• No results found

The Supplier Selection Process in the Swedish Pulp Industry: A Contingency Perspective

N/A
N/A
Protected

Academic year: 2022

Share "The Supplier Selection Process in the Swedish Pulp Industry: A Contingency Perspective"

Copied!
69
0
0

Loading.... (view fulltext now)

Full text

(1)

MASTER'S THESIS

The Supplier Selection Process in the Swedish Pulp Industry

A Contingency Perspective

Johan Elfving Jimmy Sundqvist

Master of Science in Business and Economics Master of Science in Business and Economics

Luleå University of Technology

Department of Business, Administration, Technology and Social Sciences

(2)
(3)

Acknowledgements

The authors wish to acknowledge the time devoted by the interviewees and their expressed enthusiasm towards this thesis. We would also like to express our gratitude for helpful comments, guidance and valuable thoughts received on previous drafts, especially from our tutor Anders Nilsson.

Luleå, June 2011.

Johan Elfving Jimmy Sundqvist

(4)
(5)

Abstract

This thesis aims to study the supplier selection process through a theoretical lens grounded in contingency theory. The increasing importance of supplier selection as a tool for developing competitive advantages motivated the subject for this study. Contingency factors influence on the supplier selection process was studied through a quantitative survey. The Swedish pulp industry, involving sixteen organizations, constituted the object of study. In order to facilitate the empirical research, a theoretical framework based on the supplier selection process and contingency factors was developed. Previous research and practical studies provided inspiration for the identification of the three contingency categories (buying-specific, organizational and inter-organizational). Using this theoretical framework in a deductive manner, it was found that the general trend for the industry is to reduce the supplier base and utilize package-deals. Furthermore, the most important selection criteria were found to be delivery, quality and reliability. This study also confirms the importance of relationships between the supplier and the purchasing organization, where suppliers in more transactional relationships are more likely to be replaced. The majority of the correlations confirm or partly confirm findings in previous research concerning other industries.

Key words: Supplier selection, Purchasing, Sourcing strategies, Supply chain, Contingency

theory, Pulp industry, Chemicals

(6)
(7)

Table of Content

1

!

INTRODUCTION ... 1

!

1.1

!

BACKGROUND ... 1

!

1.2

!

PROBLEM DISCUSSION ... 2

!

1.3

!

RESEARCH PROBLEM ... 3

!

1.4

!

RESEARCH QUESTIONS ... 3

!

2

!

LITERATURE REVIEW ... 4

!

2.1

!

THE SUPPLIER SELECTION PROCESS ... 4

!

2.2

!

BUYING-SPECIFIC FACTORS ... 7

!

2.3

!

ORGANIZATIONAL FACTORS ... 7

!

2.4

!

INTER-ORGANIZATIONAL FACTORS ... 10

!

2.5

!

SUMMARY OF LITERATURE REVIEW ... 11

!

3

!

METHODOLOGY ... 12

!

3.1

!

RESEARCH PURPOSE ... 12

!

3.2

!

RESEARCH APPROACH ... 12

!

3.3

!

RESEARCH STRATEGY ... 13

!

3.4

!

SAMPLE SELECTION ... 13

!

3.5

!

DATA COLLECTION METHODS ... 14

!

3.6

!

DATA ANALYSIS ... 14

!

3.7

!

METHOD PROBLEMS ... 17

!

4

!

DATA PRESENTATION ... 19

!

4.1

!

THE SUPPLIER SELECTION PROCESS ... 19

!

4.2

!

BUYING-SPECIFIC FACTORS ... 22

!

4.3

!

ORGANIZATIONAL FACTORS ... 24

!

4.4

!

INTER-ORGANIZATIONAL FACTORS ... 27

!

5

!

ANALYSIS ... 29

!

5.1

!

THE SUPPLIER SELECTION PROCESS ... 29

!

5.2

!

INFLUENCES OF BUYING-SPECIFIC FACTORS ... 32

!

5.3

!

INFLUENCES OF ORGANIZATIONAL FACTORS ... 34

!

5.4

!

INFLUENCES OF INTER-ORGANIZATIONAL FACTORS ... 39

!

6

!

CONCLUSIONS AND DISCUSSION ... 42

!

6.1

!

CONCLUSIONS ... 42

!

6.2

!

DISCUSSION ... 45

!

LIST OF REFERENCES

!

APPENDIX

!

APPENDIX I:INTERVIEW GUIDE

!

APPENDIX II:SCALE OF MEASUREMENT AND CODING OF VARIABLES

!

APPENDIX III:DESCRIPTIVE STATISTICS FOR CONTINUOUS VARIABLES

!

APPENDIX IV:DESCRIPTIVE STATISTICS FOR DISCRETE VARIABLES

!

APPENDIX V:SPEARMANS RANK CORRELATION COEFFICIENT

!

APPENDIX VI:ANALYZED RANK CORRELATIONS

!

APPENDIX VII:EXAMPLE OF SPEARMANS RANK CORRELATION COEFFICIENT

!

APPENDIX VIII:EXAMPLE OF FUZZY LOGIC ALGORITHM

!

(8)
(9)

1 Introduction

This chapter gives the reader an introduction and background to the research area, which focuses on the supplier selection process in the Swedish pulp industry. At the end of the chapter the research problem is discussed which leads to the formulation of the research questions.

1.1 Background

The purchasing process has throughout most of our industrial history been viewed as a set of operational activities incapable of contributing to overall organizational effectiveness (Henderson 1975). In recent years this view has changed and today considerable attention is given to the strategic contribution that the purchasing process can make to an organizations total performance (Ellram & Carr 1994). Executive managers are relying on the purchasing process to contribute to and develop competitive advantages in areas such as cost savings, quality improvements, waste reductions, reliable delivery, and technology innovations (Monczka & Morgan 1992). In industrial organizations the purchasing’s share in the total turnover ranges between 50 and 90 percent, making the purchasing process one of an organizations primary determinants of an organizations profitability (de Boer, Labro

& Morlacchi 2001; Degraeve & Roodhooft 2000).

In the last two decades globalization of trade and information technology have enlarged the purchaser’s choice set. Rapidly changing customer preferences regarding products requires a broader and faster supplier selection and more cross-functional thinking (de Boer et al.

2001). Simultaneously, an ongoing trend is to streamline the number of suppliers to obtain a more efficient supply chain. This means that more long-term relationships are established with fewer suppliers, and in the long run provides opportunities of significantly reducing costs and improving quality (Monczka et al. 1993). With a reduced supplier base, organizations’ are more dependent on their suppliers and the consequences of poor decision-making become more severe (de Boer et al. 2001). Furthermore, due to a reduced supplier base in combination with higher risks associated with supplier selection, the selection criteria used by purchasing managers have developed and today encompasses more than price as selection criteria (Bevilacqua & Petroni 2002).

These developments urge for a more systematic approach to handle purchasing-decisions, especially regarding the area of supplier selection (Carter et al. 1998). However, the nature of these types of decisions is often unstructured and highly complex (Degraeve &

Roodhooft 2000). Supplier selection can be seen as a multicriteria problem that

incorporates both qualitative and quantitative factors (de Boer et al. 2001; Kleinsorge,

Schary & Tanner 1992). In order to make the decision about choosing the best suppliers it

is necessary to make a trade off between these tangible and intangible factors that

sometimes conflicts with each other (Ghodsypour & O’Brien 1998). Therefore, how

purchasing decisions are reached and justified should receive much attention in future

research (de Boer 1998). Consequently, it is important to determine which factors influence

and affects the process of selecting the most suitable suppliers.

(10)

1.2 Problem Discussion

Increasing customer demands, innovations in technologies, higher competition in a global environment, decreased governmental regulation and increased environmental consciousness have forced organizations to nurture their supply chain and put more effort into the supplier selection process (Tracey & Tan 2001). These trends align well with the contingency theory, advocating that there exists “no best way” to operate an organization.

To maximize performance, organizations have to adapt to different contingency factors. In the supplier selection process these contingencies can be of internal and external character, such as buying-specific factors, organizational factors and inter-organizational factors (Wind & Thomas 1980). Furthermore, de Boer et al. (1998) strengthens this claim arguing that contingency factors are highly determinative of how the supplier selection process is structured. The fit between how organizations select suppliers’ and contingency variables continuously changes and organizations need to adapt to the new level of the contingencies to avoid loss of performance (Donaldson 2001). Organizations become shaped by the contingencies and it is therefore interesting to study how these contingency factors are shaping organizations supplier selection processes.

To study how contingency factors influence the supplier selection process the Swedish pulp industry is examined. Sweden has the world’s leading manufacturers of pulp intended for paper production and a large proportion of all pulp used in Europe originates from pulp mills located in Sweden (Bredström 2004). In 2009 Swedish pulp mills produced over 11.5 millions tones of pulp and accounts for a major part of the Swedish trade balance where a total of 3.4 million tones was exported (Swedish Forest Industries Federation 2009). Moreover, there is reason to believe that the supplier selection process have changed in this industry due to the increased consolidation and the advent of higher global competition (Waitt 1994). Today, the twelve largest pulp mills in Sweden accounts for 70 percent of the total production. The size of the industry, the importance to the trade balance and the changing business logic indicates that the pulp industry is interesting to study.

Furthermore, the composition of the pulp depends on the manufacturing process, but the

main ingredient is always fibers from pulpwood. The majority of the pulp mills in Sweden

have concentrated their manufacturing to sulphite pulp, sulphate pulp and semi-chemical

pulp (Swedish Forest Industries Federation 2009). Mechanical pulp and CTMP pulp are

also produced but in smaller volumes. In the manufacturing processes to produce pulp,

chemicals are used to remove the lignin and improve the quality of the fibers (Swedish

Chemicals Agency 2011). The ability to choose the most suitable chemical suppliers is

therefore highly determinative of a pulp mills success. Therefore, to further define the

research area this study will focus on the selection process of chemical suppliers within the

Swedish pulp industry.

(11)

1.3 Research Problem

In summarization the supplier selection process is today one of the key activities in organizations and is affected by contingency factors, the Swedish pulp industry has always had a strong position and chemicals is of great importance to the pulp mills manufacturing processes. Therefore, the research problem that this thesis attempts to answer have been stated as follows:

How can the supplier selection process of chemical suppliers and the key influences of well-recognized buying- specific factors, organizational factors and inter-organizational factors in the Swedish pulp industry be characterized?

1.4 Research Questions

In order to further clarify the scope of this thesis the research problem is digested into four research questions. First it is important to determine the structure of the supplier selection process in the Swedish pulp industry to establish a basis for further comparison.

RQ1: How can the supplier selection process in the Swedish pulp industry be characterized?

Then each contingency category is individually examined to determine the influence on the supplier selection process. The first category is buying-specific factors referring to factors that are directly associated with the supplier selection process.

RQ2: How can the key influences of buying-specific factors on the supplier selection process in the Swedish pulp industry be characterized?

The second category is organizational factors, involving factors that are indirectly related to the supplier selection process.

RQ3: How can the key influences of organizational factors on the supplier selection process in the Swedish pulp industry be characterized?

The third category is inter-organizational factors, examining influences on the supplier selection process of factors between the buying and selling organizations.

RQ4: How can the key influences of inter-organizational factors on the supplier selection process in the

Swedish pulp industry be characterized?

(12)

2 Literature Review

In this chapter a literature review is presented to give the reader a theoretical background of selected theories within the research area. The literature review is structured by: the supplier selection process, buying-specific factors, organizational factors and inter-organizational factors.

2.1 The Supplier Selection Process

According to de Boer et al. (2001) the supplier selection process is divided into different phases that take the diversity of procurement situations in terms of complexity and importance into account. Based on an extensive search in literature the authors have positioned the contributions in the area of supplier selection in a framework that covers four phases in the supplier selection process (see Figure 1). These phases are recognized as problem definition, formulation of criteria, qualification and final selection. The most visible phase in this process is final selection and has therefore received far more attention from researchers in the area. However, the quality of the final selection phase is highly dependent of the prior steps and those steps should therefore not be disregarded (de Boer et al. 2001).

Figure 1: Illustration of the different phases in the supplier selection process adapted from de Boer (1998) Problem Definition

The first phase is characterized by the objective to determine the ultimate problem of why supplier selection is an appropriate line of action. It can also be described as finding out what an organization wants to achieve by selecting a supplier (de Boer et al. 2001). More specific the phase consists of deciding whether a problem can be solved or not by selecting one or more suppliers. This phase is useful to aid the purchaser in solving the “right problem” and refraining from dropping a supplier for wrong reasons, e.g. the suppliers delivery problems was actually cased by feeding them with outdated information.

Generally, the problem definition phase is an underdeveloped area in supplier selection literature, but still important to recognize (de Boer et al. 1998). According to de Boer et al.

(2001) the four questions that might be asked during the phase are: Should a supplier or the resources within the organization be used to solve the problem? Is it necessary to use more than one supplier to deal with the problem? Should fewer suppliers be used than before to solve the problem? Is it necessary to replace a supplier and how should the organization

Acceptable Suppliers

Selected Suppliers All Available Suppliers

1. Problem Definition

2. Formulation

of Criteria 3. Qualification

4. Final Selection

(13)

deal with the potential supplier? Furthermore, the main motive of replacing an existing supplier is unsatisfactory supply performance. In these cases there are no need to modify the specifications in the supplier selection process (de Boer 1998). However, when finding new ways of dealing with current suppliers’ modifications of the specifications in the supplier selection process has to be done (ibid).

Formulation of Criteria

The second phase is the formulation of criteria where relevant criteria are defined. These criteria should reflect what the buying organization demands from a potential supplier (de Boer et al. 2001). If a previous relationship exists with the supplier, selection criteria may already be available. Otherwise new criteria have to be formulated (ibid). These supplier selection criteria are used in the selection process of suppliers, but can also be used as key performance indicators later to evaluate the suppliers (ibid). In general, many quantitative and qualitative factors such as quality, price, flexibility and delivery performance must be considered to determine suitable suppliers (Bevilacqua & Petroni 2002). However, due to the complex nature of the supplier selection process an abundance of criteria might occur and not all decision makers might consider the same criteria (de Boer et al. 2001).

Table 1: Ranking of the most important selection criteria first conducted by Dickson (1966) and later reviewed and compared by Cheraghi et al. (2001)

Rank (2001) Rank (1966) Factor

1 3 Quality

2 2 Delivery

3 1 Price

4 10 Repair service

5 5 Technical capability

6 4 Production facilities and capacity

7 9 Financial position

8 7 Management and organization

9 New Reliability

10 New Flexibility

11 8 Attitude

12 13 Communication system

13 11 Performance history

14 6 Geographical location

As seen in Table 1, the most important supplier selection criteria complied in the mid

1960s was price and delivery (Dickson 1966). More recent studies have discovered a shift

away from price as a primary determinant of supplier selection (Bevilacqua & Petroni

2002). The set of relevant supplier selection criteria changes over time as a natural adaption

to changing business climates and competitive environments (ibid). According to Scott and

Westbrook (1991) is it likely that a strategic approach of purchasing influence this process

and creates new sets of criteria for supplier selection. For example, it is more likely that

buyers who are involved in a strategic partnership not solely rely on traditional selection

criteria and instead involves new selection criteria of a more soft character (Ellram 1990).

(14)

Many previous studies have been focused on determining what criteria should be used in this phase of the supplier selection process. Cheraghi, Dadashzadeh and Subramanian (2001) conducted a cluster analysis of 110 research papers regarding critical success factors in supplier selection. The study identified the fourteen most popular selection criteria to:

quality, delivery, price, service, technical capability, production facilities and capacity, financial position, management and organization, reliability, flexibility, attitude, communication system, performance history and geographical location (Cheraghi et al.

2001). However, the relative importance of each criterion varies largely due to the nature of the situation and what industry the organization operates within. This implies that there probably cannot be a generalized consensus on how to weight the relative relevance of the different criteria (de Boer et al. 2001).

According to Cheraghi et al. (2001), the changes in rank of selection criteria between the two studies can be explained by an abundance of reasons. Firstly, the changing business environment and the accelerating globalization are considered as central explanations.

Secondly, Just in time (JIT) and obtaining a more competitive supply chain are mentioned to have an increasing importance (ibid). Thirdly, core competencies are discussed where organizations conduct competence-related activities internally and outsource activities beyond their core competency. Cheraghi et al. (2001) imply that there is an increased focus on core competencies in order to obtain competitive advantages. Buying organizations often prefer relationships that provide opportunities to enjoy each other’s unique core competencies (Cheraghi et al. 2001).

Qualification

The third phase is not always easily distinguished from final selection, and may therefore not always be easily identified in a real life context per definition (de Boer et al. 2001).

Nevertheless, the qualification phase refers to the process of reducing the set of all available suppliers to a more narrow set of acceptable suppliers (ibid). The decisions in this phase can be based on both historical data and new data that have been collected for the specific processes in hand (ibid). Theoretically speaking, the qualification phase can always be deduced to the activity where suppliers are asked to submit a quotation (ibid).

According to Crow, Olshavsky and Summers (1980), 80 % of industrial buyers send

quotation requests to three or four suppliers in average. Furthermore, Crow et al. (1980)

found that none of the industrial buyers send less than three quotations to a group of

potential suppliers. Dasgupta and Hashinto (2004) claim that four quotation requests is the

characteristic number in industrial buying of products that are not sold by none

monopolistic companies. In more sophisticated qualification procedures previously defined

criteria are considered (de Boer 1998). Judgments about the importance of the defined

criteria and how well a supplier performs regarding the criteria are also included (ibid). This

process is often carried out in more than one step. However, the first step always consists

of defining acceptable supplier. Subsequent steps tries to reduce the number of suppliers to

consider, in order to minimize the time spent on evaluating less good suppliers (de Boer et

al. 2001).

(15)

Final Selection

The last phase in the supplier selection process is referred to as final selection or choice (de Boer 1998). In this phase the buying organization selects a supplier or a number of suppliers to solve the problem stated in the first phase (ibid). The final selection phase takes the magnitude of the actual selection into consideration. More specific, the final selection phase can be characterized as single-deals, package-deals or multiple-deals (ibid).

Single-deals only consider one product at a time whereas package-deals consider multiple products within the same product line. Multiple-deals refer to multiple products to different product lines (ibid). According to de Boer et al. (2001), almost two-thirds of the existing supplier selection processes can be characterized as single-deals or package-deals, whereas multiple-deals is not as frequently occurring.

2.2 Buying-specific Factors

Buying-specific factors refers to factors that might affect the decision-making within an organization and furthermore the supplier selection process. The factors that are considered in this section are type of purchase and time pressure.

Type of Purchase

According to Robinson, Faris & Wind (1967) a certain purchase can be classified into separate types of purchases often referred to as buy classes. Three factors that help determine the buy class are: the newness of the problem, the information requirements, and the consideration of new alternatives. Jobber (2001) has simplified the definitions and describes the three classes as following: New task refers to the first time purchase of a product or input by an organization and this buy class that requires most time for evaluation in order to solve the problem that has instituted the purchase. Modified rebuy refers to a purchase where a regular requirement for the type of product exists and the buying alternatives are known but sufficient changes (e.g. a delivery problem) have occurred to require some alteration to the normal supply procedure. Straight rebuy refers to a purchase by an organization from a previous approved supplier of a previously purchased item and is the buy class that requires least time of the three buy classes. Straight rebuy is often the most common buy class among industrial companies.

Time Pressure

A third factor to consider regarding supplier decisions is time pressure. If the decision is made on an emergency basis or a great deal of time pressure, the buying center is likely to be smaller in size and the buying decision is delegated to one individual (Robinson et al.

1967). A high degree of decentralization provides greater opportunities to deal with emergencies and thereby is the incitement to avoid these types of situations smaller than for more flexible organizations (ibid). The degree of planned and unplanned purchases is a measure of the time pressure that a buying center is exposed to over a certain period (ibid).

2.3 Organizational Factors

Organizational factors refer to factors that might affect the different phases within the

supplier selection process. The topics that are presented in this section are: different

measures of size, sourcing strategy, degree of decentralization and the dimensions of the

buying center.

(16)

Size

According to Bonoma and Johnstone (1971), the size of a company has an effect on the activities undertaken in the supplier selection process. The measure of size can also be referred to as the structure of a company, involving factors as number of employees, annual turnover, number of departments and costs (such as value of purchases).

Sourcing Strategy

According to Quayle (1998), the three most common sourcing strategies are: single sourcing, multiple sourcing and parallel sourcing. Backward vertical integration and supplying yourself are less common sourcing strategies, partly covered in different sections of this literature review. According to Goffin, Szwejczewski and New (1997) the trend in many industries is to decrease the use of multiple sourcing in favor for establishing closer relationships with a smaller group of suppliers. Furthermore, vertical integration is less common due to a higher focus on core competency and outsourcing the tasks that are not included in the own core competencies (ibid). Quayle (1998) has provided definitions and characteristics of the three sourcing strategies single sourcing, multiple sourcing and parallel sourcing, which are presented below.

Single Sourcing can be defined as using one supplier to provide the organization with a certain product. The sourcing strategy enables concentrated purchasing on a long-term basis. If only one supplier is used more effort and resources can be dedicated to the relationship and cooperation to the supplier, for example through quality improvement programs. Single sourcing do normally imply on a stronger bargain power for the buying organization in relation to the other sourcing strategies due to that the purchases are consolidated to one supplier. A major disadvantage with a single sourcing strategy is that an organization might be highly dependent to a supplier, depending on the importance of the products that are sourced from the supplier.

Multiple Sourcing has been seen as a way of maintaining flexibility and a response to the increased uncertainties of relying on the deliveries of only one supplier. The flexibility arises from the possibility of having several suppliers available and to choose from different products. Another advantage of multiple sourcing is that the products from several research and development departments can be available. In addition to these advantages are the tactical aspects such as altering different divisions of business between the suppliers in order to encourage a high performance. Items that are especially suitable for a multiple sourcing strategy are items of critical character to the production processes. Multiple sourcing does also provide the opportunity to contract the supplier with most favorable offer for the moment. However can the higher cost that might be a result of dividing the business of purchasing items from different suppliers can be seen as an insurance against the cost that might arise due to stopped production.

Parallel Sourcing can be defined as using one source for a component at one assembly plant

while another source is used at a different plant. In other words are only one supplier the

provider of a certain product to one of a company’s facilities whether another supplier are

providing another of the company’s facilities with the same product. The suppliers are

however not providing the same facilities with the same product. Parallel sourcing can also

(17)

be explained as that there is two or more supplier who have similar capabilities and currently are sourcing products with substantial similarities. One reason to why a parallel sourcing strategy is chosen is that an organization wants to compare and evaluate different potential suppliers. Furthermore the parallel sourcing provide incentives for suppliers to exert oneself due to that other suppliers might be relatively easy take their place.

Risks of Sourcing Strategies are thoroughly described by Treleven and Bergman Schweikhart (1988). Organizations that are relying on single sourcing are more likely to encounter a disruption in the supply than organizations practicing multiple sourcing. A higher number of suppliers to rely on decrease the risk of running out of supplies (ibid). The risks of single sourcing are often balanced against the advantages of improved communications, cooperative relationships and higher bargain power (ibid). A closer communication can provide early warnings of potential future problems and as a consequence, increased possibilities to help solve, avoid or plan around them (ibid). However, the development of close relationships requires time, resources and mutual trust in order to be successful (ibid).

Degree of Decentralization

According to Webster and Wind (1972) the degree of decentralization corresponds directly to the extent of the buying authority that has been delegated to operating levels in the organization. The degree of decentralization in the decision process depends on to which extent the authority structure interacts with the communication structure (ibid). In a decentralized organization local departments often execute the decision-making. In contrary, a centralized organization often has a central decision unit that handles the decision-making (David & Hwang 2002). Organizations that focus on cost leadership strategies should rely on a centralized purchasing department to gain economics of scale (Fisher 1997). An organization with a differentiation strategy should rely on a more decentralized purchasing organization in order to obtain high flexibility, speed and coordination (ibid). The general case is that centralized organizations have a need for long- term planning and stable processes in order to utilize the possible advantages whereas decentralized organizations can cope with changes more quickly (ibid).

The Buying Center

The concept of the buying center refers to all members of an organization that are involved

in the buying process for a particular product or service (Robinson et al. 1967). No buying

center are likely to be operated exactly the same, however general patterns of interaction

and behavior will be the same, even across moderately dissimilar purchase situations

(Johnston & Bonoma, 1981). From their theoretical perspective, four dimensions of the

buying center could be specified and measured. (1) Vertical involvement is referring to the

number of authority levels involved and influencing the communication within the buying

center. (2) Lateral involvement can be described as the number of separate departments

involved in the purchase decision. (3) Extensivity relates to the total number of individuals

that are involved in the buying communication network. (4) Formalization of the

communications in the buying center, e.g. the degree to which the members of the buying

center has formal or informal connections concerning the purchase. Johnstone and

Bonoma (1981) argue that a higher degree of formalization makes employees keener to

fulfill the objectives of the organization due to that the responsibilities are clearer.

(18)

2.4 Inter-organizational Factors

Types of relationships refer to all possible relationships between a buying and a selling organization (Webster 1992). These relationships affect how business is conducted within an organization. Hence, relationships that might affect the organization might also affect the supplier selection process (ibid). The number of available suppliers and the number of relationships to suppliers can also affect the rest of the organization (ibid). In inter- organizational relationships, the trend has been to decrease the supplier base in order to gain advantages from the more evolved types of relationships established with fewer suppliers (Parmar, Wu, Callarman, Fowler and Wolfe 2010). Terminations of relationships and switching of suppliers are normally associated with costs and might lead to adverse relations lasting longer than what is optimal from a total cost point of view (Klemperer 1987). Furthermore, the switching costs are generally higher for relationships with higher interdependencies (ibid). According to Webster (1992) there are seven possible types of relationships: transactions, repeated transactions, long-term relationship, buyer-seller partnership, strategic alliances, network organizations and vertical integration.

Types of Relationships

The terminology and typology for describing organization forms and types of relationships can sometimes be confusing. Webster (1992) has provided a model describing the range of relationships categorized in seven different phases. This model should be seen as a continuum rather then explicit phases where transactions forms the beginning and vertical integration the end due to the sometimes unclear distinctions in real life.

Transactions are a one-time exchange of value between two parties with no prior interaction with each other. Price is established in the competitive marketplace and contains all of the information necessary for both parties to conclude the exchange. In pure transaction, there is no brand name, no recognition of the customer by seller, no credit extension, no preference, no loyalty, and no differentiation of one producer’s output from that of another. This pure form of market marks the beginning of the continuum and provides a useful starting point for theoretical analysis.

Repeated Transactions are often called the precursors of a relationship. The actual sale is the end result of the activity although repeated purchases are important to the organization. In repeated transaction there is a presence of other factors than just price and organizations are aggressively marketing their products to win the customer’s preference and loyalty.

Long-term Relationships involve long-term contractual commitments, but even here a distance between the buyer and marketer exists. In these types of relationships the buying organization promotes competition among the vendors by competitive bidding and tight product specifications. The buying organization also wants to keep many vendors involved to keep the pressure on price and ensure alternative sources of supply.

Buyer-seller Partnerships is moving away from the transactional perspective and toward an

understanding of the need to develop partnerships with customers in order to achieve a

high quality at a low cost. Other characteristics of buyer-seller partnerships are a small

number of supplier and early supplier involvement in product development.

(19)

Strategic Alliances are a form of expanded partnerships where both parties strive to achieve common long-term strategic goals. There are multiple types of strategic alliances;

organizations can cooperate in developing new technology, new products, and new markets. A typical strategic alliance involves commitment of capital and management resources with the objective to enhance the partners’ competitive position. Joint venture is a form of strategic alliance where the unique feature is that a new organization is created.

Network Organizations are complex, multifaceted organization structures that incorporates multiple types of strategic alliances usually combined with cross-collaboration between divisions, subsidiaries, and re-sellers. The unique feature of a network organization is a loose and flexible coalition guided from a central hub where the key functions exists.

Network organizations often emerge in large multi-national organizations that want to specialize in their core competencies.

Vertical Integration is compared to network organizations an organization that focus on controlling the whole value-chain in order to be more effective and make improvements.

By doing this supply and demand can be synchronized along the chain of products and in some cases the possibility to monopolize a market can emerge.

2.5 Summary of Literature Review

The supplier selection process is the center of this study and is represented by the first research question. This process is presented as a box containing four different phases that is recognized by the literature in the area. Furthermore, the three contingency categories described in this literature review (see Figure 2) – (1) Buying-specific factors, (2) – Organizational factors, (3) – Inter-organizational factors – are thought to impact the shape of the supplier selection process, each representing a research question. Appendix I and II shows the interview questions and variables that emerged from the literature review.

Figure 2: Summary of literature review

Organizational

factors Inter-organizational factors The Supplier Selection Process

Problem

definition Formulation of

criteria Qualification Final selection

Buying-specific factors

RQ1

RQ2 RQ3 RQ4

(20)

3 Methodology

This chapter provides a description of the research process and the methodologies used in this thesis. The chapter covers the research purpose, research approach, research strategy, sample selection, data collection methods, data analysis and method problems.

3.1 Research Purpose

The classification of research purposes are often separated into three categories:

exploratory, descriptive and explanatory; which one to use depends on how the research questions are formulated (Saunders 2009). An exploratory purpose is often used to seek new insights when no or little information exists regarding a particular problem (Saunders 2009). A descriptive purpose focuses on finding key phenomena’s in order to identify patterns and trends in a situation with the aim to draw conclusions from the data that are described (Yin 2003). The emphasis with an explanatory purpose is to study a situation or a problem in order to explain the relationships between variables (Saunders 2009).

Based on these three categories and the formulation of this thesis research questions, this study adapted a descripto-explanatory research purpose. According to Saunders (2009) descriptive studies should be thought of as a means to an end rather than an end in itself and therefore this thesis utilized description as a precursor to explanation. First the supplier selection process was described to create a basis for further analysis. Subsequently, casual relationships were established between the supplier selection process and contingency variables that justified the explanatory purpose.

3.2 Research Approach

There are different approaches possible when deciding how the research should be conducted. The first steps is to determine if the research should have a deductive approach or an inductive approach and then decide whether the approach should be of quantitative or qualitative character (Saunders 2009). The deductive approach means that you first develop a hypothesis or a problem formulation and then design a research strategy to test it in real life. With the inductive approach you collect data and then develop a theory based on the data collected (Saunders 2009).

The idea of this thesis was to gather theories concerning the studied area and then

empirically tests them towards representatives within the Swedish pulp industry in order to

accurately describe the situation and establish correlations. A literature review was

conducted prior the data collection in order to incorporate existing theory from accessible

studies which points to the direction of a deductive approach. To differentiate analysis

procedures regarding data collection techniques and data analysis the terms quantitative

and qualitative data can be applied. Quantitative data refers to numerical data whilst

qualitative refers to non-numerical data (Saunders 2009). This study mostly used

quantitative data in order to establish correlations with use of statistics.

(21)

3.3 Research Strategy

The types of research strategies are usually divided into five different categories:

experimental, survey, archival analysis, historical and case study. When deciding what type of strategy to choose the easiest way is to evaluate the possibilities through three different conditions: (1) Type of research questions, (2) Extent of control over behavioral events and (3) The degree of contemporary events. Since this thesis focused on contemporary events and not had control over all behavioral event the experiment strategy and history strategy was excluded. Archival research was not suitable due to the fact that the thesis aimed to collect new information. The survey strategy was chosen over a case study to enable the use of a larger sample and achieve higher generalizability. It also emphasis on the use of quantitative data followed by statistical analyses, which was in line with the formulation of the research questions (Saunders 2009). Furthermore, the study was conducted with a snapshot time horizon and can therefore be attributed as a cross-sectional study.

3.4 Sample Selection

In general there are two main types of sample selections to use: probability or non- probability sampling methods. A probability sample is a selection of sampling techniques in which the probability of each case being selected is known and is often equal for all cases.

A non-probability sample is on the contrary a selection of techniques where the probability of each case selected is unknown (Saunders 2009).

Table 2: Overview of the organizations in the Swedish pulp industry (Swedish Forest Industries Federation 2009)

# Organization Pulp Mills

1 Billerud AB Karlsborg, Gruvön & Skärblacka 2 Cascades Djupafors Djupafors

3 Domsjö Fabriker AB Örnsköldsvik

4 Holmen Iggesund, Hallstavik & Braviken 5 Kinnevik Korsnäs, Frövifors & Rockhammer

6 M-Real Husum

7 Mondi Packaging Dynäs

8 Munksjö Billingsfors & Aspa 9 Nordic Paper Säffle & Bäckhammar 10 Rottneros AB Vallvik, Utansjö & Rottneros

11 SCA Östrand, Munksund, Ortviken & Obbola 12 Smurfit Kappa Group Lövholmen

13 Stora Enso Skutskär, Kvarnsveden, Skoghall, Hylte, Fors & Nymölla 14 Södra Värö, Mösterås & Mörrum

15 Waggeryd Cell Vaggeryd

16 Vida Papper Lessebo

Occasionally it may be possible to collect data from every possible case or group member

in the population: this is termed census (Saunders 2009). The sample in this thesis can be

considered a census, considering that every organization within the Swedish pulp industry

was included. Table 2 shows the 16 organizations that were included in this thesis.

(22)

3.5 Data Collection Methods

Primary data is gathered from first-hand experiences and is collected directly by observations while secondary data is gathered from second-hand experiences e.g. articles and books (Walliman 2001). In a survey there are three primary collection methods to consider, questionnaire, structured observation and structured interviews (Saunders, 2009).

To maximize response rates and minimize the time spent on data collection the primary data was collected through structured telephone interviews. This method utilized the benefits from predetermined and standardized questions to facilitate the subsequent data analysis whilst avoiding poor response rates using a questionnaire when the interview participants willingly answered the interview questions.

The interview participants consisted of one purchasing manager from each organization within the Swedish pulp industry. They were chosen under the assumption to be the most proficient regarding supplier selection processes and contingency factors. The structure of the organizations entailed that purchasing managers should be chosen from organizations instead of from mills with the knowledge that one purchasing managers often had responsibility for more than one pulp mill. Therefore, no emphasis was put on which mill the purchasing manager belonged to as long he held an appropriate title and was employed by the organization.

The interview guide used in this study was developed based on the formulation of the research questions in combination with the literature review. Both academics and purchasing managers reviewed the interview guide beforehand to ensure relevant content and clarity in the interview questions. Prior to the actual interview the interview participant was introduced to the study via an introductory e-mail that later was followed up with a telephone call where a meeting was set up. All of the interview participants received the interview guide beforehand so they could prepare their answers. This resulted in interviews that took no more than 30 minutes to complete. The secondary data used in this study is mostly based on articles that have been found in different publications within the studied area.

3.6 Data Analysis

Quantitative data conveys little or no meaning to most people before it has been processed and analyzed (Saunders 2009). Furthermore, Robson (2002) argues that quantitative data analysis is a field where it is not at all difficult to carry out an analysis which is simply wrong, or inappropriate for the research questions. This chapter aims to clarify the data analysis procedure used in this thesis, how the data were explored, presented, described and how relationships and trends within the data were analyzed.

Preparation of Collected Data

During the development of the interview guide each interview question were assigned a

scale of measurement and a code. The scale of measurement was used to determine the

range of techniques available for further analysis (Saunders 2009). Most of the questions

were determined to be of discrete and ordinal character. Discrete data represents a finite

number and can be measured precisely compared to ordinal data where the relative

position of each data is known but the actual numerical measures on which the position is

(23)

based on are not recorded (Saunders 2009). However, ordinal data often have similar gaps between data values, and can therefore often be analyzed as discrete data. Each question was assigned codes that showed the number of the question, the variable name and from what concept the variable originated in the literature review. The coding was primary used to avoid conflation in the data input and to facilitate the data analysis procedure. The full list of variables and scale of measurements can be seen in Appendix I and Appendix II.

When the data collection was finished all data was checked for errors. By calculating the sum of each ordinal variable it could be concluded that each variable ended up in one hundred percent. The discrete data was double-checked against annual reports to make sure the interview representatives gave accurate numbers. Furthermore, no corrections for missing data needed to be complemented in hindsight.

Presentation of Collected Data

When the data had been digitalized, checked for errors and the scale of measurement and codes had been assigned it was time to determine how the data should be presented. In total the interview guide consisted of 15 questions and 47 variables. However, some of the variables are interrelated and it is therefore more suitable to present the data question by question. As stated before most of the data are either discrete or ordinal, and ordinal can often be handled as discrete data. Also, the data should not be presented so that each question can be attributed to a specific organization. With this in mind the most appropriate diagram was found to be a box plot. Therefore, the average, median, minimum, maximum, upper percentile and lower percentile were calculated and can be seen in Appendix III and Appendix IV.

Determine Correlations

To narrow down the potential relationships Spearman’s rank correlation coefficient was used, which is one of the oldest and best known of nonparametric analysis procedures (Zar 1972; Saunders 2009). It is a measure of statistical dependence between two variables that describes the strength of the relationship using a monotonic function. The correlation coefficient results in a value within the range -1 and 1 (see Figure 3). It was chosen to further analyze the relationships outside the range of -0.4 and 0.4. This number filtrated out the weakest correlations and made the number of possible correlations to further examine more manageable. A more detailed example of Spearman’s rank correlation coefficient is available in Appendix VII.

Figure 3: Values and labels of the correlation coefficients (Saunders 2009)

Furthermore, the relevant variables that had correlations within the accepted range were studied further. To make a conclusion about the correlation another literature review was conducted. If no explanation of the correlation could the found in the literature the conclusion was based on logical reasoning. In some cases, neither literature references nor logical explanation could be found to explain the correlation and was therefore not discussed thoroughly.

-1 -0.7 -0.3 0 0.3 0.7 1

Perfect

negative Strong

negative Weak

negative Perfect

independence Weak

positive Strong

positive Perfect

positive

(24)

Ranking of Selection Criteria

To evaluate the relative importance of different selection criteria in the supplier selection process a systematic approach based on fuzzy logic was used. Fuzzy logic is a powerful tool that can quantify imprecise data and handle vague and incomplete data where absolute values cannot be determined (Klir & Yuan 1995). Conventional logic is composed by true or false sets whilst Fuzzy logic attempts to establish a partial truth similar to what is accomplished by probability theory. The application of a fuzzy logic implies that the relative importance of each selection criterion can be defined using linguistic variables instead of numerical. This is particularly effective in decisions where fuzzy expressions are more natural to describe the phenomena than mathematical equations (Bevilacqua &

Petroni 2002). Every purchasing manager determined an importance level of each criterion using a linguistic variable. Five importance levels were used: very low (VL), low (L), medium (M), high (H) and very high (VH).

Figure 4: Triangular membership function to weight the linguistic importance of the selection criteria, with the fuzzy numbers stated by Liang & Wang (1993): VL(0,0,0.3); L(0,0.3,0.5); M(0.2,0.5,0.8);

H(0.5,0.7,1); VH(0.7,1,1)

The linguistic variables were then translated into fuzzy numbers through the definition of appropriate membership functions. The membership function was built on the assumptions and prescriptions following Liang & Wang (1993). Triangular fuzzy numbers was chosen due to the benefits of an uncomplicated computational procedure compared to other membership functions (Leung & Cao, 2000). Figure 4 is an example of the membership function adopted from the study conducted by Liang & Wang (1993). If a criterion is weighted “0.6”, the degrees of membership to the “low” and “medium”

functions are found as the projection of the horizontal intersection originating from the value “0.6” on the x-axis. When the weight matrix was established an approach described by Facchinetti, Ghiselli Ricci and Muzioli (1998) was used to determine the final ratings of the selection criteria. The method prescribes the calculation of a score for each selection criteria simply using the formula: (c

1

+2c

2

+c

3

)/4. An example with detailed calculations is available in Appendix VIII.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.1 1

0 1

VL L M H VH

(25)

Analysis of Selection Criteria

The influence of contingency factors on phase two in the supplier selection process, formulation of criteria, was analyzed separately. As mentioned above, the rank of the selection criteria is obtained by fuzzy logic. The rank of the criteria was compared with two earlier studies in the area, that both focused on ranking selection criteria. These studies were literature reviews that involved many types of industries and the methodology of how the ranking was accomplished may differ. Furthermore, a regression analysis was conducted to determine the influence of buying-specific factors, organizational factors and inter-organizational factors on the selection criteria. The influencing factors were grouped together to determine the relative change in importance when comparing high and low levels. This methodology was chosen mainly due to the fact that examining each criterion against each influencing factors would be too time consuming.

3.7 Method Problems

In order to reduce method problems two particular phases in the research must be considered, reliability and validity. In literal sense, it is very difficult to ensure that the research result is both valid and reliable. However it is possible to reduce the risk regarding these factors (Saunders 2009).

Reliability

Reliability refers to the stability in the measurement and analysis processes. The data collection and analysis procedure should generate consistent findings in order to ensure the research reliability. According to Robson (2002) there are four different threats to reliability related to interviews: subject or participant error, subject and participant bias, observer error and observer bias. Subject or participant error implies that interviews performed at different times of the week may generate different results. Therefore the interviews took place on a neutral time of the week, Tuesdays, Wednesdays and Thursdays, in order to try to standardize the participants’ moods.

Subject and participant bias refers to the threat of employment insecurity, for example that the interview participants may have been saying what they thought their bosses wanted them to say. To avoid this, the interviews were completely anonymous and in some cases a letter of confidentiality was signed. To further ensure the interview participants anonymity, no data were presented in such a way that it could be traced back to a specific organization.

Furthermore, two interviewers often have different ways of asking questions and interpreting replies; this is called observer error and observer bias. The same interviewer was used during the whole data collection procedure to minimize the impact from these factors.

Validity

Validity is concerned with whether the findings are really about what they appear to be

about (Saunders, 2009). It is important to reduce the risk to lose orientation and produce

outcome that does not answer the research questions. The concept validity can be divided

in internal and external validity. Internal validity refers to the flaws within the study itself

and external validity refers to the ability to generalize the findings to a larger context

(Saunders 2009).

(26)

Most of the concepts used in the literature review was critically analyzed and triangulated from multiple authors before included into the study. This was done to avoid dependence on a single source and incorporate different authors’ views. A threat to validity in this thesis is the ambiguity about the casual directions, described by Saunders (2009). Due to the fact that this study did not have control over all behavioral events it is hard to determine if factor (A) causes factor (B). The literature review points to the direction used in this study but in reality it may be factor (B) that causes factor (A), an unknown factor (C) that causes both factor (A) and factor (B), a combination of all factors or just an coincident.

Correlation does not directly imply causation but it gives a good hint on what to focus on.

(27)

4 Data Presentation

This chapter will present the data collected through interviews with representatives at the organizations in the Swedish pulp industry. The data are structured according to the literature review and provides a foundation for the upcoming analysis.

4.1 The Supplier Selection Process

The data collected for the supplier selection derives from questions one to five in the interview guide. The four phases in the supplier selection process of problem definition, formulation of criteria, qualification and final selection are presented in this section.

Problem Definition

The first phase in the supplier selection process is problem definition. The aim of this phase is to determine the ultimate problem or what an organization wants to achieve by selecting one or more suppliers. The first interview questions aims to determine how often a supplier selection process is initiated compared to developing own chemical facilities when facing a problem regarding chemical supplies. This assumes that the organization have some sort of in-house chemical production or research and development department.

The collected data shows that most organizations prefer to consult suppliers instead of solving the problem with in-house resources (see Figure 5).

Figure 5: Box plot illustrating the distribution of the two approaches to solve problems regarding chemical supplies

Nine of the sixteen respondents always initiate a supplier selection process when facing a problem with the chemical supplies where only three respondents consult their own chemical facilities more than half of the times. The average of initiating a supplier selection process is 85 % and the median is 100 %, skewing the data series in an upward direction.

Furthermore, the lowest percentage of initiating a supplier selection process results in an outlier at 20 % percent, while the second lowest percentage takes a leap up to 50 %. These data shows that it is almost six times more likely that an organization initiates a supplier selection process instead of developing own chemical facilities.

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Developing the own chemical facilities Initiating a supplier selection process

(28)

The next step in the problem definition phase is to determine the result of initiating a supplier selection process. According to the literature the aim to initiate a supplier selection process is to get fewer suppliers, more suppliers, replacing current suppliers or new way of dealing with the current suppliers. Figure 6 presents the data collected from interview question two and shows that the actual result of an initiated supplier selection process often is to reduce the supplier base.

Figure 6: Box plot illustrating the distribution of the four possible results of the supplier selection process The general trend is that organizations reduce their supplier base, with an average of 41 % and median 30 %. However, the data has rather high variance ranging from 8 % all the way up to 80 %. The second most common result is a new way of dealing with current suppliers with an average on 25 % and a median of 20 %, which ranges between 0 % and 70 %. The third most common result is replacing suppliers with an average on 22 % and a median of 20 %. This result ranges between 5 % and 50 %. The least common result is more suppliers with an average on 12 % and a median on 10 % with a range between 0 % and 30 %.

Formulation of Criteria

The next step in the supplier selection process is to formulate and weight selection criteria that the suppliers are judged by. All of the criteria have been formulated from the literature review to ensure relevance and later weighted by the interview participants. Table 3 shows that the five most important selection criteria in the Swedish pulp industry are delivery, quality, reliability, technical capacity and flexibility. Price is recognized as the sixth most important criterion. Geographical location and management and organization are considered the least important factors of the fourteen that has been included in this study.

The weight for each criterion ranges between 0.91 (delivery) to 0.50 (geographical location), where the possible theoretical range is between 1.00 and 0.00. The largest difference of two adjacent criteria is 0.13, involving reliability and technical capability. Furthermore, the fifth and sixth criteria, and the seventh and eight criteria have the same weight, which implies an equal rank.

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Fewer suppliers More suppliers Replacing suppliers New way with

suppliers

(29)

Table 3: Relative importance of fourteen selection criteria weighed with fuzzy logic

Rank Factor Weight

1 Delivery 0.91

2 Quality 0.90

3 Reliability 0.89

4 Technical Capability 0.76

5 Flexibility 0.75

6 Price 0.74

7 Service 0.73

8 Attitude 0.72

9 Performance History 0.71

10 Production Facilities and Capacity 0.67

11 Communication System 0.60

12 Financial Position 0.54

13 Management and Organization 0.52

14 Geographical Location 0.50

Qualification

The aim of the qualification phase is to reduce the total number of supplier further in order to find acceptable supplier to solve the problem. This step is measured in how many quotations a supplier selection process involves. The average number of quotation requests that are sent to potential suppliers in the qualification phase of the supplier selection process is presented below (see Figure 7).

Figure 7: Box plot illustrating the distribution of the number of quotation requests sent to suppliers

Figure 7 are based on the answers regarding the average number of quotation requests sent to suppliers that are considered as acceptable, in other words the suppliers that qualify to the next phase, final selection. The average numbers of quotation requests sent to acceptable suppliers ranges between two and six with an outlier with fifteen quotations.

Furthermore, the average is four quotations and the median is three quotations.

0 2 4 6 8 10 12 14 16

Quotations

(30)

Final Selection

Final selection is the step where the final selection of the suppliers or the group of suppliers that solves the ultimate problem is determined. This can result in a supplier that either delivers one product (single-deals), multiple products within the same product category (package-deals) or multiple products from different product categories (multiple- deals). As seen in Figure 8 package-deals is the most common procedure, followed by single-deals and multiple-deals.

Figure 8: Box plot illustrating the distribution of the three types of deals in the final selection phase

The average for package-deals is 47 % and the median is 50 %, with the data ranging between 0 % and 90 %. Single-deals are the second most common result in the phase final selection with an average on 30 % and a median of 20 % and a range equal to package- deals. Multiple-deals are least common with an average of 20 % and a median of 18 %, with a range between 0 % and 65 %.

4.2 Buying-specific Factors

The data collected for the buying-specific factors derives from questions six and seven in the interview guide. The factors that are presented in this chapter are type of purchase and time pressure.

Type of Purchase

New task, modified rebuy and straight rebuy refer to the type of purchase that is associated with the supplier selection process. In a new task both the supplier and the product is new, in modified rebuy either the supplier or the product is new and in straight rebuy neither the supplier nor the product is new to the organization. Figure 9 shows that straight rebuy is by far the most common buy class, followed by modified rebuy and new task for the last twelve months.

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Single-deals Package-deals Multiple-deals

References

Related documents

The primary contribution of this study is that it proposes and tests a series of hypotheses pertaining to the factors that influence adoption of CRM technology at organizational

Supplier development elements like; communication, collaboration, product development, top management involvement and long term commitment are considered as high

Therefore, the purpose of this study was to explain and contribute to an increased understanding of the issues small hardware manufacturing companies face when selecting

Based on our research and understanding of Volvo Group’s supplier selection process, we will further attempt to understand how the supplier evaluation model is actually being

(2013) provided a new group of supplier evaluation techniques, known as a multi- criteria decision-making approach and stated that these techniques can be considered an

Keywords: Sustainable supplier evaluation; Logistics industry; Sourcing; Benchmarking; Technique for Order of Preference by Similarity to the Ideal Solution (TOPSIS);

Company 2 does not conduct a proper analysis of the distributor, which is in compliance with the muddling through process (Schweizer, 2012). The difference between the

In order to study the phenomena of the sensemaking during crisis this article will address the restructure process that occurred in the Migration board due to the refugee