In this paper, we provide an update to the previously published research on our triangulation framework and how the set of banks in our select peer group performed relative to expectations of increased reserve build. We compiled the actual Q2 2020 estimates, superimposed them on the research results, and provided commentary on the released results for each bank relative to the peer level benchmark. As expected,1 most banks kept building aggressively toward the March upper-bound indicator (Figure 1) as economic conditions from March to June continued to deteriorate.
We observed that the peer group weighted average allowance for credit loss (ACL) increased from 1.98% to 2.59% from Q1 2020 to Q2 2020, which is a dramatic increase of more than 30% on average. We expect that our lower bound computed as of Q2 will probably rise by a similar amount once we complete the Q2 analysis with call report data.
Our initial research2 from March 31, 2020 led us to the conclusion that most banks would continue to build their allowance for Q2 2020. The research results helped create a triangulation index upper- and lower-bound range based on a few different metrics outlined in the initial paper. The peer group average for the lower bound was 1.64% and the upper-bound average was 3.24%. We now review each bank’s results compared to our expectation to understand whether our triangulation index provided a good benchmark indicator going into Q2 results. Based on the economic condition assumptions for Q3, readers can form their own conclusions on the potential direction of reserves for any given bank within the peer group.
Our research relied principally on applying historical experience and net charge-off (NCO) forecast-based expected credit loss using a few different methodologies. We used these metrics to create a composite index to build a range of estimates for the peer group as well as each bank within the peer group.
We relied on quantitative metrics but also used riskiness indicators to tweak our judgment based on how well historical performance fits the current portfolio loss profile. The main riskiness indicator was based on the latest Dodd-Frank Act Stress Testing (DFAST) results (a DFAST riskiness scaler) to establish the notion of relative riskiness across banks. The riskiness indicator table in Figure 2 is meant as a refresher (a green highlighted cell means the bank was below the peer group average).
Comparison of results – averages
At first glance (Figure 3), we can see that most banks’ Q2 2020 estimates fall directly within the established range as opposed to Q1, when our analysis showed several banks that seemed to be under-reserved. Based on our initial commentary from our March 31, 2020 research, we will now add commentary after the fact and determine why some banks may have built more or less than we expected. This review should help us understand how our triangulation index performed and help validate its use as a tool to address internal management challenges.
Figure 4 presents the details of our analysis as well as the percentage build between Q1 and Q2 2020. We will base most of our commentary on this data and earnings release commentary provided by each bank.
Starting with JP Morgan, its allowance build from 2.50% to 3.32% of loans is in line with its early disclosure that the Q2 build would be substantial. We stipulated that if its allowance build was in line with early disclosure, it would be right in line with our triangulation index upper bound, and we would see it as well reserved based on our estimates of where JPM should be at this point in the cycle. We will update our triangulation index once the new call report data is out; we expect that it will fall above the lower bound of the index with some of the other more aggressively reserved banks. It is worth mentioning that JPM was very aggressive this quarter in ramping up reserves for the commercial real estate (CRE) portfolio and continuing to bulk up its card portfolio as well.
Bank of America’s Q2 2020 allowance build from 1.63% to 1.96% is a 20% increase. This is smaller than the average increase for the peer group but BofA’s DFAST riskiness indicator is the lowest of the peer group. We expected it to build more this quarter and get closer to its upper bound of 2.56%, which is lower than the peer group given the relative riskiness of its portfolio. Its Q2 build was slightly larger than the 19% build in the Q1 post true-up of 31%. As with many of its peers, BofA saw a reduction in NCO on the consumer side and an increase in NCO on the commercial side. It remains to be seen if the support offered to consumers through its deferral program will eventually end up generating a larger-than-expected NCO increase. At close to 2% ACL, we expect BofA to be right on the peer group lower bound once we update the analysis with Q2 call report data.
Wells Fargo’s Q2 2020 allowance build from 1.19% to 2.19% represents an 84% build in Q2, compared to a 25% build in Q1 2020. This is the largest Q2 build of all banks in the peer group. The increase in its Q2 allowance is largely driven by economic conditions that have worsened significantly compared to the prior quarter’s outlook. We surmise that the low level of allowance up to now was in part due to higher-than-average recoveries that will likely not materialize. Though customer forbearances and deferrals will delay the recognition of charge-offs, delinquencies, and non-accruals, these have been considered in its loan loss forecast. Commercial criticized assets increased in Q2 for downward credit migration, particularly in the retail, entertainment, and recreation sectors. We expected Wells to have a large build as the original analysis had it under-reserved compared to peers, being below the lower-bound index. Now, it sits in a much more reasonable range.
Citigroup's Q2 2020 allowance build from 3.14% to 3.89% is a 24% increase. This is below the average build for the peer group, but Citi was already the most reserved bank of the entire group at 3.14%, and its Q2 build is slightly larger than that of Q1. Citi’s DFAST riskiness indicator is also much lower than the peer group average—as such, a smaller build for Q2 makes sense. Overall, as discussed in the original paper, Citi’s card and CRE portfolio are driving it to reserve a larger amount than its peers and its range estimate is much larger than the peer group. It is worth noting that Citi’s average NCO during the Great Recession was the highest among all peers, which drives the much-larger upper bound shown in Figure 4. We expect Citi to be well above the peer group lower bound once we update our analysis with Q2 data, and we expect it to be right in line with its lower-bound estimates based on our triangulation methodology.
US Bank’s Q2 2020 allowance build from 2.07% to 2.54% represents a 23% build in Q2, compared to an 11% build in Q1 2020. The Q2 build was in response to changes in economic conditions but is smaller than the average build of the peer group. In the original paper, we had viewed it as already well reserved given that it was well above its lower-bound index and it had a DFAST riskiness indicator slightly better than the peer group average; therefore, a substantial build in Q2 was not needed. Its Q2 ending allowance would place it above its previous upper-bound index. Therefore, it is very well prepared for future charge-offs.
Truist’s Q2 2020 allowance build from 1.78% to 1.81% is the smallest build of all banks in the peer group. Note that it was the most aggressive in its Day 1 build, which was 200% of the ACL as of December 31, 2019. Truist was above both the peer group and above its lower-bound range estimate, but we expect it to fall below the peer group once we update the Q2 analysis. Historically, Truist has had a very conservative portfolio and performed better than most based on its DFAST riskiness indicators. On its earnings call, management discussed that the credit mark, accrued based on the merger of equals, gives it a lot of additional cushion if things take a turn for the worse. We believe Truist continues to be well prepared for the coming emergence of losses as forecasted by its lifetime potential loss experience.
PNC Bank’s Q2 2020 allowance build from 1.66% to 2.29% represents a 38% build in Q2, compared to a 19% build in Q1 2020. This increase was primarily due to changes in its economic scenarios and to the weightings assigned to its scenarios. Another factor noted in its earnings presentation was changes in its portfolio, including new loans, credit quality, and draws. The Q2 build was much higher than peers; in our March triangulation analysis, it had already been well above its lower-bound index. We expected it to build in Q2 to put it near its upper-bound index—which is exactly what happened.
Fifth Third Bank’s Q2 2020 allowance build from 2.02% to 2.50% represents a 24% build for Q2, compared to an 18% build in Q1 2020. Fifth Third’s build brings it well above the peer group average but it remains below its estimated lower bound, which gives us pause. Fifth Third has one of the relatively riskier portfolios overall based on the DFAST riskiness indicators and its Great Recession average NCO. At 2.50%, it will easily surpass the lower-bound estimate for the peer group once we update our analysis for Q2 data. Fifth Third does have some help with the unamortized discount for the MB Financial acquisition and Paycheck Protection Program (PPP) loans which, if accounted for, would bring its ACL ratio to 2.76%. It also has plenty of coverage for nonperforming loans (NPLs) and non-performing assets (NPAs) at 410% and 385%, respectively. Fifth Third may be one of the banks that has to ramp up ACL once the loss emergence of the lifetime loss estimate starts materializing, given its relative risk profile. However, we believe that Fifth Third is well positioned entering the downturn.
Ally Bank’s Q2 2020 allowance build from 2.54% to 2.85% represents a 12% build in Q2, compared to a 25% build in Q1 2020. The Q2 build was well below the peer group but before Q2, it was the second most well-reserved bank in the peer group. In our original paper, we believed it was appropriately reserved, so a larger build for Q2 was not needed as it built its allowance largely for the Day 1 CECL adoption. The auto financing lender has taken a more conservative loss forecasting approach as it excludes any stimulus-related benefits or assumptions within its modeling, which could have substantial benefits given US government measures. We still believe it to be adequately reserved.
Citizens Financial Group’s Q2 2020 allowance build from 1.73% to 2.01% represents a 16% build in Q2, compared to a 21% build in Q1 2020. CFG moves quite a bit above the peer group lower-bound threshold and is likely to be right on the lower bound once we update our analysis. However, CFG’s range was higher than that of its peers, which may indicate that it would be less well reserved. The earnings call in Q2 2020 revealed that, like Regions, it is de-risking the portfolio. It is moving an almost $1 billion student loan portfolio as held for sale, freeing up $100 million of ACL being reassigned to the remaining portfolio. Like other peer banks, ex-PPP ACL would rise to 2.09%, which gives it an additional cushion. We posited that CFG was going to be an institution that had to do much more reserve build this quarter; we observed a combination of reserves build and portfolio de-risking that leaves it better prepared than in Q1. We still believe that CFG may be one of the banks that needs to keep building ACL going forward if the recovery drags on more than expected.
Key Bank’s Q2 2020 allowance build from 1.28% to 1.80% represents a 41% build in Q2, compared to a 9% build in Q1 2020. The Q2 build was much higher than peers but note that Key was one of two banks that fell below the peer group lower-bound estimate in March; therefore, a larger build was necessary to put it into more alignment among its peers. Its Q2 ending allowance now places it slightly above the lower-bound estimate but still low relative to peers. In its earnings call, Key management cited a strengthened portfolio, underlying credit quality, and PPP loans included in its C&I category with minimal future loss assumptions as drivers for lower allowance rates. We expected Key to have a higher-than-average loan loss reserve build in Q2, which it did, so it should now be considered as more adequately reserved. If there is a prolonged downturn, expect it to have a larger-than-average ACL build.
Regions Bank’s Q2 2020 allowance build from 1.89% to 2.51% represents a Q2 build of 33%, in line with the peer group average. This amount of ACL puts it well above the lower bound for the peer group and well above its lower bound, which was much higher than the peer group due to the risk profile of the bank. In our original paper, we stated that Regions had the highest DFAST riskiness indicators and that a much larger build was going to be expected. We were pleased to see that they did just that and more. As part of the Q2 strategy, Regions proceeded to de-risk its overall portfolio by aggressively charging off more than $182 million and applying downgrades to loan portfolios significantly affected by COVID-19, driving reserves higher ($382 million). The impact of these actions drove Regions net income into negative territory (like Wells) but positions it well for the waves of anticipated loss emergence. We expect Regions will be much better positioned when we update our analysis with Q2 2020 call report data.
M&T Bank’s Q2 2020 allowance build from 1.47% to 1.68% represents a 14% build in Q2, compared to a 13% build in Q1 2020. The increase—much like all peer banks in Q2—was due to more adverse economic scenarios than Q1, and it was also proactive with borrower rating downgrades. Though its Q2 build was lower than the peer average and its Q2 ending allowance is also relatively low among peers, its reserve level in our triangulation analysis had it as adequately reserved in March because it was the only bank to exceed its upper-bound threshold. Its upper-bound index was low due to its very low historical loss experience. PPP loans were also cited as contributing to a lower total allowance rate. We expect M&T to be below the peer group lower bound once we update our analysis with Q2 data. However, we still expect it to be near its upper-bound estimates based on our triangulation methodology. Unless M&T replicates its historically low loss experience, it will have to ramp up reserves very aggressively as the loss emergence period materializes.
Huntington Bank’s Q2 2020 allowance build from 2.05% to 2.12% represents a 9% increase over Q1 2020, which is about one-third of the peer group average. The bank is still well above the peer group average lower bound and much higher than its lower bound, which is 1.41%. We expect it to be well positioned and in line with the peer group average when we update our research with Q2 call report data. Its total ACL cumulates to 2.27% and ex-PPP rises to 2.45%, putting it among the better-prepared banks. It is worth nothing that its DFAST riskiness indicators are on par with our peer group but much lower than other super-regionals. We noted in our previous paper that its concentration in mortgage attracts lower reserves due to the resilient housing market; if this changes, we expect it to have to build more aggressively on the retail side. We expect Huntington to be well prepared if the economic environment stabilizes over the next quarter.
Summary and takeaways
The main reason for undertaking this research was to understand whether 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. 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.4
After reviewing the earnings release from each bank and comparing those to our previous commentary on where each bank stood within the peer group, we believe our benchmark confirms the build action taken by most institutions within the group, especially Wells, US Bank, Key, and Regions. Our analysis does surface that a few banks in the peer group may still have to build more allowance as we enter the loss emergence period. Key and M&T will be two to watch after we update our analysis. The other exception worth mentioning is Truist, which had the smallest build of all banks but can rely on its credit mark from the recent merger as a backstop.
The same methodology can be repeated using portfolio-level call report information to narrow the index range even more and to refine the precision level. In times of rapidly developing and extreme uncertainty, we find that combining heuristics can offer
much-needed reassurance when trying to look for probable outcomes. The tools used in this paper to conduct this type of benchmarking are available from Moody’s Analytics. We can offer executive management a point of view on the array of possible results, especially when internal model reliance is brought into question.
Additional resources from Moody’s and Moody’s Analytics
1 Initial research conducted prior to Q2 disclosures: https://www.moodysanalytics.com/-/media/article/2020/cecl-build-is-it-enough.pdf
2 “CECL build – Is it Enough?” https://www.moodysanalytics.com/-/media/article/2020/cecl-build-is-it-enough.pdf
3 Note: the upper and lower bound were estimated based on Q1 call report data; we compared those to actuals for Q1 and Q2 2020.
4 We expect that once the economic environment stabilizes, the benchmark range can be valid for more than one to two quarters, making it a lasting