CECL Benchmark Q1 2021

In this paper, we continue the research analysis that has been performed for more than a year, which lets us establish a point of view on whether banks will keep building, maintain, or start releasing allowance into the next quarter. The analysis also shows where they stand on the range of preparedness compared to peers in the context of an evolving economic outlook.

This quarter’s analysis continues with the expansion provided in our last paper to include a differentiation by product type. By further breaking down our analysis into this greater detail, we can provide readers with significantly more insightful results. We also now have the benefit at reviewing multiple quarters together at this more detailed level.

In this quarter, we have also added an enhancement to the analysis by providing an over-reserved/under-reserved percentage to each of our bank-level tables. This percentage represents the amount of the benchmark’s midpoint for which the bank is reserved.

For Q4 2020, we observed that the peer group weighted average allowance for credit loss (ACL) was 2.45%. Going into Q1 2021, we establish that our new upper- and lower-bound indexes for the peer group are respectively 2.11% and 0.66%. The movement in this quarter’s upper and lower bounds continued the trend we saw at the end of 2020 with continued drops in both metrics. This coincides with continued improvement in the economic variable outlook.

Despite this, quarter over quarter there still remains a significant amount of uncertainty on the eventual impact of the pandemic. Therefore, while banks are beginning to release reserve balances in some asset classes, those releases are still typically below expectation when compared to an analysis such as the one we have performed. This is evident in Figure 1, which shows most of the banks within this peer group are still, on an overall basis, reserved above or near their individual benchmark upper-bound. Consistent with the previous studies, we offer peer group and bank-specific commentary.

Introduction

Our benchmark analysis began March 31, 2020 and has continued on a quarterly basis throughout the past year. Each previous study was able to conclude on a general direction banks would take when either building their allowance (as seen early in the pandemic) or holding it consistent (as noted in the past two quarters). As we move into 2021, the question will now focus on whether the time has come to begin releasing the reserves that were built up in 2020. The level of release and which product types are seeing the most change will be the focus of this quarter’s study.
In addition to reviewing these trends for our portfolio level upper- and lower-bound summaries, for this quarter we will once again present our analysis at an asset-type level. This provides additional insight into the changes we are seeing in the industry and offer a better benchmark for how this portfolio mix will affect the way the overall summary analysis has looked.

Overall, the storyline of the economic scenarios in this quarter includes a continued strengthening of many of the variables that had been so volatile the past year, as shown in Figure 2. Based upon this overall effect, we observe that both the upper and lower bounds continued their steady decrease. The peer group average for the lower bound is 0.66% and the upper-bound average was 2.11%, compared to 0.81% and 2.39% from the prior quarter’s analysis. While these changes in bounds in this quarter were not as significant as the Q4 2020 study, note that the banks in the peer group did not release that level of reserves at that time. This leads to the conclusion that overall reserve releases continue to be the expected outcome over the next couple of quarters.

For a summary review of the methodology used to generate this benchmark study, see our original whitepaper. In this paper, we break down the overall results of our benchmark study at a portfolio level and based on asset type. Then, we review each bank’s Q4 2020 results and where those stand compared to our triangulation index. Based on the economic condition assumptions for Q1 2021, we will give commentary on what we expect of the banks in the peer group—which ones will build, maintain, or release allowance.

Comparison of results – portfolio averages

The intent of the updated study is to help gauge what the reserves level will be for Q1 and beyond. As previously mentioned, we have seen a stabilization in the economy and economic variables continuing to show signs of strengthening; the focus now shifts to whether some banks may be prepared for larger reserve releases. This is further supported upon review of the results (Figure 3), as we see that most of the banks within peer groups fell near or above their upper bounds per the analysis. Even those that were not at this level remain significantly distanced from the lower bounds at an individual or peer level, as opposed to earlier studies from 2020 when our analysis showed several banks appeared under-reserved.

This once again highlights the importance of asset-type analysis to showcase how the outlook differs based on asset class—and the continued uncertainty in some markets will slow some of the reserve releases, even at the overall level.

Figure 4 presents the details of our overall analysis, as well as the percentage increase above the peer group lower bound. In addition, this table also includes the new metric to determine how much a bank is reserved when compared to its midpoint per the benchmark. This percentage is determined by dividing the ACL rate over the midpoint rate. Based on this calculation, all of the banks in the peer group remain reserved well above their midpoint. We look further into this metric as we review the asset-type analysis as it provides additional explanation of current quarter expectations.

As was the case in closing out 2020, the big question for early 2021 remains, will banks release reserves and if so, when? Based on our triangulation index and the improving economic forecast, it would appear that reserve releases are upcoming—and in some cases may begin to occur in a material level at some point in the next few quarters. Despite that, with the continued uncertainty around the epidemiology of the pandemic and related vaccination efforts, and the delayed onset of losses by continued government support actions, it seems most of the banks in our peer group will not be releasing significant amounts of reserves until later in 2021. We expect a weighted average reserve release for the peer group in the range of 10-15% overall, which is about three times the release level we observed in Q4 2020.

Comparison of results – results per asset type

In each of the following sections, we present the peer group upper and lower bound graphed against each bank’s Q4 2020 ACL reserves, and then show the Q3 to Q4 changes as well as the bank’s own upper and lower bound. In addition, we show the dispersion of the results that were used to create the benchmark. The gist is that the more disperse the results for a given bank, the more volatile the portfolio performance will tend to have been over time. Finally, we present a view of the peer group indicators to help the reader understand where a bank stands among its peer group in terms of riskiness profile. For each product segment, we offer summary commentary for the banks within their peer group based on Q1 2021 earnings release information.

Commercial real estate

The CRE portfolio has garnered the most attention since the beginning of the COVID-19 pandemic. Business closures, empty offices, and hotels shutdowns have made these portfolios susceptible to increased credit deterioration. However, some property types and locations have fared relatively well, and the government has provided much-needed support that trickles down to the net operating income of CRE properties.

Figure 5 presents the peer group’s current ACL reserve ratio for Q4 2020 (blue bars), as well as the peer group upper and lower bound (red and green lines) and the bank’s own upper and lower bound (red and green dots). We observe that most banks sit above the upper bound and expect releases for most in Q1 2021.

The peer group’s upper and lower bounds are 1.66% and 0.24%, respectively. These bounds have remained relatively stable since the prior quarter’s analysis. Despite improved economic conditions, the uncertainty surrounding this particular product type has yielded a different outcome than others.

To exemplify this point, Figure 6 shows the evolution of each bank’s reserves from Q3 to Q4 as well as their own upper- and lower-bound values. The results in the table show that ACL increases were still prevalent across most banks in Q4 for CRE, while the benchmark itself is showing expected improvements. The combination of these two items can be exemplified by the amount in which most of these banks are over-reserved when compared to their midpoint, including many that are more than three times greater.

Figure 7 gives a sense of the dispersion of the metrics used to build the benchmark. Such information can be used to interpret which banks may be exposed to more or less volatility. The tighter range of results means that for all of the different metrics computed that are part of the benchmark, some have a much smaller range of sensitivity (we use different models but also different scenarios), leading us to interpret that banks with small ranges should have less volatile reserves over time. CRE is a case in point; for example, we observe that JP Morgan Chase and M&T Bank have a very tight range of metrics and their outliers are not too far off from the box plot when compared to their peers.

Figure 8 shows the riskiness indicators for the CRE portfolio in terms of maximum and average net charge-off (NCO) experienced during the last financial crisis, and ACL coverage in quarters for each bank compared to its peers. Note that green is better than average.

This information can be used alongside the benchmark and idiosyncratic knowledge of the portfolio evolution over time to determine whether a given bank is better reserved than a given peer.

Credit cards

The credit card portfolio is one where reserves were hiked very early in the pandemic when we saw unemployment and job losses skyrocket. Because they are unsecured, credit cards usually experience much higher levels of defaults—and we observed a reserve build commensurate with that expected behavior.

Figure 9 illustrates the peer group’s current ACL reserve ratio for Q4 2020 (blue bars), as well as the peer group’s upper and lower bound (red and green lines) and the bank’s own upper and lower bound (red and green dots). The peer group’s upper and lower bounds are 8.80% and 3.2%, respectively, as opposed to CRE. The improved economic conditions have lowered the boundaries but to a lesser extent, while banks have been maintaining reserves at more conservative levels given the still uncertain but more positive economic outlook. Most banks find themselves at or above their own upper bound, which points to potential releases in the future.

Figure 10 presents how each bank’s reserves have evolved from Q3 to Q4 and their own upper- and lower-bound values. The results show that ACL decreases were prevalent for all banks but one in Q4. That being said, even with some of the releases all of the banks in the peer group remain at reserve levels that are at or near their upper bound. Key Bank continues to be the most aggressive in releasing reserves as it relates to its card portfolio, and is now the closest of the banks in its peer group to its midpoint. With the economic outlook improving, we anticipate more broad-based reductions in allowance for cards in upcoming quarters.

Figure 11 gives a sense of the dispersion of the metrics that were used in building the benchmark. Such information can be used to interpret which banks may be exposed to more or less volatility. The tighter range of results means that for all of the different metrics computed that are part of the benchmark, some have a much smaller range of sensitivity (we use different models but also different scenarios), leading us to conclude that banks with small ranges should have less-volatile reserves over time. Cards is a case in point; for example, we observe that JP Morgan Chase and Citigroup have a broad range of metrics (think riskier portfolios) and their outliers are in the 20%-plus range. Our expectation is that those banks would sit well above the peer group average as a result, which they do.

Figure 12 shows the riskiness indicators for the card portfolio in terms of ACL coverage in quarters as well as the maximum and average NCO experienced during the last financial crisis compared to their peers. Here, staying with our two example banks (JP Morgan Chase and Citigroup), we observe that these two banks have riskiness indicators that are worse than average for the peer group, which could justify why they have some of the highest card reserves among the peer group. Note that green is better than peer average.

This information can be used alongside the benchmark and idiosyncratic knowledge of the portfolio evolution over time to determine whether a given bank is better reserved than a given peer.

Mortgage

Contrary to the credit card portfolio analysis above, residential mortgages has been a portfolio where reserves stayed relatively low throughout the pandemic. Despite unemployment spikes and significant job losses that remain at heightened levels even now, other factors have remained steady, leaving this industry without the loss projections others have seen. In particular, strong, consistent house prices and house price projections have kept this market stable in terms of loss projections. In addition, various government and regulatory initiatives have helped leave delinquency and default levels for this market very low.

Figure 13 depicts the peer group’s current ACL reserve ratio for Q4 2020 (blue bars), as well as the peer group upper and lower bound (red and green lines) and the bank’s own upper and lower bound (red and green dots). The peer group’s upper and lower bounds are 1.36% and 0.36%, respectively, which is significantly lower than the other consumer loan portfolios (that is, credit cards and other). This asset type is also showing a continued reduction in both upper and lower bounds from the prior quarter benchmark. As a result, most banks in this study now find themselves squarely in the middle of their individual upper and lower bounds, with Ally Bank as the only dramatic exception.

Figure 14 shows how each bank’s reserves have evolved from Q3 to Q4, as well as their own upper- and lower-bound values. The added column for residential mortgages on the far right shows a very different picture than the prior two asset types. Here we can see about half of the banks in the peer group that are reserved below the midpoint of their benchmark, and no bank is greater than 141%. This is in stark contrast to the CRE portfolio where many banks were over 300%, and all were above 100%. This analysis leads us to conclude that there will be a smaller expectation of potential future reserve releases from this category, and the rest of this year will most likely see stable results.

Figure 15 gives a sense of the dispersion of the metrics that were used in building the benchmark. Such information can be used to interpret which banks may be exposed to more or less volatility. The tighter range of results means that for all of the different metrics computed that are part of the benchmark, some have a much smaller range of sensitivity (we use different models but also different scenarios), leading us to interpret that banks with small ranges should have less-volatile reserves over time. A good example of this can be seen in the results of Fifth Third Bank, which has a significantly broader range in the residential mortgage space than its peers—leading it to hold a much higher level of reserves per the benchmark.

Figure 16 illustrates the riskiness indicators for the mortgage portfolio in terms of ACL coverage in quarters, as well as the maximum and average NCO experienced during the last financial crisis compared to their peers. Based on the results, there is a consistent outcome in which Ally Bank appears to have lower reserve levels and higher riskiness factors than its peers in this loan type. Note that green is better than peer average.

This information can be used alongside the benchmark and idiosyncratic knowledge of the portfolio evolution over time to determine whether a given bank is better reserved than a given peer.

CRE construction

CRE construction portfolios saw relatively large increases in reserves on a percentage basis in 2020. The construction industry was hit hard by the pandemic overall as employment declined greatly at the end of Q1 but has since partially rebounded. The pandemic caused construction delays and much uncertainty, especially around future demand in sectors such as retail and office. These factors contributed to reserve increases throughout much of 2020.

Figure 17 presents the peer group’s current ACL reserve ratio for Q4 2020 (blue bars), as well as the peer group upper and lower bound (red and green lines) and the bank’s own upper and lower bound (red and green dots). The peer group’s upper and lower bounds are 3.45% and 1.31%, respectively. As overall economic conditions have continued to improve over the past couple of quarters, the upper and lower bounds continue to reduce, and reserves in this sector have now begun following suit. Some of the banks in the peer group released reserves rather aggressively in this product type, including Bank of America and Huntington Bank, which released over 40% of their reserve. This drew most of the banks far closer to their midpoint than in past quarters. We should note that historically the construction portfolio was one of the hardest-hit portfolios during the last recession. Since then, banks have adapted their processes to better mitigate the risk of loss, which could also be the reason why most are far below their own upper-bound benchmark.

The results in Figure 18 show just how aggressively some banks began reducing reserves in this asset type during Q4. Interestingly, some in this peer group did not release quite as aggressively and now appear as outliers having reserves well above their benchmark midpoint, such as M&T Bank. However, despite the large releases Bank of America and JP Morgan Chase remain significantly over-reserved per this new metric.

The dispersion of the metrics in Figure 19 shows construction tends to have a tighter range of results than other portfolios, as our model results produce fairly elevated results under the Baseline scenario. Also, under an S3 scenario, the results have a more moderate increase compared to other portfolios.

Figure 20 illustrates the riskiness indicators for the construction portfolio in terms of ACL coverage in quarters, as well as the maximum and average NCO experienced during the last financial crisis compared to their peers. This table continues to highlight that despite the reserve decreases, both JP Morgan Chase and Bank of America have riskiness indicators that are better than the average for the peer group.

This information can be used alongside the benchmark and idiosyncratic knowledge of the portfolio evolution over time to determine whether a given bank is better reserved than a given peer.

Other retail

As expected, other consumer or retail loans continue to find their reserving thresholds ranging somewhere between residential mortgages and credit cards. We also find that given the large variety of loan structures and collateral types that underlie this category, there are larger variations in both the benchmark and allowance rates per bank. For example, banks in which this portfolio comprises mostly auto loans will see different results from the study as banks with mostly short-term lending. The variety of loan types in this category can also make finding one or two key drivers of change or differences a bit more challenging. That said, similar to residential mortgages and despite a continually high unemployment rate, losses in this portfolio remain low, which we project as the outlook going forward. This results in a relatively small allowance rate for loans in this category.

Figure 21 depicts the peer group’s current ACL reserve ratio for Q4 2020 (blue bars), as well as the peer group’s upper and lower bound (red and green lines) and the bank’s own upper and lower bound (red and green dots). The peer group’s upper and lower bounds are 2.11% and 0.82%, respectively. As shown in Figure 21, however, those upper and lower bounds change significantly as you review the individual banks within the peer group. For example, the upper and lower bounds for Regions Bank are 5.63% and 1.85%, respectively, which amounts to more than double the average for the overall peer group. That said, Regions Bank remains reserved at a level well above its upper-bound benchmark.

Figure 22 shows how each bank’s reserves have evolved from Q3 to Q4 as well as their own upper- and lower-bound values. The results in the table show that ACL changes were slight when compared to other asset types for all banks as they relate to the other retail loan portfolio. Other than Citigroup, which is near its lower-bound benchmark, all other banks in the study remain reserved at levels well above their midpoint benchmark, in contrast to the retail mortgage loans. This leads us to believe that additional reserve releases are likely in upcoming quarters.

Figure 23 gives a sense of the dispersion of the metrics that were used in building the benchmark. Such information can be used to interpret which banks may be exposed to more or less volatility. The tighter range of results means that for all of the different metrics computed that are part of the benchmark, some have a much smaller range of sensitivity (we use different models but also different scenarios), leading us to conclude that those banks with small ranges should have less-volatile reserves over time. On this box plot, note that despite having the larger reserve percentage in this category among its peers, Regions Bank comparatively shows very little volatility with a few large outliers.

Figure 24 illustrates the riskiness indicators for the mortgage portfolio in terms of ACL coverage in quarters, as well as the maximum and average NCO experienced during the last financial crisis as compared to their peers. Note that green is better than peer average.

This information can be used alongside the benchmark and idiosyncratic knowledge of the portfolio evolution over time to determine whether a given bank is better reserved than a given peer.

Commercial and industrial

Commercial and industrial (C&I) portfolios saw increases in reserves in the first half of 2020, as smaller businesses were especially hard hit by the pandemic. Despite economic conditions improving in the latter half of the year, it was not until Q4 that we started seeing significant reserve releases in this asset type.

Figure 25 shows the peer group’s current ACL reserve ratio for Q4 2020 (blue bars), as well as the peer group’s upper and lower bound (red and green lines) and the bank’s own upper and lower bound (red and green dots). The peer group’s upper and lower bounds are 1.26% and 0.37%, respectively. As overall economic conditions continue to improve, the upper and lower bounds have reduced again. Despite that, however, all but Key Bank still remain reserved above their upper-bound benchmark.

The results in Figure 26 show that some banks in this peer group are now starting to more aggressively release reserve balances on their C&I portfolio in Q4. As all of the banks remain at reserve levels well above their midpoint benchmark, including those that released more than 10% of their reserve, we would expect to see additional reserve releases across the peer group in this asset type.

The dispersion of the metrics in Figure 27 shows C&I tends to have a uniform range of results, except for Ally Bank. Our CECL forecaster model produces low results for Ally Bank even when applying more adverse scenarios. This leads to a tighter range of results than the different metrics computed. As in the CECL forecaster model that is part of the benchmark, there is a much smaller range of sensitivity. Therefore, these smaller ranges should lead to less-volatile reserves over time.

Figure 28 illustrates the riskiness indicators for the commercial portfolio in terms of ACL coverage in quarters, as well as the maximum and average NCO experienced during the last financial crisis compared to their peers. One of the important considerations shown here is that even those banks that released larger percentages of reserves in this quarter remain low from a riskiness perspective. Note that green is better than peer average.

This information can be used alongside the benchmark and idiosyncratic knowledge of the portfolio evolution over time to determine whether a given bank is better reserved than a given peer.

Summary and takeaways

The main reason for undertaking this research was to understand if there was a practical way to produce an upper- and lower-bound index that could provide a reasonable indicator of the level of reserves across a set of peer banks. This fifth paper in the benchmark series shows that a triangulation index built on different heuristic measures for both a peer group and a bank can give management the necessary guardrails to understand where they are relative to peers in their reserve practices. We found evidence that by combining top-down methodologies and riskiness indicators, we can better grasp the position in which each individual bank finds itself among its peer group for the current and following quarters.

Comparing ACLs from different call reports and historical experiences during the Great Recession is almost impossible given the underlying assumptions that remain undisclosed (that is, weighted average portfolio life) and the difference in portfolio composition. Furthermore, economic environmental uncertainty, lack of clarity on the timing of NCOs, and the impact of government support continue to affect different banks in various ways. Thus, it is vital to understand the parameters of your allowance and know where you stand with respect to your peer group, whether you are above the upper bound or below the lower bound. Such knowledge is crucial for management teams.

We have now automated the process of building the triangulation index based on the outlined measures. This means that we can conduct this analysis on any peer group and at a portfolio level within days. If you have tried to find a reliable benchmark range for you and your peers, feel free to contact us. We offer executive management a view on the array of possible results, especially when internal model reliance is brought into question.