Predicting Earnings: CECL’s Implications for Allowance Forecasts

A financial institution’s allowance for loan and lease losses (ALLL) is an important estimate with significant impacts on an organization’s overall earnings and capital. While this reserve calculation has always had the potential to be quite complex, the new accounting procedures brought by the current expected credit loss model (CECL) and International Financial Reporting Standard 9 (IFRS 9) change the important elements of the process. With these new regimes, allowances must be updated on every reporting date to reflect more than current credit conditions; credit quality will need to be measured from a forward-looking perspective which, by definition, will vary through time. The resulting overall portfolio loss allowance, and thus earnings, can exhibit substantial volatility.

The industry has already had a taste of the potential impacts of using expected cash flows for allowances with acquisitions of distressed loans and purchase loan accounting. This fair value accounting on acquired loans exhibited incredible volatility when compared to other assets. In CECL and IFRS 9, this forward-looking approach applies to the entire institution, and the expected patterns at the portfolio and sub-portfolio levels will be much more sensitive to the economic cycles, portfolio composition, and calculation assumptions.

This shift in predictability of losses and earnings will demand significant time from senior management not only to explain differences period over period, but also to accurately and confidently communicate expected patterns given anticipated strategy choices and market conditions.

There are two main decision types which drive the ability to accurately forecast allowances and overall earnings:

1. Framework and methodology choices – data granularity, model selection, scenario narrative, and a wide array of smaller elements
2. Business and strategy choices – loan structure, type, industry, and geographic distribution, as well as potential for clustered defaults and downgrades (concentration)

Clearly, there are methodology choices that impact overall results; however, it is also clear that the economic dynamics of the portfolio and its composition will now have an increasingly important effect on outcomes.

The predictability of losses is mostly driven by the true economic relationships in the portfolio, which are best described by concentration effects (e.g., name, sector, product, and geography). Some of the dynamics are quite intuitive; for example, an institution heavily invested in California real estate would have losses closely related to statewide housing prices as well as important commercial sectors in California. However, it is clear that more diversified institutions will find a systematic approach helpful in fully understanding, anticipating, and communicating outcomes over time.

While there are multiple approaches that can be considered, we find that using a simulation to determine credit earnings volatility provides a useful measure. Credit earnings volatility provides quantification and insight that will help senior management anticipate what parts of the portfolio, management actions, and scenarios most impact predictability. This measure encapsulates the credit risk in earnings for the entire institution, as well as the contribution by portfolio segment, sector geography, etc.

Armed with an understanding of the dynamics within the portfolio, management can take actions to reduce portfolio credit earnings volatility and better communicate the anticipated volatility, given a market outlook or set of strategic choices. Figure 1 provides the basic formula for calculating credit earnings.

Figure 1 Credit earnings generated by a portfolio from time t₀ to t₁

Source: Moody's Analytics

In the following study, we isolate the impact of shifting from expected credit loss (ECL) over a 12-month horizon to the lifetime ECL allowance framework required by CECL for a sample global corporate loan portfolio created by Moody’s Analytics. Our example analysis is not intended to directly proxy current reserves since the loss emergence period varies by portfolio, but to quantify the shifts caused by concentration effects with a simple example. We use the same portfolio to ensure that the true economics and performance of the portfolio remain the same – the study demonstrates how the attractiveness of particular portfolio segments is impacted by a shift in calculation horizon. Figure 2 shows the top countries and industries represented in the sample portfolio.

Figure 2 Portfolio characteristics: Top five countries and industries represented

Source: Moody's Analytics

The portfolio was analyzed twice with the same starting default probabilities and an analysis horizon of one year. As a straightforward example of the potential dynamics of increasing the ECL to lifetime, allowances were calculated using one-year ECL in the first run, and lifetime ECL in the second run. Using a correlation-based model, we simulated the credit earnings at horizon to determine the expected credit earnings value and volatility over the next year. We also calculated a new measure known as the credit earnings sharpe ratio, which provides a way to quantify profitability with consideration given to the new allowance requirements. Our quantitative measure ranks both segments and instruments by assessing their marginal contributions to credit earnings volatility or the credit earnings sharpe ratio.

Figure 3 Analysis results at the portfolio level

Source: Moody's Analytics

Results from the two runs match intuitive expectations that the overall portfolio allowance level and volatility will increase when applying a lifetime loss metric. Further, we see intuitive patterns where particular loan characteristics are more or less attractive when considering longer loss horizons. For example, for the entire portfolio of approximately 6,000 instruments, the weighted average time to maturity was approximately 3.5 years. The 1,000 top-ranked instruments based on expected loss allowances over a 12-month horizon have a longer average time to maturity, while the top-ranked instruments under lifetime allowances have a significantly shorter average time to maturity. This broad pattern supports the expectation that the new accounting standards will incentivize institutions to favor shorter-term instruments.

Forward-looking credit considerations impact allowances under the new standards, so we are not surprised to find that many of the highest contributors to volatility of credit earnings are exposures that have some of the highest default probabilities. However, when comparing the two runs, there were several areas in the portfolio where assets ranked poorly based on credit earnings volatility – despite the fact that they had smaller default probabilities in the 12-month analysis.

The analysis becomes much more insightful once we look more deeply into segment dynamics and individual instrument impacts. Portfolio diversification plays a much larger role when looking at longer periods of time, which encourages institutions to consider the relative benefit of an instrument or segment and look more closely at overall portfolio composition.

The relative benefits of certain sectors clearly change based on the required allowance horizon. We see in this analysis that the top-ranked exposures when using ECL over a 12-month horizon for allowances are different than the top-ranked exposures when considering lifetime allowances. In Figure 4, we see the patterns within the portfolio. It is important to remember that the economics of the portfolio are the same in both runs, so our simulation correctly reflects that many of the best performers under a 12-month horizon are the same under lifetime allowances. At the same time, there are clear cases where sectors are ranked significantly differently.

Figure 4 Top 20 projections for lifetime allowances by credit earnings sharpe ratio

Source: Moody's Analytics

In our study, it becomes clear that interactions of various segments within the overall portfolio can play an important role in outcomes. For example, we see that the Swiss machinery and equipment segment is very attractive when looking over a single-year period; however, when we consider the full life of the loan, that segment becomes significantly less attractive due to the expected volatility of allowances in this category. Conversely, all of the real estate categories broadly increase in relative attractiveness when we evaluate our portfolio with a lifetime perspective.

We find that there is value in quantifying the risk and profitability of not only the portfolio as a whole, but also the interaction of individual elements within. Segment-level insights provide a quantitative basis for understanding dynamics, as well as hard numbers for reference when communicating strategy, expectations, and policy shifts to internal and external stakeholders. In our example, the analysis indicates a clear justification for increased investments in real estate in a lifetime allowance environment and decreasing focus (or shorter durations) in some industrial categories.

It is also worth noting that the above analysis is based on a benign credit environment. The impact of using a forward-looking default probability will have a significant impact in the negative part of the credit cycle. There will be even greater costs and uncertainty for organizations holding risky instruments, as a simple change in default probabilities alone will cause significant volatility in earnings.

As CECL rolls out across financial institutions in the US, and IFRS 9 takes effect for much of the world, managers must adopt new ways to manage risk, compare instruments, and communicate expected outcomes and dynamics. As we have shown in this simple study, these considerations must be worked into business as usual for institutions and should be addressed at origination and in strategy to ensure organizations are following strategic and lucrative business practices given a new set of dynamics introduced by CECL.