Buyer–Supplier and Supplier–Supplier Alliances: Do They Reinforce or Undermine One Another?
Sergio G. Lazzarini, Associate Professor of Organization and Strategy, Ibmec São Paulo, R. Quata 300, São Paulo, SP Brazil 04546-042 (SergioGL1@isp.edu.br).
abstract Previous research has portrayed buyer–supplier and supplier–supplier alliances as important mechanisms to foster learning and exchange efficiencies. Controversy remains, however, as to how these alliances interact. While some propose they reinforce one another (e.g. learning in horizontal ties generates positive spillovers to vertical ties), others propose a negative interplay (e.g. when increasing vertical-tie intensity, suppliers may weaken horizontal ties to avoid retaliation from buyers who wish to preserve bargaining power). We empirically test these competing views using survey data from the Brazilian auto-parts industry. In an attempt at reconciliation, we propose that the positive or negative interaction between vertical and horizontal alliances depends on the level of technological uncertainty of goods exchanged. Vertical ties seem to inhibit horizontal ties when technological uncertainty is low; when technological uncertainty is higher, vertical and horizontal ties do not seem to have any meaningful form of interaction. We discuss implications for theory and practice.
The intensity of alliance ties to partners is often deemed an imperative strategic activity for suppliers of manufactured goods. For instance, management scholars have devoted substantial attention to the study of buyer–supplier alliances and how they impact supplier (and buyer) performance; there is a great deal of evidence, for example, that through such vertical collaboration, suppliers and buyers are able to revamp production processes, reduce transaction costs and deliver better products to consumers (e.g. Dyer, 1997; Helper, 1991; Kalwani and Narayandas, 1995; Kotabe et al., 2003; Martin et al., 1995; Srinivasan and Brush, 2006). More recently, scholars have also underscored the importance of supplier–supplier alliances, i.e. horizontal ties among suppliers themselves. According to this literature, suppliers are more and more engaging in value-enhancing collective efforts such as the exchange of best practices and joint product development. As a result, many inter-organizational networks have turned into ‘netchains’, that is, layers of horizontally-linked suppliers which are also associated with buyers through vertical collaborative ties (Lazzarini et al., 2001).
Although scholars have increasingly recognized the importance of suppliers increasing the intensity of both their buyer–supplier and supplier–supplier alliance ties, there is some controversy about how these horizontal and vertical ties interact with one another and what consequences such an interaction brings. Some consider the intensity of supplier–supplier alliances as a beneficial organizational effort that enhances the effectiveness of buyer–supplier collaboration or vice versa. The canonical example supporting this view is Toyota's network of suppliers, which has been studied in great detail (Dyer and Nobeoka, 2000; Nishiguchi and Beaudet, 1998; Sako, 2004). By forging stronger ties and learning teams with other Toyota suppliers, each individual supplier is able to benefit from positive network spillovers, such as the diffusion of valuable knowledge and practice. Because Toyota – their common buyer – ultimately benefits from the faster knowledge sharing and capabilities improvement in its supplier base, it stimulates intense collaboration among its suppliers. The resulting improvements in processes and product configurations allow suppliers to be on the cutting edge in terms of production efficiency and innovation, which is in the best interest of Toyota. Moreover, collaboration between a knowledgeable buyer, such as Toyota, with individual suppliers (e.g. through problem-solving teams) helps these suppliers develop competencies that can be disseminated in their horizontal network. In this sense, supplier–supplier and buyer–supplier alliances reinforce one another.
There is, however, an opposite possibility: stronger supplier–supplier alliances may associate with weaker buyer–supplier alliances. For example, instead of engaging in value-enhancing practices that would otherwise benefit the network as a whole, suppliers pursue intensive horizontal ties to balance their relative dependence and power in negotiations with buyers (Choi et al., 2002). Observing that suppliers are collaborating with each other more intensively, for that purpose, buyers may attempt to constrain the buyer–supplier ties with that particular supplier; alternatively, buyers can also discourage supplier–supplier alliance intensity when stronger vertical alliances are in place. For fear of retaliation from buyers, suppliers themselves may then weaken their horizontal ties, so as to preserve their valuable vertical relationships. Reve (1992) finds empirical evidence that supplier–supplier and buyer–supplier alliances are negatively associated, and concludes that ‘rather than advocating the development of tightly structured alliances both horizontally and vertically, a trade-off between horizontal and vertical alliances is suggested’ (p. 254). If this negative effect occurs, we should expect either supplier–supplier or buyer–supplier collaborations, not both (as in Toyota's case).
These diverging views invite further work attempting to theoretically and empirically examine whether supplier horizontal alliances (i.e. those with another supplier of the same buyer) and vertical alliances (i.e. those with the buyer) reinforce or undermine one another. This is the goal of our paper. We begin by providing theoretical underpinning to the opposing patterns of interaction between supplier–supplier and buyer–supplier alliances, leading to the development of testable competing hypotheses. We next attempt to reconcile these conflicting views by outlining a contingency that is likely to moderate the interaction between the intensities of horizontal and vertical alliances: technological uncertainty (i.e. the extent to which products being exchanged exhibit constant change in specifications and new technologies). In a nutshell, our argument is as follows: subject to high technological uncertainty, supplier–supplier alliances are likely to pursue intense learning and innovation activities among suppliers, thus generating a positive spillover on the vertical exchange. When technological uncertainty is low, however, this effect should be minimal; any supplier–supplier association may be interpreted by buyers as a way to balance power in the network, thereby yielding a negative interaction between horizontal and vertical alliances. We then test our hypotheses using survey-based data in the context of the auto-parts industry in Brazil. This country has a history of alliances not only between suppliers and automobile manufacturers, but also among suppliers themselves (Addis, 1999). We conclude the paper by outlining implications for theory and practice, and suggesting some directions for further research.
THEORY AND HYPOTHESES
Inter-organizational alliances are commonly conceptualized as collaborative, interdependent efforts between two or more firms (Contractor and Lorange, 1988; Gulati, 1998). Moving beyond traditional arm's-length transactions or even formal links to other firms, intense alliances are characterized by the deeper sharing and joint deployment of financial resources, knowledge, and infrastructure in such a way to increase the performance of products and processes in a particular exchange. As each individual firm intensifies its links to several other partners, these firms end up forming a network of alliances. Research has found that, through alliances, firms can benefit from knowledge generated by exchange partners (Kale et al., 2000; Liebeskind et al., 1996; Powell et al., 1996) and promote relationship-specific investments without the need of costly vertical integration (Dyer, 1997; Holmstrom and Roberts, 1998; Jarillo, 1988). As partnering firms devote more time, effort and resources in their joint activities, the alliance increases in intensity.
Stimulated by the success of Japanese manufacturing companies, scholars have paid particular attention to the management of buyer–supplier or vertical alliances along the supply chain (Asanuma, 1989; Dyer and Ouchi, 1993; Helper and Sako, 1995; Kalwani and Narayandas, 1995; Srinivasan and Brush, 2006). For instance, suppliers may collaborate with buyers to reduce inventory and promote timely delivery (e.g. just-in-time production). Also, a supplier can collaborate with the buyer in the process of designing a new product or improving existing ones. Intense vertical alliances typically exhibit considerable amount of joint effort in the activities in which partners are involved. Such intense vertical alliances have been found to positively influence the performance of buyers and suppliers, in terms of both production efficiencies and innovation (Clark, 1989; Cusumano and Takeishi, 1991; Kotabe et al., 2003; Srinivasan and Brush, 2006).
Another parallel research stream has analysed the emergence of supplier–supplier or horizontal alliances between peers. The idea that firms in the same industry can create competitive advantage through mutual collaboration has been extensively discussed by the literature on regional clusters and industrial districts, which has explained the emergence of such patterns of collaboration as a way to economize on the provision of collective resources such as skilled labour, service providers, capital, infrastructure and other factors of production (e.g. Marshall, 1920; Mesquita and Lazzarini, forthcoming; Piore and Sabel, 1984; Schmitz and Nadvi, 1999; Wong-Gonzalez, 1992). Horizontal collaboration, however, may occur beyond the bounds of a particular location. Suppliers may create broad, representative associations providing collective services such as market information, lobbying with governments, and training (Altenburg and Meyer-Stamer, 1999; Harrison, 1992; Lane and Bachmann, 1996).
Suppliers may also go beyond their simple participation in such associations to pursue higher intensity alliances (Dyer and Nobeoka, 2000; Stuart et al., 1998) – which is the focus of our study. Thus, suppliers can build up committees and learning groups to exchange best practices and help one another to solve specific problems in their production processes In addition, suppliers can support joint market products, develop new distribution channels, and share resources (such as contacts and infrastructure) to reach new markets. Several authors have noted that intense ties among competitors have become increasingly frequent in a broad range of industries (Dussauge et al., 2000; Gomes-Casseres, 1994; Hamel et al., 1989).
More recently, scholars have tried to integrate these distinct research streams by offering a simultaneous assessment of buyer–supplier and supplier–supplier alliances. Reve (1992) analyses the emergence of horizontal and vertical relationships in distribution channels. Brown and Hendry (1998) infuse elements of supply chain analysis to discuss forms of inter organizational learning in industrial districts, often said to exhibit collaborations of a horizontal kind. Lazzarini et al. (2001) discuss how supply chains have evolved into ‘netchains’ comprised of horizontal networks of suppliers vertically linked with buyers. They propose that the analysis of horizontal and vertical ties cannot be divorced, because the pursuit of intense supplier–supplier alliances may have consequences for the performance of buyer–supplier alliances, and conversely. Choi et al. (2002) offer a taxonomy of supplier–supplier alliances and deliver several propositions regarding how such alliances may impact the performance of buyers and suppliers.
Despite these recent advances in the literature, there is still some controversy regarding how buyer–supplier and supplier–supplier alliances interact with each other. In fact, distinct theoretical arguments accommodate opposing views. On the one hand, intense supplier–supplier alliances may reinforce existing buyer–supplier alliances, or vice versa. In this case, we can expect the joint occurrence of buyer–supplier and supplier–supplier alliance intensity in a particular industry context. On the other hand, vertical collaborations may associate with weakened horizontal ties among suppliers, or vice versa. If this effect holds, then we can expect less intense supplier–supplier alliances when buyer–supplier alliances deepen. We discuss each possibility next.
Buyer–Supplier and Supplier–Supplier Alliances Reinforcing One Another
The view that buyer–supplier and supplier–supplier alliances reinforce one another can be supported by at least two distinct theoretical arguments. First, the inter-organizational learning that occurs in buyer–supplier and supplier–supplier alliances may be complementary. Research in organizational learning has stressed the importance of acquiring knowledge and innovative ideas based on a network of relationships (Appleyard, 1996; Liebeskind et al., 1996; Powell, 1996). For example, by tapping into the external knowledge of its buyers, a supplier can improve the performance of its production system, as well as develop new product configurations. To do so, however, both the supplier and the buyer will need to jointly invest in knowledge generation and sharing through successive interactions (Dyer and Nobeoka, 2000; Kale et al., 2000). Therefore, suppliers involved in collaborations with buyers will likely create themselves routines and know-how that can be useful to support other relationships, including relationships with horizontal peers. Using Cohen and Levinthal's (1990) term, alliances with buyers will equip suppliers with an ‘absorptive capacity’ through which they can value and understand external knowledge from horizontal partners. Consequently, high intensity of a vertical tie with a buyer is likely to be associated with high intensity of its collaborative ties with horizontal peers.
The argument also works in reverse: the intensity of supplier–supplier alliances may increase the inter-organizational learning that occurs in buyer–supplier alliances. Even in cases where suppliers are competitors, they will likely present differentiated knowledge based on their particular experiences and investments in product design and process improvements (Hamel et al., 1989; Schilling, 2000). Suppliers can therefore increase the intensity of alliances with horizontal peers, with whom they compete otherwise, as a way to exploit their knowledge diversity, and consequently benefit from improved performance with their common buyer (Feldman and Audretsch, 1999; Kogut, 2000). For instance, a supplier may see its performance with a buyer improve if that supplier learns from another supplier how to manage process flows and promote continuous improvements in components. Therefore, as suppliers share knowledge about best practices and get involved in collaborations among themselves, they become more and more capable of intensifying collaborative ties with buyers and meeting high standards of quality and innovation. This argument has been used to explain why Toyota encourages suppliers to engage in horizontal ties with each other (Dyer and Nobeoka, 2000; Sako, 2004).
A second theoretical argument supporting a complementary interaction between vertical and horizontal alliances is based on the idea that intense relationships among suppliers may increase the commitment of the buyer to its vertical relationships, therefore favouring the intensification of buyer–supplier alliances. Fundamentally, this argument is based on Coleman's (1988) view of ties as conduits of information about the behaviour of partners. Because vertical alliances may require substantial relationship-specific investments, including the knowledge that will be generated through inter-organizational learning (Bureth et al., 1997), suppliers may be reluctant to intensify their vertical collaborative agreements if they fear that their clients will adversely renegotiate terms of those agreements or switch to alternative suppliers (Holmstrom and Roberts, 1998; Williamson, 1985). In this case, more intense horizontal ties among suppliers may guarantee that any opportunistic behaviour by the buyer will likely be disseminated across its supplier base, creating a negative reputation and triggering retaliation (e.g. suppliers may avoid transacting with or devoting high effort in exchanges with that buyer). Anticipating this effect, buyers are likely to avoid reneging existing agreements or switching suppliers at will. In contrast, in the context of less intense supplier–supplier alliances, any past opportunistic action by the buyer may remain undetected (except by the offended supplier). Also using Toyota's production network as an example, Kreps (2004, pp. 605–8) employs this logic to explain why buyers may find it advantageous to stimulate supplier–supplier alliances as a way to develop and commit to a cooperative buyer–supplier alliance.
Collectively, these distinct theoretical arguments lead to:
Hypothesis 1: There is a positive association between the intensity of vertical (buyer–supplier) and horizontal (supplier–supplier) alliances: if a supplier increases the intensity of a vertical alliance with a buyer, it will more likely increase intensity of a horizontal alliance with another supplier of that buyer, or vice versa.
Buyer–Supplier and Supplier–Supplier Alliances Undermining One Another
Essentially, the arguments above consider more intense buyer–supplier and supplier–supplier alliances on the matters of efficiency – i.e. partners strengthen their ties either to increase inter-organizational learning, or to create incentives for cooperation in the exchange (Dyer and Nobeoka, 2000). Rather than a means to increase efficiency in the system, however, vertical and horizontal collaborations may influence the position of buyers and suppliers in their negotiations. Resource dependence theory (e.g. Oliver, 1991; Pfeffer and Salancik, 1978) and social exchange theory (e.g. Blau, 1964; Emerson, 1962) provide underpinning for this view, given their emphasis on how actors in an exchange balance their dependence and power. Such considerations support the prediction that more intense buyer–supplier alliances will constrain supplier–supplier alliances, or conversely, for two reasons.
First, most learning and performance-enhancing investments that occur in inter-organizational alliances tend to be relationship-specific. Although those specific assets have greater value in that particular relationship (Dyer, 1997; Madhok and Tallman, 1998), by definition they are less valuable when redeployed to alternative uses or with alternative partners (Williamson, 1985). Thus, if a supplier builds up a vertical alliance to engage in product design, the knowledge generated in the exchange may be mostly applicable to the buyer with whom the supplier is collaborating, thereby constraining the use of other ties (Blau, 1964; Galaskiewicz, 1985; Oliver, 1991) – including ties of a horizontal kind. Conversely, intense supplier–supplier alliances may undermine vertical collaborations. If suppliers anticipate that increasing the intensity of vertical alliances will increase their dependence on a particular buyer (Laamanen, 2005; Singh and Mitchell, 1996), they may reduce their resource commitments to that particular buyer and, instead, focus on horizontal collaborations which may allow them to apply the resulting know-how to a broader range of clients. Thus, Stuart et al. (1998, p. 91) suggest that horizontal networks may focus on ‘holistic learning and organizational advancement, not necessarily tied to the products sold to a particular buyer’.
A second and related argument suggests that suppliers may build up their relationship to horizontal peers to deliberately increase their bargaining power relative to buyers. For instance, suppliers engaged in horizontal alliances may jointly negotiate the terms of exchanges with common buyers (such as prices or exclusivity clauses). Consequently, when building up their vertical partnerships, buyers may attempt to inhibit the development of horizontal alliances as a way to avoid supplier collusion (Choi et al., 2002). This logic implies that more intense buyer–supplier alliances weakens supplier–supplier alliances. It is also possible, however, that less intense vertical alliances prompt the build up of supplier–supplier alliances. Lacking a collaborative orientation, buyer–suppliers negotiations will tend to be strictly price-driven and, to a large extent, impersonal (Dwyer et al., 1987; Macneil, 1978). This environment is likely to prompt suppliers to strengthen alliances among themselves to create, using Galbraith's (1956) terminology, countervailing power in their vertical negotiations. By doing so, suppliers are likely to reduce their direct competition and pose constraints on buyers’ ability to benefit from multiple, unarticulated actors (Burt, 1992; Emerson, 1962). As proposed by Reve (1992, p. 238), ‘organizations facing external constraints in one direction (e.g., vis-à-vis their customers) may strengthen their external ties in other directions (e.g., vis-à-vis their competitors) with the objective of influencing the first sector as well’.
These arguments lead to our second hypothesis, which competes with the first hypothesis outlined before:
Hypothesis 2: There is a negative association between the intensity of vertical (buyer–supplier) and horizontal (supplier–supplier) alliances: if a supplier increases the intensity of a vertical alliance with a buyer, it will less likely increase intensity of a horizontal alliance with another supplier of that buyer, or vice versa.
The Contingent Effect of Technological Uncertainty
The two competing views outlined before identify a trade-off for the intensification of horizontal and vertical alliances. From the point of the view of buyers, stimulating its supplier partners to intensify their horizontal alliances is likely to augment inter-organizational learning and promote cooperation; but it is also likely to reduce buyers’ bargaining position in the vertical tie. From the point of view of suppliers, more intense collaborative vertical alliances are likely to allow them to develop know-how that can be applied to other exchanges; but it is also likely to mandate a substantial commitment of specific resources, possibly constraining the intensification of alternative ties. Therefore, the interplay between buyer–supplier and supplier–supplier alliances may be negative or positive depending on the relative magnitude of those opposing effects. This discussion invites the use of some contingency that might influence the relative benefits and costs of intensifying horizontal alliances when vertical alliances are in place, or conversely (Choi et al., 2002).
We adopt an environmental variable that has been widely studied in contingency-based theory (Burns and Stalker, 1961; Thompson, 1967) and in the literature on inter-organizational alliances (Afuah, 2000; Harrigan, 1988; Poppo and Zenger, 2002): technological uncertainty. Supply components which display constant change in specifications and new technological breakthroughs will create an environment where technological uncertainty is at a high degree. For instance, in the auto industry, carburettors are normally considered as supplies involving constant innovations, while clutches constitute an example of standardized product with mature, well-known technology (Swaminathan et al., 2002). Thus, depending on the type of products and components being exchanged, a network will exhibit heterogeneity in terms of the technological uncertainty that surrounds transacting partners. We submit that this heterogeneity is likely to moderate the interplay between horizontal and vertical alliances.
When technological uncertainty is high, the need for inter-organizational learning at the supplier level will escalate. Exchange of best practices and even joint development of technologies through supplier–supplier alliances will be crucial to keep suppliers on the cutting edge and generate valuable product or process innovations (Stuart et al., 1998). To develop innovations that are useful for buyers, suppliers will also likely need to promote relationship-specific investments, which will demand high commitment levels in the vertical exchange. As discussed before, horizontal alliances among suppliers should increase their confidence that buyers will not renege on existing agreements. Thus, from the point of view of suppliers, vertical collaborations with buyers are likely to improve their ability to strengthen their alliances into valuable, innovative relationships. This effect also makes sense from the point of view of buyers: the effect of intensifying supplier–supplier alliances is likely to outweigh the negative effect associated with increased supplier power, when technological uncertainty is high. In contrast, when technological uncertainty is low, there should be no need for inter-supplier learning and accompanying investments in relationship-specific assets. Aware that buyers may be reluctant to accept favourably the intensification of supplier–supplier alliances because of bargaining power asymmetries, suppliers may intensify their buyer–supplier alliances at the expense of existing or potential supplier–supplier relationships. Consequently, lack of technological uncertainty should lead to a negative effect of supplier–supplier alliances on buyer–supplier alliances (or vice versa). In other words:
Hypothesis 3a: When technological uncertainty is high, there is a positive association between the intensity of vertical (buyer–supplier) and horizontal (supplier–supplier) alliances.
Hypothesis 3b: When technological uncertainty is low, there is a negative association between the intensity of vertical (buyer–supplier) and horizontal (supplier–supplier) alliances.
DATA AND METHODS
We tested our hypotheses by surveying suppliers of automobile parts in Brazil. The auto industry in Brazil is an appropriate empirical setting for our study. This industry has a history of alliances not only between suppliers and automobile manufacturers, but also among suppliers themselves (Addis, 1999). We focused our survey instrument on suppliers (instead of manufacturers) because their population is much larger than the population of auto manufacturers. Moreover, they are well-organized through a business association, Sindipeças (Sindicato Nacional da Indústria de Componentes para Veículos Automotores), whose members represent over 90 per cent of Brazil's parts sector revenues. Sindipeças granted us access to its directory of members and helped us in the procedures to implement the survey.
Research Design and Data Collection
Our survey data collection processes mostly followed Dillman's (2000) procedures. We initially developed a questionnaire by identifying construct items used in previous studies. We also obtained the help of other academics and managers to develop items where the literature was silent, to refine survey wording, and to check the overall validity of questions vis-à-vis the industry environment. We compiled a mailing list of approximately 450 firms, using a list of Sindipeças members. Through Sindipeças, we identified the key respondents – either general or division manager. These managers are most knowledgeable about their firm's relationships with other suppliers and with auto manufacturers, as well as about company-specific information. Our response rate was just above 40 per cent (184 responses) – though, because of missing values, a total of 105 observations were effectively used in our regressions. We assessed whether the sub-sample of 105 suppliers used in our analysis is significantly different from the larger sample through t-tests on key dimensions (see, e.g. Armstrong and Overton, 1977). No significant differences were found. We also assessed non-respondent bias by comparing early with late respondents through the same procedure. Again, we found no significant differences. Lastly, we checked for the reliability of the self-reported data (i.e. performance, size and sales) though public data and telephone contact. The 12 companies approached by phone were randomly chosen and the information was checked with the responses of the questionnaire. The remaining 10 companies were chosen because of the availability of public data. We identified no biases or other related issues with the data provided by informants.
In the questionnaire, we asked respondents about the prior three years of their firm's activities to avoid biased responses due to specific aberrant experiences. In the survey, we asked respondents to assess vertical and horizontal alliances. In the case of vertical alliances, we asked respondents to focus on ‘a customer that the respondent was most knowledgeable about’. We also asked questions related to each firm's relationship with another supplier. Respondents were asked to select a company that supplies a similar or complementary product to the buyer (customer) that the respondent previously identified. To facilitate responses, our questionnaire focused on a particular product that represented most of the firm's revenues.
Our sample includes both large and small firms, as well as domestic and multinational ones. Nearly half (45 per cent) of the firms surveyed were subsidiaries of foreign multinationals. Moreover, our sample appears to exhibit a diversity of firms in terms of size. Measured by sales volume in millions of dollars, 7 per cent of them had sales that were less than US$2.5 million, 25 per cent had sales between $2.5 million and $10 million, 36 per cent had sales between $10 million and $40 million, 22 per cent had sales between $40 million and $200 million, 6 per cent had sales between $200 million and $400 million, and 8 per cent had sales of more than $400 million.
We describe below four sets of variables: (1) alliance-based variables (used to measure the intensity of vertical and horizontal alliances); (2) determinants of the intensity of vertical alliances; (3) determinants of the intensity of horizontal alliances; and (4) common controls (for both vertical and horizontal ties). Variables based on measurement scales are described in Table I. All multi-item scales are based on a 5-point Likert scale, where 1 represents ‘low degree’ and 5 represents ‘high degree’. Table II presents the descriptive statistics of our variables.
Table I. Description of variables involving multi-item measurement scales
|Buyer–supplier alliance intensity (α = 0.70) and supplier–supplier alliance intensity (α = 0.81)*|
|To what extent your company develops the activities (listed below) together with your selected customer:|
| 1. Marketing and export-related activities|
| 2. New product and process development|
| 3. Sharing of equipment and other production machinery|
| 4. Joint purchases|
| 5. Joint negotiations with governmental agencies|
| 6. Sharing of investments, responsibilities and efforts|
| 7. Sharing the burden and the outcomes of joint initiatives|
|Relational norms in the vertical exchange (α = 0.82) and in the horizontal exchange (α = 0.95)*|
|To what extent your relationship with your selected supplier/customer is characterized by the following elements:|
| 1. Fluid exchange of information|
| 2. Exchange of information about production costs|
| 3. Exchange of information about plans, programmes and schedules|
| 4. Transparent negotiations|
| 5. Helping each other in implementing programmes of inventory management (e.g. just-in-time)|
| 6. Problem-solving approach rather than mutual conflict|
| 7. Mutual trust and commitment|
| 8. Flexibility to make adjustments when difficult situations unfold|
|Participation in supplier associations (α = 0.81)|
|The degree to which your company uses specialized support from the business association (Sindipeças) for:|
| 1. Technical training|
| 2. Technical and economic information|
| 3. Social events|
| 4. Political lobbying for the industry as a whole|
| 5. Political lobbying for your company or group of companies|
| 6. Support for collective actions among associates (e.g. collective procurement, marketing)|
| 7. Advising and monitoring collective norms and standards (e.g. quality)|
|Specific investments in the vertical exchange (α = 0.73)|
|To what extent your company invested in the following routines and specific assets to satisfy the exclusive and unique needs of your customer:|
| 1. Research and development, product design and facilities for R&D|
| 2. Production process, machinery and equipment|
| 3. Sales and procurement systems, logistics and inventory management|
| 4. New production plants and warehouses|
| 5. Computer hardware and software|
|Technological uncertainty (α = 0.81)|
|To what extent your business is characterized by:|
| 1. Frequent introduction of new products or innovations in the existing products|
| 2. Constant need to fulfil the technological needs of customers|
| 3. Frequent variation in demanded over the years|
| 4. Frequent technological change|
Table II. Descriptive statistics
| 1. Buyer–supplier alliance intensity||1.00|| || || || || || || || || || || |
| 2. Supplier–supplier alliance intensity||−0.13||1.00|| || || || || || || || || || |
| 3. Relational norms (vertical)||0.54||−0.15||1.00|| || || || || || || || || |
| 4. Specific investments||0.16||0.04||0.06||1.00|| || || || || || || || |
| 5. Relational norms (horizontal)||−0.11||0.75||−0.11||0.07||1.00|| || || || || || || |
| 6. Participation in supplier association||0.17||0.15||0.16||−0.04||0.14||1.00|| || || || || || |
| 7. Technological uncertainty||0.10||−0.05||0.17||0.21||−0.02||0.05||1.00|| || || || || |
| 8. R&D||−0.16||0.08||−0.20||0.08||0.06||0.00||0.12||1.00|| || || || |
| 9. Multinational||0.06||0.03||−0.17||−0.06||−0.06||−0.01||−0.17||0.01||1.00|| || || |
|10. Firm age||−0.02||0.06||−0.12||−0.05||0.06||−0.09||−0.06||0.10||0.05||1.00|| || |
|11. Value-added||−0.03||−0.01||−0.07||−0.06||0.10||0.05||−0.06||−0.11||−0.02||−0.05||1.00|| |
|12. Firm size||0.25||−0.08||0.05||0.10||0.02||0.00||0.01||−0.03||0.06||0.00||−0.08||1.00|
(1) Alliance-Based Variables
We measured alliance intensity through the actions of parties in fostering joint activities (Heide and Miner, 1992; Lusch and Brown, 1996). Specifically, we gauged the extent to which the respondent's firm has engaged in joint activities with its selected customer (buyer) and peer (another supplier of the selected customer). We asked the respondent to report the degree to which their relationships with partners involved coordinated action in areas such as: marketing and exports related activities; new product and process development; sharing of equipment and other resources; joint purchase of inputs (in the case of buyer–supplier ties only); and joint representation at governmental agencies. Because we had no prior reason to suppose that one particular activity was more important than another in a particular relationship, we created the variables buyer–supplier alliance intensity and supplier–supplier alliance intensity as the unweighted average of the items (Cronbach alphas, respectively, equal to 0.70 and 0.81). Therefore, our alliance-based variables measure the intensity of collaboration among parties in the activities involved in their vertical or horizontal exchange.
(2) Variables Influencing Buyer–Supplier Alliance Intensity
Relational norms in the vertical and horizontal exchange. The variable relational norms in the vertical exchange measures the degree to which a buyer and a supplier have developed relational commitments in their alliance (Anderson and Narus, 1990). In general, we tried to measure the respondents’ perception of whether negotiations are fair, whether the resulting commitments are likely to be sustained, and whether there is trust and collaborative orientation among parties. As widely discussed in the literature, relational norms support the emergence and stability of inter-organizational alliances because they allow parties to economize on the costs of contracting and mutually adapt to changing circumstances (Baker et al., 2002; Dwyer et al., 1987; Dyer, 1997). Since those relational norms likely emerge through repeated interaction in past transactions (Gulati, 1995; Poppo and Zenger, 2002), they are commonly considered as determinants of inter-organizational collaborations. The score of the perceptions about relational norms was computed as the unweighted average of the measurement items (Cronbach alpha = 0.82). In regards to relational norms in the horizontal exchange, we used the same measurement dimensions as above, except that respondents were asked to refer to another supplier who transacts with the selected customer (Cronbach alpha = 0.95). As in the case of vertical alliances, we expect that relational norms among suppliers should be positively associated with the intensity of horizontal alliances.
Specific investments. This variable measures the extent to which investments, made in the past five years, were made to serve the exclusive and unique needs of the focus customer (buyer). The investments could be related to production processes, sales, procurement or logistics, research and development, and physical assets (e.g. hardware and industrial facilities). Specific investments of the types mentioned above, according to the literature, can influence the intensification and stability of alliances because they not only signal commitments to the exchange (Anderson and Weitz, 1992) but also support joint action (Dyer, 1997). For instance, if the supplier invests in facilities close to its client, inter-partner exchange of knowledge will be largely facilitated. We created the variable for specific investments through the unweighted average of the scores for the items indicating possible investments as listed above (Cronbach alpha = 0.73).
(3) Variables Influencing Supplier–Supplier Alliance Intensity
Participation in supplier associations. Several authors have noted that, through their participation in business associations, firms can forge specific contacts and alliances (Lane and Bachmann, 1996; Rosenkopf et al., 2001). Thus, the effective involvement of suppliers in their main business association (Sindipeças) could be an antecedent of stronger horizontal ties. Sindipeças was singled out for special focus because both the literature and preliminary work on the survey confirmed that this organization is the dominant institution for parts firms (e.g. Addis, 1999). Because we sent questionnaires to all Sindipeças’ members, we needed to measure not only whether suppliers were involved in their association, but also the intensity of their involvement. In this sense, we tried to gauge respondents’ assessments of the effectiveness of Sindipeças in supporting their activities as a way to assess the extent to which they were engaged in the association. We thus crafted a multi-dimensional construct, where respondents indicated on a 5-point Likert scale the degree to which they agreed that Sindipeças was effective to organize and supply technical training, develop economic/technical studies, hold social events to gather members, and support collective negotiations with the government (Cronbach alpha = 0.81).
(4) Common Control Variables Influencing Alliance Intensity
Technological uncertainty. As a result of volatile market demand and technological breakthroughs, some components are likely to exhibit changing specifications and embodied technologies. An environment with high technological uncertainty should not only favour the intensification of alliances in general (Rowley et al., 2000), but also, as discussed before, moderate the interplay between horizontal and vertical alliance intensities. We thus asked respondents to assess the degree to which their main product has been subject to constant technological change and variations in customers’ needs. The score of the perceptions about technological uncertainty was calculated as the unweighted average of the items (Cronbach alpha = 0.76).
R&D. We asked in the questionnaire the percentage of annual revenues that the company usually spends in R&D. Firms with large investments in R&D should have greater capacity and/or propensity to intensify alliances so as to improve processes and products, both horizontally and vertically (Powell et al., 1996).
Multinational. This dummy variable is coded 1 if the supplier is foreign-controlled, and 0 otherwise. One might suggest that foreign-owned companies are more likely to intensify alliances, given their global influence and reach.
Firm age. Older firms may be more knowledgeable of other firms in the industry, thus favouring their alliance intensity. Literature suggests that older firms benefit from accumulated experience and the establishment of routines with an alliance partner (Carroll and Hannan, 2000; Powell et al., 1996). We asked respondents to indicate the number of years their companies have been working in the auto-part sector.
Valued-added. We asked respondents to report the three-year average yearly revenues of their firm, and then divided this number by the total number of employees to compute the variable Value-added. This variable intends to control for differences across industry segments and its impact on the propensity of a firm to intensify alliances (Gulati, 1995). For instance, high-tech firms that do not employ intensive labour in the production processes may have superior capabilities to intensify alliances in general (Gulati and Gargiulo, 1999). Valued-added firms are more likely to have complementary capabilities that can make them interdependent and lead to alliances between them.
Firm size. Larger firms may possess a superior pool of resources and scale necessary to invest in cooperative efforts with other suppliers and customers (March and Simon, 1958). Martin et al. (1995) found that firm size increases collaborative Japanese automobile buyer–supplier links in North America because of organizational and financial resources available to undertake joint activities. Thus, we asked respondents to indicate the current total number of employees in their company as a measure of supplier size.
A straightforward way to test our hypotheses would be to run simple regressions where buyer–supplier alliance intensity is a dependent variable and supplier–supplier alliance intensity is an independent variable, and then vice versa. A critical problem with this approach is that vertical and horizontal alliance intensities are likely to be endogenously determined, based on a host of unobserved factors. For instance, a firm may have distinctive capabilities to manage alliances in general, thus increasing its propensity to increase the intensity of both vertical and horizontal alliances for a reason that is distinct from the arguments supporting Hypothesis 1. After observing a positive, significant association between buyer–supplier and supplier–supplier alliance intensities, one might be tempted to conclude that their interaction is positive. However, this effect may be spuriously caused by a third, unobserved factor: a firm's superior ability to manage alliances in general. Alternatively, the interaction between buyer–supplier and supplier–supplier alliance intensities may be negative not due to the effect of one type of alliance on the other per se, as proposed by Hypothesis 2, but due to firms’own limitations in physical, financial or human resources impeding the simultaneous management of a portfolio of stronger vertical and horizontal alliances. Failing to control for this joint determination should lead to biased and inconsistent estimates. Indeed, a Wu-Hausman test applied to our data revealed the existence of simultaneity involving vertical and horizontal alliances.
To control for endogeneity, we begin by specifying two regressions: one where buyer–supplier alliance intensity is the dependent variable and supplier–supplier alliance intensity is an independent variable, and another where these variables switch their position. Support for Hypothesis 1 can be found either if the coefficient of supplier–supplier alliance intensity is significantly positive in the regression where buyer–supplier alliance intensity is in the left-hand side, or if the coefficient of buyer–supplier alliance intensity is significantly positive in the regression where supplier–supplier alliance intensity is the dependent variable. Hypothesis 2 is supported if one of the alliance intensity variables exhibits a significantly negative coefficient. The estimation of the system of equations, however, requires the use of instrumental variables: variables that influence the intensity of one type of alliance (e.g. vertical), but that are uncorrelated with the error term (residual) of the other type of alliance (e.g. horizontal). More specifically, when buyer–supplier alliance intensity is on the left-hand side we use as instruments the variables influencing vertical alliance intensity (Relational norms in the vertical exchange and Specific investments), and exclude them in the other regression. We do the same in the regression where supplier–supplier alliance intensity is on the left-hand side, using as instruments the variables influencing horizontal alliances (Relational norms in the horizontal exchange and Participation in supplier association). In both equations, we add the common control variables discussed before. Therefore, our system of equations is specified as follows:
Buyer–supplier alliance = f(supplier–supplier alliance intensity, Relational norms in the vertical exchange, Specific investments, Technological uncertainty, controls)
Supplier–supplier alliance = f(buyer–supplier alliance intensity, Relational norms in the horizontal exchange, Participation in supplier association, Technological uncertainty, controls)
A possible way to estimate the system is, for instance, to first perform the second regression, generate predicted (fitted) values for supplier–supplier alliance intensity, and then use these predicted values in the first regression by replacing the original variable coding the intensity of the horizontal alliance. This is the so-called two-stage least squares method (2SLS). A shortcoming of the 2SLS method, however, is that the error terms involved in each equation may be correlated with each other. Failing to account for this correlation may yield inefficient estimates. The three-stage least squares method (3SLS) is designed to account for such possible correlation. The procedure is as follows. In the first stage, we generate 2SLS estimates, as discussed before. The 2SLS residuals we obtain in the first stage are then used to estimate a covariance matrix of errors. In the final stage, we re-estimate the system of equations, taking into account the covariance matrix we obtained in the second stage. We thus use the 3SLS estimates resulting from this third stage. The 3SLS method is found to generate more efficient estimates than the simpler 2SLS method (e.g. Greene, 2000; Poppo and Zenger, 2002).
To test Hypotheses 3a and 3b, in turn, we need to assess whether the coefficients of buyer–supplier and supplier–supplier alliance intensities vary depending on the level of the variable Technological uncertainty. A usual way to observe this effect is to create interactions involving the contingency variable and the main variables of interest. The problem of using this approach in our context is that our 3SLS routine assumes linearity of the endogenous variables, therefore failing to accommodate interactions. To overcome this limitation, we adopt a simple approach of splitting the data into two sub-samples: one in which Technological uncertainty is above or equal to its mean, and another in which this variable is below its mean. This procedure allows us to examine how the coefficients of the alliance intensity variables change according to two conditions: high and low technological uncertainty. Hypothesis 3a and 3b are supported, respectively, if the coefficients of the alliance-based variables are positive in the sub-sample where technological uncertainty is perceived to be high, and negative in the sub-sample where technological uncertainty is perceived to be low.
RESULTS AND DISCUSSION
Table III presents the results of the three-stage least squares estimation. Models (1a) and (1b) are jointly estimated based on the whole sample of suppliers, and hence are used to test the competing Hypotheses 1 and 2. Although the coefficient of supplier–supplier alliance intensity in the regression using buyer–supplier alliance intensity as a dependent variable is insignificant, the coefficient of buyer–supplier alliance intensity in the regression where supplier–supplier alliance intensity is in the right-hand side (1b) is significant (p < 0.05) and indicates a negative effect of the intensity of the vertical alliance on the intensity of the horizontal alliance. Thus, Hypothesis 2 is supported whereas Hypothesis 1 is not. On average, buyer–supplier and supplier–supplier alliances appear to display a negative interaction. Furthermore, the effect is unidirectional: vertical alliances constrain horizontal ones, but not conversely. With intense vertical alliances, suppliers commit specific resources to their joint collaboration with buyers and, doing so, they apparently reduce their autonomy to intensify other ties (including ties with peers). Alternatively, buyers may deliberately influence their vertical partners not to pursue an intense horizontal tie in order to avoid a loss in bargaining power. These two explanations are consistent with the results presented in Models (1a) and (1b).
Table III. Determinants of buyer–supplier (vertical) and supplier–supplier (horizontal) alliances: three-stage least squares estimation
|Buyer–supplier alliance intensity|| ||−0.201*|| ||−0.392**|| ||−0.065|
| ||(0.120)|| ||(0.167)|| ||(0.154)|
|Supplier–supplier alliance intensity||−0.050|| ||0.191|| ||−0.195|| |
|(0.098)|| ||(0.145)|| ||(0.159)|| |
|Relational norms in the vertical exchange||0.457**|| ||0.513**|| ||0.490**|| |
|(0.066)|| ||(0.109)|| ||(0.077)|| |
|Specific investments||0.123*|| ||−0.011|| ||0.200**|| |
|(0.061)|| ||(0.094)|| ||(0.071)|| |
|Relational norms in the horizontal exchange|| ||0.461**|| ||0.508**|| ||0.434**|
| ||(0.041)|| ||(0.055)|| ||(0.074)|
|Participation in supplier association|| ||0.103*|| ||0.114|| ||0.092|
| ||(0.060)|| ||(0.080)|| ||(0.079)|
Hypotheses 3a and 3b imply differential effects of buyer–supplier on supplier alliances (or conversely) depending on the level of technological uncertainty involving supply components. As discussed before, the test of two differential effects is performed by splitting the sample into two conditions: where Technological uncertainty is below its mean (Models (2a) and (2b)), and where it is above its mean (Models (3a) and (3b)). Supporting Hypothesis 3b, we find a significantly negative effect of buyer–supplier alliance intensity on supplier–supplier alliance intensity (p < 0.01, Model (2b)) when technological uncertainty is low, in a magnitude that is apparently larger than in the regression involving the whole sample. This effect, however, is insignificant in the sub-sample where technological uncertainty is high (Model (3b)). Thus, Hypothesis 3a is not supported. Specifically, there is no positive interaction between horizontal and vertical alliance intensities when components are subject to high technological change. As before, we find no significant result involving the effect of supplier–supplier alliance intensity on buyer–supplier alliance intensity.
Together, these results provide mixed support for our contingency-based prediction on the interplay between buyer–supplier and supplier–supplier alliances. On the one hand, the negative interaction among those two types of alliances is found only when technological uncertainty is low, which supports our argument that the efficiency-based benefits of supplier–supplier alliances should decline when the need for supplier inter-organizational learning is scant. In this condition, when strengthening vertical alliances with suppliers, buyers will likely view the strengthening of horizontal alliances in a negative way. On the other hand, we do not find a positive interaction between horizontal and vertical alliances when technological uncertainty is high. Indeed, we find no significant interaction in this condition whatsoever. Apparently, when technological uncertainty is high, vertical and horizontal alliances have no direct consequence for each other. Because we fail to find any significant and positive interaction between buyer–supplier and supplier–supplier alliances when technological uncertainty is high, the discriminating effect of this contingency is weaker than we initially hypothesized.
Some comments about the instrumental and control variables are in order. As expected, Relational norms in the exchange is statistically significant in both models; because it positively influences the intensity of both buyer–supplier and supplier–supplier alliances (p < 0.01), it helps us confirm that trust and a general propensity to collaborate support deeper inter-organizational ties, either vertical or horizontal. This effect remains significant even in the split-sample regressions.
Also as expected, specific investments established by suppliers to support their exchanges with buyers are positively correlated with more intense vertical collaborations. Interestingly, the effect of specific investments varies according to the different sub-samples. This variable displays a strongly significant and positive coefficient when technological uncertainty is high (p < 0.01); however, it displays an insignificant coefficient when technological uncertainty is low. A possible explanation is that conditions involving high technological uncertainty mandate a broader range of specific investments to support inter-organizational learning, which is by itself specific to the exchange. For instance, suppliers may need to invest in facilities close to the buyer to support the joint design of products. In contrast, when technological uncertainty is low, no profound inter-organizational learning needs to occur, thereby reducing the need of other accompanying investments.
The last instrumental variable, Participation in supplier association, exhibits mixed results. In the whole sample, the statistically significant and positive coefficient of that variable indicates that Brazilian auto-parts suppliers’ involvement in their supplier association (Sindipeças) positively influences the intensity of horizontal ties (p < 0.05). This result confirms that business associations serve not only as a way to supply collective resources, but also as a networking mechanism through which members can forge specific alliances with one another. Participation, however, becomes insignificant in the split-sample regressions. Given that this variable is statistically significant and positive in the whole sample, its lack of significance in the sub-samples may be a result of the reduced statistical power in Models (2) and (3).
As a final note, it is worth observing that control variables are in general insignificant, indicating that most of the explanatory power of the models comes from the hypothesized and instrumental variables. Strong effects are found only with Firm age in Model (2a) (indicating that older suppliers are more likely to pursue higher vertical alliance intensity when technological uncertainty is low) and Multinational in Model (3b) (indicating that foreign-owned suppliers are more likely to pursue horizontal collaborations when technological uncertainty is high).
In sum, our results are more aligned with the literature proposing a negative interplay between horizontal and vertical alliances. As in Reve (1992), we find that intense ties in one dimension reduce the intensity of ties in the other dimension. This result is consistent with the argument that firms consider how their bargaining position and dependence may change when they intensify vertical and horizontal collaborations (e.g. Oliver, 1991; Pfeffer and Salancik, 1978). In particular, the unidirectional effect of vertical alliance intensity on horizontal alliance intensity suggests that, in our setting, suppliers are constrained by buyers in their horizontal ties. Buyers may either attempt to preserve their bargaining power by discouraging the intensification of ties among their suppliers, or demand a high level of relationship-specific investments in their vertical collaborations, thereby reducing suppliers’ autonomy to intensify alternative ties.
Therefore, our results are inconsistent with existing work observing a positive interplay between vertical and horizontal alliances (e.g. Dyer and Nobeoka, 2000; Stuart et al., 1998). A firm's attempts to keep its bargaining power apparently outweigh possible learning effects that would otherwise occur through more intense vertical and horizontal collaborations. This explanation is also partially supported by our result that the negative interplay between vertical and horizontal alliances especially occurs in conditions of low technological uncertainty, where the benefits associated with inter organizational learning are minimal. We note, however, that no positive association between supplier–supplier and buyer–supplier alliances is found in the condition involving high technological uncertainty. Apparently, even in the case of supply components exhibiting higher technological change, the learning benefits associated with extensive collaborations do not compensate for the loss in bargaining power that is likely to occur when a firm's partners pursue intense ties among themselves. To our knowledge, no previous study in the literature has observed the interplay between horizontal and vertical alliances taking into account contingencies, such as technology, that should influence the trade-offs involved in the formation of supplier–supplier alliances when intense buyer–supplier alliances are in place, or vice versa.
Our paper contributes to the literature by theoretically and empirically examining the interplay between buyer–supplier (vertical) and supplier–supplier (horizontal) alliance intensities in production systems. Although scholars have increasingly stressed the importance of the simultaneous assessment of vertical and horizontal alliances, distinct theoretical arguments have led to different predictions regarding the nature of their interaction. In this paper, we not only outline those distinct arguments, but also provide an attempt at reconciliation. Basically, we propose that the interaction between buyer–supplier and supplier–supplier alliances will be positive or negative depending on a particular contingency: the extent to which supply components are subject to uncertain technological change. Our data reveal that vertical alliance intensity negatively affects horizontal alliance intensity, but only when technological uncertainty is relatively low. When components are subject to higher technological change, those two types of alliance do not appear to significantly affect one another.
Although aligned with the results reported in Reve (1992), the absence of a positive effect of supplier–supplier on buyer–supplier alliances (and vice versa) in our data is inconsistent with existing work suggesting that horizontal alliances complement vertical forms of collaboration in supply chains (e.g. Dyer and Nobeoka, 2000; Stuart et al., 1998). On the contrary, our results indicate that, through the development of ties in one dimension, actors apparently want to maintain their autonomy in the other dimension – especially when supply components are not subject to high technological change. Thus, in general, deep horizontal and vertical alliances should not be frequently observed coexisting in inter-organizational networks. Given the growing interest in Japanese production systems emphasizing a mix of horizontal and vertical alliances, the practical relevance of our results becomes salient.
Our study, however, presents several limitations. Our database is focused on a single industry and a single country, therefore lacking heterogeneity in terms of technology and institutional features. The literature on emerging economies shows how country-level differences in terms of institutions, transaction costs and availability of resources may influence firm strategy (Hoskisson et al., 2000; Wright et al., 2005). For instance, although inter-organizational networks have been commonly observed in the Brazilian auto industry, their social role appears to have declined over the years, in part due to increasing entry of foreign firms replacing existing, well-connected actors (Addis, 1999). The disruption of existing networks may have reduced firms’ overall propensity to pursue extensive ties and their level of trust, thereby leading to a negative association between horizontal and vertical alliances. In contexts involving higher levels of societal trust, the interaction between those types of alliances may differ. Thus, it may be the case that intense horizontal and vertical alliances coexist to a higher degree in Japan because of the greater role of networks as conduits of information about partners’ behaviour in that country (e.g. Hagen and Choe, 1998). Therefore, the generalization of our results should be carried out with caution, taking into account possible effects from the institutional environment in consideration. Future research should attempt to evaluate the interplay between horizontal and vertical alliances in a broader range of industries and institutional contexts.
A related limitation of our study refers to the way we treat technological uncertainty. Since we are focusing on a single industry and a single country, the only feature that guarantees variation with respect to this variable in our study is the apparent heterogeneity of supply components in the sample. Other industries, however, may exhibit even greater variation in technological uncertainty. Thus, the products in our sample may not have a rate of technological change that is high enough to augment the benefits of inter-organizational learning and lead to a positive interaction between vertical and horizontal alliances. A possible way to increase the heterogeneity with respect to this contingency in future work is to collect data from different industries showing polar levels of perceived uncertainty (e.g. Rowley et al., 2000).
The interplay between vertical and horizontal alliances might vary because of the dynamics of alliances overtime (Anderson and Narus, 1990). Alliances evolve in terms of the degree of maturation and experience that may impact the interplay in different ways. Moreover, it is important to control for the company that initiated the alliance. The initiator might take the lead of several activities and gain some advantages. Future research should consider the alliance stage of evolution and the initiator of the alliance.
Finally, we focus on the interplay between buyer–supplier and supplier–supplier alliances without any consideration about the impact of those alliances on the performance of buyers and suppliers. Yet, performance considerations are critical to inform managers about the possible costs and benefits to intensify vertical and horizontal alliances. Thus, in conditions where those alliances are negatively related, suppliers must ascertain whether they should invest in vertical alliances with buyers (which will likely constrain the development of horizontal ties) or invest in horizontal alliances with peers (thus refraining from pursuing deep vertical ties). This choice will depend on the relative contributions of buyer–supplier and supplier–supplier alliances for the economic and operational performance of firms. Future research examining how horizontal and vertical alliances interact with each other and influence firm performance would be welcome.
We thank seminar participants at HEC-Paris, and at the Academy of Management Conference (2006), as well as two anonymous referees and the editor (Mike Wright) for comments and suggestions. We are also thankful to Fabio Wrobel Zausner, Felipe Fenelon Jacob, Fernanda Taís Nagamatsu, Guilherme de Moraes Attuy, Luciana Shawyuin Liu, Marcelo de Biazi Goldberg, Michel Tridico Tortelli, Pedro Alberto Puhler and Vanessa Lima Gonsalez for their research assistance on data collection and analysis, as well as William Mufarej, Flávio Del Soldato and Vanderlei Bueno from Sindipeças (Brazil's auto-parts firms association) for helping us contacting members. This research received funds from the CIBER at the Thunderbird-Garvin International School of Management, the Center for Research in Business Strategy at Ibmec São Paulo, and CNPq (Brazil's National Council for Scientific and Technological Development).
We also performed a Chow test by comparing directing the coefficient of ‘buyer–supplier alliance intensity’ in the two sub-samples involving high and low technological uncertainty (e.g. the coefficients of ‘buyer–supplier alliance intensity’ in Table III, Models (2b) and (3b)). As mentioned earlier, our 3SLS routine does not accommodate non-linear specifications such as interactions involving endogenous variables. So, we decided to perform the test with the simpler two-step 2SLS model in the following way. First, we run the regression where ‘buyer–supplier alliance intensity’ is the dependent variable and then generate fitted values for this variable. We then run the regression where ‘supplier–supplier alliance intensity’ is the dependent variable and then include in this regression the fitted values of ‘buyer–supplier alliance intensity’ from the first stage plus an interaction involving this variable and a dummy variable assuming value 1 in the sub-sample where ‘Technological Uncertainty’ is low and 0 otherwise. (For robustness, we also include in this regression the dummy variable alone as a ‘main effect’.) We observe that the coefficient of such interaction is negative and statistically significant (p < 0.05). Therefore, the effect of ‘buyer–supplier alliance intensity’ is significantly lower in the case where technological uncertainty is low.