+0.20%

ABSTRACT

THE RELATIVE PREDICTIVE CAPACITY OF TWO BANK
EARNINGS MEASURES: AN EMPIRICAL EVALUATION

BY

Hugh A. Hoyt

A controversy exists between bankers and accountants over which
of two earnings measures should be presented as the final reported earnings
figure on a bank's income statement. Bankers favor a figure, described as
Net Operating Earnings, that excludes any provision for loan losses or any
security gains and losses. Accountants favor a figure, described as Net
Income, that includes some provision for loan losses and all security gains
and losses.

An investigation into the nature and causes of the controversy
revealed that although there would be no controversy if it were not for
the existence of the two disputed items (i.e., the provision for loan
losses and security gains and losses), it remains unresolved for reasons
unrelated to the arguments presented by both sides in support of their
treatment of the two items. The investigation disclosed that the primary
reason that the controversy remains unresolved is the belief, held by
both bankers and accountants, that investors have single-figure fixation.
According to this belief investors focus almost entirely on the final
reported earnings figure in evaluating the current and potential performance
of a firm while other financial data included in the published financial
statements are largely ignored. Granted the existence of the belief in
single-figure fixation, it was found that a secondary reason for the
controversy is a disagreement between the opposing sides as to which of

the two earnings measures is the most relevant to users of publicly

Hugh A. Hoyt
reported bank earnings.

Initial efforts to obtain empirical evidence that would aid in
resolving the controversy were focused on attempts to perform a study to
determine the validity of the belief in single-figure fixation. Such a
study was not found to be feasible. Subsequent efforts were focused on
performing a study to determine which of the two earnings measures was
the more relevant. Such a study was found to be feasible once "the
capacity to predict future earnings" was specified as a surrogate for
relevance.

While the general purpose of the ”predictive" study was to obtain
empirical evidence that would aid in resolving the controversy, the
specific purpose was to determine the relative capacity of the two earnings
measures to predict future all-inclusive earnings. For, it was argued, if
one earnings measure were found to have a predictive capacity superior to
the other there would be new evidence to support the presentation of the
measure with the superior predictive capacity as the final reported
earnings figure on the grounds that it was more relevant. All-inclusive
earnings was specified as the relevant forecast objective on the grounds
that it provides the best measure of a bank's future overall performance.

Net Operating Earnings and Net Income per share were computed for
each year of the twelve-year period ending in 1968 for each of 26 large
commercial banks on the basis of data contained in the published annual
reports of these banks. These per-share figures served as the sole input(s)
to six linear forecast models which were used to obtain estimates of all-
inclusive earnings per share one year in the future. The forecast models
employed prior years' earnings figures for periods ranging from one to

five years. The difference between forecast earnings and actual all-

Hugh A. Hoyt
inclusive earnings served as the measure of forecast error. Aggregations'
of the annual forecast errors associated with each earnings measure over
forecast-error-periods ranging from one to seven years served as the
basis for determining the relative predictive capacity of the two earnings
measures. A

An analysis of the results of this study disclosed that Net
Operating Earnings (the bankers' earnings measure) showed a superior
predictive capacity for forecast-error-periods of two years or less in
length whereas Net Income (the accountants' earnings measure) showed a
superior predictive capacity for forecast-error-periods of four years or
more in length with neither measure showing a superiority for the three-
year forecast-error-period. Since both earnings measures were found to
have superior predictive capacities over one or more forecast-error-
periods considered to be relevant to users of a final reported earnings
figure, neither measure could be unequivocably designated as being more
relevant than the other. Consequently it was concluded that no new
evidence was found to support the presentation of either Net Operating
Earnings or Net Income as the final reported earnings ligure on a bank's

income statement.

THE RELATIVE PREDICTIVE CAPACITY OF
TWO BANK EARNINGS MEASURES:
AN EMPIRICAL EVALUATION

By
J“ ‘1

Hugh A: Hoyt

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Accounting and Financial Administration

1970

Copyright
by
Hugh A. Hoyt

1970

~——-\
'\ ,l

ACKNOWLEDGMENTS

Many people were responsible in one way or another for my
successfully completing the Ph.D. program. The following represents a
weak attempt to thank those to whom I will be forever indebted.

First, appreciation is expressed to the members of my dissertation
committee, namely Professors James Don Edwards (chairman), Alvin Arens,
and Leo Erickson for the time and effort spent in reviewing and construc-
tively criticizing my thesis. It is a far better product for their
efforts. Special thanks is also due Professor Edwards as chairman of the
department for seeing to it that financial aid was available throughout
my graduate career.

I am also indebted to several faculty members and fellow doctoral
candidates for providing needed encouragement and helpful suggestions
during the formative and conclusive stages of this research effort. Those
deserving special credit include Robert May, James Parker, Michael Crooch,
Gerald St. Amand, and Elba Baskin, William Morris, Michael Lawrence and
Professor Geraldine Dominiak.

Without financial support completion of the Ph.D. program would
not have been feasible. Therefore, my appreciation is expressed to
Ernst & Ernst, the Earhart Foundation, the American Accounting Association,
and the Michigan Accountancy Foundation for financing my graduate career.

There is one other group of individuals who, while they did not
play a direct role in helping me complete the Ph.D. program, were
instrumental in providing me with the internal and external resources

ii

without which I would have been unable to begin, let alone complete, an
undertaking of this nature. Specifically included in this group are
Margaret Miller, Mary and Frank McHenney, Catherine and Harry Miller,
Albert Neukom, Professors Russell Ogden and Oscar Collins, and, of course,
my parents and brothers and sisters.

Finally, I express my gratitude to my wife, Joan, who, in addition
to doing all of the typing on my dissertation, provided constant support
throughout the ups and downs of the Ph.D. program and to my sons, Mark

and David. They were the sine qua non of my efforts.

iii

TABLE OF CONTENTS

Page
ACKNOWLEDGMENTS. . . . . . . . . . . . . . . . . . . .'. . . . . ii
LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . vi

LIST OF FIGURES. . . . . . . . . . . . . . . . . . . . . . . . . viii

Chapter

I. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . l
1.1 Purpose of Research . . 1
1.2 The Controversy . . . . . . . . . . . 1
1.3 Importance of the Controversy . . . . . . . . . 8
1.4 Current Status of the Controversy . . . . . . . 12
1.5 Scope and Methodology . . . . . . . . . . . . . 15
II. THE ITEMS INVOLVED IN THE CONTROVERSY. . . . . . . . . . 17
2.1 Introduction. . . . . . . . . . . . . . . . . 17
2. 2 Security Gains and Losses . . . . . . . . . . . 17
2.2.1 The Bankers' Position . . . . . . . . . 18
2.2 2 The Accountants' Position . . . . . . . 20

2.2.3 Proposed Alternative Accounting
Treatment. . . . . . . . . . . . . . 22
2.3 The Provision for Loan Losses . . . . . . . . . 24
2.3.1 The Bankers' Position . . . . . . . . . 25
2.3.2 The Accountants' Position . . . . . . . 31
2.4 Summary and Conclusion. . . . . . . . . . . . . 33

III. FORMULATION OF AN EMPIRICAL STUDY DESIGNED TO AID 1N

RESOLVING THE CONTROVERSY . . . . . . . . . . . . . . 34
3.1 Introduction. . . . . . . . . . . . 34
3.2 The Basic Reasons for the Controversy . . . . . 34
3.2.1 The Primary Reason. . . . . . . . . . . 34
3.2.2 The Secondary Reason. . . . . . . . . . 40
3.2.3 Summary . . . . . . . . . . . 42
3.3 Proposed Means of Resolving the Controversy . . 43
3.3.1 Test of Single-Figure Fixation. . . . . 43
3.3.2 Test of Relevancy . . . . . . . . . . . 45
3.4 Summary . . . . . . . . . . . . . . . . . . . . 48

iv

IV.

v.

VI.

APPENDIX

A.

B.

METHODOLOGY . . . . . . . . . . . . . . . . . . . . . .

4.1 Introduction . . . . . . . . . . . .
4.2 Specification of the Forecast Objective. . .
4.3 Specification of Forecast Models .
4.4 Sample Selection and Data Source . . . . . . . .
4.5 Calculation of Earnings Per Share. . . . . . . .
4.5.1 Computation of Net Operating Earnings.
4.5.2 Computation of Net Income. . . . . . . .
4.5.3 Computation of a Normal Provision for
Loan Losses . . . . . . . . . . . . .
4.5.4 Adjustment for Stock Dividends and
Splits. . . . . . . . . . . . . .
4.6 Specification of a Forecast Error Measure.
4.7 Summary. . . . . . . . . . . . . . . . . . . .
EMPIRICAL FINDINGS. . . . . . . . . . . . . . . . . .
5.1 Introduction . . . . . . . . . . . .
5.2 Identification of Evaluation Criteria.
5.3 Results Analyzed According to the Short- Run
Criterion . . . . . . . . . .
5.3.1 Aggregation Procedures .
5.3.2 Observations . . . . . .
5.4 Results Analyzed According to the Long-Run
Criterion . . . . . . . . . . . . . . . .
5.4.1 Aggregation Procedures . . . . . . . .
5.4.2 Observations . . . . .
5.5 A Plausible Explanation of One Trend in the
Forecast Errors . . . .
5.6 Summary. . . . . . . . . . . . . . . . . . . . .

SUMMARY AND CONCLUSIONS . . . . . . . . . . . . . . .

Introduction . . . . . . . . . . .
Review of Research Purpose and Design. . .
Limitations. . . . . . . . . .

Conclusions and Implications . . . . . . . . . .
6.4.1 Major Conclusion . . . . . . . . . . . .
6.4.2 Minor Conclusion . . . . . . . . .

Suggestions for Future Research. . . . .

Banks Included In The Study . . . . . . . . . . . . .

Supplementary Tables To Chapter V . . . . . . . . . . .

SELECTED BIBLIOGRAPHY . . . . . .

Page

50

50
50
56
59
60
61
62

.63

68
68
69

70

7O
70

71
71
78

80
80
84

90
92

97

97
97
99
100
100
105
105

107

108

115

Table

II.

III.

IV.

VI.

VII.

VIII.

IX.

LIST OF TABLES

Page

Selected Data on Federal Reserve Member Banks,
1957-68 . . . . . . . . . . . . . . . . . . . . . . . 9

Selected Information About the Forecast
Models. I O O O O O O O O O O O O O O O 0 O c O 0 O O 73

Mean Absolute Percentage Forecast Error:
Forecast-Error-Period Equal to One Year . . . . . . . 75

Mean Deviation of Absolute Percentage Forecast
Errors: Forecast-Error-Period Equal to
One Year. . . . . . . . . . . . . . . . . . . . . . . 76

Number of Banks for Which Each Earnings Measure
Produced the Lowest Absolute Forecast Error:
Forecast-Error-Period Equal to One Year . . . . . . . 77

Number of Time Periods Contained in the Total
Period Covered by Each Forecast Model for Each
Multi-Year Forecast-Error-Period. . . . . . . . . . . 83

Earnings Measure with the Lowest Total-Period
Mean Percentage Forecast Error and Lowest
Total-Period Average Mean Deviation for
Selected Forecast-Error-Periods . . . . . . . . . . . 93

Earnings Measure with Lowest Common-Period Mean
Percentage Forecast Error and Lowest Common-
Period Average Mean Deviation for Selected
Forecast-Error-Periods. . . . . . . . . . . . . . . . 94

Earnings Measure with Lowest Total-Period Mean
Percentage Forecast Error and Lowest Total-
Period Average Mean Deviation for a Majority
of Forecast-Error-Periods Covered by Each
Forecast Model for Selected Forecast-Error-
Periods . . . . . . . . . . . . . . . . . . . . . . . 95

Mean Absolute Percentage Forecast Error and
Mean Deviation of Absolute Percentage Forecast
Errors: Forecast-Error-Period Equal to Two
Years . . . . . . . . . . . . . . . . . . . . . . . . 109

vi

Page

XI. Mean Absolute Percentage Forecast Error and
Mean Deviation of Absolute Percentage Forecast
Errors: Forecast-Error-Period Equal to Three
Years . . . . . . . . . . . . . . . . . . . . . . . . 110

XII. Mean Absolute Percentage Forecast Error and
Mean Deviation of Absolute Percentage Forecast
Errors: Forecast-Error-Period Equal to Four
Years . . . . . . . . . . . . . . . . . . . . . . . . 111

XIII. Mean Absolute Percentage Forecast Error and
Mean Deviation of Absolute Percentage Forecast
Errors: Forecast-Error-Period Equal to Five
Years . . . . . . . . . . . . . . . . . . . . . . . . 112

XIV. Mean Absolute Percentage Forecast Error and
Mean Deviation of Absolute Percentage Forecast
Errors: Forecast-Error-Period Equal to Six
Years . . . . . . . . . . . . . . . . . . . . . . . . 113

XV. Mean Absolute Percentage Forecast Error and
Mean Deviation of Absolute Percentage Forecast
Errors: Forecast-Error-Period Equal to Seven
Years . . . . . . . . . . . . . . . . . . . . . . . . 114

vii

Figure

10.

LIST OF FIGURES

Income Statement Format Advocated By Bankers.

Income Statement Format Advocated by Accountants.

Graphic Presentation of Two Earnings Patterns
Graphic Illustration of Bias and Efficiency .

Total-Period Mean Absolute Percentage Forecast
Errors for Selected Forecast-Error-Periods:
MOdel 81- o o o o o o o a o o o

Total-Period Mean Absolute Percentage Forecast
Errors for Selected Forecast-Error-Periods:
Model 82. . . . . . . . . . . . . . . . .

Total-Period Mean Absolute Percentage Forecast
Errors for Selected Forecast-Error-Periods:
MOdel S30 0 o o o o o o o o o o o o o o

Total-Period Mean Absolute Percentage Forecast
Errors for Selected Forecast-Error-Periods:
Model R1. . . . . . . . . . . . . .

Total-Period Mean Absolute Percentage Forecast
Errors for Selected Forecast-Error-Periods:
Model R2. . . . . . . . . . . .

Total-Period Mean Absolute Percentage Forecast
Errors for Selected Forecast-Error-Periods:
MOdE]. R3 0 o o o o o o o o o o o o o o 0

viii

qu5e

38

54

101

101

102

102

103

l()3

CHAPTER I

INTRODUCTION

1.1 Purpose of Research
The purpose of this research is to obtain empirical evidence that
will aid in resolving the controversy that exists between bankers and

accountants over the form and content of bank income statements.

1.2 The Controversy

The controversy between bankers and accountants essentially
involves a disagreement over which of two earnings measures should be
presented as the final reported earnings figure on a bank's income state-
ment. Bankers favor a figure, described as Net Operating Earnings, that
excludes any provision for loan losses or any security gains and losses.
Accountants favor a figure, described as Net Income, that includes some
provision for loan losses and all security gains and losses. lankers
contend that the presentation of Net Income as a final reported earnings
figure would give a distorted indication of a bank's "earning power"
since such a figure includes two items (i.e., a provision for loan losses
and security gains and losses) that they consider to be of a nonoperating
nature. Accountants argue that security gains and losses and loan losses
are a normal part of a bank's operations. Therefore, they contend that a
failure to include security gains and losses and some provision for loan
losses in the determination of a final reported earnings figure is mis-
leading. Edward T. Shipley, former chairman of the Accounting Principles
Committee of the American Bankers Association and past president of The

1

Association for Bank Audit, Control, and Operations (NABAC) has summarized
the controversy in the following manner:

Among the subjects of continuing discussion in bank finan-
cial reporting is the possible adoption of the so-called "all-
inclusive" income Statement. This concept means simply that
the income statement should reflect all income and expense
items, including even extraordinary, nonrecurring items; the
alternative approach reflects the idea that the income state-
ment should reveal current operating performance and that
extraordinary items, unrelated to operations for the period,
should by-pass the income statement and be directly credited
to or charged against the capital accounts. The SEC and the
AICPA have tended to favor the all-inclusive income statement,
although extraordinary items are expected to be presented
"below the line," as additions to or deductions from net
operating income in arriving at net income. In banking, the
debate has centered on two important items, the creation of
a bad debt reserve and the treatment of gains or losses on
securities transactions. There are substantial reasons why
the generally preferred all-inclusive income statement would
be detrimental to the best interests of banks and investors
in bank securities.1

Figures 1 and 2 illustrate the income statement format advocated
by bankers and accountants, reSpectively.2 The underlined captions in
the two examples represent the major areas of difference. Identical
amounts are used in both of the examples for comparable revenue and
expense items with the exception of the provision for loan losses. The
major differences in the two illustrated income statements are described
below. All references to amounts pertain to the figures listed in the
"current year" column of Figures 1 and 2.

The first difference in the two statements occurs in the treat-

ment of a provision for loan losses. Figure 2 (the accountants' state-

 

1Edward T. Shipley, "Bank Accounting Principles: A Progress
Report," Law and Contemporarerroblems, Vol. 32: No. 1 (Winter, 1967),
p. 144.

2The format for Figure 1 was synthesized from an analysis of the
1968 annual reports of the fifty largest commercial banks. The format for
Figure 2 was extracted, with minor modifications, from Audits of Banks:
Supplement (Committee on Bank Accounting and Auditing of the American

Institute of CPA'S, [New York: AICPA, Inc., 1969] , p. 5.).

3

STATEMENT OF INCOME
For the Years Ended December 31, 19__ and 19__

 

 

 

 

 

 

 

 

 

 

 

 

Current Preceding
Year Year
Operating income:
Interest on loans $1,000,000 $ 900,000
Interest and dividends on:
U.S. Government securities 180,000 200,000
State and municipal securities 60,000 60,000
Other securities 10,000 8,000
Trust Department income 100,000 90,000
Service charges on deposit accounts 50,000 45,000
Other operating income 60,000 55,000
Total $1,460,000 $1,358,000
Operating expenses:
Salaries $ 220,000 $ 200,000
Other employee benefits 15,000 15,000
Interest 480,000 460,000
Net occupancy expense of bank premises 55,000 52,000
Other operating expenses 56,000 65,000
Total $ 826,000 $ 792,000
Operating earnings before income taxes $ 634,000 $ 566,000
Less applicable income taxes 277,000 244,000
Net operating_earning§ $ 351,000 § 322,000
Net opeggting earnings per share $3.57 $3.22

 

Nonoperatingiadditions and (deductions):
Securities gains (losses), less related
income tax effect of $50,000 in 19__
and $100,000 in 19__ $ 150,000 $ (100,000)
Provision for loan losses, less income
tax reduction of $19,000 in 19___

 

 

 

 

 

 

 

 

and $21,000 in 19__ (19,000) (21,000)
Other, less income tax reduction of
$30,000 $28,000) ---
Total § 103,000 $ (121,000)
Transferred to undivided_profits $ 460,000 $, 201,000
FIGURE 1

INCOME STATEMENT FORMAT ADVOCATED BY BANKERS

4

STATEMENT OF INCOME

For the Years Ended December 31, 19__

Operating income:
Interest on loans
Interest and dividends on:
U.S. Government securities
State and municipal securities
Other securities
Trust Department income
Service charges on deposit accounts
Other operating income
Total

Operating expenses:
Salaries
Other employee benefits
Interest
Net occupancy expense of bank premises
Provision for loan losses
Other operating expenses
Total

Income before income taxes and securities
gainsg(losses)

Less applicable income taxes

Income before securities gains (losses)

Securities gains (losses), less related
income tax effect of $50,000 in 19__
and $100,000 in 19__.

Income before extraordinary item
(Loss) on sale of branch bank
building, less related reduction
in income tax of $30,000

Net income

 

 

Earnings data per share:
Income before extraordinary,item
Exgggordinary item,,less related
reduction in income tax
Net Income

 

Current
Year

 

$1,000,000

180,000
60,000
10,000

100,000
50,000
60,000

$1,460,000

 

$ 220,000
15,000
480,000
55,000
15,000

M

§ 841,000

$ 619,000

269,000

350,000

150 000

A

500,000

'(28,000)
g 472,000

$5.00

(.28)
$4.72

and 19__

Preceding
Year

 

$ 900,000

200,000
60,000
8,000
90,000
45,000
55,000
$1,358,000

 

$ 200,000
15,000
460,000
52,000
15,000
65,000

 

$ 807,000

$ 551,000
236,000

 

315,000

(100,000)
215,000

 

 

215 000

 

$2.15

$2.15

 

FIGURE 2

INCOME STATEMENT FORMAT ADVOCATED BY ACCOUNTANTS

ment) includes a provision for loan losses of $15,000 in ”Operating
expenses" whereas Figure 1 (the bankers' statement) contains no such
item. The $15,000 figure represents what accountants would call a
"normal" provision for loan losses in that it excludes that portion of
the total tax-deductible provision considered to be of a contingency
nature by them. Further discussion of the concepts of a "normal"
provision and a contingency portion of the total tax-deductible provision
is deferred until Chapter II. The inclusion of a normal provision for
loan losses, adjusted for the related income-tax effect ($15,000 - $8,000 =
$7,000), accounts for the only difference between the figure reported as
"Net operating earnings" ($357,000) by bankers and the figure reported as
"Income before securities gains (losses)" ($350,000) by accountants.

Figure 1 does contain a provision for loan losses. However, the
amount is shown as a nonoperating deduction from "Net operating earnings"
and represents the total tax-deductible provision of $38,000 allowed the
bank. Since the "Nonoperating additions and (deductions)" are shown net-
of-tax, the amount of the provision for loan losses shown in Figure l is
$19,000 ($38,000 - $19,000). The $12,000 difference between the provision
for loan losses shown in Figure 1 ($19,000) and the net-of-tax provision
deducted in Figure 2 ($7,000, as derived above) represents the contingency
portion of the total tax-deductible provision. No such amount is contained
anywhere in Figure 2 since the contingency portion of the total tax-deduct-
ible provision for loan losses is considered by accountants to be a
segregation of retained earnings. The $12,000 difference between the
two provisions for loan losses accounts for the difference between the
amount reported by bankers as "Transferred to undivided profits" ($460,000)

and the amount reported by accountants as "Net income" ($472,000).

The next noticeable differences in the two statements pertains
to the location and identification of security gains or losses and
extraordinary items. Figure 1 shows securities gains of $150,000 along
with the provision for loan losses and an "other" deduction of $28,000
as "below the line" adjustments to "Net operating earnings". Figure 2
treats security gains or losses as an adjustment to "Income before
securities gains (losses)" in arriving at ”Income before extraordinary
itemV. 'In the absence of any extraordinary item, security gains or losses
would represent the final adjustment in arriving at "Net income". Figure 2
describes in greater detail what Figure 1 shows as an "other" deduction
and treats it as an extraordinary item in arriving at "Net income”.
Incidentally, little attention has been focused on the treatment of
extraordinary items in the controversy, presumably because such items
do not loom large in relation to security gains and losses and the
provision for loan losses.

The final difference, of course, concerns the designation of a
final reported earnings figure. Figure 1 shows "Net operating earnings"
both as a gross and earnings-per-share figure to illustrate the bankers'
preference for it as the final reported earnings figure even though it is
not the final figure on the income statement. Figure 2 shows "Net income"
as the final figure on the income statement in illustrating the accountants'
preference for it as the final reported earnings figure. Although
Figures 1 and 2 are merely examples, they do serve to illustrate that
there can be a considerable difference between the final reported earnings
figures advocated by both parties involved in the controversy.

It is worth noting that although a controversy still exists over

the appropriate form and content of bank income statements, significant

advances have been made in the overall financial reporting practices of
banks in the past ten years.

Prior to the Securities Acts Amendments of 1964, not a
great deal was done to provide bank shareholders and the
investment community with needed information. Some larger
banks had begun to publish income statements, but most bgnks
limited their disclosure of their financial affairs to the
dissemiggtion of the reports of condition [i.e., balance
sheets] required by the various supervisory authorities.

These reports, in forms prescribed by the Comptroller of

the Currency, the Federal Reserve Board, the Federal Deposit
Insurance Corporation (FDIC), or state supervisory agencies,
were largely identical in format. They were filed quarterly
with the agencies but were typically published, as required,
twice a year. The similarity in the statements' appearance
belied the lack of uniformity in the accounting principles
employed in their preparation. Statements of one bank could
not be safely compared with those of another, and the
prescribed forms made no provision for footnotes to explicate
the figures and the accounting techniques followed in arriving
at them. Statements of fipgncigl,conditign did not revggl
earnings eggept,§§_they might perhgps be deducible by
cggparison of surplus gpd undivided pgofit figpres with the
same acgpunts disclosed ingggrlier statements, comparative
figures, showing changes from one statement to the next,,were
neyer required and seldom provided. While most supervisory
authorities did require the filing of annual reports of
earnings and dividends, these were unavailable to the public.
(Emphasis added.)

3

The Securities Acts Amendments of 1964 vested the regulatory
authority for bank financial reporting with the respective federal agency
responsible for supervising the bank generally. Thus, the Comptroller
of the Currency became responsible for national banks, the Federal
Reserve Board for state member banks, and the FDIC for insured state
nonmember banks.

Subsequently the Federal Reserve joined forces with the FDIC to

produce substantially identical regulations in 1965.4 The Federal

 

3Shipley, p. 132.

4Code of Federal Regplations: Title 12 - ngks and Banking,
sections 206.1-.71 (January 1, 1970 revision).

Reserve's is the version most often referred to and is identified as
Regulation F. The Comptroller of the Currency did not issue comparable
regulations until 1967.5 Among other things, these regulations stipulated
that banks begin providing their stockholders with annual financial
statements which included an income statement and disclosed amounts
relating to security gains and losses and the provision for loan losses.
The income statement illustrated in Figure 1 is indicative of the effect
that these new reporting requirements had on the published income state-
ments of most large, publicly-held banks by the end of 1968.

Thus, no longer is the argument between bankers and accountants
so much over disclosure as it is over format and terminology. Both sides
are willing to include the total amount of security gains and losses and
some provision for loan losses in the income statement; they disagree as
to the location of these items in the income statement and the designation

of a final reported earnings figure.

1.3 Importance of the Controversy

Two types of evidence are presented to indicate the importance of
the controversy: (1) the size of the items involved in the controversy
and (2) the publicity generated by the controversy.

Table I contains data relating to Net Operating Earnings, security
gains and losses, the provision for loan losses, and actual loan losses of
Federal Reserve member banks for the twelve-year period from 1957-68. The
data reveal that security gains and losses and/or the provision for loan
losses have been material in relation to Net Operating Earnings individually

or collectively for each of the twelve years. Expressed as a percentage

 

51bid., sections 10.1-.4 and l8.7-.7.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.Lmq .a .momH .zm: .cHuoHLam u>nommm ”append .zmoaq .mEOUCH xcmm nonsmzz “moon oouaow:
\o.N~ Nm.m~ Nm.qm gm.m No. Hm.m xw.m x~.oL Nn.w~ aq.q~ m.wH ua.mm mmcwcumm wcHuauomo Lo ucoonom
amq W ammmvw mHo.Ew mam kw Om w. mmN w. aim w «on w. ems w soo.nw cm“ W msaqu lawman a 00L
cofimw>oum ecu can ”mew UV ammo;
xuwuauwm ou one mmcwcumu mcfiucuoao
uoz CH Ammmococwv wwmmuooq uoz
em.N so.a Am.n mm.m no.m em.s an.o aq.m so.“ wo.m ~.m .q.o mwcacumu weasacmao co Susanna
mm w as w mm w and w “ma 1w NML w 50H w mm“ w mmm w mmm m “mm m mmm w mommog coon umz Hmauu<
as.o so.m am.a No.0n an.~n as.mE em.~n Ne.sn e~.m~ an.sa .o.oH ¢~.ss mwcacumm mcaumsoao Lo “assume
cam (W oom WI Hmm w nmm w Hmm w nosllw mom m Now w coo w. mom 1w. woo w 0mm w Lmuow
“mg mnd «ma mag qmm nmm con man moo Mom Ham “mm o>mewx an smooLoCH aux
mm m Ne w mm m and m an m mmfi w hog m mm“ m mmw m Nmm m ”mm m mmm w moauo>oomm mo box .muwonomumzu
mommoq smog new coamw>oum
im.m em.am No.k~ w~.~ NN.HH No.4 em.~ sm.n em. »H.oH gm. no.w mwcactam measmtmao no “nausea
new a lammmvw Mme w ammv w Asmmqw Amnaqw fines w Ne 1w. cm w seq m mm w mes w Assauev mommoa auatsomm
osm.~w cam.~w. nmm.mw mk~.mw msa.mw ~444mw mm~.mw sac.mw mm04mw Ona.sm mmm.sm Hooqu mmcactmm weaumrmao s97
5mm“ mnma mama coma Emma Noaa mood «cog mood coma mood mama
Amoxma usoucH awuovom ouowun mouzwww Admv
Aucoafiaws ca nuances uuafioov
amounmma .mxz<m mmmZmz m>mmmmd A<mmnmm zo «Hro omwomqmm

H mqm<H

10

of Net Operating Earnings, the combined items ranged from a low of .9%
to a high of 34.5% with a mean percentage of 16.0. Individually, security
gains and losses ranged from .5% to 27.0% with a mean percentage of 8.3;
the provision for loan losses ranged from 7.5% to 18.2% with a mean
percentage of 12.5.

That the controversy has received public attention is shown by
the following quotations taken from media with a national audience. The

January 8, 1969, Wall Street Journal carried the following front page

 

caption:

Frustrated CPA's
Accounting Body Fails
In Attempt to Change
Some Firms' Reporting

Three Proposals Withdrawn
Or Weakened Could Have
Reduced '68 Profit Figures

A Dare From Banking Group
The text of the article included the following statements:

It [the American Institute of Certified Public Accountants]
has delayed making a final ruling that banks' reported profits
should reflect losses on uncollectible loans and gains and losses
on sales of securities; it is now negotiating a compromise with
the American Bankers Association.

The American Bankers Association went beyond criticizing the
AICPA ruling that banks should reflect bad-loan losses and
securities profits or losses in a figure clearly labeled as
net profit. In effect, it dgred the AICPA to make the ruling
stick. (Emphasis added.)

Banks now report figures on bad-debt losses and on
securities gains and losses but not as part of a figure labeled
net profit; the only figure labeled profit on most reports is
operating profit from interest on loans and investments and
from fees for banking service. The ABA said that regulatory
agencies permit banks to report this way and advised member
banks to continue doing so "even though such action could
result in qualified opinions from CPA's (that is, opinions
containing exceptions to profit figures on the grounds that
the AICPA ruling wasn‘t followed).

11

The New York Times referred to the controversy in the following manner:

 

[L]ast week the accounting profession - through its senior
rule-making body, the Accounting Principles Board - called on
bankers to stop overstating their profits and start issuing
their annual reports in accordance with "generally accepted
accounting principles".

But whether the accountants' demand, which was part of a
long-term drive to bring bank reporting in line with that of
other industries, will do much to resolve the muddle over
bank accountigg is still an open question. (Emphasis added.)

The fact is that accountants themselves - let alone bankers
and bank regulatory agencies - are far from agreed as to what
these "generally accepted principles" ought to be in certain
key areas.

 

Even Time magazine devoted some attention to the dispute:

After a protracted wrangle with bankers, the board
[Accounting Principles Board] last month demanded that banks
include in their reported profits the losses on collectible
loans as well as the gains or losses on securities transactions.

Even though great strides have been taken to improve bank finan-
cial reporting practices in the past few years, the above disclosure of
the amounts involved in and the attention given to the controversy
indicates that there is a need to find a way to resolve the dispute.

The aforementioned Mr. Shipley suggested the need for continued research
efforts in order to resolve the controversy in his progress report on
bank accounting principles.

Regulation F did not purport to resolve the debate over the
proper function of the income statement. It compromised on the
question of presentation, showing first "net operating earnings"
and ultimately, after "nonoperating" items, a figure labelled
"transferred to undivided profits"; no "net income" figure is
shown. More important, Regulation F does not satisfactorily
resolve the problems raised by bad debt reserves and securities

transactions, assigning both to "below the line" status even
though each may have elements of current expense or income.

 

6H. Erich Heinemann, "Banks and Profits: Whether Accountants'
Plea Will Help Resolve Reporting Muddle is Uncertain," New York Times,
March 10, 1969, p. 63.

7"Cooking the Books to Fatten Profits," Time, April 11, 1969,
p. 96.

12

Both issues are thus intertwined with the functional concept
of the income statement.8

He concludes that Regulation F represents a step toward industry-wide
accounting uniformity but that:

A considerable controversy continues as the AICPA seeks to

establish its "generally accepted accounting principles for

banks" in certain areas where the accountants' approach

differs from the philosophy of a large group of the major

banks throughout the country. It would appear, therefore,

that the immediate objective is to obtain uniform and

comparable reporting on the part of all banks and subsequently

to reconcile such reporting practices with the proposals of

the financial analysts and the CPAs. Admittedly this will

require time, but, with the increasing study currently being

devoted to the subject, we can look for continuing progress

in this direction.9
1.4 Current Status of the Controversy

Although the disagreement between bankers and accountants over

income reporting requirements has existed for over a decade, the dispute
did not begin to reach its current proportions until the AICPA issued
Accounting Principles Board Opinion No. 9 on Repgrting the Results of
Operations in December, 1966.10 In essence, this opinion, had it been
applicable to banks, would have required banks whose financial statements
were audited by CPA's to include gains or losses on sales of securities
and some provision for loan losses in the determination of a final
reported earnings figure to be identified on the income statement as Net

Income. Failure to follow this procedure would have resulted in the

expression of a qualified opinion in the auditor's report.

 

8Ship1ey, pp. 144-145.

9Ibid, p. 148.

10Accounting Principles Board of the American Institute of Certified
Public Accountants, Opinions of the Accounting_Principles Board - No. 9:
Repggging the Results of Operations (New York: AICPA, Inc., 1967). Here-
after referred to as APB 9.

13

However, because of the disagreement between bankers and account-
ants, the applicability of APB 9 to banks was deferred pending further

11 After two years

study and recommendation by a committee of the AICPA.
of study and discussions with representatives of the commercial banking
industry the AICPA issued, in March, 1969, Accounting Principles Board
Opinion No. 1312 which specifically brought commercial banks under the
requirements of APB 9.

It was the proposed and ultimate issuance of APB 13 which prompted
the reactions disclosed in the quotations presented in section 1.3 above.
The threat by bankers to ignore APB 13, as alluded to in those quotations
was not an idle one. Until July 8, 1969, there was no indication that
the vast majority of banks would be affected by APB 13. First, large,
publicly-held banks, unlike their corporate industrial counterparts, are
not legally required to have their financial statements audited by CPA's.
Until 1964 banks were excluded from the Securities Acts of 1933 and 1934.
The Securities Acts Amendments of 1964 extended the prior acts to cover
large, publicly-held banks. However, instead of placing these banks under
the jurisdiction of the Securities and Exchange Commission, which requires
audited financial statements, the amendment left them responsible to their
designated regulatory authority. At the present time, neither the Federal
Reserve, Federal Deposit Insurance Corporation, nor the Comptroller of
the Currency require CPA audited financial statements.

Second, since most large publicly-held bank stocks are traded

over-the-counter, there is no influential body, such as the New York Stock

 

111b1d., p. 108.

12Accounting Principles Board of the American Institute of Certi-
fied Public Accountants, Opinions of the Accounting Principles Bogrd -
No. 13: Amending_Paragraph 6 of APB Opinion No. 9, Applicgpion to Commer-
cial Banks (New York: AICPA, Inc., 1969). Hereafter referred to as APB l3.

14

Exchange, which might be concerned if these banks (1) did not have their
financial statements audited by CPA's or (2) received qualified opinions
from CPA's for failure to adhere to APB 13.13

However, at a meeting on July 8, 1969, representatives of the
AICPA, the American Bankers Association, the three Federal banking
regulatory authorities and the SEC reached an understanding as to
acceptable standards of reporting for commercial banks. This agreement
essentially embodied the tenets of APB 13. In accordance with instruc-
tions issued by the three bank regulatory agencies subsequent to this
meeting banks began reporting Net Income in their 1969 annual report.

Thus, a reporting impasse between bankers and accountants was
averted. The agreement can be praised for primarily one reason --
uniformity; bank financial statements are now more comparable with those
of other industries. However laudible this improvement may be from the
standpoint of uniformity, it should not be interpreted as meaning that
the controversy has been resolved. Past compromises in the interest of
uniformity have not always proved successful in resolving controversies
dealing with accounting matters. The investment credit controversy
provides the most notable example. Prior to the issuance of AFB 2,
accountants were unable to reach agreement as to what accounting treat-
ment should be accorded the investment credit. After much debate, APB 2
was issued in the interest of achieving uniformity. However, it did not
resolve the controversy. The issuance of APB 4 attests to that fact.

Indeed, the controversy remains unresolved.

 

13Four bank stocks were listed on the New York Stock Exchange as
of July 31, 1969.

15

The investment credit controversy still exists because the
disputants have been unable to find a common theoretical or empirical
basis on which to evaluate the various proposed accounting treatments.
Perhaps no common ground exists. Yet this has not precluded the need
for research efforts to discover such a basis.

A similar case can be made for the bank income reporting
controversy. For nearly two-and-one-half years (considerably longer if
one considers the fact that most large banks have been reporting Net
Operating Earnings for over fifteen years) bankers and accountants stood
firm in their respective positions. Then, on July 8, 1969 the banking
regulatory agencies abandoned their previous position and sided with the
accountants. Since no new arguments had been presented and no new
empirical evidence uncovered, this change in position must be attributed
to a desire for more uniform financial statements on the part of the
regulatory agencies. However, since no new evidence was presented in
reaching the compromise it is debatable as to whether the arguments will
abate and the controversy die or pressure be brought by dissident bankers
to have the reporting regulations changed. In fact, recent statements in
the financial press indicate that some bankers remain dissatisfied with

the new reporting format.14

Thus, there is still need for research to
determine if a common basis can be found upon which the opposing positions

can be judged and the controversy resolved.

1.5 Scope and Methodolpgy

Since the purpose of this research is to obtain empirical evidence

 

14See, for example, the January 16, 1970 (p. 23) and April 15, 1970
(p, 1) issues of The Wall Street Journal.

16

that will aid in resolving the controversy, the scope of this disserta-
tion is limited primarily to an analysis of past earnings data of several
large banks. More specifically, the earnings data that is analyzed
consists of Net Operating Earnings and some measure of Net Income - i.e.,
Net Operating Earnings adjusted for security gains and losses and some
provision for loan losses.

A review and discussion of the theoretical arguments surrounding
the accounting treatment of security gains and losses and the provision
for loan losses is presented in Chapter II. The purpose of such discussion
is twofold. First, some review of the items involved in the controversy
is necessary before any additional analysis can be undertaken. Second,
the discussion serves to point out that logical arguments exist supporting
the positions taken by both bankers and accountants. Presumably, this is
why the controversy exists. Therefore, there is a need to develop some
other basis for settling the dispute since logical reasoning has failed
to provide a solution to the controversy. Chapter III is devoted to the
development of just such a basis. More specifically, it (1) contains an
analysis of the basic reasons for the controversy and (2) lays the founda-
tion for the development of an empirical study the results of which could
provide evidence that will aid in resolving the controversy. The method-
ology used to operationalize the empirical study developed in Chapter III
is presented in Chapter IV. Chapter V contains the empirical findings
while Chapter VI discusses the limitations and presents the conclusions

and implications of the study.

CHAPTER II

THE ITEMS INVOLVED IN THE CONTROVERSY

2.1 Introduction

Previous discussion has already pointed out that the principal
items around which the controversy centers are security gains and losses
and the provision for loan losses. Without these items there would be no
controversy since there would be little or no difference between the
amount of the final reported earnings figure advocated by bankers and
the figure supported by accountants.

This chapter presents the arguments advanced by bankers and
accountants in support of their positions for or against including
security gains and losses and some provision for loan losses in a final
reported earnings figure. As indicated in Chapter 1, the purpose of
such a presentation is (l) to provide background information about the
controversy and (2) to show that there is a need to develop a new basis
for evaluating the positions taken by the two opposing sides to the

controversy.

2.2 Securipy Gains and Losses1

 

1The following sources were consulted extensively in synthesizing
the arguments presented in this section: The Accounting Commission, NABAC
Accountipngulletin No. 1,,Revised -- Securipies Accounting (NABAC, The
Association for Bank Audit, Control and Operation, 1967). Walter C.
Johnson, "Financial Reporting To Stockholders By Commercial Banks" (unpub-
lished Master's Thesis, Stonier Graduate School of Banking, Rutgers -- The
State University), pp. 43-74. Shipley, pp. 146-147.

17

18

2.2.1 The Bankers' Position

Bankers contend that security gains and losses (SG&L) should be
treated as nonoperating items and excluded from the determination of any
measure of current earnings for the following reasons.

1) SG&L do not result from activities that are a normal part
of bank operations because the magnitude of the amounts and the timing of
the transactions are often influenced by other than usual operating trans-
actions. This argument stems from the preferential tax treatment accorded
banks regarding the treatment of SG&L. Banks are allowed to claim the
capital gains rate on profits on securities held longer than six months.
They are also permitted to deduct losses on securities sold at ordinary
tax rates. However, the tax rate applicable in any given year depends
upon whether there is a net gain or net loss after gains and losses have
been offset against each other. The capital gains rate is applied if
there is a net gain, the ordinary rate if there is a net loss. Consequently,
prudent financial management requires that, in order to obtain the maximum
tax benefit, banks must avoid taking security gains and losses in the
same year and must strive to concentrate losses in one tax year and gains
in another.

Bankers point out that such tax-inspired transactions produce wide
fluctuations in SG&L from one year to the next. These wide fluctuations,
they argue, are not reflective of the normal operations of a bank.

2) Because of the wide fluctuations in SG&L from one year to the
next, the inclusion of SG&L in the determination of current year earnings
would distort any comparison of year to year earnings figures and could
have two deleterious effects. First, it could lead to serious misinter-

pretation by users of the reported earnings figure who do not understand

19

the nature of SG&L. Second, it could lead "Statement conscious"
managements to comingle gains and losses in a given year to offset the
wide fluctuations presently existing. Such a management decision would
lose for the bank and for the shareholders the benefits available under
the tax laws and would be to their joint detriment.

3) The bulk of SG&L for any one year represents an adjustment
of prior years' earnings and therefore should not be included in the
determination of current earnings. Since the great majority of the
securities in a bank's portfolio are held for periods in excess of one
year, bankers argue that it is not proper to include in the earnings of
the year of disposition a gain or loss that is partly attributable to
one or more prior years. They argue further that the entire amount
of a given year's SG&L should be excluded from current earnings since
no acceptable method has been developed for apportioning SG&L to either
prior to future periods.

4) Current disclosure requirements, when coupled with the above
arguments, provide the financial statement users with sufficient informa-
tion to determine the impact of SG&L on the current and future operations
of the bank.

Past practice in banks' treatment of security gains and

losses has included the carrying of the results of such
transactions directly to the undivided profits account or
to reserves specifically provided for this purpose. If the
annual reports did not include a reconciliation of such
reserves or of the undivided profits account, the investor
or analyst would have no opportunity to ascertain the impact
of such transaction on the equity portion of the balance
sheet. Under the provisions of Regulation F and of the
Comptroller's new regulation, however, the results of such
transactions must be shown as a "below-the-line" item. In
addition, Regulation F and the Comptroller's regulations

each require a reconciliation of the undivided profits
account and all reserves, thereby providing further dis-

20

closure as to the accumulated results of the investment trans-
actions.2

2.2.2 The Accountants' Position

 

Accountants argue that SG&L should be included in the determina-
tion of a final reported earnings figure for the following reasons.

1) SG&L do result from activities that are a normal part of
bank operations. This conclusion is derived from the following arguments.
Banks maintain security portfolios in order to earn a return on funds
that must be maintained in a liquid state to meet any sudden deposit
withdrawals. Gains or losses on dispositions of securities result for
one of two reasons. First, a gain or loss may be incurred because the
securities portfolio was reduced to meet a deposit run off. Second,
securities could be disposed of in an effort to improve the long-term
earnings of the bank by such activities as (1) realigning the portfolio
in order to obtain an improved long-term yield, or (2) granting loans at
higher yields, or (3) reducing interest costs on borrowed funds. The
size of a given period's gain or loss is indicative of (1) management's
ability to forecast fluctuations in demand deposits and (2) the degree
of risk management is willing to accept in its efforts to obtain invest-
ment income. Management could minimize security losses by purchasing
short-term, low-yield treasury bills. Changes in interest rates would
have little or no affect on the market price of these shortly maturing
securities. With the exception of dispositions incurred to meet sudden
deposit withdrawal demands, losses could also be minimized by holding all
securities to maturity.

The consistent incurrence of large gains and losses over past

 

ZShipley, p. 147.

21

years and the continued inclusion of long-term, high yield securities

in banks' security portfolios indicates that SG&L are a part of a bank's
normal and expected operating pattern. Therefore, it is illogical to
include the income resulting from a management decision to use the funds
provided by the disposition of securities in current earnings and at the
same time exclude the results of the securities transactions that made
the funds available.

2) The true measure of performance of a security is its yield.
Yield is determined by the difference between the purchase price and the
selling price of a security plus any interest received during the holding
period. Therefore, it is inconsistent to include interest, premium
amortization, and discount accretion in current earnings but to exclude
SG&L.

3) Excluding SG&L from the determination of a final reported
earnings figure allows banks to manipulate current earnings and prevents
comparison with other banks. The following two examples illustrate the
basis for this argument.

Example 1. Assume that a bank holds $1,000,000 of 3% treasury
notes at par maturing in five years and that the current yield on treasury
notes maturing in five years is 7%. Annual interest income on these notes
would amount to $30,000 per year. Discounting the currently held treasury
notes at 7% indicates that they should have a current market price of
approximately $836,000. By selling the notes the bank would realize an
after-tax loss of $82,000 assuming an ordinary tax rate of 50%. This loss,
of course, would be excluded from current earnings. By reinvesting the
after-tax proceeds of $918,000 ($836,000 plus $82,000 tax reduction) in

new five-year 7% treasury notes the bank will receive $64,260 (.07 x

22

$918,000) in interest income before taxes each year to maturity. The net
effect of this transaction is that net operating earnings before taxes
would be increased $34,260 (from $30,000 to $64,260). Yet the actual
earnings performance of the bank would not have changed materially. A
loss would have been realized currently in order to increase future
earnings. It seems clear that if this loss is never reflected in income,
the reported earnings of this bank would be overstated.

Example 2. Assume that a bank holds a $1,000,000 3% bond at par
maturing in five years and that its current market value iS $900,000. By
selling the bond the bank would realize a loss of $100,000 which would be
excluded from current earnings. The proceeds could then be reinvested in
a different $1,000,000 3% bond for approximately $900,000. If the discount
of $100,000 is accrued over the remaining life of the bond, the net effect
would be to increase net operating earnings by $100,000. A stockholder
would be no better off as a result of this transaction and, yet, he would
be lead to believe that the earnings performance of the bank was improving.

(For purposes of simplicity, income taxes were ignored in this example.)

2.2.3 Proposed Alternative Accounting Treatment

Some bankers and accountants have proposed an alternative method
of accounting for the bulk of SG&L. Basically, this method involves the
deferral and subsequent amortization of SG&L resulting from portfolio
realignment (as distinguished from gains or losses resulting from sales
of securities to obtain proceeds to meet new loan demands or deposit
withdrawals). Under this method the amortized amounts would be included
in the determination of current earnings as an adjustment of interest

income from securities.

Proponents of this method view the sale and subsequent reinvest-

23

ment in similar grade securities as constituting no substantive change
in the investment portfolio. They argue that amortizing the gain or
loss to the maturity date of the security sold results in the inclusion
of gains and losses in the fiscal period to which they relate.

They [the proponents of deferral and amortization] point out

that the same decline or rise in prevailing interest rates

which results in a gain or loss also results in a commensu-

rately lower or higher interest earnings (that is, coupon

rate adjusted for premium or discount amortization) on the

security purchased with the proceeds of sale during the

period from the date of sale to the maturity date of the

security sold. Accordingly, they maintain that the deferral

and amortization of gains and losses to the maturity dates

of the securities sold result in the fairest presentation

of investment earnings.3

Advocates of the deferral method also argue that deferral and
subsequent amortization would eliminate the deficiencies illustrated in
the two examples presented in Section 2.2.2 above.
Opponents of the deferral-and-amortization method maintain that

a sale constitutes a completed transaction and therefore no part of the
gain or loss should be attributed to subsequent securities purchased or
subsequent years' operating earnings. "They point out further the lack
of precedent for such treatment, either in practice in other industries

or in authoritative literature."4

They also question the desirability of adopting a practice that
tends to delay the reflection in interest earnings of changes
in prevailing interest rates when, at the same time, such
changes are reflected relatively currently in the interest
rates banks pay for interest deposits and borrowed funds.5

 

3Committee on Bank Accounting and Auditing of the American Institute

of CPAs, Audits of Banks, (New York: AICPA, Inc., 1968), p. 38.

4Ibid, p. 39.

51bid, p. 39.

24

Those opposing the deferral method also argue that application
might not be consistent, either by a given bank over a period of time or
across several banks for a given period, because of the inability to
objectively define activities related to portfolio realignment and those
related to the demand for more liquid funds.

The Committee on Bank Accounting and Auditing of the AICPA is the
principal authoritative source which supports the use of the deferral-and-
amortization method. The Committee, however, recognizes the acceptability
of both the completed-transaction (conventional) and the deferral methods.6
The Accounting Commission of the Association of Bank Audit, Control and
Operation (NABAC), although it has published a summary of the arguments
supporting both the completed-transaction and the deferral-and-amortiza-
tion methods, has indicated that it does not ”endorse, oppose, or comment
on the validity of the arguments presented."7 The principal source of
opposition to the deferral-and-amortization method comes from the three
federal regulatory agencies who have taken the position that the deferral
concept does not conform with current regulations.8 As a consequence,
the deferral method has been adopted by only a few banks for use in
determining SG&L in their published financial statements and has not been

a significant factor in the controversy.

2.3 The Provision for Loan Losses9

 

 

Committee on Bank Accounting. . ., Audits of Banks: Supplement, p. 1.
7The Accounting Commission, p. 8.

8Committee on Bank Accounting. . ., Audits of Banks: Supplement, p. l.

 

9The following source was consulted extensively in developing the
material presented in this section: Ray H. Garrison,"Accounting for Bad
Debt Reserves in Commercial Banks" (unpublished Ph.D. dissertation,
Graduate School of Business, Indiana University, 1966).

25

2.3.1 The Bankers' Position

The bankers' position that no provision for loan losses (PFLL)
should be included in the determination of current earnings is based
primarily on tradition. In order to fully understand their position it
is necessary to review the events and activities that have lead to the
development of the treatment advocated by them.

Initial impetus for the exclusion of either actual loan losses
or some provision for loan losses occurred in the early 1930's. Prior to
and during the depression most banks followed the direct charge-off method
of accounting for loan losses.10 Under this method, loan losses were
charged against current earnings in the year the loan was determined
to be uncollectible. As a consequence of the huge loan and security
losses that were incurred during the depression most banks were showing
net deficits instead of net profits. According to one source, it was at
this time that the few banks that published any earnings data began
publishing "net current operating earnings" (i.e. - earnings that excluded
SG&L and a PFLL) for reassurance as to the existence of normal earning
power which would reassert itself once these "extraordinary" losses
returned to more normal levels.11

The federal regulatory authorities lent support to this new
reporting practice by requiring, in 1933, reporting banks to isolate and
detail in their annual Report of Income and Dividends their loan loss

activities in a special "losses and recoveries" section below "net current

 

10The Accounting Commission, NABAC Accounting Bulletin No. 3 -
Loan Loss and Related Reserve Accounting (NABAC, The Association for Bank
Audit, Control and Administration, 1966), p. 1.

11F. L. Garcia, How To Analyze a Bank Statement (Boston: Bankers
Publishing Company, 1966), p. 50.

26

operating earnings".12

One researcher has suggested that the purpose of
this requirement was "(1) to eliminate the distorting effect of depres-
sion losses from the results of normal operating activities, and (2) to
note clearly the portion of current earnings being set aside as protective
reserves."13

Little public attention was focused on bank earnings until many
large banks began publishing detailed financial statements, including an
income statement, in the middle 1950's. By this time, of course, loan
losses and SG&L had had ample time to return to normal levels. However,
instead of returning to pre-depression income determination methods, these
banks reported "net current operating earnings" as the measure of current
earnings.

By this time two new reasons for reporting ”net current operating
earnings" had emerged. First, the federal regulatory agencies continued
to require the annual Report of Income and Dividends to be filed in

4 Although banks

essentially the same format as established in 1933.1
were free to choose a different reporting format for published reports
for their own circulation, they chose to emulate the federal agency
reports.

The second, and by far the most important, reason can be attributed
to the issuance and subsequent revision by the Treasury Department of an

arbitrary formula for the determination of the annual PFLL deductible for

tax purposes. The first formula was introduced by the Internal Revenue

 

12Garrison, p. 151.

13Ibid.

141b1d.

27

Service in 1947 in the form of Mimeograph 6209.15 According to Mimeo-
graph 6209, a commercial bank was allowed to accumulate loan loss reserves
equal to three times its average percentage loan loss experience over
the preceding twenty-year period, on a moving average basis. This formula
was modified in 1954 to allow banks to use any twenty-year period after
1927 as a basis for establishing its loan experience.l6
Adoption of the Internal Revenue Service formula allowed many
banks to accumulate substantial loan loss reserves. However, some banks
were able to accumulate only small reserves because they had incurred
only minor loan losses during the twenty-year experience period. In
essence, the formula discriminated against banks with a good loan loss
record in favor of those with a poor record; since, the greater a bank's
historical loan loss experience, the greater was the size of the reserve
allowable under the formula.
As a result of this discrimination, the formula prescribed in
Mimeograph 6209 was replaced with a new formula contained in Revenue
Ruling 65-92 issued by the Internal Revenue Service on March 15, 1965.17
This new formula, which remains in effect today, allows a bank to accumu-
late a loan loss reserve equal to 2.4% of outstanding loans at the end of
the year. The reserve can be built up over a ten-year period resulting in
an allocation each year that may substantially exceed the actual or average
loss experience of a given bank since the 2.4% bears no relationship to

an individual bank's current or past loan loss experience. The 2.4% figure

 

15Mimeograph 6209 (reproduced in full in Garrison, pp. 214-216.)

16Revenue Ruling 54-148 (reproduced in full in Garrison, pp. 217-218.)

17Revenue Ruling 65-92 (reproduced in full in Garrison, pp. 221-225.)

28

does have some historical significance in that it is "equal to three times
the average percentage loss experience of all operating commercial banks
over the twenty-year period 1928-1947."18

Bankers were, and continue to be, reluctant to include the total
tax-deductible PFLL in the determination of current earnings because of
the distortiVe effects it could have on current earnings. This distortion
is due to the following factors. First, since the 2.4% reserve-to-loan
ratio bears no relationship to an individual bank's experience, amounts
deductible under the formula can be extremely large -- ”equalling 50% or
more of before-tax income in many cases."19 Second, amounts deducted
under the formula tend to vary widely from year to year. This variation
is caused by two factors. The first is an option in the IRS regulations
that allows bankers to pass up deductions in one year and make up the
deficiency in a subsequent year.20 This option "permitted a bank to alter
its bad debt deduction according to the current year's taxable income,
and even omit a deduction entirely if a loss year was sustained. However,
in a subsequent year when earnings were more favorable a massive deduction
could be taken, if needed, to reduce the taxable earnings to the desired

1."21 The second factor was the requirement that once a bank reached

leve
its reserve ceiling, no further deductions were allowable until the

reserve-to-loan ratio dropped below 2.4%.22 This requirement, coupled

with the requirement that, if adopted, the formula had to be used for

 

18Garrison, p. 5.

191bid, p. 153.

20Ibid, p. 44.

lebid.

ZZIbid.

29

23 meant that in some years there was no PFLL

both book and tax purposes,
to report in the financial statements.

Therefore, bankers concluded that treatment of the total tax-
deductible PFLL as an operating expense would produce artificial and,
perhaps, unfavorable trends in current earnings. To them, the format
suggested by the federal regulatory authorities' Report of Income and
Dividends provided the most suitable means of eliminating these distortive
effects.

Accountants have generally agreed with bankers, albeit for different
reasons, that it is improper to include the total tax-deductible PFLL in
the determination of current earnings, unless it happens to equal a
normal PFLL. Accountants base their arguments for excluding a portion
of the total tax-deductible PFLL from current earnings on the following
grounds. Over the long-run, the total tax deductible PFLL contains both
a valuation element and a contingency element. The valuation element is
designed to allow banks to build a reserve to absorb loan losses that can
be expected to be incurred as a result of normal lending activities. The
contingency element is designed to allow banks to accumulate a reserve
to absorb loan losses which occur at infrequent, irregular, and unpredict-
able intervals such as during a depression or severe recession. Accountants
contend that only the valuation element should be included in the deter-
mination of current earnings while the contingency element should be
excluded since it represents nothing more than a segregation of retained
earnings.

The basis of their contention that a portion of the total tax-

deductible PFLL contains a contingency element stems from the preferential

 

23Ibid. 45.

30

tax treatment granted banks in recognition of the important social
welfare role played by them. The essence of this social welfare function
is embodied in the following statement included in a staff paper prepared
by the American Bankers Association and submitted to the Assistant
Secretary of the Treasury in regard to the revision of Treasury guide-
lines for bad debt reserves.
Although the specific function of a bad debt reserve is to
absorb losses on loans, the public policy implications of the
existence of such reserves are extremely important. Commercial
banks occupy a central place in the economy as providers of
credit for business and individuals. When adequate provision
has been made to absorb loan losses, banks are in a better
position to continue to make the kinds of loans which are
essential to the economic growth and well-being of their
respective communities, even during periods of economic
adversity. Thus, in a broad sense, the function of bad debt
reserves should be viewed as that of enabling the commercial
banking system, under varying economic conditions, to play
its full role in sustaining the nation's economic growth.24
Accountants contend that proof of the recognition of the social
welfare role played by banks is inherent in the fact that the Treasury
Department allows banks to accumulate a ratio of reserve-for-loan-losses
to loans that is three times greater than the average ratio over a twenty-
year period that included the worst years in the history of banking.
The following discussion by the former chairman of the Accounting
Principles Committee of the American Bankers Association illustrates why
bankers have not been willing to segregate a normal portion of the tax-

deductible PFLL and treat it as a determinant of current earnings.

While there can be no quarrel with the theory that some
charge for possible losses in the loan account should be made

 

24The American Bankers Association, Commercial Bank Bad Debt
Reserves: A Re-examination and Appraisal of Treasury Guidelines, A staff
paper submitted to the Assistant Secretary of the Treasury (Washington,
D.C., 1964), p. 7, quoted in Johnson, pp. 84-85.

2

31

against current operating earnings, there have not, at the
present time, been established any criteria for the determina-
tion of the proper amount of this charge. Suggestions have
been made that the charge to current operating earnings should
represent the average loss experience for a five-year period,

a ten-year period, or a twenty-year period. Those advocating
the five-year period make the point that only by using the most
recent data will the content of the loan portfolio approximate
the same type of loans as those existing during the current
accounting period; thus, the loss ratio encountered during the
most recent time interval is that most likely to continue with
each successive accounting period. On the other hand, those
advocating the longer time interval point out that a five-year
span is too short a time to include the effect of the extremes
of the economic cycle. In the case of catastrophic losses,
such as those resulting from the recent vegetable oil scandal
or those of banks participating in a recent substantial credit
transaction with a large national industrial company, the five-
year average can be substantially distorted. In the case of
the loans to the industrial company referred to, recovery was
obtained within a five-year period, which would have the effect
of minimizing the charge against current operating earnings in
those years following the recovery after the charge-off was
excluded from the average. It would even be possible under
such circumstances for the five-year average to result in a
net credit, which on its face would be ridiculous. For one
bank affected by this particular loan, the five-year average
ranged from .08 per cent to .0024 as the effect of the recovery
was realized.

Another approach advocated by some has been that a determina-
tion should be made by management at the end of each accounting
period of those loans which represent potential losses and that a
charge should be made against current operating earnings in such
an amount. Opposing this, of course, is the weight of authority
that financial statements should be based upon objective
accounting and not upon subjective interpretation. As is
pointed out in the AICPA's Accounting Research Bulletin No. 43,
"an important objective of income presentation should be the
avoidance of any practice that leads to income equalization."
Certainly, a subjective determination of potential loan losses
could easily lead to income equalization.25

.3.2 The Accountants' Position

Accountants base their contention that some PFLL should be included

in the determination of current earnings on the following arguments.

 

25Shipley, p. 145.

32

First, they point out that the interest rate charged a loaning customer
is determined by four factors: namely (1) the cost of money, (2) the cost
of servicing, (3) an allowance for losses and, (4) a reasonable profit.
Therefore, they contend, it is inconsistent to include interest from
loans in current earnings and to exclude some provision for loan losses.

Furthermore, they argue that failure to include a normal PFLL in
current earnings could result in misleading conclusions being drawn when
comparing the performance of different banks. The interest charged a
loan customer reflects the degree of risk associated with collecting the
loan - i.e., the greater the risk, the higher the interest rate. The
higher interest rate is necessary to provide income to offset subsequent
loan losses that would be greater than those incurred with loans of a
lesser risk. Given two banks with an equal amount of loans outstanding,
the bank with a portfolio of high risk loans will have higher interest
income than the bank that lends only to low risk customers. The "high
risk" bank is also likely to have larger loan losses over time than the
"low risk" bank. Under the reporting practice advocated by bankers, the
”high risk" bank will show higher current earnings, ceteris paribus, than
the "low risk" bank even though their overall long-run all-inclusive
earnings may be identical.

Accountants contend that a normal PFLL should be based on (1)
past loss experience, adjusted for such factors as known changes in the
character of the loan portfolio, (2) management credit policies, and (3)
economic conditions.26 They argue that, despite the element of subjectiv-

ity involved, bank management is in the best position to weigh each of

 

26Committee on Bank Accounting. . ., Audits of Banks, p. 47.

33

these three factors in determining a PFLL that will "result in systematic

loan loss charges to operating earnings on a consistent basis.27

2.4 Summary and Conclusion

The preceding sections present the reasoning used by bankers and
accountants in support of their respective positions on the treatment of
SG&L and a PFLL. The arguments presented in those sections can be
summarized briefly as follows. Bankers favor excluding SG&L and PFLL
from a final reported earnings figure on the grounds that they are not
a normal part of bank operations because of their artificial and widely
fluctuating nature. Accountants argue that SG&L and loan losses are
incurred as a part of normal bank operations. Therefore, they contend
that SG&L and some PFLL should be included in the determination of the
final reported earnings figure.

An analysis of the arguments presented by both sides indicates
that there is logical support for each position. Presumably, this is why
the opposing groups have been unable to reach agreement as to the proper
treatment of each of the items. As is the case with so many issues in
accounting today, resolution as to which treatment is "better" on
theoretical grounds has become impossible because it is useless to argue
which is the "more logical".28 Therefore, there is a need to search for
and obtain other means of evaluating the alternatives. The results of

such a search are presented in the following chapter.

 

271b1d.

28Consider, for example, the following issues: (1) all-inclusive
vs. current operating concept, (2) treatment of investment credits, (3)
purchase vs. pooling, and (4) historical vs. current costs.

CHAPTER III

FORMULATION OF AN EMPIRICAL STUDY
DESIGNED TO AID IN RESOLVING THE CONTROVERSY
3.1 Introduction
The purpose of this chapter is to present the results of a search
for new means of resolving the controversy. To meet this end, the chapter
is divided into two major sections. Section 3.2 analyzes the controversy
in a new light in order to show that there is a primary and secondary
reason for its existence. By identifying these reasons the foundation
is laid for the development of two empirical studies the results of which
could provide new evidence that would aid in resolving the controversy.
Section 3.3 introduces these studies and explores the feasibility of

performing each.

3.2 The Basic Reasons for the Controversy

3.2.1 The Primagy Reason

Further analysis of the controversy indicates that although there
would be no controversy if security gains and losses and a provision for
loan losses did not exist (the difference between Net Operating Earnings
and Net Income would be insignificant), the primary reason that it remains
unresolved is the belief that investors have single-figure fixation.
According to this belief investors focus almost entirely on the final
reported earnings figure in evaluating the current and potential performance

of a firm while other financial data included in the published financial

34

35

statements are largely ignored.1
The essence of this belief is perhaps best described in the
following quotation taken from Forbes magazine:

In our complicated, changing world it is not really possible
to distill a whole year's operations of a vast business enter-
prise into a single absolute figure. The public, however,
doesn't always appreciate this. Thus, the annual net earnings
figure tends to have a magical significance - not only for the
ordinary investor but for security analysts and even acquisition
minded managements. It becomes in effect what grades are for
the student - a measure of excellence, of progress or lack of
progress.

People tend to demand this kind of simplicity, and the single
conveniently packaged net earnings figure has always seemed to
fill the bill perfectly. Companies report them frequently -
annually, quarterly and sometimes even monthly - and they carry
the blessings of certified public accountants. So it is that
reported earnings have become the common denominator of the
stock market. On their reliability, billions of investment
dollars are wagered.2

The antithesis of this belief holds that investors are of a more
rational nature. They realize that many financial factors must be
considered in evaluating a firm's performance. Consequently, it does
not matter so much what amount is presented as the final reported earnings
figure as it does that all pertinent financial facts are disclosed in the
financial statements.

That the existence of the belief in single-figure fixation is
widespread is shown not only by the above quotation but also by the
following references to its existence:

The board's [Accounting Principles Board] most complex
decision came as it struggled to divulge what Savoie [Leonard

 

1Single figure fixation, in the sense used here, pertains only to
information in the financial statements. It does not mean that investors
ignore such external factors as gross national product, industry trends,
sales forecasts, etc. in making investment decisions about a particular
stock.

2”WhatIAEE’Earnings? The Growing Credibility Gap." .Forbes,
May 15, 1967, p. 28.

36

M. Savoie, exeuctive vice-president of the AICPA] calls "ersatz-
earnings" -- per-share profits derived from fancy financial
footwork. This is g sensitive mggter becppse many investors
mistakenly believe that they can gauge a stock's merit simply
by checking,per-share earnings. (Emphasis added.)3

"Net Income" figures are widely published, and in spite
of admonitions against the use of a single figure in assessing
the operating results of an enterprise, there is evidence that
such a figure is, in fact, widely used and relied upon by
investors and others.4

Evidence that accountants believe in the existence of single-
figure fixation is found in the following statements taken from AICPA
publications. Accounting Research Bulletin No. 43 makes the following
reference to it:

In its deliberations concerning the nature and purpose of

the income statement, the committee has been mindful of the
disposition of even well-informed persons to attach undue

importance to a sipgle net income figure and to earningsgper
share shown for a particular year. (Emphasis added.)5

 

In advocating the presentation of earnings-per-share figures before and
after extraordinary items, APB 9 concludes that,
not only will this format increase the usefulness of the reports
of results of operations of business entities, but it will also
help eliminate the tendency of many users to place undue emphasis
on one amount reported as earnings per share.
While this researcher has found no explicit statements by bankers

indicating that they believe in the existence of single-figure fixation,

it is not difficult to reach this conclusion. First, the bankers'

 

3Time, p. 96.

4Leopold A. Bernstein, Accounting,for Extraordingrinains and Losses
(New York: The Ronald Press Company, 1967), p. 15.

5AccountingResearch and Terminology Bulletins (Final Edition;
New York: AICPA, 1961), Accounting Research Bulletin No. 43, p. 65.
Subsequently referred to as Bulletin 43.

6APB 9, p. 119.

37

contention that investors will make misleading inferences if Net Income
is published as the final reported earnings figure indicates that they
do not believe that investors look beyond the final reported earnings
figure.

Second, the existence of single-figure fixation provides the only
plausible explanation as to why bankers have been reluctant to present
some measure of Net Income as the final reported earnings figure. As
support for this conclusion, consider the affect that a change from Net
Operating Earnings to Net Income is likely to have on the market price of

a bank's stock if sipgle-figure fixation exists. The price would fall for

 

two reasons.

First, Net Income is likely to be lower over time than Net Operating
Earnings because of the annual provision for loan losses included in the
determination of Net Income. As David Gates, the bank stock analyst at
Loeb, Rhodes & Company puts it:

Many banks hesitate to be the leader in lowering the level of
reported earnings, the natural consequence to be anticipated from
"All-Inclusive" accounting. This reluctance is understandable, to
the extent banks compete for prestige, a high market price for
distribution of stock, or to attract merger candidates with high-
priced shares. [sic.] 7

Second, Net Income is likely to fluctuate more than Net Operating
Earnings because of the nature of security gains and losses. Earlier it
was pointed out that bankers lump all security gains in one year and all
losses in another. The affect of this practice is to amplify any swings

in Net Operating Earnings. Figure 3 illustrates this point as it relates

to one bank - the First National Bank in Dallas. In this case, Net Income

 

7David C. Gates, "Developing Directions in Bank Accounting,"

Bankers Magazine, CL (Summer, 1967), P. 36.

$ Per
Share

3.00

.20

.00

38

 

 

57

58

59 60 61 62 63 64 65

Year

 

FIGURE 3

GRAPHIC PRESENTATION OF TWO EARNINGS PATTERNS

39

clearly is more volatile (i.e., fluctuates more widely) than Net Operating
Earnings. Aside from the fact that Net Income is likely to be lower
than Net Operating Earnings, publishing Net Income as the final reported
earnings figure is likely to have a downward affect on the bank's stock
price because investors prefer, ceteris paribus, smoothly rising earnings
to widely fluctuating earnings even though the two may be equal in the long
run.8 The fluctuating earnings are considered to be of greater risk and
are given a lower market value.

Note, however, that the reduced amount and increased volatility of
the final reported earnings figure will have an affect on a bank's stock

price only if single-figure fixation exists. For the past three to five

 

years most large, publicly-held banks have been reporting actual loan
losses, the total tax-deductible provision for loan losses, and security
gains and losses in their published financial statements either as
supplementary information directly below Net Operating Earnings in the
income statement or in supplementary reserve reconciliation schedules.
Hence, even though these banks have not included these items in the
determination of their final reported earnings figure (i.e. - Net Operating
Earnings), information about them has been available to the interested
investor. Therefore, in the absence of single-figure fixation, the sudden
switch to reporting Net Income would have little affect on the market price
of a bank's stock since investors would already have taken the provision

for loan losses and security gains and losses into consideration in

 

8See, for example, Jerome B. Cohen and Edward D. Zinbarg, Invest-
ment Anglysis and Portfolio Management, (Homewood, Illinois: Richard D.
Irwin, 1967), pp. 235-43, for a further discussion of this generally
recognized investment principle.

40

valuing the stock of the reporting bank. The new final reported earnings
figure would not serve as an unwarranted signal to them that there had
been a sudden change in a given bank's performance because, ceteris
paribus, there would have been no change in performance, only a change
in reporting format.

A review of the literature on the subject of single-figure fixation
indicates that little empirical evidence has been obtained in support of
its existence. However, whether single-figure fixation exists is irrelevant.
What really matters is whether the belief in its existence exists. The
above discussion indicates that the belief does exist and that it is

widespread.

3.2.2 The Secondapy Reason

Belief in the existence of single-figure fixation leads to a
secondary reason for the controversy -- a disagreement between bankers
and accountants as to which of the two earnings measures is the most
relevant single measure of a bank's performance. Bankers, of course,
favor Net Operating Earnings while accountants argue for Net Income.

A disagreement of this nature is by no means unusual. Accounting
Research Bulletin No. 43 discusses the matter as follows:

The question of what constitutes the most practically useful
concept of income for the year is one on which there is much
difference of opinion. On the one hand, net income is defined
according to a strict proprietary concept by which it is presumed
to be determined by the inclusion of all items affecting the net
increase in proprietorship during the period except dividend
distributions and capital transactions. The form of presentation
which gives effect to this broad concept of net income has sometimes
been designated the all-inclusive income statement. On the other
hand, a different concept places its principal emphasis upon
relationships of items to the operations, and to the year, excluding
from the determination of net income any material extraordinary
items which are not so related or which, if included, would impair
the significance of net income so that misleading inferences might

 

41

be drawn therefrom. This latter concept would require the
income statement to be designed on which might be called a
current operatipg performance basis, because its chief purpose
is to aid those primarily interested in what a company was able
to earn under the operating conditions of the period covered by
the statement.

Bankers, of course, have taken a current operating concept approach
toward income reporting by arguing for the exclusion of security gains and
losses and some provision for loan losses from current earnings. Account-
ants essentially favor the all-inclusive concept. Their position is
espoused in APB 9, which now covers banks, as follows:

Net Income should reflect all items of profit and loss

recognized during the period except for prior period adjustments

[which, it explains, are expected to be rare], with extraordinary

items to be shown separately as an element of net income of the

period . . .10

Whether bankers are justified in defending their position from a
current operating viewpoint is open to question. The basic difference
between the all-inclusive and current operating concepts is the financial
statement treatment of extraordinary items. However, just what constitutes
an extraordinary item is not altogether clear. In his authoritative study
on Accounting for Extraordinary Gains and Losses, Leopold Bernstein has
described four types of extraordinary items: "(1) items affecting results
of prior years; (2) non-recurring operating gains or losses; (3) recurring
non-operating gains or losses; and (4) non-recurring, non-operating gains
and losses."11

Given this classification scheme, bankers presumably would place

security gains and losses and the provision for loan losses in category

 

gBulletin 43, p. 60.
IOAPB 9, p. 107.

11Bernstein, pp. 11-13.

42

(l) or (3) on the basis of their arguments presented in sections 2.2.1
and 2.3.1 of Chapter II. As has already been indicated, accountants
contend that, with the exception of the contingency element included in
the tax-deductible provision for loan losses, security gains and losses
and a provision for loan losses are really a normal part of a bank's
operations and should not be treated as extraordinary. This disagreement
remains unresolved because, as Bulletin 43 so aptly puts it, "the ultimate
distinction between operating income and non-operatipg gains and losses,
terms having considerable currency in the accounting profession, has not
been established."12

The important point, however, is not whether bankers are justified
in identifying themselves as advocates of the current operating concept,

but whether they have done so. Their actions indicate that they have.

3.2.3 Summary

The purpose of this section was to show that there is a primary
and secondary reason for the controversy. The primary reason is the
belief in the existence of single-figure fixation. The secondary reason
is the disagreement between bankers and accountants as to which of two
earnings measures is the single most relevant measure of a bank's perform-
ance. The latter reason is identified as being secondary because it would
not exist if it were not for the existence of the former. For, if bankers
and accountants did not believe in the existence of single-figure fixation,
the controversy could be resolved by reporting both Net Operating Earnings
and Net Income in the income statement. The final reported earnings

figure would lose its significance because investors would be considered

 

lzBulletin 43, p. 60.

43

sophisticated enough to take each of the two earnings measures, along
with security gains and losses, a provision for loan losses and other
financial data into consideration in evaluating the performance of a
given bank. There would be no need to debate over which of the two

earnings measures was the most relevant.

3.3 Proposed Mepps of Resolving_the Controverpy

Given that there is a primary and secondary reason for the
existence of the controversy, this researcher suggests that evidence
which might aid in resolving the controversy could be obtained in one of
two ways. The first would be to test for the existence of single-figure
fixation. The second would be to evaluate the relevancy of each of the
two advocated earnings measures. If single-figure fixation was shown
to be nonexistent then, of course, the primary reason for the controversy
would have been found to be without basis. If single-figure fixation was
found to exist or if it were not possible to test for its existence, then
the development and performance of a test of the relevancy of the two
earnings measures could prove useful in resolving the controversy. For,
if one earnings measure was found to be more relevant than the other,
new evidence would exist in support of its publication as the final

reported earnings figure.

3.3.1 Test of Single-Figure Fixation

 

In an attempt to develop a means of testing for the existence of
single-figure fixation, this researcher discovered three empirical studies
dealing with this matter. Each of the studies centered on the public
utilities industry. Each employed market prices as the operational

variable used to measure the affect of alternative accounting practices

44

on investors.13 More specifically, each study attempted to determine if
investors were able to recognize and account for the difference in
earnings reported by utility companies which were classified either as
"normalizing" or "flow-through". All firms in each study computed
depreciation on an accelerated basis for tax purposes and on a straight-
line basis for book purposes. Firms which followed deferred tax
accounting principles were classified as "normalizing". Firms which
deducted the entire amount of the current income-tax liability in
computing reported net income were classified as "flow-through".

A priori, it was argued that if single-figure fixation existed
there would be no difference between the average price earnings ratios
of the "normalizing" companies and the "flow-through" companies in any
given year. On the other hand, if single-figure fixation did not exist,
the average price-earnings ratio of the "normalizing" firms would be
expected to be higher than that of the "flow-through9 firms. This latter
argument was basedwon the fact that, ceteris paribus, in any given year,
the reported earnings of the "flow-throug " firms would be higher than
those of the "normalizing" firms because of the large accelerated
‘depreciation charges incurred by the utility companies over the period
covered by the study. In the absence of single-figure fixation, investors
would recognize that there was no difference between the performance of

the two groups of firms and would therefore capitalize the earnings of

 

13John L. O'Donnell, "Relationships Between Reported Earnings and
Stock Prices in the Electric Utility Industry," The Accounting Review,
January, 1965, pp. 135-143. John L. O'Donnell, "Further Observations on
Reported Earnings and Stock Prices," The Accounting Review, July, 1968,
pp. 549-553. Francis A. Mlynarczyk, Jr., "An Empirical Study of Accounting
Methods and Stock Prices," Empirical Research in Accounting: Selected Studies,
1969 (published as a supplement to Vol. 7 of the Joppnal pf Accounting

Research), pp. 63-89.

45

the flow-through firms at a lower rate.

In general, the results of each of the studies indicated that
investors in public utility stocks did not have single-figure fixation
as related to the use of alternative accounting principles -- i.e., they
were able to correctly differentiate between "flow-through" and "normalizing"
companies in valuing reported earnings.

This researcher had originally planned to use methodology similar
to that described above to determine whether investors in bank stocks

have single-figure fixation as related to alternative rgporting practices.

 

However, in order to perform such a study it would be necessary to obtain
adequate samples of two types of banks -- those that reported Net Income

and those that reported Net Operating Earnings. Utilizing methodology
similar to that employed in the utility studies, it would then be possible
to compare, after certain adjustments, the average price-earnings ratios

of the sample banks included in each reporting category. Unfortunately,

too few banks, for which earnings data could be obtained, have ever reported
Net Income as the final reported earnings figure. Furthermore, the reported
Net Income figures of these banks were not comparable because varying
accounting and reporting practices were employed in determining Net Income.
Hence, it was not possible to obtain an adequate sample of comparable "Net
Income" banks. Therefore, such a study could not be performed. Further
attempts to develop a reasonable and practical means to test for the

existence of single-figure fixation proved to be fruitless.

3.3.2 Test of Relevgncy
Since a test of the existence of single-figure fixation was not

feasible, efforts were focused on the second approach toward obtaining

46

evidence that might aid in resolving the controversy -- the development
and performance of an empirical study designed to determine which of
the two earnings measures was the more relevant.

Before determining the feasibility of performing such a study it
was first necessary to specify an operational definition of relevance.
Consequently, the prediction of future earnings was identified as being
the most relevant need of the primary users of a bank's final reported
earnings figure. This conclusion was based on the following pronouncement
by the American Accounting Association's Committee to Prepare a Statement
of Basic Accounting Theory.

Almost all external users of financial information reported
by a profit-oriented firm are involved in efforts to predict
the earnings of the firm for some future period. Such predictions
are the most crucial in the case of_present_§nd prospective equipy
investors and their reppesenpgtives -- considered by mgny to be
the most impoppppt of the user groups. Future earnings are the
chief determinant of future dividends and future market prices of
shares (given some predetermined price earnings ratio), which,
when taken together, are generally considered to provide the
primary basis for establishing a subjective value for the shares
in the mind of the user. The pgpt earnipgs of the firm are
considered to be thegppst important single item of information
releygnt to the prediction of future pgrnipgs. It follows from
this thatépést eggnipgs should be mpgsured and disclosed in such
a manner as to,give a user as much aid as practicable in efforts
to make this ppediction with a minimum of uncerpginty. (Emphasis
added.)

I O C O O O ' O C O O O O O O O O O O O O O O O O O O O O O I O O O

A person using financial statements as an aid in predicting
future earnings has a right to demand from the accountant measure-
ments of past earnings that supply as much relevant information as
possible.14

Additional support was found in the following statement made by Professor

Robert Sprouse:

 

1l'American Accounting Association, Committee to Prepare a Statement
of Basic Accounting Theory, A Statement of ngic Accounting Theory
(Evanston, Illinois: 1966), pp. 23-24.

47

The primary purpose of the measurement of last year's income
reported to investors is to provide a basis for predicting future
years' income.15

The primary users of a bank's final reported earnings figure were
assumed to be (1) "present and prospective equity investors and their
representatives" and (2) current and prospective depositors. Although
the primary concern of depositors is the safety of their deposits and not
the prediction of future earnings, they were included with investors as
primary users of reported earnings on the grounds that they share the
same interest in reported earnings as investors. This conclusion was
reached via the following argument. The best current indicator of the
degree of protection of depositors' funds is the bank's financial position
as described in the current statement of condition (balance sheet).
However, future financial position, and therefore future deposit protec-
tion, is, among other things, dependent on future earnings. Hence, it is
argued that, even though depositors may be most interested in the state-
ment of condition, the most relevant use that they make of reported earnings
is the prediction of future earnings.

Once "the prediction of future earnings" was specified as the
operational definition of relevance, it was possible to conceive of an
empirical study to determine which, if any, of the two earnings measures
was the more relevant. The purpose of such a study would be to determine
the rplétive capacipy of the two pgrnings measures to predict future

earnings. The outcome of such a study could provide new evidence that

 

15Robert T. Sprouse, "The Measurement of Financial Position and
Income: Purpose and Procedure," Research in Accounting Measurement, ed.
Robert K. Jaedicke, Yuji Ijiri.and Oswald Neilsen (Evanston, Illinois:
American Accounting Association, 1966), p. 106.

48

would aid in resolving the controversy if one earnings measure were found
to have a predictive capacity superior to the other. For, given such an
outcome, new support would exist for the presentation of the earnings
measure with the superior predictive capacity as the final reported
earnings figure on the grounds that it was the more relevant earnings
measure. If, of course, neither earnings measure was found to have
superior predictive capacities, no new evidence for resolving the contro-
versy would have been uncovered. Additional research would still be
needed in order to find an equitable solution to the controversy.

Further investigation indicated that it was possible to obtain
the data necessary to perform such a study as it relates to large,
publicly-held commercial banks. The methodology employed to perform
the study is described in the following chapter while the empirical

findings and conclusions are presented in Chapters V and VI, respectively.

3.4 Summagy

The purpose of this chapter was to present the results of a search
for new means of resolving the controversy. This goal was achieved as
follows. The existence of the belief in single-figure fixation and a
disagreement as to which earnings measure is the most relevant were shown
to be the primary and secondary reasons, respectively, why the controversy
remains unresolved. Specification of these reasons provided new criteria
that could be used in an attempt to resolve the controversy. For, (1) if
single-figure fixation could be found to be nonexistent there would be no
logical grounds for the controversy or (2) if one earnings measure were
found to be more relevant than the other there would be new evidence to
support the presentation of that earnings measure as the final reported

earnings figure.

49

Hence, two empirical studies were proposed, the first of which was
designed to test for the existence of single-figure fixation and the second
of which was designed to measure the relative relevance of the two earnings
measures. Unfortunately, implementation of the first study was found not
to be feasible. However, implementation of the second study was found
to be feasible once "the prediction of future earnings" was specified as
a surrogate for relevance. The second empirical study was formalized as
a study to determine the relative capacity of Net Operating Earnings and
Net Income to predict future earnings. The methodology and results of

such a study are presented in the chapters that follow.

CHAPTER IV

METHODOLOGY

4.1 Introduction

 

The purpose of this chapter is to present the methodology employed
in an empirical study to determine the relative capacity of Net Operating
Earnings and Net Income to predict future earnings.

In order to operationalize such a test it was first necessary to
perform the following steps: (1) specify the forecast objective to be
used as a basis for determining the relative predictive capacity of each
earnings measure; (2) specify the forecast models to be used in predicting
future earnings; (3) select a sample of banks from the population of large
banks and obtain the necessary data; (4) calculate the earnings per share
for each bank for each year; and (5) specify and obtain a measure of
forecast error. These steps are described in detail in Sections 4.2

through 4.6 as follows:

4.2 Specification of the Forecast opjective

The forecast objective used as a basis for determining the relative
predictive capacity of the two earnings measures has previously been
referred to only as the prediction of "future earnings". A more specific
definition was needed in order to operationalize a test of relevancy.
Consequently, all-inclusive earnings (i.e. - Net Income) was specified

as the relevant forecast objective since it provides the best measure of
a bank's overall performance. This conclusion was based on the contention

50

51

that, subject to the limitations of the accounting process, the economic
success or failure (i.e. - overall performance) of a corporate enterprise
is best measured by the change in retained earnings, adjusted for capital
and dividend transactions, over some period of time. The earnings
measure that most accurately measures the change in retained earnings
from one year to the next is all-inclusive earnings.

The above conclusion should not be interpreted as an indictment
against Net Operating Earnings as a measure of current year earnings.
The principle argument advanced in support of Net Operating Earnings as
a measure of current earnings is that by excluding certain items from
the measurement of a bank's current performance a better indication is
obtained as to the future performance of the same bank. Future performance,
however, involves more than just Net Operating Earnings. It encompasses
all items affecting stockholders' equity exclusive of capital and dividend
transactions. For, a bank's ultimate ability to earn a return for its
owners is dependent upon all revenue and expense items regardless of
whether they are identified as being nonoperating or extraordinary in
nature.

The following example should serve to illustrate the above point.
The illustrative income statements presented in Figures 1 and 2 (Section
1.2 of Chapter I) contained an extraordinary item identified in Figure 2
(the accountants' income statement) as a loss on the sale of a branch
building. The amount of this loss was excluded from "Net operating earnings"
as reported in Figure 1 (the bankers' income statement). Presumably the
reason for excluding it was that it represented a loss or expense that
was not likely to be incurred in the future. [Since a future occurrence

was unlikely, it could be argued that including such an amount in current

earnings could result in misleading conclusions being drawn about the

52

bank's future performance by those who were unaware that such an item was
included in the reported earnings figure.

Excluding an extraordinary loss from current earnings does not
mean, however, that a current or prospective investor is not concerned
about the bank's incurrence of extraordinary gains or losses. For example,
assume that one or two years prior to the incurrence of the extraordinary
loss a current or prospective investor was in the process of evaluating
the bank's potential. He certainly would have been more interested in a
predictive earnings measure that forecasted an earnings figure that
included the loss than one that did not. For, such a loss will undoubtedly
have some affect on the bank's performance subsequent to the year of its
incurrence since, had the loss not been incurred, the bank would have had
a greater net asset base from.which to expand or maintain future operations.

Therefore, it was concluded that the only rational objective of a
strictly predictive earnings measure such as Net Operating Earnings is the
prediction of all-inclusive earnings (i.e. - Net Income) not the prediction

of a predictive income figure.1

 

1One argument for predicting Net Operating Earnings rests on the
notion that income cannot exist except as it is reported. Furthermore, it
may be argued that market values are determined on the basis of reported
earnings since, to the investing public, income i§_the amount reported by
the bank as earnings per share. In the short run there may be some validity
to this position. Therefore, it might be considered worthwhile to compare
the ability of each income measure to predict future earnings of the same
series. It could then be argued that the measure determined to be the better
predictor should be the reported earnings figure.

However, one wonders about the value of obtaining empirical evidence
in support of an income measure that may knowingly mislead investors. Such
could be the case if Net Operating Earnings was found to be the best self
predictor and at the same time a poor predictor of all-inclusive earnings.
In this case, it would be entirely possible for Net Operating Earnings to
show steady improvement while retained earnings were being dissipated by
large security and loan losses. Ultimately, perhaps through a reduction in
dividend payments, the market would recognize the bank's "real" financial
position. Losses to current stockholders could be severe. Since the purpose
of this research is to obtain an empirically supported equitable solution
to the reporting controversy, no attempt will be made to predict future Net
Operating Earnings.

53

It may appear at first glance that, a priori, Net Income will be
selected as the better predictor since by predicting itself it would be
considered an unbiased predictor whereas Net Operating Earnings would
not. Net Income would be considered an unbiased predictor on the basis
that, ceteris paribus, over a large number of consecutive annual forecasts
for a given bank, the expected value of the forecast error associated
with Net Income as a predictor is equal to zero.2 Net Operating Earnings,
on the other hand, would not be considered an unbiased predictor since
the expected value of the forecast errors associated with it as a predictor
is not equal to zero due to the fact that it does not contain certain items,
namely security gains and losses and a provision for loan losses, that are
included in Net Income -- the figure it is being used to predict.

Although the above statements as to the biasedness or unbiasedness
of the two earnings measures are correct, the following statistical
analogy is presented to show that lack of bias alone is not sufficient
grounds for drawing the a priori conclusion that one earnings measure is
superior to the other as a predictor.

Suppose we have two estimators y and z of the populapion mean

X. Let y be an unbiased and 2 be a biased estimator of X. Which
estimator should we use? __

We should use the one that we may expect to be closer to X.

Let us explain this statement with the following illustration.
Suppose the distribution of y and 2 may be shown diagrammatically
as in Figure 4. As shown, the dispersion of z is much smaller
than that of y, except that z is distriputed around E(z) = X + b
whereas y is distributed around E(y) = X. However, we can see
intuitively that if the bias b is small, even though 2 is biased,

because its dispersion is much smaller than that_of y, it will
probably give us an estimator that is closer to X.

 

2The "expected value of the forecast error for a given bank over
a large number of consecutive annual forecasts" should not be confused with
the "expected value of the forecast error for a given year over a large
number of banks." The former concept relates only to a single bank while
the latter is concerned with several banks.

 

 

 

 

b
_ _ Z
x E(z)=x+b
_, y
E(y)=X
FIGURE 4

GRAPHIC ILLUSTRATION OF BIAS AND EFFICIENCY

The implication of this reasoning is that a biased estimator
may be better than pp unbiased one, dppendin on the magnitude of
the variance and the bipp, (Emphasis added.) .
Since it is known that Net Operating Earnings is a biased predic-

tor of Net Income, the question of whether it is a better predictor
depends upon how much more efficient a predictor it is than Net Incomw --
i.e., how much lower is the variance associated with it than Net Income.4

A priori, Net Operating Earnings would appear to be a more efficient

predictor for short-run (i.e. - one-year) forecasts since, as has been

 

3Taro Yamane, Elementary Sampling Theory (Englewood Cliffs, N.J.:
Prentice Hall, Inc., 1967), p. 61.

4Once again it should be noted that the variance referred to here
pertains to the "variance of the forecast errors for a single bank over a
large number of years" and pp£_"the variance of the forecast errors for
a given year over a large number of banks."

55

pointed out in Chapter III, the inclusion of security gains and losses and
a provision for loan losses in the determination of Net Income is likely
to make Net Income a more volatile figure than Net Operating Earnings.
Granted that Net Income is likely to be the more volatile of the two
earnings measures, it follows that the variance of the annual forecast
errors associated with it as a predictor are likely to be greater than the
variance of the errors associated with Net Operating Earnings. However,
whether the presumed greater efficiency of Net Operating Earnings as a
predictor is sufficient to overcome its biasedness as a predictor can be
determined only by empirical test and not by a priori reasoning.

When the annual algebraic forecast errors for a given bank are
aggregated over some long-run period (i.e., two or more years) in arriving
at a multi-year forecast error it would be expected that as the number of
years included in the multi-year forecast-error-period became longer and
longer Net Income would eventually become, if it already was not, the
superior predictor. This conclusion is based upon the following argument.
As the number of years in the forecast-error-period is increased the
variance of the multi-year forecast errors should decline since more and
more annual overforecasts and underforecasts are allowed to offset each
other thereby reducing the size and range of the multi-year forecast errors.
As the efficiency of each earnings measure improves, bias becomes a more
important factor in determining the better predictor. As an unbiased
predictor, Net Income should, if it already is not, become a better
predictor when the efficiency of Net Operating Earnings as a predictor is
not sufficient to overcome the biasedness of Net Operating Earnings.
However, whether Net Income will become, if it already is not, a superior

predictor when the multi-year forecast-error-period contains two years,

56

four years, or ten years can only be determined by empirical test.

4.3 Specificglion of Forecgpt Models

In order to obtain forecasts of future all-inclusive earnings it
was necessary to specify a forecast model or models to which earnings
data from prior years could be input. Six such forecast models were
employed in this empirical study. The purpose of each model was to predict
"current" year earnings-per-share. These forecast models are identified
and described as follows:

81: Forecast earnings equal earnings for last year.

32: Forecast earnings equal earnings for last year
plus the dollar change in earnings of last year

over the preceding year.

S3: Forecast earnings equal earnings for last year
plus the same percentage change in earnings of

last year over the preceding year.

R1: Forecast earnings is the dependent variable derived
from the linear equation y = a + bt where t is the
current year and a and b are parameters determined
by regressing actual earnings on time for the ppppg

preceding years.

R2: Same as R1 except that the parameters are determined
by regressing earnings on time for the four preceding

years.

57

Same as R1 and R2 except that the parameters are
determined by regressing earnings on time for the

five preceding years.

In formula form these forecast models appear as follows:

81:

FEt = Et-l
where FE = forecast earnings-per-share
E = actual earnings-per-share
t = current year
(E )
FE = E __t_'.l_

t t-I (Et-Z)

FE a + bt

where: a and b = parameters determined from a
simple, least squares linear regression
of actual earnings on time for the years
t-4 through t-l

For example, using this forecast model, earnings

data for the years 1959-61 served as inputs to

determine the parameters of the forecast equation
which in turn was used to predict 1962 earnings-

per-share.

Same as R1 except that the regression parameters
are based on actual earnings for the years t-5

through t-l.

Same as R1 and R2 except that the regression

58

parameters are based on actual earnings for the

years t-6 through t-l.

Four of the six forecast models were suggested by other research
studies. Models 81, $2 and S3 were developed and used by Professors
Green and Segall in their studies of the predictive power of first quarter
earnings reports.5 Model R3 was used by Professor James Parker in his
study of the predictive capacity of current operating and all-inclusive
earnings.6 Models R1 and R2 are merely variations of Parker's model.

A model or models that assumed more sophistication on the part of
investors could have been developed. For example, models employing longer
base periods or affixing different weights to the various years in the
base period could have been designed. Or models that employed additional
independent variables such as deposit size, interest income, security
gains and losses, etc. could have been devised. However such models were
rejected as being inconsistent with the notion of the naive investor as
implied in the single-figure fixation concept.

Earlier it was shown that there was a primary and secondary reason
for the controversy. The primary reason was the existence of the belief
that investors in bank stocks have single-figure fixation. The secondary
reason was the disagreement between bankers and accountants as to the most

relevant single earnings measure. Furthermore it was disclosed that

 

5David Green, Jr. and Joel Segall, "The Predictive Power of First-
Quarter Earnings Reports," The Journpl of Business, XL, No. 1 (January,
1967), pp. 44-55, and "The Predictive Power of First-Quarter Earnings
Reports: A Replication," Empirical Research in Accounting: Selected Studies,
1966, (supplement to Volume 4 of the Journal of Accountinngesearch),
pp. 21-43.

6James E. Parker, "A Study of the Predictive Significance of Several
Income Measures Relative to the Accounting for Extraordinary Items and Prior
Period Adjustments" (unpublished Ph.D. dissertation, Graduate School of

Business Administration, Michigan State University, 1969), p. l.

59

neither the controversy nor the secondary reason would exist if it were
not for the primary reason. Hence, any attempt to resolve the controversy
by evaluating the secondary reason must be based on the assumption that
single-figure fixation exists. For, as was pointed out in Chapter III,

if it does not exist, there is no need to test for relevancy since it no
longer matters which of the two measures is the more relevant.

Granted the assumption as to the existence of single-figure
fixation, it seems logical to conclude that investors who focus almost
solely on a single earnings-per-share figure merely extrapolate the trend
of earnings from some prior period. Therefore, the forecast models were
purposely kept "simple".

Mention of one limitation is appropriate at this point. The results
of any study employing the criterion of predictive capacity are dependent
in part upon the particular forecast model(s) employed. Positive results
represent a joint confirmation of both the model(s) and the measurement
under study. Negative results, however, may be due to an error in either
or both elements. Hence, it is possible to argue that use of a different
model could have turned the results of a given test around. In fact,
regardless of how many models are employed, it is always possible to argue
that some untested model would give the opposite results. The only defense
against this argument is an intuitive appeal based upon the reasonableness
of the forecast model(s) used. Such an appeal is made in the case of the

models discussed above.

4.4 Sample Selection gnd Data Source
Data necessary to compute Net Operating Earnings and Net Income
were obtained from 26 large, publicly-held commercial banks for the

twelve-year period ending December 31, 1968. The period was limited to

6O

twelve years because of a lack of publication of annual reports containing
detailed financial statements prior to 1957. These banks were selected in
descending order of asset size on the basis of the following criteria:
(1) publication during each year in the study of an annual report
containing detailed financial statements, and (2) availability of data
in the annual report or through correspondence to compute Net Operating
Earnings and Net Income. Appendix A contains a list of the banks included
in this study showing their ranking in Fortune's "50 Largest Banks" for
1968 and their total assets at December 31, 1968.2

Because all but the largest banks were rejected for failure to
meet one or both of the above criteria, the sample was not random. Of
course, this means that the conclusions of the study are not generalizable
to the entire population of commercial banks. However, since the largest
banks account for a significant portion of total commercial bank assets,
the results should have a significant bearing on the resolution of the
controversy. The following data illustrate this point. There were
13,698 commercial banks in the United States at the end of 1968.7 Total
assets of these banks amounted to $504,637,017,000.8 The total assets at
December 31, 1968 of the 26 used in this study were $144,222,000,000.
Hence, while the banks used in this study represent less than .2% of all

commercial banks they accounted for over 28% of total bank assets in 1968.

4.5 Calculation of Earnings Per Share

Since Net Operating Earnings per share and Net Income per share

 

7Annual Report of the Federal Deposit Insurance Corporation: 1968
(Washington, D.C.: Federal Deposit Insurance Corporation, May 1, 1969),
p. 179.

81bid, p. 184.

61

were the basic inputs of the forecast models it was necessary to convert
all the original financial data obtained into comparable earnings per
share figures. The first two steps in this conversion process consisted
of developing a working definition of Net Operating Earnings and Net
Income. The third step consisted of the computation of a "normal
provision for loan losses" since such an amount was not provided by the
banks. The fourth and final step eliminated the effects of stock splits
and stock dividends on the average shares outstanding during the period

of the study.

4.5.1 Computation of Netlgperatinngarnings

The figure described as Net Operating Earnings or Net Operating
Income in the income statement contained in the annual report was used as
the basis for the determination of Net Operating Earnings as used in this
study. For most banks for most years there was no need to make any
adjustments to this reported figure. However, each annual report was
reviewed to determine if the reported Net Operating Earnings figure
included amounts considered to be (1) appropriations of retained earnings,
(2) security gains or losses, (3) amortization of security gains and losses,
(4) a provision for loan losses, and/or (5) tax adjustments related to any
of the previous four items. In each incident where amounts of the above
nature were discovered, reported Net Operating Earnings was adjusted to
exclude such amounts.

Fortunately, in all instances where items (1) and (3) were
encountered, the amount of the applicable income tax adjustments, if any,
was disclosed in the annual report. Hence, it was not necessary to make
any estimates as to the tax adjustment. In most instances where items (2)

and (4) were discovered, the applicable tax adjustment was provided. On

62

the few occasions when the tax adjustment was not given it was estimated
as follows:
(A) For security gains, it was determined by multiplying
the amount of the gain by the capital gains rate in
effect during the year of the reported gain.
(B) For security losses and loan loss provisions, it
was computed by multiplying the amount of the
security loss or loan loss provisions by the
marginal tax rate in effect during the year the
amounts were reported.
The reasons for excluding each of these five items was to make the
annual Net Operating Earnings of all banks included in the study as

comparable as possible in terms of the treatment of each of the five items.

4.5.2 Computapipp of Net Income

Net Income was computed by adjusting Net Operating Earnings as
defined in the preceding section to include (1) security gains or losses,
(2) a "normal" provision for loan losses,9 (3) other appropriate nonopera-
ting revenue and expense items excluded from Net Operating Earnings but
reported in the annual report, and (4) tax adjustments related to any of
the preceding three items.

"Other appropriate nonoperating revenue and expense items" included
all remaining noncapital adjustments to stockholders equity except amounts
determined to be appropriations of retained earnings. Such noncapital
adjustments included extraordinary items and prior period adjustments.
Also included were revenue and expense items such as pension past service

costs which had previously been excluded from Net Operating Earnings.

 

9A definition of ”normal" will be presented later.

63

Examples of appropriations of retained earnings which were ignored in
determining Net Income are transfers to and from both unspecified
"contingency" reserves and security valuation reserves.

On the rare occasion when the tax adjustment attributable to item
(1) was not disclosed in the annual report it was computed on the same
basis as disclosed in (A) and (B) of the preceding section. The determina-
tion of the tax adjustment attributable to item (2) is discussed in the
following section. When the tax adjustment applicable to item (3) was
not disclosed, it was computed by applying the marginal tax rate in

effect in the year of the adjustment.

4.5.3 Computation of a Normal Provision for Loan Losses

Previous discussion in Chapter II pointed out that the total tax
deductible provision for loan losses is intended to allow banks to accumu-
late a reserve for two types of loan losses. The first type was identified
as those of a normal recurring nature. The second type was identified as
those which occur at infrequent, irregular, and unpredictable intervals
such as during a depression or severe recession.

Both bankers and accountants have agreed that the portion of the
provision applicable to the second or contingency type of loss should not be
charged against current earnings. Of course, the means which each side pro-

poses to utilize in excluding this amount differ, with bankers favoring the

exclusion of the total provision for loan losses while accountants prefer

to exclude only the contingency portion. Aside from the arguments presented
by both sides it is apparent that (1) to include the total tax deductible
provision in Net Income would result in a consistent understatement of
earnings and (2) to exclude the total tax deductible provision from Net

Income would result in a consistent overstatement of earnings. This

64

conclusion is supported by the fact that the total tax deductible provi-
sion for loan losses of Federal Reserve member banks was more than double
the actual loan losses of these banks for the period from 1957-68.

(See Table I).

Therefore, for purposes of this research, the net of tax excess
of the tax deductible provision for loan losses over a "normal" provision
was excluded from the determination of Net Income. This excess was
treated as an appropriation of retained earnings since its purpose is
to provide a contingency reserve for some indeterminable future loss.

The tax reduction applicable to the excess was also excluded from Net
Income since it was treated as a deferred tax credit.

While it made the determination of Net Income more difficult and
involved the utilization of estimates, the computation and use of a normal
provision for loan losses was considered necessary if a realistic Net
Income figure was to be computed. It is recognized that the use of
estimates places limitations on the conclusions that can be drawn from
the study. However, failure to use a Net Income figure that excluded the
excess provision would subject the results to the far greater criticism
of being irrelevant. Therefore, the normal provision for loan losses for
a given year was considered to be the average of the actual loan losses
for a three-year period encompassing (l) the current year, (2) the preceding
year, and (3) the subsequent year.

One exception was made to this procedure. Since the actual loan
losses for 1969 were not known at the time this study was implemented, the
normal provision for loan losses for 1968 was considered to be the average
of the actual loan losses for 1968 and 1967.

This method of estimating a normal loan loss provision falls

65

somewhere between the two methods originally proposed by bankers and the
bank regulatory agencies following the July 8, 1969 compromise. The
first method required that the minimum charge be "equivalent to the five-
year average ratio of losses computed on the basis of net charge-offs to
total loans."10 An additional amount based on management's judgment
could be provided so long as "adequate disclosure of such discretionary
action was included in a referenced footnote." The second method allowed
banks that were on the charge-off basis of recognizing loan losses to
charge current operations with an amount equal to the actual charge-offs
for the year.11

Subsequent to the issuance of instructions describing these two
methods, a third method has been approved. This method allows banks to
determine the minimum charge on the basis of a five-year average moving
forward. This five-year average would involve using actual charge-offs
for 1969, a two-year average in 1970, and so on until a five-year average
was built up in 1973.12 Presumably, support for this method came from
those banks that have had heavy loan losses in the past five years, but
have improved the quality of their loan portfolio.

Each of these three methods and the one proposed for use in this
study are designed to provide a surrogate for actual loan losses that will
result from loans currently outstanding. Obviously, some estimate is
necessary since the actual loan losses will not be known until sometime

in the future.

 

10Letter from J. T. Watson, Acting Comptroller of the Currency, to
the presidents of all national banks, Washington, D.C., July 24, 1969.

llIbid.

12Merrill Lynch, Pierce, Fenner & Smith Inc. wire flash A 918,
November 4, 1969.

66

Sound arguments can be presented in support of each proposed
method. For example, if a bank has had little or no change in its
lending policy or the quality of its loans outstanding for the past
five years, then the use of a past five-year ratio based on actual losses
may well provide the best estimate of future losses. However, if there
has been a change in the bank's lending policy during the past five years
and the new policy is expected to remain in effect for the next few years,
then the use of the five-year average moving forward is likely to prove
the better estimator. The use of actual losses for the current year
can be justified when the balance in the loan account has remained fairly
constant over the last few years and there has been little or no change
in the quality of outstanding loans.

The three-year average of the current, preceding and subsequent
years used in this study represents a compromise of the three methods
discussed above. The three-year average can be defended on the following
grounds.

1) The period involved is long enough to diffuse the

effects of unusually large or small write-offs that
may occur in any given year.

2) The period involved is short enough to give effect
to changes in the size of the balance of loans
outstanding.

3) The use of the "subsequent" year gives some
recognition to the effect that changes in (l)
the size of the loan balance and (2) lending
policy would have on future loan losses.

4) It recognizes that management is in the best

67

position to determine what a normal provision for
loan losses should be. The use of the "subsequent"
year serves as a proxy for management's estimate 0f
future loan losses.

While the three-year method can be justified on theoretical grounds
as well as any of the other proposed methods, practical considerations
also dictate its usage in this study. Due to the lack of reported data
for years prior to 1957 this study was limited to the twelve-year period
ending in 1968. Excluding the method used to determine the normal provision
for loan losses, the number of forecast years varies from seven to eleven
depending on the forecast model used.

By obtaining the actual loan losses for each bank for 1956 and
by using the two-year average method for 1968 instead of the three-year
average it was possible to keep the range of forecasts at seven to eleven
years.

Use of either of the five-year methods would have reduced the
range of forecast years by two years to five to nine years. Naturally,
the validity of this study is dependent upon the number of years for
which results can be obtained. The larger the number of years the more
valid are the conclusions. Since the arguments for using either of the
five-year methods are not superior to those of the three-year method,
the three-year method was used since it resulted in a greater number of
forecast years.

The tax reduction attributable to the normal provision for loan
losses was deducted from the computed normal provision to arrive at the
net of tax normal provision. The amount of the tax adjustment was

determined by applying the marginal corporate income tax rate in effect

68

in the year for which the provision was computed.

4.5.4 Adjustment for Stock Dividends and Splits

Since all earnings data used in the forecast models were in the
form of earnings per share it was necessary to adjust for stock dividends
and stock splits that occurred during the period of the study. Accordingly,
an adjustment factor was computed for each bank for each year in the study
and was applied to the amounts previously reported as (or determined to
be) earnings-per-share for the year. The year 1969 serves as the standard
of comparison. The following formula was used to determine the adjustment

factor for a given year:

 

 

 

AF = 1 o 1 o o o 1 o 1
19Y-1 1 + P1963 1 + P19Y+1 1 + Plgy
where: AF = Adjustment factor
Y = year in which the stock dividend or split
occurred
P = number of new shares issued as a stock

dividend or stock split expressed as a
percentage of the number of shares
outstanding prior to the dividend or

split

4.6 Specification of a Forecast Error Measure

The measure of forecast error used in subsequent evaluations of
the relative predictive capacity of the two income measures was the percent-
age (relative) forecast error. The use of a percentage forecast error
measure was selected over an absolute forecast error measure (forecast

earnings-per-share minus actual earnings-per-share) in order to avoid

69

distortions due to differences in bank size and income levels when
aggregating error measures. One weakness of the percentage error measure
is that it assigns great weight (when aggregating error measures) to small
absolute errors when earnings-per-share are close to zero. This deficiency
in the percentage error is not too serious if earnings-per-share are close
to zero because the number of outstanding shares is large; but if earnings-
per-share are close to zero because of an abnormally bad year, heavy
weights for small errors can be misleading. Fortunately, none of the
earnings-per-share figures used in this study came close to approaching
zero nor were any of the earnings-per-share figures negative. Therefore,
the deficiency associated with the use of a percentage error measure was

not of concern.

4.7 Summary

The purpose of this chapter was to present the methodology employed
in an empirical study to determine the relative capacity of Net Operating
Earnings and Net Income to predict future earnings. Since such a purpose
has been achieved it is now possible to turn to a discussion of the
empirical findings derived from the implementation of this methodology.

The following chapter presents such a discussion.

CHAPTER V

EMPIRICAL FINDINGS

5.1 Introduction

 

The purpose of this chapter is to present the empirical findings
of this research study. To meet this end, the chapter is divided into
five major sections. Section 5.2 identifies the two criteria used in
evaluating the forecast errors. Sections 5.3 and 5.4 present the results
according to each criteria. Section 5.4 discusses the possible cause of
an observable trend in the forecast errors. Finally, Section 5.5
summarizes the discussion and observations about the empirical findings
presented in the preceding sections. All conclusions are reserved until

Chapter VI.

5.2 Identification of Evaluation Criteria

The relative predictive capacity of the two earnings measures was
evaluated according to two concepts of user needs. Under the first or
short-run concept, the forecaster of future bank earnings is viewed as
being most interested in that earnings measure which is most likely to
produce the lowest forecast error for a given bank for the subsequent
year. Under the second or long-run concept, the forecaster of future bank
earnings is viewed as being most interested in that earnings measure which
is most likely to product the lowest aggregate forecast error for a
given bank for some future long-run period ranging, for purposes of this

research, from two to seven years.

70

71

Two specific criteria were developed from these two concepts in
order to evaluate the percentage forecast errors generated by each of the
six forecast models employed in this empirical study.

Short-run criterion -- Under the short-run criterion, the

earnings measure which produced the lowest absolute annual
forecast error was designated the best predictor of future Net
Income.

Long-run criterion -- Under the long-run criterion, the
earnings measure which produced the lowest absolute forecast
error over a given multi-year time period was designated as
the best predictor of future Net Income.1

As indicated in each of the criteria, the absolute percentage
forecast error served as the basis for comparing the predictive capacity
of the two earnings measures. Accordingly, an earnings measure which
produced an individual or aggregate percentage forecast error of -5%
(i.e. - an underforecast of 5%) was considered to be no better or no
worse than the earnings measure which produced an individual or aggregate
forecast error of +5% (i.e. - an overforecast of 5%). In other words,
there was assumed to be no difference in the penalties associated with
overforecasts and underforecasts of comparable size.

The empirical findings of this study are presented separately

according to each criterion in Sections 5.3 and 5.4 as follows.
5.3 Results Analyzed According to the Short-Run Criterion

5.3.1 Aggregation Procedures

The basic measure used to evaluate the relative predictive capacity

 

1The terms "long-run" and "short-run" are used in this chapter
solely to facilitate the presentation and analysis of the results of this
study. It is recognized that the terms are relative in nature and that
what is considered to be "long-run" by one person or in one instance may
be considered to be "short-run" by another person or in another instance.
Consequently, it should be noted that pp assertion is made that the short-
run is one-year or that the long-run varies from two to seven years.

72

of the two earnings measures under the short-run criterion was the
absolute value of FEij - the percentage forecast error for bank 1 in

forecast year j. FEij was calculated according to the following formula:

FEi. = F1 ' Ai'
J Aij
where: Fij = forecast earnings for bank i in year j
Aij = actual Net Income for bank i in year j

Table III shows, for the 26 banks included in the study, the
mean absolute percentage forecast error for (1) each forecast year, (2)
selected three-year periods, (3) the 1962-68 period, and (4) the total
period covered by each forecast model. The latter two periods were
included for the following reasons. Although the data used in the study
encompassed twelve years (1957-68), the number of forecast years covered
by the forecast models varied because of differences in the number of
base years employed by the six models. For example, model 32 required
two-years' input before it could be used to forecast current earnings
whereas model R3 required five-years' input. Hence, while model 32 could
be used to forecast earnings beginning in 1959, model R3 could not be
used until 1962. The 1962-68 period contains the maximum number of
forecast periods that was common to all of the forecast models. Data for
the 1962-68 period were included in Table III and subsequent tables for
use when comparing performances across all forecast models. Data for
the total period covered by each forecast model were presented for use
when analyzing a single forecast model.

Table II shows the number of base years and forecast years asso-

ciated with each forecast model along with additional information about

73

TABLE II

SELECTED INFORMATION ABOUT THE FORECAST MODELS

 

 

 

 

Number of Base Number of Number Total Number
Forecast Years Employed Forecast of of Forecasts Period
Model by Forecast Model ngrs Banks (col.3 x col.4) Covered
SI 1 ll 26 286 1958-68
32 2 10 26 260 1959-68
S3 2 10 26 260 1959-68
R1 3 9 26 234 1960-68
R2 4 8 26 208 1961-68
R3 5 7 26 182 1962-68

 

each model. However, for a description of the forecast models themselves,
the reader should refer to Section 4.3 of Chapter IV.
Each annual mean absolute percentage forecast error figure (Mj)

shown in Table III was determined according to the following formula:

 

m
_ , E IFEijl
M, = 1 - l
J m
where: m = 26, the number of banks
FEij = the percentage forecast error for bank i

in forecast year j

The mean absolute percentage forecast error for each three-year period,
the 1962-68 period, and the total period covered by each forecast model
was calculated by averaging the annual means (M5) for the period in question.
For example, the total-period mean absolute percentage forecast error for
model R1 was computed by averaging the annual mean forecast errors (M3)

for the years 1960 through 1968 whereas the 1962-68 mean was obtained by

74

averaging the annual means for the years 1962 through 1968.

Since the reliability of an average forecast error depends not
only on the size of the forecast error but also on the degree of dispersion
of the individual forecast errors around this average, the mean deviation
of the absolute percentage forecast errors was computed. Table IV shows,
for the 26 banks included in the study, the mean deviation for periods
identical to those presented in Table III. The annual mean deviation

figures (M5,) were determined according to the following formula:

 

m —
1: lAFE.. - M
MD =
j m
where: m = 26, the number of banks
AFEij = [FEijl , the absolute percentage forecast
error for bank i in forecast year j
M5 = mean absolute percentage forecast error in

forecast year j

The mean deviation for each three-year period, the 1962-68 period, and the
total period covered by each forecast model was calculated by averaging
the annual mean deviations (M5,) for the period in question.

One other way of analyzing the results of this study is to tabulate
the number of banks, in any given period, for which each of the two earnings
measures produced the lowest annual absolute forecast error. Table V
presents the results of such a tabulation for periods identical to those
presented in Tables III and IV. The principal advantage of this analytical
method is that it is simple, straight-forward and presents the results in
a manner that is more easily comprehended and assimilated than the two

previously mentioned methods. The primary disadvantages are (1) that it

fails to provide any information about the magnitude of the forecast errors

 

'75

N m N o N n 0 od 0 OH m w owoucoouom umo3oq
sum: muse» mo sagasz

 

 

 

 

 

 

 

sm.e~ ue.aa so.oa N¢.- x~.- eq.m~ so.m3 N~.na so.mm wa.3s _a.m~ s~.na cam: nonsma-amsoa
m.sa e.a~ a.na ~.- w.sa m.~a ~.m~ «.ma o.s~ m.ma s.m1 4.0 wo-~oaa
n.3n n.- 3.mE m.n_ a.¢~ a.m~ m.m~ n.8a m.- o.ma “.3a a.a mo-soaa
3.oa ~.m~ o.os 1.3a m.sa o.3a m.e~ ~.ms a.o~ o.~a a.ss o.m moms
k.ea o.~ m.oa o.a m.o~ o.oa s.m~ o.m~ o.3~ a.m~ n.ss o.m “can
k.ms o.ma «.mH ~.ms n.8a m.as ~.e~ ~.ma o.- n.ms m.mH a.nz coma
3.3n e.os m.na m.oa m.m~ ~.sa o.ma k.aa m.mn m.aa m.0a m.a mo-nooa
m.3s m.aa m.ma s.os m.na a.os ~.oa o.m~ n.8a m.~a s.~a 3.0s moms
a.ws ~.a_ m.~s m.~a o.ns m.an 3.mc k.aa w.q~ m.aa s.m m.oa some
s.os o.m o.ms m.m o.ns ~.as ~.3~ m.o~ ~.m~ ~.oa o.m o.m moms
a.nm c.3a 0.38 8.4a H.3n n.3a m.on H.mH No-08aa
an.ma a~.az m.ma m.~a s.mm o.aa n.m¢ m.oa «.ms m.oa N.ma m.m Noah
pa.o~ am.ss a.mm ~.ma «.ma n.8a q.um o.ma m.k~ n.3a soon
e~.0m sm.ma m.~m o.oa o.ko n.8a 4.3m 0.8s coon
ak.¢mH N~.m~ ea.soa Ne.e~ a.ao 3.8N ammo
e~.m~ so.o~ mass
ceases
Hz hex Hz - moz Hz moz Hz moz Hz moz Hz moz magmas: «sanctum
me .e He mm Nm am specs announce

 

 

m<my uzo OH A<Dom QOHMmmImOMdmon<Umeh "woman Hm<om¢0h mu<H2mUmwm MHDAOmm< z<mz

HHH m4m<H

 

 

m a n m
nk.a em.“ , an.oa Na.w
s.o w.“ m.m o.~
N.a m.m m.m m.m
a.“ m.w a.m o.o

m.o o.m m.m o.m
o.oa m.aa x.oa m.an

m.m ~.m k.o 8.x
o.m 0.x a.“ m.w
m.aa m.o o.o ~.m
o.m m.n ~.~a o.m
N¢.Ma x3.n k.m m.o
e3.ka rm.m
: moz Hz moz

me we

 

 

H m
wq.m~ No.0
~.- 0.0
0.0L 0.0
0.0 0.0
0.0 0.5
o.m~ q.~H
L.w 0.0
m.“ ~.0
0.0 0.0
n.m 0.x
m.HN A.m
0.HN 0.0
o.Hm o.o~
NN.~N wo.o~
Hz mOz

am

N0.mm No.0H

0.5H m.0
o.om o.-
~.HN 0.“
~.¢H N.H~
¢.q~ 0.m~
o.- N.0
0.0 0.0
L.m 0.m
q.q 0.0
m.m1 0.0
o.0~ 0.0
o.om o.HH
~.oN 0.0

No.00 wo.0~

 

H:

m x
mm

 

H o
Na.m~ N0.o
n.0fi 0.0
o.- m.o~
~.m~ m.m
o.0H ~.oH
m.m~ m.m~
m.- N.w
¢.~H m.0
0.0 0.m
0.0H 0.0
0.00 0.x
0.0m m.q
0.0m 0.0L
0.0m 0.0
$0.00 N0.NH
Hz moz

Nm

N0.m~ No.0

m.0 o.0
~.w m.“
0.0 H.0
m.0 ~.m
0.0 m.o~
~.0 o.“
m.0 0.0
m.m 0.5
0.0 o.¢
¢.mH ¢.m
o.m~ H.m
m.om 0.0L
0.0a ~.-
H.0m n.md

N~.ma N0.~H

 

owouCoouom amazon
Lugs muse» mo nersz

couuow>eo coax
omouo>< vowuomufiouOH

00u~00~

m0:00m~

00ma
~0ma
00mH

m0um0m~

n0m~
q0mH
m0m~

m0uo0m~

Noma
H0mH
O00H

ommfi
mmofi

 

nownom

nonsmooz awnucumm
dove: umooouom

 

 

m<mw mzo OH A<Dom OOmem momdm Hm<0md0m "mmoxdw Hm<omdom mu<Hzmummm MHDAOmm< mo ZOHH<H>MQ z<m2

>H mam<w

77

 

 

N 0
0mm.» Nu m
0“ mod
0m 000
on «0
00 La
0 ON
Ma 00
NM 00
AH 00
m 00
«H NH
ML 00
Hz mOz

me

 

m 0
,0m vac
00 mm“
mm ROM
00 mm
50 0
0 0H
0H 0
mm mm
0 NN
«H NA
m 0H
m 0H
m 0N
Hz 002

he

 

N m
m0m N00
N0 N00
00 Nod
00 Nm
0L 0
OH 0“
0H 0
Lm n0
0 m~
HA m0
AH 00
mg 00
m mN
0 AN
n 00
Hz 002

 

O 0
NON Nam
cm 004
N0 ONL
an no
B 0g
LL 00
00 ML
0N mm
HA ma
0 0H
0 ON
00 m0
0 ON
0 ON
0 MN
0 MN
Hz 002

mm

 

O 0
no 00A
O0 NNH
0N 00
0 0L
0 00
mg 00
0N cm
AL 00
N 00
0 ON
00 00
m 0H
0 ON
N «N
m MN
Hz 002

~m

am

 

H 0
N00 yo:
00 00A
N0 mad
RN #0
0 ma
0 NN
«a NH
Hm mo
0 ON
00 00
NH 0H
0H 00
0 NH
0 00
N 0N
q NN
R 0a
Hz 002

.snsafiam :ons
momma mo peasam

ucounum o no
maouoe ocmno

mamuOH fiasco

001N00H

0010000

0000
NO0L
0000

0010000

000H
0000
M000

N01O00H

NO0H
H000
O00H

000A
0m0~

 

possum

sesame: mmnacuom
#0002 umooouom

 

 

m<mw 020 OH 44000 OOmemnmommuuHm<quow “00000 Hm<ommoh 0H34000< Hmm304 NIH 00030000 mdbm<mz moszm<m 1040 :UHm3 000 mMde 00 mmmZOZ

> 040<H

78

and (2) that it gives no weight to the difference between the two forecast
errors being compared (i.e. - no consideration is given to the fact that

a forecast error associated with a given earnings measure may be lower

by .1% or 100%). As a result of the second deficiency, the final results
obtained under this method could be misleading if the number of low
forecast errors associated with one earnings measure are attributable to
small margins while those attributable to the other measure are based upon
large differences. A comparison of the annual and periodic results
presented in Table V with those shown in Tables III and IV indicates

that such a deficiency does not exist since, from a relative standpoint,

17") . '

 

the results presented in all three tables give rise to comparable conclu- 3
sions.
Since the primary purpose of this study is to determine the
relative predictive capacity of the two earnings measures and since the
results presented in Table V are not biased, Table V is presented as a
useful summary of the results of this study when evaluated under the

short-run criterion.

5.3.2 Observations

The following observations are based upon the data presented in
Tables III, IV and V and are divided into four categories: (1) overall
results, (2) annual results, (3) comparison of forecast models, and (4)
relationship between the rank of the means and the rank of the mean
deviations.

Oveggll results-- On gn oveggll basis,_Net Operpting,E§;pipgs
showed the best_predictive capacity. This statement is based on the fact

that the total-period means and 1962-68 means ot Table III and the total-

period average mean deviations and 1962-68 mean deviations of Table IV

79

were lower for forecast errors associated with Net Operating Earnings
regardless of the forecast model employed. Similarly, the grand totals

and the 1962-68 totals of Table V show that a majority of annual individual
bank forecast errors were lowest for all forecast models when Net Operating
Earnings was used to predict Net Income.

Annual results-- Although the annual results do not favor Net

 

Operating Earnipgs over Net Income as overwhelmingly as do the overall
results, Net Operating Earnings still showed the best predictive capacity

on an annual basis. This observation is based upon a review of the annual

 

forecast errors for each model and is supported by the data presented on
the bottom line of Tables III, IV and V. For Tables III and IV, the
bottom line indicates the number of forecast years for a given forecast
model for which each of the two earnings measures produced the lowest
mean absolute percentage forecast error and mean deviation, respectively.
For Table V, the bottom line indicates the number of forecast years for
a given model for which each earnings measure was associated with the
majority of banks having the lowest absolute forecast error. The bottom
line of the three tables also reveals that the annual superiority of Net
Operating Earnings was less for the forecast models employing the longer
base periods (i.e., R1, R2, and R3) than for those employing the shorter
base period(s) (i.e., 31, S2 and 83).

Comparison of foregggt models-- Model S;was superior to the other
forec t m 1 ' t rms o the size nd v r' tion of forec st errors.
This statement is based solely on a comparison of the 1963-65, 1966-68,
and 1962-68 aggregate data of Tables III and IV. The data for periods
prior to 1962 were ignored because in order to make a meaningful comparison

between forecast models it was necessary to include only those forecast

8O

periods covered by all models. The mean absolute percentage forecast
errors for the 1963-65 and the 1962-68 periods and the average mean
deviations for the 1963-65, 1966-68 and 1962-68 periods were less for
model 81 than for any of the other forecast models. Only the mean absolute
percentage forecast errors for models R2 and R3 for the 1966-68 period
were lower than those of model 81 for the periods compared.

Rank relationship between means and mean deviations-- The annual
mean absolute percentage forecast error gave almost as good an indication

2 of the predictive capacity of the two earnings

of the relative reliability
measures by itself as it did when used in conjunction with the annual

mean deviation of the absolute percentage forecast errors. The basis

for this statement rests on a comparison of the ranks of the mean absolute
percentage forecast errors and the mean deviations for all the annual
forecast periods contained in Tables III and IV. For forty-six of the
fifty-five3 annual periods, the earnings measure with the lowest mean in
Table III was the same as the earnings measure with the lowest mean
deviation in Table IV. Furthermore, in six of the nine remaining periods

the difference between the earnings measure with the highest mean deviation

and the one with the lowest mean deviation was less than one percent.
5.4 Results Analyzed According to the Long-Run Criterion

5.4.1 éggregation Procedures

Whereas the absolute value of FEij - the percentage forecast

 

2Where reliability is determined by considering both the size and
the variation of the annual forecast errors.

3Fifty-five equals the sum of the annual forecast periods covered
by each forecast model, i.e. eleven years for model 31 plus ten years for

32 etc.

 

81

error for bank i in foreggptlyear j - served as the basic measure for
evaluating the performance of the two earnings measures under the short-
run criterion, the absolute value of FEik - the percentage forecast error
for bank i in time ppriod k - was used as the basic measure for evaluating
the relative predictive capacity of the two earnings measures under each
of six separate forecast-error-periods associated with the long-run

criterion. FEik was calculated according to the following formula:

 

h
2: (Fij'Aij)
FE, =j=1
1k h
z (A .)
i=1 ‘3
where: h = 2, 3, 4, 5, 6 or 7, the number of annual
forecasts included in the forecast-error-
period

Fi = forecast earnings for bank i in the jth year
1 of time period k

A. = actual Net Income for bank i in the jth year
ij . .
of t1me period k

The length of a forecast-error-period varied from two to seven
years depending upon the number of annual forecast errors algebraically
aggregated in arriving at the percentage forecast error for a given bank
in a given time period. For example, when the forecast-error-period was
three years, the percentage forecast error was calculated by averaging
the algebraic amount of each of three consecutive annual dollar forecast
errors for a given bank and dividing by the total actual earnings of that
bank for the three-year period. By varying the length of the forecast-
error-period it was possible, of course, to vary the number of years
included in the "long-run". By aggregating the algebraic amount of the

annual dollar forecast errors and dividing by the actual earnings for the

82

number of years included in the forecast-error-period, it was possible
to obtain a measure of the cumulative forecast error for a time span
as long as the forecast-error-period. The length of the forecast-error-
period was limited to no more than seven years because seven years was
the longest forecast-error-period that was common to all six forecast
models.

The number of time periods covered by the forecast models varied E
according to the length of the forecast-error-period and the number of ,
base years employed by the forecast model. For example, there were

three time periods covered by forecast model R2 when the forecast-error-

1~

period was six years. The number of time periods covered was limited

to three because model R2 employed four base years (starting with 1957,
58, 59 and 60) and the period covered by this research study was limited
to twelve years (1957-68). Consequently, the only time periods for
which results were available were 1961-66, 1962-67 and 1963-68. Table VI
shows the number of time periods associated with each forecast model

for each multi-year forecast-error-period.

The empirical findings aggregated under the long-run criterion
are contained in Tables X through XV of Appendix B. Each table contains
the results obtained from utilizing a forecast-error-period of a
different length. Each table shows, for the 26 banks included in the
study, the mean absolute percentage forecast error and the mean deviation
of the absolute percentage forecast errors for (1) each time period,

(2) the common number of time periods covered by all forecast models

and (3) the total number of time periods covered by each forecast model.

83

TABLE VI

NUMBER OF TIME PERIODS CONTAINED IN THE TOTAL PERIOD
COVERED BY EACH FORECAST MODEL FOR EACH MULTI-YEAR
FORECAST-ERROR-PERIOD

 

 

Number of Forecast

 

 

 

Years in Forecast- Forecast Model
Error-Period 81 S2 S3 R1 R2 R3
2 10 9 9 8 7 6
3 9 8 8 7 6 5
4 8 7 7 6 5 4
5 7 6 6 5 4 3
6 6 5 5 4 3 2
7 5 4 4 3 2 1

 

The mean absolute percentage forecast error for each time period

(Mk) was determined according to the following formula:

m
2 IFEikl

 

z:
5‘
m
H
m
B
I

- 26, the number of banks

nu
{'11
II

the percentage forecast error for bank 1
in time period k

The mean deviation of the absolute percentage forecast errors for

each time period (MDk) was computed according to the following formula:

m —
Z [AFEik ' Mk

 

m

where: m = 26, the number of banks

 

 

84

AFE.

1k IFE. [ , the absolute percentage forecast

error for bank 1 in time period k

mean absolute percentage forecast error
in time period k

The mean absolute percentage forecast error and the average mean
deviation of the absolute percentage forecast errors for the common

number of time periods covered by all the forecast models and for the

'1 ii
1

total number of time periods covered by each forecast model was determined
by averaging the Mp's and the MDk's, respectively, for the number of time
periods in question. The common-period and total-period means and mean 9
deviations were calculated for each of the six multi-year forecast-error-

periods.

5.4.2 Observations

The following observations are based upon the data contained in
Tables X through XV of Appendix B and are divided into the same four
categories used in conjunction with the short-run criterion: (1) overall
results, (2) annual results, (3) comparison of forecast models, and (4)
relationship between the rank of the means and the rank of the mean
deviations. The observations relating to each of the two-, three- and
four-year forecast-error-periods (Tables X, XI and XII) are presented
separately in Subsections 5.4.2a, b and c while the observations relating
to the five-, six- and seven-year forecast-error-periods (Tables XIII, XIV,

and XV) are presented as one group in Subsection 5.4.2d.

5.4.2a Two-Year Forecast-Error-Period
Overall results-- Observation: Although on an overall basis, Net
Operating Earnings showed the best predictive capacity, its superiority

declined from what it was under the short-run criterion. Basis: Net

85

Operating Earnings had the lowest total-period mean for all six forecast
models. However, it was superior for only three of the six common-period
means and two of the six common-period average mean deviations. A
comparison of the total-period means for forecast-error-periods of one
year (Table III) and two years (Table X) reveals that the difference
between the total-period means of the two earnings measures declined from

an average (across all models) of 13.3% favoring Net Operating Earnings

-O.v-r1;*7.1 A hl

when the forecast-error-period was one year to an average of 2.5%
favoring Net Operating Earnings when the forecast-error-period was

lengthened to two years. A similar decline was noted when comparing

 

the common-period means.

Annual results-- Observation: Neither earnings measure was

 

superior on an annual basis. Basis: Net Operating Earnings had the
lowest mean for a majority of the two-year time periods covered by each
forecast model for four of the six forecast models. However, it had the
lowest mean deviation for only two of the six forecast models. It is
interesting to note that there appeared to be a decided trend toward Net
Income as the superior predictor in the last few periods covered by each
forecast model. For example, a review of the means for the last four
time periods covered by each forecast model indicates that Net Income
was lowest for twenty of the twenty-four total time periods (six
forecast models times four time periods) whereas it was lowest for only
three of the twenty-five preceding total time periods. This reversal in
superiority will be commented upon in a subsequent section.

Compprison of forecast models-- Observation: Model 51 was, from
a relative standpoint, superior to the other forecast models in terms of

the size and variation of forecast errors. Basis: The common-period

86

means and average mean deviations were lower for both earnings measures
for model 31 than for any of the other forecast models. It should be
noted, however, that the common-period means and average mean deviations
for both earnings measures for all forecast models were within 3.9% and
2.8% of each other, respectively.

Rank relationship between megpslgpdlmean deviations-- Observation:
The mean absolute percentage forecast error for each two-year time period
gave almost as good an indication of the relative reliability of the
predictive capacity of the two earnings measures by itself as it did
when used in conjunction with the mean deviation of the absolute percentage
forecast errors. Basis: A comparison of the ranks of the mean and the
mean deviations for all of the two-year time periods covered indicates
that the earnings measure with the lowest mean was the same as the earnings
measure with the lowest mean deviation for forty-five of the forty-nine

total time periods.

5.4.2b Three-Yegr,Forecgst-Error-Period

Overall results-- Observation: Neither earnings measure could be

 

considered distinctly superior although, when<:ompared to the overall
results when the forecast-error-period was one and two years, there appears
to be a trend toward Net Income as the superior predictor. Basis: Net
Income had the lowest total-period means and average mean deviations for
three of the six and four of the six forecast models, respectively. An
analysis of the common-period means and average mean deviations indicates
that Net Income was superior for four of the six and six of the six fore-
cast models, respectively.

Annual results-- Observation: Neither earnings measure could be

 

considered distinctly superior. Basis: Each earnings measure had the

87

lowest mean for a majority of the three-year time periods covered by
each forecast model for two of the forecast models with two ties. How-
ever, Net Income did have the lowest mean deviation for four of the six
forecast models with the outcome of the other two models ending in ties.
The trend toward Net Income as the superior predictor in the latter time
periods was still existent as is shown by the fact that Net Income had
the lowest mean for eighteen of the final twenty-four total time periods
covered by all of the forecast models.

Comparison of forecast models-- Observation: Model S1 was, from
a relative standpoint, superior to the other forecast models in terms of
the size and variation of forecast errors. Basis: The common-period
means and average mean deviations were lower for both earnings measures
for model 31 than for any other forecast model. However, just as was the
case when the forecast-error-period was two years, it is worth noting that
the common-period means and average mean deviations for both earnings
measures for all forecast models were within 3.4% and 2.9% of each other,
respectively.

Rank relationship between means and mean deviations-- Observation:
As was the case when the forecast-error-period was one and two years, the
mean absolute percentage forecast error gave a good indication, by itself,
of the relative reliability of the predictive capacity of the two earnings
measures. Basis: A comparison of the ranks of the means and mean deviations
shows that they were the same for thirty-three of the forty-three total time

periods covered by all of the forecast models.

5.4.2c Four-Year Forecast-Error-Period
Overall results-- Observation: Net Income showed the best predictive

capacity. Basis: Net Income had the lowest total-period means and average

88

mean deviations for all forecast models except model 83 and the lowest
common-period means and average mean deviations for all six forecast
models.

Annual results-- Observation: Although the annual results do not

 

favor Net Income over Net Operating Earnings as significantly as do the

overall results, Net Income still showed the best predictive capacity on
an annual basis. Basis: Net Income had the lowest means and lowest é
mean deviations for a majority of the four-year time periods covered by ,
each forecast model for four of the six and five of the six forecast p

models, respectively.

 

.Comparison of foreggst models-- Observation: Models 31 and S2
are superior, from a relative standpoint, to the other forecast models
in terms of the size and variation of forecast errors. Basis: Model S1
had the lowest common-period mean for Net Operating Earnings and the
lowest common-period average mean deviation for Net Income. Model S2
had the lowest common-period mean for Net Income and was tied with
model 81 for the lowest common-period average mean deviation for Net
Operating Earnings. Once again, however, it is worth noting that the
common-period means and average mean deviations for both earnings measures
for all forecast models were within 3.3% of each other.

Rank relationship between mean§=§pd mpgp devigplons-- Observation:
The mean absolute percentage forecast error continued to give a good
indication, by itself, of the relative reliability of the predictive
capacity of the two earnings measures. Basis: A comparison of the ranks
of the means and mean deviations reveals that they were the same for twenty-

eight of the thirty-seven total time periods covered by all of the forecast

models.

89

5.4.2d Five-, Six-,‘and Seven-Year Foreggst-Error-Periods

Overall results-- Observation: Net Income showed the best predictive

 

capacity. Basis: Net Income had the lowest total-period means and average
mean deviations over all three forecast-error-periods for all forecast
‘models except model S3. Net Income also had the lowest common-period means
and average mean deviations over all three forecast—error-periods for all
six forecast models with one exception. The lone exception was model 31
when the forecast-error-period was six years where the common-period mean
for Net Operating Earnings was lower than for Net Income.

Annual results-- Observation: Net Income showed the best_predictive

 

capacity. Basis: Net Income had the lowest means and mean deviations for
a majority of the time periods covered by each forecast model for all
forecast models except model S3 over all three forecast-error-periods.

Compgrison of forepgst models-- Observation: No one forecast
model is superior to the others. Basis: Model S1 had the lowest common-
period means and average mean deviations for forecast errors related to
Net Operating Earnings for all forecast models for all three multi-year
forecast-error-periods. However, with the exception of models 83 and R1,
all of the forecast models had the distinction of at least sharing the
lowest common-period mean or average mean deviation for forecast errors
related to Net Income for one or more of the three multi-year forecast-
error-periods. The latter outcome is indicative of the fact that,
excluding the results of model S3, the common-period means and average
mean deviations for forecast errors related to Net Income are within 3.4%
and 1.8% of each other, respectively, for all forecast models for all of
the three forecast-error-periods.

Rank relationship,between means and mean deviations-- Observation:

fLIh-‘n '1- hrnafl-A-rfir. m .1

. n .J.
.....~.-...r

5‘
I, _.

sn-v -

90

The mean absolute percentage forecast error continued to give a good
indication, by itself, of the reliability of the relative predictive
capacity of the two earnings measures. Basis: The ranks of the means
and mean deviations were the same for at least 80% of the total time
periods covered by all of the forecast models for each of the three

forecast-error-periods.

5.5 A Plausible Explanation of One Trend in the Forecast Errors

The purpose of this section is to present a plausible explanation
for the occurrence of one of two trends that are observable in the data
contained in the tables presented in the preceding two sections. Both of
these trends indicate a switch from Net Operating Earnings to Nethncome
as the superior predictor. The first appears as a trend across the single
and multi-year forecast-error-periods while the second appears as a trend
within the total time periods covered by each of three multi-year forecast-
error-periods. The first or across-period trend away from Net Operating
Earnings toward Net Income occurs as the forecast-error-period is lengthened.
Further discussion of this trend is deferred until Chapter VI. The second
or within-period trend away from Net Operating Earnings toward Net Income
occurs in the last few time periods covered by each of the first three
multi-year forecast-error-periods. This trend was most observable when
the forecast-error-period was two years (Table X) and to a lesser extent
when it was three and four years (Tables XI and XII).

A plausible explanation of why the within-period trend occurred
can be found in a comparison of the security gains and losses incurred in
the last five years (1964-68) covered by this study with those incurred in
the preceding five years (1959-63). A review of the security gains or

losses incurred by all of the 26 banks included in this study shows for

91

the 1964-68 period that 97 out of a possible 130 total bank-years (26
banks x 5 years) were years in which security losses were incurred
whereas for the 1959-63 period only 53 out of 130 total bank-years were
loss years.

Given the above analysis, it may be argued that the shift from a
less consistent incurrence of gains (77 out of 130) to a more consistent
incurrence of losses (97 out of 130) may well have caused the shift
toward Net Income as the superior predictor. This argument is based

upon the following logic. Referring to the graph in Figure 4 on page 54

-:.“'*"‘.‘- T. ‘. 1.21:. -Lt
,
H

of Chapter IV and letting 2 refer to Net Operating Earnings and y to Net

0-“

Income it can be argued that when security losses become consistent over
some long-run period the dispersion of Net Income (y) can be expected
to become smaller and the bias (b) of Net Operating Earnings (2) larger.
The dispersion of Net Income can be expected to become smaller because
the fluctuations in Net Income over time will be less due to the consistent
incurrence of security losses. The bias will become larger due to the
fact that the difference between Net Income and Net Operating Earnings
will be larger since there are no security gains to offset the losses.
At some point, as the dispersion of Net Income is reduced and the bias
of Net Operating Earnings increased, the superior efficiency (i.e., lower
dispersion) of Net Operating Earnings, if any, will no longer be sufficient
to overcome its biasedness. At that point, Net Income would become a better
predictor.

Such a point seems to have occurred during the latter time periods
covered by the first three multi-year forecast-error-periods. Apparently,
however, there was not as great an increase in the efficiency of Net Income

or the bias of Net Operating Earnings when the forecast-error-period was

92

one year (i.e. - the short-run). For, although Table III shows that Net
Income has the lowest mean absolute percentage forecast errors for some
forecast models for some forecast years after 1963, no apparent trend is

observable in the annual forecast errors.

~5-6 1119411231

The purpose of this chapter was to present the empirical findings
of this research study. In order to do so, however, it was first necessary
to specify (1) the criteria used in evaluating and (2) the procedures used
in aggregating the forecast errors. Both a short-run and a long-run
evaluation criterion were established. Under both criteria, the earnings
measure associated with the lowest absolute percentage forecast error for
a given time period was designated as having the superior predictive
capacity. The two criteria differed in that for the short-run criterion
a forecast-error-period was specified as being one year in length whereas
for the long-run criterion it was allowed to vary from two to seven years.

Three different aggregation procedures were utilized in applying
the evaluation criteria. The mean absolute percentage forecast error and
the mean deviation of the absolute percentage forecast errors were used in
applying both the short-run and long-run criteria. These two aggregation
procedures were considered to be the most accurate means of determining,
on the average, the relative predictive capacity of the two earnings
measures. The third aggregation procedure consisted of tabulating for each
time period or combination of time periods the number of banks for which
each earnings measure produced the lowest absolute percentage forecast
error. Such an aggregation procedure was considered somewhat inferior to

the two previously described procedures but was utilized in conjunction

93

TABLE VII

EARNINGS MEASURE WITH THE LOWEST TOTAL-PERIOD MEAN PERCENTAGE,
FORECAST ERROR AND LOWEST TOTAL-PERIOD AVERAGE MEAN DEVIATION
FOR SELECTED FORECAST-ERROR-PERIODS

 

T
T

 

 

 

 

 

Forecast- Forecast Details
Error- Error in Forecast Model
Period Measure Table-- 81 32 83 R1 R2 R3
One Year Mean III NOE NOE NOE NOE NOE NOE
Mean Deviation IV NOE NOE NOE NOE NOE NOE
Two Years Mean X NOE NOE NOE NOE NOE NOE 1
Mean Deviation x NI NOE NOE NOE NOE NOE 1
Three Years Mean XI NI NI NOE NI NOE NOE ;
Mean Deviation XI NI NOE NOE NI NI NI 9
Four Years Mean XII NI NI NOE NI NI NI
Mean Deviation XII N1 N1 NOE NI NI NI
Five Years Mean XIII NI NI NOE NI NI NI
Mean Deviation XIII NI NI NOE N1 N1 NI
Six Years Mean XIV NI NI NOE NI NI NI
Mean Deviation XIV NI NI NOE N1 N1 N1
Seven Years Mean XV NI NI NOE NI NI NI
Mean Deviation XV NI NI NOE N1 N1 NI

 

with the short-run criterion only, in order to aid the reader in assimilating
the empirical findings of this research study.

Once the evaluation criteria and aggregation procedures were specified
it was possible to present the empirical findings in a manner suitable for
meaningful analysis. Accordingly, Tables III, IV, V of this chapter and
Tables X through XV of Appendix B contain the results upon which the conclu-
sions of this study will be based. In order to provide support for these
conclusions, observations based upon the data included in the above mentioned

tables were presented. These observations, for both the short-run and long-

94

TABLE VIII

EARNINGS MEASURE WITH LOWEST COMMON-PERIOD MEAN PERCENTAGE
FORECAST ERROR AND LOWEST COMMON-PERIOD AVERAGE MEAN DEVIATION
FOR SELECTED FORECAST-ERROR-PERIODS

 

 

 

 

 

 

 

Forecast- Forecast Details
Error- Error in Forecast Model
Period Measure Table-- S1 32 S3 R1 R2 R3
One Year Mean III NOE NOE NOE NOE NOE NOE
Mean Deviation IV NOE NOE NOE NOE NOE NOE
Two Years Mean X NOE NI NOE NI NOE NOE
Mean Deviation X NI NI NI NI NOE NOE
Three Years Mean XI Tie NI NI NI NI NOE
Mean Deviation XI NI NI NI NI NI NI
Four Years Mean XII NI NI NI NI NI NI
Mean Deviation XII NI NI NI NI NI NI
Five Years Mean XIII NI NI NI NI NI NI
Mean Deviation XIII NI NI NI NI NI NI
Six Years Mean XIV NOE NI N1 N1 N1 NI
Mean Deviation XIV NI NI NI N1 N1 N1
Seven Years Mean XV N1 N1 N1 N1 NI NI
Mean Deviation XV NI NI NI N1 N1 N1

 

run criterion were grouped into four categories: (1) overall results, (2)
annual results, (3) a comparison of the size and variation of forecast errors
across all forecast models, and (4) the relationship between the rank of the
mean absolute percentage forecast errors and their related mean deviations.
The latter two categories of observations were presented not because
they were considered to provide necessary support for drawing conclusions
about the relative predictive capacity of the two earnings measures, but
because they were considered to be useful by-products of this research

effort. The observations about the overall results and annual results were,

95

TMEEIX

EARNINGS MEASURE WITH LOWEST TOTAL-PERIOD MEAN PERCENTAGE
FORECAST ERROR AND LOWEST TOTAL-PERIOD AVERAGE MEAN DEVIATION
FOR A MAJORITY OF FORECAST-ERROR-PERIODS COVERED BY EACH
FORECAST MODEL FOR SELECTED FORECAST-ERROR-PERIODS

 

 

 

 

 

 

Forecast- Forecast Details
Error- Error in Forecast Model
Period Measure Table-- S1 S2 S3 R1 R2 R3
raga.
One Year Mean III NOE NOE NOE NOE NOE NOE
Mean Deviation IV NOE NOE NOE NOE NOE NOE
Two Years Mean X NOE NOE NOE N1 N1 NOE
Mean Deviation X NI NOE NOE NI NI Tie
Three Years Mean x1 NI NOE Tie NI Tie NOE f
Mean Deviation XI NI Tie Tie NI NI NI ’
Four Years Mean XII NI NI NOE NI NI Tie
Mean Deviation XII NI NOE N1 N1 NI NI
Five Years Mean XIII N1 N1 NOE NI NI NI
Mean Deviation XIII NI NI Tie NI NI NI
Six Years Mean XIV NI NI NOE N1 N1 N1
Mean Deviation XIV NI NI NOE NI NI NI
Seven Years Mean XV NI NI NOE NI NI NI
Mean Deviation XV NI NI NOE NI NI NI

 

of course, presented for the primary purpose of providing a basis for
determining which, if any, of the two earnings measures showed the superior
predictive capacity.

The four categories of observations are summarized, in reverse order,
as follows:

Rank relationship between megps and mgppideviptions-- The mean
absolute percentage forecast error gave almost as good an indication by
itself of the relative reliability of the two earnings measures as it did

when used in conjunction with the mean deviation.

96

Comparison of forecast models-- Forecast model 81 showed a predictive
capacity superior to all of the other forecast models when the length of the
forecast-error-period was less than four years.

Overall results and Annual results-- The superior predictive

 

capacity of Net Operating Earnings was superseded by Net Income as the
length of the forecast-error-period increased. Because of the important

role that this statement and the data upon which it was based will play

s . '

in forming a basis for the conclusions that will be presented in the
following chapter, Tables VII, VIII and IX are presented to reinforce the

observations made in earlier sections of this chapter. Tables VII and VIII .

 

summarize the empirical findings as described by the total-period and
common-period means and mean deviations. Such findings were used in making
observations about the overall results. Table IX summarizes the empirical
findings that were used in making observations about the annual results.
The empirical findings as presented in the tables of this chapter
and Appendix B and as summarized in the observations provide the basis for

the conclusions that will be presented in the following chapter.

CHAPTER VI

SUMMARY AND CONCLUSIONS

6.1 Introduction I
The primary purpose of this chapter is to present the conclusions
reached as a result of this study and to discuss their implications. In

achieving this goal the chapter is divided into four major sections. '

 

 

Section 6.2 contains a brief review of the purpose of this research and
the research design. Section 6.3 discusses two limitations of the study.
Section 6.4 presents the conclusions and their related implications.

Finally, Section 6.5 suggests some avenues for future research.

6.2 Review of Research Purposgippd Design

The primary purpose of this research study has been ot obtain
empirical evidence that would aid in resolving the controversy that exists
between bankers and accountants over which of two different earnings
measures should be presented as the final reported earnings figures on a
bank's income statement. The two earnings measures referred to above
differ in respect to two items. Bankers favor a final reported earnings
figure, described as Net Operating Earnings, that excludes any provision
for loan losses or any security gains or losses. Accountants, on the other
hand, favor a figure, described as Net Income, that includes some provision
for loan losses and all security gains and losses.

An investigation into the nature and causes of the controversy

revealed that although there would be no controversy if it were not for

97

98

the existence of the two disputed items, it remains unresolved for reasons
unrelated to the arguments presented by both sides in support of their
treatment of the two items. This conclusion was reached after an analysis
of the arguments presented by both sides revealed that there was logical
support for both positions. Further investigation into the nature of the
controversy revealed that the primary reason for its existence is the
belief that investors focus almost entirely on the final reported earnings
figure in evaluating the current and potential performance of a firm while
largely ignoring other financial data included in the published financial
statements. This penchant for focusing on a single figure is referred to
as single-figure fixation. Granted the existence of the belief in single-
figure fixation, it was found that a secondary reason for the controversy
is a disagreement between the opposing sides as to which of the two earnings
measures is the most relevant to users of publicly reported bank earnings.
Initial efforts to obtain empirical evidence that would aid in
resolving the controversy were focused on attempts to perform a study to
determine the validity of the belief in single-figure fixation. Such a
study was not found to be feasible. Subsequent efforts were focused on
performing a study to determine which of the two earnings measures was the
more relevant. Such a study was found to be feasible once "the capacity
to predict future earnings" was specified as a surrogate for relevance.
While the general purpose of the "predictive" study was to obtain
empirical evidence that would aid in resolving the controversy, the
specific purpose was to determine the relative capacity of the two earnings
measures to predict future all-inclusive earnings. For, it was argued, if
one earnings measure were found to have a predictive capacity superior to

the other there would be new evidence for presenting the measure with the

 

99

superior predictive capacity as the final reported earnings figure on the
grounds that it was more relevant. All-inclusive earnings was specified
as the relevant forecast objective on the grounds that it provides the
best measure of a bank's future overall performance.

Net Operating Earnings per share and Net Income per share,
respectively, served as inputs to the six linear forecast models used to
obtain an estimate of Net Income per share one year in the future. The
resulting forecast errors served as the basis for determining the relative
predictive capacity of the two earnings measures over forecast-error-periods

varying from one to seven years in length.

ow" . ‘34:)...- mm“.- A I

6.3 Limitations

Prior to presenting any conclusions it would be worthwhile to point
out two limitations associated with this study. The first relates to the
fact, as was pointed out in Chapter IV, that the findings of any predictive
study depend in part upon the particular model(s) employed. Therefore, the
validity of any conclusions made as a result of the findings of this or any
other predictive study are dependent upon the reader's acceptance of the
reasonableness of the forecast models.

The second limitation relates to an assumption made by any study
that uses historical data as a basis for making inferences about the future.
The assumption, of course, is that the conditions which gave rise to past
results will continue to prevail in the future. For example, as relates to
this study, if Net Income were found to have had superior predictive
capacities over the period covered by this study, a basic conclusion would
be that in the future it should be designated the final reported earnings
measure on the grounds that it is the more relevant earnings measure. Such

a conclusion assumes that Net Income will continue to have superior predictive

100

capacities in the future. Any change in future conditions that would
impair the validity of this assumption would, of course, impair the

validity of the conclusion.

6.4 Conclusions ppd Implications
There are two basic conclusions to this study. The first or

major conclusion relates to the primary objective of this research whereas

11m
1
l

the second or minor conclusion relates to the individual forecast models

it

employed by the study.

6.4.1 Majpr Conclusion

As relgtes to the primgry purpose of this study, this researcher

waT-fi *9 '7

concludes that sufficient new evidence to sppport the presentation of
either Net Operating Earnings or Net Income as the final rpported earnings
figpre on a bank's income statement has not been found. This conclusion

is based on the fact that neither earnings measure showed a superior
predictive capacity across a significant number of the forecast-error-
periods specified in this study. Figures 5 through 10 illustrate this
fact. Each of the figures show the trend line of the total-period mean
absolute percentage forecast errors for each earnings measure for each
forecast model as the forecast-error-period is increased from one to seven
years. The figures show, for all forecast models, that the lowest forecast
errors were associated with Net Operating Earnings when the forecast-error-
period was two years or less. The figures also show for all forecast models
except model 83 that the lowest forecast errors were associated with Net
Income when the forecast-error-period was four years or longer. Thus, it
is concluded that the designation as to which earnings measure is the more

relevant depends upon the length of the forecast-error-period deemed most

101

mm ammo: "maoHemm-eomem-am<ommoe
amaomamm eon meomem sm<emaon
moeazmoemm mHDAOmm< z<mz aoHemm-q<Hos

0 MMDOHm

Hm ammo: “mnonmm.eommw-am<ummom
amsomamm eon mmomae am<omeom
eo<azmemee mHDAOmm< z<mz QOHmmn-a<Hoa

m MMOOHm

 

 

 

oowuomnuOHHMuummomuom
CH oooSHonH mummoonom Hosnnm mo nooaoz

n .0 0 0 m N H

OH

NH

00

0H

00

ON

NN
unmouom

poauomnuouHMuummoouom
:H woosaonH mummoouom Hosnn< mo Honenz

N 0 m 0 m N H

 

OH

NH

«0

0H

0H

ON

NN
unmouom

102

 

as ammo: "moonmm-moemw-am<omeom
amsomamm mom meoaee em<omeon
mueszmoeme mssaomm< z<mz oonme-a<Hos

0 HMOUHm

mm amaoz "maoHeee-mommmuem<ommoe
amaomamm mom mmomem am<ommoe
moaazmoeme mHDQOmm< 24m: oonme-a<Hoa

m MMDOHm

 

 

mowuomnuouHMnumooouom
aw oomsaonH mumooouom Hosnn< mo monasz

N 0 m s m N H

 

unmouom

OH

NH

0H

0H

00

ON

NN

oomuomnuouumuummoouom
a“ oomsaocH mummoouom Hosan< mo monasz

m 0 m 0 m N H

 

OH

NH

0H

0H

0H

ON

NN

unwouom

103

mm ammo: "maonmn-aommm-am<ommoe
ameomamm mom mmoeem Hmeommoe
mo<azmoame mHDAOmma z<ma eoHeee-a<soa

OH MMOUHm

Ne ammo: "moonwm-momem-em<ommoe
oesomamm eon mmoeem Hmeomeoe
moeazmomee masaomm< z<mz ooHeme-a<Hoa

0 MMOOHM

 

 

ooHuomuHOHHMIumooeuom
CH moosHonH mumooouom Hodnnm Ho nonesz

n 0 m 0 m N H

 

ucouuom

OH

NH

0H

0H

0H

ON

NN

oOHHomnuouumuummoouom
CH ooosHocH mummoouom Hoana< mo monasz

m 0 m 0 m N H

 

OH

NH

0H

0H

0H

ON

NN
unmouem

104

relevant to the user of a final reported earnings figure. For periods of

two years or less Net Operating Earnings would receive the designation.

For periods of four or more years Net Income would receive the designation.
It was noted in Section 4.2 of Chapter IV that as the forecast-

error-period was lengthened Net Income would eventually, if it did not from

the start, show a superior predictive capacity. This observation was

based on the fact that Net Income was considered to be an unbiased predictor

of Net Income whereas Net Operating Earnings was not. It was also noted

that only by empirical study could it be determined when (i.e., over a

forecast-error-period of what length) Net Income would show a superior

‘2““Qffisfid‘Ji ...n«‘-n..1fir-:v 37F

predictive capacity. Figures 5 through 10 indicate for all but one of the
six forecast models employed that Net Income replaced Net Operating Earnings
as the superior predictor somewhere between forecast-error-periods of two
and four years in length. Had the change occurred between forecast-error-
periods of greater length, say six or seven years, it would have been
possible to argue that Net Operating Earnings should be unequivocably
designated the most relevant earnings measure on the grounds that few users
of a final reported earnings figure are interested in a forecast-error-period
longer than five or six years. However, this researcher has found no evidence
to indicate that users of a final reported earnings figure are more interested
in a forecast-error-period of one or two years than one of three, four or
five years. Consequently, neither earnings measure can be unequiVOCably
designated as being the more relevant. Such a designation depends on the
length of the forecast-error-period considered to be most relevant to the
user.

The major implication of this conclusion is that additional research

efforts are still needed in order to find an equitable means of resolving

105
the controversy.

6.4.2 Minor Conclusion

As relates to the individual forecast models, this researcher
concludes that increasing the number of base years employed in a linear
forecast model does not improve the capacity of either earnings measure
to predict future bank earnings. This conclusion is based on the fact
that forecast model SI, the "simplest" of the forecast models in that
forecast earnings were set to equal earnings for the preceding year, was
consistently superior to or as good as the five other "more sophisticated"
forecast models with longer base periods.

The major implication of this conclusion is that, granting the
existence of single-figure fixation and given that users of a bank's final
reported earnings figure linearly extrapolate an estimate of future
earnings, earnings figures contained in comparative income statements and
historical summaries are of little or no incremental value to the forecaster

of future bank earnings.

6.5 Suggestions for Future Research

The fact that the predictive capacity of both earnings measures
was found to be relevant for forecast-error-periods of different lengths
points out more than ever the need for investors to look at more than just
a single figure in attempting to make inferences about the future performance
of a firm. In the absence of the belief in the existence of single-figure
fixation, one recommendation of this study would be that both Net Operating
Earnings and Net Income be presented in both the income statement and

published earnings announcements. Such a presentation has been advocated

 

106

by the AICPA since it issued APB 9.1 However, such a recommendation
would beg the issue since the existence of the belief in single-figure
fixation was shown to be the primary reason for the existence of the
controversy. Consequently, one obvious suggestion that evolves from the
results of this study is that additional research efforts be made to
determine the validity of the belief in single-figure fixation.

In the event that such studies confirm the existence of single-
figure fixation or prove to be unfeasible or inconclusive, continued
efforts should be made to educate investors as to the importance of judging

a bank on more than just a single final reported earnings figure.

 

gin. rv-. .

A final suggestion is that efforts continue to be directed toward
discovering and defining the needs of users of accounting information.
More specifically, as relates to this study, such efforts could be directed
toward the determination, if there is one, of the most relevant forecast-

error-period.

 

IAPB 9, p. 119.

APPENDICES

 

APPENDIX A

BANKS INCLUDED IN THE STUDY

(Ranked According To Asset Size)

 

 

 

Fortune Assets2 At
Bank Ranking 12/31/68
19681 (000,000)
Bank of America National Trust & Savings Assn. 1 $ 23,961
First National City Bank 2 19,355
The Chase Manhattan Bank, N.A. 3 19,014
Manufacturers Hanover Trust Co. 4 10,439
Morgan Guaranty Trust Co. of New York 5 10,370
Chemical Bank 6 8,968
Bankers Trust Co. 7 7,653
Security Pacific National Bank 10 6,288.
Irving Trust Co. 12 5,070
Northwest Bancorporation * 3,955
Franklin National Bank 18 2,868
First Pennsylvania Banking and Trust Co. 20 2,507
Cleveland Trust Co. 21 2,449
Republic National Bank of Dallas 22 2,176
The Detroit Bank & Trust Co. 23 2,068
Philadelphia National Bank 24 2,065
Manufacturers National Bank of Detroit 26 1,970
Harris Trust & Savings Bank 27 1,918
Girard Trust Bank 30 1,750
First National Bank in Dallas 34 1,714
Wachovia Bank & Trust Co., N.A. 37 1,618
The Fidelity Bank 43 1,379
First Wisconsin Bankshares Corp. * 1,338
Mercantile Trust Co., N.A. 45 1,291
First National Bank in St. Louis * 1,092
First National Bank of Atlanta * 946
Total Assets at December 31, 1968 $144,222

 

1Source: "The 50 Largest Commercial Banks," Fortune, LXXIX, No. 6

(May 15, 1969), pp. 190-191.

2Source: Moody's Bppk & Financial Manual, April, 1969.

*Not ranked by Fortune.

107

 

APPENDIX B

SUPPLEMENTARY TABLES TO CHAPTER V

108

 

Ur

E o a . . . I .

 

 

 

 

 

 

 

 

 

m m a n n n m n a m 5 m owoucwuumm unseen
:uHa vaHuom mo nonesz
$5.0 $0.0 $0.0 $H.0 $0.5 $N.0 $0.NA $0.0 $0.5 $N.0 $0.0 $0.0 coHuoH>oo soot
ommuo>< oOHLomnHouoH
5.0 0.0 0.0 0.0 N.0 0.0 H.0 0.0 0.0 5.0 H.¢ N.0 :oHuoH>eo coo:
owmuo>< voHuoNICOEEoo
0.0 0.0 «.0 0.0 0.0 0.5 0.0 H.5 0.0 0.0 0.0 0.0 00-500H
0.0 N.0 5.0 0.0 0.0 5.0 0.0 5.0 0.0 H.0 0.0 0.0 50:000H
H.5 0.0 0.0 N.0 0.0 0.5 H.0 H.0 H.0 0.0 5.0 0.0 00:000H
5.0 0.0 0.0 0.5 0.0 0.5 0.0 0.0 0.0 0.0 0.0 0.5 00-000H
0.5 N.0 0.0 0.0 5.0 0.0 0.0 0.0 0.0 0.0 0.0 N.0 00-000H
$0.0 $0.0 H.0H 0.0 0.0H 0.0 0.0 0.0 H.0 0.0 N.0 0.0 001N00H
$N.5 $0.0 m.HH 0.0 0.00 0.0 N.HH 5.0 0.0 0.0 N01H00H
$0.HH $0.0 0.0m 0.0 o.HH 0.0 0.5 0.5 H01000H
$0.0N $0.0 $5.0H $0.0 0.0 N.5 001000H
mH.0 $0.0 001000H
coHuoH>oo com:
9
nu
11 N 0 q 0 0 0 q 0 q 0 q 0 owmucouuom umo30H
euH3 moOHuom mo Hananz
$0.LH $H.0 $N.HH $0.0 $m.NH $0.0H $N.5H $0.HH $¢.NH $5.0H $0.0H $H.0 coo: uoHuomuHmuoH
0.HH H.0 0.0H 0.0 H.00 0.0H 0.0 ¢.HH «.0 0.0H H.0 0.5 noes voHuomICOEEoo
5.HH 0.0 0.00 0.0H 0.0 0.HH «.0 0.0H 0.0 5.HH q.NH 0.0 00-500H
m.NH 0.0 0.0 0.0H 0.0 H.NH 0.0 N.0H 0.0 5.0H 0.0H 0.5 50:000H
0.0 0.0 0.0 0.0H «.5 H.0H 0.0 0.0H 0.5 0.0H 5.0 N.0 00:000H
0.0 0.0H 0.0 0.0 0.5 5.0 0.5 0.0 0.5 0.0 5.0 0.0 00-000H
0.HL 0.0 0.HH N.0 0.0 0.0 0.0 5.5 0.0 0.5 0.0 0.5 00:000H
$0.NH $0.0 0.0H 0.0 0.0H 0.0 0.NH 0.5 0.0H H.0 H.5 0.0 001N00H
$N.mH $0.0 N.0H 0.0 0.00 0.0 N.NN H.0 5.0 0.0 N01H00H
$0.NN 00.0 0.0N 0.0 0.0N 0.0 5.0H o.NH H01000H
$0.0N $0.0H $0.5H $0.0H N.NH 0.HH 00:000H
$N.LH $0.NH 00-000H
coo:
Hz moz Hz 002 Hz moz Hz 002 Hz 002 Hz 002 ensues: mNCchom
me we as mm Nm am seeoz announce

 

 

00400 039 CH H0000 00H0001000001Hm<00000 "000000 Hm<00000 mu<Hzmommm mHOH000< 00 20HH<H>00 z<mz 02¢ 00000 Hm<ommom mu<Hzmommm 090H000< z<mz

x 0H0<H

 

 

III)

 

 

 

 

 

 

 

 

0 0 0 N 0 N 0 0 0 0 0 0 mmmucmuuom 000300
2003 wvowuwm 00 000852
50.0 00.0 00.0 00.0 00.0 00.0 00.00 00.0 $0.0 $0.0 m0.0 00.0 000000>00 0002
mmmum>< vowuwm-HmuOH
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 000000>m0 0002
mmmum>¢ UOHHmmICOEEOU
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.5 0.0 0.0 0.0 5.0 00-0000
0.0 0.0 0.0 0.0 0.0 5.0 0.0 0.0 0.0 0.0 0.0 5.0 50-0000
0.0 0.0 0.0 0.0 5.0 0.5 0.0 0.5 0.0 0.5 0.0 0.5 00-0000
0.0 0.0 0.0 5.0 0.0 0.0 5.0 5.0 0.0 0.0 0.0 ~.0 00-0000
m5.5 00.0 5.0 5.0 0.0 0.0 0.5 ~.0 0.0 0.0 0.0 0.0 00-0000
no.5 00.0 0.0 0.0 0.00 5.0 0.0 0.0 0.0 0.0 00-0000
00.0 00.0 0.00 0.0 0.0 0.0 0.0 0.0 00-0000
00.00 u0.0 00.0 00.0 0.0 0.0 00-0000
00.0 00.5 00-0000
COwumH>wQ cme
0 0 0 0 0 0 0 0 0 0 0 0 0000000000 ummaoq
£00.03 mvowumm mo HQQEDZ
00.0 00.0 00.0 00.0 00.5 05.0 00.00 00.0 00.0 00.0 00.5 00.0 000: voHumm-HmuoH
0.0 0.0 0.0 0.0 0.5 0.0 0.5 0.00 0.5 0.0 0.5 0.5 0002 uoqumm-coEEoo
0.0 0.0 0.0 0.00 0.0 0.00 0.0 0.00 0.0 5.00 0.0 ~.0 00-0000
0.00 0.0 0.0 0.0 0.5 5.0 0.0 0.00 0.0 0.00 0.00 0.0 50-0000
0.0 0.0 0.0 0.0 0.0 5.0 0.0 0.00 0.0 0.00 0.0 0.0 00-0000
5.0 0.0 0.0 0.0 0.0 0.0 5.0 0.0 0.5 0.5 0.0 0.0 00-0000
.00.~0 00.5 0.00 0.0 0.00 0.0 0.00 0.5 0.00 0.5 0.0 0.0 00-0000
00.0 00.0 0.5 0.0 5.00 0.0 0.00 0.0 0.0 0.5 00-0000
00.5 00.5 5.00 0.0 0.00 0.5 0.00 0.00 00-0000
m0.00 M0.0 00.00 00.0 0.0 0.0 00-0000
00.00 00.00 00-0000
cmmx
z 002 Hz 002 Hz 002 Hz 002 Hz 002 Hz 002 augmmmz mmcflcuum
0x mm 00 0m 00 00 0000: ummoouoh

 

mm<m% mmxzh OH A<zom oonmm-mom0muhm<om00h "mmommm Hm<om00h mu<Hzmummm MHOA000< mo onH<H>mQ z<mx 02¢ mOMMM Hm<omxom mo<HzmoMmm MHQAOmmd Kdmz

Hx mqmde

 

111

 

 

 

 

 

 

 

q o 0 0 m A q n m a m o uwnuCQUHum umuzo;
:uflz mvowpom mo nonssz

Sm.q No.0 fim.m x¢.m Nm.m $0.0 xm.m No.0 $0.0 wo.n Ho.m -#.m :ofiunw>oa :69:
owwuw>< wofiuomnfiauOH

m.q 0.0 ¢.m «.0 m.m 0.m 0.0 0.m ¢.m 0.0 m.~ «.0 ceaumfi>mo cam:
owMMw>< nowpwmnCOEEoo

0.0 0.0 m.m o.m H.m 0.0 m.m m.m m.N 0.0 0.m 0.0 00nm0mH

m.m 0.0 0.m ~.0 o.m N.0 0.0 0.0 «.0 ~.0 m.~ m.m n0u¢0m0

w.m o.0 o.~ «.0 m.~ 0.0 n.¢ 4.0 m.m m.0 0.~ m.m 00cm0m~

x~.m Nm.m n.m 0.0 0.0 0.0 m.m ~.¢ N.¢ 0.0 ~.m 0.0 m0um0ma

$0.0 no.0 ¢.¢ 0.0 m.mH ~.¢ 0.0 0.4 m.m 0.0 c0u~0o0

Nm.m 00.0 0.00 m.m m.¢ 0.0 w.~ o.¢ m0uo0md

Nm.mH wm.¢ u¢.0 Nm.¢ o.m 0.0 ~0ummmH

wm.¢ $0.0 ~0uwma~
:ofiumw>uo cam:

N N m m ¢ N m c c m m m ammucmuumm amaze;

Law: mwoflumm mo “mnezu
No.n NH.“ $0.0 Na.m No.0 Sm.“ No.~0 N~.m Nm.m No.m $0.0 Nm.m cmwx vofiummufimuow
o.m 0.x «.0 «.5 0.0 N.w m.m 0.m 0.0 o.m m.0 «.0 cme 0000mmucoEEoo

m.w «.0 ~.m «.0 ~.m 0.00 ~.m m.~a 0.0 ~.~0 «.0 0.0 m0um0ma

0.0 0.0 m.m 0.0 o.“ 0.0 m.¢ 0.00 n.m «.00 ¢.m 0.0 n0u¢0¢0

0.0 0.0 0.0 o.m ~.m 0.5 0.0 ~.0 0.0 m.m 0.0 0.~ 00um0oH

mm.m No.m m.m 5.0 m.m 5.0 m.0 «.0 ~.0 m.0 n.¢ ~.0 m0u~0m~

Nm.0 Hm.m m.m 0.0 ~.m~ m.m H.0H 0.0 0.0 0.0 «0-0090

Nm.m mw.0 0.00 m.m o.m 0.0 o.a n.w m0uo0mfi

Nm.~m $0.5 m~.oa No.5 0.0 o.m n0nmmm~

Nm.o 30.00 ~0n0mo~

cam:

Hz moz Hz 002 Hz moz Hz moz Hz moz Hz 002 «usmcw: mwcwcumm
mm mm Hm mm Nm Hm Hmvoz ummowpom

 

 

mmdmy mDOw OH A<Dom QOHmMLIMOMMMIHm<Ummom "mmomMM Hm<umm0h mu<HszMmm MHDAOmm¢ mo ZOHH<H>MQ z<m2 024 mommm Hm<omm0m wu<Hszmmm NHpAOmmd z<mfi

HHx mam<H

 

 

1H

mm.q

m.¢

v—‘Ox?
WQW

.7\°

 

Hz

a3c>~w
-n\o|n

i“

 

0 o
Rm.~ $0.0
m.~ 0.0
¢.n m.m
q.~ 0.0
m.~ «.0
..o.~ .03..»
c o
$5.0 $0.0
0.0 0.0
0.0 o.0
m.n 0.0
m.q o.n
Nm.q No.0
Hz moz

x~.~

 

Hz

N0.q

moz

 

m m
$5.0 Xm.¢
«.m «.0
o.m ~.m
m.~ m.m
0.0 o.m
n.0H 0.n
~.o ¢.m
o.- Nn.n
N e
Nm.- Nm.m
m.m 0.0
0.0 ~.~
5.0 n.m
0.0 «.0
n.50 0.0
n.m 5.0
Nq.¢~ N~.0
Hz mOz

$0.0

0.0

O‘ONU‘ON
WOWMON

.\°

 

Hz

ééééééé

ONOOQQON

o
b\

OVONVDVOOI-n

Ox‘Yx'fNQ—‘O

6

 

ammucwoumm umuaoa
Lug: avouuwm mo punssz

cowumw>mo cmmz

mmmum>< vowummuawuow

acquww>wo cam:

mwmuv>< noHummucoEEou

m0nc0®H
~0um0mH
00um0oa
m0u~0ofi
q0n00m~
m0ummm~
woummmfi

 

sewuww>mo cam:

mwmuchHmm ummzoa
nu“: mnoaumm we “unasz

cam: vowummnHMuoe

cam: vowummncoEEoo

m0n¢0o~
m0nm0m~
00um0m~
m0u~0m~
¢0a00m~
m0nammH
~0ummmfi

 

cmoz

madmmmz meHCumu
Amvoz ummumuom

 

 

mm<mw m>um OH 4<Dow Qonmmlmomxmnhm<umMOm ”mmomMm Hm<omxom mu<Hszxmm mabqommd mo onH<H>mQ z<mz Qz< MOMMM Hm<ommom mQ<Hzmommm mhzqomm< z<mx

HHHX mqm<a

 

 

113

n 0 0 0 0 0 N 0 0 n 0 0 ommucwuuwm ””9300
:u03 m000umm wo amnesz

 

 

 

 

 

 

 

 

00.0 00.0 $0.0 $0.0 $0.0 $0.0 $~.0 $0.0 $0.0 $0.0 $0.0 w0.0 500um0>m0 cam:
mmmgm>< 0000mmu0wuow

0.0 0.0 0.~ 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 c00um0>m0 cmmz
mmmum>< vo0ummucoaeou

0.0 0.0 0.0 0.0 ~.N 0.0 0.0 0.0 0.0 0.0 $.0 0.0 00:0000

00.» M0.0 0.0 0.0 0.0 $.0 $.0 0.0 0.0 0.0 0.~ 0.0 $0u~0m0

$0.0 $0.0 0.0 0.0 0.00 0.0 0.0 0.0 0.0 0.0 00:0000

$$.~ 00.0 0.0 0.0 0.0 0.0 0.0 0.0 00-0000

$0.00 $0.0 $0.0 m0.0 0.0 0.0 00:0000

$0.0 $0.0 00-0000
cowum0>wo :mmz

N o m o q o N m m m 0 N ammucmuuom “maze;
zu03 mnowumm mo “massu
$0.0 $0.0 $0.0 $0.0 00.0 $~.0 H0.00 $$.0 $$.0 m0.0 $0.0 $0.0 cam: nofiuwmu0muow
0.0 0.0 0.0 0.0 0.0 0.0 ~.0 0.$ “.0 0.0 0.0 0.0 cum: vo0ummncoesou

0.0 0.0 0.0 0.0 0.0 $.0 0.0 0.0 0.0 0.$ 0.0 0.0 0010000

$0.0 $0.0 0.~ 0.0 0.0 0.0 N.0 $.0 0.0 0.0 0.0 0.0 $0u~0m0

$0.0 00.0 0.0 0.0 0.$0 0.0 0.$ 0.0 0.0 0.0 00:0000

m$.0 $0.0 0.0 0.0 0.$ 0.0 0.$ 0.$ 00:0000

00.00 00.0 $0.0 No.0 0.0 0.0 00-0000

w$.0 $0.0 00:0000

cmmz
0x moz Hz 002 Hz moz 02 002 02 : 002 Hz moz musmmmz m0:0cuwm
mm mm 0m 00 0m 0m 0wvoz ummuwuom

 

 

mm<m> x00 OH A<Dom QOHMmmlmomxmnHm<ommOm umMOMMm Hm<ommom mu<Hzmommm MEDAOmm< ho ZOHH<H>mQ z<mx Q24 momxm Hm<omem mo<Hzmomum MEDAOmmd z<mz

>Hx m0m¢H

 

0 o 0 0 0 o 0 0 0 0 0 o wmmucooumm umw300
Lu03 m000uwm we “mosaz

 

 

 

 

 

 

 

00.0 00.0 $$.0 $0.0 $0.0 $0.0 $0.0 $0.0 00.0 ,$.0 00.0 00.0 :000m0>wa cam:
~00um>< 0000mmu0muOH
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 c00umw>wn cam:
m0~um>¢ 0000mmucoEEoo
$0.0 $0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.\ 0.0 0.0 00:0000
$$.0 No.0 $.0 0.0 $.00 0.0 0.0 0.0 0.0 0.0 $0n0000
00.0 $0.0 $.0 0.0 $.0 0.0 0.0 0.0 00-0000
$0.00 $0.0 $0.0 00.0 0.0 0.0 00-0000
30.0 00.0 00:0000
c00u00>wo cmoz
L4
1
1L
0 o 0 o m o 0 m r 0 m o ammucouumm amazoq
Luau—3 mfiOthm MO .Hwn—EDZ
00.0 00.0 $0.0 $0.0 $0.0 $0.0 $0.00 $0.0 «0.0 -0.0 $0.0 00.0 cam: nowuomu0muow
0.0 0.0 0.0 0.0 0.0 $.0 0.0 0.$ 0.0 0.0 0.0 0.0 cmwz 0000mmncoEEoo
«0.0 00.0 0.0 0.0 0.0 $.0 0.0 0.$ 0.0 0.0 0.0 0.0 00-0000
00.0 00.0 0.0 $.0 0.00 0.0 0.0 0.0 0.0 0.0 $0-0000
$0.0 $$.0 0.$ $.0 0.0 0.0 0.0 0.0 00:0000
$0.00 $0.0 p0.0 .0.0 0.0 0.0 00:0000
$0.0 w0.$ 00-0000
cmwz
0x 002 02 002 Hz 002 Hz 002 0% z 00x 02 002 whammwz mmcflcumm
mm mm 0x mm .m 0m 0mnoz ammumuom

 

 

0040» zm>mm OH 0400a QOHmmm|0000wnHmduudOm "000000 Hm<ommom mo<Hzmommm mHDAOmm< mo ZOHH40000 7 H0 004 000000 Hmdomxom momHzmummm mHDAOmmd Z<Ha

>x m00<H

SELECTED B I BLIOGRAPHY

 

SELECTED BIBLIOGRAPHY

Reference Works

The Accounting Commission. NABAC Accounting Bulletin No. 1, Revised --
Securities Accounting. NABAC, The Association for Bank Audit,
Control and Operation, 1967.

NABAC Accountipg Bulletin No. 3 - Loan Loss and Releted
Reserve Accounting. NABAC, The Association for Bank Audit,
Control and Operation, 1966.

Accounting Principles Board of the American Institute of Certified Public
Accountants. Opinions of the Accounting Principles Board - No. 9:
Reporting the Results of Operations. New York: AICPA, Inc., 1967.

Opinions of the Accountipg Principles Board - No. 13:
Amending Paragraph 6 of APB Opinion No. 9, Application to Commer-
cial Banks. New York: AICPA, Inc., 1969.

 

Accounting Research and Terminology Bulletins. Final Edition; New York:
AICPA, 1961. Accounting Research Bulletin No. 43.

American Accounting Association, Committee to Prepare a Statement of Basic
Accounting Theory. A Statement of Basic Accounting;Iheory.
Evanston, Illinois: 1966.

Annual Report of the Federal Deposit Insurance Corporation: 1968.
Washington, D.C.: Federal Deposit Insurance Corporation, May 1,
1969.

Bernstein, Leopold A. Accounting for Extraordinggy Gains and Losses.
New York: The Ronald Press Company, 1967.

Code of Federal Regulations: Title 12 - Banks and Banking. Sections
206.1-.71 and 335.1-.71. January 1, 1970 revision.

. Sections 10.1-.4 and 18.7-.7. January 1, 1970 revision.

 

Cohen, Jerome B. and Zinbarg, Edward D. Investment Analysis and Portfolio
Management. Homewood, Illinois: Richard D. Irwin, 1967.

Committee on Bank Accounting and Auditing of the American Institute of
CPA's. Audits of Banks. New York: AICPA, Inc., 1966.

115

 

116

Audits of Banks: Sppplement. New York: AICPA, Inc., 1969.

 

Garcia, F. L. How To Analyze a Bank Statement. Boston: Bankers Publishing
Company, 1966.

Garrison, Ray H. "Accounting for Bad Debt Reserves in Commercial Banks."
Unpublished Ph. D. dissertation, Graduate School of Business,
Indiana University, 1966.

Johnson, Walter C. "Financial Reporting To Stockholders By Commercial
Banks." Unpublished Master's thesis, Stonier Graduate School of
Banking, Rutgers -- The State University.

Parker, James E. "A Study of the Predictive Significance of Several
Income Measures Relative to the Accounting for Extraordinary Items
and Prior Period Adjustments." Unpublished Ph.D. dissertation,
Graduate School of Business Administration, Michigan State
University, 1969.

Sprouse, Robert T. "The Measurement of Financial Position and Income:
Purpose and Procedure," Research in AccountingeMeasurement. g
Edited by Robert K. Jaedicke, Yuji Ijiri and Oswald Neilsen.
Evanston, Illinois: American Accounting Association, 1966.

 

Yamane, Taro. Elementepy SamplingeTheory. Englewood Cliffs, N.J.:
Prentice Hall, Inc., 1967.

Periodical Literature

 

Gates, David C. "Developing Directions in Bank Accounting," Bankers
Magazine, CL (Summer, 1967).

"Cooking the Books to Fatten Profits," Time, April 11, 1969, p. 96.

Green, David, Jr. and Segall, Joel. "The Predictive Power of First-
Quarter Earnings Reports," The Journal of Business, XL, No. 1
(January, 1967).

"The Predictive Power of First-Quarter Earnings Reports: A
Replication," Empirical Research in Accounting: Selected Studies,
1966, (published as a supplement to Volume 4 of the Journal of
Accountipg_Research).

 

Heinemann, H. Erich. "Banks and Profits: Whether Accountants' Plea Will
Help Resolve Reporting Muddle is Uncertain," New York Times,
March 10, 1969, p. 63.

"Member Bank Income, 1968," Federal Reserve Bulletin, May, 1969.

117

Mlynarczyk, Francis A., Jr. "An Empirical Study of Accounting Methods and
Stock Prices," Empirical Research in Accounting: Selected Studies,
1969,(published as a supplement to Volume 7 of the Journal of
AccountingiResearch).

O'Donnell, John L. "Relationships Between Reported Earnings and Stock
Prices in the Electric Utility Industry," The Accounting_Review,
January, 1965.

. "Further Observations on Reported Earnings and Stock Prices,"
The Accounting Review, July, 1968.

 

Shipley, Edward T. "Bank Accounting Principles: A Progress Report," Law
and Contemporaperroblems, Volume 32: No. 1 (Winter, 1967)..

The Wall Street Journal. January 16, 1970 (p. 23) and April 15, 1970 (p. 1).

”What Are Earnings? The Growing Credibility Gap." Forbes, May 15, 1967.