University of Groningen Not lean by default Ziengs, Nick

(1)

Not lean by default Ziengs, Nick

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.

Document Version

Publisher's PDF, also known as Version of record

Publication date: 2018

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Ziengs, N. (2018). Not lean by default: Exploring practices, their design, and underlying mechanisms driving performance. University of Groningen, SOM research school.

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

(2)

517640-L-bw-Ziengs-SOM 517640-L-bw-Ziengs-SOM 517640-L-bw-Ziengs-SOM 517640-L-bw-Ziengs-SOM Processed on: 14-3-2018 Processed on: 14-3-2018 Processed on: 14-3-2018

Processed on: 14-3-2018 PDF page: 121PDF page: 121PDF page: 121PDF page: 121

113

Chapter 4

Motivational Mechanisms in

Work-In-Progress Restricted Production

Systems: An Experimental Study

4.1 Introduction

Manufacturers often use pull production systems to control the flow of orders on the shop floor (González-R et al., 2012; Kumar & Panneerselvam, 2007; Lage Junior & Godinho Filho, 2010). These systems restrict work-in-progress on the shop floor (Hopp & Spearman, 2004, 2008). A work-in-progress restriction is beneficial because it makes shop floor throughput times shorter and predictable. However, restricting work-in-progress is not without its downsides as it increases dependencies within the production system and makes the system as a whole more susceptible to increases in idle time, or coordination losses,

(3)

114

4

in the form of blocking and starving thereby reducing throughput (Conway et al., 1988; Hudson, McNamara, & Shaaban, 2014).

A number of studies suggest that these detrimental effects are overstated in production systems where workers are the primary determinant of processing times (Doerr, Mitchell, Klastorin, & Brown, 1996; Schultz et al., 1998; Schultz, Juran, & Boudreau, 1999). Individuals in work-in-progress restricted production systems have been shown to adjust their effort in order to prevent idle time (Schultz et al., 1998, 1999). Although this adjustment of effort has been demonstrated in prior studies, the mechanisms which drive such behavior are less well understood thereby making it difficult to design a production system which is able to exploit these behaviors. In this study, we explore these underlying mechanisms.

Social psychology provides two explanations for the adjustment of effort, or motivation gains, of individuals in work-in-progress restricted production systems (Hertel, Kerr, & Messé, 2000; Weber & Hertel, 2007). The first explanation is derived from social comparison theory (Festinger, 1954) which suggests that individuals working in close proximity tend to derive performance standards by observing others. Studies have shown that slower individuals increase their effort to match these standards (Hertel et al., 2000; Weber & Hertel, 2007). Restricting work-in-progress is argued to facilitate the development of these performance standards or productivity norms because work-in-progress restriction makes the relative speed of others more salient (Schultz et al., 1998, 1999). The second explanation is derived from

(4)

115

4

social indispensability theory (Hertel et al., 2000; Weber & Hertel, 2007) which suggests that individuals working in groups tend to adjust their effort based on the importance of their effort to the group outcome. Those who believe their effort to be dispensable tend to decrease their effort (Karau & Williams, 1993), whereas those who believe their effort to be indispensable tend to increase their effort (Hertel et al., 2000; Weber & Hertel, 2007). Restricting work-in-progress increases interdependencies thereby making a slower individual’s contribution to the group outcome more pronounced.

Social comparison and social indispensability are supposed to be additive, although it is recognized that the degree to which the two mechanisms drive the adjustment of effort depends both on context as well as individual differences (Hertel et al., 2000; Weber & Hertel, 2007). Unlike social comparison, social indispensability requires task or outcome dependencies between individuals. Therefore, increased effort, or motivation gains, resulting from social indispensability and its facilitation may come at the expense of increased idle time or coordination losses. As such, there is a potential trade-off between motivation gains and coordination losses. Previous literature provides limited means to navigate this trade-off.

In this study, we aim to identify which mechanism – social comparison or social indispensability – drives the adjustment behavior in work-in-progress restricted production systems observed in previous studies (Doerr, Freed, Mitchell, Schriesheim, & Zhou, 2004; Doerr et al., 1996; Schultz et al., 1998, 1999). Additionally, we explore the

(5)

116

4

consequences of different types of coordination losses – caused by blocking or starving – on the adjustment behavior of individuals. To date, we have a limited understanding of the role coordination losses and motivation gains play when it comes to performance in work-in-progress restricted production systems.

To address these aims, we conducted a computerized experiment in which participants worked on the assembly of printed circuit boards. The experiment allowed us to distinguish between the effects of social comparison and social indispensability and their presupposed triggers, namely blocking and starving.

This chapter makes the following contributions. First, we study the adjustment of effort previously observed in production (Schultz et al., 1998, 1999; Schultz, Schoenherr, & Nembhard, 2010) and service systems (Hopp, Iravani, & Yuen, 2007; Shunko, Niederhoff, & Rosokha, 2017; Tan & Netessine, 2014) from a novel perspective, namely social comparison and social indispensability which allows us to provide recommendations which take both coordination losses and motivation gains into account. Second, unlike other studies which address social comparison and social indispensability, we differentiate between different types of dependencies. Third, we add to the existing literature on state-dependent behavior which explores how the productivity of individuals is affected by the state of the system workers find themselves in (see Delasay, Ingolfsson, Kolfal, & Schultz, 2015 for a review). The study of state-dependent behavior facilitates the development of more sophisticated mathematical and simulation

(6)

117

4

models which take behaviour of individuals into account (see also Bertrand & Van Ooijen, 2002; Heimbach, Grahl, & Rothlauf, 2012; Öner-Közen, Minner, & Steinthaler, 2017; Powell & Schultz, 2004 for examples of studies which integrated state-dependent behaviour in simulation models).

The remainder of this chapter is organized as follows. In section 4.2, we discuss motivation and coordination gains and losses. In addition, we discuss how motivation and coordination gains and losses are affected by the design of work-in-progress restricted systems. In section 4.3, we detail the design of our experiment. The results are presented in section 4.4. The implications for theory and practice are presented in section 4.5. Section 4.6 concludes and provides suggestions for future research.

4.2 Background

In recent years, the study of behavior in production and service systems has gained considerable traction (see Croson, Schultz, Siemsen, & Yeo, 2013 for an overview). The aim of these behavioral operations management studies is to provide guidelines for the design of operations which take human behavior into account (Gino & Pisano, 2008). In this chapter, we address the interdependencies in production systems and how they influence worker behavior. In the current section, we provide an overview of the literature on pull production systems and work-in-progress restriction, the adjustment behavior of individuals in work-in-progress restricted production systems, and the mechanisms

(7)

118

4

that drive this behavior. At the end, we conclude and list our expectations.

4.2.1 Work-in-progress restriction

Production planning and control systems are used to govern the flow of orders to and on the shop floor (González-R et al., 2012; Kumar & Panneerselvam, 2007; Lage Junior & Godinho Filho, 2010). These systems govern order acceptance, release, and dispatching (Land & Gaalman, 1996; Stevenson et al., 2005; Thürer, Stevenson, & Silva, 2016). Order acceptance refers, amongst others, to handling order selection, due date setting, and capacity adjustment. Order release refers to how and when orders are released to the shop floor. Order dispatching refers to how and when orders progress on from one resource to the next after they have been released to the shop floor.

Pull production systems control the release and dispatching of orders by restricting the amount of work-in-progress inventory that is allowed on the shop floor (Hopp & Spearman, 2004, 2008). How and to what degree pull production systems restrict work-in-progress is determined by their structure and configuration (Gaury et al., 2001). The structure refers to the number of work-in-progress restrictions (single, multiple), the number of resources (e.g. machines, operators) placed within the confines of each restriction (single, multiple), and the placement of each restriction relative to the other restrictions (non-overlapping, overlapping). The structure, therefore, also determines the interdependencies between resources on the shop floor. In the case of individuals working on the shop floor, the structure determines who

(8)

119

4

depends on whom. In card-based pull production systems, cards are used to control the amount of work within each of the work-in-progress restrictions. The configuration refers to the number of cards associated with each restriction. As such, the configuration determines the extent of the work-in-progress restriction. The configuration, therefore, determines to what degree individuals depend on each other. Together, the structure and configuration of the pull production system shape the nature of the dependencies between individuals on the shop floor.

Increased interdependencies between individuals on the shop floor increase the likelihood of individuals becoming idle (Conway et al., 1988; Hudson et al., 2014). High levels of work-in-progress acts as a buffer thereby safeguarding against the propagation of variability throughout the production system thereby reducing idle time or coordination losses (Hopp & Spearman, 2004, 2008). Idle time is caused by blocking or starving. Individuals who are unable to pass on work are blocked. Blocking occurs when downstream inventory buffers are at capacity. Individuals who are unable to commence work because no work is available are starved. Starving occurs when upstream inventory buffers are empty.

The following examples illustrate how the structure and configuration of a pull production system determine the interdependencies in a serial production line. Figure 4.1 shows a serial production line with separate work-in-progress restrictions. KANBAN (Lage Junior & Godinho Filho, 2010; Sugimori et al., 1977) is a well-known example of a pull production system that uses separate

(9)

work-in-517640-L-bw-Ziengs-SOM 517640-L-bw-Ziengs-SOM 517640-L-bw-Ziengs-SOM 517640-L-bw-Ziengs-SOM Processed on: 14-3-2018 Processed on: 14-3-2018 Processed on: 14-3-2018

120

4

progress restrictions. In the example, each combination of a workstation and inventory buffer is covered by a separate work-in-progress restriction (Gaury et al., 2001). The structure consists of three work-in-progress restrictions each restricted by a number of cards. If the upstream inventory buffer is empty, the downstream individual is starved of work. If the downstream inventory buffer is filled up to the restriction, as indicated by the number of cards, the upstream individual is blocked and unable to pass on work to the downstream inventory buffer. Both the release and dispatching of orders can be prohibited. The configuration, or the number of cards used to regulate work-in-progress within the restrictions, determines the likelihood that individuals will become either blocked or starved. The lower the number of cards, the more tightly coupled the individuals within the line, the higher the change that line imbalances will result in blocking and starving. Separate work-in-progress restrictions create sequential interdependencies where upstream individuals can starve downstream individuals by not working fast enough and downstream individuals can block upstream individuals by not working fast enough. In short, separate work-in-progress restrictions create tightly coupled production systems.

Figure 4.2 shows the same serial production line with a shared work-in-progress restriction. CONWIP (Prakash & Chin, 2015; Spearman et al., 1990) is a well-known example of a pull production system that uses a shared work-in-progress restriction. In the example, the structure consists of a single work-in-progress restriction which spans multiple workstations (Gaury et al., 2001). If the upstream

(10)

121

4

inventory buffer is empty, the downstream individual is starved of work. Blocking, however, is prohibited once an order has been released.

A shared work-in-progress restriction creates sequential

interdependencies between individuals where only upstream individuals are able to starve downstream individuals. Shared work-in-progress restrictions, therefore, create less tightly coupled production systems compared to separate work-in-progress restrictions.

Figure 4.3 shows the same serial production line with no work-in-progress restriction. A system with no work-work-in-progress restriction or work-in-progress restrictions in which the configuration does not restrict work-in-progress acts as a push production system (Hopp & Spearman, 2004). MRP (Orlicky, 1976) is a well-known example of a push production system which does not explicitly restrict progress on the shop floor. In the example, the system with no work-in-progress restriction allows work-in-work-in-progress to build up in the inventory buffers and blocking and starving will not occur provided that enough work-in-progress is available. Although sequential interdependencies are still present, these interdependencies between individuals will only materialize in the long term.

In general, modeling and simulation studies show that pull production systems with a shared work-in-progress restriction (e.g. CONWIP) outperform pull production systems with separate work-in-progress restrictions (e.g. KANBAN) in terms of throughput (Gaury, Pierreval, & Kleijnen, 2000). However, production systems with no

(11)

122

4

Figure 4.1. Three-stage production line with separate work-in-progress restrictions

Figure 4.2. Three-stage production line with a shared work-in-progress restriction

(12)

123

4

work-in-progress restriction tend to outperform systems with shared or separate work-in-progress restrictions in terms of throughput (Khojasteh-Ghamari, 2009). The primary benefits of pull production systems, however, are lower work-in-progress levels and shorter shop floor throughput times. Secondary benefits include, for example, reduced inventory cost, increased flexibility, and improved quality (Hopp & Spearman, 2004; Khojasteh-Ghamari, 2009). These benefits come at the expense of decreased throughput and increased idle time. To achieve the advantages and avoid the disadvantages associated with work-in-progress restriction, the structure and configuration are generally set to restrict work-in-progress at the lowest possible level without increasing the likelihood of idle time through the increased incidence of blocking and starving (Conway et al., 1988). As such, pull production systems are designed with the intent to avoid interdependencies thereby ensuring continuous flow with minimal work-in-progress. The intent to avoid interdependencies, however, does not take into account how workers act when these interdependencies materialize.

4.2.2 Work-in-progress restriction and adjustment behavior

A number of studies suggest that there is no need to avoid dependencies in production systems where workers are the primary determinants of processing times (Doerr et al., 2004, 1996, Schultz et al., 1998, 1999; Schultz, McClain, & Thomas, 2003). Doerr et al. (1996) were the first to demonstrate in a field experiment that individuals work faster when

(13)

124

4

working in a production system governed by a pull production system which restricts work-in-progress. Doerr et al. (1996) suggested that individuals worked faster because idle time created by work-in-progress restrictions allowed individuals to rest and recuperate. Schultz et al. (1998) replicated these results in a controlled experiment and provided a different explanation which suggests that work-in-progress restrictions facilitated the development of standards or productivity norms within a group of individuals by making the relative performance of individuals working in the production system more salient. The authors argued that relative performance becomes more salient because it is easier to derive productivity standards from a change in a few orders than a large number of orders. As such, production systems with a work-in-progress restriction provide more task-related feedback. In addition, Schultz et al. (1999) also argued that individuals are motivated not to violate these implicit productivity norms. Idle time can be avoided by adjusting their effort and increasing and decreasing their speed when appropriate. In later studies, Schultz et al. (2003) and Doerr et al. (2004) indeed demonstrated that the absence of information related to relative performance prohibits the development of these productivity norms. As a consequence, individuals did not adjust their effort appropriately when information related to relative performance was not available. Despite these advances, the mechanisms which drive the development of productivity norms in work-in-progress restricted production systems have not been carefully examined.

(14)

125

4

4.2.3 Social comparison and social indispensability

The literature on social psychology provides two explanations for the adjustment in the effort of individuals in work-in-progress restricted production systems (Hertel et al., 2000; Weber & Hertel, 2007).

The first explanation relies on social comparison (Festinger, 1954). Social comparison theory suggests that in the absence of explicit standards individuals tend to derive standards or productivity norms by observing others work. A large number of studies, synthesized in a meta-analysis (Weber & Hertel, 2007), have shown that slower individuals tend to adjust their effort upwards to match standards set by faster individuals. The upward adjustment of effort is referred to as motivation gain whereas the downward adjustment of effort is referred to as motivation loss. Slower individuals receive these standards by observing faster individuals. If the slower individuals believe the required adjustment in effort to be attainable, they will increase their effort in order to match the standards set by faster individuals (Hertel et al., 2000; Weber & Hertel, 2007). If the distance towards the standard is too large and seems unattainable or if the distance is too small and the individual believes the standard has already been attained, no upward adjustment of effort is likely to occur. A similar argument can be made for faster individuals who tend to adjust their effort downward when they derive performance standards from slower individuals (Suls, Martin, & Wheeler, 2002). The motives associated with social comparison are individualistic in nature. Individuals aim to maintain self-esteem and a positive self-image.

(15)

126

4

Social comparison requires a performance standard which can be obtained through the observation of others, but social comparison does not require task or outcome interdependencies. Studies have shown that individuals adjust their effort upward in additive as well as co-active tasks (Weber & Hertel, 2007). In additive group tasks, the group outcome is the sum of individual performance. As such there are no task dependencies, only outcome interdependencies. Additive tasks, however, are susceptible to social loafing and free riding because faster individuals can compensate for slower individuals (Steiner, 1972). The group outcome creates the outcome dependency. In co-active tasks individuals, independently work on a similar task in close proximity to others and, due to a lack of task and outcome dependency, faster individuals cannot compensate for slower individuals (Seta, 1982).

The second explanation relies on social indispensability (Hertel et al., 2000; Weber & Hertel, 2007). Social indispensability theory suggests that individuals adjust their effort based on the importance of their effort to the group outcome. Social indispensability therefore requires outcome interdependence and is commonly observed in conjunctive tasks where the slowest individual determines group performance (Weber & Hertel, 2007). Individuals who believe their effort to be dispensable for the group outcome will tend to decrease their effort (Karau & Williams, 1993). Individuals who believe their effort to be indispensable, on the other hand, tend to increase their effort (Weber & Hertel, 2007). The indispensability of effort is predicted to result in increased effort. Motives are no longer individualistic, such as avoiding becoming idle, but also collectivistic such as preventing others

(16)

127

4

from becoming idle. Individuals are no longer just trying to outperform others, but are also attempting to improve the group outcome. The perceived indispensability of effort results in increased effort because the effort the individual exerts not only affects their own task performance and outcome, but also the outcome and task performance of their group members. Studies have shown that weaker or slower individuals adjust their effort upwards in conjunctive tasks where group performance depends on the weakest or slowest individual (Steiner, 1972). The adjustment of effort can be explained by instrumentality x value models of motivation (Karau & Williams, 1993; Van Eerde & Thierry, 1996; Vroom, 1964) which suggest that motivation requires a valued goal and the expectation that the valued goal can be achieved. Table 4.1 provides an overview of task and outcome interdependencies in additive, coactive, and conjunctive tasks and whether social loafing, social comparison, or social indispensability can occur.

Table 4.1. Social loafing, social comparison, and social indispensability in additive, coactive, and conjunctive tasks

Task structure Social loafing Social

comparison Social indispensability Additive X X Coactive X Conjunctive X X

(17)

128

4

4.2.4 Social comparison, social indispensability, and work-in-progress restriction

Social comparison and social indispensability theories explain the adjustment behavior of individuals in work-in-progress restricted production systems observed in earlier studies (Doerr et al., 2004, 1996, Schultz et al., 1998, 1999, 2003). However, as stated, these earlier studies do not distinguish between social comparison and social indispensability. The distinction is important because social comparison can be facilitated by merely providing performance standards, whereas social indispensability requires interdependencies thereby risking coordination losses. Earlier studies on social comparison and social indispensability (Hertel et al., 2000; Wittchen, Schlereth, & Hertel, 2007) do make a distinction between social comparison and indispensability. However, these studies neglect the trade-off between motivation gains and coordination losses that is likely to occur in work-in-progress restricted production systems caused by, for instance, blocking and starving. In this study, we address these omissions by bringing these two streams of literature together.

To do so, we compare individual performance in conjunctive production systems to individual performance in a coactive production system. The conjunctive production systems allow us to evaluate the relative importance of the social indispensability explanation. The coactive production system allows us to evaluate the importance of the social comparison explanation. The coactive production system consists of three independent workstations. For the conjunctive production systems, three serial production lines are used, each with a

(18)

129

4

different structure and comparable configuration, similar as depicted in Figure 4.1, Figure 4.2, and Figure 4.3.

The first production line has no work-in-progress restriction (see Figure 4.3), the second has a shared work-in-progress restriction (see Figure 4.2), and the third has separate work-in-progress restrictions for each combination of a workstation and an inventory buffer (see Figure 4.1). The different conjunctive production systems allow us to evaluate different types of coordination losses, namely blocking and starving. Although the possibility of coordination losses should facilitate motivation gains, if the coordination losses are too extensive, assessment of relative performance becomes more difficult thereby making it more difficult to establish performance standards. As such, coordination losses might obscure the contribution of individuals to the group outcome thereby making the contribution of individuals more dispensable. Finally, outcome interdependence, present in sequential production lines, emphasizes the contribution of the last individual in the line, thereby making the contribution of the other individuals in the line more dispensable which in turn might result in motivation losses (Wittchen et al., 2007) and not motivation gains as suggested in previous studies (Doerr et al., 2004; Schultz et al., 1998, 1999, 2003).

Based on the previous discussion, we expect the following. First, due to social comparison processes, we expect slower individuals to show increased effort or motivation gain in coactive tasks due to the development of productivity norms. Faster individuals are expected to maintain or decrease their effort to match the norm. Second, we expect

(19)

130

4

individuals to show an additional increase in effort or motivation gain in conjunctive tasks due to social indispensability processes unless the sequential nature of the serial production line results in motivation loss. In that case, the increased effort of slower individuals caused by the conjunctive demands resulting from work-in-progress restrictions could potentially mitigate these coordination losses to a degree. In that case we expect more tightly coupled production lines with separate work-in-progress restrictions to be most beneficial. Figure 4.4 summarizes the relationships under investigation.

Figure 4.4. Social comparison and social indispensability in work-in-progress restricted production systems

(20)

131

4

4.3 Methodology

To explore the interplay between motivation gains and coordination losses, we used a controlled experiment which allows us to create different conditions in which the mechanisms, described in the previous section, can be isolated. An experiment is particularly suitable to study the mechanisms involved because it provides us with a controlled environment in which we can rule out other potential causes (Shadish, Cook, & Campbell, 2002). In the following sections, the sample, design, task, measures, and procedures are discussed.

4.3.1 Sample

418 undergraduate business administration students participated in the experiment. The students were enrolled in two different Bachelor level introductory operations management courses, each of which was taught

at a different large Dutch university. 250 male (Mage = 19.472 years,

SDage = 1.047 years) and 168 female (Mage = 20.851 years, SDage =

2.043 years) students participated in the experiment. The students participated in partial fulfillment of course requirements and for a chance to earn a monetary reward.

4.3.2 Design

Table 4.2 provides an overview of the experimental design. The experiment employed a 2 (i.e. individual baseline; experimental) by 5 (i.e. individual, coactive, conjunctive with no work-in-progress restriction, conjunctive with separate work-in-progress restrictions for

(21)

132

4

individual buffers, conjunctive with a shared work-in-progress restriction for the production line as a whole) experimental design. The first factor was within-subjects and the second factor was between subjects. Participants participated in an individual condition before participating in one of five experimental conditions. In the first round, participants worked in isolation. In the second round, participants worked either in isolation (i.e. individual condition) or in one of the group settings (i.e. coactive, conjunctive with no work-in-progress restriction, conjunctive with a separate work-in-progress restriction, conjunctive with a shared work-in-progress restriction). The inclusion of the individual and coactive conditions allows us to vary the working condition to isolate the relative importance of the social comparison and social indispensability mechanisms.

4.3.3 Task

The experimental task was designed to closely resemble the assembly of a product. As production lines are often used to assemble products, the task itself (i.e. assembling products) and the context in which the task was conducted (i.e. within a production line) were recognizable to the participants. The task itself mimicked the assembly of printed circuit boards (PCBs). Printed circuit boards consist of a circuit board and a number of electronic components. The experimental task required participants to select a number of electronic components and position them correctly on a circuit board. The participants completed the tasks using a web interface. Figure 4.5 shows the interface.

(22)

133 T ab le 4. 2. E xp er im en tal an d ass ig ned co nd itio ns Ex peri m enta l co nd it io n s Ass ig ne d co nd it io n s F irst _roun d Seco nd ro un d P os it io n Wo rk -in -pro gre ss re st rict io n Av aila ble perf or m ance st and ard T as k inte r-depend ency O utc om e inte r-depend ency 1 1 In div id ua l In div id ua l No No No No No 2 2 In div id ua l C oac tiv e 1 No Yes No No 3 In div id ua l C oac tiv e 2 No Yes No No 4 In div id ua l C oac tiv e 3 No Yes No No 3 5 In div id ua l C on ju nc tiv e 1 No Yes No Yes 6 In div id ua l C on ju nc tiv e 2 No Yes No Yes 7 In div id ua l C on ju nc tiv e 3 No Yes No Yes 4 8 In div id ua l C on ju nc tiv e 1 Sh ar ed Yes Yes Yes 9 In div id ua l C on ju nc tiv e 2 Sh ar ed Yes Yes Yes 10 In div id ua l C on ju nc tiv e 3 Sh ar ed Yes Yes Yes 5 11 In div id ua l C on ju nc tiv e 1 Sep ar ate Yes Yes Yes 12 In div id ua l C on ju nc tiv e 2 Sep ar ate Yes Yes Yes 13 In div id ua l C on ju nc tiv e 3 Sep ar ate Yes Yes Yes

(23)

134

4

The circuit board consisted of 128 notches positioned on a 16 by 8 grid. The circuit board was displayed at the center of the computer screen (see Figure 4.5; Circuit board). The components had to be positioned on one or more of the notches in accordance with the specification which was presented as a schematic drawing positioned above the circuit board (see Figure 4.5; Schematic drawing). The drawing indicated which component to select from the component stores and where to place the component on the circuit board. The component stores were positioned on the left and right-hand side of the screen (see Figure 4.5; Component store).

Each schematic drawing showed a different combination of components. Although each schematic drawing requires participants to select and place 6 components on the circuit board, the type of component and their placement were different for each printed circuit board they assembled. The component stores held 6 different component types (i.e., Resistors - R; Transistors - T; LEDs - L; Input jacks - I; Processors -P; and Memory - M). Each component type had a different shape. The shape of the component determined the number of notches each component covered (R - 3; T - 2; L - 1; I - 4; P - 8; and M - 6). Each component type had 6 variations (e.g. R1, R2, R3, R4, R5, and R6). Each variation had a different color. Each component was also clearly identifiable by the combination of a letter and a number (e.g. Transistor variation 1 - T1; Resistor variation 2 - R2; or Processor variation 3 - P3) which matched the information provided in the drawing.

(24)

135

4

The participants had to place components on the circuit board by selecting them and dragging and dropping them on the appropriate notches. Incorrectly selected components could be removed from the circuit board by dragging and dropping them back into the component store. The incorrectly placed component could also be relocated by dragging and dropping them to the correct notch. To start working on a printed circuit board, the participants had to manually retrieve the empty circuit board by clicking on a 'retrieve' button (see Figure 4.5; Retrieve). Once the participants had finished the assembly, they had to manually deposit the completed printed circuit board by clicking on a 'deposit' button (see Figure 4.5; Deposit). After the printed circuit board was deposited, no additional changes could be made any more.

Individual condition. In the first round and the second round of the individual condition, the participants assembled the printed circuit boards by themselves in isolation. The participants were working on an individual task with no task and outcome interdependencies. In addition to the elements of the interface outlined above, the participants were also able to see their first name, the number of printed circuit boards assembled (total throughput - tt), the number of printed circuit boards assembled according to specification (throughput correct - tc), and a representation of the shop floor (see Figure 4.5 and Figure 4.6). There were no upstream and downstream inventory buffers in the individual condition. That is, participants were always able to retrieve and deposit printed circuit boards. The lack of upstream and downstream inventory

(25)

136

4

buffers prohibited blocking and starving from occurring and allowed participants to place the components on the circuit board at their own pace.

Coactive condition. In the coactive condition, the participants also had to assemble printed circuit boards by themselves, but they no longer worked in isolation. Participants were not dependent on other participants for the completion of their task. Although the participants were also working on an individual task, the participants in the coactive condition were able to observe the work of 2 other participants. Participants were provided with the first names of the 2 other participants in their group, the number of printed circuit boards assembled by the others participants (tt), and the number of printed circuit boards assembled according to specification (tc) (see Figure 4.7). The display allows participants to observe the real-time movement and placement of components by the two other participants assigned to the same group. There were no upstream and downstream inventory buffers and participants were always able to retrieve empty circuit boards and deposit completed circuit boards at any given time. Blocking and starving could not occur, which allowed participants to dictate their own pace.

Conjunctive conditions. In the conjunctive conditions, the participants assembled printed circuit boards together with their group members. Participants no longer worked by themselves but worked together with

(26)

137

4

two others on an assembly. Each group member was responsible for a distinctive part of the assembly. The progression of work was sequential. Finished assemblies were passed downstream. Similar to the coactive condition, the participants were able to observe each other’s work, but in the conjunctive conditions, each participant was also sequentially dependent on the next. The production system used resembled an unbalanced line (Hudson et al., 2014). As before, the participants retrieved subassemblies from their upstream inventory buffer and deposited subassemblies in their downstream inventory buffer. However, in the conjunctive conditions, the participants were sequentially dependent and work was passed on through the intermediate buffers from position 1 to position 2 and finally to position 3 within the line. The participants were shown their own performance as well as the performance of their group members. In addition, the participants were able to see the dependencies by observing the number of subassemblies in each of the inventory buffers before, between, and after each workstation (see Figure 4.8, Figure 4.9, and Figure 4.10).

There were three conjunctive conditions (i.e. conjunctive with no progress restriction, conjunctive with separate work-in-progress restrictions, conjunctive with a shared work-in-work-in-progress restriction for the production line as a whole). The conjunctive condition with no work-in-progress restriction, conjunctive with a separate work-in-progress restriction, and conjunctive with a shared work-in-progress restriction each represent a different structure and matching configuration.

(27)

138

4

Conjunctive condition with no work-in-progress restriction. In the conjunctive condition with no work-in-progress restriction, the inventory buffers were unrestricted (see Figure 4.8). The participants were always able to retrieve and deposit subassemblies in their respective up and downstream inventory buffers such that the intermediate buffers contained more work-in-progress that could be finished by the participants within the timeframe of the experiment. As such, participants were never blocked or starved.

Conjunctive condition with separate work-in-progress restrictions. In the conjunctive condition with a work-in-progress restriction for individual buffers and the conjunctive condition with a work-in-progress restriction for the production line as a whole, the inventory buffers were restricted. In the conjunctive condition with a separate work-in-progress restriction for individual buffers, each intermediate inventory buffer had its own restriction on work-in-progress (see Figure 4.9) and, therefore, resembled a KANBAN system (Sugimori et al., 1977). If the upstream inventory buffer was filled up to its limits, the upstream participant would be blocked and not able to deposit a circuit board. If the upstream inventory buffer was empty, the participant would starve. Both blocking and starving could occur.

Conjunctive condition with a shared work-in-progress restriction. In the conjunctive condition with a shared work-in-progress restriction for the production line as a whole, the two intermediate inventory buffers

(28)

139

4

were governed by a single limit (see Figure 4.10) and closely resembled a CONWIP system (Spearman et al., 1990). In this condition, only starving could occur.

Figure 4.5. Interface; Shop floor, schematic, circuit boards, and component stores

(29)

140

4

Figure 4.6. Shop floor; Individual condition

Figure 4.7. Shop floor; Coactive condition

Figure 4.8. Shop floor; Conjunctive condition; No Work-In-Progress restriction

Figure 4.9. Shop floor; Conjunctive condition; Separate work-in-progress restrictions (a maximum number of circuit boards per intermediate inventory buffer)

Figure 4.10. Shop floor; Conjunctive condition; Shared work-in-progress restriction (a maximum number of circuit boards for the line as a whole)

(30)

141

4

4.3.4 Measures

To assess the impact of production system design on motivation gains and coordination losses, we relied on three performance scores, namely (1) processing time (seconds per subassembly); (2) idle time (seconds between consecutive subassemblies); and (3) quality (average number of correctly placed components per subassembly). The performance scores in the first round (i.e. individual round) were compared with the performance scores in the second round (i.e. coactive, or conjunctive round) to assess whether participants showed signs of motivation gains and coordination losses compared to the first round when performance standards were present (i.e. coactive) or when both performance standards and task and outcome interdependencies were present (i.e. conjunctive).

Processing time refers to the average time the participant worked on a subassembly and is measured as the average time between the retrieval and finalization of the subassembly. A decrease in processing time compared to the first round indicates that the participant put in more effort to complete the subassembly and is taken as a sign of motivation gain, whereas an increase in processing time is seen as a sign of motivation loss.

Idle time refers to the average time the participant waited between subassemblies and is measured as the average time between the finalization of a subassembly and the retrieval of the next subassembly. An increase in idle time compared to the first round indicates that the participant was waiting and is considered to be a sign

(31)

142

4

of coordination loss. An increase in idle time is considered as a sign of coordination loss because increases in idle time are caused by increased incidence of blocking and starving due to variation in processing time. Quality refers to the average number of components that were placed correctly per subassembly. A decrease in quality compared to the first round suggests either a motivation loss or a redirection of effort. Individuals faced with increased demands often sacrifice quality in order to meet those demands (see for instance de Freitas, Costa, & Ferraz, 2017; Kc & Terwiesch, 2009; Tan & Netessine, 2014).

Processing time, idle time, and quality were also used as a randomization check by comparing the performance scores of the participants in the first round across assigned conditions.

Two measures were used as manipulation checks. First, a single item was used to assess whether changes to the task demands, from individual to coactive or conjunctive, resulted in additional task difficulty as perceived by the participants. Participants were asked to rate on a 7-point Likert scale to rate to which degree they agreed with the following statement: The task was difficult. The perceived difficulty should remain the same across the various conditions to rule out if task demands increased in the group settings. Second, a single item for perceived pressure was used to assess whether changes across conditions, from individual to coactive or conjunctive, resulted in an increase in perceived pressure. Participants were asked to rate on a 7-point Likert scale to which degree they agreed with the following statement: I felt pressured when working on the task. Perceived pressure

(32)

143

4

should increase due to additional task demands in the coactive and conjunctive conditions. These questions were asked after the first and second round. Similar as before, the differences between the answer in the first and the answer in the second round was used for the manipulation checks.

4.3.5 Procedures

Upon arrival, the participants were welcomed and showed to a waiting area. Three to five groups of three participants arrived simultaneously. The participants were randomly assigned to a group and a role (i.e. first, second, or third position within the production line). After participants were assigned to a group and a role, group members were asked to introduce themselves to each other. Groups were randomly assigned to one of the five experimental conditions.

The participants were informed that the focus of the study was to assess the performance implications of different working conditions. In addition, the participants were told that they would participate in two consecutive rounds, each lasting roughly 15 minutes. As is common in most behavioral operations management experiments, no deception was used (Katok, 2011). Participants were not, however, informed of the hypotheses being tested. After the introduction, the participants were asked to sign an informed consent form, switch off their phones, remove their watches, and proceed to answer a few demographic questions.

(33)

144

4

Next, the participants were asked to watch an instruction video in which the experimental task and the reward structure were explained. The reward scheme was based on a previous study (Hertel, Deter, & Konradt, 2003). The instruction video was shot at a printed circuit board manufacturer to ensure the relevance for the participants. The participants had a chance to earn up to 10 euros in each round. In the individual and coactive conditions, each correct subassembly had a value of 25 cents. In the conjunctive conditions, each correct finished assembly had a value of 25 cents. Individual rewards work better when participants are independent of others, whereas group rewards work better when participants are dependent on others (Mitchell & Silver, 1990). In addition, a bonus of 25 percent was awarded if the groups belonging to the same university outperformed the groups belonging to the other university. After the experiment, one in three participants was randomly selected to receive the reward based on their individual or group performance depending on the condition to which they were assigned. The instruction video was followed by a short five-minute practice round. The practice round allowed participants to familiarize themselves with the interface.

In the first round, all participants worked in isolation. The participants were informed that they would be working in isolation and that their performance would determine the outcome. Before the start of the first round, participants were reminded of the reward structure and were provided with a summary of the instructions. The first, individual, round followed.

(34)

145

4

In the second round, participants worked either in isolation (i.e. individual condition) or in a group setting (i.e. all other experimental conditions). In the individual condition, participants were informed that they would be working in isolation and that their own effort would determine their performance. In the coactive condition, the participants were told that they would be working alongside the other members of their group, but that they were not dependent on each other. Consequently, their own effort would also determine their performance. In the conjunctive conditions, the participants were told that they would be working alongside the other members of their group and that they would be dependent on each other. Consequently, the effort of the slowest group member would determine the outcome. After the instruction, participants were reminded of the reward structure and were provided with a summary of the instructions. The second round followed.

Both the first and the second round were followed by questions used to assess if the manipulation succeeded. After the final set of questions, the participants were thanked, debriefed, and informed if they were eligible for a reward and, if so, where to collect the reward.

4.4 Analysis

In the following section, the various checks used to ensure validity are presented. First, we assess learning effects that have occurred and correct the learning effects found. Assessment of learning effects is important because of the within-subjects’ design and associated

(35)

146

4

carryover effects. Second, we assess whether randomization was successful. Third, we assess whether our manipulation succeeded. The learning checks, randomization checks, and manipulation checks are used to ensure validity. The procedure used was based on a previous study which incorporated similar learning, randomization, and manipulation checks to ensure validity (Hertel et al., 2003).

4.4.1 Learning checks

To determine if learning or fatigue occurred, the performance scores of the first round were compared to the corresponding performance scores of the second round in the individual condition (experimental condition 1; see also Table 4.2). The individual condition consisted of two consecutive individual rounds where participants worked in isolation and the first round was identical to the second round. The individual condition was used to correct for learning or fatigue.

A series of dependent t-tests were conducted to evaluate whether or not learning took place. There was a statistically significant difference in processing time (t(34) = 9.467, p < .01) from the first

round (Mpt = 46.514, SDpt = 11.814) to the second round (Mpt = 38.457,

SDpt = 11.567). Participants were faster in the second round which

indicates a learning effect. There was also a statistically significant

difference in idle time (t(34) = 3.174, p < .01) from the first round (Mit

= 1.657, SDit = .674) to the second round (Mit = 1.314, SDit = .522).

Participants spent less time idling in the second round which also suggests a learning effect as the participants learned to more quickly retrieve new subassemblies. Finally, there was a statistically significant

(36)

147

4

difference in quality (t(34) = .256, p < .05) from the first round (Mq =

5.772, SDq = .351) to the second round (Mq = .588, SDq = .244). Quality

was higher in the second round which suggests that participants did not achieve the improvements in processing and idle time by sacrificing quality. The participants showed signs of improvement in all performance scores. Therefore, we concluded learning took place and correspondingly decided to correct all performance scores using the ratio of the first round to the second round.

To evaluate whether the same correction can be applied to all participants, we conducted a number of simple linear regressions. Simple linear regression indicated that processing time in the first round

does not predict the processing time difference scores (R2_{= .078, p >}

.10). Faster participants did not improve more than slower participants. Simple linear regression indicated that idle time in the first round does

not predict the idle time difference scores (R2_{= .0001, p > .10). Simple}

linear regression also indicated that quality in the first round does not

predict the quality difference scores (R2_{= .044, p > .10). That is, the}

size of the learning effects did not depend on the speed or initial skill level of the participant. Faster participants did not improve more than slower participants, more idle participants did not improve more than less idle participants, and more accurate participants did not improve more than less accurate participants. Therefore, the same correction for learning effects was applied to all participant across experimental conditions.

(37)

148

4

4.4.2 Randomization checks

To determine if randomization was successful, we compared the first round of the 13 assigned conditions participants were randomly assigned to (see Table 4.2). To determine if there were no differences in processing time, idle time, and quality performance scores in the first round, three one-way ANOVAs were conducted. No differences in performance scores were expected. There was no statistically significant difference in processing times between assigned conditions as determined by a one-way ANOVA (F(12,403) = 1.326, p > .10). There was also no statistically significant difference in idle time between assigned conditions as determined by a one-way ANOVA (F(12,403) = .701, p > .10). Finally, there was no statistically significant difference in quality between assigned conditions as determined by a one-way ANOVA (F(12,403) = 1.512, p > .10). In conclusion, no significant differences between the assigned conditions implies that randomization was successful.

4.4.3 Manipulation checks

To assess whether the manipulation was successful, two manipulation checks were conducted. Increased social comparison and increased indispensability were hypothesized to increase pressure to perform. Perceived difficulty should remain the same, however.

Similar as before, t-tests were used to determine whether perceived pressure and difficulty remained stable across rounds. There was a no statistically significant difference in perceived difficulty (t(34)

(38)

149

4

= -.329, p > .10) or perceived pressure (t(34) = .442, p > .10) in the individual condition. As such, no correction was required.

Although participants in the conjunctive conditions evaluated the perceived task difficulty higher than in the individual condition, a one-way ANOVA showed the differences in perceived task difficulty difference scores not to be significant (F(4, 412) = 1.830, p > .10). The task difficulty participants perceived in the different experimental groups were comparable which suggests that the conjunctive condition did not impose any additional burden on the participants. In conclusion, the first manipulation succeeded.

As expected, the participants in the coactive and conjunctive experimental conditions considered the pressure to be higher in the second round of the experiment. A one-way ANOVA showed a statistically significant difference for the perceived pressure difference scores (F(4, 412) = 4.575, p < .01). A Tukey post hoc test showed that the perceived pressure difference scores of participants in the individual

condition (Mpp = .08, SDpp = .759) was statistically significantly higher

than participants in the coactive condition (Mpp = -.91, SDpp = 1.786, p

<.01), conjunctive with no work-in-progress restriction (Mpp = -.85,

SDpp = 1.276, p > .10), conjunctive with separate work-in-progress

restrictions (Mpp = -1.19, SDpp = 1.679, p < .01), conjunctive with a

shared work-in-progress restriction (Mpp = -.92, SDpp = 1.570, p < .01).

There were no statistically significant differences between the coactive and conjunctive experimental conditions (p>.10). As expected, the

(39)

150

4

coactive and conjunctive task conditions increase perceived pressure. In conclusion, the second manipulation succeeded as well.

4.5 Results

In the following section, the results are presented. First, we present the average processing times, average idle times, and average quality scores observed in each of the five experimental conditions. Second, we present the average processing times, average idle times, and average quality scores observed in each of the five experimental conditions, partitioned according to the speed of the participants in the first round which allows us to assess whether the adjustment behavior of slower and faster individuals differs as expected.

4.5.1 Performance across experimental conditions

Table 4.3 provides an overview of the average performance scores for both rounds of the experiment. Figure 4.11, Figure 4.12, and Figure 4.13 provide an overview of the average performance difference scores. A number of paired contrasts were used to compare the performance scores of individuals in the first round to the corrected performance scores of individuals in the second round (see also 2003 for a similar procedure).

The average processing times of participants in the coactive condition were lower in the second round compared to the first round. The average processing times of participants in the conjunctive conditions were higher in the second round compared to the first round.

(40)

151

4

As shown in Table 4.3, the increase was significant for the conjunctive with no in-progress restriction, conjunctive with a shared work-in-progress restriction, but not for the conjunctive condition with a separate work-in-progress restriction even though the increase was the largest (see also Figure 4.11). The lack of significance might be explained by the increase in variation observed in the conjunctive condition with a separate work-in-progress restriction. Participants in the coactive condition demonstrated motivation gains, whereas participants in the conjunctive condition consistently showed motivation losses.

There were no statistically significant differences in the average idle times of participants in the coactive and conjunctive with no work-in-progress restriction conditions. There were statistically significant differences in the average idle times of participants in the conjunctive with separate work-in-progress restrictions and conjunctive with a shared work-in-progress restriction conditions. As expected, participants in the coactive and conjunctive with no work-in-progress restriction conditions were not subject to coordination losses, whereas participants in the remaining conjunctive conditions were subject to coordination losses (see also Figure 4.12).

There was a statistically significant difference in the average quality scores of participants in the coactive condition and a marginally significant difference in the average quality scores of participants in the conjunctive with no work-in-progress restriction conditions. There were no statistically significant differences in the average quality scores