This is to certify that the

dissertation entitled

Income Management After
Initial Public Offerings

presented by \

Lawrence H. Bajor

has been accepted towards fulfillment
of the requirements for

Doctorate degreein Accountancy

 

 

Major professor

 

SM 1 on

MS U i: an Affirmative Action/Equal Opportunity Institution 0-12771

 

INCOME MANAGEMENT AFTER
INITLAL PUBLIC OFFERINGS

By

Lawrence H. Bajor

A Dissertation

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

DOCTOR OF PHILOSOPHY

Department of Accounting and Information Systems
2002

ABSTRACT

INCOME MANAGEMENT AFTER
INITIAL PUBLIC OFFERINGS

By
Lawrence H. Bajor

Abstract: This paper examines earnings management during the
underwriter- imposed lock-up period immediately afier an I PC. It predicts and
finds the use of income increasing accruals during the lock-up period. A sample
of 190 firms having IPOs in 1995 and a sample of matching control firms are
selected. Using the Modified-J ones, the Hribar/Collins and the Healy definitions
of abnormal accruals, earnings management during the lock-up period is
demonstrated. The paper also examines the effect of another suspected earnings
management vehicle during this period, the valuation allowance under FA S
109.

This study contributes to the literature in that it extends Miller and Skinner
to the post IPO lock-up period. It indicates an accounting driven earnings
management explanation of the post-1P0 behavior observed by Tech et al..
(1998) and Ritter (1991). In extending Miller and Skinner (1998) the study
provides a strong test of the deferred tax asset valuation allowance as an
earnings management tool in an environment in which earnings management is
demonstrated. I finds that the valuation allowance is not a component of
earnings management and that the valuation allowance bears an inverse relation

to income from operations.

TABLE OF CONTENTS

LIST OF TABLES .................................................................................................. iv
INTRODUCTION ................................................................................................... 1

LITERATURE REVIEW ........................................................................................ 3
Initial Public Offerings ............................................................................................ 5

FAS 109 ................................................................................................................... 7
HYPOTHESIS DEVELOPMENT ......................................................................... 12
SAMPLE SELECTION AND RESEARCH METHOD ....................................... 17
MODEL AND TEST OF DISCRETIONARY ACCRUALS ............................... 23'
MODEL AND TEST OF THE VALUATION ALLOWANCE ........................... 27
CONCLUSION ...................................................................................................... 36
REFERENCES ...................................................................................................... 38
TABLES ................................................................................................................ 41

iii

TABLE 1

TABLE 2

TABLE 3

TABLE 4a

TABLE 4b

TABLE 4c

TABLE 5

TABLE 6

TABLE 7

TABLE 8

TABLE 9

TABLE 10

LIST OF TABLES

General Descriptive Statistics ................................................ 41 — 42
Descriptive Statistics of the Accrual Positions .............................. 43
Descriptive Statistics of the Deferred Tax Positions .............. 44 - 45

Association Between Accruals and Lock-Up Periods, I-Iribar ....... 46
Association Between Accruals and Lock-Up Periods, Jones ........ 47
Association Between Accruals and Lock-Up Periods, Healy ........ 48

Association Between Valuation Allowance and Lock-Up

Periods ............................................................................................ 49
Descriptive Statistics Other Independent Variables ...................... 50
Correlations Between Independent Variables ................................ 51
OLS Levels Regressions ................................................................ 52
Panel Data Time Series Cross-Sectional Regressions ................... 53
Panel Data Time Series Cross-Sectional Regressions ................... 54

iv

Introduction

This study examines eamings management in the period immediately after an
initial public offering (IPO). Because managers are heavily invested in their own firms, I
predict that managers will take steps to increase earnings through the period of time
during which the lock-up expires. The study predicts and finds that compared to a
matched control sample, income-increasing accruals are employed during the lock-up
period to produce higher earnings. Once the lock-up periods have passed this study
predicts and finds no difference between the experimental and matched conuol group
with regard to the nature of the accruals observed. This study also evaluates the impact
of the valuation allowance has on earnings management in the post-[PO period.

This study employs an experimental sample of 190 firms and a matched control
group. The experimental sample consists of firms having an IPO in 1995. Test results
indicate that, relative to the control group, the experimental sample has a significantly
larger proportion of firms exhibiting income-increasing accruals during the lock-up
period. The mean of the income increasing accruals is significantly more positive for the
experimental group than the control group. Similar tests comparing the experimental and
control groups in the years after the lock-up expiration do not reveal differences in either
the proportion of firms demonstrating positive accruals or in the mean of those accruals.
A test comparing the experimental group with itself, before and after the lock-up period,
demonstrates a significantly larger proportion of firms showing positive accruals with a
significantly larger mean during the lock-up period.

In terms of income increasing behavior, a test of the valuation allowance showed

no difference in the ratio of the valuation allowance to deferred tax assets in the lock-up

and post lock-up periods for the experimental group. It is doubtful that the valuation
allowance is employed to generally increase income across a broad spectrum of firms.

Using panel data for the experimental group a test of income smoothing
employing the valuation allowance is conducted. This test reveals that the change in the
valuation allowance was inversely proportional to the change in income fi‘om operations.
This indicates that as income from operations becomes smaller, a greater proportion of
the firms deferred tax assets are thought not to be recoverable. This finding is contrary to
what one would expect if the valuation allowance were used to smooth income.

This study makes several contributions. Ritter (1991) observes that IPOs,
generally, are underpriced. Measured fi'om the offering price to the market price at the
end of the first day of trading IPOs produce an average initial return estimated at 16.4
percent. Ritter’s most important contribution is his demonstration that IPOs in the long
run appear to be overpriced. He finds a fall ofi‘ of returns in the third year. He observes
that these firms significantly underperform a set of comparable firms matched by size and
industry. Tech (1998) explains Ritter’s findings in terms of accounting information
specifically the manipulation of accruals. This study explains the observations of Ritter
(1991) and Tech et a]. (1998) in terms of income increasing accounting accruals that
correspond to the expiration of the lock-up period on management share holdings. The
accrual patterns found by this study help to explain Ritter’s and Teoh’s post—[PO
observations.

The accounting community has expressed concern with the income
management potential provided by FAS 109. Miller and Skinner (1998) list three

instances of probable earnings management using FAS 109 involving Fortune 500 firms.

This study reproduces Miller and Skinner (1998), in part, in a setting in which income
management using other means is demonstrated. Consistent with Miller and Skinner, the
study finds that the valuation allowance is a function of past earnings, the size of net
deferred tax assets, and total carryforwards. However, with regard to income smoothing,
I find an inverse relation between first differences in income before extraordinary items
and first differences in the valuation allowance, a relation not observed in Miller and
Skinner and one which is contrary to that expected if there were income smoothing using
FAS 109. In an environment marked by income increasing accruals, manipulating the
valuation allowance does not appear to contribute to the earnings management effort.
This study proceeds as follows: in the next section I provide a literature review.
The third section provides hypothesis development. Section four contains sample
selection. Section five describes the research methods and a test of the model related to
accruals. Section six describes the research methods and a test of the model related to the

valuation allowance. Section seven presents the conclusion.

Literature Review

Earnings Management

Earnings management and its varied forms have been a staple of accounting
research for many years. Schipper (1989) defines earnings management as purposeful
intervention in the external financial reporting process, with the intent of obtaining some
private gain. Healy (1985) examined the effect of bonus schemes on accounting
decisions. He found that managers manipulate earnings to maximize bonus plan

compensation conditional on whether earnings before accruals were above, between, or

below that required to receive a bonus. Jones (1991) examined the effect of foreign trade
regulations on accounting choices. She introduced a sophisticated accrual expectation
model to detect earnings management as a response to accounting-based regulations
formulated by the US. Foreign Trade Commission. Dechow, Sloan, and Sweeney (1995)
evaluated alternative discretionary-accrual-based models for detecting earnings
management. They evaluated tests created by Healy (1985), Jones (1991), and DeAngelo
(1994) and they found that all of the tests were of low power for detecting earnings
management. They cited a modified version of the Jones (1991) as having the most
power to detect earnings management. Beneish (1997) studied actual instances of
earnings management. From his work he presented a model to detect earnings
management among firms experiencing extreme financial performance. He compared the
model’s performance to that of discretionary accrual models.

Burgstahler and Dichev (1997) provide evidence that firms manage reported
earnings to avoid earnings decreases and losses. In cross-sectional distributions of
earnings changes and earnings, they found unusually low fi'equencies of small decreases
in earnings and small losses. They also offer theories about the motivation for avoidance
of earnings decreases and losses. Burgstahler and Dichev (1997) document eamings
management without estimating abnormal accruals. Their main contribution is that
earnings management does not occur in a vacuum. Management must perceive a benefit
from the actions taken.

Guay, Kothari, and Watts (1996), hereafter GKW, criticize the discretionary-
accrual models used in much of the income-management research, Healy (1985);

DeAngelo (1986); Jones (1991); Jones as modified in Dechow, Sloan, and Sweeny

(1995) and the industry model proposed by Dechow and Sloan (1991). Their study is a
joint test of earnings-price models. Its findings similar to those in Dechow, Sloan, and
Sweeny (1995) in that GKW find that all of the models estimate discretionary accruals
with imprecision. GKW indicate that the Jones and modified Jones models have the most
potential to provide the best chance of estimating discretionary accruals. Their overall
results suggest that caution should be exercised in interpreting the research on
management’s use of accruals motivated by opportunism or performance measure
improvement.

Healy (1996) discusses the GKW (1996) paper and finds it to be flawed. He finds
the strength of their findings to be at odds with the strength of their study. Healy finds
fault with the noise in their market research as it is devised. Healy asks as to whether
should one reject the accrual models as GKW suggest or should one reject the GKW
model and its predictions. Healy questions the use of a market study in this context
because the difference between discretionary and non-discretionary accruals would not be
public information. He doubts that the market has the information available to
distinguish between the discretionary and non-discretionary accruals. Healy concedes
that as with many areas of academic accounting inquiry the method of accrual
measurement is flawed. He does not believe that the imperfection of the tools should
deter additional efforts in the area of income management but rather that this
imperfection provides opportunity for additional research. Healy indicates that the key to
research in this area is to select firms and situations in which accruals management is
likely to occur. These Situations will provide Sufficient power so that even crude proxies

for earnings management are likely to detect an effect.

Collins and Hribar (1999), question the use of a balance sheet approach to test for
earnings management. The Healy (1985) and the Jones and modified Jones tests are
balance Sheet methods. Collins and Hribar find in particular that these tests of earnings
management can be biased in the presence of mergers, acquisitions, and discontinued
operations. They recommend computing accruals as follows:

Accruals, = Earnings, - CFOi
where Eamings is earnings from operations (COMPUSTAT #76) and CFO is the cash
flow from continuing operations (COMPUSTAT #108- #78).
Dechow et al.. (2001) further adapt the Jones model to better explain
accrual behavior. They drop the implicit assumption that all credit sales are
discretionary. They also attempt to capture the predictable component in a firm’s

accruals and its future sales growth.

Initial Public Offerings

Ritter (1991) observes that IPOs, generally, are underpriced. Measured from the
offering price to the market price at the end of the first day of trading, IPOS produce an
average initial return estimated at 16.4 percent. Ritter’s most important contribution with
regard to this study, is his demonstration that IPOs in the long run appear to be
overpriced. He finds a fall off of returns in the third year. IPO firms significantly
underperforrned a set of comparable firms matched by size and industry. In an analysis
of individual accounting items, Tech (1998) found that IPO firms adopt income
increasing depreciation policies before the IPO when they deviate from a matched control

group of similarly performing non-issuing industry peers. IPO firms also provide

significantly less for uncollectible accounts receivables than the control group of non-
issuers. Ritter and Tech both examined years before the advent of SFAS No. 109.

Beaty, Riffe, and Thompson (2000) examine the relation between IPO stock
values and available accounting information. In explaining Ritter, they find that
accounting book value, earnings and revenue, in conjunction with several other firm and
market characteristics, explain a large portion of IPO offer values. They find that first
day stock return fi'om offer price to closing market is also significantly correlated with

firm characteristics and accounting information.

FAS 109

Statement of Financial Accounting Standards No. 109, Accounting for Income
m, was issued in 1992, effective for fiscal years beginning after December 15, 1992.
Deferred taxes represent the future tax consequences of events already recognized in
either the financial statements or tax returns. Under FAS 109 businesses recognize
deferred tax liabilities for temporary differences that may result in taxable amounts in
future years. If in management’s judgement future taxable income will not be sufficient
to realize a tax benefit from a deferred tax asset, a valuation allowance, offsetting the
deferred tax asset by the projected unused portion, must be created. The FASB provides
the following guidance in the determination whether or not a valuation allowance is
needed. All available evidence should be considered as to whether or not a valuation
allowance is needed. The decision is to be made on the basis of the weight of that
evidence, both positive and negative. After considering both positive and negative

evidence management must decide if it is more likely than not that the deferred tax asset

will be realized. The Board defines mere likely than not to mean a level of likelith
that is more than 50 percent. To the extent that it is not more likely than not a deferred
tax asset will be realized a valuation allowance must be set up. There is no
corresponding impairment for deferred tax liabilities. The more likely than not standard
is viewed as sufficiently broad to enable aggressive managers to manipulate the size of
the valuation allowance and subsequently earnings.

FAS 109 has been strongly criticized in that it appears to provide an additional
opportunity for managerial interpretation and income management. Ayers (1998)
examines the statement from an information content perspective. He concludes that the
changes made by FAS 109 provide value relevant firm data beyond that provided by APB
11.

Espahbodi, Espahbodi and Tehranian (1995) observed that SFAS 109 generally is
an income-increasing standard. They studied the market reaction observed around the
Exposure Draft dates and they found significant positive abnormal returns. The authors
also hypothesized that the equity price reaction should correspond to the income effects
and the consequences of a given income effect. Their results are consistent with
contracting and political cost hypotheses.

Behn, Eaton, and Williams (1998) examined the association between the
recognized deferred tax asset valuation allowance and certain variables put forth as
sources of evidence in F AS 109. They found that taxable income in prior years, future
reversals of deferred tax liabilities, and the potential for future income are factors that are

strongly associated with the relative level of deferred tax asset valuation allowance.

Miller and Skinner (1998) also examined the components of the valuation
allowance mandated by the standard as it applies to large stable firms. Consistent with
F AS 109, they found that the valuation allowance is larger for firms with relatively more
deferred tax assets. They also found that the allowance is smaller for firms with higher
levels of expected future taxable income. The size of a firm’s tax credit and tax loss
carryforwards are the most important determinants of the size of the valuation allowance.
Miller and Skinner did not find any relation between the size of the valuation allowance
and income from operations. Miller and Skinner hypothesized the existence of income
management employing the valuation allowance but found no evidence of it. The authors
theorized that the reason that their research found no income management was that the
power of their tests was insufficient to capture the effect. The sample in the Miller and
Skinner paper consisted of 200 large, well established, and publicly traded firms. The
only requirement was that they have net deferred tax assets on their balance sheets.
Examining firms in an environment known to foster income management as suggested in
Healy (1996) could provide a more powerful test.

Brav and Gompers (2000) examined the role of investment bankers and lock-up
provisions in the market for new equity issues. They found an average abnormal return at
the lock-up expiration. They found that earnings forecasts made by both affiliated and
unaffiliated analysts were more optimistic.

This study will first establish the existence of earnings management in a sample
of firms that is representative of the post-IPO environment. It will then analyze what
contributions the valuation allowance under FAS 109 makes toward earnings

management.

Hypothesis Development

The important question being asked is what is the nature of the income
management observed prior to the IPO lock-up expiration. A second question explored is
whether or not the valuation allowance is being used to manage income. Observing
companies in the time frame immediately post-IPO is thought to present an Opportunity
for management either to artificially increase earnings or to smooth them at artificially
high levels. Managers typically have a large portion of their wealth tied. up in their
company’s equity. The median insider ownership of IPO shares in this sample is 37
percent. The mean insider ownership is 39 percent. The median capitalization of the
firms in this sample on their IPO date is $139 million dollars. Because of the lock-up of
up to 3 years mandated by investment bankers, managers cannot sell their shares at the
IPO or immediately after. Beneish (1999) finds that debt covenants and the cost of
external financing do not determine earnings management. He finds that the monitoring
of managers’ trading behavior can be informative about the likelihood of earnings
management. Gombola, Lee, and Liu (1999) note significant net selling by insiders prior
to seasoned equity offering announcements. Anecdotally, web Sites such as
mmipolockupcom and www.unlockdates.com are numerous. Their sole purpose is to provide
information about expiring lock-ups. They maintain that stock prices can fluctuate
greatly when Shares unlock. There is little academic accounting work on the expiration
of lock-ups. The suspicion is that there is an incentive for management to control
earnings during and just after the lock-up period.

The SEC monitors the trading behavior of insiders. Trading by this group

immediately after an earnings surge or before an earnings drop might invite SEC

10

scrutiny. If managers use accounting to artificially increase income they may also use
accounting to smooth income for a time to avoid scrutiny.

The smoothing hypothesis is well known in accounting literature]. It concludes
that managers use accounting discretion to reduce fluctuations in reported income. Healy
(1985) indicates that managers manipulate earnings by increasing and decreasing accruals
to match the requirements of bonus contracts. The findings of Ritter (1991) are
inconclusive as to whether income increasing or income smoothing transpires in the post-
IPO environment. He observes a pattern of high returns at the IPO date and negative
returns three years afier the IPO date. Tech, Welch, and Rao (1998) used the Ritter
database to examine earnings management during initial public offerings. They found
that IPO firms have high positive issue-year earnings and abnormal accruals. The Ritter
and Tech et al.. efforts do not specifically make this statement, but these papers leave
open the possibility that the earnings management observed in the year of the IPO is
generated to increase or substantiate the offer price.

In examining the sample of IPOs gathered for this study one looks to important
dates in the process. First the company must register with the SEC. As part of the
registration document, a prospectus is provided. For many companies the registration
document will also constitute a comprehensive disclosure document about the company.
After the securities are registered, the next important date for purposes of this study is the
issue date. At this time the shares become available to the public. With regard to the
firms examined in this study, the issue date is usually 60 days after the registration date.
After the issue date, the date on which the lock-up of insider shares expires becomes the

next most important event.

11

Closer inspection of the companies in this sample reveals an interesting
occurrence. Afier comparing the fiscal year ends with the registration date and the issue
date, it is apparent that the 1995 year-end financial statements were available to investors
in only 5 percent of the [POS.2 A typical company had the following time line. Total
Renal Care is a firm in the sample. Its fiscal year end is 12/31. Its registration date is
3/14/95. The issue date is 4/19/95, and the lock-up expiration date is 5/13/96.
Consequently, the year-end 1995 financial data were not available to influence investors
in the initial public offering. If income is being managed to influence the IPO offer price,
then evidence of earnings manipulation should be evident in the year before the IPO.
Tech et al.. (1998) found evidence of earnings management in the year of the offering.
Earnings management in the year of the offering is typically too late to influence the offer
price. What then is management’s motivation? The only motivation remaining is the
lock-up expiration date. In this sample 130 of 190 firms (68 percent) have a 1996 lock-
up expiration date. None of the firms has a 1997 lock-up expiration. The lock-up period
for this sample varies fi'om 180 to 360 days.

To test for earnings management the study will use three definitions of accrual,
modified-Jones (1995), Collins Hribar (1999) and Healy (1985). Each will be employed
to test hypothesis one and two separately. First the study will use the modified Jones

model as described by Dechow, et a1. (2001)3.

TACCit = a + [31(ASales“ " AReCit) + BZPPEit+§it (1)

 

' Watts and Zimmerman (1986).

2 Since most firms have 12-31 year ends this fact should generalize across all IPOs.

3 The enhanced version of the modified-Jones model was not used because its requirements caused so many
data points to be lost that the power available was nil.

12

where ASalesn = the change in finni’s sales (Compustat item#12) fiom year H to t;
AReci. = the change in firm i’s accounts receivable fi'om operating activities for
years M to t (Compustat item #302.)
PPEit = firm i’s year gross property, plant and equipment (Compustat item #7);
and
£1, = error term.
AReci, is the change made to the Jones (1991) model so that only credit sales are assumed
to be discretionary. The variables are scaled by ending year total assets. Equation (1) is
estimated separately for both the lock-up and post lock-up periods. The error term
represents abnormal accruals computed using this model and is referred to as MJacc.
The second accruals model that I estimate is the Collins-Hribar model of accruals:
Accruals“ = Earnings" - CFO“ (2)
where Earnings is earnings from operations (COMPUSTAT #76) and CFO is the cash
flow from continuing operations (COMPUSTAT #108- #78). All variables are scaled by
ending year total assets. This is a non-balance sheet model and reportedly is robust to
large corporate transitions such as mergers and acquisitions. One of the purposes of this
study is to examine the use of the valuation allowance in managing income. This
particular method of capturing accruals is chosen because it reflects the income statement
effects of reducing the valuation allowance. There would be book income without a
corresponding increase in cash flow.

The third accrual model employed is Healy’s definition of accrual:

l3

ACC" = -DEPit - XI" + AARit + AINV it - AAP it - (ATPit "l' Dit) CC].(3)4

Where
DEPn = depreciation in year t;
XI“ = extraordinary items in year t;
AARi, = accounts receivable in yeart less accounts receivable in year t-l;
AAPu = accounts payable in year t less accounts payable in year t-l;
ATP“ = income taxes payable in yeart less income taxes payable in year M;
Di, = deferred income tax expense (credit) for year t.
ACC“ = NA,l + DA“, where NA" represents non-discretionary accruals and DA,t
represents discretionary accruals. When ACC5 is positive Healy postulates that it is
indicative of the fact that managers chose income increasing discretionary accruals
whereas a negative Sign on ACC indicated that either managers choose negative accruals
or that they no longer have the ability to manage reversals. Healy defined discretionary
accruals as accruals that arise because management exercises discretion in reporting.
Non-discretionary accruals are those that arise from the business firndamentals of a firm.
Healy dislikes the terminology and would use the term “unexpected” accruals to describe
discretionary accruals and “expected” accruals to label non-discretionary accruals. The
terms discretionary and non-discretionary, however, are part of the lexicon of the
literature and are employed here.

Healy examined bonus contracts and predicted that managers would manipulate

the sign of ACC in response to the income reporting incentives of their bonus contracts.

This study predicts that managers will use income-increasing accruals during the period

 

’ Note: The dummy variables D1 and D2 are omitted from this equation. Their purpose in the Healy paper
was to distinguish between different types of bonus contracts. That issue is moot in this effort.
5 To control for size ACC is scaled by total book assets.

14

of time leading up to the expiration of the lock-up period. These droughts lead to the
following hypotheses:
H1: Sample firms will exhibit a higher proportion of positive
accruals and significantly more positive average accruals than a
matched sample during the lock-up period.
H2: Sample firms will exhibit a higher proportion of positive
accruals and a significantly more positive average accruals during
the lock-up period than they will in the post lock-up period years.

Miller and Skinner (1998) looked for income management in a general way.
There were no compelling circumstances for eamings management in their sample or
timefrarne. The only restriction on the Miller and Skinner sample was that the firms in it
have deferred tax assets. They did not examine them in the context conclusive to income
management of Burgstahler and Dichev (1997) or Tech et al. ( 1998). Looking at the
immediate post-IPO timeframe, because of the potential for income management,
provides a stronger test of the valuation allowance as an income management tool.

The realization of deferred tax assets depends principally on the existence of
taxable income of sufficient amount and appropriate character. The two major sources of
such taxable income are (1) future reversals of existing taxable temporary differences and
(2) future taxable income exclusive of reversing temporary differences. These sources of
taxable income suggest that the appropriate level of the valuation allowance under FAS
109 also depends on three propositions: (1) the allowance will be smaller the larger the
firm’s available taxable (deferred tax liabilities) temporary differences; (2) the allowance
will be smaller for firms with larger expected future taxable income; (3) the allowance

will be larger for firms with large net operating losses and credit carryforwards. The term

propositions is used to indicate that one is testing issues that involve compliance with

15

FAS 109. Both the occurrences of the taxable temporary differences and the taxable
income must occur in periods that coincide with reversals of the deductible differences
(deferred tax assets). The tax law also places restrictions on the benefits associated with
carryforward items. These limits usually are in the form of “character” of the loss
carryforward, time limits on its use, and jurisdictional differences. For example, capital
loss carryforwards can only be used against realized capital gains. Net operating loss
(N 0L) carryforwards under the US. Internal Revenue Code are limited to 20 years after
the year of loss,6 and NOLs generated in other countries are not deductible against
income earned in the United States. In general, there are likely to be important
restrictions on the extent to which carryforward items can be realized. Therefore
proposition (3), the amount of valuation allowance will be greater for firms with large
carryforwards, seems appropriate.

The test of the three propositions employs the levels analysis used in Miller and
Skinner (1998). The information these tests provides is a baseline to understand if the
determinants of the valuation allowance in the period immediately after the IPO date are
similar to those found by Miller and Skinner.

With regard to the valuation allowance this research first analyzes what
financial elements may be determinants of the valuation allowance and follows the form
of the Miller and Skinner study using recent IPOS as the subject population. Finally, the

study looks for evidence of income management using the valuation allowance. The

 

6 These rules are found in Internal Revenue code section 172. The two-year carryback and 20 year

. carryforward provisions are effective for years beginning after August 5, 1997. For years beginning before
that date the provision is a carryback to three years and or a carryforward to fifteen years. The provision
for foreign tax credits is less generous. Under Code section 904(c) a carryback to two years and/or a
carryforward to 5 succeeding years is permitted.

l6

following hypotheses will be tested to determine if managers use the valuation allowance

to manage income.
H3: The valuation allowance is used to increase income during the
lock-up period the ratio of the valuation allowance to deferred tax
assets will be smaller during the lock-up period than it is during the
post-lock-up period.
H4: The valuation allowance is used to smooth income there will

be a positive relation between changes in the valuation allowance
and changes in before tax income from operations.

Sample Selection and Research Method

The sample consists of firms that made initial public offerings in 1995. That year
was selected because it was sufficiently removed fi'om the implementation date of FAS
109 so that the financial community would be beyond any implementation uncertainty
with regard to the standard. Using 1995 as opposed to a more recent year permits ample
post-IPO information to be readily available. Both the Ritter (1991) and the Tech (1998)
papers look at a timeframe that includes three years past the initial public offering. This
study looks at the year of going public and three years past the IPO date and therefore
includes data up through 1998. The sample was constructed by obtaining a file fiom
Thompson Securities Data Corporation (TSDC). This file contained 2,046 entries
representing IPOS occurring in the years 1993 through 1995. A search of the CRSP
database using the cusip numbers provided by TSDC revealed trade dates for 1,873 firms.
Of these firms 581 had 1995 initial public offerings. This list of firms was run against the
Research Insight (COMPUSTAT) database and 553 firms were found. Wishing to
examine only firms of substance, an additional cut was made by retaining only firms

having assets greater than or equal to $20 million by the end of 1996. This resulted in

17

209 firms remaining. From this sample, publicly traded foreign companies (ADRs) were
eliminated, leaving 198 firms. The result was a sample of companies having an IPO in
1995. Again using TSDC as a data source, the sample was examined for firms reporting
lock-up dates. Lock-up dates are not features of all IPOs. Some firms that have them do
not report them. Eliminating the firms not reporting lock-up dates resulted in a final
sample of 192 firms (hereafier, the “experimental group”). Data were collected on each
company for the year of going public and three subsequent years.

A control group was selected using matching firms based on the four-digit SIC
code and total assets. Reasonable size matches could not be found within the four-digit
SIC code for 3 firms. In these instances a three-digit SIC code match was used. An
additional constraint placed on control group firms, was that they had not undergone an
IPO in the last 3 years. As a whole the managers of the control group should not have the
same incentive to use income-increasing accruals during the 1995 —— 1996 lock-up period
as do the managers of the experimental group.

Table 1 provides basic statistics on the sales, total assets, and earnings from
operations for the experimental and control group firms for each year of the study, 1995
through 1998, individually. The 1996 mean total assets (COMPUSTAT item A6) for the
experimental group are $561 million and the median was $87 million. The largest
company has assets of approximately $55 billion, and the smallest reported total assets of
$20 million. For the control group the 1996 mean total assets are $674 million. The
median total assets are $87 million. For the experimental group the mean income from
operations in 1996 is $43 million with a median of $8.9 million. Mean sales for the

control group is $263 million with a standard deviation of $807 million. The median

18

sales are $81 million. For the control group in 1996 the mean income fi'om operations is
$46 million. The mean sales are $337 million with a median of 88 million. The
experimental and control samples appear to be well matched in terms of size. In any case
the variables of interest used to test the hypothesis are all ratios, which should minimize
size concerns. The companies in this sample are substantial but smaller in general than
those employed by Miller and Skinner ( 1998).

Table 2 contains statistics on the major variable used to test for income
management, ACC. ACC is total accruals as defined in Healy (1985) scaled by total
assets. The mean of ACC for the experimental group in 1996 is 0.01862 with a median
of 0.01650 and a standard deviation of 0.1308. For the control group in 1996 the mean is
—0.01394 with a median of—0.01760 and a standard deviation of0. 1224.

I Miller and Skinner in limiting their sample also concern themselves with the
OPEB charge to sales in the adoption year. Because the firms in this study are all lPOs
few had any charges related to Statement of Financial Standards No. 106. Given the time
frame of this study there are no issues concerning the adoption of FAS 106.

Miller and Skinner chose to collect a sample of firms that have relatively large
deferred tax assets because they rationalized that these firms would have the largest
valuation allowances. This study looks at recent lPOs because these new firms might
have greater incentive to use F AS 109 to manage income. Because new firms represent a
different environment than those examined by Miller and Skinner it is possible that the
relation between various components of deferred taxes and the valuation allowance might
demonstrate a different dynamic. Not all of the firms in this sample have deferred tax

assets. Not all of the firms with deferred tax assets retain them for the entire length of the

19

study. Over the life of the study the number of firms with deferred tax assets varies from
a high of 185 to a low of 134. Some of this change in firms reporting deferred tax assets
is due to attrition in the sample. Of the 190 firms in the Study, 66 report no valuation
allowance for the entire length of the study.

In the accounting literature it is customary to use variables deflated by some
element of the firm’s financials related to size such as total assets, sales, or some other
measure appropriate to the task7. Scaling is used to mitigate the effect of size on results.
It is reasonable to assume that a larger firm may have a larger deferred tax asset account
because of its activities. All of the firms in the Miller and Skinner study were selected
because they have deferred tax assets. Miller and Skinner use this number to deflate
many of the variables in their study. They refer to deferred tax assets as a natural deflator
for the study.8 Table 3 provides descriptive statistics on the deferred tax positions of
sample firms using both total assets and deferred tax assets as a deflator. These data were
obtained by examining the tax footnotes as provided in the 10K reports for each firm.
The mean deferred tax assets to total assets for the first year of the study is .0922. The
mean of the difference between deferred tax assets and deferred tax liabilities scaled by
deferred tax assets is .1859, and this again is comparable to the Miller and Skinner study.
The mean of tax carryforwards scaled by deferred tax assets is .2811, which is
considerably higher than that in the earlier work. In this study’s first year 93 firms
recorded a valuation allowance of zero. In the first year of the Miller and Skinner study
72 firms reported a valuation allowance of zero. The other years of this study were also

comparable along these parameters. Of course Miller and Skinner covers 3 years and this

 

7 Barth and Kallapur (1996) is a definitive work on the effects of scale differences.

20

study creates a panel consisting of 4 years. Note that with the variable DIFFAT that the
maximum exceeds one in a number of years. The net deferred tax assets exceed the total
Of all assets. This is not an error. It does not reflect gamesmanship on the part of
management. Many firms in this sample take the net of the deferred tax assets and
deferred tax liabilities and create a valuation allowance for the difference. This
effectively removes any net deferred tax asset from the balance sheet. The net deferred
tax asset may well be larger than the remaining assets. This effect is also due to the types
of firms having 1995 initial public offerings. Many “new” firms have substantial
intangible assets that have been expensed not capitalized under GAAP. For example in
1998: the maximum value for DIFFAT is 3.1235. The maximum value for VAAT is also

3.1235.

Model and Test of Discretionary Accruals

The study employs three definitions of accrual to determine if there is eamings
management during the lock-up period, the modified-Jones, the Collins and Hribar, and
the Healy methods. Each is used to test hypothesis H1 by comparing both the proportion
of experimental group firms reporting positive accruals during the lock-up period and the
mean accrual during this period with the same parameters in the control group firms. It
also compares both mean accruals and the proportion of experimental group firms
reporting income-increasing accruals during the post lock-up period with control group
firms. If H1 is to be rejected, both the mean of ACC and the proportion of firms
reporting income-increasing accruals must be Statistically the same during the lock-up

period and statistically different during the post lock-up period. By using a matched

 

8 The author has also nm the test found in this paper using total assets as a deflator. The results presented

21

control group, the effects of confounding factors due to the direction of the economy and
requirements of the industries represented during the time period studied are muted. Any
effects observed should be due primarily to the difference in management incentives.
Because the firms in the control group have been selected precisely because they are not
recent IPOs, their managers should not have, as a whole, the same income increasing
incentives.

Table 4a part A illustrates the Collins Hribar definition of accrual. Observe that
during the lock-up period the proportion of firms in the experimental group reporting
income-increasing accruals is 70 percent. The proportion of firms in the control group
reporting income-increasing accruals is 61 percent. A test of the inequality of these
proportions generates a z score of 1.923, p-value of < 0.05 indicating that the proportion
of firms in the experimental group reporting income-increasing accruals is larger than the
proportion of firms in the control group reporting income-increasing accruals during the
lock-up period. A test of the means during the lock-up period between the study and the
control group generates a t of 2.87, p-value of < 0.05. This indicates that the mean
accrual is significantly greater for the experimental group than it is for the control group.

To insure that this finding is not an artifact of the method by which samples were
selected the same tests are run for the period after lock-ups expire. See Table 4a section
B. If the two groups are good matches, one would expect no difference in the income
increasing accruals once the lock-up period has expired. Also, one would expect no
difference in either the proportion of firms booking income increasing accruals or the size
of the mean accrual when comparing the experimental and control groups in the post-

lock-up period. A test comparing the mean accruals for equality between groups returns

 

are robust to the deflator utilized.

22

a t of 1.96. The null hypothesis stating that the mean accruals are the same cannot be
rejected at any traditional level of significance. A test of the equality of the proportion of
firms reporting positive accruals between the two groups returns a z score of 5.345,
which is significant at the 0.05 level. The null hypothesis stating that the proportions are
the same cannot be rejected.

A test of H2 involves a comparison of the experimental group with itself across
the two periods being tested. To reject the null of H2 the proportion of firms reporting
income-increasing accruals would have to be the same in both the lock-up and post-lock-
up periods. See Table 4a part C. During the lock-up period, 70 percent of the firms
report income-increasing accruals. After the lock-up period, only 64 percent of the
experimental firms report income-increasing accruals. A test of these proportions
generates a Z score of 1.37, p value of < 0.05. The null hypothesis that these proportions
are the same cannot be rejected. A comparison of the mean accruals between periods for
the experimental group indicates that the mean accrual is higher during the lock-up
period. The t statistic is 3.61. It permits the rejection of the null hypothesis that the
means are the same at p < 0.05.

Neither H1 nor H2 can be rejected by these tests. Note Table 4a aggregates both
the two lock-up and the two post-lock-up years. The results are not sensitive to the level
of aggregation. Testing years individually does not alter the findings. The aggregation is
an act of parsimony to benefit the reader. .

Table 4b part A illustrates the test of the modified-Jones definition of accrual.
Observe that during the lock-up period the proportion of firms in the experimental group

reporting income-increasing accruals is 54 percent. The proportion of firms in the control

23

group reporting income-increasing accruals is 37 percent. A test of the inequality of
these proportions generates a Z score of 3.74, p-value of < 0.05 indicating that the
proportion of firms in the experimental group reporting income-increasing accruals is
larger than the proportion of firms in the control group reporting income-increasing
accruals during the lock-up period. A test of the means during the lock-up period
between the study and the control group generates a t of —0.0028, p-value of > 0.05. This
indicates that the mean accrual may not greater for the experimental group than it is for
the control group.

To insure that this finding is not an artifact of the method by which sarnples were
selected the same tests are run for the period after lock-ups expire. See Table 4b section
B. If the two groups are good matches, one would expect no difference in the income
increasing accruals once the lock-up period has expired. Also, one would expect no
difference in either the proportion of firms booking income increasing accruals or the size
of the mean accrual when comparing the experimental and control groups in the post-
lock-up period. A test comparing the mean accruals between groups returns a t of
0.0520. The null hypothesis stating that the mean accruals are the same can be rejected at
the traditional level of significance. A test of the equality of the proportion of firms
reporting positive accruals between the two groups returns a Z score of 6.84, which is
significant at the 0.05 level. The null hypothesis stating that the proportions are the same
cannot be rejected.

A test of H2 involves a comparison of the experimental group with itself across
the two periods being tested. To reject the null of H2 the proportion of firms reporting

income-increasing accruals would have to be the same in both the lock-up and post-lock-

24

up periods. See Table 4 part C. During the lock—up period, 54 percent of the firms report
income-increasing accruals. After the lock-up period, only 41 percent of the
experimental firms report income-increasing accruals. A test of the inequality of these
proportions generates a Z score of 2.97, p value of < 0.05. The null hypothesis that these
proportions are the same can be rejected. A comparison of the mean accruals between
periods for the experimental group indicates that the mean accrual is higher during the
lock-up period. The t statistic is -.O718. It does not permit the rejection of the null
hypothesis that the means are the same at p < 0.05.

Under the modified-Jones definition of accrual, neither H1 nor H2 can be
completely rejected. The modified-Jones model rejects the hypothesis with regard to the
average accruals. It cannot reject these hypotheses with regard to the proportions of
firms demonstrating positive accruals. Note Table 4b aggregates both the two lock-up
and the two post-lock-up years. The results are not sensitive to the level of aggregation.
Testing years individually does not alter the findings.

Table 4c part A, Healy definition of accrual, observe that during the lock-up
period the proportion of firms in the experimental group reporting income-increasing
accruals is 60 percent. The proportion of firms in the control group reporting income-
increasing accruals is 42 percent. A test of the inequality of these proportions generates a
Z score of 4.110, p-value of < 0.05 indicating that the proportion of firms in the
experimental group reporting income-increasing accruals is larger than the proportion of
firms in the control group reporting income-increasing accruals during the lock-up

period. A test of the means during the lock-up period between the study and the control

25

group generates a t of 5.4038, p-value of < 0.05. This indicates that the mean accrual is
greater for the experimental group than it is for the control group.

To insure that this finding is not an artifact of the method by which samples were
selected the same tests are run for the period after lock-ups expire. See Table 4c section
B. If the two groups are good matches, one would expect no difference in the income
increasing accruals once the lock-up period has expired. Also, one would expect no
difference in either the proportion of firms booking income increasing accruals or the size
of the mean accrual when comparing the experimental and control groups in the post-
lock-up period. A test comparing the mean accruals between groups returns a t of
0.8270. The null hypothesis stating that the mean accruals are the same cannot be
rejected at any traditional level of significance. A test of the proportion of firms
reporting positive accruals between the two groups returns a Z score of 1.725, which is
not significant at the 0.05 level. The null hypothesis stating that the proportions are the
same cannot be rejected.

A test of H2 involves a comparison of the experimental group with itself across
the two periods being tested. To reject the null of H2 the proportion of firms reporting
income-increasing accruals would have to be the same in both the lock-up and post-lock-
up periods. See Table 4c part C. During the lock-up period, 60 percent of the firms
report income-increasing accruals. After the lock-up period, only 41 percent of the
experimental firms report income-increasing accruals. A test of the equality of these
proportions generates a Z score of 4.501, p value of < 0.05. The null hypothesis that
these proportions are the same can be rejected. A comparison of the mean accruals

between periods for the experimental group indicates that the mean accrual is higher

26

during the lock-up period. The t statistic is 8.5992. It permits the rejection of the null
hypothesis that the means are the same at p < 0.05.

Neither H1 nor H2 can be rejected by these tests. Note Table 4 aggregates both
the two lock-up and the two post-lock-up years. The results are not sensitive to the level
of aggregation. Testing years individually does not alter the findings. The aggregation is

an act of parsimony to benefit the reader.

Model and Test of the Valuation Allowance

VADTA is a percentage indicating what proportion of the deferred tax assets have
been recognized as being unrecoverable. If the valuation allowance is employed as an
income increasing vehicle during the lock-up period one would expect that the mean of
the variable VADTA would be smaller during the lock-up period than it is in the post-
lock-up period. See Table 5. A test of H3 involves a comparison of the experimental
group with itself across the two periods being tested. To reject the null of H3 the mean
VADTA, a test of the equality of these means across the lock-up and post lock-up periods
is required. A comparison of the mean of VADTA between periods for the experimental
group indicates that the mean of VADTA is 0.32027 during the lock-up period. The
mean of the variable VADTA is 0.32344 during the post lock-up period. The t statistic
testing the equality of these means is -0.0947. The null H3 that these means are the same
cannot be rejected at any meaningfirl level of significance.

Because the test of H3, the use of the valuation allowance to increase income is

negative, perhaps there is something unusual about the composition of the valuation

27

allowance in this sample. To remove this concern, the study replicates Miller and
Skinner (1998) in this post-IPO environment.

To test the three compliance propositions listed earlier, one requires measures of
these firms’ net deferred taxes, their expected future taxable income, and the extent to
which their deferred tax assets are comprised of the tax benefits of tax loss and tax credit
carryforwards. Measures of net deferred tax credit and loss carryforwards come directly
from the tax footnotes of the 10K. Measures of managers’ expectations of firture taxable
income are less obvious. Beaver in Market Efficiency, Financial Renortirlg: An
Accounting Revolution indicates that the best guess of future earnings is those of the
most recent year past. I employ income from operations before the adjustment for the
change in deferred tax asset valuation account. Another variable commonly used as a
method of predicting future earnings is the ratio of the firm’s market value divided by the
book value of equity. This ratio expresses the market’s opinion as to a firm’s future
earnings. This variable is COMPUSTAT’S mnemonic MKBK.

A measure of financial distress is Altrnan’s Z score measure of bankruptcy
(Altman 1983). It measures the likelihood that the corporation will survive. With a
normal cross section of firms, bankruptcy is a rare event. In the IPO environment initial
losses resulting from start-up expenses are common. The potential to encounter firms
with poor financial standing would appear to be greater in the IPO environment. If a
firm’s Z score is less than 1.8 there is a high probability of bankruptcy. If the Z score is
greater than 3.0, then there is a low probability of bankruptcy. This item is designed to
forecast failure in the short-term (within two years). The mean Z score for this sample in

the first year of the study is 9.68 with a median of 5.95. The average firm, based on

28

Altman’s test, does not appear to be in danger of bankruptcy. However, 10 percent of the
1995 sample is below the 1.8 barrier, indicating a high probability of bankruptcy.

Table 6 presents descriptive statistics on the proposition variables. Table 7 shows
the correlation between independent variables. The correlations, in general, are not high
among these variables, indicating that they are measuring different aspects of
performance. There is no danger of violating MLR assumption number 4, which requires
that the independent variables be orthogonal. There are no linear relations among the
independent variables.

I test the first three propositions using the following model:

VADTA = a + 131 DIFDTA - 132 PROA - p3 MKBK +pizs+135 CFDTA + it
Eq (4)

Where VADTA is the valuation allowance scaled by deferred tax assets. PROA is the
average of past ROAs. MKBK is the market-to-book ratio. ZS is the COMPUSTAT Z
score, and CFDTA represents the sum of NOL and tax credit carryforwards scaled by
deferred tax assets. I test this model first by using cross—sectional levels by year.9 The
Breusch-Pagan test indicates the probability of heteroskedasticity at p = 0.05.
Consequently all of the t-tests reported in subsequent tables will employ standard errors
computed in a manner to make them robust to heteroskedasticity.

The first proposition implies that as the quantity (deferred tax assets - deferred

tax liabilities)/tota1 assets decreases, DIFDTA, the valuation allowance, falls. A positive

 

9 Various studies have found sensitivity to management adventurism and audit quality. These studies
frequently bifurcate the world of auditors into Big 5 and Non-Big Five. However, of the 190 firms in the
sample, 184 employed either a Big 5 auditor or a firm that was merged into the Big 5. This variable when
added to the model is not significant and does not change the results.

29

coefficient on DIFDTA is expected. The second proposition indicates that the valuation
allowance should become smaller as evidence of future profitability increases. Negative
coefficients are expected for PROA and MKBK. The third proposition predicts a positive
coefficient on CFDTA. As net operating loss carryforwards and credit carryforwards
increase it becomes less likely that they will be realized and the valuation allowance will
increase.

The estimates using cross-sectional OLS by year are in Table 8. The coefficient
on DIFDTA is positive and significant at p- value of 0.05 for all years as was predicted
by the first proposition. Over the time period fiom 1995 to 1998 the median value for
CFDTA moved from 28 percent to 37 percent of deferred tax assets. During this time
period DIFDTA moved fiom 18 percent to 9 percent of differed tax assets. It is possible
that as these firms mature the value of the carryforwards will be a greater predictor of the
valuation allowance than the difference between deferred tax assets and deferred tax
liabilities.

Proposition 2 predicts a negative coefficient on all measures of profitability.
PROA is negative and significant for all years for which it is available.

As predicted by proposition 3, CFDTA is positive and significant in all years. The
sum of net operating loss carryforwards and credit carryforwards appears to dominate the
other variables, as was the case in the Miller and Skinner study. This is not surprising
because for many of the frrrns the valuation allowance is set at the sum of the net
operating loss carryforwards and credit carryforwards. There are 760 data points across
the 4 years of the study. For 462 entries the valuation allowance was set equal to the

carryforwards.

30

MKBK is negative and significant only in the first year of the study. It is possible
that given the volatility of the IPO market that MKBK is not as good a predictor of future
profitability in an IPO environment as it is in the general population of firms.10 Note the
dependent variable VADTA is limited. It is roughly continuous over strictly positive
values, but it is zero for a nontrivial fraction of the firms in the sample. Wooldridge
(1999, Chapter 17) indicates that in this instance the appropriate statistical tool is Tobit.
OLS statistics are reported in the body of this study. However, the results are robust to
the use of a model employing Tobit.

Hypothesis 4 theorizes that managers will use the valuation allowance as a means
of smoothing reported income. The valuation allowance will be used as a hidden reserve
to be employed when income from operations falls below expectations. Using the
random walk theory, the best predictor of next year’s earnings is last year’s earnings. To
the extent that this year’s earnings fall below last year’s earnings, one might expect the

valuation allowance to fall to offset the shortfall. The model is as follows:

AVADTA = a + B] ADIFDTA+ [52 APROA + B3 AMKBK + [is ACFDTA +
[36 AIOAT + Ll
Eq.(5)

The measure of earnings is IOAT (COMPUSTAT number A178). This item represents
the operating income of a company after deducting expenses for cost of goods sold,
selling, general and administrative expenses, and depreciation. It is scaled by total assets

(COMPUSTAT number A6). This item is not net of taxes. Any change in the valuation

 

'0 The regressions testing hypothesis one were examined for the effects of outliers. The covratio first
employed by Besley, Kuh and Welsch (1980) was used in the evaluation. Eight data points identified as

31

allowance would affect the after-tax number only. Miller and Skinner found no
correlation of any kind between changes in the valuation allowance and changes in
income from operations. From a reading of FAS 109 one would expect a negative
correlation between the valuation allowance and IOAT. That is, the coefficient on IOAT
will be negative. If an effort to smooth earnings is employed, one would expect the
coefficient on IOAT to be positive. The data collected for this study represent a panel,

' The nature of FAS 109 is to cause management to

190 firms followed over 4 years.1
make a link between years. To set (or not set) a valuation allowance in the current year,
management must predict future profitability sufficient to use deferred tax assets before
they expire. This connection between years makes the use of panel data relevant to the
situation. The data have both time series and cross-sectional dimensions; one cannot
assume that the observations are independently distributed across time. The decision to
create or change a valuation allowance considers aspects of the firm’s financial condition
that may well span more than one year. There may also be unobserved differences
between firms that are time constant and affect the dependent variable. These differences
might be location, industry, management philosophy, etc. To control for these
unobserved effects, fixed effects panel data analysis is employed. Note Miller and
Skinner used a technique called first differences in which the difference between the

years of a variable become the data points. First differencing also controls for omitted

variables; however, it is not robust to serial correlation.

 

outliers were dropped from the regressions without adversely affecting the results presented.

”A panel data set has both a cross—sectional and a time series dimension. It could be called longitudinal
data. I follow or attempt to follow the same firms over the life of the study. Obviously, this is an
unbalanced panel. Several firms drop out of existence before the end of the research period. The statistics
employed are robust to this eventuality. Wooldridge (1999, Chapter 14).

32

This is not to imply an error on the part of Miller and Skinner. To circumvent the
serial correlation issue, they bifurcate their test period into two parts effectively making
the issue of serial correlation moot. When there are only two time periods in a sample,
all test statistics and estimates are identical when first differencing and fixed effects are
compared. This manner of testing the hypothesis sacrifices the power obtained from
using the entire sample of four periods.

A test for serial correlation in this sample of IPOS reveals a significant positive
relation between the current and lagged error terms. The finding of serial correlation
makes it necessary to use a cross-sectional time-series regression model, (Wooldrich
1999, Chapter 14).

The helpful aspects of first differencing are retained by the panel data statistics,
the elimination of the effects of unobserved omitted variables. There are two different
methods of evaluating panel data, fixed effects and random effects. The differences
between these two methods are subtle and beyond the scope of this study. For
completeness, results using both methods are reported in Table 9.12 Please note that
although much is made of the differing management mindset in the lock-up vs. the post-
lock-up years the results shown for the test of H4 are robust to the time frame examined.

The coefficient on PROA is negative and significant as expected. As the change
in past return on assets is positive the change in VADTA is negative. The coefficient on
the change in CFDTA is positive and significant as expected. When net operating losses
and credit carryforwards accumulate the valuation allowance increases. Finally, the
coefficient on IOAT is negative and significant. As the change in income from

operations increases, the valuation allowance becomes smaller. IPOs may be firms with

33

a brief history. Early year losses are not uncommon. As earnings from operations
increase year-to-year management gains confidence in the quality of those eamings.

Bradshaw (2000) attempts to evaluate how analysts use earnings forecasts in
generating stock recommendations. He finds that analysts incorporate their earnings
forecasts into their recommendations in a manner consistent with earnings based
heuristics rather than a theoretically based valuation model. The earnings favored by
analysts appear to be income before extraordinary items and not income from operations.
This model was tested for sensitivity by using the variable PIAT. PIAT is
COMPUSTAT’S annual data number A170 deflated by total assets, annual data number
A6. PIAT represents operating and non-operating income before provisions for income
taxes and minority interest. It specifically excludes extraordinary items and discontinued
operations. The correlation between PIAT and IOAT is high, .8512. See Table 7. The
model tested is identical to that in equation 5 except that PIAT is substituted for IOAT.
See Table 9 Part 2. The sign on the coefficient for PIAT is negative and significant. The
coefficients and the t-scores are Significantly larger for PIAT. The result is, as one would
expect. If the valuation allowance is driven primarily by loss carryforwards, then, any
income of the appropriate source13 will serve as a means of reducing the NOL.

One of the basic rules in experimentation is that if there is no variation in a
variable, its correlation with other variables does little to inform. Certain firms within the
sample demonstrate persistent zero valuation allowances. Of the 190 firms in the sample,

61 do not report any change in the valuation allowance. It is persistently zero for the life

 

'2 An adjusted R squared is not available for this statistic.
'3 Source rules are important in using NOL’S and credits under the US. Tax Code. A NOL generated in a
foreign jurisdiction cannot be offset by income generated in the US. The [RC distinguishes between

34

of the study. Because there is no variation in the valuation allowance it is obvious that
the valuation allowance is not being used to smooth income. These firms are removed
from the sample. It is possible that their presence is affecting the Sign or the significance
or the variables testing for income smoothing. The model is otherwise unchanged from
equation 5. See Table 10. The coefficients on DIFDTA, PROA, CFDTA, IOAT, and
PIAT do not change in either sign or significance. Importantly, IOAT and PIAT remain
negative and significant. Note that when compared with Table 9, the coefficients on
IOAT and PIAT are smaller. The R squared for the model presented in Table 10 also is
smaller.

Unlike the Miller and Skinner study, a change in earnings appears to be a factor ,
considered by management when evaluating the required size of the valuation allowance.
The Miller and Skinner study found no correlation between first differences in the
valuation allowance scaled by deferred tax assets and the first differences in income from
operations before tax scaled by total assets. Both studies found huge coefficients and t-
scores on the variable measuring the sum of net operating losses and deferred tax credits
deflated by deferred tax assets. The difference in the findings probably results from the
difference in the life cycles of the firms studied. In the Miller and Skinner study the
firms are well established. The net operating losses and or deferred tax credits may have,
in effect, become structural components of the balance sheet, with little or no hope of
recovery. In the Miller and Skinner paper, management may set a valuation allowance,
not because future profitability is in doubt, but rather because the net operating losses and

deferred credits are near expiration. If these deferred tax assets are near expiration there

 

ordinary income and capital gains. It does not distinguish between income from operations and income
from administrative activities.

35

may be little that even a profitable firm can do to recover them. In the present study,
because of the newness of the firms in question, management may not only have
concerns about the continued profitability of the operation but the size of those profits.
Demonstrated profitability is not a stated requirement in F AS109 for non-recognition of a
valuation allowance. However, it is difficult to assume future profitability if profitability

has never been demonstrated or if continued profitability is in doubt.

CONCLUSION

This study explores the nature of earnings management in the post-IPO lock-up
period. It extends Healy (1985) in that it demonstrates management’s willingness to.
employ income increasing accruals in a situation in which doing so has the potential to
enrich those managers. The study extends Teoh, Wong and Rao (1998) and Ritter (1991)
in that it uses institutional knowledge of the IPO process to explain the reasons for the
income management they first observed. Given the timing of the prospectus, fiscal year
ends, and the issue date, it is unlikely that the financial statements of the IPO year play
any significant role for those investing in the IPO. The observed propensity for income
increasing accruals in the IPO year seem to be a matter of managerial self-interest.
PTOportionally, more managers employ income-increasing accruals during the lock-up
period than do managers for whom no Similar incentive exists.

The valuation allowance under FAS 109 has been a suspected tool of income
management since its inception. This study does not support this suspicion. In
examining the determinants of the valuation allowance for deferred tax assets under F AS

109 in the context of recent initial public offerings, the study finds the major

36

determinants of the valuation allowance to be the level of net deferred tax assets and the
level of tax loss and credit carryforwards. The study also finds a Strong negative relation
between a firm’s average past return on assets and the level of the valuation allowance.
These relations reflect general compliance with the letter of FAS 109. The findings are in
agreement with Miller and Skinner (1998).

This study also demonstrates a negative correlation between changes in the
valuation allowance and changes in income from operations before taxes and income
from all sources before extraordinary items and taxes. Miller and Skinner did not
observe this effect. With IPOS, the fact that there is an inverse relation between changes
in the valuation allowance and earnings could reflect management’s uncertainty about
future profitability and its magnitude.

The data are also used to determine if the valuation allowance is used to increase
income during the lock-up period. This test also was negative. I conclude that the
valuation allowance under FAS 109, when examined in an environment in which there is

evidence of income management, is not used to either increase or to smooth income.

37

References

Accounting Principles Board (APB). 1967, Accounting for Income Taxes, Accounting
Principles Board Opinion No. 11, New York, NY: American Institute of Certified Public
Accountants.

Altman, E. I. 1983, Corporate Financial Distress, New York, NY: John Wiley & Sons.

Baiman, S., 1990, Agency Research in Managerial Accounting - A second look,
Accounting Organizations and Society 15, pp. 341 -371.

 

Beaver, W., Market Efficiency. Financial Reporting: An Accounting Revolution. (2"d
Edition): pp. 130-175.

Beaver, W., M. McNichols, and K. Nelson, DO firms issuing equity manage their
earnings? Evidence from the property-casualty insurance industry. Working paper,
January 2000.

Behn, B., T. Eaton, and J. Williams, 1998 The determinants of the deferred tax allowance
account under SFAS No. 109. Accounting Horizons, March, Vol. 12 No.1, pp. 63-78.

Beneish, M., 1997, Detecting GAAP violation: Implications for assessing earnings
management among firms with extreme financial performance. Journal of Accounting
and Public Policy. Vol. 16, pp.271-309.

Bradshaw, M., How do analysts use their earnings forecasts in generating stock
recommendations? Working Paper, January 2000.

Brav, A., and P. Gompers, Insider trading subsequent to initial public Offerings: Evidence
from expirations of lock-up provisions. Working paper February 2000.

Burgstahler, D. and I. Dichev, 1997, Earnings management to avoid eamings decreases
and losses, Journal of Accounting & Economics, Dec., Volume 24, pp. 99-126.

Collins, D and P. Hribar, 1999, Errors in estimating accruals: Implications for empirical
research. Working Paper, University of Iowa.

Dechow, P., R. Sloan, and A. Sweeney. 1995. Detecting earnings management. The
Accounting Review, Vol. 70 No.2, April, pp.l93-225.

Dechow, P. and R. Sloan, 1991. Executive incentives and the horizon problem: an
empirical investigation. Journal of Accounting and Economics, July 1994, pp. 3-42.

38

Dechow, P., S Richardson, and AI. Tuna, 2001 Earnings management and costs to
investors from firms meeting or slightly exceeding benchmarks. Working paper,
University of Michigan.

Espahbodi, H., P. Espahbodi and H. Tehranian. Equity price reaction to the
pronouncements related to accounting for income taxes. The Accounting Review,
October, Vol. 70, Issue 4, pp. 655-669.

Financial Accounting Standards Board (F ASB), 1987. Accounting for Income Taxes.
Statement of Financial Accounting Standards No. 96, Norwalk, CT: F ASB.

Financial Accounting Standards Board (F A83), 1992. Accounting for Income Taxes.
Statement of Financial Accounting Standards No. 109, Norwalk, CT: FASB.

Guay, W., S. Kothari, and R. Watts, 1996. A market-based evaluation of discretionary
accrual models. Journal of Accounting Research, Vol. 34 Supplement 1996, pp. 83-105.

Hayn, C. 1995. The information content of losses. Journal of Accounting and Economics,
20. September: 125-153.

Healy, P.,1985, The effect of bonus schemes on accounting decisions, JAE, April, pp. 85-
107.

, 1996. Discussion of a market-based evaluation of discretionary accrual models.
Journal of Accounting Research, Vol. 34 Supplement 1996, pp. 107-115.

Jones, J ., 1991, The effects of foreign trade regulation on accounting choices, Journal of
Accounting Research, pp. 193-228.

Miller, G. S. and D. J. Skinner, 1998, Determinants of the valuation allowance for
deferred tax assets under SFAS No. 109. The Accounting Review, (April): 213-233.

Peavey, DE, and H. Numberg, 1993, F ASB 109: Auditing considerations of deferred tax
assets, Journal of Accountancy (May): 77-81.

Petree, T. R., G. J. Gregory, and R. J. Vitray, 1995, Evaluating deferred tax assets.
Journal of Accountancy (March): 71 -77.

Phillips, J ., M. Pincus, S. O. Rego, 2001, Earnings management: New Evidence Based on
Dewferred Tax Expense. Working paper, University of Iowa.

Ritter J. R.,1991, The long run performance of initial public offerings, Journal of
Finance. No. 46, pp. 3-27.

Schipper, K. 1989, Commentary on earnings management. Accounting Horizons 3: pp.
91-102.

39

Teoh, S. H., T. J. Wong and G. R. Rao, 1998, Are accruals during initial public offerings
opportunistic?, Review of Accounting Studies, 3, pp. 175-208.

Willenborg, M. 1999, Empirical analysis of the economic demand for auditing in the
initial public offerings market. Journal of Accounting Research, Vol. 37 No.1, pp. 225-
238.

Wooldrich, J ., Introductory Econometrics: A Modern Aopmggh. South-Western College
Publishing, 1999.

40

omp

o: I-
woos. Fun
msnmom
smoo
known

oo..
2.2m-
oooooow
ommoow
ono
wnomv

mop

vow p N-
ooowovw
Nmmovm
omoo

ozov
25:92.0

E9“. chcc.
maOZ.

oo..

o
ooosbos P
39%:
Novoow
enooom

oow

o
ooomoo:
mooosmw
o 500
oomown

oo.

o
ooooooop
oovmsww
v .. nmo
Evoow

mofim
ww..<m

9.90 .9300

oo.

oo: ..
onomookm
unvuoov
no to
Ewnooms

oo..

mo Pop
own? wmmm
pnwmwvv
oomno
vommvso

mop

oNoN
oowmsbmv
mnoomov
3 Foo
odovo .. o

.20.. £32
...<

2.83.2. 8.2.. 82 .5225 32 28>

Goo. :.

oop
KNQNN-
oonnnov
Ymommvn
pm wow
noNop pm

oow
mvnoo..-
oooommn
ooéoowv
oooo
ooéoowv

oo—
swoon-
oowooow
oooovww
.. «mm
5.880

20:92.0

EOE oEocc.

waOZ.

oo..

o
oooXKo
.mmoomoo
vowpow
woman 5

oow

saw
ooooooo
oowhoo
mKom ..o
Qommnow

oo P

o
ooovoNo
véommws
goon
unwoou

”one
$35

25.“. 035mm

82 c. mOa. uses: were. com 3 «2.2.2.0.. 9.3.88 .9950

.. mom/SP

oow
ooh;
ooooooox.
mow room
ovoow P
«Koovoh

oo.

oo Pow
ooooommm
ms 5 wow
omuuo
Ehoopoo

oo p

Nome
oooooovv
acnvown
mowmm
fivoouvw

.20... 832
._.<

.30
232.22
232.x<2
om
Z<.Ow2
Z<m2
58..

.30
232.2.2
232322
ow
Z<.Dw2
Z<m2
ooop

.waO
232.2.2
232322
Om
ZSow2
Z<w2
moor

m2<z
w..m<_m.<>

41

oop
ode?
msvommm
op ..v 5
oomo

mmomm
30:92.0
E2“. cha...

9.02.

.3 E... E9. Simodzoo .83. 20.2. .3 :23:er 5.3.853 .23 2.2.82.0 Ea... oEooc. .1. 9.02.
N: a E2. .83an0 60.3 .9. u mm..<m
.ofi. E2. .2»:an0 .238 .82 u h<

.2523. an occcoe 2m mm.nc..m>

.333 .90. one coco O... .3 coscEE mg... 8.. 3 amazon. 9.96 .9250 9?

.59»... 5.83m E9. Em End .82 :. men. 9..me mE.... om. moatanu ancm

om. om— om. om.
o meme 888.. new
82.8: mama 3mm .88 8 ~36me
mnmmE «Rom: 12.3mm ZEN Em
comm F P omen. . momma Noun?
85:. 38:. 8.89%. 9:35
39.980
3.3 .23 232 E2“. 930... 83m
wm4<m h< mnoz. mw._<m

9.0.0 6:50 9...... 0...an

38. ...
82 c. 8... 9..me 2...". 8. .o 8.33% 9.3.88 .9880

noncocoo w w..m<._.

oow

owns
ooooomms
3:.on
mioom:
meow 3

.30... 232
._.<

.30
232.2.2
2D2.X<2
0m
Z<.Dw2
2(m2
ooow

w2<Z
w..m<.m<>

42

Descriptive Statistics on the Accrual Positions of 190 Firms Making IPOs in 1995

1995
MEAN
MEDIAN
SD
MAXIMUM
MINIMUM
Obs.

1996
MEAN
MEDIAN
SD
MAXIMUM
MINIMUM
Obs.

1997
MEAN
MEDIAN
SD
MAXIMUM
MINIMUM
Obs.

1998
MEAN
MEDIAN
SD
MAXIMUM
MINIMUM
Obs.

Sample Firms

0.051048
0.02573
0.13559
0.71066

-0.38459
141

0.018662
0.016505
0.130889
0.41 1 195
-0.606972
179

-0.056091
-0.01746
0.304371
0.28254
-3.522041
176

-0.042293
-0.025525
0.153917
0.290875
-1 .300136
170

TABLE 2

and the Control Group

The reported accruals are all scaled by total assets.

43

Control Group

-0.017936
-0.020268
0.1 15233
0.35222
-0.456596
141

-0.013935
-0.017603
0.122433
0.361479
—0.403878
179

-0.027837
-0.020843
0.1 82686
0.407227
-1 .941 646
1 76

-0.09931 9
0.032823
0.36592
0.277033
-3. 1 621 32
1 70

u:

o

P
owmmo
Noomo
noomo

Now

o

P
5.56
Sud
vmwmo

v:

o

P
Iowo
Pomoo
....ow.o

(HO

oo ..

o

..
ooovo
NNooo
moomo

oo..

o

..
Pmomo
o
ooowo

o:

o

..
enovo
o
vopmo

(#0

so?
oovoo-
w
mono...
Nnono
mooNo

mow
mmvdw-
F
mmomd
n «moo
good

#2.
ooovfl?
e

momoé
owmoo
omo...o

<..O

 

womflngu. w02<>>0....< .Eb-<._.o

>mm<0

(kn—“.0

<...D<>

<._.D...=D

Q?

o

on Poo
NNoNo
....wo.o
omooo

mow

o
momoo
ooooo
.booo
mnvoo

2...

o
«mow...
Powwo
Food
vvvoo

<._.

mop

o

3 5.?
ommo
..ooo.o
omooo

oow

o
voooo
«or «o
o
ammoo

o:

o
No9...
oov...o
o
Numoo

(h

 

womi<>>mOm wOZ<>>Oifid ._._.D..<.PD

>mm<0

...<...O

m m..m<._.

.525

no?
Sofia-
.1. 5.?
0036
..m~o.o
good

Now

vow ...o-
oo :6
N3.” .. .o
mnwoo
mouooo

Nov
mommo-
some. ..
moo .. .o
mmomoo
ongoo

<h

2&5 :25
89 a. no... 9..me we... 8. .o 29.8.. E 8:98 9.. co 8.6.35 2.38me

oo..

ooooo
3 5...
mNmNo
vmvoo
now _..o

mop

noooo
oo :o
nomwo
Novoo
Noooo

Now

ooooo
homv...
void
Povoo
NNooo

<._.

(PD

.30
5.32.2.5.
5.3.2593).
Ow
250m}
Z<m_2
hoop
.mnO
5.22.2.5.
23.2.5}...
om
Z<.Dm$.
Z<w_2
coop
.30
23.2.2.2
$.35..X<S.
0w
250w...
Z<m_2
moor

w2<2
w..m<.m<>

82,32,835”. .60... xm. ucm ..02 ME... m .0 E8 o... «.5353. u $5353.50
.2058. xm. ME... on. E9. 933 xm. 8.8.2. .0. 8526.3 co..m=.m> u moz<>>O....<
.2oEoo. xm. ME... 9.. E9. .399 $2.53. xm. 8.8.2. u ...b
.399 Somme x9 8.8.2. .i. 5.0

.326. mm 8&8 2.... mo_€..m> .9... Va.

ME... .o 8.058. xm. 0... E9. 2m «ED 68.. :. mOn: 9..me mE... om? mcmtano oEEmw

45

 

 

N: For Now mt. Now Now 3.. .30
o o voNN- o o hoovo- Nwooo 232.2.2
.. _. _. mova mowfim one...” one...” 2D2_x<2
v Fond Nmovo vommd omwmo oonno hwomo mound Om
memo «oooo osmoo on ..o.o omooo gmoo ovmoo Z<_Dm2
oohmo owmmo omooo op ...o oomfio owgo ooo...o Z<w2
moor
(PO (#0 (#0 <... <._. (h <._.

mom/14310“. wOZ<>>O._._¢ ...PD-<...D massmou. mozi<>>044|¢ ohm—42.0 a

>mm<o >mm<0

(hDuU <...O<> (ham—D ...<...0 ._.<<> ...(mua ._.<<.PD w2<z

m..m<.m<>

moor :. mOa. 9.3m.)— mE..n. oo.. .0 mcoEmoa xm... notmomo 9.. co 8.6.65 o>..a..omm0
8:5...00 m m...m<._.

TABLE 4a
Summary of the Association Between Accruals and Lock-up Periods

Hribar Definition of Accrual
Proportion of Accruals t-test for
With a Given Sign Mean Difference
Portfolio POSITIVE NEGATIVE Years Accruals in Means

A. Comparison of the Study and Control portfolios during the lock-up period.

Study .70 .30 1995/1996 .08002
Control .61 .39 1995/1996 .04225 267'
test of proportion S>C
differences 1 .923“
S>C
B. Comparison of the Study and Control portfolios after the lock-up period.
Study .64 .36 1997/1998 .03252
Control .64 .36 1997/1998 .05793 1.96'
test of proportion S=C
differences 5.345”
S=C
C. Comparison of the lock-up period with the post lock-up period for the Study portfolio alone.
Study .70 .30 1995/1996 .08002
Study .64 .36 1997/1998 .03252 361'
test of proportion St1 >St2
difi‘erences 1.37
St1>St2

Portfolio: Study consists of a sample of 190 firms having an IPO in 1995.

Portfolio: Control consists of firms matching the Study portfolio on size and SIC code.
None of the control firms has undergone an ipo within 3 years of 1995.

Accruals are deflated by the book value of total assets.
" Indicates a t significant at p < .05.
” Indicates a z-score significant at p < .05.

These tests are not sensitive to the IeveI of aggregation. Tests of individual years provide
similar results.

46

TABLE 4b
Summary of the Association Between Accruals and Lock-up Periods
Modified-Jones Definition of Accrual

Proportion of Accruals t-test for
With a Given Sign Mean Difference
Portfolio POSITIVE NEGATIVE Years Accruals in Means

A. Comparison of the Study and Control portfolios during the lock-up period.

Study .54 .46 1995/1996 -.0025
Control .37 .63 1995/1996 -.0023 -0.0028
test of proportion S>C
differences 3.74“

S>C

B. Comparison of the Study and Control portfolios after the lock-up period.

Study .54 .46 1997/1998 .0001
Control .54 .46 1997/1996 -.0003 0.0520
test of proportion S=C
differences 6.83”

S=C
C. Comparison of the lock-up period with the post lock-up period for the Study portfolio alone.
Study .54 .46 1995/1996 -.0025
Study .41 .59 1997/1998 .0001 -0.0718
test of proportion St1>St2
differences 2.97“

St1>St2

Portfolio: Study consists of a sample of 190 firms having an IPO in 1995.

Portfolio: Control consists of firms matching the Study portfolio on size and SIC code. None
of the control firms has undergone an ipo within 3 years of 1995.

These accruals are by definition a residual.
" Indicates a t significant at p < .05.
” Indicates a z-score significant at p < .05.

These tests are not sensitive to the level of aggregation. Tests of individual years provide
similar results.

47

TABLE 4c
Summary of the Association Between Accruals and Lock-up Periods

Healy Definition of Accrual
Proportion of Accruals t-test for
With a Given Sign Mean Difference
Portfolio POSITIVE NEGATIVE Years Accruals in Means

A. Comparison of the Study and Control portfolios during the lock-up period.

Study 0.60 0.40 1995/1996 0.03427
Control 0.42 0.58 1995/1996 -0.01627 5.4038'
test of proportion S>C
differences 4.1 1 0“

S>C

B. Comparison of the Study and Control portfolios after the lock-up period.

Study 0.41 0.59 1997/1998 0045822
Control 0.35 0.65 1997/1998 0049657 0.8270
test of proportion S>C
differences 1 .725

S>C
C. Comparison of the lock-up period with the post lock-up period for the Study portfolio.
Study 0.60 0.40 1995/1996 0.03405
Study 0.41 0.59 1997/1998 004582 8.5992‘
test of proportion St1>St2
differences 4.501”

St1>St2

Portfolio: Study consists of a sample of 190 firms having an IPO in 1995.

Portfolio: Control consists of firms matching the Study portfolio on size and SIC code.
None of the control firms has undergone an ipo within 3 years of 1995.

Accruals are deflated by the book value of total assets.
‘ Indicates a t significant at p < .05.
” Indicates a z-score significant at p < .05.

These tests are not sensitive to the level of aggregation. Tests of individual years provide
similar results.

48

TABLE 5

Summary of the Association Between
the Valuation Allowance
and Lock-up Periods

t-test for
Mean Difference
Portfolio Years VADTA in Means
Comparison of the lock-up period with the post lock-up period for the Study portfolio.
Sample 1995/1996 0.3202699
Sample 1997/1998 0.3234425 -0.0947'
Proportion Z-test of
With VADTA Proportions
Sample 1995/1996 .55
Sample 1997/1998 .58 -0.99"

Portfolio: Study consists of a sample of 190 firms having an IPO in 1995.
VADTA is the valuation allowance deflated by deferred tax assets.

’ P > m = 0.09247. The null hypothesis that the means are equal cannot be
rejected at any meaningful level.

“ P > z = 0.8410. The null hypothesis that the means are equal cannot be rejected
at ant meaningful level.

This test is not sensitive to the level of aggregation. Testing separate years provides
similar results.

49

 

3x9 9.6 mEE. ...cEEomzxc 992. mEooc. .93. u ._.<.m

3x9 993 39333 9. cEcoE u ...<0.

<0”. .8. 8992 u <9...

xoom o. .93.). u xmxs.
$6390.39 .o 3:598... 26. 6 3.8.9.. .235 .o od .o o:.m> < $02.95.:ch .o
3.339.. no... a 3.36... 5.. :65 m3. .o o:.m> < $039ch 9.20.35 .0 oo...cE m m. ..
.293. 5.33m. c. 325:. 5.3.298 .o 35.9 o... m. 2.; .933 3.2.3.63 moo. +
.8..._.nm__ .0 was, 582.38 .o 2......) .9222. o. + .988 99:3. 3 +

.933 9239.93 3599. v4 .1233 3859.36 9.2.95. N... u mN

326:9 3 amazon 9m 3.3..m>
.593. 5.33m E9. 9m Sun. .32. £601. 9..me «E... om. 3m..oEoc cEEmm

oow oo.. mm? 8.. oow mZO...<>mwme
omoo.m- ooono- 90.5mm- Fmodmw- oofis- 232.2.2

oommo oovmo ..ov.vo moo: omdmw 2D2.X<2

Noowo vmopo omnwfim oooo. Pm oNovow ow

o ..uoo ovooo and ice memo...“ Z<.Om2

onvoo vooo mm ..oN- 3.2.... o ..mo. 3 Z<w2

...(E .55. (Can. ¥m¥2 wN m2<z m..m<_m.<>
moo? :.

33m..m> E33335 350 :0 3.6.35 9.3.630
o m..m<..

50

.2003 3.0. B 00300 .mE0.. 0050.00.30 0.3 33. 0.0.00 E... 0... .0 0E8... .0: 0... 0. u .2...
0.003 3.0. >0 00.000 .33. 0.0.00 2.0.3.000 E0.. 0Eoo... 0... m. n ...(O.
.288 8. 8.8.0.. .3 82>...
0033.03.30 ..00.0 .3. 0.. 0:0 00.03.030.00 30. .3. 0E... 0 .0 E00 0... 2:03.00. 0...... u 5.9.0
0.3. .000 0. .0...0E ME... 0... 0. u xmv...‘
.>0.0:.....00 .0 5.300... 2.0. 0 330.0... 330.0 .0 0.0 .0 33> < .>0.0:....30 .0
$.88... .5... m 8.8.2. 5.. 8... 80. .o 029, < sensing 8.0.8... .o 350... a m. ..
$0.8. 68080 c. 82.6... 5.8.358 .0 850E 9.. m. a...» .288 8.2.8.8. 000. +
.8.....8.. .0 23> .3038 .o 8.9, .032. 0. + .288 8.2.000. on +
3.003 3.0.305E00 000.03.. vé + 3.003 3.0.3.300 00....03. N. u mN
.0303 ..0 E20. .30 30.00 ME... 0 .0 000.03 0..» u (Cm...
.0303 .3. 00.3.00
.5 00.000 3.....3: x0. 00.3.00 03 0.003 .3. 00.3.00 ME... 0 0003.00 0000.00... 0.; u (.505

. 0.00.0 «000.0- N090 033.0 «~30 mm. ...0- .52...
P 0006.0- 2.3.0 000.0 0000.0 030.0- ..(O.
F 0. P ..0- 09.0.0 ~90- 03:0 (.500
. 003.0 033.0 m-0.0- xmxs.
P 3; 9.0- 83.0 mN
P v0.2.0. (can.
P <hou..o

ESQ 5.0. 5.9.0 xmxs. wN (Gm... (#005
0.00.00.000 00.3.0000 00030...

3.03 30000005 03

0 0.3... 0. 000: 3.3.3) 30000000... 0353 30.3.0000
\. m..m<...

51

 

.mEE >cmE 5. 03.8.85 8: cm; (Om .mBom 63 >37. £5 MB 9.5: mE: m5 520 2.

.532 .o no. u a 8 8:85:98 38855 xmtfimm :5

8828338906: 2 .259 29.6 Bane—Em :o 823 828252 9: 9m. £35583 5 888:: 9:.
.m new m 83m... :0 83% 33.355 235 B 8:055..me 9;

.3888 388-8 :89 .851 .3888 $288

-88 8:8 888- 8.88 588. 8888 838 E 888.
.2858 :8: 68.3 .8888 58.: 39.8

8388 8:8 888- «~88 .888- 388 88:8 3. 88.
.8888 388-8 £888 .3881 beta GEE

888 8888 $888- 8888 ~88? 888 ~88 v: 88.
.3888 .853 88.: .888 38.

888 .888 -88- -88 : :88 ~88 B. 882

88:8 m 580 x882 mN <98 5.9.5 .8225 .30 88>

355mm 22:“. new «man. .0 82:23.2 :0

2082 xmh 8.580 S 858 mac—826:4 co=m2m> 9: .o mcofimemmm 8.93 m._0
m w4m<H

52

 

. .89. corona. .9: E9. =82

:85 5 E222. 2.3.288 .o: 995 8039...; 305 9.6205 95.8053: .328 cams. 9:
£5. maniacs“. EOE : 38E 2 c2835.... 9: E2. mint? 8 mm 8&2: mm; wN

.830. .5 me. u a .m cosmozEgm «286:. 8.8.23 5...

.0 new m moan-r :o 53% 338:? $85 .0 8:02:ch 9F

53

.888: :38: .33.: .33-: .68.: 8.88: 288
888 ~m-8- 828 8888 :88- 8.88 ~88 :8 588m
2.88-: 28...: 88.: ..~8.-: :88: :8: 8.88m
888 88.8- ~88 .888 888- 8.88 8~.8 :8 8x...
88:8 m 2E <58 xmxz <omn_ <Bn=o 38.2... 8:0 8:85:82
.88..-. .688: :8: .8888: ..~8~.~: 888: £88
.888 ~.-8- 8:8 .888 .888- 8..88 88.8 :8 588m
.8....~-: .85.: 88.: ..B~.~-: 88:8: 88...: 8.85
8.88 8....8- 88:8 .888 ~88- 8~88 ~.8.8 :8 8.8
88:8 m 20. <58 xmxs. <0”... 59.5 .8225 .30 8:85:82

888. 8:95 888. 28>
m5£ooEm 9:85 .0 89:83.2 :0 £32 xmk 8.580

3 :2QO 8526.2 co_.m:_m> 9: .0 308853. 38289390 85..» 0E; Ema .055
m wage...-

..:0.0...0 2.0.8.0.: .0: 0.03 00.00:? 000... 05020:. 05.00050”. $00.0 mam—2 0:.
5.3 0.00.0058 0.0:. .. 0.0:. 0. 8.00050. 0:. 80.. 0.00:? 0 00 00000.0 002. mN
.0o:.:.00 500.8 0. 00:02.08 :0..0:_0> 0:. 0:.0:
.0: 0.0 00:02.0..0 :0..0:.0> 0 505.3 0...... 50.8.2.0 .00:0.<.0..0 :0..0:_0> 0... .0 .05. 0... 0:...00

0. 0.000. ....3 .:000... 0. 8550080. .:0E0o0:0.: 005500 .. 05.50.00 0. 003 .00. 0.... .0 02.0030 0: .-
.:0.000.m0. 0... 50.. 0205.58 0.02. 00:02.0:0 :0..0:.0> 0 000.000. .0>0: .0... 08......

802.0. .0 no. u 0 .0 00:03:90 00.00.05 0.0.8.00 :<.
.0 0:0 0 00.00.? :0 .00000 00.00..0> 000... .0 0:05:03 0...-

.888
.088 8.88-
«A80.Nuv
8.08 ~88-
88:8 m .3:
..08.~-.
.088 88.8-
.08. 7.
80.8 8. .8»
8.0:... 0 .<0_

..08. . :
8.88

..8...:
..08

<._.Dn.0

.33. P C
ammo

.83....
.208

(HO-...o

.8... :
0888

.08. ..
moooo

xmv=2

C00. 5
ooooo

$8....
0888

xmxs.

.831.
0888-

«A08. an
chod-

<Omn.

88.1
888-

.88..-.
8.88-

<01...

.88.
.838

.38.
0888

(ham—D

Amend.
vmwoo

..08.~.
8888

(...o-...o

mom

.08....
0.~.8 80

2.00.0.5 .000

.08.:
~88 80
.08....
0~.8 80

.000.0.:_ .000

..00:0>:0=< :0..00_0> 0.0N .:0.0.0.0n. ....>> 00...“. 00.0:.:.._w 0.00.0.w

o:.....00.:w 00.00:. .0 00.00005. :0 0.000< x0.- 02.0.00

.3 00.00.... 00:030.? :0_.0:_0> 0... .0 08.00050”. .0:0..00w-000.o 00:0m 0:... .- 0.00 .0:0n.

ow m...m<-_.

0.00am
0.00:0...

088
8....

.0..0.::00<

0.00.5
0.00:3.

0.00am
cox...

.0..0.::00<

54

       

IHHI H: mm

1293 02327 1020