The effect of risk taking by decision making
teams on the order stability in supply chains:
Exploring risk attitude as a behavioural factor influencing the
bullwhip effect
Researcher
Imke L.J. Gommans (4786446), Master’s student Business Analysis
and Modelling, Radboud University, Nijmegen
Supervisor
Dr. Hubert P.L.M. Korzilius, Radboud University, Nijmegen
Second examiner
Dr. Inge L. Bleijenbergh, Radboud University, Nijmegen
Abstract
Literature widely explores factors causing the bullwhip effect. However, less is known about behavioural factors influencing this phenomenon. Previous research indicated that risk attitude might be an important behavioural factor. Therefore, this experimental research assessed the effect of risk attitude of a team of decision makers on the variance of orders placed by a team (i.e. the bullwhip effect), using the Beer Game. Additionally, it investigates whether team diversity and risk perception influence this relation. Teams who played the Beer Game consisted of decision makers with the same risk attitude, with different compositions in terms of gender and country of origin. Evidence was found that decision making teams who were risk averse created a larger bullwhip effect than risk seeking decision making teams. This was especially true upstream the supply chain and when the team consisted of culturally diverse decision makers. Risk attitude seems to be an important behavioural factor influencing the bullwhip effect, indicating that it is worthwhile to take the risk attitude of supply chain employees into consideration, as a selection criterion in application processes.
Keywords: risk attitude, expected utility theory, bullwhip effect, Beer Game, decision making
teams, supply chain
Acknowledgements
I would first like to thank my supervisor Dr. Hubert Korzilius for his guidance and support throughout the whole research and writing process. He steered me in the right direction whenever he thought I needed it and helped me to grow as a researcher. Additionally, I would like to thank my second examiner Dr. Inge Bleijenbergh for her sharp feedback and encouragement. Furthermore, I would like to thank Prof. Dr. Etiënne Rouwette for providing the board games, which were needed to play the Beer Game. Executing this experiment and writing this thesis would have been impossible without the help of my classmates and friends. A special thank you to (in alphabetical order): Alec Eckert, Andres Montellano Zuna, Bram Stevens, Bryan van den Brink, Elena Shevyakova, Evi Voesten, Gian Hildebrandt, Giulietta Heusinkveld, Jasper Remmerswaal, Jesper Slaats, Ruud Jilesen and Sergio Alzate. Last, but definitely not least, I would like to thank my family and my partner for providing me with unfailing support and encouragement, not only during the process of researching and writing this thesis, but also throughout my years of study. This accomplishment would not have been possible without them, thank you.
Introduction
Supply chain management is recognized as a crucial element in the currently high competitive global business environment (Silvestre, 2015). Supply chain management focusses on seeking linkages and coordination between processes of suppliers, customers and the organisation itself, within the supply chain (Christopher, 2011; Tokar, 2011). Managing these processes lead to both strategic and operational advantages for organisations (Silvestre, 2015). One of the goals of supply chain management is the reduction or even elimination of buffers of inventory (Christopher, 2011; Tokar, 2010). Tokar (2010) additionally stresses the importance of behavioural issues in supply chain management, especially in terms of decision making and judgements. He argues that more research is needed to develop knowledge regarding human behaviour and decision making in supply chains (Tokar, 2010). This research focusses on investigating human behaviour and decision making in supply chains by investigating the influence of the risk attitude (i.e. “a person’s standing on the continuum from risk aversion to risk seeking” (Weber, Blais, & Betz, 2002: 264) of a team of decision makers on the bullwhip effect. “The bullwhip effect describes the tendency for the variance of orders in supply chains to increase as one moves upstream from consumer demand” (Croson, Donohue, Katok, & Sterman, 2014: 176). In addition, this research investigates whether team diversity (i.e. “…differences between individuals on any attribute that may lead to the perception that another person is different from self” (Van Knippenberg, De Dreu, & Homan, 2004) and the risk perception (i.e. someone’s belief about potential harm; someone’s perception of their own risk taking (Brewer et al., 2007) of a team have a moderating effect on the relation between the risk attitude of a team of decision makers and the bullwhip effect.
Multiple factors influencing the bullwhip effect have been researched. Croson et al. (2014), for example, tested if controlling environmental factors, such as unknown demand and fluctuating demand, which lead to coordination risk, have an influence on the bullwhip effect. When controlling for these factors the bullwhip effect decreased but still existed. Furthermore, other research discovered that even when the demand distribution is stationary and commonly understood, the bullwhip effect still exists (Croson & Donohue, 2006). This is partly due to the underweighting of the supply line; people tend to exclude orders they have placed but are not delivered yet (Croson & Donohue, 2006). Both researches indicate that human behaviour is an important driver for the bullwhip effect. Hung and Ryu (2008) agree with this and focus their research specifically on managers’ risk preferences as a behavioural factor. They argue that “…changing the risk preferences of managers with respect to demand changes and supplier
failures is a significant behavioural factor in explaining deviations in ordering decisions” (Hung and Ryu, 2008: 770). As a result, they state that risk attitude is an additional behavioural cause for the bullwhip effect, which provides insights on the challenges of reducing this effect. Qorbani (2017) confirms this, his research shows that the risk attitude of decision makers in the supply chain is influencing the bullwhip effect. He suggests that it could be worthwhile for companies to test applicants on the characteristic risk. Further research is needed to investigate if, and if so, how the risk attitude of decision makers is a cause of the bullwhip effect. This could give theoretical insights in a behavioural factor influencing the bullwhip effect and could potentially lead to insights for managerial and/or organisational practice (Colquitt & George, 2011).
To investigate whether risk attitude of a team of decision makers has an influence on the bullwhip effect (i.e. variance of orders placed by a team), this research used an experimental research method by simulating the supply chain via the Beer Game. The Beer Game is “… a training exercise developed at MIT in the early 1960s, offers an easy-to use tool for creating a common awareness of the fundamental issues in a supply chain” (Hieber & Hartel, 2003: 122). By using the Beer Game the influence of risk attitude of people participating (decision makers in real life) is measured and compared in a controlled manner (Hieber & Hartel, 2003). This research focusses on the risk attitude of teams. It does not only look at the relationship between risk attitude of a specific team of decision makers and the bullwhip effect, but also if team diversity on gender and on culture, and risk perception of decision makers has an influence on that relationship.
Research on team diversity based on gender, specifically in supply chains, shows that gender diversity in teams is a factor which influences performance of decision makers and other stakeholders in the supply chain (Brauner et al., 2013). The reason for this could be that women are generally less risk seeking than men, causing differences in performances of decision makers (Powell & Ansic, 1997). The current research investigates whether the performance (i.e. variance of orders placed by a team) in the supply chain is indeed influenced by gender diversity in teams or that risk attitude is the basis for that. Moreover, team diversity on culture could be a factor influencing the relationship between risk attitude of a team of decision makers and the variance of orders placed by a team. Cultural diverse teams are seen by Stahl et al. (2010) as both an asset and a liability, which fits the general thoughts within the literature. While Watson, Kumar, and Michaelson (1993) argue that cultural diverse teams have difficulty in solving process problems effectively in the short term, however score the same as cultural homogeneous
teams in the long term. The current research tries to investigate to what extent team diversity on culture influences the relation between risk attitude and performance (i.e. variance of order placed by a team).
Besides team diversity, risk perception could be a factor influencing the relation between risk attitude of a team of decision makers and the variance of orders placed by a team. Risk perception is different from the risk attitude of someone, since risk attitude is seen as a personality characteristic (Brewer et al., 2007; Weber, Blais, & Betz, 2002). There is no agreement in existing literature on the exact definitions of risk attitude and risk perception; some definitions have a lack of concreteness, a few have overlap between the two concepts and some are even contradictory (e.g. Fellner & Maciejovsky, 2007; Goodwin & Wright, 2014; Kahneman & Tversky, 1979; Pennings & Garcia, 2001; Sjöberg, 2000; Slovic & Peters, 2006; Weber, Blais, & Betz, 2002; Weber & Brachinger, 1997). Moreover, previous studies do not focus on the differences between risk perception and risk attitude, which is partly due to the inconsistent definitions used in research involving risk. This study could give an indication whether risk perception (as a separate concept) is of an influence.
To conclude, Qorbani (2017) supposes that risk averse people will create a larger bullwhip effect than risk seeking people. Apart from that, he suggests that risk attitude of possible employees is worth to test before assigning an applicant to a position in the supply chain. Therefore, the aim of this study is to give insights whether teams of risk seeking people indeed take better decisions in the supply chain (bullwhip effect is smaller), and could give insights in other opportunities to minimise the bullwhip effect using a simulation game representing the supply chain (i.e. the Beer Game). Accordingly, this thesis displays a concise literature review on the following topics: the bullwhip effect, risk attitude, team diversity on both gender and culture, and risk perception, followed by the hypotheses. Furthermore, it explains the methodology, after which the results are presented. Finally, the thesis discusses the interpretation of the results, the limitations of the study, the suggestions for further research and the practical implications.
Research objective & research questions
This research introduces the concept of risk attitude of decision makers in relation to the bullwhip effect and the possibility that this relation is moderated by team diversity and/or risk perception. The main objective is to contribute to existing literature on the bullwhip effect, by gaining more insight in risk attitude as a behavioural factor of the bullwhip effect, using the
Beer Game as an analogy of the real-world supply chain. This research answers the following research questions:
Research question 1: To what extent does risk attitude of decision making teams has an impact on the bullwhip effect in the supply chain using the Beer Game?
Research question 2: To what extent do diversity of decision making teams - both on gender and on culture - and risk perception influence the relationship between risk attitude of a team of decision makers and the bullwhip effect?
Theoretical background
Bullwhip effect
“The bullwhip effect refers to the tendency of orders to increase in variation as one moves up a supply chain” (Croson & Donohue, 2006: 323). Forrester (1958) coined this effect and its causes, using industrial dynamic approaches to show the bullwhip effect (Lee, Padmanabhan, & Whang, 2004a). Decision makers involved in the supply chain affect the bullwhip effect, particularly the ones who have an influence on the placement of orders (Lee, Padmanabhan, & Whang, 1997). The bullwhip effect is seen as a major difficulty in supply chain management; it increases stock levels, causes higher holding costs and inefficient use of resources, and eventually leads to poor profitability and poor customer service (Paik & Bagchi, 2007). Consequently, the inventory level accumulates followed by a phase of inventory shortage (backlogs), leading to high costs followed by the loss of the possibility to sell (Machuca & Barajas, 2003). This process repeats itself, in which the amplitude of the oscillations gradually increase, which is harmful to the efficiency of the supply chain and can even paralyze it (Lee et al., 2004a). Therefore, supply chain management tries to reduce both the oscillations caused by inventory changes as well as the total costs of the whole supply chain. Identifying the causes of the bullwhip effect is of vital importance in this respect, since it leads to directions for reducing the impact of this phenomenon (Lee, Padmanabhan, & Whang, 2004b).
Scholars see the bullwhip effect, on the one hand, as a consequence of exogeneous unpredictability; “variability is entering the supply chain system in the form of random demand, and the combination of forecasting and ordering procedures works to ensure that orders will be increasingly variable as they progress up the supply chain” (Kim & Springer, 2007: 172-173). On the other hand, they argue that the effect also can exist in the absence of exogenous oscillatory, meaning that endogenous outcomes play an important role (Kim & Springer, 2007). Lee et al. (2004a) acknowledge this. They see the bullwhip effect as a consequence of
endogenous variability of a firm’s processes, policies and industry characteristics. Causes of the bullwhip effect can therefore be “….ordering frequency, return policy, frequency and depth of price promotion, the degree of information sharing, demand forecasting methods and so forth” (Lee, Padmanabhan, & Whang, 2004a: 1888). Chen et al. (2000) argue that demand forecasting is indeed one of the important causes of the bullwhip effect. However, even if decision makers have complete knowledge about the demand, the bullwhip effect still exists (Chen, Drezner, Ryan, & Simchi-Levi, 2000). Croson and Donohue (2006) came to the same results; even when the demand distribution is stationary and commonly understood, the bullwhip effect exists. The literature therefore indicates, not surprisingly, that not only operational factors but also behavioural factors are a cause of the bullwhip effect (Croson & Donohue, 2006; Lee et al., 1997; Chatfield et al., 2004; Bhattacharya & Bandyopadhyay, 2010). Related to the behavioural factors causing the bullwhip effect, is the research done by Nienhaus, Ziegenbein and Schoensleben (2006), they were the first to address the role of human performance in a supply chain. They discovered that some aspects of human behaviour are a cause of the bullwhip effect. Important results where the finding of two types of extreme behaviour of humans which have a negative impact on the performance of the supply chain; safe harbour – “they order more than actually necessary and by that increase their safety stock” (Nienhaus et al., 2006: 553); and panic – “to empty the stock before the end customer’s demand increases” (Nienhaus et al., 2006: 553). Besides that, Bendoly, Donohue and Shultz (2006) argue that neglecting delays is also a cause of the bullwhip effect, since people have the tendency to discount feedback delays. Furthermore, Senge (1994) adds that learning disabilities and our ways of thinking are part of the cause of the bullwhip effect. People tend to not oversee how their actions affect others and do not experience learning because the consequences of their actions are often felt somewhere else in the supply chain.
In summary, three main behavioural causes of the bullwhip effect exist: “…(1) neglecting time delays in making ordering decisions; (2) lack of learning and/or training; and (3) fear of empty stock” (Bhattacharya & Bandyopadhyay, 2010: 1246; Donohue & Schultz, 2006; Nienhaus et al., 2006; Senge, 1994). A review concerning the causes of the bullwhip effect reveals that there are just a few studies specifically focusing on behavioural causes of this phenomenon. To get better knowledge about the behavioural factors causing the bullwhip effect, additional research is necessary (Bhattacharya & Bandyopadhyay, 2010).
Qorbani (2017) indicates that risk attitude could be an important behavioural factor for explaining the bullwhip effect. The results of the study by Hung and Ryu (2008) also show that
risk attitude might influence the bullwhip effect. In their research, they specifically looked at individual changes in risk attitude over time and concluded that decision makers in supply chains could change their risk preference.
Risk attitude
This research focusses on the influence of risk attitude of a team of decision makers on the variance of orders placed by a team (i.e. the bullwhip effect). Risk attitude is “a person’s standing on the continuum from risk aversion to risk seeking” (Weber, Blais, & Betz, 2002: 264) and it can be seen as a personality characteristic. Qorbani (2017) indicates that risk averse people, when exposed to the loss-of-rewards incentives, cause more oscillations and a larger bullwhip effect within the supply chain. MacCrimmon and Wehrung (1990) affirm these thoughts, since they found that the most successful executives are risk seeking. However, they also found that the most mature executives were the most risk averse, which looks contradictory. Additionally, Brauner et al. (2013) found that risk taking has an impact on the variance of orders placed in supply chains; it improved performance (i.e. the bullwhip effect).
The literature shows that risk attitude of individuals is measured in multiple ways (e.g. Cano & Salzberger, 2017; Harrison et al., 2005; MacCrimmon & Wehrung, 1990). The current study focuses on the expected utility theory as a means of measuring the risk attitude of individuals. The theory allows researchers to take the attitude of decision makers towards risk into account, when explaining if the risk attitude of a team has an effect on the variance in orders placed by a team (Goodwin & Wright, 2014). Risk attitude in the expected utility theory is “… a descriptive label for the shape of the utility function presumed to underlie a person’s choices. A person’s risk attitude describes the shape of his or her utility function (derived from a series of risky choices) for the outcomes in question” (Weber et al., 2002: 264). In other words, a utility function can specifically be used to assess not money related aspects, in this case the attitude towards risk of a decision maker (Goodwin & Wright, 2014). The expected utility theory has four substantive assumptions: cancellation (i.e. choice of options should depend on states which have different outcomes), transitivity (i.e. the transitivity of preference), dominance (i.e. principle of rational choice, someone always chooses the best option) and invariance (i.e. different ways of framing the same question should give the same results). Invariance and dominance seem to be essential assumptions, while transitivity could be questioned. Cancellation is rejected by many actors, which makes that most models using expected utility theory focus on dominance, invariance and transitivity (Tversky & Kahneman, 1986). Apart from this, there is a number of arguments against utility: (1) judgment is not
always rational - an option can be the best choice according to the expected utility theory, however another option is chosen (i.e. Allais paradox) (Goodwin & Wright, 2014); (2) the understanding and motivations of people underlying their responses is sometimes questionable (MacCrimmon & Wehrung,1990); and (3) individuals appear not to be consistently risk averse or risk seeking, this depends on the situation and domain (Weber, Blais, & Betz, 2002). A consequence of the criticism on expected utility theory is the development of the prospect theory. Prospect theory derives utility from gains and losses and states that people have an aversion to loss, meaning that people value the loss of €100 much higher than a gain of €100 (Barberis, 2013). This causes differences in behaviour and choices of decision makers, possibly leading to different ordering patterns in supply chains (Tokar, 2010). The prospect theory is multiplying the value of an uncertain outcome by a decision weight, while the expected utility theory is weighting each outcome purely with a probability (Tversky & Kahneman, 1986). This indicates that the prospect theory has a more nuanced and less rational view on decision making under risk than the expected utility theory. But the difficulty with the prospect theory is that “…it is often unclear what a gain or loss represents in any given situation” (Barberis, 2013: 192), which makes application of the theory difficult, especially outside the field of economics (Barberis, 2013).
Despite the criticisms and development of the prospect theory, the expected utility theory is still used for explaining a person’s risk attitude. Goodwin and Wright (2014) argue that the expected utility theory is valuable when looking at decision making incorporating problems involving a high level of uncertainty and risk. An important note is that the decision maker must be familiar with the concept of probability.
To contribute to literature, this research further explores risk attitude of decision makers as a behavioural factor explaining the bullwhip effect. Not by focussing on the changes in risk attitude of the individual, but by looking at teams and in which way teams, with certain predetermined risk attitudes, influence the bullwhip effect. These teams are composed by bringing together individuals with the same risk attitude. Qorbani (2017) indicates that risk averse people cause a larger bullwhip effect than risk seeking people. Besides that, Bendoly et al. (2006) argue that natural risk aversion could have a negative impact on operational success. This research wants to look if, and if so, how this is the case in supply chains; does risk attitude of a team of decision makers affect the performance (i.e. variance of order placed by a team) of the supply chain. Leading to the following hypothesis:
Hypothesis 1: The risk attitude of a team of decision makers has a negative effect on the variance of orders placed by a team.
Moderating variables
Team diversity on gender and on culture, and risk perception may influence the probable effect of risk attitude of a team of decision makers on performance – variance of orders placed by a team. These possibly moderating variables are explained in the following two paragraphs.
Team diversity on either gender and culture
Diversity can be described as “…differences between individuals on any attribute that may lead to the perception that another person is different from self” (Van Knippenberg et al., 2004: 1008). Team diversity can be seen as both an opportunity and a threat; it can create operational synergy, but also miscommunication, intragroup1 conflict and lack of trust (Horwitz, 2005). Collective intelligence could be a factor causing opportunities as well as threats for performance. Wooley et al. (2010) argue that people in a group create collective intelligence when working together. However, this depends on the composition of the group and the way a group interacts (Wooley et al., 2010). This indicates that team diversity could influence performance. An important note to make is that the members of a team in the current experiment are not allowed to talk to each other. Meaning that they are a team, but not actively working together, they only communicate via placing orders and receiving crates of beer. Since most literature on team diversity is focussing on teams which are allowed to communicate, results of the current research could be different from existing knowledge.
Horwitz and Hortwitz (2007) reviewed existing literature on team diversity and concluded that researchers reported inconsistent findings related to team diversity and its effect on team outcomes. Their research shows that task-related diversity has a positive outcome on team performance, but bio-demographic diversity has not (bio-demographic diversity concerns gender, age, race etc.). Although this may be true, Brauner et al. (2013) indicate that gender does have an influence on performance of decision makers and other stakeholders in supply chains. The results suggest that women (individually) perform worse within supply chains (i.e. they have a larger stock spread) than men, which is due to their, on average, lower self-efficacy levels (Brauner et al., 2013). On contrast, Wooley et al. (2010) argue that groups including women have higher collective intelligence; indicating that groups including women do have a positive effect on the performance of decision makers in supply chains. Another study, looking
1 Groups and teams are both used in existing literature. When the literature refers to groups instead of teams, but
at gender differences in risk behaviour in financial decision-making, shows that women are generally less risk seeking than men (Powell & Ansic, 1997), which could indicate that risk attitude could be the underlying factor influencing the performance of decision makers. To assess if team diversity on gender has a moderating influence on the relation between the risk attitude of a team of decision makers and the variance of orders placed by a team, the following hypothesis is tested:
Hypothesis 2a: The more diverse teams of decision makers are in terms of gender, the larger the effect of team risk attitude on the variance of orders placed by a team.
Team diversity on culture could also be seen as a factor influencing the performance of teams. “Culture consists of a commonly held body of beliefs and values that define the ‘shoulds’ and the ‘oughts’ of life …and guide the meaning that people attach to aspects of the world around themselves. Cultures provide a source of identity for their members” (Stahl et al., 2010: 691). Stahl et al. (2010) argue that cultural diversity can be seen both as an asset and a liability, indicating that the influence can be significantly positive as well as negative. McDaniels and Gregory (1991) agree with this, their results show that cultural characteristics can have an effect on decision making, since it influences the evaluation of options, especially in uncertain situations. While Watson, Kumar and Michaelson (1993) argue that a cultural diverse team only scores different on solving process problems in the short term. In the long term, culturally homogeneous and heterogeneous groups score the same. In short, current literature indicates that team diversity on culture could have an effect on the performance of decision makers. To operationalize and measure this effect, country of origin is used since culture often relates to country-based cultures (especially in international business) (Stahl et al., 2010). Which leads to the following hypothesis:
Hypothesis 2b: The more diverse teams of decision makers are in terms of country of origin, the larger the effect of team risk attitude on the variance of orders placed by a team.
Risk perception
Risk perception is someone’s belief about potential harm, in other words, someone’s perception of their own risk taking (Brewer et al., 2007). Individual, group, and cultural differences, as well as situational differences can influence this perception (Weber, Blais, & Betz, 2002). The difference between risk attitude and risk perception is that risk attitude is seen as a personality
characteristic, while risk perception is how an individual perceives his own risk taking. The definitions of risk attitude and risk perception in current literature vary. Risk attitude is defined as a stable personality trait (Fellner & Maciejovsky, 2007), as a stable construct which does not change by stimuli (Pennings & Garcia, 2001) and contradictory as something that does change over stimuli ranges (Kahneman & Tversky, 1979). Besides, risk attitude is seen as nothing more than a descriptive label showing the shape of the utility function of a person’s choice (Weber, Blais and Betz, 2002), but also as someone’s natural preference for a more or less risky alternative when making a decision (Goodwin & Wright, 2014). Risk perception is seen as peoples’ perception which determines the definition of risk (Weber & Brachinger, 1997) and as “a function of properties of the hazards” (Sjöberg, 2000: 9) in which attitude is a crucial factor. Furthermore, risk perception is seen as something that includes risk as feelings – the intuitive and instinctive reactions towards danger – and risk as analysis – the logic, reason and scientific considerations to assess risk and making decisions (Slovic & Peters, 2006). Multiple definitions for either risk attitude and risk perception are used, some are more specific than others, some are contrary and a few have overlap. This makes it difficult to compare results and shows the importance of a clear definition of risk attitude and risk perception.
Most research, either in- and outside supply chain management, focusses on either risk attitude or risk perception and sometimes these concepts are intertwined. This study focusses on risk attitude and risk perception as two separate concepts. There is no study which specifically looks at the moderating effect of risk perception on the relation between risk attitude and another variable, except for Weber, Blais and Betz (2002), to some extent. In their research, they mainly focus on measuring risk perception and risk behaviour which fall, according to them, both under a domain-specific risk-attitude scale.
To understand the willingness (i.e. the perception) of a person (e.g. decision maker) to take risks, multiple tools are developed by different researchers (Cano & Salzberger, 2017; Gilliam, Chatterjee, & Grable, 2010; Kuzniak et al., 2015). This research uses the 13-item Grable and Lytton financial risk-tolerance scale, since it is seen as one of the most reliable and valid methods to measure risk perception (Cano & Salzberger, 2017; Kuzniak et al., 2015). Risk-tolerance is explained, in this scale, as the inverse of risk aversion (Grable, 2016). To conclude, there is no research which clearly indicates whether risk perception has a moderating effect on the relation between risk attitude of a team of decision makers and the variance of orders placed by a team, and if so, in which direction this effect exists. However, since other research on risk indicates that risk averse decision makers influences performance negatively (e.g. Bendoly et al., 2006; Qorbani, 2017), the following hypothesis is developed, in
which a higher perception of risk taking means a higher willingness to take risk (risk seeking); a team is more risk tolerant, which is the inverse of risk aversion (Grable, 2016).
Hypothesis 3: The higher the perception of risk taking by a team, the smaller the effect of team risk attitude on the variance of orders placed by a team.
Conceptual model
Figure 1 depicts the conceptual model of this study and the developed hypotheses (H1, H2 and H3). The independent variable is risk attitude of decision makers, whereas the dependent variable is performance; variance of orders placed (Croson & Donohue, 2005). The possible influence of the moderating and control variables are displayed. All variables are measured and analysed per team.
Method
Design
The research used an experimental research method. In experimental research one variable is manipulated to see the effect on other variables (Field, 2013). The experiment was conducted in a controlled environment (Denscombe, 2012), which was useful because only then the risk attitude of a team could be manipulated, while the other factors were controlled for (Field, 2018). This was important to determine if the risk attitude of a team of decision makers affects the supply chain, and in which way.
Risk attitude of decision makers Performance Variance of orders placed Moderating variables
- Team diversity; on gender and on country of origin - Risk perception
Control variables
- Time of participation - Pairs of game leader(s) - Experience Beer Game
H1
H2/H3
The experiment conducted within this research had a between-subjects design; i.e. “…different teams of entities take part in each experimental condition” (Field, 2018: 18). Risk attitude of a team of decision makers was the independent variable in this study and performance - variance of orders placed by a team (Croson & Donohue, 2005) – the dependent variable. The greater the variance in performance, the larger the bullwhip effect. The risk attitude of each decision maker was measured before the experiment, using a questionnaire based on the expected utility theory. The effect of risk attitude of a team of decision makers on the variance of orders placed by a team (i.e. performance) was measured via the Beer Game, a simulation game representing the supply chain of beer.
This research investigated, besides the effect of risk attitude of a team of decision makers on the variance of orders placed by a team, if moderator variables affected the relationship between the independent and dependent variable. A variable is moderating when the strength or direction of the relationship between the independent and dependent variable is changed (Field, 2018). In this study, three moderating variables were taken into account: (1) team diversity on gender, (2) team diversity on country of origin, and (3) risk perception of teams of decision makers. Furthermore, control variables were used to see if certain extraneous factors had an influence on the causal relation between risk attitude of decision making teams and the variance of orders placed by a team (Field, 2018). In other words, is there an alternative explanation for the results than the causal relation between the independent and dependent variable suggests. Three control variables were taken into account; time of participation (early morning, late morning or afternoon at the 14th of March and early morning at the 21st of March); pairs of game leaders; and having experience with playing the Beer Game.
An experiment mainly focusses on internal validity, which “refers to the approximate validity with which we infer that a relationship between two variables is causal or that the absence of a relationship implies the absence of cause” (Cook and Campbell, 1979: 37). The internal validity in the current research was high, because the experiment primarily looks at risk attitude of a team of decision makers and leaves other factors constant, this made it possible to measure exactly the relationship that was aimed to be measured. The data collected from the participants (i.e. decision makers) were therefore valid for the purpose of this research. Furthermore, to create construct validity, factor analysis was used to determine if the questionnaire assessing risk perception of decision makers (13-item financial risk-tolerance scale of Grable and Lytton, 2003) was valid. This analysis identifies cluster of variables and helps in “…understanding the structure of a set of variables; …constructing a questionnaire to measure an underlying variable;
… and reduces a data set to a more manageable size while retaining as much of the original information as possible” (Field, 2018: 779).
The strategy used in this research was reliable, which means that the research produces the same data if the research is repeated (Denscombe, 2012). This study consisted of determining the risk attitude, gender, country of origin and risk perception of the decision makers – risk attitude and risk perception was first measured individually and then combined in team scores. The Beer Game – a simulation game representing the supply chain – was used to determine how these concepts influenced a team of decision makers in the supply chain specifically looking at the orders placed by teams of decision makers. All decision makers filled in a questionnaire before the experiment (to determine risk attitude of a team of decision makers, gender and country of origin) and after the experiment (to determine risk perception of decision makers). The questions and the results were carefully designed and documented, which makes it possible to ask respondents the same questions again and receive the same results. The composition of the teams within this experiment was documented to make it possible to undertake the same experiment again. The steps taken during the experiment were performed multiple times in other researches, and were executed in the same way in this experiment. The experiment was reliable since a strict set of protocols were followed in playing and managing the Beer Game. This makes it possible to execute the same experiment again. In addition, Cronbach’s alpha was used to measure scale reliability of financial risk-tolerance scale (Field, 2018).
Participants
The experiment was done in the Business Analysis for Responsible Organisations (BAfRO) course given at the Radboud University, Nijmegen, in which 314 students participated (this was part of the second year bachelor’s program of (International) Business Administration and the pre-master’s programme of Business Administration). The participants (i.e. decision makers) consisted of men and women from the Netherlands or a foreign country. This research focussed on teams. In total 59 teams (i.e. 246 individuals) participated in the Beer Game, from which 40 teams had valid scores on the risk attitude. In total 10 risk averse teams and 30 risk seeking teams were used in the analyses (see Procedure & Measures for further explanation). From the 40 teams, 37 teams had a valid score on the variance of orders placed by a team. This sample size enables a medium to large effect with a statistical power of at least .80, using a significance level (α) of .05 (Cohen, 1992).
Materials
Pre-experimental, the risk attitude of individuals was measured, after which teams were formed of participants (i.e. decision makers) with the same risk attitude. To measure risk attitude the expected utility theory was applied, which uses a questionnaire with lottery based questions.
Figure 2: Lottery based question
As shown in Figure 2, a question had two options, one getting €50 for certain and one a lottery ticket offering a probability of getting €100 versus a probability of getting nothing. The questionnaire had five questions with five probabilities to win €100; 0.5 as shown in Figure 2, and 0.1, 0.3, 0.7, and 0.9, respectively, in the other four questions. When the expected payoffs of the two options in one question were the same and the participant was indifferent between the two options, the participant was risk-neutral. When the participant either chose the first or the second option the participant was risk averse or risk seeking; risk averse when going for the option which gives more certainty in getting a reward and risk seeking when going for the opposite. The expected utility theory states that the utility assessments is sensitive to the values used (Goodwin & Wright, 2014). Students could relate to the values used and could therefore understand the risk associated with that. Weber et al. (2002) argue that the degree of risk taking is highly domain-specific. Since students generally do not have any experience in decision making in supply chains, it was chosen to state the questions in general terms. Furthermore, the expected utility theory states that a decision maker needs to be able to place all lotteries in order of preference (Goodwin & Wright, 2014). Therefore, the questionnaire displayed the five lottery questions at the same page, which made it possible for the decision maker to consider and place the lotteries in order. Moreover, the questionnaire consisted of questions indicating gender and the country of origin of a decision maker, to arrange the teams on diversity. To carry out the questionnaire, an online survey was used (Appendix A).
The experiment consisted of executing the Beer Game. “The Beer Game is a replica of a system for producing and distributing a single brand of beer” (Goodwin & Franklin, 1994: 7). To conduct the Beer Game, a board which represents a typical supply chain was used (Sterman, 2000). The game consisted of a deck of cards which represented customer demand, red coins which represented crates of beer, small blank notes and the game record sheets. The game record sheets showed the oscillations in inventory, backlog and orders placed, which was used
to calculate the variance in orders placed by a team (Appendix B).
Post-experimental, a questionnaire was administered which consisted of three components: the inventory or backlog (by week) sheet; the estimated retail customer orders (by week) sheet (System Dynamics Society, 1998); and the Grable and Lytton (2003) financial risk-tolerance scale. The sheets showed the order patterns of the decision makers and their thoughts about how much the ‘real’ customer ordered - via the customer order cards (System Dynamics Society, 1998) (Appendix C & D). The Grable and Lytton financial risk-tolerance scale consists of thirteen questions measuring the financial risk-tolerance of an individual (Grable, 2016) and was used to assess the risk perception of each participant (Appendix E). The individual scores were added up to establish the risk perception a the team of decision makers. The questionnaire including a hardcopy of the two sheets and the Grable and Lytton financial risk-tolerance scale, was filled in by each decision maker after the Beer Game.
Procedure & measures
A questionnaire was administrated to determine the risk attitude of each participant. This questionnaire consisted of ten questions; five lottery based questions; and five questions containing the topics gender, country of origin, student number, study program and experience with the Beer Game. Decision makers were asked via Blackboard and email to fill in the questionnaire, via an internet link, two weeks before the experiment. The results of the questionnaire were collected and the answers were analysed, after which the decision makers were divided into teams on the basis of their risk attitude – very risk averse, risk averse, risk neutral, risk seeking, very risk seeking – and study program to make sure that no one had to miss a lecture. The composition of the teams were as follows: (1) only women/men, (2) three women/men and one man/woman, (3) two women and two men, (4) two Dutch and two foreign persons, and (5) three Dutch and one foreign person. In total 59 teams participated in the Beer Game.
To execute the Beer Game, pairs of game leaders were asked to help with the experiment. The game leaders were students of either the European Master of System Dynamics (EMSD) or the Master Business Analysis and Modelling (BAM). All the game leaders got a preparatory meeting led by the researcher, which consisted of explaining the rules of the game, sharing and demonstrating the PowerPoint that needed to be used, and introducing the questionnaire which had to be handed out at the end of the game. Game leaders who did not had any experience with the Beer Game, played the game themselves, together with the researcher, before the experiment.
Following the preparations, the decision makers took part in the Beer game. “There are four positions at each game board: Factory, Distributor, Wholesaler and Retailer” (Goodwin & Franklin, 1994: 7). One person was assigned to each position, he or she had to play the game and keep the score. Customer demand at each position drove the system. Every position’s placed order was only revealed to the player on their right and the factory placed an order from itself (i.e. at his/her brewery) and had an unlimited supply. As Figure 3 shows, the game required two time periods (i.e. weeks) for an order/shipment to reach its destination, except for the factory which required, in total, three weeks to receive their order (Goodwin & Franklin, 1994; Sarkar & Kumar, 2015). Figure 3 shows the delays present in the supply chain (the double lines).
Figure 3: Schematic representation of the Beer Game
The goal for every player was to minimise costs during the whole game. Carrying inventory costed €0.50 per crate of beer/per period and backlog €1 per crate/per period (Sterman, 2000). Since the decision makers were asked to keep the records manually, this experiment was influenced by human-involvement, which makes mistakes unavoidable. To detect these mistakes, a developed computer model (i.e. Excel sheet) was used to test the records on consistency. Mistakes in the record sheets of more than a few cases per week in any of the four positions, led to exclusion from the analysis for the total team (Sterman, 1989). Furthermore, when a record sheet from one or multiple positions was missing or not fully filled in, the team was discarded from further analysis. This also applied to teams that played less than 25 weeks (Qorbani, 2017). It was possible that teams played less or more than 25 weeks due to two factors: (1) teams that were rather slow where separated from the group and individually helped by one of the game leaders, and (2) teams that where rather fast where allowed to play the game on their own - under the condition that the factory named the week number at the start of each new week.
three components: the inventory or backlog (by week) sheet, the estimated retail customer orders (by week) sheet (System Dynamics Society, 1998) and the Grable and Lytton financial risk-tolerance scale (Grable & Lytton, 2003).
The risk attitude of the decision makers in the supply chain was measured via the questionnaire in February 2018. The experiment and the questionnaire measuring the risk perception of decision makers took place in March 2018.
The data collected from the questionnaires and the Beer Game were used for analyses. Before the analyses were executed, the presence of the bullwhip effect was tested using two methods. The first method displayed the presence of the bullwhip effect by depicting the average of orders placed by each decision maker per week (Sarkar & Kumar, 2015). The second method displayed the presence of the bullwhip effect by using the order rate variance ratio. The ratio was calculated by dividing the variance of orders placed by a decision maker in a position, by the variance of orders placed by a decision maker in its immediate lower position. When the ratio was greater than one, the bullwhip effect existed (Chen et al., 2000; Croson & Donohue, 2006). The variance of orders placed per team was used in further analyses to measure the effect of risk attitude of teams of decision makers on the bullwhip effect.
Validity and reliability analyses were performed on the 13-item financial risk-tolerance scale. Before executing these analyses, the scores on question 9 and 10 were averaged to obtain a composite score (Kuzniak et al., 2015). Besides that, scores were standardised to a scale of one till four, to get more reliable results (Appendix F). After that, factor analysis was applied to measure the construct validity of the scale. First an exploratory factor analysis of the 13-item scale was carried out and resulted in a two factorial solution. However, too many items loaded on both factors. Therefore, it was decided to run a fixed, one factorial design. The one factorial analysis showed that both item 3 and 7 needed to be excluded from the analysis. Question 9 and 10 combined and question 11 had communalities < .20, but since their factor loadings were above .30 and the total explained variance was 34.3%, these items were included in further analyses (KMO = .847 and Bartlett’s Test of Sphericity is significant since p < .05) (Appendix G). Furthermore, Cronbach’s α was used to measure the reliability. A scale is considered to be reliable when Cronbach’s α is above .70 and when it is under .60 it is inadequate (Field, 2013; Grable, 2017). The reliability analysis was executed on the ten remaining items from the 13-item financial risk-tolerance scale. Cronbach’s α = .77, meaning that the risk perception scale is reliable. These ten items were used in further analyses (Appendix H).
Statistical analysis
All data collected via the questionnaires and the Beer Game were processed in SPSS. Table 1 shows the distribution of the sample for teams with valid scores, considering the restrictions mentioned in the former paragraph. There were 40 teams with a valid score on risk attitude; one team was very risk averse, nine teams were risk averse, 24 teams were risk neutral and six teams were risk seeking.
Table 1: Distribution of the sample – independent, dependent and possible moderating variables
Since not every risk category was big enough to perform an analysis, the risk categories were combined and two groups were composed; the very risk averse teams and the risk averse teams were combined in one risk averse team, and the risk neutral and the risk seeking teams were combined in the risk seeking team. In total 10 risk averse teams and 30 risk seeking teams were used for the analyses. Besides that, 37 teams had a valid score on both risk attitude and the variance of orders placed by a team and 40 teams had a valid score on both risk attitude and risk perception of a team of decision makers. Team diversity had 38 teams with valid scores on both gender and risk attitude of decision making teams; four teams with only men, three teams with only women, 10 teams with three men and one woman, eight teams with three women and one man, and 13 teams with two men and two women. Furthermore, team diversity had 39 teams with valid scores on both country of origin and risk attitude of decision making teams; 23 teams had only decision makers born in the Netherlands and 16 teams had at least one decision maker not born in the Netherlands.
To investigate hypothesis 1, regression analysis was used. Since the research had one categorical independent variable (risk attitude of a team of decision makers) and one quantitative, unbounded and continuous dependent variable (variance of orders placed by a team), simple regression was used to analyse the data. Because the risk attitude of a team of decision makers was categorical, it was transformed into a dummy variable (i.e. risk averse decision making teams had a score of one and risk seeking decision making teams had a score of zero), which made both the independent and dependent variable of an interval level. Furthermore, the dependent variable was modified using inverse transformation to improve
N Minimum Maximum Mean Median Std. Deviation
Risk attitude - team 40 1.00 4.00 2.88 3.00 0.69 Variance of orders placed - team 37 8.47 2055.51 161.45 36.12 385.13 Team diversity on gender 38 1.00 5.00 3.61 4.00 1.33 Team diversity on culture 39 1.00 2.00 1.41 1.00 0.50 Risk perception - team 40 71.00 140.99 98.47 97.67 15.72 Valid N (listwise) 34
normality. The main assumptions of a regression analysis were met; linearity, independent errors, homoscedasticity, normality, non-zero variance, and multicollinearity (Appendix I).
To investigate hypothesis 2a, 2b and 3, moderation analysis (i.e. interactions in regression) was used to analyse the data. A moderation effect is present when there is a significant interaction between the predictor (in this research risk attitude of a team of decision makers) and the moderator (in this research team diversity and risk perception of a team of decision makers) in a regression (Field, 2013). This can be determined by looking at (1) the effect of the predictor on the outcome variable, (2) the effect of the moderator on the outcome variable and (3) the effect of the predictor multiplied by the moderator on the outcome variable (Baron & Kenny, 1986). Before executing the moderation analysis the 13-item financial risk-tolerance scale was analysed. All answers given by an individual were assigned a score and led to total scores between the 13 and 47 points (Appendix E). According to the research of Kuzniak et al. (2015) the total scores have a normal curve, the higher the scores the higher the risk-tolerance of a person, meaning that the mean score was used to determine the degree of risk perception of each individual. The individual scores were summed up to create the team score. Within this 13-items scale, three questions had the same design as the questions which determined the risk attitude of decision makers. To investigate whether there was conceptual overlap between these questions and the questions asked in the questionnaire determining the risk attitude of decision makers, Pearson’s correlation was used. A Pearson’s correlation of .10 indicates a small effect, .30 a medium effect and .50 a large effect (Cohen, 1992) (Appendix J). After which, the moderation analysis was executed; the PROCESS tool in SPSS was used (Field, 2018) (Appendix K).
At last, control variables – time of participation, pairs of game leaders and experience with the Beer Game – were analysed to see if they have an influence on the causal relation between risk attitude of a team of decision makers and the variance of orders placed by a team.
Table 2: Distribution of the sample - control variables
Table 2 shows that all control variables had a valid score when there also was a valid score on the risk attitude of decision making teams (N = 40). The teams participated in the experiment in four timeslots (i.e. time of participation), from which the first three timeslot were on the 14th of March and the last timeslot was at the 21st of March; 15 teams participated in the early
N Minimum Maximum Mean Median Std. Deviation
Timeslot 40 1.00 4.00 2.15 2.00 1.05
Pairs of game leaders 40 1.00 10.00 5.20 5.00 2.58
Experience Beer Game - team 40 1.00 2.00 1.03 1.00 .16
morning, eight teams participated in the late morning, 13 teams participated in the afternoon (14th of March) and four teams participated in the early morning (21st of March). There were ten different pairs of game leaders, in which the valid scores ranged between leading one and seven teams. Furthermore, only one person had experience with the Beer Game, making it impossible to analyse the possible influence of this control variable. Consequently, only time of participation and pairs of game leaders were used in the analyses. ANOVA was applied, but since the assumptions of homogeneity and linearity could not be met by both control variables, the Kruskal-Wallis test was used. Kruskal-Wallis test is a non-parametric test which was applied “to look for differences between groups of scores when those scores have come from different entities” (Field, 2013: 236). For all statistical analyses α was set at .05.
Research ethics
This research was conducted in accordance with the Netherlands Code of Conduct on Scientific Practice. The first part of the study contained a questionnaire to determine the risk attitude, gender and country of origin of the participants (i.e. decision makers) in order to divide them in teams. This meant that the decision makers could not participate in the experiment fully anonymously. A decision maker was not made aware of the risk category he or she, or others were in. Furthermore, student numbers provided by the decision makers were only used for linking the data of the two questionnaires and the data from the Beer Game, after which only the team outcomes were used for the results. In this way anonymity could be guaranteed to some extent. This to prevent the decision makers from harm and respect their privacy (Denscombe, 2012).
The results of the questionnaires are only used for this research and were/will be treated in the strictest confidence. Only the researchers had access to the data and any third party did and will not receive access (Denscombe, 2012).
Participation was not fully voluntary because it was an element of the BAfRO course; while this is not a mandatory part of the course, students could have the feeling that it is obliged to participate. When a decision maker had reasons not to participate, he or she was allowed to withdraw from the experiment at any time. This was possible via: (1) informing the researcher per email, (2) not filling in the first questionnaire, leading to no participation in the Beer Game, since the results of the questionnaire gave a participant access to play the game, and (3) by informing the game leaders during the game. The participant did not lose course credits when withdrawing from the experiment. There were no inducements, besides the learning aspect for the BAfRO course, to encourage participation. The decision makers gave written consent to use the results by filling in the first questionnaire (Denscombe, 2012).
All the decision makers were informed about the experiment via the BAfRO course manual. The necessary information was distributed via BlackBoard and Radboud University email, including a debriefing after the Beer Game. Decision makers had the opportunity to subscribe to receive the results of this research.
Results
Figure 4 presents the average orders placed per decision maker in the supply chain for both risk averse teams and risk seeking teams. It shows that the variance of orders placed increases as one moves up a supply chain; the bullwhip effect. Besides that, the graph shows that the difference between orders placed by risk averse and risk seeking decision makers was relatively small for the retailer and wholesaler compared with the distributor and the factory. Two methods were used to demonstrate the presence of the bullwhip effect (Chen et al., 2000; Croson & Donohue, 2006; Sarkar & Kumar, 2015).
Figure 4: Average orders placed per decision maker in the supply chain
Figure 5: Average order placed by each risk attitude team per position per week
The first method displays the presence of the bullwhip effect by depicting the average of orders placed by each decision maker per week (Sarkar & Kumar, 2015). Figure 5 visualises this for both risk averse and risk seeking teams. Both risk attitude teams show that as one
moves up the supply chain the orders placed by a decision maker (on average) increases, seeming to create the bullwhip effect. The graphs indicates that a team of risk averse decision makers creates a larger bullwhip effect than a team of risk seeking decision makers (i.e. the amplitude of the bullwhip effect is higher in risk averse teams than in risk seeking teams). The second method displays the presence of the bullwhip effect by using the order rate variance ratio (Chen et al., 2000; Croson & Donohue, 2006). Figure 6 presents the ratios of the current research, and of previous studies. Both risk averse and risk seeking teams had ratios above one, meaning that the bullwhip effect seemed to exist, this was also the case when combining the results of both teams.
Figure 6: The ratio of average variances between roles in this study and previous studies
The current research showed a similar pattern as Sterman (1989) when looking at teams including risk seeking decision makers. Teams including risk averse decision makers deviated from previous research and even showed (in larger numbers) the opposite behaviour (Croson & Donohue, 2006; Qorbani, 2017; Sterman, 1989).
Testing the hypotheses
To statistically test whether the risk attitude of a team of decision makers does have an influence on the variance of orders placed by a team, regression analysis was applied. First, a regression analysis was executed with the original variables; the independent variable was the dummy variable for the risk attitude of a team of decision makers and the dependent variable was the variance of orders placed by a team. Second, a regression analysis was executed using the (inverse) transformed variable; the independent variable was the dummy variable for the risk attitude of a team of decision makers and the dependent variable was the transformed variable of the variance of orders placed by a team. Table 3 shows the results of both regression analyses. Since the adjusted R2 in the analysis including the original variables was higher than de adjusted
original variables was used.
The regression analysis showed that a risk averse attitude of a team of decision makers deviates significantly from a risk seeking attitude of a team of decision makers (p = .006). The variance of orders placed by risk averse teams was on average 390 more than the variance of orders placed by risk seeking teams. The overall fit was significant and the explanatory power indicates a medium to large effect (Cohen, 1988; Field, 2013) (Appendix I).
Table 3: Results regression analysis
Moderating variables
Three moderating variables were analysed: team diversity on gender, team diversity on country of origin and risk perception of a team of decision makers. Before the moderation analysis was executed, the correlation between risk attitude of a team of decision makers and risk perception of a team of decision makers was analysed. Since three questions in the 13-item financial risk-tolerance scale had the same design as the questionnaire used to measure the risk attitude of decision makers, Pearson’s correlation was used to analyse whether there was a correlation. Pearson’s correlation was .23, meaning that the correlation is relatively weak. Therefore, all three questions were included in further analyses (Appendix J).
Table 4 shows the results of the moderation analysis. Team diversity on country of origin had a moderating influence on the relation between risk attitude of a team of decision makers and the variance of orders placed by a team. Additionally, the conditional effect of risk attitude of decision making teams on the variance of orders placed by a team at the values of team diversity on country of origin showed the following results: when teams consisted of decision makers all born in the Netherlands, there was a non-significant negative relationship between the risk attitude of a team of decision makers and the variance of orders placed by a team (b = -0.417,
b SE B β p
Constant 66.52 66.26 p = .322
(32.41, 115.55)
Risk attitude team - risk averse (dummy) 390.27 134.35 .441 p = .006 (18.84, 939.26)
Constant .04 .01 p < .001
(.03, .05)
Risk attitude team - risk averse (dummy) -.02 .01 -.254 p = .129 (-.04, .01)
Including transformed variable of variance of orders placed by a team Including original variables
Note. Adjusted R2 = .17 for regression including original variables; Adjusted R2 = .04 for regression including transformed variable
95% CI [-133.008, 179.756], t = 0.46, p = .645); when teams consisted of at least one decision maker not born in the Netherlands, there was a significant positive relationship between the risk attitude of a team of decision makers and the variance of orders placed by a team (b = 0.583, 95% CI [274.273, 2861.835], t = 2.47, p = .019). These results show that the risk attitude of a team of decision makers had a larger effect on the variance of orders placed by a team, in teams consisting of at least one decision maker not born in the Netherlands than in teams consisting of decision makers only born in the Netherlands; teams consisting of at least one decision maker not born in the Netherlands had a greater difference in the variance of orders placed by risk averse and risk seeking teams than teams of decision makers only born in the Netherlands. Both team diversity on gender and risk perception of a team of decision makers had no moderating effect (Appendix K).
Table 4: Results moderation analysis
b SE B t p 178.00 84.73 2.10 p = .044 (5.19, 350.82) 45.63 40.26 1.13 p = .226 (-36.49, 127.74) 473.59 325.00 1.46 p = .155 (-189.27, 1136.45) 76.82 144.13 .53 p = .598 (-217.15, 370.79) 239.60 69.57 3.44 p = .002 (97.88, 381.32) 327.20 163.17 2.01 p = .053 (-5.17, 659.57) 672.82 267.94 2.51 p = .017 (127.03, 1218.62) 1534.68 639.20 2.40 p = .022 (232.64, 2836.72) 119.84 78.72 1.52 p = .137 (-40.31, 279.99) -6.08 9.45 -.64 p = .524 (-25.31, 13.14) 221.27 316.30 .70 p = .489 (-422.25, 864.80) -25.98 38.78 -.67 p = .507 (-104.88, 52.91) Team diversity on country of origin * Risk
attitude team - risk averse
Constant
Risk perception (centred)
Risk attitude team - risk averse (dummy)
Risk perception * Risk attitude team - risk averse
Note. R2 = .229 for team diversity on gender; R2 = .835 for team diversity on country of origin; R2 = .304 risk perception of a team of decision makers
Team diversity on gender (centred)
Risk attitude team - risk averse (dummy)
Team diversity on gender * Risk attitude team - risk averse
Constant
Team diversity on country of origin (centred)
Risk attitude team - risk averse (dummy) Team diversity on gender
Team diversity on country of origin
Risk perception of a team of decision makers Constant
Control variables
ANOVA and Kruskal-Wallis test was applied to analyse whether there was a difference between the teams within the control variables, on the dependent variable – variance of orders placed by a team. Table 5 shows that both time of participation and pairs of game leaders had no statistically significant influence on the variance of orders placed by a team (Appendix L).
Table 5: Results ANOVA and Kruskal-Wallis test control variables
Conclusion & discussion
In this experimental research, the Beer Game (i.e. a simulation game representing the supply chain of beer) was used to analyse the impact of risk attitude of decision making teams on the variance of orders placed by a team (i.e. the bullwhip effect). The study also tested to what extent team diversity – either on gender or country of origin – and risk perception of decision making teams had an influence on that relationship. In addition, this research assessed whether a set of control variables – time of participation in the Beer Game, the pairs of game leaders and experience with the Beer Game – confounded these effects.
Empirical support was found for the first hypothesis, therefore it can be concluded that risk attitude of decision making teams does have an influence on the bullwhip effect. The variance of orders placed by risk averse teams was on average 390 more than the variance of orders placed by risk seeking teams, meaning that risk averse decision making teams created a larger bullwhip effect than risk seeking decision making teams. The bullwhip effect present in supply chains including risk seeking decision makers showed a similar pattern as Sterman (1989). Risk averse decision making teams showed an opposite pattern, compared to previous studies (Croson & Donohue, 2006; Qorbani, 2017; Sterman, 1989). The difference between the two risk attitude decision making teams was especially large upstream the supply chain (i.e. distributor and factory), while at the consumer side of the supply chain the risk averse retailer even had, on average, a lower variance in orders placed than a risk seeking retailer. This caused the observed bullwhip effect, present in risk averse decision making teams, to deviate from previous studies. Furthermore, the current study was the first to apply expected utility theory, as a means of measuring the risk attitude of decision makers, to research possible behavioural factors causing the bullwhip effect. The concept appeared to be useful since it showed the importance of risk attitude as a behavioural factor. Nevertheless, prospect theory as a means of
ANOVA Kruskal-Wallis test
Time of participation F(3, 33) = 1.17, p = .336 X2(3) = 5.57, p = .134 Pairs of game leaders F(9, 27) = 1,18, p = .344 X2(9) = 10.73, p = .295
