Third, the two dimensions can be separated empirically. For example, in many organizations, the group initiating inter-firm relationships and involved in the crafting of the original contracts is quite distinct from the group in charge of the implementation of the agreement. The contractual coordination mechanisms are frequently negotiated by top-management and a group of lawyers, while the setting up of procedural coordination is left to business-unit managers, who have usually been involved in similar alliances in the past (Haspeslagh and Jamison 1991; Sjoberg 1992; Thomas and Trevino 1993). Whenever such functional separation is not carefully bridged, however, the negotiation and the implementation aspects of inter-firm relations are de facto detached, increasing the chances that the relationship will fail.

In the first part of this paper, we build on different theoretical perspectives to articulate the characteristics and specificities of contractual and procedural coordination. We argue that they are influenced - at least partly - by the same underlying constructs, but we stress how they describe complementary aspects of the relationship. We focus on how the mix of contractual and procedural coordination can be determined by characteristics of the tasks to be addressed in the relationship, and how this relates to the relationship performance. While other factors such as the partners' characteristics and the larger context can be expected to impact the structures and processes of a relationship, keeping the environment and firm-specific factors constant and concentrating on task characteristics allows us to focus on those explanatory variables that are more likely to be directly under partners' control.

In the second part of the paper, the third and fourth sections, we use a quantitative meta-analysis on a sample of 32 studies on inter-firm relationships published during the last 15 years to investigate whether the proposed link between task characteristics, contractual and procedural coordination, and relationship outcome is reflected in previous empirical studies. The meta-analysis reveals that previous research investigated both dimensions of a relationship structure. None of the studies, however, analyzed the two dimensions simultaneously. Task characteristics drive both contractual and procedural coordination, and these, in turn, appear to drive the performance of the relationship. Yet, since the empirical studies reviewed did not investigate contractual coordination and procedural coordination simultaneously, neglecting one of the dimensions might lead to spurious results, and validity concerns have to be raised. We conclude, in the fifth section, by discussing the implications of our results for the design of inter-organizational research.

Two Dimensions of Inter-firm Relationships: Contractual and Procedural Coordination

Several streams of research investigated inter-firm relationships, concentrating on different characteristics of the relationships (Oliver 1990; Ring and Van de Ven 1992; Grandori and Soda 1995). Two fundamental dimensions appear to be of key importance for the governance and management of such relationships: contractual and procedural coordination. Both dimensions characterize solutions to the coordination problem that arises when activities are distributed among several actors. These dimensions describe the distribution of rights and the coordination of information flows.

Contractual coordination refers to the mutual exchange of rights between the parties involved in a relationship in order to govern the combination of agents or functions towards the production of results. The nature and characteristics of these rights may vary along the notion of the hierarchical contractual elements proposed by Stinchcombe (1985). They can encompass the definition of a command structure and an authority system used to govern the exchange, as well as some predetermined incentive systems. Moreover, these rights define the setting up of operating procedures to govern the exchange and resolve possible disputes among the partners.

Whether related to the distribution of decision power or of information among partners, however, the formal definition of a governance structure based on the distribution of rights underscores the role played by the actual implementation of the relationship as a necessary complementary element to be designed. Procedural coordination relates to the mutual exchange of information for the combination of agents or functions towards the production of results. These exchange opportunities and mechanisms might be structurally identified by the form of contractual mechanisms chosen and someone might say that they are still governed by contracts, in the form of so-called internal or psychological contracts. Yet, the implementation of how, when, by which means, and to what extent information is exchanged among the partners typically represents an area of direct organizational influence of the process, whose discretional enactment can sensibly affect the outcome of the relationship.

In the remainder of this section, we characterize these two dimensions, building on four different theoretical perspectives: transaction cost economics, structural contingency, organizational learning, and resource dependency. On the one hand, contractual coordination is a core construct of transaction cost economics (Williamson 1975; Williamson 1985), whose research concentrates on relationship governance mechanisms (Anderson 1985; Joskow 1985; Levy 1985; Joskow 1987; Masten et al. 1991). On the other hand, procedural coordination is strongly embedded in structural contingency (Burns and Stalker 1961; Woodward 1965; Lawrence and Lorsch 1967; Thompson 1967; Galbraith 1974; Duncan 1976) and organizational learning approaches (Fiol and Lyles 1985; Levitt and March 1988; Kogut and Zander 1993; Nonaka 1994), which are both primarily concerned with the differentiation, integration and adaptation of actions within organizations. Finally, resource dependency appears to link contractual and procedural coordination (Pfeffer and Salancik 1978; Astley and Van de Ven 1983; Ulrich and Barney 1984). However, the relating empirical work tends to concentrate primarily on the contractual dimension (Pfeffer and Nowak 1976; Shan 1991).

Contractual Coordination

The use of interfirm relationships implies that at least some activities are divided up among the parties. This division of labour entails the need for coordination and re-integration. The distribution of rights among the partners is a central determinant of how coordination can occur. It affects the possibilities for each partner to control the coordination of activities performed within the relationship. When entering a relationship, each partner gives up some of its rights and gains others through either explicit or implicit contracts. The resulting allocation of rights and the institutions relating to these rights determine the governance structure.

The choice of the governance structure to minimize the sum of production and transaction costs for a given transaction is the core issue investigated by transaction costs economics (Williamson 1975; Williamson 1985). Although the term 'governance structure' is frequently used in a very broad sense to encompass every measure to organize, structure, and guide economic behaviour, the theoretical definitions (Williamson 1979) and empirical operationalizations (Armour and Teece 1980; Joskow 1987) primarily refer to the contractual dimension. That is, the governance structure consists of contractually determined means to coordinate the behaviour of the partners in the relationship.

According to this understanding, the governance structure encompasses how the partners obligate themselves to a specific course of action or establish a general commitment to a specific relationship via contractual mechanisms, which define such aspects as how equity (and thus decision rights) is exchanged among the partners, and whether or how instruments such as joint board meetings, personnel exchange, and performance-based incentive systems are to be conducted to govern information rights. When setting up a governance structure, the partners have to choose between either prescribing and enforcing specific actions or using means to create a general commitment between the partners from which desirable actions evolve (Williamson 1983). In some cases, a general commitment to the relationship between the transaction partners can serve as a substitute and provide the necessary incentives to perform the desired actions (Parkhe 1993b).

Transaction costs economics propose that the necessary degree of relationship-specific commitment depends on two aspects of the underlying transaction: asset specificity and uncertainty (Williamson 1989). Asset specificity defines the irreversibility of the investments involved. It can be related to physical as well as human capital investments. The construct uncertainty is applied to a variety of phenomena ranging from individual tasks to market conditions. One common denominator appears to be that uncertainty is negatively related to the ability to bind actors in a meaningful way to a specific course of action through contracts (Joskow 1985; Joskow 1987).

A slightly different perspective on the question of how to govern transactions or relationships is offered by resource dependency (Pfeffer and Salancik 1978). Similar to transaction costs economics, resource dependency investigates which type of governance structure to choose in a specific situation.

Each interaction, though varying in legality, represents an attempt to stabilize the transactions of organizations through some form of interfirm linkage (...). (...) organizations attempt to establish linkages with elements in their environments and use these linkages to access resources, to stabilize outcomes, and to avert environmental control.' (Pfeffer and Salancik 1978: 144)

Differences between these two approaches to the determinance of the appropriate governance structure depend on their epistemological roots. Transaction costs economics inherits from neoclassical economics the assumption that any transaction object is perfectly transferable (Conner 1991). In other words, it investigates only the enforceability of a specific transaction, taking its feasibility as given. Moreover, a main assumption is that the output of a specific transaction is not influenced by how the transaction is structured. For example, the design of a new chassis will not differ whether it is entirely developed by a car manufacturer or jointly developed with a supplier. Any consequences of task partitioning choices will only be reflected in the final costs.

Resource dependency theory, on the contrary, is based on an organizational perspective rooted in the work of Thompson (1967). Inter-firm relations are seen primarily as organizational issues (Ulrich and Barney 1984), and transactions are not assumed to be perfectly feasible in any environment. Two aspects of the feasibility are discussed (Pfeffer and Salancik 1978: 143). First, for any given transaction, the existence of external constraints may inhibit its occurrence, even within the most appropriate governance structure. Reasons may vary from a lack of resources to institutional decisions about agents' separation. Second, the nature of the task itself might limit the options available to govern its completion. For example, the acquisition of some assets, such as technical know-how, may require dense communication between the transaction partners. A governance structure not allowing for such communication flows would make the transaction unfeasible. Contractual coordination mechanisms, therefore, have to be aligned with procedural coordination mechanisms.

Procedural Coordination

Contractual coordination mechanisms provide institutions for achieving the alignment of incentives among the panners. However, from the availability of these institutions, it is impossible to deduce how they are actually employed to coordinate the activities of the partners in the relationship. Even if two organizations have contractually agreed on powerful institutions for coordination, it does not imply that they necessarily do coordinate their actions. Doz, Hamel and Prahalad argue that the actual coordination is not achieved through contractual mechanisms but, rather, is realized by the day-to-day communication of the employees involved in the activities of the relationship: 'Top management puts together strategic alliances and sets the legal parameters for exchange. But what actually gets traded is determined by day-to-day interactions of engineers, marketers, and product developers' (1989: 136).

These 'day-to-day interactions' are at the core of the construct 'procedural coordination'. This describes the extent to which the parties coordinate their processes by exchanging information, thereby making them learn to adjust their activities to each other. Procedural coordination does not refer to institutions that may be in place to govern the relationship, and which might as well include specific agreements about information rights, but asks how these institutions are used. In other words, the institutional perspective of the contractual coordination dimension is supplemented by a process-oriented perspective.

The problem of how to coordinate and re-integrate the activities of several actors is tied to the core research question of structural contingency theories: How should an organization structure and coordinate its activities among different units, given specific task characteristics and other contextual factors? Tasks can be characterized along a multitude of dimensions, such as the resources needed and the interdependency with other tasks. Building on information processing models (Galbraith 1974), the structural contingency framework focuses on task-related uncertainty, which has been operationalized in different ways such as, for example, the volatility of the task environment (Bums and Stalker 1961; Tushman and Anderson 1986) or the degree of differentiation in the task environment (Lawrence and Lorsch 1967).

Among the different definitions proposed, Thompson's (1967) dual notion of uncertainty seems appropriate in this context. He distinguished between uncertainty about the actions needed to achieve a certain goal and uncertainty about the goal itself. This distinction is implicit in several contributions following the decision-making tradition (Tversky and Kahneman 1974) and has also been formulated as a distinction between uncertainty and ambiguity (Schrader et al. 1993). Taken together, these two dimensions are useful in identifying the extent of complexity of the project observed, and where attention should be addressed in order to obtain the expected outcome.

Depending on the combinations of the different levels of uncertainty, different organizational structures are deemed appropriate and are articulated along dimensions such as degree of formalization, level of specialization, and direction of influence (Lawrence and Lorsch 1967; Duncan 1972). These organizational design variables characterize the mechanisms used to achieve procedural coordination between actors, given a specific task partitioning. They all influence patterns of information exchange within an organization (Galbraith 1974; Allen 1986; Larson and Gobeli 1988).

Despite its focus on the intra-organizational level, the structural contingency perspective can be transferred to the inter-organizational level (see, e.g. Rebentish and Ferretti 1993; Schrader 1994). In fact, the empirical evidence stemming out of recent studies on the impact of inter-organizational relations on actors' performance (see, e.g. Clark 1989) points to the interrelationship between level of task uncertainty, measures of procedural coordination, and performance.

The coordination mechanisms proposed by structural contingency frameworks are based on information exchange. For such an information exchange to be an effective coordination mechanism, the interaction partners have to adjust their cognitive frameworks and have to develop a joint language, that is, they have to engage in learning processes. This dynamic perspective has been neglected by the structural contingency frameworks. However, several studies on inter-firm relationships indicate how they evolve over time (Boari 1993). Nanda and Williamson (1994), for example, show that joint ventures can serve as an instrument for companies to learn about each others' skills and resources. This learning enables the companies to determine which of the partner's resources are of interest and how to value these resources.

Organizational learning theory explicitly provides such a dynamic perspective. Although extremely varied in its focus and level of analysis (Levitt and March 1988; Huber 1991), the basic organizational leaming question can be formulated as: How can learning processes be structured or enabled, given the nature of the knowledge to be learned? The emphasis is once more on the nature of the task, interpreted in this case as the transfer of knowledge between two or more agents. Nonaka (1994) uses two dimensions to describe knowledge: its manifestation (tacit vs. codified) and its location (individual vs. group). The first dimension can be assimilated to Thompson's (1967) distinction between the different types of uncertainty, which can be found in the other three perspectives reviewed. The emphasis is on the specific characteristics of the task/object/action observed.

The second dimension introduces, from a theoretical perspective, the importance of the localization of information in any transaction. Learning possibilities exist whenever knowledge is unevenly distributed. However, as von Hippel (1994) points out, an uneven distribution per se does not imply that two tasks need to be strongly coupled, since some information is easily transferable, whereas other information is 'sticky' - i.e. context specific. Only in those cases in which information is sticky, is it necessary to invest in high levels of procedural coordination mechanisms. Consequently, the amount of task-related sticky information determines the choice of the procedural coordination mechanisms to be used to structure the relationship. Thus, the information location issue is isomorphic to the issue of task uncertainty.

The main difference between structural contingency and organizational learning is the acknowledgment by the latter that the nature of tasks may change during learning processes. This implies that the appropriate level of procedural coordination is not a constant, but needs to change as well. In other words, the search for the 'right' coordination structure is doomed to be fruitless as long as it is not taken into consideration that task and structures need to adapt to each other and thus change constantly in a continuous feedback system. A similar notion has been suggested by Cainarca et al. (1992), who found that the occurrence of different types of inter-firm relations changes during the technology life-cycle. At any given moment, however, task and structure need to be in accordance to each other. Thus, using a static view, the underlying structures of structural contingency and the organizational learning frameworks are similar.

In conclusion, both theoretical approaches help to identify mechanisms allowing parties to coordinate their processes. These mechanisms are based on the exchange of information. The higher the level of task uncertainty, the greater the need for procedural coordination. The purpose of procedural coordination is that actors exchange sufficient information so that they can adjust their mutual behaviour in a meaningful way for any given associated distribution of rights among the partners. In other words, the level of procedural coordination can be described through the quantity and complexity of the information exchanged. The organizational learning framework indicates, however, that relationships should not be viewed as static entities. Rather, through ongoing learning, task characteristics change constantly. Therefore, the mechanisms used for procedural coordination have to be adjusted accordingly.

Linking the Different Perspectives

With contractual and procedural coordination two different dimensions of inter-firm relations have been identified. The first dimension relates to coordination through the exchange of rights, while the second one describes coordination through the exchange of information. Several authors assume that contractual and procedural coordination are closely linked to each other (Pfeffer and Salancik 1978; Helper 1991). Empirical evidence, however, suggests that this is not necessarily the case. There are several examples of firms having established a high level of procedural coordination without intense contractual coordination (Piore and Sabel 1984; Best 1990). Similarly, firms might set up institutions for considerable contractual coordination without establishing significant procedural coordination (Joskow 1987).

The choice along both dimensions is related to the characteristics of the tasks performed in the relationship. Combining the different perspectives reviewed, the task domain can be articulated by focusing on three aspects (Table 1). Asset specificity determines the extent to which the activities performed in the relationships have some economic value per se or not. The higher the asset specificity, the lower the chances for the partners to benefit from their activities outside the relationship. The level of task uncertainty can be referred to the action or to the goal domain. In the first case, the partners have agreed on the objective of their relationships, but have multiple options to achieve the goal. Ring and Rands (1989), for example, document how NASA and 3M were able to specify up front their goal for common projects on microgravity experiments and subsequently worked through the implementation of their collaboration. In the second case, the goal itself is unclear. These situations have been documented, for example, by studies of collaborative R&D projects, where the partners might initiate the relationship for some generic strategic reason, but lack an operational objective (Tripsas et al. 1995).

Given the different characteristics of the tasks to be performed within the relationship, the partners structure their interaction by (a) articulating the legal conditions governing the transaction and (b) identifying the mechanisms to transfer information among themselves in order to implement the transaction. Contractual coordination mechanisms are used to define the legal boundaries of the relationships (Table 1). They involve the choice of the legal form governing the agreement (ex. joint venture or strategic alliance, equity and non equity joint venture, etc.), [TABULAR DATA FOR TABLE 1 OMITTED] the length of the agreement, the extent to which partners are bound to the agreement (ex. exclusivity clauses, penalties, etc.) and the degree to which these choices are specific to the agreement or not (ex. standardized vs. personalized contracts), and the distribution of information rights.

Procedural coordination mechanisms, instead, are used to put into place the relationships within the institutional boundaries defined by means of the contractual coordination mechanisms chosen. They are targeted to the structuring of the information flows between the partners (Table 1). Decisions on the frequency, timing and directionality of information flows, as well as the means through which these flows occur, identify the operational dimensions of procedural coordination mechanisms. For a given institutional setting, defined by a specific combination of contractual coordination mechanisms, alternative choices of procedural coordination mechanisms are going to impact the outcome of the relationship differently. Carter and Miller (1989), for example, show how frequent and bilateral communication between vendors and buyers limited the occurrence of quality problems in the materials delivered.

To some extent, investments in procedural coordination mechanisms may substitute for contractual coordination mechanisms. Procedural coordination does not come for free, though (Sobrero and Roberts 1996). Leaming has to take place, interfaces need to be established, lines of communication reorganized, etc. The investments necessary for these changes may already sufficiently commit the relationship partners to each other. Consequently, additional contractually based commitment mechanisms might not be necessary anymore. In other words, while the task characteristics may suggest that the partners should establish a high level of contractual coordination, it may actually be unnecessary to do so if the partners are already committed to each other through investing in procedural coordination.

In general, therefore, the simultaneous inclusion of the characteristics of the task partitioned among the partners, the choices along the structuring dimensions (contractual and procedural coordination) and the consequences of such choices should guide the design of inter-organizational research to address the following question: Do combinations of contractual and procedural coordination, given specific task characteristics, explain performance differentials? Not recognizing the complementarity of these dimensions might lead to different and narrower research questions.

Figure 1 summarizes the three constructs emerging from the analysis of the four theoretical perspectives on inter-firm relationships, and how they are linked. Moreover, it identifies a set of six distinct research questions which can be formulated to focus on one specific portion of the larger framework. Although clearly offering a partial view of a more complex set of interactions among the different constructs, each question reflects the main concern of the theoretical perspectives reviewed. In the following sections, we use this set of questions to guide the quantitative meta-analysis on previous research on inter-firm relationships to classify the studies reviewed. Our objective is twofold. First, we want to formally summarize the empirical evidence on the relationships among the theoretical constructs identified. Second, we want to examine the consequences of a partial research focus on specific aspects of the linkages between task characteristics, contractual and procedural coordination, and outcome of the relationship on the validity of the results.

Quantitative Meta-analysis of Inter-firm Relationship Research

Sample

To review the empirical results on inter-firm relations, we conducted an electronic search of the articles published in the major international management and economics journals from January 1987 to April 1994. After carrying out a review of the electronic sources by collecting and indexing abstracts of social science journals, we decided to rely on ABI-Inform only. Certain databases are too specific (ex. Economic Abstracts), while others are too generic (ex. Lexis/Nexis). ABI-Inform, on the contrary, lists all the major social science journals, covering a broad range of disciplines.

The list of the journals used for the search is reported in Table 2. Although certainly not exhaustive, it is a representative sample considering the range of disciplines covered, the differences in the targeted audiences (managerial vs. academic), and in the empirical approaches used. Moreover, the use of the file drawer procedure described below (see p. 600 below) will serve as an additional check for a possible bias in the selection process.

The search was conducted using an informed set of keywords derived from the theoretical analysis and from reviewing some older seminal articles. It included the words 'collaboration', 'cooperation', 'partnership', 'alliance', 'joint venture', 'inter-firm' and was specifically addressed towards empirical pieces by also including among the keywords 'data', 'empirical', and 'survey'. The search generated a list of 118 articles. based on the abstracts, all articles involving the analysis of interactions among non-profit organizations or focusing on university-industry relations were excluded. Articles based on secondary empirical sources to formulate prescriptive indications were also excluded. These first filtering processes reduced the original list to 46 empirical articles on inter-firm relations published between 1987 and 1994 in the major academic journals. This list was completed by scanning the references of the selected articles and looking for those past works which were widely quoted, but not included in the sample. The whole process resulted in a total of 56 articles.

After a careful reading of each of the 56 articles, we retained 32 of them for the meta-analysis. In addition to the previously discussed exclusion criteria, several had to be excluded, because they reported results from surveys which were not based on any theoretical framework (e.g. Kleinknecht and Reijen 1992). Others approached inter-firm relationships by relating them to their external environment conditions, such as market structure, industry life cycle and the like (e.g. MacDonald 1985; Cainarca et al. 1992), but they did not take into account task-related factors. Finally, other researchers continued to move on a theoretical ground, limiting their empirical contribution to a purely speculative one (e.g. Hamel 1991).

The selected studies were grouped along the six questions presented in Figure 1. While some of the studies addressed more than one question, all considered each question independently. In other words, interaction effects were not investigated. Within each group of studies, we recorded information concerning the theoretical framework used, the relevant dependent and independent constructs included in the analysis, their operationalization, the sample size, and the significance level of the results.

Methods

Meta-analytic procedures are used to transform literature reviews from purely qualitative realms into quantitative ones. Each study reviewed is treated as an observation, making it possible to compare or combine the empirical evidence emerging from the studies reviewed, using the reported significance level or effect size as a starting point (Hunter et al. 1982). The first purpose of any meta-analysis is to partition the overall variance observed in the reported results to partial out study-specific variance (Hunter and Schmidt 1990). After any spurious source of variation has been partialled out, meta-analysis can be used essentially for three purposes (Rosenthal 1991): (1) to help summarize the evidence emerging from several studies in which the relationship between two or more variables are investigated, as we do in this study; (2) to isolate a set of moderator variables and verify their overall impact on the relation being studied; and (3) to generate hypotheses by clustering the examined studies along variables not directly observed or measured.

Despite its merits, however, meta-analysis suffers from some substantial and computational limitations (Hedges and Olkin 1985; Rosenthal 1991), among which three are particularly relevant in our case. First, to fully benefit from all potential outcomes of meta-analytic techniques, experimental studies with reported effect size estimates are needed. While the use of experiments is rather common in certain disciplines (i.e. experimental psychology), social sciences often focus on levels of analysis which make experiments unfeasible - the analysis of inter-firm relations is such a case. Regardless of the perspective used, whether rooted in economics or in organization theory, we are dealing with non-experimental studies, where different covariates are often included.

Since we have to accept the non-experimental nature of studies at the firm level, we are fairly limited in any meta-analysis involving the comparison and combination of effect sizes. Obtaining indicators partialling out the effect of covariates might be impossible, depending on the type of results reported. In addition, the presence of difference covariates in the different studies reviewed will increase the magnitude of the problem (Rosenthal 1991: Chapt. 2). We will therefore focus on the directionality and significance of the effects rather than on their magnitude, and concentrate accordingly on significance levels.

Second, the same theoretical constructs are frequently operationalized and measured in different ways. The effect that weaknesses in construct validity and reliability might have on the observed results is a common concern for meta-analysts. We address these limitations computationally in the following pages, presenting how we combined multi-indicator constructs and how we used and interpreted the results of multivariate procedures.

Third, meta-analysis is frequently criticized as being based on biased data sets, since the published studies are only a fraction of all the studies performed on a certain topic (the so called 'file drawer problem'). Typically, the objection is that studies reporting non significant results rarely get published. While there is accordance on the other problems of meta-analysis discussed, this specific point is still harshly debated (Hunter and Schmidt 1990; Rosenthal 1991). On the one hand, it is argued that unpublished studies are such because inherent methodological flaws make their results unreliable and account for the weakness of their findings. On the other hand, one might argue that deviants to well established 'paradigms' are more likely to encounter resistance within the scientific community, and therefore are less likely to be published.

Both positions focus on the possible reasons for the presence of a high number of unpublished studies. Whatever these reasons are, however, one would like to estimate their potential impact on the external validity of the results of the meta-analysis. Rosenthal (1991: 103-109) has proposed approaching this issue by calculating how many studies reporting non-significant results, or even results contradicting theoretical predictions, need to be laying in some drawers waiting publication for the conclusions reached by the meta-analysis to be invalidated. The higher the number of unpublished studies needed to invalidate such conclusions, the smaller the selection bias and the greater the generalizability of the results (see also Hunter and Schmidt 1990: Chapt. 13). In this analysis we use the statistical procedures introduced by Rosenthal to assess the external validity of the results.

Data Coding

For each study, we recorded the one-tailed p-value associated with the reported test and found the corresponding standard normal deviate (z). The sign of the Z-score was determined by whether the empirical evidence supported the underlying theory or not. If the theory was supported, the Z-score was recorded as positive. If it was not supported, the Z-score was recorded as negative. For example, transaction cost economics predicts that if asset specificity is high, task partitioning will be low. Results showing a negative association between the two constructs were coded as supporting the theory, while results showing a positive association were coded as not supporting the theory.

An alternative coding procedure is to use the indications emerging from the whole set to make choices as to the directionality issue. For example, if the majority of the study reviewed shows a negative association between X and Y, all the positive ones will be coded as negative, and vice versa. Since we were testing specific theoretical predictions, however, we did not use this coding schema. Finally, if p-values were not available or reported only as threshold levels (e.g. p [less than] .05), we used the reported t-statistic and found the exact p-value using the associated distribution.

Whenever more than one indicator was used in the study reviewed for a specific construct, we first determined the standard normal deviate (Z) corresponding to the p-value associated with each indicator. The Z's were then averaged to calculate the corresponding p-value (Rosenthal 1991: 27-28). The same procedure was used whenever more than one analysis was performed on the same functional relationship within the same study and on the same sample. For example, a standard procedure in economic research is to present and estimate different functional forms associated with the model developed, and compare the emerging results. In this case, we considered each estimate as a single set of results and combined them all (Hunter et al. 1982: Chapt. 5). Analyses on different samples performed within the same study, however, were treated as fully replicated designs, and therefore as independent observations.

Statistical Tests

To compare the studies reviewed along their probability levels, we tested for the heterogeneity of the corresponding Z's (Rosenthal and Rubin 1979). In particular, given N equal to the number of studies reviewed, [Z.sub.i] the standard normal deviates computed as described above, [Z.sub.[Mu]] the average of the [Z.sub.i], we computed the following test, which is distributed as [[Chi].sup.2] statistic with N-1 degrees of freedom:

[summation of] [([Z.sub.i] - [Z.sub.[Mu]]).sup.2] where i = 1 to N Test of Heterogeneity of the Probability Levels (1)

If results were not homogenous, the studies were carefully investigated individually to identify possible causes for the discordance of the results. Where there was accordance, we tested whether it was highlighted because of chance alone or rather because of an underlying pattern in the population. When finding accordance with respect to the results obtained, we were then interested in assessing whether the accordance was systematic or random. For this purpose, we combined the results.

The logic of the combining step is the usual one of statistical testing. The purpose is to get an overall estimate of the probability that the p-values of the studies reviewed might have been obtained if the null hypothesis of no relation between the variables observed were true. Different tests are available for combining independent probabilities obtained from two or more studies testing the same directional hypothesis (Rosenthal 1978). Each one presents advantages and disadvantages. Given the limited number of studies for each research question, we avoided relying on a single test, and used more than one to combine probabilities obtained from the review. Table 3 reports the tests used, their computation and a brief description of their advantages and disadvantages.

All these tests require that the studies reviewed employ quantitative analysis methods. Any evidence emerging from qualitative studies would therefore [TABULAR DATA FOR TABLE 3 OMITTED] be lost. Yet, in social science research, qualitative methodologies are frequently used. In this sample, for example, studies use firm-level cases (ex. Bertodo 1990) or industry-level longitudinal analysis (ex. Pisano 1991). To benefit from their results as well, we included each of these studies in the appropriate group/s and coded their result with respect to their theoretical expectations. Combining this information with that previously recorded for the quantitative studies, we derived for each group the number of studies that showed results in accordance with the theory and the number of studies that did not. We then used the binomial probability theorem to find the probability of k successes (with k equal to the number of studies supporting the theory) in n trials (with n equal to the number of studies reviewed) to assess the likelihood that the distribution of results could be obtained by chance alone. While less accurate, this procedure allows the inclusion of qualitative results in the analysis.

Finally, to approach the 'file drawer problem' we followed the procedure recommended by Rosenthal (1991: 103-109). The purpose was to estimate the number of unretrieved studies averaging null results which should exist if the results obtained from the retrieved studies were due to chance alone. Traditionally, meta-analyses consider unretrieved studies as those which have not yet been published. The procedure, however, can logically be extended to consider the generalizability of the results, which might as well be affected by the sampling criteria used. The file-drawer test therefore also becomes a way to control for potential biases introduced by our decisions to exclude some papers, as explained above at the beginning of the third section, and to have not included other journals in our analysis.

Technically we needed to calculate the number of unlocated studies averaging null results to bring the significance level of the combination test down to 0.05. Given that N is the number of studies reviewed, [Z.sub.[Mu]] the average of the [Z.sub.i] obtained from each of the retrieved studies, we used the following to compute x, the number of studies to be exceeded to invalidate the results:

x = N/2.706 [N[(Z.sub.[Mu]]).sup.2] - 2.706] File Drawer Test (2)

Results

A first reading of the empirical studies collected for the meta-analysis reveals the absence of an integrated perspective that simultaneously considers the relationships between task characteristics, the contractual and procedural dimensions of inter-firm relationships, and the outcome of the relation. While studies on inter-firm relations are numerous, they are parsimonious and diversified in aim and scope. Table 4 lists all the studies reviewed, the dependent and independent constructs, the theoretical predictions as to the directionality of the effects, the directions of the effects observed in the empirical analysis, the corresponding p-values and the degrees of freedom used in the empirical analysis. The studies are grouped by the six questions previously discussed, each one addressing a separate relation highlighted by the theoretical framework. In particular, among the studies using quantitative methodologies, six examined the relationship between task characteristics and level of contractual coordination, nine the relationship between task characteristics and procedural coordination, five the relationship between level of contractual coordination and performance of the relation, and eight the relationship between organizational coordination and performance of the relation.

The remaining 20 studies using quantitative methodologies, while often declaring to be approaching a specific research question, present results on the more general questions linking task characteristics and task partitioning decisions (18 studies), and task partitioning and performance (2 studies). The first evidence emerging from the analysis is the heterogeneity of the significance levels reported in the studies investigating the relationship between task characteristics and task partitioning. A visual inspection of the data reported in Table 4 for the studies addressing Question I shows a general accordance between theoretical predictions and empirical observations. This is confirmed by the binomial probability test (p [less than] .001). The heterogeneity test, however, reveals that the p-values are statistically different ([[Chi].sup.2] = 11.82, df = 17, p [less than] .01). The separate use of several different constructs to assess task characteristics might account for the detected heterogeneity. A further analysis at the construct level confirms this intuition. [TABULAR DATA FOR TABLE 4 OMITTED] While significance levels of studies investigating the relationships between complexity and task partitioning ([[Chi].sub.2] = 5.03, df = 4, n.s.), similarity and task partitioning (Z = 0.94, n.s.) and ambiguity and task partitioning (Z = 0.74, n.s.) are homogenous in magnitude, heterogeneity strongly emerged with regard to the relationship between asset specificity and task partitioning ([[Chi].sup.2] = 18.56, df = 6, p [less than] .01) and uncertainty and task partitioning (Z = 2.76, p [less than] .01).

Although there seems to be a diffused agreement about the role played by asset specificity and uncertainty in determining how and if tasks are split among multiple actors, the results do not show such a strong relationship. The few studies reviewed for each of the constructs, however, suggests caution in the interpretation of these findings. Yet, such theoretical agreement is often challenged on an operationalization ground where different types of asset specificity (i.e. human resource, location and physical asset specificity) are lumped in the same indicator.

The observed heterogeneity might therefore derive from operationalizations which fail to distinguish the different components of asset specificity. Similarly, in the case of uncertainty, the resulting contradictory evidence might derive from an unclear distinction between environment-related and task-related uncertainty. Indeed, while Balakrishnan and Wernerfelt (1986) carefully distinguish between environmental and task-related uncertainty, focusing only on the second one, Walker and Weber (1984) confuse an environmental variability indicator (the expected rate of future technological change) with an indicator of task-related uncertainty (the amount of expected changes in components), using the former to measure the latter. More generally, the analysis at the construct level suggests the need for a closer look at the individual studies reviewed whenever heterogeneity is detected.

The evidence for a relationship between task characteristics and level of contractual coordination, addressed by the studies grouped under Question 2, seems to be similar to what emerged with regard to Question 1. While all studies show agreement between expected and observed direction of effect (binomial probability test, p [less than] .001), the comparison of the significance levels reported by each study detects heterogeneity of such significance levels ([[Chi].sup.2] = 11.24, df = 5, p [less than] .05). A further inspection of the p-values reported by each study shows that this result is caused by the magnitude of the discrepancies among the four studies reviewed. In four out of six studies, the p-values are at or just below the 0.05 level, while two are much smaller (around 0.00001).

Therefore the heterogeneity detected does not prevent the combination of studies, since there is agreement in the directionality of the effect, and each study also statistically confirms the presence of an effect. Despite the small number of studies reviewed (6), all the different combining procedures applied confirm the generalizability of the results observed in the sample (see Table 5), pointing to a relationship between task characteristics and level of contractual coordination. Moreover, according to the 'file drawer test' there should be a very large number (more than 87) of unpublished [TABULAR DATA FOR TABLE 5 OMITTED] studies reporting findings that question the theoretical predictions invalidating this conclusion.

The meta-analysis, however, does not provide evidence for a relationship between task characteristics and procedural coordination mechanisms. Already a visual inspection of the studies listed in Table 4 shows discrepancies between theoretical predictions and empirical results. The binomial probability test confirms these indications with a value of 0.05. In addition, the p-values of five out of nine studies are all fairly large, ranging from roughly 0.4 to roughly 0.8. Accordingly, the heterogeneity test finds a significant difference among the p-values reported by the studies ([[Chi].sup.2] = 24.44, df = 8, p [less than] .01). Further analysis at the construct level shows that the heterogeneity detected also holds for the relationship between asset specificity and procedural coordination ([[Chi].sup.2] = 11.18, df = 2, p [less than] .05), and for the relationship between dependency and procedural coordination ([[Chi].sup.2] = 13.38, df = 2, p [less than] .01).

The concerns previously raised about the operationalization of asset specificity find here a first tentative confirmation. In addition, the studies addressing Question 3 rely on a fairly low number of observations which limit their power. The lack of power might also account for the results regarding the relationship between ambiguity and procedural coordination, which are statistically homogeneous, but clearly not significant, although generally in the predicted direction (Z = 0.29, n.s.). Finally, the discrepancies might suggest the presence of more fundamental issues related to the unfeasibility of a clear unidirectional relationship between task characteristics and procedural coordination mechanisms, due to the role of context-specific elements such as the emergence of heterogeneity in the partners' experience and 'learning-by-doing' processes.

The relationship between task partitioning in general and performance addressed by the studies investigating Question 4 is supported by the metaanalysis. Both Armour and Teece (1980) and Clark (1989) independently conclude, working at different levels of analysis, that task partitioning increases performance levels (innovativeness in the first case, product development efficiency in the second case). These results were confirmed when we examined the relationship between each of the two dimensions of task partitioning and performance levels at a more detailed level.

In answering Question 5, we found evidence for homogeneity of the findings reviewed regarding the relationship between levels of contractual coordination and performance ([[Chi].sup.2] = 2.25, df = 4, n.s.). Combining tests confirmed the generalizability of that conclusion and the file drawer test indicated that 30 unpublished studies reporting results in the opposite direction would be needed to falsify this conclusion (see Table 5).

Similar results emerged when reviewing the studies investigating the relationship between procedural coordination and performance. The p-values of the eight studies addressing Question 6 were not heterogeneous ([[Chi].sup.2] = 2.27, df = 7, n.s.) and could therefore be combined to test for their generalizability. The different combination tests strongly confirm that the significance levels of the studies reviewed are not homogeneous in magnitude and direction by chance alone. Furthermore, the file drawer test indicates that a large number of unpublished studies, more than 85, reporting results in the opposite direction would be needed to falsify this conclusion. We can therefore conclude that there is strong evidence for a relationship between procedural coordination and performance.

Discussion and Conclusions

The quantitative meta-analysis of 32 studies presented in this paper supports the distinction between contractual and procedural coordination mechanisms as two separate and complementary dimensions for the structuring of inter-firm relationships. Moreover, it confirms the links between task characteristics, contractual and procedural coordination, and relationship outcome, although none of the analyzed studies simultaneously investigated all these different aspects.

Contractual coordination mechanisms are an important dimension in structuring inter-organizational relationships. Their choice depends on the taskrelated characteristics and it affects final performance. The evidence about the role of procedural coordination mechanisms, on the contrary, is mixed. They clearly appear to influence the performance of the relationship. The empirical findings regarding the relationship between task characteristics and the level of procedural coordination, however, is more ambiguous. First, it is statistically problematic to combine the studies relating task characteristics and procedural coordination due to magnitude differences in the reported significance levels. Second, substantive concern relating to the variance in the operationalization of the theoretical constructs used within the different studies needs to be raised. Finally, the limited results on the procedural coordination dimension might also depend on the characteristics of our sample of studies. Due to the choice of journals used for the analysis, the papers reviewed tend to have a stronger focus on economics and strategy, where more attention is usually paid to contractual coordination mechanisms, than in sociology or political science, where the focus tends to be on procedural coordination mechanisms (Oliver and Ebers 1998).

Neither the studies focusing on contractual coordination mechanisms nor those focusing on procedural coordination mechanisms, however, simultaneously considered the relationship between task characteristics, coordination, and performance. The research focusing on how task characteristics impact coordination, i.e. the structuring of a relationship, takes the performance implications as given; the research investigating how the structuring impacts performance, assumes that choices among structuring alternatives are not constrained by task characteristics. This separation between the determinants of the choice among structural alternatives and the effects of such a choice is unsatisfactory. Performance implications are neglected and the variables omitted raise considerable concerns with regard to validity.

(a) Neglected performance implications: Research investigating the correspondence between task characteristics and contractual or procedural coordination mechanisms use the same assumption as traditional context-structure models (Van de Ven and Drazin 1985; Scott 1990). Typically, these models test the fit between context and structure through correlation analysis of these constructs, not including any measures on performance or any interaction terms (Joskow 1987; Heide and John 1990; Shan 1991; Heide 1994). Decision makers involved in the choice among different structural alternatives are therefore not provided with any evidence on whether certain solutions are chosen because they are indeed the most appropriate ones, or rather because they are, for example, an institutionalized response. In other words, this research does not give any guidance regarding the structuring of relationships that goes beyond the repetition of existing patterns.

(b) Omitted variables: Research investigating the influence of contractual or procedural coordination on performance typically limits its investigation to one of the coordination mechanisms, thereby neglecting the other. From this, omitted variable biases may arise. Empirically, contractual and procedural coordination are not orthogonal (Parkhe 1993b). Consequently, limiting the research to only one coordination mechanism at a time is an inappropriate simplification, and it is not clear whether the results are valid. From a managerial perspective, this approach legitimizes an apparently inappropriate separation of the contractual and procedural aspects of relationships.

Although addressing separate needs of the structuring problems, contractual and procedural coordination mechanisms might be considered as complementary dimensions of investments in a longitudinal perspective. For example, specific investments in procedural coordination mechanisms such as specific interfaces, lines of communication and the like may already sufficiently commit the relationship partners to each other. Consequently, additional contractually based commitment mechanisms, once deemed appropriate, might not be necessary any more. Similarly, relationships might start out with tight contractual coordination and a need for complex procedural coordination. In the course of the relationship, however, the parties might develop trust or learn sufficiently so as to be able to loosen their contractual bonds or they may consider the information transfer to be embedded and relax their efforts to achieve procedural coordination (Roberts and Berry 1985; Nanda and Williamson 1994).

This suggests that it is not possible to determine the optimal combination of contractual and procedural coordination, given specific task characteristics alone. Several models of fit have been used to determine levels of coordination (Scott 1990). On the one hand, models which limit their analysis to univariate context-structure relations fail to account for the multidimensional nature of the context and of the structure, and they do not explicitly consider outcome implications (Van de Ven and Drazin 1985). On the other hand, models of fit which explicitly include outcome indicators typically fail to consider that alternative combinations of structural dimensions may have the same performance implications. They determine each structural dimension independently of the others, without investigating the possible substitution and trade-off effects.

Instead, it is more realistic to assume that several combinations of contractual and procedural coordination are equally well fitted for a given task. This notion is similar to the interpretation of fit in the systems approach, where Van de Ven and Drazin (1985) use the concept of equifinality to define fit as 'any feasible set of equally effective, internally consistent context-structure combinations'. The relationship between task characteristics and structuring alternatives needs therefore to be articulated to incorporate a multidimensional combination of context and structure. Task characteristics are varied and can result in multiple combinations, depending on the actors involved and the context. Similarly, alternatives along the contractual and procedural coordination dimensions of structuring decisions can result in different outcomes, although identified to address similar task characteristics.

The cumulative evidence offered by our meta-analysis on these issues has several implications for the design of inter-organizational research. First, the single relationship rather than a firm, or a set of firms, should be the unit of analysis. Multiple relationships can occur among the same partners (Wellman and Wortley 1990). Moreover, different approaches can be used by the same partner to structure a relationship with different partners (Schrader 1991). Second, sampling on the legal dimension as the distinctive character of the relationship should be avoided, thereby limiting selection biases. Focusing on joint ventures, strategic alliances, or consortia is a typical approach in many studies, which severely limits ex-ante the variance in the contractual coordination dimension and might have similar effects on the procedural coordination dimension.

Third, while a relational perspective has been introduced in the analysis of contracts (MacNeil 1978), it is a promising and rather underexplored area of development for inter-organizational research. Despite all the work conducted in sociology within the social network tradition, only recently have these approaches and tools been rigorously extended to the analysis of inter-firm relationships (Gulaty 1995; Uzzi 1996), proposing interesting developments about the concepts and meanings of structure and coordination in a relational context (Sobrero and Grandi 1997).

Finally, the systemic notion of fit underlying the analysis of inter-firm relationships is also helpful in gaining a longitudinal perspective on how the structuring problem changes during the life of the relationship (Cainarca et al. 1992). Over time, consistency between the characteristics of the task and the structure of the relationship is achieved through a realignment of the structuring alternatives. Indeed, the very meaning of success of a relationship seems to depend on the partners' capability to constantly assess the congruence between the nature of the task and the coordination mechanisms chosen (Haspeslagh and Jamison 1991; Boari 1993; Lorenzoni and Baden Fuller 1995).

Table 2

Journals Selected for the Electronic Search between 1987 and 1994 on ABI-Inform

Academy of Management Executive Academy of Management Journal Academy of Management Review Administrative Science Quarterly American Journal of Economics and Sociology American Economic Review Brookings Papers on Economic Activities Business History Review California Management Review Columbia Journal of World Business Decision Science Econometrica Harvard Business Review IEEE Transactions on Engineering Management International Journal of Industrial Organization International Journal of Technology Management Journal of Economic Behavior and Organizations Journal of Engineering and Technology Management Journal of Industrial Economies Journal of International Business Strategy Journal of Law and Economics Journal of Management Journal of Marketing Journal of Marketing Research Journal of Product Innovation Management Long Range Planning Management Science Managerial and Decision Economics Marketing Science Organization Science Organization Studies Quarterly Journal of Economics R&D Management Rand Journal of Economics Research Policy Research/Technology Management Sloan Management Review Strategic Management Journal Technology Review Technovation

Note

A previous version of this paper was presented at the European Science Foundation EMOT workshop, Program IV in Como, October 1994. We would like to thank Massimo Colombo, Anna Grandori, Richard Lester, Ed Roberts, Peter Smith Ring, Antonello Zanfei and three anonymous reviewers for their valuable comments. Stephan Schrader, an exceptional mentor and friend, died of a brain tumor on July 2nd, 1997.

References

Allen, Thomas J. 1986 'Organizational structure, information technology, and R&D productivity'. IEEE Transactions on Engineering Management EM-33/4: 212-217.

Alter, Catherine 1990 'An exploratory study of conflict and coordination in interorganizational service delivery systems'. Academy of Management Journal 33/3: 478-502.

Anderson, Erin 1985 'The sales person as outside agent or employee: A transaction cost analysis'. Marketing Science 4/3: 234-254.

Anderson, Erin, and David C. Schmittlein 1984 'Integration of the sales force: An empirical examination'. Rand Journal of Economics 15/3: 385-395.

Anderson, Erin, and Burton Weitz 1989 'Determinants of continuity in conventional industrial channel dyads'. Marketing Science 8/4: 310-323.

Anderson, James C., and James A. Narus 1990 'A model of distributor firm and manufacturer firm working partnerships'. Journal of Marketing 54/1: 42-58.

Armour, Henry Ogden, and David J. Teece 1980 'Vertical integration and technological innovation'. The Review of Economics and Statistics 62 (August): 490-494.

Astley, G. W., and A. H. Van de Ven 1983 'Central perspectives and debates in organization theory'. Administrative Science Quarterly 28/245-273.

Balakrishnan, Srinivasan, and Birger Wernerfelt 1986 'Technical change, competition and vertical integration'. Strategic Management Journal 7/347-359.

Bertodo, R.G. 1990 'The strategic alliance: Automotive paradigm for the 1990's'. International Journal of Technology Management 5/4: 375-408.

Best, M. 1990 The new competition: Institutions of industrial restructuring. Cambridge, MA: Harvard University Press.

Boari, Cristina 1993 Le Joint Venture di lunga durata. Milan: ETAS Libri.

Bucklin, Louis P., and Sanjit Sengupta 1993 'Organizing successful co-marketing alliances'. Journal of Marketing 57/2: 32-46.

Burns, Tom, and G. M. Stalker 1961 The management of innovation. New York: The Free Press.

Cainarca, Gian Carlo, Massimo G. Colombo, and Sergio Mariotti 1992 'Agreements between firms and the technological life cycle model: Evidence from information technologies'. Research Policy 21/1: 45-62.

Carter, Joseph R., and Jeffrey G. Miller 1989 'The impact of alternative vendor-buyer communication structures on the quality of purchased materials'. Decision Sciences 20:759-775.

Clark, Kim B. 1989 'Project scope and project performance: The effect of pans strategy and supplier involvement on product development'. Management Science 35/10: 1247-1263.

Conner, Kathleen R. 1991 'A historical comparison of resource-based theory and five schools of thought within industrial organization economics: Do we have a new theory of the firm?'. Journal of Management 17/1: 121-154.

Contractor, F. J., and P. Lorange 1988 'Why should firms cooperate? The strategy and economics basis for cooperative ventures' in Cooperative strategies in international business. F.J. Contractor and P. Lorange (eds.), 3-28. Lexington, MA: D.C. Heath.

Doz, Y. L., G. Hamel, and C. K. Prahalad 1989 'Collaborate with your competitor and win'. Harvard Business Review 67/1: 133-139.

Duncan, Robert B. 1972 'Characteristics of organizational environments and perceived environmental uncertainty'. Administrative Science Quarterly 17/3: 13-327.

Duncan, Robert B. 1976 The ambidextrous organization: Designing dual structures for innovation. New York: North-Holland.

Edgington, E. S. 1972a 'An additive method for combining probability values from independent experiments'. Journal of Psychology 80: 351-363.

Edgington, E. S. 1972b 'A normal curve method fro combining probability values from independent experiments'. Journal of Psychology 82: 85-89.

Fiol, Marlene C., and Marjorie A. Lyles 1985 'Organizational learning'. Academy of Management Review 10/4: 803-813.

Galbraith, J. R. 1974 'Organization design: An information processing view'. Interfaces 4(May): 28-36.

Goshal, Sumantra, and Philippe C. Haspeslagh 1990 'The acquisition and integration of Zanussi by Electrolux: A case study'. European Journal of Management 8/4: 414-433.

Grandori, Anna, and Giuseppe Soda 1995 'Inter-firm networks: Antecedents, mechanisms and forms'. Organization Studies 16/2: 183-214.

Granovetter, Mark 1973 'The strength of weak ties'. American Journal of Sociology 78: 1360-1379.

Granovetter, Mark 1985 'Economic action and social structure: The problem of embeddedness'. American Journal of Sociology 91: 481-510.

Gulaty, Ranjay 1995 'Social structure and alliance formation patterns'. Administrative Science Quarterly 40: 619-652.

Hakanson, Lars 1993 'Managing cooperative research and development: Partner selection and contract design'. R&D Management 23/4: 273-285.

Hamel, Gary 1991 'Competition from competence and interpartner learning within international strategic alliances'. Strategic Management Journal 12: 83-103.

Harrigan, Kathryn R. 1986 'Matching vertical integration strategies to competitive conditions'. Strategic Management Journal 7: 535-555.

Harrigan, Kathryn R. 1988 'Joint ventures and competitive strategy'. Strategic Management Journal 9: 141-158.

Harrison, Benneth 1991 'Industrial districts: Old wine in new bottles'. Regional Development 26/5: 469-483.

Harrison, Benneth 1994 Lean and mean: The changing landscape of corporate power in the Age of Flexibility. New York: Basic Books.

Haspeslagh, Philippe C., and David B. Jamison 1991 Managing acquisitions: Creating value through corporate renewal. New York: The Free Press.

Hedges, L. V., and I. Olkin 1985 Statistical methods for meta-analysis. New York, NY: Academic Press.

Heide, Jan B. 1994 'Interorganizational governance in marketing channels'. Journal of Marketing 58/1: 71-85.

Heide, Jan B., and George John 1990 'Alliances in industrial purchasing: The determinants of joint action in buyer-supplier relationships'. Journal of Marketing Research 27/1: 24-36.

Heide, Jan B., and Anne S. Miner 1992 'The shadow of the future: Effects of anticipated interaction and frequency of contacts on buyer-seller cooperation'. Academy of Management Journal 35/2: 265-291.

Helper, Susan 1991 'How much has really changed between U.S. automakers and their suppliers'. Sloan Management Review 32/4: 15-28.

Huber, George P. 1991 'Organizational learning: The contributing processes and the literatures'. Organization Science 2/1: 88-115.

Hunter, J. E., and F. L. Schmidt 1990 Methods for meta-analysis: Correcting error and bias in research findings. Newbury Park, CA: Sage.

Hunter, John E., Frank L. Schmidt, and Gregg B. Jackson 1982 Meta-analysis: Cumulating research findings across studies. Beverly Hills, CA: Sage.

Joskow, P. 1985 'Vertical integration and long-term contracts: The case of coal-burning electric-generating plants'. Journal of Law, Economics and Organization l(Spring): 33-80.

Joskow, Paul L. 1987 'Contract duration and relationship-specific investments: Empirical evidence from coal markets'. American Economic Review 77/1: 168-185.

Kleinknecht, Alfred, and Jeroen O. N. Reijen 1992 'Why do firms cooperate on R&D? An empirical study'. Research Policy 21/4: 347-360.

Kogut, Bruce 1988 'Joint ventures: Theoretical and empirical perspectives'. Strategic Management Journal 9/2:319-332.

Kogut, Bruce, and Udo Zander 1993 'Knowledge of the firm and the evolutionary theory of the multinational corporation'. Journal of International Business Studies 24/4: 625-645.

Larson, Erik W., and David H. Gobeli 1988 'Organizing for product development projects'. Journal of Product Innovation Management 5: 180-190.

Lawrence, Paul R., and Jay W. Lorsch 1967 Organization and environments.' Managing differentiation and integration. Homewood, IL: Irwin.

Levitt, Barbara, and James March 1988 'Organizational learning'. Annual Review of Sociology 14:319-340.

Levy, David T. 1985 'The transactions cost approach to vertical integration: An empirical examination'. The Review of Economics and Statistics 67: 438-445.

Lorenzoni, Gianni, and Charles Baden Fuller 1995 'Creating a strategic center to manage a web of partners'. California Management Review 37/3: 146-162.

MacDonald, James M. 1985 'Market exchange or vertical integration: An empirical analysis'. The Review of Economics and Statistics 67: 327-331.

MacNeil, Ian R. 1978 'Contracts: Adjustments for long-term economic relations under classical, neoclassical and relational contract law'. Northwestern University, Law Review 72: 854-906.

Masten, Scott E., James W. Jr Meehan, and Edward A. Snyder 1991 'The costs of organization'. Journal of Law, Economics and Organizations 7/1: 1-26.

McMillan, John 1990 'Managing suppliers: Incentive systems in Japanese and U.S. industry'. California Management Review 32/4: 38-55.

Monteverde, K., and D. Teece 1982 'Supplier switching costs and vertical integration in the automobile industry'. Bell Journal of Economics 13 (Spring): 206-213.

Morris, Deigan, and Michael Hergert 1987 'Trends in international collaborative agreements'. Columbia Journal of World Business 22/2: 15-21.

Mosteller, F. M. and R. R. Bush 954 'Selected quantitative techniques' in Handbook of Social Psychology. G. Lindzey (ed.), 289-334. Cambridge, MA: Addison-Wesley.

Nanda, Ashish, and Peter J. Williamson 1994 'Joint ventures as transitory learning organizations'. 2nd EMOT Workshop 'The Changing Boundaries of the Firm' Como, 21-23 October.

Nonaka, Ikujiro 1994 'A dynamic theory of organizational knowledge creation'. Organization Science 3/1: 14-37.

Noordewier, Thomas G., George John, and John R. Nevin 1990 'Performance outcomes of purchasing arrangements in industrial buyer-vendor relationships'. Journal of Marketing 54/4: 80-93.

Oliver, Amalya, and Mark Ebers 1998 'Networking network studies: An integrative framing of conceptual dimensions in the study of interorganizational relationships'. Organization Studies 19/4: 549-583.

Oliver, Christine 1990 'Determinants of interorganizational relationships: Integration and future directions'. Academy of Management Review 15/2: 241-265.

Parkhe, Arvind 1993a 'Patter nationality and structure-performance relationship in strategic alliances'. Organization Science 4/2: 301-324.

Parkhe, Arvind 1993b 'Strategic alliance structuring: A game theoretic and transaction cost examination of interfirm cooperation'. Academy of Management Journal 36/4: 794-829.

Pfeffer, Jeffrey, and Phillip Nowak 1976 'Joint ventures and interorganizational interdependence'. Administrative Science Quarterly 21 (September): 398-418.

Pfeffer, Jeffrey, and Gerald R. Salancik 1978 The external control of organizations: A resource dependence perspective. New York: Harper & Row.

Piore, M., and C. Sabel 1984 The second industrial divide. New York: Basic Books.

Pisano, Gary P. 1991 'The governance of innovation: Vertical integration and collaborative arrangements in the biotechnology industry'. Research Policy 20/3: 237-249.

Porter, Michael E., and Mark B. Fuller 1987 Coalitions and global strategy. Boston, MA: Harvard Business School Press.

Rebentish, E., and M. Ferretti 1993 'A knowledge-asset based view of technology transfer'. Working Paper WP#86-93, International Center for the Research in the Management of Innovation, MIT.

Ring, Peter Smith, and Gordon P. Rands 1989 'Sensemaking, understanding, and committing: Emergent interpersonal transaction processes in the evolution of 3M's microgravity research program' in Research in the management of innovation: The Minnesota studies. A. H. Van de Ven, H. Angle and M. S. Poole (eds.), 337-366. New York: Ballinger/Harper Row.

Ring, Peter Smith, and Andrew van de Ven 1992 'Structuring cooperative relationships between organizations'. Strategic Management Journal 13: 483-498.

Roberts, Edward B. 1980 'New ventures for corporate growth'. Harvard Business Review (April): 134-142.

Roberts, Edward B., and Charles A. Berry 1985 'Entering new businesses: Selecting strategies for success'. Sloan Management Review (Spring): 3-17.

Rosenthal, Robert 1978 'Combining results of independent studies'. Psychological Bulletin 85: 185-193.

Rosenthal, Robert 1991 Meta-analytic procedures for social research: London: Sage.

Rosenthal, Robert, and D. B. Rubin 1979 'Comparing significance levels of independent studies'. Psychological Bulletin 86: 1165-1168.

Schrader, Stephan 1991 'Informal technology transfer between firms: Cooperation through information trading'. Research Policy 20: 153-170.

Schrader, Stephan 1994 'Organization of the supplier-manufacturer interface: The importance of domain specific communication'. Unpublished Manuscript Alfred P. Sloan School of Management, Massachusetts Institute of Technology.

Schrader, Stephan, William M. Riggs, and Robert P. Smith 1993 'Choice over uncertainty and ambiguity in technical problem solving'. Journal of Engineering and Technology Management 10: 73-99.

Scott, Richard W. 1990 'Technology and structure: An organizational-level perspective' in Technology and organizations. Paul Goodman and Lee S. Sproull (eds.). San Francisco, CA: Jossey-Bass.

Shan, Weijian 1991 'Environmental risks and joint venture sharing arrangements'. Journal of International Business Studies 22/4: 555-578.

Sjoberg, Per-Olof 1992 The management of inter-firm alliances. Cambridge, MA.: MIT Sloan School of Management.

Sobrero, Maurizio, and Alessandro Grandi 1997 'Multiple dimensions of networking activities: Scale, scope and structural patterns of interaction in innovation processes'. Academy of Management Conference, Boston, MA.

Sobrero, Maurizio, and Edward B. Roberts 1996 'The trade-off between efficiency and learning in inter-organizational relationships'. Working Paper WP 3896, Alfred P. Sloan School of Management, Massachusetts Institute of Technology, Cambridge, MA.

Stinchcombe, Arthur 1985 'Organizing information outside the firm: Contracts as hierarchical documents' in Information and organizations. A. Stinchcombe (ed.), 194-239. Berkley, CA: University of California Press.

Thomas, James B., and Linda Klebe Trevino 1993 'Information processing in strategic alliance building: A multiple case approach'. Journal of Management Studies 30/5: 779-814.

Thompson, James D. 1967 Organizations in action. New York: McGraw-Hill.

Tripsas, Mary, Stephan Schrader, and Maurizio Sobrero 1995 'Discouraging opportunistic behavior in R&D consortia: A new role for the government'. Research Policy 24/3: 367-390.

Tushman, L. Michael, and P. Anderson 1986 'Technological discontinuities and organizational environments'. Administrative Science Quarterly 31: 439-465.

Tversky, Amos, and Daniel Kahneman 1974 'Judgment under uncertainty: Heuristics and biases'. Science 185: 1124-1131.

Ulrich, David, and Jay Barney 1984 'Perspectives in organizations: Resource dependence, efficiency, and population'. Academy of Management Review 9/3: 471-481.

Uzzi, Brian 1996 'The sources and consequences of embeddedness for the economic performance of organizations'. American Sociological Review 61: 674-698.

van de Ven, Andrew, Gordon Walker, and J. Liston 1979 'Coordination patterns within interorganizational networks'. Human Relations 32:19-36.

van de Ven, Andrew H. and Robert Drazin 1985 'The concept of fit in contingency theory' in Research in organizational behavior. L. L. Cummings and Barry M. Staw (eds.), 333-366. Greenwich, CT: JAI Press.

von Hippel, Eric 1994 '"Sticky information" and the locus of problem solving'. Management Science 40/4: 429-439.

Walker, Gordon, and David Weber 1984 'A transaction cost approach to make-or-buy decisions'. Administrative Science Quarterly 29: 373-391.

Wellman, Barry, and Steven Wortley 1990 'Different strokes from different folkes: Community ties and social support'. American Journal of Sociology 96: 558-588.

Williamson, Oliver E. 1975 Markets and hierarchies. New York: Free Press.

Williamson, Oliver E. 1979 'Transaction cost economics: The governance of contractual relations'. Journal of Law and Economics 22: 233-262.

Williamson, Oliver E. 1983 'Credible commitment: Using hostages to support exchange'. American Economic Review 73: 519-540.

Williamson, Oliver E. 1985 The economic institutions of capitalism. New York: The Free Press.

Williamson, Oliver E. 1989 'Transaction costs economics' in Handbook of industrial organization. R. Schmalensee and R. D. Willig (eds.), 136-182. Amsterdam: Elsevier Science Publishers.

Woodward, Joan 1965 Industrial organizations: Theory and practice. London: Oxford University Press.