Regulators: CECL Implementation Considerations for the Smaller Less Complex Institutions Small Banks Must Find The Data

Regulators: CECL Implementation Considerations for the Smaller Less Complex Institutions Small Banks Must Find The Data Peter Cherpack, EVP & Partner, Ardmore Banking Advisors, Inc April 18, 2018 CECL A Non Event for Smaller Community Banks? The implementation of the new CECL accounting standard has proven to be difficult to understand for banks of all sizes, though it is the smaller institutions that have proven to be the slowest to get started getting prepared for the rule s implementation. During a CECL informational webinar given by Ardmore Banking Advisors in early March 2018, 44% of the banks under $500 million in assets attending responded to a poll about where they were in their preparation for CECL with the answer doing research. The month before, when talking with a CEO of a smaller community bank in the Midwest about CECL, he told me I am in a group of some 25 CEO s of local community banks, and we all agreed that we weren t going to do anything about CECL until the FDIC or the State tells us we have to. Until February 27 th s Ask the Regulators webinar titled Practical methods that smaller, less complex community banks can use for a starting place for CECL the general message from the regulatory agencies had been: CECL is scalable for smaller institutions, if you are smaller and less complex you can probably pretty much do as you do your ALLL today. For example, on April 7th of 2016, the FDIC Advisory Committee on Community Banking presented thoughts on CECL, and Robert Storch, Chief Accountant of the FDIC s Division of Risk Management Supervision stated the following: If you have a fairly straightforward approach today, even using Excel spreadsheets and so forth, there is an expectation that you should be able to continue to use that type of an approach there's no expectation that community institutions will need to apply a complex model or use costly outside vendors in order to implement the new model. Webinars and presentations since then from the Agencies typically repeated that same line of thinking, with FAQ s issued supporting it. Buoyed by the agencies stated view that the implementation of CECL was essentially a non event for smaller community banks, most smaller institutions did little or nothing to prepare for CECL through early 2018, other than gather industry information, attend vendor webinars and worry about what they may have to do in 2021 to be compliant. A Game Changer for Smaller Institutions Or Non Starter? Ask the Regulators CECL Webinar February 27, 2018 The seemingly justified paralysis of the smaller community banks relating to CECL was finally jolted by the February 2018 webinar titled: Practical methods that smaller, less complex community banks can use for a starting place for CECL which featured speakers from the FDIC, Fed, Conference of State Bank Supervisors, FASB and SEC addressing smaller bank CECL compliance. It was notable that it took the agencies almost two years from the CECL rule s implementation to directly address smaller bank concerns. April, 2018 Ardmore Banking Advisors, Inc www.ardmoreadvisors.com 1

While most of the speakers at that webinar tried to convey an understanding that there are lower expectations for CECL compliance for smaller less complex institutions than larger banks, several of the comments from the speakers indicated that CECL compliance was still going to be difficult and costly for even the smallest institutions. Some key concepts articulated during the webinar included: A bank can do the actual CECL loss rate calculations many ways, and some banks may be able to use various spreadsheet models to do so The agencies expect all banks, no matter how small, to give their best efforts to being CECL compliant, but that these efforts need to get better over time Creating a reasonable and supportable life of loan estimate is very difficult and challenging particularly for a smaller institution Most smaller banks will need to have some type of data warehousing to hold and manage in a controlled fashion the data needed for CECL A bank s internal data will often be insufficient for CECL loss modeling and smaller banks will probably need to acquire 3rd party or peer data All banks should start now collecting risk characteristic data for CECL and must develop a data retention plan No bank, no matter how small, is or will be exempt from CECL While these concepts may not be new, this was the first time we heard from the Agencies that smaller less complex institutions DO have to worry about CECL, may have to change loan segmentation, will probably have to acquire 3 rd party peer data and a data warehouse. That is a far cry from the previous message to the smaller banks: you can probably keep doing what you have been doing. CECL Compliance Requires Additional Resources and New Practices for All Banks The topics discussed at the February Ask the Regulators webinar show that the Agencies are concerned that smaller banks are doing very little about CECL, and further that they understand that smaller banks are particularly challenged to create meaningful life of loan loss rates and justifiable reasonable and supportable projections based on market economic conditions. One section of the webinar was dedicated to discussing the challenges smaller banks, in particular, have with managing the credit data needed to support CECL, and was presented by Robert Storch of the FDIC. Mr. Storch walked through issues including how difficult it would be to establish meaningful and supportable CECL life of loan loss calculations when loan losses are minimal or nonexistent for particular segments of loan portfolio, when loan losses are sporadic so there are no observable, predictable patterns, and when loan pools or portfolio segments have a limited number loans and loss data only available for a short historical period. He commented: Which typically means management needs to look to loss data from external sources such as peer data. April, 2018 Ardmore Banking Advisors, Inc www.ardmoreadvisors.com 2

While much of the webinar was spent demonstrating different spreadsheet methods to calculate life of loan loss rates, other components of CECL compliance with more complexity were also emphasized by the presenters. Several times it was stated that it was not recommended that a bank, even a smaller less complex one, store their credit data in spreadsheets, some warehousing is needed though that could be with the assistance of the bank s core accounting system provider. While the CECL loss calculations themselves can be based in spreadsheets, they emphasized that the data that would eventually need to be input into spreadsheets would likely be too voluminous to adequately control. At one point it was stated that the smaller institution s current ALLL collective asset pooling and segmentation methods, typically by regulatory call code category may be appropriate for CECL or it may not though no specific criteria was explained. Kyle Thomas of CSBS suggested: CECL allowances are bases on life time loan losses, this is the starting point if you don t have lifetime historical loss data you will need to turn to other sources of data for your starting point. Mr. Thomas underlined credit data management challenges for smaller institutions including: The methods you use may be determined by how much data is available to you, you may start out thinking one method is better than the other and find out that it just does not work either because of data issues or improperly skewed results We are not suggesting that you maintain loan data in a spreadsheet, in fact as Joanne [Wakim of the Federal Reserve] mentioned you may need additional warehousing to accommodate the volume of data required You may need to consider peer data, other external data, you may need to consider proxies, and there s other extrapolations, interpolations Where Can Smaller Less Complex Institutions Go From Here? It is obvious to the industry and regulatory bodies that smaller less complex institutions typically have fewer resources, less staff, data and automation than their larger sized brethren. There is also a stated desire by the Agencies that CECL should not cause an undue burden or add significant additional expenses for the smaller institutions. At the same time, as was evidenced at the recent Ask the Regulators webinar, most smaller bank s data, data management automation and controls will likely be inadequate for CECL compliance. Where do they go from here? Historically community banks have been behind in technology investments and data management compared to larger financial institutions. Typically, their tools, practices and systems have been added over time only as needed, without a holistic view of current data management practices. The it s not broken, why fix it? legacy of small bank operations and data management has made compliance with data dependent regulations like CECL a significant challenge. That understood, CECL compliance may be a great way to justify implementing long needed advances in credit data management by community banks. April, 2018 Ardmore Banking Advisors, Inc www.ardmoreadvisors.com 3

Let s take a look at what historical and current credit data and resources are available to most community banks, even the smallest ones: Individual historical loan loss events and by asset pool (current ASC 450 20 method ALLL pools, usually call code) FDIC/UBPR peer loss data Current and recent (6 12 month) loan performance data from the core accounting system Future enhanced core risk characteristic and loan origination data Other third party provided data sources There are clearly some viable components of credit data and loss history by asset class that can be readily accessed by a smaller institution. Would piecing together historical data from multiple sources at different levels of detail and categorization be acceptable to the Accounting Firms and Regulators for CECL? For example, all banks have saved their ALLL spreadsheets for years, and they contain valuable bank specific loss and segmentation information but unfortunately the data is typically limited to more blunt credit categorizations like call code, account number, charge off amount and date. This alone is not sufficient to create meaningful loss patterns or curves with any real credit risk characteristics, asset performance or meaningful trending. All banks can access the FDIC peer group web site and create aggregate historical portfolio performance data at the call code level. This data can be aggregated based on certain bank characteristics to make the data more relevant for a particular bank as a peer group. Every bank typically has at least about a year of their core data available at no significant additional cost, and this data is linked to some risk characteristic data categories like origination date, industry and purpose codes, risk rating and possibly even borrower financial data like LTV, DSCR and Credit Score (quality and consistency not withstanding). Based on the information shared by speakers in February s Ask the Regulators webinar, most smaller banks will realize that they need to augment their core accounting system booking process and collect and retain more risk characteristic data starting now, regardless if portfolio segmentation can be done by call code. On a go forward basis the institution can do a much better job creating historical loss patterns and loss curves with more granular underlying risk driver related data. As these data collection practices become the new normal, over time even the smallest banks will have better historical data to use for CECL loss modeling. As the years pass, the oldest and least specific data will fall away and be replaced with newer, better data. While third party peer data about portfolio performance at the asset level can be purchased today, smaller banks as usual have some additional challenges if they pursue this avenue. First, the cost can be significant, as to date most vendors have positioned their data services to larger institutions with significantly deeper pockets and in house analytical resources. Second, the type of data available April, 2018 Ardmore Banking Advisors, Inc www.ardmoreadvisors.com 4

contain typically national and agency rated credits that may or may not be relevant to smaller community bank borrowers and credit risk. While some industry organizations like the Risk Management Association are creating peer data sharing groups, these are usually limited to actual loss history, and do not the loan performance history over time needed to establish meaningful loss curves and patterns for CECL justifications. The RMA CECL Loss Data service is the only moderately priced source of credit data readily accessible to smaller institutions, but is limited to those banks that are willing to share their data and are members of RMA. A Modest Proposal for Smaller Community Institution CECL Loss Estimations Based on the resources readily available to smaller less complex financial institutions, it is possible that loss history and justification for CECL loss estimations could be put together this way: Under the scenario depicted above, an institution would start by collecting historical peer data from the FDIC website for as far back as practical and available, hopefully through a full economic cycle (10 year or more). At the same time, they would gather their own historical loss history with as much associated risk characteristic data as possible. In theory the bank could blend and smooth these data sources together to come up with a bare bones loss history for a full economic cycle. Other free sources of viable historical credit loss data could include Fannie and Freddie data for residential portfolios. Some community bankers have suggested that the Agencies themselves could make their loan performance data available to help out smaller institutions. For most smaller community banks CRE is a significant exposure, but currently there is little publicly available CRE historical data outside of the FDIC database at the call code level. As a next step the institution would layer in current and most recent loan performance data, likely the most recent 12 months, and include more risk characteristic data available in their core. (It is likely that for most community banks credit data will suffer from the garbage in garbage out syndrome April, 2018 Ardmore Banking Advisors, Inc www.ardmoreadvisors.com 5

with their core credit data but even with inconsistent quality, it is somewhat better than no data at all). With this current and recent portfolio data some loss patterning can start, looking at origination dates, asset types and risk ratings. Finally, on a go forward basis, better, more detailed and controlled credit data will be collected and archived by the institution as soon as they complete their internal data assessment and data management practice reviews. This data would be only available from some point in 2018 but would be used to better pattern portfolio asset performance and loss curves. The amount of this data would increase as time passes and become more meaningful and justifiable as the old data drops away. There are a number of real concerns surrounding using a methodology like this for CECL compliance. Using FDIC Call Code level peer data is not really risk characteristic based which is a stated key tenant of all CECL estimations. Changes in quarterly loss rates would not be reasonably supported by the bank s own data, or any actual explainable credit activities. Historical call code based pools can represent a mix of origination dates, maturities, and risk rated/non rated credits, which makes it difficult to estimate loss curves or patterns. The institution would have to come up with a supportable method to blend peer UPBR loss data with the institution s own internal data, and use extensive Q factors to account for adjustments. Would this method qualify for a good faith effort for smaller community banks in 2021? How small and less complex would a bank have to be to be allowed to use a methodology like this one? Conclusion The CECL Journey Has Some Complexity for Banks of All Sizes While best practices related to CECL compliance are still emerging, most information available on CECL data management are based on the large bank early adopters and other larger financial institutions. Most CECL methodologies and tools are based on assumptions of extensive data availability and the institution having available accounting and modeling resources in house. Considering the amount of subjectivity and flexibility purposely build in to CECL, the CECL journey is long and offers many twists and turns that require testing, research and support. The smallest institutions are those that are the least equipped to cope with uncertainty and open standards, and are now looking for answers from their auditors, accounting firms, regulators and vendors. Based on the revelations recently brought to light in the February Ask the Regulator webinar addressing CECL loss estimation for smaller less complex institutions even the smallest banks and credit unions need to start now, collect and retain more of their own credit risk characteristic data and likely obtain 3 rd party peer data as well. All of this data needs to be archived in some type of auditable data warehouse, with appropriate controls and data management practices. Can the smaller less complex institutions use data and resources readily available like FDIC peer group and other free resources to build their CECL life of loan loss data patterns? While there are ways to do it, such practical best practices have not yet emerged. It is time for the industry to look hard at what really is reasonable and supportable to expect from our smallest banks and credit unions April, 2018 Ardmore Banking Advisors, Inc www.ardmoreadvisors.com 6

relating to CECL as these companies represent the largest number of institutions in our financial marketplace today. For now, they know they need to collect more data, but there is a responsibility of those that oversee the implementation of the CECL accounting standard to continue to help establish practical guidelines and best practices that are readily achievable by our smallest financial institutions. Perhaps reasonable and achievable practices like those outlined in this paper can be a way to increase the dialogue on this important idea. April, 2018 Ardmore Banking Advisors, Inc www.ardmoreadvisors.com 7