The much-anticipated and -deliberated changes to the accounting for financial instruments ushered in by ASU-2016-13, Financial Instruments–Credit Losses, happened to take effect during an unprecedented economic crisis brought about by the COVID-19 pandemic. This coincidence provides insight into management’s decision making in a crisis and a test of whether the guidance meets the needs and expectations of users. The authors’ analysis of selected CECL disclosures from last year concludes that management’s judgment often influences and even overrides the results generated by CECL modeling; although conservatism may have been warranted under extraordinary circumstances, users might justifi-ably wonder whether such judgment is consistent with the goals of the standard.
The United States and its economy are struggling to cope with the COVID-19 pandemic, and there is great uncertainty regarding its ultimate impact. This article provides some insight into the effects of the pandemic’s economic crisis on the implementation of ASU 2016-13, which requires estimates of current expected credit losses (CECL) that reflect the macroeconomic environment at the time of reporting. No one could have predicted that the required implementation of this long-awaited standard would coincide with a devastating pandemic. Reflecting the uncertainty of today’s world in accounting estimates poses an immense challenge for financial statement preparers as well as auditors.
This article presents insights from a review of CECL disclosures in earnings releases and Form 10-Qs from the second quarter of 2020 using a sample of 25 large and smaller banking entities (Exhibit 1). The evidence suggests that a significant inconsistency has developed between management’s loss provisioning estimates as calculated and reported under CECL and the seemingly stabilizing macroeconomic situation. Furthermore, management’s judgment may have significantly influenced and overridden the provisioning results generated by CECL modeling techniques.
Sampled Banking Institutions
In June 2016, FASB issued Accounting Standards Update (ASU) 2016-13, Financial Instruments–Credit Losses (Topic 326), which represented a significant change in the financial accounting model for credit losses applicable to all financial instruments other than those measured at fair value. This new standard required a change to the “current expected loss” model from the previous “incurred loss” model and became effective for fiscal years beginning on or after December 15, 2019—which happened to coincide with the coronavirus (COVID-19) pandemic that began in early 2020. In ASU 2016-13, CECL calculations reflect the macroeconomic environment at the time of reporting; this key attribute is no easy feat in the face of unprecedented economic uncertainty due to a pandemic.
Before the ASU became effective, banking institutions’ 2016 Form 10-K disclosures revealed that implementation was expected to require significant effort; new techniques and analyses would need to be developed; new loss models would be required; and the impact of the new guidance was likely to be material (Pinello and Puschaver, “Accounting for Credit Losses Under ASU 2016-13: Anticipating the Impact on Reporting and Disclosure,” The CPA Journal, February 2018). These were the expectations before anyone knew that the eventual implementation would be marked by the unique macroeconomic uncertainty brought about by the COVID-19 pandemic.
As the adoption date neared and then became reality, a review of current expected credit loss disclosures in 2018 Form 10-Ks, 2019 Form 10-Ks, and first-quarter 2020 earnings releases and Form 10-Qs of sampled banking entities revealed that CECL was making a significant difference (Pinello and Puschaver, “CECL Encounters a Perfect Storm,” The CPA Journal, July/August 2020). Specifically, CECL enhanced the degree and timeliness of credit loss provisioning and significantly increased the levels of the allowance for credit losses (ACL). The increases in ACL levels arose from both the “transition adjustment” recorded effective January 1, 2020, upon adoption and from the very significant CECL provisioning reported in Q1-2020 just as the potential depth and breadth of the stresses arising from the pandemic were emerging and generating immense uncertainty. A comparison of the resulting March 31, 2020, ACL levels determined under CECL, with the December 31, 2019, levels determined under the previous “incurred loss” model, showed a universal increase. The magnitude of increases was diverse; however, a few entities reported an ACL increase in the range of 15–20%, while others reported increases in the range of 200%.
Auditors must audit CECL estimates that, unlike other types of estimates, are a function of the sometimes highly uncertain macroeconomic environment. Auditing estimates is generally one of the most challenging areas auditors face. The implementation of CECL estimates has happened at a time when the auditing standards regarding estimates have been enhanced by the PCAOB’s issuance of Auditing Standard (AS) 2501, Auditing Accounting Estimates, Including Fair Value Measurements (effective for audits of financial statements ending on or after December 15, 2020), and by the AICPA’s issuance of Statement on Auditing Standards (SAS) 143, Auditing Accounting Estimates and Related Disclosures (effective for audits of financial statements ending on or after December 15, 2023). These changes are being implemented to help auditors more effectively audit estimates, a task that has proven to be difficult in itself. The incorporation of the predictable, but very challenging, macroeconomic factors into CECL estimates increases the complexity of the estimation process and thereby renders both management’s and the auditor’s tasks that much more difficult.
Fundamental Modeling Challenges
The estimating process required by CECL must incorporate two very uncertain variables: First, what do various models and consensuses indicate about the current status of and overall evolving macroeconomic environment—and how reliable is that forecasting model? Second, how is the credit portfolio as it exists at a point in time (although its composition is fairly stable quarter to quarter) going to react to the forecasted economic environment? Essentially, how applicable are past historical data and experience to the current situation? The interaction of these two dimensions compounds the uncertainty involved. (The challenges for auditors relating to this complexity are very significant, as outlined by Matthew Clohessy, “Impacts and Challenges in Auditing CECL,” The CPA Journal, July/August 2020.)
Although there is likely great complexity within the models as they balance a myriad of inputs, the disclosures provided by companies do not provide such detail.
Management’s need to integrate these two variables is a core requirement of the CECL modeling challenge and would be inherent to the underlying algorithms developed. But as the current environment deviates from past experience, the inherent utility of historical data decreases. For example, if peak unemployment were 6% during the periods of historical data, then its use in predicting behavior within a 15% unemployment environment becomes increasingly speculative. Similarly, the forecasting utility of historical experience during modest swings in GDP growth/contraction over business cycles is of decreasing utility when one is forecasting a dramatic GDP contraction in the 20% range, hopefully to be followed by a robust rebound enhanced by government support policies and programs.
Although there is likely great complexity within the models as they balance a myriad of inputs, the disclosures provided by companies do not provide such detail. Management teams did offer disclosures about the assumptions involving GDP and unemployment forecasts, indicating that they are key modeling inputs. Furthermore, in discussing how Q2-2020 CECL provisions were developed and how they compared to those in Q1, management teams did not indicate that the changes were due to changes in the credit portfolio composition, which is reasonable given that portfolio composition does not dramatically change quarter to quarter; nor did they cite fundamental changes in the modeling algorithms.
The Estimating Environment in Q1-2020
Although there still remains uncertainty about the state of the economy, there is now more clarity regarding what is likely on the horizon. The CECL estimating required in Q1-2020 (the first quarter under the new guidance) came at a time of unparalleled uncertainty. The pandemic was just taking hold and there were very few known variables. The stock market evidenced this incredible uncertainty very quickly, hitting dramatic lows on March 23, 2020, falling about 30% in five weeks just as Q1 results were being finalized. Forecasts for the economy, including loss of jobs and unemployment, were very wide-ranging and universally depressing. The incredible stresses and conflicting insight of the CECL estimating effort in the first quarter of 2020 were captured by the Wall Street Journal on April 15, 2020: “[JPMorgan] said the provision was based, in part, on the assumption that U.S. gross domestic profit [Q2] would fall an annualized 25% and unemployment would rise to more than 10% in the second quarter. But JPMorgan economists have recently amended their forecast to a 40% decline in GDP in the quarter and a 20% unemployment rate” (“JP Morgan, Wells Prep For Fallout,” p. A8). The worsening economic forecast was released on April 9, 2020, just days before JPMorgan’s earnings release on April 14. AS 2501.14 alerts auditors to this type of situation, cautioning auditors to evaluate whether “the data [used by the company to develop an estimate] is internally consistent with its use by the company in other significant accounts and disclosures.”
The economic trend turned out better than at first expected, and the forecasting consensuses evidenced a better grasp of what would likely happen. For example, as reported by CNBC.com, the Bureau of Labor Statistics reported a 2.5 million gain in jobs for May 2020, which was much better than the consensus among economists of a loss of 8.3 million jobs. For June, the disparity between the forecasted and actual jobs impact started to decline, with a reported 4.8 million gain in jobs that was still better than the consensus forecast of 2.9 million gain. For July, the Bureau reported a 1.8 million increase in jobs, which was significantly closer to the consensus forecast of 1.5 million. For August, the consensus forecast was a 1.32 million gain compared with the 1.37 million gain reported; and for September, the 661,000 gain was only slightly below the 800,000 consensus. Similarly, the June 2020 unemployment rate was reported at 11.1%, better than the 12.4% estimate (down from a peak of 14.2% in April and just over half the 20% predicted by JPMorgan in early April); the July unemployment rate was reported at 10.2% in line with the consensus of 10.6%; for August, the unemployment rate dropped to a reported 8.4%, which is better than, but in the range of, the 9.8% forecasted; and for September, the unemployment rate continued to improve to 7.9%, which was slightly better than the 8.2% estimate. Lastly, after many earlier dire and wide-ranging forecasts, the GDP results for Q2 were reported at an annual contraction rate of 32.9%, which was close to and slightly better than the 34.7% consensus at the time. Estimates per the Conference Board’s September 9, 2020, updated Base Case Forecast were for a Q3 rebound in the range of 33% and for an overall 3.8% contraction in 2020 GDP.
Clearly, the economic outlook had improved from the dire views during Q1-2020. It is important to keep in mind, however, that some of the above information was not available as management developed its CECL estimate and there is wide disparity among economic forecasts. Furthermore, there are many sources management can look to for perspective, as well as its own internal forecast. For example, the Conference Board’s June 9, 2020, forecast (which would have been the latest available before the end of Q2) projected an annualized GDP contraction rate of 39.5% in Q2, to be followed by a 32.2% rebound in Q3 (which remains consistent with its September 9, 2020 outlook) and 7.8% growth in Q4. In addition, management would have been gaining insight on the effectiveness of its own estimating process as various data became available during April, May, and June.
The government’s response was very aggressive, with various programs aggregating trillions of dollars put in place. Many of these initiatives were targeted toward support for individuals and small businesses, but there were also programs to support larger businesses, especially ones hit hardest by societal “stay-in-place” programs, such as the travel and leisure industries including the airlines. Last, but not least, the stock market recovered and the major indices climbed back up near and beyond the February 2020 highs reached just before the effects of the pandemic hit.
Regarding the second major variable of assessing how an individual credit portfolio will react to the evolving microenvironment, it is too early to form any insights as to how it might differ from historical experience, especially with regard to consumer-related matters. This estimating challenge was further muddled by government support programs, such as the Paycheck Protection Program and enhanced unemployment benefits, companies that furloughed employees while maintaining their medical coverage, and efforts to accommodate borrowers’ needs through payment waivers, forbearance, eviction moratoriums, and so on. Nevertheless, because consumer spending is estimated at 68% of GDP, the related GDP forecasts are inherently making judgments about consumers’ state of mind. Although there remains uncertainty, there is also politicians’ inherent inclination to help their voting base. In addition, the entities’ nonper-forming assets at the end of the second quarter did not increase dramatically: The largest entities in the authors’ sample averaged an approximately 15% increase, not unduly alarming and not unexpected, led by increases of 39% for Citigroup and 31% for JPMorgan. The smaller entities averaged an approximately 11% increase, with Western Alliance reporting a 54% increase followed by Bank of the Ozarks and United Bankshares with increases slightly over 20%. These trends are not alarming, but may represent a recognition delayed by government support programs.
Q2-2020 Credit Loss Provisioning
As would be expected, in Q1-2020 the 25 sampled entities consistently reported very significant credit loss provisioning compared with that at Q4-2019—not only due to the nature of the new CECL guidance changing from an “incurred loss” to a “current estimated loss” model, but also due to the dire and uncertain outlook at the time. Those large loss provisions also would be expected to have a relatively cautious nature given the dramatic environment at that time, as exemplified by these two discussions of Q1 provisioning: 1) Bank of the Ozarks noted, “If future economic conditions align with our projections, then our provision expense in future quarters should primarily … reflect … loan growth.” 2) US Bancorp commented, “Expected loss estimates consider both the decrease in economic activity and the mitigating effects of government stimulus and industrywide loan modification efforts designed to limit long term effects of the pandemic event.”
Although there still remains uncertainty about the state of the economy, there is now more clarity regarding what is likely on the horizon.
As management prepared for the CECL loss provisioning required for Q2 reporting, it appeared that the macroeconomic backdrop had not only improved from the incredible unknowns of Q1—the apparent dependability of the forecasting had also improved. In such a context, one might expect that Q2 loss provisioning would return to more normal levels because the ACL levels had been significantly bolstered by both the very significant Q1 loss provisioning as well as the increases arising from the January 1, 2020, transition charge upon adopting CECL requirements. But this is not what happened.
By a significant majority, the entities reported Q2 credit loss provisions comparable with or significantly greater than those in Q1, although there were also instances where provisioning for Q2 actually was meaningfully less than that in Q1 (i.e., a provision multiple lower than 1×; see Exhibit 2). Comparing Q2’s provision with Q1, among the larger banks, PNC Financial increased Q2 by 2.7×, Wells Fargo by 2.5×, and US Bancorp by 1.7× (note its Q1 comment above). However, Ally Financial, a major lender within the automotive-related industry, which had disclosed that $602 million (two-thirds!) of its first-quarter $903 million provision was “qualitatively determined” based on management judgment—and thereby apparently overrode the CECL modeling results to a significant extent—reported a Q2 provision 68% lower than Q1 (the greatest decrease among the reviewed entities). In Q2, Ally reported a $287 million provision, which was in line with previous quarters other than Q1-2020, and also reported a 26% decline in nonper-forming loans. Management still noted that $128 million of the Q2 provision was “attributable to the macroeconomic environment and other factors,” implying a continuing concern. Arguably, this provisioning pattern is what one would expect under CECL, notwithstanding the large management qualitative Q1 portion, and noting that sustaining the ACL level inclusive of that large judgmental addition evidences management’s perception of a continuing need.
Sampled Banking Institutions’ Q2-2020 Credit Loss Provisions Relative to Q1-2020 Provisions, Presented as a Multiple
There was wider disparity among smaller banks. Flagstar reported a sizable Q2-2020 provision of $102 million, compared with $14 million in Q1, a multiple of 7.3×, with the next highest multiples being significantly smaller at 1.8× for Western Alliance and at 1.7× for United Bankshares. Flagstar’s CEO commented, “This was a provision largely driven by the uncertainty around the pandemic and the conservative approach we took within the CECL framework of modeling-in economic variables.” Flagstar’s Q2 press release also noted: “The increase was primarily driven by our forecast of economic conditions. These forecasts reflect our view that the economy will continue to be challenged by the response to the COVID-19 pandemic, especially in the commercial real estate sector, for an extended period of time.” In the press release’s supplemental information, it disclosed that, of the $102 million provision, $39 million was a qualitative factor, and $31 million arose from its economic forecast. It also disclosed that it was now using an assumption of 10% unemployment and only a slight economic recovery at year-end (differing significantly from the Conference Board’s economic forecasts). In Q1, Flagstar disclosed it was using an assumption of an 18% decrease in Q2 GDP (compared with the 32.7% experienced) and peak unemployment of 9% (compared with the 14.2% peak reported for April and 11.1% for June). So it appears management had ample reasons to change its assumptions (“modeling-in” per the CEO), but the changes are very dramatic; coincidentally, Flagstar also reported that a “second quarter 2020 net gain on loan sales increased $213 million, to $303 million, as compared to $90 million in the first quarter 2020.” In comparison, JPMorgan disclosed that for Q2 provisioning, it used a base case year-end unemployment estimate of 10.9%, up from 6.6% used for Q1 provisioning, and its Q2 provision was 1.3× that used for Q1. In contrast to the larger banks, six of the smaller entities reported relative reductions in their Q2 provisioning, with the lowest entities reporting a 56% reduction. Those lowering entities included Bank of the Ozarks (note its Q1 comment above), which decreased its Q2 provision by 39%.
The stresses of implementing CECL within the pandemic environment are exemplified by Wells Fargo, a large entity with significant resources to devote to the estimating task. At December 31, 2016, its ACL appeared conservative among the larger banks, at 3.3× the level of its 2016 provision (which at December 31, 2019, increased to 3.9× the level of its 2019 credit loss provision). Yet in its 2016 Form 10-K, it had warned that the new guidance could have a material impact. At the time of adoption on January 1, 2020, however, it actually recorded a 12% reduction to its ACL; after reporting a $3.8 billion Q1 loss provision, the overall increase to its ACL at Q1-2020 compared with Q4-2019 was 15%, among the lowest of all the entities surveyed. In Q2-2020, however, it reported a $9.5 billion loss provision, one of the largest increases compared with its Q1 provision of all the large entities. (It also suspended its common stock dividend.) In the supplemental information, Wells Fargo stated that “economic conditions worsened significantly compared to prior expectations” and disclosed several key modeling assumptions: that the economy will only grow 6% in the second half of the year (similar to Flagstar’s assumption above of minimal growth, but significantly below the Conference Board’s forecast) and that year-end unemployment would be at 10% (it did not disclose comparative Q1 assumptions). Without disclosing an amount, it also commented that management did “apply some weighting on a downside scenario to reflect the uncertainty in the economic forecast.” In addition, under “Critical Accounting Policies” of its Form 10-Q’s Management Discussion and Analysis section, Wells Fargo presented four paragraphs of discussion about the “sensitivity” related to performing CECL estimates. It highlighted that its ACL, which was $20.4 billion, might have increased by $5 billion under certain scenarios, but concluded with: “Management believes that the estimate for the ACL for loans was appropriate at the balance sheet date. Because significant judgment is used, it is possible that others performing similar analyses could reach different conclusions.”
Even as the macroeconomic trends discussed above were improving, many of the banks conversely commented that the additional provisioning in Q2 was due to “worsening conditions and increasing uncertainty.”
Why Is There Such Significant Q2 Provisioning?
One would have thought that with all the uncertainty existing at the end of Q1-2020 and the first-time recording of credit loss provisions under CECL, bank management teams would have been especially cautious in their outlook. From the commentary regarding the Q2 provisioning, however, it appears that such was not the case. Even as the macroeconomic trends discussed above were improving, many of the banks conversely commented that the additional provisioning in Q2 was due to “worsening conditions and increasing uncertainty.” Such could mean one of two fundamental technical situations—either the macroeconomic assumptions used in Q1 were not conservative enough, or management now had less comfort in its ability to formulate the required estimates. For example, as noted above, Ally Financial’s disclosure evidenced that in Q1, management significantly overrode the CECL modeling results. Similarly, in Q2, Citigroup disclosed that its significant provisioning included “an additional qualitative management adjustment” but did not disclose the amount, as noted earlier, Wells Fargo did the same. Trustmark pointed out this emerging issue in its Form 10-Q: “However, due to the COVID-19 pandemic, the macroeconomic variables used for reasonable and supportable forecasting have changed rapidly. At the current levels, it is not clear that the models currently in production will produce reasonably representative results since the models were originally estimated using data beginning in 2004 through 2017.”
There is a third alternative: Because the marketplace was still uncertain about how to evaluate the potential impact of the COVID-19 pandemic, management realized that it had an opportunity to increase ACL levels without being viewed adversely. Finally, there was the very important but unknown aspect of what reserving pressure may have existed behind the scenes from the various banking regulators.
As the second half of 2020 evolved and continued into 2021, it became apparent that government programs had in fact helped to stabilize the economy. The most dire modeling inputs that had been used by management teams in some instances to generate very significant CECL reserves were shown to have been outside the range of reasonableness. Beginning in Q3-2020, the sampled entities were generally no longer building ACL levels and in fact began to reverse some of the excess provisioning, with reversals continuing into Q4-2020 and Q1-2021. The Wall Street Journal headlined the situation (“Banks Eye Cash Reserves for Profits” March 16, 2021, p. A1; “Banks to Free Up Stockpiles,” p. A2), noting:
Banks set their loan loss-expectations using broad economic gauges, particularly [GDP] and unemployment rates. From that starting point, they factor in hundreds of other variables, with some leeway built in for how optimistic or pessimistic executives feel … bank executives have tended toward pessimism … As a result, their internal models offer a more muted outlook than broader economic forecasts. The Organization of Economic Cooperation and Development recently said it expects US [GDP] to grow 6.5% this year, double its December forecast. Economists are projecting unemployment, now down to 6.2%, to keep falling to 5% by December according to a [WSJ] survey.
And such reversals have continued throughout the year and into Q3-2021, as evidenced by CNBC.com reports on October 13, 2021, regarding JPMorgan and on October 14, 2021 for Bank of America and Wells Fargo.
The authors’ research outlined above implies that management influence in Q1-2020 and Q2-2020 CECL reserving represented a good deal more than just “some leeway,” as evidenced by the wide disparity in key inputs for GDP and unemployment rates used and additional “judgment” provisioning. Furthermore, in the PCAOB’s October 2021 “Spotlight—Staff Update and Preview of 2020 Inspection Observations,” it noted that it had extended its 2020 review to include “a sample of audits of public companies with fiscal years ended primarily between March 31, 2020, and June 30, 2020” (p. 6) in addition to its normal review of 2019 related audits, in order to gain insights about the impact from the COVID-19 pandemic. Among its observations, it noted that the auditing of accounting estimates needed improvement and singled out the auditing of allowances for loan losses: “While we have observed improvement in auditing accounting estimates, deficiencies continue to occur, particularly in auditing the allowance for loan losses (ALL) … Auditors reviewed management’s memorandum describing assumptions used in determining the ALL but did not evaluate evidence supporting certain assumption changes from the prior year, or lack of changes, when evaluating the reasonableness of such assumptions” (pp. 9–10). In addition, regarding deficiencies in auditing other estimates, it noted: “Auditors did not obtain sufficient appropriate audit evidence to support the assumptions used, or perform procedures to resolve any known contradictory evidence, when evaluating the reasonableness of financial statement forecasts” (p. 10).
Clearly, COVID-19–related matters and auditing allowances for loan losses already have the PCAOB’s attention, with the board concerned about auditors not adequately addressing the reasonableness of assumptions or considering contradictory evidence. The forthcoming next round of PCAOB inspections will include the 2020 adoption and implementation of CECL within the COVID-19 environment and will deal with the potential aberrations the authors’ research has highlighted.
The Possible Evolution of CECL Reporting
The implementation of CECL as detailed in ASU 2016-13 required many new techniques and the use of enhanced analytical and forecasting models, as well as significant management effort. As FASB noted when it issued the ASU, it will “broaden the information that an entity must consider in developing its expected credit loss estimate for assets measured either collectively or individually.” Although the data output from CECL modeling represents a significantly enhanced “baseline” for management to assess the adequacy of the ACL, ultimately the determination of related critical modeling assumptions and the overall level of the ACL reflects management’s judgment in fulfilling its responsibility to prepare the financial statements.
What may appear as relatively small changes in underlying assumptions could alter CECL modeling outputs dramatically. AS 2501.15 specifically cautions “significant assumptions include those that … are sensitive to variation, such that minor changes in the assumption can cause significant changes in the estimate.” How can accounting theory and auditing standards contend with the changing assumptions about year-end unemployment rates disclosed by JPMorgan and Flagstar? Although disclosed, does Flagstar’s assumption that the economy would only be flat in the second half of the year appear reasonable when the general consensus per the Conference Board in June was for a surge of 32.2% for Q3? Where can one look to establish a benchmark for reasonableness within the context of such differing and changing views by various management teams? As Trustmark commented in its Form 10-Q: “During the second quarter of 2020, Trustmark revised its ACL methodology in order to prevent the econometric models from extrapolating beyond reasonable boundaries of their input variables. Trustmark chose to establish an upper and lower limit process when applying the periodic forecasts.” Of course, other entities could experience the same stresses.
Furthermore, because the underlying basis of financial reporting is often considered to embrace a convention of conservatism, it is not surprising that management might determine that an additional judgmental increase to the ACL is warranted during times of significant macroeconomic stress and uncertainty. Making such management adjustments, however, will likely be more challenging when encountering a situation in which the CECL baseline analysis indicates that a particular ACL level is warranted, but management instead judges that a significantly lower level is more appropriate. On the other hand, what if management just tweaked the underlying assumptions within the modeling and generated a more agreeable CECL baseline? Much of the discussion and guidance in the recently issued SAS 143 and AS 2501 caution auditors to be alert for such discrepancies and critical estimating junctures—and such tweaking of assumptions does not seem consistent with the rationale behind ASU 2016-13 and FASB’s inherent desire to bring more rigor, insight, and timeliness into the estimating process after enduring the stresses using the “incurred loss” model during the 2008/09 financial crisis.
What may appear as relatively small changes in underlying assumptions could alter CECL modeling outputs dramatically.
Although disclosing the assumptions used can provide insight into an ACL’s conservatism and management’s thinking, presently such disclosures are being made in the supplemental information provided by management at the time of the earnings release. It might evolve that such disclosures of key assumptions will become part of the financial statements—perhaps included in the notes with commentary regarding their impact on the sensitivity of the estimate calculation. Wells Fargo’s comments in its Form 10-Q cited above are an example of what might transpire. Although increased disclosure may be useful, however, the variability in modeling processes and critical assumptions might lead to an offsetting loss of comparability.