University of Southern California Dissertations and Theses

Cross-sectional tests of tax and regulatory costs of a change in depreciation methods in the railroad industry

(USC Thesis Other)

Cross-sectional tests of tax and regulatory costs of a change in depreciation methods in the railroad industry

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content CROSS-SECTIONAL TESTS OF TAX AND REGULATORY COSTS OF
A CHANGE IN DEPRECIATION METHODS IN THE RAILROAD INDUSTRY
by
Sharon S. Lassar
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Business Administration)
May 1989
Copyright 1989 Sharon S. Lassar
UMI Number: DP22659
All rights reserved
INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a com plete manuscript
and there are missing pages, th e se will be noted. Also, if material had to be removed,
a note will indicate the deletion.
Dissertation Publishing
UMI DP22659
Published by ProQ uest LLC (2014). Copyright in the Dissertation held by the Author.
Microform Edition © ProQuest LLC.
All rights reserved. This work is protected against
unauthorized copying under Title 17, United S tates Code
ProQuest LLC.
789 E ast Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 4 8 1 0 6 -1 3 4 6
UNIVERSITY OF SOUTHERN CALIFORNIA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CALIFORNIA 90089
p a .
Com
i f
i ^ w >
3*3o f j o
This dissertation, w ritten by
..............
under the direction of h A . < %. D issertation
Committee, and approved by all its members,
has been presented to and accepted b y The
Graduate School, in partial fulfillm ent of re
quirem ents for the degree of
D O C TO R OF PH ILOSOPH Y
Dean of Graduate Studies
D a te J.222
DISSERTATION COMMITTEE
Chairperson
Table of Contents
/
List of Tables................................................... iv
1. INTRODUCTION AND LITERATURE REVIEW ..................... 1
1.1 The issues and events ............................... 1
1.2 Accounting theory ...................................... 5
2. HISTORICAL REVIEW AND POTENTIAL COSTS OF TRACK STRUCTURE
ACCOUNTING ............................................. 8
2.1 Description and history of RRB accounting ................ 8
2.2 Tax costs............................................... 9
2.3 Regulatory costs ...................................... 9
3. CHRONOLOGY OF EVENTS.....................................12
3.1 The 4-R Act ............................................... 12
3.2 Codification of R R B ' . . 14
3.3 Economic Recovery Tax Act of 1981...........................14
3.4 Transition rules ...................................... 15
3.6 Inflation Accounting.................................... 17
4. COST IDENTIFICATION AND HYPOTHESES DEVELOPMENT ......... 18
4.1 Introduction............................................. 18
4.2 Tax costs................................................. 19
4.3 Regulatory costs.........................................21
4.4 Codification of R R B........................................23
4.5 Transition rule tax benefits..............................24
4.6 Inflation accounting .................................. 25
5. EVENT STUDY...............................................26
5.1 Methodology for event study.............................. 26
5.2 Abnormal returns.............................. 26
5.3 Sample selection ...................................... 27
6 . CROSS-SECTIONAL TESTS ................................... 29
6.1 Cross-sectional tests ................................... 29
6.2 Tax costs.................................................29
6.3 Regulatory costs ...................................... 31
6.4 Codification of RR B ........................................33
6.5 Transition rule tax benefits..............................33
6.6 Inflation accounting...................... 34
7. RESULTS...................................................35
APPENDIX A: HISTORICAL NOTE ON R R B .............................47
APPENDIX B: REGULATORY BODIES ................................. 50
APPENDIX C: CHRONOLOGY OF EVENTS ............................ 53
ii
Table of Contents, Continued
APPENDIX D: SAMPLE COMPANIES ................ 55
APPENDIX E: METHODOLOGICAL ISSUES ............................ 57
BIBLIOGRAPHY ................................................... 70
iii
List of Tables
Table 1 Estimated tax cost of switch to ratable depreciation . . 3
Table 2 Summary of events tested.............................36
Table 3 Descriptive statistics on independent variables .... 37
Table 4 Cross-sectional regression parameter estimates ....... 38
Table 5 Structural stability test statistics .................. 59
iv
1. INTRODUCTION AND LITERATURE REVIEW
1.1 The issues and events
This study tests security price reaction to mandated accounting
change events with regulatory cost variables. The accounting change
requires that railroads switch from retirement-replacement-
betterment (RRB) accounting for track structure to ratable
depreciation. Before the test period, RRB had been used for book,
tax and regulatory purposes. The events studied led to a different
form of ratable depreciation being required for each purpose. Under
RRB, expenditures made to replace track are immediately expensed
rather than capitalized and depreciated. These expenditures average
close to 20% of operating expenses each year. Switching from
immediate expensing (RRB) to ratable depreciation causes a
significant increase in income for several years following the
change. The accounting change is hypothesized to be costly because
of potential wealth transfers caused by Interstate Commerce
Commission (ICC) intervention, income taxes, and contract payoffs
contingent on net income.
This study makes several significant contributions. First, tax
laws and rate making policies are used to construct tests for
regulatory costs of a mandatory accounting change. Secondly, the
accounting change increases net income of the firms affected whereas
most mandatory changes reduce current net income.-*- Although the
■*-FASB Statement No. 19 mandated successful-efforts accounting
of oil and gas wells. In 1974 the SEC imposed a moratorium on firms
adopting interest capitalization. FASB Statement No. 2 required
1
change is income increasing, negative abnormal returns are predicted
because of higher expected regulatory and contracting costs. This
setting separates valuation influences (naive investor hypothesis)
from regulatory and contracting costs. Finally, the events studied
involve different regulatory agencies that influence accounting
methods for distinct purposes with the ultimate result being that a
different method of ratable depreciation was mandated for each
purpose. This setting is unique in the way it distinguishes between
tax and other regulatory costs and provides independent event tests
for specific costs.
The largest regulatory cost of the change to ratable depreciation
is the increase in federal tax if taxable income is calculated under
the new method of accounting. Measurement of expected tax cost is
addressed in.Chapter 6 . However, to gain initial insight into the
potential magnitude of the additional tax burden, an ad hoc
calculation was made. Table 1 shows selected 1980 data for ten of
the largest railroads affected by the change. The estimated present
value tax cost shown in Table 1 assumes the increase in taxable
income is constant for six years then becomes zero, all firms
experience a 46% marginal tax rate and a 10% cost of capital with no
tax benefit carryovers. Under these illustrative assumptions, the
potential present value federal tax cost of the change is between 6%
firms to expense research and development expenditures. APB Opinion
16 restricted use of pooling-of-interests accounting. See Watts and
Zimmerman (1986) for a review of studies that look at these mandated
accounting changes.
2
Table 1
Estimated tax cost of switch to ratable depreciation
(in thousands)
Net
income
Carrier (RRB)
Better- Increase
RRB ments in 1980 Estimated
replacement and taxable tax
property additions income cost
ATSF $162,930
BO 34,629
■BN 222,715
|
jOO 131,620
MP 123,388
NW 232,416
SCL 175,369
SOU 185,283
SP 72,812
UP 164,738
$139,352 $16,726
82,503 4,070
272,612 59,746
64,650 9,957
117,069 21,266
57,852 13,029
94,367 13,449
94,326 9,533
36,719 15,076
144,890 33,882
$134,149 $379,172
79,617 225,038
261,533 739,224
62,163 175,704
112,458 317,863
55,490 156,843
90,773 256,570
87,718 247,935
34,992 98,905
138,931 392,689
Cost
Book value as a
of percent
equity of equity
$1,622,717 23.4
843,805 26.7
1,597,626 46.3
643,838 27.3
849,298 37.4
1,578,881 9.9
1,030,250 24.9
1,417,988 17.5
1,645,747 6.0
1,591,681 24.7
Source : Boberg (1983)
Estimated present value tax cost assumptions:
1. Marginal tax rate is 46%.
2. Ratable depreciation is 30 year straight line.
3. Timing difference is a 6 year annuity.
These ad hoc assumptions were chosen for illustrative purposes only.
Seme roads are owned by common parent corporations. To detect a stock reaction,
cost as a percent of market value of the traded corporation is relevant.
Parent
companies
Estimated
tax cost
Market value
common stock
Tax cost
as percent
market value
ATSF 380,172 2,985,415 12.7
CSX 657,313 1,949,346 33.7
MP 112,458 1,591,415 7.1
NW 156,843 1,297,236 12.1
SOU 247,935 1,237,034 20.0
SP 98,905 1,109,388 8.9
UP 392,689 7,552,519 5.2
3
and 46% of the book value of equity. The cost ranges from 5% to 33%
of the market value of parent corporation equity. More carefully
measured tax variables presented in Chapter 7 are of similar
magnitude.
Because RRB had been allowed for tax purposes as long as it was
t
required by the ICC, railroad response to an ICC proposal to switch
to ratable depreciation included lobbying for tax relief.
Unprecedented tax relief was ultimately granted in the form of
transition rules added to the Economic Recovery Tax Act of 1981.
The following section reviews accounting theory as it relates to
this study. Chapter 2 begins with a description of retirement-
replacement-betterment accounting and its historical background,
then identifies costs of changing to ratable depreciation. Chapter
3 chronologically traces the series of events that led from RRB
being used for book, tax, and regulatory purposes to a separate
accounting method being used for each purpose. Chapter 4 models
costs associated with the events. Chapters 5 and 6 explain the
research methodology. Results are presented in Chapter 7.
1.2 Accounting theory
Accounting theory has evolved from the observation that accounting
numbers are used in contracting and regulatory (political) processes
(Watts and Zimmerman, 1979). Because accounting numbers are used by
regulatory participants, a change in those numbers alters the
probability of wealth transfers taking place through taxes,
4
subsidies, etc. A mandated change in accounting standards, which
restricts the set of available procedures, will increase contracting
and regulatory costs to a firm if the mandate moves the firm away
from its optimal set of accounting procedures. The increase in cost
reduces firm value.
Tests of accounting theory involve predicting either an accounting
choice or its effect on firm value based on size and contracting
variables. Several tests of accounting theory acknowledge the
existence of regulatory cost incentives of accounting choice by
including a size variable. This study does not rely on size as an
explanatory variable. A more powerful test is constructed by
identifying specific regulatory costs of the accounting change.
Contracts create a demand for accounting numbers to determine
payoffs or restrict investment/financing decisions so that specified
payoff schedules are likely to be met. Powerful test of contracting
hypotheses are challenging to construct for the following reasons.
1. Accounting choices are not made independently. A firm's
portfolio of accounting procedures must be considered
(Zmijewski and Hagerman, 1981; Healy, 1985).
2. Contract details must be considered (Healy, 1985).
3. The size hypothesis generally leads to predictions opposite
to the contracting hypothesis.
The change from replacement-retirement-betterment to ratable
depreciation has two advantages over prior studies.
1. The change is large and occurs in an industry where
accounting procedures are strictly regulated. It is unlikely
that a firm could make accounting choices that offset the
effect of the depreciation change.
2. All hypotheses predict a reduction in firm value. The change
would be costly to firms with contingent interest
arrangements.
Proxy statements reveal little cross-sectional variation in
compensation and few recent bond offerings. Recent borrowings are
f y
in the form of equipment obligations such as capitalized leases.
However, some roads have contingent interest obligations where
interest payments increase with the level of earnings. The
existence of contingent interest contracts with less than maximum
interest being paid would seem to a be significant variable in
explaining negative abnormal returns. However, the two firms paying
less than maximum interest are not in danger of having interest
payments increased by the accounting change because the accumulated
deficit that must be eliminated before interest is paid is large and
growing by operating losses. Even though operating loss tax
benefits are absorbed by the parent holding companies, the parent
companies did not guarantee the contingent interest debt.
Because there is little cross-sectional variation in compensation
contracts and cross-sectional variation in debt contracts does not
increase expected debt contracting costs, this study does not
include bonus plan and leverage variables traditionally found in
tests of accounting theory. By reviewing contract details, it is
^Railroads have few modern day general debt obligations because
of restrictions imposed by contracts written nearly a century ago.
For example, railroad mortgages, secured by after-acquired property,
prevent railroads from issuing secured debt to finance track
additions.
6
concluded that such variables would be collinear with the intercept
in the test model developed below.
In the next chapter a brief discussion of the historical
background of RRB and costs of switching to ratable depreciation is
presented. In Chapter 3, the chronology of events under scrutiny is
| presented. Chapter 4 develops models of the effect the events have
on firm value. Chapter 5 discusses the event study methodology
used, Chapter 6 describes the cross-sectional tests performed, and
results are presented in Chapter 7.
7
2. HISTORICAL REVIEW AND POTENTIAL COSTS OF TRACK STRUCTURE
ACCOUNTING
2.1 Description and history of RRB accounting
This section provides a brief historical review of retirement-
replacement-betterment (RRB) accounting. Sections 2.2 through 2.5
describe potential costs associated with a change in depreciation
method.
Under RRB accounting, track structure is capitalized when
initially laid and not depreciated until the line is retired, when
it is written-off entirely. Any repairs or replacements made to
track are immediately expensed whereas the improvement portion of
betterments is capitalized. For example, if 100-pound track is
replaced with 1 2 0-pound track, the current cost of a 1 0 0-pound rail
is written off and the difference in cost between the 1 0 0-pound and
the 120-pound rail is capitalized. Each year approximately 70% of
all track expenses classified as maintenance under RRB would be
capitalized expenditures under ratable depreciation.^
RRB accounting was developed by the railroads before the ICC
developed the first Uniform System of Accounts in 1914 and before
the imposition of a federal income tax. A historical review
suggests that RRB accounting evolved as an efficient solution to
state tax and rate making influences (see Appendix A). A change in
accounting for track structure moves a railroad away from its
optimal accounting technique set and reduces firm value by the net
^This estimate was calculated based on data disclosed by the
ICC in Transport Statistics in the United States: Part 1, Annual.
8
increase in contracting and regulatory costs. Costs associated with
the change from RRB to ratable depreciation for track structure are
discussed in the following sections.
2.2 Tax costs
The largest expected cost of a change from RRB to ratable
depreciation of track structure was the threat that the Internal
Revenue Service (IRS) would require tax depreciation to conform with
book depreciation. The IRS had informally allowed RRB accounting as
long as it had been used for regulatory purposes. If the ICC were
to switch to ratable depreciation, the IRS could be expected to
follow suit. Suddenly, 70% of track maintenance expenditures being
expensed as replacements under RRB would be recovered over a
substantially longer period of time. Furthermore, replacements
qualified for the 10% investment tax credit under I.R.C. Section
48(a)(9) even though the costs were immediately expensed.
The tax cost to an individual railroad would depend on the
extent to which it had taxable income, tax benefit carry forwards,
and planned track expenditures.
2.3 Regulatory costs
In addition to the tax costs discussed above, the railroad
industry would face regulatory costs with a change from RRB to
ratable depreciation because of its potential impact on Commission
actions. The ICC has jurisdiction over rates only when a carrier is
9
said to have market dominance and a shipper has filed a petition
requesting intervention. The sole determinant of market dominance
is the cost recovery percentage (CRP), which is the lowest revenue-
to-cost ratio at which all profitable movements could be handled to
allow a carrier to break even. If a rate returns revenues greater
than a fixed percentage of costs, the ICC may consider rate
reasonableness. Rate flexibility would also be affected because
even controlled rates may freely increase by a rail cost adjustment
factor (RCAF). The RCAF is developed using basic cost categories
such as material, supplies, labor, and depreciation. Lastly,
carriers not earning adequate revenues are given greater flexibility
in raising rates and/or cancelling service. Return on investment
(ROI) is the standard used to determine revenue adequacy. Under
ratable depreciation, the ROI would generally increase, the RCAF may
increase at a slower rate, and the CRP may decrease, principally
because of the reduction in reported expenses.*^
A change from RRB to ratable depreciation would decrease railroad
reported expenses in two ways. First, new track and abandonments
are expensed all at once under RRB, and only gradually under ratable
depreciation. Thus, railroads that are conducting extensive
maintenance or abandonment programs will show lower costs and higher
revenue-to-costs ratios than they would under RRB. Secondly, under
ratable depreciation, depreciation charges are based on historical
^Alternative Methods of Accounting for Railroad Track
Structures. Docket No. 36988, 367 ICC 157 (January 26,1983).
10
cost rather than replacement costs as they are under RRB. Also, to
the extent that reduced expenses would offset the expansion of the
investment base (resulting from recording track investment at the
cost of the most recent installation rather than at original cost),
a railroad's ROI would appear higher under ratable depreciation than
under RRB. Higher income and ROI would increase a carrier's rate
regulation exposure and potential regulatory costs.
This chapter discussed the evolution of RRB accounting and
identified potential costs of the switch to ratable depreciation.
The next chapter traces the events that ultimately resulted in the
demise of RRB accounting and the rise of three different methods of
ratable depreciation for IRS, SEC and ICC purposes.
11
3. CHRONOLOGY OF EVENTS
3.1 The 4-R Act
The Railroad Revitalization and Regulatory Reform Act5 of 1976 (4-
R Act) provided the impetus to reconsider the appropriateness of RRB
accounting. Prior to the 4-R Act, the ICC had exclusive
jurisdiction over prescription of railroad accounting practices.
The 4-R Act called for the Department of Transportation (DOT) to
review the Uniform System of Accounts and gave the Securities
Exchange Commission (SEC) authority to prescribe reporting rules and
regulations for railroad companies. See Appendix B for a
description of the roles these agencies play in regulating
railroads.
In conjunction with a cost accounting modernization of the Uniform
System of Accounts, the DOT filed a formal petition with the ICC
seeking a change to ratable depreciation.5 The DOT expressed
concern that betterment accounting may provide management with an
incentive to defer maintenance. On April 19, 1977, the ICC denied
the DOT's petition.7
The SEC then exercised its new authority over railroads by issuing
an Advance Notice of Proposed Rulemaking (ANFR) inviting comment as
5PL 94-210, 94th Cong., 2nd Sess., 1976.
^Uniform System of Accounts for Railroads. 361 ICC 120 (1977).
A review of the Uniform System of Accounts was requested by Congress
in view of the government's huge financial commitment to revitalize
railroads. Account revisions proposed by the DOT were generally
cost accounting modernizations.
7361 ICC 120 (1977).
12
to whether betterment accounting should continue to be an acceptable
accounting principle for railroads reporting to the SEC.® Boberg
(1983) reviews written submissions filed in response to the ANPR.
The ICC submission asked the SEC to postpone its ruling on
depreciation until issues were resolved, such as; estimated life and
salvage value, retroactive restatement, and units of property
definition. Boberg states:
Even if the issues were to be resolved, the Commission foresaw
troubling implications of a change to ratable depreciation. Of
primary concern was the potential for an increased tax burden on
the railroad industry without a corresponding increase in
financial resources. The Commission also realized that a change
to ratable depreciation would affect areas of railroad rate
regulation such as the rate base, reported earnings, and rate of
return on net investment [emphasis added],®
After reviewing comments submitted, the SEC took no further action
with respect to RRB. The SEC later refused a General Accounting
Office (GAO) recommendation that the SEC complete its hearings.1®
On February 15, 1978, the ICC Bureau of Accounts held an informal
conference to study RRB vs ratable depreciation with particular
interest in the tax and regulatory issues identified above.11 Later
in the year, the ICC served a Notice of Study (NOS) to request
formal comment, again emphasizing the need to resolve the above
®SEC Release Nos. 33-5824 and 34-13479, April 28,1977.
^Boberg (1983) summarizing written submission of A. Daniel
O'Neal, Chairman, Interstate Commerce Commission, in SEC File No.
S7-692, June 17, 1977.
1®General Accounting Office, Accounting Changes Needed in the
Railroad Industry (Washington, D.C.: GPO, February 1981).
11Docket No. 36557, February 15-16, 1978.
13
issues before taking action.^ Railroads lobbied for relief from
tax and regulatory costs of the change to ratable depreciation. Tax
relief was first to be granted.
3.2 Codification of RRB
Congress legislated RRB accounting into the Internal Revenue Code
on December 13, 1980.-^ The codification of RRB eliminated the tax
i
i
cost of a change in accounting method by preventing the IRS from
requiring ratable depreciation. Resolution of the tax issue also
increased the probability that a change to ratable depreciation
would be required for regulatory purposes. This expectation was
reinforced in February 1981, when the General Accounting Office
called for a change to ratable depreciation.^ The ICC intensified
its review of the subject. Refer to Appendix C for event dates.
3.3 Economic Recovery Tax Act of 1981
The tax situation reversed less than 9 months later when the
Economic Recovery Tax Act of 1981 (ERTA) repealed the recently
legislated RRB and replaced it with ACRS. Initial ERTA hearings
reveal that a change from RRB to ACRS ratable depreciation was
-*-2"Alternative Methods of Accounting for Railroad Track
Structure", 43 F.R. 50717, October 31, 1978.
■^Miscellaneous Changes - Tax Laws, P.L. 96-613, 96th Cong.,
2nd Sess., 1980.
^General Accounting Office, Accounting Changes Needed in the
Railroad Industry (Washington, D.C.: GPO, February 1981).
14
viewed as a loss by the railroad industry. At best, the industry
would trade an immediate write-off of track replacements for
recovery over 5 years.-'--’ Transition rules were later added to the
tax bill which would be expected to cause a revaluation of the
railroads.
3.4 Transition rules
Phase-in provisions were subsequently added to proposed
legislation. The provisions provided for RRB to be phased out over
4 years and replaced with ACRS. With this rule there was no
immediate cost, rather a first year benefit, of a change from RRB to
ACRS. All replacement property was still expensed in 1981, but all
betterments and additions became 5 year property. A second
transition rule allowed railroads to recover costs of previously
capitalized track over any period between 5 and 50 years. The
"frozen base" cost of existing track could now be written off using
any depreciation method listed in IRC Section 167(b), including the
^It was not clear that track would be treated as 5-year
property. A longer life was possible. A 10-year life was proposed
for other rail assets such as tank cars.
-*-^The recovery percentages for replacement property during the
transition period would be as follows:
Percentage of Cost Deductible
Year placed in Service_______1981_____1982_____1983____1984
Ownership year:
1 100 50 33 25
2 50 45 38
3 22 25
4 12
15
double declining balance method for the early years and switching to
the Siam of the years digits method at a time to maximize the
acceleration of deductions.^ A change of this magnitude was
unprecedented. I.R.C. Section 481 normally applies to tax
accounting method changes. A Section 481 restatement would have
resulted in an adjustment factor that would have increased taxable
income by the amount of previous expenditures that had been expensed
under RRB, but would have been capitalized under ratable deprecia
tion and not yet fully depreciated.^-® Such adjustments may be taken
into income over 10 years. Under the transition rule, no
restatement was required. The "frozen base" write-off generated
large tax deductions as seen with this excerpt from Burlington
Northern Inc.'s 1981 annual report.
The Act [ERTA] provided that costs capitalized under the RRB
method and not recovered through retirement as of December 31,
1980, should be ratably depreciated. Primarily as a result of
•^For example, a railroad could recover the so-called "frozen
asset base" over a five-year period using the following schedule of
deductions:
Year Percentage of Basis Deductible
1981 40
1982 24
1983 18
1984 12
1985 6
l®Under Section 481 the book value of the track would have been
restated to reflect (a) the historical cost of the track actually in
place, including what had been previously expensed as replacements,
and (b) the accumulated depreciation to the year of the change that
would have resulted if the ratable method had been used.
■^The adjustment factor would have been at least partially
offset by depreciation deductions on the restated track base.
16
this change, we incurred a current tax net operating loss of
$104,736,000, which will be carried forward.
3.6 Inflation Accounting
After ERTA, the only significant issues remaining regarding a
change to ratable depreciation were implementation and impact on
regulatory proceedings. A change to ratable depreciation would have
expanded ICC authority to set rates and control service.
Deregulatory forces and railroad lobbying led the ICC to adopt a
third method of accounting for regulatory purposes -- inflation
adjusted asset base and depreciation charges.
In March 16, 1982, the ICC requested comment in an attempt to find
an appropriate solution to the impact of inflation in regulatory
matters.20 The final rulemaking called for all assets and related
depreciation charges to be adjusted for inflation.21
This Chapter discussed the events which led to the adoption of
ratable depreciation. In the next Chapter the tax and regulatory
costs of the change in depreciation method are modeled.
2047 FR 11539 (March 17, 1982).
21367 ICC 157 (1983).
17
4. COST IDENTIFICATION AND HYPOTHESES DEVELOPMENT
4 .1 Int roduc tion
Accounting theory suggests that a mandated accounting change will
have an impact on the value of firms affected if the change has an
associated cost or benefit. Using notation adapted from Leftwich
*
(1981) , let MVj[ be the market value of firm i if investors are
certain that there will be no mandated accounting changes. Thus, at
time t,
J
MVit = MVi - 2 Pjt C-j-j (4.1)
j-1
where
MVit = market value of firm i at time t,
Pjt = investors' estimate of the probability of state j
at time t,
Gjj = present value cost of the change in accounting
rules in state j to firm i, and
J = number of states.
The present value cost of the change, C^j, is conditional upon state
j occurring. If investors receive information in period t+1 which
causes them to revise their expectations of an accounting change
being mandated, they will revise their assessment of the market
value of the firm. Thus,
J
A MVi = 2 (Fjt - Fjt+l) cij (4-2)
j=l
where
A MV± = MVit+l - MVit
This model illustrates that an event which causes a change in
either the cost of an accounting change or the likelihood that the
change will be endorsed will cause an adjustment in the market value
18
of a firm, unless the changes offset one another.22 Chapter 3
identified tax and regulatory costs associated with a change from
RRB to ratable depreciation. Tax costs were later compensated for
by transition rule tax benefits. In the following sections tax and
regulatory costs/benefits are put into equation form. Chapter 6
addresses issues involving measurement of the variables in the
following equations.
4.2 Tax costs
It is posited that if the ICC mandated a change to ratable
depreciation, railroads would face a high probability of increased
tax payments because the IRS could be expected to conform to ICC
track accounting rules. Ratable depreciation would result in lower
tax deductions in early years following the change. Tax cost is
defined as the present value change in tax liability. The change in
taxable income in year t equals replacements for year t, which are
capitalized instead of expensed, less the depreciation deduction
allowed for replacements and betterments made for all year's
subsequent to the depreciation change to year t. The tax cost to a
firm can be expressed as follows.
22For example, an announcement that an accounting change is
more likely because opposing parties have reached a compromise on
implementation issues may leave firm value unchanged. Investors may
revise probability assessments upwards, but cost assessments downward
19
t_1 iRk-±-lkl
)
T
TC
-k=l-
(4.3)
t
t=l
(1 + i2)
z=l
where
TC = present value tax cost of change to ratable depreciation
Rt = track replacements in year t
Bt = track betterments and additions in year t
L = life over which track is depreciated
i = discount rate
rt = marginal tax rate in period t
T = life of the firm
If replacements are greater than the depreciation charge, the
accounting change is income increasing. In present value terms, a
change to ratable depreciation bears a tax cost in all but the most
unusual circumstances.^ In absolute dollars, the depreciation
charge will become greater than year t replacements at some point in
the future and result in a tax savings. This is because a change
from RRB to ratable depreciation changes only the timing, not the
amount of depreciation charged. However, increasing costs,
expansion, and changing tax rates could indefinitely postpone the
realization of future tax savings arising from timing differences.
Even if timing differences do turnaround, a modest recovery period
2^For example, a firm in the position of having net operating
losses expire without benefit or present value of future tax rates
exceeding current tax rates would experience a benefit of such a
change. One firm is found to experience a tax benefit due to
rapidly contracting operations combined with NOL carryforwards.
20
causes the net change in firm value to be negative.^ From this
discussion, the following hypotheses are inferred:
H]_: An unexpected event that increases investors' probability
assessments of a change to ratable depreciation being
required for tax purpose would reduce the value of the firms
affected.
H2 - The decline in value is an increasing function of the
expected tax costs.
To test these hypotheses, investor expectations of the tax cost of
a change to ratable depreciation, as specified in Equation 4.1, must
be identified. Measurement of these variables is discussed in
Chapter 6.2.
4.3 Regulatory costs
A change from RRB to ratable depreciation would decrease railroad
costs, increase reported revenue to cost ratios, and increase return
on investment. A carrier's rate regulation exposure and potential
regulatory costs would increase. Regulatory costs consist of lost
revenues due to rate intervention, the cost of fighting
intervention, and losses incurred on unprofitable lines because of
forced continuation of service. Therefore, it is hypothesized that:
H3 : An unexpected event that increases investors' probability
assessments of a change to ratable depreciation being
required for regulatory purposes would reduce the value of
the firms affected.
^The General Accounting Office used 30 year depreciable lives
in its proposal recommending a change to ratable depreciation.
There is no indication what depreciable life the IRS would allow.
Track was not included in the Asset Depreciation Range (ADR)
guidelines because of the longstanding use of RRB. However, a 10-
year life had been assigned to most other rail assets.
21
H4 - The decline in value is an increasing function of the
increase in expected regulatory costs.
Regulatory costs of an accounting change, like opportunity costs,
cannot be observed directly. However, rate intervention will occur
only in market dominant situations and potential abandonments are
identified as such up to three years before abandonment requests are
filed with the ICC. Costly service intervention is less likely to
take place when a carrier is earning inadequate return on investment
(ROI). The cost of disputing intervention is assumed to be a
constant multiple of other regulatory costs and will be reflected in
the scale of coefficients on regulatory variables. The increase in
expected regulatory costs is the increase in probability that
intervention will occur times the cost of the increased
intervention. If an investor receives information in t+1 that
leads him to revise his estimate of expected regulatory cost of a
change in depreciation methods, he will revise his estimate of the
firm. This revision in firm value equals the change in expected
regulatory costs. The change in probability that regulatory
intervention will occur is, in part, a function of the increase in a
firm's ROI caused by the change in depreciation methods. The change
in probability is also a function of the change in the Cost Recovery
Percentage and the Rail Cost Adjustment Factor caused by the switch
to ratable depreciation. Regulatory costs (RC) equals present value I
income lost due to ICC intervention. The change in firm value is
expressed below.
22
J
where
A MV£ = S (PJt - Pjt+l) RCij (4.4)
j=l
A M V = MVit+1 - MVit
MVj[ = Market value of firm i
Pjt = investors' estimate of the probability of
state j,
RCij = regulatory cost of the change in accounting
rules in state j to firm i,
t+ 1 = time of event,
J = number of states, and
J
S (Pit - P-jt+l) = f (Event, ROI, RCAF, CRP)
j-1
RCij = / (LOSTREV, LOSSES) (4.5)
where
Event =
ROI =
RCAF =
CRP =
LOSTREV =
LOSSES -
event which changes probability of mandated
accounting change,
change in return on investment,
change in rail cost adjustment factor,
change in cost recovery percentage,
present value losses due to rate intervention, and
present value losses sustained on unprofitable
lines that may not be abandoned.
The equations above show that the change in firm value is a function
of several variables, some of which are unobservable. Furthermore,
these variables interact multiplicatively. Measurement of these
variables is discussed in section 6.3.
4.4 Codification of RRB
The first form of tax relief legislated by Congress was the
codification of RRB so that the IRS would be prohibited from
23
enforcing tax conformity to any new book depreciation method. The
codification of RRB significantly increased the probability that a
change to ratable depreciation would be mandated by the ICC but it
removed the threat of federal tax costs of the change. The market
value of the firms affected would increase to the extent that the
expected decrease in tax cost to zero was not offset by the increase
in expected regulatory cost caused by the probability revision.
Reversals will be discussed in Chapter 6.4.
4.5 Transition rule tax benefits
Phase-in provisions added to proposed ERTA legislation provided
for an immediate benefit of a change from RRB to ACRS. The
unexpected tax benefit provided by these rules is the excess tax
deduction that results when the phase-in provisions are applied to
replacement property instead of ACRS. The tax benefit to a firm is
expressed as follows:
TB = ri (.85R!) + r2 (.35 R2 - .22Ri) + r3 (.I8R3 + .28R2 -
.21R!> + r4 (.IOR4 + .23R3 - ,21R2 - .21Ri) + r5 (.I6R4
+ ,01R3 - .21R2 - .21RX) + r6 (.03R4 - ,21R3 -,21R2) +
r7 (-.09R4 - .21R3) + r8 (-.21R4) (4.6)
where
TB = tax benefit
rt = present value marginal tax rate of firm i in the tt^
year of the change
Rt = replacements made in year t of the change
A second transition rule allowed railroads to recover track
previously capitalized under RRB over any period from five to 50
24
years using any depreciation method listed in IRC Section 167(b).
Assuming the fastest method is chosen, the resulting tax benefit is
expressed as follows:
TB = FB (.4r]_ + . 2r2 + .18r3 + . 12r4 + ,06r5) (4.7)
where
FB = frozen asset base
TB = tax benefit
rt = present value marginal tax rate of firm i in the tt^
year of the change
4.6 Inflation accounting
The adoption of inflation accounting would have eliminated the
expected regulatory cost of ratable depreciation. Investors would
be expected to reverse their previous devaluations of firms caused
by expected regulatory costs.
25
5. EVENT STUDY
I
5.1 Methodology for event study
To test the hypotheses enumerated above, a two-step procedure is
used. First, several events are identified where information is
released regarding either the cost or probability of a mandated
accounting change. The market model is used to calculate the
abnormal returns for each railroad for a period surrounding each
event. Hypotheses 2 and 4 are then tested by estimating a cross-
sectional regression for each event. This chapter describes the
methodology used for the event tests. Chapter 6 describes the
cross-sectional tests.
5.2 Abnormal returns
Hypotheses 1 and 3 state that an unexpected announcement that
increases either the likelihood or cost of a change to ratable
depreciation will cause a decrease in the market values of the firms
affected. To test whether the value of a firm changes conditional
on new information, efficient markets and an equity valuation model
must be assumed. Christie (1987) shows that a returns model is
economically equivalent to the traditional equity valuation model.
A returns model is employed for this study. The market model is
used to generate expected returns. To allow for nonsynchronous
trading of securities, lead and lag market returns are included in
the market model during the estimation period. These parameter
estimates are then used to construct consistent estimators of each
26
firm's intercept and beta. Hereafter, these consistent estimators
are referred to as compound alphas and betas. See Scholes and
Williams (1977) for derivation of the consistent estimators. The
unexpected return for the event period is the sum of the daily-
unexpected returns from the day before to two days after the event
where each day's unexpected return is as follows.
Several events tested are regulatory announcements by the ICC and
legislative actions by Congress. The information in these
announcements may be relayed to the firms affected before the
official pronouncement. Since the exact day that the information
reaches the market cannot be identified four day event periods are
constructed.
5.3 Sample selection
All Class I railroads for which daily stock returns are supplied
by the University of Chicago's Center for Research on Security
Prices (CRSP) are included in the sample. Class I railroads are
defined as those carriers with freight revenues in excess of $50
million. Class I roads carry 97% of all freight moved by rail in
the United States. At the beginning of the test period there were
uit _ rit ” ( “i + Pi rmt )
(5.1)
where
rit
rmt
Realized return to firm i in period t
Realized return to market in period t
Compound market beta for firm i
Compound alpha for firm i
Unexpected return for firm i in period t
27
34 Class I carriers, many of whom are owned by parent corporations.
There are 21 parent corporations for which stock market data is
available. During the event period, mergers in the industry reduced
the number of parent corporations to 15. The 21 parent corporations
with their largest railroad subsidiaries are listed in Appendix D.
28
6. CROSS-SECTIONAL TESTS
6.1 Cross-sectional tests
This chapter describes the cross-sectional tests performed and
discusses measurement of the independent variables used. To test
hypotheses 2 and 4, which state that the abnormal returns are a
function of tax and regulatory costs associated with the change, a
cross-sectional regression for each event is estimated.
uit = «i + «]_(TCit/Sit) + S2W Dit/Sit) + *3(Ait/sit) + eit (e*1)
where
Uj[t = Unexpected return for firm i in event period t
TC^t = Tax cost/benefit for firm i in event t
MDit = Market dominant freight revenue for firm i at t
Ait = Miles of abandonable line for firm i at t
S^t = Opening market value of equity
Measurement of these variables is discussed in Sections 6,2 and
6.3. Some methodological issues are addressed in Appendix E.
6.2 Tax costs
The expected tax costs of a change from RRB to ratable
depreciation equals the resulting timing difference multiplied by
the present value of the tax rate. Expected track expenditures,
that give rise to expected timing differences, are measured assuming
3-year perfect foresight. Track expenditures beyond three years are
expected to increase by the average annual increase. This
methodology incorporates rising prices. The present value tax rate
is the marginal tax rate the firm is subject to on the resulting
timing difference. If the tax on the timing difference is paid in
29
the future, the future marginal rate must be discounted to the
present.
Because the few firms in the railroad industry are required to
report significant tax data to the ICC, marginal tax rates can be
calculated with less measurement error than for other industries.
Railroads are required to report net operating loss and investment
tax credit carryover amounts and the year the benefits are expected
to be used. In instances where a carrier files a consolidated tax
return with an affiliated group, the marginal tax rate for the group
is the relevant variable. The relevant marginal tax rate is the
combined federal and state rate. It is assumed that the state rate
is simply a multiple of the federal rate that will scale the
coefficients. This constraint will be relaxed in the future since
railroads disclose accrued taxes by state. Finally, to calculate
the marginal tax rate, figures must be adjusted "as-if" the
accounting change of interest had taken place.
"As-if" taxable income is calculated as the current period book
income before taxes, adjusted for the change to ratable deprecia
tion, and adjusted for items for which deferred taxes have been
provided. "As-if" income is also calculated assuming 3-year perfect
foresight of annual report data. This methodology mitigates the
impact of alternative tax strategies available to many of the
railroads.^ A switch from RRB to ratable depreciation would
^An alternative assumption of "as-if" income following a
random walk process, where the last observation is the best
predictor of future observations, was found to be grossly inadequate
30
increase taxable income. Firms in an "as-if" tax paying situation
will be assumed to have a marginal rate of 46% for all future years,
discounted to the present by the cost of capital. For firms with
positive "as-if" income but no current tax expense because of an NOL
carry forward, the year the firm is expected to be in a tax paying
situation is computed if it is not within the 3-year foresight
period. The NOL is divided by "as-if" taxable income to determine
the year in which the firm will be in a tax paying situation. If a
firm is in a loss "as-if" situation for the foresight period, it is
assumed that it will not be paying taxes in the foreseeable future.
Final determination of when taxes are paid must also consider
existing investment tax credits (ITC). The interaction of ITC and
NOL carry forwards and ITC limitations complicate the determination
of the year that taxes are paid. Because of the small sample size,
all tax calculations are done firm by firm. "As-if" calculations
are carried out for a moving 10-year period. After 10-years, the
difference between ratable depreciation and RRB is small and firms
will have had plenty of opportunity to revise tax strategies.
6.3 Regulatory costs
Regulatory costs are identified in Section 4.3 to be the product
of the change in probability that ICC intervention will occur and
the cost of that intervention. The cost of intervention consists of
because of the large degrees of freedom most carriers experience in
controlling taxable income. For example, taxable income can be
generated by developing land or mineral holdings, cutting timber, etc
31
revenues lost due to rate intervention and losses sustained on lines
for which approval to abandon cannot be obtained. These costs
cannot be observed. It is assumed that the costs are cross-
sectionally linear in freight revenues subject to intervention and
miles of abandonable track. Abandonable lines are miles that a
railroad has identified as such. Identification is a matter of
record and takes place up to 3 years before permission to abandon is
requested. Freight revenues subject to intervention are movements
where the cost recovery percentage is exceeded and the ICC has
jurisdiction under the definition of market dominance. The cost
recovery percentage is generally exceeded for shipments of coal.
Serving these "captive shippers" exposes a carrier to rate
investigations. The impact of these rate investigations on a firm
can be material, as indicated in Norfolk Southern's 1982 contingent
liability footnote.
The ICC is engaged in a proposed rulemaking proceeding ... in
which it is considering guidelines to determine the
reasonableness of coal rates nationwide. The ICC originally
proposed for consideration fully allocated costs as the standard
of reasonableness. After additional comments by the parties,
the ICC . . .modified the cost methodology originally proposed
and expressed the ICC's conviction that rate reasonableness
cannot be determined solely by a strict cost approach. This
decision was appealed to the US Court of Appeals.. . .On July
30, 1982, the ICC reopened its interim decision ... On
February 24, 1983, the ICC served its further decision proposing
a maximum rail rate policy applicable to captive coal traffic
which adopts a differential price standard. Railroads with
inadequate revenues would be free to maintain and increase coal
rates which do not exceed "stand-alone" costs, on a depreciated
current cost basis . . .
The above excerpt reflects the complexity of rate making and other
ICC regulatory proceedings. Information used by the ICC to reach
32
decisions are generally available only to the parties involved. No
one, including the ICC, knows what percentage of freight carried is
under ICC market dominant jurisdiction. Abandonment proceedings are
also vague. Because regulatory costs cannot be estimated, it is
assumed that they are linear in ton miles of coal carried and miles
of potential abandonments.
The change in return on investment, the rail cost adjustment
factor, and the cost recovery percentage are identified in Section
4.3 as affecting the probability of ICC intervention. The change in
probability of ICC intervention is impounded in the coefficients of
the independent variables as is the linear relationship required to
transform the included independent variables to dollar costs of the
regulatory intervention.
6.4 Codification of RRB
The codification of RRB would have caused investors to reverse
prior tax cost expectations while increasing regulatory cost
expectations due to the increased probability that a change would be
required for regulatory purposes now that there were no federal
income tax consequences. The reversal in tax cost is simply modeled
as the negative of the tax cost calculated above.
6.5 Transition rule tax benefits
The transitional rules provided by ERTA gave rise to tax benefits
for which the schedule of marginal tax rates must be identified. An
33
analysis similar to the one performed in Section 6.2 above will be
performed with additional consideration for carryback potential.
The tax benefit will first offset current tax expense for the
current and three prior years. The remaining tax benefit will then
be assumed to offset tax-affected and deferred tax adjusted "as-if"
operating income for future periods. The transition rule tax
benefit was so large that most firms disclosed it separately in the
tax footnote.This excerpt from Missouri-Kansas-Texas 1981 tax
footnote illustrates the magnitude by which the frozen base writeoff
exceeded all other ERTA provisions.
The Economic Recovery Tax Act of 1981 (ERTA) resulted in $25
million of additional deductions in 1981 because of a provision
which permits the company to depreciate the unrecovered cost of
track structure at December 31, 1980 over a period of five years
beginning in 1981 using accelerated methods and an estimated
$700 thousand because of increased depreciation deductions under
the new Accelerated Cost Recovery System (ACRS) relating to
properties placed in service subsequent to December 31, 1980.
6.6 Inflation accounting
The adoption of inflation accounting for regulatory purposes
caused investors to reverse prior regulatory cost expectations. The
reversal is accounted for by changing the sign of the regulatory
cost variables.
2®An attempt was made to calculate the transition rule tax
benefits by applying the transition rules to track account balance
details reported by each railroad to the ICC. These calculations
resulted in estimated benefits much larger than those disclosed by
the railroads in tax footnotes. This indicates that the tax bases
for track assets were lower than the book bases. The cause of this
difference cannot be determined in all cases. One firm disclosed
that it had depreciated track structure for several years while it
was also expensed replacement track. The resulting tax savings were
disclosed as a contingent liability because of IRS assessments.
34
7. RESULTS
To test the hypotheses enumerated in Chapter 4, it was necessary
to identify event dates, generate unexpected returns for the event
period, and regress them on unexpected tax and regulatory
costs/benefits associated with each event. Event dates identified
in Appendix B are summarized in Table 2 with a predicted sign as to
whether the event has a cost (-), benefit (+), or neither (0)
because of taxes (T), potential market dominant coal ton-miles (C),
or potential abandonments (A). Summary statistics on the absolute
value of these benefits are presented in Table 3. Cross-sectional
event test results are presented in Table 4. All cross-sectional
costs were included in the cross-sectional tests as negative
numbers. So, all parameter estimates presented in Table 4 are
predicted to be positive, regardless of event.
Identification of event date is often difficult. Ideally, one
would like to know the exact time of day news of the event of
interest reaches market participants. Even when events are reported
in the financial press, it is generally not possible to isolate one
day as an event period. For industry specific regulatory events not
reported in financial newspapers event date identification is
particularly difficult. This limitation applies to all but three of
the 22 events studied. However, the events studied were published
in government agency releases. If the semi-strong theory of
efficient markets holds, the four day event periods constructed in
this study should have captured the market reaction.
35
Table 2
Summary of events tested
Event Description T c A
1 4/12/77 ICC denies petition to change from RRB + + +
2 4/28/77 SEC calls for change from RRB - - -
3 2/15/78 Informal hearings at ICC + + +
4 10/31/78 Invitation to comment issued by ICC + + +
5 7/9/79 Bill to codify RRB introduced + - -
6 7/27/79 House hearings + - -
7 10/22/79 Senate hearings + - -
8 12/13/79 Senate version of bill dies - + +
9 5/22/80 House recommends new RRB bill + - -
10 11/24/80 Senate recommends passage of HR 7171 + - -
11 12/13/80 PL 96-613 passes with RRB - f * - -
12 2/4/81 GAO publishes calls for change from RRB 0 - -
13 5/18/81 Tax hearings confirm that RRB property
would not qualify as recovery property - 0 0
14 6/12/81 Transition rules introduced + 0 0
15 6/22/81 ICC NOPR to change method 0 - -
16 6/25/81 Senate Finance approves RR provisions + 0 0
17 7/15/81 Ways and Means agree to RR provisions + 0 0
18 7/29/81 ERTA passes house + 0 0
19 8/3/81 ERTA passes Senate + 0 0
20 8/10/88 Reagan signs + 0 0
21 3/17/82 ICC Supplemental NOPR 0 + +
22 2/17/83 ICC Rulemaking 0 + +
T = Tax cost
C = Regulatory cost measured by coal ton-miles
A = Regulatory cost measured by miles of potential abandonments
36
Table 3
Descriptive statistics on independent variables
Standard
Mean Deviation Minimum Maximum Medial
1977
Tax .1965 .1215 .1047 .3953 .1752
Coal .0511 .0547 0 .1545 .0261
Abandon .0013 .0015 0 .0046 .0007
1978
Tax .1979 .1246 .0057 .4486 . 1948
Coal .0567 .0668 0 .2006 .0336
Abandon .0013 .0014 0 .0039 .0008
1979
Tax .1755 .1187 .0065 .3779 .1609
Coal .0585 .0645 0 .2013 .0459
Abandon .0010 .0011 0 .0030 .0006
1980
Tax .1735 .1251 .0108 .4190 .2798
Coal .0583 .0650 0 .2072 .0349
Abandon .0010 .0012 0 .0037 .0006
1981A
Tax .0513 .0351 - .0021 .1288 .0429
Coal -.0331 .0368 0 .1123 .0147
Abandon -.0006 .0009 0 .0032 .0003
19 8 IB
Tax .1337 .0663 .0100 .2129 .1484
1982
Coal .0414 .0441 0 .1239 .0814
Abandon .0007 .0010 0 .0040 .0002
37
Table 4
Cross-sectional regression parameter estimates
Event Average
Date df Reaction Intercept Tax Coal Abandon
(t statistics)
9/12/77 15 .027
(2.29)
.0391
(3.87)
.0023
(.04)
.0001
(.0 0 2)
-9.77
(-2.37)
4/28/77 15 .0067
(.65)
-.0039
(-.29)
-.0334
(-.79)
-.1546
(-1.18)
3.521
(.70)
2/15/78 15 -.0177
(-1.57)
-.0211
(-1.63)
.0646
(.82)
-.2415
(-1.93)a
3.416
(.60)
10/31/78 15 -.0032
(-.52)
-.0141
(-.92)
.0725
(.77)
- .2335
(-1.62)b
7.606
(1.03)'
7/9/79 14 .009
(.79)
.0145
(1.94)
-.0484
(- .96)d
.0029
(.04)
-4.984
(—.97)d
7/27/79 14 .0099
(.50)
.0015
(.15)
.0707
(1 .0 2)d
.0819
(.76)
4.407
(-61)
10/22/79 14 -.0220
(-1.56)
- .0299
(-2.28)
-.0644
(-.76)
-.2249
(-1.75)b
-4.921
(-.56)
12/13/79 14 -.0079
(-.18)
.0271
(1.63)
.2705
(2.35)a
.0728
(.42)
9.772
(-84)
5/22/80 14 .0099
(.61)
.0259
(1.85)
-.1023
(-1.18)c
-.0625
(-.40)
1.75
(-2 1)
11/24/80 11 -.0012
(-.03)
-.0078
(-.60)
.0473
(.33)
.233
(.71)
-14.09
(-1.26)'
12/13/80 11 .0119
(.61)
.0041
(.27)
.1922
(1 .2 2)c
.5726
(1.59)b
-4.134
(-.34)
2/4/81 13 .0085
(.40)
-.0060
(-.51)
.3529
(1.87)a
3.010
(.38)
5/18/81 14 -.0100
(-.52)
-.0063
(-.48)
.0683
(.33)
6/12/81 14 -.0035
(-.05)
-.0055
(-.35)
.0163
(-16)
38
Table 4, Continued
Event
Date df
Average
Reaction Intercept Tax Coal Abandon
(t statistics)
6/17/81 14 .0023
(.23)
-.0014
(-.1 1)
.0247
(.27)
6/22/81 13 .0305
(1.87)
.0154
(1.69)
.1476
(,89)d
9.991
(1 .0 0)
7/18/81 14 .0127
(.61)
-.0006
(-.03)
.1098
(.82)
7/29/81 14 .0032
(.19)
-.0066
(-.36)
.0704
(.58)
8/3/81 14 .0074
(.52)
.0334
(1.44)
- .1981
(-1.28)c
8/10/81 14 -.0079
(-.51)
-.0250
(-1.46)
.1279
(l.H)c
3/17/82 14 -.0179
(-1.09)
-.0196
(-1.33)
.0715
(.27)
-5.67
(-.51)
df = degrees of freedom
a = significant at .05
b = significant at .10
c = significant at .15
d = significant at .20
39
An interesting observation in Table 2 is that events that a casual
observer might expect to be bad news events were actually good news
events. For instance, events three and four were ICC actions
regarding switching from RRB to ratable depreciation. Regulatory
actions are generally expected to be bad news. However, a careful
reading of the ICC releases show that these events gave the ICC
opportunity to publicly support the railroads by stating that very
serious issues would have to be resolved before depreciation methods
could be changed. This situation illustrates that identification of
good verses bad news may be a matter of investor and researcher
interpretation of the event details.
The cross-sectional results presented in Table 4 yield two
additional observations. First, tax events appear to be closer to
significance levels in the right direction upon a resolution of
legislation rather than during Congressional negotiations or
hearings.^ Secondly, the statistical significance of the average
event reaction is not a good indication of the importance of a
particular event to an industry. The cross-sectional variation in
the tax and regulatory variables causes some firms to react
positively and others negatively to the same event. If cross-
sectional variation is expected, it may not be wise to select which
events to test in cross-section based on the significance of the
2^0n December 13, 1979 the bill to codify RRB was killed in the
Senate. On December 13, 1980 a new version of the same bill passed.
ERTA reaction is marginally significant in the predicted direction
only the day the President signs the bill.
40
average event reaction. This selection process was used by Leftwich
(1981) to omit 12 out of 21 events from his cross-sectional tests.
In this study, all events are tested,
The results presented in Table 4 generally are not significant.
Christie (1989) notes that insignificant test statistics is common
in studies with small sample sizes and suggests an asymptotic test
that aggregates test statistics. Using this methodology, the
aggregate z statistic for the tax variable was computed to be 1.055,
significant only at the p=.15 level for a one tailed test. The
aggregate statistic yields weak support for market reaction to
expected tax costs/benefits. The coal ton-miles and miles of
potential abandonments aggregate test statistics are insignificant
and the coal statistic is in the wrong direction.
For events prior to ERTA legislation insignificant parameter
estimates on tax variables might not be surprising. The parameter
estimate for the tax variable can be interpreted as the revision in
probability that a switch from RRB to ratable depreciation would be
required for tax purposes. Investor's probability of increased tax
costs to railroads may have been low in light of the federal
government's large and increasing investment in railroads. Tax laws
that were introduced and ultimately passed were simply codification
of existing practice. Insignificant results for events involving
ERTA transition rule benefits are harder to understand. The
magnitude of the benefits were large enough to swamp any confounding
tax changes; the benefits were unexpected and they were negotiated
41
separately from general tax reform provisions. Possible explana
tions are that news of the transition rules did not reach market
participants in a timely manner, or that the news was not
understood. The Wall Street Journal first mentioned transition rule
tax benefits on October 27, 1981, nearly 3 months after passage of
ERTA. Railway Age did not report on the tax bill until November
1981. In this article, Richard E. Briggs, then executive vice
president of American Association of Railroads, voiced concern over
losing RRB'and downplayed the transition rule benefits.
. . . the big fault I find in those who talk only about the
advantages in the early years of writing off the so-called
'frozen base' [is that] they don't remember that the railroad
industry is, so far as I know, the only major industry that had
regular tax depreciation for one of its largest assets reduced.
The effect of going from one to the five year write-off is going
to be very significant--and adverse.
Perhaps the industry considered it politically unwise to publicly
announce its windfall gains under ERTA.
The insignificance of the parameters might be caused by errors in
the variables or correlated omitted variables. Both problems result
in biased estimators. Expected tax costs/benefits are carefully
measured using perfect foresight assumptions for annual report data.
However, the miles of abandonable track variable may measure the
regulatory cost of service intervention with error. This variable
was chosen as a measure of regulatory costs expected to be borne by
carriers who may be forced to continue service on lines because
ratable depreciation would result in lower reported costs of
operating such lines. One marginal firm disclosed over 7300 miles
42
of potential abandonments. It is unreasonable to expect the ICC to
force such a marginal carrier to continue service on unprofitable
lines indefinitely. This carrier consistently showed up as an
outlier in the cross-sectional tests. After several attempts to
come up with more reasonable measures for this firm failed, it was
omitted from the tests. The firm eventually entered bankruptcy
proceedings. Ton-miles of coal carried is included as a measure of
regulatory costs due to rate intervention. Again, this variable
measures regulatory costs with error. Also, coal carried may be
correlated with other variables that explain unexpected returns but
are not included in the model. For example, relative price of
alternative energy sources causes some variation in operating costs
of all carriers but a wide fluctuation in revenues of coal carriers.
Labor unrest by the United Mine Workers, weather conditions,
progress of coal slurry pipelines and waterway conditions all may
help explain unexpected returns and be correlated with coal ton-
miles . Although no announcements of these sort were identified as
taking place on the same dates as the events studied in this paper,
it is possible that such announcements ocurred but were not
published in the Wall Street Journal and railroad trade magazines.
Another possible explanation as to why the regulatory cost
parameter estimates are insignificant and often the wrong direction
comes back to the problem of interpretation of event news and
formation of expectations. Although the ICC and the industry
claimed that potential regulatory costs associated with a change in
43
depreciation methods were significant, market expectations of these
costs may have been nominal. The events may have been interpreted as
political statements voiced with the intent for railroads and the
ICC to retain control over accounting policies. Perhaps the major
determinant of investor expectations of regulatory costs is whether
the ruling Commissioners generally act with deregulatory intent,
despite of official releases.
In short, this study finds weak support for the existence of tax
costs when'all events are aggregated, but no support for the
existence of regulatory costs. Because tax costs have direct cash
flow impact, market reaction is expected. In particular, market
reaction to unexpected transition rule tax benefits that average 13
percent of market value of the firms affected are expected to be
significant. However, cross-sectional parameter estimates were not.
This result has implications for future tax research. It would be
interesting to know whether market reaction to similar unexpected
tax legislation amendments or transition rules are also insig
nificant. Is a market reaction more readily detectible when the
financial press reports expected tax costs/benefits rather than when
no estimates are readily available? Are obscure or complex tax
ramifications reflected in market prices gradually rather than
instantaneously? If so, does this market inefficiency result
because of complexity in the tax law or because of investor
inability to form rational expectations of tax costs/benefits?
Before progress can be made in understanding market reaction to tax
44
events, a model of expected taxable income must be developed and
tested.
This study also raises more questions as to the existence of
political costs of accounting methods other than taxes. A better
understanding of the political process is needed before we can
determine how accounting methods fit into this process. Perhaps
accounting numbers are used so seldomly in the political process
that expectations of political costs are immaterial. Perhaps
accounting numbers are used only to help distressed industries and
to punish flourishing ones. This study hypothesized that a
distressed industry would suffer regulatory costs of a change in
accounting method. If accounting numbers are used in the political
process only to help distressed industries there would be no
regulatory costs of a mandatory accounting change on railroads.
Future research efforts may be enhanced by evidence on when
accounting procedures enter the political process.
An extension of this research may be to incorporate substitute or
competing industries into an analysis of market reaction to tax
changes. For example, good news to railroads may yield negative
abnormal returns to trucking companies. Sufficiently large and
relatively unexpected tax changes that lend themselves to market
test are likely to be industry specific. However, the small number
of firms in any one industry may make cross-sectional tests
difficult unless there are multiple events. Incorporating competing
industries into the model may help overcome sample size problems.
45
This chapter presented the results of the study and suggested
areas of future research. In summary, only weak support was found
for changes in expected taxes when test statistics were aggregated,
and no support was found for changes in expected regulatory costs of
a mandatory accounting change.
46
APPENDIX A: HISTORICAL MOTE OM RRB
Two reasons have been offered for the development of retirement-
replacement-betterment (RRB) accounting (Boockholdt, 1978). The
first, and traditional, reason offered for RRB is one of asset
valuation. Because the railroads were among the first companies to
require major quantities of long-lived assets and massive amounts of
outside capital, their accountants were the first to deal with the
problems of asset valuation and disclosure. Littleton (1953)
discusses early methods railroads used to record assets. Methods
can be classified as periodic revaluation, actuarially computed
reserves for future replacement, and direct expensing of replace
ments . The replacement method was sometimes accompanied by a
renewal reserve for the excess of repairs and replacements that
should have been made over actual expenditures. The reserve would
be drawn down in future periods when previously deferred maintenance
would be undertaken. Reserves were presented the same way as modern
day accumulated depreciation. The fundamental tenet of the
replacement (or retirement) method is that as long as property is
maintained in proper repair, no decline in value occurs. Only when
property is abandoned does it lose functional capacity and value.
The replacement method was originally used for all assets.
The second reason for the development of RRB offered by Boockholdt
is management policy in pursuing the best interests of the
railroads. He notes that in 1846, the Commonwealth of Massachusetts
47
expropriated any earnings in excess of 10% by way of a tax. In
1876, the Massachusetts Railway Commission issued a revised set of
instructions regarding railway accounts acknowledging the retirement
method by making no mention of depreciation, but calling for the
reporting of new assets charged to operating expense to make good
original numbers. Boockholdt observes that
of all the railroads referred to by Littleton and Mason in their
discussions of early depreciation methods, the only ones which are
known to have repeatedly reflected such charges in their accounts
operated in the State of Massachusetts.^®
Boockholdt suggests income fluctuations and excess earnings taxes
could be mitigated by increasing the renewal account. Citing
Ferguson (1916), Boockholdt notes that in the 30 years in which
Virginia had a similar law in effect, it never collected an excess
profits tax from a single railroad. If other states followed
Massachusetts in taxing earnings in excess of a statutory level,
railroads had incentive to use replacement-renewal reserve
accounting.
State regulation expanded into rate setting with the passage of
The Granger Laws giving states the power to regulate commerce. Rate
regulation coincided with the demise of renewal reserves and the
tendency to capitalize assets other than track. Capitalization
increases the investment base. Boer (1966) explains that rates were
typically set to provide for a set return on investment. Regulation
2®Boockholdt, p.15.
48
then expanded into accounting practices. RRB persisted for track
structure and ratable depreciation evolved for other assets.
RRB accounting was developed by the railroads before the ICC
developed the first Uniform System of Accounts in 1914 and before
the imposition of a federal income tax. The historical review
suggests that RRB accounting evolved as an efficient solution to
state tax and rate making influences. These influences were
reinforced by the development of interstate rate regulation and
federal taxation. A change in accounting for track structure would
move a railroad away from its optimal accounting technique set and
reduce firm value by the net increase in contracting and regulatory
costs. Costs associated with the change from RRB to ratable
depreciation for track structure are discussed Chapter 3.
49
APPENDIX B: REGULATORY BODIES
The Interstate Commerce Commission (ICC) was created as an
independent regulatory agency by act of February 4, 1987 (49 U.S.C.
1-22), now known as the Interstate Commerce Act, to regulate
commerce. The ICC regulates interstate surface transportation,
including trains, trucks, buses, water carriers, freight forwarders,
transportation brokers and coal slurry pipelines. In broad terms,
Commission regulation encompass transportation economics and
service.
In the economics area, the Commission settles controversies over
rates. It rules on applications for mergers and acquisitions. It
prescribes accounting rules and administers laws relating to
railroad bankruptcy. The ICC has jurisdiction over the use,
control, supply and movement of railroad equipment.
In the transportation service area the ICC grants operating rights
and approves applications to construct and abandon railroad lines.
It is authorized to direct the handling and movement of traffic over
a railroad and its distribution over other lines of railroads. It
has the authority to prescribe inclusion of marginal railroads into
stronger systems.
Although these regulatory powers are quite broad, the
Commissions's statutory mandate has been reduced directly and
indirectly in the past 16 years. The June 1970 bankruptcy of Penn
Central, once the country's largest transportation company, and the
subsequent financial commitment by Congress to Conrail, led Congress
50
to evaluate ICG policies and procedures. The following passage
indicates the extent of Congressional dissatisfaction.
Of all the government regulatory agencies involved in the
Penn Central fiasco, the ICC stands out as the most inefficient
and ineffective of them all. Here was the agency primarily
responsible for overseeing the railroad's collapse, the ICC did
not investigate or issue any public statements regarding either
the financial or the operating condition of the railroad.
According to statements of various ICC officials, the Commission
was as surprised as everyone else when the Penn Central went
under.
Where was the ICC while the railroad was collapsing? The
ICC was certainly in the best position to know the internal
problems of the Penn Central and its critical financial
condition. If the ICC did not have such information about the
largest single corporation under its regulation, the agency must
be charged with gross incompetence. If it did have the
information, but failed to act and failed to inform the public,
it was sadly remiss--if not legally negligent--in carrying out
its functions.29
As the government's investment in railroads grew other agencies,
some newly created, were given power to distribute money and oversee
rail activities. The Railroad Revitalization and Regulatory Reform
Act of 1976 (45 U.S.C. 801) explicitly shifted some regulatory
jurisdiction from the ICC to the Department of Transportation (DOT)
and Securities Exchange Commission (SEC). The DOT's role in
railroad regulation is discussed next.
The Department of Transportation establishes the Nation's overall
transportation policy. There are nine operating administrations
under its umbrella, including the Federal Railroad Administration
(FRA). The purpose of the FRA is to promulgate and enforce rail
29 The Penn Central Failure and The Role of Financial
Institutions. Staff Report of the Committee on Banking and
Currency, House of Representatives. 92nd Congress, 1st Session,
June 3, 1972.
51
safety regulations, administer railroad financial assistance
programs, conduct research and development, provide for the
rehabilitation of Northeast Corridor rail passenger service, and
consolidate government support of rail transportation activities.
It exercises jurisdiction overall areas of rail safety under the
Rail Safety Act of 1970, such as track maintenance. The FRA also
administers programs to develop, implement, and administer rail
system policies, plans, and programs in support of applicable
provisions of the 4-R Act and the Rail Passenger Service Act (45
U.S.C. 501), and related legislation.
The third agency concerned with railroad regulation is the
Securities Exchange Commission. Until the 4-R Act, the ICC
exercised sole jurisdiction over railroad financial disclosures.
The 4-R Act removed this power from the ICC and placed it with the
SEC.
52
APPENDIX C: CHRONOLOGY OF EVENTS
April 30, 1976 - DOT petitioned ICC to institute a rulemaking
examining the merits of RRB vs ratable depreciation. The petition
was not announced or published.
April 12, 1977 - ICC denies DOT petition.
April 28, 1977 - SEC invited comment on "Railroad Industry
Disclosure Guidelines, Deferred Maintenance and Betterment
Accounting."
February 15-16, 1978 Docket No. 36557. Informal hearings by ICC's
Bureau of Accounts
October 31, 1978 - ICC served Notice of Study on "Alternative
Methods of Accounting for Railroad Track Structure"
Legislative history of miscellaneous tax changes to codify RRB
accounting for tax purposes by adding I.R.C. Section 167(r).
July 9, 1979 - Bill introduced.
July 27, 1979 - House hearings.
October 22, 1979 - Senate hearings.
December 13, 1979 - Senate Committee recommends passage but
bill is not placed on Senate agenda.
May 22, 1980 - House recommends passage.
November 24, 1980 - Senate recommends passage.
December 13, 1980 - P.L. 96-613 passed.
February 4, 1981 - GAO publishes "Accounting Changes needed in the
Railroad Industry" recommending switch to ratable depreciation
and requesting SEC action. SEC denies request.
May 18, 1981 - Tax hearings confirm that RRB property would not
qualify as recovery property
June 12, 1981 - H.R. 3849 introduced with RRB transition rules and
frozen base write-off.
June 17, 1981 - Hearings discuss details of the original
administration proposal and the newly endorsed H.R. 3849 that
53
repeals IRC Sec. 167(r) and makes railroad track structure subject
to ratable depreciation as 5 year property with a phase in
provided by transitional rules.
June 22, 1981 - ICC Notice of Proposed Rulemaking. Notice of
intent to change depreciation methods.
July 18, 1981 - Ways and Means Committee agree to railroad
provisions.
July 29, 1981 - ERTA passes House.
August 3, 1981 - ERTA passes Senate.
August 10, 1981 - ERTA signed by President Reagan.
October 27, 1981 - Wall Street Journal runs a front page article
on the transition rule tax benefits.
March 17, 1982 - ICC Supplemental NOPR to address issues raised
by comments on NOPR.
February 17, 1983 - ICC Rulemaking.
54
APPENDIX D: SAMPLE COMPANIES
Burlington Northern, Inc.
Burlington Northern Railroad
Colorado & Southern Railway
Fort Worth & Denver Railway
St. Louis - San Francisco Railway
Canadian Pacific Limited
Chicago Milwaukee Corporation
Chicago, Milwaukee, St. Paul and Pacific Railroad
Chessie System, Inc.
Baltimore & Ohio Railroad
Chesapeake & Ohio Railway
Western Maryland Railway
Seaboard Coast Line Industries, Inc
Seaboard Coast Line Railroad
Clinchfield Railroad
Louisville & Nashville Railroad
Flordia East Coast Industries
Flordia East Coast Railway
Chicago Milwaukee Corporation
Chicago, Milwaukee, St. Paul & Pacific
IC Industries
Illinois Central Gulf Railroad
Kansas City Southern Industries
Kansas City Southern Railway
Louisiana & Arkansas Railway
Katy Industries
Missouri-Kansas-Texas Railroad
Missouri Pacific Corporation
Missouri Pacific Railroad
Norfolk & Western Railway
Delaware 6c Hudson Railway
Southern Railway System
Southern Railway
Alabama Great Southern Railroad
55
Central of Georgia Railway
Cincinnati, New Orleans & Texas Pacific Railroad
Rio Grande Industries
Denver & Rio Grande Western Railroad
Sante Fe Industries
Atchinson, Topeka & Sante Fe Railway
Southern Pacific
Southern Pacific Transportation Co
St. Louis Southwestern Railway
Northwestern Pacific Railroad
Union Pacific Corp
Union Pacific Railroad
Western Pacific Industries, Inc.
Western Pacific Railroad
Soo Line Railroad
USX, Inc.
Bessemer & Lake Erie
Elgin, Joliet & Eastern
Duluth, Missabe & Iron Range
Mergers During Test Period
Burlington Northern acquires St. Louis - San Francisco Railway
CSX is formed by merger of Chessie System and Seaboard Coast Lines
Norfolk Southern formed by merger of Norfolk & Western and
Southern Railway
Union Pacific Corp merges with Western Pacific and Missouri Pacific
56
AFPEEEDIX E: METHODOLOGICAL ISSUES
1. Introduction
This appendix addresses some methodological concerns frequently-
raised in a clustered event study and explains why ordinary least
squares (OLS) is chosen as the estimation method for this study
instead of the seemingly unrelated regression (SUR) technique.
When all sample firms are from the same industry and are affected
on the same date by the event of interest, randomization across
industries and over time is not possible. Under these
circumstances, returns may be cross-sectionally correlated and
heteroscedastic (Collins and Dent, 1984). One way to handle this
problem is with seemingly unrelated regressions (SUR). For the
covariance matrix in a SUR to be invertible, the number of time
periods must be greater than the number of firms. Researchers can
often satisfy this constraint by combining estimation and test
periods (Schipper and Thompson, 1983).
The SUR methodology has three significant limitations. These
limitations concern nonsynchronous trading, structural stability,
and sampling error introduced by SUR. First, simultaneous
estimation of market and event parameters does not correctly adjust
for nonsynchronous trading. Lead and lag market returns can be
included in the regression, but the common method of dealing with
nonsynchronous trading calls for computation of compound coeffi
cients which adjust for the autocorrelation in the market return
(Scholes and Williams, 1977). Second, structural stability is
57
required over the estimation period. In this study the estimation
period spans 6 years. During this time the structure of the
industry and competing industries change drastically because of
deregulation. Table 5 reports F statistics that test the hypothesis
that the intercept, lag, contemporaneous, and lead parameter
estimates from the April 28, 1977 estimation period are the same as
the estimates from the August 8, 1981 estimation period for the
firms that stay in the sample. The hypothesis is rejected at the
one percent level in 11 out of 15 cases. Lastly, the use of SUR
introduces an additional source of sampling error and SUR's finite
sample properties are unknown. This point is elaborated below.
The following section summarizes the reason for concern over
cross-sectional correlation and discusses a methodology commonly
employed when this correlation exists. Section 3 summarizes some
recent work on the bias in OLS variance estimates when OLS is used
despite the existence of cross-sectional correlation.
2. The Effect and Treatment of Cross-Sectional Correlation
Cross-sectional correlation is likely to occur if an industry
specific event, unrelated to the event of interest, occurs at the
same time. Statistical tests that assume independently identically
distributed (iid) errors are not valid. Similarly, heteroscedastic
errors have the same consequences. Generally, in the case of
nonspherical disturbances, OLS produces unbiased but inefficient
estimates and reported standard errors are biased. The concern over
58
Table 5
Structural stability test statistics
F test of Hq: p2 = P20
where /3t =
aT
P r
P T
Burlington Northern, Inc 3.18
Canadian Pacific Limited .23
Flordia East Coast Industries 13.27*
IC Industries 12.11*
Kansas City Southern Industries .31
Katy Industries 4.70
Missouri Pacific Corporation 5.22*
Norfolk 6c Western Railway .70
Southern Railway System 5.20*
Rio Grande Industries 3.72*
Sante Fe Industries 4.23
Southern Pacific 21.83*
Union Pacific Corporation 4.60*
Soo Line Railroad 6.57
USX, Inc. 8.38*
* = significant at .01
errors in inference when events are either clustered in calendar
time or over certain industries has been addressed by several
researchers with sometimes conflicting opinions. Christie (1987)
finds no evidence that cross-sectional dependence in the data causes
serious bias in standard errors for three cross-sectional returns
studies he reviews. Alternatively, Collins and Dent (1984) and
Sefcik and Thompson (1986) describe hypothetical situations where,
when cross-sectional dependence is ignored, true standard errors of
estimates exceed reported standard errors by several orders of
magnitude. Bernard (1987) notes this discrepancy and provides a
rigorous analysis of the problem that attempts to reconcile the
performance observed by Christie with hypothetical situations
described by other authors. More importantly, Bernard describes
when OLS standard errors are likely to be biased. Because Bernard's
paper applies directly to this study, some of his most relevant
points are elaborated below.
First, it is helpful to distinguish between "event studies" and
"cross-sectional returns studies", and to illustrate the methodolog
ical problem. An event study is one where the researcher is testing
for a stock price reaction conditional only upon the occurrence of
an event. With reasonable assumptions, Bernard shows that the
standard error bias under OLS is at its maximum. Although the
improvement over OLS is data-dependent, alternative estimation
procedures are generally available for event studies. A cross-
sectional study is one where a stock return metric is regressed
60
against firm specific variables. Intuitively, we expect the bias in
OLS standard errors to be less for cross-sectional studies than for
event studies. To elaborate, recall that clustering may cause
correlation in the unexpected returns generated by the market model
in step 1. An event study simply tests for a change in mean
conditional upon an event by regressing firm unexpected returns on a
dummy variable with the coefficient constrained to be the same
across firms.
U-£ = 7 D + (1)
where
Uj[ = unexpected return on event date for
firm i, generally estimated with the market
model in "stage 1"
7 = estimated event parameter
D = dummy variable indicating event
e^ = estimated residual for firm i
J
In practice, the estimation and test periods are often combined. To
the extent cross-sectional dependence is not explained by the dummy
variable, it persists in equation 1 residuals.
It is reasonable to assume that firms would systematically differ
in relative share price sensitivity to specific events. A cross-
sectional study tries to capture systematic sensitivity by
regressing unexpected returns on firm specific variables. Although
the error terms will be smaller after accounting for cross-sectional
I
variations, they still may not be identically, independently,
distributed (iid).
To illustrate potential cross-sectional dependence in (1), it is
helpful to look at its residual variance/covariance matrix, z.
!
61
2
C °X Q2° 1
ala 1 a2
2 =
(2)
r 2
n
If the diagonal terms are not the same, heteroscedasticity exists.
In cross-sectional studies, the most likely source of
heteroscedasticity is removed by scaling the independent variables.
If the off-diagonal terms are nonzero, then contemporaneous
correlation exists. This occurs when sample residuals move
together, such as when an unrelated industry specific event occurs.
When s is not diagonal, estimation techniques other than OLS are
generally employed. The most widely used estimation technique in
the presence of nonspherical disturbances is estimated generalized
least squares (EGLS). Prior information is used to estimate the
disturbance variance matrix. The estimated disturbance parameters
are then used to transform the original variables. OLS on the
transformed model produces estimates that have all the properties of
OLS estimators and can be subjected to the usual inference
procedures. EGLS is consistent, but its small sample properties are
generally unknown and caution is suggested on making inferences
(Johnston, 1984).
To elaborate, recall that OLS on (1) when 2 is nondiagonal is
unbiased but inefficient and produces biased standard errors. EGLS
on (1) is efficient and consistent, but introduces additional
sampling error. Suppose that 2 is in fact diagonal, but sampling
error causes estimated s to be nondiagonal. In this case, OLS is
62
preferred over EGLS. Even if z is not diagonal but its off-diagonal
elements are sufficiently small, OLS may be preferred to EGLS. In
general, the nonsphericalness of the error terms must be quite
severe to make EGLS superior to OLS. When EGLS is applied to a set
of equations (SUR) the gain in efficiency over OLS increases with
the correlation between disturbances from different equations and
decreases as the correlation between the different sets of
explanatory variables increase (Theil, 1971).
Researchers have started using SUR to deal with the possibility of
contemporaneous correlation. Schipper and Thompson (1983) apply
this methodology to monthly returns, where contemporaneous
correlation is found to be most severe by Bernard (1987). Tax event
studies have extended this methodology to daily returns (Madeo and
Pincus, 1985; and Karlinsky and Manegold, 1988). However, Bernard
and Christie agree that contemporaneous correlation is generally not
significant for short return intervals. On a daily return basis,
the additional sampling error introduced by SUR may more than offset
possible efficiency gains of allowing nonspherical errors. The
potential dangers of using SUR in a daily return setting are
illustrated below. The following model combines estimation and test
periods in a SUR framework.
63
rit = “i + P i“ t-l + + P+irmt+l + i D * Xi + eit (3)
where
rit =
return for firm i on day t,
“i =
estimated intercept for firm i,
P+i =
estimated lag, contemporaneous, and lead beta
for firm i,
7 =
estimated event parameter,
D dummy variable D=1 on event date, 0 otherwise
Xi =
firm specific independent variable,
ei =
estimated residual for firm i,
In the case where there is only one event available to estimate the
event parameter gamma, it is constrained to be the same across
firms. For sample size 2, the right hand side of the model is
represented by the following matricies:
1
rm0
rml
rm2
1 rmi rm2
rm3
1 rm2 rm3 rm4
1 rmt.! rmt rmt+1
0
0
0
0
0
*1
0
0
0
0
0
1 tmt_^ rmt rmt+2 . X2
1 rm0 rm]_ rm2
1 rm2 rm3
1 rm2 rm3 rm 4
r
' eir
ai e12
P l
Pi
P+i
+
eit
a 2
P~ 2
P 2
P + 2
_
e2t
(4)
Under the usual assumptions of zero noncontemporaneous correlation
and within firm homoscedasticity, the covariance matrix is
z =
o o
o o
a ! 2 0
0 a ^ '
0 0
0 pa 1 ® 0
0 0 pa ±a 2 0
0
0
0
pa ±a 2
a 1
0
0
0
0 p a ^ a 2
2
2
0 0 0
0 a 22 0 0
0 0
* 2 *
0
0 0 0
a 2
(5)
I
64
Typical SUR routines run OLS on each equation separately. The OLS
mean squared error is the estimated an2 for SUR. OLS residuals are t
also used to estimate paLa^ as 1/n z e^tejt. If we assume that
p=0, then SUR collapses to weighted least squares (WLS) on each
equation. With this in mind, I estimated each cross-sectional
equation using WLS, using the estimation period OLS mean squared
errors to develop appropriate weights. With the exception of the
event period residuals not entering into the calculation of the mean
square error, this estimation procedure is essentially the same as
SUR where p=0. It is interesting to note that WLS produced grossly
misspecified models when tested with White's specification test
(White, 1980). The weights introduced heteroscedasticty into a
model that was otherwise well specified. Since WLS resulted in
misspecified models, OLS was chosen as the estimation procedure for
this study. Although WLS applied to a two-stage model may be
misspecified, this does not mean that SUR applied to a single stage
model would be similarly misspecified. The issue of the propriety
of SUR in a daily returns arena is an empirical one.
Results presented by econometricians raise more questions. It is
a well known result that SUR produces no benefit over least squares
(LS) when the independent variables across equations are the same
(perfectly correlated). Efficiency gains increase as the
contemporaneous correlation across equation increases and the
correlation of explanatory variables across equation decrease. The
65
most favorable situation is where independent variables are
orthogonal across equations (Theil, 1971). In a situation where a
200 day estimation period and a 5 day event period are included for
each firm, each equation has 200 out of 205 observations with
identical independent variables. The explanatory variables across
equations would seem to be almost perfectly correlated. Again, this
is an empirical question that is data dependent.
The extent of contemporaneous correlation is also an empirical
issue on which accounting researchers have expended considerable
energy. Most recently, Bernard (1987) finds mean intraindustry
cross-sectional correlation in market model residuals to be .04 for
daily returns and .09 for weekly returns. Under the most favorable
conditions of orthogonal independent variables in a two equation
setting, Zellner (1963) compared the use of SUR to LS. In such a
setting, LS is superior to SUR with contemporaneous correlation as
high as .20 and reasonable sample sizes. Only when contemporaneous
correlation reaches .40 does SUR become unambiguously superior to
LS. Again, this result is data dependent.
This section provides some thoughts regarding the use of SUR
estimation in a daily returns setting. Because of the limitations
imposed by using SUR and its questionable efficiency gains, OLS was
chosen as the estimation procedure for this study. However, if
contemporaneous correlation does exist, LS variance estimates will
be biased. The following section discusses this bias.
3. Bias in OLS-based variance estimates.
66
To better understand the bias in OLS-based variance estimates,
Bernard develops a model of the bias attributable to Greenwald
(1983) and reproduced below.
N N N N
Bias = 1 + 1/N 2 2 PijWij + 1/N 2 2 (ai o. - 1)0^ , wA *
i - j i^j i = J
N
+ 1/N 2 (C7i2 - 1) Ui/U (6)
i - j
The first term in equation 6 represents the bias due to cross
correlation which depends on the residual cross-correlations p i j ,
and certain functions of the regressors Wjj . Bernard refers to Wjj
as regressor cross-correlation. The last term in (6) captures bias
due solely to residual heteroscedasticity. The term in the middle
captures bias due to an interaction between cross-correlation and
heteroscedasticity. After further decomposition Bernard states:
We can immediately make the following observations about the
bias due solely to cross-correlation. First, in order for bias
to exist, there imst be cross-sectional dependencies in not only
the residuals but also in the regressor. That is, both p and
Wjj must be nonzero for at least some pairs of firms. Second,
the bias depends not only on the magnitude of the residual
cross-correlations pij and the regressor cross-correlations w^j
but also their covariability; higher covariability translates
into greater bias. ... The implication is that the bias will be
most serious when unidentified factors (e.g., industry effects)
that lead to a large correlation between residuals for some
firm-pairs also lead to a large correlation between values of
the regressor for the same firm-pairs. Third, it is only cross
correlation in the orthogonal component of the regressor that
ultimately affects the bias. ... An implication is that if there
is cross-sectional correlation among values of each of two or
more regressors only because of common industry effects that do
not persist in the orthogonal components of the regressors, then
no bias (from this source) would exist in the OLS-based variance
estimates. Fourth, equations (6) and (7) suggest a relation
between sample size and bias. In the special case where
Cov[p£j , Wj_j ] , p, and w are held constant and positive as we
67
increase the sample size, the relation is simple; the bias
increases linearly in N.
Two empirical results that are not evident in equation 6 is that
the degree of cross-correlation increases dramatically as the return
interval lengthens and the degree of interindustry correlation is
small relative to intraindustry correlation.
Based on Bernard's analysis, my a priori expectation is that the
degree of bias in this study will not be serious. First, a short
return interval is used. Bernard and Christie agree that cross-
sectional correlation is generally not a significant problem for
short return intervals. Secondly, the sample consists of a small
number of firms. Bernard shows that bias due to cross-sectional
correlations is a monotonic function of sample size. Third,
heteroscedasticity is not expected to be a problem because of proper
deflation of all variables. In fact, the deflated model is well
specified under White's specification test (White, 1980)). Finally,
there is no reason to expect that regressor cross-correlations will
vary with residual cross-correlations. This covariance drives
residual cross-correlations, the first term in equation 6.
To expand on the last point, recall that all explanatory variables
included in this study are carefully specified measures of firm tax
and regulatory costs. Previous cross-sectional studies have relied
upon broadly interpreted proxy variables such as leverage and size.
Jain (1986) provides evidence that size and leverage also proxy for
the covariance of a security's residual return with a portfolio of
all firms in the analysis. Thus, if leverage and size variables are
68
included in a model, researchers might expect high covariance of
regressor correlations with residual correlations. There is no
reason to expect such covariance to exist in this study. However,
this issue is again an empirical one, the testing of which is beyond
the scope of this paper.
69
BIBLIOGRAPHY
Bernard, V.L., "Cross-Sectional Dependence and Problems in Inference
in Market-Based Accounting Research," Journal of Accounting
Research (Spring 1987), pp. 1-48.
Boberg, K.B., "Changes in Railroad Track Structure Depreciation
Recognition in the United States: Their Potential Consequences,"
Ph. D. Dissertation (Pennsylvania State University, 1983).
Boer, G., "Replacement Cost: A Historical Look," The Accounting
Review (January 1966), pp. 92-97.
Boockholdt, J.L., "Influence of Nineteenth and Early Twentieth
Century Railroad Accounting of the Development of Modern
Accounting Theory," The Accounting Historians Journal (Spring
1978), pp. 9-28.
Christie, A.A., "Aggregation of Test Statistics," Journal of
Accounting and Economics (Forthcoming 1989).
Christie, A.A. "On Cross-Sectional Analysis in Accounting
Research," Journal of Accounting and Economics (December 1987),
pp. 231-258.
Collins, D. and W. Dent, "A Comparison of Alternative Testing
Methodologies Used in Capital Market Research," Journal of
Accounting Research (Spring 1984), pp. 48-84.
Fahrenwald, W., "What the New Tax Law Means to the Railroads,"
Railway Age (November 1981), pp. 36-42.
General Accounting Office, Accounting Changes Needed in the Railroad
Industry (Washington, D.C.: Government Printing Office, 1981).
Greenwald, B.C., "A General Analysis of Bias in the Estimated
Standard Errors of Least Squares Coefficients," Journal of
Econometrics (August 1983), pp. 323-38.
Healy, P., "The Impact of Bonus Schemes on the Selection of
Accounting Principles," Journal of Accounting and Economics
(April 1985), pp. 85-107.
Holthausen, R.W. and R.W. Leftwich, "The Economic Consequences of
Accounting Choice: Implications of Costly Contracting and
Monitoring," Journal of Accounting and Economics (August 1983),
pp. 77-117.
70
— 1
"ICC Staff Calculates Depreciated Value of Rail Track Structure,"
Traffic World (July 11, 1983) pp. 4-5.
Interstate Commerce Commission, Depreciation Charges of Steam
Railroad Companies 177 I.C.C. 351 (1931).
Interstate Commerce Commission, "Notice of Study," in Alternative
Methods of Accounting for Railroad Track Structure 43 F.R. 50717
(1978).
__________, "Notice of Proposed Rulemaking," in Alternative Methods
of Accounting for Railroad Track Structure 46 F.R. 32289 (1981).
__________, "Supplemental Notice of Proposed Rulemaking," in
Alternative Methods of Accounting for Railroad Track Structure 47
F.R. 11539 (1982).
__________, Alternative Methods of Accounting for Railroad Track
Structure Docket No. 36988, 367 I.C.C. 157 (1983).
Jain, P.C., "Analysis of the Distribution of Security Market Model
Prediction Errors for Daily Returns Data," Journal of Accounting
Research (Spring 1986), pp. 76-96.
Johnston, J., Econometric Methods (New York, NY: McGraw-Hill Book
Company, 1984).
Leftwich, R., "Evidence of the Impact of Mandatory Changes in
Accounting Principles on Corporate Loan Agreements," Journal of
Accounting and Economics (March 1981), pp. 3-36.
Littleton, A.C., Accounting Evolution to 1900 (New York: American
Institute Publishing Company, 1953).
Madeo, S.A., and M. Pincus, "Stock Market Behavior and Tax Rule
Changes: The Case of the Disallowance of Certain Interest
Deductions Claimed by Banks," The Accounting Review (July 1985),
pp.407-429.
Securities and Exchange Commission, "Railroad Industry Disclosure
Guidelines, Deferred Maintenance and Betterment Accounting," File
No. 57-692, (Washington, D.C.: 1978).
Schipper, K., and R. Thompson, "The Impact of Merger-Related
Regulations on the Shareholders of Acquiring Firms," Journal of
Accounting Research (Spring 1983), pp. 184-221.
Schipper, K., R. Thompson, and R. L. Weil, "Disentangling
Interrelated Effects of Regulatory Changes on Shareholder Wealth:
The Case of Motor Carrier Deregulation," Journal of Law and
Economics (April 1987), pp. 67-100.
71
Scholes, M., and J. Williams, "Estimating Betas From Nonsynchronous
Data," Journal of Financial Economics (December 1977), pp. 309-
327.
Sefcik, S., and R. Thompson, "An Approach to Statistical Inference
in Cross-Sectional Regression with Security Abnormal Returns as
Dependent Variable," Journal of Accounting Research (Autumn 1986),
pp. 316-34.
Theil, H., Principles of Economics (New York, NY: John Wiley and
Sons, Inc., 1971).
U.S. Congress, House, Committee on Ways and Means, Tax Aspects of
the President's Economic Program. Hearings Before the House
Committee on Wavs and Means 97th Congress, 1st Session, 1981.
_________ , The Bankruptcy Act and Minor Tax Bills: Hearings Before
the Subcommittee of Select Revenue Measures of the House Committee
on Wavs and Means 96th Congress, 1st Session, 1979.
U.S. Congress, House, Subcommittee on Transportation and Commerce,
Congressional Symposium Railroads - 1977 and Bevond: Problems and
Promises 95th Congress, 1st Session, 1977.
U.S. Congress, Senate, Committee on Finance, Miscellaneous Tax
Bills III. Hearings Before the Subcommittee on Taxation and Debt
Management Generally of the Senate Committee on Finance 96th
Congress, 1st Session, 1979.
Watts, R.L. and J.L. Zimmerman, "The Demand and Supply of Accounting
Theories: The Market for Excuses," Accounting Review (April 1979),
pp.273-305.
Watts, R.L. and J.L. Zimmerman, Positive Accounting Theory
(Englewood Cliffs, NJ: Prentice Hall, Inc., 1986).
White, H, "A Heteroskedasticity-Consistent Covariance Matrix
Estimator and A Direct Test for Heteroskedasticity," Econometrica
(May 1980), pp.817-28.
Zimmerman, J.L., "Taxes and Firm Size," Journal of Accounting and
Economics (August 1983), pp. 119-149.
Zmijewski, M. and R. Hagerman, "An Income Strategy Approach to the
Positive Theory of Accounting Standard Setting/Choice," Journal of
Accounting and Economics (August 1981), pp. 129-149.
72

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A conceptual framework of distribution intensity

PDF

The effects of regulation on coporate tax shifting: An analysis of the property and casualty insurance industry

PDF

A contingent claims analysis of the equity interest in property-casualty insurance companies

PDF

An empirical analysis of off-balance-sheet financial instrument disclosures in the banking industry.

PDF

Adoption of regulatory accounting principles by stockheld savings and loan associations: An event study of discretionary accounting choice

PDF

Cointegration and the demand for crude petroleum

PDF

Commercial loan default risk assessment: Predictive accuracy of alternative accounting structures in an inflationary environment

PDF

Communications medium and product class: Their effect on negotiation outcome

PDF

A quantitative model for the analysis of warranty policies

PDF

A biogeographic theory of industrial market structure and competitive dynamics

PDF

Accounting policy choice in firms subject to Superfund environmental costs

PDF

Career advancement in managerial hierarchies in the United States firms: A multi-theoretic model and empirical tests of the determinants of managerial promotion

PDF

An examination of the effect of environmental cues on audit judgments

PDF

An investigation of information requirements determination and analogical problem solving

PDF

A study of the English Canadian feature film industry 1977-1981

PDF

A study on the content and determinants of close exchange relationships in channels of distribution

PDF

Census, map, museum: Wordsworth and the writing of the nation.

PDF

Charles Ives - master songwriter: The methods behind his madness

PDF

An appraisal of alternative spline methods for estimating the term structure of interest rates.

PDF

An empirical investigation of variables related to management consulting engagement effectiveness as perceived by small business clients and their 'Big Eight' CPA consultants

Asset Metadata

Creator Lassar, Sharon S (author)

Core Title Cross-sectional tests of tax and regulatory costs of a change in depreciation methods in the railroad industry

Degree Doctor of Philosophy

Degree Program Business Administration

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Economics, Commerce-Business,economics, general,OAI-PMH Harvest,transportation

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Moore, Michael (committee chair), [illegible] (committee member), Manegold, James (committee member), Weiss, Andrew (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-756765

Unique identifier UC11348886

Identifier DP22659.pdf (filename),usctheses-c17-756765 (legacy record id)

Legacy Identifier DP22659.pdf

Dmrecord 756765

Document Type Dissertation

Rights Lassar, Sharon S.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA