University of Southern California Dissertations and Theses

Who manages the firm matters: the incremental effect of individual managers on accounting quality

(USC Thesis Other)

Who manages the firm matters: the incremental effect of individual managers on accounting quality

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

WHO MANAGES THE FIRM MATTERS:
THE INCREMENTAL EFFECT OF INDIVIDUAL MANAGERS ON ACCOUNTING
QUALITY

by
Kara Wells

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BUSINESS ADMINISTRATION)

August, 2013

Copyright, 2013 Kara Wells
2

ACKNOWLEDGEMENTS

I gratefully acknowledge the support and guidance of my dissertation chair, K.R.
Subramanyam, and committee members Mark DeFond and Kevin Murphy. I thank seminar
participants at the University of Southern California, Florida State University, Drexel University,
Southern Methodist University, Texas A&M, and Penn State University for helpful comments.
I also thank my fellow classmates: Kelsey Dworkis, Derek Horstmeyer, Jeff McMullin,
Suresh Nallareddy, Kari Olsen, Bryce Schonberger, Jim Stekelberg, Karen Ton, Biqin Xie, and
Alicia Yancy. Lastly, I thank my family for their unconditional support throughout my academic
career.
3

TABLE OF CONTENTS
ACKNOWLEDGEMENTS……………………………………………………………..……..… 2

LIST OF TABLES....……………………………………………………………………..…….... 5

LIST OF GRAPHS………………………………………………………………………...…….. 6

LIST OF FIGURES……………………………………………………………………..……….. 7

ABSTRACT………………………………………………………………………………..…….. 8

SECTION 1: INTRODUCTION……………………………………………………………....… 9

SECTION 2: MOTIVATION, PRIOR LITERATURE, AND HYPOTHESIS
DEVELOPMENT……………………………………………………….………. 16

2.1 Determinants of Accounting Quality……………………………………………..…... 16

2.2 The Role of Individual Managers in Corporate Decision Making …………………....17

2.3 Heterogeneous Managers with Non-Pecuniary Preferences………………………….. 23

2.4 Hypothesis Development……………………………………………………………... 24

SECTION 3: RESEARCH DESIGN, DATA, SAMPLE, AND VARIABLE DEFINITIONS…27

3.1 Methodology for Isolating Manager-Specific Effects…………………………………27

3.1.1 The Manager Mobility Method…………………………………………………..27

3.1.2 The Manager Connectedness Method……………………………………………31

3.1.3 The Spell Method………………………………………………………………...34

3.2 Sample and Data Selection ……………………........................................................... 35

3.3 Variable Definitions...………………………………………………………………….37

3.3.1 Measuring Accrual Quality………………………………………………………37

3.3.2 Measuring Innate Economic Determinants………………………………………39

SECTION 4: RESULTS…………………………………………………………………………41

4.1 Individual Managers and the Cross-Sectional Variation in AQ……………………….41

4

4.2 Change in AQ When Firms’ Switch Manager Type………………………………….. 44

4.3 The Role of Contracts and Governance………………………………………………. 45

SECTION 5: ROBUSTNESS TESTS………………………………………………………….. 47

5.1 Direct Approach to Calculating Accruals…………………………………………...... 47

5.2 Scaled Accounting Quality…………………………………………………………… 48

5.3 Placebo Tests…………………………………………………………………………. 48

5.4 CEOs and CFOs vs. All Top Managers………………………………………………. 49

5.5 Signed Discretionary Accruals……………………………………………………….. 49

5.5 Changes in Firm AQ around Manager Departures…………………………………… 50

SECTION 6: CONCLUSION………………………………………………………………….. 51

REFERENCES…………………………………………..……………………………………... 53

APPENDIX……………………………………………………………………………………... 82

5

LIST OF TABLES

TABLE 1: Data Availability......................................................................................................... 58

TABLE 2: Summary Statistics...................................................................................................... 59

TABLE 3: Correlations for Accounting Quality Measures and Economic
Determinants of Accounting Quality.......................................................................... 61

TABLE 4: Testing Individual Managers' Fixed Effects on Accounting Quality
– Manager Mobility Method....................................................................................... 62

TABLE 5: Testing Individual Managers' Fixed Effects on Accounting Quality
- SPELL Method......................................................................................................... 64

TABLE 6: Change in Firm Measures and Executive Compensation When Firms
Switch Manager Type................................................................................................. 65

TABLE 7: Change in Accounting Quality as Managers Move To a 3rd Firm............................. 66

TABLE 8: Firm and Manager Characteristics.............................................................................. 67

TABLE 9: Testing Individual Managers' Fixed Effects on Accounting Quality
Controlling for Governance and Compensation......................................................... 68

TABLE 10: Testing Individual Managers' Fixed Effects on Accounting Quality
– Statement of Cash Flows...................................................................................... 69

TABLE 11: Testing Individual Managers' Fixed Effects on Scaled Accounting
Quality...................................................................................................................... 70

TABLE 12: Testing Individual CEO Fixed Effects on Accounting Quality................................ 71

TABLE 13: Testing Individual CFO Fixed Effects on Accounting Quality................................ 72

TABLE 14: Testing Individual CEO or CFO Fixed Effects on Accounting Quality................... 73

TABLE 15: Testing Individual (Non-CEO or CFO) Managers' Fixed Effects on
Accounting Quality.................................................................................................. 74

TABLE 16: Testing Individual Managers' Fixed Effects on Discretionary Accruals
(3 year average)........................................................................................................ 75

TABLE 17: Testing Individual Managers' Fixed Effects on Discretionary Accruals
(5 year average)........................................................................................................ 76

6

LIST OF GRAPHS

GRAPH 1: Firm Accounting Quality Over Time......................................................................... 77

GRAPH 2: Firm Discretionary Accruals Over Time.................................................................... 78

GRAPH 3: Firm Discretionary Accruals (3 year average) Over Time......................................... 79

GRAPH 4: Firm Accounting Quality Around Manager Changes (1 year)................................... 80

GRAPH 5: Firm Accounting Quality Around Manager Changes (3 year)................................... 81

7

LIST OF FIGURES

FIGURE 1: Timeline of Data Included in the Regression Analysis............................................. 29

FIGURE 2: Manager Connectedness Example............................................................................. 31

8

ABSTRACT

I investigate whether individual managers have an incremental effect on their firms’
accounting quality (AQ) after controlling for known determinants of AQ, time fixed effects and
firm fixed effects. To identify the manager-specific effect on firm AQ, I construct a data set that
tracks the movement of 720 managers across firms over the period 1992-2011. Results indicate
that individual manager fixed effects explain a statistically and economically significant
proportion of the cross-sectional variation in AQ, which is three times larger than that of industry
fixed effects and comparable to that of firm fixed effects. Moving from the quartile of most
aggressive managers to that of the most conservative managers, results in a 41% improvement in
the firms’ AQ. I also examine the consequences of switching between conservative and
aggressive managers on the firms’ AQ. Results show that a firm that switches from an aggressive
(conservative) manager to a conservative (aggressive) manager increases (decreases) the firm’s
AQ by 32% (94%). My study underscores the importance of understanding individual manager
styles in the determination of the firms’ AQ.

9

SECTION 1: INTRODUCTION

In a survey by Dichev, Graham, Harvey, and Rajgopal (2013), the authors asked CFOs to
comment on how they would investigate the quality of accounting (earnings) information. One
CFO responded as follows: “intensive fundamental analysis of the backgrounds of the top people
running the company. I would like to look at the experience of the people behind a lot of the
numbers” (p. 43). This candid response underscores the practitioner perspective that the
individuals that manage firms are important in influencing not only strategic decisions but also in
determining accounting quality. While practitioners accept that who manages the firm is
important, academics have only recently begun to diverge from the neoclassical economic
assumption of homogenous managers to examine whether individual managers of the firm have
an economically significant effect on the manner in which the firm conducts its business
(Bertrand & Schoar, 2003).
In this study, I investigate whether the individual manager has an incremental impact on
the firm’s accounting quality (hereafter “AQ”) after controlling for known determinants of AQ,
time fixed effects and firm fixed effects.
1
While prior literature has examined various
determinants of quality, little attention has been focused on the role of the individual manager;
managers were assumed to be homogenous or inconsequential in determining AQ.
2
In contrast, I

1
While I acknowledge that accounting quality and earnings quality encompass various dimensions of “quality” that
are decision and context specific, this study focuses on the dimension of quality captured the modified Dechow and
Dichev (2002) accrual quality measure.

2
Dechow, Ge, and Schrand (2010) review the earnings quality literature and note there are over 100 published
papers that use some variation of the AQ measure used in this study. The authors note that that the determinants
studied in the earnings quality literature include firm characteristics, governance, auditors, and capital market
incentives, and that the literature has mixed results depending on the dimension of earnings quality examined.
Francis, Olsson, and Schipper (2006) note that managers make decisions that impact financial reporting. However,
this stream of literature does not examine the individual impact of a specific manager on financial reporting. One
notable exception is Ge, Matsumoto, and Zhang (2011) who look at the impact of individual CFOs on various
10

consider the possibility that the individual who manages the firm can have an incremental effect
on AQ that is distinct and different from that of the firm.
I find that the role of individual managers on AQ is both statistically and economically
significant. In the cross-section, I find that the inclusion of manager fixed effects to a regression
of AQ on economic determinants, time and firm fixed effects results in a 3% increase in the
adjusted R
2
, and that the AQ for a firm that employs a manager in the 25
th
percentile
(conservative manager) is 41% higher than a firm that employs a manager in the 75
th
percentile
(aggressive manager). Using an alternative methodology to decompose the relative importance of
each determinant of AQ, I find that the relative importance of manager fixed effects in
explaining AQ is even comparable to that of firm fixed effects. I also find that firms that switch
from an aggressive (conservative) manager to a conservative (aggressive) manager experience an
increase (decrease) in AQ of 32% (94%). Overall, my results underscore the important role that
individual managers play in determining AQ.
My empirical analysis is comprised of two sets of tests. In the first set, I examine
whether, and to what extent, individual managers incrementally explain the cross-sectional
variation in AQ. To do this, I regress AQ on known determinants of AQ, time, firm, and manager
fixed effects. Following prior literature, I measure accounting quality as the inverse of the
standard deviation of abnormal accruals using the modified Dechow and Dichev model (Dechow
& Dichev, 2002; Francis, LaFond, Olsson, & Schipper, 2004, 2005). Also, as explained later, I
use several alternative methodologies to address unique econometric problems that arise with
each method and to ascertain the robustness of my results.

financial reporting decisions, such as off-balance sheet decisions for operating leases and rates of return on pension
assets.
11

First, I use the manager mobility method (Bertrand & Schoar, 2003), which separately
identifies the manager-specific effects on AQ from those of the firms-specific effects by tracking
a set of managers that change firms. Using a sample of 880 firms and 720 managers who are
employed by at least two firms in my sample, I find that adding manager fixed effects to a
regression of AQ on economic determinants, time and firm fixed effects increases the adjusted
R
2
by 3%, which is statistically and economically significant.
3
One criticism of this result is that
it is not surprising that manager fixed effects can explain a significant portion of the variation in
AQ due to the large number of manager fixed effects. To address this criticism, I perform two
types of analysis. First, I document that the frequency of significant coefficient estimates for the
manager fixed effects exceeds the threshold that would be observed under the null hypothesis
that the significant coefficients are due to chance. Also, a joint test of the manager fixed effects
being equal to zero is rejected even after bootstrapping the distribution to allow for potential
biases in the F-statistic, suggesting that manager fixed effects are significant in determining AQ.
4

Second, I examine the cross-sectional variation in the coefficient estimates of the manager fixed
effects and find that moving from the 75
th
percentile (aggressive manager) to the 25
th
percentile
(conservative manager) results in a 41% improvement in AQ, which appears to be economically
significant.
Second, to address limitations of the manager mobility method due to the stringent data
requirements, I implement two alternative methodologies: the manager connectedness method

3
For example, the incremental adjusted R
2
is only 0.6% when I include industry fixed effects in a regression of AQ
on economic determinants and time fixed effects.

4
Fee, Hadlock, and Pierce (2013) note that randomization or bootstrapping data and comparing the bootstrapped
distribution to the F-stat calculated from the joint hypothesis test that the fixed effects are jointly equal to zero is a
way to alleviate the concern of biased F-tests as explained in Wooldridge (2002).

12

and the spell method (Abowd, Kramarz, & Margolis, 1999).
5
Using the manager connectedness
method, I rerun the main regression: AQ on previously documented known determinants, time,
firm, and manager fixed effects and find that manager fixed effects are still important
determinants of AQ. Specifically, I decompose the explained variation of AQ to understand the
relative importance of each determinant of AQ in explaining the model R
2
. I find that the
manager-specific component accounts for 20% of the overall explained variation in AQ, which is
greater than the 15% attributable to firm fixed effects, suggesting that the manager has a
relatively more important role in determining AQ than the firm. Finally, using the spell method, I
run a modified version of the main regression and find that managers have a distinct impact on
AQ that is different from the impact of the firm, even in a larger setting, when manager mobility
is not imposed.
Thus far, I have only quantified the extent of cross-sectional variation in AQ that can be
attributed to individual managers. In my second set of tests, I more directly examine the effect of
managerial style on AQ by examining what happens to AQ when firms switch managers with
differing styles. Specifically, I classify managers as aggressive, moderate, or conservative based
on the quartile ranks of their fixed effects coefficients under the manager mobility method
regression. Then, I identify those firms that switched managers across the extreme styles (i.e.,
aggressive to conservative or vice versa) and examine the change in firms’ AQ after the manager

5
The manager connectedness method was developed by Abowd et al. (1999) to study employee-employer
relationships in labor economics. They formally prove that “connectedness” is a necessary and sufficient condition
to identify firm and manager fixed effects. This method allows me to calculate the incremental effect that individual
managers have on AQ for all managers employed by firms that have at least one manager that moves to another firm
in the dataset (and not just managers that change firms, as in the manager mobility method). Because of the
decreased data restrictions, my manager connectedness sample consists of the same 880 firms used in my manager
mobility sample, but includes 12,946 managers: the 720 managers in the manager mobility sample plus an additional
12,226 managers that do not move to a different firm in my sample. The spell method does not require manager
mobility but does not provide separate identification of the fixed effects of managers. As such, my spell sample
consists of 2,235 unique firms and 20,011 unique managers, a much larger sample than used in my other tests.

13

change. I find that, for the 14 firms (28 firms) that switch from conservative (aggressive)
managers to aggressive (conservative) managers there is a decrease (increase) in AQ of 94%
(32%) in the subsequent period. I find no significant changes in AQ for the 192 manager
switches when a moderate manager is involved.
6

The classification of managerial style (i.e., aggressive or conservative) using the manager
mobility method requires the use of data for the full time period, i.e., before and after the switch.
As a result, my test is in-sample and could potentially suffer from ex-post bias. To rule out ex-
post bias, I use a subsample of 29 managers that switch firms for a second time in my sample and
determine manager types using data prior to the second switch. This allows me to implement an
out-of-sample analysis for the 29 instances where the managers switch for the second time. The
findings from this analysis are directionally similar to my prior analysis although not statistically
significant, potentially due to low power.
Finally, I examine if the incremental manager-specific effect on AQ is driven by
fundamental differences in firm governance and/or compensation of the manager. To control for
these differences, I include measures of governance and compensation in the regression of AQ
on manager fixed effects, using the manager mobility method. My results indicate that the
incremental manager-specific effects persist even after explicitly controlling for governance and
compensation of the manager. This result shows that individual differences between managers
that incrementally impact firm AQ are not fully monitored or contracted away, suggesting that

6
This test cannot rule out a dynamic self-selection wherein companies undergoing structural change seek managers
better suited to the new company objectives. I look at a handful of firms that switch managers again, and find that
the firm’s AQ subsequently shifts again. While this cannot rule out self-selection, it is rare that a firm would shift
structurally and then switch back.

14

other manager characteristics unrelated to contracting explain much of the variation in AQ
attributable to individual managers.
7

My study contributes to the literature by documenting that the individual who is the
manager of a firm significantly affects AQ, and this effect is as significant as the individual firm
itself. This result underscores the importance of studying the firms’ managers for a complete
understanding of the accounting information environment of a firm. In addition, this finding
complements a current trend that examines managerial characteristics as likely determinants of
AQ (Demerjian, Lev, Lewis, & McVay, 2012, 2013; Koh, 2011; Schrand & Zechman, 2012). In
contrast to these studies, I utilize several alternative methodologies to separately identify the
magnitude of the incremental effect that managers have on AQ after controlling for firm fixed
effects and other firm characteristics, emphasizing the importance of understanding individual
manager styles in the determination of the firms’ AQ.
My study also contributes to the growing managerial “style” literature in finance and
accounting, which has provided evidence that managerial style impacts corporate decisions
(Bertrand & Schoar, 2003), compensation (Graham, Li, & Qiu, 2012) voluntary disclosure
(Bamber, Jiang, & Wang, 2010; Davis, Ge, Matsumoto, & Zhang, 2012; Yang, 2012), tax
avoidance (Dyreng, Hanlon, & Maydew, 2010) and various reporting choices (Ge et al., 2011).
The contribution to this literature is twofold. First, I document that who the manager is impacts
the overall accounting quality of a firm and that this impact is as large as that of the firm itself. I
also document that the cross-sectional variation in managers’ impact on accounting quality is not
fully explained by firm governance or compensation contracts as implied by traditional agency
theory. These results are an important addition to the managerial style literature because they

7
While compensation contracts and monitoring do not eliminate the idiosyncratic impact of the manager on firm
accounting quality, this does not imply anything about the optimality of governance or compensation contracts.
15

show that managers’ effect extends beyond specific investment choices to also impact the firms’
accounting quality. Second, I also use alternative methodologies such as the manager
connectedness and the spell methods (Abowd et al., 1999) when examining AQ effects.
8
This
new method furthers the managerial “style” literature by allowing researchers to study the impact
of managers on firm decisions for manager who do not switch firms.
In the next section I motivate the study, review the relevant literature, and develop formal
hypotheses. In Section 3, I discuss the alternative methodologies employed in this study, data,
sample, and variable definitions. Section 4 presents the results. Section 5 discusses various
robustness checks, and Section 6 concludes.

8
There are two papers in the finance literature; Graham et al. (2012) and a 2013 working paper by Coles and Li that
use the manager connectedness method to study manager fixed effects in a compensation context. To my knowledge
there are no papers in the accounting literature that employ the manager connectedness method.

16

SECTION 2: MOTIVATION, PRIOR LITERATURE, AND HYPOTHESIS
DEVELOPMENT

2.1 Determinants of Accounting Quality
According to the Statement of Financial Accounting Concepts No. 1 and following the
definition used in Dechow et al. (2010, p. 1), earnings quality is defined as “higher quality
earnings provide more information about the features of a firm’s financial performance that are
relevant to a specific decision made by a specific decision-maker.”
This definition acknowledges that quality is conditional on the decision-relevance of the
information provided. The decision-relevance of the information is user specific, and there are
multiple users of financial information with varying objectives (Kothari, Ramanna, & Skinner,
2010). In addition, reported earnings are a function of a firm’s financial performance during a
reporting period which varies not only with the accounting measurement system but also with the
implementation of the accounting system (Dechow et al., 2010).
Many studies in the accounting quality literature focus on the accounting measurement
system by modeling the accrual process of earnings, using the abnormal component as a proxy
for earnings quality (Dechow, 1994; Dechow & Dichev, 2002; Jones, 1991). Here the premise is
that if we model the accrual process correctly, then any deviations or abnormal accruals erode
decision usefulness and result in lower earnings quality.
A widely accepted abnormal accrual model in this field is a modified version of the
Dechow and Dichev model (Dechow & Dichev, 2002). In this model accruals serve a noise
reduction role for mapping reported earnings into firm performance (Dechow, 1994). Subsequent
research adds to this model by adding growth in revenue and gross property plant and equipment
17

to capture performance and expand the type of accruals modeled (Francis et al., 2005;
McNichols, 2002). Accrual quality is then the standard deviation of the residuals from the
regression of working capital accruals on cash flows from operations, growth in revenue and
gross property plant and equipment.
Francis et al. (2004) note the standard deviation of the residuals is comprised of innate
estimation errors and discretionary estimation errors. To isolate the discretionary estimation
errors they model the accrual quality as a function of firm characteristics identified in the
literature as varying with the accrual quality measure (Dechow & Dichev, 2002). These firm
characteristics include: firm size, cash flow variability, sales variability, length of the operating
cycle, incidence of negative earnings realizations, absence of intangibles, intangibles intensity,
and capital intensity.
My study contributes to the literature on the determinants of accrual quality by
decomposing the discretionary component of accrual quality. I document that another important
determinant of accounting quality (specifically accrual quality) is the individual manager of a
firm. By including the incremental manager-specific effect in the accrual quality model, after
controlling for innate determinants of accrual quality, I show that individual managers impact the
accounting measurement system which affects the accounting quality of a firm. In the next
section I provide evidence in the literature for why an individual manager would or would not be
able to uniquely affect firm decisions and more specifically, accounting choices.

2.2 The Role of Individual Managers in Corporate Decision Making
In the finance and accounting literatures, researchers have recently begun to examine the
potential existence and impact of manager-specific effects on various firm decisions (e.g.,
18

Bamber et al., 2010; Bertrand & Schoar 2003; Dyreng et al., 2010; Ge et al., 2011; Graham et al.,
2012). This recent development in the literature challenges traditional economic theory, which
provides no role for idiosyncrasies in managers, by documenting that manager characteristics do
have an impact on certain firm decisions. The seminal study in this emerging line of research,
Bertrand and Schoar (2003), refer to the manager-specific effect as “managerial style.”
9

Bertrand and Schoar (2003) investigate the impact of managerial style on three types of
firm policies: investment policy (capital expenditures, acquisitions, etc.), financial policy
(dividend payout, interest coverage, etc.), and organizational policy (diversification, R&D
expense, etc.). They find that managerial style does have an incremental effect in explaining the
variation in some firm policies, with the greatest effect for large decisions such as acquisitions.
The authors also note that managerial style is systematically related to manager characteristics
such as if the manager was born prior to World War II and if the manager received an MBA
degree. Expanding on Bertrand and Schoar, Graham et al. (2012) find that managerial style
impacts managers’ compensation and that the managerial style component associated with
compensation is correlated with the managerial style of corporate decisions.
In the accounting literature several recent studies adopt the manager mobility
methodology of Bertrand and Schoar (2003) to understand the impact of managerial style on
accounting choices (e.g., Bamber et al., 2010; Davis, Ge, Matsumoto, & Zhang, 2012; Demerjian
et al., 2013; Dyreng et al., 2010; and Ge et al., 2011).
10
Bamber et al. (2010) find that a
managerial style impacts voluntary disclosure for managers whose last position is at the level of

9
The manager mobility methodology proposed by Bertrand and Schoar (2003) estimates a fixed-effects regression,
controlling for known determinants of the dependent variable along with firm fixed effects. The estimated manager
fixed effect is therefore the incremental contribution of a manager’s style on the firm decision of interest. This
methodology is explained in greater detail in the research design section.

10
While these studies are similar methodology, the fundamental research questions that they examine are different
from this current study. In addition, Demerjian et al. (2012) propose a new measure using data envelopment analysis
(DEA) to specifically capture the manager-specific component that is attributable to ability.
19

CEO, CFO, or general counsel. In addition, they find that managers with military experience or
who are born prior to World War II have more conservative disclosure styles with outside
investors. Dyreng et al. (2010) find an incremental manager-specific effect when the accounting
variable of interest is tax avoidance. However, when the authors try to explain the manager-
specific effect using manager characteristics, they find no systematic relationship between a
manager’s firm-level tax avoidance style and a manager’s biographical characteristics. They
attribute this difference to the manager setting the “tone at the top” rather than being directly
involved in tax avoidance activities. Ge et al. (2011) examine whether the CFO’s style impacts
accounting decisions such as off-balance sheet operating lease activity and the expected rate of
return on pension assets. The paper finds mixed evidence that CFO style impacts these specific
accounting decisions and that CFO style is not linked to observable characteristics such as
functional career track, gender, or age. In a current working paper, Davis et al. (2012) find that
managerial style can be linked to the tone (measured by word choice) of conference calls. This
paper also finds that optimistic tone is positively associated with charitable giving. Using a
model of manager ability developed in Demerjian et al. (2012), Demerjian et al. (2013) find that
firm accounting quality is linked to a manager’s ability.
In contrast to the recent managerial style literature, neoclassical economic theory
provides no role for managerial style since all individuals in the economy are assumed to be
homogenous. Under this view of the world, managers are perfect substitutes, and in the extreme
case, are irrelevant for firm decisions.
11
A less extreme view of the world allows managers to be
heterogeneous in very limited ways. Specifically, traditional agency theory allows managers or
agents to vary in two dimensions: risk preferences and effort or ability. Here managers want to

11
Bertrand and Schoar (2003) note that assuming a neoclassical view of the firm, “different managers are regarded
as perfect substitutes for one another. An even more extreme assumption is that top managers simply do not matter
for what is going on within a firm” (p. 1173).
20

maximize their utility function which incorporates risk aversion and effort aversion. In the
simplest case, the firm (i.e., the board of directors) can monitor the manager to ensure that he
takes the correct action and exerts the desired level of effort. Additionally, to the extent that the
firm can use contracts (specifically compensation contracts) to incentive-align the manager, the
manager will act as an agent of the firm and his specific idiosyncrasies will play no role in firm
decisions (albeit they may play a role in the contract that the manager receives). The effective
use of monitoring and contracting would then imply that we can think of all managers in terms of
a “representative agent” (e.g., Jensen & Meckling, 1976).
12

Consistent with the classic research in economics and finance, the classic research in
strategic management also concludes that managers are substitutable. One specific stream of
literature in this field focuses on the socialization and selection processes for becoming a
manager. This stream of literature finds that the process to become a top manager at a large firm
increases the homogeneity among managers in similar positions (DiMaggio & Powell, 1983; Hitt
& Tyler, 1991; Salanick & Pfeffer, 1977).
13
This decrease in heterogeneity for top managers is
attributed to the mechanism by which individuals are selected to rise through the ranks and
become a top manager (DiMaggio & Powell, 1983).
A divergence from the classic views in the strategic management literature occurred with
the introduction of upper echelons theory (Hambrick & Mason, 1984). Upper echelons theory

12
The corporate governance literature allows for managerial discretion to exist and create variation in firm decisions
when the monitoring of the manager is poor (i.e., poor corporate governance). This literature primarily attributes the
variation to the monitoring capabilities of the firm, not to idiosyncrasies of managers. Hermalin and Weisbach
(1998) are one exception in this stream of literature. Here the authors specifically model manager discretion and
show a feedback relationship between the level of discretion a manager can implement and the level of corporate
governance.

13
In addition to this particular stream of research in strategic management, other streams of literature in
management reach the same conclusions that top managers are interchangeable due to entrenched norms and
cultures (Hannan & Freeman, 1977) and due to managers imitating other managers (Hambrick, Geletkanycz, &
Fredrickson, 1993; Spender, 1989).
21

states that individuals use bounded rationality when making complex or ambiguous decisions
and are influenced by past experiences and values (Finkelstein & Hambrick, 1996; Hambrick,
2007; Hambrick & Mason, 1984). More recent research in management has begun to study the
empirical implications of upper echelon theory and has largely found that managers can impact
firm decisions (Thomas, Litschert, & Ramaswamy, 1991; Tihanyi, Ellstrand, Daily, & Dalton,
2000; Smith & White, 1987). In sum, prior literature provides mixed evidence that manager-
specific effects impact firm decisions.
My study contributes to two streams of literature in accounting: the association between
manager characteristics and accounting quality literature and the “managerial style” literature.
The accounting quality or financial reporting and manager characteristics literature (e.g., Dyreng,
Mayew, & Williams, 2009; Francis, Huang, Rajgopal, & Zang, 2008; Hutton, Jiang, & Kumar,
2013; Schrand & Zechman, 2012) use different measures of accounting quality (e.g., AAERs)
and document associations with CEO characteristics such as overconfidence, ability, and
reputation. In contrast, my study differs from this prior literature by separately identifying the
magnitude of the impact of all top managers on accrual quality. My study adds to this literature
by documenting that managers have a distinct impact on AQ that is different from the firm and
which seems to be more important than the impact of the industry and comparable to the impact
of the firm fixed effect. My findings are important and incremental to the AQ and manager
characteristics literature, because I am able to distinguish the magnitude of the impact of the
manager from the firm, which prior studies using observable manager characteristics could not
do since many inferred managerial traits from firm level observations, making it difficult to
disentangle the firm effect from the manager effect. In addition, my study captures both
observable and unobservable time invariant traits of the manager and does not rely on inferring
22

managerial characteristics from outsiders (e.g., press coverage and awards) or by using firm level
measures to infer manager characteristics. In addition, I provide out-of-sample evidence that
managers impact AQ in a consistent manner and that it is the manager per se that impacts AQ.
The second stream of literature related to my study is the “managerial style” literature
(e.g., Bamber et al., 2010; Davis et al., 2012; Dyreng et al., 2010; Ge et al., 2011; Yang, 2012).
This stream of literature uses the manager mobility methodology to document that “managerial
style” is present in settings such as voluntary disclosure choices, tax avoidance behaviors, tone of
conference calls, choice of operating leases, etc. In contrast, my study uses a summary measure
of accounting quality, to capture the overall impact of the many choices of managers, which are
not made in isolation, and which can have an interactive effect on the overall accounting
information environment of the firm. My study adds to this literature by examining a summary
measure of managers’ accounting choices, showing who the manager is impacts the overall
accounting quality of a firm and that this impact is three times as large as the impact of industry
fixed effects and comparable to the firm fixed effect. I also document that the cross-sectional
variation in managers’ impact on accounting quality is not fully explained by firm governance or
compensation contracts as implied by traditional agency theory. These results are an important
addition to the managerial style literature because they show that managers’ impact to firm
choices extends beyond specific investment choices to impact the overall accounting information
environment of the firm and that these results are robust to bias correction in the F-Statistic (the
main test statistic reported in these studies). Second, this study adds to the managerial style
literature in accounting by introducing an alternative methodology; the manager connectedness
method (Abowd et al., 1999) and using a second alternative method, the spell method, to
increase generalizability without imposing manager mobility.
23

2.3 Heterogeneous Managers with Non-Pecuniary Preferences
In this section I will describe and provide intuition for how two managers with different
utility functions arrive at different firm level decisions given the same level of monitoring and
choice of compensation contracts. The example included in this section serves as an illustration
for how traditional models of expected utility theory can be modified to show that the individual
preferences of a manager can impact firms and are not irrelevant (i.e., managers are not
homogenous agents).
In the traditional economic framework individuals or agents want to maximize their
utility function. A utility function is comprised of all activities by which an agent derives utility
or pleasure from consuming. While these activities may be non-pecuniary, all activities can be
mapped to a monetary value such that a relationship between activities can be determined. Under
this framework an agent can then maximize his utility by maximizing his personal wealth. If
wealth maximization is synonymous with utility maximization then a firm can provide monetary
incentives to get the manager to take the correct action.
If however, managers’ utility functions include preferences which are unable to be
mapped to wealth maximization, then it becomes unclear that a firm will be able to incentivize
the manager to take the correct action if the correct action (according to the firm) does not
maximize the manager’s utility function. Some examples of personal preferences could include
such things as honesty, integrity, ability, or risk attitudes.
14

To clarify how differences in managers’ preferences can lead to differences in firm
accounting quality, consider the following illustrative example. Consider two managers
(Manager A and Manager B). Manager A values personal integrity and the loss of his integrity

14
Khalil (2004) notes that integrity is a non-ordinary commodity as it is a by-product of a choice over alternatives
taken within a context and can therefore not be reduced to be one of the alternatives.
24

would give him infinitely negative utility. Manager B does not value personal integrity in his
utility function, and is only concerned with personal wealth maximization.
Both managers are employed at firms where they are responsible for making decisions
that impact the overall accounting quality of the firm. Some decisions, such as managing
reported earnings, would decrease accounting quality but could potentially increase the expected
wealth of the manager. Manager A would never chose to engage in activities that would impact
his personal integrity. Manager B would engage in activities that decrease AQ if those activities
increase his expected personal wealth.
If firms do not want managers to engage in activities that decrease firm accounting
quality firms can 1) monitoring the actions of managers 2) offer compensation contracts that
incentive align the manager with the firm 3) hire only managers that value personal integrity. If
firms can effectively monitor managers and prevent them from engaging in activities that
decrease firm accounting quality then I would not expect individual managers to impact firm
accounting quality. In addition, if engaging in this activities is not wealth maximizing for
manager B then I would expect manager A and manager B to take the same actions and thus not
differentially impact firm accounting quality. However, if firms are unaware of managers’
individual preferences or if the cost to constrain managers’ individual preferences is too high,
then I would expect individual preferences to impact firm decisions, including the overall
accounting quality of the firm.

2.4 Hypothesis Development
Top managers are an integral part of a firm and are involved in daily operations by setting
“the tone at the top” as well as direct involvement in activities that require major strategic
25

decision making such as acquisitions (Bertrand and Schoar, 2003; Dyreng et al., 2010). This vital
role of top managers, working together to make key firm decisions, is paramount for the health
and operations of the firm. All of these decisions, whether large or small, ultimately need to be
recorded in the financial statements, which for the firms in my study are all bound by US GAAP.
While regulators mandate financial statements in accordance with US GAAP and auditors
oversee adherence to these standards, judgments and estimates in the accounting system provide
an opportunity for managerial discretion in determining some of the numbers reported in
financial statements. If the manager per se matters, i.e. individual managers have differential
impact on the overall accounting quality of the firm then I would expect that accounting quality
would be associated with the manager’s individual style. This leads to my first hypothesis:

H1: Individual managers are determinants of accounting quality.

While prior literature suggests that there is cross-sectional variation across managers
(e.g., Schrand & Zechman, 2012) it is an open question as to whether this variation in managers
leads to incremental explanatory power in understanding the cross-sectional variation in
accounting quality.
If managers incrementally impact accounting quality and apply a consistent style across
firms, then when a new manager is hired by a firm that previously employed a manager with a
different style, the firm should subsequently exhibit changes in accounting quality due to the new
manager imposing a different style. To capture the AQ style of the managers, I classify managers
as either AQ conservative, aggressive, or moderate (explained in details in the research design
section). If firms switch from employing aggressive (conservative) managers to conservative
26

(aggressive) managers, then I would expect the firm’s AQ to change in response to the new type
of manager. This leads to my next set of hypotheses:

H2a: Firms that initially hire conservative (aggressive) managers and subsequently hire
aggressive (conservative) managers will have a decrease (increase) in accounting
quality

H2b: Firms that initially hire moderate managers and subsequently hire moderate
managers will have no change in accounting quality

My last sets of hypotheses explore the mechanism by which managers’ individual style is
captured in a firm. It could be that although managers are individually different that their
differences are either not observed in the data or are attenuated because the firm offers
compensation contracts or has a level of monitoring that limits the individual manager effects
from impacting the overall AQ of the firm as explained in section 2.3. To study this link I test the
following non-directional hypotheses:

H3a: Conservative managers will have compensation contracts that are statistically
different from the compensation contracts of aggressive managers

H3b: The governance at firms which employ conservative managers will be statistically
different from the governance at firms which employ aggressive managers.

27

SECTION 3: RESEARCH DESIGN, DATA, SAMPLE, AND VARIABLE DEFINITIONS

In this section I provide details regarding the three methods used in this study to identify
the incremental impact of individual managers on firms’ accounting quality. After explaining the
three methodologies, I then describe the sample and data selections followed in this study as well
as define its key variables.

3.1 Methodology for Isolating Manager-Specific Effects
To examine whether individual managers impact accounting quality, I employ three
alternative methods designed to isolate the impact of managers on firm policies. Specifically, I
use the manager mobility method developed by Bertrand and Schoar (2003), the manager
connectedness method developed by Abowd et al. (1999), and the spell method introduced by
Abowd et al. While each method has distinct advantages, each method also has limitations.
These advantages and limitations, in addition to details regarding the implementation of each
method are described below.

3.1.1 The Manager Mobility Method
The primary methodology employed in this study is the manager mobility method.
15
This
method was first popularized in the finance literature by Bertrand and Schoar (2003), who use
the manager mobility method to study the incremental impact of managers on various corporate
decisions. In short, by exploiting the movement of managers across firms, the manager mobility
method allows researchers to separately identify fixed effects due to time invariant
characteristics of the firm and fixed effects due to time invariant characteristics of the manager.

15
Graham et al. (2012) define the manager mobility method as “the mover dummy variable (MDV) method”.
28

To more fully develop the intuition behind the manager mobility method, imagine that a
researcher would like to answer the question: how does manager A impact the firm decision Y?
To examine this question, the researcher would like to run the following regression:

(1)

Where

is the decision variable of interest for firm i at time t,

is a vector of
known determinants of the decision variable, and

is a dummy variable that equals 1 for
Manager A (i.e., a manager at firm i at time t) and zero otherwise. The problem with this model
lies in the interpretation of the coefficient on the fixed effect

. Under the above specification it
is unclear whether the fixed effect is due to Manager A or, alternatively, if this fixed effect is
actually a firm fixed effect and Manager A is irrelevant to the firm decision Y.
16
In order to
disentangle the firm fixed effect from the manager fixed effect, the researcher needs to include
both firm and manager fixed effects in the regression specification as follows:

(2)

Where

is an indicator variable that equals 1 for firm i and zero otherwise, and all other
terms are as defined in equation 1. Under this regression specification, the estimated coefficient
on

can be interpreted as the incremental effect of manager A on the decision variable Y at
firm i, after controlling for the fixed effect of firm i on the decision variable Y. Empirically,
however, this specification is problematic because, if Manager A is only a manager at firm i,
then

and

will be perfectly collinear.

16
This is the assumption maintained by neoclassical economic theory as explained in section 2.
29

Bertrand and Schoar (2003) provide a solution to the above problem by tracking
managers as they change firms over time. To be included in their study’s sample, Bertrand and
Schoar require that a manager must move firms at least once in his career and that he must be at
each firm for at least three years in order to have sufficient time to imprint his own managerial
“style” at the firm. Additionally, to increase the strength of estimating the firm fixed effects, the
authors include all firm-years in the regression, for all firms that have a manager that is in the
dataset.
To illustrate this point more concretely consider the following example, illustrated in
figure 1, from a manager that is included in the dataset used in my study. Charles Conaway was
the president and COO of CVS from 1996 to 1999. When he left CVS, he became the CEO of
Kmart (now Sears holding company) from 2000 to 2002. In the data set the following firm-year
observations are included for CVS and Kmart:

FIGURE 1
Timeline of Data Included in the Regression Analysis

Replacing F
i
in equation 2 with F
CVS
and F
Kmart
and replacing M
i
with M
Conaway
:
F
CVS
= 1 for all available years of data for CVS from 1975-2011

Firm = CVS
t =1975-1995 1996-1999 2000-2010

Firm = Kmart
t =1975-1999 2000-2002 2003-2010

CONAWAY
is COO
CONAWAY
is CEO
2000-2011
2003-2011
30

F
Kmart
= 1 for all available years of data for Kmart from 1975-2011
M
Conaway
= 1 from 1996-2002
Here the fixed effect for Charles Conaway is measured from 1996 to 2002 (over his
tenure at CVS and Kmart) and is interpreted as the incremental fixed effect of Charles Conaway
on firm accounting quality after controlling for the fixed effect of CVS and Kmart (estimated
using all available firm-years of data).
17
The estimated coefficient for M
Conaway
is therefore
Conaway’s “style”.
18

The primary benefit of the manager mobility method is that it allows the researcher to
separately identify the incremental impact of the manager on the decision variable of interest
while controlling for time invariant firm characteristics. On the other hand, the primary
limitation of the manager mobility method is that only managers who move firms and stay in the
data sample (i.e., are employed by two firms covered in the Execucomp universe) can be
included in the sample. While this approach provides the cleanest sample for testing the
incremental impact of managers, to the extent that managers who move between Execucomp
firms are fundamentally different from managers who either stay at one firm or who move to a
firm not in the sample, the generalizability of the results of tests that use the manager mobility
method may be limited. The next two methods employed in this study help to alleviate this
limitation of the manager mobility method by increasing the sample of managers that can be
included in the dataset.

17
In this study I also control for the innate economic determinants of accrual quality as discussed in Dechow and
Dichev (2002) and implemented in Francis et al. (2004). Additionally, I also control for time fixed effects.

18
Interestingly, the incremental manager-specific effect for Conaway is in the top quartile of aggressive “styles.” In
2010, Conaway settled with the SEC for $5.5 million in response to allegations of fraudulent activity associated with
Kmart Corp (Wachtel, 2010).
31

3.1.2 The Manager Connectedness Method
The second method employed in this study is the manager connectedness method. The
manager connectedness method was developed by Abowd et al. (1999), who use this method to
study employer -employee relationships in a labor economics context. This method utilizes the
mobility of managers and expands the sample to include all managers that are “connected”.
Abowd et al. (1999) prove that connectedness is a necessary and sufficient condition for separate
identification of person and firm fixed effects. Thus the individual mobility of any one person
while sufficient is not a necessary condition.

FIGURE 2
Manager Connectedness Example

The manager connectedness method works by grouping all managers and firms in
separate groups based on commonalities between managers and firms. For example, in figure 2,
Firm 1-3 and Managers A-G are connected by Manager A and Manager E. This group is labeled
“Group 1”. To form groups the researcher can implement the following procedure:
1) Choose an arbitrary manager in the dataset - Manager A

2) Identify all of Manager A’s employers - Manager A: Firm 1 and Firm 2

32

3) At each firm identify all other managers - Firm 1: Manager B and Manager C, Firm 2:
Manager D and Manager E

4) Repeat step 2 for Manager B - E - Manager B: Firm 1, Manager C: Firm 1, Manager
D: Firm 2, Manager E: Firm 2 and Firm 3

5) Repeat Step 3 for Firm 3 - Firm 3: Manager F and Manager G

6) Repeat steps 2 and 3 until no new managers or firms can be added - This creates the
first group (Group 1)

7) Repeat steps 1 – 6 until the entire dataset is categorized into disjoint groups.
From the diagram it is apparent that while there is movement between members of the
same group, there is no movement across groups such that every manager and every firm belongs
to exactly one group.
19

The benefit of the manager connectedness method is that managers do not have to move
firms in order to identify their fixed effect on the dependent variable of interest. This is because
the firm fixed effect can be calculated using the managers that move between firms, allowing the
researcher to disentangle the firm fixed effect for all manager in the sample irrespective of
whether the manager moves to another firm. Managers must simply be categorized in one of the
groups, i.e., at least someone at the managers’ firm must move between firms. This relaxation of
the manager mobility methodology allows a significantly larger number of manager fixed effects
to be identified (although the number of firms in the sample stays the same). In the above
diagram, under the manager mobility method, only Firms 1-5 and Managers A, E, and I (only
those managers that change firms) would be included in the sample. On the other hand, under the
manager connectedness method, Firms 1-5 and all of the Managers A-K would be included in the
sample.

19
Practically, this method can be implemented in STATA using the felsdvreg command developed by Cornelissen
(2008). Note that for each group, one fixed effect with be dropped in the estimation of the all other fixed effects.

33

To illustrate this methodology using the example from the discussion of the manager
mobility method, consider the connectedness associated with Charles Conaway. Through his
employment at CVS Caremark, Conaway can be connected to 11 other individuals (none of
whom move to another firm in my sample). Through his employment at Kmart, Conaway is
connected to 30 individuals, of which 5 other individuals move to other firms and remain in my
manager mobility sample.
20
These 6 individuals are directly connected to 10 different firms.
Next, I would look at all individuals at these other firms and see if they are connected to still
more firms. When no more individuals can be added to the group, the connectedness group
which started with Charles Conaway would be complete.
21

The primary advantage of the connectedness method over the manager mobility method
is that the connectedness method allows the researcher to separately identify the incremental
impact of individual managers for a larger number of managers, including managers that never
move firms and managers who move to firms that are not in the researcher’s sample. This
method could be especially useful if the researcher is interested in studying within-firm
movement such as the impact of promotion on firm decisions.
22
While this method helps to
increase the generalizability of the results documented in the manager mobility sample, the
connectedness method also has a few limitations. The first potential limitation of this method is
that it works best when there is a large amount of connectedness in the sample.
23
The second

20
I require non-movers to be present in the firm for at least three years to be consistent with the data requirements of
the manager mobility sample.

21
Note that starting with any individual (mover or non-mover) in the group would lead to the same connected group.

22
Graham et al. (2012) note that using the connectedness method allows them to separately identify the pay increase
due to promotion after controlling for the person-specific effect on compensation.

23
Andrews, Gill, Schank, and Upward (2008) note that in order to limit estimation bias in the firm fixed effects
(which are intuited by the mobility sample) the mobility sample should be relatively large. This limited mobility bias
is present in any method that relies on movement of individuals to different firms.
34

limitation is also a limitation in the manager mobility method, namely, only firms that have
managers who move to another firm in the sample are included in the analysis. Thus, while both
the manager mobility method and the connectedness method are limited by the fact that they
require managers (or at least some managers) to move between firms, the last method considered
in this paper, the spell method, does not make this restriction.

3.1.3 The Spell Method
The final method employed in this study is the spell method as introduced by Abowd et
al. (1999). The spell method includes dummy variables for each firm-manager combination in a
dataset. Each firm-manager combination is defined as the “spell” for a given manager at a given
firm. Thus, the coefficient on the “spell” can be interpreted as the time invariant fixed effect of
the manager-firm combination on the dependent variable of interest. While this method allows
the researcher to capture the time invariant impacts of firms and managers on the outcome
variable of interest, it does not allow the researcher to document the relative importance of
managers because the manager and firm fixed effects are perfectly collinear.
The primary benefit of the spell method is that it provides the researcher the largest
sample to study. In contrast to the manager mobility and connectedness methods, the spell
method does not require that managers move between firms, nor does it require the firms in the
sample to hire a manager that has moved between firms. If the researcher is primarily interested
in studying the effect of an independent variable on the dependent variable and wants to control
for time-invariant heterogeneities among firms and managers, rather than study the specific
heterogeneities per se, then the spell method provides a solution to controlling for unobserved
time-invariant characteristics without limiting the generalizability of the sample. Wooldridge
35

(2002) notes that including fixed effects limits the problem associated with omitted variable
leading to consistent, unbiased and generally efficient estimates of coefficients on the
independent variable.
24

The primary limitation of the spell method is that the specific impact of the manager is
not separable from the impact of the firm on the dependent variable. Since I am interested in
identifying the specific impact of managers (incremental to the firm) on accounting quality, I use
the first two methods to conduct my primary analysis and, in supplementary analyses, use the
spell method to provide evidence of my study’s generalizability to a larger group of managers
and firms.

3.2 Sample and Data Selection
Most of the data in this study comes from the intersection of Compustat and ExecuComp
data. Compustat contains all of the firm specific variables used in the primary analysis for this
study. In subsequent tests I use institutional ownership and block holder data from Thompson
Reuter’s 13F holdings. I gather board characteristics from RiskMetrics and I use SDC Platinum
to identify merger activity.
In order to be included in the data set firms must have available data for each measure
(this is explained in detail for each variable in the variable measurement section). ExecuComp
begins in 1992 and tracks the CEO and the top 4 highest paid executives in the firm. Starting in
2006, firms must also report the compensation of the CFO. I include manager information from

24
It is important to note that any fixed effects only control for time invariant or slow moving heterogeneity (either
firm or manager). Fixed effects regressions do not capture heterogeneity from effects which are time variant. I
attempt to control for this in later analyses.
36

ExecuComp from 1992-2011 and firm information from Compustat from 1975-2011.
25
Table 1
outlines the data availability for each method used in this paper as compared to the overall
ExecuComp database. The primary differences between my samples and the universe of
ExecuComp occur because there are 1) missing data from Compustat and 2) data restrictions
require every manager to be in employed at the firm for at least three years.
For the spell method, I include all firm-year observations for firms with available data
and which employ at least one manager covered in the ExecuComp database for at least three
years. As reported in Table 1, the spell sample covers more firms and managers than the other
two methods. Specifically the spell method contains 139,837 observations with 2,235 unique
firms and 20,011 unique managers.
The manager mobility sample (MDV) contains 720 unique managers and 880 firms with
a total of 20,615 observations. The average manager in this dataset was at each firm for 5 years
and was in the dataset for approximately 10 years. To identify this group each manager must be
employed in at least two firms for at least three years.
26
The manager mobility sample was then
merged with information from SDC Platinum to eliminate managers that were classified as
changing firms when in fact the two firms had merged. The connectedness sample (AKM)
consists of the same 880 firms as the manager mobility sample, but is expanded to include all
managers employed in those 880 firms who do not move to another firm in my dataset. The
inclusion of these “non-movers” increases the number of managers to 12,946 and the total
number of observations to 72,401.

25
Consistent with Bertrand and Schoar (2003) I include all firm years with available data to better estimate the firm
fixed effects.

26
There are 29 managers in this sample that are employed at three firms for at least three years at each firm.

37

3.3 Variable Definitions
The primary dependent variable of interest in this study is accrual quality as defined by
Dechow and Dichev (2002) and modified by Francis et al. (2004, 2005). Accrual quality often
referred to in the literature as accounting quality or earnings quality is just one measure of
earnings quality as discussed in the motivation section.
27
In addition, prior literature has noted
that the accrual quality measure is systematically related to certain firm characteristics (Dechow
& Dichev, 2002; Francis et al., 2004). I describe the measurement for each variable in the
subsequent section.
28

3.3.1 Measuring Accrual Quality
Accrual quality is measured using a modified version of the Dechow and Dichev (2002)
measure of accrual quality as specified in Francis et al. (2005):

(3)

where:

= firm j’s total current accruals in year t (change in current assets less current
liabilities less cash plus short term debt)
29

= firm j’s average total assets in year t and t-1

= Cash flow from operations in year t (net income before extraordinary items
less total current accruals plus depreciation)

27
In unreported results, I also look at different dimensions of earnings quality including earnings persistence,
predictability and smoothness. Results are qualitatively similar.

28
Detailed descriptions of all variables used in this study are in the Appendix.

29
I use the indirect or balance sheet approach to estimate accruals since I require data prior to 1988 when the
statement of cash flow became available. I rerun the main analysis in the robustness section using the direct
approach for calculating cash flows and find similar results (Hribar & Collins, 2002).
38

= firms j’s change in sales revenues between year t-1 and year t

= firm j’s gross property plant and equipment in year t

The original Dechow and Dichev (2002) model focuses on short-term working capital
accruals and models them as a function of past, present, and future cash flows. The intuition
behind the original model is that accruals anticipate future cash collection/payment and reverse
when cash previously recognized in accruals is received/paid (Dechow & Dichev, 2002; Dechow
et al., 2010). The last two terms in equation 3 are the modification to the Dechow and Dichev
model (Francis et al., 2005; McNichols 2002). Here the change in sales revenue is included to
capture growth and gross PPE allows for depreciation.
30

I calculate equation 3 in the cross-section by year and by industry using the Fama and
French (1992) 48 industry classification for all industries with at least 20 firms in year t.
31
The
accrual quality for firm j in year t is then the standard deviation of the firm’s residuals in
equation 3 from year t-4 to year t (

32
Under this specification of accruals quality,
large values of AQ indicate poorer accrual quality.
Table 2 Panel A reports summary statistics for the Accrual Quality measure for the entire
sample of ExecuComp with available information to calculate the Accrual Quality measure over
the sample period from 1975 to 2011. The mean AQ for these firms is 0.047 and the median is
0.034. These numbers are consistent with the numbers reported by Francis et al. (2005) who
report a mean of 0.0442 and a median of 0.0313 for a sample of firms from 1970-2001. The

30
Francis et al. (2005) show that including growth in revenues and PPE increases the R
2
from a mean of 39% to a
mean of 50% for equation 3.

31
Defond and Jiambalvo (1994) estimate a modified version of the Jones model (Jones 1991) by industry rather than
firm to lessen firm-year requirements. Francis et al. (2005) also use this approach in estimating the regression in
equation 3.

32
I winsorize AQ at the 1
st
and 99
th
percentiles.
39

mean and median Accrual Quality for the manager mobility sample is 0.031 and 0.025
respectively. The sample of firms included in the manager mobility sample do have a statistically
significantly lower mean AQ or on average higher accounting quality than the overall sample of
firms in ExecuComp. This finding is consistent with the finding in Dechow and Dichev (2002)
who note that AQ is inversely related to firm size, as the firms included in the manager mobility
sample are slightly larger than the average firm in ExecuComp.

3.3.2 Measuring Innate Economic Determinants
Dechow and Dichev (2002) show that firms with larger standard deviations of residuals
from the accruals model (i.e. higher abnormal accruals) are smaller firms with less persistent
earnings, longer operating cycles, larger accruals, more volatile cash flow, accruals and earnings,
and are more likely to report a loss. To control for these firm-specific characteristics that impact
the accrual quality measure, Francis et al. (2004) controls for these “innate economic
determinants” by regressing accounting quality (

on the following eight innate
economic determinants:
33

(4)

where:

= Accounting quality from equation 3

= the log of total assets for firm j at time t

33
Francis et al. (2004) augment the original measures proposed by Dechow and Dichev (2002) by including
measures of intangibles intensity, absence of intangibles, and capital intensity to capture cross sectional differences
documented in prior literature between these firm characteristics and various measures of earnings quality (Baginski,
Branson, Lorek, & Willinger, 1999; Francis & Schipper, 1999; Lev, 1983; Penman & Zhang, 2002).
40

= Cash flow variability: standard deviation of firm j's rolling five year
cash flow from operations, scaled by assets

= Sales variability: standard deviation of firm j's rolling five year sales
revenue, scaled by assets

= Operating cycle: log of the sum of firm j's days accounts receivables
and days inventory

= Incidence of negative earnings realizations: proportion of losses for
firm j over the prior five years

= Intangible intensity: sum of firm j’s R&D Expenses and Advertising
Expenses as a proportion of sales revenue at time t

= Intensity Dummy: 1 if intangibles intensity is 0

= Capital intensity: net book value of PPE scaled by total assets for
firm j at time t

Table 2 Panel B provides descriptive statistics for the above innate economic
determinants for the full ExecuComp sample. These means are largely consistent with the means
reported in Francis et al. (2005). Table 2 Panel B also provides descriptive statistics for the
innate economic determinants for the sample of firms used in this study. In general the innate
economic determinants across the various subsamples of ExecuComp are consistent with the full
ExecuComp database, with the exception of the firms used in my study having lower incidence
of negative earnings realizations. Table 3 provides correlation tables between accounting quality
measures and economic determinants of accounting quality which are consistent with prior
literature.

41

SECTION 4: RESULTS

4.1 Individual Managers and the Cross-Sectional Variation in AQ
To understand whether, and to what extent, individual managers incrementally explain
the cross-sectional variation in AQ (H1), I use the three methods described in section 3: the
manager mobility method, the manager connectedness method, and the spell method. Using the
manager mobility sample, I estimate the following model:

(5)

Where the dependent variable

is the accrual quality measure as defined in Francis et
al. (2005) and described in the variable measurement section. Economic Determinants are the
time-varying firm specific economic determinants discussed in the variable measurement
section, Year
t
is an indicator variable that equals 1 in year t and 0 otherwise, Firm
i
is an indicator
variable equal to 1 for firm i and 0 otherwise, and the variable of interest is Manager
m
is an
indicator variable equal to 1 for manager m and 0 otherwise and which captures a manager’s
incremental impact on firm accounting quality.
The results from running various regression specifications of equation 5 using the
manager mobility method are presented in Table 4. To see if manager-specific effects impact AQ
incrementally to known economic determinants, I run regressions with and without controlling
for manager fixed effects and use robust standard errors. I find that the adjusted R
2
for the
regression without manager fixed effects is 63.80%. Once manager fixed effects are included in
the regression specification, the adjusted R
2
increases to 66.80%. This 4.75% increase in the
adjusted R
2
(or a 3.03% increase in the prior adjusted R
2
) provides evidence that the manager
42

specific effect is providing additional explanatory power. This result is striking when comparing
the incremental increase in the adjusted R
2
(3.03%) to the incremental adjusted R
2
from a
regression of AQ on economic determinants, time and industry fixed effects (untabulated) which
results in a small increase in the adjusted R
2
of 0.6%. This finding suggests that the economic
importance of including manager fixed effects in as a determinant of AQ is more important than
including industry fixed effect. Another way to evaluate the importance of manager fixed effects
versus the importance of firm fixed effects in explaining AQ is to look at a Theil decomposition
(Theil, 1967). Comparing the explanatory power of a regression of AQ on economic
determinants, year and firm fixed effects to a regression of AQ on economic determinants, year,
and manager fixed effects to a regression of AQ on economic determinants, year, firm, and
manager fixed effects is that the firm fixed effects account for 20% of the explained variation
while the manager fixed effects account for 3% of the explained variation and the economic
determinants and year fixed effects account for 44% of the explained variation.
Table 4 also reports the F-statistic that tests whether or not the manager fixed effects are
jointly different than zero. The F-stat of 4.73 indicates that the manager fixed effects are jointly
different than zero. One concern with using the F-statistic as a measure of statistical significance
is the possibility of biased test statistics (Wooldridge, 2002). To alleviate this concern, I
randomize the data and bootstrap the resulting distribution of manager fixed effects and find that
my results still hold.
It is important to document that these manager fixed effects do not occur by chance.
Table 4 Panel B shows that 51.88% of the observations are significantly different at the 10%
level and that 45.20% of the observations are significant at the 5% level. This is higher than
expected by chance at the 10% and 5% level respectively. If the coefficients on the manager
43

fixed effects are capturing the managers’ impact on AQ, then it is important that then cross-
sectional variation in the coefficients capture a large variation in AQ. This is indeed the case; the
mean effect is 0.0004 while the median effect is -0.0021 and the difference between the 25
th

percentile and the 75
th
percentile is 0.0126 which indicates that there is significant variation in
the distribution of the manager fixed effects and represents about 41% of the mean AQ.
One concern with the manager mobility method is that managers that move to another
firm may be different from managers that do not move firms. To address this concern, I rerun the
above analysis using the spell method and the manager connectedness method. Table 5 reports
the adjusted R
2
for regressions that include firm fixed effects and firm and manager fixed effects.
Including “spell” effects or firm-manager effects increases the adjusted R
2
by 6.39% as
compared to a regression of accounting quality on economic determinants, time, and firm fixed
effects. This provides additional evidence that manager style is an important determinant of AQ
for a much larger sample of firms and managers.
Table 4 Panel C reports results from the manager connectedness sample, which allows for
wider identification of manager fixed effects as explained in section 3. I find that the F-test that
person and firm effects are equal to zero is rejected and that both firm and person fixed effects
are jointly different than zero. In addition, using this method I am able to decompose the
explained variation (the adjusted R
2
) from the regression of AQ on economic determinants, time,
firm, and manager fixed effects to understand the relative importance of each factor in
determining AQ. I find that about 20% of the explained variation in AQ is attributable to the
manager fixed effect, which is slightly higher than the 15% attributable to the firm fixed effect.
Taken together, all three methods provide evidence of the importance of manager-specific effects
impacting firm AQ.
44

4.2 Change in AQ When Firms’ Switch Manager Type
The first set of tests provided evidence that managers do incrementally impact AQ. The
next set of analyses is to investigate if managers apply a consistent style across firms and if that
style leads to changes in AQ as firms switch manager types over time (H2). To classify the style
of the manager, I take the coefficient estimates from the manager fixed effects in equation 5 and
sort them from largest to smallest. The managers with the largest positive impact on AQ (75
th

percentile and higher) are classified as “aggressive” and the managers with the largest negative
impact on AQ (25
th
percentile and lower) are classified as “conservative”. Likewise, the middle
50% of managers are classified as “moderate”. If firms switch from employing aggressive
(conservative) managers to conservative (aggressive) managers, then AQ should increase
(decrease) in response to the new type of manager.
Table 6 reports the results when firms switch manager types. When firms previously
employed a conservative manager and then hire an aggressive manager there is a subsequent
increase in AQ with the new manager (i.e., there is a decrease in the accounting quality of the
firm). This change (-0.0645) is statistically different than zero and represents a 94% decrease in
accounting quality.
34
Likewise, when a firm initially employs an aggressive manager and
subsequently hires a conservative manager there is a decrease in AQ (i.e., and increase in
accounting quality) that is statistically significant and represents a change of 0.0259 or a 32%
increase in accounting quality.
35
To illustrate that changes in managers do not always lead to
changes in firm accounting quality, I look at what happens when a firm switch managers but
both managers are moderate managers. Table 6 shows that when firms switch from one moderate

34
The mean AQ for firms that switch from a conservative manager to an aggressive manager goes from 0.0685 to
0.1331, resulting in a decrease in AQ by 94%.

35
The mean AQ for firms that switch from an aggressive manager to a conservative manager goes from 0.0798 to
0.0539, resulting in an increase in AQ by 32%
45

manager to another moderate manager that there is no statistically significant change in
accounting quality.
36

The classification of managerial style requires the use of data for the full time period and
as a result, the in-sample test results in Table 6 could suffer from ex-post bias. To rule out ex-
post bias, I use a subsample of 29 managers that move to a third firm as a holdout sample. Table
7 provides results for managers that move to a third firm. Here “style” is measured over the first
two firms and the 3
rd
firm is used as a holdout sample to see if the manager imposes a consistent
style at the 3
rd
firm. Results indicate that conservative manager have lower AQ or higher
accounting quality at the 3
rd
firm than aggressive managers, however, these results need to be
interpreted with caution since they are not statistically significant potentially due to the very
small sample size (29 managers) thus resulting in a low power test.

4.3 The Role of Contracts and Governance
The final set of analyses investigates the mechanism by which managers’ individual style
is captured in a firm (H3). Specifically, I examine differences in compensation and governance
for conservative and aggressive managers. In Table 8, I provide summary statistics on firm,
governance, and compensation information between conservative and aggressive managers to
see if there are any discernible differences between these types of managers and the firms that
employ them. Table 8 shows that conservative managers tend to work at slightly larger firms
with higher accounting quality. In addition, these firms pay out slightly more in dividends but
have the same level of monitoring. One other difference is that conservative managers have
slightly higher compensation in the form of salary and bonus.

36
The mean AQ for firms that switch from one moderate manager to another moderate manager goes from 0.0187 to
0.0212. This change is not statistically significant.
46

In Table 9, I rerun the main analysis of equation 5, but now I include controls for
governance (Gindex)
37
and compensation (log of total salary + bonus + equity based
compensation). Table 9 confirms that variation in firm governance practices does not explain the
relationship between manager fixed effects and accounting quality. Adding governance
(measured as the Gindex) as an independent variable, decreases the adjusted R
2
from 74.75% to
74.06%. However, including compensation in the model does increase overall model fit from
62.97% to 73.15%, but does not nullify the impact of manager fixed effects on accounting
quality and the adjusted R
2
increases to 74.75%. This analysis provides some evidence that
managers with different AQ styles do receive different compensation, but that compensation
contracts to not fully mitigate the relationship between individual managers and their role in
determining firm AQ.

37
Gindex is a composite index developed by Gompers, Ishii and Metrick (2003) to measure firm governance.
47

SECTION 5: ROBUSTNESS TESTS

In this section I describe alternative specifications for measuring accruals as described in
Hribar and Collins (2002), Ogneva (2008), and Dechow et al. (2010). I also describe in detail
placebo tests conducted to ensure that significance of manger fixed effects is not due to bias in
the F-Stat. Lastly, I consider the relative importance of different types of managers along with
examining AQ around the event of a manager leaving a firm.

5.1 Direct approach to calculating accruals
Hribar and Collins (2002) find that the measurement of accruals is sensitive to whether
accruals are measured using an indirect approach (balance sheet approach) or a direct approach
(statement of cash flows approach). They find that the difference is especially magnified for
firms that engage in mergers and acquisitions or have a large amount of discontinued operations
or foreign currency translation. This difference arises because these non-operating events impact
current assets and liabilities but have no earnings impact and so are erroneously classified as
accruals under the balance sheet approach (Collins & Hribar, 2000; Hribar & Collins, 2002).
This difference could be a problem for the firms in this study, as they are large firms that most
likely engage in these non-operating activities.
Many prior studies that calculated accruals using the balance sheet approach note that the
statement of cash flows is only available starting in 1988 and would therefore significantly
restrict the sample size (Francis et al., 2004). While using the direct approach to calculating
accruals decreases my sample size by more than half, I rerun equation 4 using the manager
mobility sample restricted by the new data requirements. Table 10 reports the results and shows
48

that there is an overall increase in adjusted R
2
of 3.39% and that the F-Stat rejects that the
manager effects are jointly equal to zero.

5.2 Scaled Accounting Quality
Ogneva (2008) notes that it is important to scale the Dechow and Dichev (2002) measure
of accrual quality by the average over that last five years of a firms absolute value of accruals.
Table 11 reports the results from using this alternative measure of accounting quality. The
overall adjusted R
2
from these regressions is lower that than using the traditional measures of
accounting quality. However, there is a 6.93% increase in adjusted R
2
when manager fixed
effects are included in the regression of economic determinants, firm and time fixed effects.

5.3 Placebo Tests
Fee et al. (2013) note that randomization or bootstrapping data and comparing the
bootstrapped distribution to the F-stat calculated from the joint hypothesis test that the fixed
effects are jointly equal to zero is a way to alleviate the concern of biased F-tests as explained in
Wooldridge (2002). Specifically, there could be bias in the F-stats if there are significant
amounts of serial correlation in the data. To address this problem I take the distribution of the
AQ measure and of the economic determinants and use the distributional properties of the data to
create a new data set. I then randomize managers at firms and bootstrap the distribution of
manger fixed effects. I find that compared to the bootstrapped data my manager fixed effects
from the main analysis are still jointly different than zero.
As a second robustness check I use the manager mobility sample and randomly assign
managers to firms within this sample. I then calculate the F-stat for these placebo mangers (and
49

repeat the process 5,000 times). I find that compared to the resulting distribution of F-Stats that
my main analysis still holds.

5.4 CEOs and CFOs vs. All Top Managers
Section 4 considers the importance of all top managers at a firm. This section looks at
differences in the job title of the top managers to understand if some types of managers are more
important in impacting the AQ of the firm. I classify a manager as a CEO or a CFO if the
manager was ever a CEO or a CFO at any firm in my manager mobility sample. I find that of the
managers in my dataset that 447 managers were either a CEO or a CFO and 267 were never a
CEO or CFO. Tables 14 through 17 report the regression results when the managers are CEOs,
CFOs, CEO or CFO, Neither a CEO or CFO respectively. The increase in adjusted R
2
for
including the manager fixed effects in AQ regression are 4.4%, 5.0%, 4.6% and 3% respectively.
When the manager was ever a CFO at either firm, the increase in the adjusted R
2
is the largest,
however manager fixed effects increase the adjusted R
2
regardless of the title of the top manager.
This lends support to the idea that all top managers are important in determining the AQ of a
firm.

5.5 Signed Discretionary Accruals
This section considers signed discretionary accruals rather than the five year rolling
standard deviation of discretionary accrual. Discretionary accruals are calculated as the residual
from the regression:

(6)

where:
50

= firm j’s total current accruals in year t (change in current assets less current
liabilities less cash plus short term debt)

= firm j’s average total assets in year t and t-1

= firms j’s change in sales revenues between year t-1 and year t

= firm j’s gross property plant and equipment in year t

Tables 16 and 17 report the results of discretionary accruals (averaged over 3 or 5 years
respectively) regressed on economic determinants, year, firm, and manager fixed effects. Table
16 shows that adding manager fixed effects to a regression of average discretionary accruals
(calculated over 3 years) on economic determinants, time and firm effects increases the adjusted
R
2
from 23.88% to 26.67% for a 2.79% increase. Table 17 shows that adding manager fixed
effects to a regression of average discretionary accruals (calculated over 5 years) on economic
determinants, time and firm effects increases the adjusted R
2
from 38.34% to 42.71% for a
4.37% increase.

5.5 Changes in Firm AQ around Manager Departures
This section examines what happens to the AQ of a firm when a manager leaves a firm.
Graphs 1 through 5 look at the AQ and discretionary accruals of a firm centered on the departure
of a manager. Graph 1 shows the annual AQ calculation for firms three years prior to a manager
leaving and three years after a manager departs. Accounting quality increases for firms after the
departure of an aggressive manager, while the accounting quality for the average firm does not
change drastically after the departure of a manager. Graphs 2 and 3 tell a similar story for the
signed discretionary accruals of a firm.

51

SECTION 6: CONCLUSION

In this study I examine whether individual managers play a unique and significant role in
determining accounting quality (AQ). I find that managers do have a statistically and
economically significant role in determining AQ, as evidenced by the incremental impact of
manager fixed effects being of similar magnitude to the relative impact of firm fixed effects. I
also provide evidence that changes in AQ are driven by idiosyncratic differences between
managers rather than in response to a structural shift in firm policy. In addition, I show that that
compensation and governance mechanisms do not fully attenuate the relationship between AQ
and individual managers. These findings add to the literature on the determinants of accrual
quality, by separating the firm characteristics from the unique characteristics of the manager and
showing that who manages the firm impacts firm AQ.
This study also contributes to the growing “managerial style” literature in finance and
accounting, which has provided evidence that managerial style impacts corporate decisions
(Bertrand & Schoar, 2003), compensation (Graham et al., 2012), voluntary disclosure (Bamber et
al., 2010; Davis et al., 2012; Yang, 2012;), tax avoidance (Dyreng et al., 2010), and various
reporting choices (Ge et al., 2011). The contribution to this literature is twofold. First, I
document that who the manager is impacts the overall accounting quality of a firm and that this
impact is three times as large as the impact of industry fixed effects and comparable to firm fixed
effects. In addition, I find that the overall explained variation of accounting quality is due more
to the manager fixed effect than the firm fixed effect. I also document that the cross-sectional
variation in managers’ impact on accounting quality is not fully explained by firm governance or
compensation contracts as implied by traditional agency theory. These results are an important
52

addition to the managerial style literature because they show that managers’ impact to firm
choices extends beyond specific investment choices to impact the overall accounting information
environment of the firm. Second, this study adds to the managerial style literature in accounting
by introducing an alternative methodology; the manager connectedness method (Abowd et al.,
1999). This new method provides a way for researchers to capture the individual effects of
managers on decision outcomes without requiring the manager to move firms. This improvement
to the manager mobility method of Bertrand and Schoar (2003) makes it possible to study the
dynamics of managers within the same firm even if some managers are only employed by one
firm.

53

REFERENCES

Abowd, J., Kramarz, F., & Margolis, D. (1999). High wage workers and high wage firms.
Econometrica, 67, 251-333.
Andrews, M., Gill, L., Schank, T., & Upward, R. (2008). High wage workers and low wage
firms: Negative assortative matching or limited mobility bias? Journal of the Royal
Statistical Society: Series A (Statistics in Society), 171(3), 673-697.
Baginski, S., Branson, B., Lorek, K., & Willinger, G. (1999). The relationship between economic
characteristics and alternative annual earnings persistence measures. The Accounting
Review, 74, 105-120.
Ball, R., & Shivakumar, L. (2006). The role of accruals in asymmetrically timely gain and loss
recognition. Journal of Accounting Research, 44, 207-242.
Bamber, L., Jiang, J., & Wang, I. (2010). What's my style? The influence of top managers on
voluntary corporate financial disclosure. The Accounting Review, 85(4), 1131-1162.
Basu, S. (1997). The conservatism principle and the asymmetric timeliness of earnings. Journal
of Accounting and Economics, 24, 3-37.
Bebchuk, L. A., Cohen, A., & Ferrell, A. (2009). What matters in corporate governance? The
Review of Financial Studies, 22(2), 783-827.
Bertrand, M., & Schoar, A. (2003). Managing with style: The effect of managers on firm
policies. Quarterly Journal of Economics, 118(4), 1169-1208.
Brainard, W. C., & Tobin, J. (1968). Pitfalls in financial model building. American Economic
Review, 58(2), 99-122.
Chalmers, K., & Godfrey, J. (2004). Reputation costs: The impetus for voluntary derivative
financial instrument reporting. Accounting, Organizations and Society, 29(2), 95-125.
Coles, J., & Li, Z. F. (2013). Managerial attributes, incentives, and performance. Unpublished
manuscript. Retrieved from http://ssrn.com/abstract=1680484
Collins, D. W., & Hribar, P. (2000). Earnings-based and accrual-based market anomalies: One
effect or two? Journal of Accounting and Economics, 29(1), 101-123.
Cornelissen, T. (2008). The Stata command felsdvreg to fit a linear model with two high-
dimensional fixed effects. The Stata Journal, 8(2), 170-189.
54

Davis, A., Ge, W., Matsumoto, D., & Zhang, J. (2012). The effect of managerial “style” on the
tone of earnings conference calls. CAAA Annual Conference 2012. Retrieved from
http://ssrn.com/abstract=1982259
Dechow, P. (1994). Accounting earnings and cash flows as measures of firm performance: The
role of accounting accruals. Journal of Accounting and Economics, 18, 3-42.
Dechow, P., & Dichev, I. (2002). The quality of accruals and earnings: The role of accrual
estimation errors. The Accounting Review, 77, 35-59.
Dechow, P., Ge, W., & Schrand, C. (2010). Understanding earnings quality: A review of the
proxies, their determinants and their consequences. Journal of Accounting and
Economics, 50(2-3), 344-401.
DeFond, M., & Jiambalvo, J. (1994). Debt covenant violation and manipulation of accruals.
Journal of Accounting and Economics, 17, 145-176.
Demerjian, P., Lev, B., Lewis, M., & McVay, S. (2012). Quantifying managerial ability: A new
measure and validity tests. Management Science, 58(7), 1229-1248.
Demerjian, P., Lev, B., Lewis, M., & McVay, S. (2013). Managerial ability and earnings quality.
The Accounting Review, Forthcoming.
Dichev, I., Graham, J., Harvey, C., & Rajgopal, S. (2013). Earnings quality: Evidence from the
field. Unpublished manuscript. Retrieved from http://ssrn.com/abstract=2103384
DiMaggio, P., & Powell, W. (1983). The iron cage revisited: Institutional isomorphism and
collective rationality in organizational fields. American Sociological Review, 48(2), 147-
160.
Dyreng, S., Hanlon, M., & Maydew, E. (2010). The effects of executives on corporate tax
avoidance. The Accounting Review, 85(4), 1163-1189.
Dyreng, S., Mayew, W., & Williams, C. (2009). Religious social norms and corporate financial
reporting. AAA 2010 Financial Accounting and Reporting Section (FARS) Paper.
Retrieved from http://ssrn.com/abstract=1444839
Fama, E. F., & French, K. R. (1992). The cross-section of expected stock returns. The Journal of
Finance, 47(2), 427-465.
Fee, C. E., Hadlock, C. J., & Pierce, J. (2013). Managers with and without style: Evidence using
exogenous variation. The Review of Financial Studies, 26(3), 567-601.
Finacial Accounting Standards Board. (1978). Statement of Financial Accounting Concepts No.
1, Objectives of Financial Reporting by Business Enterprises. Norwalk, CT: FASB.
55

Finkelstein, S., & Hambrick, D. (1996). Strategic leadership: Top executives and their effects on
organizations. St. Paul, MN: West Publishing Company.
Francis, J., Huang, A. H., Rajgopal, S., & Zang, A. Y. (2008). CEO reputation and earnings
quality. Contemporary Accounting Research, 25, 109-147.
Francis, J., LaFond, R., Olsson, P., & Schipper, K. (2004). Costs of equity and earnings
attributes. The Accounting Review, 79, 967-1010.
Francis, J., LaFond, R., Olsson, P., & Schipper, K. (2005). The market pricing of accruals
quality. Journal of Accounting and Economics, 39, 295-327.
Francis, J., Olsson, P., & Schipper, K. (2006). Earnings quality. Foundation and Trends in
Accounting, 1, 259-340.
Francis, J., & Schipper, K. (1999). Have financial statements lost their relevance? Journal of
Accounting Research, 37, 319-352.
Ge, W., Matsumoto, D., & Zhang, J. (2011). Do CFOs have style? An empirical investigation of
the effect of individual CFOs on accounting practices. Contemporary Accounting
Research, 28(4), 1141-1179.
Gompers, P., Ishii, J., & Metrick, A. (2003). Corporate governance and equity prices. The
Quarterly Journal of Economics, 118(1), 107-155.
Graham, J., Li, S., & Qiu, J. (2012). Managerial attributes and executive compensation. The
Review of Financial Studies, 25(1), 144-186.
Hambrick, D. (2007). Upper echelons theory: An update. Academy of Management Review,
32(2), 334-343.
Hambrick, D., & Mason, P. (1984). Upper echelons: The organization as a reflection of its top
managers. Academy of Management Review, 9(2), 193-206.
Hambrick, D., Geletkanycz, M., & Fredrickson, J. (1993). Top executive commitment to the
status quo: Some tests of its determinants. Strategic Management Journal, 14(6), 401-
418.
Hannan, M., & Freeman, J. (1977). The population ecology of organizations. American Journal
of Sociology, 82(5), 929-964.
Hermalin, B. E., & Weisbach, M. S. (1998). Endogenously chosen boards of directors and their
monitoring of the CEO. American Economic Review, 88, 96-118.
Hitt, M., & Tyler, B. (1991). Strategic decision models: Integrating different perspectives.
Strategic Management Journal, 12(5), 327-351.
56

Hribar, P., & Collins, D. (2002). Errors in estimating accruals: implications for empirical
research. Journal of Accounting Research, 40, 105-134.
Hutton, I., Jiang, D., & Kumar, A. (2013). Corporate policies of republican managers. Journal of
Financial and Quantitative Analysis, Forthcoming. Retrieved from
http://ssrn.com/abstract=1573
Jensen, M., & Meckling, W. (1976). Theory of the firm: Managerial behavior, agency costs and
capital structure. Journal of Financial Economics, 3, 305-360.
Jones, J. (1991). Earnings management during import relief investigations. Journal of
Accounting Research, 29, 193-228.
Kaplan, S., & Zingales, L. (1997). Do investment-cash flow sensitivities provide useful measures
of financing constraints? The Quarterly Journal of Economics, 112(1), 169-215.
Khalil, K. (2004). Integrity, Shame and Self-Rationalization. Unpublished manuscript. Retrieved
from http://EconPapers.repec.org/RePEc:vas:papers:55
Koh, K. (2011). Value or glamour?An empirical investigation of the effect of celebrity CEOs on
financial reporting practices and firm performance. Accounting and Finance, 51(2), 517-
547.
Kothari, S., Ramanna, K., & Skinner, D. (2010). Implications for GAAP from an analysis of
positive research in accounting. Journal of Accounting and Economics, 50(2-3), 246-286.
Lev, B. (1983). Some economic determinants of time-series properties of earnings. Journal of
Accounting and Economics, 5, 31-48.
McNichols, M. (2002). Discussion of ''The quality of accruals and earnings: The role of accrual
estimation errors". The Accounting Review, 77, 61-69.
Ogneva, M. (2008). Accrual quality, realized returns, and expected returns: The importance of
controlling for cash flow shocks. Unpublished manuscript. Retrieved from
http://ssrn.com/abstract=1306598
Penman, S., & Zhang, X. (2002). Accounting conservatism, the quality of earnings, and stock
returns. The Accounting Review, 77, 237-264.
Salanick, G., & Pfeffer, J. (1977). Constraints on administrator discretion: The limited influence
of mayors on city budgets. Urban Affairs Quarterly, 12(4), 475-498.
Schrand, C., & Zechman, S. (2012). Executive overconfidence and the slippery slope to financial
misreporting. Journal of Accounting and Economics, 53(1-2), 311-329.
57

Smith, M., & White, M. (1987). Strategy, CEO specialization, and succession. Administrative
Science Quarterly, 32(2), 263-280.
Spender, J. (1989). Industry recipes: An enquiry into the nature and sources of managerial
judgment. Oxford, U.K.: Blackwell.
Theil, H. (1967). Economics and information theory. Chicago: Rand McNally and Company.
Thomas, A., Litschert, R., & Ramaswamy, K. (1991). The performance impact of strategy-
manager coalignment: An empirical examination. Strategic Management Journal, 12(7),
509-522.
Tihanyi, L., Ellstrand, A., Daily, C., & Dalton, D. (2000). Composition of the top management
team and firm international diversification. Journal of Management, 26(6), 1157-1177.
Wachtel, K. (2010, November 17). Former Kmart CEO is ordered to pay $5.5 million by SEC
for lying to investors. Retrieved from http://www.businessinsider.com/former-kmart-ceo-
ordered-to-pay-55-million-by-sec-for-misleading-investors-2010-11
Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data. Cambridge,
MA: MIT Press.
Yang, H. (2012). Capital market consequences of managers’ voluntary disclosure styles. Journal
of Accounting and Economics, 53, 167-184.

58

TABLE 1
Data Availability

This table provides a summary of data availability for the full sample (i.e., the ExecuComp
universe) with that of the data used in this study. “Change” represents the sample of managers that are
employed by at least two firms in the ExecuComp database. “SPELL” represents the data used to implement
the spell method. “AKM” represents the data used to implement the manager connectedness method (i.e., all
managers employed by firms with at least one manager that moves firms). “MDV” represents the data used
to implement the manager mobility method (i.e., managers employed by at least two firms in the dataset).
Specifics on the construction of each dataset are provided in Section 3.
ExecuComp Data used in this study
FULL Change SPELL AKM MDV
Total Observations 219,156 27,887 139,837 72,401 20,615

Firm-Manager 42,974 6,354 20,764 14,035 1,469

Firms 3,316 2,976 2,235 880 880

Managers 39,596 2,224 20,011 12,946 720

59

TABLE 2
Summary Statistics

This table provides summary statistics of the variables in the full sample (i.e., the ExecuComp universe), the spell method sample (SPELL), the
manager mobility sample (MDV), and the manager connectedness sample (AKM). The details of the definitions and the measurement of all of the variables are
reported in the Appendix. Panel A reports summary statistics on firm level variables. Sample size refers to the largest number of observations available for each
method where the unit of analysis is at the firm-year level. Panel B provides summary statistics for the innate economic determinant of AQ. Panel C reports
summary statistics on firm governance and executive compensation variables.
Panel A

FULL EXECUCOMP SPELL AKM/MDV
MEAN MED STD MEAN MED STD MEAN MED STD
Assets 2306 191 7898 3302 770 7757 5020 1399 11475
R&D 0.097 0.000 0.673 0.036 0.000 0.104 0.033 0.000 0.070
Leverage 0.230 0.220 0.179 0.223 0.220 0.160 0.233 0.233 0.153
Cash Holdings 0.281 0.071 0.751 0.197 0.064 0.371 0.180 0.059 0.346
Dividend Yield 0.016 0.000 0.025 0.019 0.011 0.025 0.023 0.015 0.027
Dividend Indicator 0.438 0.000 0.496 0.625 1.000 0.484 0.665 1.000 0.472
Investment 0.309 0.193 0.482 0.286 0.209 0.321 0.388 0.201 2.953
AQ 0.047 0.034 0.044 0.034 0.027 0.026 0.031 0.025 0.025
Persistence -0.315 -0.310 0.368 -0.368 -0.382 0.360 -0.356 -0.371 0.363
Predictability 1.089 0.597 1.701 1.272 0.707 1.746 1.209 0.794 1.272
Smoothness 0.734 0.700 0.381 0.685 0.645 0.359 0.676 0.641 0.351
Asymmetric TLR 0.253 0.187 0.880 0.275 0.206 0.746 0.314 0.283 0.671
ROE -0.028 0.088 0.640 0.079 0.110 0.265 0.088 0.111 0.261
ROA 0.096 0.120 0.165 0.147 0.143 0.091 0.149 0.143 0.082
Tobin's Q 1.673 1.233 1.379 1.711 1.340 1.144 1.689 1.319 1.148
KZ Index -1.546 0.459 18.187 -1.120 0.089 9.158 -0.949 0.041 7.809
Sample Size 91746 38789 17415

60

Summary Statistics
Panel B

FULL EXECUCOMP SPELL AKM/MDV
MEAN MED STD MEAN MED STD MEAN MED STD
Firm Size 5.297 5.213 2.362 6.745 6.687 1.701 7.277 7.276 1.643
Cash Flow Variability 0.104 0.076 0.098 0.072 0.058 0.056 0.067 0.054 0.049
Sales Variability 0.244 0.188 0.202 0.197 0.158 0.146 0.182 0.148 0.135
Operating Cycle 4.665 4.733 0.735 4.618 4.674 0.650 4.621 4.650 0.605
Incidence of Negative Earnings 0.614 0.008 2.368 0.047 0.000 0.217 0.027 0.000 0.143
Intangibles Intensity 0.095 0.011 0.507 0.045 0.013 0.093 0.043 0.014 0.072
Intensity Indicator 0.337 0.000 0.473 0.334 0.000 0.472 0.323 0.000 0.468
Capital Intensity 0.331 0.271 0.243 0.358 0.304 0.237 0.374 0.322 0.240

Summary Statistics
Panel C

FULL EXECUCOMP SPELL AKM MDV
MEAN MED STD MEAN MED STD MEAN MED STD MEAN MED STD
GINDEX 9 9 3 9 9 3 9 9 3 9 9 3
EINDEX 2 2 1 2 2 1 2 3 1 2 3 1
Institutional Ownership 0.595 0.640 0.293 0.672 0.695 0.238 0.689 0.716 0.229 0.689 0.716 0.229
Blockholder 0.086 0.081 0.049 0.090 0.085 0.045 0.090 0.085 0.046 0.090 0.085 0.046
Board Size 9 9 3 9 9 3 10 10 3 10 10 3
Board Meetings 8 7 3 8 7 3 8 8 4 8 8 4
Independence 0.664 0.667 0.170 0.664 0.688 0.169 0.691 0.714 0.160 0.691 0.714 0.160
Total Compensation 6.961 6.879 1.040 6.976 6.892 1.043 7.144 7.075 1.054 7.412 7.374 1.115
Salary and Bonus 6.075 6.026 0.714 6.107 6.054 0.708 6.163 6.121 0.719 6.315 6.292 0.810
Equity Based Compensation 5.981 5.979 1.497 5.986 5.982 1.499 6.209 6.202 1.498 6.530 6.530 1.487

61

TABLE 3
Correlations for Accounting Quality Measures and Economic Determinants of Accounting Quality

This table provides Pearson correlations (top) and Spearman correlations (bottom) for AQ and the innate economic determinants of AQ. The
details of the definitions and the measurement of all of the variables are reported in the Appendix.

AQ Firm Size
Cash Flow
Variability
Sales
Variability
Operating
Cycle
Incidence of
Negative
Earnings
Intangibles
Intensity
Intensity
Indicator
Capital
Intensity
AQ 1 -0.2832 0.5649 0.3221 0.0986 0.0158 0.0656 -0.1571 -0.3546
Firm Size -0.2979 1 -0.3087 -0.1773 -0.1178 -0.0287 -0.0506 0.0988 0.2280
Cash Flow
Variability 0.6093 -0.3647 1 0.3255 0.0845 0.0165 0.1469 -0.1510 -0.3178
Sales
Variability 0.3816 -0.2288 0.4524 1 -0.1322 0.0089 0.0082 -0.0440 -0.2473
Operating
Cycle 0.1937 -0.1766 0.2032 -0.0853 1 0.0079 -0.0424 -0.2362 -0.3688

Incidence of
Negative
Earnings 0.3682 -0.2964 0.4935 0.2198 0.0881 1 0.0002 -0.0049 -0.0087
Intangibles
Intensity 0.2566 -0.1178 0.2691 0.0537 0.4382 0.1398 1 -0.0250 -0.0175
Intensity
Indicator -0.2261 0.1043 -0.2030 -0.1351 -0.3271 -0.0586 -0.7884 1 0.4058
Capital
Intensity -0.4224 0.2276 -0.4091 -0.3025 -0.3606 -0.1695 -0.4256 0.3551 1

62

TABLE 4
Testing Individual Managers' Fixed Effects on Accounting Quality – Manager Mobility Method
This table provides the adjusted R
2
for the regression of

as described in Section 3.
Economic determinants include: firm size, cash flow variability, sales variability, operating cycle, incidence of negative earnings realizations, intangibles intensity, intensity
indicator, and capital intensity. Variable descriptions are provided in the Appendix. Results reported in this table are determined using the Manager Mobility Method. “Yes”
indicates that the variable(s) is/are included in the regression. Coefficient estimates are not reported in this table, but all regressions include corrections for robust standard errors.
The F-Statistic is for the test that all of the manager fixed effects are jointly equal to 0. Panel B reports summary statistics on the manager fixed effects from the regression in
Panel A column 8 (the full regression specification)
Panel A

Dependent Variable = AQ (n = 20,615)
Economic Determinants

YES YES

YES YES
Year Fixed Effects YES YES YES YES YES YES YES YES
Firm Fixed Effects

YES

YES

YES

YES
Manager Fixed Effects

YES YES YES YES

Testing Manager Fixed Effects =0

F-statistics

25.83 5.98 8.74 4.73
(p-value)

(<0.001) (<0.001) (<0.001) (<0.001)

Adjusted R
2
10.68% 65.43% 40.26% 63.80% 22.63% 69.31% 46.88% 66.83%
Improvement in Adjusted R
2

raw %

11.95% 3.88% 6.62% 3.03%
% of comparable model

111.89% 5.93% 16.44% 4.75%
Panel B

Manager Fixed Effects (n=719)

% Significant at the 10% 51.88%

% Significant at the 5% 45.20%

Mean Effect 0.0004

Median Effect -0.0021

25th Percentile 0.0053

75th Percentile -0.0072

Difference between the 25th and 75th percentile 0.0126

Difference as a percentage of the mean AQ 41.59%
63

Panel C

Relative importance in explaining the variation in AQ (Shapley Decomposition)

Manager Mobility Method Manager Connectedness Method
Cov (AQ, Econ.Deter.)/ Var (AQ) 43.04% 41.28%
Cov (AQ, MFE)/ Var (AQ) 6.18% 20.53%
Cov (AQ, FFE)/ Var (AQ) 20.33% 15.09%
Cov (AQ, Residual) / Var (AQ) 30.45% 23.09%

64

TABLE 5
Testing Individual Managers' Fixed Effects on Accounting Quality - SPELL Method
This table provides the adjusted R
2
for the regression of

as described in Section 3.
Economic determinants include: firm size, cash flow variability, sales variability, operating cycle, incidence of negative earnings realizations, intangibles intensity, intensity
indicator, and capital intensity. Variable descriptions are provided in the Appendix. Results reported in this table are determined using the Spell Method. “Yes” indicates that the
variable(s) is/are included in the regression. Coefficient estimates are not reported in this table, but all regressions include corrections for robust standard errors.

Dependent Variable = AQ (n = 139,837)
Economic Determinants

YES YES

YES
Year Fixed Effects YES YES YES YES YES YES
Firm Fixed Effects

YES

YES

Firm and Manager Fixed Effects

YES YES

Adjusted R
2
3.84% 59.65% 36.25% 64.91% 67.14% 71.30%
Improvement in Adjusted R
2

raw %

7.49% 6.39%
% of comparable model

12.56% 9.84%

65

TABLE 6
Change in Firm Measures and Executive Compensation When Firms Switch Manager Type
This table provides a summary of the difference in the mean values when a firm switches manager types. Column 1 reports the difference in the means when a firm switches
from a conservative manager to an aggressive manager. Column 2 reports the difference in the means when a firm switches from an aggressive manager to a conservative manager.
Column 3 reports the difference in the means when a firm switches from a moderate manager to a different moderate manager. Aggressive, moderate and conservative are defined as
the 75
th
, middle 50
th
and 25
th
percentile of coefficients on manager fixed effects from the regression of

. Variable definitions are in the Appendix. *, **, *** represents statistical significance at the 10%, 5%, and 1% level respectively.

(1) (2) (3)

Conservative to Aggressive Manager Aggressive to Conservative Manager Moderate to Moderate Manager
# of observations 30 60 102
AQ -0.0645*** 0.0259*** -0.0024
ROA -0.0242 0.0168 0.0009
GINDEX 0.2667 -0.8426 -0.8800
EINDEX 0.4154 -0.2130 -0.1615
R&D -0.0219 0.0016 -0.0003
LEVERAGE 0.1640** 0.0503 -0.0089
Cash Holdings -0.2262 -0.0992 0.0512
Dividend Yield 0.0000 -0.0002 0.0055
Dividend Indicator 0.0000 0.0176 0.0000
Equity Based Compensation 2.1127** 0.4992 0.2364
Total Compensation 0.4650 -0.1405 -0.1923
Salary + Bonus -0.1917 -0.1727 0.0740

66

TABLE 7
Change in Accounting Quality as Managers Move To a 3rd Firm (n=29)
This table reports the mean AQ and the difference in the mean AQ when a manager switches to a 3
rd
firm. The manager type is calculated using the first two firms that
employ the manager. Aggressive, moderate 2, moderate 1, and conservative are defined as the 4
th
, 3
rd
, 2
nd
, and 1
st
quartile of coefficients on manager fixed effects from the
regression of

. *, **, *** represents statistical significance at the 10%, 5%, and 1% level
respectively.
Manager Classification Firm 1 and 2 AQ Firm 3 AQ Difference
Conservative 0.0169 0.0261 -0.0093
Moderate 1 0.0221 0.0336 -0.0115
Moderate 2 0.0269 0.0322 -0.0053
Aggressive 0.0603 0.0356 0.0247

67

TABLE 8
Firm and Manager Characteristics
This table provides a summary statistics for aggressive and conservative managers and the difference in the means. Aggressive, moderate and conservative are defined
as the 75
th
, middle 50
th
and 25
th
percentile of coefficients on manager fixed effects from the regression of

. Variable definitions are in the Appendix. *, **, *** represents statistical significance at the 10%, 5%, and 1% level respectively for the difference in the means of
conservative and aggressive managers and their respective firms.
Conservative Managers Aggressive Managers
MEAN MED STD MEAN MED STD Difference
AQ 0.034 0.026 0.027 0.053 0.044 0.036 -0.020***
Assets 6690 2502 12140 4923 1673 13170 1767***
Firm Age 22 19 13 22 19 13 -0.169
Manager Age 52 52 8 52 52 7 0.435
Tobin's Q 1.787 1.401 1.384 1.953 1.496 1.709 -0.166*
ROE 0.012 0.082 1.005 -0.261 0.101 7.889 0.273
ROA 0.125 0.120 0.095 0.136 0.130 0.100 -0.010*
KZ Index -1.552 0.747 22.214 -0.290 0.568 15.799 -1.261
GINDEX 9 9 3 9 9 3 -0.308**
EINDEX 2 2 1 3 3 1 -0.046
R&D 0.047 0.003 0.086 0.049 0.001 0.119 -0.001
Leverage 0.218 0.221 0.156 0.196 0.190 0.147 0.022
Cash Holdings 0.224 0.078 0.415 0.287 0.086 0.651 -0.064
Dividend Yield 0.011 0.000 0.024 0.009 0.000 0.025 0.002**
Dividend Indicator 0.460 0.000 0.499 0.456 0.000 0.498 0.004*
Equity Based Compensation 6.582 6.547 1.577 6.590 6.603 1.397 -0.008
Total Compensation 7.384 7.359 1.231 7.261 7.250 1.167 0.124*
Salary and Bonus 6.304 6.294 0.970 6.174 6.166 0.995 0.130**
SMBE 0.167 0.000 0.373 0.162 0.000 0.369 0.005
Blockholder 0.102 0.092 0.066 0.097 0.092 0.051 0.005
Institutional Ownership 0.722 0.751 0.219 0.728 0.751 0.222 -0.005
Board Meetings 8.421 8.000 3.586 8.227 7.000 4.718 0.194
Independence 0.668 0.667 0.170 0.686 0.707 0.168 -0.018

68

TABLE 9
Testing Individual Managers' Fixed Effects on Accounting Quality Controlling for Governance and Compensation
This table provides the adjusted R
2
for the regression of

as described in Section 3. Economic determinants include: firm size, cash flow variability, sales variability, operating cycle, incidence of negative earnings
realizations, intangibles intensity, intensity indicator, and capital intensity. Governance is the Gindex and Compensation is the log of salary plus bonus plus equity compensation.
Variable descriptions are provided in the Appendix. Results reported in this table are determined using the Manager Mobility Method. “Yes” indicates that the variable(s) is/are
included in the regression. Coefficient estimates are not reported in this table, but all regressions include corrections for robust standard errors. The F-Statistic is for the test that
all of the manager fixed effects are jointly equal to 0.

Dependent Variable = AQ (n = 11,381)
Economic Determinants YES YES YES YES YES YES
Governance YES

YES YES

YES
Compensation

YES YES

YES YES
Year Fixed Effects YES YES YES YES YES YES
Firm Fixed Effects YES YES YES YES YES YES
Manager Fixed Effects

YES YES YES

Testing Manager Fixed Effects =0

F-statistics

5.67 14.96 14.34
(p-value)

(<0.001) (<0.001) (<0.001)

Adjusted R
2
62.97% 73.15% 72.18% 66.58% 74.75% 74.06%
Improvement in Adjusted R
2

raw %

3.61% 1.60% 0.91%
% of comparable model

5.73% 2.19% 1.24%

69

TABLE 10
Testing Individual Managers' Fixed Effects on Accounting Quality – Statement of Cash Flows
This table provides the adjusted R
2
for the regression of

as described in Section 3.
AQ is calculated using cash flows from operations from the Statement of Cash Flows (Hribar and Collins, 2002). Economic determinants include: firm size, cash flow variability,
sales variability, operating cycle, incidence of negative earnings realizations, intangibles intensity, intensity indicator, and capital intensity. Variable descriptions are provided in
the Appendix. Results reported in this table are determined using the Manager Mobility Method. “Yes” indicates that the variable(s) is/are included in the regression. Coefficient
estimates are not reported in this table, but all regressions include corrections for robust standard errors. The F-Statistic is for the test that all of the manager fixed effects are
jointly equal to 0.

Dependent Variable = AQ (Statement of Cash Flows) (n = 6,923)
Economic Determinants

YES YES

YES YES
Year Fixed Effects YES YES YES YES YES YES YES YES
Firm Fixed Effects

YES

YES

YES

YES
Manager Fixed Effects

YES YES YES YES

Testing Manager Fixed Effects =0

F-statistics

22.04 19.92 8.34 11.12
(p-value)

(<0.001) (<0.001) (<0.001) (<0.001)

Adjusted R
2
2.61% 64.64% 35.27% 70.46% 37.99% 70.43% 52.53% 73.85%
Improvement in Adjusted R
2

raw %

35.38% 5.79% 17.26% 3.39%
% of comparable model

1355.56% 8.96% 48.94% 4.81%

70

TABLE 11
Testing Individual Managers' Fixed Effects on Scaled Accounting Quality
This table provides the adjusted R
2
for the regression of

as described in Section 3.
AQ is scaled by average over that last five years of a firm’s absolute value of accruals (Ogneva, 2008). Economic determinants include: firm size, cash flow variability, sales
variability, operating cycle, incidence of negative earnings realizations, intangibles intensity, intensity indicator, and capital intensity. Variable descriptions are provided in the
Appendix. Results reported in this table are determined using the Manager Mobility Method. “Yes” indicates that the variable(s) is/are included in the regression. Coefficient
estimates are not reported in this table, but all regressions include corrections for robust standard errors. The F-Statistic is for the test that all of the manager fixed effects are
jointly equal to 0.

Dependent Variable = scaled AQ (n = 13,626)
Economic Determinants

YES YES

YES YES
Year Fixed Effects YES YES YES YES YES YES YES YES
Firm Fixed Effects

YES

YES

YES

YES
Manager Fixed Effects

YES YES YES YES

Testing Manager Fixed Effects =0

F-statistics

9.97 6.64 9.06 6.49
(p-value)

(<0.001) (<0.001) (<0.001) (<0.001)

Adjusted R
2
17.76% 44.63% 21.05% 44.97% 36.00% 51.70% 37.46% 51.90%
Improvement in Adjusted R
2

raw %

18.24% 7.07% 16.41% 6.93%
% of comparable model

102.70% 15.84% 77.96% 15.41%

71

TABLE 12
Testing Individual CEO Fixed Effects on Accounting Quality
This table provides the adjusted R
2
for the regression of

as described in Section 3.
Economic determinants include: firm size, cash flow variability, sales variability, operating cycle, incidence of negative earnings realizations, intangibles intensity, intensity
indicator, and capital intensity. Variable descriptions are provided in the Appendix. Results reported in this table are determined using the Manager Mobility Method. “Yes”
indicates that the variable(s) is/are included in the regression. Coefficient estimates are not reported in this table, but all regressions include corrections for robust standard errors.
The F-Statistic is for the test that all of the manager fixed effects are jointly equal to 0. Panel B reports summary statistics on the manager fixed effects from the regression in
Panel A column 8 (the full regression specification)
Panel A

Dependent Variable = AQ (n = 12,753)
Economic Determinants

YES YES

YES YES
Year Fixed Effects YES YES YES YES YES YES YES YES
Firm Fixed Effects

YES

YES

YES

YES
Manager Fixed Effects

YES YES YES YES

Testing Manager Fixed Effects =0

F-statistics

19.51 6.58 8.24 5.80
(p-value)

(<0.001) (<0.001) (<0.001) (<0.001)

Adjusted R
2
9.08% 52.17% 38.02% 58.76% 26.04% 58.29% 47.70% 63.24%
Improvement in Adjusted R
2

raw %

16.96% 6.12% 9.68% 4.48%
% of comparable model

186.78% 11.73% 25.46% 7.62%
Panel B

Manager Fixed Effects (n=406)

% Significant at the 10% 46.06%

% Significant at the 5% 39.16%

Mean Effect 0.0004

Median Effect -0.0005

25th Percentile 0.0070

75th Percentile -0.0073

Difference between the 25th and 75th percentile 0.0143

Difference as a percentage of the mean AQ 47.39%
72

TABLE 13
Testing Individual CFO Fixed Effects on Accounting Quality
This table provides the adjusted R
2
for the regression of

as described in Section 3.
Economic determinants include: firm size, cash flow variability, sales variability, operating cycle, incidence of negative earnings realizations, intangibles intensity, intensity
indicator, and capital intensity. Variable descriptions are provided in the Appendix. Results reported in this table are determined using the Manager Mobility Method. “Yes”
indicates that the variable(s) is/are included in the regression. Coefficient estimates are not reported in this table, but all regressions include corrections for robust standard errors.
The F-Statistic is for the test that all of the manager fixed effects are jointly equal to 0. Panel B reports summary statistics on the manager fixed effects from the regression in
Panel A column 8 (the full regression specification)
Panel A

Dependent Variable = AQ (n = 4,821)
Economic Determinants

YES YES

YES YES
Year Fixed Effects YES YES YES YES YES YES YES YES
Firm Fixed Effects

YES

YES

YES

YES
Manager Fixed Effects

YES YES YES YES

Testing Manager Fixed Effects =0

F-statistics

16.83 13.03 10.49 10.63
(p-value)

(<0.001) (<0.001) (<0.001) (<0.001)

Adjusted R
2
11.36% 50.12% 37.76% 55.32% 23.89% 57.06% 45.35% 60.32%
Improvement in Adjusted R
2

raw %

12.53% 6.94% 7.59% 5.00%
% of comparable model

110.30% 13.85% 20.10% 9.04%
Panel B

Manager Fixed Effects (n=140)

% Significant at the 10% 45.71%

% Significant at the 5% 35.00%

Mean Effect 0.0003

Median Effect -0.0003

25th Percentile 0.0076

75th Percentile -0.0063

Difference between the 25th and 75th percentile 0.0140

Difference as a percentage of the mean AQ 46.63%
73

TABLE 14
Testing Individual CEO or CFO Fixed Effects on Accounting Quality
This table provides the adjusted R
2
for the regression of

as described in Section 3.
Economic determinants include: firm size, cash flow variability, sales variability, operating cycle, incidence of negative earnings realizations, intangibles intensity, intensity
indicator, and capital intensity. Variable descriptions are provided in the Appendix. Results reported in this table are determined using the Manager Mobility Method. “Yes”
indicates that the variable(s) is/are included in the regression. Coefficient estimates are not reported in this table, but all regressions include corrections for robust standard errors.
The F-Statistic is for the test that all of the manager fixed effects are jointly equal to 0. Panel B reports summary statistics on the manager fixed effects from the regression in
Panel A column 8 (the full regression specification)
Panel A

Dependent Variable = AQ (n = 13,494)
Economic Determinants

YES YES

YES YES
Year Fixed Effects YES YES YES YES YES YES YES YES
Firm Fixed Effects

YES

YES

YES

YES
Manager Fixed Effects

YES YES YES YES

Testing Manager Fixed Effects =0

F-statistics

19.57 6.99 8.36 6.08
(p-value)

(<0.001) (<0.001) (<0.001) (<0.001)

Adjusted R
2
9.48% 51.91% 38.61% 58.79% 26.13% 58.26% 48.11% 63.39%
Improvement in Adjusted R
2

raw %

16.65% 6.35% 9.50% 4.60%
% of comparable model

175.63% 12.23% 24.61% 7.82%
Panel B

Manager Fixed Effects (n=447)

% Significant at the 10% 46.73%

% Significant at the 5% 39.73%

Mean Effect 0.0006

Median Effect -0.0006

25th Percentile 0.0070

75th Percentile -0.0068

Difference between the 25th and 75th percentile 0.0138

Difference as a percentage of the mean AQ 45.68%
74

TABLE 15
Testing Individual (Non-CEO or CFO) Managers' Fixed Effects on Accounting Quality
This table provides the adjusted R
2
for the regression of

as described in Section 3.
Economic determinants include: firm size, cash flow variability, sales variability, operating cycle, incidence of negative earnings realizations, intangibles intensity, intensity
indicator, and capital intensity. Variable descriptions are provided in the Appendix. Results reported in this table are determined using the Manager Mobility Method. “Yes”
indicates that the variable(s) is/are included in the regression. Coefficient estimates are not reported in this table, but all regressions include corrections for robust standard errors.
The F-Statistic is for the test that all of the manager fixed effects are jointly equal to 0. Panel B reports summary statistics on the manager fixed effects from the regression in
Panel A column 8 (the full regression specification)
Panel A

Dependent Variable = AQ (n = 8,139)
Economic Determinants

YES YES

YES YES
Year Fixed Effects YES YES YES YES YES YES YES YES
Firm Fixed Effects

YES

YES

YES

YES
Manager Fixed Effects

YES YES YES YES

Testing Manager Fixed Effects =0

F-statistics

30.05 7.02 12.10 6.94
(p-value)

(<0.001) (<0.001) (<0.001) (<0.001)

Adjusted R
2
10.18% 55.97% 43.13% 63.47% 26.95% 59.47% 50.52% 66.48%
Improvement in Adjusted R
2

raw %

16.77% 3.50% 7.39% 3.01%
% of comparable model

164.73% 6.25% 17.13% 4.74%
Panel B

Manager Fixed Effects (n=267)

% Significant at the 10% 49.44%

% Significant at the 5% 43.07%

Mean Effect 0.0018

Median Effect 0.0002

25th Percentile 0.0081

75th Percentile -0.0062

Difference between the 25th and 75th percentile 0.0143

Difference as a percentage of the mean AQ 47.22%
75

TABLE 16
Testing Individual Managers' Fixed Effects on Discretionary Accruals (3 year average)
This table provides the adjusted R
2
for the regression of

as described in Section
5.5. Economic determinants include: firm size, cash flow variability, sales variability, operating cycle, incidence of negative earnings realizations, intangibles intensity, intensity
indicator, and capital intensity. Variable descriptions are provided in the Appendix. Results reported in this table are determined using the Manager Mobility Method. “Yes”
indicates that the variable(s) is/are included in the regression. Coefficient estimates are not reported in this table, but all regressions include corrections for robust standard errors.
The F-Statistic is for the test that all of the manager fixed effects are jointly equal to 0. Panel B reports summary statistics on the manager fixed effects from the regression in
Panel A column 8 (the full regression specification)
Panel A

Dependent Variable = DA (n = 20,615)
Economic Determinants

YES YES

YES YES
Year Fixed Effects YES YES YES YES YES YES YES YES
Firm Fixed Effects

YES

YES

YES

YES
Manager Fixed Effects

YES YES YES YES

Testing Manager Fixed Effects =0

F-statistics

3.95 3.37 4.26 2.69
(p-value)

(<0.001) (<0.001) (<0.001) (<0.001)

Adjusted R
2
0.63% 35.01% 5.50% 23.88% 2.51% 39.08% 9.06% 26.67%
Improvement in Adjusted R
2

raw %

1.88% 4.07% 3.56% 2.79%
% of comparable model

299.05% 11.63% 64.73% 11.68%
Panel B

Manager Fixed Effects (n=719)

% Significant at the 10% 34.91%

% Significant at the 5% 26.70%

Mean Effect -0.0011

Median Effect -0.0008

25th Percentile 0.0035

75th Percentile -0.0058

Difference between the 25th and 75th percentile 0.0092

Difference as a percentage of the mean DA 757.80%
76

TABLE 17
Testing Individual Managers' Fixed Effects on Discretionary Accruals (5 year average)
This table provides the adjusted R
2
for the regression of

as described in Section 5.5.
Economic determinants include: firm size, cash flow variability, sales variability, operating cycle, incidence of negative earnings realizations, intangibles intensity, intensity
indicator, and capital intensity. Variable descriptions are provided in the Appendix. Results reported in this table are determined using the Manager Mobility Method. “Yes”
indicates that the variable(s) is/are included in the regression. Coefficient estimates are not reported in this table, but all regressions include corrections for robust standard errors.
The F-Statistic is for the test that all of the manager fixed effects are jointly equal to 0. Panel B reports summary statistics on the manager fixed effects from the regression in
Panel A column 8 (the full regression specification)
Panel A

Dependent Variable = DA (n = 20,615)
Economic Determinants

YES YES

YES YES
Year Fixed Effects YES YES YES YES YES YES YES YES
Firm Fixed Effects

YES

YES

YES

YES
Manager Fixed Effects

YES YES YES YES

Testing Manager Fixed Effects =0

F-statistics

5.86 5.00 5.81 4.62
(p-value)

(<0.001) (<0.001) (<0.001) (<0.001)

Adjusted R
2
0.67% 50.94% 9.46% 38.34% 3.98% 56.16% 15.44% 42.71%
Improvement in Adjusted R
2

raw %

3.31% 5.22% 5.98% 4.37%
% of comparable model

494.03% 10.25% 63.21% 11.40%
Panel B

Manager Fixed Effects (n=719)

% Significant at the 10% 44.09%

% Significant at the 5% 36.30%

Mean Effect -0.0017

Median Effect -0.0014

25th Percentile 0.0059

75th Percentile -0.0086

Difference between the 25th and 75th percentile 0.0145

Difference as a percentage of the mean DA 603.43%
77

Graph 1
Firm Accounting Quality Over Time
This graph shows Firm AQ centered on the event of a manager leaving the firm. AQ is measured using equation 4.
Managers are classified as conservative, moderate or aggressive as defined in section 4.2. Time is in years where year 0
corresponds to the year the manager leaves the firm.

0
0.01
0.02
0.03
0.04
0.05
0.06
-3 -2 -1 0 1 2 3
AQ
Time
Accounting Quality (Mean)
Conservative
Moderate 1
Moderate 2
Aggressive
AQ Mean
78

Graph 2
Firm Discretionary Accruals Over Time
This graph shows firms’ discretionary accruals centered on the event of a manager leaving the firm. Discretionary
accruals are one year discretionary accruals measured as described in section 5.5. Managers are classified as conservative,
moderate or aggressive as defined in section 4.2. Time is in years where year 0 corresponds to the year the manager leaves the
firm.

-0.02
-0.015
-0.01
-0.005
0
0.005
0.01
-3 -2 -1 0 1 2 3
DA
Time
Discretionary Accruals (Mean)
Conservative
Moderate 1
Moderate 2
Aggressive
DA Mean
79

Graph 3
Firm Discretionary Accruals (3 year average) Over Time
This graph shows firms’ discretionary accruals centered on the event of a manager leaving the firm. Discretionary
accruals are the average discretionary accruals over three years and are measured as described in section 5.5. Managers are
classified as conservative, moderate or aggressive as defined in section 4.2. Time is in years where year 0 corresponds to the
year the manager leaves the firm.

-0.005
-0.004
-0.003
-0.002
-0.001
0
0.001
0.002
-3 -2 -1 0 1 2 3
DA
Time
Discretionary Accruals
(Mean 3 Year Average)
Conservative
Moderate 1
Moderate 2
Aggressive
DA Mean
80

Graph 4
Firm Accounting Quality Around Manager Changes (1 year)
This graph shows firm AQ the year prior to and the year following a manager leaving the firm. AQ is measured using
equation 4. Managers are classified as conservative, moderate or aggressive as defined in section 4.2. Time is in years where
year 0 corresponds to the year the manager leaves the firm.

0
0.01
0.02
0.03
0.04
0.05
0.06
-4 -3 -2 -1 0 1 2 3 4
AQ
Time
Accounting Quality
full
conservative
moderate 1
moderate 2
aggressive
81

Graph 5
Firm Accounting Quality Around Manager Changes (3 year)
This graph shows Firm AQ the three years prior to and three years following a manager leaving the firm. AQ is
measured using equation 4. Managers are classified as conservative, moderate or aggressive as defined in section 4.2. Time is in
years where year 0 corresponds to the year the manager leaves the firm.

0
0.01
0.02
0.03
0.04
0.05
0.06
-4 -3 -2 -1 0 1 2 3 4
AQ
Time
Accounting Quality
full
conservative
moderate 1
moderate 2
aggressive
82

APPENDIX

Variable Name Description
Assets Total assets
R&D Research and development expense/lag one year total assets
Leverage (Long term debt + debt in current liabilities)/total assets
Cash Holdings
Cash and short-term investments/(total assets – cash and short-term
investments)
Dividend Yield Dividends per share / year-end stock price
Investment Capital expenditures/lag one year net property, plant, and equipment
AQ
Five-year rolling standard deviation of the residuals from a regression of
total current accruals (change in current assets - current liabilities - cash +
short term debt) on cash flows from operations from t-1 to t+1, change in
sales revenue, and gross property, plant and equipment.
Persistence
The coefficient from a regression of current earnings on lagged earnings,
using a ten-year rolling window
Predictability
The square root of the error variance from a regression of current earnings
on lagged earnings, using a ten-year rolling window
Smoothness
The standard deviation of net income before extraordinary items and
discontinued operations scaled by average total assets / the standard
deviation of cash flows from operations scaled by average total assets
Asymmetric TLR
The coefficient on the interaction term of change in operating cash flows
scaled by assets and an indicator variable that is 1 when the change in
operating cash flows is negative, from the regression of accruals on change
in cash flows from operations, an indicator variable if the change is negative
and the interaction term. (Ball & Shivakumar, 2006; Basu, 1997)
ROE Net income /total book value of common equity
83

ROA
Net income before extraordinary items and discontinued operations/ total
assets
Tobin's Q
Market value of equity plus the book value of liabilities / Total assets
(Brainard & Tobin, 1968)
KZ Index
-1.00*(net income before extraordinary items and discontinued operations
plus dividends paid /lagged net property, plant and equipment) + 0.28 *
TOBIN'S Q + 3.13*(Liabilities/lagged net property, plant and equipment) -
39.36*(Dividends/lagged net property, plant and equipment) -1.31*(Cash
and short term investments/lagged net property, plant and equipment)
(Kaplan & Zingales, 1997)
Firm Size Log of total assets
Cash Flow Variability
Five-year rolling Standard deviation of cash flow from operations, scaled by
assets
Sales Variability Five-year rolling standard deviation of sales revenue, scaled by assets
Operating Cycle The log of the sum of days accounts receivables and days inventory
Incidence of Negative
Earnings
Proportion of losses over the prior 10-yr
Intangibles Intensity (R&D Expenses plus Advertising Expenses)/ sales revenue
Intensity Indicator Indicator variable equal to 1 if intangibles intensity is greater than 0
Capital Intensity The net book value of property, plant and equipment / total assets
GINDEX Gompers, Ishii, and Metrick (2003) antitakeover index
EINDEX Bebchuk, Cohen, and Ferrell (2009) entrenchment index
84

Institutional Ownership The percentage of stock held by institutional owners
Blockholder The percentage of stock held by the largest block holder
Board Size Annual number of board members
Board Meetings Annual number of board meetings
Independence The proportion of independent directors over the total number of directors
Total Compensation
The log of the annual salary plus bonus plus stock based compensation
awarded to the executive
Salary and Bonus The log of the annual salary and bonus awarded to the executive
Equity Based
Compensation
The log of the annual equity based compensation awarded to the executive

Abstract (if available)

Abstract I investigate whether individual managers have an incremental effect on their firms’ accounting quality (AQ) after controlling for known determinants of AQ, time fixed effects and firm fixed effects. To identify the manager-specific effect on firm AQ, I construct a data set that tracks the movement of 720 managers across firms over the period 1992-2011. Results indicate that individual manager fixed effects explain a statistically and economically significant proportion of the cross-sectional variation in AQ, which is three times larger than that of industry fixed effects and comparable to that of firm fixed effects. Moving from the quartile of most aggressive managers to that of the most conservative managers, results in a 41% improvement in the firms’ AQ. I also examine the consequences of switching between conservative and aggressive managers on the firms’ AQ. Results show that a firm that switches from an aggressive (conservative) manager to a conservative (aggressive) manager increases (decreases) the firm’s AQ by 32% (94%). My study underscores the importance of understanding individual manager styles in the determination of the firms’ AQ.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Accrual quality and expected returns: the importance of controlling for cash flow shocks

PDF

Can I borrow your footnotes? Learning and network benefits of footnote similarity

PDF

Do shared auditors improve audit quality? Evidence from banking relationships

PDF

Real asset liquidity and asset impairments

PDF

Share repurchases: how important is market timing?

PDF

Does fair value accounting exacerbate the pro-cyclicality of bank lending?

PDF

Does mandatory adoption of International Accounting Standards reduce the cost of equity capital?

PDF

Effectiveness of the SEC’s comment letters in initial public offerings

PDF

Does mandatory adoption of international financial reporting standards decrease the voting premium for dual-class shares: theory and evidence

PDF

Essays in asset pricing

PDF

Evolution of returns to scale and investor flows during the life cycle of active asset management

PDF

Three essays on young entrepreneurial firms

PDF

Three essays on board structure

PDF

The effects of accounting performance and professional relationships on promotion, dismissal, and transfer decisions in a conglomerate

PDF

Does differential sensitivity to aggregate earnings shocks drive post-earnings-announcement drift?

PDF

Executive compensation: the trend toward one size fits all

PDF

Quality investment and advertising: an empirical analysis of the auto industry

PDF

Gone with the big data: institutional lender demand for private information

PDF

The effects of a customer's comparative processing with positive and negative information on product choice

PDF

Private disclosure contracts within supply chains and managerial learning from stock prices

Asset Metadata

Creator Wells, Kara (author)

Core Title Who manages the firm matters: the incremental effect of individual managers on accounting quality

School Marshall School of Business

Degree Doctor of Philosophy

Degree Program Business Administration

Publication Date 05/29/2013

Defense Date 04/30/2013

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

Tag accounting quality,accrual quality,manager fixed effects,OAI-PMH Harvest

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Subramanyam, K.R. (committee chair), DeFond, Mark L. (committee member), Murphy, Kevin J. (committee member)

Creator Email kara.wells.2012@marshall.usc.edu,karawell@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-256752

Unique identifier UC11294582

Identifier etd-WellsKara-1703.pdf (filename),usctheses-c3-256752 (legacy record id)

Legacy Identifier etd-WellsKara-1703.pdf

Dmrecord 256752

Document Type Dissertation

Format application/pdf (imt)

Rights Wells, Kara

Type texts

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

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

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA