Assessments

Federal RegisterJul 13, 2015

Ask Donna

What actually matters in this document.

Text

FEDERAL DEPOSIT INSURANCE CORPORATION

12 CFR Part 327

RIN 3064-AE37

Assessments

AGENCY:

Federal Deposit Insurance Corporation (FDIC).

ACTION:

Notice of proposed rulemaking (NPR) and request for comment.

SUMMARY:

The FDIC is proposing to amend 12 CFR part 327 to refine the deposit insurance assessment system for small insured depository institutions that have been federally insured for at least 5 years (established small banks) by: revising the financial ratios method so that it would be based on a statistical model estimating the probability of failure over three years; updating the financial measures used in the financial ratios method consistent with the statistical model; and eliminating risk categories for established small banks and using the financial ratios method to determine assessment rates for all such banks (subject to minimum or maximum initial assessment rates based upon a bank's CAMELS composite rating). The FDIC does not propose changing the range of assessment rates that will apply once the Deposit Insurance Fund (DIF or fund) reserve ratio reaches 1.15 percent; thus, under the proposal, as under current regulations, the range of initial deposit insurance assessment rates will fall once the reserve ratio reaches 1.15 percent. The FDIC proposes that a final rule would go into effect the quarter after a final rule is adopted; by their terms, however, the proposed amendments would not become operative until the quarter after the DIF reserve ratio reaches 1.15 percent.

DATES:

Comments must be received by the FDIC no later than September 11, 2015.

ADDRESSES:

You may submit comments on the notice of proposed rulemaking using any of the following methods:

•

Agency Web site: http://www.fdic.gov/regulations/laws/federal/.

Follow the instructions for submitting comments on the agency Web site.

•

Email: comments@fdic.gov.

Include RIN 3064-AE37 on the subject line of the message.

•

Mail:

Robert E. Feldman, Executive Secretary, Attention: Comments, Federal Deposit Insurance Corporation, 550 17th Street NW., Washington, DC 20429.

•

Hand Delivery:

Comments may be hand delivered to the guard station at the rear of the 550 17th Street Building (located on F Street) on business days between 7 a.m. and 5 p.m.

•

Public Inspection:

All comments received, including any personal information provided, will be posted generally without change to

http://www.fdic.gov/regulations/laws/federal.

FOR FURTHER INFORMATION CONTACT:

Munsell St.Clair, Chief, Banking and Regulatory Policy, Division of Insurance and Research, 202-898-8967; Nefretete Smith, Senior Attorney, Legal Division, 202-898-6851; Thomas Hearn, Counsel, Legal Division, 202-898-6967.

SUPPLEMENTARY INFORMATION:

I. Policy Objectives

The Federal Deposit Insurance Act (FDI Act) requires that the FDIC Board of Directors (Board) establish a risk-based deposit insurance assessment system.

Pursuant to this requirement, the FDIC adopted a risk-based deposit insurance assessment system effective in 1993 that applied to all banks.

A risk-based assessment system reduces the subsidy that lower-risk banks provide higher-risk banks and provides incentives for banks to monitor and reduce risks that could increase potential losses to the DIF. Since 1993, the FDIC has met its statutory mandate and has pursued these policy goals by periodically introducing improvements in the deposit insurance assessment system's ability to differentiate for risk. The primary purpose of the proposals in this NPR is to improve the risk-based deposit insurance assessment system applicable to small banks to more accurately reflect risk.

12 U.S.C. 1817(b). A “risk-based assessment system” means a system for calculating an insured depository institution's assessment based on the institution's probability of causing a loss to the DIF due to the composition and concentration of the institution's assets and liabilities, the likely amount of any such loss, and the revenue needs of the DIF. See 12 U.S.C. 1817(b)(1)(C).

As used in this NPR, the term “bank” is synonymous with the term “insured depository institution” as it is used in section 3(c)(2) of the FDI Act, 12 U.S.C 1813(c)(2).

On January 1, 2007, the FDIC instituted separate assessment systems for small and large banks. 71 FR 69282 (Nov. 30, 2006). See 12 U.S.C. 1817(b)(1)(D) (granting the Board the authority to establish separate risk-based assessment systems for large and small insured depository institutions).

As used in this NPR, the term “small bank” is synonymous with the term “small institution” as it is used in 12 CFR 327.8. In general, a “small bank” is one with less than $10 billion in total assets.

II. Background

Risk-Based Deposit Insurance Assessments for Small Banks

Since 2007, assessment rates for small banks have been determined by placing each bank into one of four risk categories, Risk Categories I, II, III, and IV. These four risk categories are based on two criteria: capital levels and supervisory ratings. The three capital groups—well capitalized, adequately capitalized, and undercapitalized—are based on the leverage ratio and three risk-based capital ratios used for regulatory capital purposes.

The three supervisory groups, termed A, B, and C, are based upon supervisory evaluations by the small bank's primary federal regulator, state regulator or the FDIC.

Group A consists of financially sound institutions with only a few minor weaknesses (generally, banks with CAMELS

composite ratings of 1 or 2); Group B consists of institutions that demonstrate weaknesses that, if not corrected could result in significant deterioration of the institution and increased risk of loss to the DIF (generally, banks with CAMELS composite ratings of 3); and Group C consists of institutions that pose a substantial probability of loss to the DIF unless effective corrective action is taken (generally, banks with CAMELS composite ratings of 4 or 5). An institution's capital and supervisory group determine its risk category as set out in Table 1 below.

The common equity tier 1 capital ratio, a new risk-based capital ratio, was incorporated into the deposit insurance assessment system effective January 1, 2015. 79 FR 70427 (November 26, 2014). Beginning January 1, 2018, a supplementary leverage ratio will also be used to determine whether an advanced approaches bank is: (a) well capitalized, if the bank is subject to the enhanced supplementary leverage ratio standards under 12 CFR 6.4(c)(1)(iv)(B), 12 CFR 208.43(c)(1)(iv)(B), or 12 CFR 324.403(b)(1)(vi), as each may be amended from time to time; and (b) adequately capitalized, if the bank is subject to the advanced approaches risk-based capital rules under 12 CFR 6.4(c)(2)(iv)(B), 12 CFR 208.43(c)(2)(iv)(B), or 12 CFR 324.403(b)(2)(vi), as each may be amended from time to time. 79 FR 70427, 70437 (November 26, 2014.) The supplementary leverage ratio is expected to affect the capital group assignment of few, if any, small banks.

The term “primary federal regulator” is synonymous with the term “appropriate federal banking agency” as it is used in section 3(q) of the FDI Act, 12 U.S.C. 1813(q).

A financial institution is assigned a composite rating based on an evaluation and rating of six essential components of an institution's financial condition and operations. These component factors address the adequacy of capital (C), the quality of assets (A), the capability of management (M), the quality and level of earnings (E), the adequacy of liquidity (L), and the sensitivity to market risk (S).

Table 1—Determination of Risk Category

Capital group

Supervisory group

CAMELS 1 or 2

CAMELS 3

CAMELS 4 or 5

Well Capitalized

Risk Category I

Adequately Capitalized

Risk Category II

Risk Category III.

Under Capitalized

Risk Category III

Risk Category IV

To further differentiate risk within Risk Category I (which includes most small banks), the FDIC uses the

financial ratios method,

which combines supervisory CAMELS component ratings with current financial ratios to determine a small Risk Category I bank's initial assessment rate.

New small banks in Risk Category I, however, are charged the highest initial assessment rate in effect for that risk category. Subject to exceptions, a new bank is one that has been federally insured for less than five years as of the last day of any quarter for which it is being assessed. 12 CFR 327.8(j).

Within Risk Category I, those institutions that pose the least risk are charged a minimum initial assessment rate and those that pose the greatest risk are charged an initial assessment rate that is four basis points higher than the minimum. All other banks within Risk Category I are charged a rate that varies between these rates. In contrast, all banks in Risk Category II are charged the same initial assessment rate, which is higher than the maximum initial rate for Risk Category I. A single, higher, initial assessment rate applies to each bank in Risk Category III and another, higher, rate to each bank in Risk Category IV.

In 2011, the Board revised and approved regular assessment rate schedules. See 76 FR 10672 (Feb. 25, 2011); 12 CFR 327.10.

The financial ratios method determines the assessment rates in Risk Category I using a combination of weighted CAMELS component ratings and the following financial ratios:

• Tier 1 Leverage Ratio;

• Net Income before Taxes/Risk-Weighted Assets;

• Nonperforming Assets/Gross Assets;

• Net Loan Charge-Offs/Gross Assets;

• Loans Past Due 30-89 days/Gross Assets;

• Adjusted Brokered Deposit Ratio; and

• Weighted Average CAMELS Composite Rating.

The weights applied to CAMELS components are as follows: 25 percent each for Capital and Management; 20 percent for Asset quality; and 10 percent each for Earnings, Liquidity, and Sensitivity to market risk. These weights reflect the view of the FDIC regarding the relative importance of each of the CAMELS components for differentiating risk among institutions for deposit insurance purposes. The FDIC and other bank supervisors do not use such a system to determine CAMELS composite ratings.

To determine a Risk Category I bank's initial assessment rate, the weighted CAMELS components and financial ratios are multiplied by statistically derived pricing multipliers, the products are summed, and the sum is added to a uniform amount that applies to all Risk Category I banks. If, however, the rate is below the minimum initial assessment rate for Risk Category I, the bank will pay the minimum initial assessment rate; if the rate derived is above the maximum initial assessment rate for Risk Category I, then the bank will pay the maximum initial rate for the risk category.

The financial ratios used to determine rates come from a statistical model that predicts the probability that a Risk Category I institution will be downgraded from a composite CAMELS rating of 1 or 2 to a rating of 3 or worse within one year. The probability of a CAMELS downgrade is intended as a proxy for the bank's probability of failure. When the model was developed in 2006, the FDIC decided not to attempt to determine a bank's probability of failure because of the lack of bank failures in the years between the end of the bank and thrift crisis in the early 1990s and 2006.

See 71 FR 41910, 41913 (July 24, 2006).

The financial ratios method does not apply to new small banks or to insured branches of foreign banks (insured branches).

The manner in which assessment rates for these institutions is determined is described further below.

Insured branches of foreign banks are deemed small banks for purposes of the deposit insurance assessment system.

Assessment Rates Under Current Rules

The Dodd-Frank Wall Street Reform and Consumer Protection Act (the Dodd-Frank Act), enacted in July 2010, revised the statutory authorities governing the FDIC's management of the DIF. The Dodd-Frank Act granted the FDIC authority to manage the fund in a manner that would help maintain a positive fund balance during a banking crisis and promote moderate, steady assessment rates throughout economic credit cycles.

12 U.S.C. 1817(e) (granting the Board the discretion to suspend or limit dividends).

Among other things, the Dodd-Frank Act: (1) raised the minimum designated reserve ratio (DRR), which the FDIC must set each year, to 1.35 percent (from the former minimum of 1.15 percent) and removed the upper limit on the DRR (which was formerly capped at 1.5 percent);

(2) required that the fund reserve ratio reach 1.35 percent by September 30, 2020 (rather than 1.15 percent by the end of 2016, as formerly required);

and (3) required that, in setting assessments, the FDIC “offset the effect of [requiring that the reserve ratio reach 1.35 percent by September 30, 2020 rather than 1.15 percent by the end of 2016] on insured depository institutions with total consolidated assets of less than $10,000,000,000.”

12 U.S.C. 1817(b)(3)(B).

Public Law 111-203, 334(d), 124 Stat. 1376, 1539 (12 U.S.C. 1817(note)).

Public Law 111-203, 334(e), 124 Stat. 1376, 1539 (12 U.S.C. 1817(note)). The Dodd-Frank Act also: (1) eliminated the requirement that the FDIC provide dividends from the fund when the reserve ratio is between 1.35 percent and 1.5 percent, 12 U.S.C. 1817(e), and (2) continued the FDIC's authority to declare dividends when the reserve ratio at the end of a calendar year is at least 1.5 percent, but granted the FDIC sole discretion in determining whether to suspend or limit the declaration of payment or dividends, 12 U.S.C. 1817(e)(2)(A)-(B).

In 2011, the FDIC adopted a schedule of assessment rates designed to ensure that the reserve ratio reaches 1.15 percent by September 30, 2020.

In the near future, the FDIC plans to propose a rule to implement the Dodd-Frank Act requirement that the cost of raising the reserve ratio from 1.15 percent to 1.35

percent be paid by banks with $10 billion or more in assets.

See 76 FR 10672.

The current initial assessment rates for small and large banks are set forth in Table 2 below.

Table 2—Initial Base Assessment Rates

[In basis points per annum]

Risk category

Minimum

Maximum

III

Large &

highly

complex

institutions**

Annual Rates (in basis points)

5-35

* Initial base rates that are not the minimum or maximum will vary between these rates.

** See § 327.8(f) and § 327.8(g) for the definition of large and highly complex institutions.

An institution's total assessment rate may vary from the initial assessment rate as the result of possible adjustments.

After applying all possible adjustments, minimum and maximum total assessment rates for each risk category are set forth in Table 3 below.

A bank's total base assessment rate can vary from its initial base assessment rate as the result of three possible adjustments. Two of these adjustments—the unsecured debt adjustment and the depository institution debt adjustment (DIDA)—apply to all banks (except that the unsecured debt adjustment does not apply to new banks or insured branches). The unsecured debt adjustment lowers a bank's assessment rate based on the bank's ratio of long-term unsecured debt to the bank's assessment base. The DIDA increases a bank's assessment rate when it holds long-term, unsecured debt issued by another insured depository institution. The third possible adjustment—the brokered deposit adjustment—applies only to small banks in Risk Category II, III and IV (and to large and highly complex institutions that are not well capitalized or that are not CAMELS composite 1 or 2-rated). It does not apply to insured branches. The brokered deposit adjustment increases a bank's assessment when it holds significant amounts of brokered deposits. 12 CFR 327.9 (d).

Table 3—Total Base Assessment Rates*

[In basis points per annum]

Risk

category

Risk

category

Risk

category

III

Risk

category

Large &

highly

complex

institutions **

Initial Assessment Rate

5-9

5-35

Unsecured Debt Adjustment ***

−4.5 to 0

−5 to 0

Brokered Deposit Adjustment

N/A

0 to 10

Total Assessment Rate

2.5 to 9

9 to 24

18 to 33

30 to 45

2.5 to 45

* Total base assessment rates do not include the DIDA.

** See § 327.8(f) and (g) for the definition of large and highly complex institutions.

*** The unsecured debt adjustment cannot exceed the lesser of 5 basis points or 50 percent of an insured depository institution's initial base assessment rate. The unsecured debt adjustment does not apply to new banks or insured branches.

Before adopting the current assessment rate schedules, the FDIC undertook a historical analysis to determine how high the reserve ratio would have to have been to have maintained both a positive balance and stable assessment rates from 1950 through 2010.

The analysis shows that the fund reserve ratio would have needed to be approximately 2 percent or more before the onset of the 1980s and 2008 crises to maintain both a positive fund balance and stable assessment rates, assuming, in lieu of dividends, that the long-term industry average nominal assessment rate would have been reduced by 25 percent when the reserve ratio reached 2 percent, and by 50 percent when the reserve ratio reached 2.5 percent.

The historical analysis and long-term fund management plan are described at 76 FR at 10675 and 75 FR 66272, 66272-281 (Oct. 27, 2010).

In 2011, consistent with the FDIC's historical analysis and the FDIC's long-term fund management plan adopted as a result of the historical analysis, the Board adopted lower, moderate assessment rates that will go into effect when the DIF reserve ratio reaches 1.15 percent.

Pursuant to the FDIC's authority to set assessments, the initial base and total base assessment rates set forth in Table 4 below will take effect beginning the assessment period after the fund reserve ratio first meets or exceeds 1.15 percent, without the necessity of further action by the Board. The rates will remain in effect unless and until the reserve ratio meets or exceeds 2 percent.

See 76 FR at 10717-720.

For new banks, however, the rates will remain in effect even if the reserve ratio equals or exceeds 2 percent (or 2.5 percent).

The reserve ratio for the immediately prior assessment period must also be less than 2 percent.

Table 4—Initial and Total Base Assessment Rates *

[In basis points per annum]

[Once the reserve ratio reaches 1.15 percent]

Risk

category

Risk

category

Risk

category

III

Risk

category

Large &

highly

complex

institutions **

Initial Base Assessment Rate

3-7

3-30

Unsecured Debt Adjustment ***

−3.5 to 0

−5 to 0

Brokered Deposit Adjustment

N/A

0 to 10

Total Base Assessment Rate

1.5 to 7

7 to 22

14 to 29

25 to 40

1.5 to 40

* Total base assessment rates do not include the DIDA.

** See § 327.8(f) and (g) for the definition of large and highly complex institutions.

** The unsecured debt adjustment cannot exceed the lesser of 5 basis points or 50 percent of an insured depository institution's initial base assessment rate; thus, for example, an insured depository institution with an initial base assessment rate of 3 basis points will have a maximum unsecured debt adjustment of 1.5 basis points and cannot have a total base assessment rate lower than 1.5 basis points. The unsecured debt adjustment does not apply to new banks or insured branches.

In lieu of dividends, and pursuant to the FDIC's authority to set assessments and consistent with the FDIC's long-term fund management plan, the initial base and total base assessment rates set forth in Table 5 below will come into effect without further action by the Board when the fund reserve ratio at the end of the prior assessment period meets or exceeds 2 percent, but is less than 2.5 percent.

New small banks will remain subject to the assessment schedule in Table 5 when the reserve ratio reaches 2 percent and 2.5 percent.

Table 5—Initial and Total Base Assessment Rates*

[In basis points per annum]

[If the reserve ratio for the prior assessment period is equal to or greater than 2 percent and less than 2.5 percent]

Risk

category

Risk

category

Risk

category

III

Risk

category

Large &

highly

complex

institutions **

Initial Base Assessment Rate

2-6

2-28

Unsecured Debt Adjustment ***

−3 to 0

−5 to 0

Brokered Deposit Adjustment

N/A

0 to 10

Total Base Assessment Rate

1 to 6

5 to 20

12 to 27

23 to 38

1 to 38

* Total base assessment rates do not include the DIDA.

** See § 327.8(f) and (g) for the definition of large and highly complex institutions.

*** The unsecured debt adjustment cannot exceed the lesser of 5 basis points or 50 percent of an insured depository institution's initial base assessment rate; thus, for example, an insured depository institution with an initial base assessment rate of 2 basis points will have a maximum unsecured debt adjustment of 1 basis point and cannot have a total base assessment rate lower than 1 basis point. The unsecured debt adjustment does not apply to insured branches.

The initial base and total base assessment rates set forth in Table 6 below will come into effect, again, without further action by the Board when the fund reserve ratio at the end of the prior assessment period meets or exceeds 2.5 percent.

Table 6—Initial and Total Base Assessment Rates*

[In basis points per annum]

[If the reserve ratio for the prior assessment period is equal to or greater than 2.5 percent]

Risk

category

Risk

category

Risk

category

III

Risk

category

Large &

highly

complex

institutions **

Initial Base Assessment Rate

1—5

1-25

Unsecured Debt Adjustment ***

−2.5 to 0

−4.5 to 0

−5 to 0

Brokered Deposit Adjustment

N/A

0 to 10

Total Base Assessment Rate

0.5 to 5

4.5 to 19

10 to 25

20 to 35

0.5 to 35

* Total base assessment rates do not include the DIDA.

** See § 327.8(f) and (g) for the definition of large and highly complex institutions.

*** The unsecured debt adjustment cannot exceed the lesser of 5 basis points or 50 percent of an insured depository institution's initial base assessment rate; thus, for example, an insured depository institution with an initial base assessment rate of 1 basis point will have a maximum unsecured debt adjustment of 0.5 basis points and cannot have a total base assessment rate lower than 0.5 basis points. The unsecured debt adjustment does not apply to insured branches.

With respect to each of the four assessment rate schedules (Tables 3, 4, 5 and 6), the Board has the authority to adopt rates without further notice and comment rulemaking that are higher or lower than the total assessment rates (also known as the total base assessment rates) shown in the tables, provided that: (1) The Board cannot increase or decrease rates from one quarter to the next by more than two basis points; and (2) cumulative increases and decreases cannot be more than two basis points higher or lower than the total base assessment rates.

See 12 CFR 327.10(f); 76 FR at 10684.

III. Justification for Proposal

While the current deposit insurance assessment system effectively reflects the risk posed by small banks, it can be improved by incorporating newer data from the recent financial crisis and revising the methodology to directly estimate the probability of failure three years ahead. These improvements will allow the FDIC to more effectively price risk. The proposed improvements to the small bank risk-based assessment system will further the goals of reducing cross-subsidization of high-risk institutions by low risk institutions and help ensure that banks that take on greater risks will pay more for deposit insurance.

IV. Description of the Proposed Rule

Summary of the Proposed Rule

The FDIC proposes to improve the assessment system applicable to established small banks

(that is, small banks other than new small banks and insured branches of foreign banks) by: (1) Revising the financial ratios method so that it is based on a statistical model estimating the probability of failure over three years; (2) updating the financial measures used in the financial ratios method consistent with the statistical model; and (3) eliminating risk categories for all established small banks and using the financial ratios method to determine assessment rates for all such banks. CAMELS composite ratings, however, would be used to place a maximum on the assessment rates that CAMELS composite 1- and 2-rated banks could be charged and minimums on the assessment rates that CAMELS composite 3-, 4- and 5-rated banks could be charged.

Subject to exceptions, an established insured depository institution is one that has been federally insured for at least five years as of the last day of any quarter for which it is being assessed. 12 CFR 327.8(k).

Over 500 banks have failed since the end of 2007. These failures, together with the hundreds of failures during the banking crisis of the late 1980s and early 1990s, have generated a robust set of data on bank failures. The FDIC need no longer rely on a model that estimates a proxy for failure—the probability that a bank with a CAMELS composite rating of 1 or 2 will be downgraded to a CAMELS composite rating of 3, 4, or 5 within 12 months; rather, the FDIC can base small bank deposit insurance assessments on a statistical model that estimates a bank's probability of failure directly.

In addition to estimating probability of failure directly, the proposal improves the small bank deposit insurance assessment system in other ways. First, it allows the assessment system to better capture risk when the risk is assumed, rather than when the risk has already resulted in losses. The statistical model on which the proposed deposit insurance assessment system for small banks is based estimates the probability of failure within three years, balancing the need to capture risk when it is assumed with the need for accurate failure predictions. (The longer the prediction period, the less accurate a model's predictions will tend to be; so, for example, the FDIC cannot create a model that predicts failure ten years in the future with sufficient accuracy.) The risk-based assessment system established in 2011 for large banks is also designed to capture performance over a period longer than one year. The FDIC would update the financial measures used in the financial ratios method to be consistent with the proposed statistical model. All of the proposed measures were statistically significant in predicting a bank's probability of failure within a three-year period.

Second, because the model allows the FDIC to estimate the probability of failure directly, it allows the FDIC to apply the model to all established small banks, not just those in Risk Category I. In part because CAMELS ratings can incorporate information that the model cannot, the FDIC proposes to apply minimum or maximum initial base assessment rates that will depend on a bank's CAMELS composite rating. Thus, as it has with large banks, the FDIC would eliminate risk categories for small banks (other than new small banks and insured branches of foreign banks).

Third, because the model predicts the probability of failure three years ahead using data on hundreds of failures (including failures during the recent crisis), it better reflects banks' actual risks and provides incentives to banks to monitor and reduce risks that increase potential losses to the DIF. Because it measures risk more accurately, the model reduces the subsidization of riskier banks by less risky banks.

The FDIC intends to preserve the lower range of initial base assessment rates previously adopted by the Board. The FDIC is proposing that the new assessment system go into operation the quarter after the reserve ratio reaches 1.15 percent. At that time, under the initial base assessment rate schedules adopted by the Board in 2011, initial based assessment rates will fall automatically from the current 5 basis point to 35 basis point range to a 3 basis point to 30 basis point range, as reflected in Table 4.

The FDIC adopted this schedule of assessment rates pursuant to its long-term fund management plan as the FDIC's best estimate of the assessment rates that would have been needed from 1950 to 2010 to maintain a positive fund balance during the past two banking crises.

As under current rules, the brokered deposit adjustment would continue to apply only to established small banks that are less than well capitalized or that have a CAMELS composite rating of 3, 4 or 5.

The FDIC proposes to convert the statistical model to assessment rates within this 3 basis point to 30 basis point assessment range in a revenue neutral way; that is, in a manner that does not change the aggregate assessment revenue collected from established small banks. Specifically, the conversion would be done to ensure that aggregate assessments for an assessment period shortly before adoption of a final rule would have been approximately the same under the final rule as they would have been under the assessment rate schedule set forth in Table 4 (the rates that, under current rules, will automatically go into effect when the reserve ratio reaches 1.15 percent).

To avoid unnecessary burden, the FDIC is proposing a revised small bank assessment system that does not require small banks to report any new data in their Reports of Condition and Income (Call Reports).

Implementation of the Proposed Rule

The FDIC proposes that a final rule go into effect the quarter after a final rule is adopted; by their terms, however, the proposed revisions would not become operative until the quarter after the DIF reserve ratio reaches 1.15 percent.

Detailed Description of the Proposed Rule

Risk Differentiation

As mentioned above, the FDIC is proposing to update the financial measures used in the financial ratios method consistent with the statistical model, eliminate risk categories for all established small banks, and use the financial ratios method to determine assessment rates for all such banks. CAMELS composite ratings would be used to place a maximum on the assessment rates that CAMELS composite 1- and 2-rated banks could be charged, and minimums on the assessment rates that CAMELS composite 3-, 4- and 5-rated banks could be charged.

The financial ratios method as revised would use the measures described in the right-hand column of Table 7 below. For comparison's sake, the measures currently used in the financial ratios method are set out on the left-hand column of the table.

Table 7—Comparison of Current and Proposed Measures in the Financial Ratios Method

Current risk category I financial ratios method

Proposed financial ratios method

• Weighted Average CAMELS Component Rating

• Weighted Average CAMELS Component Rating.

• Tier 1 Leverage Ratio

• Tier 1 Leverage Ratio.

• Net Income before Taxes/Risk-Weighted Assets

• Net Income before Taxes/Total Assets.

• Nonperforming Assets/Gross Assets

• Nonperforming Loans and Leases/Gross Assets.

• Other Real Estate Owned/Gross Assets.

• Adjusted Brokered Deposit Ratio

• Core Deposits/Total Assets.

• One Year Asset Growth.

• Net Loan Charge-Offs/Gross Assets

• Loans Past Due 30-89 Days/Gross Assets

• Loan Mix Index.

All of the proposed measures are derived from a statistical analysis that estimates a bank's probability of failure within three years. Each of the measures was statistically significant in predicting a bank's probability of failure over that period. The statistical analysis used bank financial data and CAMELS ratings from 1985 through 2011, failure data from 1986 through 2014, and loan charge-off data from 2001 through 2014.

Appendix 1 to the Supplementary Information section of this notice and the proposed Appendix E describe the statistical analysis and the derivation of these proposed measures in detail.

For certain lagged variables, such as one-year asset growth rates, the statistical analysis also used bank financial data from 1984.

Two of the proposed measures—the weighted average CAMELS component rating and the tier 1 leverage ratio—are identical to the measures currently used in the financial ratios method.

The proposed net income before taxes/total assets measure is also identical to the current measure, except that the denominator is total assets rather than risk-weighted assets. The current measure nonperforming assets/gross assets includes other real estate owned. In the proposal, other real estate owned/gross assets is a separate measure from nonperforming loans and leases/gross assets.

Current rules provide that, if a Risk Category I small bank's CAMELS component ratings change during a quarter in a way that changes the bank's initial base assessment rate, the initial base assessment rate for the period before the change shall be determined under the financial ratios method using the CAMELS component ratings in effect before the change. Beginning on the date of the CAMELS component ratings change, the initial base assessment rate for the remainder of the quarter is determined using the CAMELS component ratings in effect after the change. 12 CFR 327.9(a)(4)(iv)(B). Under the proposal, this rule would remain essentially unchanged, but would apply to all established small banks rather than just banks within Risk Category I.

The remaining three proposed measures—core deposits/total assets, one-year asset growth, and the loan mix index—are new.

Two measures in the current financial ratios method—net loan charge-offs/gross assets and loans past due 30-89 days/gross assets—are not used in the statistical analysis and are not among the proposed measures.

Under the proposal, the core deposits/total assets and the one-year asset growth measures would replace the adjusted brokered deposit ratio currently used in the financial ratios method. The adjusted brokered deposit ratio increases a Risk Category I small bank's assessment rate only if the bank has both large amounts of brokered deposits and high asset growth.

Few banks have both, so the ratio affects few banks.

One of the proposed replacement measures—core deposits/total assets—will tend to lower assessment rates for most small banks. The other proposed replacement measure—one-year asset growth—will tend to raise assessment rates for small banks that grow significantly over a year (other than through merger or by acquiring failed banks).

The adjusted brokered deposit ratio can affect assessment rates only if a bank's brokered deposits (excluding reciprocal deposits) exceed 10 percent of its non-reciprocal brokered deposits and its assets have grown more than 40 percent in the previous 4 years. 12 CFR 327 Appendix A to Subpart A.

As of December 31, 2014, the adjusted brokered deposit ratio affected the assessment rate of 81 banks.

The loan mix index is a measure of the extent to which a bank's total assets include higher-risk categories of loans. Each category of loan in a bank's loan portfolio is divided by the bank's total assets to determine the percentage of the bank's assets represented by that category of loan. Each percentage is then multiplied by that category of loan's historical weighted average industry-wide charge-off rate. The products are then summed to determine the loan mix index value for that bank.

The loan categories in the loan mix index were selected based on the availability of category-specific charge-off rates over a sufficiently lengthy period (2001 through 2014) to be representative. The loan categories exclude credit card loans.

For each loan category, the weighted average charge-off rate weights each industry-wide charge-off rate for each year by the number of bank failures in that year. Thus, charge-off rates from 2009 through 2014, during the recent banking crisis, have a much greater influence on the weighted average charge-off rate than charge-off rates from the years before the crisis, when few failures occurred. The weighted averages assure that types of loans that have high

charge-off rates during downturns have an appropriate influence on assessment rates.

Credit card loans were excluded from the loan mix index because they produced anomalously high assessment rates for banks with significant credit card loans. Credit card loans have very high charge-off rates, which the loan mix index can capture, but they also tend to have very high interest rates to compensate. In addition, few small banks have significant concentrations of credit card loans. Consequently, credit card loans are omitted from the index.

Table 8 below

illustrates how the loan mix index is calculated for a hypothetical bank.

As discussed above, the loan mix index uses loan charge-off data from 2001 through 2014. As discussed in greater detail below, if financial, failure and charge-off data from later years is available at the time the FDIC adopts a final rule pursuant to this proposal, the FDIC may update the statistical model, including the loan mix index, using the methodology described in Appendix E.

The table shows industry-wide weighted charge-off percentage rates, the loan category as a percentage of total assets and the products to two decimal places. In fact, the FDIC proposes to use seven decimal places for industry-wide weighted charge-off percentage rates, and as many decimal places as permitted by the FDIC's computer systems for the loan category as a percentage of total assets and the products. The total (the loan mix index itself) would use three decimal places.

Table 8—Loan Mix Index for a Hypothetical Bank

Weighted charge-off rate percent

Loan category as a percent of hypothetical bank's total assets

Product of two columns to the left

Construction & Development

4.50

1.40

6.29

Commercial & Industrial

1.60

24.24

38.75

Leases

1.50

0.64

0.96

Other Consumer

1.46

14.93

21.74

Loans to Foreign Government

1.34

0.24

0.32

Real Estate Loans Residual

1.02

0.11

Multifamily Residential

0.88

2.42

2.14

Nonfarm Nonresidential

0.73

13.71

9.99

1-4 Family Residential

0.70

2.27

1.58

Loans to Depository banks

0.58

1.15

0.66

Agricultural Real Estate

0.24

3.43

0.82

Agriculture

0.24

5.91

1.44

SUM (Loan Mix Index)

70.45

84.79

The weighted charge-off rates in the table are the same for all small banks. The remaining two columns vary from bank to bank, depending on the bank's loan portfolio. For each loan type, the value in the rightmost column is calculated by multiplying the weighted charge-off rate by the bank's loans of that type as a percent of its total assets. In this illustration, the sum of the right-hand column (84.79) is the loan mix index for this bank.

As in the current methodology for Risk Category I small banks, under the proposal the weighted CAMELS components and financial ratios would be multiplied by statistically derived pricing multipliers, the products would be summed, and the sum would be added to a uniform amount that would be: (a) Derived from the statistical analysis, (b) adjusted for assessment rates set by the FDIC, and (c) applied to all established small banks. The total would equal the bank's initial assessment rate. If, however, the resulting rate were below the minimum initial assessment rate for small banks, the bank's initial assessment rate would be the minimum initial assessment rate; if the rate were above the maximum, then the bank's initial assessment rate would be the maximum initial rate for small banks. In addition, if the resulting rate for a small bank were below the minimum or above the maximum initial assessment rate applicable to banks with the bank's CAMELS composite rating, the bank's initial assessment rate would be the respective minimum or maximum assessment rate for a small bank with its CAMELS composite rating. This approach would allow rates to vary incrementally across a wide range of rates for all small banks (other than new small banks and insured branches). The conversion of the statistical model to pricing multipliers and uniform amount are discussed further below and in detail in the proposed Appendix E. Appendix E also discusses the derivation of the pricing multipliers and the uniform amount.

Adjustments to Initial Base Assessment Rates

As under current rules: (1) The DIDA would continue to apply to all banks; (2) the unsecured debt adjustment would continue to apply to all banks except new banks and insured branches; and (3) the brokered deposit adjustment would continue to apply to all small banks except those that are well capitalized and have a CAMELS composite rating of 1 or 2.

As under current rules, if, during a quarter, a bank's supervisory rating changes from a CAMELS composite 1 or 2 rating to a CAMELS composite 3, 4 or 5 rating or vice versa, the bank would be subject to the brokered deposit adjustment for the portion of the quarter that it did not have a CAMELS composite 1 or 2 rating.

As under current rules, however, no adjustments would apply to bridge banks or conservatorships. These banks would continue to be charged the minimum assessment rate applicable to small banks. As under current rules, the brokered deposit adjustment would not apply to insured branches.

If the bank were less than well capitalized, it would be subject to the brokered deposit adjustment for the whole quarter.

Proposed Assessment Rates

As described above and as set out in the rate schedule in Table 9 below, for established small banks, the FDIC proposes to eliminate risk categories, but maintain the range of initial assessment rates (3 basis points to 30 basis points) that the Board has previously determined will go into effect starting the quarter after the reserve ratio reaches 1.15 percent and include a maximum assessment rate that would apply to CAMELS composite 1- and 2-rated banks and the minimum assessment rates that would apply to CAMELS composite 3-rated banks and CAMELS composite 4- and 5-rated banks.

Unless revised by the Board, these rates would remain in effect so long as the reserve ratio is less than 2 percent.

See 12 CFR 327.10(b); 76 FR at 10718.

Table 9—Initial and Total Base Assessment Rates *

[In basis points per annum]

[Once the reserve ratio reaches 1.15 percent]

Established small banks

CAMELS Composite

1 or 2

4.or 5

Large & highly complex

institutions **

Initial Base Assessment Rate

3 to 16

6 to 30

16 to 30

3 to 30

Unsecured Debt Adjustment ***

−5 to 0

Brokered Deposit Adjustment

0 to10 ****

0 to10

0 to 10

Total Base Assessment Rate

1.5 to 26

3 to 40

11 to 40

1.5 to 40

* Total base assessment rates in the table do not include the DIDA.

** See § 327.8(f) and (g) for the definition of large and highly complex institutions.

*** The unsecured debt adjustment cannot exceed the lesser of 5 basis points or 50 percent of an insured depository institution's initial base assessment rate; thus, for example, an insured depository institution with an initial base assessment rate of 3 basis points will have a maximum unsecured debt adjustment of 1.5 basis points and cannot have a total base assessment rate lower than 1.5 basis points.

**** The brokered deposit adjustment applies to established small banks with CAMELS composite ratings of 1 or 2 only if they are less than well capitalized.

discussed above, the FDIC adopted the range of assessment rates in this rate schedule pursuant to its long-term fund management plan as the FDIC's best estimate of the assessment rates that would have been needed from 1950 to 2010 to maintain a positive fund balance during the past two banking crises. This assessment rate schedule remains the FDIC's best estimate of the long-term rates needed. Consequently, and as discussed in greater detail further below and in detail in Appendix E, the FDIC proposes to convert its statistical model to assessment rates within this 3 basis point to 30 basis point assessment range in a revenue neutral way.

The reserve ratio for the immediately prior assessment period must also be less than 2 percent.

The FDIC proposes to maintain the range of initial assessment rates, set out in the rate schedule in Table 10 below, that the Board has previously determined will go into effect starting the quarter after the reserve ratio reaches or exceeds 2 percent and is less than 2.5 percent. Unless revised by the Board, these rates would remain in effect so long as the reserve ratio is in this range. Table 10 also includes the maximum assessment rates that will apply to CAMELS composite 1- and 2-rated banks and the minimum assessment rates that will apply to CAMELS composite 3-rated banks and CAMELS composite 4- and 5-rated banks.

Table 10—Initial and Total Base Assessment Rates *

[In basis points per annum]

[If the reserve ratio for the prior assessment period is equal to or greater than 2 percent and less than 2.5 percent]

Established small banks

CAMELS Composite

1 or 2

4 or 5

Large & highly complex

institutions **

Initial Base Assessment Rate

2 to 14

5 to 28

14 to 28

2 to 28

Unsecured Debt Adjustment ***

−5 to 0

Brokered Deposit Adjustment

0 to 10 ****

0 to 10

Total Base Assessment Rate

1 to 24

2.5 to 38

9 to 38

1 to 38

* Total base assessment rates in the table do not include the DIDA.

** See § 327.8(f) and (g) for the definition of large and highly complex institutions.

**** The brokered deposit adjustment applies to established small banks with CAMELS composite ratings of 1 or 2 only if they are less than well capitalized.

The FDIC proposes to maintain the range of initial assessment rates, set out in the rate schedule in Table 11 below, that the Board has previously determined will go into effect, again without further action by the Board, when the fund reserve ratio at the end of the prior assessment period meets or exceeds 2.5 percent. Unless changed by the Board, these rates would remain in effect so long as the reserve ratio is at or above this level. Table 11 also includes the maximum assessment rates that will apply to CAMELS composite 1- and 2-rated banks and the minimum assessment rates that will apply to CAMELS composite 3-rated banks and CAMELS composite 4- and 5-rated banks.

Table 11—Initial and Total Base Assessment Rates *

[In basis points per annum]

[If the reserve ratio for the prior assessment period is equal to or greater than 2.5 percent]

Established small banks

CAMELS Composite

1 or 2

4 or 5

Large & highly complex

institutions **

Initial Base Assessment Rate

1 to 13

4 to 25

13 to 25

1 to 25

Unsecured Debt Adjustment ***

−5 to 0

Brokered Deposit Adjustment

0 to 10 ****

0 to 10

Total Base Assessment Rate

0.5 to 23

2 to 35

8 to 35

0.5 to 35

* Total base assessment rates in the table do not include the DIDA.

** See § 327.8(f) and (g) for the definition of large and highly complex institutions.

*** The unsecured debt adjustment cannot exceed the lesser of 5 basis points or 50 percent of an insured depository institution's initial base assessment rate; thus, for example, an insured depository institution with an initial base assessment rate of 1 basis point will have a maximum unsecured debt adjustment of 0.5 basis points and cannot have a total base assessment rate lower than 0.5 basis points.

**** The brokered deposit adjustment applies to established small banks with CAMELS composite ratings of 1 or 2 only if they are less than well capitalized.

With respect to each of the three assessment rate schedules (Tables 9, 10 and 11), the FDIC proposes that the Board would retain its authority to uniformly adjust assessment rates up or down from the total base assessment rate schedule without further rulemaking, as long as adjustment does not exceed 2 basis points. Also, with respect to each of the three schedules, the FDIC proposes that, if a bank's CAMELS composite or component ratings change during a quarter in a way that changes the institution's initial base assessment rate, then its assessment rate would be determined separately for each portion of the quarter in which it had different CAMELS composite or component ratings.

Conversion of Statistical Model to Pricing Multipliers and Uniform Amount

As discussed above, the FDIC proposes to convert its statistical model to assessment rates set out in Table 9 in a revenue neutral manner.

Specifically, and as described in detail in Appendix E, the FDIC proposes to convert the statistical model to assessment rates to ensure that aggregate assessments for an assessment period shortly before adoption of a final rule would have been approximately the same under the final rule as they would have been under the assessment rate schedule set forth in Table 4 (the rates that, under current rules, will automatically go into effect when the reserve ratio reaches 1.15 percent).

The FDIC proposes to convert a linear version of its model, which was estimated in a non-linear manner. (See Appendix E.) The conversion using a linear version of the model preserves the same rank ordering as the non-linear model, but using the linear version of the model allows initial assessment rates to be expressed as a linear function of the model variables. The FDIC also used a linear version of its original non-linear downgrade probability statistical model when it instituted variable rates within Risk Category 1 (effective January 1, 2007).

To illustrate the conversion, Table 12 below sets out the pricing multipliers and uniform amounts that would have resulted if the FDIC had converted the statistical model to the assessment rate schedule set out in Table 9 (with a range of assessment rates from 3 basis points to 30 basis points) so that, for the fourth quarter of 2014, aggregate assessments for all established small banks under the proposal would have equaled, as closely as reasonably possible, aggregate assessments for all established small banks had the assessment rate schedule in Table 4 been in effect for that assessment period.

Partly because the actual conversion will be based upon a later quarter (and partly for the reasons discussed directly below), the pricing multipliers and the uniform amount shown in Table 12 are likely to differ somewhat from those in the final rule.

Initial assessment rates under the rate schedule actually in effect for the fourth quarter of 2014 ranged from 5 basis points to 35 basis points, since the DIF reserve ratio was under 1.15 percent.

Table 12—Pricing Multipliers and the Uniform Amount Under a Hypothetical Conversion of the Statistical Model to Assessment Rates Based on the Fourth Quarter of 2014

Model measures

Pricing

multiplier

Weighted Average CAMELS Component Rating

1.731

Tier 1 Leverage Ratio

−1.337

Net Income Before Taxes/Total Assets

−0.652

Nonperforming Loans and Leases/Gross Assets

0.924

Other Real Estate Owned/Gross Assets

0.620

Core Deposits/Total Assets

−0.139

One Year Asset Growth

0.043

Loan Mix Index

0.066

Uniform Amount

19.376

Updating the Statistical Model, Pricing Multipliers and Uniform Amount

The statistical analysis used bank financial data and CAMELS ratings from 1985 through 2011, failure data from 1986 through 2014 and loan charge-off data from 2001 through 2014. The FDIC proposes to retain the flexibility to update the statistical model from time to time using financial, failure and charge-off data from later years and publish a new loan mix index, uniform amount and pricing multipliers based on the updated model without further notice-and-comment rulemaking. Any update to the model would be done pursuant to the methodology described in Appendix E. No new financial ratios or other measures would be introduced into the model without notice-and-comment rulemaking. Because the analysis would continue to use earlier years' data as well, changes in estimations of failure probability should usually be relatively small. Similarly, if financial, failure and charge-off data from later years is available at the time the FDIC adopts a final rule pursuant to this proposal, the FDIC may update the statistical model,

including the loan mix index, using the methodology described in Appendix E.

Insured Branches of Foreign Banks and New Small Banks

The FDIC proposes to make no changes to the rules governing the assessment rate schedules applicable to insured branches or to the assessment rate schedule applicable to new small banks. The FDIC also proposes to make no changes to the way in which assessment rates for insured branches and new small banks are determined.

Insured Branches

The current risk-based deposit insurance assessment system for small banks assigns insured branches an assessment risk classification that is based on the FDIC's consideration of supervisory evaluations provided by the institution's primary federal regulator.

Within Risk Category I, each insured branch's assessment rate is based on these supervisory evaluations.

Insured branches not in Risk Category I are charged the initial base assessment rate for the risk category to which they are assigned.

Once the DIF reserve ratio reaches 1.15 percent, 2 percent, and 2.5 percent, assessment rate schedules previously adopted by the Board will go into effect and remain in place for insured branches.

These supervisory evaluations result in the assignment of supervisory ratings referred to as ROCA ratings. ROCA stands for Risk Management, Operational Controls, Compliance, and Asset Quality. Like CAMELS components, ROCA component ratings range from a “1” (best rating) to a “5” rating (worst rating). A Risk Category I insured branch generally has a ROCA composite rating of 1 or 2.

Specifically, the assessment rate depends on the insured branch's weighted average ROCA component ratings. The weights applied to individual ROCA component ratings are 35 percent, 25 percent, 25 percent, and 15 percent, respectively.

No insured branch in any risk category is subject to the unsecured debt adjustment or brokered deposit adjustment. Insured branches are subject to the DIDA.

The FDIC does not propose changing the way assessment rates applicable to insured branches are determined.

Insured branches do not report the information that the FDIC would need to apply the financial ratios method to them.

Moreover, because insured branches operate as extensions of a foreign bank's global banking operations, they pose unique risks, which the financial ratios method may not be able to capture. An insured branch operates without capital of its own (capital is held by the foreign bank), its business strategies are typically directed by the foreign bank, it relies extensively on the foreign bank for liquidity and funding, and it often has considerable country and transfer risk exposures not typically found in other insured institutions of similar size. Insured branches also present potentially challenging concerns in the event of failure.

As of March 31, 2015, there were only 9 insured branches that file regulatory financial submissions (FFIEC Form 002). (One of these branches, however, files for itself and another branch of the same foreign bank that does not file separately.)

For example, insured branches of foreign banks do not report earnings and report only limited balance sheet information in FFIEC Form 002.

New Small Banks

New small banks are currently assigned to risk categories in the same manner as all other small banks. All new small banks in Risk Category I, however, are charged the maximum rate applicable to Risk Category I. New small banks not in Risk Category I are charged the initial base assessment rate for the risk category to which they are assigned.

Once the DIF reserve ratio reaches 1.15 percent, new small banks will be charged initial rates under the previously adopted rate schedule that automatically goes into effect then. This rate schedule will remain in place even if the reserve ratio equals or exceeds 2 percent or 2.5 percent.

After applying all possible adjustments, minimum and maximum total assessment rates for new small banks in each risk category are set forth in Table 13 below.

New small banks are subject to the DIDA. New small banks in Risk Categories II, III, and IV are subject to the brokered deposit adjustment. New small banks are not subject to the unsecured debt adjustment.

As with other assessment rates, the Board has the ability to adopt actual rates that are higher or lower than these total assessment rates without the necessity of further notice and comment rulemaking, provided that: (1) The Board cannot increase or decrease rates from one quarter to the next by more than two basis points; and (2) cumulative increases and decreases cannot be more than two basis points higher or lower than the total base rates.

Table 13—Total Base Assessment Rates, New Small Banks *

[In basis points per annum]

Risk category

Risk category II

Risk category III

Risk category IV

Initial Assessment Rate

Brokered Deposit Adjustment (added)

N/A

0 to 10

Total Assessment Rate

12 to 22

19 to 29

30 to 40

* The unsecured debt adjustment does not apply to new banks. Total assessment rates do not include the DIDA.

The FDIC does not propose changing the way assessment rates applicable to new small banks are determined.

The financial data on which the financial ratios method is based tends to be harder to interpret and less meaningful for new small banks. A new bank undergoes rapid changes in the scale and scope of operations, often causing financial ratios to be fairly volatile. In addition, a new bank's loan portfolio is often unseasoned, and therefore it is difficult to assess credit risk based solely on current financial ratios.

Further, on average, new banks have a higher failure rate than established institutions.

Current rules provide that: (1) under specified conditions, certain subsidiary small banks will be considered established rather than new, 12 CFR 327.8(k)(4); and (2) the time that a bank has spent as a federally insured credit union is included in determining whether a bank is established, 12 CFR 327.8(k)(5). If a Risk Category I small bank is considered established under these rules, but has no CAMELS component ratings, its initial assessment rate is 2 basis points above the minimum initial assessment rate applicable to Risk Category I (which is equivalent to 2 basis points above the minimum initial assessment rate for established small banks) until it receives CAMELS component ratings. Thereafter, the assessment rate is determined by annualizing, where appropriate, financial ratios obtained from all quarterly Call Reports that have been filed, until the bank files four quarterly Call Reports. For small banks that are considered established under these rules, but do not have CAMELS component ratings, the FDIC proposes the following:

1. If the bank has no CAMELS composite rating, its initial assessment rate would be 2 basis points above the minimum initial assessment rate for established small banks until it receives a CAMELS composite rating; and

2. If the bank has a CAMELS composite rating but no CAMELS component ratings, its initial assessment rate would be determined using the financial ratios method by substituting its CAMELS composite rating for its weighted average CAMELS component rating and, if the bank has not yet filed four quarterly Call Reports, by annualizing, where appropriate, financial ratios obtained from all quarterly Call Reports that have been filed.

Empirical studies show that new banks exhibit a “life cycle” pattern, and it takes close to a decade after its establishment for a new bank to mature.

Despite low profitability and rapid growth, banks that are three years or newer have, on average, a probability of failure lower than established banks, perhaps owing to large capital cushions and close supervisory attention. However, after three years, new banks' failure probability, on average, surpasses that of established banks. New banks typically grow more rapidly than established banks and tend to engage in more high-risk lending activities funded by large deposits. Studies based on data from the 1980s showed that asset quality deteriorated rapidly for many new banks as a result, and failure probability (conditional upon survival in prior years) reached a peak by the ninth year. Many financial ratios of new banks generally begin to resemble those of established banks by about the seventh or eighth year of their operation. See Chiwon Yom, “Recently Chartered Banks” Vulnerability to Real Estate Crisis,” FDIC Banking Review 17 (2005): 115 and Robert DeYoung, “For How Long Are Newly Chartered Banks Financially Fragile?” Federal Reserve Bank of Chicago Working Paper Series 2000-09.

V. Expected Effects of the Proposed Rule

Effect on Assessment Rates

To illustrate the effects of the proposal on small bank assessment rates, the FDIC compared actual assessment rates of established small banks as of the end of 2014, using a range of initial assessment rates of 5 basis points to 35 basis points with hypothetical assessment rates under Table 9 of the proposal (which has an overall range of assessment rates of 3 basis points to 30 basis points).

The proportion (and number) of established small banks paying the minimum initial assessment rate would have increased significantly, from 23.3 percent in actuality (1,493 small banks) to 56.0 percent under the proposal (3,584 small banks). The proportion (and number) of established small banks paying the maximum assessment rate would have decreased from 0.7 percent of established small banks in actuality (43 small banks) to 0.1 percent of established small banks under the proposal (7 small banks). Most established small banks (5,922 or 92.5 percent) would have had rate decreases. On average, Risk Category I established small banks would have had a rate decrease of 2.4 basis points, and Risk Category II, III, and IV established small banks would have had a rate decrease of 6.5 basis points. Of the Risk Category II, III, and IV established small banks, 96.3 percent would have had rate decreases; the average decrease would have been 6.8 basis points. 481 established small banks (7.5 percent of established small banks) would have had rate increases. Of the Risk Category I established small banks, 8.0 percent would have had rate increases; the average increase would have been 1.6 basis points.

The proposal assumes a range of initial assessment rates from 3 basis points to 30 basis points. For purposes of determining assessment rates for the illustration, the FDIC converted the statistical model to a range of assessment rates from 3 basis points to 30 basis points so that, for the fourth quarter of 2014, aggregate assessments for all established small banks under the proposal would have equaled, as closely as reasonably possible, aggregate assessments for all established small banks under the rate schedule in Table 4 (the rates that, under current rules, will automatically go into effect when the reserve ratio reaches 1.15 percent). Initial assessment rates under the rate schedule actually in effect for the fourth quarter of 2014 ranged from 5 basis points to 35 basis points, since the DIF reserve ratio was under 1.15 percent.

Chart 1 below graphically compares the distribution of established small bank initial assessment rates under this illustration. The horizontal axis in the chart represents established small banks ranked by risk, from the least risky on the left to the most risky on the right. Because actual risk rankings under the current small bank deposit insurance assessment system differ from risk rankings under the proposal, a particular point on the horizontal axis is not likely to represent the same bank for the current system and the proposal. Thus, the chart does not show how an individual bank's assessment would change under the proposal; it simply compares the distribution of assessment rates under the current system to the distribution under the proposal.

EP13JY15.147

To further illustrate the effects of the proposal on small bank assessment rates, the FDIC compared hypothetical assessment rates under the proposal with the assessment rates established small banks would have been charged as of the end of 2014 if the assessment rate schedule that, under current rules, will go into effect when the reserve ratio reaches 1.15 percent had been in effect. The proportion of established small banks paying the minimum initial assessment rate would also have increased from 23.3 percent in actuality to 56.0 percent under the proposal and the proportion of established small banks paying the maximum assessment rate would also have decreased from 0.7 percent of established small banks in actuality to 0.1 percent of established small banks under the proposal. Most established small banks (3,814 or 59.5 percent) would have had rate decreases. On average, Risk Category I established small banks would have had a rate decrease of 0.4 basis points, and Risk Category II, III, and IV established small banks would have had a rate decrease of 3.7 basis points. Of the Risk Category II, III, and IV established small banks, 90.9 percent would have had rate decreases; the average decrease would have been 4.4 basis points. 1,268 established small banks (19.8 percent of established small banks) would have had rate increases. Of the Risk Category I established small banks, 21.4 percent would have had rate increases; the average increase would have been 1.9 basis points.

Chart 2 below graphically compares the distribution of established small bank initial assessment rates under this illustration.

EP13JY15.148

Effect on Capital and Earnings

Appendix 2 to the Supplementary Information section of this notice discusses the effect of the proposal on the capital and earnings of small established banks in detail. Annualizing fourth quarter 2014 balance sheet data, Appendix 2 analyzes the effects of the proposal on capital and income in two ways: (1) The effect of the proposal compared to the current small bank deposit insurance assessment system under the rate schedule in Table 3 (with an initial assessment rate range of 5 basis points to 35 basis points) (the first comparison); and (2) the effect of the proposal compared to the current small bank deposit insurance assessment system under the rate schedule in Table 4 (with an initial assessment rate range of 3 basis points to 30 basis points; this rate schedule is to go into effect the quarter after the DIF reserve ratio reaches 1.15 percent) (the second comparison).

Under either comparison, the proposal would cause no small banks to fall below a 4 percent or 2 percent leverage ratio that would otherwise be above these thresholds. Similarly, the proposal would cause no small banks to rise above a 2 percent leverage ratio that would otherwise be below this threshold. Two established small banks facing a decrease in assessments under the first comparison and one established small bank facing a decrease in assessments under the second comparison would, as a result of the proposal, have their leverage ratios rise above 4 percent, when they would have been below 4 percent otherwise.

In the first comparison, only approximately 7 percent of profitable established small banks and approximately 6 percent of unprofitable small banks would face a rate increase; all but a very few (26) banks would have resulting declines in income (or increases in losses, where the bank is unprofitable) of 5 percent or less. As discussed above, assessment rates for approximately 92 percent of established small banks would decline, resulting in increases in income (or decreases in losses), some of which would be substantial.

In the second comparison, approximately 20 percent of profitable established small banks and approximately 14 percent of unprofitable established small banks would face a rate increase; all but 111 established small banks would have resulting declines in income (or increases in losses, where the bank is unprofitable) of 5 percent or less. As discussed above, assessment rates for approximately 60 percent of established

small banks would decline, resulting in increases in income (or decreases in losses), some of which would be substantial.

In sum, because the proposed revisions are intended to generate the same total revenue from small banks as would have been generated absent the proposal, the revisions should, overall, have no effect on the capital and earnings of the banking industry, although the revisions will affect the earnings and capital of individual institutions.

VI. Backtesting

To evaluate the proposed revisions to the risk-based deposit insurance assessment system for small banks, the FDIC tested how well the revised system would have differentiated between banks that failed and those that did not during the recent crisis compared to the current small bank deposit insurance assessment system.

Table 14 compares accuracy ratios for the proposed system and the current small bank deposit insurance assessment system. An accuracy ratio compares how well each approach would have discriminated between banks that failed within the projection period and those that did not. The projection period in each case is the three years following the date of the projection (the first column), which is the last day of the year given. Thus, for example, the accuracy ratios for 2006 reflect how well each approach would have discriminated in its projection between banks that failed and those that did not from 2007 through 2009.

A “perfect” projection would receive an accuracy ratio of 1; a random projection would receive an accuracy ratio of 0.

The current small bank deposit insurance assessment system did not exist at the end of 2006 and existed in somewhat different forms in years before 2011. The comparison assumes that the small bank deposit insurance assessment system in its current form existed in each year of the comparison.

A “perfect” projection is defined as one where the projection rates every bank that fails over the projection period as more risky than every bank that does not fail. A random projection is one where the projection does no better than chance; that is, any given percentage of banks with projected higher risk will include the same percentage of banks that fail over the projection period. Thus, for example, in a random projection, the 10 percent of banks that receive the highest risk projections will include 10 percent of the banks that fail over the projection period; the 20 percent of banks that receive the highest risk projections will include 20 percent of the banks that fail over the projection period, and so on.

Table 14—Accuracy Ratio Comparison Between the Proposal and the Current Small Bank Deposit Insurance Assessment System

Year of projection

Accuracy ratio for the proposal *

Accuracy ratio for the current small bank assessment system

Accuracy ratio for the proposal—

accuracy ratio for the current system

(A)

(B)

(A-B)

2006

0.7029

0.3491

0.3539

2007

0.7779

0.5616

0.2163

2008

0.8930

0.7825

0.1105

2009

0.9398

0.9015

0.0383

2010

0.9657

0.9394

0.0262

2011

0.9485

0.9323

0.0161

* The accuracy ratio for the proposal is based on the conversion of the statistical model as estimated through 2014.

The table reveals that, while the current system did relatively well at capturing risk and predicting failures in more recent years, the proposed system would have not only done significantly better immediately before the recent crisis and at the beginning of the crisis, but also better overall.

In the early part of the crisis, when CAMELS ratings had not fully reflected the worsening condition of many banks, the proposed system would have recognized risk far better than the current system, primarily because the rates under the proposed system are not constrained by risk categories. As the crisis progressed and CAMELS ratings more fully reflected crisis conditions, the superiority of the proposed system decreased, but it still performed better than the current system.

As implied in the footnote to Table 14, the accuracy ratios in the table for the proposed system are based on in-sample backtesting. In-sample backtesting compares model forecasts to actual outcomes where those outcomes are included in the data used in model development. Out-of-sample backtesting is the comparison of model predictions against outcomes where those outcomes are not used as part of the model development used to generate predictions. Out-of-sample backtesting, discussed in Appendix 1 of the Supplementary Information section of this notice, also shows that, while the current assessment system for small banks did relatively well at predicting failures in more recent years, the proposed system would have done significantly better immediately before the recent crisis and at the beginning of the crisis, but also better overall.

Appendix 1 to the Supplementary Information section of this notice contains a more detailed description of the FDIC's backtests of the proposal.

VII. Alternatives Considered

Alternative Minimum and Maximum Assessment Rates Based on CAMELS Composite Ratings

The FDIC considered imposing no minimum or maximum initial assessment rates based on a bank's CAMELS composite rating, which would have allowed initial assessment rates to vary between the minimum and maximum initial assessment rates of the entire rate schedule without regard to a bank's CAMELS composite rating (the unbounded variation). Thus, for example, under the 3 basis point to 30 basis point initial assessment range, a CAMELS composite 5 rated bank could, in principle, have paid a 3 basis point initial rate and a CAMELS composite 1 rated bank could, in principle, have paid a 30 basis point initial rate. As Table 15 shows, the accuracy ratios for this unbounded variation would have been similar to the accuracy ratios for the proposal.

Table 15—Accuracy Ratio Comparison Between the Proposal and the Unbounded Variation

Year of projection

Accuracy ratio for the unbounded variation

Accuracy ratio for the proposal *

Accuracy ratio for the unbounded variation—accuracy ratio for the proposal (A-B)

(A)

(B)

2006

0.6959

0.7029

−0.0070

2007

0.7779

0.0001

2008

0.9121

0.8930

0.0191

2009

0.9407

0.9398

0.0010

2010

0.9670

0.9657

0.0013

2011

0.9514

0.9485

0.0029

* The accuracy ratios for the variation and for the proposal are based on the conversion of the statistical model as estimated through 2014.

The FDIC decided not to propose the unbounded variation, however. Other than taking into account weighted average CAMELS component ratings, the statistical model uses historical financial data to estimate average relationships between financial measures and the risk of failure. The statistical model does not take into account idiosyncratic or unquantifiable risk or risk mitigators (

e.g.,

entering or exiting a risky line of lending; having inexperienced or experienced management, reducing or tightening underwriting requirements), again except through weighted average CAMELS component ratings. The model does take into account weighted average CAMELS component ratings, but it assigns the same weight to them for each bank. Thus, for banks that have significant idiosyncratic or unquantifiable risk or risk mitigators, the model may not assign an assessment rate that reflects their actual risk. The proposal, however, ensures that the assessment system takes idiosyncratic and unquantifiable risks and risk mitigators into account to the extent that they are reflected in CAMELS composite ratings, and prevents the assessment system from assigning a rate that reflects either too little risk (for a bank with a CAMELS composite 3, 4 or 5 rating) or too much risk (for a bank with a CAMELS composite 1 or 2 rating). As a result, under the proposal, initial assessment rates for small banks that are well rated (those with CAMELS composite ratings of 1 or 2) would not overlap with initial assessment rates for troubled small banks (those with CAMELS composite ratings of 4 or 5), except at the maximum initial rate for CAMELS composite 1- and 2-rated banks and the minimum initial rate for CAMELS composite 4- and 5-rated banks.

In seeking the proper balance between maintaining the accuracy of the assessment system overall and reducing the risk that a particular bank's assessment rate might be inappropriate, the FDIC considered many other variations of minimum and maximum initial assessment rates based on a bank's CAMELS composite rating. Some variations with lower (or no) minimums for CAMELS 3- and/or CAMELS 4- and 5-rated banks and/or higher (or no) maximums for CAMELS 1- and/or CAMELS 2-rated banks had slightly higher accuracy ratios, but would have increased the risk of inappropriate assessment rates for some banks. Some variations with higher minimums for CAMELS 3- and/or CAMELS 4- and 5-rated banks and/or lower maximums for CAMELS 1- and/or CAMELS 2-rated banks had somewhat lower (or significantly lower) accuracy ratios. The maximums and minimums in the proposal represent the FDIC's best judgment on the proper balance. The FDIC is requesting comment on whether the proposal achieves the proper balance and whether the final rule should, instead, use alternative (or no) maximums and minimums based on CAMELS composite ratings. Because the FDIC intends that the effect of the proposal be revenue neutral, any reduction in the maximum initial assessment rate applicable to CAMELS composite 1- or CAMELS 2-rated banks that lowers some banks' assessment rates will increase the assessment rates of other banks.

To be revenue neutral, using different maximums or minimums will lead to different uniform amounts and pricing multipliers from the proposal when the new statistical model is converted to assessment rates.

The FDIC is particularly interested in comment on two alternatives to the proposal, both of which would distinguish between CAMELS composite 1- and 2-rated small banks. The first alternative would maintain the assessment rate schedule that would go into effect starting the quarter after the reserve ratio reaches 1.15 percent (with a range of initial assessment rates of 3 basis points to 30 basis points) and include the same maximum and minimum assessment rates based upon banks' CAMELS composite ratings (see Table 9), except that it would lower the maximum initial assessment rate for a CAMELS composite 1-rated bank from 16 basis points to 12 basis points.

As reflected in Table 16 below, compared to the proposal, this alternative would have virtually no effect on accuracy (that is, on how well the assessment system would have differentiated between banks that failed and those that did not during the recent crisis); the alternative, like the proposal, is also significantly more accurate than the current small bank deposit insurance assessment system. On the other hand, the FDIC has never before distinguished between CAMELS composite 1-rated banks and CAMELS composite 2-rated banks for deposit insurance assessment purposes.

Similarly, the first alternative would maintain the proposed assessment rate schedule that would go into effect the quarter after the reserve ratio reaches or exceeds 2 percent, but is less than 2.5 percent, and include the same maximum and minimum assessment rates determined by CAMELS composite ratings (see Table 10), except that it would lower the maximum initial assessment rate for a CAMELS composite 1 rated bank from 14 basis points to 10 basis points. Also, the first alternative would maintain the proposed assessment rate schedule that would go into effect the quarter after the reserve ratio reaches or exceeds 2.5 percent, and include the same maximum and minimum assessment rates determined by CAMELS composite ratings (see Table 11), except that it would lower the maximum initial assessment rate for a CAMELS composite 1 rated bank from 13 basis points to 9 basis points.

Table 16—Accuracy Ratio Comparison Between the First Alternative, the Proposal and the Current Small Bank Deposit Insurance Assessment System

Year of projection

Accuracy ratio for the alternative *

Accuracy ratio for the proposal *

Accuracy ratio for the alternative—accuracy ratio for the proposal (A-B)

Accuracy ratio for the current small bank assessment system

Accuracy ratio for the alternative—accuracy ratio for the current system (A-C)

(A)

(B)

(C)

2006

0.7045

0.7029

0.0016

0.3491

0.3555

2007

0.7770

0.7779

−0.0009

0.5616

0.2154

2008

0.8895

0.8930

−0.0035

0.7825

0.1070

2009

0.9398

0.0000

0.9015

0.0383

2010

0.9657

0.0000

0.9394

0.0262

2011

0.9485

0.0000

0.9323

0.0161

* The accuracy ratios for the alternative and for the proposal are based on the conversion of the statistical model as estimated through 2014.

The second alternative is the same as the first, except that, for the rate schedule that would go into effect the quarter after the reserve ratio reaches 1.15 percent, the minimum initial assessment rate applicable to CAMELS composite 4- and 5-rated banks would be lowered from 16 basis points to 12 basis points.

54 55

As reflected in Table 17 below, compared to the proposal, this alternative would also have little effect on accuracy and, like the proposal, is significantly more accurate than the current small bank deposit insurance assessment system.

The second alternative would have the same assessment rate schedule go into effect the quarter after the reserve ratio reaches or exceeds 2 percent, but is less than 2.5 percent, as the first alternative and include the same maximum and minimum assessment rates determined by CAMELS composite ratings, except that it would lower the minimum initial assessment rate for a CAMELS composite 4 and 5 rated banks from 14 basis points to 10 basis points. Also, the second alternative would have the same assessment rate schedule go into effect the quarter after the reserve ratio reaches or exceeds 2.5 percent as the first alternative, and include the same maximum and minimum assessment rates determined by CAMELS composite ratings (see Table 11), except that it would lower the minimum initial assessment rate for a CAMELS composite 4- and 5-rated banks from 13 basis points to 9 basis points.

Under either alternative, if a bank's CAMELS composite or component ratings changed during a quarter (other than a change in CAMELS composite rating from a 4 to a 5 or a 5 to a 4 with no change in component ratings), including a change in CAMELS composite rating from a 1 to a 2 or a 2 to a 1, its assessment rate would be determined separately for each portion of the quarter in which it had different CAMELS composite or component ratings.

Table 17—Accuracy Ratio Comparison Between the Second Alternative, the Proposal and the Current Small Bank Deposit Insurance Assessment System

Year of projection

Accuracy ratio for the alternative *

Accuracy ratio for the proposal *

Accuracy ratio for the alternative-

accuracy ratio for the proposal (A-B)

Accuracy ratio for the current small bank assessment system

Accuracy ratio for the alternative-

accuracy ratio for the

current system (A-C)

2006

0.7061

0.7029

0.0032

0.3491

0.3570

2007

0.7779

0.0000

0.5616

0.2163

2008

0.8903

0.8930

−0.0027

0.7825

0.1078

2009

0.9407

0.9398

0.0009

0.9015

0.0392

2010

0.9671

0.9657

0.0014

0.9394

0.0276

2011

0.9504

0.9485

0.0019

0.9323

0.0180

* The accuracy ratios for the alternative and for the proposal are based on the conversion of the statistical model as estimated through 2014.

In addition to the numerous variations on minimum and maximum initial assessment rates based on CAMELS composite ratings, the FDIC also considered other alternatives when developing this proposal.

Loss Given Default

Though expected losses to the DIF are a function of both the probability of a failure (or probability of default (PD)) and the loss given failure (or loss given default (LGD)), the new statistical model estimates only the PD. As discussed in Appendix 1 to the Supplementary Information section of this notice, the FDIC did not model LGD. Actual losses for many failed banks during the recent crisis are still estimated, primarily because of the use of loss-sharing agreements that have not yet terminated. Until the losses are actually realized, estimating an LGD model using current data would be circular, as other FDIC models are used to estimate expected losses where losses have not yet been realized. Relying solely on realized losses would exclude much of the failure data from the recent crisis, leaving mainly failure data from the banking crisis of the late 1980s and early 1990s. However, the vast majority of the bank failures in that crisis occurred in a different regulatory regime (prior to the Federal Deposit Insurance Corporation Improvement Act of 1991) and may, therefore, not reflect expected LGD in the current environment as well. For these reasons, the FDIC considered but rejected including LGD in the new statistical model. Nevertheless, after losses from failures during the recent crisis are more fully realized, it may be appropriate to consider whether LGD should be included in a small bank pricing model.

No Change

The FDIC also considered leaving the current small bank deposit insurance assessment system in place unchanged. While the backtesting discussed in Appendix 1 revealed that the new statistical model generally performed

better than the current small bank deposit insurance assessment system, the current system performed relatively well. Nevertheless, the FDIC is proposing to change the small bank deposit insurance assessment system and base it on the new statistical model because the new model is superior to the current small bank deposit insurance assessment system. Under the proposed system, fewer riskier small banks would pay lower assessments and fewer safer banks would pay higher assessments than their conditions warrant.

VIII. Request for Comments

The FDIC seeks comment on every aspect of this proposed rulemaking, including the alternatives considered. In addition, the FDIC seeks comment on the following:

• Are there other variables, besides the eight included in the statistical model and proposal, that both predict the likelihood of bank failure with statistical significance and do not have perverse incentive effects?

• Are there variables that can be shown to predict likely losses given failure with statistical significance?

• Should the upper end of the assessment rate range decline from 35 basis points to 30 basis points as proposed or should higher assessment rates continue to apply to the riskiest banks?

IX. Regulatory Analysis

A. Regulatory Flexibility Act

The Regulatory Flexibility Act (RFA) requires that each federal agency either certify that a proposed rule would not, if adopted in final form, have a significant economic impact on a substantial number of small entities or prepare an initial regulatory flexibility analysis of the proposal and publish the analysis for comment.

Certain types of rules, such as rules of particular applicability relating to rates or corporate or financial structures, or practices relating to such rates or structures, are expressly excluded from the definition of “rule” for purposes of the RFA.

The proposed rule relates directly to the rates imposed on insured depository institutions for deposit insurance and to the deposit insurance assessment system that measures risk and determines each established small bank's assessment rate. Nonetheless, the FDIC is voluntarily undertaking an initial regulatory flexibility analysis of the proposal and seeking comment on it.

See 5 U.S.C. 603, 604 and 605.

5 U.S.C. 601.

As of December 31, 2014, of the 6,509 insured commercial banks and savings institutions, there were 5,257 small insured depository institutions as that term is defined for purposes of the RFA (

i.e.,

those with $550 million or less in assets).

Throughout this RFA analysis (unlike the rest of this NPR), a “small institution” refers to an institution with assets of $550 million or less; a “small bank,” however, continues to refer to a small insured depository institution for purposes of deposit insurance assessments (generally, a bank with less than $10 billion in assets).

For purposes of this analysis, whether the FDIC were to collect needed assessments under the existing rule or under the proposed rule, the total amount of assessments collected would be the same. The FDIC's total assessment needs are driven by the FDIC's aggregate projected and actual insurance losses, expenses, investment income, and insured deposit growth, among other factors, and assessment rates are set pursuant to the FDIC's long-term fund management plan. This analysis demonstrates how the new pricing system under the proposed range of assessment rates of 3 basis points to 30 basis points (P330) could affect small entities relative to the current assessment rate schedule (C535) and relative to the rate schedule that under current regulations will be in effect when the reserve ratio exceeds 1.15 percent (C330). Using data as of December 31, 2014, the FDIC calculated the total assessments that would be collected under both rate schedules and under the proposed rule.

The economic impact of the proposal on each small institution for RFA purposes (

i.e.,

institutions with assets of $550 million or less) was then calculated as the difference in annual assessments under the proposed rule compared to the existing rule as a percentage of the institution's annual revenue and annual profits, assuming the same total assessments collected by the FDIC from the banking industry.

For purposes of the analysis, an institution's total revenue is defined as the sum of its interest income and noninterest income and an institution's profit is defined as income before taxes and extraordinary items.

Projected Effects on Small Entities Assuming a Range of Assessment Rates Under Both the Current Established Small Bank Deposit Insurance Assessment System and the Proposed System of 3 Basis Points to 30 Basis Points (P330-C330)

Based on the December 31, 2014 data, of the total of 5,257 small institutions, one institution would have experienced an increase in assessments equal to five percent or more of its total revenue. These figures do not reflect a significant economic impact on revenues for a substantial number of small insured institutions. Table 18 below sets forth the results of the analysis in more detail.

Table 18—Percent Change in Assessments Resulting From the Proposal

[Assuming No Change in the Assessment Rate Range]

Change in assessments

Number of

institutions

Percent of

Institutions

More than 10 percent lower

5 to 10 percent lower

0 to 5 percent lower

3,296

0 to 5 percent higher

1,957

5 to 10 percent higher

More than 10 percent higher

Total

5,257

100

The FDIC performed a similar analysis to determine the impact on profits for small institutions. Based on December 31, 2014 data, of those small institutions with reported profits, 21 institutions would have an increase in assessments equal to 10 percent or more of their profits. Again, these figures do not reflect a significant economic impact on profits for a substantial number of small insured institutions.

Table 19 sets forth the results of the analysis in more detail.

Table 19*—Assessment Changes Relative to Profits for Profitable Small Institutions Under the Proposal

[Assuming No Change in the Assessment Rate Range]

Change in assessments relative to profits

Number of

institutions

Percent of

institutions

Decrease in assessments equal to more than 40 percent of profits

Decrease in assessments equal to 20 to 40 percent of profits

Decrease in assessments equal to 10 to 20 percent of profits

131

Decrease in assessments equal to 5 to 10 percent of profits

306

Decrease in assessments equal to 0 to 5 percent of profits

3,541

Increase in assessments equal to 0 to 5 percent of profits

706

Increase in assessments equal to 5 to 10 percent of profits

Increase in assessments equal to 10 to 20 percent of profits

Increase in assessments equal to 20 to 40 percent of profits

Increase in assessments equal to more than 40 percent of profits

Total

4,874

100

*Institutions with negative or no profit were excluded. These institutions are shown in Table 20.

Table 19 excludes small institutions that either show no profit or show a loss, because a percentage cannot be calculated. The FDIC analyzed the effect of the proposal on these institutions by determining the annual assessment change (either an increase or a decrease) that would result. Table 20 below shows that 27 (seven percent) of the 383 small insured institutions with negative or no reported profits would have an increase of $20,000 or more in their annual assessments.

Table 20—Change in Assessments for Unprofitable Small Institutions Resulting from the Proposal

[Assuming No Change in the Assessment Rate Range]

Change in assessments

Number of

Institutions

Percent of

Institutions

$20,000 or more decrease

170

$10,000-$20,000 decrease

$5,000-$10,000 decrease

$1,000-$5,000 decrease

$0-$1,000 decrease

$0-$1,000 increase

$1,000-$5,000 increase

$5,000-$10,000 increase

$10,000-$20,000 increase

$20,000 increase or more

Total

383

100

Projected Effects on Small Entities Assuming a Range of Assessment Rates Under the Current Established Small Bank Deposit Insurance Assessment System of 5 Basis Points to 35 Basis Points and Under the Proposed System of 3 Basis Points to 30 Basis Points (Assessment Change P330-C535)

Based on the December 31, 2014 data, of the total of 5,257 small institutions, no institution would have experienced an increase in assessments equal to five percent or more of its total revenue. These figures do not reflect a significant economic impact on revenues for a substantial number of small insured institutions. Table 21 below sets forth the results of the analysis in more detail.

Table 21—Percent Change in Assessments Resulting from the Proposal

[Assuming Assessment Rate Range Change From 5-35 Bps to 3-30 Bps]

Change in assessments

Number of

institutions

Percent of

institutions

More than 10 percent or lower

5 to 10 percent lower

0 to 5 percent lower

4,969

0 to 5 percent higher

280

More than 5 percent higher

Total

5,257

100

The FDIC performed a similar analysis to determine the impact on profits for small institutions. Based on December 31, 2014 data, of those small institutions with reported profits, eight institutions would have an increase in assessments equal to 10 percent or more of their profits. Again, these figures do not reflect a significant economic impact on profits for a substantial number of small insured institutions. Table 22 sets forth the results of the analysis in more detail.

Table 22*—Assessment Changes Relative to Profits for Profitable Small Institutions Under the Proposal

[Assuming Assessment Rate Range Change From 5-35 Bps to 3-30 Bps]

Change in assessments relative to profits

Number of

institutions

Percent of

institutions

Decrease in assessments equal to more than 40 percent of profits

119

Decrease in assessments equal to 20 to 40 percent of profits

Decrease in assessments equal to 10 to 20 percent of profits

285

Decrease in assessments equal to 5 to 10 percent of profits

603

Decrease in assessments equal to 0 to 5 percent of profits

3,513

Increase in assessments equal to 0 to 5 percent of profits

239

Increase in assessments equal to 5 to 10 percent of profits

Increase in assessments equal to 10 to 20 percent of profits

Increase in assessments equal to 20 to 40 percent of profits

Increase in assessments equal to more than 40 percent of profits

Total

4,874

100

* Institutions with negative or no profit were excluded. These institutions are shown in Table 23.

Table 22 excludes small institutions that either show no profit or show a loss, because a percentage cannot be calculated. The FDIC analyzed the effect of the proposal on these institutions by determining the annual assessment change (either an increase or a decrease) that would result. Table 23 below shows that just 11 (three percent) of the 383 small insured institutions with negative or no reported profits would have an increase of $20,000 or more in their annual assessments. Again, these figures do not reflect a significant economic impact on profits for a substantial number of small insured institutions.

Table 23—Change in Assessments for Unprofitable Small Institutions Resulting From the Proposal

[Assuming No Change in the Assessment Rate Range]

Change in assessments

Number of

institutions

Percent of

institutions

$20,000 or more decrease

262

$10,000-$20,000 decrease

$5,000-$10,000 decrease

$1,000-$5,000 decrease

$0-$1,000 decrease

$0-$1,000 increase

$1,000-$5,000 increase

$5,000-$10,000 increase

$10,000-$20,000 increase

$20,000 increase or more

Total

383

100

The proposed rule does not directly impose any “reporting” or “recordkeeping” requirements within the meaning of the Paperwork Reduction Act. The compliance requirements for the proposed rule would not exceed (and, in fact, would be the same as) existing compliance requirements for the current risk-based deposit insurance assessment system for small banks. The FDIC is unaware of any duplicative, overlapping or conflicting federal rules.

The initial RFA analysis set forth above demonstrates that, if adopted in final form, the proposed rule would not have a significant economic impact on a substantial number of small institutions within the meaning of those terms as used in the RFA.

5 U.S.C. 605.

Commenters are invited to provide the FDIC with any information they may have about the likely quantitative effects of the proposal on small insured depository institutions (those with $550 million or less in assets).

B. Riegle Community Development and Regulatory Improvement Act:

The Riegle Community Development and Regulatory Improvement Act (RCDRIA) requires that the FDIC, in determining the effective date and administrative compliance requirements of new regulations that impose additional reporting, disclosure, or other requirements on insured depository institutions, consider, consistent with principles of safety and soundness and the public interest, any administrative burdens that such regulations would place on depository institutions, including small depository institutions, and customers of depository institutions, as well as the benefits of such regulations.

12 U.S.C. 4802.

This NPR proposes no additional reporting or disclosure requirements on insured depository institutions, including small depository institutions, nor on the customers of depository institutions.

C. Paperwork Reduction Act:

No collections of information pursuant to the Paperwork Reductions Act (44 U.S.C. 3501

et seq.

) are contained in the proposed rule.

D. The Treasury and General Government Appropriations Act, 1999—Assessment of Federal Regulations and Policies on Families

The FDIC has determined that the proposed rule will not affect family well-being within the meaning of section 654 of the Treasury and General Government Appropriations Act, enacted as part of the Omnibus Consolidated and Emergency Supplemental Appropriations Act of 1999 (Pub. L. 105-277, 112 Stat. 2681).

E. Solicitation of Comments on Use of Plain Language

Section 722 of the Gramm-Leach-Bliley Act, Public Law 106-102, 113 Stat. 1338, 1471 (Nov. 12, 1999), requires the Federal banking agencies to use plain language in all proposed and final rules published after January 1, 2000. The FDIC invites your comments on how to make this proposal easier to understand. For example:

• Has the FDIC organized the material to suit your needs? If not, how could the material be better organized?

• Are the requirements in the proposed regulation clearly stated? If not, how could the regulation be stated more clearly?

• Does the proposed regulation contain language or jargon that is unclear? If so, which language requires clarification?

• Would a different format (grouping and order of sections, use of headings, paragraphing) make the regulation easier to understand?

Appendix 1—Description of Statistical Model Underlying Proposed Method for Determining Deposit Insurance Assessments For Established Small Insured Depository Institutions

This appendix provides a technical description of the statistical model (the “new model”)

underlying the proposed method for determining deposit insurance assessments for established small banks. The appendix provides background information, reviews the data and methodology used to estimate the new model underlying the proposed method, discusses estimation results and alternative specifications considered, and evaluates the results.

The preamble to the NPR refers to the new model as the “statistical model.”

I. Background

A. RRPS

The current small bank deposit insurance assessment system has been in effect, with some modifications, since January 1, 2007. The current small bank deposit insurance system assigns assessment rates in several steps. The first step assigns small banks to risk categories. The categories are jointly determined by bank capital and supervisory ratings. Well-capitalized small banks rated CAMELS 1 or 2 are placed in Risk Category I.

Small banks with lower capital or weaker CAMELS ratings are placed in either Risk Category II, Risk Category III or Risk Category IV.

Unless explicitly stated otherwise, references to CAMELS ratings are references to CAMELS composite ratings.

The second step differentiates risk further among Risk Category I small banks using the

financial ratios method,

which combines supervisory CAMELS component ratings with current financial ratios to determine a Risk Category I small bank's initial assessment rate. The contribution of these variables (the CAMELS component ratings and the financial ratios) to assessment rates is determined using a linear model (the downgrade probability model or existing model) estimating the probability that a CAMELS 1- or 2-rated bank will be downgraded to a CAMELS rating of 3 or worse within 12 months.

In November 2006, when the final rule establishing the current small bank deposit insurance system was adopted, it had been more than a decade since the United States experienced a significant number of bank failures. Consequently, historical downgrades were used as a proxy for the risk to the DIF of a bank's failure.

The data generated by the rash of bank failures since the financial crisis of 2008 suggests that the model underlying the small bank deposit insurance assessment system can be improved and updated.

B. Probability of Default

The data generated from the approximately 500 bank failures since 2008 suggests that the probability of downgrade probability model can be replaced by a probability of default (that is, a probability of failure) model. Failures are nearly always costly to the FDIC, whereas downgrades lead to DIF losses relatively infrequently, since many downgraded banks do not fail.

C. Loss Given Default

Though expected losses to the DIF are a function of both the probability of a default (PD) and the loss given default (LGD), the new model estimates only the PD. LGD was not modeled. Actual losses for many of the failed banks during the crisis are still estimated, primarily because of the use of loss-sharing agreements that have not yet terminated. Until the losses are actually realized, estimating a loss given default model using current data would be circular, as FDIC models are used to estimate expected losses where losses have not yet been realized. Relying solely on realized losses would exclude much of the failure data from the recent crisis, leaving mainly failure data from the banking crisis of the late 1980s and early 1990s. However, the vast majority of the bank failures in that crisis occurred in a different regulatory regime (prior to the Federal Deposit Insurance Corporation Improvement Act of 1991

) and may, therefore, not reflect expected LGD in the current environment as well. See Bennett and Unal (2014).

FDIC (1998), Legislation Governing the FDIC's Roles as Insurer and Receiver,” from

Managing the Crisis, https://www.fdic.gov/bank/historical/managing/history3-A.pdf,

p. 774-747.

Notwithstanding these concerns, a careful consideration of whether future rulemaking should include LGD in a small bank deposit insurance assessment model may be appropriate after most losses are realized from failures during the recent crisis.

II. Methodology

A. Variable Selection

In addition to the existing model, the FDIC relied on other existing models of bank risk, both regulatory and academic, to select candidate variables for inclusion in the new model.

1. SCOR

The Statistical CAMELS Offsite Rating (SCOR) system is one of FDIC's offsite monitoring models and is used to identify banks whose financial condition has deteriorated since their last on-site examination. SCOR is designed as a short-term model with a one-year forecast horizon, to identify institutions that are currently CAMELS 1 or 2 rated that might receive a rating of CAMELS 3, 4 or 5 at the next examination.

The SCOR model uses an ordered logistic regression to predict the composite CAMELS rating and the six CAMELS component ratings. A logistic regression allows for nonlinear relationships between each explanatory

variable and the dependent variable (the variable that depends upon the explanatory variable). In an ordered logistic regression, the dependent variable (CAMELS) can only have discrete values that are ordered. (In the case of CAMELS, the ordered values are 1 through 5.) The other variables (the explanatory variables) are then used to predict the likelihood of observing each of the possible outcomes.

SCOR uses twelve variables to measure banks' financial condition. These financial measures are (as a ratio to total assets): equity, loan loss reserves, loans past due 30-89 days, loans past due 90+ days, nonaccrual loans, other real estate owned, charge-offs, provisions for loan losses and transfer risk, income before taxes and extraordinary charges, volatile liabilities, liquid assets, and loans and long term securities.

Detailed description of the model and the variables used in SCOR can be found in “The SCOR System of Off-Site Monitoring: Its Objectives, Functioning, and Performance,” Collier, Forbush, Nuxoll, and O'Keefe (2003).

2. GMS

The Growth Monitoring System (GMS) is one of FDIC's offsite monitoring models designed to monitor banks' risk taking associated with rapid growth and heavy reliance on non-traditional sources of funds. GMS is designed to identify distress and failure before bank conditions actually weaken, thereby allowing supervisors to take preventive action.

GMS estimates the likelihood that a bank will be downgraded from a CAMELS 1 or 2 rating to a CAMELS 3, 4 or 5 rating within three years as a function of the bank's current risk characteristics. The explanatory variables include a bank's asset growth, equity ratio, loan to asset ratio, noncore funds to asset ratio, change in loan mix index, reserve coverage ratio and a binary variable indicating whether a bank is currently CAMELS 1 rated.

Detailed description of the GMS model can be found in “Bank Growth and Long Term Risk,” Hwa, Jacewitz, and Yom (May 2011).

3. Academic

There exist numerous papers discussing models that predict bank failures. In these papers, the explanatory variables predicting bank failures are largely divided into measures of (1) capital; (2) asset quality; (3) earnings; (4) liquidity; (5) sensitivity to market risk; and (6) other risk measures.

A bank's capital adequacy is an important predictor of its survival because it provides a cushion to withstand unanticipated losses. Studies have used a total equity to total assets ratio (Santoni, Ricci, and Kelshiker (2010), Betz, Oprica, Peltonen, Sarlin (2012)) or the leverage ratio (Santoni, Ricci, and Kelshiker (2010)) to measure a bank's equity position. These studies find that higher capital ratios are correlated with lower failure probability.

To measure a bank's asset quality, nonperforming loans (Wheelock and Wilson (2000), Santoni, Ricci, and Kelshiker (2010), Gilbert, Meyer, and Vaughan (1999)) and other real estate owned to total assets ratios have been used. A large volume of nonperforming loans and other real estate owned relative to total loans (or total assets) signal low credit quality in a bank's loan portfolio.

Higher bank earnings also provide a cushion to withstand adverse economic shocks and lower failure probability. To measure bank earnings, measures such as net income before taxes, interest expense (Betz, Oprica, Peltonen, Sarlin (2012)), and total operating income (Lane, Looney, and Wansley (1986)) have been used.

Loan portfolio ratios, such as commercial and industrial (C&I) loans, commercial real estate loans, construction and development (C&D) loans, and consumer loans (Cole and Gunther (1995), Whalen (1991), Lane, Looney, and Wansley (1986)), have been used to measure a bank's concentration in different loan types.

Rapid loan growth or asset growth can be indicators of a bank's aggressive risk-taking and of underwriting loans or acquiring assets with lower creditworthiness. A correlation between rapid credit growth and bank distress has been well documented in academic research (Solttila and Vihriala (1994), Clair (1992), Salas and Saurina (2002), Keeton (1999), Foos, Norden, and Weber (2009), and Logan (2001)).

Liquidity measures include a core deposits to total assets ratio (Gilbert, Meyer, Vaughan (1999)) and a liquid assets to total assets ratio (Gilbert, Meyer, Vaughan (1999), Lane, Looney, and Wansley (1986)). These measures can indicate a bank's ability to meet unexpected liquidity needs. A high loans to total deposits ratio (Gilbert, Meyer, Vaughan (1999)) or loans to total assets ratio can indicate a bank's illiquidity, since loans are typically less liquid than other assets on a bank's balance sheet.

Bank size (Gilbert, Meyer, Vaughan (1999), Wheelock and Wilson (2000)) can predict failure likelihood, since large banks can benefit from diversification across product lines and geographic regions.

Whether a bank is a part of a holding company is another measure used by some studies (Gilbert, Meyer, Vaughan (1999), Wheelock and Wilson (2000)). An indicator of holding company affiliation can predict failure probability, since a holding company can serve as a source of strength to banks.

Onali (2012) finds a positive relation between bank default risk and dividend payout ratios. This finding is consistent with the theory that dividend payouts exacerbate moral hazard. He finds, however, that the relationship is insignificant for banks that are very close to failure.

B. Variables

Table 1.1 lists and describes the variables that are included in the new model as the result of reviewing academic studies on bank risk and testing candidate variables.

Table 1.1—New Model Variable Description

Variables

Description

Tier 1 Leverage Ratio (%)

Tier 1 capital divided by adjusted average assets. (Numerator and denominator are both based on the definition for prompt corrective action.)

Net Income before Taxes/Total Assets (%)

Income (before income taxes and extraordinary items and other adjustments) for the most recent twelve months divided by total assets.

Nonperforming Loans and Leases/Gross Assets

(%)

Sum of total loans and lease financing receivables past due 90 or more days and still accruing interest and total nonaccrual loans and lease financing receivables (excluding, in both cases, the maximum amount recoverable from the U.S. Government, its agencies or government-sponsored enterprises, under guarantee or insurance provisions) divided by gross assets.*

Other Real Estate Owned/Gross Assets (%)

Other real estate owned divided by gross assets.

Core Deposits/Total Assets (%)

Domestic office deposits (excluding time deposits over the deposit insurance limit and the amount of brokered deposits below the standard maximum deposit insurance amount) divided by total assets.

Weighted Average of C, A, M, E, L, and S Component Ratings

The weighted sum of the “C,” “A,” “M,” “E”, “L”, and “S” CAMELS components, with weights of 25 percent each for the “C” and “M” components, 20 percent for the “A” component, and 10 percent each for the “E”, “L”, and “S” components. In instances where the “S” component is missing, the remaining components are scaled by a factor of 10/9.**

Loan Mix Index

A measure of credit risk described below.

Asset Growth (%)

Growth in assets (merger adjusted) over the previous year. If growth is negative, then the value is set to zero.

“Gross assets” are total assets plus the allowance for loan and lease financing receivable losses (ALLL); for purposes of estimating the statistical model, for years before 2001, when allocated transfer risk was not included in ALLL in Call Reports, allocated transfer risk was included in gross assets separately.

* Delinquency and non-accrual data on government guaranteed loans are not available for the entire estimation period. As a result, the model is estimated without deducting delinquent or past-due government guaranteed loans from the nonperforming loans and leases to gross assets ratio.

** The component rating for sensitivity to market risk (the “S” rating) is not available for years before 1997. As a result, and as described in the table, the model is estimated using a weighted average of five component ratings excluding the “S” component where the component is not available.

1. Equity

The new model includes the leverage ratio (as defined in the FDIC's capital regulations

). This variable was statistically significant across specifications (that is, it was statistically significant regardless of the other variables included in the model).

12 CFR 3.10; 12 CFR 217.10; 12 CFR 324.10.

2. Loan Mix Index

Consistent with the GMS model, the FDIC included a loan mix index (“LMI”) variable that aggregates a bank's loan portfolio and historical loan category charge-offs into a single variable. Statistically, combining the loan categories into a single index increases the explanatory power of the model.

For each loan category, the LMI assigns an industry-wide charge-off rate based on historical data. A bank's LMI value is then the sum of the products of each of that bank's loan category exposures as a percentage of total assets and the associated charge-off rate. Appendix 1.1 to the Supplementary Information section of this notice shows how the LMI is constructed for a hypothetical bank.

In constructing the LMI, many alternatives were considered, including: using the change in a bank's amount of loans in a loan category rather than simply the amount of loans in a loan category, weighting charge-offs more heavily during crises and evaluating loans in a loan category as a proportion of total loans rather than as a proportion of assets.

Both in in-sample and out-of-sample backtesting, the LMI using a bank's amount of loans in a loan category had higher forecast accuracy than using the change in a bank's amount of loans in a loan category from a previous period. In-sample backtesting compares model forecasts to actual outcomes where those outcomes are included in the data used in model development. Out-of-sample backtesting is the comparison of model predictions against outcomes where those outcomes are not used as part of the model development used to generate predictions.

In-sample, all of the explanatory power came from using the amount of loans in a loan category. Out-of-sample, including the change in a bank's amount of loans in a loan category in addition to the amount of loans in a loan category did not improve performance.

Three alternative methods of averaging yearly historical industry-wide charge-off rates were considered: an unweighted average of each year's industry-wide charge-off rate, an unweighted average of each of the recent crisis years' industry-wide charge-off rates, and an average of each year's industry-wide charge-off rate weighted by the number of bank failures in the year. Out-of-sample performance for the LMI variable using an average weighted by the number of bank failures in the year slightly outperformed the LMI variable using an unweighted average over recent crisis years and more significantly outperformed the LMI variable using an unweighted average. The LMI variable using an average weighted by the number of bank failures in a year was selected over the LMI variable using an unweighted average over recent crisis years because the latter variable requires a determination of what constitutes a crisis. No such determination is necessary using the variable selected.

The FDIC also considered using total loans as the denominator of the LMI along with a liquidity variable, but elected to use total assets as the denominator to avoid imposing excessive penalties on banks that hold few loans relative to assets. (The liquidity variable was not statistically significant when total assets were used as the denominator.) Using loans as a proportion of total assets has the advantage of not extrapolating risk exposures in loans to a bank's entire asset portfolio, although it effectively assigns zero risk to all non-loan assets, implicitly treating loans as riskier than investments in other assets. Many of these other assets, however, are liquid assets. Out-of-sample performance of the models using total assets as the denominator did not differ much from the performance using total loans as the denominator along with a liquidity variable.

3. Asset Growth

Among the variables included in the specifications was a one-year asset growth rate. The FDIC also considered a two-year growth rate and lagged one- and two-year growth rates. The one-year growth rates generally had the most explanatory power and additional growth rates did not tend to improve the model's fit.

Mergers of troubled banks into healthier banks and purchases of failed banks help limit losses to the DIF. Penalizing banks for growth that occurs through the acquisition of troubled or failed banks would create a disincentive for such mergers. Consequently, bank

asset growth was adjusted to remove growth resulting from mergers and failed bank acquisitions.

4. Income

Consistent with previous findings, net income before taxes was found to be a significant explanatory variable.

5. Core Deposits

Early test versions of the new model used noncore liabilities as a variable predictive of failure. This variable was statistically significant in-sample across all specifications with a positive correlation with failure. Subsequent versions used core deposits as the alternative variable. It provides similar predictive power, and is the variable maintained for the proposed version of the new model.

6. Nonperforming Loans and Leases

Nonperforming loans and leases are defined as the sum of total loans and leases past due 90 or more days and total nonaccrual loans and leases. This variable, which measures bank asset quality, was found to be a statistically significant predictor of failure.

7. Other Real Estate Owned

The ratio of other real estate owned to gross assets is another measure of a bank's asset quality and was a significant predictor of failure across specifications.

8. CAMELS

A weighted CAMELS component variable was included in the new model to capture examination ratings. The weighted CAMELS component variable is calculated with the following weights on the component ratings: Capital (25%), Asset quality (20%), Management (25%), Earnings (10%), Liquidity (10%), Sensitivity to market risk (10%). For model estimation, in instances where the “S” component is missing, the remaining components are scaled by a factor of 10/9.

Other specifications tested separate dummy variables for CAMELS composite ratings of 3, 4, and 5. (A dummy variable for CAMELS 2 composite ratings was not statistically significant.) However, the single weighted CAMELS component measure performed comparably in out-of-sample tests and was chosen over the dummy variable specification for both the reduction in the number of variables, for its more continuous treatment of examination ratings and for its consistency with the current financial ratios method.

C. Considered Variables

1. Loan Loss Reserves

Loan loss reserves were tested in the development of the new model and were a positive predictor of failure across all specifications. Including reserves in the new model, however, would lead to higher deposit insurance assessments for banks with higher loan loss reserves, creating a disincentive for banks to build these reserves. Because loan loss reserves protect the FDIC in the event of failure, they were ultimately excluded from the new model. (Loan loss reserves were excluded from the downgrade probability model for the same reason.) The losses to forecasting accuracy were small.

2. Lagged moving averages

To capture the possibility that changes in variables (as opposed to point-in-time values of variables) are correlated with failure, the FDIC tested the model using lagged moving averages. In theory, these lagged moving averages could also capture the effect of variables that do not change frequently. However, lagged moving averages were not consistently significant across specifications.

3. Insignificant Variables

A number of variables were also tested but ultimately not included in the model because they did not remain statistically significant across specifications. These variables are listed in Appendix 1.2 to the Supplementary Information section of this notice.

D. Excluded Variables

1. Distance to Default

Distance to default measures, which compare the amount of loss absorbing capital against the volatility of the return on underlying assets, are commonly used in failure prediction models. These variables are generally constructed with market data. However, such measures are not available for most small banks.

2. Macroeconomic Variables

Macroeconomic variables were excluded for three primary reasons. First, the assessment rates proposed are (and the rates previously adopted by the FDIC's Board were) explicitly intended to reduce procyclicality; that is, to maintain a positive reserve ratio while keeping relatively constant assessment rates.

Second, macroeconomic factors would add considerable complexity to the model. Finally, macroeconomic factors are imprecise measures of economic conditions for small banks that often operate only locally.

See

75 FR 66272, 66273-66281, 66292 (Oct. 27, 2010).

3. Holding Company Affiliation

The FDIC does not believe it is appropriate to charge a small bank a higher assessment rate because it is not part of a multi-bank holding company; consequently, the new model does not include a measure indicating whether a bank is a part of a holding company.

4. Brokered Deposits

The FDIC ultimately chose the related measure of core deposits (see above).

5. Bank Size

The FDIC is disinclined to discriminate for deposit insurance assessment purposes based on the size of an established small bank. Assessing the smallest banks at higher rates because of their size would raise the costs of many banks that are the only bank in their community. Assessing the largest of the small banks at higher rates because of their size would impair their ability to compete with large banks, which are not charged higher rates based on their size.

III. Estimation Model

A. Shumway (2001)

The FDIC chose to estimate failure using a discrete-time hazard model with a constant hazard rate. Hazard models are designed to capture the duration of time until a particular event occurs (in this case, bank failure). The defining feature of a hazard model is that at every interval of time, a bank is exposed to some risk of failure that depends on certain observed measures. If the bank fails during a period, then it is not in the sample for later periods. If the bank survives, then it remains in the sample the following period and is exposed to a new risk of failure that depends on any changes in the bank risk variables. The FDIC used a discrete time assumption because of the regular reporting schedule for Call Report data, and the simplicity and transparency of estimation. A discrete time assumption implies that only the failure or survival of the bank is modeled for a given time period. This is in contrast to a continuous time model that also considers the exact failure time within that time period.

Shumway (2001) demonstrates that if each period's probability of failure (or default probability) follows a logistic function, then the discrete-time hazard model is equivalent to a multi-period logistic model. The logistic function relates a set of variables (in this case,

measures of bank risk) to a number between 0 and 1 (in this case, the probability of bank failure). It is nonlinear, so that the effect of a change in the values of bank risk variables on the probability of bank failure depends on the level of bank risk. A multi-period logistic model estimates the probability of failure for all observations across banks and time. However, relative to a pooled logistic model in which each bank-year observation is treated as an independent event, the standard errors of the coefficients of a discrete-time hazard model require an adjustment. The adjustment is required because of the serial dependence of the failure variable; a bank that is observed in any period necessarily has not failed in any previous period and any bank that fails necessarily drops out of the sample after failing.

A multi-period model was chosen over a single time period model. A single time period failure model requires the choice of the appropriate estimation time period. Therefore, it is unable to exploit data outside of the chosen time horizon and cannot be readily adapted to include new data. For example, a single time period model could not be used to capture bank failures in the 1990s, stability in the early 2000s, and the bank failures following the 2008 financial crisis. Furthermore, there is no systematic way to choose the right sample period for a static model.

The FDIC imposed a constant hazard rate on the model. A constant hazard rate implies that the age of the bank does not affect its likelihood of future failure. This is in contrast to a non-constant hazard rate that may be more appropriate for newer banks that do not yet have an established business model or management. However, new banks are excluded from the model. Because there is no relationship between the age of an established bank (one at least five years old) and failure, a constant hazard rate is more appropriate.

C. Time Horizon

Because deposit insurance assessments should ideally reflect risks posed by banking activity as they are assumed rather than when they are realized, a three year time horizon was chosen for both the estimation and forecasting periods. To obtain predictions for the three-year forecast, the FDIC considered one-year, two-year, and three-year time horizons in estimating the new model. In each case, the FDIC used only contemporaneous data to calculate three-year forecasts. That is, the FDIC alternatively used one-year, two-year, and three-year intervals in the estimation period (1984—2010) to forecast failures out-of-sample from January 1, 2011 through December 31, 2013 based on yearend 2010 data. The three-year interval tended to outperform the one- and two-year intervals for three-year out-of-sample forecasting.

D. In-Sample Estimation

The in-sample estimation time period was chosen to be 1985 through 2011, incorporating Call Report data through the end of 2011 and failures through the end of 2014.

To avoid having overlapping three-year look-ahead periods for a given regression, each regression uses data in which only every third year is included. One regression uses insured depository institutions' Call Report and TFR data for the end of 1985 and failures from 1986 through 1988; Call Report and TFR data for the end of 1988 and failures from 1989 through 1991; and so on, ending with Call Report data for the end of 2009 and failures from 2010 through 2012. (See Table 1.2A below.) The second regression uses insured depository institutions' Call Report and TFR data for the end of 1986 and failures from 1987 through 1989, and so on, ending with Call Report data for the end of 2010 and failures from 2011 through 2013. (See Table 1.2B below.) The third regression uses insured depository institutions' Call Report and TFR data for the end of 1987 and failures from 1988 through 1990, and so on, ending with Call Report data for the end of 2011 and failures from 2012 through 2014. (See Table 1.2C below.) Since there is no particular reason for favoring any one of these three regressions over another, the actual model estimates are constructed as an average of each of the three regression estimates for each parameter.

The regressions only include observations for institutions that are at least five years of age, since younger institutions will be subject to a different assessment methodology. Also, since the model will be applied to banks with under $10 billion in assets, larger banks are not included in the regressions.

The data used for estimation is winsorized (that is, extreme values in the data are reset to reduce the effect of outliers) at the 1st percentile and 99th percentile levels for each year. For example, if a variable for a bank has a value greater than the 99th percentile value for that year, then the value for that bank is set to the 99th percentile value before estimation is made.

The test statistics applied follow the analysis of Shumway (2001). In Shumway's formulation, the standard test statistics from a logistic regression used to assess statistical significance are divided by the average number of bank-years per bank; this adjustment corrects for the lack of independence between bank-year observations. That is, an adjustment is made to account for a bank no longer being observed after failure. In tables 1.2A, 1.2B, and 1.2C below, “WaldChiSq2” shows the adjusted χ-square statistic, and “ProbChiSq2” the associated probability value. (The lower the value of ProbChisSq2, the more statistically significant is the parameter estimate. Parameter estimates with a ProbChiSq2 below .05 are considered to be statistically significant at the .05 level.)

As reported in Tables 1.2A, 1.2B, and 1.2C, banks with a higher leverage ratio are less likely to fail within the next three years. Similarly, banks' earnings before taxes and their core deposits to assets ratios are negatively correlated with failure probability. In contrast, nonperforming loans and the other real estate owned to assets ratios are positively correlated with failure probability. Moreover, banks with a higher LMI, faster asset growth, and worse weighted CAMELS component ratings are more likely to fail within the next three years.

The estimated coefficients of the variables are statistically significant at the 5% level for all three regression sets except for the asset growth rate variable. The asset growth rate is statistically significant for two out of the three regressions.

Table 1.2A.—Regression With December 2009 as Last Data Point for Independent Variables

Variable description

Estimate

WaldChiSq2

ProbChiSq2

Intercept

−2.8919

17.3025

0.000032

Tier 1 Leverage Ratio (%)

−0.3522

82.6065

0.000000

Net Income before Taxes/Total Assets (%)

−0.1197

8.0705

0.004499

Loan Mix Index

0.0152

41.9399

0.000000

Core Deposits/Total Assets (%)

−0.0265

23.7705

0.000001

Nonperforming Loans and Leases/Gross Assets (%)

0.2597

53.1450

0.000000

Other Real Estate Owned/Gross Assets (%)

0.1498

10.8676

0.000979

Asset Growth

0.0161

8.1715

0.004255

Weighted Average of C, A, M, E, L and S Component Ratings

0.4888

20.4650

0.000006

Table 1.2B—Regression With December 2010 as Last Data Point for Independent Variables

Variable description

Estimate

WaldChiSq2

ProbChiSq2

Intercept

−1.8213

7.9746

0.004744

Tier 1 Leverage Ratio (%)

−0.3603

82.0847

0.000000

Net Income before Taxes/Total Assets (%)

−0.1585

12.7807

0.000350

Loan Mix Index

0.0210

106.2229

0.000000

Core Deposits/Total Assets (%)

−0.0398

54.8076

0.000000

Nonperforming Loans and Leases/Gross Assets (%)

0.2358

39.1907

0.000000

Other Real Estate Owned/Gross Assets (%)

0.1801

17.7846

0.000025

Asset Growth

0.0046

0.5448

0.460463

Weighted Average of C, A, M, E, L and S Component Ratings

0.3432

9.9098

0.001644

Table 1.2C—Regression With December 2011 as Last Data Point for Independent Variables

Variable Description

Estimate

WaldChiSq2

ProbChiSq2

Intercept

−2.1862

10.9481

0.000937

Tier 1 Leverage Ratio (%)

−0.3410

75.4433

0.000000

Net Income before Taxes/Total Assets (%)

−0.2354

31.0665

0.000000

Loan Mix Index

0.0157

43.3664

0.000000

Core Deposits/Total Assets (%)

−0.0429

59.4956

0.000000

Nonperforming Loans and Leases/Gross Assets (%)

0.2325

37.6910

0.000000

Other Real Estate Owned/Gross Assets (%)

0.1584

12.0705

0.000512

Asset Growth

0.0133

5.5076

0.018934

Weighted Average of C, A, M, E, L and S Component Ratings

0.5318

22.3623

0.000002

The parameter estimates applied for the assessments are the average of the estimates from the three regressions above. These average values are show in table 1.2D.

Table 1.2D—Average of the Parameter Estimates Over Three Regressions

Variable description

Estimate

Intercept

−2.2998

Tier 1 Leverage Ratio (%)

−0.3512

Net Income before Taxes/Total Assets (%)

−0.1712

Loan Mix Index

0.0173

Core Deposits/Total Assets (%)

−0.0364

Nonperforming Loans and Leases/Gross Assets (%)

0.2427

Other Real Estate Owned/Gross Assets (%)

0.1628

Asset Growth

0.0113

Weighted Average of C, A, M, E, L and S Component Ratings

0.4546

When the new model is used to determine assessment rates, the variables Asset Growth and Net Income before Taxes/Total Assets are each bounded as follows:

Asset Growth ≤ 190-25 ≤ Net Income before Taxes/Total Assets ≤ 3.

For example, if Asset Growth is greater than 190 (percent) then it is reset to 190 to determine assessment rates. After the parameters shown in table 1.2D were obtained, the values of these bounds were determined by performing an iterative series of backtests covering data from 1985 to 2011, with each iteration testing a different combination of bounds; the combination of bounds that resulted in the best rank correlation (Kendall's tau) between probability of failure and actual failure is the combination of bounds selected.

IV. Validation

A. Backtest Comparison of the Proposal to the Current RRPS System

Using initial base assessment rates,

the FDIC also compared the out-of-sample forecast accuracy of the proposal in this NPR, which is based on the new model, to the current small bank deposit insurance system's financial ratios method's assessment rankings.

Comparisons were made for projections as of the end of six different years, 2006 through 2011, and are shown graphically using cumulative accuracy profile (CAP) curves. A CAP curve is illustrated in Figure 1.1. Suppose that banks are ranked on a percentile basis according to a model's predicted probability of failure, with the ranking in descending order. Thus the banks with the highest predicted probability of failure would have a percentile rank near zero, while the banks with the

lowest predicted probability of failure would have a percentile rank near 100. In Figure 1.1, the horizontal axis represents this bank percentile rank. The vertical axis represents the cumulative percentage of actual failures. For example, the point marked by “X” indicates that the 30 percent of banks with the highest projected probability of failure included 50 percent of the banks that actually failed. In general, when comparing a CAP curve for alternative models, a model with a higher CAP curve (one with more area underneath it) would be the superior model.

For the out-of-sample backtests, the parameters applied are the average of the parameters from three separate regressions, as in the new model, except with more recent three-year periods omitted. Using Table 1.3 as an example, one regression uses data from the end of 1985 and failures from 1986 through 1988; data for the end of 1988 and failures from 1989 through 1991; and so on, ending with data for the end of 2003 and failures from 2004 through 2006. The second regression uses data from the end of 1987 and failures from 1988 through 1990, and so on, ending with data for the end of 2002 and failures from 2003 through 2005. The third regression uses data from the end of 1986 and failures from 1987 through 1989, and so on, ending with data for the end of 2001 and failures from 2002 through 2004.

EP13JY15.149

Figure 1.2 shows the CAP curve for a model (dotted line) compared with two limiting CAP curves. The “random” curve (single straight line) shows what the CAP would look like if the model prediction were purely random; for example, the 30 percent of banks with the highest failure projections would include 30 percent of actual failures. At the other extreme, the two solid straight lines show a CAP curve for a model that perfectly differentiates banks that fail from banks that do not in its projections; thus, for example, assuming that 20 percent of all banks actually failed, for the “perfect” model, the 20 percent of banks with the highest projected failure probability would identify 100 percent of failures.

The accuracy ratio can be derived from the CAP curve. For the model depicted by the curved line in Figure 1.2, the area between the curved line and the dotted straight line is a measure of the superiority of the model over the random benchmark. The area between the solid line and the dotted straight line is a measure of the superiority of a “perfect” model over the random benchmark. The ratio of these two areas is the accuracy ratio for the model depicted by the curved line. The value is normalized so that it is always less than or equal to 1. An accuracy ratio of 1 occurs in the case of a perfect model, and is 0 in the case of a model that does no better than random guessing. (For the illustrative example in Figure 1.2, the accuracy ratio of the model depicted by the curved line is .396.)

EP13JY15.150

To illustrate the application of CAP curves to the assessment system, Figure 1.3 shows a CAP curve for the current small bank deposit insurance system based on its risk ranking (as reflected in assessment rates) as of 2006 and on failures over the next three years (2007 through 2009). The horizontal axis coordinates for four points on this curve, “IV”, “III”, “II”, and “I Max”, corresponding to the percentage of small banks reported in Column (A) in Table 1.3 below, and the vertical axis coordinates for the points correspond to the percentage of failures contained within these percentages of small banks, as shown in column (B) in Table 1.3. For example, the point in Figure 1.3 marked “IV” is 0.06 (percentage of small banks in Risk Category IV) on the horizontal axis and 0.65 (percentage of actual failures among small banks in Risk Category IV) on the vertical axis. Similarly, all points to the left of the point marked “III” in Figure 1.3 are Risk Category III and IV rated small banks.

The banks along the horizontal axis corresponding to the horizontal axis coordinates between the points “II” and “I Max” represent Risk Category I small banks that are assessed at the maximum assessment rate for that category. The banks corresponding to the horizontal axis coordinates between the points “I Max” and “I Var” represent Risk Category I small banks that are differentially assessed between the maximum and minimum assessment rates for Risk Category I. (Point “I Var” is not included in Table 1.3.) Banks to the right of the horizontal axis coordinate for the point “I Var” represent Risk Category I small banks that were assessed at the minimum assessment rate.

Table 1.3—Comparisons of Out-of-Sample Projection of New Model To the Small Bank Deposit Insurance Assessment System's Rankings for 2006 *

(A)

Percentage of Small Banks

in Risk