when taken as a whole (ASC 450 and ASC 310), must be within regulatory guidelines.

This article specifies the use of historical data as a part of the ASC 450 calculation. It examines the analytical procedures currently in use and outlines how linear regression analysis can be used. The regression model combines historical loss data with external metrics--in this case, U.S. unemployment and FDIC loan loss data--to forecast future losses. The purpose of this article is to help banks explore the concepts described here using their own data and discuss their results with both their regulators and accountants. …