Innovation alliance governance structure

Innovation alliances represent formal agreements between organizations to systematically manage knowledge flows across organizational boundaries, establish commercial networks, and jointly create new value (Sampson, Reference Sampson2007). An innovation alliance is formed after parties involved consider expected benefits and potential costs, while different types of alliance governance structures influence their learning efficiency and cooperation costs (Kale & Singh, Reference Kale and Singh2009; Michelino, Cammarano, Lamberti, & Caputo, Reference Michelino, Cammarano, Lamberti and Caputo2015). The equity versus non-equity constitutes the prevalent classification of alliance structures (Choi & Contractor, Reference Choi and Contractor2016; Vrande, Vanhaverbeke, & Duysters, Reference Vrande, Vanhaverbeke and Duysters2009). An equity innovation alliance refers to a cooperative innovation mode that involves equity exchanges in the forms of joint ventures or mutual shareholding with external partners. A non-equity innovation alliance involves contractual agreements that cover different types of partnerships with no equity sharing (Niesten & Jolink, Reference Niesten and Jolink2017; Yang et al., Reference Yang, Wang, Lyu and Yang2023). As we detail subsequently, both governance structures differ in terms of knowledge transfer efficiency, transaction costs, and setup costs.

According to knowledge-based view, knowledge management is a key issue that addresses the implicit embeddedness of external knowledge and determines the effectiveness of the innovation alliance (Oxley & Wada, Reference Oxley and Wada2009). As an equity innovation alliance pushes participants to share their resources and risks, strengthen their mutual trust levels, and make decisions on a long-term basis (Michelino et al., Reference Michelino, Cammarano, Lamberti and Caputo2015). The equity alliance structure represents the most effective instrument to facilitate knowledge transfer and joint creation of new knowledge (Phene & Tallman, Reference Phene and Tallman2012; Williamson, Reference Williamson2008). By comparison, non-equity governance structure, constrained by detailed contractual agreements, tends to decrease the efficiency of knowledge transfer in the collaboration process (Bagherzadeh, Gurca, & Brunswicker, Reference Bagherzadeh, Gurca and Brunswicker2019).

Scholars drawing on transaction cost theory identify that the alliance failure may arise from opportunistic behavior of members, such as knowledge misappropriation, value exploitation, or distortion of information (Das & Rahman, Reference Das and Rahman2010; Majocchi, Mayrhofer, & Camps, Reference Majocchi, Mayrhofer and Camps2013). Equity alliance governance structure, equipped with formal monitor and control mechanisms to govern alliance partners, can curb opportunism and reduce transaction costs (Michelino et al., Reference Michelino, Cammarano, Lamberti and Caputo2015; Phene & Tallman, Reference Phene and Tallman2012; Reuer et al., Reference Reuer, Arino, Poppo and Zenger2016). Conversely, non-equity innovation alliances vulnerable to moral hazards and opportunistic behaviors are usually associated with greater degrees of transaction costs (Niesten & Jolink, Reference Niesten and Jolink2017). Further, equity innovation alliances require large amounts of commitment resources including capital, personnel, and time inputs, leading to higher levels of setup costs (Howell, Reference Howell2018; Phene & Tallman, Reference Phene and Tallman2012). By contrast, non-equity innovation alliances in which the duties and interests of each participant are clearly specified in the agreement demonstrate higher cooperative flexibility and lower setup costs (Bagherzadeh, Gurca, & Brunswicker, Reference Bagherzadeh, Gurca and Brunswicker2019; Colombo, Reference Colombo2003).

The governance structures of equity and non-equity innovation alliances manifest strengths and weaknesses in the process of cooperative innovation, as shown in the following Table 1.

Table 1.

The features of equity and non-equity innovation alliances

Performance feedback

The PFT explains why and how overperformance or underperformance messages guide firms’ strategic decisions (Baum, Rowley, Shipilov, & Chuang, Reference Baum, Rowley, Shipilov and Chuang2005; Yu, Minniti, & Nason, Reference Yu, Minniti and Nason2019). Ample empirical evidence shows that performance feedback engenders a wide range of organizational actions, such as strategic change or organizational reconfiguration (e.g., Baum & Dahlin, Reference Baum and Dahlin2007; Cheng, Xie, Fang, & Mei, Reference Cheng, Xie, Fang and Mei2022). Scholars suggested that firms rely on two different types of aspirations to calibrate performance feedback: social and historical aspirations (Kotiloglu, Chen, & Lechler, Reference Kotiloglu, Chen and Lechler2021). Social aspiration is an outcome of performance of peer firms, while historical aspiration arises from previous performance of the firm itself (Cyert & March, Reference Cyert and March1963). Both types of performance feedback reflect the changes in social status of firms and the variation in organizational development trajectory respectively (Kim, Finkelstein, & Haleblian, Reference Kim, Finkelstein and Haleblian2015). Although some previous studies revealed that they seemingly exert homogeneous effects on firm’s response behaviors (Kotiloglu, Chen, & Lechler, Reference Kotiloglu, Chen and Lechler2021), recent research suggests that their impacts are different or depend on other contingencies (e.g., Dong, Reference Dong2021; Kacperczyk, Beckman, & Moliterno, Reference Kacperczyk, Beckman and Moliterno2015). Particularly, the effects of social performance feedback may not be attenuated in a short period, which in turn will encourage firms to search for long-term solutions (e.g., innovation alliance) (Tyler & Caner, Reference Tyler and Caner2016; Ye, Yu, & Nason, Reference Ye, Yu and Nason2021). Thus, in this study, we focus on social performance feedback.

Negative performance feedback and innovation alliance governance structure

According to the PFT, performance below aspirations may motivate firms to initiate problemistic search to access new knowledge and practices to solve problems and improve current performance (Furlotti & Soda, Reference Furlotti and Soda2018; Lv et al., Reference Lv, Zhu, Chen and Lan2021; Sampson, Reference Sampson2004). As problemistic search often follows a search pattern from local to nonlocal sources, firms will turn to nonlocal search when local search of old experience and internal knowledge fails to address the performance gap (Baum & Dahlin, Reference Baum and Dahlin2007; Su et al., Reference Su, Yu, Chen and Hou2023). Alliance, serving as a bridge to distant contexts, enables firms to overcome constraints of local search and to acquire new knowledge (Rosenkopf & Almeida, Reference Rosenkopf and Almeida2003). Indeed, negative performance feedback is found to increase the alliance formation intention and the number of R&D alliances (Lohrke, Kreiser, & Weaver, Reference Lohrke, Kreiser and Weaver2006; Tyler & Caner, Reference Tyler and Caner2016). Literature in partner selection also shows that firms with similar levels of research capabilities are prone to establish partnership because those firms with superior research capabilities may be not available to them (Mindruta, Moeen, & Agarwal, Reference Mindruta, Moeen and Agarwal2016).

When choosing an appropriate governance structure for innovation alliances, underperforming firms usually take into account both two aspects: the knowledge to be obtained from alliance partners and the alliance costs to be borne during joint development activities (Niesten & Jolink, Reference Niesten and Jolink2017; Tyler & Caner, Reference Tyler and Caner2016). We propose that underperforming firms increase propensity toward selection of non-equity alliance for several reasons. First, performance pressure induced by negative performance feedback encourages firms to seek improvement as soon as possible (Baum & Dahlin, Reference Baum and Dahlin2007; Lv, Zhu, Chen, & Lan, Reference Lv, Zhu, Chen and Lan2021). Since a non-equity alliance displays lower setup costs and higher flexibility than an equity one, it becomes a preferable choice that helps underperforming firms engage in innovation alliances promptly (Colombo, Reference Colombo2003; Michelino et al., Reference Michelino, Cammarano, Lamberti and Caputo2015). Second, when firms cannot achieve target performance levels, they tend to undertake risky strategies to seek major improvements (Alexy, Bascavusoglu-Moreau, & Salter, Reference Alexy, Bascavusoglu-Moreau and Salter2016; Baum et al., Reference Baum, Rowley, Shipilov and Chuang2005; Jordan & Audia, Reference Jordan and Audia2012). Despite perceived high degrees of risk caused by knowledge leakage and transaction costs in non-equity alliances, firms with performance below aspirations may still be inclined to access new knowledge through non-equity alliances (Majchrzak, Jarvenpaa, & Bagherzadeh, Reference Majchrzak, Jarvenpaa and Bagherzadeh2015). Finally, underperforming firms intend to solve specific technical problems through building non-equity alliances. Although the knowledge transfer efficiency of non-equity alliance is lower, this limitation can be offset by articulating specific knowledge items in contractual agreements (Choi & Contractor, 2019). Building on these insights, we propose the hypothesis:

Positive performance feedback and innovation alliance governance structure

The PFT holds that overfulfillment of social aspirations makes firms released from pressure of solving immediate problems, they tend to implement slack search to attain utilization of existing resource as well as creation of new value (Argote & Greve, Reference Argote and Greve2007; Martinez-Noya & Garcia-Canal, Reference Martinez-Noya and Garcia-Canal2021). Decision-makers with satisfactory performance outcomes may note and learn from others’ failures to prevent future failing and keep their success (Kc, Staats, & Gino, Reference Kc, Staats and Gino2013; Kim & Rhee, Reference Kim and Rhee2017). Indeed, it is evident that high performing firms are willing to form alliances with diverse partners to explore new opportunities beyond their existing networks (Martinez-Noya & Garcia-Canal, Reference Martinez-Noya and Garcia-Canal2021). Therefore, performance exceeding social aspiration enhances the confidence of managers when exploiting firm capabilities and new projects to create a favorable bargaining power in the cooperations (Baum et al., Reference Baum, Rowley, Shipilov and Chuang2005; Xu, Zhou, & Du, Reference Xu, Zhou and Du2019).

We argue that performance above social aspirations will lead firms to engage in equity alliances for the following reasons. First, to keep superior performance in their industry, firms are willing to continue previous businesses and practices and explore additional similar projects in their knowledge domain (Alexy, Bascavusoglu-Moreau, & Salter, Reference Alexy, Bascavusoglu-Moreau and Salter2016). An equity innovation alliance with close relationships and high-levels of trust demonstrates great knowledge transfer efficiency between firms, which enables them to achieve value creation for joint innovations (Choi & Contractor, 2019; Sampson, Reference Sampson2004). Second, performance above the aspirations increases managerial confidence and resource slacks in taking projects that are costly but related to the long-run viability of the firm (Desai, Reference Desai2013; Xu, Zhou, & Du, Reference Xu, Zhou and Du2019). Although equity alliances have higher setup costs, the excess returns of overperforming firms reduce the cost consideration when they choose alliances governance structure. Finally, the better a firm performs above social aspirations, the more capable the firm will be to create higher bargaining powers in alliance negotiations (Bosse & Alvarez, Reference Bosse and Alvarez2010; Kim & Rhee, Reference Kim and Rhee2017). An equity innovation alliance imbued with lower transaction and opportunistic costs allows outperforming firms to better utilize their advantageous situations in building alliances through which they keep the competitive edge. Thus, we hypothesize that:

The moderating role of TMT tenure

The PFT predicts that decision-makers are boundedly rational agents and their background characteristics may influence their perception and interpretation of performance feedback cues and subsequent responses (Jordan & Audia, Reference Jordan and Audia2012). Indeed, the literature of organizational change has addressed how performance feedback and decision-makers’ characteristics jointly determine the strategic choices of firms (e.g., Choi, Rhee, & Kim, Reference Choi, Rhee and Kim2019; Díaz-Fernández, González-Rodríguez, & Simonetti, Reference Díaz-Fernández, González-Rodríguez and Simonetti2016). For instance, Kolev and McNamara (Reference Kolev and McNamara2022) identified that TMT’s structural attributes influence the way in which TMTs encode poor performance information and coordinate actions. Especially, the unique role of TMTs’ job-related attributes in strategic changes or open innovations is highlighted (e.g., Lu, Liu, Xu, Liao, & Fu, Reference Lu, Liu, Xu, Liao and Fu2022; Naranjo-Gil, Reference Naranjo-Gil2009). How these characteristics affect TMTs’ interpretation of performance feedback and determine a firm’s choice of innovation alliance governance remains unaddressed. TMT tenure, capturing the time team members spend in the firm, is related to TMTs’ adherence to the status quo, their risk preference and willingness of external search (Miller, Reference Miller1991).

We expect the contingent role of TMT tenure in the relationship between performance feedback and governance structure. Non-equity innovation alliance structure, imbued with greater opportunistic costs and higher risks of alliance-failure (Majchrzak, Jarvenpaa, & Bagherzadeh, Reference Majchrzak, Jarvenpaa and Bagherzadeh2015; Niesten & Jolink, Reference Niesten and Jolink2017), posits a potential threat to TMT’s reputation and status. Thus, longer-tenured TMTs are less likely to choose such an alliance structure (Simsek, Reference Simsek2007). Consequently, TMT tenure will magnify the negative effect of negative performance feedback on non-equity innovation alliance structure. By comparison, an equity innovation alliance structure involves greater degrees of stability and higher levels of trust (Phene & Tallman, Reference Phene and Tallman2012; Williamson, Reference Williamson2008), contributing little threats on TMT’s career. Thus, the promotive effect of positive performance feedback on the adoption of an equity alliance governance will be further enhance when longer-tenured TMTs are present (Carpenter, Reference Carpenter2002). By summarizing the above points, we posit the following hypothesis:

The moderating role of TMT educational-level diversity

TMT educational-level diversity is defined as the extent to which team members differ in their educational levels (Lv et al., Reference Lv, Zhu, Chen and Lan2021). It reflects the heterogeneity of managerial cognitive bases and the diversity of TMT knowledge, which largely determines a firm’s external search ability (Kolev & McNamara, Reference Kolev and McNamara2022). In response to negative performance feedback, TMT members with heterogeneous educational levels can analyze the feedback information from various perspectives, which spurs firms into an identification of problem causes and a search for effective solutions (Lo & Fu, Reference Lo and Fu2016). As TMTs diverse in knowledge assets possess strong capabilities to gather and integrate external knowledge, they are competent to scan and obtain sophisticated information from a variety of sources. Such heterogenous knowledge bases may improve TMTs’ ability of absorbing a broader range of knowledge, help tackle technological problems arising at different levels in the knowledge integration, and partially offsets the low efficiency of knowledge transfer associated with a non-equity alliance mode (Díaz-Fernández, González-Rodríguez, & Simonetti, Reference Díaz-Fernández, González-Rodríguez and Simonetti2016). Thus, TMTs with diverse cognitive bases are more prone to choose a non-equity innovation alliance. By comparison, for those firms with positive performance feedback, heterogeneous TMTs consisting of managers with various educational levels excel in transforming knowledge to outside of the organizations. Furthermore, the knowledge breadth of TMTs offers firms with a number of choices and ensures compatibility for their cooperations with external organizations (Naranjo-Gil, Reference Naranjo-Gil2009). This advantage leads to greater transfer efficiency of innovation knowledge, which reduces the necessity for firms with positive performance feedback to use an equity-based mode for innovation alliances. We therefore hypothesize that:

Sample and data

We test our hypotheses in the context of Chinese biopharmaceutical industry where firms encounter higher levels of technological dynamism and market competition (Beers & Zand, Reference Beers and Zand2014). To address environmental challenges, the biopharmaceutical firms growingly engage in building R&D alliances with external partners to acquire complementary knowledge. Thus, innovation alliances are common in this sector (Michelino et al., Reference Michelino, Cammarano, Lamberti and Caputo2015). Our data was collected from two widely used databases: Chinese Stock Market & Accounting Research (CSMAR) database and China National Intellectual Property Administration (CNIPA) database.

As patent data has been intensively used to assess innovation, joint patents are usually considered as a measure of innovation alliances (Delerue, Reference Delerue2018). Hence, we gathered the data of joint patent to capture the selection of innovation alliance governances. The patent data of biopharmaceutical firms was primarily collected from CNIPA database, which generated an initial sample of 275 firms. We took several steps to purify the initial sample: (1) we eliminated those firms without patent data; (2) we kept the patents with two or more applicants, but omitted patents of which applicants include individuals as it was extremely difficult to identify whether those individuals were the employees of certain companies (Choi & Contractor, 2019); (3) we only retained invention patents while excluded utility and design patents because the former one sets a stricter requirement on resource inputs and innovativeness and involves greater chance of failure than the two latter ones, which thus fits our topic more appropriately. The refined sample consisted of 3,662 cooperative invention patents from 165 listed firms.

We estimated whether the alliance deals are equity or non-equity contracts through two online platforms of Qichacha and Tianyancha, which contain information of interorganizational business relationships. Specifically, we searched the relationship information via Qichacha website and then verified it by using Tianyancha website. In addition, the data of independent and control variables were collected through CSMAR database. By integrating these sources of data together, our final sample includes a total of 315 observations of 78 firms spanning the years from 2007 to 2020.

Dependent variables

Independent variables

Performance feedback

Following extant studies on performance feedback (Greve, Reference Greve2003; McConnell & Servaes, Reference McConnell and Servaes1990), we gauged this variable by capturing the gap between social aspirations and current performance. To seek performance improvement, they usually use the industry peers as the reference group to obtain accurate assessment and carry out improvement (Hu, Blettner, & Bettis, Reference Hu, Blettner and Bettis2011). Specifically, by measuring firm performance based on return on assets, we calculated social aspirational level from equation (1) and social aspirational gap from equation (2). In equation (1), median return on asset refers to the median performance of all firms within the biopharmaceutical industry in the t – 1 year, and the weight value $ \propto \,$ is 0.4. The initial social performance aspiration was obtained by weighting of the median performance in the year before (weight = 0.6) and the median performance in the previous 2 years (weight = 0.4). As shown in equation (2), social aspiration gap is measured by the discrepancy between focal firm’s return on asset and the social aspirational level. In line with Chen and Miller (Reference Chen and Miller2007), we constructed two variables: Positive performance feedback was measured by the value of the social aspiration gap when it is positive, and 0 otherwise, to capture an overperforming performance of the focal firm; Negative performance feedback was scaled by the absolute value of the social aspiration gap when this is negative, and 0 otherwise, to reflect underperforming performance of the focal firm. We then multiply performance feedback by 100 simply to make discussing the resulting coefficients easier.

(1)\begin{equation}\begin{aligned} Social\,aspirational\,leve{l_{i,t}} & = \left( {1 - \propto \,} \right)\,MRO{A_{t - 1}} \\ & \quad + Social\,aspirational\,leve{l_{i,t - 1}} \\ \end{aligned} \end{equation}

(2)\begin{equation}Social\,aspirational\,ga{p_{i,t}}\,\, = RO{A_{i,\,t}} - Social\,aspirational\,leve{l_{i,t}}\end{equation}

Moderators

Control variables

In line with prior innovation alliance studies (Lohrke, Kreiser, & Weaver, Reference Lohrke, Kreiser and Weaver2006; Tyler & Caner, Reference Tyler and Caner2016), we controlled for two sets of variables that may affect the firms’ selection of alliance governance structure. At the firm level, as small or young enterprises are likely to form innovation alliance with incumbent partners to improve their legitimacy and reduce innovation risks (Lv et al., Reference Lv, Zhu, Chen and Lan2021; Majocchi, Mayrhofer, & Camps, Reference Majocchi, Mayrhofer and Camps2013), we controlled for firm size and firm age, measured by the logarithm of the number of employees and number of years since its establishment respectively. We also included return on equity measured by equity divided by revenue, and R&D intensity scaled by the ratio of R&D expenditures to total assets. Firm ownership was included as a binary variable, coded as 1 for state owned enterprises and 0 otherwise. As firms located in the different regions of China are expected to experience different levels of difficulty in finding an appropriate partner, we used a dummy variable to control the firm’s geographic location by coding ‘1’ for firms located in Chinese eastern part (Beijing, Shanghai, Hebei, Tianjin, Shandong, Jiangsu, Zhejiang, Fujian, Guangdong, and Hainan) and ‘0’ for firms located in other cities or provinces. Following Tyler and Caner (Reference Tyler and Caner2016), firm slack was controlled by calculating the sum of different categories of slacks, including available, absorbed, and potential slack. Available slack is calculated by current assets divided by current liabilities; absorbed slack is measured by the expenses-to-sales ratio; and potential slack is measured as the equity-to-debt ratio.

We also controlled for a number of executive characteristics that may affect the firm strategic choices and alliance formation. CEO duality is measured by creating a binary indicator regarding whether or not the CEO was also the chairman of the board (1 = yes, CEO is also a chairman of the board; 0 = otherwise). We additionally considered the composition of TMT by measuring TMT size and TMT’s average age (Liu et al., 2012). Further, board independence was calculated as the percentage of the independent directors in the total number of firm’s board members (Kolev & McNamara, Reference Kolev and McNamara2022). The year effect and firm-level effect were also controlled in this study.

Analysis

In our analysis, we assess the direct impact of social performance feedback on the selection of innovation alliance structure, and the moderating effects of TMT average tenure and TMT’s educational-level diversity on the association between performance feedback and alliance structure choice. First, as many firms in our sample have selected equity innovation alliance structure solely, their propensity to choose non-equity alliance structure will approach the value of 0; vice versa. Thus, we performed 0–1-double-censored panel Tobit model where the censored condition fits our situation that many values are at a specific level of 0 or 1 (McDonald & Moffitt, Reference McDonald and Moffitt1980). Second, the fixed-effects model is more appropriate since the sample of our study is small. To obtain an unbiased standard error, we used cluster adjustment for the standard errors at the firm level when conducting analyses.

We placed a time lag for all the explanatory variables in the Tobit regression analysis to reasonably establish hypothesized causal relationships. Social aspiration gaps were set by a 3-year lag because it allows the performance feedback of a biopharmaceutical firm in t − 3 year to be translated into the cooperative applications patents in t year (Fernald, Pennings, & Claassen, Reference Fernald, Pennings and Claassen2015). Control variables were also configured as 3-year prior variables. The regression model is illustrated as follows:

(3)\begin{equation}\begin{aligned} {Y_{i,t}} & = {\beta _0} + {\beta _1}X{1_{i,t - 3}} + {\beta _2}X{2_{i,t - 3}} + {\beta _3}TMT\,tenur{e_{i,t - 3}} + {\beta _4}TMT\,educational \\ & \quad - level\,diversit{y_{i,t - 3}} + {\beta _5}X{1_{i,t - 3}} \times \,TMT\,tenur{e_{i,t - 3}} + {\beta _6}X{2_{i,t - 3}} \times TMT\,tenur{e_{i,t - 3}} \\ & \quad + {\beta _7}X{1_{i,t - 3}} \times \,TMT\,educational - level\,diversit{y_{i,t - 3}} + {\beta _8}X{2_{i,t - 3}} \times TMT\,educational \\ & \quad - level\,diversit{y_{i,t - 3}} + {\beta _9}CONTROL{S_{i,t - 3}} + {\varepsilon _{i,t - 3}} \\ \end{aligned} \end{equation}

Specifically, the dependent variables (equity innovation and non-equity innovation alliances) were regressed in accordance with the hypotheses. The independent variables (X1_{i,t −3}, X2_{i,t −3}) refer to the positive and negative performance feedback of firm_i respectively in year t − 3. The moderator variable – TMT tenure_{i,t −3} – represents the TMT’s average tenure of firm_i in t − 3 year; TMT educational-level diversity_{i ,t−3} indicates the diversity of TMT educational levels of the firm_i in t − 3 year, and $\varepsilon $ is a random distribution item.