Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
The effects of accounting performance and professional relationships on promotion, dismissal, and transfer decisions in a conglomerate
(USC Thesis Other)
The effects of accounting performance and professional relationships on promotion, dismissal, and transfer decisions in a conglomerate
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
THE EFFECTS OF ACCOUNTING PERFORMANCE AND PROFESSIONAL
RELATIONSHIPS ON PROMOTION, DISMISSAL, AND TRANSFER DECISIONS
IN A CONGLOMERATE
by
Jonghwan Kim
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 Jonghwan Kim
ii
Acknowledgements
This dissertation would not have been possible without the guidance and the help of
several individuals who in one way or another contributed and extended their valuable
assistance in the preparation and completion of this study. First and foremost, I gratefully
thank my dissertation chair, Kenneth A. Merchant for his helpful guidance and
continuous support that I will never forget. I also thank the other members of my
dissertation committee for their advice and support: David Erkens, and Hyungsik Roger
Moon.
I also appreciate the valuable comments and suggestions from Mark DeFond, Mingyi
Hung, Tatiana Sandino, Stanley Baiman, Sergio Berretta, Andrea Dossi, Mahmoud
Ezzamel, Salvador Carmona, Robert Watson, and other workshop participants at the
University of Southern California, Rutgers University-Camden, Bocconi University, IE
Business School, and ESSEC Business School. I am grateful to the Leventhal School of
Accounting and Marshall School of Business at the University of Southern California for
financial support.
Last but not least, I express my deep appreciation to Eunok, my lovely wife, for her
wholehearted support and encouragement without which I could not have completed my
degree. My gratitude also goes to my dear parents, Jungjin Kim and Kwanghee Park, and
my sister, Yoonjung, for their endless love.
iii
Table of Contents
Acknowledgements ............................................................................................................. ii
List of Tables ...................................................................................................................... v
List of Figures .................................................................................................................... vi
Abstract ........................................................................................................................ vii
Chapter 1. Introduction ...................................................................................................... 1
1.1. Relationships between Accounting Performance and Personnel Decisions ........... 1
1.2. Supervisors’ Use of Subjectivity in Personnel Decisions ....................................... 5
1.3. Contributions ........................................................................................................... 9
Chapter 2. Literature Review ........................................................................................... 13
2.1. Performance and Personnel Decisions .................................................................. 13
2.2. Supervisors’ Use of Subjectivity in Personnel Decisions ..................................... 21
Chapter 3. Accounting Performance and Personnel Decisions: Hypotheses .................. 30
3.1. Personnel Decision-Making .................................................................................. 30
3.2. Organizational Performance and Personnel Decisions ......................................... 34
3.3. Purposes and Types of Promotions ....................................................................... 35
3.4. Job Scope and Responsibility................................................................................ 37
3.5. Intra-organization Interdependencies and Accounting Performance .................... 38
Chapter 4. Discretion and Bias in Personnel Decisions: Hypotheses .............................. 41
4.1. Information Hypothesis vs. Preference Hypothesis .............................................. 41
4.2. The Effects of Presence of Professional Relationships on Personnel Decisions .. 41
4.3. The Effects of Length of Professional Relationships on Personnel Decisions ..... 43
4.4. The Effects of Group Performance on Personnel Decisions ................................. 45
4.5. The Effects of Hierarchical Levels on Personnel Decisions ................................. 45
4.6. The Effects of Promotion Competition on Personnel Decisions .......................... 47
Chapter 5. Research Setting and Data ............................................................................. 49
5.1. Characteristics of Human Resource Management in Korean Conglomerates ...... 49
5.2. Research Setting: The Careers of Executives in a Korean Conglomerate ............ 52
5.3. Sample Selection and Data Collection .................................................................. 54
5.4. Variables ............................................................................................................... 59
5.5. Description of Personnel Decisions in Research Sites .......................................... 65
Chapter 6. Accounting Performance and Personnel Decisions: Empirical Evidence ..... 79
iv
6.1. Research Design: Estimation of the Likelihood of Promotions and Dismissals ... 79
6.2. The Effects of Corporate and Segment ROAs ...................................................... 82
6.3. Other Determinants of Promotions and Dismissals .............................................. 82
6.4. Purposes and Types of Promotions ....................................................................... 84
6.5. Job Scope and Responsibility................................................................................ 90
6.6. Intra-organization Interdependency ...................................................................... 92
6.7. Additional Analysis: Cross-Unit Mobility and Promotions .................................. 96
Chapter 7. Discretion and Bias in Personnel Decisions: Empirical Evidence ............... 104
7.1. Research Model: The likelihood of Personnel Decisions ................................... 104
7.2. The Effects of Presence of Professional Relationships on Personnel Decisions 106
7.3. The Effects of Length of Professional Relationships on Personnel Decisions ... 106
7.4. The Effects of Group Performance on Personnel Decisions ............................... 107
7.5. The Effects of Performance on Personnel Decisions .......................................... 108
7.6. The Effects of Promotion Competition on Personnel Decisions ........................ 113
7.7. Additional Observations...................................................................................... 114
7.8. Additional Tests of Interaction Terms ................................................................ 115
7.9. Additional Tests Using Alternative Measure of Social Ties ............................... 116
Chapter 8. Discussion and Conclusion .......................................................................... 121
8.1. The Effects of Accounting Performance on Personnel Decisions ...................... 121
8.2. The Effects of Social Relationships on Personnel Decisions.............................. 124
8.3. Research Opportunity, Limitations, and Future Direction .................................. 125
Bibliography ................................................................................................................... 129
Appendix ...................................................................................................................... 137
v
List of Tables
Table 1: Predictions—Information-Based Discretion vs. Favoritism ............................... 42
Table 2: Sample Selection ................................................................................................ 55
Table 3: Descriptive Statistics .......................................................................................... 68
Table 4: Correlations between Variables .......................................................................... 70
Table 5: Promotion Decisions by Performance, Hierarchical Rank, and Social Tie ........ 76
Table 6: Performance, Social Tie and Promotion Decisions ............................................ 78
Table 7: Accounting Performance, Promotions, and Dismissals ...................................... 80
Table 8: Promotion Types a and Accounting Performance .............................................. 86
Table 9: Job Responsibility and Accounting Performance ............................................... 88
Table 10: Intra-organization Interdependency, Accounting Performance, and Promotions
................................................................................................................................... 94
Table 11: Cross-Unit Job Mobility and Promotions ....................................................... 101
Table 12: Determinants of Promotion Decisions ............................................................ 109
Table 13: Determinants of Stay (Dismissal) Decisions .................................................. 111
Table 14: Testing Interaction Effects ............................................................................. 117
Table 15: Titles for Hierarchical Ranks for Executives in Samsung .............................. 138
vi
List of Figures
Figure 1: An Example of an Organization Chart ............................................................. 58
Figure 2: Interaction Effects on Personnel Decisions .................................................... 119
vii
Abstract
This dissertation empirically examines the effects of accounting performance and
professional relationships between supervisors and subordinates on personnel decisions,
including promotions and dismissals, for executives working in a large conglomerate. In
the first part, this study investigates whether accounting performance measured at
corporate and reporting segment levels affects decisions to promote or dismiss executives
in sub-organizations and how the relationship varies in different contexts where
supervisors make the decisions for different types of promotions, for workers with
different job responsibilities, and in organizations with different organizational
interdependencies. The second part of this study examines the effects of subjective
evaluations in promotion decisions. In particular, I compare two competing explanations
about what appears to be the outcome of favoritism: Information vs. preference. In a
sample of 4,657 executive-years in a Korean conglomerate, the findings indicate that: (1)
when supervisors make personnel decisions, they consider organizational performance in
an accounting measure, or return on assets (ROA) in a way to make the most of benefits
of associating the performance measure with promotion-based incentives; and (2) they do
not favor subordinates simply due to an established relationship; rather, they seem to
value information that has been communicated and accumulated for a long (but not too
long) time and at a close (hierarchical) distance.
1
Chapter 1. Introduction
In this dissertation, I examine previously unexplored factors that affect personnel
decisions of executives in a conglomerate.
1
In particular, this dissertation focuses its
attention on two primary questions about organizations’ and supervisors’ behaviors in
using accounting information and subjectivity in the context of personnel decision-
making. First, I examine the relationship between accounting performance and executives’
promotions and dismissals and the moderating effects of types of promotions, promotion
candidates’ job responsibility, and level of interdependencies between business units on
the relationship. Second, I study the effects of a professional relationship established
between a subordinate and his supervisor, as a proxy either for more information
exchange or for the supervisor’s personal preference, on his promotions and dismissals.
1.1. Relationships between Accounting Performance and Personnel Decisions
Promotions are a popular incentive mechanism used in most organizations (Baker,
Jensen, & Murphy, 1988; Gibbs, 1996).
2
Previous studies on the topic, albeit few in
number, have shown evidence that promotions and dismissals are related to individual
workers’ performance as reflected in performance ratings (Medoff & Abraham, 1980;
Murphy, 1992), a subset of objective performance measures (Fee, Hadlock, & Pierce,
2006; Kwon, 2006b), accounting performance (Blackwell, Brickley, & Weisbach, 1994;
1
Personnel decisions in this paper refer to supervisors’ decisions about their subordinates’ careers—
including promotions, dismissals, and internal transfers.
2
Incentive provision is not the only purpose of promotions. Literature discusses job matching as another
important role of promotions (e.g., Baker et al., 1988; Gibbs, 1995; Milgrom & Roberts, 1992).
2
Cichello, Fee, Hadlock, & Sonti, 2009), and non-financial performance measures
(Campbell, 2008). However, it is yet unclear if the findings from these studies may be
extrapolated to the relationship between organizational performance and promotions.
This is because, unless an individual worker is held fully responsible for an
organization’s performance, organizational performance is a relatively noisy indicator of
individual workers’ qualities for promotions than individual performance, and the
efficiency of promotion-based incentives based on organizational performance may be
compromised by considerable and relatively persistent costs of promotions and
organizational constraints including organization structures and financial resources.
In explicit incentive contexts, organizational performance measures are a standard
part of incentive contracts as, per agency theory, all informative performance indicators
improve the efficiency of incentive contracts (Holmström, 1979). However, research
literature has been silent about the use of organizational performance measures in implicit
incentive contexts. In particular, it has been unanswered whether benefits from using the
limitedly informative performance measure exceed the costs and overcome the
constraints. The present study fills the void in the literature by presenting empirical
evidence of the relationship between accounting performance as an organizational
performance indicator and promotions (and other personnel decisions
3
) of workers
employed in organizations with different performance.
3
Personnel decisions in this paper include promotions, dismissals, and internal job transfers.
3
I analyze the career events of executives working in a large multinational
conglomerate, by tracking their profiles, organizational structure, and accounting
information from 2002 through 2007. The research setting identified in Korea allows
access to data not available in the United States (U.S.). It provides an excellent
opportunity to observe the career events of all executives, rather than just the top five
highest-paid executives, in companies of interest. It allows me to distinguish different
types of personnel decisions and identify relevant contextual factors while controlling for
other factors. The dataset extracted from annual reports tracks the career events of 1,251
executives (4,657 executive-years) working in the conglomerate.
I begin by providing a simple illustration of the personnel decision-making process
through which supervisors can exercise subjectivity. I posit that during the evaluation and
decision-making process, supervisors may exercise subjectivity in three ways: selection
of evaluation criteria, importance weighting, and adjustment of evaluation results. The
concept of informativeness is relevant to the selection and importance weighting
processes of performance measures and evaluation criteria (e.g., Banker & Datar, 1989;
Feltham & Xie, 1994; Holmström, 1979). Supervisors may then make some adjustments
in the provisional evaluation results to affect the likelihood of personnel decisions.
Specifically, to arrive at the decision, they may shift the evaluation results,
4
in
4
As the decision is, by nature, binary (e.g., to promote or not), it involves a certain decision threshold.
Thus, small adjustments by supervisors have a critical impact on final decision outcomes, especially for
those whose interim evaluation results are around the threshold.
4
consideration of several circumstantial factors such as organizational performance, the
level of promotion competition, and organizational capacity to feed promotions.
Taking into account the above description, I investigate the effects of corporate and
segment return on assets (ROA) on personnel decisions as a subjective adjustment factor.
Then, I consider moderating factors that determine the usefulness of corporate and
segment ROAs: promotion types, job responsibility, and intra-organization
interdependency. In addition to these analyses, I pursue an exploratory analysis of the
association between promotions and cross-unit job transfers.
The results indicate that in general supervisors associate promotions and dismissals
of their subordinates with organizational performance. In particular, I find significant
effects of accounting performance on personnel decisions of workers in an organization.
Further, while investigating the usefulness of accounting performance in personnel
decisions, I find that supervisors use accounting performance in different ways depending
on the decision-making contexts. Specifically, the findings suggest that accounting
performance of an organization is tied only to promotions awarded as incentives and to
promotions of executives of a relatively low rank or with relatively low responsibility. I
also find that the association between accounting performance and promotions is stronger
when organizational interdependency, measured as the transferability of workers’
knowledge and skills between sub-organizations, is greater. Lastly, analysis results show
that promotions are awarded in combination with cross-unit transfers from an
organization with good performance to another with poor performance.
5
1.2. Supervisors’ Use of Subjectivity in Personnel Decisions
For the second research question, I examine the effects of supervisors’ use of
subjective evaluation in promotion decisions
5
for executives. Specifically, this part of the
dissertation addresses a series of two questions about supervisors’ behaviors in using
subjectivity in the context of promotion decisions. Given the manifestation of favoritism
in practice and in academic literature, do supervisors really exercise favoritism toward
some of their subordinates over others? If not, what is the cause of such favoritism-like
behaviors? To answer the questions, I compare two potential reasons for such behaviors:
(1) supervisors’ favoritism, or showing personal preference for some subordinates over
others, that arises (simply) due to an established social tie,
6
i.e., a bias, and (2)
supervisors’ learning about their subordinates’ skills and abilities, i.e., discretion based
on information communicated through the established relationship. Understanding this
(potential) form of bias in promotion decisions and distinguishing supervisors’ use of
“sound” discretion from the bias is important for at least four reasons. First, the costs of
incorrect decisions
7
are larger than the costs of inaccurate performance evaluations for
other purposes, e.g., bonus pay decisions (Baker et al., 1988). Second, implicit incentives,
i.e., promotion-based incentives, are not insubstantial and are widely used in
5
Promotion decisions in this paper refer to supervisors’ decisions about their subordinates’ careers—
including not only promotions but also stays and dismissals. Throughout the paper, promotion decisions
and personnel decisions are used interchangeably.
6
Throughout this paper, a social tie refers to a professional relationship that is developed through a
promotion award. A social tie and a professional relationship are used interchangeably.
7
Examples include (but are not limited to) assigning a subordinate to a job that does not fit well with the
subordinate’s skill set or ability, promoting a less capable subordinate, not promoting a capable subordinate
and (as a result) losing him (due to his voluntary resignation), etc.
6
organizations (Baik, Kim, Evans III, & Yanadori, 2011; Baker et al., 1988; Ederhof, 2011;
Gibbs, 1995; Medoff & Abraham, 1980). Third, the working mechanism of subjectivity
in personnel decision contexts seems unlike the way it functions in other performance
evaluation contexts. Last, but most importantly, the two distinct causes of the supervisors’
favoritism-like behaviors offer totally incompatible measures; favoritism should be
constrained whereas (sound) discretion should be encouraged.
Performance evaluations for personnel decisions involve subjectivity. For incentive
provision purposes, the use of subjectivity is hardly different from that of other (periodic)
performance evaluations—e.g., bonus pay determination. It mitigates problems using
compensation schemes strictly based on objective or quantitative measures: for example,
correcting employees’ dysfunctional behaviors such as short-term orientation and game-
playing (Baker, Gibbons, & Murphy, 1994; Ittner, Larcker, & Meyer, 2003) and
eliminating unexpected down-side risks that employees, otherwise, would have borne
(Gibbs, Merchant, Van der Stede, & Vargus, 2004). On the other hand, evaluations of
subordinates’ skills and abilities for job matching require a higher degree of subjectivity
than those for incentive provision. By and large, the evaluations to match talents and jobs
rely on supervisors’ personal judgments of subordinates’ potential contributions to an
organization’s (future) performance.
Despite the benefits from the use of subjectivity in performance evaluation, it causes
some problems that may reduce the efficiency of compensation and job assignments.
Prior literature has documented supervisor biases arising from subjective performance
7
evaluation and its focus has been centered on two types of biases: lenience and centrality
biases (e.g., Bol, 2011; Moers, 2005; Murphy, 1992; Prendergast & Topel, 1993). In
addition to the two types of biases, another form of supervisor bias has begun to draw
researchers’ attention in economics and sociology: favoritism or social preferences (e.g.,
Bandiera, Barankay, & Rasul, 2009; Breuer, Nieken, & Sliwka, 2010; Fehr &
Fischbacher, 2002; Prendergast & Topel, 1996).
Among these types of supervisor bias, this study focuses on favoritism for two
reasons. First, compared to leniency and centrality biases, scant attention has been paid to
favoritism, especially in accounting literature. To my knowledge, no prior literature
discusses favoritism in the context of promotion decision-making. Second and more
importantly, leniency and centrality biases are not descriptive of how subjectivity leads to
biases in promotion decisions. First, unlike in other performance evaluation situations,
supervisors are “forced” to make final decisions in three categories: promote, retain, and
dismiss. However, the upward shift in evaluation ratings for all subordinates and the
compressed dispersion of ratings cannot affect the “cutoffs” between the three decision
outcome categories. In addition, personnel decisions are explicit and observable by
anyone within an organization and, oftentimes, even outside an organization. Therefore,
supervisors’ reluctance to communicate bad news (Baker et al., 1988) does not warrant
their motivations to display such biases in personnel decision-making situations because
they should make and inform their final (bad) decisions.
8
In practice, supervisors appear to favor certain subordinates with whom they have a
stronger professional relationship. Consistent with conventional belief, several studies
report the presence of favoritism (e.g., Bandiera, Barankay, & Rasul, 2008; Breuer et al.,
2010; Garicano, Palacios-Huerta, & Prendergast, 2005). Similarly, this study’s dataset
provides some evidence of “ostensible” manifestation of favoritism.
Apparently, supervisors can extract utilities from exercising favoritism (Kwon,
2006a; Prendergast & Topel, 1996). However, supervisors’ excessive exploitation of their
power raises organizational problems; corruption, collusion, disservice and demotivation
of dissatisfied subordinates. As long as supervisors themselves are rewarded for their
performance, such dysfunctional behaviors could be controlled to a minimum such that
the costs of favoritism do not exceed the benefits from it (Prendergast & Topel, 1996).
So, I contend that the real cause of the favoritism-like behavior deviates from that of
authentic favoritism. Further, I propose an alternative explanation for a supervisors’
behavioral outcome which I call the Information Hypothesis. The Information Hypothesis
appreciates the role of social ties as a channel through which more and truthful
information is communicated between the parties at the end of a node (Adler & Kwon,
2002). Accordingly, an established relationship reinforces effective communication and,
as a result, culminates in a better-informed, discreet decision that is less subject to bias.
To test the conjecture, I develop pairs of hypotheses with the two competing
explanations: discretion (based on information) vs. bias (based on preference). Overall,
this study finds that supervisors do not favor subordinates simply because they are
9
socially connected through previous promotion awards; rather, supervisors value
information that has been communicated and accumulated for a long (but not too long)
time and at a close (hierarchical) distance; and supervisors may lack resources to make
thorough evaluations of each of their subordinates where there are a large number of
contestants and, in such cases, pursue an alternative efficient decision-making strategy,
i.e., simple heuristics to reduce effort under bounded rationality (Simon, 1979). The
findings are generally consistent with the predictions made from the Information
Hypothesis.
1.3. Contributions
This study makes several contributions to existing literature. First, this dissertation
improves our understanding of the performance-promotion relationship involving mid- to
low-level executives, which has not been previously addressed. Past studies have shown
that promotions are associated with individual workers’ performance in different forms
(e.g., Campbell, 2008; Cichello et al., 2009; Medoff & Abraham, 1980). However, no
theory or empirical research discusses how organizational performance is linked to
promotion-based implicit incentives or how it affects careers in an organization. This
study bridges the gap by providing evidence of how accounting performance of an
organization affects all executives, as opposed to a few top-ranking executives in the
subunits of the organization.
Second, this dissertation contributes to the literature on promotions by identifying
new determinants of promotions and dismissals. This study fulfills this need for research.
10
Specifically, it examines how different types of personnel decisions, job responsibility,
and organizational interdependency moderate the relationship between accounting
performance and personnel decision outcomes. In addition, it identifies another correlated
omitted variable, i.e., a professional / social relationship that affects a supervisor’s
behavior in using subjectivity in promotion decisions. Typical studies about determinants
of promotions (e.g., Baker, Gibbs, & Holmstrom, 1994a, 1994b; Medoff & Abraham,
1980) consider a promotion candidate’s ability and/or performance. However, promotion
decisions involve a decision-maker’s significant effort and subjectivity, which should be
recognized as an important determinant of promotions. This paper shows a supervisor’s
behavioral factors, affected by a social relationship, have incremental effects on
promotion decisions.
Third, the study contributes to accounting literature on subjectivity, responding to
calls for research on factors affecting subjective evaluations in promotion decisions. In
his comments on Campbell (2008), Gibbs (2008) points this out : “Prior work on
promotion systems has demonstrated that subjective merit ratings are correlated with
promotions. However, very little evidence has been presented on what factors are
considered in assigning such ratings” (Gibbs, 2008, p. p. 334: emphasis is added). While
prior literature has discussed subjectivity and ensuing supervisor biases in the context of
general performance evaluations, this dissertation discusses subjectivity specifically used
in a promotion decision context. It also discusses another type of supervisor bias, i.e.,
favoritism, that has been rarely discussed in prior accounting literature and that is more
11
relevant to promotion decision-making situations than leniency and centrality biases.
Further, I propose an alternative explanation for favoritism-like decision outcomes;
compare two behavioral options that a supervisor can take in using subjectivity; and show
that blind favoritism is suppressed by incentives to improve a supervisor’s own
performance under, presumably, ordinary compensation schemes based on meritocracy.
8
The findings suggest that adequate incentives for performance may mitigate, if not
eliminate, the risk of favoritism and, further, the need for principals’ separate rewarding
and monitoring of managers’ supervisory tasks.
Fourth, this study also contributes to favoritism literature. Even though a few
empirical studies address favoritism, their research settings (e.g., referees in sports
(Garicano et al., 2005), physical laborers on a farm (Bandiera et al., 2008, 2009)) may not
be appropriate to address questions regarding promotions of managerial executives. This
is the first research in which the roles and effects of favoritism are empirically tested
using white-collar workers at managerial positions in real business organizations.
Lastly, this study contributes to the management literature on the transfer of
knowledge and best practices within an organization (e.g., Gupta & Govindarajan, 2000;
Minbaeva, Pedersen, Bjorkman, Fey, & Park, 2003). In particular, the findings from the
exploratory analysis in this study suggest that organizations may utilize promotions and
job transfers to improve the efficiency of knowledge diffusion. To the best of my
knowledge, this is one of the first studies to report the relationship between
8
The Korean conglomerate of this study stresses individuals’ competence and performance (Pucik and Lim
2001)
12
organizational performance and job transfers, which represents an interesting avenue of
future research.
The remainder of this dissertation proceeds as follows. Chapter 2 reviews prior
literature regarding the association between accounting performance and personnel
decisions and supervisors’ use of subjectivity in personnel decision-making. Chapters 3
and 4 develop the hypotheses for the two research questions. Chapter 5 introduces the
unique institutional setting that permits this research opportunity and describes data.
Chapters 6 and 7 explain the research design used for testing the hypotheses, and report
the results of the analyses. Lastly, Chapter 8 concludes the dissertation, summarizing the
findings and discussing the study’s limitations as well as potential future research
directions.
13
Chapter 2. Literature Review
2.1. Performance and Personnel Decisions
There are two explanations of why good performance lead to promotions (e.g., Baker
et al., 1988; Fairburn & Malcomson, 2001; Gibbons & Waldman, 1999; Gibbs, 1995;
Milgrom & Roberts, 1992). From the incentive mechanism perspective, promotions (and
other compensations) are awarded for outstanding “individual” performance. On the
other hand, from the human capital perspective, performance is a good indicator of an
individual’s ability and qualification for promotion to higher positions.
An important role of personnel decisions including promotions is to provide
incentives for workers’ effort allocation (e.g., Fairburn & Malcomson, 2001; Lazear &
Rosen, 1981; Malcomson, 1984) and human capital acquisition (e.g., Prendergast, 1993a).
This line of research stresses the point that promotions are associated with a large pay
raise. Undoubtedly, the large monetary compensation associated with promotions induces
workers to direct their effort toward performance or skill acquisition. In fact, many firms
use promotions as a primary incentive instrument—especially for white-collar workers
(Baker, Gibbs, et al., 1994a, 1994b; Baker et al., 1988; Murphy, 1985).
Attending to the incentive effects for effort allocation, Lazear and Rosen (1981)
introduce rank-order tournament theory. They compare tournaments and piece-rate
compensation contracts, and show that tournaments among workers for “prizes”
including fixed bonuses and promotions are equally efficient as piece-rate contracts.
Extending the Lazear and Rosen’s analysis, Mookherjee (1984) points out that a
14
tournament can create incentives even under a circumstance where an explicit contract
fails to create efficient incentives due to the lack of verifiability of individual
performance (e.g., subjective evaluation of performance).
As an alternative resolution to the problem arising from non-verifiable evaluation of
the skills collected (i.e., a worker’s effort allocation or investment toward skill
acquisition), Prendergast (1993a) and Kahn and Huberman (1988) suggest promotions
that involve a substantial increase in wage to encourage workers to acquire specific
human capital. Despite the commonality, the two studies deal with different types of
organizations—those with multiple/distinct (sole/similar) job(s) in Prendergast (Kahn and
Huberman)—where an organization’s credible promise to promote high-productivity
workers is made under different conditions. In particular, Prendergast (1993a) argues a
firm’s promise for a promotion is credible only when the substantial difference in wage
between jobs is associated with difference in tasks or the values attached to the human
capital required for the jobs. In contrast, in organizations with similar, if not sole, tasks,
Kahn and Huberman’s (1988) “up-or-out” type promotion practice
9
provides
organizations with incentives to reward workers’ investment in training because they will
lose highly skilled workers otherwise.
The human capital perspective (e.g., Becker, 1964; Rosen, 1982; Sattinger, 1993;
Waldman, 1984a) understands the promotion process as a job-assignment mechanism
with which a firm allocates human resources more efficiently. Efficient job allocation
9
For example, tenure decisions of faculty members in universities.
15
matches workers with different abilities or skills (i.e., comparative advantage) to jobs
where the workers’ abilities and skills are best suitable and can be best utilized, and thus
results in an organization’s output maximization (see Gibbons & Waldman, 1999).
10
This may well be extended to cases where jobs are at different hierarchical levels,
allowing its application to promotions. In general, an organization is structured in a
hierarchy such that jobs at higher levels require greater ability and/or higher-level skill
sets.
More importantly, it should be noted that theories in this perspectives, in general,
assume job assignments under full information (e.g., Rosen, 1982; Waldman, 1984b).
11
Therefore, as a prerequisite for efficient job assignment, it is important for an
organization to gauge workers’ true ability. However, in the real world, a worker’s ability
is often unobservable, and thus estimated with his or her productivity in different forms
of performance or even with his or her experience in a labor market.
To summarize, evaluated or measured performance makes an important indicator of
workers’ effort allocation or ability. Therefore, in either perspective, performance
indicators are expected to be associated with workers’ careers within an organization.
The subsequent subsections review relevant literature, focusing on empirical
10
Consider the following simple example. Assume that a firm has two workers (denoted in i=1,2) with
different abilities and two jobs (denoted in j=1,2) and that each worker can be assigned to a job producing
an output of a
ij
(i.e., worker i's output in job j). Then, efficient job assignment—or a set of (i,j)—maximizes
the firm’s output (Y): max
𝑗 � 𝑎 1 𝑗 + 𝑎 2~ 𝑗 � .
11
Incomplete information about a worker’s ability at the beginning of a worker’s career in an organization
may be addressed through the organization’s learning as his or her career progresses in the organization
(Murphy, 1986)
16
investigations of the association between personnel decisions and performance measures
in different forms: performance ratings, objective performance measures, accounting and
non-accounting performance measures.
Subjective Performance Ratings
Early labor economic studies investigating the association between performance and
promotions provide the evidence that promotions are related to good performance ratings
in annual performance reviews (e.g., Baker et al., 1988; Gibbs, 1995; Medoff & Abraham,
1980). In their seminal paper, Medoff and Abraham (1980) analyze the personnel records
of managerial and professional employees in two large U.S. companies. In one of their
research site company,
12
they find that supervisors’ evaluation on the performance of
their subordinates (i.e., “rated performance” or “performance rating”) affects promotion
decisions after controlling for education and experience. Subsequent studies by Baker,
Gibbs, et al. (1994a) and Gibbs (1995) confirm the positive association between the
likelihood of promotions and performance ratings, using a large longitudinal dataset from
a mid-size U.S. service firm.
Although the empirical studies supply interesting empirical regularities in labor
market dynamics, including the relationship between performance ratings and promotions,
they raise a common issue about the use of performance ratings. So to speak, they assume
that performance ratings are good indicators of employees’ productivity or ability.
12
While their original dataset includes the personnel records from two companies, only one of them (i.e.,
“Company B” in their study) provides all the necessary historical performance rating data to evaluate the
relationship between performance ratings and promotions.
17
However, the “ratings,” as being the outcomes of supervisors’ subjective evaluations by
definition, are subject to unintentional biases or deliberate manipulation of performance
reports (e.g., Baker et al., 1988; Fee et al., 2006; Medoff & Abraham, 1980; Moers, 2005;
Prendergast & Topel, 1993, 1996; Rosen, 1982).
13
The literature, in this regard, does not
warrant the relationship between performance in a given task per se and promotion
decisions.
Objective Performance Measures
Only recently has the research studied the relationships between objective measures
and personnel decisions. Kwon (2006b) and Fee et al. (2006) are among the first
empirical investigations of the relationship between objective performance measures on
careers.
Kwon (2006b) studies the incentive provision role of bonuses and promotions that
encourage workers to invest in human capital. Kwon’s (2006b) model predicts that a
worker’s human capital acquisition is rewarded and, thus, motivated through a promotion
award. He also shows that the promotion-based incentive scheme that awards promotions
solely based on the latest performance immediately prior to the promotion decision is
efficient enough to induce a worker’s investment in skill collection, even without bonuses
during the term between promotions (i.e., within a job level) and even when performance
is verifiable (i.e., objectively measurable).
14
In addition, using the personnel records
15
of
13
Potential biases and manipulation of subjective ratings will be discussed later in this section.
14
Mookherjee (1984) shows that promotion-based incentives are superior to explicit bonus contracts only
when performance is non-verifiable.
18
insurance claim processors in a large U.S. insurance company, he finds that an objective
performance measure, or specifically the weighted number of claims processed per day,
affects promotion decisions positively.
Fee et al. (2006) investigate the careers of professional football coaches at a level 2
position
16
in the National Football League (NFL) teams, and report the positive effects of
objective performance measures
17
on promotions to head coach positions. However, they
attribute the findings to a different theory from Kwon (2006b). In particular, unlike Kwon
that focuses on the incentive effects of promotions on human capital acquisition, they
instead view promotions primarily as a job-assignment mechanism and partly as an
incentive provision instrument. More importantly, it should be noted that they raise an
issue of organizational structure constraints that potentially affect the likelihood of
internal promotions and document relevant empirical findings. Specifically, they find that
in an external labor market, junior coaches with good individual performance records are
more likely to be hired to a level 1 position (i.e., a head coach) at another team. In
contrast, internal promotions of the coaches are found to be not related to the objective
performance of individual coaches.
According to their explanation, the contrast between internal and external labor
markets arises because a job opening for a head coach position scarcely occurs when a
15
The data include salaries, bonuses, job levels, and objectively measured performance.
16
Fee et al. (2006) refer to the head coach position as a level 1 position while junior coaches such as
offensive and defensive coaches are considered level 2 positions under the head coaches of each team.
17
The objective measure for an offensive (defensive) coach is the percentile rank in the points earned by
the offense (scored against the defense).
19
team performs well and the team performance is positively associated with junior coaches’
individual performance.
18
On the other hand, when a team performs poorly, the dismissal
of its head coach is likely. However, this does not necessarily lead to an internal
promotion of junior coaches because the dismissal of a head coach following poor
performance often invites another head coach from outside; see also Parrino (1997).
Taken together, the job opening process results in the lack of relationship between
individual performance and internal promotions. However, Fee et al. (2006) provide the
evidence that the promotion is awarded to coaches with better performance when a job is
opened for internal promotion.
Without question, firms use various types of objective performance measures. The
performance measures used in Kwon (2006b) and Fee et al. (2006) are only a few
examples and may not be representative of other objective measures.
19
So, there also has
been a strand of research turning attention to alternative and broadly used objective
performance measures such as accounting measures. In this vein, Blackwell et al. (1994),
Campbell (2008), and Cichello et al. (2009) examine the effects of accounting
performance on personnel decisions made to workers at managerial positions: senior
executives running subsidiaries (Blackwell et al., 1994) and divisions (Cichello et al.,
2009), and store managers in a fast food restaurant franchise (Campbell, 2008).
18
In other words, when a junior coach performs well, his team likely performs well, which reduces the
likelihood of head coach’s dismissal, or a job opening, and thus that of the junior coach’s promotion as
well—so-called “slot constraint.”
19
In addition, the research subject in Kwon (2006b) and Fee et al. (2006) are untypical in management
research: non-managerial, largely female, and relatively low-income workers in Kwon and professional
sport labor market in Fee et al. However, the fact does not devalue the implications of their findings.
20
Blackwell et al. (1994) examine the turnovers of bank subsidiary CEOs and
promotions of non-CEO senior executives at level 2 positions in bank holding companies
in Texas. Although Fee et al. (2006) do not cite, the important empirical findings of
Blackwell et al. (1994) precede Fee et al.’s. First, as to turnovers of subsidiary managers,
they find that subsidiary bank managers are more likely dismissed after a poor accounting
performance in ROA; this suggests a negative association between turnover and
accounting performance.
20
Second, given a job opening for a subsidiary bank manager,
internal promotions are more likely to be awarded to the candidates in subsidiary banks
with higher ROA. They also report that these internal promotions are often combined
with inter-subsidiary movements within the same bank holding company—being
promoted to a manager of a subsidiary bank other than his own subsidiary.
Cichello et al. (2009) inherit the central tenet of Blackwell et al. (1994) pertaining to
the influence of an organization’s ROA and personnel decision made to senior executives
responsible for the performance. Specifically, they study the personnel decisions
including promotions and dismissals of division managers in companies with multiple
reporting segments. Their findings are largely consistent with Blackwell et al.; the
negative impact of divisional ROA on managers’ dismissals and the positive impact of
relative form ROA on division managers’ promotions to corporate-level executives such
as CEOs and other C-suite executives.
20
There has been a large body of literature on the relationship between CEO or top-management turnover
and firm performance since the seminal works in mid-1980s including Coughlan and Schmidt (1985) and
Warner, Watts, and Wruck (1988); see Brickley (2003) for a brief summary of literature.
21
The similar positive effect of accounting performance measures on promotions and
dismissals is repeated in Campbell (2008). By and large, his findings regarding the effect
of accounting performance measures hardly differ from those from the above studies.
However, his study is unique at least in two points. First, he considers alternative
accounting performance measures other than ROA: such as profit, sales, sales growth,
and expenses relative to target of franchised fast food restaurants. Second, more
importantly, a groundbreaking contribution comes from his notion of the effects of
financial and non-financial performance measures on personnel decisions. In particular,
he finds that non-financial performance measures including service quality and employee
retention affect store managers’ promotions
21
even controlling for the effects of
accounting performance measures,
22
and that accordingly the positive association does
encourage managers’ effort allocation toward the non-financial performance measures.
2.2. Supervisors’ Use of Subjectivity in Personnel Decisions
Subjectivity in Performance Evaluation
Supervisors’ use of subjectivity in performance evaluations and ensuing behavioral
outcomes such as leniency bias, centrality bias, and favoritism have been well
documented in diverse areas including economics, sociology, psychology, and accounting
(e.g., Baker, Gibbons, et al., 1994; Baker et al., 1988; Bol, 2011; Gibbs, 1995; Ittner et al.,
2003; Kwon, 2006a; Medoff & Abraham, 1980; Moers, 2005; Prendergast & Topel, 1993,
21
However, they do not affect dismissals.
22
Note that non-financial objective performance measures are considered in prior studies as well but not
together with financial measures; see the discussion on Fee et al. (2006) and Kwon (2006b).
22
1996). Among the disparate studies, agency models rationalize the use of subjectivity as a
source of incremental and relevant information about employees’ actions in addition to
objective and/or quantitative measures—“informativeness principle” (e.g., Baiman &
Rajan, 1995; Holmström, 1979). Other benefits from using subjectivity include (1)
mitigating game-playing against (misspecified) compensation systems, (2) extending the
decision horizon, i.e., correcting a manager’s short-term orientation, and (3) adding
flexibility to formula-based incentive schemes and, thus, reducing uncontrollable
downside risks that employees, otherwise, would have borne (Baker et al., 1988; Gibbs et
al., 2004; Ittner et al., 2003).
Despite these advantages, subjective evaluations are unpopular among both
supervisors and subordinates (Baker et al., 1988; Lawler, 1971). This is especially so
when subordinates’ trust in their supervisor’s impartial and sound use of her discretion
has not been established (Baker et al., 1988; Gibbs et al., 2004; Lawler, 1971). Without
trust, subjective evaluations are less preferred by subordinates who are worried about
receiving unfair treatment and by supervisors who are concerned about facing potential
confrontation of dissatisfied subordinates. For example, in Ittner, Larcker, and Meyer’s
(2003) field study on the use of a balanced scorecard, the second most frequent
complaints (14 percent of responses) among managers and representatives in their
research site are relevant to subjectivity in their performance measurement system: the
system’s opacity and (potential) favoritism. Their research site company, in the end,
abolished a scorecard-based incentive plan and returned to a formula-based reward
23
system after experiencing many problems related to the use of subjectivity (Ittner et al.,
2003). This extreme case suggests that subjective evaluation systems require cautious
implementation such as the provision of a self-enforcement mechanism so that employees
are convinced of the supervisors’ fair and unbiased evaluation. So, adequate trust
between supervisors and subordinates is a prerequisite and, at the same time, a cure.
Subjectivity in Promotion Decisions
Promotion decisions involve a greater extent of subjectivity than performance
evaluations for other purposes. Given the two roles of promotions, sorting and incentive
provision, this point goes well with explanations of promotions as sorting mechanisms.
Subjectivity in promotion decisions for incentive provision complements objective
performance measures: producing additional relevant information about subordinates’
efforts and adjusting subordinates’ performance for uncontrollable adverse events. In
contrast, subjectivity in promotion decisions for sorting purposes extends beyond
complementary adjustments to objective performance measures. Gibbons (1998) notes
that when promotions are contributing to employees’ incentives for skill acquisition,
performance evaluation becomes “trickier” because “firms must now evaluate a worker’s
potential contribution to future firm value, rather than the worker’s realized contribution
to date.” So, subjectivity involving the job matching role of promotion culminates in a
supervisor’s personal judgment of subordinates’ overall abilities, skill sets, potential, and
fitness for a next higher level position. A potential problem arises from the fact that these
qualities are unobservable and non-verifiable. More importantly, unlike performance,
24
they lack appropriate reference points. So, problems of subjectivity, if any, may be graver
for sorting than for incentive provision.
Compared to well-established literature about the use of subjectivity in explicit
incentive contracts, research on subjectivity in promotion decision-making contexts is
rare. In fact, almost all promotion studies assume supervisors’ subjective evaluations of
subordinates’ abilities and skills. For example, supervisors’ performance ratings used as
an important determinant of promotion, for example, in Medoff and Abraham (1980) and
Baker, Gibbs, et al. (1994a, 1994b) involve subjectivity by definition. Prior literature is,
however, not clear about what factors are considered in assessing subordinates’
performance and how subjectivity works in promotion decisions.
Supervisor Biases Arising from Subjective Evaluations and Favoritism
Despite functioning to mitigate problems with objective measures, subjectivity is not
always effective. Prior literature has documented biases arising from the use of
subjectivity in performance evaluation and their potentially harmful influence on
incentive systems (e.g., Baker, Gibbs, et al., 1994a; Bol, 2011; Gibbs et al., 2004; Medoff
& Abraham, 1980; Moers, 2005; Murphy, 1992; Prendergast, 1999; Prendergast & Topel,
1993). In particular, it has been reported that supervisors’ subjective performance
evaluation leads to more generous ratings, “leniency bias,” and to more compressed
dispersion of ratings, “centrality bias,” than it would have been in an objective metric. In
addition, according to Moers (2005), the extensive use of combinations-of-measures, i.e.,
measurement diversity, and subjective evaluation is positively associated with the two
25
types of supervisor bias. Therefore, given that the combinations-of-measures approach,
e.g., balanced scorecard, is widely adopted in firms, the sophistication of contemporary
performance measurement systems may aggravate the problems raised by the biases,
weakening workers’ incentives because of the system’s inability to sort and reward good
performers.
Favoritism is another form of supervisor bias that has received relatively little
attention, compared to the other types of biases. Defined as supervisors’ personal
preference for certain subordinates over others, favoritism has been criticized for its
detrimental effects on people and organizations: stress, workplace conflicts, politics and
power struggles, inefficient decisions, and the loss of motivation and productivity (e.g.,
Longenecker, Sims, & Gioia, 1987; Prendergast, 1993b; Prendergast & Topel, 1996;
Tirole, 1986). In particular, literature based on agency theory has viewed favoritism as a
product of subjective performance evaluation that diminishes workers’ incentives.
Accordingly, prior studies have suggested that firms should remove the costs of
favoritism by limiting the supervisor’s power to exercise favoritism (e.g., Prendergast,
1993b; Tirole, 1986). For example, Tirole (1986) focuses his arguments on collusive
actions of supervisors and workers which harm the efficiency of contracts. In the study,
he describes supervisors’ incentives for discriminatory hierarchical coalitions, i.e.,
favoritism, and agents’ wasteful competition for attainment of favors, i.e., to be “Yes
Men” (Prendergast, 1993b) under multi-agent situations where agents’ performance is
observed only by supervisors and, at the same time, not verifiable.
26
So, a general consensus about favoritism, prior to Prendergast and Topel (1996), was
that favoritism is a supervisor’s bias that carries no positive values to organizations and,
therefore, should be eliminated. However, there are also different perspectives of
favoritism. For example, Prendergast and Topel (1996) recognize that favoritism is part
of a supervisor’s utility, and Kwon (2006a) argues that favoritism is hard to eliminate.
Prendergast and Topel’s analytical model (1996) advances the discussion of
favoritism; their model includes a supervisor’s utility from exercising bias and power to
exercise bias. More importantly, they identify the utility components of a supervisor: her
own pay and the pay of her subordinates. Each utility component represents the
supervisor’s incentive to improve performance by exerting comprehensive managerial
efforts and the influence of her (favorable and unfavorable) evaluation of the
subordinate’s performance, i.e., favoritism.
Obviously, the first component is positively associated with a supervisor’s own
managerial performance. Therefore, interest alignment between supervisors and
principals helps improve supervisors’ individual and organizational utility. On the
contrary, the favoritism or influence component has double-edged effects. First, it
enhances the incentives of supervisors and subordinates. A supervisor with influence can
extract utilities by exercising the influence itself and by mobilizing subordinates’ loyalty,
diligence, and support for her decisions. On the subordinates’ side, favoritism affects the
subordinates’ productivity and effort level. As long as a supervisor can derive utilities
from the influence, subordinates may well have incentives to pursue greater influence.
27
Second, a supervisor’s abuse of influence, such as explicitly unfair treatment against less
favored subordinates, results in loss of trust and demotivation. Influence can also cause
(potentially less able) subordinates’ blind support—more “Yes Men”
(Prendergast,
1993b). In any case, these results could eventually harm organizational performance. In
this regard, it is important that a principal maintains effective incentive and monitoring
arrangements that align supervisors’ interests with hers such that they would limit a
supervisor’s excessive exploitation of power. Kwon (2006a), to the contrary, shows that
(1) favoritism is endogenous, and (2) a strong incentive provision is not sufficient to stop
the decision-maker’s favoritism. This contrasts with the general agency literature’s view
that residual claims awarded to supervisors can generate incentives to make accurate
performance evaluations (e.g., Baker et al., 1988).
Supervisors’ Utilities from Better Performance and More Power
To summarize the above discussion, the use of subjectivity in promotion decisions
provides additional information to supervisors and allows more discretion related to how
they use this non-verifiable information. As noted in Prendergast and Topel (1996),
supervisors can derive utilities from both more information and more discretion. With
more and better information, supervisors can improve the quality of decisions about, i.e.,
better sorting of, subordinates’ abilities and qualities to enhance their own performance
and, consequently, compensation. With more discretion, they can exercise more power
and influence, which promotes subordinates’ loyalty, trust, diligence, and support.
Whereas more information, in general, serves firms’ interests as well as supervisors’
28
interests, the effects of more discretion on organizational performance depend on how it
is used. In other words, firms would benefit (suffer only) from supervisors’ desirable use
(improper exploitation) of supervisor discretion. In this regard, firms should provide
appropriate incentives to deter supervisors’ detrimental exploitation of discretion and
promote constructive utilization of more information and discretion (Prendergast & Topel,
1996). It is especially important for high-ranking executives because weak (or the
absence of) incentives to do so at the top level will “cascade down the hierarchy” (Baker
et al., 1988). However, whether supervisors’ incentives to exercise favoritism are muted
by performance-based incentive contracts is an empirical question.
To address the question, a few recent studies investigate favoritism under incentive
contracts. Exploring a very interesting natural experiment research setting
23
—favoritism
of soccer referees in English soccer leagues
24
—Rickman and Witt (2008) examine the
length of injury time
25
in “close” matches where the goal difference is one, or in that
regard the outcome could have been altered by a few more time of play. Similar to the
previous findings from Garicano et al. (2005), Dohmen (2008), and Sutter and Kocher
(2004), they find that, prior to the incentive contracts, referees allocate significantly less
injury time when home teams are ahead. However, more importantly, they also report
23
English professional soccer leagues introduced professionalism of referees in the 2001-02 season, which
provides a natural experience opportunity. Since its inception, the performances of “professional” referees
are subject to evaluation of an independent monitor and, accordingly, are tied to the referees’ income and
status.
24
Favoritism or home bias of sports referees has been studied in various academic disciplines; see Foot
Note 3 of Rickman and Witt. Subsequent studies on favoritism use the research setting frequently; for
example, Buraimo, Forrest, and Simmons (2010) and Dawson and Dobson (2010).
25
The measure of favoritism is first used in Garicano et al. (2005).
29
that the inception of professionalism involving financial incentives for performance
including fair judgment at games removes the favoritism behavior of referees. Rickman
and Witt attribute the findings to the use of financial incentives that control, if not
suppress, referees’ (excessive) utility extraction from favoritism.
Another study by Bandiera et al. (2009) performs a field experiment
26
to investigate
the interplay between managerial incentives for managers, their social preference, and
workers’ productivity. Their findings provide important implications about managers’
managerial incentives to improve organizational performance. Without incentives,
managers may display blind favoritism toward socially connected workers regardless of
ability; favoritism dominates sound discretion. However, when economic incentives are
introduced, managers are encouraged not only to increase the overall level of managerial
efforts, but also to allocate more efforts toward those with high ability subordinates;
sound discretion dominates favoritism. Despite its clear implications, Bandiera et al.
(2009) is hardly applicable to white-collar employees whose abilities are highly non-
verifiable. Moreover, in their research setting, managers’ utility from favoring connected
and less able workers is apparently smaller than the increased compensation from
directing their efforts to support more able workers. However, with white-collar workers
involved, it is difficult to figure out net benefits of favoritism or increasing managerial
efforts over the other.
26
The field experiment is performed on a fruit farm where managers and workers are hired for a fruit-
picking season. The authors manipulate the compensation scheme for managers from fixed wages to fixed
wages plus a performance bonus based on the average productivity of the workers they manage.
30
Chapter 3. Accounting Performance and Personnel Decisions: Hypotheses
3.1. Personnel Decision-Making
This section describes a model of personnel decision-making, highlighting how
supervisor’s subjectivity is incorporated into a personnel decision outcome. Essentially, a
personnel decision is the final outcome of a supervisor’s subjective evaluation of a
candidate’s qualities. In personnel decision-making contexts, subjectivity involves three
distinct components: evaluation criteria, relative weights, and an adjustment factor.
Specifically, supervisors select measures and evaluation areas and apply relative weights
to the evaluation criteria, depending on the type of personnel decision. Then, the product
sum of the relative weights and evaluations of each criterion is subjected to adjustments.
Finally, supervisors finalize their decisions in consideration of contextual issues such as
the company’s financial standing.
This can be presented in notational form as follows:
Φ
ik
= 𝛬 𝑖 × � 𝜔 𝑗 𝜃 𝑗𝑘
𝑛 𝑗 = 1
= � � 𝛬 𝑖 𝜔 𝑗 𝜃 𝑗𝑘
�
𝑛 𝑗 = 1
𝑠 . 𝑡 . ∀ 𝜔 𝑗 , ∀ 𝜃 𝑗𝑘
∈ (0,1] and ∑ 𝜔 𝑗 𝑛 𝑗 = 1
= 1,
where Φ
ik
is the likelihood of a personnel decision (e.g., promotion) for an
individual evaluatee k in an organization i
27
in which a common adjustment ( 𝛬 𝑖 =
27
An organization can be either formal or informal. It can be a group of people who share homogeneous
characteristics, for example, promotion candidates at a hierarchical level.
31
∑ 𝜆 𝑙 𝑚 𝑙 = 1
)
28
is applied to all organizational members across the board. 𝜔 𝑗 is a supervisor’s
relative weight attached to an evaluation criterion j, and 𝜃 𝑗𝑘
is evaluatee k’s rating on the
evaluation criterion j.
Evaluation Criteria and Informativeness Principle
Selection (θ)
The informativeness principle (Holmström, 1979) suggests that any performance
indicator that provides incremental information should be incorporated into contracts to
improve the efficiency of incentives. Thus, in explicit incentive contracting contexts,
supervisors select evaluation criteria based on the informativeness of each criterion to
form a set of criteria ( Θ = � 𝜃 1 0
𝜃 2 0
… 𝜃 𝑗 0
� ). By and large, the principle holds true even in
the context of personnel decisions.
However, several distinct characteristics of personnel decisions allow supervisors
more discretion in the selection of evaluation criteria. First, personnel decisions are
hardly contracted; they are predominantly implicit. Second, indicators do not necessarily
pertain to performance. They may be indicators of worker qualities, such as knowledge,
experience, or skill sets, that are informative about success in a different task. Third, they
may fulfill functions other than incentive provision. For example, promotions are known
to serve an additional role: job matching.
Relative Weighting (ω).
28
A multiplicative specification may be also possible: 𝛬 𝑖 = ∏ 𝜆 𝑙 𝑚 𝑙 = 1
.
32
With multiple indicators involved simultaneously in evaluations for personnel
decisions, supervisors ascertain the relative weights ( 𝜔 𝑗 ) of the indicators. The weight
selection is in line with the agency theory. That is, supervisors place more emphasis on
indicators that are more informative about a candidate’s eligibility for promotion. The
informativeness of a measure is associated with the measure’s sensitivity and precision
(Banker & Datar, 1989), and goal congruence (Feltham & Xie, 1994).
Individual Performance as Part of a Performance Measure Set ( 𝜣 )
Prior studies have shown that good individual performance likely leads to promotion
awards. Good performance is evaluated through several different measures. Measures
discussed previously include performance ratings in annual performance reviews (Baker,
Gibbs, et al., 1994a, 1994b; Baker et al., 1988; Gibbs, 1995; Medoff & Abraham, 1980),
accounting performance of ROA measured at subsidiary firms and business units
(Blackwell et al., 1994; Cichello et al., 2009), and non-financial performance measures of
service quality and employee retention (Campbell, 2008).
All these studies, though, only focus on individual performance. Their discussions
pertain to the selection of a specific performance indicator(s) and a non-zero weight on
the indicator(s). In particular, Medoff and Abraham (1980), in their investigation of the
performance ratings and promotions, consider the result of evaluations, or ∑ 𝜔 𝑗 𝜃 𝑗𝑘
𝑛 𝑗 = 1
, as
a whole. As a result, they ignore a common adjustment factor ( 𝛬 𝑖 ) and further
decomposition of evaluation results into separate evaluation criteria ( 𝜃 𝑗𝑘
) and weights
( 𝜔 𝑗 ). On the other hand, Blackwell et al. (1994) and Cichello et al. (2009) examine a
33
direct association between accounting performance and personnel decisions, and consider
few variables other than ROA.
29
Therefore, their research settings coincide with a special
case: (1) where there are only a few, if not one, evaluation criteria (i.e., j close or equal to
one) and (2) where the evaluation criteria ( 𝜃 𝑗𝑘
) are closely linked to ROA, or where ROA
measures are highly informative about the promotion candidates’ qualities. Similarly,
Campbell (2008) focuses on several financial and non-financial performance measures to
estimate promotion opportunity but pays little attention to supervisors’ subjective
adjustment and other dimensions of workers’ qualifications for promotions.
Common Adjustment Factor ( 𝜦 𝒊 )
The adjustment factor ( 𝛬 𝑖 ) shifts the likelihood of a personnel decision upward or
downward.
This is the case in which supervisors apply a common inflation or discount
factor to adjust the evaluation results. As implied by the word “common,” an adjustment
factor affects the likelihood of career events of all the promotion candidates of an
organization or those of homogeneous characteristics en bloc. In this regard, the
adjustment is rather contextual. Examples include accounting performance measured at
high-level organizations, the number of contestants in an organization, and an
organization’s growth potential and capacity to feed promotions. For instance, in
promotion decision-making contexts, ceteris paribus, executives in organizations with
good performance (i.e., relatively high 𝛬 ) are more likely to be promoted than those in
organizations with poor performance (i.e., relatively low 𝛬 ).
29
In Blackwell et al. (1994), the only control variable that is not related to ROA is Log of Assets. Similarly,
in Cichello et al. (2009), it is Age of an executive.
34
3.2. Organizational Performance and Personnel Decisions
The focus of this study is the effects of accounting performance acting as one of the
common adjustment factors. The existing evidence on individual performance does not
clarify the effects of organizational performance on personnel decisions. With regard to
the use and the effects of organizational performance measures in explicit incentive
contracts, agency theory literature prescribes that compensation contracts with multiple
performance measures should assign a non-zero weight to an organizational performance
indicator to align workers’ effort with the organizational performance (Datar, Kulp, &
Lambert, 2001; Feltham & Xie, 1994; Holmström, 1979). As long as promotions are
awarded as incentives ultimately to improve the performance of an organization, it
sounds plausible that implicit incentives are also tied to organizational performance.
Promotions are a strong incentive instrument (Demougin & Siow, 1994; Gibbons &
Waldman, 1999). However, the efficiency of promotions as an incentive mechanism may
be compromised. On the one hand, personnel decisions involve considerable and
relatively persistent costs, which make organizations more committed to the decisions. In
general, promotions involve a monetary raise, additional perks, and accordingly more of
the other compensations. Further, compared to the transitory nature of other incentives
such as annual bonuses, promotions are relatively persistent. Moreover, inappropriate
decisions may lead to a loss of important human capital in the long-run.
30
On the other
hand, personnel decisions are subject to organizational constraints (Baker et al., 1988;
30
For example, a supervisor may dismiss a worker who has great potential, or may pass over for promotion
a capable person who may then decide to leave the firm for better prospects.
35
Fee et al., 2006; Milgrom & Roberts, 1992). For example, the number of open positions
in an organization and the organization’s financial capability to support costly and
persistent promotions limit the number of promotions that can be awarded in the
organization.
31
Thus, considering that personnel decisions require considerable
commitment from an organization and that organizational performance is a noisy
indicator for mid- to low-level executives who only partly contribute to the organizational
performance, implicit incentives may not be tied to organizational performance.
Given the opposite predictions, it is an open empirical question whether the
likelihood of promotions and dismissals of executives in an organization is associated
with the performance of the organization. The first hypothesis tests the effects of
accounting performance as a common adjustment factor—whether ROAs measured at
corporate and segment levels affect promotions and dismissals of executives in the
organizations.
H1a: Promotions (dismissals) of executives are positively (negatively) associated
with corporate ROA.
H1b: Promotions (dismissals) of executives are positively (negatively) associated
with segment ROA.
3.3. Purposes and Types of Promotions
The literature on promotions suggests that the effects of accounting (to a lesser
extent, objective or quantitative) performance measures on promotion decisions vary with
31
In a similar vein, Milgrom and Roberts (1992) and Baker et al. (1988) point out that effective promotion-
based incentive systems cannot be sustained without organizational growth.
36
the purpose of the promotions (Gibbons, 1998; Gibbs, 1995, 2008). Depending on the
purpose of the promotions, supervisors exploit different sources of information to
estimate subordinates’ abilities and/or input of effort, and weigh them differently to make
decisions. When promotions are used as incentive mechanisms, proxies for subordinates’
effort are related to their (contribution to) past performance—demanding less subjectivity.
On the contrary, promotions awarded to match employees’ skills and qualities to specific
higher-level positions (i.e., sorting) are typically more relevant to subordinates’ “potential”
contributions to future performance—calling for more discretionary judgments (Gibbons,
1998; Gibbs, 2008) or information sources other than accounting measures (Campbell,
2008). Thus, the usefulness of accounting performance depends on the type of promotion.
Grabner and Moers (2011), from a similar perspective, show that managers use
different evaluation criteria when promoting employees for different purposes. If
promotions serve as rewards, they are likely to be significantly associated with
accounting performance measures. This is because team performance and individual
performance are generally correlated to some extent. On the other hand, if promotions are
a means to sort workers based on qualities needed for an upcoming task, supervisors need
to evaluate subordinates’ qualities, rather than past performance, which may not be
correlated with accounting performance. This will lead to a reduced relative weight being
assigned to the accounting measure.
H2a: Corporate ROA affects promotions of executives only when promotions
involve hierarchical advancement.
37
H2b: Segment ROA affects promotions of executives only when promotions
involve hierarchical advancement.
3.4. Job Scope and Responsibility
An extensive amount of literature has documented a positive (negative) association
between CEO and top-management compensation (turnover) and performance (e.g.,
Antle & Smith, 1986; Barro & Barro, 1990; Gibbons & Murphy, 1990; Murphy &
Zimmerman, 1993). Executives shoulder greater responsibility for a business unit’s
performance than middle-level managers and other employees. However, executives
below the level of CEOs may lack good accounting performance measures unless they
have full discretionary authority to affect the performance of a subunit for which separate
accounting performance measures are available. In this regard, few studies have
examined the effects of accounting performance on the promotions of mid- or low-level
executives whose contributions to firm- or business unit-level performance are less than
those of high-level executives.
In explicit contracting contexts, Aggarwal and Samwick (2003) examine a question
relevant to the issue. They show that pay-for-performance sensitivity varies with
executives’ managerial responsibilities.
32
This indicates that performances measured at
32
Dividing managers into four groups of different managerial responsibility, Aggarwal and Samwick find
that the pay-for-firm-performance sensitivity is strongest for CEOs, followed by oversight executives, then
by executives without any responsibility, and lastly, by executives with divisional responsibility. Further,
they also find that when more precise, divisional performance measures are available, the compensations
for executives with divisional responsibility are more sensitive to the divisional performance measures than
firm performance measures. They attribute the pay-performance behavior to different degrees of managerial
responsibilities.
38
different levels are reflected in the compensations of executives with different levels of
responsibility.
More importantly, their finding suggests that an executive’s managerial
responsibility is associated with a performance measure’s informativeness about the
executive’s contribution to the performance outcome. To relate this to a specific measure
of performance, ROAs of high-level organizations are more informative about the
qualities of high-level executives with greater responsibility. In contrast, ROAs may be
inadequately informative about mid- and low-level executives whose managerial
decisions contribute to the organizational performances to a lesser extent than executives
with greater levels of responsibility. Following this rationale, H3 predicts that promotions
and dismissals of executives with greater levels of managerial responsibility are more
sensitive to ROAs of high-level organizations.
H3a: The sensitivity of promotions (dismissals) to corporate ROA is stronger
when a personnel decision is made for executives with greater job
responsibility.
H3b: The sensitivity of promotions (dismissals) to segment ROA is stronger when
a personnel decision is made for executives with greater job responsibility.
3.5. Intra-organization Interdependencies and Accounting Performance
In decentralized organizations, intra-organization interdependencies may determine
the incentive structure and thus the efficiency of compensation contracts (Bushman,
Indjejikian, & Smith, 1995). In an explicit incentive contracting context, as discussed by
39
Bushman et al. (1995),
33
the incentive compensations of business unit managers are tied
to the performance aggregated at an organizational level higher than a manager’s
business unit level, to the extent that a business unit’s actions affect another business
unit’s performance and consequently the parent organization’s performance.
A pertinent question in this regard is “Will this relationship still hold even under an
implicit incentive context?” Intra-organizational interdependencies may moderate the
relationship between promotions and accounting performance aggregated at a higher-
level organization. However, the underlying mechanism that establishes the
interdependencies’ moderating effect in personnel decisions may differ from what is
described by Bushman et al. (1995). They attribute the moderating effect of
interdependencies to the interrelatedness of actions and performances between
organizations that determines the aggregate measures’ informativeness.
In contrast, I propose that in personnel decision-making contexts, it is associated
with the transferability of workers’ knowledge, which includes professional expertise and
understanding of businesses and operations. As discussed earlier, promotions are subject
to organizational constraints, including the number of promotions that can be awarded in
an organization and the organization’s financial resources to feed promotions. However,
these restrictions may be moderated if the organization can “export” its workers—who
deserve promotions but cannot be awarded—to other affiliate organizations under the
33
They find that division CEOs’ compensations are associated with corporate-level performance measures.
The findings suggest that aggregate performance measured at a higher-level organization is incorporated
into incentive contracts to provide incremental information about collaborative actions of managers at
interdependent subunit organizations and ultimately to encourage such behavior.
40
same parent firm. Conceivably, cross-unit transfers occur more frequently when the set of
requisite knowledge and skills is compatible or transferable between units. Thus,
measured as the transferability of workers’ knowledge and skills, greater interdependency
allows organizations to grant more promotions than they can accommodate, as they tie
promotions with outgoing transfers. This increases the likelihood of promotion awards in
outperforming organizations or in banner years.
To hypothesize the relationship, promotions are expected to be more strongly
associated with corporate (segment) ROA when interdependencies between reporting
segments (within a reporting segment)
34
are higher, or cross-unit job transfers are more
frequent.
35
H4a: When more executives are reassigned from one reporting segment to
another, the association between corporate ROA and promotions becomes
stronger.
H4b: When more executives are reassigned from one business unit to another
within a reporting segment, the association between segment ROA and
promotions becomes stronger.
34
In this study’s dataset, accounting performance measures are available only at the top two levels of
organizations, i.e., corporate and reporting segment levels. While accounting performance measures are
undoubtedly available even below this level of organizations, they are unobservable outside the firm.
Therefore, this paper considers the cross-unit job transfers between reporting segments and those within a
segment.
35
While Bushman et al. (1995) measured the degree of interdependencies with product-line or geographic
diversification and intersegment sales, I use the number of cross-unit job transfers as a measure for intra-
organization interdependencies, assuming that frequent cross-unit transfers indicate compatibility of
executives’ local knowledge and skills and organizational interdependency.
41
Chapter 4. Discretion and Bias in Personnel Decisions: Hypotheses
4.1. Information Hypothesis vs. Preference Hypothesis
Subjectivity in personnel decisions allows more comprehensive evaluations of
subordinates’ multifaceted qualities for compensation and job matching, but it also brings
about an undesirable behavioral bias. In this regard, I distinguish and compare two
contrasting behavioral outcomes of supervisors’ subjective evaluation in promotion
decisions: discretion vs. bias.
36
To compare supervisors’ discretion and bias, I hypothesize their different effects on
the likelihoods of promotion and stay (an event opposite to dismissal). Each of a pair is
labeled “i” or “p” to indicate “Information Hypothesis” and “Preference Hypothesis,”
respectively. Table 1 summarizes the hypotheses in the following subsections.
4.2. The Effects of Presence of Professional Relationships on Personnel Decisions
Favoritism is exercised toward a certain subset of subordinates based on some form
of social relationship (e.g., Adler & Kwon, 2002; Bourdieu, 1985; Granovetter, 1973). At
the same time, social capital literature suggests that social relationships are the source of
information and that, with stronger ties, (truthful) information is communicated through
the channel (Adler & Kwon, 2002; Coleman, 1988). Capturing the two roles of social ties,
i.e., a source of favoritism and information, the social relationship in this study refers to a
36
Discretion in this study refers to supervisors’ professional, responsible, and sound use of their ability or
power to judge subordinates’ performances, abilities, skill sets, and other qualities requisite for next higher
level positions. On the other hand, bias pertains to favoritism, rather than leniency and centrality biases,
through which a certain subset of a supervisor’s subordinates are favored due to some form of social
connections.
42
professional social relationship within an organization
37
that is developed through a
supervisor’s promotion award.
If a supervisor has made a promotion award to a favored subordinate, the subordinate
is more likely to be favored in the next round of promotion contests. On the contrary, a
supervisor’s reasonable discretion does not allow another promotion simply due to the
established professional relationship.
37
It is a subset of an individual’s entire social network and is developed based on professional relationships
rather than on kinship, friendship, neighborhood-ship, or similar personal acquaintances.
Table 1: Predictions: Information-based Discretion vs. Favoritism
Information Hypothesis
Preference Hypothesis
Promotion
Stay
Dismissal
Promotion
Stay
Dismissal
ROA
+
+
−
Same to the Left
Hierarchical Level
−
+
−
No. of Execs of Same Type − + −
Presence of
Relationship
[H5] ?
+
−
+
+
−
Length of
Relationship
[H6] +
−
+
NN
a
+
−
ROA
* Relationship
[H7] NN
a
NN
a
NP
a
+
+
−
Level
* Relationship
[H8] +
+
−
NP
a
+
−
Same Type
* Relationship
[H9] NP
a
NP
a
NN
a
+
+
−
Presence of Relationship is an indicator whether the current immediate superior has promoted an
executive. Length of Relationship is the number of years pasted since the current supervisor’s promotion
award. ROA is computed as operating profit divided by total assets measured per reporting segment.
Hierarchical Level is constructed so that a higher numerical value indicates a higher level. Number of
Executives of Same Type is the number of executives who have the same property in terms of an
established social tie with their immediate supervisor (have/not have) at a hierarchical level in a reporting
segment.
a
NN and NP indicate “Non-Negative” and “Non-Positive,” respectively.
Table 1: Predictions—Information-Based Discretion vs. Favoritism
43
H5a
i
: The presence of a professional relationship does not affect the likelihood of
promotions.
H5a
p
: The presence of a professional relationship improves the likelihood of
promotions.
Retaining favored subordinates and extending their career within an organization is
also an outcome of favoritism. Not surprisingly, even a supervisor with sound discretion
is also likely to provide another promotion opportunity to the subordinates that she had
promoted. As her previous promotion awards were also based on information about the
subordinates following a comprehensive evaluation at that time, offering another round of
promotion contest is likely unless a subordinate’s ability or skill set is found to be
inadequate for the current job. So, regarding stay (dismissal) decisions, both explanations
predict the same outcome.
H5b: The presence of a professional relationship improves (decreases) the
likelihood of stays (dismissals).
4.3. The Effects of Length of Professional Relationships on Personnel Decisions
Once the relationship is established through a promotion award, a supervisor needs
some time to confirm her previous decision about a subordinate’s qualities for the current
job and to update information about the subordinate’s qualities for a next level job. At the
same time, staying at a job for some (but not too many) years allows a subordinate more
time and opportunities to communicate his qualities to his supervisor. In this regard,
length of a relationship can capture the amount and the quality of information
communicated and, further, a strong professional relationship has positive incremental
44
effects on both amount and quality. On the other hand, favoritism affects promotions
positively regardless of a supervisor’s learning or more information about favored
subordinates. However, whether any incremental effect of the length of a relationship in
addition to the presence of the relationship exists, is unclear. However, the effect is
unlikely negative.
H6a
i
: The length of the relationship improves the likelihood of promotions.
H6a
p
: The length of the relationship does not decrease the likelihood of
promotions.
A subordinate’s stay at a job level for an extended period of time without a renewal
of the relationship, i.e., another promotion, may indicate that his supervisor could not find
qualities necessary and/or appropriate for a next higher level job from him. So, the
Information Hypothesis predicts that being passed over for promotion over many years
without a subordinate’s successful communication of his qualities adequate for a next
level job decreases the chance of his career extension in an organization. In contrast, a
supervisor with incentives for favoritism is more likely to extend a (less capable)
subordinate’s career until she finds a good opportunity to promote the favored
subordinate.
H6b
i
: The length of the relationship decreases (increases) the likelihood of stays
(dismissals).
H6b
p
: The length of the relationship increases (decreases) the likelihood of stays
(dismissals).
45
4.4. The Effects of Group Performance on Personnel Decisions
A group performance is a noisy measure for an individual’s performance. Exploiting
the drawback of a group performance measure, a supervisor with favoritism incentives
finds an opportunity to promote favored subordinates (even those with low individual
performance) when they have a good group performance. This makes the supervisor’s
favoritism less detectable and, therefore, reduces potential confrontations among
disgruntled subordinates. On the contrary, simply having an established relationship will
not have positive incremental effects on the likelihood of promotions under the
Information Hypothesis. The same expectation applies to stays (dismissals).
H7a
i
: Conditional on the same group performance, an established professional
relationship does not decrease the likelihood of promotions.
H7a
p
: Conditional on the same group performance, an established professional
relationship increases the likelihood of promotions.
H7b
i
: Conditional on the same group performance, an established professional
relationship does not decrease (increase) the likelihood of stays (dismissals).
H7b
p
: Conditional on the same group performance, an established professional
relationship improves (decreases) the likelihood of stays (dismissals).
4.5. The Effects of Hierarchical Levels on Personnel Decisions
A hierarchical level, or equivalently the hierarchical distance from a supervisor, is
associated with the amount of knowledge related to subordinates’ qualities. Having won
several rounds of promotion competitions, executives at high ranks have shown their
outstanding abilities and qualities. In this regard, promotion candidates at a high level
may well be considered almost identical with negligible differences in ability.
46
Furthermore, promotion decisions for a high level position are very important because the
costs of incorrect decisions are huge, not only to the supervisor but also to the entire
organization, and because the decisions attract organization-wide attention. It is, indeed,
where an additional piece of information for better decisions is extremely valuable and
more discreetness is required. Therefore, a supervisor is more likely to make favorable
personnel decisions for those who she has promoted before because she has once reached
a positive conclusion about the subordinate’s qualities in the previous evaluation and,
through the establishment of a strong relationship, more and better information has been
communicated. In addition, other things being equal, multiple reports of positive
feedback fortify the relationship, allowing subordinate’s loyalty and trust.
On the other hand, a supervisor’s incentives to exercise favoritism toward
subordinates at high levels decrease because her favoritism is more likely to be detected
(due to greater organization-wide attention); and because the costs of favoritism far
exceed the benefits from it—such undesirable behavior, if detected, has substantial
detrimental consequences for her and her organization. However, compared to promoting
favored (but less capable) executives, decisions to retain them are less obtrusive and
intrusive to other employees. So, favoritism works for stay decisions. As a result, the
Information and Preference Hypotheses do not differ in their prediction regarding stay
(dismissal) decisions.
H8a
i
: An established professional relationship at a higher level improves the
likelihood of promotions.
47
H8a
p
: An established professional relationship at a higher level does not improve
the likelihood of promotions.
H8b: An established professional relationship at a higher level improves
(decreases) the likelihood of stays (dismissals).
4.6. The Effects of Promotion Competition on Personnel Decisions
Favoritism may improve the efficiency in selecting promotees. In practical situations
where more subordinates are to be evaluated, narrowing the range of promotion
candidates to those who are in a supervisor’s network and are no less capable than
promotion contestants outside the network makes thorough evaluation possible. This is
valuable, especially when there are a large number of candidates because a supervisor has
limited resources, e.g., time and effort, for a first-best optimal decision. Perhaps, the
heuristics-based approach to shorten the candidate list may help to overcome bounded
rationality.
As data for the amount of time and effort used in promotion decisions is hardly
available, more likely promotion of subordinates with a professional relationship among
an increasing number of promotion contestants would indicate a supervisor’s pursuit of
more efficient decision-making. From this perspective, the presence of a social
connection with a supervisor has stronger incremental effects on the likelihood of
promotion when there are more promotion candidates. In contrast, the Information
Hypothesis does not expect the effects.
H9a
i
: When there are more promotion contestants with the same type of
relationship with their supervisor, an established professional relationship
does not improve the likelihood of promotions.
48
H9a
p
: When there are more promotion contestants with the same type of
relationship with their supervisor, an established professional relationship
improves the likelihood of promotions.
H9b
i
: When there are more promotion contestants with the same type of
relationship with their supervisor, an established professional relationship
does not improve (decrease) the likelihood of stays (dismissals).
H9b
p
: When there are more promotion contestants with the same type of
relationship with their supervisor, an established professional relationship
improves (decreases) the likelihood of stays (dismissals).
49
Chapter 5. Research Setting and Data
To test the hypotheses, I analyze a panel dataset containing 4,657 executive-years
working at six subsidiary companies in the largest Korean conglomerate
38
during the
period from 2002 to 2007. The data are manually collected from the corporate annual
reports filed in the Korean electronic disclosure filing system.
39
The dataset provides a
unique research environment that allows me to investigate diverse career events and
important contextual variables. Korean companies’ annual reports contain information
about current and previous job roles, organizations, and, in most cases, educational
backgrounds of all the executives. Therefore, with several years of such data, all the
executives’ career paths could be traced. This paper differs from the prior literature in
managers’ career events in that:
(1) The research focuses on internal labor markets rather than external labor markets,
and
(2) The research investigates career events (e.g., promotions and dismissals) rather
than compensation.
5.1. Characteristics of Human Resource Management in Korean Conglomerates
40
Human resource management practices in Korean conglomerates have similarities
and differences when compared with their U.S. counterparts (Milliman, Kim, & Von
38
Frequently referred to as Chaebols (or Jaebeols).
39
DART: Data Analysis, Retrieval and Transfer System, http://dart.fss.or.kr
40
The Fair Trade Commission annually announces a list of large conglomerates to which some regulatory
laws will apply. The Commission designates the conglomerates based on the size of total industrial assets.
The regulations generally limit the power and expansion of the conglomerates by imposing limits on total
investment and mutual assurance, etc.
50
Glinow, 1993; Milliman, Nason, Zhu, & De Cieri, 2002; Pucik & Lim, 2001). Korean
conglomerates are comparable to global competitors in their area with respect to many
aspects such as size in sales and assets
41
, operations and business practices and, at the
same time, provide a unique opportunity to study internal labor markets in depth.
There are similarities between Korean and U.S. firms in incentive programs and
other human resource management practices. Despite apparent cultural differences
between Korea and the U.S., in both countries, individuals’ competence and performance
evaluations are important determinants
42
of rewards (including promotions) (Bae &
Lawler, 2000; Milliman et al., 1993; Milliman et al., 2002; Pucik & Lim, 2001; Steers,
Shin, Ungson, & Nam, 1990). The old prejudice that Korean (and some Asian)
companies would depend primarily on seniority, and social or personal connections (e.g.,
school or regional ties) in promotion decisions does not apply to, at least, large
conglomerates (Milliman et al., 1993; Pucik & Lim, 2001). Further, there are several
reasons to expect similarities in overall human resource management (HRM) practices.
For example, the U.S. is well known for its quality business services (e.g., consulting)
and the most advanced business education (e.g., MBA programs) in distributing the
values of best business practices. Surviving competitions, Korean conglomerates often
work with these consultants. Further, Korean conglomerates hire many graduates of U.S.
41
Table 3 presents some referable financial information about the conglomerate of research. Note that the
mean total assets and net sales for a reporting segment (not even a company) are USD 22.3 Billion and
USD 24.2 Billion which are well comparable to those of large U.S. companies (e.g., Fortune 100).
42
Milliman et al. (2002) find that performance evaluations are even a stronger determinant of promotions
in Korean companies than in U.S. companies.
51
business programs (e.g., MBAs) and, as part of their professional education program,
support selected employees with MBA programs. This practice allows them to indirectly
absorb cross-cultural practices. These practices, combined, suggest that Korean
conglomerates use much the same human resource management practices as those of U.S.
firms.
43
In contrast, some characteristics distinguish Korean and U.S. conglomerates. First,
Korean conglomerates
44
elaborate and utilize their internal labor markets more than
external labor markets to manage their human capital. It is also notable that a
conglomerate runs a centralized, corporate-wide human resource management program
across constituent companies through a conglomerate headquarter-level organization.
Interestingly, Korean conglomerates have such attributes that are compatible with Pfeffer
and Cohen’s (1984) determinants of internal labor markets: emphasis on firm-specific
skills (more exactly, corporate institution), higher technological change, governmental
(interpreting it as authoritative), greater labor scarcity, larger establishment, etc.
Second, Korean conglomerates make promotion announcements typically on an
annual basis. The dates of announcement are concentrated between the calendar year-end
and the year-beginning, although this varies across conglomerates and years.
45
This
feature helps to identify promotions through annual reports and to search for relevant
43
Further arguments about this topic are beyond the scope of this study.
44
The Fair Trade Commission (FTC or the Commission) defines a conglomerate as a group of legally
independent companies that are under practical and material control of a person or an organization.
http://groupopni.ftc.go.kr/ogroup/guide/guide_01.jsp?muduCount=menu_01
45
Recently, a few conglomerates have begun making non-periodic executive promotions which are
generally minor in scale and scope, and therefore insignificant.
52
newspaper articles and press releases, and to validate the tenure and the length of
relationships with executives’ supervisors.
Third, from an employee’s perspective, executives are provided with attractive
benefits so that they prefer staying in a conglomerate as long as they can.
46
Executives
are usually re-assigned and transferred to another position across companies within a
conglomerate, which is a more frequent (and preferred) case than a move to external
companies. Related to the first point, this characteristic strengthens the efficiency of a
conglomerate’s internal labor market focus as long as executives consider external
markets as a secondary or inferior alternative.
5.2. Research Setting: The Careers of Executives in a Korean Conglomerate
The research site for this study is Samsung Group, which is the largest Korean
conglomerate of the ones that had been classified as large conglomerates by the Fair
Trade Commission (FTC). As of the end of 2009, the conglomerate consists of 64
subsidiaries, of which 28 companies filed annual reports in DART. Out of these 28
companies, six companies
47
met selection criteria. These subsidiaries operate in diverse
industries spanning semi-conductors, display panels, telecommunication devices and
46
Several reasons may explain this observation. To list a few, being an executive and remaining in the
position in a large conglomerate is considered a professional success which provides decent compensation,
a positive social reputation, and a secure retirement plan. In addition, an executive’s job movement from
one conglomerate to another, even comparable in terms of corporate size and reputation, rarely occurs.
Thus, external markets provide positions only in lower-tier companies. In other words, executives who
move outside a conglomerate are seldom provided with a career opportunity which offers social and
economic benefits greater or, at least comparable, to those in a conglomerate. Rather, it is generally a less
preferred alternative than staying in a conglomerate. Further arguments are beyond the scope of this study.
47
Six subsidiaries are Samsung Electronics, Samsung C&T, Samsung Electro-Mechanics, Samsung Heavy
Industries, Samsung SDI, and Cheil Industries.
53
equipment, consumer electronics, industrial electronic devices, electronic components,
construction, civil engineering, heavy industry goods, ship-building, trading, chemical
products, fashion, and textile, among others.
Unlike in U.S. firms, an executive’s job title in a Korean company provides two
pieces of information: his or her hierarchical rank and his or her role in an organization.
48
The titles representing executives’ hierarchical levels in the conglomerate are common to
all of its subsidiaries. In the dataset, seven titles for executives’ hierarchical ranks are
identified.
49
Executives in these companies serve a variety of roles including, but not
limited to, CEOs and chief executives at the corporate level, at the business group or
division level, or at regional headquarters or foreign subsidiaries, plant managers, and
high-level professionals—such as lawyers, researchers, and other experts.
There are several notable features of executive promotions in Samsung. First, its
HRM practices are comparable to those of U.S. firms. Several studies have documented
the conglomerate’s successful transition in HRM policies to potential competence and
performance rather than prioritizing education history and seniority (Bae & Lawler, 2000;
Kim & Briscoe, 1997; Pucik & Lim, 2001; Yu & Rowley, 2009). Second, promotions to
an executive position in conglomerates are extremely competitive. For example, the
48
Compensations largely depend on the hierarchical rank. The “dual” structure is typical in Korea and
Japan (Ariga, Ohkusa, & Brunello, 1999; Pucik & Lim, 2001). For more detail, see Appendix A.
49
They are Hoejang, Buhoejang, Sajang, Busajang, Jeonmu, Sangmu, and Sangmu-bo. There are only five
to six persons in the two highest ranks throughout the conglomerate. Moreover, Hoejang is the person who
exercises practical control over the whole conglomerate and is removed from the sample. For this reason,
the three highest levels are collapsed into a single level for analysis purposes. Further, many of these top-
level executives are eliminated from the sample as they are the ones at the peak of each corporate hierarchy
and therefore, not subject to further promotion. For more details, see Appendix B.
54
likelihood of promotion to an executive
50
in Korea’s 100 largest companies is less than
one percent and it takes an average of 21 years to acquire an executive title (KEF, 2011).
Third, both the conglomerate and its executives consider external labor markets
secondary and inferior alternatives to internal labor markets. For example, a news report
from Money Today, a prominent economic and business news provider in Korea, reflects
the conglomerate’s recent change from strict closure against an outside executive market
to exploration of the outside human resources. According to the news article, it is
extraordinary that two executives who had been hired from an external executive market
were promoted to Sajang
51
in the annual promotion announcement for 2011 (Sung,
2010).
52
5.3. Sample Selection and Data Collection
Corporate annual reports for the fiscal years of 2001 through 2008 that are available
at DART were downloaded to acquire executives’ profiles. Unlike in the U.S., Korean
companies’ annual reports present brief profiles of all the executives in a reporting
company. This unique feature enables this study’s longitudinal tracking of executives’
careers. In total, 13,301 executive-years of profiles for the 28 companies were manually
collected. As tracing executives’ career paths, changes in organizational structure, and
50
Since the beginning of a career as a new college graduate
51
Sajangs correspond to CEOs. For more details, see Appendix B.
52
Nine executives were promoted to Sajang during the year. According to the newspaper report, these two
executives have worked for the conglomerate for six and seven years, respectively, since their recruitment
as executives. This indicates that they had been recruited at a low level—highly likely to be lower than
Busajang or Jeonmu. So, they were not directly recruited from the external CEO market. It also reports that,
in the conglomerate’s history, there has been only one CEO recruited directly from the outside.
55
further data processing is a labor intensive endeavor, I limit my attention to a manageable
sample size. I restrict the sample to firms that (i) have more than, on average, 30
executives per year,
53
(ii) have at least five years of annual reports during 2001 and 2008,
(iii) are not financial institutions, (iv) are not joint ventures with companies outside the
conglomerate, and (v) have required data. Further, executive-years (i) whose hierarchical
level cannot be identified or properly inferred, (ii) who are immediate family members of
53
The number of executives of a company is calculated as its average during the period between 2001 to
2008
No. of
companies
No. of
executive-years
No. of executive-years collected
28
11,980
Executive-years in
a
- companies with less than 30 executives
per year and with less than 5 years of
measurement
17 1,762
- joint ventures 1 63
- companies for which organization charts
are not available
1 256
- financial firms 3 1,318
- Year 2008 -
836
4,235
Executives with unidentifiable
hierarchical titles
-
365
Known owner family members
-
62
Executives with other missing variables
-
2,039
Heads of Level 1 organizations
-
48
Executives for Year 2001
-
574
Final number of executive-year
observations
6
4,657
a
Companies are excluded in the following sequence.
Table 2: Sample Selection
56
the person of material control, and (iii) whose other necessary information is missing are
removed from the sample. Lastly, the data for 2001 and 2008 are removed from the
sample because they were used to identify career events for 2002 and 2007. As a result,
4,657 executive-year observations from 1,251 unique executives in six companies
comply with these conditions. Table 2 describes the sample selection process.
Executive Profiles. Typical profile data include: name, board directorship, date of
birth, hierarchical title, current and/or previous positions (responsibility/job title), and
education.
54
A unique ID created as the combination of date of birth and name allows me
to track the career changes of executives as long as they stay in a conglomerate.
Executive profiles are used to identify executives’ career events, i.e., promotions, cross-
unit transfers, and dismissals.
Organization Charts. Annual reports provide organization charts in which
organizations at the top three levels, including the president and the CEO at the top level,
can be identified. Figure 1 provides an example of an organization chart. Next, with the
information about an executive’s organization membership provided in annual reports,
the organizations identified from the charts are matched to each executive-year. Then, the
heads of organizations at each level per year can be identified. The information about
organization heads, in turn, is incorporated into the executive-year dataset, (1) matching
an executive-year’s current organization membership and the corresponding
54
I also collect newspaper articles and press releases of the conglomerate’s annual executive promotions to
reconcile discrepancies between the executives’ profiles in the annual reports and the press releases.
57
organization’s heads and (2) matching the executive-year’s organization membership and
the managers at the time of previous promotion.
55
Financial Information. In addition to the executives’ profiles, financial information
for each reporting segment is collected from annual reports. The financials are collected
for reporting segments for businesses based on product groups, with the exception of the
largest subsidiary whose reporting segments both in product groups and regions are
collected. This is because there are a significant number of executives such as CEOs,
CFOs, and plant managers in world-wide regional headquarters and subsidiaries, as
compared to other subsidiaries. These reporting segments do not necessarily correspond
to Level 2 or Level 3 organizations specified in organization charts.
56
Reporting
segments mostly correspond to Level 2 profit center organizations. For these segments,
financials are assigned to the exact-match organizations. A few other reporting segments
are the combinations of two or three Level 2/3 organizations. A likely cost center Level 2
organization is given its corporate financials. As a result, all the executives who belong to
Level 2 and lower-level organizations share the same organizational performance at a
reporting segment level that generally corresponds to Level 2 in organization charts.
55
For those executives with previous promotions prior to 2001, the best approximation is to consider their
organization heads as of 2001 as the ones in the year of previous promotions. Therefore, all executives
listed for 2001 appear to have their current organization heads along their hierarchy the same as the heads
at the time of previous promotion regardless of the actual year of previous promotions. For this reason, the
data for 2001 are removed from the sample.
56
There are two possible reasons. First, not all these Level 2 and Level 3 organizations are profit centers or
revenue generating organizations: for example, Corporate Executive Staff and CTO Strategy Office in
Figure 1. Second, even if they may generate revenues, the revenues are (1) primarily internal, (2) vested in
other businesses, or (3) insignificant, in terms of the size, to be reported in a separate segment.
58
a
The organization chart of Samsung Electronics for 2005 is shown.
Technology
System Research
Institute
Mechatronics
Research Institute
CTO Strategy
Office
Digital Media
Business
Visual Display
Div.
Digital Video
Div.
Computer
Systems Div.
Digital Printing
Div.
Digital Media
Research Inst.
Telecommunicati
on Business
Wireless Telecom
Div.
Network Div.
Telecommunicati
on Research Inst.
Appliances
Business
CEO
System Appliance
Div.
Semiconductor
Business
Memory Div.
System LSI Div.
Storage Div.
LCD Business
Corporate
Executive Staff
Domestic Sales
Div.
Global Marketing
Office
Design Center
CS Center
Digital Solution
Center
Suwon Business
Support Center
Level 1:
Level 2:
Level 3:
Figure 1: An Example of an Organization Chart
a
59
5.4. Variables
Promotions and Dismissals
The primary dependent variable of the research is the likelihood of promotions, stays,
and dismissals for an individual executive. To identify the events of interest, I analyze
each executive’s profile information provided in annual reports. Analyzing the data, three
types of promotions are identified. Type 1 promotions refer to executives’ upward
movement in the hierarchy. Type 2 and 3 promotions involve promotions to head an
organization. Type 2 (3) promotions apply to cases where executives are promoted to
head positions (profit center manager positions) from non-head positions (cost center
manager positions). The opposite types of job assignments are also observed: release
from head positions (Type 2) and from profit center manager positions (Type 3).
57
In particular, a hierarchical advancement is identified by comparing the hierarchical
level between two consecutive years. Specifically, I count events as promotions when an
executive earns a new title for a higher rank. To identify an appointment to head an
organization (from a non-head position), the role title of each executive-year is analyzed
and coded as a one if the executive takes a head role at any level (even beyond the top
three level organizations specified in the organization charts). Then, the latter type of
promotion is detected by changes from zero to one in the coded role title. On the other
hand, dismissals are identified when an executive’s role is changed to an advisory
57
Note that the release from a supervisory job is not necessarily a demotion.
60
position
58
or when an executive’s profile is no longer available. Promotions and
dismissals are coded as an indicator variable, assigning a one to promotions (dismissals)
in the following year.
It is notable that, while Type 1 promotions generally induce a non-trivial increase in
compensation, Types 2 and 3 promotions involve substantial changes in job
characteristics (i.e., increase in responsibility and job scope). However, Type 1
promotions are not exclusive of Type 2 or 3 promotions. In this regard, I add another
class of promotions, labeled as “Type 1 Only.” This type of promotion refers to
hierarchical advancements that do not involve Type 2 or 3 promotions simultaneously.
Type 1 Only promotions are associated with a non-trivial increase in compensation, but
do not involve significant changes in job characteristics.
Accounting Performance
The primary independent variable of interest in the first part of the dissertation is
accounting performance measured in ROA. In this study, ROAs are observable at two
levels: the company and the reporting segment. Corporate ROA is computed as a
company’s net income divided by total assets while reporting segment ROA is the
segment’s operating profit divided by the segment’s total assets. In addition to unadjusted
ROA, I also use a relative measure for ROA. Relative ROA is a quintile group, ranging
from one to five, based on a reporting segment’s ROA in a company by year, with a high
numerical value assigned to good performance.
58
Advisory positions are offered to “retired” executives. These retired executives continue to receive
70~90% of the salaries in their last active positions for two to three years depending on the hierarchical
ranks at their retirement.
61
Job Scope and Responsibility
Job responsibility is measured in two ways: hierarchical rank and management
position. First, I split the full sample into two groups based on the current hierarchical
rank and run separate regressions with the split samples. High-level executives are those
at the top three levels, accounting for about 22 percent (1,044 executive-years) of the full
sample, while low-level executives are the bottom two levels, of which the titles are
Sangmu and Sangmubo. Second, I identify the executives managing Level 2 or 3
organizations. They comprise about the same proportion (22%, 1,030 executive-years) of
the full sample as the high-level executives. Based on the classifications, I assume that
high-level executives and executives supervising Level 2 or 3 organizations have wider
job scope and greater responsibility in terms of their contributions to aggregate
accounting performances at high-level organizations.
Intra-organization Interdependency
Intra-organization interdependency is proxied by the frequency of cross-unit job
transfers. A cross-unit job transfer is defined as an executive’s movement from one
organization to another within the conglomerate, regardless of whether the shift involves
a promotion. There are four levels of organizations (i.e., company, reporting segment,
Level 2, and Level 3 organizations) where cross-unit job transfers can be identified.
Among these transfers, this research focuses on job transfers between and within
reporting segments. This is because a reporting segment is the lowest level organization
where accounting performances can be measured in the dataset.
62
The frequency variable is operationalized in two ways: (1) the relative frequency
computed as the number of events divided by the number of executives within a reporting
segment (Ratio), and (2) the partitioned ranges of relative frequency (Group). For
example, the relative frequency of job transfers between reporting segments (Cross-
Segment Transfer) is computed as the frequency of outgoing cross-segment transfers
from a segment divided by the number of executives in the segment. On the other hand,
that of job transfers within a reporting segment (Within-Segment Transfer) is computed as
the frequency of Within-Segment transfers within a segment divided by the number of
executives in the segment. In addition, these ratios are partitioned into three groups based
on their range: (1) zero, (2) between zero and the pooled-median of the ratios,
59
and (3)
greater than the pooled-median of the ratios. As a result, Cross-Segment Transfer and
Within-Segment Transfer in the second specification are categorical variables of three
groups.
Relationship with Subordinates
Supervisors may also consider their social relationships with their subordinates. In
Chapter 4, I compared the rationales for supervisors’ use of favoritism and discretion
based on information. For empirical comparison of these two competing hypotheses, I
use the current supervisor’s previous promotion award to an executive as a proxy for the
supervisor’s social tie that not only arouses a supervisor’s favoritism incentive but also
provides better and more information about a subordinate.
59
The pooled-median of the ratios is calculated except the observations with zero values.
63
Considering that bridging and bonding are two primary activities that determine the
structure of social networks and the intensity of social ties (Adler & Kwon, 2002;
Bourdieu & Wacquant, 1992; Fukuyama, 1997), initial (consecutive) promotion award is
an important bridging (bonding) activity that a supervisor can make to initiate (strengthen)
her social ties with promotees. So, the “presence of social ties” in this paper does not
refer to its literal definition. Rather, it represents a stronger relationship that is established
or strengthened through a specific social and professional event, or a promotion; and a
more generous subjective evaluation due to such a relationship.
To determine the presence of a social relationship, I identify the years of the last
hierarchical advancement for each executive, and then identify the superiors along the
hierarchy at the time of promotion. Presence of Relationship is constructed as an
indicator of whether an executive’s immediate supervisor at levels one through three
organizations
60
within his or her current hierarchy was included in the previous chain of
command that had promoted him or her. The heads of levels one through three
organizations are the supervisors who have formal and/or real promotion decision-
making authority. Among these supervisors in the hierarchy, the most immediate
supervisor is assumed to have the greatest influence on promotion decisions and I assume
that influence decreases as hierarchical distance increases. Following this rationale, the
variable for the presence of a social tie (i.e., potentially more generous subjective
evaluation of performance and qualities) is constructed as an indicator of whether an
executive’s immediate supervisor at levels three organizations within his current
60
The head of a Level 1 organization (i.e., CEO), in case that executive is a head of a Level 2 organization.
64
hierarchy was included in the previous chain of command that had promoted him. Length
of Relationship, measured as the number of years since the last promotion, is used as a
proxy for the amount and the quality of information that has been communicated between
a supervisor and a subordinate since the establishment or renewal of the relationship.
Other Variables
Size and Growth of a Reporting Segment. Sales and growth in sales are included
to capture a reporting segment’s capacity to feed promotions.
Hierarchical Level. The variable is included to capture the decreasing promotion
opportunities; the number of positions becomes significantly lower as the hierarchical
level increases. Numerical values are assigned to each hierarchical level, one for the top
level and increasing toward lower levels. Then, by multiplying minus one, they are
inversed for intuitive interpretation of coefficients in regression models.
Number of Executives at a Level. The number of executives, at a given level in a
reporting segment, accounts for the degree of competition for limited seats in higher-level
positions. The variable controls for the level of competition for promotions.
Education and Professional Experience. Education level has been used as a proxy
for workers’ innate abilities as education level is known to be positively associated with
the likelihood of promotion (e.g., Baker, Gibbs, et al., 1994a; Lluis, 2005). Prior literature
has also used proxies for professional experience such as age and tenure in a company or
at a job. Age is computed as the year of an executive profile subtracted by the birth year
while Tenure is the number of years since the last hierarchical advancement, i.e., Type 1
promotion.
65
Professional Background. I also consider an executive’s professional discipline.
Variables for each discipline are coded as one, indicating administration, marketing and
sales, and engineering, respectively; or zero otherwise.
Speed of Promotions. The speed of promotions to date is included to control for the
presence of the “fast track,” an empirical regularity reported in several promotion studies
(Ariga et al., 1999; Baker, Gibbs, et al., 1994b). The measure captures how fast an
executive has been promoted to the current hierarchical rank, calculated as follows:
𝑆𝑝 𝑒 𝑒 𝑑 𝑖𝑡 =
(6 − 𝐻 𝑖 𝑒𝑟𝑎 𝑟𝑐 ℎ 𝑖𝑐𝑎𝑙 𝐿 𝑒 𝑣𝑒 𝑙 𝑖𝑡 )
𝐴𝑔 𝑒 𝑖𝑡 − 𝑇 𝑒 𝑛𝑢𝑟 𝑒 𝑖𝑡
where subscripts i and t denote an individual executive and a year, respectively;
Hierarchical Level, Age, and Tenure are as defined earlier in this section.
5.5. Description of Personnel Decisions in Research Sites
Descriptive Statistics
Table 3 provides descriptive statistics for the key variables measured at the
individual executive-year (Panel A), and reporting segment and corporate levels (Panel
B). The sample contains 4,657 executive-years of profile information for 1,251 unique
executives working at six companies in a large Korean conglomerate during the period of
2002 through 2007. Therefore, given the annual promotion cycle, an executive may have
at most six opportunities for promotions. There are 906 promotions (19% of the sample)
during the period. Obviously, hierarchical advancement is more frequent (616 times) than
promotions to head positions of any level organizations (Type 2 promotion; 353 times)
and to profit center manager positions (Type 3 promotion; 161 times). Positive personnel
66
decisions including promotions and stays account for 88 percent of the cases (4,103
times). On the contrary, negative decisions, or dismissals, (12%) are less frequent than
promotions in general (19%), but almost as frequent as Type 1 promotions (13%).
The table also shows that for 62% (2,904 times) of the decision-making situations, a
supervisor makes personnel decisions over subordinates to whom she awarded a previous
promotion. Sixty nine percent of total executive-years are heads of any level
organizations. On average, executives are in their late forties and have stayed at a level
for around three years since their last personnel decision. About half of the executives
have general management jobs—not explicitly related to marketing/sales and
engineering/research. Twenty six percent of the executives have advanced degrees for
their highest education level.
Panel B presents corporate and reporting segment level variables. The mean (median)
sales of a company and a reporting segment are approximately 53.20 (73.30) billion USD
and 24.21 (14.49) billion USD,
61
respectively. The mean (median) ROAs at company and
reporting segment levels are 7.16% (8.24%) and 8.76% (8.49%), respectively.
Table 4 reports the correlations between variables. Panel A provides the key
independent variables, other than accounting/financial performance measures, used in
promotion/dismissal estimation models. On the other hand, Panel B shows the
correlations between promotions and diverse financial/accounting performances at
corporate and reporting segment levels. The correlations support that promotions are
61
Converted from South Korean Won (KRW) at an exchange rate of 1,100 KRW/USD. The average
exchange rate for the sample period from 2002 to 2007 (for October of 2012) is 1,082.80 (1,104.90)
KRW/USD.
67
positively associated with good organizational performance. Among these aggregate
performance measures, I use ROAs at corporate and reporting segment levels.
62
62
Different measures of performance would not yield significantly different results. In addition, an ROA
measure provides consistency and comparability with prior literature such as Blackwell et al. (1994) and
Cichello et al. (2009).
68
Table 3: Descriptive Statistics
Panel A: Executive-Level Variables
Mean
Std Dev
25%
Median
75%
Personnel Decisions
- Promotion
0.19
0.40
0.00
0.00
0.00
Hierarchical advancements
0.13
0.34
0.00
0.00
0.00
Appointment to a head
0.08
0.26
0.00
0.00
0.00
Appointment to a profit center head 0.03
0.18
0.00
0.00
0.00
- Release from supervisory positions
From a supervisory task 0.05
0.23
0.00
0.00
0.00
From a profit center manager position 0.02
0.13
0.00
0.00
0.00
- Dismissal
0.12
0.32
0.00
0.00
0.00
Job Responsibility
High-level executives
0.22
0.42
0.00
0.00
0.00
Head of Level 2 or 3 organizations
0.22
0.42
0.00
0.00
0.00
Head of profit centers
0.19
0.39
0.00
0.00
0.00
Cross-Unit Transfers
- Cross-segment transfer
Ratio
0.06
0.08
0.00
0.43
0.09
Group
0.97
0.83
0.00
1.00
2.00
- Within-segment transfer
Ratio
0.06
0.10
0.00
0.01
0.06
Group
0.80
0.83
0.00
1.00
2.00
Relationship
Presence of Relationship
0.62
0.48
0.00
1.00
1.00
Length of Relationship
1.50
1.72
0.00
1.00
2.00
Age
49.64
3.75
47.00
49.00
52.00
Tenure
2.97
1.83
1.00
3.00
4.00
Job Area
Administration
0.45
0.50
0.00
0.00
1.00
Marketing/Sales
0.24
0.43
0.00
0.00
0.00
Engineer/Technician/Developer
0.31
0.46
0.00
0.00
1.00
Education
College degree or below
0.74
0.44
0.00
1.00
1.00
Master’s degree
0.18
0.38
0.00
0.00
0.00
Doctorate degree
0.08
0.27
0.00
0.00
0.00
Speed of Promotions
3.96
1.87
2.22
4.00
4.55
69
Table 3 (Continued)
Panel B: Organization-Level Variables
Mean
Std Dev
25%
Median
75%
No. of Executives in a Reporting
Segment
59.69
43.72
29.00
44.00
77.00
No. of Executives in a Level 20.13 16.07 7.00 16.00 29.00
Degree of Hierarchical Levels
4.07
1.00
4.00
4.00
5.00
Corporate-Level Financials
ROA (%)
7.16
4.71
3.65
8.24
10.01
Total assets
a
49.05
34.78
10.63
62.73
73.97
Sales
a
53.20
38.58
9.72
73.30
78.03
Reporting Segment Level Financials
ROA (%)
8.76
12.85
3.45
7.49
11.07
Relative ROA 2.26 1.16 1.00 2.00 3.00
Total assets
a
22.29
27.84
3.15
9.36
28.26
Net sales
a
24.21
28.78
5.05
14.49
22.90
Sales growth
11.92
49.40
0.10
7.84
23.13
Promotion is an indicator for an executive’s hierarchical advancement (Type 1), an executive’s
appointment to a head of an organization (from a non-head position) (Type 2), or an executive’s
appointment to a head of a profit center (Type 3) in the following year. Dismissal is an indicator for an
executive’s appointment to an advisory position or for an executive’s profile being unavailable in the
following year(s). High-Level Executives is an indicator for executives at hierarchical levels of one
through three. Cross-Segment Transfers refer to executives’ job reassignments from one reporting segment
to another. Within-Segment Transfers refer to executives’ job reassignments from a Level 3 organization
to another “within” a reporting segment. Presence of Relationship is an indicator for the current
supervisor’s having awarded an executive’s previous promotion. Length of Relationship refers to the
number of years since the last promotion awarded by the current immediate supervisor. Age is an
executive’s age calculated as the year of annual reports minus the year of birth. Tenure is the number of
years since the last Type 1 promotion (i.e., a hierarchical advancement). Job Area is categorized into three
groups: executives in (1) general administration and management, (2) marketing and sales, and (3)
engineering and research. Education is 0 for college graduates or below, 1 for master’s degree holders, and
2 for doctoral degree holders. Speed of Promotion captures how fast an executive has been promoted to the
current hierarchical rank, calculated as
( 6 − 𝐻 𝑖 𝑒 𝑟𝑎 𝑟𝑐 ℎ𝑖𝑐 𝑎 𝑙 𝐿 𝑒𝑣 𝑒𝑙
𝑖𝑡
)
𝐴𝑔 𝑒 𝑖𝑡
− 𝑌𝑒 𝑎𝑟 𝑠 𝑎𝑡 𝑡 ℎ 𝑒 𝐿 𝑒𝑣 𝑒𝑙
𝑖𝑡
. Number of Executives in a Reporting
Segment and Number of Executives at a Level are the numbers of executives in a segment and at a
hierarchical level in a reporting segment, respectively. Hierarchical Level is constructed so that a higher
numerical value indicates a higher level. Corporate ROA is computed as net income divided by total assets
per company. Segment ROA is computed as operating profit divided by total assets measured per reporting
segment. Sales is measured per reporting segment. Sales growth is the growth rate in sales of a segment
between years t-1 and t. Relative ROA is the quintile rank in ROA within a company per year for which the
numerical assignment increases with the relative performance.
a
In billions USD converted assuming an approximate F/X rate of 1,100 KRW/USD. The average exchange
rate for the sample period from 2002 to 2007 (for October of 2012) is 1,082.80 (1,104.90) KRW/USD.
70
Table 4: Correlations between Variables
Panel A: Correlations between Determinants of Personnel Decisions
1
2
3
4
5
6
7
8
9
1. Promotion
2. Dismissal -0.19
***
3. Cross-Segment Transfers 0.13
***
-0.10
***
4. Within-Segment Transfers 0.11
***
-0.09
***
-0.06
***
5. Presence of Relationship -0.05
***
-0.05
***
-0.03
*
-0.01
6. Hierarchical Level -0.13
***
0.08
***
-0.02
-0.02
0.01
7. No. of Exec’s at a Level 0.02
-0.05
***
-0.06
***
-0.05
***
0.04
**
-0.35
***
8. Log(Age) -0.06
***
0.21
***
-0.06
***
0.01
-0.10
***
0.64
-0.37
***
9. Tenure -0.01
0.03
*
0.01
0.04
***
-0.06
***
0.13
-0.06
0.09
***
10. Education -0.01
0.00
-0.04
***
-0.01
0.05
***
0.09
***
-0.01
-0.10
***
-0.01
***
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests.
Promotion is an indicator for an executive’s hierarchical advancement, an executive’s appointment to a head of an organization (from a non-head
position), or an executive’s appointment to a head of a profit center in the following year. Dismissal is an indicator for an executive’s appointment to an
advisory position or for an executive’s profile being unavailable in the following year(s). Cross-Segment Transfers refer to executives’ job
reassignments from one reporting segment to another. Within-Segment Transfers refer to executives’ job reassignments from a Level 3 organization to
another “within” a reporting segment. Presence of Relationship is an indicator for the current supervisor’s having awarded an executive’s previous
promotion. Hierarchical Level is constructed so that a higher numerical value indicates a higher level. Number of Executives at a Level is the number
of executives at a hierarchical level in a reporting segment. Log(Age) is a natural logarithm of an executive’s age calculated as the year of annual reports
minus the year of birth. Tenure is the number of years since the last Type 1 promotion (i.e., a hierarchical advancement). Education is 0 for college
graduates or below, 1 for master’s degree holders, and 2 for doctoral degree holders.
71
Table 4 (Continued)
Panel B: Correlations between Promotions and Financial Performance Measures
1
2
3
4
5
6
7
8
9
10
11
12
13
Personnel Decisions
1. Promotion
2. Type 1 Promotion 0.75
***
3. Type 2 Promotion 0.55
***
0.04
***
4. Type 3 Promotion 0.36
***
0.03
**
0.18
***
5. Dismissal -0.19
***
-0.14
***
-0.11
***
-0.07
***
Corporate Level
6. Stock Returns 0.06
***
0.07
***
-0.01
0.02
-0.04
***
7. Corporate ROA 0.04
***
0.03
**
0.05
***
-0.03
*
-0.09
***
-0.10
***
8. Corporate Profit Margin 0.05
***
0.04
***
0.06
***
-0.02
-0.10
***
-0.05
***
0.98
***
Segment Level
9. Segment ROA 0.04
***
0.05
***
0.03
*
-0.02
-0.07
***
0.05
***
0.25
***
0.27
***
10. Industry-Adjusted ROA 0.03
*
0.03
*
0.02
-0.01
-0.02
0.05
***
0.08
***
0.12
***
0.93
***
11. Relative ROA 0.06
***
0.03
**
0.08
***
-0.02
-0.06
***
-0.02
0.15
***
0.15
***
0.44
***
0.36
***
12. ROA Groups 0.07
***
0.06
***
0.05
***
-0.02
-0.11
***
0.02
0.48
***
0.49
***
0.62
***
0.49
***
0.72
***
13. Segment Profit Margin 0.09
***
0.05
***
0.10
***
-0.02
-0.06
***
0.00
0.46
***
0.47
***
0.39
***
0.27
***
0.48
***
0.62
***
14 Growth in Sales 0.00
0.00
0.00
-0.01
0.00
-0.12
***
0.19
***
0.16
***
0.07
***
0.00
0.08
***
0.13
***
0.06
***
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests.
Promotion is an indicator for an executive’s hierarchical advancement (Type 1), an executive’s appointment to a head of an organization (from a non-
head position) (Type 2), or an executive’s appointment to a head of a profit center (Type 3) in the following year. Dismissal is an indicator for an
executive’s appointment to an advisory position or for an executive’s profile being unavailable in the following year(s). Stock Returns refers to the buy-
and-hold stock return for the period corresponding to a company’s fiscal year. Corporate ROA is computed as net income divided by total assets per
company. Corporate Profit Margin is computed as net income divided by total sales per company. Segment ROA is computed as operating profit
divided by total assets measured per reporting segment. Relative ROA is the quintile rank in ROA within a company per year for which the numerical
assignment increases with the relative performance. Industry-Adjusted ROA is ROA minus the industry average ROA provided at FnGuide.com. ROA
Groups are assigned based on Segment ROA, by partitioning the ranges of ROA at 0%, 5%, 10%, 15% respectively; for example, Segment ROA below 0%
is classified to Group 1 while Segment ROA above 15% is to Group 5. Segment Profit Margin is computed as operating income divided by total sales
per reporting segment. Sales growth is the growth rate in sales of a segment between years t-1 and t.
72
Personnel Decisions Made During the Sample Period
Table 5 illustrates personnel decisions (i.e., promotion, stay and dismissal decisions)
during the period. It provides the distribution of the decisions associated with
performance in absolute and relative forms, social ties with an immediate supervisor, and
hierarchical levels. Decisions for a subsequent year are either of promotion, stay, and
demotion or dismissal. ROA is partitioned into five groups of different ranges
63
and a
relative form ROA is constructed based on quintile ranks in ROA of reporting
segments.
64
The table shows that two forms of ROA are correlated.
The table suggests that good (group) performance in both absolute and relative forms
is associated with positive personnel decisions (i.e., promotions and stays). More
promotions are observed as moving from the bottom to the top within each relative ROA
block (i.e., increasing in absolute ROA) and across the whole table (i.e., increasing in
relative ROA). This may indicate that the group performance measure positively related
to the individual members’ performance (though, it is noisy) and that the firms reward
employees for good organizational performance.
More importantly, the observations from Table 5 warrant the paper’s research
question; to suspect the presence of favoritism in promotion decisions. First, more
promotions are awarded to those who have social ties with their (immediate) supervisors
than those without such ties. Second and more importantly, the differences in the number
63
The groups are split at the cutoff points 0%, 5%, 10%, and 15%. Note that these groups of ROA are
different from quintile groups ranked in ROA.
64
Reporting segments’ performances are ordered based on ROA per year so that the highest numerical
value, or five, is assigned to the highest performance. Then, the quintile rank in ROA for a reporting
segment is shared among executives in the segment.
73
of promotions between the connected and the unconnected become more substantial in
higher levels. For example, at hierarchical level 4, there are a total of 298 promotions out
of which 158 promotions (53%) are awarded to those who have established social ties
with their supervisors (i.e., Tie=Yes). Comparatively, at level 3 (2), 70 (18) promotions
out of 103 (22) are awarded to the executives with social ties. The promotions account for
68% and 82% of the total promotions made at each hierarchical level, respectively.
Apparently, the ratios increase from 53% to 82%, moving up along the hierarchy. So,
favoritism by which certain executives keep receiving favor in promotion decisions may
well be suspected, which establishes “Preference Hypothesis.”
Table 6 presents the results from ANOVA among performance, social tie, and
personnel decisions. Panels A and B present average ROA groups and quintile ranks in
ROA, respectively, for different personnel decisions over connected and unconnected
subordinates. Similar to the observations in Table 5, performances in absolute and
relative forms are the highest for those who gain promotions, then followed by those who
stay, and the lowest for those who fail to extend their career in the conglomerate. The
differences in performances are statistically significant across all cells in both panels:
Decision with F-statistics of 32.30 and 17.58, respectively. This holds when the decisions
are dichotomized into promotion vs. non-promotion
65
and into non-dismissal vs.
dismissal.
66
65
Promotions vs. stays and dismissals. Decision with F-statistics of 25.02 and 19.30 in Panels A and B,
respectively.
66
Promotions and stays vs. dismissals. Decision with F-statistics of 50.13 and 15.63 in Panels A and B,
respectively.
74
Interestingly, the table finds that performances are higher for those who have an
established social tie than those without such a relationship. Tie, indicating whether an
executive has the same immediate supervisor who has awarded his promotion, is
statistically significant in all tests. It is notable that performances of executives with
social ties are higher even among the dismissed. This carries two important implications.
First, executives with social ties may be, overall, good performers relative to those
without professional relationships and they compete among themselves to some extent.
Second, it may indicate that executives are not favored (i.e., less subject to dismissal)
blindly, or simply because of the established social ties. Favoritism would have ROAs in
both forms lower for the connected subordinates than the unconnected. The results,
however, show that the standard for personnel decisions is more stringent for the
connected than the unconnected. So, put together, the findings suggest that supervisors’
personnel decisions are quite selective; positive decisions are awarded only to good
performers and the selection process becomes stricter over socially connected
subordinates. This is consistent with supervisors’ use of sound discretion based on
information, “Information Hypothesis.”
So far, the descriptions of the dataset offer two possibilities for supervisors’
behaviors in evaluating subordinates in the personnel decision contexts. In the following
subsections, the effects of professional relationships and moderating factors are analyzed
to unravel the cause of the ostensible manifestation of favoritism. To this end, the two
competing hypotheses are tested primarily by the sign, rather than the magnitude, of
75
coefficients estimated from the logit models predicting personnel decisions in Tables 12
and 13.
76
Table 5: Promotion Decisions by Performance, Hierarchical Rank, and Social Tie
Hierarchical Rank
a
/ Decision
b
5th 4th 3rd 2nd 1st
Relative
ROA
c
ROA
d
Tie
e
Dis Stay Pro Dis Stay Pro Dis Stay Pro Dis Stay Pro Dis Stay Pro Total
Top
5 ≥15% Yes 3 111 49 4 53 17 3 32 9 1 14 2
10 2 310
No 3 30 30 6 64 15 1 22 2 1 12
2 1 189
4 [10, 15) Yes 10 26 5 8 19 7 1 7 2 4 5 1
3
98
No 6 15 4 3 10 5 6 5 1 2 4
61
3 [5, 10) Yes
2
2
4
No
3 1
4
2nd
5 ≥15% Yes 2 43 22 4 26 9 7 7 4 1 13 1
2
141
No 7 14 15 1 23 10 1 2 3 1 7 1
4
89
4 [10, 15) Yes 11 86 37 10 69 24 6 32 17
24 1 1 9
327
No 8 12 25 2 35 13 1 13 1 2 6
1
119
3 [5, 10) Yes 6 85 21 19 138 30 16 55 15 12 33 6 3 5 1 445
No 12 30 13 18 133 28 6 26 2 3 7
1 3
282
2 [0, 5) Yes 1 14 16 2 5 7 3 5 3
2
1
59
No 1 7 7 4 13 3 1 3 1 1 1
42
1 <0% Yes 3 6 5 1 4
2 1 2
24
No
8 5 1 2 1
1
18
Med
5 ≥15% Yes 2 20 1 2 11 1
5
1 3
1
47
No 2 10 3 2 7 2
4 3
33
4 [10, 15) Yes
14 3
9 2
4 2
3
37
No
1 2 1 7 3
1
15
3 [5, 10) Yes 7 126 31 6 79 15 4 25 5 3 19 2 2 6
330
No 9 41 39 13 68 17 1 23 8 2 14
4
239
2 [0, 5) Yes 8 69 16 7 60 7 3 16 3 2 19
1 3
214
No 4 24 7 7 38 5 4 8
2 2
101
1 <0% Yes 2 19 1 1 8 1
2 1 2 2
39
No 3 5 2 2 15 2 1 1
1
32
4th
5 ≥15% Yes 1 13 1 1 4 1
2 1 1 1 1
27
No
1
2
3
4 [10, 15) Yes
14
3
2
3
22
No
1
6
1
8
3 [5, 10) Yes 5 60 16 2 36 9 3 11 1 1 8 1
5
158
No 4 19 7 3 39 2 1 9 2 1 7
94
2 [0, 5) Yes 18 105 29 21 96 17 5 30 4 7 23 2 3 7
367
No 12 40 27 16 73 18 5 25 7 5 6 1 1 2
238
1 <0% Yes 4 26 4 4 11 1 3 8
3 6
70
No 4 12 5 2 10 4 2 3
1
43
Low
2 [0, 5) Yes 3 42 9
22 4 2 4 1 3 7 1 1 3
102
No 6 20 9 1 16 6
2
1 1
62
1 <0% Yes 8 24 7 3 17 6 2 3
3 5
2 3
83
No 12 19 5 14 12 4 4 4 3 2 2
81
Total 187 1,215 479 191 1,243 298 94 406 103 66 260 22 16 73 4 4,657
a
The original seven hierarchical levels are reduced to five hierarchical-level groups in each of which at
least 10 executive are available for each year. The heads of the 1
st
level organization and the 1
st
level
executives are removed from the sample. As a result, the 2
nd
hierarchical level group is collapsed into the
3
rd
group whereas the other groups remain the same.
b
Decisions for the next period (i.e., year +1) are classified into three types: promotion (Pro), stay (Stay),
and dismissal (Dis).
c
Relative performance in ROA relative to peers within a company is measured as a quintile group (higher
number labeled for higher ROA performance) in a reporting segment for a year. Reporting segments’
performances are ordered based on ROA per year so that the highest numerical value, or five, is assigned
77
to the highest performance. Then, the quintile rank in ROA for a reporting segment is shared among
executives in the segment. The table shows that two forms of ROA are correlated. For example, the top
quintile contains only those whose ROA is greater than 5%−Group 3 and higher, whereas the lowest
quintile contains only those in the bottom two ROA groups.
d
ROA is partitioned into five groups of different ranges: executives whose (organizational) ROA is (1)
below 0%−Group 1, (2) between 0% to 5%−Group 2, (3) between 5% to 10%−Group 3, (4) between 10%
to 15%−Group 4, and (5) greater than or equal to 15%−Group 5. “[” (i.e., closed interval) indicates an
interval includes the (starting) point while “)” (i.e., open interval) indicates an interval excludes the
(ending) point.
e
Tie, short for Presence of Social Tie or Relationship, indicates that an executive has the same immediate
supervisor who has awarded (or influenced) his or her previous promotion.
Table 5 (Continued)
78
Table 6: Performance, Social Tie and Promotion Decisions
Panel A: Average ROA Group
Decisions
a
Promotion vs. Stay vs.
Dismissal
c
Promotion
vs. Non-
promotion
d
Non-
dismissal vs.
Dismissal
e
Tie
b
Dis Stay Pro Total
Source F-Stat F-Stat F-Stat
Yes 2.80 3.12 3.32 3.12 Model 14.80
***
11.44
***
19.35
***
No 2.66 3.05 3.19 3.02 Decision 32.30
***
25.02
***
50.13
***
2.74 3.09 3.26 3.09 Tie 6.53
**
6.95
***
4.06
**
Tie * Decision 0.31
0.11 0.31
Panel B: Average Quintile Ranks in ROA
Decisions
a
Promotion vs. Stay vs.
Dismissal
c
Promotion
vs. Non-
promotion
d
Non-
dismissal vs.
Dismissal
e
Tie
b
Dis Stay Pro Total
Source F-Stat F-Stat F-Stat
Yes 2.14 2.24 2.51 2.28
Model
6.94
***
10.09
***
8.79
***
No 1.97 2.25 2.30 2.22
Decision
17.58
***
19.30
***
15.63
***
2.06 2.24 2.42 2.26
Tie
3.63
*
7.45
***
7.75
***
Tie * Decision
1.65
4.24
**
3.69
**
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests.
a
Decisions for the next period (i.e., year +1) are classified into three types: promotion (Pro), stay (Stay),
and dismissal (Dis).
b
Tie, short for Presence of Social Tie or Relationship, indicates that an executive has the same immediate
supervisor who has awarded (or influenced) his or her previous promotion.
c
Decisions are trichotomous: promotion vs. stay vs. dismissal.
d
Decisions are dichotomous: promotion vs. non-promotion (including stay and dismissal).
e
Decisions are dichotomous: non-dismissal (including promotion and stay) vs. dismissal.
79
Chapter 6. Accounting Performance and Personnel Decisions: Empirical
Evidence
6.1. Research Design: Estimation of the Likelihood of Promotions and Dismissals
In this section, I describe the variables incorporated in the study’s research models.
The hypotheses are tested primarily using mixed-effects logistic regression models
67
with
random intercepts that predict the likelihood of executives’ promotions and dismissals.
The personnel decision prediction models basically have the following common form that
contains corporate and segment level ROAs (CORPROA and SEGROA), size (SALES),
growth (GROWTH), hierarchical level (LEVEL), the number of executives at a level
(NOEXEC), age (AGE), tenure group (TENURECAT), education (EDU), job area (JOB),
the speed of promotions (SPEED), the presence of a social relationship between a
supervisor and a subordinate (RELATION), and the length of the relationship (LENGTH).
Moreover, as these variables are measured at different levels, random effects for multiple
levels, 𝑢 0 𝑘𝑡
and 𝑣 0 𝑡 , are included.
𝑃𝑟 � 𝐷𝑒 𝑐𝑖 𝑠𝑖 𝑜 𝑛 𝑖𝑗𝑘 , 𝑡 + 1
� = 𝛽 1
𝐶𝑂 𝑅𝑃𝑅𝑂 𝐴 𝑘𝑡
+ 𝛽 2
𝑆𝐸 𝐺𝑅 𝑂 𝐴 𝑗𝑘𝑡
+ 𝛽 3
𝐿 𝑜𝑔 � 𝑆𝐴𝐿 𝐸 𝑆 𝑗𝑘𝑡 � + 𝛽 4
𝐺𝑅 𝑂 𝑊𝑇 𝐻 𝑗𝑘𝑡 + 𝛽 5
𝐿 𝐸 𝑉 𝐸 𝐿 𝑖𝑗𝑘𝑡 + 𝛽 6
𝑁 𝑂 𝐸 𝑋 𝐸 𝐶 𝑖𝑗𝑘𝑡
+ 𝛽 7
𝐿 𝑜𝑔 � 𝐴𝐺 𝐸 𝑖𝑗𝑘𝑡 � + 𝛽 8
𝑆𝑃𝐸 𝐸 𝐷 𝑖𝑗𝑘𝑡 + 𝛽 9
𝑅 𝐸 𝐿 𝐴 𝑇 𝐼𝑂 𝑁 𝑖𝑗𝑘𝑡
+ ∑ � 𝛽 𝑙 + 1 0
𝐿𝐸 𝑁𝐺𝑇 𝐻 𝑙 , 𝑖𝑗𝑘𝑡 + 𝛽 𝑙 + 1 3
𝑇 𝐸 𝑁 𝑈𝑅 𝐸 𝐶 𝐴 𝑇 𝑙 , 𝑖𝑗𝑘𝑡 + 𝛽 𝑙 + 1 6
𝐸 𝐷 𝑈 𝑙 , 𝑖𝑗𝑘𝑡 �
2
𝑙 = 0
+ ∑ 𝛽 𝑚 + 1 9
𝐽𝑂 𝐵 𝑙 , 𝑖𝑗𝑘𝑡 3
𝑚 = 0
+ 𝛽 0
+ 𝑢 𝑜 𝑘𝑡
+ 𝑣 0 𝑡 + 𝜀 𝑖𝑗𝑘𝑡 …………… (1)
67
Mixed-effects models are used to handle the data’s longitudinal and multi-level features.
80
Table 7: Accounting Performance, Promotions, and Dismissals
Promotions
a
Dismissals
Types 1, 2, & 3
Type 1
Type 1 Only
Type 2
Type 3
Corporate ROA
4.178
***
8.096
***
8.850
***
-0.551
-0.962
-12.778
***
(3.289)
(4.996)
(5.039)
(-0.280)
(-0.439)
(-7.188)
Segment ROA
0.976
**
1.359
***
1.065
**
0.491
-0.595
-0.651
(2.309)
(2.678)
(2.208)
(0.730)
(-0.856)
(-1.055)
Log(Sales)
0.057
0.005
0.029
0.214
*
-0.082
0.196
**
Growth in Sales
-0.117
-0.190
-0.185
0.008
-0.031
0.163
*
Hierarchical Level
-2.038
***
-5.191
***
-5.081
***
-0.270
1.582
*
-2.087
***
No. of Executives at a Level
-0.030
***
-0.045
***
-0.039
***
-0.008
-0.012
-0.009
*
Log(Age)
2.746
**
9.830
***
10.247
***
-3.146
*
-2.888
17.248
***
Tenure: [3.5]
0.071
-0.167
-0.239
**
0.296
**
0.320
*
-0.067
Tenure: [6, ∞)
-0.292
*
-0.865
***
-0.859
***
0.043
0.277
-0.051
Education: Master's
-0.057
-0.077
-0.115
-0.086
0.306
0.152
Education: Doctorate
0.188
0.494
***
0.446
**
0.016
-0.369
0.262
Job Area: Marketing/Sales
0.009
-0.273
**
-0.390
***
0.326
**
0.588
***
-0.060
Job Area: Engineer/Research
-0.142
-0.377
***
-0.418
***
0.060
0.368
*
0.058
Speed of Promotion
0.671
***
2.026
***
1.977
***
0.020
-0.767
*
0.789
***
Presence of Relationship -0.784
***
-1.520
***
-1.318
***
-0.012 -0.589
***
-0.332
**
Length of Relationship: [3.5] 1.163
***
1.963
***
1.730
***
0.210 0.480
*
0.349
**
Length of Relationship: [6, ∞) 0.945
***
1.758
***
1.702
***
-0.534 0.254 0.498
*
Intercept
-23.832
***
-69.321
***
-70.860
***
4.641
18.801
-83.695
***
S.D. (u
0
)
-14.712
-16.201
-1.679
-13.823
-21.282
-0.456
S.D. (v
0
)
-0.741
***
-0.501
**
-0.716
**
-0.481
**
-1.634
**
-2.047
**
Log-Likelihood
-2,245.79
-1,546.80
-1,420.29
-1,196.20
-678.34
-1,495.08
Prob > χ
2
0.000 0.000 0.000 0.000 0.003 0.000
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests. N=4,657 for all models. t-statistics for the
coefficients on Corporate and Segment Level ROAs are reported in parentheses. t-statistics for other variables are omitted in the interest of space.
Corporate ROA is computed as net income divided by total assets per company. Segment ROA is computed as operating profit divided by total assets
measured per reporting segment. Sales is measured per reporting segment. Sales Growth is the growth rate in sales of a segment between years t-1 and t.
Hierarchical Level is constructed so that a higher numerical value indicates a higher level. Number of Executives at a Level is the number of
executives at a hierarchical level in a reporting segment. Log(Age) is a natural logarithm of an executive’s age calculated as the year of annual reports
minus the year of birth. Tenure (i.e., years in the hierarchical rank) is categorized into three groups: tenure groups of (1) less than or equal to two years,
(2) greater than or equal to three years and less than or equal to five years, and (3) greater than or equal to six years. Dummy variables are used to
81
accommodate the non-linear nature of the variable. Education is categorized into three groups based on the final degree: executives with (1) lower than
and including college degrees, (2) master’s degrees, and (3) doctorate degrees. Dummy variables are used to accommodate the non-linear nature of the
variable. The base (i.e., 0) is for the (lower than) college degree group. Job Area is categorized into three groups: executives in (1) general
administration and management, (2) marketing and sales, and (3) engineering and research. Dummy variables are used to accommodate the non-linear
nature of the variable. The base is for the general administration and management group. Speed of Promotion captures how fast an executive has been
promoted to the current hierarchical rank, calculated as
( 6 − 𝐻 𝑖 𝑒 𝑟𝑎 𝑟𝑐 ℎ𝑖 𝑐𝑎 𝑙 𝐿 𝑒𝑣 𝑒𝑙
𝑖𝑡
)
𝐴𝑔 𝑒 𝑖𝑡
− 𝑌𝑒 𝑎𝑟 𝑠 𝑎𝑡 𝑡 ℎ 𝑒 𝐿 𝑒𝑣 𝑒 𝑙 𝑖𝑡
. Presence of Relationship is an indicator of the current supervisor’s
having awarded an executive’s previous promotion. Length of Relationship referring to the number of years since the last promotion awarded by the
current immediate supervisor is categorized into three groups: (1) less than or equal to two years, (2) greater than or equal to three years and less than or
equal to five years, and (3) greater than or equal to six years. Dummy variables are used to accommodate the non-linear nature of the variable.
a
Type 1 Promotion involves a hierarchical advancement. Type 1 Only refers to a hierarchical advancement that does not involve an appointment to a
head position. Type 2 Promotion refers to an appointment to head an organization (e.g., department, division). Type 3 Promotion refers to an
appointment to head a profit center.
Table 7 (Continued)
82
6.2. The Effects of Corporate and Segment ROAs
H1 examines whether accounting performance aggregated at a high-level
organization (i.e., a company or reporting segment) affects the careers of the executives
in sub-organizations. Table 7 reports the results of the logit regressions by type of
personnel decision.
For promotions of the most comprehensive definition, both corporate and segment
ROAs are positive and statistically significant. This suggests that, other things being
equal, supervisors seem to adjust promotability of the executives in high-performing
organizations upward. Further, the table shows that corporate ROAs are more strongly
associated with promotions than segment ROAs.
68
On the other hand, dismissals are
negatively related to corporate ROAs, while segment ROAs do not affect dismissals.
69
Overall, these findings strongly support H1; ROAs at high-level organizations,
higher than the level of an organization that an executive manages or directly reports to,
affect promotions and dismissals.
6.3. Other Determinants of Promotions and Dismissals
In most models in Table 7, size (Log(Sales)) and Growth in Sales are insignificant.
However, consistent with prior literature (e.g., Ederhof, 2011; Gibbs, 1995), the
likelihood of promotions decreases with (1) Hierarchical Level due to more limited
68
The differences in the coefficients ( 𝛽 𝐶𝑂𝑅 𝑃 𝑅𝑂𝐴
− 𝛽 𝑆𝐸𝐺 𝑅 𝑂 𝐴 =2.6543 and 3.2019 for Panels A and B) are
significant with p-values of 0.0614 and 0.0263, respectively.
69
Untabulated results from an additional analysis, however, report that when estimated without the
corporate ROA variable, segment ROAs are significant determinants of both promotions and dismissals
with greater p-values (0.005 and 0.000). The inclusion of corporate ROA undermines the magnitude and
significance of the segment ROA’s effects on these decisions.
83
positions being available at higher levels, and (2) the level of promotion competition,
measured as Number of Executives at a Hierarchical Level. Interestingly, dismissals are
less likely at higher levels. This is probably because, once executives reach a high-level
position, dismissals of such high-level executives are by far more costly for firms than
dismissals of executives at lower levels, although further promotions of the executives
may become less achievable.
As expected, age (Log(Age)), the proxy for experience and local knowledge, is
positively associated with promotions and, at the same time, with dismissals. On the
contrary, another measure of experience, Tenure, negatively affects promotions. Thus,
staying longer at a hierarchical level decreases the likelihood of promotion. Education
and Job Area have different effects on promotions depending on the context in which the
personnel decision is made. Finally, Speed of Promotions is significantly associated with
promotions and dismissals. This supports the presence of “fast track,” (e.g., Baker, Gibbs,
et al., 1994a) in which executives who have reached the current ranks faster than others
are more likely to receive another promotion. In sum, the determinants of promotions
specified in this study are generally consistent with our prior beliefs about the factors
affecting personnel decisions.
The negative and significant coefficients on Presence of Relationship are at the 1
percent level except for Type 2 promotions. This contrasts with an intuitive expectation
that the relationship would positively affect the likelihood of promotions. The negative
and significant coefficient ( β=-0.332, p=0.011) in the dismissal prediction model,
however, suggests that subordinates who have good relationships with their supervisors
84
are less likely to be dismissed. The results also show that the indicators for the ranges of
Length of Relationship carry positive and significant coefficients in most models,
including the dismissal prediction model, except for Types 2 and 3 promotions. The
findings indicate that a long relationship with the current supervisor who awarded a
previous promotion to an executive increases the likelihood of the executive’s promotion
and dismissal.
6.4. Purposes and Types of Promotions
H2 tests how supervisors evaluate the usefulness of corporate and segment ROAs
depending on the purpose or type of promotion. As discussed earlier, Types 2 and 3
promotions are associated with substantial changes in tasks that require different sets of
skills, whereas Type 1 and Type 1 Only promotions involve minimum changes in tasks.
Therefore, the analysis examines how the coefficients on corporate and segment ROAs
vary with promotion types that involve different degrees of change in job characteristics.
The results in Table 7 show that the coefficients on ROA measures are positive and
significant in Type 1 and Type 1 Only columns, while they are not significant for the
other types of promotions. Given that hierarchical ranks are a predominant determinant of
monetary compensation in this conglomerate (Pucik & Lim, 2001), this finding suggests
that, consistent with H2, supervisors relate promotions to good organizational
performance only when they award promotions as incentives.
For an extended analysis, I partition the whole executive-years sample into two
groups based on managerial responsibility: non-manager executives and executives with
supervisory responsibility. I further sort profit center managers among the manager
85
executives. Then, possible promotion types are identified for each group. For example,
for cost center manager executives, hierarchical advancements (Type 1) and appointments
to profit center managers (Type 3) are the possible types. Finally, these are regressed by
each group of executives.
Table 8 shows that Type 1 and Type 1 Only promotions with non-trivial
compensation increase and a minimum level of job change are positively and
significantly associated with the ROA measures, regardless of responsibility or job type
in the current position. In contrast, promotions involving significant task changes—(1)
appointments to manager positions in any organizations for non-manager executives
(Type 2, in the fourth column) and (2) appointments to profit center manager positions for
cost center managers (Type 3, in the eighth column)—are not attributable to the ROA
measures. Again, the findings support H2 that accounting performance affects only
hierarchical advancements (Type 1 promotions).
In sum, hierarchical advancements with non-trivial compensation increases seem to
function as implicit incentive provision mechanisms. This suggests that supervisors
consider accounting performance of an organization when they make promotion
decisions for incentive provision purposes, while they find the accounting measures less
useful in matching jobs.
86
Table 8: Promotion Types a and Accounting Performance
Non-Head Positions
Head Positions
All
Cost Center Manager
PCtr Mgr
Type
1 & 2
a
Type 1
Only
a
Type 2
a
Type 1
a
Type
1 & 3
a
Type 1
Only
a
Type 3
a
Type 1
a
Corporate ROA
5.989
***
19.634
***
1.339
3.358
**
2.866
5.731
**
-1.873
7.908
**
(2.593)
(3.862)
(0.585)
(2.152)
(1.534)
(2.559)
(-0.640)
(2.303)
Segment ROA
1.169
0.292
0.948
0.874
**
0.894
*
1.242
**
-0.546
1.455
(1.220)
(0.143)
(1.086)
(2.035)
(1.810)
(2.230)
(-0.560)
(1.435)
Log(Sales)
-0.302
**
-0.001
-0.333
***
0.124
0.111
0.206
*
-0.069
0.074
Growth in Sales
-0.188
-1.568
**
0.277
-0.090
-0.341
-0.620
*
-0.092
0.114
Hierarchical Level
-0.359
-4.293
**
-0.121
-2.874
***
-1.874
**
-5.532
***
1.809
*
-6.091
***
No. of Executives at a Level
-0.016
*
-0.035
**
-0.006
-0.035
***
-0.039
***
-0.052
***
-0.012
0.001
Log(Age)
0.193
6.491
-0.088
6.363
***
5.054
***
10.940
***
-2.630
12.683
***
Tenure: [3.5]
0.055
-0.454
*
0.264
*
-0.008
-0.055
-0.291
*
0.415
*
0.134
Tenure: [6, ∞)
-0.317
-0.839
-0.057
-0.431
**
-0.469
**
-1.112
***
0.358
-0.473
Education: Master's
-0.209
-0.183
-0.196
-0.003
-0.044
-0.344
0.484
*
0.301
Education: Doctorate
-0.275
-0.511
-0.268
0.447
**
0.495
**
0.816
***
-0.507
-0.291
Job Area: Marketing/Sales
0.102
-0.639
*
0.429
**
-0.056
-0.024
-0.408
**
0.733
***
-0.247
Job Area: Engineer/Research
0.201
-1.080
***
0.592
***
-0.168
-0.136
-0.442
**
0.473
*
-0.252
Speed of Promotion
0.11
1.44
0.18
1.00
***
0.52
2.10
***
-0.82
2.60
***
Presence of Relationship -0.499
***
-1.322
***
-0.131 -1.217
***
-1.232
***
-1.436
***
-0.625
**
-1.128
***
Length of Relationship: [3.5] 1.019
***
1.784
***
0.416
**
1.568
***
1.602
***
1.825
***
0.684
**
1.553
***
Length of Relationship: [6, ∞) 0.558 2.252
***
-0.688 1.445
***
1.607
***
1.780
***
0.720 1.310
**
Intercept
1.48
-51.94
*
3.33
-43.73
***
-32.20
***
-78.20
***
18.82
-88.12
***
S.D. (u
0
)
-23.41
-13.66
-16.54
-10.66
-1.08
**
-15.40
-13.41
-15.35
S.D. (v
0
)
-0.72
*
0.08
-1.08
**
-0.90
***
-1.09
-0.51
**
-0.47
-0.86
**
Log-Likelihood
-851.05
-322.08
-740.40
-1,285.91
-970.06
-747.11
-416.62
-298.66
Observations
1,431
1,431
1,431
3,226
2,328
2,328
2,328
898
Prob > χ
2
0.000
0.000
0.000
0.000
0.000
0.000
0.011
0.000
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests. t-statistics for the coefficients on Corporate and
Segment Level ROAs are reported in parentheses. t-statistics for other variables are omitted in the interest of space.
87
Corporate ROA is computed as net income divided by total assets per company. Segment ROA is computed as operating profit divided by total assets
measured per reporting segment. Sales is measured per reporting segment. Sales growth is the growth rate in sales of a segment between years t-1 and t.
Hierarchical Level is constructed so that a higher numerical value indicates a higher level. Number of Executives at a Level is the number of
executives at a hierarchical level in a reporting segment. Log(Age) is a natural logarithm of an executive’s age calculated as the year of annual reports
minus the year of birth. Tenure (i.e., years in the hierarchical rank) is categorized into three groups: tenure groups of (1) less than or equal to two years,
(2) greater than or equal to three years and less than or equal to five years, and (3) greater than or equal to six years. Dummy variables are used to
accommodate the non-linear nature of the variable. Education is categorized into three groups based on the final degree: executives with (1) lower than
and including college degrees, (2) master’s degrees, and (3) doctorate degrees. Dummy variables are used to accommodate the non-linear nature of the
variable. The base (i.e., 0) is for the (lower than) college degree group. Job Area is categorized into three groups: executives in (1) general
administration and management, (2) marketing and sales, and (3) engineering and research. Dummy variables are used to accommodate the non-linear
nature of the variable. The base is for the general administration and management group. Speed of Promotion captures how fast an executive has been
promoted to the current hierarchical rank, calculated as
( 6 − 𝐻 𝑖 𝑒 𝑟𝑎 𝑟𝑐 ℎ𝑖 𝑐𝑎 𝑙 𝐿 𝑒𝑣 𝑒𝑙
𝑖𝑡
)
𝐴𝑔 𝑒 𝑖𝑡
− 𝑌𝑒 𝑎𝑟 𝑠 𝑎𝑡 𝑡 ℎ 𝑒 𝐿 𝑒𝑣 𝑒 𝑙 𝑖𝑡
. Presence of Relationship is an indicator of the current supervisor’s
having awarded an executive’s previous promotion. Length of Relationship referring to the number of years since the last promotion awarded by the
current immediate supervisor is categorized into three groups: (1) less than or equal to two years, (2) greater than or equal to three years and less than or
equal to five years, and (3) greater than or equal to six years. Dummy variables are used to accommodate the non-linear nature of the variable.
a
Type 1 Promotion involves a hierarchical advancement. Type 1 Only refers to a hierarchical advancement that does not involve an appointment to a
head position. Type 2 Promotion refers to an appointment to head an organization (e.g., department, division). Type 3 Promotion refers to an
appointment to head a profit center.
Table 8 (Continued)
88
Table 9: Job Responsibility and Accounting Performance
Promotions
Dismissals
Promotions
Dismissals
High
Level
a
Low
Level
b
High
Level
a
Low
Level
b
Lev 2/3
Managers
c
Others
Lev 2/3
Managers
c
Others
Corporate ROA
1.566
5.409
***
-11.757
***
-10.614
***
3.806
4.417
***
-10.213
***
-12.618
***
(0.590)
(3.794)
(-2.977)
(-5.187)
(1.270)
(3.122)
(-3.100)
(-5.797)
Segment ROA
0.451
1.104
**
-4.138
**
-0.270
-0.072
0.973
**
-0.747
-0.851
(0.549)
(2.396)
(-2.096)
(-0.444)
(-0.061)
(2.192)
(-0.577)
(-1.127)
Log(Sales)
0.068
0.062
0.297
*
0.107
-0.046
0.029
0.070
0.364
**
Growth in Sales
0.086
-0.349
*
0.355
**
0.115
0.259
-0.254
*
0.218
0.125
Hierarchical Level
-4.176
***
-2.789
***
-3.118
***
-6.029
***
-3.659
***
-1.450
**
-2.026
**
-2.249
***
No. of Executives at a Level
-0.021
-0.029
***
0.016
-0.004
0.010
-0.031
***
-0.004
-0.014
**
Log(Age)
8.095
**
3.398
**
26.889
***
20.674
***
10.931
***
1.386
18.265
***
17.896
***
Tenure: [3.5]
0.277
0.067
-0.145
-0.101
0.057
0.072
-0.044
-0.105
Tenure: [6, ∞)
-0.006
-0.308
*
-0.069
-0.222
-0.480
-0.273
0.150
-0.149
Education: Master's
-0.180
-0.037
0.106
0.163
0.023
-0.076
0.171
0.181
Education: Doctorate
-0.115
0.367
**
0.578
*
0.237
0.128
0.175
-0.094
0.449
**
Job Area: Marketing/Sales
0.092
-0.010
-0.025
-0.010
-0.004
0.025
-0.332
0.076
Job Area: Engineer/Research
-0.020
-0.183
*
-0.442
*
0.162
0.188
-0.141
-0.235
0.141
Speed of Promotion
1.61
***
0.92
**
1.28
***
2.59
***
1.43
**
0.43
0.82
*
0.85
**
Presence of Relationship -0.149
-0.886
***
-0.134
-0.279
*
-0.712
***
-0.789
***
-0.591
**
-0.283
*
Length of Relationship: [3.5] 0.564
**
1.230
***
0.362
0.125
1.045
***
1.220
***
0.759
**
0.178
Length of Relationship: [6, ∞) -0.626
1.165
***
0.318
-0.040
0.891
*
0.999
***
0.518
0.500
Intercept
-56.63
***
-30.75
***
-129.79
***
-119.18
***
-64.48
***
-14.61
*
-85.85
***
-89.71
***
S.D. (u
0
)
-21.49
-10.90
-0.61
-0.56
-1.22
*
-17.08
-0.60
-0.65
S.D. (v
0
)
-22.53
-0.66
***
-20.79
-1.19
***
-1.43
-0.78
***
-1.40
-1.15
**
Log-Likelihood
-412.11
-1,813.44
-389.61
-1,084.90
-384.97
-1,845.77
-359.68
-1,128.15
Observations
1,044
3,613
1,044
3,613
1,030
3,627
1,030
3,627
Prob > χ
2
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests. t-statistics for the coefficients on Corporate and
Segment Level ROAs are reported in parentheses. t-statistics for other variables are omitted in the interest of space.
89
Corporate ROA is computed as net income divided by total assets per company. Segment ROA is computed as operating profit divided by total assets
measured per reporting segment. Sales is measured per reporting segment. Sales growth is the growth rate in sales of a segment between years t-1 and t.
Hierarchical Level is constructed so that a higher numerical value indicates a higher level. Number of Executives at a Level is the number of
executives at a hierarchical level in a reporting segment. Log(Age) is a natural logarithm of an executive’s age calculated as the year of annual reports
minus the year of birth. Tenure (i.e., years in the hierarchical rank) is categorized into three groups: tenure groups of (1) less than or equal to two years,
(2) greater than or equal to three years and less than or equal to five years, and (3) greater than or equal to six years. Dummy variables are used to
accommodate the non-linear nature of the variable. Education is categorized into three groups based on the final degree: executives with (1) lower than
and including college degrees, (2) master’s degrees, and (3) doctorate degrees. Dummy variables are used to accommodate the non-linear nature of the
variable. The base (i.e., 0) is for the (lower than) college degree group. Job Area is categorized into three groups: executives in (1) general
administration and management, (2) marketing and sales, and (3) engineering and research. Dummy variables are used to accommodate the non-linear
nature of the variable. The base is for the general administration and management group. Speed of Promotion captures how fast an executive has been
promoted to the current hierarchical rank, calculated as
( 6 − 𝐻 𝑖 𝑒 𝑟𝑎 𝑟𝑐 ℎ𝑖 𝑐𝑎 𝑙 𝐿 𝑒𝑣 𝑒𝑙
𝑖𝑡
)
𝐴𝑔 𝑒 𝑖𝑡
− 𝑌𝑒 𝑎𝑟 𝑠 𝑎𝑡 𝑡 ℎ 𝑒 𝐿 𝑒𝑣 𝑒 𝑙 𝑖𝑡
. Presence of Relationship is an indicator of the current supervisor’s
having awarded an executive’s previous promotion. Length of Relationship referring to the number of years since the last promotion awarded by the
current immediate supervisor is categorized into three groups: (1) less than or equal to two years, (2) greater than or equal to three years and less than or
equal to five years, and (3) greater than or equal to six years. Dummy variables are used to accommodate the non-linear nature of the variable.
a
High Level refers to the likelihood of promotions or dismissals of executives at hierarchical levels of one through three.
b
Low Level refers to the likelihood of promotions or dismissals of executives at hierarchical levels of four and five.
c
Lev 2/3 Manager is an indicator for executives heading Level 2 or 3 organizations.
Table 9 (Continued)
90
6.5. Job Scope and Responsibility
To test H3, I run personnel decision prediction models with groups of executives
partitioned based on hierarchical levels and management positions. If supervisors
acknowledge individual executives’ different degrees of responsibility and contribution
to organizational accounting performance, high-level executives and managers of Level 2
or 3 organizations (low-level executives and non-Level 2/3 managers) are likely to be
more strongly (weakly) associated with accounting performance. In other words, a
stronger association between accounting performance and personnel decisions for
executives with greater responsibility may indicate that, when evaluating these executives,
supervisors place greater importance on ROAs of high-level organizations.
Table 9 provides the evidence contrary to this prediction. In the promotion models
for the two groups (columns 2, 3, 6, and 7), ROA measures are not significant
determinants of promotions for executives in High Level or Lev 2/3 Manager groups,
while they are positively and significantly related to promotions of executives classified
otherwise. The findings indicate that supervisors place considerable weight on accounting
performance at high-level organizations when they evaluate executives with less
responsibility or narrower job scope.
Several causes, solely or in combination, may explain this finding. First, firms may
motivate executives with less responsibility to collaborate with organizations higher than
the organization that they manage or directly report to, and reward them based on the
performance measured at higher-level organizations. Second, supervisors may make
91
limited efforts to evaluate executives at low levels or with less responsibility. In other
words, they may resort to performance measures that are either relatively easy-to-
measure or readily available without exploring or considering other sources of
information. The findings for executives with greater responsibility, on the contrary,
suggest that, in evaluating executives with greater job responsibility, supervisors may
include more evaluation criteria and weigh these alternative, perhaps more informative
evaluation criteria more heavily than accounting performance. Third, promotions of
executives with greater (less) responsibility may be more (less) constrained by
organizational structure (Baker et al., 1988; Gibbs, 1995; Milgrom & Roberts, 1992). For
example, given that dismissals of high-ranking executives for good performance are
unlikely, organizational structure changes to create positions at higher (lower) level
organizations is more (less) costly and infrequent. Thus, the extent to which the structure
or capacity constraint limits supervisors’ promotion awards increases with an executive’s
job responsibility. On the other hand, promotions often follow the dismissals of
superordinates when organizational performance is poor, which also offsets the positive
effects of performance on promotions. Although the current dataset does not allow me to
distinguish these potential explanations, they commonly point out that supervisors
consider the costs and constraints as well as the benefits of promotions to evaluate the
usefulness of accounting performance
In contrast, dismissals are negatively associated with ROA measures for all groups of
executives. Therefore, executives with greater responsibility actually bear greater
92
responsibility for poor performance. On the other hand, strong relationships between poor
performance and dismissal for executives with limited decision-making authority may
suggest that these executives take “undue” responsibility for the organization’s
accounting performance compared to their relatively small contribution. Alternatively,
the finding may indicate that supervisors use poor organizational performance as a good
pretext for dismissals.
In summary, the findings in Table 9 are in stark contrast to H3. The results suggest
that when supervisors make promotion decisions for executives with greater
responsibility, they seem to incorporate more alternative sources of information about
executives’ qualities for promotions. The adoption of other evaluation criteria reduces the
relative weights of the criteria correlated with ROA measures. Thus, overall correlations
between promotions and ROA diminish, which is shown in the results.
6.6. Intra-organization Interdependency
H4 predicts that intra-organization interdependency increases the sensitivity of the
likelihood of promotions to accounting performance measured at higher-level
organizations rather than the organization to which an executive directly reports. To test
this hypothesis, I operationalize intra-organization interdependency as the frequency of
executives’ cross-unit transfers in different specifications: relative frequency (Ratio) and
its partition (Group). These variables and their interaction terms with ROA measures are
incorporated into the established promotion prediction model, or Equation (1).
93
H4 is supported if the interaction terms between the frequency variable and the ROA
measures are positive and significant. Table 10 presents the results. Each column
represents different specifications of the frequency of cross-unit transfers—i.e., the
relative frequency of the event, and the partitions of the relative frequency. In Ratio
model, the coefficient of the interaction term between Cross-Segment Transfer and
Corporate ROA is significant and positive ( 𝛽 =34.256 and p = 0.007). This suggests that
if cross-segment transfers are more frequent (i.e., greater interdependencies between
segments), promotions are more strongly associated with corporate ROA. With a
categorical variable specification, the Group model provides stronger results. Both
interaction terms are positive and significant ( 𝛽 𝐼 𝑁𝑇 _ 𝐶𝑂 𝑅𝑃 𝑅𝑂 𝐴 =3.28, p = 0.009;
𝛽 𝐼𝑁 𝑇 _ 𝑆 𝐸 𝐺𝑅 𝑂 𝐴 =0.644, p = 0.089).
70
Overall, the findings in Table 10 generally support H4. In particular, with greater
interdependency between sub-organizations, or greater transferability of workers’
knowledge and skills, supervisors can improve the strength of promotion-based
incentives by associating more promotions with the accounting performance measured at
high-level organizations.
70
In addition to the analysis, I examine the combined effects of ROA measures including the main and the
interaction effect. As presented at the bottom of the table, the sums of the coefficients on ROA variables
and corresponding interaction terms ( 𝛽 𝐶 𝑂 𝑅𝑃 𝑅𝑂 𝐴 + 𝛽 𝐼 𝑛𝑡 _ 𝐶 𝑂 𝑃 𝑅𝑅𝑂 𝐴 and 𝛽 𝑆𝐸𝐺 𝑅 𝑂 𝐴 + 𝛽 𝐼 𝑛𝑡 _ 𝑆𝐸𝐺 𝑅 𝑂 𝐴 ) remain positive
and (in most cases) significant with additional cross-unit transfer variables incorporated.
94
Table 10: Intra-organization Interdependency, Accounting Performance, and
Promotions
Dependent: Pr(Promotions)
Ratio
a
Group
b
Corporate ROA
-0.137
-2.074
Cross-Segment Transfer 1.534
0.050
Interaction with Corporate ROA 34.256
***
3.28
***
Segment ROA 1.329
***
0.468
Within-Segment Transfer
3.487
***
0.30
***
Interaction with Segment ROA
1.678
0.644
*
Log(Sales)
0.274
***
0.184
**
Growth in Sales
-0.02
0.02
Hierarchical Level
-2.237
***
-2.266
***
No. of Executives at a Level
-0.031
***
-0.030
***
Log(Age)
3.574
***
3.734
***
Tenure: [3.5] 0.055
0.051
Tenure: [6, ∞)
-0.359
**
-0.367
**
Education: Master's -0.022 -0.026
Education: Doctorate
0.228
0.215
Job Area: Marketing/Sales -0.009 0.001
Job Area: Engineer/Research
-0.141
-0.126
Speed of Promotion
0.750
***
0.771
***
Presence of Relationship -0.82
***
-0.78
***
Length of Relationship: [3.5] 1.17
***
1.17
***
Length of Relationship: [6, ∞) 0.817
***
0.804
***
Intercept
-31.96
***
-31.17
***
S.D. (u
0
)
-19.70
-19.06
S.D. (v
0
) -0.61
***
-0.76
***
Log-Likelihood
-2,196.17
-2,210.81
Observations
4,657
4,657
Prob > χ
2
0.000
0.000
𝜷 𝑪 𝑶 𝑹𝑷𝑹 𝑶 𝑨 + 𝜷 𝑰 𝒏 𝒕 _ 𝑪 𝑹𝑶 𝑨 = 𝟎 34.119
***
1.207
χ
2
(8.32) (0.73)
𝜷 𝑺𝑬𝑮𝑹 𝑶𝑨
+ 𝜷 𝑰 𝒏 𝒕 _ 𝑺𝑹 𝑶𝑨
= 𝟎 3.008
*
1.112
**
χ
2
(3.38) (6.11)
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests.
N=4,657 for all models. t-statistics for all variables are omitted in the interest of space.
Corporate ROA is computed as net income divided by total assets per company. Cross-Segment Transfers
refer to executives’ job reassignments from one reporting segment to another. Segment ROA is computed
as operating profit divided by total assets measured per reporting segment. Within-Segment Transfers refer
to executives’ job reassignments from a Level 3 organization to another “within” a reporting segment.
Sales is measured per reporting segment. Sales Growth is the growth rate in sales of a segment between
years t-1 and t. Hierarchical Level is constructed so that a higher numerical value indicates a higher level.
Number of Executives at a Level is the number of executives at a hierarchical level in a reporting segment.
95
Log(Age) is a natural logarithm of an executive’s age calculated as the year of annual reports minus the
year of birth. Tenure (i.e., years in the hierarchical rank) is categorized into three groups: tenure groups of
(1) less than or equal to two years, (2) greater than or equal to three years and less than or equal to five
years, and (3) greater than or equal to six years. Dummy variables are used to accommodate the non-linear
nature of the variable. Education is categorized into three groups based on the final degree: executives with
(1) lower than and including college degrees, (2) master’s degrees, and (3) doctorate degrees. Dummy
variables are used to accommodate the non-linear nature of the variable. The base (i.e., 0) is for the (lower
than) college degree group. Job Area is categorized into three groups: executives in (1) general
administration and management, (2) marketing and sales, and (3) engineering and research. Dummy
variables are used to accommodate the non-linear nature of the variable. The base is for the general
administration and management group. Speed of Promotion captures how fast an executive has been
promoted to the current hierarchical rank, calculated as
( 6 − 𝐻 𝑖 𝑒 𝑟𝑎 𝑟𝑐 ℎ𝑖 𝑐𝑎 𝑙 𝐿 𝑒𝑣 𝑒𝑙
𝑖𝑡
)
𝐴𝑔 𝑒 𝑖𝑡
− 𝑌𝑒 𝑎𝑟 𝑠 𝑎𝑡 𝑡 ℎ 𝑒 𝐿 𝑒𝑣 𝑒 𝑙 𝑖𝑡
. Presence of Relationship is
an indicator of the current supervisor’s having awarded an executive’s previous promotion. Length of
Relationship referring to the number of years since the last promotion awarded by the current immediate
supervisor is categorized into three groups: (1) less than or equal to two years, (2) greater than or equal to
three years and less than or equal to five years, and (3) greater than or equal to six years. Dummy variables
are used to accommodate the non-linear nature of the variable.
a
In Ratio model, Cross-Unit Transfer is specified as the ratio of the frequency of a type of transfers to the
number of executives in a reporting segment. For example, Cross-Segment Transfer is computed as the
frequency of outgoing, rather than incoming, cross-segment transfers from a segment divided by the
number of executives in the segment. Within-Segment Transfer is computed as the frequency of cross-
unit transfers within a segment divided by the number of executives in the segment.
b
In Group model, the ratios specified in the Ratio model are divided into three groups: (1) zero, (2)
between zero and the pooled-median of the ratios, and (3) greater than the pooled-median of the ratios.
Thus, Cross-Segment Transfer is specified as a categorical variable of three groups.
Table 10 (Continued)
96
6.7. Additional Analysis: Cross-Unit Mobility and Promotions
The additional analysis explores the relationships among accounting performance,
promotions, and job transfers across business units. In particular, I examine how cross-
unit transfers and personnel decisions are associated. Regarding intra- and inter-firm job
transfers, the literature on human capital and knowledge management implicitly
associates job mobility with knowledge transfer or sharing and diffusion of best practices
(e.g., Boschma, Eriksson, & Lindgren, 2009; Minbaeva et al., 2003).
71
However, there is
little research dedicated to intra-firm cross-organizational job mobility—both lateral
transfers and promotions.
72
For this reason, the following analysis is driven by an
empirical motivation.
Research Model
To investigate the association, the multi-level logit regression model is specified as
follows:
𝑃𝑟 � 𝑋𝑆𝐸 𝐺𝑇𝑅
𝑖𝑗𝑘 , 𝑡 + 1
� = 𝛽 1
𝐻 𝐼 𝐺𝐻 𝑃𝐸 𝑅 𝐹 𝑗𝑘𝑡 + 𝛽 2
𝐿 𝑜𝑔 � 𝑆𝐴𝐿 𝐸 𝑆 𝑗𝑘𝑡 � + 𝛽 3
𝐺𝑅 𝑂 𝑊𝑇 𝐻 𝑗𝑘𝑡
+ 𝛴 � 𝛽 𝑙 + 3
𝐷 𝐸 𝐶 𝐼𝑆𝐼𝑂 𝑁 𝑙 , 𝑖𝑗𝑘𝑡 + 1
+ 𝛽 𝑙 + 6
𝐻 𝐼 𝐺𝐻 𝑃𝐸 𝑅 𝐹 𝑗𝑘𝑡 ∗ 𝐷 𝐸 𝐶 𝐼𝑆𝐼𝑂 𝑁 𝑙 , 𝑖𝑗𝑘𝑡 + 1
�
+ 𝛴 � 𝛽 𝑚 + 1 0
𝐽𝑂𝐵𝑅 𝐸 𝑆 𝑃 𝑚 , 𝑖𝑗𝑘𝑡 + 𝛽 2 𝑚 + 1 0
𝐻 𝐼 𝐺𝐻 𝑃𝐸 𝑅 𝐹 𝑗𝑘𝑡 ∗ 𝐽𝑂𝐵𝑅 𝐸 𝑆 𝑃 𝑚 , 𝑖𝑗𝑘𝑡 �
+ 𝛴 𝛽𝐶 𝑜 𝑛 𝑡 𝑟𝑜 𝑙𝑠 + 𝛽 0
+ 𝑢 0 𝑘𝑡
+ 𝜀 𝑖𝑗𝑡 𝑘 …………… (2)
71
The research contexts of the literature have mostly been intra-firm knowledge transfer in multi-national
corporations (MNCs) (e.g., Minbaeva et al., 2003) or inter-firm knowledge transfer through, for example,
external labor market transactions (e.g.,Song, Almeida, & Wu, 2003) or strategic alliances (e.g., Mowery,
Oxley, & Silverman, 1996). Specifically, the additional analysis investigates whether firms use promotions
and cross-unit transfers to transfer knowledge and diffuse good practices from an outperforming
organization to another as predicted in knowledge transfer arguments.
72
There are two, but not exhaustive, reasons for the lack of research about this topic. First, lateral transfers
are rare (Baker, Gibbs, et al., 1994a). Second, as the features of within- and cross-unit transfers may not be
different from each other, there has been no need for separate research.
97
where XSEGTR is cross-segment transfers at t+1, HIGHPERF as a proxy for good
organizational performance is an indicator of an executive’s segment ROA being among
the top two quintile groups, DECISION
l
—l for the type (e.g., 1, 2, and 3 for each
promotion type)—indicates the types of personnel decisions for t+1, JOBRESP
m
is an
indicator variable for job responsibility and scope—m for the number of classifications.
In addition to these key variables, I also control for the size of a reporting segment, as a
proxy for the capacity to feed promotions (Log(SALES)), and its growth (GROWTH).
In addition, I run separate regressions
73
without interaction terms for each
performance group (i.e., high and low) sorted based on the quintile rank based on
segment ROA per year. This addresses Ai and Norton’s (2003) concern about the
interpretation of interaction terms in non-linear regression models and also facilitates
more intuitive interpretations.
The Findings
Cross-Unit Transfers and Hierarchical Advancements. The first question is
whether cross-unit transfers are associated with hierarchical advancements (i.e.,
Promotion Type 1). If promotions function as rewards for good performance, they are
more likely to be made within the good performing unit than to another (probably poorly
performing) unit. To address this question, the analysis examines the interaction terms
with the indicator for High Relative Performance (HRP) group. These interaction terms
represent the difference of the coefficients for the main effects, such as Promotion: Type
73
𝑃𝑟 � 𝑋 𝑆𝐸𝐺 𝑇𝑅
𝑖𝑗 𝑘 , 𝑡 + 1
� = 𝛽 1
𝐿 𝑜𝑔 � 𝑆𝐴 𝐿 𝐸 𝑆 𝑗𝑘 𝑡 � + 𝛽 2
𝐺 𝑅 𝑂𝑊 𝑇 𝐻 𝑗𝑘 𝑡 + 𝛴 𝛽 𝑙 + 2
𝐷𝐸𝐶 𝐼 𝑆𝐼 𝑂 𝑁 𝑙 ,𝑖𝑗 𝑘𝑡 + 1
+𝛴 𝛽 𝑚 + 𝑙 + 2
𝐽 𝑂𝐵 𝑅𝐸𝑆 𝑃 𝑚 ,𝑖𝑗 𝑘𝑡 + 𝛴 𝛽 𝐶 𝑜𝑛 𝑡 𝑟 𝑜 𝑙𝑠 + 𝛽 0
+ 𝑢 0 𝑘 𝑡 + 𝜀 𝑖𝑗 𝑡 𝑘 ……… (3)
98
X
74
and Release: Type X, between HRP group and the other. In Table 11, the coefficients
for the interaction term between Promotion: Type 1 and HRP in Models 1 and 2 are
negative and significant ( 𝛽 𝐼 𝑁𝑇 _ 𝑇 1
𝑃 =-0.781 and -0.790, p=0.053 and 0.051, respectively)
while the coefficients for the main effect of Promotion: Type 1 are insignificant in both
models. The findings are consistent with the difference in the coefficients in separate
regressions for each performance group. The coefficients for Promotion: Type 1 are
negative and significant in both models for the HRP group ( 𝛽 𝑇 1
𝑃 =-0.675 and -0.627,
p=0.059 and 0.092), while they are positive and insignificant for the Low Relative
Performance (LRP) group ( 𝛽 𝑇 1
𝑃 =0.281 and 0.419, p=0.321 and 0.147). The results
indicate that (1) in relatively low-performing segments, executives’ hierarchical
advancements have nothing to do with cross-segment transfers; and (2) in relatively high-
performing segments, executives’ hierarchical advancements do not coincide with cross-
segment transfers. This suggests that hierarchical advancements occur mainly within a
segment.
Cross-Unit Transfers of Executives with Greater Responsibility. The second
question is related to the knowledge transfer perspective in human capital and knowledge
management literature. If the cross-unit transfers—especially reallocating executives
from one high-performing business unit to another—are used as a means to spread best
practices, they are more likely to occur for executives with greater job responsibility
and/or decision authority than those with less or none. This is because the efficiency of
74
X is a variable for different types of personnel decisions.
99
knowledge diffusion may be maximized when incoming executives have greater job
responsibility or decision-making authority.
To address the second question, the coefficients of the variables relevant to job
responsibility/scope, i.e., High-Level Executives, Lev 2/3 Manager, Heads of
Organizations, and Profit Center Manager, are examined. In Panel A of Table 11, I find
that, with the exception of High-Level Executives, all other indicators, proxies for greater
responsibility and/or broader job scope, are positively associated with cross-unit job
transfers of executives in the HRP group, while none of these indicators are significant
for the LRP group. According to these findings, executives who have acquired a high-
rank position in a high-performing segment are less likely to be relocated to another
(potentially poorly performing) segment. On the other hand, the talent, skills, and
knowledge of executives with supervisory roles in high-performing segments are likely to
be exported to other segments while counterparts in poorly performing segments are not.
Cross-Unit Transfers and Assignment/Release of a Supervisory Role. The third
question is an extension of the second question. Similar to the previous rationale, the
cross-unit transfers aiming at knowledge transfer are more likely to accompany increased
job responsibility and/or decision authority than the loss of a supervisory position. To
investigate whether this is the case, I compare the coefficients for Promotion: Type X ( 𝛽 𝑋 𝑃 )
with their counterparts for Release: Type X ( 𝛽 𝑋 𝑅 ), testing 𝛽 𝑋 𝑃 = 𝛽 𝑋 𝑅 . Panel B of Table 11
reports the results. In Cross-Segment Transfer Model 1, the difference between the
coefficients for Promotion: Type 2 or 3 ( 𝛽 𝑇 2 3
𝑃 ) and Release: Type 2 or 3 ( 𝛽 𝑇 2 3
𝑅 ) for the
100
HRP group is significant (p=0.002). This suggests that executives departing from high-
performing segments are more likely to be appointed to manager positions or to profit
center managers (i.e., Type 2 or 3 promotions) than to be released from such a
supervisory task. On the contrary, for the LRP group, the sign on the difference is
reversed ( 𝛽 𝑇 2 3
𝑃 − 𝛽 𝑇 2 3
𝑅 =-0.423). The negative sign suggests that executives with relatively
poor performance are more likely to lose their supervisory positions when they are
reassigned to another segment. However, the difference is not significant (p=0.277).
Separating Type 2 and Type 3 events, Cross-Segment Transfer Model 2 also
provides consistent evidence. Comparing the associations between Type 2 career events
and cross-segment transfers, the difference ( 𝛽 𝑇 2
𝑃 − 𝛽 𝑇 2
𝑅 =1.592) is significant (p=0.0118)
for the HRP group while that of the LRP group ( 𝛽 𝑇 2
𝑃 − 𝛽 𝑇 2
𝑅 =0.859) is not (p=0.1645).
Again, for executives departing from a high-performing segment, Type 2 promotions are
more likely than release from a supervisory job, while this observation does not hold true
for executives in an LRP group. In contrast, Release: Type 3 is more likely than
Promotion: Type 3 for executives transferring from a low-performing segment ( 𝛽 𝑇 3
𝑃 −
𝛽 𝑇 3
𝑅 =-2.581, p=0.000).
101
Table 11: Cross-Unit Job Mobility and Promotions
Panel A: Regressions of Cross-Unit Transfers
Dependent:
Cross-Segment Transfer
a
Cross-Segment Transfer
Model 1
b
Cross-Segment Transfer
Model 2
c
Model 1
b
Model 2
c
High Relative
Performance
d
Low Relative
Performance
d
High Relative
Performance
d
Low Relative
Performance
d
High Relative Performance
-0.385
-0.481
Promotion: Type 1
0.110
0.201
-0.675
*
0.281
-0.627
*
0.419
Interaction w/ High Performance
-0.781
*
-0.790
*
Promotion: Type 2
1.626
***
2.140
***
1.601
***
Interaction w/ High Performance
0.306
Promotion: Type 3
1.318
***
2.664
***
1.032
**
Interaction w/ High Performance
1.004
**
Promotion: Type 2 or 3
2.103
***
2.676
***
1.968
***
Interaction w/ High Performance
0.382
Release: Type 2
0.881
***
0.548
0.743
**
Interaction w/ High Performance
-0.363
Release: Type 3
3.389
***
2.160
***
3.613
***
Interaction w/ High Performance
-1.499
*
Release: Type 2 or 3
2.263
***
1.283
***
2.391
***
Interaction w/ High Performance
-1.072
***
High-Level Executives -0.177 -0.205 -0.770
**
-0.197 -0.810
**
-0.226
Interaction w/ High Performance -0.388 -0.416
Lev 2/3 Managers
0.493
**
0.800
***
0.945
***
0.059
1.102
***
0.552
Interaction w/ High Performance
0.293
0.097
Heads of Organizations -0.105 -0.110 0.870
**
0.018 0.783
*
0.018
Interaction w/ High Performance 0.588 0.563
Profit Center Managers
0.23
-0.58
*
1.07
**
0.13
0.91
*
-0.83
**
Interaction w/ High Performance
0.57
1.16
**
Log(Sales)
0.073
0.123
0.441
***
0.050
0.500
***
0.127
Growth in Sales
-0.337
-0.347
-1.455
**
0.585
-1.250
0.417
Intercept
-4.69
***
-5.41
***
-11.80
***
-4.03
**
-13.01
***
-5.35
***
S.D. (u
0
)
-0.35
*
-0.35
*
-0.16
-1.65
0.33
-1.01
102
Log-Likelihood
-1,002.25
-992.77
-370.47
-346.01
-364.14
-341.43
Observations
4,657
4,657
2,212
1,358
2,212
1,358
Prob > χ
2
0.000
0.000
0.000
0.000
0.000
0.000
Panel B: Comparisons of Coefficients
Cross-Segment Transfer Model 1 Cross-Segment Transfer Model 2
High Relative
Performance
Low Relative
Performance
High Relative
Performance
Low Relative
Performance
𝜷 𝑻𝟐 𝟑 𝑷 − 𝜷 𝑻𝟐 𝟑 𝑹 = 𝟎
1.392
***
-0.423
χ
2
(3.13)
(-1.09)
𝜷 𝑻𝟐
𝑷 − 𝜷 𝑻𝟐
𝑹 = 𝟎
1.592
**
0.859
χ
2
(6.330)
(1.930)
𝜷 𝑻𝟑
𝑷 − 𝜷 𝑻𝟑
𝑹 = 𝟎
0.503
-2.581
***
χ
2
(0.350)
(17.610)
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests. t-statistics for variables are omitted in the interest
of space.
Table 11 (Continued)
Dependent:
Cross-Segment Transfer
a
Cross-Segment Transfer
Model 1
b
Cross-Segment Transfer
Model 2
c
Model 1
b
Model 2
c
High Relative
Performance
d
Low Relative
Performance
d
High Relative
Performance
d
Low Relative
Performance
d
103
High Relative Performance is an indicator for whether a reporting segment’s ROA quintile belongs to the highest two, while Low Relative
Performance is an indicator that it belongs to the lowest two. Thus, the midst quintile does not belong to any group. Promotions or Releases Type X are
indicators for whether promotions or job re-assignments are associated with a hierarchical movement (Type 1), an appointment to or release from a head
position (Type 2), an appointment to or release from a profit center manager position (Type 3), or combinations of these. High-Level Executives is an
indicator for executives at the top three hierarchical levels. Lev 2/3 Managers indicates executives heading Level 2 or 3 organizations. Heads of
Organizations is an indicator for executives who carry “manager” title for any organizations that directly report to Levels 1 to 3 organization managers.
Profit Center Manager is an indicator for executives who supervise profit centers. Sales is measured per reporting segments. Sales growth is the growth
rate in sales of a segment between years t-1 and t. Hierarchical Level is constructed so that a higher numerical value indicates a higher level.
a
The dependent variable for all regression models in Table 8 is the likelihood of Cross-Segment Transfer.
b
Model 1 combines promotions Types 2 and 3, i.e., appointments to a manager of any type of organization.
c
Model 2 separates promotions Types 2 and 3.
d
High (Low) Relative Performance is the subset of executives whose reporting segment ROA is top (bottom) two quintiles within a company per year.
Table 11 (Continued)
104
Chapter 7. Discretion and Bias in Personnel Decisions: Empirical Evidence
7.1. Research Model: The likelihood of Personnel Decisions
The hypotheses are tested primarily by mixed-effects logistic regression models
75
with random intercepts that predict the likelihood of executives’ promotions, stays, and
dismissals. The full regression model is presented as follows:
𝑃𝑟 � 𝐷𝑒 𝑐𝑖 𝑠𝑖 𝑜 𝑛 𝑖𝑗𝑡 + 1
� = 𝛽 1
𝑅 𝐸 𝐿𝐴 𝑇 𝐼𝑂 𝑁 𝑖𝑗𝑡 + 𝛽 2
𝐿𝐸 𝑁𝐺𝑇 𝐻 𝑖𝑗𝑡
+ 𝛽 3
𝑅 𝐸 𝐿 𝑖𝑗𝑡
× 𝑅𝑂 𝐴 𝑗𝑡
+ 𝛽 4
𝑅 𝐸 𝐿 𝑖𝑗𝑡
× 𝐿 𝐸 𝑉 𝐸 𝐿 𝑖𝑗𝑡
+ 𝛽 5
𝑅 𝐸 𝐿 𝑖𝑗𝑡
× 𝑆𝐴 𝑀𝐸 𝑇 𝑌 𝑃 𝐸 𝑖𝑗𝑡
+ 𝛽 6
𝑅𝑂 𝐴 𝑗𝑡
+ 𝛽 7
𝐿𝐸 𝑉 𝐸 𝐿 𝑖𝑗𝑡
+ 𝛽 8
𝑆𝐴 𝑀𝐸 𝑇 𝑌 𝑃 𝐸 𝑖𝑗𝑡
+ 𝛽 9
𝐿 𝑜𝑔 � 𝑆𝐴𝐿 𝐸 𝑆 𝑗𝑡
�
+ 𝛽 1 0
𝐺𝑅 𝑂 𝑊𝑇 𝐻 𝑗𝑡
+ 𝛽 1 1
𝐿 𝑜𝑔 � 𝐴𝐺 𝐸 𝑖𝑗𝑡
� + Σ 𝛽 𝑥 + 𝑘 𝑇 𝐸 𝑁 𝑈𝑅 𝐸 𝐶 𝐴 𝑇 𝑖𝑗𝑡
+ Σ 𝛽 𝑦 + 𝑘 𝐸 𝐷 𝑈 𝑖𝑗𝑡
+ Σ 𝛽 𝑧 + 𝑘 𝐽𝑂 𝐵 𝑖𝑗𝑡
+ 𝛽 0
+ 𝑢 0𝑗𝑡
+ 𝜀 𝑖𝑗𝑡
……………(4)
76
The models contain the main independent variable of an established professional
relationship, an ROA for a reporting segment, the hierarchical level where an executive’s
title falls, the number of executives who have the same type of relationship with their
supervisor at a hierarchical level, and other known or plausible determinants of
promotions.
Tests of Interaction Terms
The statistical inferences in this study are drawn predominantly from the interaction
terms in non-linear models, more specifically logit regression models. However, the
interaction terms in such models may lead to incorrect inferences (Ai & Norton, 2003).
75
Mixed-effects models are used to handle the data’s longitudinal and multi-level feature.
76
REL is short for RELATION.
105
Currently, three plausible solutions
77
are available regarding the interpretation of
interaction terms: (1) statistical tests of marginal effects, (2) graphical presentations, and
(3) statistical tests using linear models.
First, statistical inferences can be drawn from tests of interaction following a
practical means to circumvent the problem embedded in computing marginal effects
78
of
interaction terms in non-linear models. Buis (2010) suggests the use of “marginal effect”
computed as the difference between the expected odds instead of computing cross partial
derivatives. Second, according to Greene (2010) and Mitchell and Chen (2005), visual
presentation of interaction effects can be used for interpretation. The interaction effects
are clearly identifiable with this approach. Third, Hellevik (2009) shows that significance
tests in logistic regressions and linear regressions are almost identical even when a linear
model is used to regress a dichotomous dependent variable. As linear models separate
interaction effects from the main effects and allow straightforward interpretations of
interaction terms, linear regressions have significant advantages in interpreting
interaction effects.
77
In addition to these solutions, Norton et al. (2004) provide the inteff command for STATA
®
to correct
coefficient estimates and standard errors of interaction terms. The command supports only one interaction
term in a model and does not fit for models with multiple interaction terms. So, this alternative is not
considered.
78
The interaction coefficient in a linear model is a cross partial derivative with respect to 𝑥 1
and 𝑥 2
, and
this equals to first derivative with respect to the multiplicative term 𝑥 1
𝑥 2
. However, in non-linear models,
cross partial derivatives are typically different from the first derivative with respect to 𝑥 1
𝑥 2
.
106
7.2. The Effects of Presence of Professional Relationships on Personnel Decisions
H5a: Information (Preference) Hypothesis predicts that an established professional
relationship does not affect (increases) the likelihood of promotion. To test its effects, the
presence of an executive’s social ties with his supervisor(s) is proxied by an indicator of
whether an immediate supervisor has awarded the executive’s previous promotion. Table
12 reports that the coefficients ( β
1
) on Presence of Relationship (RELATION) are
negative in five out of six models containing the variable, and three of them are
significant. This suggests that simply having a social connection with supervisors does
not improve, in fact reduces, the chance of promotion. Therefore, the Preference
Hypothesis is not supported. However, this does not indicate that the finding supports the
Information Hypothesis, either. Rather, it would be safer to argue that the finding is not
against the hypothesis.
H5b: Both hypotheses predict that an established professional relationship improves
the likelihood of stay. Consistently, all the coefficients ( β
1
) on Presence of Relationship
(RELATION) in Table 13 are positive, and three are significant at 1 percent level. The
results support H5b. However, the hypothesis does not differentiate the two explanations.
7.3. The Effects of Length of Professional Relationships on Personnel Decisions
H6a: The Information Hypothesis predicts a longer relationship’s positive influence
on the likelihood of promotion while the Preference Hypothesis does not. Table 12
presents strong support for this argument; all the coefficients ( β
2
) on Length of
Relationship (LENGTH) are positive and significant at 1 percent level, consistent with the
107
Information Hypothesis’s prediction. However, strong support for the Information
Hypothesis does not imply that the Preference Hypothesis is rejected. At the very least, it
does not reduce the likelihood; the presence of favoritism cannot be rejected yet. Further,
the proxy for the amount and the quality of communication, Length of relationship, may
capture another construct relevant to favoritism such as the strength of a relationship. So,
additional evidence is required to rule out the Preference Hypothesis for the Information
Hypothesis.
H6b: In Table 13, all the coefficients ( β
2
) on Length of Relationship (LENGTH) are
negative and significant at 1 percent level. If Length of Relationship captures something
relevant to favoritism, e.g., its intensity, the coefficients should not be negative for the
positive personnel decision, i.e., stays. Therefore, combined with the previous finding
from the test of H6a, the results in Table 13 provide strong support for the Information
Hypothesis and reject the Preference Hypothesis.
7.4. The Effects of Group Performance on Personnel Decisions
H7a: The Preference Hypothesis predicts that a supervisor with favoritism incentives
will exploit a weakness of a group performance measure as a performance indicator for
an individual, e.g., noisiness, to find an opportunity to promote her favored (but perhaps
less capable) subordinates. Therefore, the hypothesis expects a positive coefficient ( β
3
)
on the interaction term between Presence of Relationship and ROA (RELATION ROA).
On the contrary, with sound discretion, a supervisor does not have incentives to promote
subordinates simply because they have a connection with her or their group performance
108
is good—expecting a non-negative β
3
. The negative and insignificant coefficients in
Table 12 do not support the Preference Hypothesis, and provide weak evidence for the
Information Hypothesis.
H7b: The rationale for H7b is the same as that for H7a. Table 13 reports negative
and insignificant coefficients on β
3
, which leads to the same conclusion as in the test of
H7a.
7.5. The Effects of Performance on Personnel Decisions
H8a/b: The Information Hypothesis stresses the important of additional pieces of
information and the significance of personnel decisions (and their effects) for executives
at a high level. Accordingly, it expects an established professional relationship to
facilitate truthful communication between a supervisor and a subordinate to support a
supervisor’s better, informed decisions. So, a positive incremental effect on positive
personnel decisions (promotions (H8a) and stays (H8b)), i.e., a positive coefficient ( β
4
)
on the interaction term between Presence of Relationship and Hierarchical level
(RELATION LEVEL), would support this argument. A supervisor, acknowledging the
substantial effects of the decisions at a high level, is expected to shy away from
favoritism in making promotions.
Tables 12 and 13 present mixed results for the hypotheses. In Table 12, the
coefficients ( β
4
) on the interaction term carry positive signs, but one is not significant.
This finding weakly supports the Information Hypothesis. However, the negative (but
insignificant) β
4
’s in Table 13 are not consistent with any hypothesis.
109
Table 12: Determinants of Promotion Decisions
Prediction
b
M1 M2 M3 M4 M5 M6 M7 M8
Info / Pref
Presence of Relationship (β
1
)
(?) / (+)
0.030
-0.853
***
-0.829
***
-0.409
-1.438
***
-0.303
(0.320)
(-6.360)
(-5.700)
(-0.870)
(-8.260)
(-0.650)
Length of Relationship (β
2
)
(+) / NN
0.174
***
0.318
***
0.317
***
0.312
***
0.330
***
0.316
***
(7.360)
(9.730)
(9.710)
(9.460)
(9.980)
(9.520)
ROA * Relationship (β
3
)
NN / (+)
-0.003
-0.003
(-0.430)
(-0.440)
Level *Relationship (β
4
)
(+) / NP
0.103
0.277
***
(0.990)
(2.590)
No. of Same Type Execs
* Relationship (β
5
)
NP / (+)
0.013
***
0.015
***
(5.260)
(5.740)
ROA (%)
a
(+)
0.011
***
0.011
***
0.012
***
0.012
***
0.014
**
0.012
***
0.013
***
0.015
**
Hierarchical Level
a
(−)
-0.814
***
-0.822
***
-0.924
***
-0.791
***
-0.791
***
-0.849
***
-0.889
***
-1.048
***
No. of Executives of Same Type
a
(−)
-0.011
***
-0.011
***
-0.015
***
-0.010
***
-0.010
***
-0.009
***
-0.024
***
-0.025
***
Log(Sales)
a
0.248
***
0.250
***
0.274
***
0.224
***
0.224
***
0.218
***
0.310
***
0.306
***
Sales Growth
a
-0.066
-0.065
-0.060
-0.056
-0.056
-0.055
-0.067
-0.067
Log(Age)
a
2.000
***
2.033
***
1.654
**
0.528
0.544
0.520
0.768
0.801
Tenure: [3.5]
a
0.042
0.040
0.033
0.076
0.075
0.088
0.096
0.133
Tenure: [6, ∞)
a
-0.246
*
-0.240
*
-0.352
**
-0.615
***
-0.618
***
-0.598
***
-0.613
***
-0.574
***
Education: Master's
a
-0.078
-0.078
-0.098
-0.095
-0.095
-0.096
-0.092
-0.092
Education: Doctorate
a
0.322
**
0.321
**
0.311
**
0.345
**
0.345
**
0.351
**
0.324
**
0.342
**
Job Area: Marketing/Sales
a
0.022
0.024
0.037
-0.001
-0.002
0.001
0.008
0.015
Job Area: Engineering/Research
a
-0.238
**
-0.236
**
-0.207
**
-0.227
**
-0.226
**
-0.225
**
-0.228
**
-0.222
**
Constant
-16.239
***
-16.454
***
-15.923
***
-10.048
***
-10.110
***
-10.177
***
-12.250
***
-12.992
***
S.D. (u
0
)
0.436
***
0.445
***
0.581
***
0.426
***
0.423
***
0.417
***
0.519
***
0.505
***
Prob > chi2
<0.000
<0.000
<0.000
<0.000
<0.000
<0.000
<0.000
<0.000
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests. N=4,657 for all models. Z-statistics are reported in
parentheses.
110
N=4,657. Presence of Relationship is an indicator whether the current immediate superior has promoted an executive. Length of Relationship is the
number of years pasted since the current supervisor’s promotion award. ROA is computed as operating profit divided by total assets measured per
reporting segment. Hierarchical Level is constructed so that a higher numerical value indicates a higher level. Number of Executives of Same Type is
the number of executives who have the same property in terms of an established social tie with their immediate supervisor (have/not have) at a
hierarchical level in a reporting segment. Sales is measured per reporting segment. Sales Growth is the growth rate in sales between t-1 and t. Tenure
(i.e., years in the hierarchical rank) is categorized into three groups: tenure groups of (1) less than or equal to two years, (2) greater than or equal to three
years and less than or equal to five years, and (3) greater than or equal to six years. Dummy variables are used to accommodate the non-linear nature of
the variable. Education is categorized into three groups based on the final degree: executives with (1) lower than and including college degree, (2)
master’s degree, and (3) doctorate degree. Dummy variables are used to accommodate the non-linear nature of the variable. The base (i.e., 0) is for the
(lower than) college degree group. Job Area is categorized into three groups: executives in (1) general administration and management, (2) marketing
and sales, and (3) engineering and research. Dummy variables are used to accommodate the non-linear nature of the variable. The base is for the general
administration and management group.
a
Z-statistics for these variables are omitted in the interest of space.
b
Predictions are made for the Information Hypothesis (“Info”) and the Preference Hypothesis (“Pref”). NP and NN indicate “Non-Positive” and “Non-
Negative,” respectively.
Table 12 (Continued)
111
Table 13: Determinants of Stay (Dismissal) Decisions
Prediction
b
Info / Pref
M1
M2
M3
M4
M5
M6
M7
M8
Presence of Relationship (β
1
)
(+) / (+)
0.051
0.620
***
0.632
***
0.131
0.558
***
0.133
(0.450)
(3.760)
(3.430)
(0.300)
(2.720)
(0.290)
Length of Relationship (β
2
)
(−) / (+)
-0.082
***
-0.184
***
-0.184
***
-0.177
***
-0.182
***
-0.177
***
(-3.150)
(-4.900)
(-4.900)
(-4.680)
(-4.850)
(-4.670)
ROA * Relationship (β
3
)
NN / (+)
-0.002
-0.001
(-0.140)
(-0.040)
Level *Relationship (β
4
)
(+) / (+)
-0.122
-0.123
(-1.210)
(-1.090)
No. of Same Type Execs
NP / (+)
0.001
0.000
* Relationship (β
5
)
(0.510)
(-0.040)
ROA (%)
a
(+)
0.035
***
0.035
***
0.036
***
0.032
***
0.034
***
0.032
***
0.032
***
0.032
***
Hierarchical Level
a
(+)
0.667
***
0.652
***
0.726
***
0.617
***
0.618
***
0.685
***
0.608
***
0.687
***
No. of Executives of Same Type
a
(+)
0.006
***
0.006
**
0.008
***
0.004
**
0.005
**
0.004
*
0.003
0.004
Tenure: [3.5]
a
-0.394
***
-0.380
***
-0.449
***
-0.342
***
-0.343
***
-0.328
***
-0.328
***
-0.329
***
Tenure: [6, ∞)
a
-0.001
-0.003
0.006
-0.003
-0.004
-0.004
-0.004
-0.004
Education: Master's
a
-15.625
***
-15.576
***
-15.403
***
-14.610
***
-14.608
***
-14.619
***
-14.596
***
-14.620
***
Education: Doctorate
a
-0.032
-0.032
-0.018
0.009
0.009
-0.007
0.011
-0.007
Job Area: Marketing/Sales
a
-0.198
-0.188
-0.142
0.062
0.061
0.045
0.063
0.044
Job Area: Engineering/Research
a
-0.156
-0.158
-0.137
-0.147
-0.147
-0.144
-0.147
-0.144
Log(Sales)
a
-0.316
*
-0.317
*
-0.324
*
-0.344
*
-0.344
*
-0.362
*
-0.349
*
-0.362
*
Sales Growth
a
0.030
0.034
0.016
0.039
0.039
0.035
0.040
0.035
Log(Age)
a
-0.056
-0.052
-0.078
-0.052
-0.052
-0.056
-0.052
-0.056
Constant
71.803
***
71.324
***
72.065
***
66.771
***
66.769
***
66.883
***
66.519
***
66.908
***
S.D. (u
0
)
0.547
***
0.534
***
0.600
***
0.499
***
0.499
***
0.490
***
0.491
***
0.490
***
Prob > chi2
<0.000
<0.000
<0.000
<0.000
<0.000
<0.000
<0.000
<0.000
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests. N=4,657 for all models. Z-statistics are reported in
parentheses.
112
N=4,657. Presence of Relationship is an indicator whether the current immediate superior has promoted an executive. Length of Relationship is the
number of years pasted since the current supervisor’s promotion award. ROA is computed as operating profit divided by total assets measured per
reporting segment. Hierarchical Level is constructed so that a higher numerical value indicates a higher level. Number of Executives of Same Type is
the number of executives who have the same property in terms of an established social tie with their immediate supervisor (have/not have) at a
hierarchical level in a reporting segment. Sales is measured per reporting segment. Sales Growth is the growth rate in sales between t-1 and t. Tenure
(i.e., years in the hierarchical rank) is categorized into three groups: tenure groups of (1) less than or equal to two years, (2) greater than or equal to three
years and less than or equal to five years, and (3) greater than or equal to six years. Dummy variables are used to accommodate the non-linear nature of
the variable. Education is categorized into three groups based on the final degree: executives with (1) lower than and including college degree, (2)
master’s degree, and (3) doctorate degree. Dummy variables are used to accommodate the non-linear nature of the variable. The base (i.e., 0) is for the
(lower than) college degree group. Job Area is categorized into three groups: executives in (1) general administration and management, (2) marketing
and sales, and (3) engineering and research. Dummy variables are used to accommodate the non-linear nature of the variable. The base is for the general
administration and management group.
a
Z-statistics for these variables are omitted in the interest of space.
b
Predictions are made for the Information Hypothesis (“Info”) and the Preference Hypothesis (“Pref”). NP and NN indicate “Non-Positive” and “Non-
Negative,” respectively.
Table 13 (Continued)
113
7.6. The Effects of Promotion Competition on Personnel Decisions
H9a/b are developed to investigate whether favoring connected subordinates is
associated with a supervisor’s decision-efficiency orientation. H9 is tested by examining
the coefficients ( β
5
) on the interaction term between the number of executives who have
the same type of social connections with their supervisors, No. of Same Type Execs, and
Presence of Relationship (RELATION SAMETYPE).
H9a: The negative coefficient on the number of executives of a same type implies
that it becomes less probable for executives without a social tie to be promoted when
there are more promotion contestants outside a supervisor’s social network. On the
contrary, the positive and significant coefficients ( β
5
) on its interaction term in Table 12
indicates that when more candidates have connections with a supervisor, the supervisor is
more likely to pick subordinates within her network than to pick those outside the
network. In sum, as a supervisor has more promotion candidates to evaluate, the
likelihood of promotion is reduced for those without a connection, and it increases for
those with a connection. So, the Preference Hypothesis is supported. Accordingly, the
Information Hypothesis that expects non-positive coefficients is not supported. A
potential explanation for this outcome is that bounded rationality applies to all types of
supervisors, i.e., even those without favoritism incentives. In other words, a supervisor’s
limited resources make her narrow the promotion candidate pool first, relying on the
information acquired through a previously established relationship.
114
H9b: Table 13 presents β
5
’s not significantly different from zero. This is inconsistent
with the Preference Hypothesis, but consistent with the Information Hypothesis.
7.7. Additional Observations
M1 of Table 12 is a typical promotion prediction model in prior literature (e.g.,
Blackwell et al., 1994; Cichello et al., 2009). The model helps to see whether the dataset
is consistent with the conventional wisdom about what we expect for promotion decisions.
Not surprisingly, the results generally support the findings in the literature and our
expectation, providing the validity of the dataset to an extent. Overall, the signs and
significance of the coefficients in M1 remain unchanged in the other models.
First, group performance is a significant determinant of promotions. ROA positively
affects promotions. Second, a higher degree of competition for limited promotion
opportunities reduces the likelihood of promotion. Promotions become less likely (1) as
an executive moves up the hierarchy−negative and significant coefficients on
Hierarchical Level, or (2) when there are more contestants−negative and significant
coefficients on No. of executives of same type. Third, an organization’s capacity to feed
(more) promotions improves the likelihood of promotions—positive and significant
coefficients on Log(Sales). Fourth, human capital variables also explain promotions. Age,
in some models, positively affects promotions whereas staying at a job level for an
extended time, Tenure [6, ∞), has an adverse effect on further advancement into a next
level position. Education improves the likelihood of promotion only when candidates
have the most advanced degree. Lastly, promotions are less likely for executive level
115
engineers and researchers compared to executives in general administration and
marketing/sales jobs.
7.8. Additional Tests of Interaction Terms
As discussed in Section 7.1., the statistical inferences of this research depend on
interaction terms in non-linear models. To address Ai and Norton’s (2003) concern,
alternative means are explored. First, interaction terms are statistically tested following
Buis (2010) and the results are presented in Table 14. The table presents the predicted
odds varying in three independent variables interacting with the presence of social ties:
ROA, hierarchical level, and the number of executives of a same type. Two steps are
taken to test the effects of the interaction terms. First, differences in the expected odds
between different relationships (i.e., Presence of relationship, or RELATION) at a level in
the five variables are calculated under the fifth column (i.e., Difference (Yes – No)). Then,
the difference in differences between levels of the other component variable in an
interaction term is presented in the seventh column (i.e., Diff in Diff). Overall, differences
between different social relationships change between levels of the other interacting
variables. More importantly, the table shows that the difference in differences generally
displays similar statistical significance as in the original logit models, confirming the
interaction effects.
Second, visual presentation of interaction effects as provided in Figure 2 is be used
for interpretation. The interaction effects are clearly identifiable with this approach.
Figure 2 visualizes the marginal effects computed following Buis’s (2010) method. The
116
three pairs of graphs correspond to the three main interaction terms in M7 of Tables 12
and 13. The Y-axis represents the predicted probabilities varying with the variables on
the X-axis. The graphs show that the differences in predicted probabilities between the
two lines, representing the relationships with supervisors, vary with the changes in
variables along the X-axis. The visualized presentation is generally consistent with the
findings from M8 of Tables 12 and 13.
Third, following Hellevik (2009), the dichotomous dependent variable (i.e., an indicator
of promotion at t+1) is linearly regressed on the same independent variable set from the
corresponding logit models. Untabulated results from the linear models confirm that
statistical inferences from the interaction terms in these non-linear models remain valid.
79
7.9. Additional Tests Using Alternative Measure of Social Ties
Additionally, a proxy for different type of social connection, or school ties, is
measured as an indicator of whether alma mater of an executive and his supervisor at a
level 3 organization are the same. The additional tests using school ties (untabulated)
show that the new proxy does not produce any significant results in its main effects and
interaction effects with other relevant variables.
79
The signs and the statistical significance of the coefficients do not change in the linear models.
117
Table 14: Testing Interaction Effects
a
Panel A: Effects on Promotions
Predicted Odds
b
Group
or Value
c
RELATION
Difference
(Yes - No)
Diff.
in Diff.
d
Variable
Yes
No
z-stat
z-stat
ROA
28%
0.276
0.370
-0.094
***
-2.88
18%
0.239
0.320
-0.082
***
-3.00
-0.013
*
-1.68
8%
0.206
0.277
-0.071
***
-3.03
-0.011
*
-1.83
-2%
0.178
0.239
-0.061
***
-2.96
-0.010
**
-2.01
-12%
0.154
0.207
-0.053
***
-2.82
-0.008
**
-2.20
Hierarchical
Level
1 (Low)
0.259
0.529
-0.270
***
-4.58
2
0.221
0.168
0.053
**
2.29
-0.323
***
-5.26
3
0.134
0.142
-0.009
-0.41
0.062
***
2.92
4
0.085
0.070
0.015
0.84
-0.024
***
-2.73
5 (High)
0.074
0.027
0.047
**
2.47
-0.032
***
-2.85
No of Executives
20
1.416
0.516
0.900
**
1.99
of Same Type
45
0.767
0.279
0.488
**
2.29
0.413
*
1.70
70
0.415
0.151
0.264
***
2.67
0.224
*
1.94
95
0.225
0.082
0.143
***
3.12
0.121
**
2.23
120
0.122
0.044
0.077
***
3.57
0.066
***
2.60
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests.
a
The interaction effects are estimated based on the decision prediction model M8 specified in Table 12.
The test of interaction terms for non-linear models follows the process introduced in Buis (2010).
b
The odds are
p r( p r o mo t i o n
t + 1
)
1 − p r( p r o mo t i o n
t + 1
)
.
c
For continuous variables (i.e., ROA, the number of executives of same type), odds are computed at
specific values.
d
Difference in differences is calculated by subtracting the difference between different social tie
relationship (i.e., RELATION = 1 and 0) of the current group or specified value from that of the previous
group or specified value.
118
Table 14 (Continued)
Panel B: Effects on Stays (Dismissals)
Predicted Odds
b
Group
or Value
c
RELATION
Difference
(Yes - No)
Diff.
in Diff.
d
Variable
Yes
No
z-stat
z-stat
ROA
28%
0.231
0.311
-0.080
***
-2.91
18%
0.171
0.231
-0.059
***
-2.81
-0.021
**
-2.34
8%
0.127
0.172
-0.044
***
-2.57
-0.015
***
-2.62
-2%
0.095
0.127
-0.033
**
-2.28
-0.011
***
-2.84
-12%
0.070
0.095
-0.024
**
-2.00
-0.008
***
-2.90
Hierarchical
Level
1 (Low)
33.827
25.346
8.480
0.29
2
37.011
14.315
22.697
1.08
-14.216
-0.52
3
11.623
7.574
4.049
**
2.12
18.648
0.91
4
23.357
32.791
-9.435
-0.20
13.484
0.29
5 (High)
5.501
15.869
-10.368
**
-2.13
0.934
0.02
No of Executives
20
26.457
17.957
8.499
0.44
of Same Type
45
29.259
19.859
9.399
0.44
-0.900
-0.42
70
32.357
21.962
10.395
0.44
-0.995
-0.41
95
35.784
24.289
11.496
0.44
-1.101
-0.40
120
39.574
26.861
12.713
0.44
-1.218
-0.40
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests.
a
The interaction effects are estimated based on the decision prediction model M8 specified Table 13. The
test of interaction terms for non-linear models follows the process introduced in Buis (2010).
b
The odds are
p r( p r o mo t i o n
t + 1
)
1 − p r( p r o mo t i o n
t + 1
)
.
c
For continuous variables (i.e., ROA, the number of executives of same type), odds are computed at
specific values.
d
Difference in differences is calculated by subtracting the difference between different social tie
relationship (i.e., RELATION= 1 and 0) of the current group or specified value from that of the previous
group or specified value.
119
Figure 2: Interaction Effects on Personnel Decisions
a
Panel A: ROA × Relationship
Promotions Stays
Coefficient (z-stat): -0.003 (-0.440) Coefficient (z-stat): -0.001 (-0.040)
Panel B: Hierarchical Level × Relationship
Promotions Stays
Coefficient (z-stat): 0.277
***
(2.590) Coefficient (z-stat): -0.123 (-1.090)
0%
5%
10%
15%
20%
25%
30%
-30% -10% 10% 30%
Pr(Promotion)
ROA
0%
5%
10%
15%
20%
25%
30%
60%
65%
70%
75%
80%
85%
90%
95%
100%
-30% -10% 10% 30%
Difference
Pr(Stay)
ROA
-20%
-10%
0%
10%
20%
30%
40%
1 2 3 4 5
Pr(Promotion)
Hierarchical Level
-10%
-5%
0%
5%
10%
15%
20%
25%
30%
75%
80%
85%
90%
95%
100%
1 2 3 4 5
Difference
Pr(Stay)
Hierarchical Level
Difference
RELATION=0
RELATION=1
Difference
RELATION=0
RELATION=1
120
Figure 2 (Continued)
Panel C: No. of Same Type Executives × Relationship
Promotions Stays
Coefficient (z-stat): 0.015
***
(5.740) Coefficient (z-stat): 0.000 (-0.040)
*
,
**
,
***
Significant at 10 percent, 5 percent, and 1 percent, respectively, all based on two-tailed tests.
a
The graphs present the interaction effects on promotion decisions. The probabilities of promotion and stay
affected by interactions, coefficients and z-statistics are estimated based on the personnel decision
prediction model M8 specified in Tables 12 and 13:
𝑝 𝑟 � 𝐷 𝑒 𝑐 𝑖𝑠 𝑖𝑜 𝑛 𝑖𝑗 𝑡 + 1
�
= 𝛽 1
𝑅 𝐸𝐿𝐴 𝑇 𝐼 𝑂 𝑁 𝑖𝑗 𝑡 + 𝛽 2
𝐿 𝐸𝑁 𝐺 𝑇 𝐻 𝑖𝑗 𝑡 + 𝛽 3
𝑅𝐸 𝐿 𝑖𝑗 𝑡 × 𝑅𝑂 𝐴 𝑗𝑡
+ 𝛽 4
𝑅𝐸 𝐿 𝑖𝑗 𝑡 × 𝐿𝐸 𝑉𝐸 𝐿 𝑖𝑗 𝑡 + 𝛽 5
𝑅𝐸 𝐿 𝑖𝑗 𝑡 × 𝑆𝐴 𝑀 𝐸𝑇𝑌𝑃 𝐸 𝑖𝑗 𝑡 + 𝛽 6
𝑅𝑂 𝐴 𝑗𝑡
+ 𝛽 7
𝐿𝐸 𝑉𝐸 𝐿 𝑖𝑗 𝑡 + 𝛽 8
𝑆𝐴 𝑀 𝐸𝑇𝑌𝑃 𝐸 𝑖𝑗 𝑡 + 𝛽 9
𝐿 𝑜𝑔 � 𝑆 𝐴 𝐿𝐸 𝑆 𝑗𝑡
� + 𝛽 1 0
𝐺 𝑅 𝑂𝑊 𝑇 𝐻 𝑗𝑡
+ 𝛽 1 1
𝐿 𝑜𝑔 � 𝐴𝐺 𝐸 𝑖𝑗 𝑡 � + Σ 𝛽 𝑥 + 𝑘 𝑇 𝐸 𝑁 𝑈 𝑅𝐸𝐶 𝐴 𝑇 𝑖𝑗 𝑡 + Σ 𝛽 𝑦 + 𝑘 𝐸𝐷 𝑈 𝑖𝑗 𝑡 + Σ 𝛽 𝑧 + 𝑘 𝐽 𝑂 𝐵 𝑖𝑗 𝑡 + 𝛽 0
+ 𝑢 0 𝑗𝑡
+ 𝜀 𝑖𝑗 𝑡
0%
10%
20%
30%
40%
50%
60%
70%
80%
0 50 100 150
Pr(Promotion)
No of Executives of Same Type in a
Reporting Segment
0%
1%
2%
3%
4%
5%
6%
7%
8%
9%
10%
90%
91%
92%
93%
94%
95%
96%
97%
98%
99%
0 50 100 150
Difference
Pr(Stay)
No of Executives of Same Type in a
Reporting Segment
Difference
RELATION=0
RELATION=1
121
Chapter 8. Discussion and Conclusion
This dissertation addresses two broad research questions regarding careers in
organizations that prior literature has little discussed. First, how does accounting
performance of an organization affect personnel decisions of executives working for the
organization? Second, how a promotion candidate’s social relationship with the
supervisor affects his or her careers?
8.1. The Effects of Accounting Performance on Personnel Decisions
To address the first research question, I examine the effects of accounting
performance on the likelihood of promotions and dismissals of executives in an
organization. Existing evidence points at the association between individual performance
and personnel decisions (e.g., Blackwell et al., 1994; Campbell, 2008; Cichello et al.,
2009; Medoff & Abraham, 1980) and agency theory prescribes a non-zero weight for an
organizational performance indicator in explicit incentive contracts (Feltham & Xie, 1994;
Holmström, 1979). Despite these, we know little about the effect of organizational
performance on the personnel decisions of workers whose contribution to the
organizational performance is limited.
To fill the void in the literature, I investigate the effects of accounting performance
measured at corporate and segment levels on the personnel decisions of all executives
working in lower level organizations, i.e., subsidiaries and divisions, in a large Korean
conglomerate. This allows an empirical test of a supervisor’s adjustment of promotion
probabilities of subordinates. In addition, I identify decision-making circumstances where
122
supervisors may recognize the usefulness of accounting performance differently in
making personnel decisions of their subordinates: type of promotion, candidates’ job
responsibility, and organizational interdependencies among business divisions.
Overall, the analysis confirms that supervisors do associate personnel decisions with
accounting performance and they use accounting performance in different ways,
depending on the decision-making contexts. First, I find significant effects of accounting
performance of an organization on the probabilities of personnel decisions of executives
in the organization. The positive (negative) association between accounting performance
and promotions (dismissals) supports H1 that organizational performance in addition to
individual performance is factored in when supervisors make personnel decisions.
Second, the test of H2 shows that the effects are significant only for the promotions
involving a hierarchical advancement and accordingly, a non-trivial compensation
increase. The finding suggests that organizational performance is considered only when
promotions serve as an incentive provision mechanism. This, in other words, implies that
promotions provide incentives for executives’ managerial effort toward organizational
performance measures.
Third, the promotions of executives with greater job responsibility or at higher ranks
are less sensitive to corporate and segment ROAs. The finding is contrary to my
prediction that promotions of high-rank executives who are more responsible for
organizational performance should be more strongly tied to the accounting performance
of an organization. However, the finding may indicate that the costs and capacity
123
constraints may outweigh the marginal incentives from promotions of high-rank
executives that are awarded based on organizational performance.
Fourth, I also find that promotion opportunities are more sensitive to corporate or
segment ROAs when cross-unit transfers between or within segments are frequent.
Finally, the investigation of the association between promotions and cross-unit transfers
shows that (1) outgoing transfers from a high-performing segment are more likely for
executives in supervisory positions and (2) such job transfers are more likely to
accompany appointments to managerial positions than releases from supervisory
positions. Taken together, the findings present an interesting picture regarding cross-unit
transfers as a way to overcome an organizational capacity constraint (i.e., the number of
possible promotions that an organization can award) and to transfer knowledge across the
organization. The results show that the conglomerate often awards promotions in
combination with cross-unit transfers to relatively poorly performing business units
where dismissals are more likely and thus more open positions are expected. The
combination of promotion and outgoing transfer eases the organizational capacity
constraint, and, as a result, helps to retain the intensity of promotion-based incentives
with least organizational growth (Baker et al., 1988; Milgrom & Roberts, 1992). This
finding is also consistent with the argument that the combination helps to improve the
efficiency of intra-firm knowledge transfer or best practice diffusion.
124
8.2. The Effects of Social Relationships on Personnel Decisions
In the second part of the dissertation, I examine the two behavioral outcomes of the
use of subjective evaluations in personnel decisions: (sound) use of discretion vs.
favoritism. Prior literature suggests that subjective evaluations in general performance
evaluation contexts provide several advantages over the sole use of objective measures
and, at the same time, arouse a supervisor’s (undesirable) behaviors—“biases.” Instead of
the two most frequently discussed supervisor biases, i.e., leniency and centrality, in this
research, a third type of bias, i.e., favoritism, is being studied. Is favoritism prevalent in
organizations? Is any alternative explanation of the ostensible manifestation of favoritism
available? If any, which is more descriptive of these supervisors’ behaviors?
I conjecture that even a supervisor’s fair and sound use of discretion in evaluating
subordinates subjectively may be mistaken as favoritism. The dataset opens up two
possibilities. The distribution of personnel decisions during the sample period suggests
that socially connected executives are more likely to be promoted. It suggests that a
supervisor, at a glance, seems to favor some subordinates in personnel decisions.
However, Table 5 describing the dataset also suggests another possibility; a more
selective process is being made.
With five sets of hypotheses, I compare the two competing explanations of
supervisors’ behaviors in personnel decisions: the Information Hypothesis and the
Preference Hypothesis. Overall, the results support the Information Hypothesis with few
exceptions. Based on these findings, I argue that this behavior clearly differs from
125
favoritism that only values the current status of social connections. Rather, (strong)
professional relationships enable a more careful selection process by facilitating effective
communication between a supervisor and a subordinate and comprehensive evaluation of
overall qualifications that are hardly observed by objective and/or quantitative
performance indicators. This finding suggests that with adequate incentives for performance,
principals may constrain managers from misreporting their subordinates’ abilities for their
utility—consistent with Prendergast and Topel (1996)—and encourage an establishment of
trust along the hierarchy. Trust, in turn, reinforces the efficiency of incentive contracts using
subjective evaluations.
8.3. Research Opportunity, Limitations, and Future Direction
Researchers have developed several theoretical frameworks to explain how careers
in organizations are structured and how they are determined; see Gibbons and Waldman’s
(1999) survey of theoretical literature on careers in organizations. Despite the effort, no
single theory provides comprehensive description (Gibbons & Waldman, 1999). For
empirical part, it is important to supply more evidence of regularities from the practice
not only to confirm what theories predict but also to encourage the development of more
integrative models. However, compared to the development in theoretical literature in
this area, a relatively small set of studies such as Medoff and Abraham (1980) and Baker,
Gibbs, et al. (1994a, 1994b) has studied workers’ careers within firms. It is, however, of
no surprise that the lack of empirical research is primarily due to the inaccessibility to
confidential personnel data.
126
I employ a dataset that is collected from a publicly available data source, or
specifically annual reports of subsidiaries in a large Korean conglomerate. The dataset
offers unique features that are not available from U.S. companies and that, more
importantly, provide a clean research setting. The advantages of using the dataset are
threefold. First, the conglomerate relies primarily on the internal labor market to develop
its managerial capacity. The characteristics of the conglomerate’s human resource
development practice address the concern about the external labor market’s impact on
executives’ personnel decisions. Second, the conglomerate’s “dual career system”
(i.e.,
the separation of hierarchical titles and roles) helps to identify different types of
promotions. Third, the dataset allows me to investigate organizational contexts—job
responsibility and cross-unit transfers.
Despite the unique research opportunity, the dataset has several limitations. First of
all, a limitation pertains to the use of Korean data. Individuals’ behaviors responding to
the two types of incentives discussed in this study, i.e., sound use of discretion and
favoritism, are universal; the findings in this dissertation are generally consistent with
theories and prior empirical findings. However, admittedly culture-specific variations in
the degree of effects may exist. Unfortunately, I am unaware of any currently available
comparable data for U.S. or other national firms. So, evidence from other cultural /
institutional environments will expand the validity of the findings in Korea and their
application to other cultures.
127
Second, the dataset is comprehensive but not complete; although it provides a unique
research setting where organizational factors and diverse career events can be
investigated at the same time, it lacks performance measures for low-level organizations
(e.g., below segment levels) and individual executives (e.g., supervisors’ ratings of their
subordinates for personnel decisions) that may be correlated with personnel decisions. In
that regard, exclusive and confidential data that may complement the publicly available
information could have enriched this study. Owing to the lack of such information, I
cannot completely dismiss alternative explanations for some of the observed phenomena.
However, this limitation opens an opportunity for future, related studies, as
complementary data can be acquired.
A potential direction for future research would be to pursue field-study type research
which builds on the main findings from this archival study. The field research may focus
on other factors that may affect supervisors’ subjective evaluations, but are not
necessarily related to accounting performance. For example, one interesting topic of
investigation would be why or how, rather than whether, supervisors’ consideration of
organizational demand for knowledge transfer or social relationships between a
supervisor and subordinates affects personnel decisions. In addition, the type of research
allows access to richer and more in-depth data including a better measure for the presence
and strength of professional ties. The future study may also identify better proxies for
quality communication established through social ties and favoritism and separate two
potentially different constructs. Moreover, although a reasonable alternative proxy for
128
favoritism, i.e., school ties, is considered in this study, identifying other types of
favoritism would be a challenge for a future study as well. Another type of potentially
interesting research would examine whether promotion decisions based on social
incentives are efficient in terms of organizational performance.
129
Bibliography
Adler, P. S., & Kwon, S. W. (2002). Social Capital: Prospects for a New Concept. The
Academy of Management Review, 27(1), 17-40.
Aggarwal, R. K., & Samwick, A. A. (2003). Performance Incentives within Firms: The
Effect of Managerial Responsibility. The Journal of Finance, 58(4), 1613-1650.
Ai, C., & Norton, E. (2003). Interaction Terms in Logit and Probit Models. Economics
Letters, 80(1), 123-129.
Antle, R., & Smith, A. (1986). An Empirical Investigation of the Relative Performance
Evaluation of Corporate Executives. Journal of Accounting Research, 24(1), 1-39.
Ariga, K., Ohkusa, Y., & Brunello, G. (1999). Fast Track: Is It in the Genes? The
Promotion Policy of a Large Japanese Firm. Journal of Economic Behavior &
Organization, 38(4), 385-402.
Bae, J., & Lawler, J. J. (2000). Organizational and Hrm Strategies in Korea: Impact on
Firm Performance in an Emerging Economy. The Academy of Management
Journal, 43(3), 502-517.
Baik, B., Kim, K., Evans III, J. H., & Yanadori, Y. (2011). White Collar Incentives.
Paper presented at the AAA 2011 General Meeting, Denver, CO.
Baiman, S., & Rajan, M. V. (1995). The Informational Advantages of Discretionary
Bonus Schemes. The Accounting Review, 70(4), 557-579.
Baker, G. P., Gibbons, R. S., & Murphy, K. J. (1994). Subjective Performance Measures
in Optimal Incentive Contracts. The Quarterly Journal of Economics, 109(4),
1125-1156.
Baker, G. P., Gibbs, M., & Holmstrom, B. (1994a). The Internal Economics of the Firm:
Evidence from Personnel Data. The Quarterly Journal of Economics, 109(4), 881-
919.
Baker, G. P., Gibbs, M., & Holmstrom, B. (1994b). The Wage Policy of a Firm. The
Quarterly Journal of Economics, 109(4), 921-955.
Baker, G. P., Jensen, M. C., & Murphy, K. J. (1988). Compensation and Incentives:
Practice Vs. Theory. The Journal of Finance, 43(3), 593-616.
130
Bandiera, O., Barankay, I., & Rasul, I. (2008). Social Capital in the Workplace: Evidence
on Its Formation and Consequences. Labour Economics, 15(4), 724-748.
Bandiera, O., Barankay, I., & Rasul, I. (2009). Social Connections and Incentives in the
Workplace: Evidence from Personnel Data. Econometrica, 77(4), 1047-1094.
Banker, R. D., & Datar, S. M. (1989). Sensitivity, Precision, and Linear Aggregation of
Signals for Performance Evaluation. Journal of Accounting Research, 27(1), 21-
39.
Barro, J., & Barro, R. J. (1990). Pay, Performance, and Turnover of Bank Ceos: National
Bureau of Economic Research Cambridge, Mass., USA.
Becker, G. S. (1964). Human Capital. New York: Columbia Univ. Press.
Blackwell, D. W., Brickley, J. A., & Weisbach, M. S. (1994). Accounting Information
and Internal Performance Evaluation : Evidence from Texas Banks. Journal of
Accounting and Economics, 17(3), 331-358.
Bol, J. C. (2011). The Determinants and Performance Effects of Managers' Performance
Evaluation Biases. The Accounting Review, 86(5), 1549-1575.
Boschma, R., Eriksson, R., & Lindgren, U. (2009). How Does Labour Mobility Affect
the Performance of Plants? The Importance of Relatedness and Geographical
Proximity. Journal of Economic Geography, 9(2), 169-190.
Bourdieu, P. (1985). The Forms of Capital. In J. G. Richardson (Ed.), Handbook of
Theory and Research for the Sociology of Education (pp. 241-258). New York:
Greenwood.
Bourdieu, P., & Wacquant, L. J. D. (1992). An Invitation to Reflexive Sociology. Chicago,
IL: University of Chicago Press.
Breuer, K., Nieken, P., & Sliwka, D. (2010). Social Ties and Subjective Performance
Evaluations: An Empirical Investigation. IZA Discussion Paper Series (No. 4913).
Brickley, J. A. (2003). Empirical Research on Ceo Turnover and Firm-Performance: A
Discussion. Journal of accounting and economics, 36(1), 227-233.
Buis, M. L. (2010). Stata Tip 87: Interpretation of Interactions in Nonlinear Models. Stata
Journal, 10(2), 305-308.
131
Buraimo, B., Forrest, D., & Simmons, R. (2010). The 12th Man?: Refereeing Bias in
English and German Soccer. Journal of the Royal Statistical Society: Series A
(Statistics in Society), 173(2), 431-449.
Bushman, R. M., Indjejikian, R. J., & Smith, A. (1995). Aggregate Performance
Measures in Business Unit Manager Compensation: The Role of Intrafirm
Interdependencies. Journal of Accounting Research, 33, 101-128.
Campbell, D. (2008). Nonfinancial Performance Measures and Promotion-Based
Incentives. Journal of Accounting Research, 46(2), 297-332.
Cichello, M., Fee, C., Hadlock, C., & Sonti, R. (2009). Promotions, Turnover, and
Performance Evaluation: Evidence from the Careers of Division Managers. The
Accounting Review, 84(4), 1119.
Coleman, J. S. (1988). Social Capital in the Creation of Human Capital. The American
Journal of Sociology, 94, S95-S120.
Coughlan, A. T., & Schmidt, R. M. (1985). Executive Compensation, Management
Turnover, and Firm Performance: An Empirical Investigation. Journal of
accounting and economics, 7(1), 43-66.
Datar, S., Kulp, S. C., & Lambert, R. A. (2001). Balancing Performance Measures.
Journal of Accounting Research, 39(1), 75-92.
Dawson, P., & Dobson, S. (2010). The Influence of Social Pressure and Nationality on
Individual Decisions: Evidence from the Behaviour of Referees. Journal of
Economic Psychology, 31(2), 181-191.
Demougin, D., & Siow, A. (1994). Careers in Ongoing Hierarchies. The American
Economic Review, 1261-1277.
Dohmen, T. J. (2008). The Influence of Social Forces: Evidence from the Behavior of
Football Referees. Economic Inquiry, 46(3), 411-424.
Ederhof, M. (2011). Incentive Compensation and Promotion-Based Incentives of Mid-
Level Managers: Evidence from a Multinational Corporation. The Accounting
Review, 86, 131.
Fairburn, J. A., & Malcomson, J. M. (2001). Performance, Promotion, and the Peter
Principle. The Review of Economic Studies, 68(1), 45-66.
132
Fee, C. E., Hadlock, C. J., & Pierce, J. R. (2006). Promotions in the Internal and External
Labor Market: Evidence from Professional Football Coaching Careers. The
Journal of Business, 79(2), 821-850.
Fehr, E., & Fischbacher, U. (2002). Why Social Preferences Matter-the Impact of Non-
Selfish Motives on Competition, Cooperation and Incentives. Economic Journal,
1-33.
Feltham, G. A., & Xie, J. (1994). Performance Measure Congruity and Diversity in
Multi-Task Principal/Agent Relations. The Accounting Review, 69(3), 429-453.
Fukuyama, F. (1997). Social Capital and the Modern Capitalist Economy: Creating a
High Trust Workplace. Stern Business Magazine, 4(1), 4-16.
Garicano, L., Palacios-Huerta, I., & Prendergast, C. (2005). Favoritism under Social
Pressure. Review of Economics and Statistics, 87(2), 208-216.
Gibbons, R. S. (1998). Incentives in Organizations. The Journal of Economic
Perspectives, 12(4), 115-132.
Gibbons, R. S., & Murphy, K. J. (1990). Relative Performance Evaluation for Chief
Executive Officers. Industrial and Labor Relations Review, 30-51.
Gibbons, R. S., & Waldman, M. (1999). Careers in Organizations: Theory and Evidence.
Handbook of labor economics, 3, 2373-2437.
Gibbs, M. (1995). Incentive Compensation in a Corporate Hierarchy. Journal of
Accounting and Economics, 19(2-3), 247-277.
Gibbs, M. (1996). Promotions and Incentives. Unpublished working paper, University of
Chicago.
Gibbs, M. (2008). Discussion of Nonfinancial Performance Measures and Promotion-
Based Incentives. Journal of Accounting Research, 46(2), 333-340.
Gibbs, M., Merchant, K. A., Van der Stede, W. A., & Vargus, M. E. (2004).
Determinants and Effects of Subjectivity in Incentives. The Accounting Review,
79(2), 409-436.
Grabner, I., & Moers, F. (2011). Managers’ Choices of Evaluation Criteria in Promotion
Decisions: An Analysis of Alternative Job Assignments. Paper presented at the
AAA 2012 Management Accounting Section (MAS) Meeting, Houston, TX.
133
Granovetter, M. S. (1973). The Strength of Weak Ties. American Journal of Sociology,
78(6), 1360.
Greene, W. (2010). Testing Hypotheses About Interaction Terms in Nonlinear Models.
Economics Letters, 107(2), 291-296.
Gupta, A. K., & Govindarajan, V. (2000). Knowledge Flows within Multinational
Corporations. Strategic management journal, 21(4), 473-496.
Hellevik, O. (2009). Linear Versus Logistic Regression When the Dependent Variable Is
a Dichotomy. Quality and Quantity, 43(1), 59-74.
Holmström, B. R. (1979). Moral Hazard and Observability. The Bell Journal of
Economics, 74-91.
Ittner, C. D., Larcker, D. F., & Meyer, M. W. (2003). Subjectivity and the Weighting of
Performance Measures: Evidence from a Balanced Scorecard. The Accounting
Review, 78(3), 725-758.
Kahn, C., & Huberman, G. (1988). Two-Sided Uncertainty and" up-or-out" Contracts.
Journal of Labor Economics, 423-444.
KEF. (2011). 2011 Survey on Promitons: The Korea Employer Federation.
Kim, S., & Briscoe, D. R. (1997). Globalization and a New Human Resource Policy in
Korea: Transformation to a Performance-Based Hrm. Employee relations, 19(4),
298-308.
Kwon, I. (2006a). Endogenous Favoritism in Organizations. Topics in Theoretical
Economics, 6(1).
Kwon, I. (2006b). Incentives, Wages, and Promotions: Theory and Evidence. The Rand
journal of economics, 37(1), 100-120.
Lawler, E. (1971). Pay and Orgnizational Effectiveness: A Psychological View. New
York: McGraw Hill.
Lazear, E. P., & Rosen, S. (1981). Rank-Order Tournaments as Optimum Labor
Contracts. The Journal of Political Economy, 89(5), 841-864.
Lluis, S. (2005). The Role of Comparative Advantage and Learning in Wage Dynamics
and Intrafirm Mobility: Evidence from Germany. Journal of Labor Economics,
23(4), 725-767.
134
Longenecker, C. O., Sims, H. P. J., & Gioia, D. A. (1987). Behind the Mask: The Politics
of Employee Appraisal. The Academy of Management Executive, 1(3), 183-193.
Malcomson, J. M. (1984). Work Incentives, Hierarchy, and Internal Labor Markets. The
Journal of Political Economy, 486-507.
Medoff, J. L., & Abraham, K. G. (1980). Experience, Performance, and Earnings. The
Quarterly Journal of Economics, 95(4), 703-736.
Milgrom, P. R., & Roberts, J. (1992). Economics, Organization and Management (Vol.
7). Englewood Cliffs, NJ: Prentice-Hall International.
Milliman, J. F., Kim, Y. M., & Von Glinow, M. A. (1993). Hierarchical Advancement in
Korean Chaebols: A Model and Research Agenda. Human Resource Management
Review, 3(4), 293-320.
Milliman, J. F., Nason, S., Zhu, C., & De Cieri, H. (2002). An Exploratory Assessment of
the Purposes of Performance Appraisals in North and Central America and the
Pacific Rim. Human Resource Management, 41(1), 87-102.
Minbaeva, D., Pedersen, T., Bjorkman, I., Fey, C. F., & Park, H. J. (2003). Mnc
Knowledge Transfer, Subsidiary Absorptive Capacity, and Hrm. Journal of
International Business Studies, 34(6), 586-599.
Mitchell, M. N., & Chen, X. (2005). Visualizing Main Effects and Interactions for Binary
Logit Models. Stata Journal, 5(1), 64-82.
Moers, F. (2005). Discretion and Bias in Performance Evaluation: The Impact of
Diversity and Subjectivity. Accounting, Organizations and Society, 30(1), 67-80.
Mookherjee, D. (1984). Optimal Incentive Schemes with Many Agents. The Review of
Economic Studies, 51(3), 433-446.
Mowery, D. C., Oxley, J. E., & Silverman, B. S. (1996). Strategic Alliances and Interfirm
Knowledge Transfer. Strategic management journal, 17, 77-91.
Murphy, K. J. (1985). Corporate Performance and Managerial Remuneration: An
Empirical Analysis. Journal of accounting and economics, 7(1), 11-42.
Murphy, K. J. (1986). Incentives, Learning, and Compensation: A Theoretical and
Empirical Investigation of Managerial Labor Contracts. The Rand journal of
economics, 59-76.
135
Murphy, K. J. (1992). Performance Measurement and Appraisal: Motivating Managers
To. Performance measurement, evaluation, and incentives, 37.
Murphy, K. J., & Zimmerman, J. L. (1993). Financial Performance Surrounding Ceo
Turnover. Journal of Accounting and Economics, 16(1-3), 273-315.
Parrino, R. (1997). Ceo Turnover and Outside Succession a Cross-Sectional Analysis.
Journal of financial Economics, 46(2), 165-197.
Pfeffer, J., & Cohen, Y. (1984). Determinants of Internal Labor Markets in Organizations.
Administrative Science Quarterly, 29(4), 550-572.
Prendergast, C. (1993a). The Role of Promotion in Inducing Specific Human Capital
Acquisition. The Quarterly Journal of Economics, 108(2), 523-534.
Prendergast, C. (1993b). A Theory of "Yes Men". The American Economic Review, 83(4),
757-770.
Prendergast, C. (1999). The Provision of Incentives in Firms. Journal of Economic
Literature, 37(1), 7-63.
Prendergast, C., & Topel, R. H. (1993). Discretion and Bias in Performance Evaluation.
European Economic Review, 37(2-3), 355-365.
Prendergast, C., & Topel, R. H. (1996). Favoritism in Organizations. The Journal of
Political Economy, 104(5), 958-978.
Pucik, V., & Lim, J. C. (2001). Transforming Human Resource Management in a Korean
Chaebol: A Case Study of Samsung. In C. Rowley, T. W. Sohn & J. Bae (Eds.),
Managing Korean Business: Organization, Culture, Human Resources and
Change (pp. 137-160). London, UK: Frank Cass Publishers.
Rickman, N., & Witt, R. (2008). Favouritism and Financial Incentives: A Natural
Experiment. Economica, 75(298), 296-309.
Rosen, S. (1982). Authority, Control, and the Distribution of Earnings. The Bell Journal
of Economics, 311-323.
Sattinger, M. (1993). Assignment Models of the Distribution of Earnings. Journal of
economic literature, 31(2), 831-880.
Simon, H. A. (1979). Rational Decision Making in Business Organizations. The
American Economic Review, 69(4), 493-513.
136
Song, J., Almeida, P., & Wu, G. (2003). Learning-by-Hiring: When Is Mobility More
Likely to Facilitate Interfirm Knowledge Transfer? Management Science, 351-
365.
Steers, R. M., Shin, Y., Ungson, G., & Nam, S. (1990). Korean Corporate Culture: A
Comparative Analysis. Research in Personnel and Human Resources
Management, 247-262.
Sung, Y. (2010, Dec 3, 2010). Broken "Pure Bloodism" for Top Management Team in
Samsung (Translated), Money Today. Retrieved from
http://www.mt.co.kr/view/mtview.php?type=1&no=2010120313420035187&outl
ink=1
Sutter, M., & Kocher, M. G. (2004). Favoritism of Agents–the Case of Referees' Home
Bias. Journal of Economic Psychology, 25(4), 461-469.
Tirole, J. (1986). Hierarchies and Bureaucracies: On the Role of Collusion in
Organizations. Journal of Law, Economics, & Organization, 2(2), 181-214.
Waldman, M. (1984a). Job Assignments, Signalling, and Efficiency. The Rand journal of
economics, 15(2), 255-267.
Waldman, M. (1984b). Worker Allocation, Hierarchies and the Wage Distribution. The
Review of Economic Studies, 51(1), 95-109.
Warner, J. B., Watts, R. L., & Wruck, K. H. (1988). Stock Prices and Top Management
Changes. Journal of financial Economics, 20, 461-492.
Yu, G., & Rowley, C. (2009). The Changing Face of Korean Human Resource
Management. In C. Rowley & Y. Paik (Eds.), The Changing Face of Korean
Management (pp. 29-41). New York, NY: Routledge.
137
Appendix
Executive Titles in Korea—Positions and Roles
80
In Korean companies, a worker usually bears a job title that displays (i) a
hierarchical rank and (ii) a role in an organization.
81
The “dual” career system in Korean
conglomerates may be analogized military officers’ ranks and roles. For example, a major
general (i.e., two-star general) represents a military rank below lieutenant generals and
above brigadier generals, but it does not provide information as to the role that the person
in the rank serves. Major generals in the Army serve as division commanders or as high-
level officers at major commands and the Pentagon,
82
which indicates separation of ranks
and jobs. In a similar manner, a Sajang
83
which literally means “the head of a company”
need not be the highest ranked officer (e.g. CEO) of a company. A Sajang may be a
division head or a CFO, depending on his or her role. Therefore, executives can have a
title as “Sajang / Division Manager of ABC business,” but Sajang represents their
hierarchical position, which is typically next to Hoejang and Bu-hoejang. The subsequent
title provides information about the executive’s role and the organization to which he or
she belongs. Therefore, the title as a whole implies that an individual is one of the top
few executive officers and playing a division manager role in ABC’s business division.
80
Pucik and Lim (2001) provide almost identical descriptions of Korean companies’ job assignment and
title.
81
Pucik and Lim (2001) refer to it as a ‘dual career system.’
82
http://usmilitary.about.com/od/army/a/majgen.htm
83
“ 社長” in Chinese Characters, pronounced as [saja ŋ] in Korean. A Sajang, in most cases, is the head of a
company.
138
Similarly, job titles such as Sajang, Sangmu or Jeonmu do not provide any information
about a person’s role in an organization.
Table 15 describes the hierarchical levels used for executives in Samsung. In a
typical large Korean firm, a Hoejang and a Sajang constitute a top management team,
consisting of about five top executives. Next, executives with Bu-sajang to Jeonmu titles
are considered senior executives while Sangmu and Sangmubo are junior executives.
Table 15: Titles for Hierarchical Ranks for Executives in Samsung
Title for Rank
a
No. of
Executive-
Years
c
Proportion
Accumulated
Proportion
Hoejang ( 회장, 會長)
b
- - -
Bu-hoejang ( 부회장, 副 會長)
7 0.15% 0.15%
Sajang ( 사장, 社長)
86 1.85% 2.00%
Bu-sajang ( 부사장, 副 社長)
348 7.47% 9.47%
Jeonmu ( 전무, 專務)
603 12.95% 22.42%
Sangmu ( 상무, 常務)
1,732 37.19% 59.61%
Sangmubo ( 상무, 常務 補)
1,881 40.39% 100.00%
4,657
a
Provided in a hierarchical order (from the highest)
b
Hoejang is not included in the sample.
c
The number of executive-years at each hierarchical rank
Abstract (if available)
Abstract
This dissertation empirically examines the effects of accounting performance and professional relationships between supervisors and subordinates on personnel decisions, including promotions and dismissals, for executives working in a large conglomerate. In the first part, this study investigates whether accounting performance measured at corporate and reporting segment levels affects decisions to promote or dismiss executives in sub-organizations and how the relationship varies in different contexts where supervisors make the decisions for different types of promotions, for workers with different job responsibilities, and in organizations with different organizational interdependencies. The second part of this study examines the effects of subjective evaluations in promotion decisions. In particular, I compare two competing explanations about what appears to be the outcome of favoritism: Information vs. preference. In a sample of 4,657 executive-years in a Korean conglomerate, the findings indicate that: (1) when supervisors make personnel decisions, they consider organizational performance in an accounting measure, or return on assets (ROA) in a way to make the most of benefits of associating the performance measure with promotion-based incentives
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Enchanted algorithms: the quantification of organizational decision-making
PDF
The balanced scorecard and long-term financial performance: evidence from publicly traded companies
PDF
The interactive effects of incentive threshold and narcissism on managerial decision-making
PDF
Essays on health economics
PDF
Do shared auditors improve audit quality? Evidence from banking relationships
PDF
An examiniation of staff perceptions of a data driven decision making process used in a high performing title one urban elementary school
PDF
Two essays on the mutual fund industry and an application of the optimal risk allocation model in the real estate market
PDF
Empirical essays on alliances and innovation in the biopharmaceutical industry
PDF
Contracting for performance: examining the relationship between LAUSD and ALEKS using transaction cost economics
PDF
Incorporating social preferences into incentive system design
PDF
Sustaining arts programs in public education: a case study examining how leadership and funding decisions support and sustain the visual and performing arts program at a public high school in Cal...
PDF
Clinical decision analysis for effective supervision: a review of the supervisor-supervisee relationship
PDF
Assessing and addressing random and systematic measurement error in performance indicators of institutional effectiveness in the community college
PDF
Forgiveness: elucidating the underlying psychological processes that foster brand forgiveness and interpersonal forgiveness
PDF
Who manages the firm matters: the incremental effect of individual managers on accounting quality
PDF
Relationship between employee disengagement and employee performance among facilities employees in higher education: an evaluation study
PDF
Does it matter what people say about you: the impact of the content of buzz on firm performance
PDF
The effects of curiosity-evoking events on consumption enjoyment
PDF
Real asset liquidity and asset impairments
PDF
The effects of a customer's comparative processing with positive and negative information on product choice
Asset Metadata
Creator
Kim, Jonghwan
(author)
Core Title
The effects of accounting performance and professional relationships on promotion, dismissal, and transfer decisions in a conglomerate
School
Marshall School of Business
Degree
Doctor of Philosophy
Degree Program
Business Administration
Publication Date
06/28/2013
Defense Date
04/23/2013
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
accounting performance,dismissal,OAI-PMH Harvest,personnel decisions,professional relationship,promotion,ROA,subjectivity,transfer
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Merchant, Kenneth A. (
committee chair
), Erkens, David H. (
committee member
), Moon, Hyungsik Roger (
committee member
)
Creator Email
jonghwan.kim@usc.edu,simon.jh.kim@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-281471
Unique identifier
UC11293939
Identifier
etd-KimJonghwa-1720.pdf (filename),usctheses-c3-281471 (legacy record id)
Legacy Identifier
etd-KimJonghwa-1720.pdf
Dmrecord
281471
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Kim, Jonghwan
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
Tags
accounting performance
dismissal
personnel decisions
professional relationship
ROA
subjectivity