CFO Focus: FASB's Proposed Credit Loss Model

By Taylor Nadauld, Ph.D.

5 minutes

What's likely to change, and what steps should CUs take as a result?

The Financial Accounting Standards Board has proposed changes in the accounting treatment of credit losses--from taking largely a historical perspective toward a practice of being more forward-looking.

We begin our discussion by describing the specifics of the rule change. We then describe the steps a CU might take in calculating the expected loss in a loan portfolio.

As outlined in the Proposed Accounting Standards Update, Subtopic 825-15, current Generally Accepted Accounting Principles standards delay the recognition of credit losses until the loss is considered “probable.” By definition, this standard is backward-looking. This is because a credit loss is not likely to be deemed as probable until a credit has demonstrated some weakness in payment performance.

The proposed change detailed in the Proposed Accounting Standards Update requires that financial institutions “remove the existing ‘probable’ threshold in GAAP for recognizing credit losses and broaden the range of information that must be considered in measuring the allowance for expected credit losses.”

An expected loss standard encourages institutions to evaluate expected losses as a function of current market conditions, historical experience with similar assets, and forecasts of collateral recovery rates. In short, the rule change can be summarized as an attempt to move away from historical, delayed recognition of credit losses toward a more timely, forward-looking expectation of credit losses.

We note that the rule change emphasizes the need to incorporate a broader set of factors in analyzing expected credit losses. In addition to being more historical in nature, traditional credit loss models have taken a micro-centric approach to evaluating borrower’s credit. Key micro-centric loan characteristics include the current credit score of the borrower, the current loan-to-value ratio of the collateral (note the emphasis on the word current), income and debt measures of the borrower, interest rates, and the type of the loan.

A more macro-centric approach encouraged by the proposed change supplements the borrower-specific attributes with measures of the macroeconomic environment in which the loans exist. Broad trends in housing prices and unemployment rates are likely to influence both borrower performance and collateral recovery rates. The rule change encourages CU executives to incorporate these factors into their analysis of expected credit losses.  

Practically speaking, the proposed rule change prompts the question, “How can one create a measure of portfolio expected loss?” One approach is to use an “Expected Loss Model.” To do this, a CU must first make an assessment of the probability that a given loan will actually default, thereby subjecting itself to some potential loss on that specific credit. Second, it must assess the amount of losses that will be sustained on the loan if it does in fact default. The last step is to aggregate the expected losses on each individual loan.

More specifically, this process begins by considering the creation of an estimate that a given loan will default at some point in the future, i.e. its probability of default. A probability of default calculation is quite literally a statistical analysis of the probability—between 0 percent and 100 percent—that a certain loan will, at some point in the future, default. Fairly straightforward statistical models, such as a logistic regression or a Cox Proportional Hazard Model can be used to summarize the impact that a given set of loan attributes has on the likelihood of a loan defaulting in the future.

Estimation of default models requires a sample of data that contains two important elements—the performance of loans through time and the attributes of the loans through time. The statistical models employ the method of maximum likelihood estimation to find the correlation between realized loan performance and the multiple attributes of the loans, such as current credit score and loan-to-value.

Importantly, it has also been shown that such macroeconomic conditions as unemployment rates, interest rates, and housing market conditions in the applicable area also influence a loan’s performance. These factors should be included prominently in any probability of default calculation.

Once the likelihood of loan default has been assessed, the second step in calculating expected loss involves an assessment of the amount of loss on a defaulted credit. This term is referred to as the loss given default.

The LGD of a particular loan is the amount of loss an institution can expect to sustain after accounting for any collateral securing the loan and the costs associated with disposing of the collateral. The LGD calculation can be simplified by using a collateral recovery rate, which represents the costs of disposing of the collateral as follows: LGD= (collateral value x collateral recovery rate) – unpaid balance of the loan, including any senior liens.

The final step in calculating the expected loss for a portfolio is to sum each individual loan’s estimated probability of default multiplied by its respective calculated LGD.

More advanced analysts may consider simulation to create a distribution of expected portfolio losses. Simulation has the virtue of creating a distribution of possible loss outcomes, allowing for a statement of the value at risk within a portfolio.

We appreciate and understand the concerns many accounting professionals have in the FASB proposal and this article should not be construed as Visible Equity’s support for the proposed change. Nevertheless, from a statistical analysis point of view, an Expected Loss Model—if constructed and used properly—can provide key insight into the risk embedded in a loan portfolio, and risk managers would do well to consider such insights, regardless of what FASB decides, as they attempt to accurately identify, measure, and monitor the risks within their portfolio.

Taylor Nadauld, Ph.D., is chief economist of Visible Equity, Midvale, Utah.

CUES Learning Portal