INFORMATION ASYMMETRY AND EARNINGS MANAGEMENT:

SOME EVIDENCE

BY

VERNON J. RICHARDSON
University of Kansas
vrichardson@ukans.edu

March 30, 1998

I wish to thank members of my dissertation committee at the University of Illinois for

insightful comments and suggestions: Dick Dietrich (chairman), Don Kleinmuntz, Tom
Linsmeier, Theodore Sougiannis, Dave Ziebart and Shane Greenstein. I also thank Mike
Ettredge, Mark Hirschey, Susan Scholes, Joe Comprix and workshop participants at
University of Arizona, University of Arkansas, Brigham Young University, Georgia State
University, University of Illinois, University of Kansas, London Business School,
University of North Carolina, Purdue University, SUNY-Buffalo, Syracuse University,
Thunderbird and Tulane University. I gratefully acknowledge funding from the Deloitte
and Touche Foundation, the State Farm Companies Foundation and the Richard D. and
Anne Marie Irwin Foundation.
Abstract: This paper conducts an empirical investigation of the relationship between information

asymmetry and earnings management predicted by Dye (1988) and Trueman and Titman (1988).

When information asymmetry is high, stakeholders do not have sufficient resources, incentives, or

access to relevant information to monitor manager’s actions, which gives rise to the practice of

earnings management (Schipper (1989) and Warfield et al. (1995)). Empirical results suggest a

systematic relationship between the magnitude of information asymmetry as measured by bid-ask

spreads and analyst forecast dispersion and the level of earnings management in both a broad

sample setting and in a time-specific setting around seasoned equity offerings.

Key Words: Earnings Management, Information Asymmetry, Monitoring, Seasoned Equity

Offerings

Data Availability: Data used in this paper are available from public sources.
I. Introduction
Accounting standards allow for managerial discretion in the application of accounting

methods (commonly referred to as “earnings management”). Growing anecdotal and systematic

evidence supports the argument that earnings management is a common practice in firms.1 A

fundamental question posed for accounting research is to identify the environmental conditions

surrounding the practice of earnings management. Moreover, managers possess private information

about the firm and its earnings stream that shareholders do not have (i.e. information asymmetry

between managers and shareholders). The purpose of this paper is to investigate the relationship

between information asymmetry and earnings management.

Analytical models have demonstrated that the existence of information asymmetry between

firm management and firm shareholders is a necessary condition for the practice of earnings

management (Trueman and Titman (1988) and Dye (1988)). However, there has been little, if any,

empirical work investigating this relationship. Schipper (1989) argues that this lack of empirical

testing of the information environment surrounding earnings management represents a slippage

between analytical models and empirical tests of earnings management. Schipper also suggests the

need for empirical work considering the environmental conditions surrounding the practice of

earnings management.

This paper hypothesizes that the magnitude of information asymmetry affects the magnitude

of earnings management practiced by firm managers. When information asymmetry is high,

stakeholders do not have the necessary information to “look through” the manipulated earnings.

Earnings management may also result when shareholders have insufficient resources, incentives, or

1
access to relevant information to monitor manager’s actions, which may give rise to the practice of

earnings management (Schipper (1989) and Warfield et al. (1995)).

Test results suggest a significant, positive relationship between measures of information

asymmetry (bid-ask spreads and analysts’ forecast dispersion) and earnings management after

controlling for other previously documented determinants of earnings management such as variability

in the underlying cash flows, size, risk, and leverage. The result is consistent with the notion that the

greater the information asymmetry between management and its shareholders, the more likely the

firm is to manage accruals and earnings. The possibility of endogeneity between information

asymmetry and earnings management is also considered in the design of the empirical tests.

Earnings management around seasoned equity offerings serves as a second setting to test the

relationship between information asymmetry and earnings management. Shivakumar (1996) and

Rangan (1995) provide evidence of earnings management around seasoned equity offerings. They

suggest that management faces the incentive to manage earnings upwards around a seasoned equity

offering to maximize the offer price for its shares of stock. I test whether the presence of

information asymmetry constrains the extent of earnings management in a setting where management

faces a strong incentive to manage earnings. Results suggest a systematic relationship between

information asymmetry and earnings management around seasoned equity offerings.

The remainder of this paper is organized in the following manner: Section two develops the

hypothesized relationship between information asymmetry and earnings management. Empirical

measures which proxy for information asymmetry and earnings management are developed in section

three. The sample data, research design and empirical results are described in the fourth section.

1
Healy (1985), Perry and Williams (1994), and DeFond and Jiambalvo (1994) are examples of empirical research
suggesting that earnings management is common among firms. Some anecdotal evidence comes from an American
Express attorney, “If you tell me that it’s improper under all circumstances for management to want to smooth out
their results, adjust the level of risk, or to smooth out reserves, or to move figures from one period to another. . . I’ll
tell you, you don’t understand the way American business is conducted.” (Fortune, June 25, 1984:58-61).

2
Empirical results from the seasoned equity offering sample are discussed in section five. The final

section includes the summary and conclusions of the relationship between information asymmetry

and earnings management.

II. Hypothesis Development

Setting for Earnings Management

Figure 1 is a graphical representation of the process of communicating information about a

firm’s underlying earnings process to the investor. Assume that the manager of the firm views the
~
earnings figure ( X it ), which is earnings measured before any incremental manipulation.2 Managers

then choose their level of earnings management (MAAit), subject to the different incentives and

constraints that they face, and report accounting earnings (Ait) to investors. The relationship
~
between earnings before incremental manipulation ( X it ) and the reported accounting earnings (Ait)

is:
~
A it = X it + MAA it (1)

where:

Ait = Accounting Earnings Reported to Investors by firm i at time t

MAAit = Managed Accounting Accrual by firm i at time t
~
X it = Earnings Before Incremental Manipulation

Insert Figure 1 here

At least two factors limit the extent of earnings management (MAAit) in the firm. First,

GAAP defines limits within which reported earnings can deviate from earnings before incremental

3
manipulation (Chaney and Lewis (1995)). GAAP requires that reported earnings be a function of the
~
earnings viewed by managers (Ait = f( X it )). GAAP also serves as a disciplining mechanism by

restricting managers from consistently over- or under-reporting earnings, since over the long term
~
∑( X it )=∑Ait and E(MAAit)=0.

A second factor that affects the level of earnings management (MAAit) is the extent of

information known by outsiders about firm performance (consider figure 1). Investors have access

to the amount of earnings reported by management (Ait), but investors also have access to other

publicly reported information about the firm. Financial analysts, labor unions, large shareholders,

industry trade journals and outside board members may all monitor and report information about the

profitability and status of the firm. To the extent that such outside sources are successful in

monitoring the profitability and performance of the firm, information asymmetry between

management and shareholders will be reduced. When information asymmetry is high, outside

stakeholders do not have the necessary information to monitor management or their accounting

choices.

In analytical models of earnings management, Dye (1988) and Trueman and Titman (1988),

rely on persistent information asymmetry as a condition for earnings management. For example, Dye

(1988) assumes an overlapping generation of owners. Selling shareholders instruct management to

follow a certain earnings management strategy to create a favorable impression on the buying group.

In this model, the manager knows something about earnings that shareholders do not. A condition

that must be met for earnings management to exist in such a model is that the information asymmetry

must persist throughout the transaction. Assumptions about proprietary costs of disclosure,

accounting rules and other institutional and contractual constraints suggest there is a restriction on

2
The earnings management examined in this paper is the measure of the period-to-period (or incremental)
manipulation of accruals. The paper does not deal with longer-term accounting policy choices (e.g. FIFO vs. LIFO)

4
total communication between management and shareholders. Information asymmetry does not

dissipate over time because a form of blocked communication cannot be eliminated by changing the

contractual arrangements (Schipper (1989)).3

High levels of information asymmetry between managers and shareholders is evidence of

shareholders lacking sufficient resources, incentives or access to relevant information to monitor

manager’s actions (Schipper (1989) and Warfield et al. (1995)). For example, management of firms

with debt contracts may have incentives to manage earnings over time to avoid debt covenant

violations. When information asymmetry is high, such firms may be able to manage earnings around

those debt contracts without being detected. Therefore, the greater the information asymmetry, the

more likely the firm will not be monitored as effectively as firms with less information asymmetry

between firm management and shareholders. Figure 2 gives an example of the relationship between

incentives, constraints and potential costs of earnings management.

Insert Figure 2 here

III. Measurement of Information Asymmetry and Earnings Management

Measuring Information Asymmetry

Bhattacharya and Spiegel (1991) suggest that information asymmetry causes an unwillingness

to trade and increases the cost of capital as investors “price protect” themselves against potential

losses from trading with better informed market participants. Lev (1988) argues that observable

measures of market liquidity be used to identify the perceived level of information asymmetry facing

that are made by management.

3
Blocked communication is defined to mean that managers cannot communicate all of their private information but
some communication is permitted.

5
participants in equity markets. Bid-ask spreads are one such measure of market liquidity that have

been used extensively in previous research as a measure of information asymmetry between

management and firm shareholders. As evidence of the ability of bid-ask spreads to capture the

information environment of the firm is provided by Healy, Palepu and Sweeney (1995) and Welker

(1995) who report evidence of a negative relationship between bid-ask spreads and the firm

disclosure policy.

Market microstructure theory suggests that one persistent adverse selection problem facing

market makers is the possibility that material firm-specific information has not been publicly

disclosed by the firm (Glosten and Milgrom (1985) and Welker (1995)). Withheld information may

be privately available to select traders who invest in costly information acquisition, creating an

ongoing adverse selection problem. When uncertainty about the occurrence of information events

exists and firms consistently provide incomplete disclosures with respect to such events (see figure

1), the market maker increases the bid-ask spread to offset the potential losses of trading with

informed traders with gains from trading with liquidity traders. Therefore, over the long time

horizon, when the possibility of material firm-specific information exists and that information is not

completely disclosed, bid-ask spreads proxy for the information asymmetry between the manager and

the shareholders. In this paper, the closing bid and ask quotes for the last trading day of June for

each year of the sample are used as a proxy for baseline market liquidity.

A second measure of information asymmetry between management and shareholders used in

the literature is the dispersion in analysts’ forecasts (e.g.,Healy, Palepu and Sweeney (1995)).

Brown and Han (1992) argue that as the amount of information asymmetry decreases, there is likely

to be a higher consensus among financial analysts about the future performance of the firm.

Therefore, a measure of the dispersion in analysts’ forecasts is used as a second measure of

6
information asymmetry in this paper. Using the I/B/E/S database of analysts’ forecasts, I estimate

the dispersion in analysts’ forecasts as follows:

Dispersion in analysts’ forecasts (DISP) = Std. deviation of analysts’ forecasts4

|Median forecast|

Measuring Earnings Management

Jones (1991) offers a model to help identify firms that manage earnings. The object of the

Jones model is to segregate expected (nondiscretionary) (E(ACi)) accruals from the managed

(discretionary) accruals (MAA). Measures of the managed accounting accrual used in this paper are

estimated using Dechow, Sloan and Sweeney’s (1995) suggested modification of the Jones (1991)

model. 5 This modified Jones model estimates the level of expected accruals as a function of the

difference between the change in revenues and the change in receivables and the level of property,

plant and equipment. Therefore, the model employed in this paper is expressed as follows:

E(AC)t= α0 + α1(∆REVt - ∆RECt) +α2(PPEt) (2)

where:

E(AC)t = expected normal accrual

∆REVt = net revenues in year t less net revenues in year t-1
∆RECt = net receivables in year t less net receivables in year t-1
PPEt = property plant and equipment at time t

All variables are deflated by the beginning-of-period total assets. Both time-series and cross-

sectional methods are used to estimate two alternative measures of earnings management.

4
Deflating the standard deviation of analysts’ forecasts by year t-1 price or by actual earnings causes no changes in
inferences throughout the paper.
5
Dechow, Sloan and Sweeney (1995) test various models proposed in the literature and find the modified version of
the Jones model to provide the most powerful tests for detecting earnings management. Guay, Kothari and Watts
(1996) also find some support for the validity of the Jones and modified Jones models over the other proposed models.
Hunt, et al. (1995) provide additional support for the modified Jones model’s ability to isolate discretionary accruals.

7
 The first measure of earnings management, |MAA1i|, is estimated using a time-series

approach. 6 Estimates of α0, α1, α2 and α3 and expected normal accruals are made from 1976-1987.

These parameters are then used to predict the level of expected accruals , E(AC). Consistent with

previous studies of earnings management (Healy (1985) and Jones (1991)), the accounting accrual

(ACt) is computed as:

ACt = (∆CAt-∆CLt-∆Casht+∆STDt-Dept) (3)

where:

∆CAt = change in current assets

∆CLt = change in current liabilities
∆Casht= change in cash and cash equivalents
∆STDt = change in debt included in current liabilities
Dept = depreciation and amortization expense

All variables are deflated by the beginning-of-period total assets.

Since the hypothesis does not rely on the direction of the managed accrual, but rather on the

magnitude of the accrual adjustments, test statistics are based on the absolute value of the managed

accrual |MAAi|. Specifically, |MAA1i| is estimated in the following fashion:

|MAA1i| = |ACi-E(AC1i)| (4)

A benefit to estimating expected accruals using the time-series approach is that, presumably,

the accruals generating process is captured. However, this firm-specific approach requires a long

time-series of data and assumes that the accruals generating process is constant over time. This

assumption may not hold in periods of changing economic conditions. As an alternative to the time-

series estimation, a cross-sectional approach is used which compares the expected level of accruals

for the firms in our sample with others in the same 2-digit SIC code each year of the test period.7 A

6
This time series approach is used in Jones (1991), DeChow, et al. (1995), Hunt, et al. (1995), and Guay, et al. (1996)
7
A similar cross-sectional approach is implemented in DeFond and Jiambalvo (1994).

8
cross-sectional regression employing model (1) is estimated over all firms in the same industry and

then is used to predict the level of expected accruals (E(AC2)) for the test firms. The absolute value

of the level of managed accruals, |MAA2i|, is estimated as follows:

|MAA2i| = |ACi-E(AC2i)| (5)

AC is defined in the same manner as in equation (4).

Since both the time-series and the cross-sectional models have advantages and disadvantages,

robustness across models increases confidence in the results obtained. |MAA1i| and |MAA2i| are

thus measures of the extent to which accruals and earnings are managed each year.
~
Depending on the realization of X it (earnings before incremental manipulation) each period,

management will determine the sign and the magnitude of the needed manipulation (see example in
~
figure 2.2). In some periods, the realized X it may be exactly the amount that management needs to

~
meet its objectives. In other periods, X it may be such that a significant level of earnings

management is needed for management to meet its goals. Therefore, a prediction of any single year

may not capture the level of earnings management practiced by the firm. This study attempts to

capture the long-run propensity of firms to manage earnings. For this reason, mean values for each

firm-variable are computed over the test period and used in this study.

IV. Sample Data, Research Design and Empirical Results

Sample Data

The bid-ask spread sample of firms is identified from the Fitch Data sheets and the NYSE

Daily Sales Report which provide the closing bid-ask spreads for all NYSE firms on the last day of

June for the years 1988-1992. For simplicity and homogeneity, I require that all firms have calendar

year-ends. For the time-series estimation of the modified Jones model, firms must have the full 17

9
years (1976-1992) of Compustat data necessary to estimate the parameters of the modified Jones

model. In addition, both models require Compustat information over the 1988-92 sample period.

The sample firms are also required to have price and returns information available on CRSP. After

the data constraints are applied, the sample includes 355 (641) firms over the period 1988-1992 to

measure |MAA1i| (|MAA2i|).

The |MAA1i| and |MAA2i| samples represent samples from the time-series approach and from

the cross-sectional approach, respectively. The |MAA1i| and |MAA2i| samples have firms from a

wide cross-section of industries (41 and 46 industries respectively). In both samples the electric and

gas utilities (84 and 95 firms), chemicals and allied products (35 and 51 firms), and electrical

equipment industries (25 and 41 firms) are the industries with highest representation. Due to the use

of NYSE bid-ask information and the stringent data requirements, the average total assets for both

samples are above the 75th percentile of all Compustat observations. Therefore, the results presented

here should be interpreted as applying to large firms.

Insert Table 1

Some descriptive statistics and a correlation table are provided in panel A of table 1. The

average bid-ask spread, BIDASKi, is approximately 1.1% of share price and ranges between 0.37%

and 7.17%. Welker (1995) reports a comparable average and range in his sample of bid-ask spreads.

It is also interesting to note that the correlation between |MAA1i| and |MAA2i| is 0.46 and is

significant at a 0.01 level of significance. This suggests that both the cross-sectional and time-series

estimation of the level of abnormal accruals are both capturing the same phenomena. Other variables

appearing in panel A of table 1 are introduced and described in section V.

10
Dispersion in Analysts’ Forecasts (DISP) Sample

The dispersion in analysts’ forecasts sample of firms is identified from the I/B/E/S tapes. To

be included in this sample, each firm must have analysts’ forecasts from at least five analysts during

the month of June. Data constraints are similar to those imposed on the bid-ask spread sample. The

sample firms are also required to have price and returns information available on CRSP. After data

constraints are applied, the sample includes 422 (539) firms over the period 1988-1994 to measure

|MAA1i| (|MAA2i|).

The |MAA1i| and |MAA2i| samples represent firms from a wide cross-section of industries

(48 and 53 industries, respectively). Similar to the bid-ask sample, the utility, chemicals and allied

products, and electrical equipment industries are the industries with highest representation. Also,

due to the requirement of at least five analysts providing earnings forecasts and the stringent data

requirements, the average total assets for both samples is above the 70th percentile of all Compustat

observations. Therefore, the results presented here should be interpreted as applying to large firms.

Some descriptive statistics and a correlation table are provided in panel B of table 1. The

average dispersion in analysts’ forecasts, DISPi, is 9.74% and ranges from 0.77% to 49.42%. Similar

to the bid-ask spread sample, it is also interesting to note that the correlation between |MAA1i| and

|MAA2i| is 0.46 and is significant at a 0.01 level of significance. This suggests that both the cross-

sectional and time-series estimation of the level of abnormal accruals are both capturing the same

phenomena. Other descriptive statistics and a correlation table appear in panel B of table 1.

11
 For both the bid-ask sample and the dispersion in analysts’ forecasts samples, I trimmed all

observations below the 1st and above the 99th percentile of observations to control for the possible

influence of extreme observations.8

The correlations presented in panel A of table 1 indicate a correlation between bid-ask

spreads and the earnings management measures |MAA1i| and |MAA2i| of 0.293 and 0.262,

respectively. Panel B of table 1 shows the correlation between the dispersion in analysts forecasts

(DISPi) and the earnings management measures |MAA1i| and |MAA2i| of 0.133 and 0.174,

respectively. Each of these correlations between the measures of earnings management and the

measures of information asymmetry are significantly greater than zero at a 0.01 level of significance.

Research Design

Obviously, information asymmetry between firm managers and shareholders is not the sole

determinant of managers’ accounting choices. Figure 2 provides a graphical depiction of the

incentives, constraints and costs of managing earnings. Two incentives for managing earnings are to

reduce political costs and avoid violating debt covenants. Past research suggests that political

process theory has implications for the determination of accounting choices (Watts and Zimmerman

(1978)). For example, Zmijewski and Hagerman (1981) suggest that political costs increase with

firm size and with firm risk. Managers of large and/or high-risk firms, therefore, have greater

incentives to exploit the latitude in accounting to reduce these political costs. Firm size, the firm’s

market-to-book ratio, sales growth, and the volatility of underlying cash flows are used as proxies for

a firm’s size and risk. Moreover, Sweeney (1994) shows evidence that managers are more likely to

8
Influential observations excluded using Belsey, Kuh and Welsch (1980) criteria (not reported here) had no affect on
the inferences made.

12
exploit the latitude in accounting the closer the firm is to binding debt covenants. Therefore, a firm’s

debt-to-equity ratio is used to proxy for the firm’s proximity to binding debt covenants. Given past

research, I hypothesize that earnings management is increasing with respect to the level of firm

leverage, risk, and the volatility of underlying cash flows. Besides the positive theory implications of

firm size, the size of the firm may also be capturing a firm’s information environment and could thus

have a negative relationship with the level of managed accruals. Therefore, I do not predict the sign

of the relationship between firm size and the level of earnings management.

Using bid-ask spreads and the dispersion in analysts’ forecasts as measures of the information

asymmetry between the firm and its shareholders, I propose the following model to investigate the

relationship between earnings management and information asymmetry:

|MAAXi| = α0 + α1BIDASKi +α2 CFVARi +α3 DEBTi + α4MKTBVi + α5 SIZEi + α6 GROWTHi + εi (7)

|MAAXi| = α0 + α1 DISPi + α2 CFVARi + α3 DEBTi + α4MKTBVi + α5 SIZEi + α6 GROWTHi + εi (7’)

where:

|MAA1i| = the mean Managed Accounting Accrual under the modified Jones (1991) model
using the time-series estimation approach (see derivation in section III)
|MAA2i| = the mean Managed Accounting Accrual under the modified Jones (1991) model
using the cross-sectional estimation approach (see derivation in section III)
BIDASKi = the mean bid-ask spread at the close of trade on the last trading day of June, scaled
by the average of the bid and ask prices
DISPi = the mean standard deviation in analysts’ forecasts during the month of June, scaled
by the median forecast of year-ahead earnings for firm i over the test period
CFVARi = the standard deviation of operating cash flows over the test period divided by the
average operating cash flows over the test period
DEBTi = the mean of the long-term debt divided by the book value of equity for firm i over
the test period
MKTBVi = the mean market capitalization divided by the book value of equity for firm i over
the test period
SIZEi = Natural log of the mean market capitalization for firm i over the test period
GROWTHi = Net revenues at the end of the test period less net revenues at the beginning of the
test period scaled by net revenues at the beginning of the test period

13
 The test of the hypothesis will be tested by estimating α1. An α1 that is significantly greater

than zero would provide evidence of a positive relationship between the level of information

asymmetry and the level of earnings management.

Endogeneity

The hypothesis being tested assumes that the level of information asymmetry is a constraint

on the firm’s ability to practice earnings management. However, past research (Schipper (1989) and

Kasznik (1995)) suggests that simultaneity may exist in the determination of earnings management

and information asymmetry. For example, if the practice of earnings management (e.g. income

smoothing) provides an informative disclosure to the firm’s investors, it may reduce the information

asymmetry among the investors. Therefore, an issue which must be addressed is the possibility of

endogeneity between information asymmetry and earnings management.

An approach to establishing the direction of causality is the use of simultaneous equations. A

simultaneous equations model (two-stage least squares (2SLS)) is used to investigate whether bid-

ask spreads and earnings management are jointly determined by a set of exogenous variables. The

benefit of a simultaneous equations approach is that it potentially controls for the bias in the OLS

estimates when the error term is correlated with the independent variables, as is the case when a truly

endogenous variable is treated as an exogenous (independent) variable.

To implement this model, a first stage regression of bid-ask spreads is run on a set of

exogenous, instrumental variables which have been shown in prior research as being explanatory

determinants of bid-ask spreads. The source of instruments for the bid-ask spread is taken from

Welker (1995). These instruments include share price, average daily dollar trading volume, and the

standard deviation of daily market adjusted returns. Welker (1995) finds a non-linear relationship

between bid-ask spreads and price. To help compensate for this non-linear relationship between bid-

14
ask spreads and price, Welker suggests the use of an additional variable LOWPRICEi as a piece-wise

linear approach to compensate for the non-linearity in prices below $12.50. These determinants of

bid-ask spreads are tested using the following regression model:

BIDASKi = γ0 + γ1 PRICEi + γ2LOWPRICEi + γ3STDRETi + γ4VOLUMEi + δi (8)

where

BIDASKi= the mean bid-ask spread at the close of trade on the last trading day of June, scaled by
the average of the bid and ask prices over the test period
PRICEi = the mean of the bid and ask quotes at the close of trade on the last trading day of
June for firm i over the test period
LOWPRICEi =LOW*(PRICE-12.5), where LOW is 1 if PRICEi is less than $12.50, and zero
otherwise, for firm i over the test period
STDRETi = the standard deviation of daily returns for firm i over the test period
VOLUMEi = the mean dollar daily trading volume for firm i over the test period

Table 2 gives an idea of how good the instrumental variable is for the estimation of the bid-

ask spread. The expected signs for each of the independent variables are as predicted and the model

fit appears to be high (as indicated by the adjusted R2 of 78.2% and 79.3%).

Insert Table 2

A similar approach is used to estimate the dispersion in analysts’ forecasts. Varian (1985) and

Ajinka et al. (1991) find some evidence that the volume of trading is positively correlated with the

dispersion in analysts’ forecasts. In this paper, share turnover, TURNOVERi, is used as a proxy for

the volume of trade. The standard deviation of returns, STDRETi, is included in an attempt to

capture the disagreement or dispersion in beliefs of the market. Earnings variability, VAREARNi,

and earnings growth, EGROWTHi, are used to measure the underlying volatility of the earnings

series that the analysts are attempting to model and to make predictions of future earnings levels. As

a control variable, the number of analysts who make each forecast is included as a predictor of

dispersion in forecasts. Given prior research, it is expected that share turnover, standard deviation of

15
analysts’ forecasts, earnings variability, and earnings growth will be positively correlated with the

dispersion in analysts’ forecasts. Prior research gives no direction on the sign of the number of

analysts’ forecasts. Therefore, the following model is used to predict the dispersion in analysts’

forecasts:

DISPi=γ0+γ1NUMANALi+γ2VAREARNi+γ3STDRETi +γ4TURNOVERi+γ5EGROWTHi+δi (9)

DISPi = the mean standard deviation in analysts’ forecasts during the month of June, scaled
by the median forecast of year-ahead earnings for firm i over the test period
NUMANALi = the mean number of analysts making forecasts during the month of June over the
test period
VAREARNi = standard deviation of annual earnings over the test period scaled by earnings for
firm i
STDRETi = the mean standard deviation of daily returns for firm i over the test period
TURNOVERi = the mean dollar daily trading volume divided by the mean number of shares
outstanding for firm i over the test period
EGROWTHi = the net earnings at the beginning of the test period less net earnings at the end of the
test period scaled by net earnings at the beginning of the test period

Table 3 gives an idea of how good the instrumental variable is for the estimation of the

dispersion in analysts’ forecasts. The expected signs for each of the independent variables are as

predicted (except for the EGROWTHi variable) and the model fit appears to be reasonable as

indicated by the adjusted R2 of 30.3% and 25.7% for |MAA1i| and |MAA2i|, respectively.

Insert Table 3

Estimation Results

Table 4 shows the results of the 2SLS regression of model 7 under both methods of

estimating the level of earnings management (|MAA1i| and |MAA2i|) examined in this paper for the

bid-ask spread sample. When |MAA1i| is the dependent variable in panel A, the coefficients on the

16
control variables CFVARi and DEBTi take on the expected sign and are significantly greater than

zero. The control variables MKTBVi, SIZEi, and GROWTHi have coefficients which are not

significantly different from zero. The coefficient on the bid-ask spreads is 1.625 with a t-statistic of

4.135, which is significantly greater than zero at a 0.01 level of significance. The overall fit of the

regression equation is demonstrated by the adjusted R2, which is reasonable at 17.7%. The overall fit

of each of these regression models of managed accruals (as measured by adjusted R2) compares

favorably to those of Warfield, et al. (1995) of between 8.34% to 12.48%.

Insert Table 4

When |MAA2i| is the dependent variable (see panel B), the coefficients of the control

variables CFVARi and MKTBVi take on the expected sign and are significantly greater than zero.

The control variables DEBTi, SIZEi, and GROWTHi take on a coefficient which is not significantly

different from zero. The coefficient on bid-ask spreads is 0.600 with a t-statistic of 4.168, which is

significantly greater than zero at a 0.01 level of significance. The overall fit of the regression

equation is reasonable with an adjusted R2 of 14.0%. After controlling for other determinants of

earnings management, the results indicate that bid-ask spreads are systematically associated with the

level of earnings management practiced by managers.

Table 5 shows the results of the 2SLS regression of model 7’ for both |MAA1i| and |MAA2i|

examined in this paper using the dispersion in analysts’ forecast sample. In both samples, the control

variables CFVARi and GROWTHi take on the expected sign and are significantly greater than zero.

The coefficient on DISPi (dispersion in analysts’ forecasts) is 0.133 (0.164) with a t-statistic of 4.781

(4.689), which is significantly greater than zero at a 0.01 level of significance in the estimation of

|MAA1i| (|MAA2i|). The overall fit of the regression equation in explaining the variation in |MAA1i|

17
and |MAA2i| is indicated by an adjusted R2 of 22.1% and 14.6%, respectively. It should also be

noted that OLS estimation (instead of 2SLS estimation) causes no change in inference.

Insert Table 5

Therefore, under two different measures of earnings management (|MAA1i| and |MAA2i|) and

two different measures of information asymmetry, BIDASKi and DISPi, there is evidence of a

systematic relationship between earnings management and information asymmetry.

The analysis was also conducted without the regulated firms (firms with SIC codes between

4000 and 4999 or between 6000 and 6299) since Warfield et al. (1995) suggest that regulated firms

may be subject to different incentives than non-regulated firms. The analysis (not reported here)

reveals similar inferences regarding information asymmetry and earnings management for the non-

regulated firms. In addition, to control for different levels of earnings management across different

industries, industry dummies were included in the model. The inclusion of industry dummies (not

reported here) caused no changes in inferences regarding information asymmetry and earnings

management.

V. Tests of the Relationship Between Information Asymmetry and

Earnings Management in a Specific Setting: Earnings Management
Around Seasoned Equity Offerings

The testing performed using the broad sample suggests that in the presence of a high level of

information asymmetry, managers will manipulate earnings. As a test of robustness, the relationship

18
between information asymmetry and earnings management is tested around the period of a seasoned

equity offering.

Shivakumar (1996) argues that the need to raise capital at the lowest cost provides managers

with incentives to manage earnings upwards before an equity offering. Likewise, managers have

incentives to manage earnings after an equity offering due to the prospect of litigation if the

managers fail to meet favorable forecasts made before the offering. Shivakumar (1996) and Rangan

(1995) find evidence of income-increasing accruals management for the two quarters before, the

quarter of, and three quarters following a seasoned equity offering. Therefore, the time period

surrounding a seasoned equity offering appears to be a reasonable setting for testing the relationship

between information asymmetry and earnings management.

In a seasoned equity offering setting, potential shareholders are keenly interested in knowing

the degree of earnings management performed by management before purchasing stock. When

information asymmetry is high, shareholders lack the necessary information to monitor manager’s

actions. In the presence of high information asymmetry between management and shareholders,

management is able to manage earnings to a higher degree than in the presence of low information

asymmetry. Therefore, I hypothesize a positive relationship between the level of information

asymmetry and the extent of earnings management in a seasoned equity offering setting.

Research Design

The first test performed is to replicate the results of Shivakumar (1996) and Rangan (1995).

Using a cross-sectional estimation approach as defined in section III, the managed abnormal accruals,

MAAi, are identified by comparing the accruals of the firms making seasoned equity offerings with

the accruals of all other firms in the same industry in the fiscal year of the offering. If the accruals of

the firms making seasoned equity offerings are, on average, more income increasing than the accruals

19
of all other firms in the same industry, then it would suggest that firms have managed earnings

upwards in the fiscal year of the seasoned equity offering.

I then test to see if information asymmetry affected the level of earnings management in the

following model:

MAA2i = α0 + α1BIDASKi + α2VARi + α3DEBTi + α4MKTBVi + α5SIZEi +α6OFFSIZEi

+α7OWNi + α8GROWTHi + εi (10)

where:

MAA2i =the Managed Accounting Accrual under the modified Jones (1991) model using the
cross-sectional estimation approach in the year of the offering (see derivation in
section III)
BIDASKi =the bid-ask spread at the close of trade on the last trading day of June for the year
before the seasoned equity offering, scaled by the average of the bid and ask prices
VARi =the standard deviation of earnings divided by the average operating cash flows over
the period five years before the offering until the year before the offering
DEBTi =long-term debt divided by total assets (both from year t-1)
MKTBVi =the market capitalization divided by the book value of equity for firm i in year t-1
SIZEi =Natural log of the market capitalization for firm i in year t-1
OFFSIZEi =Relative size of seasoned equity offering defined as of gross offer proceeds divided
by the market capitalization in year t-1
OWNi =Stock ownership of the offering firm’s directors and officers as a percentage of the
total outstanding shares
GROWTHi =Net revenues at the end of the test period less net revenues at the beginning of the
test period scaled by net revenues at the beginning of the test period

The model predicting earnings management around a seasoned equity offering is derived from

equation 7 above as well as a model used by Shivakumar (1996) to analyze earnings management.

The first modification to equation 7 is that since we would expect income-increasing accruals in the

period around a seasoned equity offering, MAA2i is used as a measure of earnings management (as

opposed to the absolute value of managed accounting accruals, |MAA2i|). In addition, Shivakumar

suggests that when the stock owned by management (OWNi) is high, management has greater

incentives to maximize the proceeds from the offering through earnings management.

A third modification to equation 7 is the use of offer size as a determinant of earnings

management. Firms which expect to make a large offering relative to firm size, OFFSIZEi, stand to

20
gain more from overstating earnings than firms which expect relatively smaller offerings. VARi is the

variability in earnings over the past five years as a measure of the propensity of firms to

opportunistically manage earnings. On the one hand, the higher the level of earnings management

that was done in the past may adversely affect the level of earnings management that can be

performed in the future since it may have used up its earnings management flexibility. On the other

hand, it may simply suggest that managers have great flexibility in managing earnings and will

continue to manage earnings in the future. Therefore, I cannot predict the sign of the relationship

between VARi and the level of earnings management MAA2i.

Under the pecking order theory of corporate financing (Myers (1993)), firms will first use

internally-generated funds and then receive debt financing. These firms will use equity financing only

as a last resort. Myers (1993) argues that an equity issue becomes feasible in the pecking order only

when leverage is already high enough to make additional debt materially expensive, e.g. because of

the threat of costs of financial distress. Therefore, it may be the case that the firms that seek equity

financing may be firms that are highly leveraged and close to debt covenant violations. Shivakumar

suggests that these highly levered firms may have overstated earnings in the quarters before the

decision to issue equity was made to avoid debt covenant violations. This may restrict their ability to

manage earnings around a seasoned equity offering. Therefore, a negative relationship is predicted

between leverage, DEBTi, and the level of earnings management. As theorized above, the market-

to-book ratio, MKTBVi, sales growth, GROWTHi, and SIZEi are motivated by the presence of

political and contracting costs; firms of larger size and risk are expected to manage earnings more.

However, to the extent that SIZEi proxies for the information environment facing the firm, a negative

sign is predicted. Therefore, no sign is predicted for the relationship between SIZEi and the level of

managed accruals, MAA2i.

21
 Similar to the testing performed in the broad sample, the bid-ask spread, BIDASKi, is used as

a proxy of the information asymmetry between management and shareholders. The bid-ask spread is

measured on the last June 30 date before the equity offering takes place. The test of the hypothesis

will be testing the sign of α1 in a regression of equation 4.1. A positive sign of α1 is consistent with a

systematic relationship between information asymmetry and earnings management.

Data

The sample of seasoned equity offerings during the period June 30, 1986 to June 30, 1993

was obtained from the Securities Data Corporation. This sample includes all registered firm-

commitment offerings of stock made by NYSE firms in that period.9 10 In addition, the firm had to

have the necessary Compustat information available as well as at least ten other firms in the same 2-

digit SIC code with the necessary data.

Some summary statistics describing the remaining sample of 150 firms are included in table 6.

Panel A shows the distribution over the sample period of the seasoned equity offerings. As would be

expected, there is a decrease in the number of seasoned equity offerings around the October 1987

stock market crash. Panel B of table 6 details that the mean market value of equity of the offering

firms is $829.3 million, which is significantly larger in size than Shivakumar (1996). This is primarily

due to the requirement of bid-ask spreads from NYSE firms. Therefore, the results obtained should

9
Firms that are listed on NASDAQ are not included in the sample since there is some evidence of collusion among
market makers in setting the bid and offer prices (Christie and Schultz (1994)).
10
Following Rangan (1995), only offerings that met the following criteria were retained:
1. The offering was not made in combination with any other securities (e.g. debt, preferred stock) of the firm.
2. The offering was not an offering of common shares with warrants attached.
3. The firm offering shares had the necessary SEC proxy statements available on the Corporate Text database in the
year of the offering.
4. The offering was not made consequent to a shelf registration.
5. If there was more than one offering during the sample period, only the first offering was used in this sample.
6. The firm was not a financial services or utility firm.

22
only be generalized to large firms. Also, Panel B details that the mean offering in this sample is

$82.5 million, or 23.4% of the pre-offering market value of equity.

Insert Table 6

Results

The first issue is to determine if earnings management was performed in the period

surrounding the seasoned equity offering. Using a cross-sectional Jones model (MAA2) as derived

in section III, the level of accruals of the offering firm is compared with the level of expected

accruals. The level of expected accruals is derived from other firms in the same industry (2-digit SIC

code). After this test is performed, I find evidence that income increasing accruals were used in the

fiscal year of the seasoned equity offering ( MAA2 =0.011, p<.05). That is, income increasing

accruals were, on average, 1.1% of year t-1 assets for the offering firms. This gives additional

evidence of earnings management around seasoned equity offerings and offers support for the

findings of earnings management in this setting by Shivakumar (1996) and Rangan (1995).

Insert Table 7

Table 7 provides the results of a regression estimating the level of income-increasing accruals

used by the firm to opportunistically manage earnings.11 The model explains approximately 11.1% of

the variation in managed accruals. The model fit is reasonable compared to a similar model

attempting to explain the variation in earnings management around seasoned equity offerings by

Shivakumar (1996) with a model fit of between 9.89% and 11.89%. The significant coefficient on

the BIDASKi variable provides evidence of a positive relationship between the extent of information

23
asymmetry and the level of earnings management conducted by the firm. The control variables,

earnings variability, the debt-equity ratio, and earnings growth are also statistically significant

explanatory variables in the estimation of the level of earnings management. Therefore, this second

test lends credence to the hypothesized relationship between information asymmetry and earnings

management and particularly considers the relationship in a setting where management has the

incentive to manage earnings opportunistically.

VI. Summary and Conclusions

This paper develops and tests hypotheses of how the presence of information asymmetry

affects management incentives to manage earnings. This is a direct test of Dye (1988) and Trueman

and Titman (1988) suggesting that the presence of information asymmetry is a necessary condition

for earnings management. I extend that argument by suggesting that the level of earnings

management increases as the level of information asymmetry increases. When information asymmetry

is high, stakeholders may not have the necessary information to undo the manipulated earnings.

Another possible explanation is that the existence of firms with high levels of information asymmetry

is evidence of shareholders without sufficient resources, incentives, or access to relevant information

to monitor manager’s actions, which may give rise to the practice of earnings management.

The hypothesis is tested in both a general and a time-specific setting. Using a broad sample

of firms, multivariate tests are used to test the relationship between information asymmetry and

earnings management. Tests of the hypothesis provide evidence of the predicted positive relationship

between information asymmetry and earnings management using two different measures of earnings

management and two different measures of information asymmetry. In addition, the relationship

11
To control for outliers, all observations below the 1st and above the 99th percentile of observations were trimmed to
control the possible influence of extreme observations. Separately, influential observations were identified and

24
between information asymmetry and earnings management is tested around a seasoned equity

offering. This allows a test of the effect of information asymmetry on earnings management in a

setting where significant incentives exist for managers to opportunistically manipulate earnings.

After replicating the analysis performed by Shivakumar (1996) and Rangan (1995) and finding

evidence of earnings management around seasoned equity offerings, I analyze the extent of earnings

management and the extent of information asymmetry. I find a statistically significant relationship

between the extent of income-increasing accruals manipulation and the level of information

asymmetry.

Evidence from this paper suggests that information known about the firm and firm earnings

may limit the extent of earnings management performed by firm managers. There may also be

outside monitors who curtail management action and management’s accounting choices. Evidence

of such monitoring within a particular firm may be the proportion and strength of outside members of

the board of directors, the strength of the audit committee, focused firm stakeholders (e.g., labor

unions, firm suppliers, etc.), and shareholders that hold a large proportion of the company shares.

Research which helps us understand the information and monitoring environments faced by firm

management will enhance our understanding of why and how management makes accounting

choices.

excluded using Belsey, Kuh and Welsch (1980) criteria (not reported here). There is no effect on the inferences made.

25
Figure 1- Simplified Setting for Earnings Management: Information Flows to Investors

Industry Reports, Other Information

General Economic Intermediaries
Reports, and Labor
Reports

Earnings Before
Incremental Analyst Calls, Meetings Current and
~
Manipulation ( Χ with Management, Potential Investors
it )
Voluntary Disclosures,
etc.

~
Managers view ( Χ it )
and make accounting Reported
choice (MAA); Earnings, Ait
subject to GAAP and
information known by
shareholders

26
Figure 2 – The Effect of Incentives, Constraints and Costs on the Level of
Earnings Management (adapted from Jiambalvo(1996))

Incentives to Manage
Earnings
Avoid Violating Debt
Covenants

Maximize Management
Bonus

Maximize Share Price

(Initial Public Offering and
Seasoned Equity Offering)

Reduce Political Costs

Constraints on Earnings Earnings

Management Management Choice

Extent of Information Level of Managed

Asymmetry Accounting Accrual,
MAAit
Generally Accepted
Accounting Principles
(Auditor)

Other Factors

Potential Costs of
Earnings Management

Legal Liability

SEC Sanctions

Dispute with Auditor

27
 TABLE 1- Descriptive Statistics of the Bid-Ask Spread and Dispersion in
Analyst Forecast Samples
Panel A: Bid-Ask Spread Sample - Descriptive Statistics

Variable Mean Median Maximum Minimum Std. Dev.

|MAA1i| 0.0575 0.0396 0.3340 0.0037 0.051

|MAA2i| 0.0442 0.0348 0.1737 0.0071 0.033

BIDASKi 0.0110 0.0081 0.0717 0.0024 0.014

CFVARi 0.7223 0.5536 0.0156 6.1962 0.759

DEBTi 0.9720 0.6860 14.1010 0.0160 1.530

MKTBVi 2.0349 1.5674 13.695 0.4687 1.507

SIZEi 7.0846 7.1644 11.113 2.0731 1.603

GROWTHi 0.2160 0.1073 -1.0103 3.4289 0.470

Correlation Table

|MAA1i| |MAA2i| BIDASKi CFVARi DEBTi MKTBVi SIZEi GROWTHi

|MAA1i| 0.541 0.236 0.472 0.001 -0.028 -0.285 -0.075

|MAA2i| 0.467 0.235 0.407 0.005 0.080 -0.235 0.015

BIDASKi 0.262 0.293 0.322 0.214 -0.362 -0.802 -0.214

CFVARi 0.257 0.199 0.209 0.001 -0.161 -0.340 -0.112

DEBTi 0.127 0.141 0.121 0.209 -0.192 -0.152 -0.191

MKTBVi -0.001 0.099 -0.205 -0.115 0.169 0.408 0.461

SIZEi -0.283 -0.244 -0.597 -0.240 -0.114 0.363 0.216

GROWTHi 0.099 0.024 -0.105 -0.184 -0.058 0.360 0.363

NOTE: All of the descriptive statistics above reflect the |MAA1i| sample (with 355 observations) except for the statistics
for the variable, |MAA2i|. All correlation coefficients with an absolute value above 0.10 or greater are significant at a
0.05 level of significance (two-tailed). The Pearson correlation coefficients appear in the lower diagonal; the Spearman
rank correlation coefficients appears in the upper diagonal.

28
 TABLE 1- Descriptive Statistics (cont.)

Panel B: Dispersion in Analysts’ Forecasts Sample

Descriptive Statistics

Variable Mean Median Maximum Minimum Std. Dev.

|MAA1i| 0.0354 0.0291 0.1481 0.0034 0.032

|MAA2i| 0.0377 0.0305 0.1433 0.0001 0.034

DISPi 0.0974 0.0561 0.4942 0.0077 0.097

CFVARi 0.4124 0.2932 5.7266 0.0243 0.587

DEBTi 0.8510 0.6780 6.5070 0.0110 0.843

MKTBVi 2.2228 1.7046 12.934 0.1667 1.629

SIZEi 7.2652 7.2431 11.154 3.2626 1.413

GROWTHi 0.4613 0.2570 4.9003 -1.1363 0.658

Correlation Table

|MAA1i| |MAA2i| DISPi CFVARi DEBTi MKTBVi SIZEi GROWTHi

|MAA1i| 0.540 0.175 0.531 -0.279 0.104 -0.254 0.249

|MAA2i| 0.461 0.228 0.496 -0.210 0.168 -0.219 0.336

DISPi 0.133 0.174 0.399 0.200 -0.358 -0.252 -0.111

CFVARi 0.357 0.230 0.195 -0.007 -0.054 -0.454 0.281

DEBTi -0.106 -0.052 0.252 0.046 -0.241 -0.049 -0.236

MKTBVi 0.003 0.071 -0.092 -0.031 0.301 0.322 0.356

SIZEi -0.242 -0.218 -0.126 -0.250 -0.009 0.310 -0.188

GROWTHi 0.264 0.335 -0.167 0.039 -0.116 0.214 -0.022

NOTE: All of the descriptive statistics above reflect the |MAA1i| sample (with 422 observations) except for the statistics
for the variable, |MAA2i|. All correlation coefficients with an absolute value above 0.10 or greater are significant at a
0.05 level of significance (two-tailed).

The Pearson correlation coefficients appear in the lower diagonal; the Spearman rank correlation coefficients appears in
the upper diagonal.

29
 TABLE 1- Descriptive Statistics (cont.)

VARIABLE DEFINITIONS
MAA1i = the mean Managed Accounting Accrual under the modified Jones (1991) model using the time-series
estimation approach (see derivation in section III in the text)
MAA2i = the mean Managed Accounting Accrual under the modified Jones (1991) model using the cross-sectional
estimation approach (see derivation in section III in the text)
BIDASKi = the mean bid-ask spread at the close of trade on the last trading day of June, scaled by the average of the
bid and ask prices
DISPi = the mean standard deviation in analysts’ forecasts during the month of June, scaled by the median
forecast of year-ahead earnings for firm i over the test period
CFVARi = the standard deviation of operating cash flows over the test period divided by the average operating cash
flows over the test period
DEBTi = the mean of the long-term debt divided by the book value of equity for firm i over the test period
MKTBVi = the mean market capitalization divided by the book value of equity for firm i over the test period
SIZEi = Natural log of the mean market capitalization for firm i over the test period
GROWTHi= Net revenues at the end of the test period less net revenues at the beginning of the test period scaled by net
revenues at the beginning of the test period

30
 TABLE 2- OLS Estimation of Bid-Ask Spreads

BIDASKi = γ0 + γ1 PRICEi + γ2 LOWPRICEi + γ3 STDRETi + γ4 VOLUMEi + δi (8)

Panel A: Estimation of Bid-Ask Spreads for |MAA1i| Sample

γ0 γ1 γ2 γ3 γ4 R2 N
Predicted Sign - - + -
Coef. 0.0046 -0.00005 -0.0037 0.378 -6.8E-09 79.3% 355
(t-statistic) (4.817)** (-5.545)** (-16.757)** (8.423)** (-5.545)**

Panel B: Estimation of Bid-Ask Spreads for |MAA2i| Sample

γ0 γ1 γ2 γ3 γ4 R2 N
Predicted Sign - - + -
Coef. 0.0024 -0.00009 -0.0037 0.562 -9.9E-09 78.2% 641
(t-statistic) (1.795)** (-4.325)** (-16.355)** (10.773)** (-5.431)**

Notes: The listed R2 are the adjusted R2. The t-statistics are corrected for heteroskedasticity using White’s consistent
estimator for standard error.

**Significant at the .01 level. * Significant at the .10 level (One-sided tests)

where:

BIDASKi =the mean bid-ask spread at the close of trade on the last trading day of June, scaled by the average of
the bid and ask prices over the test period
PRICEi = the mean of the bid and ask quotes at the close of trade on the last trading day of June for firm i
over the test period
LOWPRICEi = LOW*(PRICE-12.5), where LOW is 1 if PRICEi is less than $12.50, and zero otherwise, for firm i
over the test period
STDRETi = the standard deviation of daily returns for firm i over the test period
VOLUMEi = the mean dollar daily trading volume for firm i over the test period

31
 TABLE 3- OLS Estimation of the Dispersion in Analysts’ Forecasts
DISPi=γ0+γ1NUMANALi +γ2VAREARNi +γ3STDRETi +γ4TURNOVERi+γ5EGROWTHi+δi (9)

Panel A: Estimation of Dispersion in Analysts’ Forecasts for |MAA1i| Sample

γ0 γ1 γ2 γ3 γ4 γ5 R2 N
Predicted Sign ? + + + + 30.3% 422
Coef. -0.051 -0.0003 0.00009 5.705 12.841 -0.295
(t-statistic) (-2.455)** (-0.422) (2.194)** (6.356)** (2.718)** (-4.638)**

Panel B: Estimation of Dispersion in Analysts’ Forecasts for |MAA2i| Sample

γ0 γ1 γ2 γ3 γ4 γ5 R2 N
Predicted Sign ? + + + + 25.7% 539
Coef. -0.027 -0.0008 0.0001 4.742 13.960 -0.337
(t-statistic) (-1.256) (-1.165) (2.579)** (5.368)** (3.041)** (-5.500)**

Notes: The listed R2 are the adjusted R2. The t-statistics are corrected for heteroskedasticity using White’s consistent
estimator for standard error.

**Significant at the .01 level. * Significant at the .10 level (One-sided tests)

VARIABLE DEFINITIONS

DISPi =the mean standard deviation in analysts’ forecasts during the month of June, scaled by the median
forecast
NUMANALi = the mean number of analysts making forecasts during the month of June
VAREARNi = the standard deviation of annual earnings over the test period scaled by the average earnings over
the test period for firm i
STDRETi = the standard deviation of daily returns for firm i over the test period
TURNOVERi = the mean dollar daily trading volume divided by the mean number of shares outstanding for firm i
over the test period
EGROWTHi = Net earnings at the end of the test period less net earnings at the beginning of the test period scaled
by net earnings at the beginning of the test period

32
 TABLE 4- 2SLS Regression Results of the Relationship Between
Information Asymmetry (BIDASKi) and Earnings Management
|MAAXi| = α0 + α1 BIDASKi + α2 CFVARi + α3 DEBTi + α4MKTBVi + α5 SIZEi + α6 GROWTHi + εI (7)

Panel A: Multivariate Regression with |MAA1i| as dependent variable

α0 α1 α2 α3 α4 α5 α6 R2 N
Predicted + + + + ? +
Sign
Coef. 0.030 1.625 0.0152 0.0036 -0.0003 -0.0009 0.0077 17.7% 355
(t-statistic) (1.513) (4.135)** (4.129)** (1.768)* (-0.108) (-0.382) (1.195)

Panel B: Multivariate Regression with |MAA2i| as dependent variable

α0 α1 α2 α3 α4 α5 α6 R2 N
Predicted + + + + ? +
Sign
Coef. 0.032 0.600 0.0054 0.0013 0.0017 -0.0008 0.0055 14.0% 641
(t-statistic) (3.620)** (4.168)** (3.407)** (1.164) (1.542)* (-0.741) (0.679)

Notes: The listed R2 are the adjusted R2.

**Significant at the .01 level. * Significant at the .10 level (One-sided tests)

VARIABLE DEFINITIONS
MAA1i = the mean Managed Accounting Accrual under the modified Jones (1991) model using the time-series
estimation approach (see derivation in section III in the text)
MAA2i = the mean Managed Accounting Accrual under the modified Jones (1991) model using the cross-sectional
estimation approach (see derivation in section III in the text)
BIDASKi = the predicted bid-ask spread estimated from model 7 in table 2 above
CFVARi = the standard deviation of operating cash flows over the test period divided by the average operating cash
flows over the test period
DEBTi = the mean of the long-term debt divided by the book value of equity for firm i over the test period
MKTBVi = the mean market capitalization divided by the book value of equity for firm i over the test period
SIZEi = Natural log of the mean market capitalization for firm i over the test period
GROWTHi = the net revenues at the end of the test period less net revenues at the beginning of the test period scaled
by net revenues at the beginning of the test period

33
 TABLE 5- 2SLS Regression Results of the Relationship Between
Information Asymmetry (DISPi) and Earnings Management
|MAAXi| = α0 + α1 DISPi + α2 CFVARi + α3 DEBTi + α4MKTBVi + α5 SIZEi + α6 GROWTHi + εi (7’)

Panel A: Multivariate Regression with |MAA1i| as dependent variable

α0 α1 α2 α3 α4 α5 α6 R2 N
Predicted + + + + ? +
Sign
Coef. 0.028 0.133 0.0021 -0.006 0.002 -0.0021 0.011 22.1% 422
(t-statistic) (2.438)** (4.781)** (4.222)** (-0.709) (1.335) (-1.228) (3.943)**

Panel B: Multivariate Regression with |MAA2i| as dependent variable

α0 α1 α2 α3 α4 α5 α6 R2 N
Predicted + + + + ? +
Sign
Coef. -0.001 0.164 0.0031 0.006 0.0015 0.0004 0.011 14.6% 539
(t-statistic) (-0.079) (4.689)** (2.341)** (0.706) (0.901) (0.336) (4.153)**

Notes: The listed R2 are the adjusted R2.

**Significant at the .01 level.

VARIABLE DEFINITIONS
MAA1i = the mean Managed Accounting Accrual under the modified Jones (1991) model using the time-series
estimation approach (see derivation in section III in the text)
MAA2i = the mean Managed Accounting Accrual under the modified Jones (1991) model using the cross-sectional
estimation approach (see derivation in section III in the text)
DISPi = the predicted dispersion in analysts’ forecasts as estimated from model 7 in table 3 above
CFVARi = the standard deviation of operating cash flows over the test period divided by the average operating cash
flows over the test period
DEBTi = the mean of the long-term debt divided by the book value of equity for firm i over the test period
MKTBVi = the mean market capitalization divided by the book value of equity for firm i over the test period
SIZEi = Natural log of the mean market capitalization for firm i over the test period
GROWTHi = the net revenues at the end of the test period less net revenues at the beginning of the test period scaled by
net revenues at the beginning of the test period

34
 TABLE 6- Descriptive Statistics on Seasoned Equity Offerings and
Offering Firms

Panel A: Distribution of Seasoned Equity Offerings By Half-Year

Period No. of Offerings

1986 - Q3 & Q4 10
1987 - Q1 & Q2 27
1987 - Q3 & Q4 6
1988 - Q1 & Q2 3
1988 - Q3 & Q4 6
1989 - Q1 & Q2 2
1989 - Q3 & Q4 9
1990 - Q1 & Q2 12
1990 - Q3 & Q4 2
1991 - Q1 & Q2 15
1991 - Q3 & Q4 18
1992 - Q1 & Q2 20
1992 - Q3 & Q4 7
1993 - Q1 & Q2 13

Panel B: Descriptive Statistics on Offering Firms and Offerings

Mean Median Maximum Minimum

Market Value of Equitya ($millions) 829.3 351.2 10050.9 12.4
Leverageb 0.271 0.260 0.916 0.0001
Amount offered in seasoned equity offering 82.5 57.0 517.5 3.470
($ millions)
Amount offered/ Market Value of Equity 0.234 0.162 2.221 0.007

a
The market value of equity is the market capitalization of equity at the end of fiscal year t-1 .
b
Leverage is measured as long-term debt divided by total assets (both measured at the end of fiscal year t-1)

35
 TABLE 7- OLS Regression Results of the Relationship Between
Information Asymmetry and Earnings Management around Seasoned
Equity Offerings

MAA2i = α0 + α1BIDASKi + α2VARi + α3DEBTi + α4MKTBVi + α5SIZEi +α6OFFSIZEi +α7OWNi

+ α8GROWTHi + εi (10)

Regression
Variable Predicted Parameter T-stat
Sign Estimate
Intercept 0.067 1.803
BIDASKi + 3.721 2.216
VARi ? -0.006 -2.404
DEBTi - -0.077 -2.103
MKTBVi + -0.005 -1.037
SIZEi ? -0.010 -1.458
OFFSIZEi + 0.0043 0.488
OWNi + -0.0002 -0.455
GROWTH + 0.0044 2.327
No. of Obs. 150
R-Squared 16.9%
Adj. R-Squared 11.1%

Notes: The t-statistics are corrected for heteroskedasticity using White’s consistent estimator for standard error.

MAA2i =the Managed Accounting Accrual under the modified Jones (1991) model using the cross-sectional
estimation approach in the year of the offering(see derivation in section III)
BIDASKi =the bid-ask spread at the close of trade on the last trading day of June for the year before the seasoned
equity offering, scaled by the average of the bid and ask prices
VARi =the standard deviation of earnings divided by the average operating cash flows over the period five years
before the offering until the year before the offering
DEBTi =long-term debt divided by assets (both from year t-1)
MKTBVi =the market capitalization divided by the book value of equity for firm i in year t-1
SIZEi =Natural log of the market capitalization for firm i in year t-1
OFFSIZEi =Relative size of seasoned equity offering defined as of gross offer proceeds divided by the market
capitalization in year t-1
OWNi =Stock ownership of the offering firm’s directors and officers as a percentage of the total outstanding
shares
GROWTHi =Net revenues at the end of the test period less net revenues at the beginning of the test period scaled by
net revenues at the beginning of the test period

36
 Bibliography

Ajinkya, B.B., R.K. Atiase, and M.J. Gift, 1991, Volume of trading and the dispersion in financial
analysts’ earnings forecasts, The Accounting Review 66, 389-401.

Belsey, D.A., E. Kuh, and R.E. Welsch, 1980, Regression diagnostics: Identifying influential data and
sources of collinearity (Wiley: New York, NY).

Bhattacharya, U. and M. Spiegel, 1991, Insiders, outsiders, and market breakdowns, Review of
Financial Studies 4, 255-282.

Brown L.D. and J. Han, 1992, The impact of annual earnings announcements on convergence of
beliefs, The Accounting Review 67, 862-875.

Chaney, P.K. and C.M. Lewis, 1995, Earnings management and firm valuation under asymmetric
information, Journal of Corporate Finance: Contracting, Governance and Organization 1,
319-345.

Dechow, P.M., 1994, Accounting earnings and cash flows as measures of firm performance: The role
of accounting accruals. Journal of Accounting and Economics, 17, 3-42.

Dechow, P.M., R.G. Sloan, and A.P. Sweeney, 1995, Detecting earnings management, The
Accounting Review 70, 193-225.

DeFond, M. and J. Jiambalvo, 1994, Debt covenant violation and manipulation of accruals, Journal of
Accounting and Economics 17, 145-176.

Diamond D.W. and R.E. Verrecchia, 1991, Disclosure, liquidity, and the cost of capital. The Journal
of Finance 46, 1325-1360.

Dye, R., 1988, Earnings management in an overlapping generations model, Journal of Accounting
Research 26, 195-235.

Glosten, L. and P. Milgrom, 1985, Bid, ask and transaction prices in a specialist market with
heterogeneously informed traders, Journal of Financial Economics 14, 71-100.

Healy, P., 1985, The effect of bonus schemes on accounting decisions, Journal of Accounting and
Economics 7, 85-107.

Healy, P., K. Palepu and A. Sweeney, 1995, Causes and consequences of expanded voluntary
disclosure, Working Paper (Harvard Business School, Boston, MA).

Jiambalvo, J., 1996, Discussion of “Causes and consequences of earnings manipulation: An analysis of
firms subject to enforcement actions by the SEC”, Contemporary Accounting Research,
Spring, 37-47.

37
Jones, J. 1991, Earnings management during import relief investigations, Journal of Accounting
Research 29, 193-228.

Kasznik, R. 1995, On the association between voluntary disclosure and earnings management,
Working paper (Stanford University, Stanford, CA).

Lev, B. 1988, Toward a theory of equitable and efficient accounting policy, The Accounting Review
43,1-22.

Myers, S.C., 1993, Still searching for optimal capital structure, Journal of Applied Corporate Finance
(Spring 1993): 4-14.

Perry, S.E. and T.H. Williams, 1994, Earnings management preceding management buyout offers,
Journal of Accounting and Economics 18, 157-179.

Rangan, S., 1995, Earnings around seasoned equity offerings: Are they overstated?, Working paper
(University of Pennsylvania, Philadelphia, PA).

Schipper, K., 1989, Earnings management, Accounting Horizons 3, 91-106.

Shivakumar, L., 1996, Earnings management around seasoned equity offerings, Unpublished
Dissertation (Vanderbilt University, Nashville, TN).

Sweeney, A.P., 1994, Debt-covenant violations and managers’ accounting responses, Journal of
Accounting and Economics 17, 281-308.

Trueman, B. and S. Titman, 1988, An explanation for accounting income smoothing, Journal of
Accounting Research 26 supplement, 127-139.

Varian, H. 1985. Differences of opinion and the volume of trade. Working paper, University of
Michigan.

Watts R. and J. Zimmerman, 1978, Towards a positive theory of the determination of accounting
standards, The Accounting Review 53, 112-134.

Warfield, T.D., J.J. Wild, and K.L. Wild, 1995, Managerial ownership, accounting choices, and
informativeness of earnings, Journal of Accounting and Economics 20, 61-91.

Welker, M., 1995, Disclosure policy, information asymmetry and liquidity in equity markets,
Contemporary Accounting Research 11,801-827.

Zmijewski, M. and R. Hagerman, 1981, An income strategy approach to the positive theory of
accounting standard setting/choice, Journal of Accounting and Economics 3, 129-149.

38