Business-to-Business Projects,Task Configuration,and Innovation

Innovation is inherently important to firms, as reflected in Tellis’s (2013, p. 2) observation that “perennial success belongs to those firms that innovate unrelentingly.” Consequently, understanding how innovation comes about is crucial both for academics and managers (Hauser, Tellis, and Griffin 2006). While some innovations are the result of a “flash of genius” or even accidents, as in the well-known case of the Post-it Note, ¹ most innovations have systematic and identifiable drivers.

A sizeable body of research in marketing, strategy, organization theory, and economics has documented these drivers. Historically, as noted by Baldwin and Von Hippel (2011), the focus of this research has been on a single firm and the particular characteristics that promote or impede innovation. For instance, researchers have studied organizational culture (Moorman 1995; Tellis, Prabhu, and Chandy 2009), structure (Azoulay and Lerner 2013), resources (Chandy and Tellis 1998), and demographics (Damanpour 1991). Over time, this line of research has been augmented by an interorganizational focus, including the roles of long-term relationships (Noordhoff et al. 2011), strategic alliances (Cui and O’Connor 2012; Rindfleisch and Moorman 2001), and joint ventures (Gulati 1998; Sampson 2007).

Interestingly, however, much innovation takes place outside of firms and long-term relationships, in the context of projects or so-called temporary organizations (Goodman and Goodman 1976; Hadida, Heide, and Bell 2019). A project represents a fundamental but poorly documented paradox: On the one hand, projects promote innovation by bringing together a set of specialists (Goes and Park 1997). The larger a given project's size, as reflected in its number of subcontractors (Damanpour 1991), and the greater the diversity of these subcontractors (Phelps 2010), the greater the amount of information that is brought to bear on the project (Rindfleisch and Moorman 2001), and the greater the potential for innovation.

At the same time, converting the different specialists’ inputs into innovation (Chandy et al. 2006) represents significant challenges. Kirsner (2017) notes that innovation is not limited by a lack of ideas—a greater challenge is capitalizing on them through organizational coordination. Interestingly, the very characteristics that create the potential for innovation in the first place give rise to coordination problems. Although large projects provide access to resources, they pose coordination problems due to their number of task interdependencies (Puranam and Raveendran 2013; Thompson 1967). Subcontractor diversity creates its own coordination problems due to the need to overcome differences in project members’ goals and communication codes (Gulati, Wohlgezogen, and Zhelyazkov 2012; Phelps 2010).

We argue that projects, like any organization, must be purposely organized or governed. At the same time, because of their distinct characteristics, projects require unique governance solutions. We draw on classic (e.g., March and Simon 1958; Thompson 1967) and emerging (Greenwood and Miller 2010; Joseph and Gaba 2020) work on organizational design that suggests that innovation depends crucially on a project's particular task configuration (Mintzberg 1979). Specifically, we argue that a project's general contractor's (GC’s) own task participation facilitates coordination and ultimately promotes innovation.

Our conceptual framework focuses on (1) product innovation and (2) process innovation (Chandy and Prabhu 2010), which involve significantly improved product outcomes and means of task performance, respectively (Mooi, Wathne, and Kayande 2016; Rindfleisch and Moorman 2001). Although the two may co-occur, they represent distinct forms of innovation, and they pose different challenges (Sorescu 2017).

We intend to make the following three contributions: First, we address the long-standing question of how firms should organize for innovation (Hauser, Tellis, and Griffin 2006; Lilien 2016). More specifically, we add to a nascent literature in marketing (e.g., Hadida, Heide, and Bell 2019) on organizations with defined durations.

Second, while innovation is commonly viewed as a “collective activity” (Powell et al. 2005, p. 1133), projects have been argued to represent “daunting challenges” (Greenwood and Miller 2010, p. 78). Theoretically, projects resemble a one-shot “Prisoner's Dilemma” game, whose fixed duration undermines coordinated action (Axelrod 1984). This challenge, we argue, requires a unique approach to coordination, including mechanisms that can be implemented under time compression. Task configuration, as reflected in the GC's participation in key project tasks, can play such a role.

Third, we show that a project's task configuration impacts product and process innovation differently. Indeed, we propose that participation creates a potential trade-off between the two: On the one hand, task participation produces coordination benefits that promote product innovation. At the same time, participation creates concerns about information leakage for the project's subcontractors, because a GC can misappropriate key practices and deploy them on future projects. Misappropriation concerns, we argue, discourage subcontractor engagement and ultimately impede process innovation. As such, when left unchecked, task participation has a “dark side” (Noordhoff et al. 2011). This perspective adds nuance to prior research (e.g., Damanpour 2010), which has assumed that product and process innovation have common causes.

From an empirical standpoint, we respond to calls from innovation researchers for “fine-grained data at the project level” (Azoulay and Lerner 2013, p. 576). Empirically, our Study 1 tests four hypotheses (H₁–H₄) based on a unique longitudinal data set of 429 construction projects from 2001 through 2015. As part of Study 1, we also conducted a supplementary survey across five industries to assess the generalizability of our constructs. Study 2 involves a cross-sectional survey of 230 construction subcontractors designed to test two follow-up hypotheses (H₅–H₆) regarding the key processes underlying our model, including how misappropriation concerns can be mitigated. We complement our formal hypotheses tests with a set of in-depth interviews designed to validate our key constructs. Web Appendix A illustrates our data collection process for Study 1 and Study 2.

This article is organized as follows: First, we discuss the project as an organizational form and how its unique characteristics both promote and impede innovation. Next, we present our conceptual framework, including our research hypotheses. Then, we describe the research method, the research setting, and our empirical studies. We conclude by summarizing our contributions, the study's limitations, and possible extensions.

The Project as an Organizational Form

A project is a form of temporary organization, in which a set of specialists (e.g., electricians, plumbers) collaborate to complete a task (e.g., a new building) within a predetermined time frame. Theoretically, a project represents an organizational form that does not fit into standard typologies such as “market,” “hierarchy,” and “hybrid” (Williamson 1991). To understand why, it is useful to compare a project's unique features to those of other organizational forms like permanent firms and strategic alliances on the basis of the so-called “three Ts” (Lundin and Söderholm 1995), namely “task,” “time,” and “team.” We discuss each of these next and provide concrete examples in Web Appendix B of how they manifest themselves.

First, with regard to the task, a project is typically a one-off undertaking. The Project Management Institute notes that “every project is undertaken for the first time” (Gallagher 2015). Although novel tasks may also happen within a firm or alliance, the latter two's tasks frequently involve repetition or adjustments to existing strategies. In contrast, task novelty is a defining characteristic of a project.

Second, projects are undertaken within a discrete time frame, with the beginning represented by the commitment to the focal task, and the endpoint by its completion). This contrasts with (1) ongoing firms whose members possess a common history and an expected future, and (2) partnerships that are frequently “built to last” (Bakker and Knoben 2015) and whose termination may be unrelated to task completion (Das 2006).

Finally, a project's team typically comprises a large number of specialist suppliers that are selected—frequently from scratch and in an open market—based on the characteristics of the task at hand. In contrast, a permanent organization tends to assign members from an internal labor pool (Kaufman 2013). Like a project, a strategic alliance involves purposeful partner selection, but it typically comprises only two parties, in contrast with a project, which has numerous members (Parmigiani and Rivera-Santos 2011).

Innovation as a Project Outcome

A project's capacity for innovation follows from its “task” and “team” dimensions—from the ability to bring a set of specialist suppliers’ knowledge and resources to bear on a particular task. Its “time” dimension, in contrast, represents a particular organizational constraint due to the time compression that is imposed on the project participants.

Innovation is often an explicitly sought outcome in a project context. As stated by The Project Management Institute, “A project team may not be the first to build an aircraft carrier, but they are the first to build that aircraft carrier—which must be built using this team, in that location, with those suppliers, this design, and these constraints. Project … is all about breaking new ground … this places innovation right at the heart of projects” (Gallagher 2015). At the same time, innovation is by no means an end result. For a variety of reasons, including (1) the difficulties of managing diverse subcontractors and (2) the GC's complex coordination role, a given project may or may not result in innovation.

We note that certain aspects of project performance have been studied previously, such as profits (Holloway and Parmigiani 2016) and costs (Ghazimatin, Mooi, and Heide 2021). Innovation, however, remains poorly documented. In the context of construction projects, based on our interviews (details of which we discuss in the “Method” section), the importance of innovation follows from its ability to generate savings both for contractors and buyers. Further, Laursen and Salter (2006) note how innovation goals—for instance, in the form of health, safety, and environmental standards—are often stated explicitly at a project's outset. Indeed, innovation and task completion often coincide in construction, and this represents a key source of competitive advantage for construction firms.

Conceptual Framework

As we show in Web Appendix C, a project comprises (1) a buyer, who initiates the project, (2) subcontractors, who apply their knowledge and resources to the focal tasks, and (3) a GC, who represents the buyer and coordinates the tasks of the various subcontractors. Often, the GC participates in project tasks (Knoben and Gössling 2009). A project's pattern of task allocation is described as its task configuration (Mintzberg 1979).

Baseline Predictions: Innovation Drivers and Outcomes

Our conceptual framework is shown in Figure 1. Its dependent variables are two different types of innovation, namely product and process innovation (Chandy and Prabhu 2010). Product innovation is defined as new or significantly improved product features or attributes (Rindfleisch and Moorman 2001). In the context of construction, an example would be a building with a significantly lower ecological footprint than the industry norm. Process innovation refers to a significantly improved operational process or working method (Damanpour 1991), such as a fast-track design/build scheduling methodology. We provide additional details in Web Appendix D, Panel A.

OPEN IN VIEWER

Figure 1.

Study 1 Conceptual Framework.

As shown in prior research (e.g., Cui and O’Connor 2012), innovation of any kind hinges on the availability of resources and knowledge. Our framework captures this through two particular project characteristics, namely (1) project size (the number of subcontractors involved), and (2) subcontractor diversity (the variation in the tasks performed by the different subcontractors). We discuss each of these drivers next.

Project size

As Figure 1 shows, one of our baseline predictions is that the greater the project size, as reflected in the number of subcontractors, the greater the number of product innovations. There are two reasons for this expectation: First, project size determines the amount of knowledge that is available (Rindfleisch and Moorman 2001). The larger the project, the greater the amount of available information, and the greater the potential for innovative outcomes. Second, larger projects make slack resources (i.e., resources that exceed a given project's minimum requirements) available. This allows for risk-taking and the ability to absorb failure (Sydow, Lindkvist, and DeFillippi 2004), which ultimately promotes innovation. We propose the following:

H₁: Project size is positively related to the number of (a) product innovations and (b) process innovations.

Subcontractor diversity

Project size, as we have discussed, impacts product and process innovation due to the quantity of knowledge that is generated by a large pool of members. We expect diversity to have a parallel effect, but based on a different explanatory mechanism: namely, the breadth of information and resources that are made available. The greater the diversity among a project's subcontractors, the greater the range of available skills, knowledge, and resources that are brought to bear (Edquist and Hommen 1999), and the greater the resulting number of product and process innovations (Cui and O’Connor 2012). In essence, diversity makes unique information available to a project (Rindfleisch and Moorman 2001), which allows for novel input configurations and an increased likelihood of innovative outcomes (Sampson 2007). Thus,

H₂: Subcontractor diversity is positively related to the number of (a) product innovations and (b) process innovations.

We consider H₁ and H₂ baseline predictions, based on the innovation literature's perspective on the roles of size and diversity. That said, these predictions are not entirely obvious. In fact, as we will discussed, size and diversity can give rise to various governance-related problems that undermine their innovation potential. For instance, the greater the number of suppliers, the greater the difficulty of monitoring individual performance, and the greater the potential for shirking that undermines coordination (Alchian and Demsetz 1972).

The Project Coordination Problem: Converting Innovation Drivers into Outcomes

We now add precision to our previous arguments by considering (1) whether the innovation potential that follows from size and diversity may be impeded due to coordination problems, and (2) how such problems may be solved in a way that promotes innovation.

Theoretically, a project involves a specific division of labor, namely (1) a mapping of organization-level goals into tasks and (2) an allocation of tasks to individual agents (March and Simon 1958; Puranam and Raveendran 2013). The resulting partitioning creates efficiencies (Malone et al. 1999), but it also introduces biases. One of them is a “component focus” (Heath and Staudenmayer 2000), namely a tendency by project members to focus on individual tasks to the detriment of the boundaries between them (Von Hippel 1988). In practice, this creates a coordination problem that will compromise performance unless “the collective set of interdependent tasks … is integrated” (Okhuysen and Bechky 2009, p. 463). ²

The magnitude of a given project's coordination problem depends on its particular characteristics. Consider first project size. While a large number of subcontractors increases information availability, it also increases the number of task interdependencies that must be managed (Knoben and Gössling 2009; Thompson 1967). A failure to manage them means unsynchronized tasks, likely compromises, and ultimately a failure to achieve innovation.

Subcontractor diversity creates a coordination problem of a different kind. The greater the subcontractor diversity, the higher the likelihood that the focal parties will have different goals, decision-making processes, and communication protocols (Phelps 2010). Such differences represent impediments to coordinated action, above and beyond those that follow from size and task interdependencies per se. Next, we consider how the coordination problems that follow from size and diversity may be addressed.

Task Participation as a Project Coordination Mechanism

To identify a solution to the project coordination problem, we turn to the literature on organizational design. Historically, this literature focused on coordination through permanent organizational structures (e.g., Thompson 1967), but emerging design research focuses specifically on coordination in organizational forms like projects that span organizational boundaries (Gulati, Wohlgezogen, and Zhelyazkov 2012; Joseph and Gaba 2020). This literature notes that while projects lack the support of a formal hierarchy (Greenwood and Miller 2010), hierarchical influence can be brought to bear on a project through the use of “integrators” (Oliveira and Lumineau 2017) who are uniquely qualified to assess and manage a project's interdependencies (Puranam, Raveendran, and Knudsen 2012).

A project’s GC intersects with both the buyer and the subcontractors and thus is in a unique position to serve as an integrator. In itself, however, the GC role is insufficient to ensure coordinated action. This capability resides in the GC's own participation in a range of project tasks. ³ As a design mechanism, participation has unique coordination properties because of its ability to generate and share knowledge between parties (Fang, Lee, and Yang 2015) without the rigidity of mechanisms such as plans, rules, and contracts (Thompson 1967).

Consider how a GC's task participation mitigates the coordination problems that follow from project size and diversity and thus helps realize their innovation potential. By performing key tasks internally, a GC acquires a firsthand understanding of (1) the skills needed for task execution, (2) the relevant standards for task performance, and (3) the interrelationships between individual tasks. Essentially, participation provides a GC with architectural knowledge (Sanchez and Mahoney 2013), which facilitates their ability to direct subcontractors, holding them accountable, and take corrective action to ensure that individual tasks support the larger innovation goal. ⁴

In principle, the GC's own involvement in the focal tasks generates informal authority based on expertise (Gulati, Wohlgezogen, and Zhelyazkov 2012). As such, their communications and directions will be accepted as legitimate by the subcontractors (DeFillippi and Sydow 2016). Thus, if a given project, due to its size and diversity, poses coordination challenges, the GC's own task participation ensures that the relevant project inputs are converted into product innovation with minimal friction.

The preceding discussion suggests that the expected positive effect of project size (as per H_1a) and subcontractor diversity (as per H_2a) on product innovation will be enhanced for higher levels of GC task participation. Thus,

H₃: GC task participation strengthens the effect of (a) project size and (b) subcontractor diversity on the number of product innovations.

In technical terms, H₃ expresses two positive moderation scenarios, where the effects of the project characteristics on product innovation are enhanced by the GC's task participation. Importantly, however, we expect the positive moderation effect of GC participation to be limited to product innovation. Process innovation, we argue, involves a different scenario, where GC participation actually has a negative moderation effect. Specifically, we posit that participation will weaken the effect of the two project characteristics on process innovation (as per H_1b and H_2b).

Our logic is as follows: Previous research (e.g., Slot, Wuyts, and Geyskens 2020) suggests that although task participation may improve efficiency, it may also cause transactional problems (Von Hippel 1990). Specifically, because a process innovation possesses a generalizable quality, the subcontractor's efforts and project-specific investments can be misappropriated by the GC and deployed in other projects. Indeed, a misappropriation scenario is predictable based on the GC’s ability and motivation; participation facilitates process learning (ability), which can be leveraged through subsequent redeployment (motivation). Our interviews specifically suggested that because process innovations typically result in cost and time savings, they create significant incentives for information leakage by a GC.

Once the GC adopts process innovations in one project, the innovating subcontractors may subsequently be expendable and may not benefit fully from their investments and innovation efforts. Research (e.g., Cohen, Nelson, and Walsh 2000) suggests that research and development executives have limited faith in formal safeguards like patents and often consider secrecy—purposely holding back information and efforts—to be the most effective source of protection. Our interviews also suggested that patents were “of little or no use” and that subcontractors tend to keep valuable knowledge “close to themselves” and avoid sharing it.

Thus, for process innovations, where the GC task participation poses leakage concerns, the predictable outcome is a reluctance on the part of the subcontractors to share information and engage in purposeful coordination (Susarla and Mukhopadhyay 2019). Interestingly, concerns about leakage and secrecy are less salient for product innovations. Because such innovations are tailored to a buyer's demands and become its property, they do not possess the generalizable quality that causes leakage concerns for process innovations. In summary,

H₄: GC participation weakens the effect of (a) project size and (b) subcontractor diversity on the number of process innovations.

The prediction in H₄ differs from the positive moderation scenario in H₃. Theoretically, participation by a GC has two different effects, depending on the type of innovation sought. In H₃, the unique nature of product innovation mitigates leakage concerns, and participation is expected to play a positive coordination role. In contrast, H₄ suggests that the very actions that enhance coordination for product innovation actually have a “dark side” (Noordhoff et al. 2011) that undermines process innovation.

Elaborating on The Dark Side of Task Configuration

Importantly, H₄ (and its empirical test, which involves secondary data [discussed subsequently]) can only establish whether GC participation weakens the effects of size and diversity in the context of process innovation. Next, we add precision to our arguments by considering the specific underlying processes. Our conceptual framework in Figure 2, which will be tested using survey data, explicates (1) how task participation creates leakage concerns, (2) the specific outcomes of such concerns, and (3) how these concerns can be mitigated.

OPEN IN VIEWER

Figure 2.

Study 2 Conceptual Framework.

By definition, task participation allows a GC to observe subcontractor practices and, thus, to acquire key information. Such learning, in turn, can be applied by the GC to future projects, without compensating the subcontractor(s) in question. Thus, our baseline expectation is that the greater the level of GC participation on a given project, the greater the subcontractor's leakage concern.

Our previous discussion of H₄, however, suggests that this effect is contingent on the nature of a given subcontractor's project contributions. Specifically, if the subcontractor has made process-related specific investments (Wathne et al. 2018) in the project, it increases the value of the practices that are brought to bear on it and, thus, the GC's incentive for misappropriation. From the subcontractor's standpoint, such investments represent vulnerability and, ultimately, create concern about leakage. This scenario is aptly described by Shapiro and Varian (1999, p. 3) in their description of information that is “costly to produce but cheap to reproduce.” We propose the following hypothesis:

H₅: The positive effect of GC participation on a subcontractor’s leakage concerns is amplified by the subcontractor’s process-related specific investments.

The right side of Figure 2 describes the effects of the subcontractor's leakage concern. In the absence of protection for misappropriation, a subcontractor's predictable response to leakage concern is to curtail the commitment to the focal project (Li et al. 2012)—specifically, by withholding key information (Frazier et al. 2009). Although this causes efficiency losses for the project as a whole, withholding protects the subcontractor's investments. Thus, we expect that leakage concerns in themselves will cause a subcontractor to withhold information.

This raises the question, however, of whether safeguards exist that (1) mitigate the subcontractor's concerns, and, if so, (2) promote efficient information transfer. Contracts and patents have been shown to be constrained in their ability to protect against misappropriation (Katila, Rosenberger, and Eisenhardt 2008; Mayer and Salomon 2006). We posit that this risk must be addressed in a more direct fashion—in a way that addresses the basic information problem that exists. This can be accomplished through direct monitoring, which can detect GC opportunism and thus serve as a deterrent. As a governance strategy, monitoring represents a formal communication channel that (1) signals permissible actions and (2) deters unacceptable ones like misappropriation (Murry and Heide 1998). Thus, as Figure 2 shows, we predict that the effect of leakage concern on information withholding will be mitigated in the presence of monitoring of the GC.

H₆: The positive effect of a subcontractor’s leakage concern on information withholding is weakened by monitoring of the GC.

Study 1: Method and Results

Empirical Context

Study 1 tested H₁–H₄ using data on construction projects from the Design-Build Institute of America (DBIA, https://www.dbia.org/resource-center/Pages/Project-Database.aspx), an organization that promotes design-build project delivery in the United States to its members. Our sampling frame consists of completed design-built projects that involve DBIA members. The DBIA asks its members to submit structured information, which is subsequently verified by an electronic signature certifying that “I confirm that all information provided about this project is true and accurate to the best of my knowledge.” Our data span all 429 projects that the DBIA reported on from 2001 to 2015. Typically, these projects feature a different composition of buyers, GCs, and subcontractors. However, some of them feature the same GC, enabling us to observe both between- and within-project variation. ⁵ We augmented the rich DBIA data with data from Hoovers (Company Profiles), the U.S. Census Bureau, and metrics from Google. The unit of analysis is a particular construction project and is reported on by the GC, who is in an excellent position to describe the phenomena in question. Next, we describe our measurement procedures in detail, followed by a discussion of our hypothesis testing approach.

Measurement and Descriptive Statistics

“Product innovation” is defined as new or significantly improved product features or attributes (e.g., Damanpour 1991; Rindfleisch and Moorman 2001). We multiplied the count of the number of product innovations by weight, in terms of their newness of the innovation to the market, for each innovation (Danneels and Kleinschmidt 2001). The count of product innovations was taken from the text describing the project's key achievements, from which we coded each distinct product innovation manually. Using a second independent coder, we verified that the coding was accurate, and we resolved a small number of differences through discussion. To measure product innovation newness, we used latent Dirichlet allocation (LDA) to arrive at the latent topic structure, consisting of “clusters” of words (Berger et al. 2020), which were ordered by their degree of novelty. Empirically, this involved retrieving innovation topics using LDA analysis. Similar to Jedidi et al. (2021) and Berger et al. (2020), we relied on topic interpretability and the multinomial probability of the words for each topic. This resulted in three product innovation topics: new or significantly improved (1) materials, (2) design and configuration, and (3) energy systems and infrastructure. Next, we asked our nine interviewees to rank these topics in terms of their newness, and a clear ranking pattern emerged. We assigned the least novel topic a value of 1, and the most novel one a value of 3. By multiplying the count of product innovations by their weight (1, 2, or 3), we obtained a rich measure of product innovation by simultaneously accounting for impact and breadth. For full details, see Web Appendix E.

“Process innovation” is defined conceptually as new or significantly improved operational processes or methods (Damanpour 1991). We multiplied the count of the number of process innovations by the weight, in terms of the newness of the innovation to the market, for each process innovation. We relied on the same LDA analysis and process as for product innovation to arrive at a newness measure of process innovation. We arrived at three topics of process innovation, including new or significantly improved (1) process and planning software, (2) cost and time-efficient methods, and (3) sustainable construction methods.

Project size is the count of the total number of subcontractors involved in the project. The number of subcontractors ranged from 1 to 58.

GC participation indicates the number of tasks the GC performs. There are seven generic task categories defined by the DBIA based on industry parlance, namely (1) GC, (2) team leader, (3) architecture, (4) engineering, (5) specialty consultant, (6) specialty contractor, and (7) other.

Subcontractor diversity reflects the heterogeneity of the subcontractors’ tasks. We operationalized diversity following Wuyts, Stremersch, and Dutta (2004). ⁶ We denoted the number of subcontractors performing task j in project i as n ij . We then calculated the proportion of subcontractors performing task j relative to the total number of subcontractors in project i as p ij = n ij / Σ j n ij . We squared each p ij and summed across all tasks j. To facilitate interpretation, we subtracted this from 1, thus defining subcontractor diversity as 1 − ∑ j = 1 J p ij 2 . The maximum level of diversity exists when each subcontractor performs a unique task. We note that the subcontractor tasks are a subset of the GC's, and we exclude the possibility of the subcontractor performing GC tasks.

As part of Study 1, we conducted a supplementary survey across five industries, in which we asked respondents to rate the importance of our core constructs for product and process innovation. The five industries represented project-intensive contexts, including media, advertising, engineering/construction, event management, and consulting. Most of the respondents indicated that project size and subcontractor diversity were very important to both product and process innovation. To test whether the means of our key constructs differ across these industries, we conducted an ANOVA including a pairwise Tukey comparison test. The results suggest no statistically significant differences between the importance of project size and subcontractor diversity across the five industries (p > .1).

We included an extensive set of control variables to account for potential alternative explanations of our results. We discuss these in turn:

Environmental uncertainty indicates the unpredictability of sales in the construction industry, which may impact both product and process innovation. Heeding prior work (e.g., Raassens, Wuyts, and Geyskens 2012), we use fluctuations in industry spending as a measure of uncertainty by (1) regressing total construction spending on the related years, and (2) using the standard error of the regression slope coefficient as the resulting measure.

Contracted duration is the time span (in days) between the project's contractually stipulated start and end dates. Longer-lasting projects may promote cooperation between members, which in turn may impact product and process innovation (Phelps 2010).

Buyer–project distance is the geographical distance between the buyer and the project and is measured as the number of miles between the buyer's main location and the project location (e.g., Handley and Benton 2013). Using Google Maps, we extracted the driving distance between a buyer and a project location using their respective addresses, as made available by the DBIA. We control for distance given the likelihood that geographical distance may increase coordination problems between the parties (Knoben and Gössling 2009) and ultimately reduce innovation (Tracey, Heide, and Bell 2014).

Projects involve multiple parties, some of whom may have interacted on previous projects and thus possess prior ties with each other (Eccles 1981). Prior ties can generate learning (Lane and Lubatkin 1998), which may promote product and process innovation. Prior ties may manifest themselves in different ways, and to account for this we controlled for (1) buyer–GC prior ties, a count variable indicating the number of times a buyer has done business previously with the same GC, (2) GC–subcontractor prior ties, a count variable indicating the number of times a specific GC–subcontractor pair shares a prior tie, and (3) buyer–GC–subcontractor prior ties, also a count variable indicating the number of times a specific set of buyer–GC–subcontractor shares a prior tie.

We included project category to absorb unobserved heterogeneity across different types of projects (including civil infrastructure, commercial/institutional, industrial process facility, and others). We also controlled for the GC's size, both in terms of the number of employees and annual sales. Both serve as proxies for the firm's human and financial resources, which can influence product and process innovation. We also accounted for selection and the nature of the pricing contract. Selection captures the key selection criteria used by the buyer in identifying a GC. Our specific measure indicates whether buyers in their selection emphasize (1) GCs’ capabilities and past performance (ability-based selection), or (2) project price offered (price-based selection). Pricing contract indicates whether the project was managed through a fixed or variable pricing format. A buyer's attempts to minimize costs, which may be accomplished through price-based selection and fixed pricing, will likely impact innovation. Lastly, we controlled for government and defense projects. These two are subject to (different) sets of federal and state rules plus, potentially, local rules that can impact the project and thereby, innovation.

Web Appendix F summarizes the measures and data sources used. Table 1 shows the correlations and descriptive statistics for the variable set. We note that product and process innovation, not unexpectedly, are positively correlated. However, the size of the correlation (r = .42, p < .01) is in line with previous work (e.g., Mooi, Wathne, and Kayande, 2016) and suggests that the two forms of innovation are distinct.

OPEN IN VIEWER

Table 1.

Study 1 Correlations and Descriptive Statistics.

	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18
1. Product innovation	—
2. Process innovation	.344	—
3. Project size	.167	.051	—
4. Subcontractor diversity	.106	−.030	.107	—
5. GC participation	−.063	−.023	−.170	−.219	—
6. Environmental uncertainty	.071	−.046	−.038	−.028	.024	—
7. Contracted duration	−.063	.064	−.011	.077	−.057	−.195	—
8. Buyer–project distance	−.045	−.007	−.071	−.072	.014	.056	.178	—
9. GC–buyer prior ties	−.054	−.014	−.015	−.002	−.014	.118	.053	.077	—
10. GC–subcontractor prior ties	−.012	.026	.271	.100	−.169	.007	.025	.006	.087	—
11. Buyer–GC–subcontractor prior ties	.009	−.029	.025	.060	−.039	−.026	.043	.048	.477	.247	—
12. Project category	−.110	.054	−.137	−.078	.087	−.130	.212	−.030	.004	−.008	−.043	—
13. GC total number of employees	−.007	.035	.047	−.193	.067	−.020	.017	−.015	−.021	.015	−.063	.101	—
14. GC's annual sales	.099	.191	.053	−.095	−.011	−.022	.075	−.009	.080	−.030	−.026	.202	.416	—
15. Selection	−.019	.015	.043	−.075	.177	.039	−.165	−.022	−.006	−.037	−.087	.008	.124	−.017	—
16. Pricing	−.004	−.068	.067	−.056	.145	.090	−.212	−.064	.079	−.081	−.020	−.095	.047	−.108	.281	—
17. Government	.127	.106	.101	.102	−.100	−.053	.306	−.009	−.032	.022	.021	.324	.005	.171	−.113	−.089	—
18. Defense	−.089	−.096	−.074	−.097	−.048	.001	−.011	.047	.056	−.018	.019	−.116	−.029	−.063	−.146	−.195	−.488	—
M	1.841	3.088	9.385	.508	2.491	48,138.470	664.784	184.172	.070	1.622	.308	1.469	165.628	448.378	.305	.463	.457	.210
SD	4.758	5.590	8.470	.212	.915	2,402.904	386.584	857.592	.279	3.787	1.457	.915	308.520	981.883	.461	.499	.499	.408
Minimum	.000	.000	.000	.000	1.000	44,570.560	92.000	.000	.000	.000	.000	.000	1.000	.010	.000	.000	.000	.000
Maximum	48.000	51.000	58.000	.800	6.000	51,865.810	4,856.000	7,039.000	2.000	24.000	13.000	3.000	3,000.000	5,522.310	1.000	1.000	1.000	1.000

Notes: n = 328. Correlations with an absolute value greater than .108 are significant at p < .05 (two-tailed).

Hypothesis Tests

Our data and hypotheses impose certain requirements on our analyses, namely (1) that the dependent variables (product and process innovation) can take on only count values, (2) the need to account for repeated and single observations of the same GC, (3) time-invariant regressors, (4) the possibility of unobserved heterogeneity, and (5) potential endogeneity.

We rely on the hybrid effects estimator to satisfy these requirements. The hybrid estimator parses a variable into a time-invariant mean of each regressor ( x i ¯ ) and a time-variant deviation from this mean ( x it − x i ¯ ) and relies on the random effects estimator to produce consistent estimates of both time-varying and time-invariant coefficients (Allison 2009). To test our hypotheses, we specify our hybrid effects model as follows:

y it 1 , 2 = β 0 + β 1 ( x it − x i ¯ ) + β 2 c i + β 3 x i ¯ + μ i + ϵ it .

(1)

The i subscript denotes time-invariant variables, and the t subscript denotes time-varying variables. y it 1 , 2 refers to product and process innovation, respectively. x it refers to the vector of theoretical and control variables. To implement the hybrid estimator, for each time-varying variable, we require that x i ¯ , the cluster mean of each variable and their deviations to the mean ( x it − x i ¯ ), be added as predictors. Time-invariant variables are denoted by c i . Finally, μ i denotes the combined effect of all unobserved characteristics of GCs that are constant over time (Allison 2009), and ϵ it is an individual disturbance term with a mean of zero. The parameters of interest are the deviations, with the means absorbing individual and time-invariant unobserved heterogeneity. We note that the ability of the hybrid estimator to account for time-invariant omitted variables also partially addresses concerns about endogeneity. However, this constitutes only part of our approach to address endogeneity, and we return to this point next.

Results

Table 2 contains the results of the hybrid effects models for product and process innovation. The models estimating the number of product and process innovations are significant (p < .01). Overall, the results are largely supportive of our hypotheses, with some interesting exceptions that we discuss in turn.

OPEN IN VIEWER

Table 2.

Hybrid Effects Poisson Model Estimates.

Variables	Product Innovation	Process Innovation
	B (SE)	B (SE)
Intercept	−1.352 (.746)*	1.074 (.461)**
Project size	.049 (.016)*** (H1a)	.022 (.009)** (H1b)
Subcontractor diversity	3.468 (.764)*** (H2a)	.747 (.374)** (H2b)
GC participation	−.029 (.237)	.193 (.105)**
Project size × GC participation	.081 (.026)*** (H3a)	−.013 (.013) (H4a)
Subcontractor diversity × GC participation	−3.247 (.737)*** (H3b)	−.9156 (.303)*** (H4b)
Control Variables
Environmental uncertainty	.000 (.000)	−.000 (.000)
Contracted duration	.001 (.000)	.000 (.000)
Project contracted cost	.000 (.000)***	.000 (.000)***
Buyer–project distance	.001 (.000)***	.000 (.000)***
GC–buyer prior ties	−.807 (.339)**	.048 (.255)
GC–subcontractor prior ties	−.199 (.031)***	.028 (.030)
Buyer–GC–subcontractor prior ties	.247 (.060)***	.034 (.056)
Project category
Commercial/Institution building	−2.039 (.388)***	.419 (.322)
Industrial process facility	−2.728 (.551)***	.248 (.378)
Civil infrastructure project	−3.251 (.542)***	.029 (.364)
GC total number of employees	.000 (.001)	.000 (.000)
GC's annual sales	.000 (.000)*	.000 (.000)***
Selection	−1.157 (.261)***	−.118 (.140)
Pricing	.379 (.268)	−.540 (.211)***
Government	.540 (.271)**	−.409 (.154)***
Defense	−1.477 (.404)***	−1.05 (.217)***
Wald χ2 (42)	250.910***	188.750***
Log-likelihood	−524.4623	−737.807
Observations per group
Min	1	1
Average	1.8	1.8
Max	14	14

*p < .10, **p < .05, ***p < .01.

Notes: Two-tailed tests of significance. Standard errors are reported in parentheses adjacent to coefficients. In hybrid effects models, means and deviations are estimated separately. We only show the substantively interpretable deviations for parsimony.

The results provide full support for H₁. Specifically, we obtain support for H_1a (p < .01) and H_1b (p < .05), which posited that project size is positively related to product and process innovations, respectively. We also obtain full support for H₂, which posited that subcontractor diversity is positively related to both the number of product (H_2a, p < .01) and process (H_2b, p < .05) innovations. For H₃, which suggests that GC participation strengthens the effect of (a) project size and (b) subcontractor diversity on product innovation, we find support for H_3a (p < .01) but obtain an effect opposite to our expectation for H_3b (p < .01). In Figure 3, we illustrate these contingency effects using partial derivatives. As illustrated in Panel A of Figure 3, GC participation strengthens the positive effect of project size on product innovation, whereas it weakens the relation between subcontractor diversity and product innovation (see Figure 3, Panel B).

OPEN IN VIEWER

Figure 3.

Illustrations of the Contingency Effects of Project Size and Subcontractor Diversity.

For H₄, which states that GC participation weakens the effect of project size and subcontractor diversity on process innovation, we do not find support for H_4a (p > .10), whereas H_4b is supported (p < .01). We plot the contingency effect of H_4b in Figure 3, Panel C, which shows that GC participation weakens the positive effect of subcontractor diversity on process innovation.

Turning to the control variables, we observe that neither environmental uncertainty nor contracted duration has a significant impact on product and process innovation (p > .1). Projects with higher contracted cost, however, result in more product (p < .01) and process (p < .01) innovation. Similarly, greater buyer–project distance relates positively to both product (p < .01) and process (p < .01) innovation. We also observe significant heterogeneity across project categories. Two different manifestations of prior ties, between the GC and buyer (p < .05) and between the GC and subcontractors (p < .01), have negative associations with product innovation, while prior ties between the triad of the buyer–GC–subcontractor (p < .01) correlates positively with product innovation. This shows that prior ties can help product innovation only if they have taken place between a fully matched set or “triad” of ties. No form of prior ties appears related to process innovation (p > .1).

While we found no significant effect for the GC's number of employees on either product or process innovation (p < .1), GC annual sales, a proxy for firm size, is positively related to both product (p < .10) and process (p < .01) innovations. Selection has a negative relationship with product innovation (p > .01), and has no impact on process innovation. Variable pricing correlates with process innovation (p < .01) but is unrelated to product innovation (p > .10). Lastly, we note significant heterogeneity across buyer type (government or defense).

Endogeneity

The choices with regard to project size, subcontractor diversity, and GC participation may be purposefully made to promote product or process innovation, raising the potential for endogeneity and thereby potentially biased parameter estimates (Wooldridge 2002). We rely on (1) Park and Gupta’s (2012) Gaussian copula estimator and (2) Lewbel’s (2012) heteroskedasticity-based identification estimator to assess potential endogeneity threats. The Gaussian copula approach proposed by Park and Gupta is a semiparametric method that models the joint distribution of the error term and the endogenous regressors using a copula method and makes inferences on model parameters by maximizing the resulting likelihood function. We reestimated our aforementioned model using this approach, and the results of this method (untabulated) show consistent estimates of the coefficients of interest, both in direction and level of significance, with those found in the hybrid effects model. We also estimated our model using Lewbel’s heteroskedasticity-based identification, which uses higher moments, in the form of heteroskedasticity to construct synthetic instruments. Lewbel's estimator provides results (untabulated) that are consistent with those found in the hybrid effects models (both in direction and level of significance). Both estimation approaches (1) suggest that endogeneity is not a likely threat to making correct inferences in our models and (2) demonstrate a good degree of robustness.

Study 2: Method and Results

In Study 2, we test H₅–H₆ regarding the “dark side” of task participation, specifically (1) the effect of GC participation on subcontractor leakage concern and (2) the effect of leakage concern on subcontractor information withholding.

Empirical Context

We conducted a cross-sectional survey of 230 subcontractors in the U.S. construction industry via the Qualtrics platform. To avoid self-selection, Qualtrics applied prescreening to a set of randomly selected subcontractors on certain criteria and we specifically requested the respondents to keep a construction project in mind (1) that was fully completed in the last two years and (2) for which their company acted as a subcontractor (not a GC). Our initial sampling frame consisted of 230 subcontractors. We eliminated 17 informants who reported limited project involvement.

Measures

All our focal constructs are measured using five-point multi-item scales. We adapted our measures from prior research when possible and developed new scales for the others. We conducted a pilot test with 30 U.S. construction subcontractors to examine the face validity of our measures and to assess the relevance and wording of our items. No problems were revealed. Web Appendix G presents the final scales used, and Table 3 shows the correlations and descriptive statistics for the variable set. We discuss each of the variables next.

OPEN IN VIEWER

Table 3.

Study 2 Correlations and Descriptive Statistics.

	1	2	3	4	5	6	7	8	9	10	11
1. Subcontractor withholding of information	—
2. GC participation	.255	—
3. Subcontractor process-related specific investment	.324	.606	—
4. Subcontractor information leakage concerns	.107	.478	.354	—
5. Subcontractor monitoring of the GC	.134	.456	.580	.427	—
6. GC monitoring of the subcontractor	.223	.519	.653	.409	.791	—
7. Uncertainty	.490	.493	.589	.282	.379	.496	—
8. Project duration	.063	.015	.088	−.110	.148	.075	.046	—
9. Project category	.011	.101	.105	−.059	.030	−.020	.033	.157	—
10. Subcontractor–GC prior ties	.213	.114	.053	.165	.012	.091	.099	.014	.012	—
11. Number of subcontractor's employees	.145	.351	.421	.247	.435	.396	.289	.050	.105	.130	—
M	.0246	.046	.060	.056	.044	.0279	.016	2.667	1.868	3.831	4.221
SD	.928	.595	1.06	.467	.549	.536	.653	1.258	.972	1.303	1.025
Minimum	−1.870	−1.751	−3.275	−2.120	−2.621	−1.788	−2.461	1	1	1	1
Maximum	1.459	.936	1.241	.603	.528	.543	.854	5	4	6	5

Notes: n = 213. Correlations with an absolute value greater than .134 are significant at p < .05 (two-tailed).

GC participation indicates the extent to which the GC was involved in certain tasks during project execution. The specific items were adapted from the ones used by Harmancioglu, Wuyts, and Ozturan (2021). Subcontractor information leakage concern reflects the subcontractor's initial expectation about the usefulness of its proprietary knowledge and skills to the GC. Practically, the scale captures the ex ante potential for GC misappropriation. Again, we adapted the items from Harmancioglu, Wuyts, and Ozturan (2021). Subcontractor withholding of information captures the extent to which the subcontractor purposely curtailed the sharing of new information with the GC during project execution. The items were adapted from Heide and John (1992). Subcontractor's specific investments capture the process-related investments made by the subcontractor in the project (Anderson and Weitz 1992; Wathne et al. 2018). Subcontractor monitoring of the GC indicates the extent of monitoring undertaken by the subcontractor of the GC during project execution (Heide, Wathne, and Rokkan 2007).

We included a set of control variables in our analysis. Environmental uncertainty is defined as unanticipated changes in circumstances surrounding an exchange (Noordewier, John, and Nevin 1990, p. 82), and we measured this construct based on the extent to which the market's competitive and technological conditions were difficult to predict (Noordewier, John, and Nevin 1990). Project duration, a single-item variable, reflects the number of months/years that the project had taken to complete. Shorter projects involve greater time compression, ceteris paribus, which has coordination implications (Ghazimatin, Mooi, and Heide 2021). We also account for project type to absorb unobserved heterogeneity across different types of projects (including civil infrastructure, commercial/institutional building, industrial process facility, and others). In addition, we control for GC–subcontractor prior ties which is a count variable indicating the number of times a subcontractor has collaborated previously with the same GC. We also controlled for the subcontractor's size in terms of the number of subcontractor's employees working full-time. Finally, we controlled for the possibility that the GC may monitor the subcontractor. The GC monitoring of the subcontractor scale uses similar items to the previous monitoring scale.

Construct Validity

We evaluated all our reflective multi-item scales through a series of confirmatory factor analysis models. A measurement model containing all the multi-item scales was estimated and demonstrated acceptable model fit (χ²(d.f. = 50) = 103.161, p < .001, root mean square error of approximation [RMSEA] = .071, square root mean residual [SRMR] = .048, comparative fit index [CFI] = .960, Tucker–Lewis index [TLI] = .948). Consistent with Bagozzi and Yi’s (1988) recommendations, all the alpha coefficients exceeded the threshold of .60. Using the procedure suggested by Fornell and Larcker (1981), we calculated the composite reliabilities and average variance extracted for all reflective multi-item constructs, and all of them met or exceeded the thresholds recommended by Fornell and Larcker (1981).

Further, we conducted the required analysis for formative measurements. Based on Diamantopoulos and Winklhofer’s (2001) recommendations, we first assessed the scope of our formative scales, for which we relied on construct definitions drawn from the existing literature (e.g., Anderson and Weitz 1992; Harmancioglu, Wuyts, and Ozturan 2021; Wathne et al. 2018). Web Appendix G lists the final scale items. Second, given that high multicollinearity makes the assessment of the indicator validity difficult (Diamantopoulos and Winklhofer 2001), we assessed if each indicator uniquely contributes to the relevant construct by regressing each item as the dependent variable using the remaining items as independent variables. The results suggest no evidence of collinearity as the maximum variance inflation factor for our four formative scales, monitoring of the GC, monitoring of the subcontractor, subcontractor process-related investments, and subcontractor information leakage concerns was 2.87, 2.09, 2.15, and 1.35, respectively. These levels are far lower than the conservative threshold of 5 (Mooi, Sarstedt, and Mooi-Reçi 2018) indicating that indicator multicollinearity is not a concern for our formative scales and that each indicator provides a unique measurement contribution.

Hypothesis Tests

To test our hypotheses, we must account for (1) the simultaneous estimation of multiple equations, (2) potential heteroskedasticity, and (3) potential endogeneity. Regarding the former two requirements, we rely on a seemingly unrelated estimation (SUEST), which combines heteroskedasticity-robust estimates from multiple regression equations (Wathne et al. 2018).

Results

Table 4 presents the results and robust standard errors of our SUEST estimates. In support of H₅, we find that the interaction between GC participation and subcontractor process-related specific investment has a positive and significant effect on subcontractor leakage concerns (p < .01). As hypothesized, we also obtained a negative and significant (p < .05) interaction effect between subcontractor leakage concerns and subcontractor monitoring of the GC on subcontractor withholding of information (H₆). ⁷ To understand the nature of these interaction effects (H₅ and H₆), we conducted a simple slope analysis (Aiken and West 1991). Panel As and B in Figure 4 represent the main effects across the full range of the moderator variables. For this purpose, we tested the simple slopes at their minimum, low (−1 SD), mean, high (+1 SD), and maximum values of the moderators.

OPEN IN VIEWER

Figure 4.

Illustrations of Simple Slope Analyses.

OPEN IN VIEWER

Table 4.

SUEST Estimates of Subcontractor Withholding Information Model.

Variables	Subcontractor Information Leakage Concerns	Subcontractor Withholding of Information
	β (Robust SE)	β (Robust SE)
GC participation	.278 (.071)***
Subcontractor process-related specific investment	.082 (.049)*
GC participation × Subcontractor process-related specific investment	(H5).146 (.044)***
Subcontractor information leakage concerns		−.079 (.114)
Subcontractor monitoring of the GC		−.159 (.148)
Subcontractor information leakage concerns × Subcontractor monitoring of the GC		(H6) −.750 (.317)**
GC monitoring of the subcontractor		.088 (.153)
Subcontractor information leakage concerns × GC monitoring of the subcontractor		.421 (.284)
Control variables
Uncertainty	.006 (.056)	.712 (.077)***
Project duration	−.036 (.022)	.025 (.045)
Project type:
Civil infrastructure project	−.109 (.070)	.192 (.15s6)
Commercial/institution building	−.073 (.074)	−.059(.138)
Industrial process facility	−.254 (.190)	.118 (.136)
Subcontractor–GC prior ties	.037 (.022)*	.136 (.041)***
Number of subcontractor's employees	.034 (.040)	−.019 (.056)
Intercept	−.151 (.199)***	−.474 (.304)***

*p < .10, **p < .05, ***p < .01.

Notes: Two-tailed testes of significance if hypothesized. No. observations = 213. Robust standard errors are reported in parentheses.

Figure 4, Panel A, presents the moderating effect of the subcontractor's process-related investment on the relationship between GC participation and subcontractor information leakage concerns. As the plot indicates, the effect of GC participation on subcontractor information leakage concerns is positive and significant at −1 SD (p < .01), the mean (p < .01), +1 SD (p < .01), and at the maximum (p < .01) observed value of subcontractor's process-related investment. However, this effect disappears where the subcontractor's process-related investment is absent (p > .10).

Figure 4, Panel B, presents the moderating effect of subcontractor monitoring of the GC on the relationship between subcontractor leakage concerns and subcontractor withholding of information. As the plot indicates, subcontractor monitoring has a positive but weakening effect at the minimum level (p < .05). However, from +1 SD (p < .05) to the highest (p < .05) observed values of subcontractor monitoring, this effect becomes negative and stronger.

As our particular setup involves a double moderated mediation model, we relied on Hayes (2022) and examined the conditional indirect effect of GC participation on the subcontractor's withholding of information via leakage concerns, across different levels of our two moderators. Testing this involves estimating the conditional indirect effect where we set the two moderators (subcontractor process-related investments and monitoring of the GC) at different levels. Because this test does not require normality assumptions, we relied on bootstrapping. The results show that at low levels of the moderators (−1 SD) no mediation is observed (p > .10). Yet, when both moderators are at their mean levels or higher, we start to observe significant mediation, in that leakage concerns mediate the effect of GC participation on information withholding (p < .05). Importantly, this test directly supports the inclusion of the two moderators. In a scenario in which we test for simple mediation (GC participation → subcontractor information leakage → subcontractor withholding of information) but do not account for the two moderators, a bootstrapped Sobel test, as advocated by Zhao, Lynch, and Chen (2010), suggests that no mediation is present (p > .10). Thus, we conclude that support for mediation is obtained only when we include appropriate moderators in the model.

Turning to the control variables, we found that while environmental uncertainty has no significant effect on subcontractor information leakage concerns (p > .1), it does have a positive and significant effect on subcontractor withholding of information (p < 01), implying that as environmental uncertainty, and thus adaptation, problems increase (Rindfleisch and Heide 1997), firms are more likely to withhold information as a safeguard. We found no significant effect of project duration (p > .1) or project type (p > .1) on either subcontractor information leakage concerns or withholding of information. We also found that subcontractor–GC prior ties has a positive effect on subcontractor information leakage concerns (p < .1) and subcontractor withholding of information (p < .01), which may be due to elevated opportunism concerns as parties start to understand each other’s weaknesses (Wuyts and Geyskens 2005). The number of subcontractor's employees has no significant effect on subcontractor information leakage concerns or on subcontractor withholding of information (p > .1).

Endogeneity

To address the potential for endogeneity, we used two instrument-free methods, namely (1) Park and Gupta’s (2012) Gaussian copula estimator and (2) Lewbel’s (2012) heteroskedasticity-based identification estimator, to assess the robustness of our results to endogeneity concerns. Both methods provided consistent results, both in direction and level of significance, with those found in the original ordinary least squares models, showing that endogeneity is not a likely threat.

Discussion

Most innovations are the result of “collective activity” (Powell et al. 2005, p. 1133). Important questions exist, however, about the specific context for such activity, and about the roles that different organizational forms play. Previous research in marketing has generated important insights into the innovation roles of permanent firms (Chandy and Tellis 1998) and long-term relationships (Rindfleisch and Moorman 2001), but we know relatively little about how innovation occurs within temporary organizations or projects. This gap is significant, particularly when considered against the backdrop of our interviews, which identified innovation as a crucial project outcome (Web Appendix D).

We proposed a conceptual framework with project size and subcontractor diversity as the drivers. Our results showed that size and diversity promote both product and process innovation. We also showed that the ability of project size to bring about product innovation was facilitated by the GC's task participation, due, presumably, to the GC's ability to gain architectural knowledge through participation, which facilitates coordination and promotes innovation. These findings converged with the insights from our interviews, which revealed that GC participation both (1) promotes learning about client needs and (2) helps “correct the way” subcontractors perform project tasks.

At the same time, GC participation did not play such a role relative to subcontractor diversity. In fact, GC task participation significantly weakened the effect of subcontractor diversity on product innovation. This may be due to the inherent complexity of coordinating the activities of diverse project participants. Consequently, rather than providing net coordination benefits, participation by a GC may add friction to an already complex project organization. Considered in combination, these findings suggest that the coordination effect of participation on product innovation is limited to the particular problem caused by project size.

Further, we found that GC task participation did not weaken the effect of project size on process innovation. However, GC participation did weaken the positive effect of subcontractor diversity on process innovation. Thus, in the context of process innovation, it appears that subcontractors are reluctant to share information and engage in purposeful coordination. Our interviews suggested that such reluctance is due to a lack of formal protection for process innovations.

Finally, our survey results showed that GC participation gave rise to leakage concerns when a subcontractor had made project-specific investments. This suggests that participation indeed has a “dark side.” Importantly, such concerns have negative efficiency implications by virtue of causing subcontractors to withhold information. This tendency, however, was mitigated by monitoring.

Theoretical Implications

We identify four different theoretical contributions of our findings. First, we advanced the literature on projects and temporary organizations. Somewhat surprisingly, despite the prevalence of projects in many different marketing contexts, including new product development, advertising, and service delivery, they have, with some recent exceptions (Hadida, Heide, and Bell 2019), received limited attention in the marketing literature. Indeed, reviews (Lundin et al. 2015) show that our current knowledge about projects is based almost exclusively on research in management and engineering; marketing's voice has been silent. Our study represents an attempt to rectify this and to engage marketing in the academic dialogue about projects.

Second, we generated insight into the mechanisms through which projects produce performance outcomes. Prior research in marketing (e.g., Tellis 2013) has pointed to the general role that organizational variables play in driving innovation. Much of the extant work (e.g., Chandy and Tellis 1998) has focused on organizational culture and its ability to promote (as well as constrain) innovation. It is noteworthy, however, that culture is unlikely to play a major role in projects given that (1) the relevant “team” (Lundin and Söderholm 1995) is often assembled from scratch and (2) its members interact for a short time. Indeed, these characteristics have led some organization theorists (e.g., Meyerson, Weick, and Kramer 1996) to hypothesize that projects as organizational forms are inherently unstable. Others (e.g., Bechky 2006) have challenged this view and argued that projects do not lack structure and can draw on unique sources of support. These different viewpoints suggest the need for a more fine-grained theory of coordination in a project context. Our current study represents a first step toward that goal. ⁸

Third, we provided new insights into different types of innovation. Although prior research in marketing has made a distinction between product and process innovation (Chandy and Prabhu 2010; Hauser, Tellis, and Griffin 2006), empirical evidence, with some notable exceptions (Mooi, Wathne, and Kayande 2016; Rindfleisch and Moorman 2001), is scarce. Our findings suggest that product and process innovation are different phenomena. Perhaps most importantly, task participation, which was shown to provide coordination benefits in the context of product innovation, actually impeded process innovation. This discrepancy points to an important “dark side” of coordination in a project context.

Fourth, from a governance standpoint, we showed the importance of direct or project-specific efforts in the form of monitoring. Indeed, such efforts were capable of mitigating the negative efficiency implications that leakage concerns create.

Managerial Implications

Innovation is often the result of “purposeful activity” (Drucker 1985, p. 34). However, the relevant activities often involve practical challenges. To wit, Boeing's launch of the new 737 Max airplane involved hundreds of suppliers who produced 367,000 parts, from engines to exit signs. Indeed, managing such a large and diverse supplier pool and ensuring that the end product met its innovation goals posed significant coordination problems (Macmillan and Gregg 2019).

Our findings speak to these challenges, but they also suggest possible solutions tied to the role of the GC. Specifically, our finding that GC participation plays a selective coordination role (in the context of product, but not process, innovation) suggests that buyers of complex services can make deliberate selection decisions based on the GC's project role. To a buyer, such differences in task configuration matter, because they influence a project's organizational properties and ultimately its innovation outcomes. As a consequence, for a buyer the manner in which a project's tasks are allocated, either through negotiation, contracting, or tendering, represents a key decision criterion.

Our findings suggest that decisions on task configuration must be made with specific reference to (1) project characteristics (size, diversity) and (2) desired outcomes (product or process innovation). On this note, our empirical tests uncovered a potential “dark side” to task participation by the GC. Specifically, for process innovations such participation may cause information leakage concerns since a subcontractor's learning can be exploited by the GC through subsequent redeployment on other projects. Interestingly, the World Intellectual Property Organization (Follador 2019) points out that the use of “confidentiality” is a critical protection mechanism, but one that is largely unavailable in the context of process innovations where information sharing with other parties is inherent. Indeed, based on a recent survey, Kroll (2019), a global provider of risk solutions, concludes that information leakage and IP theft were key concerns for 41% and 43%, respectively, of the firms in their sample.

Unexpectedly, we also observed a “dark side” of coordination in the domain of product innovation. Specifically, our results showed that managing subcontractor diversity through GC participation was ineffective, presumably because of the incremental organizational complexity that participation introduces.

To illustrate the practical importance of our analyses, we rely on counterfactuals to calculate “what if” scenarios. Consider the U.S. Marine Corp's project to renovate Camp Pendleton. A team of 12 subcontractors was led by a GC who did not participate in any of the key tasks. The reported outcomes reveal one process innovation but no product innovations. Had the task been configured differently, as evidenced by counterfactual analysis, our model suggests that full participation by the GC would have resulted in one product innovation (but no process innovations). Essentially, accounting for participation flips the pattern of product and process innovation between the observed pattern and the counterfactual. This has practical implications, because process innovation (as per the logic of H₄), unlike product innovation, has a generalizable and redeployable quality. As such, the shift that we observe is material, in that a product innovation falls to the buyer, whereas a process innovation can fall to any party that can observe and learn. This single counterfactual demonstrates the practical importance of task configuration.

Limitations

Our findings must be viewed in light of certain limitations. Although we attempted to assess industry differences with regard to our drivers, other types of projects may pose different organizational challenges and require different solutions. For example, Kaggle, a community that hosts data science projects, relies on (1) a clear-cut ex ante specification of desired innovative outcomes, (2) a compensation structure that is based almost solely on achievement, and (3) detailed timelines. As another example, decentralized blockchain platforms can enable hundreds of firms to collaborate safely on several (narrow) dimensions, where their proprietary information and innovation can be safeguarded by registering “evidence of creatorship and provenance authentication” (Clark and McKenzie 2018). This may lessen, but not eliminate, concerns associated with participation. We also note that our database does not speak to smaller projects that may be organized in a more ad hoc fashion, nor does it include projects without a GC. Yet, in some instances, such as when projects are fully embedded in an organization (Hadida, Heide, and Bell 2019), there is no separate GC.

Future Research

We close by suggesting four avenues for future research. One intriguing candidate is the participation construct, which has numerous manifestations in marketing contexts (e.g., Heide and John 1990; Sa Vinhas and Heide 2015). Future research can be directed toward further unpacking this construct, including its origins, effects on innovation and other performance metrics, and the resulting governance responses. Relatedly, future research may explore additional underlying processes. Our current (normative) focus involved a nuanced (double) moderated mediation mechanism and the joint effect of leakage concerns and monitoring on performance in the form of information withholding. Future research can delve deeper into individual paths that (descriptively) show firms’ actions or practices. For instance, future research may examine how participation (directly) impacts ongoing investments, and whether leakage concerns motivate firms to engage in monitoring.

Second, there are instances where project innovation may actually create undesired outcomes. For example, the construction of the Sydney Opera house, while highly innovative, also led to cost overruns, reputational damage to the architect, and significant political issues (Chua 2013). The manner in which multiple project outcomes relate to each other seems like an important future research topic.

Finally, participation may create effects that we cannot study in our context. For example, GCs may provide subcontractors with training or skills that allow them to contribute more innovatively to the project. In some industries, such as aerospace, a local GC is often used. For example, Boeing and Tata coproduced helicopters with many local subcontractors, generating coordination challenges both (1) between one subcontractor and two GCs and (2) between the GCs themselves. In general, these types of questions point to the need for more fine-grained accounts of project governance.

Supplemental Material

sj-pdf-1-mrj-10.1177_00222437221105828 - Supplemental material for Business-to-Business Projects, Task Configuration, and Innovation

Supplemental material, sj-pdf-1-mrj-10.1177_00222437221105828 for Business-to-Business Projects, Task Configuration, and Innovation by Elham Ghazimatin, Erik A. Mooi and Jan B. Heide in Journal of Marketing Research