YOU CANNOT SUPPRESS OPPORTUNISM,
IF YOU DO NOT KNOW ITS NUANCES
University of Groningen Faculty of Economics and Business Master‘s Thesis MSc. Business Administration:
Organizational and Management Control
Irene Jantje (I.J.) de Vries (S2178680) i.j.de.vries.3@student.rug.nl
Supervisor: dr. H.J. van Elten Co-assessor: S. Mukherjee
Date: 22 June 2015
Word Count (excluding Appendices): 14.399
ABSTRACT
Peculiarly, in the existing literature, the agency problem is mainly associated with the passive form of opportunism (i.e. the moral hazard problem), thereby disregarding the work of Wathne and Heide (2000). Wathne and Heide (2000) conceptualized that opportunism is composed of both active and passive forms of opportunism, systematically influencing the process and performance of the exchange relationship in a different manner. Therefore, the aim of this study is to contribute to the academic field of Management Control, by refining our understanding of the facets within the agency problem. To empirically test the models, secondary data are used, originating from the Doctoral thesis of dr. Van Elten (2012). Moreover, a subsample of one hundred respondents is used to empirically test the hypotheses. The results indicate the validation of Wathne and Heide‘s (2000) central proposition, demonstrating different antecedents as well as different weakening effects for the passive and active forms of opportunism.
TABLE OF CONTENTS
INTRODUCTION ... 1
THEORETICAL BACKGROUND ... 5
LITERATURE REVIEW ... 6
Agency Problem ... 6
Moral hazard ≠ Opportunism ... 7
Composition of Opportunism ... 7
Omission Bias ... 9
Tolerance and Payoff functions ... 9
Degree of Decentralization ... 10
Information Asymmetry ... 11
Relative performance evaluation (RPE) ... 13
METHODOLOGY ... 17 Sample ... 17 Questionnaire... 18 Measurements ... 18 Independent Variables ... 18 Dependent Variables ... 20 Moderator ... 21 Interaction effects ... 22 Control Variables ... 22 Descriptive statistics ... 24 ANALYSES ... 27 Correlations ... 27 Collinearity Statistics ... 29 Additional analyses ... 29
Multivariate Analysis: Ordinary Least Squares Regression ... 30
Analyses with Passive Managerial Opportunism Measure (full model) ... 30
Analyses with Passive Managerial Opportunism Measure (additional analysis) ... 31
Analyses with Passive Managerial Opportunism Measure (additional analysis) ... 31
Analyses with Active Managerial Opportunism Measure (full model) ... 32
Analyses with Active Managerial Opportunism Measure (additional analysis) ... 33
Analyses with Active Managerial Opportunism Measure (additional analysis) ... 33
Overall findings ... 33
Theoretical and managerial implications ... 35
Conclusion ... 36
Limitations, steering for future research ... 36
REFERENCES ... 38
APPENDIX A ... 45
Overview of Questionnaire Items ... 45
Independent variables ... 45
Dependent variables ... 46
Moderator ... 46
Control Variables ... 47
APPENDIX B... 48
Table 4.2.1: OLS Degree of Information Asymmetry: Passive Opportunism ... 48
Table 4.2.2: OLS Degree of Decentralization of Decision Rights: Passive Opportunism ... 49
Table 4.3.1: OLS Degree of Information Asymmetry: Active Opportunism ... 49
1
INTRODUCTION
The directors of such [joint-stock] companies, however, being the managers rather of other people’s money than of their own, it cannot well be expected, that they should watch over it with the same anxious vigilance with which the partners in a private copartnery frequently watch over their own. Like the stewards of a rich man, they are apt to consider attention to small matters as not for their master’s honour, and very easily give themselves a dispensation from having it. Negligence and profusion, therefore, must always prevail, more or less, in the management of the affairs of such a company.
— Adam Smith (1776)
2
Peculiarly, in the existing literature, the agency problem is mainly associated with the moral hazard problem, encompassing only passive opportunism. In the case of passive opportunism, one of the parties in an agreement is ―purposefully withholding effort‖ (Griesinger, 1990; Masten, 1988; Rousseau, 1995) or ―somehow refraining from performing the agreed-on actions‖ (Goetz & Scott, 1981). However, already in the work of Masten (1988), it became clear that the opportunism construct and the moral hazard problem cannot be treated as synonyms. Masten (1988) acknowledged that opportunism not only encompasses moral hazard, but also practices characterized as ‗blatant opportunism‘, e.g. ―lying, stealing, cheating and calculated efforts to mislead, distort, disguise, obfuscate, or otherwise confuse" (Williamson, 1985). Wathne and Heide (2000), were the first to explicitly conceptualize that opportunism is composed of both active and passive forms. Passive opportunism (i.e. moral hazard), occurs when an agent abstains obligations that were previously agreed upon, or refuses to adapt to new circumstances (i.e. opportunism by omission). Active opportunism encompasses behaviour that violates certain explicit or implicit restrictions in the relationship, and engages in forced renegotiation in response to new circumstances (i.e. opportunism by commission, blatant opportunism) (Wathne & Heide, 2000; Seggie, Griffith & Jap, 2013). Wathne and Heide (2000) specify the significance of this composition, by arguing that when the opportunism construct itself is poorly understood, its potential outcomes will remain ambiguous. Moreover if the nuances are unclear, implementing strategies for suppressing this opportunistic behaviour will become problematic (Wathne & Heide, 2000). Today‘s relevance of aforementioned distinction is underscored by Seggie, Griffith and Jap (2013), stating that even though the work of Wathne and Heide (2000) received numerous citations, their central proposition has yet to be empirically tested. The central proposition states that active and passive opportunism may systematically influence the process and performance of the exchange relationship is a different manner. When differences in these different forms of opportunism would be observed, the implications for management could be a source of a new stream of research (Seggie, Griffith & Jap, 2013).
3
mainly proves to be a case of terminology. Secondly, the question arises whether/how the preconditions of the moral hazard problem known as ‗decentralization of decision rights‘ and ‗information asymmetry‘ are really related to the passive form of opportunism, as widely acknowledged in the literature (e.g. Ross, 1973; Jensen & Meckling, 1976; Fama & Jensen, 1983; Eisenhardt, 1985, 1989; Obadia & Vida, 2006 and Navarra & Tortia, 2014), and whether/how they relate to the active form of opportunism. Finally, in answer to the statement of Wathne and Heide (2000), it will be investigated whether a strategy to suppress (or even mitigate) opportunism by a management control mechanism, will have different influences on the room for active or passive managerial opportunism. The control mechanism that will be used relates to an externally determined performance standard, identified as Relative Performance Evaluation (RPE) (Van Elten, 2012). By using a peer group benchmark in RPE, the performance of the agent can be compared with the specific reference group‘s performance, to indirectly examine the fitting performance standard (Van Elten, 2012). Since the control mechanism concerns an externally determined performance standard, the performance of the peer group cannot be influenced by the agent. Consequently, the room for managerial opportunism can be set to boundaries (Van Elten, 2012).
Wathne and Heide‘s statement (2000) encompasses not only theoretical relevance for today. This study poses that when the specific composition of opportunism and the accompanying preconditions become more clear, principals will be more effective in identifying and subsequently suppressing different nuances of opportunism by implementing a specific control mechanism. Consequently, more understanding with regards to Wathne and Heide‘s (2000) proposition also encompasses practical relevance. Moreover, relevance regarding the Management Control literature is present in the aim of the respective literature: keeping the agents‘ actions aligned with the objectives of the principals (Eisenhardt, 1989). Furthermore, by examining the use of RPE in practice, knowledge can be gained regarding the Management Control mechanism (Van Elten, 2012).
Therefore, this study has the following research question to be answered:
‘What is the role of the composition of opportunism in relation to the agency problem?’
Sub-questions:
‘Which antecedents provide room for passive and active managerial opportunism?’
4
The knowledge-generating research process that is used is the theory-testing process. Reason for the use of this process is the fact that theory testing makes use of a business phenomenon as a driver for the process. And albeit the quite elaborate academic literature regarding the agency theory(e.g. Jensen & Meckling, 1976; Fama, 1980; Eisenhardt, 1985, 1988, 1989), the evidence on the theoretical explanations of this business phenomenon is still incomplete, i.e. there exists a gap in the literature (Van Aken, Berends & Van der Bij, 2012). The model is empirically tested with data from prior research (Van Elten, 2012). The respective study used primary data collected from 325 business units. The survey instrument concerned an extensive questionnaire, filled out by business unit managers contacted through the students‘ professional networks at the Nyenrode Business University.
The unit of analysis of this study concerns the Strategic Business Unit (i.e. SBU), and more specifically, refers to the middle-management in the Netherlands (Van Elten, 2012). The relationship between the owner/top management and the middle manager/employee will be investigated, denoted in the agency terminology by ‗the principal‘ and ‗the agent‘ respectively.
5
THEORETICAL BACKGROUND
As shortly outlined in the introduction, the agency problem is a logical implication that derives from the economic exchange relationship, analysed in the agency theory. Agency theory is developed in the information economics literature, and can be seen as an expansion on organizational control literature (Ouchi, 1979; Pennings & Woiceshyn, 1987 and Thompson, 1967), by bringing in ideas with regards to information systems, risk and compensation (Eisenhardt, 1985, 1989). Moreover, Eisenhardt (1985, 1989) posed that the agency theory serves as an expansion on the key features of an organization by recognizing ‗effort-aversion‘, i.e. the discrepancy among the members of the organization with regards to their preferences. The assumption is that people prefer their own actions to maximize their own utility (i.e. Homo-economicus), being often incongruent with the organizational members‘ preferences.
6
LITERATURE REVIEW
Agency Problem
Due to the inability to write and enforce contracts in a costless manner (Fama & Jensen, 1983), Eisenhardt (1989) identified two resulting contracting problems that correlate with the agency problem: ‗moral hazard‘ and ‗adverse selection‘ (Holmstrom, 1982; Eisenhardt, 1989; Wathne & Heide, 2000, Nilakant & Rao, 1994; Obadia & Vida, 2006; Gomez-Mejia & Wiseman, 2007). The moral hazard problem refers to the situation in which the agent may not put forth the agreed-upon effort, i.e. the agent is shirking (Arrow, 1962; Eisenhardt, 1989; Nilakant & Rao, 1994). Holmstrom (1982) acknowledged that when the actions of the agent cannot be contracted for or observed in a direct manner, moral hazard refers to the problem of inducing the agent to ―supply proper amounts of productive inputs‖. In ‘62, aforementioned had already been identified by Arrow, describing moral hazard as a problem that comes into being when the principal cannot observe the agent‘s behaviour. Consequently, the agent will use this un-observability in his/her advantage, by shirking the responsibilities that he/she is held accountable for (Arrow, 1962).
7
Moral hazard ≠ Opportunism
Despite the fact that both contracting problems are incorporated in the opportunism construct, only the first aspect of the agency problem will be covered in this study, i.e. the moral hazard aspect. The underlying reason is that moral hazard is treated as a synonym for the general construct of opportunism, where it actually only constituents the passive form of opportunism (Wathne & Heide, 2000). Aforementioned is acknowledged in the study of Masten (1988), posing that opportunism cannot be treated as a synonym for moral hazard, since opportunism encompasses some ―manifestations that cannot be characterized as moral hazard‖. To get a better understanding of the distinction between the concept of opportunism and moral hazard, let us first take a closer look at the concept of moral hazard. What the manifestations of moral hazard share in common, is that they all involve passive opportunism, in the sense that one of the parties in an agreement is ―purposefully withholding effort‖ (Masten, 1988; Griesinger, 1990; Rousseau, 1995), or ―somehow refraining from performing the agreed-on actions‖ (Goetz & Scott, 1981). Wathne and Heide (2000) acknowledged that the traditional moral hazard problem is passive in nature, as it describes how information asymmetries facilitate the opportunity for one party (i.e. the agent) to ―supply lower levels of quality or output than contracted for‖. Masten (1988) posed that opportunism also encompasses forms which are more active, more inventive, and more likely to trigger responses by other parties in the context of strategy. The agents are characterized by their ability to ―circumvent rules, discover loopholes or otherwise exploiting strategic advantages‖ (Masten, 1988). Williamson (1993) argued for opportunism as a ―much wider set of phenomena‖, not only encompassing the contracting problems, i.e. moral hazard and adverse selection.
Composition of Opportunism
8
incomplete truths‖ (Anderson, 1988), ―neglecting to fulfil obligations‖ (Lee, 1998), and ―failing to provide proper notification‖ (Jap & Anderson, 2003). The opposite scenario is also identified, in which the agent actively violates the contract by providing quality levels or output in excess of what is actually needed and/or what was contractually agreed upon. In other terms, Wathne and Heide (2000) acknowledged the active form of opportunism, where an agent engages in particular behaviour that violates certain explicit or implicit restrictions in the relationship, and engages in forced renegotiation in response to new circumstances (i.e. opportunism by commission or blatant opportunism) (Wathne & Heide, 2000; Seggie, Griffith & Jap, 2013). Following the enumeration in the study of Seggie, Griffith and Jap (2013), opportunism by commission incorporates activities such as ―lying‖ (Lee, 1998), ―breaching (in)formal agreements‖ (Achrol & Gundlach, 1999), ―altering the facts‖ (John, 1984), ―making false accusations‖ (Jap & Anderson, 2003), ―exaggerating difficulties‖ (Anderson, 1988), and ―using unexpected events to extract concessions from partners‖ (Rokkan, Heide, & Wathne 2003). Williamson (1985) conceptualized the practices that are subsumed under aforementioned blatant opportunism as ―lying, stealing, cheating, and calculated efforts to mislead, distort, disguise, obfuscate, or otherwise confuse".
9
Omission Bias
So as a way of exception, Seggie Griffith and Jap (2013) did recognize the relevance previously addressed by Wathne and Heide (2000). In their theorizing for differences in outcomes and specifically in responses, Seggie, Griffith and Jap (2013) relied on two logics. The first logic is the omission bias, a cognitive psychological construct describing the tendency of people to consider. the commission of harmful acts (i.e. active opportunism) worse than equally harmful omissions‖ (i.e. passive opportunism) (Baron, 1986; Sugarman, 1986; Spranca, Minsk & Baron, 1991). Aforestated bias arises out of the moral premise that people ―avoid being the direct cause of harm‖ (Baron, 1986), and therefore envision acts of commission more abnormal than acts of omission (Kahneman & Miller, 1986). Due to the respective bias, the active form of opportunism is perceived as involving greater amounts of effort, indirectly being indicative for greater intentions (Seggie, Griffith & Jap, 2013). Consequently, these conducts are viewed and evaluated more severely than the passive conducts of opportunism, ergo having more severe repercussions. The authors acknowledged that the omission bias encourages the ―under safeguarding‖ of the performances involving acts of omission, instead of putting in place more effective safeguards (Seggie, Griffith & Jap, 2013). Therefore, it is assumed that in line with the human tendency, principal‘s provide more room (i.e. more leeway) for passive (i.e. opportunism by omission) than for active (i.e. opportunism by commission) opportunism. Room for managerial opportunism can be identified as the ―perceptual measure that relates to how respondents perceive the room to act opportunistically under their currently installed control systems‖, as primarily defined by Van Elten (2012).
Tolerance and Payoff functions
10
will lessen, which is put in practice by the more severe repercussions, i.e. a higher response
intensity (Seggie, Griffith & Jap, 2013). The former acknowledgement made the authors pose
that active opportunism will move ―more swiftly‖ along the tolerance function than the passive form of opportunism (i.e. acts of omission). Therefore, parallel with the former assumption, it can be stated that there will be more room (i.e. more leeway) for passive opportunism than for active opportunism. Indicative for aforestated are the results originating from the study of Seggie, Griffith and Jap (2013), suggesting that the ―nature of ongoing exchange provides a breeding ground for the partners to ‗cheat at the margins‘ through passive opportunism.‖
The illustration of the tolerance function derived from the study of Bucklin (1973), and the speed with which both forms of opportunism move along this line, provides a useful framework for an important assumption in this study. In this study, there will be argued for the fact that active and passive opportunism do not differ in the direction in which they systematically influence the process and the performance of an exchange relationship, but that they only differ in the degree in which they influence the process and performance in a systematic manner. This is illustrated by the fact that both forms of opportunism move in the same direction along the tolerance function, however with a different speed. This speed is based on the response intensity, and is indicative for the subsequent room that the respective type of opportunism will have.
Degree of Decentralization
11
authority to agents and the fact that agents are independent actors. This argument is backed by Jensen and Meckling (1976) and Davis, Schoorman and Donaldson (1997) stating that the crux of an agency theory is the fact that principals delegate authority to agents to act on their behalf. This specific delegation allows the agents to opportunistically maximize their own utility at the expense of the principals‘ utility (Davis, Schoorman and Donaldson, 1997). Aforestated is furthermore recognized in the study of Van Elten (2012), stating that when an organization delegates its decision rights, the agent will have more room to behave opportunistically. In line with the study of Van Elten (2012) and the findings in the prior research, it can be deduced that the respective rights assignment provides agents with more leeway to behave in an opportunistic manner. Stated more formally;
H1a: The greater the degree of decentralization of decision rights, the greater the room for
passive managerial opportunism (i.e. moral hazard).
Derived from abovementioned hypothesis, a second hypothesis could be argued for with regards to the active form of opportunism. Despite the fact that this form of opportunism has never been brought into relationship with the delegation of decision rights in isolation, this study poses that active and passive opportunism do not differ in the direction but only in the
degree in which they influence the exchange relationship with its accompanying process and
performance. Therefore, the relationship between delegation of decision rights and the room for active managerial opportunism is argued to be consistent with the first hypothesis and the above stated research. However, due to a higher response intensity, indirectly based on the omission bias, there will be less room for the agent to behave actively in an opportunistic manner, consequently expecting a weaker, but still positive effect, as compared to the passive form of opportunism. Stated more formally;
H1b: The greater the degree of decentralization of decision rights, the greater the room for
active managerial opportunism (i.e. blatant opportunism). Information Asymmetry
12
Balkin, 1992; Rajagopalan & Finkelstein, 1992). The former is backed by Van Elten (2012), who argued that information asymmetry bears upon the information gap between the principal and the agent, incorporating the limited ability to detect opportunism by the respective party. Consequently, the information gap provides a possibility to exploit the agent‘s self-interest (Van Elten, 2012). A resembling argument is found in the work of Eisenhardt (1985), by the illustration of two cases. The first case relates to the situation of observable behaviour, in which both the principal and the agent know exactly what the agent has done. The second case relates to incomplete information, in which the principal cannot determine whether the agent has behaved in a congruent manner (i.e. in the principal‘s best interest). The subsequent dilemma that arises, is one in which the principal rewards the agent based upon the behaviours that they agreed upon, without the real confirmation whether those behaviours are performed appropriately. Without this confirmation, the agent might have the incentive to shirk (i.e. the passive form of opportunism), and therefore cannot be relied upon (Eisenhardt, 1985). The relationship between information asymmetry and opportunistic behaviour is widely acknowledged in the literature (e.g. Bergen, Dutta, & Walker, 1992; Wolff, 1995; Obadia & Vida 2006 and Wathne & Heide, 2000). Specifically, information asymmetry has been brought into relation with the passive form of opportunism (Jensen & Meckling, 1976; Holmstrom, 1979; Gomez-Mejia & Wiseman, 2007; Wathne & Heide, 2000) . Consequently, the following relationship can be hypothesized for:
H2a: The greater the degree of information asymmetry, the greater the room for passive
managerial opportunism (i.e. moral hazard).
Parallel to abovementioned prior research, and as argued for above, it is posed that active and passive opportunism do not differ in the direction, but only in the degree in which they influence the exchange relationship with its accompanying process and performance. Therefore, the relationship between information asymmetry and the room for active managerial opportunism is argued to be consistent with the former hypotheses and research. However, due to a higher response intensity, which is indirectly based on the omission bias, there will be less room for the agent to behave actively in an opportunistic manner. Consequently, a weaker, but still positive effect is expected, as compared to the passive form of opportunism. Stated more formally;
H2b: The greater the degree of information asymmetry, the greater the room for active
13
Relative performance evaluation (RPE)
As earlier mentioned in this study, due to the agent‘s privately held information, principals are not able to distinguish between defection and cooperation with regards to fulfilling the principals‘ goals. Therefore, principals must incur costs of governance, because they simply cannot assume the agent‘s continued cooperation (Gomez-Mejia & Wiseman, 2007). The applicable measurements suitable for functioning as control mechanism, were thoroughly examined in the classic works of Thompson (1967) and Ouchi (1979). As argued for in the study of Ouchi (1979), the essential element underlying any form of control is the question whether it is feasible to measure the performance that is desired, accompanied by a reasonable precision. So not the costs, but the ability to measure the desired performance proves to determine when a control mechanism is considered to be a ―rational application‖(Ouchi, 1979).
Regarding aforementioned feasibility, Van Elten (2012) acknowledged that the obtainment of the required information for the performance evaluation could be costly and/or simply not available to the principal. Therefore, in line with prior research, Van Elten (2012) argued for a different control mechanism, defined as the Relative Performance Evaluation (RPE). The author acknowledged the relevance of the use of a performance standard, functioning as a norm by comparing the employees‘ performances to the standard. By using a peer group benchmark, the performance of the agent could be compared with the performance of a specific reference group, to indirectly examine the fitting performance standard (Van Elten, 2012). The crux of the respective control mechanism is the use of an externally determined performance standard, in which the peer group performance will be informative for the agent‘s performance potential (Van Elten, 2012). Because of the fact that these performance standards are externally determined, they cannot be influenced by the agent (Murphy, 2001). Consequently, the use of RPE could reduce the agent‘s room to behave in an opportunistic manner (Murphy, 2001;Van Elten, 2012). In answer to the statement of Wathne and Heide (2000), it will be investigated whether the strategy to suppress or even mitigate opportunism by the management control mechanism put forward by Van Elten (2012) will have different influences on the room for active or passive managerial opportunism.
14
that are in the best interest of the principal (Bucklin, 1973; Davis, Schoorman & Donaldson, 1997). Agents will be more likely to comply, since their performance will be benchmarked with the performance of a reference group, typically consisting out of individuals who are facing comparable tasks and circumstances (Van Elten, 2012). Therefore, a weakening effect of relative performance evaluation on the positive relationship between the decentralization of decision rights and the room for managers to behave passively in an opportunistic manner is argued for.
With regards to the asymmetric distribution of information (i.e. information asymmetry), it is exactly this information, with its accompanying information systems, that will ―curb‖ agent‘s opportunistic behaviour (Eisenhardt, 1985; 1989). The principal will be aware of the actions taken by the agent, where the latter will acknowledge that it is not possible to deceive the principal. Subsequently, the agent will behave in a more congruent manner (Eisenhardt, 1985; 1989). This is precisely the effect that Relative Performance Evaluation hopes to bring about. However, as recognized in the work of Van Elten (2012), the use of an externally determined performance standard, will not reduce the amount of information asymmetry itself. The author acknowledged that the control mechanism will work around the unit of analysis to assess its agent‘s effort, instead of obtaining the information in a direct manner (Van Elten, 2012). In this manner, the study of Van Elten (2012) poses that the control mechanism is able to reduce the target-setting consequences made possible by the information asymmetries, by determining the performance standard outside the sphere of influence of the agent. In the case of RPE, the performance target is based on the performance of a peer group instead of an internal process. Hence, the standard can be categorized as an externally determined performance standard and is consequently able to mitigate opportunistic behaviour (Murphy, 2001; Van Elten, 2012). Resulting from aforementioned, it can be posed that the management control mechanism (i.e. RPE), will weaken the positive relationships between the privately held information by the agent (i.e. information asymmetry) and the room for both passive as well as active managerial opportunism.
15
severe repercussions, i.e. a higher response intensity. Therefore, it is argued that the management control mechanism, identified as RPE, affects the relationships between decentralization of decision rights and information asymmetry on the room for passive and active managerial opportunism in the same direction, however to a differential degree. Where relative performance evaluation will weaken the positive relationships, the weakening effect will be stronger with respect to the active form of opportunism because of the higher response intensity. In contrast, the ‗under safeguarding‘ of the performance regarding the passive form of opportunism will lead to a weaker weakening effect of RPE. Stated more formally;
H3a: RPE use will negatively moderate the relationship between the degree of decentralization
of decision rights and the room for passive managerial opportunism (i.e. moral hazard).
H3b: RPE use will negatively moderate the relationship between the degree of decentralization
of decision rights and the room for active managerial opportunism (i.e. blatant opportunism).
H4a: RPE use will negatively moderate the relationship between the degree of information
asymmetry and the room for passive managerial opportunism (i.e. moral hazard).
H4b: RPE use will negatively moderate the relationship between the degree of information
asymmetry and the room for active managerial opportunism (i.e. blatant opportunism).
The aforestated hypotheses can be modelled as follows;
17
METHODOLOGY
Sample
To empirically test the models as illustrated in figure 1.1. and 1.2, secondary data are used originating from the Doctoral thesis of dr. Van Elten. The respective thesis used primary data collected from 325 business units. The survey instrument concerned an extensive questionnaire, filled out by business unit managers, which were contacted through the professional networks of students at the Nyenrode Business University. This approach yielded two advantages: respondent identification and the ability to conduct face-to-face interviews. Where the respondent identification brought about a suitable selection of Business Unit managers, through face-to-face interviews a careful responding to and a valid interpretation of the questions could be acquired (Van Elten, 2012). The unit of analysis of this study concerns the Strategic Business Unit (i.e. SBU), and more specifically, refers to the middle-management in the Netherlands (Van Elten, 2012). Within this study, a subsample of one hundred respondents is used to empirically test the hypotheses. The descriptives of the respondents‘ business units and the firm to which each business unit resides, are displayed by their size and industry in table 1.1; 1.2 and 1.3 (rounded to two decimals). As illustrated, the measurement of size is composed out of the number of FTE‘s for both the SBU (M = 209,74,
SD = 326,22) and the company (M = 15.693,98, SD = 34.652,64), as well as the Revenue,
displayed in millions for both the SBU (M = 141,31, SD = 668,88) and the company (M = 10.389,88, SD = 43.671,90) respectively. Additionally, it can be deduced that the Services industry is the most represented industry in the respondents‘ firms and their subsumed strategic business units.
N Minimum Maximum Mean Std.
Deviation
FTE SBU 100 1 2500 209,74 326,22
Revenue SBUa 96 0 6500 141,31 668,88
FTE Company 98 50 150.000 15.693,98 34.652,64
Revenue Companyb 98 0 300.000 10.389,88 43.671,90
18 N Sum Services 100 47 Production of goods 100 20 Not-for-profit 100 19 Financial sector 100 14
Table 1.2: Descriptive Statistics Industry SBU’s
N Suma
Services 100 46
Production of goods 100 27
Not-for-profit 100 19
Financial sector 100 14
a. Sums up till 106, since some business units operate in more than one sector. Multiple answers to the question were treated as valid. Table 1.3: Descriptive Statistics Industry Companies
Questionnaire
The questionnaire originating from Van Elten (2012) encompasses purpose-developed questions (e.g. for the moderator variable RPE-use), but mainly consists out of the modification of previously used and validated questions (e.g. the degree of decentralization of decision rights, the degree of information asymmetry). To ensure the understanding of the questions and the response categories, the questionnaire went through three rounds of pre-tests with twelve managers. The method that was used to increase the reliability of the instrument could be identified as the Hak‘s Three-Step Test Interview method, encompassing the observation of the actual interactions between the respondents and the instrument, where the respondent made his/her though process observable by ‗thinking aloud‘ (Hak, Van der Veer & Jansen, 2008; Van Elten, 2012). NB: The original survey questions, as stated in the Doctoral thesis of Van Elten, are demonstrated in Appendix A.
Measurements
Independent Variables
19
internally consistent (α = .788) and explaining more than half of the variance of the aggregated construct (55,035 %). Moreover, the loadings can be interpreted as significant, since they all demonstrate a higher value than the specific loading of .512, with the sample size being one hundred (Stevens, 2002). Inferring from aforestated, the factors demonstrate to encompass convergent validity (Field, 2013), where a new aggregated construct is calculated by averaging the scores on all items (see table 2.1).
Items for the degree of decentralization
Component loadings
Strategic decisions Investment decisions Marketing decisions Internal process decisions Human Resource decisions
,742 ,718 ,776 ,857 ,592 Total variance explained (in perc.) 55,035 Cronbach‘s Alpha (α) ,788 Table 2.1: Factor analysis Degree of Decentralization of Decision Rights
20
Items for the degree of information asymmetry Component loadings Activities Input/output Potential Technical Int. impact Ext. impact Achieved ,772 ,635 ,792 ,676 ,690 ,646 ,824 Total variance explained (in perc.) 52,231 Cronbach‘s Alpha (α) ,844 Table 2.2: Factor Analysis Degree of Information Asymmetry
Dependent Variables
Room for passive managerial opportunism (Q14bc); In contrast to the original use of this measurement in the study of Van Elten (2012), the factors with a passive indication are measured separately from the factors with an active indication. As can be seen in the table, the factors with a passive indication identified as ‗hiding bad performance‘ and ‗taking it easy‘, illustrate to be positively related with a significance at the 99% confidence interval (r = .655, ρ <.01), and internally consistent (α = .790). Additionally, the component loadings portray to explain a large part of the total variance of the aggregated construct (82,742 %), indicating a sound measurement for calculating a new aggregated construct by averaging the scores on the items (see table 2.3).
Items for the room for passive man. opportunism
Component loadings
Hiding bad performance Taking it easy
,910 ,910 Total variance explained (in perc.) 82,742 Cronbach‘s Alpha (α) ,790 Correlation (two-tailed) r = .655, ρ <.01
21
Room for active managerial opportunism (Q14de); Subsequently, with regards to the active form of managerial opportunism, the fourth and fifth questions regarding ‗window dressing of business unit‘s figures‘ and ‗implementing pet projects‘ respectively, demonstrate their significant positive relatedness (r = .607, ρ <.01) and internal consistency (α = .755). Additionally, the component analysis yields one component, explaining 80,337 percent of the variance in the aggregated construct as presented in table 2.4. Consequently, by calculating the mean of the former stated items, a new aggregated construct can be developed.
Items for the room for active man. opportunism
Component loadings
Window dressing SBU‘s figures Implementing pet projects
,896 ,896 Total variance explained (in perc.) 80,337 Cronbach‘s Alpha (α) ,755 Correlation (two-tailed) r = .607, ρ <.01
Table 2.4: Factor Analysis Active Managerial Opportunism
Moderator
22
Items for the Use of Relative Performance Evaluation
Component loadings
Peer performance point of ref. PE Substantially better PE
Substantially worse PE
,884 ,952
,922 Total variance explained (in perc.) 84,616
Cronbach‘s Alpha (α) ,909 Table 2.5: Factor Analysis Relative Performance Evaluation
Interaction effects
In order to analyse indirect effects (i.e. a moderation) together with the linear effects in the regression, new variables must be created by forming products of the independent and moderator variables (i.e. predictor variables), also defined as the interaction terms or interaction effects (Moosbrugger, Schermelleh-Engel, Kelava & Klein, in press; Field, 2013). The interaction effects are operationalized by z-standardizing both the independent variables (i.e. the degree of decentralization of decision rights; the degree of information asymmetry), and the moderator (i.e. use of RPE), subsequently multiplying these independent variables with the moderator. The underlying logic for z-standardizing the independent- and moderator variables could be attributed to solving for inequalities of inter-scale-variances to prevent undesired weightings (Van Elten, 2012). Despite the absence of intra-scale-variances, inter-scale-variances are present, since the independent- and moderator variables are measured on seven-point and five-point summated likert scales, respectively.
Control Variables
23
of the FTE and revenues for both units of analysis (i.e. company and SBU), the internal consistency proves to be significant lower than the threshold level (.232) to average the scores and develop two aggregated constructs (i.e. Company size (FTE and Revenue); Business Unit size (FTE and Revenue)). Therefore, this study controls for these measurements as single-item constructs, i.e. Business Unit size (FTE Ln); Business Unit size (Revenue Ln); Company size (FTE Ln) and Company size (Revenue Ln).
Comparability of SBU’s; Human Asset specificity (Q33, 34Ln, 35); Based on the study of Van Elten (2012), the importance of the comparability of the strategic business units is also acknowledged. Van Elten (2012) recognized that, for the control mechanism RPE use to be effective, the strategic business unit has to be comparable to a relevant peer group. Kruis (2008) acknowledged aforementioned, stating that the presence of asset specificity will create problems to control for opportunistic behaviour, since simply stated, there will be no market alternative available to function as a disciplining device, benchmarking the SBU‘s performance.
24
Items for the human asset specificity
Component loadings
Difficulty SBU-employee to learn ins and outs
Average weeks of extra training for new employee (Ln)
Uniqueness skills and knowledge SBU employees
,768
,784
,752
Total variance explained (in perc.) 59,031 Cronbach‘s Alpha (α) ,647 Table 2.6: Factor Analysis Human Asset Specificity
Descriptive statistics
The descriptives of aforementioned variables are illustrated in table 3.1. As acknowledged by deCarlo (1997), an assessment of any distributional assumptions is considered part of the complete statistical analysis of the study. The importance lies within the fact that departures from univariate normality are able to affect tests and confidence intervals based on normal theory methods (deCarlo, 1997). The derived statistics indicate that the original variables‘ data lies within the tolerable levels of univariate normality, since skewness and kurtosis only indicate non-normality issues when they are greater than 3.00 and 10.00 respectively (Kline, 1998; Widener, 2007).
However, as demonstrated by table 3.1, the data of the interaction variables lies not within the tolerable range with respect to the kurtosis (Kline, 1998; Widener, 2007). The interaction term regarding the degree of decentralization of decision rights illustrates a kurtosis of 11,877, where the interaction term encompassing the degree of information asymmetry displays a kurtotic value of 18,135. Moreover, the values of skew and kurtosis are indifferent to a change in the use of z-standardized values as to mean-centred values, calculated by the use of the grand mean (Field, 2013).
Descriptive statistics
25 opportunism Degree of decentralization of decision rights 1,00 7,00 4,1897 1,18703 -,228 -,011 Degree of information asymmetry 2,29 7,00 5,1693 ,79382 -,350 1,176 Use of RPE:PE ,00 5,00 3,2869 1,12802 -1,196 1,228
Dec. of decision rights * Use of RPE:PE
-4,44 7,04 ,0728 1,24162 1,485 11,877
Information asymmetry * Use of RPE:PE
-6,72 2,99 ,0155 1,04462 -2,377 18,135
Business unit size FTE (Ln)
,69 7,82 4,6933 1,12932 ,033 1,090
Business unit size Rev. (Ln)
,00 22,60 15,5816 5,52021 -2,313 4,105
Company size FTE (Ln)
3,93 11,92 7,6510 2,08469 ,360 -,573
Company size Rev. (Ln)
,00 26,43 18,4729 5,97637 -2,312 5,097
Human asset specificity 1,03 4,55 2,7697 ,67802 -,282 ,215
Table 3.1: Descriptive Statistics
In response to aforementioned values, both the Kolmogorov-Smirnov- and the Shapiro-Wilk test are conducted to see whether the distribution of the data deviates from a comparable normal distribution (Field, 2013). Both the Kolmogorov-Smirnov- and the Shapiro-Wilk test demonstrate respectively that the interaction term regarding the decentralization of decision rights proves to be significantly non-normal (D(97) = .188, ρ < .001, W(97) = .787, ρ < .001). Additionally, the interaction term encompassing the degree of information asymmetry also signifies a significant departure from normality (D(97) = .197, ρ < .001, W(97) = .754, ρ < .001), as demonstrated in table 3.2.
Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk
Statistic df Sig. Statistic df Sig.
Dec. of decision rights* Use of RPE:PE
,188 97 ,000 ,787 97 ,000
Information asymmetry * Use of RPE:PE
,197 97 ,000 ,754 97 ,000
a. Lilliefors Significance Correction
26
27
ANALYSES
Subsequent to the methodological section, this section presents the conducted analyses. An Ordinary Least Squares (OLS) regression analysis is selected, to fit a regression model to the gathered data (Field, 2013). More specifically, by the use of the OLS regression analysis, parameters are estimated that define the model for which the sum of the squared errors is the minimum it can be, given the gathered data (Field, 2013). Moreover, as discussed below, to distinguish the independent contribution to explained variance of a predictor variable (Farrar & Glauber, 1967), additional analyses are conducted in which the two independent variables (i.e. degree of decentralization of decision rights, degree of information asymmetry) and accompanying interaction effects are individually analysed. All analyses are conducted with aforementioned measurements of the variables, based on their hypothesized relationships. Besides the aim of testing the hypothesized relationships, the main effects are included to serve as a prevention for conclusions of significant coefficients of an interaction effect, when these results are in point of fact attributable to the lower-order effects, i.e. the main effects (Hartmann & Moers, 1999). The aforestated authors acknowledged that because of the fact that the interaction term is factually the product of two main effect, these constituent parts must be included to ‗partial out‘ these lower-order effects. Along with the inclusion of the main- and interaction effects, the single-item control variables are also accounted for.
In general, the Ordinary Least Squares (OLS) regression analyses showed theory-consistent results in the directional relationships. However, not every hypothesis demonstrated to be significant in the full model analysis, where the sub-analyses signified more significant results by distinguishing the independent contributions to the explained variance. Notwithstanding, some insignificant results yielded valuable insights regarding the different outcomes for the different forms of opportunism, as argued for (Wathne & Heide, 2000).
Correlations
28
signifies a theory-consistent significant association between the independent- and the dependent variables. The association between the degree of information asymmetry and the room for the active form of managerial opportunism shows to be bivariately correlated (r = .275). Additionally, two of the independent variables illustrate a significant bivariate association (i.e. the degree of decentralization of decision rights; the degree of information asymmetry), indicating potential concerns for multicollinearity since they are jointly included in the conceptual models (r = .421). As discussed below, based on the Collinearity Diagnostics, the values indeed signify an indication of multicollinearity (Field, 2013). Moreover, both forms of managerial opportunism are significantly correlated with the control variable identified as Human Asset Specificity. The Pearson correlation coefficients indicate a negative relatedness, implying theory inconsistency (see table 4.1).
29
*
Correlationis significant at the 0.10 level (two-tailed)
**
Correlation is significant at the 0.05 level (two-tailed)
***
Correlation is significant at the 0.01 level (two-tailed) Table 4.1: Correlation table
Collinearity Statistics
Prior to the analysis of the OLS regression for both dependent variables, the importance of the Collinearity Statistics is recognized. The aforestated high kurtotic values already indicated non-normality issues, with high frequencies mainly centred on the mean (deCarlo, 1997). These mean-centred values are subsequently indicative for little variance in the model, where it becomes problematic to determine which predictor variable affects the outcome and to what extent (i.e. Multicollinearity) (Field, 2013). The Collinearity Statistics identified as the Variance Inflation Factor (VIF) and Tolerance, indicate no causes for concern or indications for a bias regarding the regressions, since they signify values which are not greater than ten and less than 0.1 respectively (Myers, 1990; Menard, 1995). However, regarding the Collinearity Diagnostics, the Eigenvalues prove to be dissimilar, indicating the likelihood of the model to change by small changes in the measured coefficients (Field, 2013). Moreover, the Eigenvalues differ substantially between, on the one hand, the control variables and on the other hand the independent variables and the interaction effects. Additionally, there might be an indication of collinearity in the Condition Indices, since the values for the independent variables and the subsequent interaction terms signify a substantial difference in their size with regards to the values of the control variables (Field, 2013). Where the independent variables and the interaction effects are mainly above ten (the interaction effects are even higher than twenty), the control variables vary between one and three. Aforementioned emphasizes the potential for the inability to distinguish the independent contribution to the explained variance of a predictor variable, exhibiting little independent variation (Farrar & Glauber, 1967; Field, 2013). Hence, this detected potential serves as a motivation to measure for the independent variation of the predictor variables (i.e. independent variable and its accompanying interaction term), in addition to the full model OLS regression analyses. Additional analyses
30
in the regression analysis. By the use of the former method, the independent contribution to the explained variance of the predictor variable is measured (Farrar & Glauber, 1967; Field, 2013). Moreover, the gathered results do not only illustrate the standalone results, but signify the (in)significant enrichment or deterioration of the new model in comparison to the original full model. In these separate conducted regressions, the limitations are acknowledged regarding the exclusion of variables as stated in the work of Bowerman and O‘Connell (1990) and Field (2013). NB: The tables concerning the additional OLS regression analyses are demonstrated in Appendix B (i.e. table 4.2.1; 4.2.2; 4.3.1 and 4.3.2).
Multivariate Analysis: Ordinary Least Squares Regression
Analyses with Passive Managerial Opportunism Measure (full model)
As illustrated in table 4.2, the OLS regression analysis shows the full model being significant and fits the data. The model summary signifies an R2 of .196, indicating a significant explained variability by the predictor variables of almost twenty percent, accompanied by a significant ANOVA F-statistic (F = 1.955, df = 10.80, ρ < .05). The information asymmetry variable demonstrates a significant effect on the passive form of managerial opportunism (t (90) = 1.762, ρ < .10), thereby confirming hypothesis H2a. Concerning the control variables,
both the company size (Revenue Ln) and the human asset specificity prove to be significant, representing negative b-coefficients where the latter‘s direction is theory-inconsistent and counterintuitive.
Ordinary Least Squares (OLS) regression Model 1
Room for Passive Managerial Opportunism B SE T-stat. Prob.
Independent variables
Degree of decentralization of decision rights (H1a)
Degree of information asymmetry (H2a)
Use of RPE: PE
Interaction effects
Degree of decentralization of decision rights * Use of RPE:PE (H3a)
Degree of information asymmetry * Use of RPE:PE (H4a)
Control variables
Business unit size FTE (Ln) Business unit size Rev. (Ln)
31 Company size FTE (Ln)
Company size Rev. (Ln) Human asset specificity
,073 -,091 -,718 ,101 ,049 ,240 ,726 -1,860 -2,994 ,470 ,066* ,004*** R square ,196 Δ R square ,196** F-Statistic 1,955 Prob. F-stat. ,049** * ρ < .10 ** ρ < .05 *** ρ < .01
Table 4.2: OLS Full model
Analyses with Passive Managerial Opportunism Measure (additional analysis) The model summary indicates an insignificant decrease in the total variance explained (Δ R2 = -.007, ρ >.10), when the predictor variable encompassing decentralization of decision rights is excluded (table 4.2.1). Deriving from aforestated, it is argued that the excluded independent variable and its product with the moderator (i.e. the interaction term) do not account for a significant variance in the model. Aforementioned inference is strengthened by the ANOVA‘s F-statistic; signifying that the independent variable and the interaction term regarding information asymmetry do account for a significant effect in the model (F = 2.395, df = 8.82, ρ < .05). When controlled for the predictor variable‘s contributions individually, the hypothesized main effect proves to be significant (t (90) = 2,048, ρ < .05), implying confirmation of H2a. Moreover, the OLS regression output demonstrates the confirmation of
the weakening moderating effect as argued for in Hypothesis H4a (t (90) = -1,880, ρ < .10).
32
Analyses with Active Managerial Opportunism Measure (full model)
As illustrated in table 4.3, the OLS estimation shows the full model being significant and fits the data. The model summary signifies an R2 of .255, accompanied with a significant ANOVA F-statistic (F = 2.745, df = 10.80, ρ < .01). Regarding the model parameters, the degree of information asymmetry demonstrates a significant effect on the active form of managerial opportunism (t (90) = 2.919, ρ < .01), thereby confirming hypothesis H2b. Concerning the
control variables, both the company size (Revenue Ln) and the human asset specificity prove to be significant, representing negative b-coefficients where the latter‘s direction is theory-inconsistent and counterintuitive. Moreover, the business unit size (FTE Ln) proves to be significant at the 95% confidence interval, signalling a positive relationship (b = .294).
Ordinary Least Squares (OLS) regression Model 1
Room for Active Managerial Opportunism B SE T-stat. Prob.
Independent variables
Degree of decentralization of decision rights (H1b)
Degree of information asymmetry (H2b)
Use of RPE: PE
Interaction effects
Degree of decentralization of decision rights * Use of RPE:PE (H3b)
Degree of information asymmetry * Use of RPE:PE (H4b)
Control variables
Business unit size FTE (Ln) Business unit size Rev. (Ln) Company size FTE (Ln) Company size Rev. (Ln) Human asset specificity
,070 ,507 ,227 -,172 -,054 ,294 ,006 ,084 -,098 -,619 ,166 ,174 ,161 ,139 ,165 ,146 ,037 ,099 ,048 ,234 ,421 2,919 1,414 -1,242 -,329 2,008 ,173 ,850 -2,060 -2,643 ,675 ,005*** ,161 ,218 ,743 ,048** ,863 ,398 ,043** ,010*** R square ,255 Δ R square ,255*** F-Statistic 2,745 Prob. F-stat. ,006*** * ρ < .10 ** ρ < .05 *** ρ < .01
33
Analyses with Active Managerial Opportunism Measure (additional analysis)
As demonstrated in table 4.3.1, with regards to the predictor variables encompassing the presence of private information, the model summary indicates an insignificant decrease in the total variance explained (Δ R2 = -.017, ρ >.10), where the remaining components demonstrate to be significant to the model (F = 3.214, df = 8.82, ρ < .01). Inferring from aforestated, it is posed that the excluded predictor variable and accompanied interaction term do not account for a significant variance in the model. Regarding the model parameters, the information asymmetry variable accounts for a significant main effect (t (90) = 3.462, ρ <.01) as hypothesized for in H2b.
Analyses with Active Managerial Opportunism Measure (additional analysis)
In addition to controlling for the presence of private information, an OLS regression analysis is conducted with regards to the individual contribution of the predictor variable encompassing decentralized decision making. Despite the insignificance of the interaction term concerning information asymmetry in the latter analysis (ρ >.10), a significant decrease can be found in the total variance explained when the predictor variable is excluded (Δ R2 = -.086, ρ <.05). In response to aforementioned, the results indicate that by only excluding the interaction effect, the decrease in total variance of .1 percent proves to be insignificant (Δ R2 = -.001, ρ >.10). This is strengthened by the significance signified in the ANOVA‘s F statistic (F = 2.096, df = 8.82, ρ < .05). Therefore, with regards to the active form of opportunism, it becomes clear that the main effect with information asymmetry as independent variable has a significant stake in the model. However, the interaction term with regards to the information asymmetry proves to be insignificant. The model parameters indicate a significant effect of the degree of decentralization of decision rights on the active form of managerial opportunism (t (90) = 1.928, ρ <.10), thereby confirming H1b. Moreover, the moderation effect proves to be
marginally significant just below the cut-off rate of 90% (t (90) = -1,633, ρ > .10), with the b-coefficient illustrating the negative relationship as theorized for in Hypothesis H3b (b = -.206).
Overall findings
34
The OLS regression output indicates the confirmation of the former statement where both antecedents (i.e. main effects), as theorized for in H1b and H2b, demonstrate to be significant
for the blatant form of opportunism. Notwithstanding, where H2b demonstrates significance
both in the full model as well as in the additional analysis, H1b is solely significant in the
additional analysis. In contrast, albeit the confirmation of H2a regarding the main effect of
information asymmetry on the passive form of opportunism, Hypotheses H1a encompassing
the delegation of decision rights, proves to be of insignificant meaning for the moral hazard problem.
Regarding the interaction terms, both moderation effects encompassing the degree of decentralization of decision rights (H3a; H3b), demonstrate to be marginally significant just
below the cut-off rate of the 90% confidence interval (ρ > .10). Contrary to the former output, a difference can be identified with regards to the moderation effects encompassing private information (H4a; H4b). Where a significant T-statistic is demonstrated for the passive form of
opportunism after conducting an additional analysis (H4a), the moderation effect does not
signify a significant output regarding the active form of managerial opportunism.
35
DISCUSSION AND CONCLUSION
Theoretical and managerial implications
A possible explanation for the insignificant effect of the degree of decentralization of decision rights on the passive form of opportunism, can be found in the rights assignments, associated with an organization‘s internal ‗regulatory or common law traditions‘ (Jensen & Meckling, 1995). The firm puts regulatory constraints on delegated behaviour by circumscribing in advance the opportunity-set from which a decision-maker can choose (Jensen & Meckling, 1995). Consequently, these regulations will suppress inferior courses of action (Jensen & Meckling, 1995). This argumentation is strengthened by Tommasi and Weinschelbaum (2007), acknowledging this prevention by stipulating the level of effort that has to be put forward by the agent, also known as ‗contractable effort‘.
A plausible explanation for the moderation effects‘ significance balancing on the edge when concerning the degree of decentralization, is attributable to a limitation of the control mechanism itself (i.e. RPE-use). Dekker, Groot and Schoute (2012), recognized problems regarding the determination of the specific performance target and an exact aspiration level for subordinate managers. The authors posed that both issues are subject to internal and external firm contingencies and significantly vary across firms. It is argued that due to these determinations, agents have some leeway to provide less (i.e. passive opportunism) or more (i.e. active opportunism) output than was contracted for (Wathne & Heide, 2000).
36
room for active managerial opportunism, with information asymmetry being the antecedent, can steer future research.
Concerning the control variables, a possible (part) of the explanation for the significant negative directional effect of human asset specificity, can be found in the positive side-effect of an exchange relationship where the principal is dependent on the agent (Chiou & Droge, 2006). Chiou and Droge (2006) defined aforestated control variable as a ―Specific Asset Investment (SAI)‖, and subsequently found that these SAIs have direct effects on agent‘s attitudinal- and behavioural loyalty. Additionally, the authors recognized that when the principal makes an idiosyncratic investment for the agent, the agent is unlikely to engage in opportunistic behaviour, i.e. creating a mutual exchange relationship (Chiou & Droge, 2006).
Conclusion
In answer to the statement of Wathne and Heide (2000) and the subsequent recognition of its importance by Seggie, Griffith and Jap (2013), this study had the aim to contribute to the academic field of Management Control by refining our understanding of the facets in the agency problem. Aforestated is operationalized by clarifying the relevance of the composition of opportunism by making three specific contributions. Firstly, by conducting factor analyses, it became clear that the factors significantly loaded onto their accompanying aggregated constructs, indicating a confirmation of the relevance for composition. Secondly, the relevance proved even stronger in the subsequent OLS regression analyses, where different main effects were found regarding the antecedents for the different forms of opportunism, subsequently providing valuable theoretical- and managerial implications. Lastly, in answer to the statement of Wathne and Heide (2000), the strategies to suppress or even mitigate opportunism demonstrated different moderation effects by the use of the OLS regression analyses, functioning as the final validation of Wathne and Heide‘s (2000) proposition, i.e. the relevance of the composition of the opportunism construct.
Limitations, steering for future research
37
regression analyses (Montgomery & Askin, 1981). Notwithstanding, the study solely indicated a potential for collinearity issues, since the Collinearity Statistics (i.e. the Variance Inflation Factor (VIF) and Tolerance) proved not to indicate multicollinearity concerns. When the problems would be more severe, future research would be wise to use a Robust estimation method as an alternative to the Least Squares regression (Montgomery & Askin, 1981).
Moreover, the conducted additional analyses are not free from limitations. As acknowledged by, among others, Field (2013) and Bowerman and O‘Connell (1990), there are no statistical grounds to omit one variable over another. Moreover, Field (2013) acknowledged that the resulting theoretical conclusions are meaningless, since, statistically speaking, any of the variables could be removed. Recognizing aforestated limitations, it was decided to measure both predictor variables individually, instead of choosing the predictor variable that indicated more significance in the full model. Moreover, the results originating from the additional analysis were mainly used for interpretation and insights, instead of their use as hard, factual evidence. Future research could follow the advice from Bowerman and O‘Connell (1990), posing that in situations of multicollinearity, a removed predictor must be replaced by an equally important predictor not having such strong collinearity problems.
38
REFERENCES
Achrol, R. S., & Gundlach, G. T. (1999). Legal and Social Safeguards Against Opportunism in Exchange. Journal of Retailing,15 (1): 107-121.
Akerlof, G. A. (1970). The market for 'Lemons': Quality uncertainty and the market for mechanism. Quarterly Journal of Economics, 84: 488-500.
Anderson, E. (1988). Transaction Costs as Determinants of Opportunism in Integrated and Independent Sales Forces. Journal of Economic Behavior and Organization, 9 (3): 247-64. Anderson, E., & Schmittlein, D. C. (1984). Integration of the sales force: an empirical examination. RAND Journal of Economics, 15: 385–395.
Arrow, K. J. (1962). Economic welfare and the allocation of resources for invention. In The
rate and direction of inventive activity. Princeton, NJ: Princeton University Press.
Baron, R.M. (1986). Trade-offs Among Reasons for Action. Journal for the Theory of Social
Behavior, 16 (2): 173-95.
Bergen, M., Dutta, S., & Walker, O. C. Jr. (1992). Agency Relationships in Marketing: A Review of the Implications and Applications of Agency and Related Theories, Journal of
Marketing, 56: 1–24.
Bowerman, B, L., & O‘ Connell, R. T. (1990). Linear statistical models: An applied approach (2nd ed.). Belmont, CA: Duxbury.
Bucklin, L. (1973). A Theory of Channel Control. Journal of Marketing, 37: 39-47.
Chiou, J. S., & Droge, C. (2006). Service quality, trust, specific asset investment, and expertise: Direct and indirect effects in a satisfaction-loyalty framework. Journal of the
Academy of Marketing Science, 34 (4): 613-627.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the
behavioral sciences. Hillsdale: Erlbaum.
Cordes, C., Richerson, P., McElreath, R., & Strimling, P. (2011). How does opportunistic behavior influence firm size? An evolutionary approach to organizational behavior. Journal of
39
Cortina, J. M. (1993). What is coefficient alpha? An examination of theory and applications.
Journal of applied psychology, 78: 98-104.
Dahlstrom, R., & Nygaard, A. (1999). An Empirical Investigation of Ex Post Transaction Costs in Franchised Distribution Channels. Journal of Marketing Research, 36:160-70.
Davis. J. H., Schoorman. F. D. & Donaldson, L. (1997). Toward a Stewardship Theory of Management. The Academy of Management Review, 22 (1): 20-47.
DeCarlo, L. T. (1997). On the Meaning and Use of Kurtosis. Psychological methods, 2 (3): 292-307.
Dekker, H. C., Groot, T. L. C. M., & Schoute, M. (2012). Determining performance targets.
Behavioral Research in Accounting, forthcoming.
Demsetz, H. (1988). The theory of the firm revisited. Journal of Law, Economics, &
Organization, 4 (1): 141-161.
Dunk, A. S. (1993). The effect of budget emphasis and information asymmetry on the relation between budgetary participation and slack. The Accounting Review, 68: 400-10.
Eisenhardt, K. M. (1985). Control: Organizational and Economic Approaches. Management
Science, 31 (2): 134-149.
― (1988). Agency- and Institutional-Theory Explanations: The Case of Retail Sales Compensation. The Academy of Management Journal, 31 (3): 488-511.
― (1989). Agency theory; An assessment and review. The Academy of Management Review, 14 (1): 57-74.
Fama, E. (1980) Agency problems and the theory of the firm. Journal of Political Economy, 88: 288-307.
Fama, E. F., & Jensen, M. C. (1983). Separation of ownership and control. Journal of law and
economics, 301-325.
