October 7, 2016
VIA EDGAR AND ELECTRONIC MAIL
Stephanie L. Sullivan
Senior Assistant Chief Accountant
Office of Financial Services
Securities and Exchange Commission
Mail Stop 4720
Washington, D.C. 20549
Re: Wells Fargo & Company
Form 10-Q for the Quarter Ended June 30, 2016
Filed August 3, 2016
File No. 001-02979
Dear Ms. Sullivan:
In response to the comments by the staff (“Staff”) of the Securities and Exchange Commission (“Commission”) contained in the Staff’s letter dated September 23, 2016, to Wells Fargo & Company (“Wells Fargo,” “Company,” “we,” or “our”), we submit the following information. The Staff’s comments, indicated in bold, are followed by Wells Fargo’s responses.
Form 10-Q for the Quarter Ended June 30, 2016
Risk Management – Credit Risk Management, page 22
Purchased Credit-Impaired Loans, page 23
1. | We note your disclosure that the carrying value of PCI loans at June 30, 2016, which included $1.0 billion from the GE Capital acquisitions, totaled $19.3 billion, compared with $20.0 billion at December 31, 2015. These balances compare to an accretable difference of $15.7 billion and $16.3 billion as of June 30, 2016 and December 31, 2015, respectively. Please respond to the following: |
• | Tell us why the accretable difference relative to the carrying value of PCI loans continues to increase each period. For example, it appears from disclosures in your prior year filings that the accretable difference relative to the carrying value of PCI |
Stephanie L. Sullivan
October 7, 2016
Page 2
loans has increased from 73% at December 31, 2013, to 76% at December 31, 2014, to 82% at December 31, 2015, and to 89% at June 30, 2016.
Wells Fargo response:
As detailed in the table below, the relationship of the accretable yield balance to PCI carrying value was 65%, 76%, 82% and 81% (85% excluding the GE Capital portfolio acquisitions) at December 31, 2013, 2014, 2015 and June 30, 2016, respectively. 
Changes in the carrying value of PCI loans are primarily driven by loan amortization and resolutions, either through loan liquidations or voluntary customer prepayments. The accretable yield represents the excess of cash flows expected to be collected over carrying value. The rate of accretion into interest income is based on the lifetime yield of our expected cash flows. While we recognize interest income based on an effective yield method, the projected yield embedded in our cash flow estimates for every period fluctuates based on the expected timing of loan resolutions, both liquidations and customer prepayments. As such, the expected ratio of the accretable yield to the carrying value will vary somewhat throughout the contractual life of the portfolio based on the projected timing of the expected cash flows. Aside from the normal accretion into interest income, changes in the accretable yield are impacted by reclassifications from nonaccretable difference for lower expected losses on loans with improving credit-related cash flows and other changes in expected cash flows that do not affect nonaccretable difference, such as changes in interest rate indices for variable rate PCI loans and changes in prepayment assumptions. The estimated accretable yield is updated when we have a significant change in our expected cash flows associated with our PCI portfolios and we adjust our prospective yield accordingly.
The following table provides a rollforward of the changes in the ratio of the accretable yield to the carrying value from January 1, 2014 through June 30, 2016.
Stephanie L. Sullivan
October 7, 2016
Page 3
Based on the above, the increase in the ratio of accretable yield balance to PCI carrying value from December 31, 2013 through December 31, 2015, was primarily due to the reclassifications from nonaccretable difference. During the six months ended June 30, 2016, the reduction in the ratio of accretable yield to PCI carrying value was primarily attributable to the impact of the GE Capital portfolio acquisitions, partially offset by the reduction in carrying value due to prepayments. Prepayments decrease our PCI carrying value at the time of prepayment while the related change in the accretable yield for the prepaid loans occurs when there is a significant change in expected cash flows. Estimated cash flows are updated quarterly to assess when a significant change has occurred. As such, the reduction in carrying value during the six months ended June 30, 2016, contributed to a 3% increase in the ratio of the accretable yield to carrying value as detailed in the table above. For more information about prepayment impacts see discussion below in response to the question regarding the reduction in carrying value. The nonaccretable difference reclassifications and impacts on carrying value of our Pick-a-Pay PCI portfolio caused by prepayments were the result of observed and forecasted economic strengthening, particularly in housing prices, and our loan modification efforts, both of which lowered expected defaults and losses, and increased prepayments.
• | In regards to the PCI portion of the Pick-a-Pay portfolio discussed in more detail on page 30, please tell us why the weighted average life has not decreased significantly from prior periods. For example, the weighted average life was 12.7 years at December 31, 2013, decreased to 11.7 years as of December 31, 2014, and has remained fairly consistent with that level through June 30, 2016, when the weighted average life was 11.5 years. |
Stephanie L. Sullivan
October 7, 2016
Page 4
Wells Fargo response:
The estimated weighted average life of the Pick-a-Pay PCI portfolio is influenced by passage of time, loan modifications, and our expectations for both defaults and prepayments included in our cash flow estimation process, which utilizes internally developed models for this portfolio. The weighted average life is an output from our quarterly cash flow estimation process. Since December 31, 2013, our weighted average life has declined by 1.2 years. While passage of time would have resulted in a decline of 2.5 years, the following factors partially offset the expected decline by 1.3 years during this period:
• | Expected default rates significantly improved from 2013, reflecting observed and forecasted economic strengthening, primarily in housing prices, and when combined with the impacts of our loan modification efforts, lengthened our estimate of the portfolio’s weighted average life and increased our estimate of accretable yield as noted above. |
• | Expected prepayments increased from 2013 and, when combined with passage of time, offset the lengthening in the weighted average life from the improvement in expected defaults. We have calibrated our model estimates for prepayments two times during this period to reflect the observed trends in prepayment activity in this portfolio. |
While we calibrate our model prepayment estimates as necessary to reflect improved prepayment and default experience, historically, we have expected that the credit-stressed borrower characteristics and distressed collateral values associated with our Pick-a-Pay PCI loans would limit the ability of these borrowers to prepay their loans. However, in second quarter 2016, we observed a sharp increase in prepayments in the modified segment of our Pick-a-Pay PCI portfolio, (i.e. approximately 70% has been modified) and some leveling off of defaults. While we typically observe seasonal increases in prepayments in the second quarter, the elevated prepayment level continued in third quarter 2016, indicating a trend. Unlike the relationships we have previously observed (as described above) where improvements in defaults and loan modification efforts have more than offset our updated prepayment expectations, we believe there will be limited further improvement in defaults such that increases in updated prepayment expectations are expected to result in a net reduction in our estimated weighted average life.
• | Tell us in more detail the factors driving the $2.7 billion reduction (representing 13%) in the PCI portion of the Pick-a-Pay portfolio carrying value during the second quarter of 2016. As part of your response, please tell us how this reduction in carrying value was factored into your analysis to not change the accretable yield during the second quarter of 2016. |
Wells Fargo response:
As detailed in the table below, the carrying value of the Pick-a-Pay PCI portfolio declined by $835 million or 4.5% (18% annualized) in second quarter 2016, $642 million or 3.4% (14% annualized) in first quarter 2016, and $2.5 billion or 12% during 2015.
Stephanie L. Sullivan
October 7, 2016
Page 5
Reductions in carrying value are driven primarily by loan amortization and resolutions, either liquidations or voluntary customer prepayments. The higher rate of carrying value reduction in second quarter 2016 was primarily attributed to increased levels of observed customer prepayments, which was consistent with our typical experience that prepayments are normally elevated in the second quarter due to seasonal factors impacting the home purchase market. Prepayments in second quarter 2016 increased 45% compared to first quarter 2016, which was consistent with the relative percentage increase we observed in second quarter 2015 compared to first quarter 2015. In addition, we have historically expected that the credit-stressed borrower characteristics and distressed collateral values associated with our Pick-a-Pay PCI loans would limit the ability of these borrowers to prepay their loans. Since the reduction in second quarter 2016 carrying value was not significantly inconsistent with our expectations, we did not change our accretable yield at that time.
• | Tell us the process you perform to evaluate the weighted average life and accretable yield for this portfolio. As part of your response, describe how frequently this process is performed, and describe any internal controls over the process. |
Wells Fargo response:
On a quarterly basis, we update our estimates of cash flows expected to be collected for our Pick-a-Pay PCI portfolio. This process involves using models that estimate the expected payments, defaults, losses, and prepayments over the estimated life of the portfolio. These models are validated by a separate model validation team. In reviewing the quarterly cash flow estimates, we review the trend of actual cash collected during the preceding periods to what was projected by the model estimates. We also review model performance for all key model projections such as prepayments, defaults, and losses on a quarterly basis to identify when a model may need to be calibrated. The weighted average life and accretable yield are outputs from this cash flow estimation process. Changes in the projected cash flows and weighted average life are rolled forward and attributed to key drivers and updates on a quarterly basis. Estimates, including any calibrations, associated with the Pick-Pay PCI portfolio are reviewed and challenged quarterly, prior to finalization of the PCI accounting and disclosures, by an executive level committee comprised of representatives from Corporate Finance (including the Corporate Controller and members of his team), Corporate Risk (including the Chief Risk Officer, the Chief Credit Officer and members of their teams) and the Home Mortgage line of business.
Stephanie L. Sullivan
October 7, 2016
Page 6
Model performance directly impacts our estimates of the weighted average life and accretable yield of the portfolio. The performance of the models are monitored and reviewed by subject matter experts from a cross-functional group of team members (e.g. corporate controller and credit monitoring groups). When deterioration in model performance outside of our pre-defined thresholds for tolerance occurs, it signals that a model needs to be replaced. Deterioration in model performance can occur when there are rapid and significant changes in economic variables such as house price appreciation or loan level borrower characteristics that were not observable for inclusion in the development of the model. The new model development process can take 6 to 12 months to complete and another 3 to 6 months for the model to be validated by key stakeholders, including validation by our internal model validation team, according to regulatory requirements. During this time period, depending on the magnitude and persistence in the trend of variances we observe in our model performance monitoring and cash flow back-testing results, we may calibrate the existing under-performing model in order to improve the estimate during the period of model re-development. As noted above, we have continually calibrated our models to reflect our best estimate of defaults, prepayments and losses based on the observable trends and expectations for each quarter. As part of our normal ongoing process, we will be implementing a new model in third quarter 2016 that will incorporate our most recent experience. We believe this new model will provide a better estimate of the expected defaults, prepayments and losses for the Pick-a-Pay PCI portfolio.
• | Explain how the significant improvements in the credit quality of the borrowers (as evidenced by higher FICO scores) and significantly lower LTVs for the Pick-a-Pay portfolio are factored into your analysis to evaluate the weighted average life and accretable yield for this portfolio. |
Wells Fargo response:
We rely on models to project the defaults, losses and prepayments associated with this Pick-a-Pay Portfolio. When a model is developed, historical loan level performance data, loan level attributes, and economic variables are evaluated to determine the correlation amongst all of these variables. Prior to the recent financial crisis, the industry did not have significant loan modification activities. As a result of the financial crisis, we have worked to estimate the expected performance of these customers after they receive a loan modification. Over time we continue to observe and collect more performance data on these loans and evaluate economic variables such as housing prices and unemployment. As models are developed, the sensitivity and correlation of the projected defaults, losses and prepayments to each loan level attribute such as LTV and FICO and each economic variable such as housing price index, may change as we are able to build upon the historical data observations and derive updated correlations to each of these variables to enhance our estimates. Since fourth quarter 2014, we have not observed significant improvement in the weighted average FICO score of this Pick-a-Pay PCI portfolio (632 in fourth quarter 2014; 640 in fourth quarter 2015; 643 in second quarter 2016). We have, however, observed significant improvement in the weighted average LTV ratios as calculated based on the total unpaid principal balance of the borrower
Stephanie L. Sullivan
October 7, 2016
Page 7
to the estimated collateral value (83% at December 31, 2014; 79% at December 31, 2015; and 74% at June 30, 2016).
The LTVs reported in table 19 of our second quarter of 2016 Form 10-Q are calculated based on our ‘adjusted unpaid principal balance’ which is net of write-downs to collateral value on severely delinquent loans. The LTVs based on the gross unpaid principal balance, as well as the current FICO scores are inputs to our model to estimate the projected defaults, losses and prepayments, which ultimately drive the estimated weighted average life and accretable yield.
Given the minimal improvement in borrower FICO scores, the distressed credit profile of these borrowers, and the historical data set used to develop our current models; our models to date have estimated that these borrowers will have limited ability to refinance their loans, despite the recent improvement in house prices. While we have continually calibrated the models as necessary to reflect the observable trends and expectations, more recently, with the reduction in the average borrower LTV to below 80%, which is a key refinancing inflection point, and observed continued elevated prepayments in third quarter 2016, we believe these borrowers now have more opportunity to refinance their loans. As a result of these trends, we believe the prepayments will remain elevated and accordingly we have updated the accretable yield and weighted average life estimates to reflect the significant change in expected cash flows. As described above, the accretable yield and weighted average life were estimated using a new model implemented in third quarter 2016 that incorporates the observed price appreciation and prepayments.
________________________________
The Company acknowledges that:
•the Company is responsible for the adequacy and accuracy of the disclosure in the filing;
•Staff comments or changes to disclosure in response to Staff comments do not foreclose the Commission from taking any action with respect to the filing; and
•the Company may not assert Staff comments as a defense in any proceeding initiated by the Commission or any person under the federal securities laws of the United States.
Stephanie L. Sullivan
October 7, 2016
Page 8
Questions concerning the information set forth in this letter may be directed to me at
(415) 222–3119.
Very truly yours,
/s/ RICHARD D. LEVY
Richard D. Levy
Executive Vice President and Controller
(Principal Accounting Officer)
cc: John G. Stumpf, Chairman and Chief Executive Officer
John R. Shrewsberry, Senior Executive Vice President and Chief Financial Officer