1
Personality Traits, Supply Chain
Complexity and Supply Chain
Performance
--for behavioral causes of the bullwhip effect research
Master thesis, MscBA specialization Operations & Supply Chains University of Groningen, Faculty of Economics and Business
July, 2010
LI CHEN
Student number: 1801767 Email: chenlizjmier@gmail.com
2
PREFACE
Firstly of all, I want to thank my supervisor Dirk Pieter van Donk, associate professor of University of Groningen, for providing me this internal project. His useful advice helped me a lot to conduct a good research in a scientific way. Furthermore, his constant encouragement gave me more courage to challenge myself frequently during this research.
Great thanks for my second supervisor Gwenny Ruël, associate professor of University of Groningen. Her previous research provided valuable insight into my master thesis. Additionally, her great supports in data processing really helped me to make reasonable analyses for my paper.
I also want to thank Dr. Xiaoming Liu, Zernike Institute for Advanced Materials, University of Groningen. His rich mathematical knowledge helped me to understand the equation of the entropy H, which makes application of information-theoretic measure possible in my paper.
Last but not least, I would like to express my gratitude for my parents who supported me during my study in Netherlands. Special thanks for my friends who ever cooked, chatted or played with me to let me have a colorful and happy life here.
Li Chen
3
ABSTRACT
We examine, in an experiment with the Beer Distribution Game, a complementary behavioral cause of the bullwhip effect which has been overlooked in the previous research: personality traits. Specially, the effects of supply chain complexity, which is measured by using information-theoretic approach, on the relations between personality traits and performance, are also fully investigated in this research. The results confirm that differences in personality traits – risk taking, ambiguity, self-efficacy and locus of control – have different impacts upon supply chain performance (backorder costs and/or inventory costs), but the way and/or magnitude of those relations are affected by supply chain complexity. The positive effect of risk taking on backorder performance was not only partially mediated by complexity, also by moderated complexity. Ambiguity and locus of control influenced backorder performance through the moderating effect of complexity. Self-efficacy influenced inventory performance through the moderating effect of complexity. According to these first evidences, we propose a conceptual model to describe all relations among personality traits, supply chain complexity and performance. This model provides insights on the challenges for reducing the bullwhip effects from personality traits perspective, and also suggests that the ability to manage complexity can be significant to improve performance in a supply chain.
Keywords: beer game, bullwhip effect, human behavior, personality traits, supply chain
4
Table of Content
PREFACE ... 2
ABSTRACT ... 3
1. Introduction ... 6
2. Theoretical background and proposed hypotheses ... 8
2.1 Personality characteristics and its relationship with supply chain performance ... 8
2.1.1 Risk taking and supply chain performance ... 9
2.1.2 Ambiguity and supply chain performance ... 10
2.1.3 Self-efficacy and supply chain performance ... 11
2.1.4 Locus of control and supply chain performance... 12
2.2 Supply chain complexity and its relationship with supply chain performance ... 13
2.2.1 Complexity ... 13
2.2.2 Supply chain complexity ... 14
2.2.3 Supply chain complexity and its performance ... 15
2.3 Personality traits, supply chain complexity and supply chain performance ... 16
3. Methodology ... 18
3.1 Experiment design and implementation ... 18
3.2 Supply chain complexity ... 20
3.2.1 Characteristics of supply chain complexity in the beer game ... 20
3.2.2 Approach to measure supply chain complexity ... 20
3.2.3 Information-theoretic measure ... 20
3.2.3.1Definition of entropy H from mathematics perspective ... 21
3.2.3.2 Application of information-theoretic measure to quantify supply chain complexity in the beer game ... 22
3.2.3.3 Example of calculation of supply chain complexity in the beer game ... 23
3.3 Data analysis ... 24
4. Results ... 25
4.1 Preliminary reliability analysis ... 25
4.2 The bullwhip effect ... 25
5
4.4 Correlations and descriptive statistics... 27
4.5 Testing the hypotheses ... 27
4.6 Results for exploratory analysis ... 31
4.6.1 Taking complexity as mediator (testing the mediating model) ... 31
4.6.2 Taking complexity as moderator (testing the moderating model) ... 33
5. Discussion ... 38
5.1 Findings ... 38
5.1.1 Personality traits and its relation with supply chain performance ... 39
5.1.2 Supply chain complexity and its relation with supply chain performance ... 41
5.1.3 Exploratory findings between personality traits, supply chain complexity and supply chain performance ... 41
5.1.3.1 Risk taking, supply chain complexity and supply chain performance ... 41
5.1.3.2 Ambiguity, supply chain complexity and supply chain performance ... 42
5.1.3.3 Self-efficacy, supply chain complexity and supply chain performance ... 43
5.1.3.4 Locus of control, supply chain complexity and supply chain performance ... 43
5.2 A conceptual model for the empirical findings ... 44
5.3 Conclusion ... 45
5.4 Managerial implications ... 46
5.5 Limitations and future research ... 47
REFERENCE ... 49
6
1. Introduction
A supply chain is an inherently complex system which includes multiple entities (i.e., suppliers, manufacturers, distributors, and retailers) involving in activities of moving goods and adding value from the beginning supplier to the end customer. Due to several types of uncertainties (e.g. demand uncertainty, production and supply uncertainty), and inherent characteristics (e.g. indirect or delayed feedbacks; time delays), it is difficult to manage this system efficiently. One well-known inefficient outcome of a supply chain is the much studied bullwhip effect.
The bullwhip effect, first noted by Forrester (1958), describes the variation of demand increased at each level of a supply chain from retailer to manufacturer and the further away an entity from the end customer, the larger is this variation (Lee et al., 1997a). This effect is a frequent and costly problem in supply chains, causing excessive capacity investment, unnecessary inventory, lower capacity utilization and poor customer service (Armony et al., 2005; Gonçalves, 2003; Lee et al., 1997a; Sterman, 2000). For instance, Hewlett-Packard lost millions of dollars in unnecessary capacity and excessive inventory caused by order amplification and fluctuations from resellers to the company's integrated circuit division in comparison to customer demands (Lee et al., 1997b).
Most of the previous research on the bullwhip effect has focused on understanding its causes and countermeasures. Two categories of explanations have been offered. Lee et al. (1997a, 1997b) identified four key operational causes of the effect: (1) demand signal processing, (2) rationing and shortage gaming, (3) order batching, and (4) price fluctuations caused by promotions. These causes have been recorded in practice, and techniques or strategies to alleviate them are considered as an important part of the tool kit for supply chain design (Simchi-Levi et al., 2003).
7 yet received). Moreover, many later studies further demonstrate that supply line underweighting is highly robust - it does not vary significantly by supply chain role (Croson et al., 2004; Wu et al., 2006). In general, these studies have confirmed that, after removing most of the operational causes, behavioral causes are still influencing the decisions. However, supply line underweighting can not clearly explain why decision-makers treat backorder and inventory differently. Also, this identified behavioral explanation only emphasizes human players' bounded rationality or cognitive limitations, less attention is paid to other directions of human behavior (such as motivation, communication methods). More recently, a new direction is touched on the personal attributes. Hung and Ryu (2008) stated that changing risk preferences of managers concerning demand changes and supplier failures is a significant behavioral factor in explaining the bullwhip effect. As a whole, human behavior needs comprehensive research in future.
8 effect research.
The remainder of this paper is structured as follows. The next section presents a theoretical background of the four personality variables and the concept of complexity in a supply chain context, followed by statements of hypotheses. Then, detailed experimental design and methodology are discussed. Next, the results of this analysis are reported. Section five begins with a discussion upon the results, which is followed by a proposed conceptual model. After that, a conclusion is given. The paper concludes with managerial implications, limitation and future research.
2. Theoretical background and proposed hypotheses
2.1 Personality characteristics and its relationship with supply chain performance
Personality refers to a person‘s dynamic and organized set of characteristics that uniquely influences his or her cognition, incentive and behavior in various situations (Ryckman, 2004). In daily life, personalities of individuals to some extent indicate what we can expect of them (Carver and Scheier, 2000). For example, we are more willing to ask a help from a person whom we believe to be warm-hearted than indifferent. Followed by this line, we propose to investigate those traits which are considered to be more relevant to expected behaviors or abilities in managing a supply chain.
9 "panic" strategy. Therefore, we would investigate to what extent the trait of risk taking of a player is related to both decision making strategy and his or her performance in this research.
Overall, relations of each personality trait, as noted above - risk taking, self-efficacy, ambiguity and locus of control, with supply chain performance are examined and formulated into hypotheses in the following sections.
2.1.1 Risk taking and supply chain performance
According to decision theory, risk is taken by the decision-makers when they presume results of future events deviating from their expected value. Early research (e.g. Pratt, 1964; Arrow, 1965) defined risk averse as one's preference for an outcome with certainty over a gamble of the same expected value. For example, a risk adverse investor might save his or her money with a low but guaranteed interest rate, rather than buy stocks with high returns, but also had a chance of losing all money. Risk seeking is the opposite behavior to risk averse. Risk neutral behavior is between risk aversion and risk seeking. According to different attitudes toward risk, people can be classified into three categories: low risk taking, middle risk taking and high risk taking. From traditional decision theory perspective, decision makers with low risk taking prefer relatively low risks and are inclined to sacrifice some expected returns in order to decrease the variation in possible outcomes, while high risk taking decision makers prefer relatively high risks and are inclined to sacrifice some expected returns in order to increase the variation (March and Shapira, 1987).
10 beer game. According to above-mentioned behavior, low risk taking decision makers will try to reduce order deviation so as to decrease the variation in the cumulated cost. Therefore, it is reasonable that low risk taking decision makers will have a better performance.
However, it is uncertain whether differences in risk taking can account for performance differences. Some research stated that risk taking is considered as a stable property of people, which is related to aspects of personality development or culture (e.g. Douglas and Wildavsky, 1982 ), while other research indicated that decision maker's risk preference is path dependent (e.g. Hung and Ryu, 2008). That means his or her risk preference may change with past experience or training. So it is possible that participants may have more experience in the demand pattern and supplier capacity after several rounds and change their risk preference to reduce order fluctuations, as suggested by the findings of previous research (Hung and Ryu, 2008). In order to reduce this uncertainty, the following two hypotheses are intended to examine this relationship again in this paper.
H1a: Risk taking is negatively related to backorder/ service performance H1b: Risk taking is negatively related to inventory performance
2.1.2 Ambiguity and supply chain performance
Tolerance of ambiguity is defined as the way an individual distinguishes and processes information about ambiguous situations when confronted by an array of unfamiliar, complex or incongruent clues (Furnham and Ribchester, 1995). According to the findings of Furnham and Ribchester (1995), individuals with low tolerance of ambiguity (or intolerance of ambiguity) feel stress, response prematurely, and avoid ambiguous situations, while individuals with high tolerance for ambiguity perceive ambiguous situations as desirable, challenging, and interesting and neither deny nor distort their complexity of incongruity. Also, some articles discovered that persons with low tolerance of ambiguity (or intolerance of ambiguity) are predicted to report higher levels of perceived uncertainty in decision situations that provide very limited (or low quality) information than those with high tolerance of ambiguity (e.g. Gifford et al., 1979).
11 performance. For the existing literature of ambiguity, more attentions are paid on the relationship between role ambiguity and job performance from both cognitive and motivational perspectives, but less studied from a personality trait view. Only few works indirectly support such link and suggest that tolerance of ambiguity is important to performance in a number of business domains from decision making (Dollinger et al., 1995) and decision confidence (Ghosh and Ray, 1997), to entrepreneurial behavior (Schere, 1982), effectiveness in negotiations (Ghosh, 2007), as well as the capability to deal with change (Judge et al., 1999), and complexity (Gupta and Fogarty, 1993).
Definitely, the link with performance remains unclear. In order to clarify this link, a test will be conducted in our study with the beer game setting. According to above findings, it is expected that in the beer game which is qualified as ambiguous (Sterman, 1994), participants with high tolerance of ambiguity are prone to feel less stress and more actively cope with complex situations than those with low tolerance of ambiguity. Thus, we proposed that participants with high tolerance of ambiguity have better performance than those with low tolerance of ambiguity, as states formally the following two hypotheses.
H2a: Tolerance of ambiguity is positively related to backorder/service performance H2b: Tolerance of ambiguity is positively related to inventory performance
2.1.3 Self-efficacy and supply chain performance
Self-efficacy is defined by Bandura (1997, p3) as ―belief in one‘s capabilities to organize and execute the course of action required to produce given attainments‖. According to the research of Bandura (1997), people avoid activities that they believe exceed their coping capability, but they readily undertake and perform assuredly those that they judge themselves capable of handling. This finding is also supported by other previous research (e.g. Brown and Inouye, 1978; Phillips and Gully, 1997). For example, when confronting with difficulties, people who have serious doubts about their capabilities tend to make less effort or give up; people who are high self-efficacy tend to persist in and make more effort to conquer challenges.
12 (1998) stated that self-efficacy is positively and strongly related to work- related performance, and this relation is moderated by task complexity. Several years later, however, Chen et al. (2001) extended findings by Stajkovic and Luthans (1998) by stating that self-efficacy does not relate to complex task performance. The discrepancy is explained by Chen et al. (2001) that they conducted performance research in field settings, while Stajkovic and Luthans (1998) did in lab settings. Moreover, the results of Chen et al. (2001) revealed that self-efficacy is related to performance on simple task.
In addition, for two decades of empirical research which demonstrated that self-efficacy has a positive relationship with performance in an organizational setting (e.g. Bandura, 1988; Wood and Bandura, 1989), their performance measures included adaptability to advanced technology, managerial performance etc., but did not take supply chain performance respects into account.
Overall, there is confusion about the relationship between self-efficacy and performance. Some researchers highlight the situation in which there is a relation, while other researches do not consider. In this paper, it is thought that people with high self-efficacy can achieve good performance as they are prone to be more perseverant and it is well known that high perseverance usually produces high performance achievements. If that is the case, then self-efficacy will influence supply chain performance, as the following two hypotheses suggests.
H3a: Self-efficacy is positively related to backorder/service performance H3b: Self-efficacy is positively related to inventory performance
2.1.4 Locus of control and supply chain performance
13 Previous research suggests that internals and externals have different capacities for effective information processing. Internals are more inclined to seek out and acquire relevant information than externals (Lefcourt, 1983), and seem to learn more from feedback and past experiences than externals (Phares, 1976). Additional, people who have an internal locus of control in the face of an unfamiliar situation are likely to involve in extensive trial-and-error behavior (Boone et al., 1996).
According to these findings, we expect that internals perform better than externals for two reasons. First, individuals, who believe in their ability to control the environment, are inclined to be more persistent in solving problems, even confronted with adversity. Second, the different ability in information processing between internals and externals may have significant impact on their performance, as suggested by the statement of Spector (1982) that internals are better at collecting and processing information than externals, and would be better at performing complex tasks. Moreover, the findings of Booone et al. (1996) demonstrated that firms headed by internal CEOs perform better than firms led by external CEOs. This leads to next two hypotheses.
H4a: Internal locus of control is positively related to backorder/service
performance
H4b: Internal locus of control is positively related to inventory performance
2.2 Supply chain complexity and its relationship with supply chain performance
2.2.1 Complexity
14 and the emergent behaviors that can not be predicted from the individual system components (Arteta and Giachetti, 2004).
2.2.2 Supply chain complexity
Supply chain complexity should relate to the number of organizations, the number of relationships among these organizations and the types of relationships between the organizations, etc., according to above these characteristics of complexity. Several previous works have shed light on supply chain complexity. Wilding (1998) defined the interaction between deterministic chaos, parallel interactions and demand amplification as a ‗supply chain complexity triangle‘ so as to identify the generation of uncertainty in a supply chain. Calinescu et al. (2001) divided complexity in manufacturing intro three respects: decision-making complexity, structural complexity and behavioral complexity to analyze the manufacturing complexity. Milgate (2001) developed a conceptual model of supply chain complexity, composing of three dimensions: uncertainty (upstream and downstream), technological intricacy (product and process) and organizational systems (internal and external). Frizelle (1998) defined supply chain complexity as the variety and uncertainty associated with internal or external causes by the information and/or material flows along the supply chain that are expected, unexpected, predicted or unpredicted. Many studies have been done on the base of the Frizelle's (1998) definition over the last ten years and made contributions in the practical field. For instance, by studying the trends in transmitting complexity both inside and at the interface between companies, activities and actions, which hinder decision makers in making optimal decisions, or potential obstacles to integration, could be identified (e.g. Sivadasan et al., 1999; Sivadasan et al., 2002). In this research, we empirically examine supply chain complexity based on the Frizelle's (1998) definition.
Blackhurst et al. (2004) identified four respects that cause supply chain complexity: Large network of interlinked participants (information, material and financial flows)
including suppliers, distributors and manufacturers across multiple organizations. Each of these participants may be a member of a large number of other supply
15 Dynamic and uncertain nature of the supply chain.
Each participant may have differing objectives.
It can be seen that the reasons causing complexity can originate from inside or from outside of a supply chain system. Thus, supply chain complexity is classified into three main categories: internal, upstream and downstream (external) complexity (Bozarth et al., 2009). Each organization in a supply chain can have its own internal and external complexity level.
Internal complexity is observed with the uncertainty and variety of material and information flows within an organization of a supply chain. The causes for increased internal complexity are: the types of processes, the number of products, the stability of schedules, management techniques etc. (e.g. Bozarth et al., 2009; Isik, 2009). Upstream complexity is associated with the uncertainty and variety of material and
information flows exported from upstream supply base. Potential drivers of upstream complexity include the number of suppliers, the delivery lead time and reliability of suppliers etc. (e.g. Bozarth et al., 2009; Isik, 2009).
Downstream complexity is defined as the uncertainty and variety of material and information flows from downstream organizations. Examples for increased complexity might be the increased number of customers, the variability of demand and so on (e.g. Bozarth et al., 2009; Isik, 2009).
2.2.3 Supply chain complexity and its performance
16 delivery performance.
Although, above research found a relation between supply chain complexity and performance, it is questionable what can be concluded from the results. Hoole (2006) and Vickers and Kodarin (2006) focused on complexity measured by the number of organizations (e.g. suppliers, customers and distributors), manufacturing points and products/services. Perona and Miragliotta (2004) focused on complexity measured by average of supply relationship duration, the procurement policies and the number of products. It is difficult to compare these two different sets of variables used by these researchers.
According to the possible reasons causing complexity, it is thought complexity is negatively related to supply chain performance. For example, if suppliers are variable and unpredictable, the manufacturer can accommodate it by keeping stock while bearing additional storage costs, which in turn results in poor inventory performance. As explained in the section 2.2.2, this unreliable delivery, which can be described by the difference between purchasing orders and actual dispatches, would lead to an increase in external (or upstream) complexity of the manufacturer. Overall, this relation will be tested again in our study but with different variables. This results in the following two hypotheses:
H5a: Supply chain complexity has a negative relation with backorder/service
performance
H5b: Supply chain complexity has a negative relation with inventory performance
2.3 Personality traits, supply chain complexity and supply chain performance
17 Two different mechanisms through which personality traits influence supply chain performance are found according to the previous discussion of each personality trait. The impact of risk taking on performance is through influencing ordering decision making process. High risk takers are expected to have different ordering policies in comparison to low risk takes. However, the effects of the other three personality traits on performance are more triggered by personal emotions, which characterize the problem solving behavior. For instance, persons with high tolerance of ambiguity will feel pleased, challenged and interested when facing with ambiguous situations (Furnham and Ribchester,1995); individuals with lower self-efficacy might perceive the task as more difficult by feeling less confidence in their ability to perform the task (e.g. Brown and Inouye,1978); people who have an external locus of control tend to be overwhelmed by the complex task because high task complexity may incur more feeling of helplessness for them (e.g. Perrewe and Mizerski, 1987). We argue that the difference in mechanism might lead to different model paths among personality traits, complexity and performance.
There are three sources of supply chain complexity in the beer game. Internal complexity is determined by the individual's ordering decision making process, which is described with information flow and the way the individual actually transforms downstream (customer) orders into upstream (supplier) orders. Upstream complexity is addressed by delivery reliability and downstream complexity is described by demand variability. External complexity (upstream complexity and downstream complexity) will lead to the increase in the number of information flows, which in turn creates problems in planning, scheduling and control to affect internal complexity.
18
3. Methodology
3.1 Experiment design and implementation
Our experiment follows the standard protocol of the Beer Distribution Game. The game represents a serial supply chain with exogenous final customer demand, consisting of four echelons: retailer, wholesaler, distributor and manufacturer. Each participant makes decisions - based on his or her current inventory level, backorder position, the amount which will be delivered next period - about the number of items to order from the upstream supplier for replenishment so as to meet demands of downstream customer over multiple of periods.
Within each period, every participant executes the following series of events: 1) an incoming shipment from an immediate upstream facility and as a result, an increase in inventory; 2) an incoming order from an immediate downstream facility, which is either filled with available inventory or placed in backlog. Except for retailers, which receive orders outside the system; 3) a new order placed to the upstream facility. Except for manufacturers, this is a production schedule. No orders will be ignored, and all orders must eventually be met. The sequence of events implies the existence of time delays in the supply chain: shipment delays (two periods) or production delays (only for the manufacturer, a lag of three periods and infinite production capacity assumed) and order processing delays (two periods). Donohue and Croson (2002) and Sterman (1989a) provide further details of the game.
19 the system. After the completion of the game, statistics and graphs documenting the performance of each team were immediately available. More information about an internet implementation of the game is provided by Jacobs (2000).
We first tested a first version of questionnaire on risk-taking and other personality traits among PhD students in our university. Participants responded on a 5-point Likert-type scale which ranged from 1 (fully disagree) to 5 (fully agree) to indicate their level of agreement with statements such as "If I don't succeed immediately in doing something, I give up easily " and " Buying stocks attracts me‖. After analyzing the data, we revised the questionnaire by reducing items from 42 to 32 (Appendix A). Then about 120 first year undergraduate students filled in the reduced questionnaire and finally 56 of them took part in an internet version of the game that we discussed above. These participants were divided into 14 homogeneous groups based on the amount of risk taking behavior: 5 low risk taking groups, 3 middle risk taking groups and 6 high risk taking groups.
20
3.2 Supply chain complexity
3.2.1 Characteristics of supply chain complexity in the beer game
Supply chain complexity is characterized with the number of interlinked organizations, the number of products etc. as discussed in the section of 2.2.2. The system in the beer game is relatively simple for several reasons. First, there are only four organizations dealing with one kind of product in the system. Next, the possible sources of complexity include both inside and outside of a supply chain, but reduce to few. For example, external supply chain complexity might only depend on the demand from downstream and delivery from upstream. Internal supply chain complexity might only relate to the ordering decision process of a decision maker. Also retailers have no complexity from the downstream, while manufacturers have no complexity from the upstream.
3.2.2 Approach to measure supply chain complexity
Generally, two approaches are applied in the existing literature to measure supply chain complexity (Wu et al., 2007). Some researchers used the metrics approach to measure separate respects of supply chain complexity, as we introduced above (e.g. Milgate, 2001; Wilding, 1998). Another method is from an information-theoretic perspective, which is widely used in many industrial case studies and simulations (e.g. Deshmukh et al., 1998; Sivadasan et al., 2002; Sivadasan et al., 2006). The study of Calinescu et al. (1998) shows that the information-theoretic approach is a valuable formal tool for assessing the complexity of simple systems by gathering objective data to provide more insightful information on the system. Considering the simple system displayed by the beer game, therefore, we apply the information-theoretic approach to measure supply chain complexity in this research.
3.2.3 Information-theoretic measure
21 engineering, provides a means of quantifying complexity. From an information-theoretic perspective, entropy is a key measure of uncertainty and variety associated with the state of the system and can be expressed by the amount of information required to describe the state of the system (Shannon, 1948). In addition, information-theoretic measure can quantify complexity of a system in one aspect, which can be used to compare different aspects of system.
3.2.3.1 Definition of entropy H from mathematics perspective
The entropy H of a system S is used to describe the complexity of the system by using the equation below:
i n i i p p H 2 , 1 log
(1)Where i 1,n is the number of state in the system and pi is the probability of i
th
state. In the equation, we use the logarithm base 2, since one bit is the required information to make a binary choice. If the system consists of a number of independent flows, then the entropy H can be measured by the summation of all entropy Hi associated with each independent flow.
Detailed explanations of the measure from mathematics perspective are described next:
1) H reaches maximum in the case of an equi-probable system, in which all pi are the same value as (1/n). That means each state is equally unpredictable, and this is the most uncertain situation for a given number of states. Its value increases with the increasing number of possible states (n).
2) H reaches its minimum in the case of absolute certainty, in which one of the states (i=1… n) gives pi 1, and all the others give pi 0. With only one absolute
certainty, observations of the system provide the least information by the fact that H is equal to 0.
22 (observations) in information and material flows.
3.2.3.2 Application of information-theoretic measure to quantify supply chain complexity in the beer game
According to above discussion, first we have to identify specific measurement variables in order to use the information-theoretic measure in our study. The detailed measurement variables and requirements considered in this system are shown as follows:
1) Material flow variation monitored. As shown in Fig. 1, this system has information
and material flows, but the material flow variation is addressed to quantify the supply chain complexity. According to rules of the beer game discussed in the section of 3.1, manufacturers have infinite capacity to produce products, which means there is no difference between ―Actual production‖ and ―Scheduled production". Also, orders placed to the upstream can not be changed. Thus, there is no information flow variation across organizations. In the present assumption, material flow variation only exits between "Shipment" and "Demand".
2) Quantity variations measured across each material flow variation. We use quantity
variation in material flow to address the supply chain complexity. In the beer game, there is quantity variations between "Shipment "and "Demand" and since lead time is set (two periods for shipment delays, except for manufacturer, it is three periods.), we don‘t take into account the time variations between "Shipment" and "Demand".
3) The states of material flow variation in quantity. A system shows many states, the
way how to define each state is described below: we defined one of states as zero to address that there is no quantity variation between "Shipment‖ and ―Demand". Other states were defined based on the frequency distribution of all the data (40 periods for "Shipment - Demand" of all players). This set of states remained same for all four organizations so as that consistency and comparability of flows existed between different organizations and make further analysis across organizations available.
23 accurate the estimate of the state of the system, and the more precise the calculated value of supply chain complexity.
Figure 1. Measured variations in quantity across the supply chain
3.2.3.3 Example of calculation of supply chain complexity in the beer game
A small example is given so as to further learn about how to use information-theoretic measure in the beer game. Table 1 displays a greatly simplified sample of data collection to be used for measurements for each participant in the beer game (based on a record of a participant as retailer of group 1.1 with 8 periods). This simple example in Table 1 records quantity variation values for each period. These example measurements will be used in the following analysis of supply chain complexity.
24 The general procedure for data analysis is described next (with reference to Table 2): 1) A set of states is defined and sorted according to grouped variations in quantity, as
shown in the second column.
2) The value of variation that falls within each state is identified from Table 1, as recorded in the frequency columns. The probability of each state is calculated based on the frequency.
3) Complexity for each player is calculated using the equation (1).
Table 2: Example of complexity calculation with states grouped
3.3 Data analysis
We use Statistical Package for Social Sciences (SPSS) to analyze data. The first four hypotheses regarding the relationship between certain personality with different levels and performance were examined by the means of analysis of variance (ANOVA). We use the Pearson correlation coefficient to test the relationship between complexity and performance. The moderating effect of complexity were further analyzed by a hierarchical multiple regression analysis. According to the extensive discussion made by Fiske et al. (1982), an ANOVA provides a limited test of a meditational hypothesis. Thus, to test the mediating effect of supply chain complexity, we use a procedure described by Baron and Kenny (1986) to examine the following three regression models:
1) Regression analysis of the mediator (complexity) on the independent variable (e.g. risk taking).
25 independent variable.
3) Regression analysis of the dependent variable on both independent variable and on the mediator.
According to the statement of Baron and Kenny (1986), if all these regression models are displayed in the predicted direction, then the effect of the independent variable on the dependent variable must be less in the third model than in the second one. There is a perfect mediation if the independent variable has no effect on the dependent variable when the mediator is included (Baron and Kenny, 1986). A partial mediation exists if the relation between the independent variable and dependent variable is significantly smaller when the mediator is added but still be greater than zero (Frazier and Tix, 2004).
4. Results
4.1 Preliminary reliability analysis
We first conducted reliability analyses concerning the scale‘s internal consistency of each personality trait. An indicator of internal consistency is Cronbach‘s alpha coefficient, which should be in general above .70 (Pallant, 2005). ―Risk taking‖, which is our main experimental variable, had a Cronbach alpha coefficient of .82. The value for tolerance of ―ambiguity‖ is also good (α=.75). However, the internal consistency for "self-efficacy" and "locus of control" were somewhat disappointing, with a Cronbach alpha coefficient reported of .52 and .54 respectively.
4.2 The bullwhip effect
26
Figure 2. Bullwhip effect: average variance in demand experienced by each role with three levels of risk taking
4.3 Performance for different positions
The indicators of supply chain performance in the beer game are backorder costs and inventory costs. Figure 3 displays the average performance of players with different levels of risk taking in all positions. People with low level of risk taking have average higher backorder costs and lower average inventory costs than people with middle or high level of risk taking. And players in wholesaler and distributor have higher average backorder costs than players in retailer and manufacturer.
0 500 1000 1500 2000 2500 3000 3500 Average Backorder costs Average inventory costs Average Backorder costs Average inventory costs Average Backorder costs Average inventory costs Average Backorder costs Average inventory costs Retailer Wholesaler Distributor Manufacturer
Low Middle High
27
4.4 Correlations and descriptive statistics
Table 3 displays the means, standard deviations and Pearson correlations of all variables we discussed in this study. Negative and significant correlations were found between risk taking and backorder costs (r=-.38, p<.05) and between ambiguity and backorder costs(r=-.29, p<.01). There were positive and significant correlations generated between self-efficacy and inventory costs (r=.34, p<.05). No correlation was found between locus of control and supply chain performance. Complexity had strong correlations with both backorder costs (r=.59, p<.05) and inventory costs (r=-.46, p<.05).
Table 3: Means, standard deviations and Pearson correlations
*. Correlation is significant at the 0.05 level (1-tailed). **. Correlation is significant at the 0.01 level (1-tailed). For complexity, backorder costs and inventory costs, N=48; For risk taking, ambiguity and self-efficacy, N=47;
For internal lotus of control, N=46.
4.5 Testing the hypotheses
28 measured by backorder costs and inventory costs. For each personality trait, participants were divided into low, middle and high level.
Table 4 shows the results for the dependent variable supply chain performance: backorder costs and inventory costs.
H1a: Risk taking is negatively related to backorder/service performance is tested. There was a statistically significant difference at the p<.05 level in backorder costs for the three levels of risk taking (F=5.32, p=.01, see table 4). Low risk taking players have more backorder costs than middle and high risk taking players. Also the actual difference in mean backorder costs between levels of risk taking is big. Hypothesis 1a does not hold, however, it shows that risk taking is positively related to backorder performance (the higher level of risk taking, the lower backorder costs). H1b: Risk taking is negatively
related to inventory performance was examined. No relation was found for the relation
between risk taking and inventory performance (F=.84, p=.44, see table 4). This means that hypothesis 1b is not confirmed.
Table 4: Results of ANOVA with risk taking as independent variable and backorder costs, inventory costs as dependent variable, and associated means
It was predicted in H2a that tolerance of ambiguity is positively related to backorder/service performance. In Table 5, a significant effect was found between
29
has a positive relation with inventory performance. No significant result was found for
the relation between ambiguity and inventory costs (F=.15, p=.87). This means that hypothesis 2b is not confirmed.
Table 5: Results of ANOVA with ambiguity as independent variable and backorder costs, inventory costs as dependent variable, and associated means
H3a: efficacy is positively related to backorder/service performance. H3b: Self-efficacy is positively related to inventory performance. As showed in table 6, no
significant result was found (F=.72, p=.49 and F=.92, p=.41 respectively). People with low self-efficacy have higher backorder costs, but lower inventory costs than people with high self-efficacy. Theses results point out that H3a and H3b do not hold.
30
It was hypothesized that internal locus of control has positive relation with backorder/service performance in H4a and internal locus of control has positive relation with inventory performance in H4b. Table 7 presents that people who have a low internal
locus of control tent to have lower backorder costs (mean "low"=1195.25) and higher inventory costs (mean "low"=2012.25) than people who have higher internal locus of control (mean "middle"=1296.44; mean "high"=1694.38 for backorder costs and mean "middle"=1607.00; mean "high"=1006.71 for inventory costs). The actual difference in mean backorder costs or mean inventory costs between the levels was quite clear, however, no significant results were found between internal locus of control and backorder costs or inventory costs (F=.55, p=.58; F=2.08, p=.14 respectively). This means that both of H4a and H4b are not supported.
Table 7: Results of ANOVA with locus of control as independent variable and backorder costs, inventory costs as dependent variable, and associated means
H5a: Supply chain complexity is negatively related with backorder/service performance and H5b: Supply chain complexity is negatively related with inventory performance, were tested in table 3. The relation between complexity and backorder costs
31 Overall, of the 10 hypotheses, two hypotheses were supported (H2a and H5a), six were not confirmed (H1b, H2b, H3a, b, H4a and H4b), and our results were contrary to two of the hypotheses (H1a and H5b). Our findings of the hypothesized relations can be concluded as follows:
Two hypotheses concerning backorder performance were confirmed. Tolerance of ambiguity was positively related to backorder performance, while supply chain complexity was negatively related to backorder performance.
Six of our hypotheses did not hold. Risk taking did not negatively relate to inventory performance, and also tolerance of ambiguity did not have positive relation with inventory performance. All the hypotheses concerning self-efficacy and locus of control were not supported. Self-efficacy did not have positive effects on backorder or inventory performance. Similarly, internal locus of control did not have positive effects on backorder or inventory performance.
Two relations occurred which were in contrast with our predictions, in that risk taking had a positive relation with backorder performance, and supply chain complexity had a positive relation with inventory performance.
4.6 Results for exploratory analysis
4.6.1 Taking complexity as mediator (testing the mediating model)
To test the mediating effect of complexity, we followed the procedure recommended by Baron and Kenny (1986) building three regression models. As we mentioned in the section 3.3, three conditions all must hold in the predicted direction.
32 Table 7: Regression analyses to test whether complexity mediates between risk taking
and supply chain performance
We found a significant mediating effect of complexity on the relation between risk taking and backorder performance in Table 7. In the first model, complexity - the expected mediator- was regressed on risk taking (β=-.32, p<.05). In the second model, backorder costs was regressed on risk taking (β=-.42, p<.005). In the third model, backorder costs were regressed on both risk taking and complexity. The third model shows that when complexity is added (with a significant effect of β=.50; p<.0005), the regression coefficient of risk taking drops its significance (β=-.26; p=.045). All conditions for a partial mediation were met because risk taking (the independent variable) affected complexity (the mediator) as well as backorder costs (the dependent variable), because complexity affected backorder costs, and because the effect of risk taking upon backorder costs reduced a lot and only had a very weak significant effect when complexity was included (Baron and Kenny, 1986). In other words, risk taking affected backorder costs, partly and indirectly via its negative impact on complexity.
33 second model (β=.10; n.s.). It can not meet the condition, which we mentioned in the section 3.3, that is the independent variable must be shown to affect the dependent variable. This means complexity can not serve as a mediator between risk taking and inventory costs.
In conclusion, only the partial mediating effect of complexity on the relation between risk taking and backorder performance was found in this study.
4.6.2 Taking complexity as moderator (testing the moderating model)
To test the moderating effect of complexity, we proposed interaction terms between risk taking, ambiguity, self-efficacy, lotus of control and complexity. The independent variables and the moderator (complexity) were standardized before constructing the interaction terms (the independent variable time the moderator) so as to reduce the possibility of multicollinearity (Aiken and West, 1991). Only the significant effects of interactions will be presented in figure.
34 The moderating effect of complexity on the relation between risk taking and performance was tested and significant interaction was found on the relation between risk taking and backorder performance (model 1 in table 8: β=-.28, p<.05). The plot revealed that risk taking had a strong and positive relation with backorder performance when complexity was high and a weak and negative relation when complexity was low (See figure 4). No significant effect was found when the moderation effect of complexity was tested on the relation between risk taking and inventory performance (model 2 in table 8: β=.26, n.s.). 0 500 1000 1500 2000 2500 3000 Low High
Risk Tak ing
B ac k or d er C os ts Complexity high Complexity low
Figure 4. Interaction effects of risk taking and complexity on backorder costs
35 Table 9: Results of moderated hierarchical regression analyses among ambiguity,
complexity and performance
0 500 1000 1500 2000 2500 3000 Low High Tolerance of Ambiguity Ba c k o r d e r c o sts Complexity high Complexity low
Figure 5. Interaction effects of tolerance of ambiguity and complexity on backorder costs
36 the relation between self-efficacy and inventory costs (see figure 6). Moreover, the negative effect between self-efficacy and inventory costs is stronger in a low complex situation. No significant interaction effect was found when complexity serves as a moderator upon the relation between self-efficacy and backorder costs (model 5: β=-.05, n.s.).
Table 10: Results of moderated hierarchical regression analyses among self-efficacy, complexity and performance
0 500 1000 1500 2000 2500 3000 3500 Low High Self-efficacy In v e n to r y C o sts Complexity high Complexity low
37 Table 11 presents the result of testing the moderating effect of complexity on the relation between internal locus of control and performance. There was a significant interaction between internal locus of control and backorder costs (model 7: β=.37, p<.01). It is important to indicate that, as described in figure 7, the interaction effect of internal locus of control and backorder costs was very large. Internals were associated with more increase in the backorder costs in a situation with higher lever of complexity, but with more decrease in the backorder costs in a situation with lower lever of complexity. And there was a slightly stronger positive effect. No significant effect was found for the moderating effect of complexity on the relation between internal locus of control and inventory performance (model 8: β=.14, n.s.).
38 -1000 -500 0 500 1000 1500 2000 2500 3000 3500 Low High
Internal Locus of Control
B ac k or d er C os ts Complexity high Complexity low
Figure 7. Interaction effects of internal locus of control and complexity on backorder costs
In sum, our results provide supports for the moderating role of complexity in some respects. Risk taking, ambiguity and locus of control did interact with complexity to predict backorder performance, while self-efficacy did interact with complexity to predict inventory performance.
5. Discussion
According to existing literature it is expected that both personality traits (e.g. risk taking, ambiguity, self-efficacy and locus of control) and supply chain complexity would have impact on supply chain performance and that complexity might mediate or moderate the relations between personality traits and supply chain performance. First, key findings are discussed and summarized by a proposed conceptual model. Next, a conclusion is given. After the conclusion, managerial implications are described. This chapter concludes with research limitations and directions for future research.
5.1 Findings
39 hypotheses were formulated to analyze the relationship between specific personality traits and supply chain performance. The other two hypotheses were defined to test the effect of supply chain complexity on performance. Moreover, we conducted the exploratory analysis about the mediating or moderating effect of complexity on the relationship between personality traits and performance. The findings regarding each relationship will be discussed next.
5.1.1 Personality traits and its relation with supply chain performance
In this research, it is not confirmed that risk taking has a negative relation with supply chain performance (backorder or inventory performance). Conversely, the
positive relation between risk taking and backorder costs was found in our data. A possible explanation could be the under-reacting behavior of people when facing with outstanding orders. People with low risk taking were too careful about their orders and responded slowly in changing their order level to the increase in demand after some periods. Also, they ignored the supply line and ordered required quantity again and again, which was higher than necessary inventory levels. As a result, they have a high amount of both stockouts and stocks.
Next, the average performance in different positions also suggests this point (see the section of 4.3). For low risk taking, the wholesalers and the distributors have more backorder costs than the retailers and manufacturers, the underlying reason might be that wholesalers and distributors also experience the side effect of their upstream whose orders lag behind demand.
It is expected that tolerance of ambiguity is positively related to both backorder performance and inventory performance in this research. The positive relation was
40 stress, more challenged and interested to deal with stockouts than inventories, which resulted in a low amount of backorders. Also, this finding is consistent with the indications of previous research, which presents the importance of tolerance of ambiguity to performance in many business domains such as decision making (Dollinger et al., 1995), and the capability to deal with change (Judge et al., 1999) and complexity (Gupta and Fogarty, 1993).
Contrary to our expectation, the positive relation between self-efficacy and supply chain performance (backorder costs and inventory costs) is not supported.
One of possible reasons might be the different measures of performance used. The previous two decades of empirical research found that self-efficacy has a positive relationship with performance (e.g. Bandura, 1988; Wood and Bandura, 1989). Their performance measures, however, only included adaptability to advanced technology, managerial performance etc, not taking supply chain performance (backorder costs and inventory costs) respects into account.
Another reason for not finding a relationship might be because of other factors such as sample size, the proportion of people with different level of self-efficacy, etc. that can also have an effect upon the self-efficacy - supply chain performance relationship. In our research the average score of self-efficacy was 3.82 (SD=.30, see table 3 p.27), which means that players are on average middle self-efficacy and the standard deviation is low. However, among all 47 players, most of players (45%) were high self-efficacy, 25% of players were middle self-efficacy and the other players (30%) were low self-efficacy. Obviously, the distinction in score of self-efficacy is fine. Thus, if this research would be conducted by more players with even proportion and clear difference in self-efficacy, results might be different.
Overall, it should not be concluded that no relation was found between self-efficacy and supply chain performance. More study is needed to confirm our findings.
In our research, no positive relation is found between internal locus of control
and backorder performance or inventory performance. This is in contrast with
41 locus of control. However, nearly half players (n=21) were divided into high internal locus of control, few players (n=9) were defined as middle internal locus of control. This big difference in size seems to significantly affect the results. As shown in table 7, there is evident trend in supply chain performance but without significance in relation. Still, more future research on internal locus of control and supply chain performance is needed to confirm our finding.
5.1.2 Supply chain complexity and its relation with supply chain performance
The relation between supply chain complexity and its performance is expected to be negative in this study. This negative relation is only found between complexity and
backorder performance. This means the more uncertainty or variety there is in a system, the more complex the supply network, and the higher backorder costs (the poorer service performance). This finding is in line with the previous research which stated that complexity would hamper (delivery) service performance (e.g. Milgate, 2001; Perona and Miragliotta, 2004).
However, it is not confirmed that higher complexity is related to more inventory carrying costs. Probably, the reason for not finding a relation could be the set of complexity drivers. The previous cited research (e.g. Vickers and Kodarin, 2006) includes many variables inside and outside, such as the number of suppliers, the number of products or services etc. However, in our research, only three variables (the internal ordering decision making process, downward demanding and upward supplying) are thought to affect the complexity. More research on complexity is needed to confirm our finding.
5.1.3 Exploratory findings between personality traits, supply chain complexity and supply chain performance
5.1.3.1 Risk taking, supply chain complexity and supply chain performance
42
taking and backorder performance is found in this research. This finding further
explains that risk taking positively influences supply chain performance (backorder costs) through the partial mediating effect of supply chain complexity. Probably, players with low risk taking deviate from the optimal ordering policies, which lead to the increase in the amount of internal complexity as Sivadasan et al. (2002) stated "poor decision making can contribute to complexity generation", and then the increased complexity hampers the performance (backorder costs).
The moderating effect of complexity on the relation between risk taking and backorder performance is supported. This means risk taking could directly influence
the backorder costs without changing the amount of complexity, but the level of complexity would determine the direction and strength of the relation between risk taking and backorder performance. According to the findings of Hung and Ryu (2008), risk preferences of individuals may change with past experiences of demand deviation and supplier failures. Thereby, one possible reason might be that players change their risk preferences after several experiences of unpredictable orders from downstream or unreliable deliveries from upstream. From another point of view, this finding helps to support our earlier mentioned finding that complexity only partially mediates the effect of risk taking on backorder performance, not fully.
Not finding the mediating or moderating effect of complexity on the relation between risk taking and inventory performance is consistent with the failure in the establishment of hypothesis 1b.
5.1.3.2 Ambiguity, supply chain complexity and supply chain performance
The mediating effect of complexity upon the relation between tolerance of ambiguity and performance does not hold in this research. It is only confirmed that the positive relation between tolerance of ambiguity and backorder performance is stronger among organizations with higher level of complexity. Ashford and Cummings (1985)
43 performance. However, indirect or delayed feedback within the supply chain is instrumental in increasing the supply chain complexity. Thus, it is much more difficult for intolerants to seek feedback and enhance the performance among organizations with higher level of complexity.
Moreover, this finding improves our understanding of the relation between tolerance of ambiguity and backorder performance (H2a), by revealing that the impact of tolerance of ambiguity upon backorder performance is dependent of complexity.
5.1.3.3 Self-efficacy, supply chain complexity and supply chain performance
It is not confirmed that self-efficacy is related to supply chain performance through the mediating effect of complexity. The strong and negative relation between self-efficacy and inventory performance is found among organizations with lower level of complexity, and the weak and positive relation among organizations with higher level of complexity. Probably, with higher level of complexity observed with variable
demands and unreliable shipments, players with low self-efficacy will feel more difficult and doubt about their capabilities to perform. As a result, they exert less effort to make decisions. However, players who are high self-efficacy would tend to be persistent in this situation and make more effort to improve performance. In other words, people with higher self-efficacy could have better performance in the higher level of complexity, but not in the lower level of complexity. This is in line with the statement of Stajkovic and Luthans (1998) that self-efficacy relates to complex task performance, as we mentioned earlier.
5.1.3.4 Locus of control, supply chain complexity and supply chain performance
Like self-efficacy, the mediating effect of complexity on the relation between internal of locus of control and performance is not found. It is determined that the negative relation between internal of locus of control and backorder performance is slightly weaker among organizations with higher level of complexity. This is in
44 better than externals especially in the complex problem-solving situation. A possible reason might be different measure of the complex situation. Spector (1982) described the complex situation with tasks requiring complex information collection or processing, however, we observed the complex situation with uncertain and diverse material flow in our research, which might lead to different results.
In particular, our results show that internals could behave in totally different direction, dependent on the level of complexity. This interesting finding illustrates why internal locus of control is not positively related to backorder performance (H4a).
5.2 A conceptual model for the empirical findings
This section displays a new conceptual model derived by a synthesis of our above empirical observations (see figure 8).
45 According to this model, supply chain complexity is characterized by internal complexity and external complexity. External complexity is expected to influence internal complexity to some extent. Internal complexity mediates the relation between risk taking and backorder performance. On the other hand, external complexity moderates the relation between risk taking and backorder performance. Similarly, the direct effects of ambiguity and locus of control on backorder performance are moderated by external complexity. With respect to inventory performance, only the interaction effect of external complexity and self-efficacy exists.
Overall, this model clearly shows the way through which each personality trait influences performance (backorder costs or inventory costs), and highlights the importance of clarifying the role of complexity in the effects of personality traits on performance.
5.3 Conclusion
The goal of this research was to undertake an exploratory study of the relationship between personality traits and supply chain performance (backorder costs and inventory costs) with varying levels of supply chain complexity in the beer game setting. This research makes certain significant contributions to the understanding of behavioral causes of the bullwhip effect by studying the effect of human personality traits upon supply chain performance. Previous research has studied some human behavioral causes but mainly focus on the human cognitive ability. This research studies personality traits, which explores a new direction in this field. And to some extent, it explains why participants treated backorder and inventory differently, which is highlighted by Oliva and Gonçalves (2005) for future study. The present paper is one of the first to study supply chain complexity, which is measured by an information-theoretic approach, in the context of beer game experiments. Theoretical and empirical conclusions of other researchers were empirically examined again in the lab setting, which is an important contribution to the existing literature.
46 tolerance of ambiguity were found to have positive relations with supply chain performance, and only backorder performance, not inventory performance. No evidence was found that self-efficacy and locus of control were related to supply chain performance.
Also, this research provided evidences for the relationship between complexity and supply chain performance. That is, complexity is negatively related to backorder performance, however, a positive relationship was found between complexity and inventory performance, which was in contrast to our hypothesis. The finding of wholesaler with highest complexity having poorest performance indicated that wholesaler is the key position to control the amplification of orders further up the chain.
Supply chain complexity played an important role in the effects of personality traits on supply chain performance in this study. The partial mediating effect of complexity on relation between risk taking and backorder performance is supported. That is low risk takers led to the increase in the amount of complexity, which in turn decreased performance. In addition to an indirect effect found through complexity, a direct positive effect of risk taking on backorder performance is also found stronger in a certain higher level of complexity. Our results also reveal that tolerance of ambiguity had the stronger positive effect on backorder performance among organizations with higher level of complexity. The moderating effect of complexity on the relation between self-efficacy and inventory performance is also confirmed. Higher level of complexity had a strong and positive effect on the relation between self-efficacy and inventory performance. Finally, our research provides the evidence that complexity moderated the relation between internal locus of control and backorder performance. Internal was positively related to performance in less complex situation, and negatively related to performance in more complex situation.
5.4 Managerial implications
