1
The consequences of supply base
characteristics on the number of drug
recalls
Master’s Thesis Supply Chain Management
Rijksuniversiteit Groningen
Faculty of Economics and Business
June 24
rd, 2019
Author: R. (Rik) Stoeten
Student Number: S3523292
E-mail:r.stoeten@student.rug.nl
Word count: 9314
Supervisor: dr. X. (Bruce) Tong
University of Groningen, Faculty of Economics & Business
Co-assessor: dr. ir. T. (Thomas) Bortolotti
2
Abstract
This study investigates whether and to what extent supply base characteristics are associated with the number of drug recalls in the pharmaceutical industry. Prior research has linked supply base complexity to a decrease in firm performance and to supply chain disruptions. However, empirical evidence is missing that describes whether and to what extent supply base complexity could potentially influence the number of drug recalls in the pharmaceutical industry. In this research the direct relationship between horizontal supply base complexity (defined as the number of first-tier suppliers), spatial supply base complexity (the geographic diversity between first-tier suppliers) and the number of drug recalls is investigated. Furthermore, this study explores the moderation effect of spatial supply base complexity. Data from United States pharmaceutical buying firms were collected and analysed using a negative binominal regression analysis. The results of this research not only suggest that both horizontal and supply base complexity directly increase the number of drug recalls, but also that spatial supply base complexity moderates the direct relationship between horizontal supply base complexity and the number of drug recalls in opposing ways depending on the level of horizontal supply base complexity (the number of first-tier suppliers). At high levels of horizontal supply base complexity, it is better to have a dispersed supply base, while at low levels of horizontal supply base complexity, it is better to have a concentrated supply base.
3
Content
1. Introduction ... 5
2. Background Information and Literature Review... 7
2.1. Drug recalls and the link to supply chain quality management ... 7
2.1.1. Quality management in the pharmaceutical industry ... 7
2.1.2. Quality management in supply chains ... 8
2.2. Supply base complexity ... 8
2.2.1. Supply base ... 8
2.2.2. System complexity ... 9
2.2.3. Supply base complexity ... 9
3. Conceptual Framework and Hypothesis Development ... 11
3.1. Horizontal supply base complexity and the number of drug recalls ... 11
3.2. Spatial supply base complexity and the number of drug recalls ... 12
3.3. The moderating effect of spatial supply base complexity ... 13
4. Method... 14
4.1. Data and procedure ... 14
4.2. Variable definitions and measures ... 16
4.2.1. Measure of the dependent variable ... 16
4.2.2. Measure of horizontal supply base complexity ... 16
4.2.3. Measure of spatial supply base complexity... 16
4.2.4. Measure of the moderator variable ... 17
4.2.5. Measurement control variables ... 17
4.3. The empirical model ... 17
5. Analysis and Results ... 19
5.1. Hypothesis testing ... 20
5.2. Horizontal supply base complexity results (hypothesis 1) ... 21
5.3. Spatial supply base complexity results (hypothesis 2)... 21
4
6. Discussion ... 24
6.1. Interpretation of results ... 24
6.2. Theoretical implications ... 26
6.3. Managerial implications ... 27
6.4. Limitations and future research directions ... 27
Literature ... 29
Appendices ... 33
Appendix A - Frequency table of the dependent variable... 34
Appendix B – SPSS output model 1 ... 35
Appendix C - SPSS output model 2 ... 36
Appendix D - SPSS output model 3 ... 37
5
1. Introduction
In 1999, the drug-manufacturing giant Merck introduced Vioxx, a prescriptive drug for treating pain associated with osteoarthritis (Krumholz, Ross, Presler, and Egilman, 2007). In 2004, the drug, which achieved a peak sales revenue of approximately 2.5 billion dollars a year, was recalled from the market because research revealed that the drug increased the risk of a heart attack by 400% (Union of Concerned Scientists, 2017). The consequences were horrendous. It is expected that the drug has resulted in thousands of premature deaths, ca. 100,000 heart attacks (Union of Concerned Scientists, 2017), and nearly 27,000 people filed a lawsuit against Merck. In 2007, these lawsuits were settled for 4.85 billion dollars (Wadman, 2007). In the aftermath of the recall announcement, the stock value of Merck dropped from ca. 44 dollars per share to ca. 12 dollars per share, resulting in a market decapitalization of 25 billion dollars (Neilan, 2004). This example demonstrates that the consequences of a recall can be severe, not only for the people using the drug but also for the manufacturing firm and its shareholders. It is therefore necessary to keep the number of drug recalls as low as possible. Despite the consequences, the number of drug recalls has steadily increased since 2013 (Jayaraman, Alhammadi, Can, and Simsekler, 2018; Nagaich and Sadhna, 2015; Smith, 2018). Surprisingly, no empirical evidence is found that identifies factors that explain this increase in drug recalls. However, it is commonly believed that supply base1 complexity is one of the primary factors causing performance
problems across a wide variety of industries (Choi and Krause, 2006). Gray, Roth and Leiblein (2011) found that a greater geographic distance between buyer and supplier increases supply base complexity, which in turn compromises product quality. Bode and Wagner (2015) found evidence that complexity in the supply chain increases the likelihood of supply chain disruptions. While these studies provide valuable insights into the consequences of supply base complexity, no empirical evidence is found that links supply base complexity to the observed increase of drug recalls in the drug manufacturing industry. It is surprising that this question was not asked previously, despite the benefits it could bring for patient safety and social responsibility. Therefore, this research attempts to address this gap by answering the following question: ‘Whether and to what extent do structural characteristics of the
supply base of drug manufacturing firms impact the number of drug recalls?’
Studies regarding product recalls in the pharmaceutical industry demonstrated the negative impact that a drug recall has on financial metrics such as shareholder wealth, brand equity and company reputation (Zhao, Li, and Flynn, 2013). Research revealed also that a drug recall negatively impacts supply shortages and patient safety (Azghandi, Griffin, and Jalali, 2018; Ball, Shah, and Donohue, 2018). Some researchers have investigated factors that increase the likelihood of drug recalls, but these studies focussed either on the impact of product scope and price competition (Ball, Shah, and Wowak, 2018) or on the effect of previous recall experience (Thirumalai and Sinha, 2011). Due to the increase
1The group of suppliers with whom the focal company has a direct relationship are called the ‘supply base’
6 in outsourcing and globalisation, many companies become more dependent on a diverse set of suppliers worldwide. This has increased the complexity and managerial effort to manage them. According to Choi and Krause (2006), supply base complexity can be defined using three dimensions: ‘(1) the number
of suppliers in the supply base, (2) the degree of differentiation of these suppliers [in terms of different geographical distance, cultural distance, organisational cultures and political situations], and (3) the level of inter-relationships among the suppliers’ (p.641). These dimensions have been studied by
scholars and are associated with changes in quality performance in the consumer goods industry (Steven, Dong, and Corsi, 2014) and to decreases in plant performance (Bozarth, Warsing, Flynn, and Flynn, 2009). As drug recalls are linked to violations of the Current Good Manufacturing Practices (cGMP) (FDA, 2019), a framework developed by the United States Food and Drug Administration (FDA) to guarantee drug quality, the link to quality management is essential. Due to globalisation and the increased levels of outsourcing, quality should be managed across the entire supply chain and by every company within it. However, I argue that quality management across the supply chain (known as supply chain quality management) is heavily complicated by the increase in supply base complexity. This in turn is expected to comprise drug quality and increases the number of drug recalls.
This research offers two contributions. First, this research contributes to supply chain literature by proposing and empirically investigating linkages between supply base characteristics and drug recalls assessed through supply chain quality management. Second, as globalisation in the pharmaceutical industry is expected to continue rising (Torreya, 2017), the information found in this research can be used by managers to define and (re)shape their outsourcing strategies. To answer the research question, I rely on secondary and publicly available news announcements concerning the buyer-supplier relationships in the drugs-manufacturing industry. This data is combined with drugs recalling data from the FDA database. Via a statistical regression analysis, I test my theory that supply base (complexity) characteristics are linked to the number of drug recalls.
7
2. Background Information and Literature Review
This chapter presents the theoretical background of this research. It consists of a thorough review of the literature that is used in Chapter 3 to develop the hypotheses.
2.1. Drug recalls and the link to supply chain quality management
2.1.1. Quality management in the pharmaceutical industry
Companies that operate in the pharmaceutical industry are heavily regulated by agencies. In the United States, that agency is the U.S. Food and Drug Administration (FDA). The FDA has the essential task of ensuring patient health and safety (Atella, Bhattacharya, and Carbonari, 2011). The FDA is responsible for both the assessment and approval of new drugs and for monitoring the risk, quality and effectiveness of drugs on the market (FDA, 2016). Field inspectors of the FDA verify a firm’s operation on a regular basis to ensure that the firm complies with Current Good Manufacturing Practices (cGMP). The cGMP is ‘the main regulatory standard for ensuring pharmaceutical quality’ (FDA, 2018) and ensures ‘that quality is built into the design and manufacturing process’ (FDA, 2018). It requires that ‘manufacturers of medications adequately control manufacturing operations. This includes
establishing strong quality management systems, obtaining appropriate quality raw materials, establishing robust operating procedures, detecting and investigating product quality deviations, and maintaining reliable testing laboratories’ (FDA, 2018). Regardless of these regulations, it remains
possible that a drug violates the laws (cGMP) of the FDA. Whenever a violation is detected, the manufacturer should, on their own initiative or by the request of the FDA, remove the drug from the market (a recall). Such a recall is performed to protect the public from a defective or potentially harmful product (FDA, 2019). Recalls are likely to occur whenever a drug ‘cause health hazard, is mislabelled
or packaged poorly, is potentially contaminated, is not what it says, is poorly manufactured’ (Nagaich
& Sadhna, 2015: 17) Three of the five reasons for a product recall are linked to the quality and the production of the drug, which provides the importance of safeguarding the quality during production and transportation. Although cGMP is the primary regulatory and obligatory framework for drug manufacturers, it should be considered as part of quality management (World Health Organization, 2006). This indicates that cGMP is interrelated with the concepts of quality assurance and quality control (World Health Organization, 2006). The FDA states the following about this interrelationship:
8
2.1.2. Quality management in supply chains
Current Good Manufacturing Process describes in detail the practices that pharmaceutical manufacturers are obligated to honour. However, a large limitation of the framework is that it neglects practices to incorporate suppliers into quality management. This is surprising, because research has discovered that increased performance is not solely achieved by improving internal processes but that the focus should be also on inter-organisational processes and relationships (Mellat-Parast, 2013). This is magnified by the globalised business environment in which increased outsourcing made firms dependent on the quality of their suppliers (Soares, Soltani, and Liao, 2017). An external view towards supply chain members is therefore essential to guarantee the desired quality of the end product. This need inspires scholars to research quality practices within a supply chain setting. These scholars argue that quality-related tools can be well combined with supply chain management activities (Mellat-Parast, 2013). This relationship between quality management and supply chain management is, in the literature, known as ‘Supply Chain Quality Management (SCQM)’ (Soares et al., 2017).
The most used definition of SCQM in literature is provided by Robinson and Malhotra (2005: 319) and defined as follows:
SCQM is the formal coordination and integration of business processes involving all partner organizations in the supply channel to measure, analyze and continually improve products, services, and processes in order to create value and achieve satisfaction of intermediate and final customers in the marketplace. (Robinson and Malhotra, 2005: 319)
From this definition, it becomes clear that the coordination and integration of all inter-firm processes, products, services and work cultures within the supply chain are the key constructs of SCQM. Close relationships between supply chain members are essential enablers of SCQM, and key constructs hereby are coordination and integration (Kaynak and Hartley, 2007). This is because close relationships between supply chain members can enable communication, participation and collaboration, which in turn can facilitate the development of specific inter-organisational processes (Mellat-Parast, 2013). Such close relationships are based on trust, open information sharing and co-makership and are indeed found to increase the quality of products (Flynn and Flynn, 2005).
2.2. Supply base complexity
2.2.1. Supply base
9 (Choi and Krause, 2006). The decision to actively manage suppliers is often strategic and related to (1) acquiring important (tacit) resources, (2) the possibility of managing the quality and safety of the materials and (3) the possibility of achieving stability by increased trust and interdependence (Mena, Humphries, and Choi, 2007). As Section 2.1 concluded, the relationships with suppliers are important to guarantee product quality. However, due to increased complexity in the supply base, this seemingly simple task of connecting to first-tier suppliers can be greatly complicated.
2.2.2. System complexity
Complexity is a term used across a great number of academic disciplines and research fields (Bode and Wagner, 2015). Its conceptualisation and measurement greatly depend on the context in which the term is used, but it appears that all authors attempt to use complexity as a construct to explain, predict and control chaotic systems (Caridi, Crippa, Perego, Sianesi, and Tumino, 2010). System complexity is often used as an overarching construct from which more specific types of complexity, such as supply base complexity, are derived. Bozarth et al. (2009) performed a detailed literature review regarding the origins and development of system complexity and concluded that no formal definition of system complexity exists. However, many authors use the same overarching building blocks to describe system complexity. These building blocks are (1) the structure and the number of components/elements in a system and (2) the way components/elements connect, intertwine and interact with one another. To state this more simply, complexity concerns the ‘structure’ and ‘behaviour’ of elements within a system (Andersons; Burnes; Perow; Senge, in Bode and Wagner, 2015). The structural part of complexity is often termed detail complexity, while the behavioural portion is often termed dynamic complexity (Bode and Wagner, 2015; Bozarth et al., 2009). It should be noted that these two ‘building blocks’ of complexity are often seen as interrelated. This occurs because when the structure of a system expands, the number of possible interactions between elements also expands, which increases the variety and behaviours within a system and makes it more complex (Bode and Wagner, 2015).
2.2.3. Supply base complexity
Supply chain scholars used system complexity theory to describe the complexity within the upstream part of the supply chain. Choi and Krause (2006) defined supply base complexity as a concept associated with three dimensions, namely: ‘(1) the number of suppliers in the supply base, (2) the degree of
differentiation of these suppliers [in terms of different geographical distance, cultural distance, organisational cultures and political situations], and (3) the level of inter-relationships among the suppliers’ (Choi and Krause, 2006: 637). These authors argued that the three dimensions (structural
10 of the upstream supply chain. These dimensions are (1) horizontal complexity, (2) vertical complexity and (3) spatial complexity. Horizontal complexity is defined as the total number of first-tier suppliers, vertical complexity as the total number of upstream tiers and spatial complexity as the dispersion between supply chain members (measured as the average geographical distance between members) (Choi and Hong, 2002). Bode and Wagner (2015) researched whether supply chain complexity increases the frequency of supply chain disruptions. These authors used horizontal, spatial and vertical
supply chain complexity as explanatory variables and concluded that especially horizontal and spatial
complexity are important elements in predicting disruptions.
Table 2.1 summarises the possible elements of supply base complexity together with the corresponding literature.
As can be observed in Table 2.1, most evidence is found to suggest that horizontal complexity (defined as the number of first-tier suppliers) and spatial complexity (defined as the geographical diversity between suppliers) are the most relevant drivers of supply base complexity. Therefore, the decision is made to employ these two elements as the basis for developing the hypothesis in the following chapter.
11
3. Conceptual Framework and Hypothesis Development
Based on the literature review regarding drug recalls, supply chain quality management and supply base complexity, the following hypotheses can be developed.
3.1. Horizontal supply base complexity and the number of drug recalls
Two arguments are used to explain the link between horizontal supply base complexity (number of suppliers in the supply base) and the number of drug recalls. First, enlarging the supply base by involving a large number of suppliers (increased horizontal complexity) is linked to ‘increased number
of information flows, physical flows and relationships that must be managed’ (Bozarth et al., 2009: 81).
This results in an increase in the coordination and administrative tasks of the focal firm and increases the risk of misinformation and supply chain disturbances (Bode and Wagner, 2015; Steven et al., 2014). By having many suppliers, it becomes much more difficult to maintain a close relationship with all suppliers (Choi and Krause, 2006). This could have disastrous consequences for the product quality. Without the close supplier relationships, it is unlikely to arrange an effective integration and collaboration, the cornerstones of quality management in the supply chain (SCQM), among supply chain members (Kaynak and Hartley, 2007; Robinson and Malhotra, 2005; Zu and Kaynak, 2012). Without SCQM (and integration and collaboration) in place, it is unlikely that all information is shared willingly and openly among members. This increases information asymmetry between the supplier and buyer and decreases the control that the buyer has over the quality of the supplier (Steven et al., 2014). This in turn increases the potential for opportunistic behaviour of the supplier (Jensen and Meckling, in Steven et al., 2014), who can therefore progressively lower the quality of raw materials without the buyer noticing (known as quality fade2) (Midler, 2007) and thereby compromising drug quality. This
link between opportunistic behaviour and quality problems is made extensively in research across numerous industries and is revealed to be relevant (Gray et al., 2011; Steven et al., 2014; Whipple and Roh, 2010; Zu and Kaynak, 2012). Furthermore, many researchers have found that effective SCQM practices lead to increased quality performance (Sila, Ebrahimpour, and Birkholz, 2006; Soares et al., 2017). This causes me to believe that SCQM is necessary to ensure drug quality.
Second, by enlarging the supply base (increased horizontal supply base complexity), suppliers experience higher degrees of freedom. This occurs because the focal firm cannot control all the elements in the system. This decreases supply chain visibility and traceability and results in the need to selectively conduct quality control (Choi and Krause, 2006; Steven et al., 2014). This makes it much more difficult to enforce the quality control program of the FDA (CGMP), that obligates manufacturers to guarantee the quality of all raw materials. When suppliers experience more freedom, it is likely that some suppliers
2 ‘a deliberate and secret habit of widening profit margins [by decreasing costs] through a reduction in
12 will use the situation to develop opportunistic behaviour (Steven et al., 2014). Subsequently, opportunistic behaviour of suppliers is linked to potential breaches of product (drug) quality (Midler, 2007).
The two arguments discussed above lead me to expect the following:
Hypothesis 1: Horizontal supply base complexity is positively related to the number of drug
recalls.
3.2. Spatial supply base complexity and the number of drug recalls
Researchers seems to be clear about the definition of spatial supply base complexity, namely: the geographical and cultural spread of suppliers (Bode and Wagner, 2015; Choi and Hong, 2002; Choi and Krause, 2006; Gray et al., 2011). Two arguments support my believe that increased levels of spatial complexity will increase the number of drug recalls. First, to guarantee and improve the quality of end products, SCQM practices are found to be essential (Sila et al., 2006; Soares et al., 2017). Effectively implementing these practices across firm boundaries requires in-depth tacit knowledge, great coordination and day-to-day adherence (Gray et al., 2011). Effective implementation is found to be impeded by the distances between supply chain members (Gray et al., 2011). Especially cultural distance, rather than geographical-, educational- and firm-level differences, is linked to difficulties in implementing SCQM practices along the chain and in offshore plants (Gray et al., 2011). This is because cultural distance is related to language nuances and communication difficulties (Gray et al., 2011). In addition, establishing close relationships with suppliers is found to require substantial resources and long-term commitment from both sides, which greatly increases when large geographic and cultural distances separate supplier and buyer (Steven et al., 2014). With fewer possibilities to enable close relationships with supply chain members, SCQM practices are less likely to be effective, and suppliers are less likely to consistently adhere to quality control programmes (the cGMP framework of the FDA). This could result in compromises of drug quality.
Second, when the supply base is dispersed across numerous geographic regions, it becomes increasingly difficult to coordinate the various cultures, strategies and processes (Huo, Ye, Zhao, and Zhu, 2019). Furthermore, globalisation is found to expose companies to factors that make efficient and effective supplier management even more challenging. These factors are defined as: ‘different
import/export laws, fluctuations in currency valuations, cultural differences and longer and more uncertain lead times’ (Cho and Kang in Bozarth et al., 2009: 82). Again, differences between suppliers
are expected to compromise product quality.
These two arguments result in the following expectation:
13
3.3. The moderating effect of spatial supply base complexity
In hypotheses 1 and 2, I argue that a direct relationship exists between both horizontal supply base complexity, spatial horizontal complexity and the number of drug recalls. However, prior research has found evidence that ‘the effect of horizontal complexity on the number of [supply chain] disruptions is
weak at low and medium levels of vertical and spatial complexity, but strong at high levels of vertical and spatial complexity’ (Bode and Wagner, 2005: 223). This view appears logical and sparked the idea
to test this statement in an adapted way in this research. What Bode & Wagner (2015) argue is that it is unlikely that elements of a system (a supply base is a system, as argued in subsection 2.2.2.) act solely on their own. Instead, due to the nature of elements, elements within a system form complementary and reinforcing relationships (Ennen and Richter in Bode & Wagner, 2015). This view appears reasonable, as I consider that one of the building blocks of system complexity describes that components/elements of a system connect, intertwine and interact with one another. In this manner, one element can potentially change the behaviour of another element. Seen from this perspective, it appears that it is more difficult to manage and coordinate many suppliers from different geographic regions, than it is to manage many suppliers from the same geographic regions. A supply base within several countries is more complex to manage and causes monitoring to be more difficult and costly. This increases the possibility for suppliers to disobey quality management practices, which in turn could decrease drug quality and increase the number of drug recalls.
It should be noted that the idea to test this hypothesis comes from reading the paper of Bode and Wagner (2015). However, there are some great differences between their research and mine, which makes this hypothesis still relevant to research. First, they operationalized their dependent variable
supply disruption as follows: ‘In 2006, approximately how many supply disruptions did you experience?’ (Bode and Wagner, 2015: 225) and collected the responses in a self-administered survey.
A supply disruption is defined as any situation that has a negative consequence for the firm. I find this question quite vague, the term supply disruption rather broad defined and for these reasons rather subjective. Because of this approach, different supply chain disruptions, severe and minor, are aggregated to form one large database. Second, Bode & Wagner (2015) have not focused on a specific industry, but have received responses from 396 firms operating in 13 industries. They have combined all responses and have not drawn industry specific conclusions. However, I belief that industry specific differences are an important element that could influence the dependent variable. I have decided to specifically focus on the number of drug recalls in the U.S. pharmaceutical industry and I use the sum of all drug recalls over a 10-year period. I belief that this approach differs enough from that of Bode & Wagner (2015) and is therefore still relevant to investigate in this research.
Hypothesis 3: The positive relationship between horizontal supply base complexity and the
14
4. Method
This section develops the empirical model to test the hypotheses proposed above. Figure 4.1 illustrates the conceptual model for this study. This model is based on the relationship between drug recalls and horizontal and spatial supply base complexity. I expect that both horizontal (H1) and spatial (H2) supply base are directly and positively related to the number of drug recalls. Furthermore, I expect that spatial supply base complexity not only has a direct affect, but also moderates the relationship between horizontal supply base complexity and the number of drug recalls (H3). This moderating effect is indicated by the dashed line in figure 4.1.
4.1. Data and procedure
This study is limited to publicly traded companies within the U.S. drug manufacturing industry. Several reasons exist for doing so. First, drug recalls are often due to manufacturing malfunctions (such as, among others, contamination, incorrect labelling and lack of sterility (Nagaich & Sadhna, 2015), so companies encountering drug recalls are most often manufacturing companies. Second, more information is available for public firms. This is because public firms are required to make regular U.S. Securities and Exchange Commission (SEC) filings, which contains valuable information to construct the variables. Private firms are not required to make SEC filings, which makes them unsuitable for data collection and are thus excluded for this research.
Data for this analysis was gathered from two sources. The data to construct the two independent variables was collected from the database Factiva. This is a database that gathers global news and business information from more than 32,000 licensed and free sources into a single database. With use of the search string in Figure 4.2, which is provided by my supervisor, 24,737 announcements regarding buyer-supplier relationships are found over the period 1980–2017. My supervisor tested the validity of
15 the search string and concluded that the search string showed a high efficiency in capturing valid samples.
The initial plan was to conduct the coding process by five master’s students, but two students stopped their thesis projects hallway through. The coding of all 24,737 news announcements by the remaining three students would require too much time, and therefore, the collective decision was made to focus on the period 2008–2017. This period consists of ca. 9,000 news announcements. All 9,000 news announcements were first screened and considered applicable when they satisfied the following criteria:
• The announcement should be made from a company that operates in the drug industry; • The supplier-buyer relationship should concern drugs;
• When an announcement was published more than once, the earliest announcement was considered as applicable.
Whenever a news announcement was considered applicable, it was coded using the coding scheme in Table 4.1. In total, 2340 supplier-buyer relationships were found and coded. Hereafter, the buying firms (‘buyer’ in Table 4.1) were separated into public and private firms; only buyer-supplier data for public buying firms were retained for this study. As a result, 1,263 relationships were kept for further analysis. Analysis revealed 544 unique public buying firms, together accounting for the 1,263 supplier-buyer relationships.
TABLE 4.1—CODING SCHEME
The data for the dependent variable was compiled from the publicly available FDA database on drug recalls. Subtracting and coding the data from this database was performed by students from Hong Kong. The dataset includes all drug recalls in the United States from 1980 onwards, but only the recalls that occurred in the period 2008–2017 (the same timespan as that for the independent variables) were retained for this study. A preliminary analysis of the FDA database revealed that multiple drugs produced by the same manufacturer were often recalled on the same day. Each product was individually archived in the dataset, but they had an identical recall event number, identical reason for recall and occurred on the same day. See Figure 4.3 for an example. Because those recalls had an identical reason for recall, the decision was made to count such a recall event as one, not as four. By deleting duplicates based on ‘recall event’, a new dataset was compiled in which only unique recall events were listed. In total, 2,698 unique recall events were documented. Of this total number, 1,043 recall events could be
News title Start time End time Supplier Parent firm Private or Public country buyer Parent firm Private or Public2 country (of buying firm)
(("suppl!" or "provide!" or "purchase!" or "buy!" or "bought" or "sell!" or "sold" or "acquire!" or "offer" or "distribut!") and ("contract" or "agreement"))
16 linked to the 544 public buying firms that were found to have one or more supplier relationships. A count if function was used to link the number of recalls to the applicable buying firm.
FIGURE 4.3—EXAMPLE OF PRELIMINARY ANALYSIS FDA DATA SET
Data on the control variables were collected from their annual SEC 10k filing (annual report). Information was collected on each buying firm’s annual turnover and the firm foundation date. Fifty-nine buying firms did not make any revenue to date. These firms were in the drug research and development stage and had no drug to sell. I decided to exclude these firms from the data set because a recall event was not possible to happen to these firms, as they did not had a drug to market. After excluding these firms, 485 unique and valid buying firms (samples) remained. These buying firms were used for the analysis.
4.2. Variable definitions and measures
4.2.1. Measure of the dependent variable
To measure the dependent variable number of drug recalls, I followed Steven et al. (2014) and used the total number of recalls that a buying firm experienced in the period 2008–2017. When a buying firm did not experienced a recall, a 0 was stated.
4.2.2. Measure of horizontal supply base complexity
In congruence with Bode and Wagner (2015) and Bozarth et al. (2009), horizontal supply base complexity can be measured as the number of first-tier suppliers of the buying firm. This was performed by counting the unique buyer-supplier relationships found in the Factiva database for each buying firm over the period 2008–2017.
4.2.3. Measure of spatial supply base complexity
To measure spatial supply base complexity, I followed Stock, Greis and Kasarda in Bode and Wagner (2015) and calculated how the supply base of the buying firm is spread over the following four geographic regions of the world: North America (Canada, Mexico, United States), Europe, Asia/Pacific and the rest of the world. Stock et al. (2000) designed a formula (Figure 4.4) to indicate how evenly the supply base is spread geographically. The outcome of the formula is a range between 0 (concentrated suppliers within one region) to 1 (evenly spread over the four regions). The formula works in the following way. When the supply base is concentrated in one region, the percentage of that region will be 100% and the percentage of the three other regions will be 0%. Because the formula uses the absolute value of each geographic region, the outcome for the numerator, using the example above, equals [|100-25| + |0-[|100-25| + |0-[|100-25| + |0-[|100-25|] = 150. Subsequently, 150 divided by 150 equals 1 and 1-1 is 0. On the
Recall Event Recalling Firm Product Description Product Quantity Reason for Recall Recall Initiation Date
17 other side, when the supply base is spread evenly over all four geographic regions, the numerator equals [|25-25| + |25-25| + |25-25| + |25-25|] = 0. Subsequently, 0 divided by 150 equals 0 and 1-0 is 1.
FIGURE 4.4—FORMULA SPATIAL SUPPLY BASE COMPLEXITY (ADOPTED BY STOCK ET AL.[2010:540])
4.2.4. Measure of the moderator variable
To test the moderator variable, first the independent variables and the control variables were centred using the grand mean centring method. This is performed because the presence of a moderator variable can make the main variables uninterpretable (Field, 2009). Centring the variables solves this problem. Thereafter, the moderator variable was created by multiplying both centred independent variables into
Horizontal_SupplyBaseComplexity*Vertical_SupplyBaseComplexity.
4.2.5. Measurement control variables
The first control variable used in this research was firm size. Firm size was used by Steven et al. (2014) and by Bode and Wagner (2009) as a control variable because these authors argued that larger firms often sell various products (diversified product base) and are, therefore, more complex. With a more diversified product base, the probability on a recall is higher than when only one product must be managed. Firm size was measured as the yearly turnover per firm obtained from the latest SEC filing (annual report).
The second control used in this research was firm age and is adopted by Bode and Wagner (2015). These authors argued that firm age can function ‘as a proxy for knowledge, experience, and familiarity
with supply chain processes which may have an attenuating effect of the frequency of disruptions [recalls]’ (Bode and Wagner, 2015: 220). Firm age was measured ‘as the difference between the founding year and the year of data collection [2019]’ (Bode and Wagner, 2015: 220)
4.3. The empirical model
The dependent variable in the model is the buying firm’s number of recalls within the period 2008– 2017. The independent variables are horizontal supply base complexity and spatial supply base
complexity. The dependent variable can be categorised as a count variable because it consists solely of
integer data that is zero or greater as can be seen in the frequency table in Appendix A. The descriptive statistics demonstrated that the distribution of the data is positively skewed (µ = 2.13; σ =11.181; skewness = 10.636) and has a minimum of 0 and a maximum of 171 (Table 4.2).
18 TABLE 4.3—DESCRIPTION AND MEASUREMENT OF VARIABLES
Since the variance (σ2 = 125.025) is much larger than the mean (µ = 2.13), a high overdispersion
of the data can be assumed. The most suitable statistical test in such a case is the negative binominal regression analysis (a Poisson regression assumes an equal mean and variance and is therefore not suitable). In the case of a negative binominal regression analysis, ‘the log of the outcome is predicted
with a linear combination of the predictors’ (UCLA: Statistical Consulting Group, n.d.). This leads to
the following equation:
Because the log of the dependent variable (Recall_Event) can be deleted by taking the exponential of the predictors (UCLA: Statistical Consulting Group, n.d.), the equation can be rewritten as:
19
5. Analysis and Results
First, it should be noted that hypotheses 1 and 2 are tested with a model that does not contain the moderator variable. This is decided because centring variables, which should be performed when testing a moderator variable, changes the effect of the lower-order predictors (the direct effect of horizontal and spatial supply base complexity) and makes them less trustworthy (Field, 2009). Therefore, the decision is made to add the moderator variable only when the moderation is actually tested (in Section 5.4). This means that all the results discussed up to Section 5.4 are based on the model without the moderator variable.
Before conducting the negative binominal regression analysis, the data set is checked on the potential for multicollinearity between the independent variables. The independent variables should not correlate with one another because this weakens the statistical power of the regression model and makes the p values less trustworthy (Frost, n.d.). Table 5.1 provides the Pearson correlations between the variables. The highest Pearson correlation is .469, between Spatial_SupplyBaseComplexity and Horizontal_Supply BaseComplexity. This is quite high but is not above 0.7, which is considered the upper level, and is therefore unlikely to cause problems. This high value can be justified also because it is expected in hypothesis 3 that both variables are interrelated in some way.
TABLE 5.1—CORRELATION BETWEEN VARIABLES
Furthermore, Table 5.1 presents the variance inflation factor (VIF). The largest VIF value is 1.515 for Horizontal_SupplyBaseComplexity. This suggests that the effect of multi-collinearity is limited in this model, as each variable VIF estimate is less than 10, which is widely accepted as the upper limit (UCLA: Statistical Consulting Group, n.d.).
20 In Model 2 (M2), the outliners are identified by assessing the standardised residuals of the negative binominal regression analysis. It is widely accepted that standardised residuals greater than 3 and less than -3 can be considered as outliners (Laerd Statistics, n.d.); The samples concerning this criteria are deleted. As observable in Table 5.2, the overall fit of the model has increased by approximately 48% while maintaining a significant model (p ≤ 0.05). Therefore, the negative binominal regression analysis is performed using Model 2 (with the outliners deleted and N = 464).
TABLE 5.2—THE FIT OF THE MODELS
5.1. Hypothesis testing
To test the hypotheses the model (model 2: N = 464) was statistically tested using the statistical software package SPSS. Table 5.3 presents the results of the negative binominal regression analysis. First, I tested a model with only the control variables in Model 1 (M1). Afterwards, I added the two independent variables in the model to come to Model 2 (M2). The complete model M2 (AIC = 660.938; BIC = 681.637; p = .000) has a better fit (+32%) than the model with only the control variables M1 (AIC = 968.906; BIC = 981.326; p = .000). This reveals that the complete model better fits the data, is more reliable and can therefore be used for further analysis. The SPSS outputs of the models can be found in Appendices B and C, respectively.
21
5.2. Horizontal supply base complexity results (hypothesis 1)
Hypothesis 1 holds that higher levels of horizontal supply base complexity is positively related to the number of drug recalls experienced by buying firms. The results in Table 5.3 reveal that horizontal supply base complexity positively and significantly impacts the number of recalls experienced by the buying firm (β = .203; p = .000). This means that buying firms that increase their horizontal supply base by one supplier could experience an increase in the log of the number of drug recalls of .203 units (95% CI, .153 to .254) while holding the other variables in the model constant. By means of explanation, if two buying firms are compared that have the same spatial supply base complexity, the same yearly revenue and the same years of operations, the model expects the buying firm with the higher horizontal supply base to experience more recalls. The exponentiated value of the beta value (β = .203) is 1.225 (95% CI, 1.165 to 1.289). This implies that the number of recalls will rise by 1.225 (or 22.5%) for each additional first-tier supplier added to the supply base. Figure 5.1 displays the corresponding plot. As a result of the analysis, sufficient evidence is found to support hypothesis 1.
FIGURE 5.1-EFFECT OF HORIZONTAL SUPPLY BASE COMPLEXITY
5.3. Spatial supply base complexity results (hypothesis 2)
22 FIGURE 5.2-EFFECT OF SPATIAL SUPPLY BASE COMPLEXITY
5.4. The moderator effect of supply base complexity (hypothesis 3)
To test the moderator variable, the negative binominal regression analysis is performed again. However, now all variables (except the dependent variable) are centred using the grand mean centring method. The results of the analysis are presented in Table 5.4. First, a model is tested with only the centred control variables and the two centred independent variables (M3). Afterwards, the moderator variable was included in Model 4 (M4). Model 4 (AIC = 643.31; BIC = 668.149; p = .000) has a slightly better fit (+3%) compared to the model without the moderator variable, M3 (AIC = 660.938; BIC = 681.637; p = .000). This reveals that the model with the moderator variable has a slightly better fit and can therefore be used for further analysis. The SPSS outputs of the models can be found in Appendices D and E, respectively.
The results in table 5.4 reveal that spatial supply base complexity is a significant but negative moderator effect (β = -.513; p = .000). This implies that when the supply base gets more dispersed
23 worldwide (increasing spatial supply base complexity), the effect of the number of first-tier suppliers (horizontal complexity) on the number of drug recalls reduces. This is surprising, as the opposite was expected. To further investigate this effect, the measurement for spatial supply base complexity is divided into five categories (see Table 5.5). The first category equals a supply base that is not dispersed, while category five represents a supply base that is highly geographically dispersed. It should be noted that the frequencies of categories three and four are low compared with those of categories one and two. This leads to fewer data points that can be plotted and can make the results less trustworthy. Despite this drawback, the results can still provide valuable insights.
When plotting the moderator effect (Figure 5.3), several interesting insights are identified. First, at high levels of horizontal supply base complexity (many first-tier suppliers), the effect of low levels of spatial supply base complexity is more severe than high levels of spatial supply base complexity. This implies that a supply base that consists of many suppliers from the same geographic region is expected to experience more drug recalls than a supply base that has many suppliers dispersed worldwide. Second, firms with only a few first-tier suppliers (low horizontal supply base complexity) would experience more drug recalls when these suppliers are highly geographically dispersed. Third, when a firm has a supply base that is widely dispersed worldwide (high levels of spatial supply base complexity), it is better for such a firm to have many suppliers. Last, firms that have not yet decided to globalise their supply base (low levels of spatial supply base complexity) can better have a few supplier relationships (low horizontal supply base complexity).
FIGURE 5.3—MODERATING EFFECT AT DIFFERENT LEVELS OF SPATIAL SUPPLY BASE COMPLEXITY
24
6. Discussion
6.1. Interpretation of results
This research aimed to address which and to what extent supply base characteristics of pharmaceutical public firms affects the number of drug recall events occurring at these firms. This research built upon three hypotheses and an empirical study for testing them. Four key findings are identified.
Key insight one
The empirical results from this study reveal that increased complexity in the supply base is associated with an increase in the number of drug recalls. It is found that an increase in horizontal supply base complexity, or the number of first-tier suppliers, results in a significant increase in the number of drug recall events (β = .203; p = .000). It is expected that adding an additional supplier to the supply base results in a 22.5% increase in the number of drug recalls. This finding aligns with the research of Steven et al. (2014), who discovered that ‘concentrating the supply base is associated with better quality
performance’ (p. 251). In addition, my results align with system complexity theory. System complexity
argues that the number of elements in the system is an important factor in making the upstream supply chain more complex (Bozarth et al., 2009). This is because it is associated with an increase in the number of information flows and the number of relationships that must be managed by the focal firm. Since not all elements can be simultaneously controlled in such a complex system (Choi and Krause, 2006), supply chain disruptions (Steven et al., 2014; Bode and Wagner, 2015) and negative plant performance (Bozarth et al., 2009) are likelier to occur. In addition, the results can be discussed in the light of transaction cost economics. Having numerous first-tier suppliers is related to an increase in transaction cost due additional searching, contracting, monitoring, enforcement, and coordination costs (Steven et al., 2014). Since the objective of most organisations is to minimise these costs, it is unlikely that all suppliers are monitored extensively because this will be very costly (Ketchen and Hult, 2007). This is found to lower supply chain visibility and decrease quality performance. My findings indeed identify that the number of suppliers acts as a manifestation of such negative consequences.
Key insight two
25 FIGURE 6.1-MATRIX DISPLAYING VARIABLES
quality. Evidence is found to support that companies that have a highly geographically dispersed supply base indeed experience problems with product quality, measured as a recall event.
Key insight three
An important new insight, one not expected from theory, is that the results reveal a significant but negative moderating effect (β = -.513; p = .000). Previous research had suggested ‘that the effect of
horizontal complexity on the number of disruptions is weak at low and medium levels of vertical and spatial complexity, but strong at high levels’ (Bode and Wagner, 2015: 223). The explanatory
26 To explain the findings, theory regarding power in the supply chain can be used. Firms that use only a few suppliers are greatly dependent upon these suppliers. When something unwanted occurs concerning these suppliers or should they begin behaving opportunistically, the possibility for supply chain disruptions arises. In this case, it is easier to manage and coordinate suppliers from the same geographic region, as a one-plan-fits-all approach is applicable. In addition, when the buying firm has only a few suppliers, the firm can use coercive power to obtain (high-quality) resources from such suppliers (Pulles, Veldman, Holger, and Sierksma, 2014). It could be the case that coercive power is less effective over distance and different geographic regions, as different cultures are likely to differently experience the use of power. Another interesting theoretical explanation is found in
preferred customer theory. Research found that achieving a preferred customer status is more difficult
when suppliers are located in another geographic region (Steinle and Schiele, 2008). These authors argued that when a preferred customer status is essential, the buying firm can best form close relationships with firms located in the same geographic region. This could explain the discovery that low horizontal supply chain complexity has the best fit with low spatial supply base complexity regarding the number of drug recalls. When a buying firm has only a few first-tier suppliers, the dependency on these suppliers is high. This makes it desirable to have a preferred customer status, as the supplier is, in that case, less likely to behave opportunistically, and the likelihood of supply chain disruptions is therefore lower. However, becoming a preferred customer is likelier to occur for suppliers located in the same geographic region (Steinle and Schiele, 2008). On the other side of the continuum, when a buying firm has many suppliers, a preferred customer status is less important, as materials can be sourced from many geographic locations. Without the need of a preferred customer status, suppliers can be contracted worldwide.
6.2. Theoretical implications
27 and supply chain theory. Research has proposed a relationship between the elements, but contradicting results have been found. The matrix designed in Figure 6.1 (p. 25) could be a first attempt at describing the interplay between supply base complexity dimensions and supply chain disruptions (when argued that a recall is indeed a disruption).
6.3. Managerial implications
This study’s primary message for practice is that adding additional complexity to the supply base is associated with unintended consequences, such as an expected increase in the number of drug recalls. The first implication is that the results demonstrated that having a supply base consisting of (1) many suppliers or (2) worldwide suppliers is expected to increase the number of drug recalls. The second implication is that the results revealed the moderating effect of spatial supply base complexity. Managers in the pharmaceutical industry can use the matrix in Figure 6.1 (p. 25) to define their sourcing strategy. Based on this matrix, it is suggested that managers either build close relationships with only a few suppliers located in the same geographic region or source materials from many suppliers dispersed worldwide. Important in making this decision is whether a preferred customer state is sought. This can be the case when, for example, materials are sold by only a few suppliers. In this case, it is better to connect with a supplier located in the same geographic region because this increases the likelihood of becoming a preferred customer. On the other extreme, the buying firm’s sourcing strategy could be aimed on sourcing materials as cheap as possible. In this case, becoming a preferred customer is less important, and is it better to have many suppliers dispersed worldwide. Suppliers can than compete against one another for the cheapest price.
6.4. Limitations and future research directions
It should be mentioned that this study has limitations. First, the data collection is primarily focussed on the United States. It could be interesting for future research to explore this topic across different countries.
28 relationship had stopped previously. For future research, I suggest constructing a database in which the number of recalls and the number of supplier-buyer relationships are counted per year. In this way, a more elaborated model can be built.
Third, I can only assume how the mechanism between both complexity elements and the number of product recalls works. I expect that the managerial effort to manage suppliers increases whenever the number of suppliers, or the diversity between them, increases. This offers suppliers the opportunity to compromise quality management practices and product quality. However, future research could focus on how the mechanism between an increase in supply base complexity and the number of recalls exactly works. Interviews with supply chain managers (qualitative research) could be conducted to investigate this topic and should offer interesting views.
29
Literature
Atella, V., Bhattacharya, J., & Carbonari, L. (2011). Pharmaceutical Industry, Drug Quality and Regulations: Evidence from US and Italy. American Society of Health Economists, 1–22.
Azghandi, R., Griffin, J., & Jalali, M. S. (2018). Minimization of Drug Shortages in Pharmaceutical Supply Chains: A Simulation-Based Analysis of Drug Recall Patterns and Inventory Policies.
Complexity, 1–14. https://doi.org/10.1155/2018/6348413
Ball, G. P., Shah, R., & Donohue, K. (2018). The decision to recall: A behavioral investigation in the medical device industry. Journal of Operations Management, 62, 1–15. https://doi.org/https://dx.doi.org/10.2139/ssrn.1316805
Ball, G. P., Shah, R., & Wowak, K. D. (2018). Product competition, managerial discretion, and manufacturing recalls in the U.S. pharmaceutical industry. Journal of Operations Management,
58–59(April), 59–72. https://doi.org/10.1016/j.jom.2018.04.003
Bode, C., & Wagner, S. M. (2015). Structural drivers of upstream supply chain complexity and the frequency of supply chain disruptions. Journal of Operations Management, 36, 215–228. https://doi.org/10.1016/j.jom.2014.12.004
Bozarth, C. C., Warsing, D. P., Flynn, B. B., & Flynn, E. J. (2009). The impact of supply chain complexity on manufacturing plant performance. Journal of Operations Management, 27, 78–93. https://doi.org/10.1016/j.jom.2008.07.003
Caridi, M., Crippa, L., Perego, A., Sianesi, A., & Tumino, A. (2010). Do virtuality and complexity affect supply chain visibility? International Journal of Production Economics, 127, 372–383. https://doi.org/10.1016/j.ijpe.2009.08.016
Choi, T. Y., & Hong, Y. (2002). Unveiling the structure of supply networks: case studies in Honda, Acura, and DaimlerChrysler. Journal of Operations Management, 20, 469–493.
Choi, T. Y., & Krause, D. R. (2006). The supply base and its complexity: Implications for transaction costs, risks, responsiveness, and innovation. Journal of Operations Management, 24, 637–652. https://doi.org/10.1016/j.jom.2005.07.002
Claeskens, G. (2011). Groningen-Short course Model selection and model averaging. Retrieved May 29, 2019, from http://www.math.rug.nl/stat/models/files/claeskens.pdf
FDA. (2016). CDER: The Consumer Watchdog for Safe and Effective Drugs. Retrieved March 19, 2019, from https://www.fda.gov/Drugs/ResourcesForYou/Consumers/ucm143462.htm
30
2019, from
https://www.fda.gov/drugs/developmentapprovalprocess/manufacturing/ucm169105.htm
FDA. (2019). Drug Recalls. Retrieved February 11, 2019, from https://www.fda.gov/Drugs/DrugSafety/DrugRecalls/default.htm
Field, A. (2009). Moderation, mediation and more regression. In Discovering Statistics Using IBM SPSS
Statistics (3rd ed., pp. 392–428). London: SAGE Publications Ltd.
Flynn, B. B., & Flynn, E. J. (2005). Synergies between supply chain management and quality management: emerging implications. International Journal of Production Research, 43(16), 3421–3436. https://doi.org/10.1080/00207540500118076
Frost, J. (n.d.). Multicollinearity in Regression Analysis: Problems, Detection, and Solutions. Retrieved May 29, 2019, from https://statisticsbyjim.com/regression/multicollinearity-in-regression-analysis/
Gray, J. V, Roth, A. V, & Leiblein, M. J. (2011). Quality risk in offshore manufacturing: Evidence from the pharmaceutical industry. Journal of Operations Management, 29, 737–752.
https://doi.org/10.1016/j.jom.2011.06.004
Huo, B., Ye, Y., Zhao, X., & Zhu, K. (2019). Supply chain quality integration: A taxonomy perspective.
International Journal of Production Economics, 207, 236–246. https://doi.org/10.1016/j.ijpe.2016.05.004
Jayaraman, R., Alhammadi, F., Can, M., & Simsekler, E. (2018). Managing Product Recalls in Healthcare Supply Chain. Proceeding of the 2018 IEEE International Conference on Industrial
Engineering and Engineering Management (IEEM), 293–297.
Kaynak, H., & Hartley, J. L. (2007). A replication and extension of quality management into the supply chain. https://doi.org/10.1016/j.jom.2007.06.002
Ketchen, D. J., & Hult, G. T. M. (2007). Bridging organization theory and supply chain management: The case of best value supply chains. Journal of Operations Management, 25(2), 573–580. https://doi.org/10.1016/j.jom.2006.05.010
Krumholz, H. M., Ross, J. S., Presler, A. H., & Egilman, D. S. (2007). What have we learnt from Vioxx?
BMJ, 334(7585), 120–123. https://doi.org/10.1136/bmj.39024.487720.68
Laerd Statistics. (n.d.). Multiple regression in SPSS Statistics | Laerd Statistics Premium. Retrieved May 29, 2019, from https://statistics.laerd.com/premium/spss/mr/multiple-regression-in-spss-13.php
31 perspective. International Journal of Quality & Reliability Management, 30(5), 511–529. https://doi.org/10.1108/02656711311315495
Mena, C., Humphries, A., & Choi, T. Y. (2007). Toward a Theory of Multi-Tier Supply Chain Management. Journal of Supply Chain Management, 49(2), 58–77. https://doi.org/https://doi.org/10.1111/jscm.12003
Midler, P. (2007). “Quality Fade”: China’s Great Business Challenge. Retrieved March 1, 2019, from http://knowledge.wharton.upenn.edu/article/quality-fade-chinas-great-business-challenge/ Nagaich, U., & Sadhna, D. (2015). Drug recall: An incubus for pharmaceutical companies and most
serious drug recall of history. International Journal of Pharmaceutical Investigation, 5(1), 13–19. https://doi.org/10.4103/2230-973X.147222
Neilan, T. (2004, September 30). Merck Pulls Vioxx Painkiller From Market, and Stock Plunges. Retrieved February 21, 2019, from https://www.nytimes.com/2004/09/30/business/merck-pulls-vioxx-painkiller-from-market-and-stock-plunges.html
Pulles, N. J., Veldman, J., Holger, S., & Sierksma, H. (2014). Pressure or Pamper? The Effects of Power and Trust Dimensions on Supplier Resource Allocation. Journal of Supply Chain Management,
50(3), 16–36. https://doi.org/10.1111/jscm.12049
Robinson, C. J., & Malhotra, M. K. (2005). Defining the concept of supply chain quality management and its relevance to academic and industrial practice. International Journal of Production
Economics, 96, 315–337. https://doi.org/10.1016/j.ijpe.2004.06.055
Sila, I., Ebrahimpour, M., & Birkholz, C. (2006). Quality in supply chains: an empirical analysis. Supply
Chain Management: An International Journal, 11(6), 491–502. https://doi.org/10.1108/13598540610703882
Smith, M. (2018). Drug Recalls Put Spotlight on Drug Supply Chain. Retrieved February 21, 2019, from https://www.webmd.com/a-to-z-guides/news/20181119/drug-recalls-put-spotlight-on-drug-supply-chain
Soares, A., Soltani, E., & Liao, Y.-Y. (2017). The influence of supply chain quality management practices on quality performance: an empirical investigation. Supply Chain Management: An
International Journal, 22(2), 122–144. https://doi.org/10.1108/SCM-08-2016-0286
Steinle, C., & Schiele, H. (2008). Limits to global sourcing?. Strategic consequences of dependency on international suppliers: Cluster theory, resource-based view and case studies. Journal of
Purchasing and Supply Management, 14, 3–14. https://doi.org/10.1016/j.pursup.2008.01.001
32 outsourcing-supplier concentration-product recalls linkages. Journal of Operations Management,
32(5), 241–253. https://doi.org/10.1016/j.jom.2014.04.003
Stock, G. N., Greis, N. P., & Kasarda, J. D. (2000). Enterprise logistics and supply chain structure: the role of fit. Journal of Operations Management, 18(5), 531–547. https://doi.org/10.1016/S0272-6963(00)00035-8
Thirumalai, S., & Sinha, K. K. (2011). Product Recalls in the Medical Device Industry: An Empirical Exploration of the Sources and Financial Consequences. Management Science, 57(2), 376–392. https://doi.org/10.1287/mnsc.1100.1267
Torreya. (2017). The Future of the Global Pharmaceutical Industry.
UCLA: Statistical Consulting Group. (n.d.). Introduction to Regression with SPSS Lesson 2: SPSS
Regression Diagnostics. Retrieved May 29, 2019, from
https://stats.idre.ucla.edu/spss/seminars/introduction-to-regression-with-spss/introreg-lesson2/#s5
Union of Concerned Scientists. (2017). Merck Manipulated the Science about the Drug Vioxx. Retrieved February 21, 2019, from https://www.ucsusa.org/manipulating-science-about-drug-vioxx#.XG6h-vrPyUk
Wadman, M. (2007). Merck settles Vioxx lawsuits for $4.85 billion. Retrieved February 21, 2019, from http://www.nature.com/doifinder/10.1038/450324b
Whipple, J. M., & Roh, J. (2010). Agency theory and quality fade in buyer-supplier relationships. The
International Journal of Logistics Management, 21(3), 338–352. https://doi.org/10.1108/09574091011089781
World Health Organization. (2006). Quality assurance of pharmaceuticals: A Compendium of
Guidelines and Related Materials: v. 2: Good Manufacturing Practices and Inspection (2nd ed.).
Geneva: World Health Organization.
Zhao, X., Li, Y., & Flynn, B. B. (2013). The financial impact of product recall announcements in China.
International Journal of Production Economics, 142, 115–123. https://doi.org/10.1016/j.ijpe.2012.10.018
Zu, X., & Kaynak, H. (2012). An agency theory perspective on supply chain quality management.
International Journal of Operations & Production Management, 32(4), 423–446.
33
34
35
36
37
Appendix D - SPSS output model 3
38
