Master Thesis Project

Master Thesis Project

Mimetic behaviour in relation to operational performance

University of Groningen Faculty of Economics and Business

MSc Supply Chain Management Master Thesis Project University of Groningen Advisor: Dr. ir. T. Bortolotti

Co-assessor: Dr. X. Tong Student: Rick Boersma – S3459349

Date: 24-6-2019

2

ABSTRACT

Lean Manufacturing seemingly led Toyota to superior operational performance. After which many organizations decided to also adopt Lean. However, researchers found that the results of Lean Manufacturing on operational performance varied. This difference has partly been explained using contingency theory, claiming that management approaches, such as Lean, are dependent on contextual factors. Researchers have been successful in finding contingencies explaining some of the differences. Yet, in their discussion some mention that the mimicry of others, which is proposed to increase when uncertainty increases, can also affect the performance of the copied practice. Furthermore, no distinction has been made in the types of uncertainty which would increase mimetic behaviour. Therefore, by means of the High-Performance Manufacturing dataset, this thesis tested how different types of uncertainty affect mimetic behaviour, and how Mimetic behaviour might affect the relationship between the implementation of lean management practices and operational performance. For this purpose, Multiple linear regressions were used. The results show that different types of uncertainty affect Mimetic behaviour differently. Furthermore, the relationship between Hard lean practices and operational performance is positively moderated by Mimetic behaviour, whilst Soft lean and operational performance is negatively moderated.

Keywords: Lean; Hard Lean; Soft Lean; Lean manufacturing; Manufacturing; Operational performance;

3

Table of Contents

4.4 The Measurement model ... 18

4.5 Measurement scales, constructs unidimensionality, -reliability and -validity ... 18

4.6 Statistical tests ... 24

5. RESULTS ... 25

5.1 The relation between Uncertainty and Mimetic behaviour ... 25

5.2 Mimetic behaviour as a moderator ... 27

6. DISCUSSION ... 30

6. 1 The effect of Uncertainty on Mimetic behaviour ... 30

6.2 The effect of Mimetic behaviour on the ‘Lean-operational performance relationship’ ... 32

6.3 Wholistic model perspective ... 33

7. CONCLUSION ... 34

8. LIMITATIONS & FUTURE RESEARCH ... 35

REFERENCES ... 37

APPENDIX ... 44

Appendix A ... 44

Appendix B ... 45

4

1. INTRODUCTION

The increasing competition in markets has heightened the interest of organizations for continuous improvement. One of the biggest industry success stories of improvement is that of Japanese car manufacturer Toyota. They developed a management approach called: Toyota Production System, which later grew into the widely known concept of Lean Manufacturing. The book “The Machine that Changed the World” by Womack et al. (1990) telling this story inspired numerous organizations to mimic Toyota. Nowadays, one of the possible methods to mimic competitors is the use of benchmarking (Huang, Gattiker, & Schroeder, 2010).

However, the result from Lean implementations are not always positive, with only two percent of Lean initiatives reaching their set out goals, and only 24 percent reporting to have received ‘significant results’ (Liker & Rother, 2011; Pay, 2008). Researchers have found that many of those differences are explained by contextual factors such as e.g. plant size, plant age (Shah & Ward, 2003), and national- and organizational culture (Bortolotti, Boscari, & Danese, 2015). Henceforth, supporting the view that management practices are not universally applicable, as described in contingency theory (Sousa & Voss, 2008). In the discussion of papers who link Lean management with operational performance, some also argue that mimetic behaviour of organizations might have affected their findings (Bortolotti et al., 2015).

The reason why contextual factors, and mutations in practices, might affect the results of such studies is that the samples which are used might be ‘diluted’ with organizations who have used ‘uncritical mimicry’ of others to adopt a management approach e.g. Lean management (Ketokivi & Schroeder, 2004). Authors propose that, ill motivated mimetic behaviour can lead to ‘inefficient adoption or a superficial and incomplete use of practices’, which in turn, might provide suboptimal results (Bortolotti et al., 2015). Thus, raising the question when Mimetic behaviour is uncritical or ill motivated.

This debate is most prevalent in Institutional theory, form which the the concept of Mimetic behaviour by organizations originates. The theory explains the increasing heterogeneity of organizational practices, structures and strategies within markets over time (DiMaggio & Powell, 1983; Haunschild & Miner, 1997). Within Institutional theory, two perspectives are adopted: the economic perspective, which views mimetic behaviour as an organizations’ desire for efficiency, and the social perspective, which views it as a desire for legitimacy. Oftentimes researchers view the desire for legitimacy as negative, and the economic desire as positive (Kauppi, 2013). Research shows that managers feel the pressure of both perspectives, where they want to obtain legitimacy for their practices, but also are expected to perform (Barreto & Baden-Fuller, 2006; Rogers, Purdy, Safayeni, & Duimering, 2007). Therefore, in this thesis both perspectives are viewed as complimentary, as opposed to mutually exclusive.

5 when practices are copied mutations are created between the imitators and imitated practice, due to a lack of information, causing different results (Nelson & Winter, 1982). Whilst others concede in the usefulness of Mimetic behaviour to find solutions to problems without incurring large financial costs (Cyert & March, 1963).

In both the social and the economic perspective of institutional theory, uncertainty is seen as the main driver of Mimetic behaviour (DiMaggio & Powell, 1983; Miemczyk, 2008). Because, when making a decision where uncertainty exists about the possible options and their outcomes, decision makers look for ‘good enough’ answers, within their own ‘immediate environment’, and consequently might copy others (Cyert & March, 1963).

As for the gaps in literature of interest to this thesis, no empirical research has been performed which could give context to the proposed effect of mimetic isomorphism on the results of papers linking management approaches with performance. Thus, forming the first gap. The second gap exists in the research claiming a positive relationship between uncertainty and mimetic behaviour, no distinctions are made between the types of uncertainty which organizations face.

In response, this thesis aims to reduce the aforementioned gaps in literature using statistical data gathered from the third round of the High-Performance Manufacturing survey. This survey includes data on 317 manufacturing plants, across 10 countries, from three different industries. In line with the aforementioned gaps in research the following research questions are formulated:

1. How is uncertainty related to Mimetic behaviour?

2. How is Mimetic behaviour related with operational performance?

With the answering of the research questions, this thesis contributes to literature by exploring the cause and effect of Mimetic behaviour for Manufacturing plants whether it attributes to the explanation of the variation in results gained from lean implementation.

6

2. LITERATURE REVIEW

2.1 Lean Manufacturing

As mentioned in the introduction, the roots of Lean Manufacturing lay within Toyota, best known for manufacturing cars (Schonberger, 2007). After the second world war Toyota was struggling and looking for ways to improve their efficiency. Toyota knew that an American worker produced much more than a Japanese worker and decided to investigate. Thus, Taiichi Ohno, an industrial engineer and businessman working for Toyota studied the American production system at Ford. He found that Ford relied heavily on high production volumes and large lot sizes. However, because the Japanese car market was characterized by a lower overall demand, from a wider range of products, he realized these practices would not be beneficial for Toyota. During Taiichi Ohno’s stay in America he also noticed the replenishment system used by supermarkets, which at the time were not common in Japan. These supermarkets would only produce upstream what was used downstream, in order to replenish stock. With this shift in thinking Taiichi Ohno returned to Japan and together with other engineers set out to create Toyota’s own production system, partly using this same principle.

The resulting production philosophy was termed the Toyota Production System (TPS), which contained multiple production methods. These methods were aimed at eliminating six types of wastes that Taiichi Ohno had formulated after observing the American production methods (Ohno, 1978).

The adoption of TPS created a large performance gap between the Toyota and other car manufacturers (Hines, Holweg, & Rich, 2004). In the following years a book was published named: ‘The Machine That Changed the World’ by Womack and Jones (1990), which translated the production methods. Although not the first to do so (e.g. Shingo, 1981), it introduced the concept to a much broader audience, describing the production methods as ‘Lean’. At first, the authors were concerned that readers would disregard the findings as something only applicable within Japan (Womack, Jones, & Roos, 1990). However, the books notoriety soon grew, and is now recognized as one of the most influential pieces of literature for the proliferation of Lean Manufacturing in the western world.

7 Hard and Soft Lean practices

This thesis uses the definition for Lean management as created by Shah & Ward, (2007): “Lean production is an integrated socio-technical system whose main objective is to eliminate waste by concurrently reducing or minimizing supplier, customer, and internal variability” In the first part of the definition Shah & Ward, (2007) talk about Lean management being a socio-technical system. This is best described by Hadid, Mansouri, & Gallear (2016), who state that “The technical system includes equipment, tools, techniques and processes, while the social system comprises people and relationships among them.” The technical aspect could be viewed as the hard side of lean, whereas the social aspect can be seen as the soft side. Lean management consists of multiple bundles of practices including: Just-In-Time management, Total Quality Management and Human Resource Management. Just-Just-In-Time management and Total Quality Management are often attributed to the Hard lean side, whereas Human Resource Management is attributed to Soft lean management (Bortolotti et al., 2015). As also found by Negrão, Godinho Filho, & Marodin (2017), in the existing literature Hard and Soft lean practices have been operationalized in a wide variety of ways. The distinction between hard and soft lean practices within this thesis are based both on the paper of Bortolotti et al. (2015) and statistical methods, which will be discussed further in the method section.

The relationship between Lean management and operational performance

Many studies provide evidence of the link between the adoption of management approaches and operational performance. Perhaps most famous, is the study of Shah & Ward (2003). In their study they also showed that 23 percent of the variance between operational performance could be explained by lean bundles, when controlling for the effects of plant size, plant age and unionization status. In subsequent years academics searched for further explanations regarding the gap in results between organizations implementing Lean. In these studies they found that: organizational culture, demand variability, the factory size and the implementation stage of lean effect the performance gained by lean implementation (Bortolotti et al., 2015; Negrão et al., 2017; Netland, 2016). Furthermore, in the study of Bortolotti et al., (2015) the differences between successful, and unsuccessful implementers of lean are analysed, showing that the adoption of Soft lean practices is key for high operational performance.

The effect of uncertainty on Lean implementation

8 2.2 Institutional theory

Institutional theory finds that as markets develop, more heterogeneity arises, in the structures, strategies, and processes characterizing organizations due to ‘institutional pressures’ (David L . Deephouse, 1996; DiMaggio & Powell, 1983). This phenomenon is also called ‘institutional isomorphism’. By developing the field of Institutional theory researchers try to identify and understand causes for heterogeneity and are using its insights to explain changes in organizational practices, structures and strategies. In the application of the theory within the Operations Management and

Supply Chain Management field of research, organizations are often seen as ‘institutions’ even though

the meaning is interpreted much broader in other fields of study, such as social sciences. Within Institutional theory, both social and the economic arguments are provided as explanations for the occurrence of isomorphism.

The social perspective of institutional isomorphism claim that isomorphism in a market can be explained by the desire of organizations to be seen as legitimate by their stakeholders (DiMaggio & Powell, 1983). In turn, this legitimacy is achieved through the imitation of other organizations. When imagining these stakeholders, both traditional- and societal stakeholders should be thought of, including board members, investors but also customers and non-profit organizations (Grewal and Dharwadkar 2002).

The economic perspective is more dependent on the desire of organizations to maximize their profit, and in doing so borrow practices from other organizations which they deem beneficial in this pursuit (Ketokivi & Schroeder, 2004). As mentioned in the introduction, research shows that managers feel the pressure of both perspectives, where they want to obtain legitimacy for their practices, but also are expected to perform (Barreto & Baden-Fuller, 2006; Rogers et al., 2007). Therefore, in this thesis both perspectives are viewed as complimentary, as opposed to contrary.

Furthermore, the social perspective of institutional theory provides three pressure for isomorphism, namely: Coercive-, Normative-, and Mimetic isomorphism (DiMaggio & Powell, 1983; Haunschild & Miner, 1997). Since this thesis only focusses on Mimetic behaviour, Coercive and Normative isomorphism will first briefly be discussed as to provide context on the social perspective of Institutional theory, moving on to a more in-depth analysis on the literature regarding Mimetic isomorphism comparing the similarities and differences between both the social and economic perspective of institutional theory concerning mimetic behaviour.

9 that powerful firms can use their power to make their partners adapt in ways that are of interest to them (Liu, Gu, Chen, Ke, & Wei, 2010).

Normative isomorphism comes from the professionalization of industries, where norms and standards are formulated to create meaning and consistency (Sherer, Meyerhoefer, & Peng, 2016). These norms create a degree of legitimacy by defining conditions and approaches of labour (Gopal & Gao, 2008). Norms and standards are often created by the employees of organizations/industries and evolve from the education they have received (DiMaggio & Powell, 1983). Extending on this is the literature written by researchers over the industry and other knowledge-based networks. Therefore, the academic world is also viewed as a strong source of normative pressure (Kauppi, 2013).

Mimetic isomorphism is the imitation of other organizations. As mentioned before, the social perspective of intuitional theory attributes mimicry to the desire for legitimacy (DiMaggio & Powell, 1983), whilst the economic perspective attributes such behaviour to a desire for economic benefit (Haunschild & Miner, 1997). The economic perspective divides mimetic isomorphism further into three categories. The first category is Frequency-based imitation, which occurs when the decision to copy is based on the high frequency of the copied aspect in the environment. The second category is

Trait-based imitation, which is the imitation of others based on their traits which are comparable or

different, such as an organizations size. The third category is Outcome-based imitation, which is the copying of organizations based on an outcome such as product quality, or delivery speed.

The cause of mimetic behaviour is in both perspectives stimulated by uncertainty (DiMaggio & Powell, 1983; Haunschild & Miner, 1997; Miemczyk, 2008). From the social perspective this uncertainty is more on how to achieve a goal with the current resources. The economic perspective focusses more on the contextual uncertainties, thus measurable in the context of the organization. In both cases more social comparison is used as input for the decisions-making process (Cyert & March, 1963). These solutions are often sought in the organizations proximity and should be ‘good enough’. A tool to make such comparisons is benchmarking, which is a process by which organizations compare themselves with the most successful in their market (Liu et al., 2010).

Further information on the benefits and impairments of mimetic behaviour are covered in the hypothesis section, such that repetition of information is minimized.

10 Using Mimetic behaviour as a moderator

Using the insights of the institutional theory literature review by Kauppi et al., (2013), it becomes clear that there are some divergences in the application of the theory. In the majority of articles institutional theory is seen as an antecedent, having a direct effect on a dependant variable. However, there are also papers that use institutional theory as a moderating effect on a relationship. Most notable is the paper of (Wu, Ding, & Chen, 2012), which looks at the implementation of Green Supply Chain Management practices. In this paper the authors find that ‘the moderating effects of institutional pressures between implementation of sustainable practices and green performance are conflicting, specifically between the different pressures’ (Kauppi, 2013). The different pressures articulated within the study are market-, competitive- and regulatory pressure (Wu et al.market-, 2012).

This thesis will test for a moderation effect of mimetic behaviour on the relationship between Hard- or Soft lean implementation on organizational performance. The reason being, literature explains that mimetic behaviour does not necessarily have a direct effect on either lean implementation or operational performance. Rather, the literature suggests that if practices are adopted using mimetic behaviour, this might have a negative effect on the results gained from lean implementation on operational performance (Nelson & Winter, 1982; Sousa & Voss, 2008).

2.3 Uncertainty

Research on uncertainty first started back in the sixties, within the organizational theory literature and the book of J.D. Thompson, (1967). The message of the book was that all organizations must devise ways of handling uncertainty. In its wake, more research started to dissect the different aspects of uncertainty.

Since uncertainty is a broad and complex concept, measurement proved to be quite difficult. Consequently, the discussion arisen whether to measure uncertainty as a ‘perceptual phenomenon’ or as a ‘property of organizational environments’ (Milliken, 1987). Some papers adopt the view that uncertainty is the ‘property of organizational environments, such as: Tomas, Hult, Craighead, & Ketchen (2010), whilst others see uncertainty as a perceptual phenomenon such as: Flynn, Koufteros, & Lu (2016). In the paper of Flynn et al. (2016), they argue that, since they want to know the influence of uncertainty on organizations behaviour the experienced uncertainty is more important, thus using a perceptual measure. Since this thesis has the same intent, their definition is seen to be fitting, more on this in the method chapter.

11 Micro-level uncertainty is defined by Flynn et al. (2016) as: “the variability in technical aspects of the core supply chain based on the variability of inputs to the technical core of a supply chain, corresponding to the traditional operationalization of uncertainty in the supply chain and operations management literature.” Another term used to describe this type of uncertainty is operational complexity as used by the paper of Sivadasan, Efstathiou, Frizelle, Shirazi, & Calinescu (2002), cited by Flynn et al (2016) to formulate the definition. In their paper they claim that when the operational complexity increases, the more information is needed to ‘monitor and manage that system’, which brings us to the next type of uncertainty.

Meso-level uncertainty is defined by Flynn et al., (2016) as: ‘is the lack of information needed by a supply chain member, corresponding to the information processing theory perspective.’ This definition shows that to operate, information is needed by supply chain members, but the absence of this information creates uncertainty. However, in a supply chain, not all members necessarily strive towards the same goals, or are trust other members enough act in their best interest, making information sharing difficult (Lee, Whang, & Hau L. Lee, 2000).

12

3. RESEARCH HYPOTHESES

In Figure 1, the conceptual model of this thesis project is presented. The upper half of the model displays how the concept of uncertainty, consisting of its three types (Flynn et al., 2016), has a proposed effect on Mimetic behaviour. Furthermore, the model shows a hypothesized moderation effect of mimetic behaviour on the relationship between Hard lean and operational performance and Soft lean and operational performance.

Figure 1

Conceptual model

3.1 Hypotheses formulation

The conceptual model consists of multiple constructs and relationships. In the following paragraph this model is divided into hypotheses, which are tested, the results are given in the ‘RESULTS’ chapter.

Effect of Uncertainty on Mimetic behaviour

As described in the paper of DiMaggio & Powell (1983), Haunschild & Miner (1997), and Miemczyk (2008), when uncertainty increases Mimetic isomorphism is more likely to occur. Thus, this suggests a positive relationship between Uncertainty and Mimetic behaviour.

13 Micro-, Meso- and Macro-level uncertainty differ from each other in the degree to which the information needed to reduce uncertainty is attainable/predictable (Flynn et al., 2016). Where Micro-level uncertainty could be explained through a distribution, e.g. demand variability, Meso-Micro-level uncertainty concentrates on the demand for information that is currently not attained, and Macro-level uncertainty is created in a setting where the contexts is so vague that the formulation of questions that would reduce uncertainty is challenging (Flynn et al., 2016). Therefore, the impact of the uncertainty types on Mimetic behaviour would logically increase from Micro- towards Macro-level uncertainty, due to its increasing difficulty to predict. However, assuming that mimetic behaviour is at least partly due to managers decisions, they would also consider the impact of the uncertainty if left unchecked (Zsidisin, Melnyk, & Ragatz, 2005). Since the perceived impact of the types of uncertainty could differ between manager and other contingency factors such as industry, some ambiguity remains in the degree to which each type of uncertainty would affect Mimetic behaviour relative to the other. Therefore, the hypotheses are formulated without differentiating in the size of effect between the uncertainty types. Thus, only the positive effects are hypothesized in accordance with the literature of DiMaggio & Powell (1983), Haunschild & Miner (1997) and Miemczyk (2008).

H2: Micro-level uncertainty has a positive influence on Mimetic behaviour. H3: Meso-level uncertainty has a positive influence on Mimetic behaviour. H4: Marco-level uncertainty, has a positive influence on Mimetic behaviour.

Benefits and impairments of Mimetic behaviour

As mentioned in the introduction, drawbacks of mimetic behaviour are claimed to include reduced operational performance. This is a consequence of practices being dependent on contextual factors, as outlined in contingency theory. Therefore, when organizations transfer practices which work outside of their context, they might see negative results (Sousa & Voss, 2008). Furthermore, in the process of imitation, practices can also mutate between the imitating and imitated organization, due to a lack of information on the practices. The mutation of the practice is seen a result of the imitator not having an open line of communication with the imitated, and the imitated not wanting to be copied. Thus, causing incomplete/incorrect information on the practice intended for copying. Subsequently, the mutations in the practice might cause unintended negative effects (Nelson & Winter, 1982).

14 facilitate practitioners in sharing knowledge on practices to achieve a greater understanding on how certain success are attained. Also, organizations gain valuable information on their position relative to their competitors.

Therefore, performance gains by organizations employing benchmarking cannot be solely attributed to copying, but rather, the better understanding of the strategic fit, between practices, market conditions and organizations. Consequently, when only looking at copying behaviour as a board concept occurring in practice, it is expected that the findings of this thesis reflects the realty, which seem to be rather negative, as documented by Pay (2008). Consequently, the following hypotheses are formulated:

H5: Mimetic behaviour negatively moderates the relationship between the implementation of Hard lean practices and operational performance.

H6: Mimetic behaviour negatively moderates the relationship between the implementation of Soft lean practices and operational performance.

Further differences in the benefits and impairments of Mimetic behaviour are often argued using late and early adopters of practices. However, since the measure in this thesis does not distinguished between early and late adopters no hypothesis can be formulated. Nevertheless, for the completeness of the literature review, these are also included.

15

4. METHOD

Research can only be considered to be good research if it is both reliable and valid. The reliability of research is dependent on whether it can be redone by another researcher, and they would find the same results. This relates to the manner in which data was collected and treated afterwards. The validity of research is judged on whether it has measured what it has intended to. This is done with various statistical test, but also theoretical argumentation (Karlsson, 2016). Therefore, this thesis will also discuss both its reliability and validity. Starting with the reliability aspects pertaining to the collection and treatment of the data in paragraphs 4.1, 4.2 and 4.3. Followed by an explanation of the validity of the research in paragraphs 4.4, 4.5 and 4.6, but also contributing to the reliability of the research by describing how scales were combined, and tests were performed.

4.1 Data

16

Table 1

Data Characteristics

Country Industry Total Electronics Transportation components Machinery

Austria 10 4 7 21 Italy 10 7 10 27 Spain 9 10 9 28 Sweden 7 7 19 24 Germany 9 19 13 41 Japan 10 13 12 35 Finland 14 10 6 30 South Korea 10 11 10 31 China 21 14 16 51 USA 9 9 11 29 Total 109 104 104 317

Table 2

Respondent Characteristics

Country Number of respondents per plant

Plant accounting manager 1

Direct labour 10

Human resource manager 1

Information system manager 1

Production control manager 1

Inventory manager 1

Member of product development team 1

17 4.2 Data cleaning

In order to determine whether a plant had implemented Lean manufacturing, this thesis focused on the following question within the survey: “Please circle the number that indicates your opinion about how your plant compares to its competition in your industry, on a global basis.” below this question were multiple practices which could be rated. Lean was also among the list of practices. When sorting the answers ascendingly, only two plants had not filled in this question. This was interpreted this as that these plants did not use Lean Manufacturing and should not be considered in the analysis. Furthermore, the other plants all had answers above one, which indicated that they at least adopted Lean Manufacturing to some degree. Additionally, due to a large number of missing values concentrated in a select number of plants, effecting mostly the scale for operational performance, it was decided to remove 29 plants form the dataset. This was found to be more appropriate than replacing the missing values with the mean of the series, since the missing values were highly concentrated in specific plants, and the number plants effected was relatively small (27/317). All 29 deleted plants are listed in Appendix A. All other missing values within the data have been replaced with the mean value of the series.

4.3 Data usage

All scales in this thesis are answered using a Likert-scale. This type of answer registration asks the respondent to fill in to which extent they agree with a statement. As a result, the findings of this thesis are based on perceptual data. In this thesis both 5-point and 7-point scales are used. This range is good, since lower than 5 would provide little distinction between categories, but higher than 7 makes it hard for a human to make meaningful distinction between the categories (Dawes, 2018). The lowest number always correspond to the most extreme form of disagreement in that specific scale, whereas the highest number corresponds to the opposite. In order to compare 5-point scales with 7-point scales, these variables have been turned in to standardized variables.

18 A drawback from using a 7 or 5 point Likert scale is that a neutral option is provided, which can be interpreted as an actual neutral answer, the respondents inability to formulate an answer, or the lack of knowledge possessed by the respondent to formulate answer (Chimi & Russell, 2009).

4.4 The Measurement model

The measurement model is based on prior research reported in the literature section and designed to answer both of the research questions. Therefore, it consists of two parts. The first part analyses the relationship between the different uncertainty types and mimetic behaviour. The second part of the model analyses how Mimetic behaviour influences the relationship between Hard and Soft lean practices and operational performance. The reason for positioning the variables, as shown in the conceptual model, is a consequence of the existing research and has been discussed in chapter 2. The next step is to validate the concepts themselves. The appropriateness of the scales measuring or forming each concept is again argued using the literature, whilst their validity is simultaneously analysed using statistical tests.

4.5 Measurement scales, constructs unidimensionality, -reliability and -validity

To elaborate, these tests analyse the construct validity, which is an umbrella term containing both

convergent validity and discriminant validity. The assessment of convergent validity looks if each item

in a scale converge to measure one construct. Within this thesis, a Cronbach’s Alpha test is used for this purpose and reported where necessary. Furthermore, discriminant validity refers to the degree to which the scales/items of one concept are distinct, and thus, not measuring other latent constructs. To guarantee discriminant validity an Exploratory Factor Analysis (EFA) will be performed, using the varimax rotation method. The loading of all factors should be higher than 0.500 (J. C. . Anderson & Gerbing, 1998). Also, the cross-loading of items on factors should not be too high, having at least a difference of 0.2. The performed test will be elaborated upon per measurement scale, in the following paragraphs.

4.5.1 Operational performance measurement

19 When testing for discriminant validity using an EFA, each item measured the same latent construct. Therefore, it is assumed that all items should be combined without first combining the item representing one competitive priority, as can be seen in the computation below table 3. Only unit cost of manufacturing falls below the 0.5 threshold. However, since this is a formative measure, no further action was taken.

By combining the separate items into one scale for operational performance, an organization performing well on all items would have the highest operational performance. This way of measuring performance is congruent with the idea of an efficiency frontier, where excelling in one dimension is seen to having a trade-off in other dimensions (Baik, Chae, Choi, & Farber, 2010; Eccles & Serafeim, 2015).

Table 3

Operational performance scale

Factor loading (Mean, S.D.)

PER1 Unit cost of manufacturing .414 (3.2606, 0.88880)

PER2 Quality conformance .528 (3.8746, 0.68707)

PER3 On time delivery performance .788 (3.8258, 0.84575)

PER4 Fast delivery .733 (3.7369, 0.82948)

PER5 Flexibility to change product mix .670 (3.8491, 0.75711)

PER6 Flexibility to change volume .740 (3.7917, 0.80395)

The items were combined as follows:

20 4.5.2 Uncertainty measurement

As mentioned before, uncertainty could be measured both as the volatility in activities, or as a perceptual measure. In this thesis perceptual measures are used, since action of organizations are not necessarily rational, but also based on their perception of a situation (Flynn et al., 2016). In the article of DiMaggio & Powell (1983), multiple sources of uncertainty are given which induce mimetic behaviour. Namely: ‘When organizational technologies are poorly understood (March and Olsen, 1976), when goals are ambiguous, or when the environment creates symbolic uncertainty’. When comparing these with the measures from Flynn et al., (2016): Micro-, Meso-, and Macro-level uncertainty, in that order, they seem to be very similar.

The measurement of uncertainty was adopted from the paper of Flynn et al. (2016), published in the Journal of Supply Chain Management. To guarantee the convergent validity an Exploratory Factor Analysis was performed. The analysis showed low Cronbach’s Alpha’s values for both Meso- and Macro-level uncertainty, 0.293 and 0.461, respectively. In contrast, the Cronbach’s Alpha’s for Micro-level uncertainty was high, with a value of 0.855.

Furthermore, to guarantee the discriminant validity an EFA was performed. The analysis showed considerable cross loading of the items measuring Meso-uncertainty with the items of other scales. In order to increase the Cronbach’s Alpha of Meso-level uncertainty, and reduce cross-loading, only one original item from the scale was kept, and another was added. The items which remained was: ‘Manufacturing management is not aware of our business strategy’, the added item was: ‘We need better accuracy in our demand forecast’. As can be seen from these statements, both items measure some form of uncertainty that could be reduced by communication between internal and external members of the supply chain. Which is in line with the definition given to Meso-level uncertainty by Flynn et al., (2016). Since the new scale Meso-level uncertainty consists of two items, each representing a dimension of the type of uncertainty no Cronbach’s Alpha test was performed.

The same applies to the Macro-level uncertainty which measure the uncertainty of the market, but seemingly in different situation, again making the Cronbach’s Alpha not a very good indicator if the scale is appropriate. It is assumed that Flynn et al., (2016) came to the same conclusion and therefore did not report it in their study either. From the Factor loading however, it can be seen that all items are loading above 0.685 per scale. The indication of cross-loading in the model was suppressed up to the 0.3 level. After running the analysis, no cross-loading appeared model, which indicates unidimensionality of all the scales. Therefore, due to the selected items being in line with the literature, and Factor loading conforming that each scale measures one concept, the scales are used in this thesis.

21

Table 4

Uncertainty scales

In the event that all types of uncertainty have the same direction of effect, the following computation will be made of the overall uncertainty measure:

((𝑀𝐼𝐶1 + 𝑀𝐼𝐶2)/2) + ((𝑀𝐸𝑆1 + 𝑀𝐸𝑆2)/2) + ((𝑀𝐴𝐶1 + 𝑀𝐴𝐶2 + 𝑀𝐴𝐶3)/3) 3

Item Statements Factor Loading Cronbach’s Alpha

Micro-level uncertainty:

MIC1 (reversed) Manufacturing demands are stable in our firm.

0.912 0.855

MIC2 (reversed) Our total demand, across all products, is relatively stable.

0.924 0.855

Meso-level uncertainty:

MES1 We need better accuracy in our demand forecast.

0.813 -

MES2 Manufacturing management is not aware if our business strategy.

0.813 -

Macro-level uncertainty:

MAC1 Our competitive pressures are extremely high.

0.701 -

MAC2 Competitive moves in our market are slow and deliberate, with long time gaps between different companies’ reactions.

0.689 -

MAC3 The needs and wants of our customers are changing fast.

22 4.5.3 Mimetic behaviour measurement

Since Mimetic behaviour has been used in other studies, there are no measurement scales to be adopted from prior research. Therefore, in this thesis, Mimetic behaviour is measured with items that are perceived to be in line with the existing literature on Mimetic behaviour.

Congruent with the literature on mimetic behaviour, two items were selected. The selection of items was based on the research of Sila (2007). In their study they suggest that organizations first adopt practices which are expected of them by their constituency, and after further development start to imitate practices from industry leaders.

As such, the first item measures the dimension of Mimetic behaviour where the imitator feels that it is required to copy others behaviour, which is tested with the following question: ‘When our competitors make a move, we are forced to quickly follow it’. One could argue that this seems to be similar with the coercive pressure of Institutional theory. However, this is arguably not true, because coercive pressures rely on their being an entity that enforces those practices, instead of it being a consequence of the desire to stay competitive in the market.

The second dimension of mimetic behaviour is where the imitator is self-motivated in its copying of others. This is measured with: ‘We try to copy our competitors’. Since these two items sufficiently load onto one latent construct but measure two different dimensions no further testing has been applied. The scale, containing its items, and the respondent per item is shown in table 5.

As mentioned before, using the literature from Barreto & Baden-Fuller (2006), it is gathered that managers feel the pressure of both perspectives. Meaning that the perspectives of Institutional theory are not necessarily competing but can be seen as complementary. Furthermore, both perspectives are seen to be used in organizations, where the efficiency argument was more common for the early adopters, and the legitimacy argument more for the late adopters (Westphal et al., 2006).

Table 5

Mimetic behaviour scale

Mean (SD) Respondents MIM1 When our competitors make a move, we are

forced to follow it

0.768 (3.9996, 0.8754) PE, PM, PS

23 4.5.4 Hard and Soft lean implementation measurements

For the measurement of Lean implementation, the idea is adopted that Lean management is a social-technical system, comprising of both hard and soft lean practices (Shah & Ward, 2007). Therefore, in this thesis, the implementation of lean is divided onto, both Hard and Soft Lean implementation, as done by Bortolotti et al. (2015).

However, some additional practices have been added, and others removed after Factor reduction method. This method insured that all practices within each bundle is not measuring both the implementation of Hard-, and for Soft lean implementation. The latent construct 1, as can be seen in table 6 is considered to be the measurement of Human Resource Management practices. The latent constructs 2 and 3 are considered to be Just-In-Time management and Total Quality Management, respectively. The maximum amount of cross-loading in the resulting variables was lower than 0,30. In table 12, the items corresponding to each scale their factor loadings, and their Cronbach’s Alpha and other descriptive statistics are provided. Because of its length this table can be found in Appendix B.

In table 12, none of the Cronbach’s Alpha’s were lower than 0.741. The EFA showed that for many practices, the items measuring it had factor loading higher than 0.7. However, some items loaded a bit lower, with the lowest being 0.564.

Table 6

Configuration of Hard and Soft lean practices factor loading

Latent construct

Practice 1 2 3

Hard lean practices

HD1 Equipment layout for continuous flow JIT .317 .716

HD2 Just in time delivery by suppliers JIT .830

HD3 Kanban JIT .745

HD4 Setup reduction JIT .741

HD5 Maintenance support TQM .740

HD6 Cleanliness and organization TQM .330 .516

HD7 Feedback TQM .784

Soft lean practices

SF1 Top Management Leadership for Quality: HRM .559 .304

SF2 Small Group Problem Solving: HRM .713

SF3 Continuous Improvement and Learning HRM .755

SF4 Multi-Functional Employees: HRM .690

24 4.6 Statistical tests

Descriptive statistics and independent sample t-test

Since prior literature has stated the positive effect of uncertainty on Mimetic behaviour, this thesis will show the mean degree of Mimetic behaviour in low, medium and high uncertainty groups for each type of uncertainty. Thus, the increase in mimetic behaviour can be visualized. To accomplish this, the experienced degree of e.g. Micro-level uncertainty is sorted ascendingly. From the in total 288 plants, the plants ranking from 1-96 will be considered the low uncertainty group, the plants from 97-192 are considered as the medium uncertainty group, and the 193-288 are considered to be the high uncertainty group. This was also done for the other two types of uncertainty.

Furthermore, in order to analyse whether the average degree of Mimetic behaviour is significantly different between the low and high category of uncertainty, an independent sample t-test will be performed. The results are shown in the next chapter in table 7.

In the following paragraphs Multiple linear regressions are used. In order to control for multicollinearity a VIF statistic was run, indicating no value above the 4.0 threshold, thus no multicollinearity was assumed (Miles & Shevlin, 2001). Normality of the data was assumed by testing the skewness and kurtosis, both stayed within the 0,5 to -0,5 and 2 to -2 threshold, respectively (George & Mallery, 2003). These values can be found in Appendix C, table 14.

Multiple linear regression

To further investigate the relationship between the uncertainty types and Mimetic behaviour, this thesis aims to measure their effect and overall predicting value. These desires are met using a linear regression. Since uncertainty in measured using multiple uncertainty types (Micro-, Meso- and Macro-level uncertainty), the Multiple linear regression option was used in SPSS, which provides the possibility of selecting multiple independent variables for prediction. The results of the statistical test can be viewed in table 8.

Multiple linear regression with interaction terms

25

5. RESULTS

5.1 The relation between Uncertainty and Mimetic behaviour

The first part of this research sets out to answer the question how uncertainty and Mimetic behaviour are related. In practice this meant finding out whether when a high/low degree of Mimetic behaviour was measured, this also corresponds to a high/low overall level of uncertainty/uncertainty types. This shown using a mean comparison between the degree of Mimetic behaviour in the various uncertainty categories shown in table 7.

Table 7

Descriptive statistics of Mimetic behaviour across Uncertainty categories

1.00 (low) 3.7239 96 .64334

2.00 (medium) 4.0420 96 .64821

3.00 (High) 4.6389 96 .71508

Total 4.1349 288 .76791

Micro-level uncertainty categories

1.00 (low) 3.6567 96 .61925

2.00 (medium) 4.0099 96 .58532

3.00 (High) 4.7382 96 .66395

Total 4.1349 288 .76791

Meso-level uncertainty categories

1.00 (low) 4.2028 96 .77165

2.00 (medium) 4.0995 96 .76426

3.00 (High) 4.1025 96 .77134

Total 4.1349 288 .76791

Macro-level uncertainty categories

1.00 (low) 3.9207 96 . 76429

2.00 (medium) 4.1474 96 . 73152

3.00 (High) 4.3367 96 . 75800

Total 4.1349 288 . 76791

Distribution of categories:

26 of Mimetic behaviour. The highest group has the lowest degree of Mimetic behaviour. The middle group is in-between. Thus, being reversed compared to the other types.

Findings independent sample t-test: equal variances can be assumed for all Uncertainty measures since the Levene’s test is not significant in any of the tests. For Micro-, and Macro-, and overall Uncertainty the group with the lowest uncertainty score (group 1) has a significant lower score on Mimetic behaviour than the group with the highest uncertainty (group 3) (p<.001). Again, for Meso-level uncertainty this was the opposite, where the group with the lowest uncertainty score (group 1) has a significant higher score on Mimetic behaviour than the group with the highest uncertainty (group 3) (p<.05).

Multiple regression analysis:

In the figure below, the effect of each uncertainty type on mimetic behaviour is shown, along with the predicting value of the overall model.

Table 8

Simple linear regression model of the uncertainty types

Findings: when Mimetic behaviour was predicted it was found that Micro-level uncertainty (b=.445, p<0.01), Meso-level uncertainty (b=-.107, p<0.01) and Macro-level uncertainty (b=.173, p<0.01) were significant predictors. The overall model fit was R2_{=0.460. Both Micro-level uncertainty and}

Macro-level uncertainty have a positive coefficient whilst Meso-Macro-level uncertainty has a negative coefficient.

The results of the analysis suggest that Micro-level uncertainty and Macro-level uncertainty have a positive effect on Mimetic behaviour, supporting hypothesis 2 and 4. The effect of Meso-level uncertainty on Mimetic behaviour was found to have a significant negative effect on Mimetic behaviour (b=-.107, p<0.01). Therefore, hypothesis 3 is rejected.

27 Since not all types of uncertainty have the same effect direction, the measure for overall uncertainty cannot be created. Therefore hypothesis 1 is not tested, and the effect of the ‘overall uncertainty’ on Mimetic behaviour will not be further discussed in this thesis.

5.2 Mimetic behaviour as a moderator

To use Mimetic behaviour as a moderator means to test whether it effects the strength of a relationship between two variables (Goldsby, Michael Knemeyer, Miller, & Wallenburg, 2013). In this paper the moderation effect is analysed using the relationship between the constructs Hard- and Soft lean

implementation and Operational Performance.

Table 9

Linear regression model with the added interaction effect

Model Variables b F t Sig. R2 _{(adjusted) Model Sig. VIF}

1 16.252 .147 (.138) .000

Mimetic behaviour .028 .508 .613 1.023

Hard lean implementation .250 3.773 .000 1.466

Soft lean implementation .186 2.798 .005 1.460

2 10.887 .162 (.147) .000

Mimetic behaviour .045 .810 .418 1.050

Hard lean implementation .178 2.681 .008 1.481

Soft lean implementation .265 4.002 .000 1.475

Hard lean * Mimetic behaviour .131 2.045 .042 1.458

Soft lean * Mimetic behaviour -.134 -1.930 .055 1.430

The addition of the interaction terms, which represent the moderation effect, increase the predicting value of the overall model from R2 _{= .147 to R}2 _{= .162, constituting a 1.5 percent increase in explanatory}

28 finding will still be discussed in the ‘DISCUSSION’ chapter. The slopes of the interaction effects are shown in figure 2 and 3. The slopes were plotted using the paper of Dawson (2014).

Using the PROCESS plugin for SPSS model 1 was used to run another analysis. This analysis revealed that Hard and Soft lean implementation had different effects on operational performance at different levels of Mimetic behaviour.

Table 10

The effect of Hard lean implementation on operational performance at different level

of Mimetic behaviour

Z-score Mimetic behaviour Effect Se t p

-1.0000 .1252 .0838 1.4954 .1359

0.0000 .1892 .0664 2,8496 .0047

1.0000 .2532 .0875 2,8949 .0041

Table 11

The effect of Soft lean implementation on operational performance at different level

of Mimetic behaviour

Z-score Mimetic behaviour Effect Se t p

-1.0000 .3090 .0899 3.4386 .0007

0.0000 .2520 .0663 3.8022 .0002

1.0000 .1950 .0871 2.2404 .0258

29

Figure 2

Plot of Interaction effect between Hard lean implementation and mimetic behaviour

Figure 3

30

6. DISCUSSION

As mentioned in the introduction, the goal of this thesis is twofold. The first goal is to answer the question: “How is uncertainty related to Mimetic behaviour?” whilst the second goal focusses on answering the question of: “How is Mimetic behaviour related with operational performance?” In this chapter, each research question will first be answered separately, after which, a more holistic interpretation is given of the whole conceptual model. Starting with a visualization of all the results in figure 4.

Figure 4

Visualization of results

6. 1 The effect of Uncertainty on Mimetic behaviour

Before moving on to the first research question , it must be noted that due to the use of Factor Reduction Analysis, the adopted measures from Flynn et al. (2016), were reformed as described in the method section. The reforming assured that each item per scale measures one concept, improving the measures discriminant validity, thus contributing to the measurement of uncertainty, as uncertainty types in literature.

Answering the second research question, “How is Uncertainty related with Mimetic

behaviour? The results provide support for Micro-level uncertainty (b=.445, p<0.01), Meso-level

31 types of uncertainty was measured to be quite high, at a value of R2_{=0.460. Furthermore, both}

Micro-level uncertainty and Macro-Micro-level uncertainty have a positive effect, whilst Meso-Micro-level uncertainty has a negative effect on Mimetic behaviour.

Hence, by considering different types of Uncertainty as identified by Flynn et al. (2016), this thesis goes beyond the general claim made by the papers of DiMaggio & Powell (1983), Haunschild & Miner (1997) and Miemczyk, (2008) that uncertainty increases mimetic behaviour, and adds additional complexity by examining different uncertainty types and their influence.

Furthermore, finding a type of uncertainty that has a negative effect on Mimetic behaviour does not contradict the works of DiMaggio & Powell (1983), Haunschild & Miner (1997) and Miemczyk, (2008). Instead, it signifies that, even though the authors find that uncertainty in many cases has led to mimetic behaviour in organizations, seemingly not all types of uncertainty contribute to this outcome.

However, the negative effect of Meso-level uncertainty is still contrary to its hypothesized positive effect. This positive effect was assumed, since uncertainty or unavailability of information in the supply chain could contribute to the perceived overall uncertainty of possible options and their outcomes (Cyert & March, 1963).

Using the same literature, it can also be argued that Meso-level uncertainty did not only increase the uncertainty, but also lowered the awareness of the organization for possible options and their outcomes. This argument relies on the assumption that Meso-level uncertainty indicates an organization struggle to properly access and process information. This could be further supported by papers which state that mimicking is often accomplished through the benchmarking of competitors (Huang et al., 2010). Adding to this is the Supply Chain literature, indicating that awareness of competitors actions is seen to be of influences on organizational actions, which may include mimetic behaviour (Capron & White, 2008).

Moving on to the positive effect of Micro-, and Macro-level Uncertainty. The results suggest that Micro-level uncertainty, relating to the technical aspects of the organization, (b=.445) has a larger effect on Mimetic behaviour when compared to Macro-level uncertainty (b=.173), which relates to the uncertainty in the market.

During the hypothesis formulation it was argued that both the unpredictability and the perceived impact of resulting from unaccounted for uncertainty affected Mimetic behaviour (Zsidisin et al., 2005). Since the impact is subjective to a manager’s perception, the degree to which each uncertainty type effected Mimetic behaviour, compared to the other types, was not hypothesized. Using the same argument, it could be reasoned that, the unpredictability combined with the perceived impact of uncertainty, is larger for Micro-level uncertainty, as compared to Marco-level uncertainty.

32 Since micro-level uncertainty measures this and has an effect on Mimetic behaviour the findings support this proposition.

6.2 The effect of Mimetic behaviour on the ‘Lean-operational performance relationship’ Answering the second research question, “How is Mimetic behaviour related with operational

performance? The results provide support that Mimetic behaviour significantly moderates the

relationship between both Hard lean practices and operational performance (b= .131, p<0.05), and almost significantly moderates the relationship between Soft lean practices and operational performance (b= -.134, p=0.055) (for readability purposes referred to as the ‘Hard lean-performance relationship’ and ‘Soft lean-performance relationship’). The addition of the moderation effects of Mimetic behaviour raises the predicting value of the overall model from R2_{= .147 to R}2_{=.162. Thus,}

increasing the explanatory value of the model by 1.5 percent. This is arguably not much. However, only 23 percent of the variance in operational performance could be explained by Hard and Soft lean practices, after controlling for plant unionization, size and age, in the study of (Shah & Ward, 2003). Thus, making the 1.5 percent increase in predicting value of the overall model relatively more significant.

The almost significant moderation of mimetic behaviour on the ‘Soft lean-performance relationship’, is still relevant to discuss, because the operationalizations of Hard and Soft lean practices bundles vary between studies (Negrão et al., 2017), but also how the individual practices are used in organizations vary (Lewis, 2000), consequently changing the moderation effect of mimetic behaviour. The suggestion of a moderation effect on ‘Hard performance relationship’ and ‘Soft lean-performance relationship’ contributes to the existing literature by further supporting the claim of Ketokivi & Schroeder (2004), that process-performance studies should consider the effects of mimetic behaviour. When controlling for this effect, studies could add additional explanation for the often reported discrepancy in results gained from lean implementation among organizations (Pay, 2008).

Furthermore, the results suggest differences between how Mimetic behaviour effects the relationships of Hard and Soft lean practices towards operational performance. Firstly, where the ‘Hard lean-performance relationship’ is positively moderated (b= .131, p<0.05), the ‘Soft lean-performance relationship’ is almost significantly negatively moderated (b= -.134, p=0.055). Secondly, when looking at the moderation effects at different intervals for Mimetic behaviour. At the lower interval for Mimetic behaviour Hard lean practices do not significantly contribute to operational performance. Suggesting that Mimetic behaviour might be a prerequisite for the successful implementation of Hard lean.

33 Consequently, the positive and negative finding of this thesis require an alternative explanation. Starting with Sousa & Voss, (2008) who claim that copied practices have a lower performance effect due to contextual factors. It might be said that contextual factors for Hard and Soft lean practices are different, or the ability to account for them whilst copying is different, or both. Furthermore, using the literature of Nelson & Winter (1982) it might be that the availability of information differ between Hard and Soft lean practices. Since the authors claim that the availability of information influences the degree to which practices mutate between the imitated and imitator, causing lower performance. Alternatively, it can also be said that Soft lean practices are benchmarked less, or less effectively due to these supposed differences in the characteristics of Soft- and Hard lean practices. Thus, hindering the benchmarking process of Soft lean practices due to a lower tangibility and information availability.

In the literature, these differences between Hard and Soft lean are seemingly supported. In the context of a Manufacturing plant, technical aspects of the organization are visible, often extensively monitored, and quantify-able increasing its accessibility (Gunasekaran, Patel, & McGaughey, 2004). On the contrary, Soft practices are less visible, e.g. customer involvement, less monitored e.g. culture and dependent on less tangible/quantify-able contextual factors (Gunasekaran et al., 2004).

The use of these measurements are also low (Gunasekaran et al., 2004). Subsequently, imitating organizations are more likely experience the bad effects of mimetic behaviour as proposed by Nelson & Winter (1982) and Sousa & Voss (2008), with the implementation of Soft practices. Consequently, experiencing a negative or less than expected result on their operational performance (Pay, 2008).

6.3 Wholistic model perspective

34

7. CONCLUSION

The first aim of this thesis was to develop a more in-depth understanding of how uncertainty would affect Mimetic behaviour, as claimed in Institutional theory. In doing so, multiple types of uncertainty were tested for their effect on Mimetic behaviour. It was found that not all types of uncertainty increase Mimetic behaviour, thus nuancing the claim of institutional theory, and providing a more detailed indication of the effect. To elaborate, whilst uncertainty in the technical aspects of an organization (Micro-level uncertainty), and the uncertainty of the market (Macro-level uncertainty), increased Mimetic behaviour, uncertainty in the supply chain, due to an absence of information (Meso-level uncertainty), showed a negative effect on Mimetic behaviour.

The second aim of this thesis was to test the proposed effects of Mimetic behaviour on the contribution of Lean management towards operational performance. The results show two moderations, where the relationship between Hard lean implementation and operational performance is positively moderated by Mimetic behaviour, the relationship between Soft lean implementation and operational performance is almost significantly negatively moderated by Mimetic behaviour. This thesis suggests from literature that the different effect directions can be a signal of the different characteristics and contextual factors related to Hard and Soft lean practices, the manner in which they are treated equally or differently, and the effectiveness of such approaches. The results show that the current manner in which this occurs is beneficial for Hard lean practices, but not necessarily for Soft lean practices.

Furthermore, the findings provide evidence that organizations should at least display medium to high levels of Mimetic behaviour to receive significant benefits of implementing Hard lean practices on operational performance. Finally, these results suggest that Mimetic behaviour could add explanatory value for the discrepancy in results of organizations who have implemented lean. Therefore, future studies should take into account the effect of Mimetic behaviour when linking Lean management with operational performance.

Managerial implications

35 both the uncertainty in the technical aspects of an organization (Micro-level uncertainty), and the uncertainty of the market (Macro-level uncertainty).

It must be noted that, since the results in this research direction are very nascent, providing managers with very detailed advice would neglect the complexity of the underlying concepts. As has been shown in many cases apart from this thesis, there is not one exclusive method of achieving success. Therefore, advice should always be critically considered, paying mind to all the potential factors which might influence its effectiveness.

8. LIMITATIONS & FUTURE RESEARCH

As with all research, this thesis also has its limitations, and due to nascence of this research direction, ample opportunity is created for future researchers to expand upon this thesis.

First and foremost, the generalizability of the research must be taken into account. All used measurement relies on perceptual data, while some view this as appropriate for studying organizational behaviour (Flynn et al., 2016), also using more objective measure could make for a more balanced, and even more accurate measurement. Furthermore, the unit of analysis only included manufacturing plants with more than 100 employees. This is of importance since research indicates that plant size is an important contingency between lean implementation and operational performance (Shah & Ward, 2003). Also, larger organizations feel a greater need to legitimize (DiMaggio & Powell, 1983), and therefore might display higher levels of mimetic behaviour. Thus, whether the findings change for smaller manufacturing plants must be tested in future research.

Secondly, future research should expand on the existing Institutional theory literature by finding even more appropriate ways of measuring the mimetic behaviour of organizations. In such a pursuit, multiple distinction can be made in the measurement. For instance, mimetic behaviour caused by the social or economic institutional pressures, but also early and late adopters of practices. As such, a better distinction can be made between what types of imitations and types of motivations lead to possible positive or negative moderation effects. Thus, deploying both social and economic perspectives of institutional theory. Also, researchers should take into account how Mimetic behaviour is performed by organizations, and how the difference in approach to mimetic behaviour might lead to greater or reduced benefit from the implementation of Hard and Soft lean practices.

Thirdly, due to the scope of the research, contextual factors which affect the relationship between lean management and operational performance have not been taken into account. Future research should improve upon this thesis by adding these contextual factors, such e.g. as plants size, organizational culture, national culture thus reducing the variance in the sample.

36 before, since the effect of mimetic behaviour is at least opposite of the suggested effect of uncertainty, mimetic behaviour can be seen as having its own independent effect on the relationship between Lean management and operational performance. However, not testing for the possibility is still a shortcoming in this thesis and should be improved upon in future research.

Fifthly, a distinction could be made in the measurement between the different types of imitation, (frequency, outcome, trait) as defined by (Haunschild & Miner, 1997). It is possible that each type of imitation or its combination would provide different findings, since frequency and trait-based imitation are seen to be less rational, thus reflecting on the thoroughness of the decision-making process. The testing Mimetic behaviour with these distinctions could thus contribute to better defining what constitutes positive and negative mimetic behaviour.

37

REFERENCES

Anderson, J. C. ., & Gerbing, D. W. (1998). Structural equation modelling by Anderson and Gerbing 1988. Psychological Bulletin,.

Anderson, J. C., Cleveland, G., & Schroeder, R. G. (1989). Operations strategy: A literature review. Journal of Operations Management. https://doi.org/10.1016/0272-6963(89)90016-8

Baik, B., Chae, J., Choi, S., & Farber, D. B. (2010). Changes in Operational Efficiency and

Firm Performance: A Frontier Analysis Approach. SSRN.

https://doi.org/10.2139/ssrn.1681748

Barreto, I., & Baden-Fuller, C. (2006). To Conform or To Perform? Mimetic Behaviour,

Legitimacy-Based Groups and Performance Consequences*.

https://doi.org/10.1111/j.1467-6486.2006.00620.x

Bloom, N., Sadun, R., Reenen, J. Van, & Van Reenen, J. (2009). The Organization of Firms Across Countries. The Quarterly Journal of Economics, 127(4), 1663–1705. https://doi.org/10.1093/qje/qje029

Bonaccorsi, A., Carmignani, G., & Zammori, F. (2011). Service Value Stream Management (SVSM): Developing Lean Thinking in the Service Industry. Journal of Service Science

and Management, 04(04), 428–439. https://doi.org/10.4236/jssm.2011.44048

Bortolotti, T., Boscari, S., & Danese, P. (2015). Successful lean implementation: Organizational culture and soft lean practices. International Journal of Production

Economics, 160, 182–201. https://doi.org/10.1016/j.ijpe.2014.10.013

Capron, L., & White, H. C. (2008). Competitors’ Resource-Oriented Strategies: Acting on Competitors’ Resources Through Interventions in Factor Markets and Political Markets.

Academy of Management Review, 33(1), 97–121. https://doi.org/Article

Chimi, C. J., & Russell, D. L. (2009). The Likert Scale: A Proposal for Improvement Using Quasi-Continuous Variables. In Isecon (Vol. 26, pp. 1–10).

Cyert, R. M., & March, J. G. (1963). A behavioral theory of the firm. Englewood Cliffs, NJ,