Exploring and yet failing less: learning from past and current exploration in R&D

Exploring and yet failing less: learning from past and current exploration in R&D D’Este, Pablo;Marzucchi, Alberto;Rentocchini, Francesco 2017-11-17 00:00:00 Abstract Exploration is both an important part of a firm’s innovation strategy and an activity that involves a high degree of uncertainty. This article investigates a duality in the exploratory component of R&D activity with regard to innovation failure: while exploration is likely to increase firms’ exposure to failure, it might also provide learning opportunities to reduce failure. Our study contributes to the innovation management and organizational learning literatures by demonstrating the value of exploratory R&D for enabling two types of learning mechanisms. The first, experience-based learning, is based on the learning opportunities derived from accumulated experience in exploratory R&D: it involves improvements to procedures associated with experimentation and provides guidance for current exploration and to navigate the search space. The second, inferential-based learning, is based on the learning opportunities derived from current exploratory R&D efforts, which are associated with improved interpretation of ill-defined problems and timely responses to unstructured information. We draw on a longitudinal data set of 2226 Spanish manufacturing companies and show that, when past experience is associated with current exploration, innovation failure in the conception phase is reduced. We also find an inverted U-shaped relation between current exploratory R&D and innovation failure, in both the conception and implementation phases of innovation activities, showing that increasing levels of investment in current exploration activities attenuate the initial positive association between exploratory R&D and failure. 1. Introduction Exploration is an important part of the strategies of firms striving to bring to market new products and processes that promise large revenues if they are difficult to imitate by competitors. In this sense, firms’ exploration activities are essential to achieve sustained competitive advantage (March, 1991). However, while firms need to explore to build and retain their competitive edge, they need also to learn how to manage the greater level of uncertainty and risk involved in highly explorative innovation activity (Edmondson, 2011). This is not an easy balance to achieve. Although firms want to minimize failure related to innovation-related activities such as exploration (Desai, 2010), they may be willing to tolerate some level of failure if the exploration activity provides valuable new knowledge and learning opportunities that are aligned to their innovation strategy (Leonard-Barton, 1995; Edmondson, 2011). However, despite the increasing interest in organizational learning from failure (Haunschild and Sullivan, 2002; Madsen and Desai, 2010; Desai, 2015) and the organizational costs associated with the termination of R&D projects (Shepherd et al., 2014), very little research focuses specifically on the relationship between exploratory R&D and innovation failure. The present study investigates the relationship between exploratory R&D and failure in innovation activities. We examine the learning potential from two sources of information and knowledge gained in firms’ exploration activities that can contribute to lowering the risks of failure associated with exploratory R&D. The first is the firm’s accumulated experience in exploration (i.e., experience-based learning from past exploration). Prior experience affects the articulation of search routines and helps to identify the relevant search space (Nelson and Winter, 1982). Moreover, when applied to current activities, previously elaborated schemes can be refined to improve understanding about the underlying causes of problems in innovation-related activities (Argote and Miron-Spektor, 2011). The second source is the learning opportunities derived from concurrent exploration efforts, which emerge when a significant amount of (unstructured) information collected en route enables better interpretation of ill-defined problems and more timely response to them (inferential-based learning from current exploration) (Cook and Brown, 1999). Specifically, we investigate the following interrelated questions. First, we investigate whether, and to what extent, firms can benefit from accumulated experience in exploration activities and, in particular, whether there are positive complementarities with current exploration activities in terms of reducing innovation failure. Second, we investigate whether the scale of current explorative R&D contributes to reducing the likelihood of innovation failure. To address these research questions, we use longitudinal data, covering the period 2007–2012, on a sample of 2,226 manufacturing firms derived from the Spanish Technological Innovation Panel (PITEC) and exploit information on firms’ expenditure on the exploratory component of R&D, and the probability of failed innovation activity. This article contributes to the organizational learning and innovation management literatures by shedding new light on the mechanisms that allow firms to gain valuable learning from exploratory R&D. To the best of our knowledge, the relation between exploratory R&D activities and innovation failure has not been tested explicitly. We aim to fill this gap and to challenge the conventional wisdom that recognizes exploration as a high-risk activity but ignores the learning opportunities that it provides. These learning opportunities improve the effectiveness of the exploration activities and attenuate the risks of innovation failure. We show that exploratory R&D can improve organizational performance by reducing the risk of abandonment of innovation activities. We use information related to two temporal dimensions of exploratory R&D: current tasks and accumulated experience (Argote and Miron-Spektor, 2011). We find evidence of a beneficial interplay between past experience of explorative research and current exploratory R&D, although the effect is significant only for reducing failure in the conception phase of the innovation process. This suggests that information based on previous experience, when applied to current exploration, can help to identify problems and possible solutions mainly in the conception phase, when innovation activities are characterized by structured experimentation and are less dependent on context-specific issues. We show also that the intensity of investment in current exploration contributes to reducing the risk of failure in both the conception and implementation phases of innovation activities. This suggests that although current exploration is likely to generate unstructured information, which could lead to attribution and inferential errors, a substantial amount of efforts devoted to current exploration can help the firm to interpret ill-defined problems and to respond to them in a timely way. The article is organized as follows. Section 2 reviews the relevant literature and presents our hypotheses. Section 3 describes the data and methods. Section 4 presents our findings and some robustness checks. Section 5 concludes by discussing the implications of our findings for management and future research. 2. Exploration and innovation failure 2.1 Exploratory R&D: buffer to myopic learning and driver of innovation failure? Organizational learning processes can induce myopia due to over-attention to immediate performance and a tendency to ignore exploration and distant search (Levinthal and March, 1993). Adaptive learning, in which firms prioritize effectiveness and attention to innovation in the familiar territory of current experience, limits the opportunities for learning (Gupta et al., 2006; Fang et al., 2010). This is because firms tend to focus on reliable and predictable performance at the expense of the search for new ideas or new markets, which has less certain outcomes, longer time horizons, and more diffuse effects (March, 1991; Arora et al., 2015). These adaptive learning processes potentially can be self-destructive and endanger the organization’s long-term survival (March, 1991, 2006). In contrast, organizations may try deliberately to counterbalance the bias toward myopic learning by committing to exploration activities. Exploration includes activities related to search, variation, experimentation, flexibility, and discovery (March, 1991). It contributes to building capabilities outside current competencies and niches (Rosenkopf and Nerkar, 2001; Katila and Ahuja, 2002; Nerkar, 2003), helps to develop new cognitive frames (Chiva et al., 2010), and enables learning from failure (Madsen and Desai, 2010; Edmondson, 2011). Thus, exploration activities can be conceived as essential for long-term firm survival and growth. Exploration typically is embedded in the innovation strategies of firms committed to formal R&D investment. Firms’ investment in R&D nurtures expertise in fields that represent new opportunities and contributes to the building of in-house competencies to develop new products and processes (Cohen and Levinthal, 1990; Tidd et al., 1997; Dodgson et al., 2008). In particular, exploration-oriented R&D is likely to contribute to a radical departure from the organization’s current knowledge-base and avoid myopic learning. In fact, R&D is not a homogenous activity (Czarnitzki et al., 2011; Barge-Gil and López, 2014). We contend that the “R,” the research component, is the main channel for exploration within R&D activity, while the “D,” the development component, is associated mainly with exploitation activities. Research (R) can be disentangled further into basic and applied elements. Basic research is experimental or theoretical work aimed at the acquisition of new knowledge on foundations and observable facts, not aimed at any particular application (OECD, 2002). This activity, which is characterized by uncertainty and a long time horizon, is clearly exploratory in nature since it refers to the search for new ideas and relations among constructs (March, 1991). Applied research shares with basic research the search for original outcomes—i.e., new pieces of knowledge or new operational methods—while it differentiates from basic research for the specific objective of addressing practical aims (OECD, 2002). In contrast to research, the outcomes of the development component of R&D are more predictable, have a shorter time horizon, and are based on existing ideas and relations (March, 1991). Development makes use of existing knowledge that emerged from the research phase or from practical experience and is directed toward the introduction of new materials, products, and devices or improvements to existing ones (OECD, 2002). In sum, to use March’s (1991) terminology, research can be considered as closely related to the “exploration of new possibilities,” while development is concerned mainly with the “exploitation of old certainties.” While the exploratory component of R&D is necessary to avoid myopic learning, it increases the chances of innovation failure due to the inherently complex nature of the activities involved in exploration and the uncertainty related to its potential outcomes. Project failure is more frequent in companies engaged in exploration activities since this type of activity is more likely to face unpredictable contingencies. Indeed, the notion of innovation failure is typically associated with deviation from the expected and desired results from an R&D project and involves “both avoidable errors and the unavoidable negative outcomes of experiments and risk taking” (Cannon and Edmondson, 2005: 300). We define innovation failure as the realization of unacceptably low performance in R&D projects that results in a managerial decision to terminate the activity, which can occur at any stage of the innovation development process (Green et al., 2003; Shepherd and Cardon, 2009; Shepherd et al., 2014). We consider two major stages associated with the development of firm innovation projects: the conception phase and the implementation (or execution) phase (Tidd et al., 1997). In the conception phase, firms search for alternative innovation choices (identifying technological and market opportunities) and select ideas that are proposed and tested to establish an intended design or proof-of-concept prototype. The implementation phase corresponds to the stage in the innovation process when innovation efforts are oriented to executing the project to achieve a functional or working prototype that can be used to scale up manufacture and deliver a marketable product. Since exploratory R&D involves expanding the current set of capabilities, firms are likely to be exposed to a wide range of possible failures in both the conception and implementation phases of the innovation process. We consider project innovation failures associated with these two phases separately, since we expect this distinction to allow us to identify similarities and differences in the relationship between exploratory R&D and innovation failures. Based on this discussion, we hypothesize that: H1: Exploratory R&D is positively associated to the probability of innovation failure in the conception and implementation phases of the innovation process. H1 is our baseline hypothesis and is in line with prior contributions, although, to our knowledge, no direct test of the effect of exploratory R&D on the probability of innovation failure has been proposed in the literature. In what follows, we test this H1 to challenge two fundamental aspects of its underlying proposition. First, we contend that the expectation of a positive association between exploratory R&D and innovation failure is not uniform; instead, this effect may depend on both the scale of current exploratory R&D and the interplay with accumulated experience of exploratory R&D. Second, we investigate whether the effects of exploratory R&D on innovation failure are different in the innovation process phases of conception and implementation. 2.2 Learning from R&D exploration activity The above discussion refers to the increased exposure to failure associated to exploratory R&D and raises questions about whether and to what extent firms can manage this conflict. Drawing on organizational research which maintains that learning begins with experience (Fiol and Lyles, 1985; Argote and Miron-Spektor, 2011), we argue that experience in R&D exploration activity can provide learning opportunities that help organizations to attenuate the risk of innovation failure. Search processes are not unstructured activities (Nelson and Winter, 1982). Firms engaging in exploratory search activities establish programmed procedures and standard practices to increase the returns from experimentation and to increase the efficiency of the process by reducing the chances of errors and mistakes. Consequently, firms can improve the performance of their exploration-related activities based on accumulated experience. However, although accumulated experience is important, the transfer of existing practices to specific contexts is not always either easy or feasible, particularly in contexts that involve the identification of actionable responses related to radically new lines of inquiry—which is what exploratory search is about (Cook and Brown, 1999; Majchrzak et al., 2004). This leads to an acknowledgement of the temporal dimension of the knowledge creation and learning processes associated to exploration activity (Nerkar, 2003), and the distinction between learning from accumulated experience and learning in practice (Cook and Brown, 1999). Building on this discussion, we suggest that exploration in R&D offers two sources of learning opportunities, which could contribute to balancing the conflict between more intensive exploratory search and increased rates of failure: experiential-based learning from past exploration and inferential-based learning from current exploration. We discuss these aspects below. 2.2.1 Experiential-based learning from past exploration Firms’ accumulated experience of exploratory R&D can be embedded in routines and standard procedures, which contribute to the development of intelligence, monitoring, and surveillance capabilities associated with experimentation activity. This accumulated experience and sustained exploratory R&D efforts can lead to learning opportunities to reduce the likelihood of failure in innovation activities. This is described as “experiential-based learning from past exploration.” First, we argue that programmed exploration can be particularly effective for reducing instances of preventable failure caused by deviance from rules, inattention, or lack of ability in the conduct of routine and predictable operations (Edmondson, 2011). Along similar lines, Thomke (2001) suggests that building experimentation capabilities implies avoiding two types of mistakes: those resulting from badly conducted experiments, which produce ambiguous or not valuable information, and those that involve repeating prior failures. The extant literature suggests that longitudinal information on past performance, different perspectives on what went wrong in the past, the sharing of ideas with managers able to implement change, and careful consideration of whether change is having the desired effect contribute to building the firm’s capacity to anticipate, analyze and act upon failure (Tucker and Edmondson, 2003; Cannon and Edmondson, 2005). In addition, we contend that past experience in exploration activity can be beneficial if it leads to opportunities for more effective future exploration activities. Organizations can benefit from the interaction between current and accumulated experience in exploration activity, through a moderating effect that reduces the probability of failure. Past exploration activities can help to map and navigate the relevant search space and, thereby, contribute to making current exploration activities more efficient and to developing a learning process associated to current exploration (Eisenhardt and Martin, 2000; Zollo and Winter, 2002). Majchrzak et al. (2004) report that past experience can support different stages of current radical and exploratory innovation: from the conceptualization of new ideas, through the use of analogies and extensions linked to previously developed concepts, to final product development, in which experience is shared in the course of testing, review, and improvements on prototypes. Past experience can provide an appropriate lens to analyze new information collected en route and to improve the accuracy of inferences related to current exploration processes (Denrell et al., 2004). This can be achieved by applying information elicited from both successful and failed past exploratory activities, to current R&D exploration (Leonard-Barton, 1995; Edmondson, 2011). Accordingly, we acknowledge knowledge retention as central to the firm’s learning process and as depending on the organizational memory embedded in employees, the tools available, and the tasks to be performed (Argote and Ingram, 2000). In this framework, past experience provides meta-knowledge about who knows and does what (i.e., transactive memory), which, in turn, increases ongoing performance by improving: (i) task assignment (better matching of researchers to research tasks); (ii) problem-solving and coordination (researchers know whom to approach for advice); and (iii) division of intellectual labor which leads to specialized learning of different pieces of information. In sum, past experience of exploratory R&D, if properly retained and reused in current endeavors (Argote and Miron-Spektor, 2011), can improve organizational performance and reduce the risk of failure. However, it is important to note that applying previously elaborated data and schemes to current, and possibly different, activities requires refinement of insights from prior experience. Although the relevant information may be available in the organization through accumulated experience, it can be difficult to find solutions to complex problems in a changed use environment (Von Hippel and Tyre, 1995). In our context, this implies that past exploration experience becomes an organizational asset if applied appropriately to current exploration. Firms can make use of previously acquired information to improve identification of the causes of problems in their innovation activities and allow their correction before they can lead to failure. Such beneficial interplay between past and current exploration resonates with the prescriptions in Schilling et al. (2003: 50), according to which learning opportunities are maximized in the case of related variation, that is, when “working on different but similar types of problems over time.” This type of variation, which is likely to occur in subsequent exploratory activities within the same firm, allows a deeper understanding of the performed tasks by creating associations among similar concepts in different contexts. These arguments lead to the following hypothesis: H2: Past experience in exploratory R&D negatively moderates the relation between current exploratory R&D and the probability of failure in the conception and implementation phases. 2.2.2 Inferential-based learning from current exploration We next discuss the learning potential from current exploration activities, which we describe as inferential learning in the context of practice (Cook and Brown, 1999). Efforts oriented toward exploration tend, by definition, to challenge the organization’s existing knowledge base. Indeed, the context that characterizes exploration activities involves features that, typically, defy current routines and existing capabilities. New exploration activities involve causal ambiguity and resolution of ill-defined problems, since firms have a partial understanding of the causal relationships among multiple stages in search processes associated with radical new product development (Lippman and Rumelt, 1982; Segal, 2004). Current R&D exploration is characterized also by delayed performance feedback and credit assignment, since there is a temporal lag between action and observed performance (Denrell et al., 2004; Fang, 2012). These activities often involve the formulation of new agendas to define novel product portfolio and business model orientations (Fiol and Lyles, 1985; He and Wong, 2004). In this context, well-established routines may not only provide limited guidance but also may hinder the development of new cognitive frames that provide adequate representations of the nature of the problem at hand. However, we contend that learning from current exploration can occur especially if the scale of the research expands the experimentation space and increases the number of observations to draw upon. Learning from current exploration might be even more important than learning based on past experience, whose value might depreciate over time (Argote et al., 1990; Benkard, 2000; Argote and Miron-Spektor, 2011). In the course of their current exploratory R&D, firms can devise adaptive problem-solving reactions, even if the possibility for a radical revision of the prevailing rules of action and routines is limited in the short-run (Argyris and Schön, 1978; Fiol and Lyles, 1985; Sadler-Smith et al., 2001). These learning outcomes can be guided by information elicited from projects that are ongoing. Partial information, in the form of unstructured pieces of knowledge in the problem space or intermediate clues and signals within a multistage problem, may be useful to guide exploration and avoid “random walks” (Denrell et al., 2004; Fang and Levinthal, 2009). In organizing their exploration activities, firms generally acknowledge that more radical experiments are likely to provide both learning opportunities and also greater exposure to failure in the short term. Leonard-Barton (1995) and Edmondson (2011) point to instances of “good” or “intelligent” innovation failure that are associated with deliberate experimentation and exploration oriented to providing valuable opportunities for acquiring new knowledge in the short term, and contributing to the interpretation of ongoing problems. Intelligent failures are likely to be particularly important in the project conception phase when firms can explore different paths and reject unfeasible alternatives at comparatively low cost. This applies especially to new simulation technologies, which have continuously reduced the costs of generating critical data on virtual, as opposed to physical prototypes (Thomke, 2001). Moreover, current R&D exploration efforts can trigger new heuristics and interpretive skills at any point in the innovation process, contributing to the adoption of new frames of reference and problem-definition schemes to avoid preventable and complexity-related failures. New heuristics and interpretive skills increase inference accuracy and compensate for the small-scale experimentation and limited experiential-based sources of learning associated with new exploration (Kim et al., 2009). Therefore, as a consequence of inferential-based learning from exploration activities, we argue that the probability of innovation failure is likely to decline with an increased scale of current exploration efforts. Accordingly, we hypothesize a curvilinear inverted U-shaped relationship between current exploratory R&D and innovation failure. The rationale for this hypothesis is that, while current R&D exploration is likely to challenge existing routines and expose firms to higher risks of innovation failure (as put forward in H1), current R&D exploration efforts are likely to enhance the firm’s capacity to interpret unstructured information once a threshold scale of exploration is achieved, resulting in learning opportunities that contribute to reduce innovation failure in both the conception and implementation phases. Thus, we hypothesize that: H3: The degree of current exploratory R&D has an inverted U-shape relationship with the probability of experiencing innovation failure in the conception and implementation phases. 3. Data and methods 3.1 Data and sample frame Our analysis uses data from PITEC which is co-managed by the Spanish National Statistics Institute (INE), the Spanish Foundation for Science and Technology (FECYT), and the Foundation for Technical Innovation (COTEC). PITEC is a Community Innovation Survey (CIS)-type, firm-level anonymized data set based on survey waves (for a review of innovation surveys, see Mairesse and Mohnen, 2010).1 PITEC’s core sections are similar to sections in the CIS, a harmonized questionnaire developed by Eurostat based on the Organisation for Economic Co-operation and Development (OECD)’s Oslo Manual (OECD, 2005). Due to its reliability, open-access policy and range of innovation-related information, PITEC is being used increasingly as the data source for empirical studies of firm-level innovation.2 PITEC data are particularly appropriate for our study because they allow identification of the exploratory character of R&D, and they provide information on innovation failures in the conception and implementation phases of the innovation process. We focus on the period 2007–2012, aggregating information from six PITEC annual survey waves. Similar to other CIS-type surveys, some of the questions (e.g., about failure in innovation projects) refer to a 3-year period (e.g., for the 2012 survey wave, some questions refer to the period 2010–2012), while other questions (e.g., on R&D spending, among others) refer to a single year (e.g. 2012). For consistency, we use three consecutive waves of the PITEC survey to build measures for our key variables (i.e., 2007, 2008, and 2009; and 2010, 2011, and 2012). This allows us to build a set of variables referring to two non-overlapping periods (2007–2009 and 2010–2012). The sample contains full information for the variables of interest for 2,226 manufacturing firms over two time periods (2007–2009 and 2010–2012), resulting in a quasi-balanced panel of 4,191 firm-period observations. Our main focus is on firms that engage in innovation-related activities and therefore are susceptible to experience different rates of innovation-related failure. Table 1 shows the proportion of firms included in PITEC that have experienced overall failure and failure in the conception and implementation phases of innovation projects, distinguishing between those engaged in innovation-related activities and those that are not. Table 1 shows that, as expected, the probability of failure is higher for companies that engage in innovation activities (the difference between the two probabilities is always statistically significant at the 1% confidence level). Given the focus of our analysis, we restrict the sample to manufacturing firms that are exposed to the risk of experiencing innovation failure, that is, companies that engage in innovation activities.3 Table 1. Probability of failure in R&D projects: engagement versus non-engagement in innovation activities Engaging in innovation-related activities Not engaging in innovation-related activities Pearson χ2 % failure overall 31.91 9.38 563.23 (1)*** % failure conception 24.4 5.71 482.62 (1)*** % failure implementation 20.84 6.03 336.77 (1)*** Observations 3136 5628 Engaging in innovation-related activities Not engaging in innovation-related activities Pearson χ2 % failure overall 31.91 9.38 563.23 (1)*** % failure conception 24.4 5.71 482.62 (1)*** % failure implementation 20.84 6.03 336.77 (1)*** Observations 3136 5628 Degrees of freedom are in brackets. The sample refers to all manufacturing companies in the period 2007–2012. Firms engaging in innovation-related activities refer to companies which, in the period 2007–2012, invested in at least one of the following activities: intramural R&D; extramural R&D; acquisition of machinery, equipment and software; acquisition of other external knowledge; training; market introduction of innovations; other activities. ***P<0.01 Table 1. Probability of failure in R&D projects: engagement versus non-engagement in innovation activities Engaging in innovation-related activities Not engaging in innovation-related activities Pearson χ2 % failure overall 31.91 9.38 563.23 (1)*** % failure conception 24.4 5.71 482.62 (1)*** % failure implementation 20.84 6.03 336.77 (1)*** Observations 3136 5628 Engaging in innovation-related activities Not engaging in innovation-related activities Pearson χ2 % failure overall 31.91 9.38 563.23 (1)*** % failure conception 24.4 5.71 482.62 (1)*** % failure implementation 20.84 6.03 336.77 (1)*** Observations 3136 5628 Degrees of freedom are in brackets. The sample refers to all manufacturing companies in the period 2007–2012. Firms engaging in innovation-related activities refer to companies which, in the period 2007–2012, invested in at least one of the following activities: intramural R&D; extramural R&D; acquisition of machinery, equipment and software; acquisition of other external knowledge; training; market introduction of innovations; other activities. ***P<0.01 3.2 Variables and measures 3.2.1 Dependent variables Our dependent variables, Failure overall, Failure conception, and Failure implementation, proxy for failure in innovation activities, reflected in the decision to terminate an innovation project, and the decision to terminate it in the stages of conception and implementation (Green et al., 2003; Shepherd and Cardon, 2009; Shepherd et al., 2014). Other studies based on CIS or PITEC data (e.g., Mohnen et al., 2008; Lhuillery and Pfister, 2009; D’Este et al., 2016; Leoncini, 2016) adopt a similar approach to capture innovation failures using information on abandoned activities. Our dependent variables are built relying on firms’ responses corresponding to the relevant 3-year periods: 2007–2009 and 2010–2012. Failure overall is a dummy that takes the value 1 if the firm faces a failure event in an innovation project, i.e., if the firm reported abandonment of an innovation project in the conception or implementation phases. Following the arguments in Section 2, we consider two phases in which failures can occur: firms were asked to “locate” the event of a failure in either the conception or the implementation phases of innovation activities. Failure conception is a dummy that takes the value 1 if the firm faces a failure event in the conception phase of an innovation project, that is, if a project was abandoned in the conception phase. Failure implementation is a dummy that takes the value 1 if the firm faces a failure event in the implementation phase of an innovation project, that is, if the firm reported abandoning an innovation project in the implementation phase. 3.2.2 Explanatory variables We draw on previous studies and distinguish between exploratory and exploitation activities in the R&D process to build our explanatory variables (Czarnitzki et al., 2009; Czarnitzki et al., 2011; Barge-Gil and López, 2014, 2015). In Section 2.1 we discussed the connection between the characterization of exploration and exploitation proposed in March (1991), and the standard definitions of R&D (OECD, 2002). PITEC data allow us to quantify the firm’s engagement in the different components of R&D investment, in a way that is consistent with established measurement of innovation activity practices (OECD, 2002, 2005). In particular, PITEC data allow us to differentiate among investment in basic research, applied research, and development. Basic research relates to “experimental or theoretical work undertaken to obtain new knowledge on the foundation of phenomena or observable facts, without any particular application in view.” Also, applied research “consists of original work undertaken to obtain new knowledge” and differs from basic research in being “aimed at a particular objective.” Finally, development is defined as “systematic work based on existing knowledge, derived from research or practical experience, that is directed to the production of new materials, products or devices, or to installing new processes, systems and services or to improving those already existing or installed.”4 Based on these definitions, we construct our main explanatory variables for engagement in exploratory R&D. R&D exploration is obtained by averaging real expenditure on basic and applied research in 2007–2009 and in 2010–20125 and dividing the figures obtained by the average number of employees in the corresponding periods. To reduce distribution skewness, we apply a natural logarithmic transformation (adding +1 to avoid dropping zeros). We use R&D exploration to test our baseline H1 on the positive relationship between current exploration activities and the firm’s exposure to innovation failure. H2 refers to the combined effect of past exploration experience and current exploratory R&D. We test this hypothesis by including the interaction R&D exploration × R&D exploration-1. R&D exploration-1 captures past exploratory R&D: in 2007–2009 for the probability of failure in the period 2010–2012, and in 2004–2006 for the probability of failure in the period 2007–2009. We obtained information on investment in explorative R&D activities during 2004–2006 from three previous waves of the PITEC survey, which maintains the panel nature of our data set.6 Finally, H3 predicts an inverted U-shaped relationship between explorative R&D and the probability of experiencing innovation failure. To capture this nonlinear effect, we include in our econometric specification the squared term R&D exploration sq. 3.2.3 Control variables We tried to minimize omitted variables bias by including a set of controls in the econometric specification. We include a variable for the amount of investment in exploitation activities, R&D exploitation, measured as the natural log-transformed ratio of average expenditure on development activities normalized by the average number of employees in the same period. We control for factors that might affect the likelihood of failure during innovation activities. Following the literature, we consider both financial and non-financial barriers (e.g., D’Este et al., 2012, 2014; Blanchard et al., 2013). The PITEC data do not provide continuous information on the relevance of barriers: hampering factors are recorded on Likert scales that reflect the relevance of these issues for the firm. To define our barriers-related variables, we follow D’Este et al. (2012, 2014) and Antonioli et al. (2017) and create three dummies that take value 1 if the firm faced at least one highly relevant obstacle related to cost, knowledge, or market. Specifically, Cost barriers captures whether the firm faced at least one major obstacle in the form of innovation costs, and internal or external funding for innovation. Knowledge barriers reflects whether the firm reported knowledge barriers as highly important, including items such as problems related to the availability of skilled personnel, information on technology, information on markets, and availability of appropriate innovation partners. We consider also the potential effect of markets perceived as being dominated by incumbents (Market barriers—dominated) and characterized by uncertain demand (Market barriers—uncertainty). We control for different forms of engagement in innovation not included in our exploration and exploitation measures, that might be particularly relevant to small- and medium-sized enterprises (SMEs) and non-R&D intensive industries (e.g., Sterlacchini, 1999; Rammer et al., 2009). Other innovation expenditures, defined over the reference period as the natural log-transformed average sum (adding +1 to avoid dropped zeros), include expenditure per employee on external R&D, machinery, equipment and software, external knowledge, training, introduction of innovation in the market, design, and other activities. To further capture the complexity of the firm’s innovation profile, we control for adoption of open innovation strategies (e.g., Chesbrough, 2003). Building upon Laursen and Salter (2006), we use two variables proxying for external knowledge search breadth (Knowledge breadth) and depth (Knowledge depth) in the firm’s innovation strategy. Knowledge breadth takes values from 0 to 10, based on the number of sources of information for innovation used by the firm in the period considered. Knowledge depth ranges from 0 to 10, based on the number of information sources the firm rated as highly important in the period under consideration.7 To control for cutting-edge and risky innovation activities, we include a dummy, Radical innovation. In the absence of information on the number of innovations introduced, this variable takes the value 1 if the firm introduced at least one innovation that was new to the market (e.g., Monjon and Waelbroeck, 2003; Duguet, 2006). Also, we control for whether the company undertook any relevant organizational innovation in the period considered. We exploit information from responses to PITEC survey questions on organizational innovation to create the dummy Organizational innovation, which takes the value 1 if the firm introduced new or improved management systems, changes to the firm’s work organization (with reference to decision-making and distribution of responsibilities), or changes to relations with other organizations, and zero otherwise.8 We also include firm age (natural log transformed) (Firm age), which may be related to the propensity to introduce disruptive and risky innovations, and to the likelihood of higher obstacles to innovation (e.g., Schneider and Veugelers, 2010). We consider a set of characteristics that might influence innovation resources, incentives, and, in turn, the likelihood of engagement in unsuccessful innovation activities: number of employees dedicated to R&D activities (Number of R&D researchers), group affiliation (Group), and firm size (Number of employees). Number of R&D researchers is the (log transformed + 1) number of Full Time Equivalent (FTE) researchers in the firm’s R&D department and takes account of a possible R&D size effect. Group affiliation is measured as a dummy variable, while firm size is measured as the natural logarithm of the average number of employees in the reference period (+1). Finally, we include four dummies for the effect of industry characteristics, based on Pavitt’s (1984) industry sectors: science-based, specialized supplier, scale-intensive, and supplier-dominated.9 Table 2 presents descriptive statistics for the variables used in the study; Table 3 reports the correlation matrix of these variables. Apart from variables that are transformations of other variables (e.g., R&D exploration sq), correlation among the independent variables is generally weak and mean variance inflation factors range between 1.36 and 3.73 (well below the threshold value of 5), suggesting the absence of multicollinearity problems. 10 Table 2. Descriptive statistics (N = 4191) Variable Mean SD Min Max Dependent variables  Fail overall 0.312 0.463 0 1  Fail conception 0.237 0.425 0 1  Fail implementation 0.202 0.401 0 1 Explanatory variables  R&D exploration 1702.3 4196.19 0 76,390.03  R&D exploration-1 1848.7 4513.98 0 76,390.03 Controls  R&D exploitation 2490.204 6060.781 0 114,585  Number of R&D researchers 0.723 0.862 0 6.06  Other innovation expenditures 5.934 2.509 0 12.34  Cost barriers 0.167 0.373 0 1  Knowledge barriers 0.012 0.109 0 1  Market barriers uncertainty 0.266 0.442 0 1  Market barriers dominated 0.191 0.393 0 1  Knowledge breadth 6.406 3.302 0 10  Knowledge depth 1.279 1.579 0 10  Radical innovation 0.414 0.493 0 1  Organizational innovation 0.547 0.498 0 1  Firm age 3.284 0.572 1.61 5.18  Group 0.450 0.498 0 1  Number of employees 4.290 1.307 0.69 9.19 Industry dummies (Pavitt Taxonomy)  Supplier dominated 0.343 0.475 0 1  Scale intensive 0.271 0.445 0 1  Specialized suppliers 0.211 0.408 0 1  Science based 0.175 0.380 0 1 Variable Mean SD Min Max Dependent variables  Fail overall 0.312 0.463 0 1  Fail conception 0.237 0.425 0 1  Fail implementation 0.202 0.401 0 1 Explanatory variables  R&D exploration 1702.3 4196.19 0 76,390.03  R&D exploration-1 1848.7 4513.98 0 76,390.03 Controls  R&D exploitation 2490.204 6060.781 0 114,585  Number of R&D researchers 0.723 0.862 0 6.06  Other innovation expenditures 5.934 2.509 0 12.34  Cost barriers 0.167 0.373 0 1  Knowledge barriers 0.012 0.109 0 1  Market barriers uncertainty 0.266 0.442 0 1  Market barriers dominated 0.191 0.393 0 1  Knowledge breadth 6.406 3.302 0 10  Knowledge depth 1.279 1.579 0 10  Radical innovation 0.414 0.493 0 1  Organizational innovation 0.547 0.498 0 1  Firm age 3.284 0.572 1.61 5.18  Group 0.450 0.498 0 1  Number of employees 4.290 1.307 0.69 9.19 Industry dummies (Pavitt Taxonomy)  Supplier dominated 0.343 0.475 0 1  Scale intensive 0.271 0.445 0 1  Specialized suppliers 0.211 0.408 0 1  Science based 0.175 0.380 0 1 Descriptive statistics for R&D exploration; R&D exploration-1 and R&D exploitation refer to the variables before natural log-transformation and are measured in euros per employee. Descriptive statistics for the following variables are reported with natural log transformation: Number of R&D researchers, Other innovation expenditures, Firm age, and Number of employees. Table 2. Descriptive statistics (N = 4191) Variable Mean SD Min Max Dependent variables  Fail overall 0.312 0.463 0 1  Fail conception 0.237 0.425 0 1  Fail implementation 0.202 0.401 0 1 Explanatory variables  R&D exploration 1702.3 4196.19 0 76,390.03  R&D exploration-1 1848.7 4513.98 0 76,390.03 Controls  R&D exploitation 2490.204 6060.781 0 114,585  Number of R&D researchers 0.723 0.862 0 6.06  Other innovation expenditures 5.934 2.509 0 12.34  Cost barriers 0.167 0.373 0 1  Knowledge barriers 0.012 0.109 0 1  Market barriers uncertainty 0.266 0.442 0 1  Market barriers dominated 0.191 0.393 0 1  Knowledge breadth 6.406 3.302 0 10  Knowledge depth 1.279 1.579 0 10  Radical innovation 0.414 0.493 0 1  Organizational innovation 0.547 0.498 0 1  Firm age 3.284 0.572 1.61 5.18  Group 0.450 0.498 0 1  Number of employees 4.290 1.307 0.69 9.19 Industry dummies (Pavitt Taxonomy)  Supplier dominated 0.343 0.475 0 1  Scale intensive 0.271 0.445 0 1  Specialized suppliers 0.211 0.408 0 1  Science based 0.175 0.380 0 1 Variable Mean SD Min Max Dependent variables  Fail overall 0.312 0.463 0 1  Fail conception 0.237 0.425 0 1  Fail implementation 0.202 0.401 0 1 Explanatory variables  R&D exploration 1702.3 4196.19 0 76,390.03  R&D exploration-1 1848.7 4513.98 0 76,390.03 Controls  R&D exploitation 2490.204 6060.781 0 114,585  Number of R&D researchers 0.723 0.862 0 6.06  Other innovation expenditures 5.934 2.509 0 12.34  Cost barriers 0.167 0.373 0 1  Knowledge barriers 0.012 0.109 0 1  Market barriers uncertainty 0.266 0.442 0 1  Market barriers dominated 0.191 0.393 0 1  Knowledge breadth 6.406 3.302 0 10  Knowledge depth 1.279 1.579 0 10  Radical innovation 0.414 0.493 0 1  Organizational innovation 0.547 0.498 0 1  Firm age 3.284 0.572 1.61 5.18  Group 0.450 0.498 0 1  Number of employees 4.290 1.307 0.69 9.19 Industry dummies (Pavitt Taxonomy)  Supplier dominated 0.343 0.475 0 1  Scale intensive 0.271 0.445 0 1  Specialized suppliers 0.211 0.408 0 1  Science based 0.175 0.380 0 1 Descriptive statistics for R&D exploration; R&D exploration-1 and R&D exploitation refer to the variables before natural log-transformation and are measured in euros per employee. Descriptive statistics for the following variables are reported with natural log transformation: Number of R&D researchers, Other innovation expenditures, Firm age, and Number of employees. Table 3. Correlation matrix (N = 4191) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) (1) Failure 1 (2) Failure conception 0.82* 1 (3) Failure implementation 0.74* 0.46* 1 (4) R&D exploration 0.12* 0.14* 0.1* 1 (5) R&D exploration-1 0.09* 0.10* 0.08* 0.61* 1 (6) Explor × Explor-1 0.1* 0.12* 0.09* 0.83* 0.81* 1 (7) R&D exploration sq 0.11* 0.13* 0.09* 0.96* 0.62* 0.87* 1 (8) R&D exploitation 0.07* 0.1* 0.03* 0.11* 0.08* 0.08* 0.11* 1 (9) Knowledge breadth 0.12* 0.18* 0.09* 0.22* 0.15* 0.20* 0.22* 0.23* 1 (10) Knowledge depth 0.07* 0.11* 0.06* 0.14* 0.1* 0.14* 0.15* 0.15* 0.41* 1 (11) Other innovation exp 0.04* 0.05* 0.03* 0.06* 0.05* 0.09* 0.11* 0.08* 0.17* 0.14* 1 (12) Firm age 0.05* 0.04* 0.04* −0.05* −0.06* −0.06* −0.08* −0.05* 0.04* 0.01 −0.05* 1 (13) Group 0.04* 0.04* 0.05* 0.01 −0.01 0.01 0.01 −0.01 0.08* 0.03* 0.05* 0.11* 1 (14) Cost barriers −0.02 −0.02 −0.02 0.01 0.02 0.01 0.01 −0.01 −0.01 0.06* 0.01 −0.06* −0.13* 1 (15) Knowledge barriers 0.01 0.01 0.01 0.01 0.01 0.01 0.01 −0.01 −0.01 −0.01 −0.01 −0.01 −0.06* 0.12* 1 (16) Mkt barriers uncertainty 0.07* 0.06* 0.07* 0.03* 0.01 0.03* 0.03* 0.03* 0.04* 0.08* 0.03* −0.01 −0.06* 0.15* 0.09* 1 (17) Mkt barriers dominated 0.06* 0.06* 0.04* 0.06* 0.02 0.06* 0.07* 0.04* 0.05* 0.11* 0.03* −0.03* −0.08* 0.13* 0.12* 0.37* 1 (18) Number of employees 0.06* 0.07* 0.06* −0.03* −0.08* −0.05* −0.07* −0.04* 0.14* 0.06* 0.01 0.35* 0.55* −0.15* −0.06* −0.11* −0.1* 1 (19) Radical innovation 0.1* 0.14* 0.07* 0.15* 0.11* 0.14* 0.15* 0.21* 0.17* 0.12* 0.15* 0.01 0.04* −0.03* −0.01 0.03* −0.01 0.05* 1 (20) Organizational innovation 0.13* 0.17* 0.09* 0.09* 0.07* 0.1* 0.08* 0.09* 0.23* 0.12* 0.14* 0.05* 0.10* −0.04* −0.01 0.04* 0.03* 0.16* 0.14* 1 (21) Number R&D employees 0.13* 0.14* 0.12* 0.34* 0.27* 0.39* 0.40* 0.37* 0.29* 0.21* 0.19* 0.07* 0.28* −0.04* −0.03* −0.01 0.03* 0.4* 0.21* 0.16* (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) (1) Failure 1 (2) Failure conception 0.82* 1 (3) Failure implementation 0.74* 0.46* 1 (4) R&D exploration 0.12* 0.14* 0.1* 1 (5) R&D exploration-1 0.09* 0.10* 0.08* 0.61* 1 (6) Explor × Explor-1 0.1* 0.12* 0.09* 0.83* 0.81* 1 (7) R&D exploration sq 0.11* 0.13* 0.09* 0.96* 0.62* 0.87* 1 (8) R&D exploitation 0.07* 0.1* 0.03* 0.11* 0.08* 0.08* 0.11* 1 (9) Knowledge breadth 0.12* 0.18* 0.09* 0.22* 0.15* 0.20* 0.22* 0.23* 1 (10) Knowledge depth 0.07* 0.11* 0.06* 0.14* 0.1* 0.14* 0.15* 0.15* 0.41* 1 (11) Other innovation exp 0.04* 0.05* 0.03* 0.06* 0.05* 0.09* 0.11* 0.08* 0.17* 0.14* 1 (12) Firm age 0.05* 0.04* 0.04* −0.05* −0.06* −0.06* −0.08* −0.05* 0.04* 0.01 −0.05* 1 (13) Group 0.04* 0.04* 0.05* 0.01 −0.01 0.01 0.01 −0.01 0.08* 0.03* 0.05* 0.11* 1 (14) Cost barriers −0.02 −0.02 −0.02 0.01 0.02 0.01 0.01 −0.01 −0.01 0.06* 0.01 −0.06* −0.13* 1 (15) Knowledge barriers 0.01 0.01 0.01 0.01 0.01 0.01 0.01 −0.01 −0.01 −0.01 −0.01 −0.01 −0.06* 0.12* 1 (16) Mkt barriers uncertainty 0.07* 0.06* 0.07* 0.03* 0.01 0.03* 0.03* 0.03* 0.04* 0.08* 0.03* −0.01 −0.06* 0.15* 0.09* 1 (17) Mkt barriers dominated 0.06* 0.06* 0.04* 0.06* 0.02 0.06* 0.07* 0.04* 0.05* 0.11* 0.03* −0.03* −0.08* 0.13* 0.12* 0.37* 1 (18) Number of employees 0.06* 0.07* 0.06* −0.03* −0.08* −0.05* −0.07* −0.04* 0.14* 0.06* 0.01 0.35* 0.55* −0.15* −0.06* −0.11* −0.1* 1 (19) Radical innovation 0.1* 0.14* 0.07* 0.15* 0.11* 0.14* 0.15* 0.21* 0.17* 0.12* 0.15* 0.01 0.04* −0.03* −0.01 0.03* −0.01 0.05* 1 (20) Organizational innovation 0.13* 0.17* 0.09* 0.09* 0.07* 0.1* 0.08* 0.09* 0.23* 0.12* 0.14* 0.05* 0.10* −0.04* −0.01 0.04* 0.03* 0.16* 0.14* 1 (21) Number R&D employees 0.13* 0.14* 0.12* 0.34* 0.27* 0.39* 0.40* 0.37* 0.29* 0.21* 0.19* 0.07* 0.28* −0.04* −0.03* −0.01 0.03* 0.4* 0.21* 0.16* Dummies for Pavitt (1984) four categories of industrial firms are not included for space reasons but correlation coefficient is never above 0.3. * P < 0.05. Table 3. Correlation matrix (N = 4191) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) (1) Failure 1 (2) Failure conception 0.82* 1 (3) Failure implementation 0.74* 0.46* 1 (4) R&D exploration 0.12* 0.14* 0.1* 1 (5) R&D exploration-1 0.09* 0.10* 0.08* 0.61* 1 (6) Explor × Explor-1 0.1* 0.12* 0.09* 0.83* 0.81* 1 (7) R&D exploration sq 0.11* 0.13* 0.09* 0.96* 0.62* 0.87* 1 (8) R&D exploitation 0.07* 0.1* 0.03* 0.11* 0.08* 0.08* 0.11* 1 (9) Knowledge breadth 0.12* 0.18* 0.09* 0.22* 0.15* 0.20* 0.22* 0.23* 1 (10) Knowledge depth 0.07* 0.11* 0.06* 0.14* 0.1* 0.14* 0.15* 0.15* 0.41* 1 (11) Other innovation exp 0.04* 0.05* 0.03* 0.06* 0.05* 0.09* 0.11* 0.08* 0.17* 0.14* 1 (12) Firm age 0.05* 0.04* 0.04* −0.05* −0.06* −0.06* −0.08* −0.05* 0.04* 0.01 −0.05* 1 (13) Group 0.04* 0.04* 0.05* 0.01 −0.01 0.01 0.01 −0.01 0.08* 0.03* 0.05* 0.11* 1 (14) Cost barriers −0.02 −0.02 −0.02 0.01 0.02 0.01 0.01 −0.01 −0.01 0.06* 0.01 −0.06* −0.13* 1 (15) Knowledge barriers 0.01 0.01 0.01 0.01 0.01 0.01 0.01 −0.01 −0.01 −0.01 −0.01 −0.01 −0.06* 0.12* 1 (16) Mkt barriers uncertainty 0.07* 0.06* 0.07* 0.03* 0.01 0.03* 0.03* 0.03* 0.04* 0.08* 0.03* −0.01 −0.06* 0.15* 0.09* 1 (17) Mkt barriers dominated 0.06* 0.06* 0.04* 0.06* 0.02 0.06* 0.07* 0.04* 0.05* 0.11* 0.03* −0.03* −0.08* 0.13* 0.12* 0.37* 1 (18) Number of employees 0.06* 0.07* 0.06* −0.03* −0.08* −0.05* −0.07* −0.04* 0.14* 0.06* 0.01 0.35* 0.55* −0.15* −0.06* −0.11* −0.1* 1 (19) Radical innovation 0.1* 0.14* 0.07* 0.15* 0.11* 0.14* 0.15* 0.21* 0.17* 0.12* 0.15* 0.01 0.04* −0.03* −0.01 0.03* −0.01 0.05* 1 (20) Organizational innovation 0.13* 0.17* 0.09* 0.09* 0.07* 0.1* 0.08* 0.09* 0.23* 0.12* 0.14* 0.05* 0.10* −0.04* −0.01 0.04* 0.03* 0.16* 0.14* 1 (21) Number R&D employees 0.13* 0.14* 0.12* 0.34* 0.27* 0.39* 0.40* 0.37* 0.29* 0.21* 0.19* 0.07* 0.28* −0.04* −0.03* −0.01 0.03* 0.4* 0.21* 0.16* (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) (1) Failure 1 (2) Failure conception 0.82* 1 (3) Failure implementation 0.74* 0.46* 1 (4) R&D exploration 0.12* 0.14* 0.1* 1 (5) R&D exploration-1 0.09* 0.10* 0.08* 0.61* 1 (6) Explor × Explor-1 0.1* 0.12* 0.09* 0.83* 0.81* 1 (7) R&D exploration sq 0.11* 0.13* 0.09* 0.96* 0.62* 0.87* 1 (8) R&D exploitation 0.07* 0.1* 0.03* 0.11* 0.08* 0.08* 0.11* 1 (9) Knowledge breadth 0.12* 0.18* 0.09* 0.22* 0.15* 0.20* 0.22* 0.23* 1 (10) Knowledge depth 0.07* 0.11* 0.06* 0.14* 0.1* 0.14* 0.15* 0.15* 0.41* 1 (11) Other innovation exp 0.04* 0.05* 0.03* 0.06* 0.05* 0.09* 0.11* 0.08* 0.17* 0.14* 1 (12) Firm age 0.05* 0.04* 0.04* −0.05* −0.06* −0.06* −0.08* −0.05* 0.04* 0.01 −0.05* 1 (13) Group 0.04* 0.04* 0.05* 0.01 −0.01 0.01 0.01 −0.01 0.08* 0.03* 0.05* 0.11* 1 (14) Cost barriers −0.02 −0.02 −0.02 0.01 0.02 0.01 0.01 −0.01 −0.01 0.06* 0.01 −0.06* −0.13* 1 (15) Knowledge barriers 0.01 0.01 0.01 0.01 0.01 0.01 0.01 −0.01 −0.01 −0.01 −0.01 −0.01 −0.06* 0.12* 1 (16) Mkt barriers uncertainty 0.07* 0.06* 0.07* 0.03* 0.01 0.03* 0.03* 0.03* 0.04* 0.08* 0.03* −0.01 −0.06* 0.15* 0.09* 1 (17) Mkt barriers dominated 0.06* 0.06* 0.04* 0.06* 0.02 0.06* 0.07* 0.04* 0.05* 0.11* 0.03* −0.03* −0.08* 0.13* 0.12* 0.37* 1 (18) Number of employees 0.06* 0.07* 0.06* −0.03* −0.08* −0.05* −0.07* −0.04* 0.14* 0.06* 0.01 0.35* 0.55* −0.15* −0.06* −0.11* −0.1* 1 (19) Radical innovation 0.1* 0.14* 0.07* 0.15* 0.11* 0.14* 0.15* 0.21* 0.17* 0.12* 0.15* 0.01 0.04* −0.03* −0.01 0.03* −0.01 0.05* 1 (20) Organizational innovation 0.13* 0.17* 0.09* 0.09* 0.07* 0.1* 0.08* 0.09* 0.23* 0.12* 0.14* 0.05* 0.10* −0.04* −0.01 0.04* 0.03* 0.16* 0.14* 1 (21) Number R&D employees 0.13* 0.14* 0.12* 0.34* 0.27* 0.39* 0.40* 0.37* 0.29* 0.21* 0.19* 0.07* 0.28* −0.04* −0.03* −0.01 0.03* 0.4* 0.21* 0.16* Dummies for Pavitt (1984) four categories of industrial firms are not included for space reasons but correlation coefficient is never above 0.3. * P < 0.05. 3.2.4 Methods To estimate the factors that influence three different types of innovation failure events, we employ probit models based on panel data: P(Failure overallit=1| xit, zit, μi) =Φ(xit' β+zit' γ+μi) P(Failure conceptionit=1| xit, zit, μi) =Φ(xit' β+zit' γ+μi) P(Failure implementationit=1| xit, zit, μi) =Φ(xit' β+zit' γ+μi) where Φ(.) is the cumulative normal distribution function; xit' is the vector of our key explanatory variables for firm i at time t; zit' is the vector of controls for firm i at time t; and μi is the time invariant unobserved heterogeneity term. The models are estimated by employing a random effects specification. We prefer random effects to fixed effects because a large proportion of the firms in our sample is characterized by no within-variation in the relevant dependent variables,11 which would have significantly reduced the number of companies available for a fixed-effects estimation. We preferred a larger and more representative sample and, thus, implement only random effects.12 4. Results 4.1 Main results Table 4 presents our baseline results, examining the relation between current engagement in exploratory R&D and the probability of innovation failure—overall and in the conception and implementation phases. The coefficient of R&D exploration is always positive and highly significant, which provides evidence of the risky nature of exploration. Increasing the exploration orientation of R&D activities increases the overall risk of failure in innovation activities. Most significantly for our hypotheses, an increase in exploratory R&D activities increases the probability of failing in the conception and in the implementation phases of the innovation project life cycle as shown by the results in Table 4, Columns 2 and 3. This supports H1. The relation is relevant from an economic perspective: an increase of 1% in exploratory R&D increases the odds of failure in the conception (implementation) phase by around 12% (6%).13 Table 4. Innovation failures determinants: current exploration activities Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.044*** 0.063*** 0.033*** (0.010) (0.012) (0.011) R&D exploitation 0.026** 0.039*** 0.005 (0.011) (0.014) (0.012) Knowledge breadth 0.013 0.062*** 0.008 (0.012) (0.014) (0.012) Knowledge depth 0.019 0.037 0.016 (0.023) (0.026) (0.024) Other innovation expenditures 0.000 −0.002 −0.001 (0.014) (0.016) (0.015) Firm age 0.154** 0.115 0.127* (0.068) (0.077) (0.074) Group 0.042 0.023 0.137 (0.086) (0.098) (0.091) Cost barriers −0.081 −0.128 −0.105 (0.093) (0.108) (0.099) Knowledge barriers 0.233 0.332 0.315 (0.308) (0.346) (0.316) Market barriers—uncertainty 0.226*** 0.146 0.256*** (0.081) (0.093) (0.084) Market barriers—dominated 0.099 0.176* 0.026 (0.091) (0.104) (0.096) Number of employees 0.049 0.066 0.023 (0.040) (0.046) (0.043) Radical innovation 0.135** 0.269*** 0.105 (0.068) (0.076) (0.072) Org innovation 0.311*** 0.516*** 0.212*** (0.068) (0.079) (0.073) Number of R&D researchers 0.056 −0.003 0.107* (0.058) (0.063) (0.060) Wald Χ2 126.855 (18)*** 172.878 (18)*** 90.606 (18)*** Firm-period observations 2226 2226 2226 Firm observations 4191 4191 4191 Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.044*** 0.063*** 0.033*** (0.010) (0.012) (0.011) R&D exploitation 0.026** 0.039*** 0.005 (0.011) (0.014) (0.012) Knowledge breadth 0.013 0.062*** 0.008 (0.012) (0.014) (0.012) Knowledge depth 0.019 0.037 0.016 (0.023) (0.026) (0.024) Other innovation expenditures 0.000 −0.002 −0.001 (0.014) (0.016) (0.015) Firm age 0.154** 0.115 0.127* (0.068) (0.077) (0.074) Group 0.042 0.023 0.137 (0.086) (0.098) (0.091) Cost barriers −0.081 −0.128 −0.105 (0.093) (0.108) (0.099) Knowledge barriers 0.233 0.332 0.315 (0.308) (0.346) (0.316) Market barriers—uncertainty 0.226*** 0.146 0.256*** (0.081) (0.093) (0.084) Market barriers—dominated 0.099 0.176* 0.026 (0.091) (0.104) (0.096) Number of employees 0.049 0.066 0.023 (0.040) (0.046) (0.043) Radical innovation 0.135** 0.269*** 0.105 (0.068) (0.076) (0.072) Org innovation 0.311*** 0.516*** 0.212*** (0.068) (0.079) (0.073) Number of R&D researchers 0.056 −0.003 0.107* (0.058) (0.063) (0.060) Wald Χ2 126.855 (18)*** 172.878 (18)*** 90.606 (18)*** Firm-period observations 2226 2226 2226 Firm observations 4191 4191 4191 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Column 1, the probability to fail at the conception phase in Column 2 and the probability to fail at the implementation phase in Column 3. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels. * P < 0.10; ** P < 0.05; *** P < 0.01. Table 4. Innovation failures determinants: current exploration activities Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.044*** 0.063*** 0.033*** (0.010) (0.012) (0.011) R&D exploitation 0.026** 0.039*** 0.005 (0.011) (0.014) (0.012) Knowledge breadth 0.013 0.062*** 0.008 (0.012) (0.014) (0.012) Knowledge depth 0.019 0.037 0.016 (0.023) (0.026) (0.024) Other innovation expenditures 0.000 −0.002 −0.001 (0.014) (0.016) (0.015) Firm age 0.154** 0.115 0.127* (0.068) (0.077) (0.074) Group 0.042 0.023 0.137 (0.086) (0.098) (0.091) Cost barriers −0.081 −0.128 −0.105 (0.093) (0.108) (0.099) Knowledge barriers 0.233 0.332 0.315 (0.308) (0.346) (0.316) Market barriers—uncertainty 0.226*** 0.146 0.256*** (0.081) (0.093) (0.084) Market barriers—dominated 0.099 0.176* 0.026 (0.091) (0.104) (0.096) Number of employees 0.049 0.066 0.023 (0.040) (0.046) (0.043) Radical innovation 0.135** 0.269*** 0.105 (0.068) (0.076) (0.072) Org innovation 0.311*** 0.516*** 0.212*** (0.068) (0.079) (0.073) Number of R&D researchers 0.056 −0.003 0.107* (0.058) (0.063) (0.060) Wald Χ2 126.855 (18)*** 172.878 (18)*** 90.606 (18)*** Firm-period observations 2226 2226 2226 Firm observations 4191 4191 4191 Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.044*** 0.063*** 0.033*** (0.010) (0.012) (0.011) R&D exploitation 0.026** 0.039*** 0.005 (0.011) (0.014) (0.012) Knowledge breadth 0.013 0.062*** 0.008 (0.012) (0.014) (0.012) Knowledge depth 0.019 0.037 0.016 (0.023) (0.026) (0.024) Other innovation expenditures 0.000 −0.002 −0.001 (0.014) (0.016) (0.015) Firm age 0.154** 0.115 0.127* (0.068) (0.077) (0.074) Group 0.042 0.023 0.137 (0.086) (0.098) (0.091) Cost barriers −0.081 −0.128 −0.105 (0.093) (0.108) (0.099) Knowledge barriers 0.233 0.332 0.315 (0.308) (0.346) (0.316) Market barriers—uncertainty 0.226*** 0.146 0.256*** (0.081) (0.093) (0.084) Market barriers—dominated 0.099 0.176* 0.026 (0.091) (0.104) (0.096) Number of employees 0.049 0.066 0.023 (0.040) (0.046) (0.043) Radical innovation 0.135** 0.269*** 0.105 (0.068) (0.076) (0.072) Org innovation 0.311*** 0.516*** 0.212*** (0.068) (0.079) (0.073) Number of R&D researchers 0.056 −0.003 0.107* (0.058) (0.063) (0.060) Wald Χ2 126.855 (18)*** 172.878 (18)*** 90.606 (18)*** Firm-period observations 2226 2226 2226 Firm observations 4191 4191 4191 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Column 1, the probability to fail at the conception phase in Column 2 and the probability to fail at the implementation phase in Column 3. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels. * P < 0.10; ** P < 0.05; *** P < 0.01. H2 refers to the interaction between past exploration and current engagement in exploratory R&D. Table 5 shows that the value of previous R&D exploration activities is never significant per se (Columns 1, 3, and 5). Also, the results reported in Table 5 Column 2 point to the absence of a moderating effect (R&D exploration × R&D exploration-1 negative but not statistically significant) in the case of the overall probability of failure. Interestingly, if we split the probability of failure into its constituent parts (see Table 5 Columns 3 and 4 for conception and Columns 5 and 6 for implementation), we find that in the case of conception failure, there is a direct effect, captured by the coefficient of R&D exploration-1, which is positive and significant, and a moderating effect, captured by the interaction R&D exploration × R&D exploration-1, which is negative and significant (see Table 5 Column 4). The effect of the interaction between past experience and current exploration is small, but non-negligible: the effect of current exploration on innovation failure at the conception phase is decreased by 1.1% for a 1% increase in investment in past exploration. In contrast, we find no evidence of a negative interplay between past experience and current experimentation in the case of implementation failure, which reduces support for H2. We reflect on these different results in the discussion section. Table 5. Innovation failures determinants: interaction effects Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.039*** 0.049*** 0.059*** 0.086*** 0.028** 0.030* (0.011) (0.017) (0.013) (0.019) (0.012) (0.018) R&D exploration-1 0.013 0.023 0.009 0.037** 0.014 0.016 (0.011) (0.016) (0.013) (0.019) (0.012) (0.018) R&D exploration × R&D exploration-1 −0.002 −0.006** −0.001 (0.003) (0.003) (0.003) R&D exploitation 0.027** 0.026** 0.040*** 0.037*** 0.006 0.006 (0.012) (0.012) (0.014) (0.014) (0.012) (0.012) Knowledge breadth 0.013 0.013 0.062*** 0.061*** 0.007 0.007 (0.012) (0.012) (0.014) (0.014) (0.013) (0.013) Knowledge depth 0.019 0.019 0.037 0.038 0.017 0.017 (0.023) (0.023) (0.026) (0.025) (0.024) (0.024) Other innovation expenditures 0.000 0.001 −0.002 −0.000 −0.001 −0.001 (0.014) (0.014) (0.016) (0.016) (0.015) (0.015) Firm age 0.154** 0.154** 0.115 0.117 0.126* 0.126* (0.068) (0.068) (0.077) (0.077) (0.074) (0.074) Group 0.041 0.041 0.022 0.024 0.136 0.136 (0.086) (0.086) (0.098) (0.098) (0.091) (0.091) Cost barriers −0.083 −0.084 −0.129 −0.131 −0.107 −0.107 (0.093) (0.093) (0.108) (0.108) (0.099) (0.099) Knowledge barriers 0.231 0.235 0.328 0.339 0.314 0.315 (0.308) (0.308) (0.345) (0.346) (0.315) (0.315) Market barriers—uncertainty 0.227*** 0.227*** 0.147 0.147 0.258*** 0.258*** (0.081) (0.081) (0.093) (0.093) (0.084) (0.084) Market barriers—dominated 0.102 0.104 0.178* 0.182* 0.029 0.030 (0.091) (0.091) (0.104) (0.104) (0.097) (0.096) Number of employees 0.055 0.053 0.070 0.063 0.029 0.028 (0.041) (0.041) (0.047) (0.047) (0.043) (0.043) Radical innovation 0.135** 0.134** 0.269*** 0.268*** 0.105 0.104 (0.068) (0.068) (0.076) (0.076) (0.072) (0.072) Org innovation 0.308*** 0.308*** 0.513*** 0.515*** 0.209*** 0.209*** (0.068) (0.068) (0.079) (0.080) (0.073) (0.073) Number of R&D researchers 0.046 0.054 −0.009 0.012 0.098 0.099 (0.058) (0.059) (0.064) (0.064) (0.061) (0.062) Wald Χ2 127.516 (19)*** 127.76 (20)*** 172.709 (19)*** 175.43 (20)*** 90.861 (19)*** 90.96 (20)*** Firm-period observations 2226 2226 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 4191 4191 Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.039*** 0.049*** 0.059*** 0.086*** 0.028** 0.030* (0.011) (0.017) (0.013) (0.019) (0.012) (0.018) R&D exploration-1 0.013 0.023 0.009 0.037** 0.014 0.016 (0.011) (0.016) (0.013) (0.019) (0.012) (0.018) R&D exploration × R&D exploration-1 −0.002 −0.006** −0.001 (0.003) (0.003) (0.003) R&D exploitation 0.027** 0.026** 0.040*** 0.037*** 0.006 0.006 (0.012) (0.012) (0.014) (0.014) (0.012) (0.012) Knowledge breadth 0.013 0.013 0.062*** 0.061*** 0.007 0.007 (0.012) (0.012) (0.014) (0.014) (0.013) (0.013) Knowledge depth 0.019 0.019 0.037 0.038 0.017 0.017 (0.023) (0.023) (0.026) (0.025) (0.024) (0.024) Other innovation expenditures 0.000 0.001 −0.002 −0.000 −0.001 −0.001 (0.014) (0.014) (0.016) (0.016) (0.015) (0.015) Firm age 0.154** 0.154** 0.115 0.117 0.126* 0.126* (0.068) (0.068) (0.077) (0.077) (0.074) (0.074) Group 0.041 0.041 0.022 0.024 0.136 0.136 (0.086) (0.086) (0.098) (0.098) (0.091) (0.091) Cost barriers −0.083 −0.084 −0.129 −0.131 −0.107 −0.107 (0.093) (0.093) (0.108) (0.108) (0.099) (0.099) Knowledge barriers 0.231 0.235 0.328 0.339 0.314 0.315 (0.308) (0.308) (0.345) (0.346) (0.315) (0.315) Market barriers—uncertainty 0.227*** 0.227*** 0.147 0.147 0.258*** 0.258*** (0.081) (0.081) (0.093) (0.093) (0.084) (0.084) Market barriers—dominated 0.102 0.104 0.178* 0.182* 0.029 0.030 (0.091) (0.091) (0.104) (0.104) (0.097) (0.096) Number of employees 0.055 0.053 0.070 0.063 0.029 0.028 (0.041) (0.041) (0.047) (0.047) (0.043) (0.043) Radical innovation 0.135** 0.134** 0.269*** 0.268*** 0.105 0.104 (0.068) (0.068) (0.076) (0.076) (0.072) (0.072) Org innovation 0.308*** 0.308*** 0.513*** 0.515*** 0.209*** 0.209*** (0.068) (0.068) (0.079) (0.080) (0.073) (0.073) Number of R&D researchers 0.046 0.054 −0.009 0.012 0.098 0.099 (0.058) (0.059) (0.064) (0.064) (0.061) (0.062) Wald Χ2 127.516 (19)*** 127.76 (20)*** 172.709 (19)*** 175.43 (20)*** 90.861 (19)*** 90.96 (20)*** Firm-period observations 2226 2226 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 4191 4191 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Columns 1 and 2, the probability to fail at the conception phase in Columns 3 and 4, and the probability to fail at the implementation phase in Columns 5 and 6. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels where the start date of the panel refers to the dependent variable (i.e., if t = 2007–2009, so t-1 = 2004–2006). * P < 0.10; ** P < 0.05; *** P < 0.01. Table 5. Innovation failures determinants: interaction effects Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.039*** 0.049*** 0.059*** 0.086*** 0.028** 0.030* (0.011) (0.017) (0.013) (0.019) (0.012) (0.018) R&D exploration-1 0.013 0.023 0.009 0.037** 0.014 0.016 (0.011) (0.016) (0.013) (0.019) (0.012) (0.018) R&D exploration × R&D exploration-1 −0.002 −0.006** −0.001 (0.003) (0.003) (0.003) R&D exploitation 0.027** 0.026** 0.040*** 0.037*** 0.006 0.006 (0.012) (0.012) (0.014) (0.014) (0.012) (0.012) Knowledge breadth 0.013 0.013 0.062*** 0.061*** 0.007 0.007 (0.012) (0.012) (0.014) (0.014) (0.013) (0.013) Knowledge depth 0.019 0.019 0.037 0.038 0.017 0.017 (0.023) (0.023) (0.026) (0.025) (0.024) (0.024) Other innovation expenditures 0.000 0.001 −0.002 −0.000 −0.001 −0.001 (0.014) (0.014) (0.016) (0.016) (0.015) (0.015) Firm age 0.154** 0.154** 0.115 0.117 0.126* 0.126* (0.068) (0.068) (0.077) (0.077) (0.074) (0.074) Group 0.041 0.041 0.022 0.024 0.136 0.136 (0.086) (0.086) (0.098) (0.098) (0.091) (0.091) Cost barriers −0.083 −0.084 −0.129 −0.131 −0.107 −0.107 (0.093) (0.093) (0.108) (0.108) (0.099) (0.099) Knowledge barriers 0.231 0.235 0.328 0.339 0.314 0.315 (0.308) (0.308) (0.345) (0.346) (0.315) (0.315) Market barriers—uncertainty 0.227*** 0.227*** 0.147 0.147 0.258*** 0.258*** (0.081) (0.081) (0.093) (0.093) (0.084) (0.084) Market barriers—dominated 0.102 0.104 0.178* 0.182* 0.029 0.030 (0.091) (0.091) (0.104) (0.104) (0.097) (0.096) Number of employees 0.055 0.053 0.070 0.063 0.029 0.028 (0.041) (0.041) (0.047) (0.047) (0.043) (0.043) Radical innovation 0.135** 0.134** 0.269*** 0.268*** 0.105 0.104 (0.068) (0.068) (0.076) (0.076) (0.072) (0.072) Org innovation 0.308*** 0.308*** 0.513*** 0.515*** 0.209*** 0.209*** (0.068) (0.068) (0.079) (0.080) (0.073) (0.073) Number of R&D researchers 0.046 0.054 −0.009 0.012 0.098 0.099 (0.058) (0.059) (0.064) (0.064) (0.061) (0.062) Wald Χ2 127.516 (19)*** 127.76 (20)*** 172.709 (19)*** 175.43 (20)*** 90.861 (19)*** 90.96 (20)*** Firm-period observations 2226 2226 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 4191 4191 Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.039*** 0.049*** 0.059*** 0.086*** 0.028** 0.030* (0.011) (0.017) (0.013) (0.019) (0.012) (0.018) R&D exploration-1 0.013 0.023 0.009 0.037** 0.014 0.016 (0.011) (0.016) (0.013) (0.019) (0.012) (0.018) R&D exploration × R&D exploration-1 −0.002 −0.006** −0.001 (0.003) (0.003) (0.003) R&D exploitation 0.027** 0.026** 0.040*** 0.037*** 0.006 0.006 (0.012) (0.012) (0.014) (0.014) (0.012) (0.012) Knowledge breadth 0.013 0.013 0.062*** 0.061*** 0.007 0.007 (0.012) (0.012) (0.014) (0.014) (0.013) (0.013) Knowledge depth 0.019 0.019 0.037 0.038 0.017 0.017 (0.023) (0.023) (0.026) (0.025) (0.024) (0.024) Other innovation expenditures 0.000 0.001 −0.002 −0.000 −0.001 −0.001 (0.014) (0.014) (0.016) (0.016) (0.015) (0.015) Firm age 0.154** 0.154** 0.115 0.117 0.126* 0.126* (0.068) (0.068) (0.077) (0.077) (0.074) (0.074) Group 0.041 0.041 0.022 0.024 0.136 0.136 (0.086) (0.086) (0.098) (0.098) (0.091) (0.091) Cost barriers −0.083 −0.084 −0.129 −0.131 −0.107 −0.107 (0.093) (0.093) (0.108) (0.108) (0.099) (0.099) Knowledge barriers 0.231 0.235 0.328 0.339 0.314 0.315 (0.308) (0.308) (0.345) (0.346) (0.315) (0.315) Market barriers—uncertainty 0.227*** 0.227*** 0.147 0.147 0.258*** 0.258*** (0.081) (0.081) (0.093) (0.093) (0.084) (0.084) Market barriers—dominated 0.102 0.104 0.178* 0.182* 0.029 0.030 (0.091) (0.091) (0.104) (0.104) (0.097) (0.096) Number of employees 0.055 0.053 0.070 0.063 0.029 0.028 (0.041) (0.041) (0.047) (0.047) (0.043) (0.043) Radical innovation 0.135** 0.134** 0.269*** 0.268*** 0.105 0.104 (0.068) (0.068) (0.076) (0.076) (0.072) (0.072) Org innovation 0.308*** 0.308*** 0.513*** 0.515*** 0.209*** 0.209*** (0.068) (0.068) (0.079) (0.080) (0.073) (0.073) Number of R&D researchers 0.046 0.054 −0.009 0.012 0.098 0.099 (0.058) (0.059) (0.064) (0.064) (0.061) (0.062) Wald Χ2 127.516 (19)*** 127.76 (20)*** 172.709 (19)*** 175.43 (20)*** 90.861 (19)*** 90.96 (20)*** Firm-period observations 2226 2226 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 4191 4191 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Columns 1 and 2, the probability to fail at the conception phase in Columns 3 and 4, and the probability to fail at the implementation phase in Columns 5 and 6. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels where the start date of the panel refers to the dependent variable (i.e., if t = 2007–2009, so t-1 = 2004–2006). * P < 0.10; ** P < 0.05; *** P < 0.01. The results related to H3 are reported in Table 6. The sign and significance of the coefficients of R&D exploration (positive and significant) and R&D exploration sq (negative and significant) suggest the presence of an inverted U-shaped relationship, which supports H3.14 Notably, the positive relationship between exploration and failure at conception (implementation) is decreased by 3.29% (2.09%) for every additional 1% increase in current exploration. We can conclude that the intensity of engagement in exploratory R&D activity attenuates the initial positive relation between current exploratory R&D and failure. This beneficial effect occurs in the same 3-year period. Hence, our results seem to confirm that firms can learn how to reduce failures from the performance of current and novel tasks if they have accumulated sufficient information from ongoing activities. Overall, our evidence points to the presence of a learning effect from exploratory R&D along the entire innovation project life cycle. Table 6. Innovation failures determinants: curvilinear effects Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.146*** 0.206*** 0.121*** (0.040) (0.046) (0.044) R&D exploration sq −0.014*** −0.019*** −0.012** (0.005) (0.006) (0.006) R&D exploitation 0.021* 0.032** 0.001 (0.012) (0.014) (0.012) Knowledge breadth 0.014 0.063*** 0.008 (0.012) (0.014) (0.012) Knowledge depth 0.019 0.037 0.017 (0.023) (0.025) (0.024) Other innovation expenditures 0.006 0.006 0.004 (0.014) (0.016) (0.015) Firm age 0.146** 0.105 0.120 (0.068) (0.077) (0.073) Group 0.054 0.039 0.146 (0.086) (0.097) (0.090) Cost barriers −0.086 −0.136 −0.109 (0.092) (0.107) (0.098) Knowledge barriers 0.230 0.326 0.311 (0.306) (0.343) (0.313) Market barriers—uncertainty 0.218*** 0.133 0.247*** (0.081) (0.092) (0.084) Market barriers—dominated 0.102 0.179* 0.028 (0.090) (0.103) (0.096) Number of employees 0.018 0.020 −0.004 (0.042) (0.049) (0.045) Radical innovation 0.135** 0.269*** 0.106 (0.067) (0.075) (0.072) Org innovation 0.308*** 0.512*** 0.208*** (0.068) (0.079) (0.072) Number of R&D researchers 0.109* 0.073 0.153** (0.061) (0.068) (0.065) Wald Χ2 133.07 (19)*** 181.62 (19)*** 93.71 (19)*** Firm-period observations 2226 2226 2226 Firm observations 4191 4191 4191 Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.146*** 0.206*** 0.121*** (0.040) (0.046) (0.044) R&D exploration sq −0.014*** −0.019*** −0.012** (0.005) (0.006) (0.006) R&D exploitation 0.021* 0.032** 0.001 (0.012) (0.014) (0.012) Knowledge breadth 0.014 0.063*** 0.008 (0.012) (0.014) (0.012) Knowledge depth 0.019 0.037 0.017 (0.023) (0.025) (0.024) Other innovation expenditures 0.006 0.006 0.004 (0.014) (0.016) (0.015) Firm age 0.146** 0.105 0.120 (0.068) (0.077) (0.073) Group 0.054 0.039 0.146 (0.086) (0.097) (0.090) Cost barriers −0.086 −0.136 −0.109 (0.092) (0.107) (0.098) Knowledge barriers 0.230 0.326 0.311 (0.306) (0.343) (0.313) Market barriers—uncertainty 0.218*** 0.133 0.247*** (0.081) (0.092) (0.084) Market barriers—dominated 0.102 0.179* 0.028 (0.090) (0.103) (0.096) Number of employees 0.018 0.020 −0.004 (0.042) (0.049) (0.045) Radical innovation 0.135** 0.269*** 0.106 (0.067) (0.075) (0.072) Org innovation 0.308*** 0.512*** 0.208*** (0.068) (0.079) (0.072) Number of R&D researchers 0.109* 0.073 0.153** (0.061) (0.068) (0.065) Wald Χ2 133.07 (19)*** 181.62 (19)*** 93.71 (19)*** Firm-period observations 2226 2226 2226 Firm observations 4191 4191 4191 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Column 1, the probability to fail at the conception phase in Column 2, and the probability to fail at the implementation phase in Column 3. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels. * P < 0.10; ** P < 0.05; *** P < 0.01. Table 6. Innovation failures determinants: curvilinear effects Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.146*** 0.206*** 0.121*** (0.040) (0.046) (0.044) R&D exploration sq −0.014*** −0.019*** −0.012** (0.005) (0.006) (0.006) R&D exploitation 0.021* 0.032** 0.001 (0.012) (0.014) (0.012) Knowledge breadth 0.014 0.063*** 0.008 (0.012) (0.014) (0.012) Knowledge depth 0.019 0.037 0.017 (0.023) (0.025) (0.024) Other innovation expenditures 0.006 0.006 0.004 (0.014) (0.016) (0.015) Firm age 0.146** 0.105 0.120 (0.068) (0.077) (0.073) Group 0.054 0.039 0.146 (0.086) (0.097) (0.090) Cost barriers −0.086 −0.136 −0.109 (0.092) (0.107) (0.098) Knowledge barriers 0.230 0.326 0.311 (0.306) (0.343) (0.313) Market barriers—uncertainty 0.218*** 0.133 0.247*** (0.081) (0.092) (0.084) Market barriers—dominated 0.102 0.179* 0.028 (0.090) (0.103) (0.096) Number of employees 0.018 0.020 −0.004 (0.042) (0.049) (0.045) Radical innovation 0.135** 0.269*** 0.106 (0.067) (0.075) (0.072) Org innovation 0.308*** 0.512*** 0.208*** (0.068) (0.079) (0.072) Number of R&D researchers 0.109* 0.073 0.153** (0.061) (0.068) (0.065) Wald Χ2 133.07 (19)*** 181.62 (19)*** 93.71 (19)*** Firm-period observations 2226 2226 2226 Firm observations 4191 4191 4191 Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.146*** 0.206*** 0.121*** (0.040) (0.046) (0.044) R&D exploration sq −0.014*** −0.019*** −0.012** (0.005) (0.006) (0.006) R&D exploitation 0.021* 0.032** 0.001 (0.012) (0.014) (0.012) Knowledge breadth 0.014 0.063*** 0.008 (0.012) (0.014) (0.012) Knowledge depth 0.019 0.037 0.017 (0.023) (0.025) (0.024) Other innovation expenditures 0.006 0.006 0.004 (0.014) (0.016) (0.015) Firm age 0.146** 0.105 0.120 (0.068) (0.077) (0.073) Group 0.054 0.039 0.146 (0.086) (0.097) (0.090) Cost barriers −0.086 −0.136 −0.109 (0.092) (0.107) (0.098) Knowledge barriers 0.230 0.326 0.311 (0.306) (0.343) (0.313) Market barriers—uncertainty 0.218*** 0.133 0.247*** (0.081) (0.092) (0.084) Market barriers—dominated 0.102 0.179* 0.028 (0.090) (0.103) (0.096) Number of employees 0.018 0.020 −0.004 (0.042) (0.049) (0.045) Radical innovation 0.135** 0.269*** 0.106 (0.067) (0.075) (0.072) Org innovation 0.308*** 0.512*** 0.208*** (0.068) (0.079) (0.072) Number of R&D researchers 0.109* 0.073 0.153** (0.061) (0.068) (0.065) Wald Χ2 133.07 (19)*** 181.62 (19)*** 93.71 (19)*** Firm-period observations 2226 2226 2226 Firm observations 4191 4191 4191 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Column 1, the probability to fail at the conception phase in Column 2, and the probability to fail at the implementation phase in Column 3. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels. * P < 0.10; ** P < 0.05; *** P < 0.01. 4.2 Robustness checks We check the robustness of our results to three problems. First, we consider an alternative measure for past experience in exploration activities: R&D exploration stock-1, which reflects the cumulated stock of expenditure on exploratory R&D in the previous period, which accounts also for obsolescent information on past engagement in exploratory R&D. Table A2 shows that the results are confirmed if we use an alternative measure of past experience, cumulated stock of exploration in R&D (R&D exploration stock-1 and R&D exploration × R&D exploration stock-1).15 Second, there might be problems related to the firm-level nature of our data. To investigate the determinants of innovation failure, the ideal analytical level is the project, with information on project resources (e.g., different types of R&D), obstacles, and termination or success. Although we do not have project-level information, we can check the robustness of our results to this problem by estimating our models for the subsample of small companies (less than 50 employees based on the European Commission definition).16 The rationale is that smaller sized companies, due to time and resource constraints and the absence of a formal R&D department, are more likely to have fewer (ideally no more than 1) innovation projects per period. The results reported in Table A3, Table A4, and Table A5 mostly support our main results. We also consider the interdependence between conception and implementation phase failures in a bivariate setting, since the likelihood of these types of failure were found to be correlated (Pearson and tetrachoric correlation coefficients are 0.46 and 0.7, respectively). In the absence of well-established models to control for correlation among the error terms in a nonlinear panel setting, we run a set of “pooled” bivariate probit models and control for clustering of the error terms for each firm. To further corroborate our results, we build upon Roodman’s (2011) recent work on fully observed recursive mixed-process models and estimate a probit Seemingly Unrelated Regression (SURE). Tables A6 and A7 provide evidence of the robustness of our main results in both bivariate and probit SURE specifications. 5. Discussion and conclusions This article investigated the dual effects of exploration in R&D: while exploration increases firms’ exposure to disruption and abandonment of innovation activities, it provides potential learning opportunities to mitigate the risk of innovation failure. Investigation and disclosure of the learning returns from exploration, and understanding the underlying working principles, are the core contributions of this study. Our empirical analysis confirms that exploratory R&D increases the firm’s exposure to innovation failure in both the conception and implementation phases of the innovation process. This supports the argument that engagement in exploratory research is associated with higher exposure to innovation failure. However, our findings also qualify this assertion in three critical aspects. First, we highlight the presence of learning opportunities from exploratory research with regard to the capacity of firms to mitigate innovation failure. Second, we uncover two mechanisms through which learning from exploration operates: (i) the interplay between past exploration experience and current exploration practice; and (ii) the enhanced inferential capacity from current exploration efforts. Third, we show that learning opportunities differ in distinct phases of the innovation process (i.e., the conception and implementation phases). These contributions have implications for both theory and practice, which we discuss below. This work contributes to organizational learning theory by examining learning processes from firms’ exploratory research. Our results provide support for the presence of a learning process from exploration, which triggers a positive change in organizational performance by reducing innovation failure. The main conceptual contributions of the study can be summarized as follows. First, this study identifies two potential sources of learning from exploration: (i) experiential-based learning from past exploratory research; and (ii) inferential-based learning from current exploratory research. The former is based on the learning opportunities derived from accumulated experience in exploration and is associated with the formation of routines and standard procedures that enhance functional performance associated with experimentation. The latter is based on learning opportunities derived from ongoing exploration efforts and is associated with improved interpretation of ill-defined problems and enhanced inference in the presence of causal ambiguity. The study provides empirical support for the claim that these sources of learning can contribute to mitigate innovation failure, and provides a new lens to understand how exploratory R&D can contribute positively to firms’ innovation performance. Second, our analysis highlights the intricacies associated with the capacity to realize the learning opportunities from exploration, especially in relation to the role of the temporal dimensions of learning from exploratory research activities (Argote and Miron-Spektor, 2011). On the one hand, we find that accumulated experience from past exploration activities contributes to reducing the probability of innovation failure. However, our results run contrary to the standard learning curve which predicts that performance outcomes are an automatic by-product of accumulated experience (Zangwill and Kantor, 1998). We find that past experience, per se, does not reduce the risk of abandonment of innovation projects. Our results show that accumulated experience in exploration activities has a beneficial effect on the link between current exploration efforts and innovation failure: it reduces the firm’s exposure to innovation failure through the interplay with current R&D exploration activity. We suggest that past experience in exploration provides critical learning for current activities by helping to navigate the relevant search space for exploration and building surveillance systems to identify deviations from established standards. In addition, when applied to current exploration, previously elaborated information can be refined and increase understanding about the causes of problems and their possible solutions. This result resonates with the “generative dance” proposed by Cook and Brown (1999) between knowledge embedded in organizational routines and knowledge gained in the context of practice, as a source of organizational learning and innovation. It should be noted also that the moderating role of past exploration experience is significant only for reducing failure in the conception phase of the innovation process, not in the implementation phase. This is likely a consequence of the distinct challenges posed by exploration activities in the two phases of the innovation process. The conception phase is generally characterized by well-structured experimentation and standards, which exploit the analytical skills and capacity to interpret causal attribution and avoid inferential errors, of R&D unit researchers. The complementarities between past and current exploration experience seem to be less effective at reducing innovation failure in the implementation stages of the innovation process. Here, context specificities associated with the scaling up of the manufacturing processes and the generation of new working prototypes seem to benefit less from the interaction between accumulated experience and current exploration efforts. On the other hand, we find evidence of a curvilinear relationship between current exploration and innovation failure. In particular, we found support for an inverted U-shaped relationship between current investment in exploration activity and the probability of innovation failure in the conception and implementation phases of the innovation process. We found that, initially, exploration increases the chances of experiencing failure in the conception and implementation phases of the innovation process since generally it involves committing resources to search activities beyond current competencies and expertise. However, we also observed learning returns from exploratory R&D, measured by a reduced probability of failure with increasing levels of investment in current exploration activities. We argue that this source of learning from ongoing exploration efforts is associated with the firm’s improved capacity to interpret ill-defined problems, based on a better understanding of the causal relation mechanisms underlying the phenomenon being explored. This nonlinear, inverted-U-shaped relationship between current exploration efforts and the probability of failure, reflects the complexity of the processes involved in exploration activities. These include difficulties related to interpreting unstructured information, attribution errors due to the absence of a precise understanding of the underlying causal relationships and inferential errors based on a restricted sample of experimental results. By identifying these beneficial learning effects, we have highlighted an additional economic payoff from the exploratory component of R&D (David et al., 1992): fewer wasted resources resulting from the lower propensity to abandon innovation projects. Our findings provide valuable insights for innovation management. Although firms tend to prefer development-based exploitation to research-based exploration (Arora et al., 2015), we find that exploration offers two sources of learning that have the potential to reduce the risk of a failed innovation project. These sources emerge after a certain level of exploration effort and when current efforts are combined with accumulated experience in exploration activities. Hence, planning periods of peak engagement in exploratory activities, which previous research shows have a positive effect on firm performance (Mudambi and Swift, 2011, 2014), might be beneficial for reducing exposure to innovation failure. However, “compartmentalized” periods of sustained exploration could hamper the interaction between ongoing and past exploration. Our results point to the need for persistent commitment to exploration over time, avoiding major discontinuities to capitalize on retention, integration, and reuse of previously acquired knowledge in ongoing efforts. The extent to which firms are able to connect ongoing exploration and past experience is essential to conceptualize problems and to scan, evaluate, and implement reusable ideas—especially in the presence of time and resource constraints (Cook and Brown, 1999; Majchrzak et al., 2004). Moreover, our results provide finer-grained results on the interplay between accumulated experience in exploratory R&D and current exploration, showing that this interaction is more beneficial in the conception than in the implementation phases. The capacity to learn from exploration and act upon unsuccessful innovation activities is greater in the conception phase, given that both transactive memory (Argote and Miron-Spektor, 2011) and related variation (Schilling et al., 2003) enable larger learning opportunities in settings that provide better conditions for controlled experimentation and structured planning, and greater flexibility in terms of experimentation alternatives due to the lower opportunity costs of failure. Hence, it is in the early stage of a project life cycle that firms can maximize the reliance on previously developed standards and information. This beneficial connection between experience and current practice is reduced in the implementation phase when context specificities are more pronounced. Additionally, sources of organizational rigidity such as entrapment situations (Balachandra et al., 1996; Jani, 2011) or escalation of commitments (Staw and Ross, 1987; Schmidt and Calantone, 1998; Jani, 2011), which are particularly likely to emerge in the implementation phase, could further weaken the connection between prior experience and current exploration activities. The article has some limitations that suggest avenues for future research. First, our definition of innovation failure forces us to measure it as a binary variable (whether the focal firm abandoned an innovation project or not in the period of reference). Providing a measure of the intensity of innovation failure at the firm level would enrich the analysis by showing the relative importance of innovation failure to firms experiencing different levels of research exploration efforts. Second, we rely on data from one country only, i.e., Spain. Future work should extend our analysis to a range of countries to allow the results obtained to be generalized. Third, although our longitudinal analysis tries to minimize problems of omitted variables bias and unobserved heterogeneity, the absence of a pure experimental setting (or instrumental variables approach) suggests caution in interpreting the results in a causal way. Finally, future work could investigate the exploration–failure relation using project-level data, which would allow greater scrutiny of the mechanisms related to knowledge retention, codification, and integration that might be at the basis of the beneficial interplay between past experience and current exploration. Despite these limitations, we believe that the insights from our study could serve as a guide and foundation for future work investigating the important role of firms’ exploration strategies for reducing innovation failure and achieving sustained competitive advantage. Footnotes 1 PITEC data for the different waves are available at: http://icono.fecyt.es/PITEC/Paginas/descarga_bbdd.aspx (last accessed April 2017). 2 See for instance: Molero and García, 2008; Vega-Jurado et al., 2009; Santamaría et al., 2012; Trigo and Vence, 2012; Audretsch et al., 2014; Barge-Gil and López, 2014; Busom et al., 2014; Afcha and García-Quevedo, 2016; Marzucchi and Montresor, 2017. A list of PITEC-based publications is available at: https://icono.fecyt.es/PITEC/Documents/2016/Utilizaci%C3%B3n%20del%20PITEC%20(Nov.%202016).pdf (last accessed July 2017). 3 Note that the firms in our sample are likely to suffer innovation failures not necessarily related only to intra-mural R&D, which should mitigate any sample selection problems in the construction of our sample. Table A1 shows the non-negligible proportion of companies that experience all three types of failure (overall, conception, and implementation) and engage in different types of innovation-related activities: (i) only intramural R&D; (ii) only other innovation-related activities (i.e., non-intramural R&D); and (iii) both intramural R&D and other innovation-related activities. 4 See https://icono.fecyt.es/PITEC/Documents/2014/cuestionario12.pdf (translated from Spanish, last accessed April 2017). 5 Real exploration expenditure is nominal exploration expenditure deflated by the consumer price index for the corresponding period. 6 Due to missing information for some relevant independent variables, such as organizational innovation, we could not include this time period, 2004–2006, in the estimation of our baseline model. 7 PITEC data include information on the following sources of information for innovation: suppliers; customers; competitors; consultants and private R&D institutes; universities; public research institutes; technological transfer offices; conferences, trade fairs, exhibitions; scientific journals and trade/technical publications; professional and industry associations. 8 Organizational innovation takes the value 1 if the firm replied “yes” to any of the following questions: (i) “Has the firm introduced new business practices in the organization of work or company procedures?”; (ii) “Has the firm introduced new methods for the organization of workplaces with the aim of distribution of responsibilities and decision-making?”; and (iii) “Has the firm introduced new methods of managing external relations with other companies or public institutions?” 9 Each firm was assigned to a unique Pavitt category according to its two-digit industry class (NACE rev.2 classification) following the Revised Pavitt taxonomy in Bogliacino and Pianta (2016). We estimated our empirical models using the set of two-digit industry dummies. Although the results hold, the test of joint significance of industry dummies is always far from significant. For this reason, we prefer to rely on broader (but generally jointly significant) industry controls based on Pavitt categories. 10 As the correlation table shows (Table 3), the correlations between our focal variables (R&D exploration, R&D exploration-1, and R&D exploration sq) are high and, in some cases, above an acceptable threshold (0.7). We ran some additional diagnostic tests to check for multicollinearity problems. Notably, the condition index for the three model specifications adopted in the empirical part of the paper ranges between 28 and 31: all values are close to the threshold of 30 (Hair et al., 1998: 220). Also, the Theil R2 multicollinearity effect for the three models ranges between 0.01 and 0.08, which is well below the value indicating multicollinearity, i.e., 1 (Theil, 1971). As further corroboration, we followed Cronbach (1987) and subtracted the mean from our core explanatory variables and then re ran our main estimates. The results confirmed the estimates obtained using the untransformed variables and reported in the results section. 11 For example, 55% of the companies that experienced failure in the conception phase in period t faced same types of failure in period t + 1. 12 We also performed Hausman and Mundlak tests (Wooldridge, 2010) to check whether a fixed effects specification was preferred to a random effects specification. Both tests in all model specifications (Tables 4–6) cannot be rejected at the standard confidence levels, confirming our choice of random effects. 13 Interpreting the coefficients in the nonlinear models is difficult, especially in relation to the squared and interaction effects. Recent studies using cross-sectional data suggest graphing the marginal effects to obtain a better understanding of the point in the distribution when they become significant (Ai and Norton, 2003; Zelner, 2009; Greene, 2010). In our case, the longitudinal nature of our empirical models adds complexity. Specifically, the presence of an unobserved heterogeneity term does not allow application of these graphing approaches without strong assumptions concerning the behavior of the heterogeneity term. To overcome this problem, we adopt the option suggested by Buis (2010) and interpret the coefficients as odds ratios. The main advantage of interpreting interaction and curvilinear effects as odds ratios is that this interpretation (unlike use of marginal effects) does not depend on the value of the independent variable being considered. 14 When comparing the unconstrained with the constrained specification for the three different specifications, we found the difference between the log-likelihoods of the constrained and unconstrained models was always statistically significant [respectively Χ2(1) = 7.22, P-value = 0.0072; Χ2(1) = 11.04, P-value = 0.0009; and Χ2(1) = 4.72, P-value = 0.0298]. Thus, we conclude that the less restrictive model (the one with the squared term included) fits the data significantly better than the more restrictive model (our baseline model with only the linear term). Furthermore, the presence of an inverted U-shaped effect is confirmed by the test suggested in Lind and Mehlum (2010): we reject the null hypothesis of a monotonic or U-shaped relationship at the 5% significance level, in all the three specifications. 15 The stock variable is computed using the perpetual inventory method based on the following formulas: Explor stockt= Explor stockt-11-δ+Explort and Explor stockt0= Explort0/(g+δ). A 15% depreciation rate is assumed. As a robustness check, a range of depreciation rates (from 15% to 75% in intervals of 20 points) was used drawing on insights from the relevant literature on R&D stocks (Hall, 1993) and organizational learning (Argote et al., 1990; Baum and Ingram, 1998). The results are robust to the different specifications adopted. As for the pre-sample growth rate g, we assume it to be equal to the average growth rate of exploration at the two-digit industry level (NACE rev.2 classification) in the period 2004–2008. We further tested the robustness of H2 by controlling for persistence in organizational commitment to exploratory R&D by including a third variable, which takes the value 1 if the firm invested in exploration in all 3 years in the reference period and 0 otherwise. The results are robust to this further specification and are available from the authors upon request. 16 European Commission, “Commission Recommendation of 6 May 2003 concerning the definition of micro, small and medium-sized enterprises (Text with EEA relevance) (notified under document number C(2003) 1422)”, available at: http://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32003H0361 (accessed April 2017). References Afcha S. , García-Quevedo J. ( 2016 ), ‘ The impact of R&D subsidies on R&D employment composition ,’ Industrial and Corporate Change , 25 ( 6 ), 955 – 975 . Ai C. , Norton E. C. ( 2003 ), ‘ Interaction terms in logit and probit models ,’ Economics Letters , 80 ( 1 ), 123 – 129 . Google Scholar CrossRef Search ADS Antonioli D. , Marzucchi A. , Savona M. ( 2017 ), ‘ Pain shared, pain halved? Cooperation as a coping strategy for innovation barriers ,’ The Journal of Technology Transfer , 42 ( 4 ), 841 – 864 . doi: 10.1007/s10961-016-9545-9. Google Scholar CrossRef Search ADS Argote L. , Beckman S. L. , Epple D. ( 1990 ), ‘ The persistence and transfer of learning in industrial settings ,’ Management Science , 36 ( 2 ), 140 – 154 . Google Scholar CrossRef Search ADS Argote L. , Ingram P. ( 2000 ), ‘ Knowledge transfer: a basis for competitive advantage in firms ,’ Organizational Behavior and Human Decision Processes , 82 ( 1 ), 150 – 169 . Google Scholar CrossRef Search ADS Argote L. , Miron-Spektor E. ( 2011 ), ‘ Organizational learning: from experience to knowledge ,’ Organization Science , 22 ( 5 ), 1123 – 1137 . Google Scholar CrossRef Search ADS Argyris C. , Schön D. ( 1978 ), Organizational Learning: A Theory of Action Perspective , Addison-Wesley Publishing Company : Reading, MA . Arora A. , Belenzon S. , Patacconi A. ( 2015 ), ‘Killing the golden goose? The decline of science in corporate R&D,’ National Bureau of Economic Research No. w20902. Audretsch D. B. , Segarra A. , Teruel M. ( 2014 ), ‘ Why don't all young firms invest in R&D? ,’ Small Business Economics , 43 ( 4 ), 751 – 766 . Google Scholar CrossRef Search ADS Balachandra R. , Brockhoff K. , Pearson A. W. ( 1996 ), ‘ R&D project termination decisions: processes, communication, and personnel changes ,’ Journal of Product Innovation Management , 13 ( 3 ), 245 – 256 . Google Scholar CrossRef Search ADS Barge-Gil A. , López A. ( 2014 ), ‘ R&D determinants: accounting for the differences between research and development ,’ Research Policy , 43 ( 9 ), 1634 – 1648 . Google Scholar CrossRef Search ADS Barge-Gil A. , López A. ( 2015 ), ‘ R versus D: estimating the differentiated effect of research and development on innovation results ,’ Industrial and Corporate Change , 24 ( 1 ), 93 – 129 . Google Scholar CrossRef Search ADS Baum J. A. , Ingram P. ( 1998 ), ‘ Survival-enhancing learning in the Manhattan hotel industry, 1898–1980 ,’ Management Science , 44 ( 7 ), 996 – 1016 . Google Scholar CrossRef Search ADS Benkard C. L. ( 2000 ), ‘ Learning and forgetting: the dynamics of aircraft production ,’ American Economic Review , 90 ( 4 ), 1034 – 1054 . Google Scholar CrossRef Search ADS Blanchard P. , Huiban J. P. , Musolesi A. , Sevestre P. ( 2013 ), ‘ Where there is a will, there is a way? Assessing the impact of obstacles to innovation ,’ Industrial and Corporate Change , 22 ( 3 ), 679 – 710 . Google Scholar CrossRef Search ADS Bogliacino F. , Pianta M. ( 2016 ), ‘ The Pavitt Taxonomy, revisited. Patterns of innovation in manufacturing and services ,’ Economia Politica , 33 ( 2 ), 153 – 158 . Google Scholar CrossRef Search ADS Buis M. L. ( 2010 ), ‘ Stata tip 87: interpretation of interactions in non-linear models ,’ The Stata Journal , 10 ( 2 ), 305 – 308 . Busom I. , Corchuelo B. , Martínez Ros E. ( 2014 ), ‘ Tax incentives or subsidies for business R&D? ,’ Small Business Economics , 43 ( 3 ), 571 – 598 . Google Scholar CrossRef Search ADS Cannon M. D. , Edmondson A. C. ( 2005 ), ‘ Failing to learn and learning to fail (intelligently): how great organizations put failure to work to innovate and improve ,’ Long Range Planning , 38 ( 3 ), 299 – 319 . Google Scholar CrossRef Search ADS Chesbrough H. ( 2003 ), Open Innovation: The New Imperative for Creating and Profiting from Technology . Harvard Business Press : Boston, MA . Chiva R. , Grandío A. , Alegre J. ( 2010 ), ‘ Adaptive and generative learning: implications from complexity theories ,’ International Journal of Management Reviews , 12 ( 2 ), 114 – 129 . Google Scholar CrossRef Search ADS Cohen W. M. , Levinthal D. A. ( 1990 ), ‘ Absorptive capacity: a new perspective on learning and innovation ,’ Administrative Science Quarterly , 35 ( 1 ), 128 – 152 . Google Scholar CrossRef Search ADS Cook S. D. , Brown J. S. ( 1999 ), ‘ Bridging epistemologies: the generative dance between organizational knowledge and organizational knowing ,’ Organization Science , 10 ( 4 ), 381 – 400 . Google Scholar CrossRef Search ADS Cronbach L. ( 1987 ), ‘ Statistical tests for moderator variables: flaw in analyses recently proposed ,’ Psychological Bulletin , 102 ( 3 ), 4114 – 4417 . Google Scholar CrossRef Search ADS Czarnitzki D. , Hottenrott H. , Thorwarth S. ( 2011 ), ‘ Industrial research versus development investment: the implications of financial constraints ,’ Cambridge Journal of Economics , 35 ( 3 ), 527 – 544 . Google Scholar CrossRef Search ADS Czarnitzki D. , Kraft K. , Thorwarth S. ( 2009 ), ‘ The knowledge production of ‘R’and ‘D ,’ Economics Letters , 105 ( 1 ), 141 – 143 . Google Scholar CrossRef Search ADS D’Este P. , Amara N. , Olmos-Peñuela J. ( 2016 ), ‘ Fostering novelty while reducing failure: balancing the twin challenges of product innovation ,’ Technological Forecasting and Social Change , 113 , 280 – 292 . Available online: doi: 10.1016/j.techfore.2015.08.011. Google Scholar CrossRef Search ADS D’Este P. , Iammarino S. , Savona M. , von Tunzelmann N. ( 2012 ), ‘ What hampers innovation? Revealed barriers versus deterring barriers ,’ Research Policy , 41 ( 2 ), 482 – 488 . Google Scholar CrossRef Search ADS D’Este P. , Rentocchini F. , Vega-Jurado J. ( 2014 ), ‘ The role of human capital in lowering the barriers to engaging in innovation: evidence from the Spanish innovation survey ,’ Industry and Innovation , 21 ( 1 ), 1 – 19 . Google Scholar CrossRef Search ADS David P. A. , Mowery D. , Steinmueller W. E. ( 1992 ), ‘ Analysing the economic payoffs from basic research ,’ Economics of Innovation and New Technology , 2 ( 1 ), 73 – 90 . Google Scholar CrossRef Search ADS Denrell J. , Fang C. , Levinthal D. A. ( 2004 ), ‘ From T-mazes to labyrinths: learning from model-based feedback ,’ Management Science , 50 ( 10 ), 1366 – 1378 . Google Scholar CrossRef Search ADS Desai V. ( 2010 ), ‘ Learning to learn from failures: the impact of operating experience on railroad accident responses ,’ Industrial and Corporate Change , 25 ( 2 ), 199 – 226 . http://dx.doi.org/10.1093/icc/dtq019. Google Scholar CrossRef Search ADS Desai V. ( 2015 ), ‘ Learning through the distribution of failures within an organization: evidence from heart bypass surgery performance ,’ Academy of Management Journal , 58 ( 4 ), 1032 – 1050 . Google Scholar CrossRef Search ADS Dodgson M. , Gann D. , Salter A. ( 2008 ), The Management of Technological Innovation. Strategy and Practice . Oxford University Press : Oxford, UK . Duguet E. ( 2006 ), ‘ Innovation height, spillovers and TFP growth at the firm level: evidence from French manufacturing ,’ Economics of Innovation and New Technology , 15 ( 4–5 ), 415 – 442 . Google Scholar CrossRef Search ADS Edmondson A. C. ( 2011 ), ‘ Strategies for learning from failure ,’ Harvard Business Review , 2011 , 48 – 55 . Eisenhardt K. M. , Martin J. A. ( 2000 ), ‘ Dynamic capabilities: what are they? ,’ Strategic Management Journal , 21 ( 10–11 ), 1105 – 1121 . Google Scholar CrossRef Search ADS Fang C. ( 2012 ), ‘ Organizational learning as credit assignment: a model and two experiments ,’ Organization Science , 23 ( 6 ), 1717 – 1732 . Google Scholar CrossRef Search ADS Fang C. , Lee J. , Schilling M. A. ( 2010 ), ‘ Balancing exploration and exploitation through structural design: the isolation of subgroups and organizational learning ,’ Organization Science , 21 ( 3 ), 625 – 642 . Google Scholar CrossRef Search ADS Fang C. , Levinthal D. ( 2009 ), ‘ Near-term liability of exploitation: exploration and exploitation in multistage problems ,’ Organization Science , 20 ( 3 ), 538 – 551 . Google Scholar CrossRef Search ADS Fiol C. M. , Lyles M. A. ( 1985 ), ‘ Organizational learning ,’ Academy of Management Review , 10 ( 4 ), 803 – 813 . Google Scholar CrossRef Search ADS Green S. G. , Welsh M. A. , Dehler G. E. ( 2003 ), ‘ Advocacy, performance, and threshold influences on decisions to terminate new product development ,’ Academy of Management Journal , 46 ( 4 ), 419 – 434 . Greene W. ( 2010 ), ‘ Testing hypotheses about interaction terms in nonlinear models ,’ Economics Letters , 107 ( 2 ), 291 – 296 . Google Scholar CrossRef Search ADS Gupta A. K. , Smith K. G. , Shalley C. E. ( 2006 ), ‘ The interplay between exploration and exploitation ,’ Academy of Management Journal , 49 ( 4 ), 693 – 706 . Google Scholar CrossRef Search ADS Hair J. F. Jr , Tatham R. L. , Anderson R. E. , Black W. ( 1998 ), Multivariate Data Analysis , 5th edn. Prentice-Hall : Englewood Cliffs, NJ . Hall B. H. ( 1993 ), ‘ The stock market's valuation of R&D investment during the 1980's ,’ The American Economic Review , 83 ( 2 ), 259 – 264 . Haunschild P. R. , Sullivan B. ( 2002 ), ‘ Learning from complexity: effects of prior accidents and incidents on airlines’ learning ,’ Administrative Science Quarterly , 47 ( 4 ), 609 – 643 . Google Scholar CrossRef Search ADS He Z. L. , Wong P. K. ( 2004 ), ‘ Exploration vs. exploitation: an empirical test of the ambidexterity hypothesis ,’ Organization Science , 15 ( 4 ), 481 – 494 . Google Scholar CrossRef Search ADS Jani A. ( 2011 ), ‘ Escalation commitment in troubled IT projects: influence of project risk factors and self-efficacy on the perception of risk and the commitment to a failing project ,’ International Journal of Project Management , 29 ( 7 ), 934 – 945 . Google Scholar CrossRef Search ADS Katila R. , Ahuja G. ( 2002 ), ‘ Something old, something new: a longitudinal study of search behavior and new product introduction ,’ Academy of Management Journal , 45 ( 6 ), 1183 – 1194 . Kim J. Y. , Kim J. Y. , Miner A. S. ( 2009 ), ‘ Organizational learning from extreme performance experience: the impact of success and recovery experience ,’ Organization Science , 20 ( 6 ), 958 – 978 . Google Scholar CrossRef Search ADS Laursen K. , Salter A. ( 2006 ), ‘ Open for innovation: the role of openness in explaining innovation performance among UK manufacturing firms ,’ Strategic Management Journal , 27 ( 2 ), 131 – 150 . Google Scholar CrossRef Search ADS Leonard-Barton D. ( 1995 ), Wellsprings of Knowledge: Building and Sustaining the Sources of Innovation . Harvard Business School Press . Boston, MA . Leoncini R. ( 2016 ), ‘ Learning-by-failing. An empirical exercise on CIS data ,’ Research Policy , 45 ( 2 ), 376 – 386 . Google Scholar CrossRef Search ADS Levinthal D. A. , March J. G. ( 1993 ), ‘ The myopia of learning ,’ Strategic Management Journal , 14 ( S2 ), 95 – 112 . Google Scholar CrossRef Search ADS Lhuillery S. , Pfister E. ( 2009 ), ‘ R&D cooperation and failures in innovation projects: empirical evidence from French CIS data ,’ Research Policy , 38 ( 1 ), 45 – 57 . Google Scholar CrossRef Search ADS Lind J. T. , Mehlum H. ( 2010 ), ‘ With or without u? The appropriate test for a U‐shaped relationship ,’ Oxford Bulletin of Economics and Statistics , 72 ( 1 ), 109 – 118 . Google Scholar CrossRef Search ADS Lippman S. A. , Rumelt R. P. ( 1982 ), ‘ Uncertain imitability: an analysis of interfirm differences in efficiency under competition ,’ The Bell Journal of Economics , 13 ( 2 ), 418 – 438 . Google Scholar CrossRef Search ADS Madsen P. M. , Desai V. ( 2010 ), ‘ Failing to learn? The effects of failure and success on organizational learning in the global orbital launch vehicle industry ,’ Academy of Management Journal , 53 ( 3 ), 451 – 476 . Google Scholar CrossRef Search ADS Mairesse J. , Mohnen P. ( 2010 ), ‘Using innovations surveys for econometric analysis,’ National Bureau of Economic Research., No. w15857. Majchrzak A. , Cooper L. P. , Neece O. E. ( 2004 ), ‘ Knowledge reuse for innovation ,’ Management Science , 50 ( 2 ), 174 – 188 . Google Scholar CrossRef Search ADS March J. G. ( 1991 ), ‘ Exploration and exploitation in organizational learning ,’ Organization Science , 2 ( 1 ), 71 – 87 . Google Scholar CrossRef Search ADS March J. G. ( 2006 ), ‘ Rationality, foolishness, and adaptive intelligence ,’ Strategic Management Journal , 27 ( 3 ), 201 – 214 . Google Scholar CrossRef Search ADS Marzucchi A. , Montresor S. ( 2017 ), ‘ Forms of knowledge and eco-innovation modes: evidence from Spanish manufacturing firms ,’ Ecological Economics , 131 , 208 – 221 . Google Scholar CrossRef Search ADS Mohnen P. , Palm F. C. , Van Der Loeff S. S. , Tiwari A. ( 2008 ), ‘ Financial constraints and other obstacles: are they a threat to innovation activity? ,’ De Economist , 156 ( 2 ), 201 – 214 . Google Scholar CrossRef Search ADS Molero A. , García A. ( 2008 ), ‘ The innovative activity of foreign subsidiaries in the Spanish innovation system: an evaluation of their impact from a sectoral taxonomy approach ,’ Technovation , 28 ( 11 ), 739 – 757 . Google Scholar CrossRef Search ADS Monjon S. , Waelbroeck P. ( 2003 ), ‘ Assessing spillovers from universities to firms: evidence from French firm-level data ,’ International Journal of Industrial Organization , 21 ( 9 ), 1255 – 1270 . Google Scholar CrossRef Search ADS Mudambi R. , Swift T. ( 2011 ), ‘ Proactive R&D management and firm growth: a punctuated equilibrium model ,’ Research Policy , 40 ( 3 ), 429 – 440 . Google Scholar CrossRef Search ADS Mudambi R. , Swift T. ( 2014 ), ‘ Knowing when to leap: transitioning between exploitative and explorative R&D ,’ Strategic Management Journal , 35 ( 1 ), 126 – 145 . Google Scholar CrossRef Search ADS Nelson R. R. , Winter S. G. ( 1982 ), An Evolutionary Theory of Economic Change . Harvard University Press : Cambridge, MA . Nerkar A. ( 2003 ), ‘ Old is gold? The value of temporal exploration in the creation of new knowledge ,’ Management Science , 49 ( 2 ), 211 – 229 . Google Scholar CrossRef Search ADS OECD ( 2002 ), Frascati Manual: Proposed Standard Practice for Surveys on Research and Experimental Development , 6th edn . OECD Publishing : Paris, France . OECD ( 2005 ), The Measurement of Scientific and Technological Activities. Oslo Manual: Guidelines for Collecting and Interpreting Innovation Data , 3rd edn. OECD Publishing : Paris, France . Pavitt K. ( 1984 ), ‘ Sectoral patterns of technical change: towards a taxonomy and a theory ,’ Research Policy , 13 ( 6 ), 343 – 373 . Google Scholar CrossRef Search ADS Rammer C. , Czarnitzki D. , Spielkamp A. ( 2009 ), ‘ Innovation success of non-R&D-performers: substituting technology by management in SMEs ,’ Small Business Economics , 33 ( 1 ), 35 – 58 . Google Scholar CrossRef Search ADS Roodman D. ( 2011 ), ‘ Estimating fully observed recursive mixed-process models with cmp ,’ Stata Journal , 11 ( 2 ), 159 – 206 . Rosenkopf L. , Nerkar A. ( 2001 ), ‘ Beyond local search: boundary-spanning, exploration, and impact in the optical disk industry ,’ Strategic Management Journal , 22 ( 4 ), 287 – 306 . Google Scholar CrossRef Search ADS Sadler-Smith E. , Spicer D. P. , Chaston I. ( 2001 ), ‘ Learning orientations and growth in smaller firms ,’ Long Range Planning , 34 ( 2 ), 139 – 158 . Google Scholar CrossRef Search ADS Santamaría L. , Nieto M. J. , Miles I. ( 2012 ), ‘ Service innovation in manufacturing firms: evidence from Spain ,’ Technovation , 32 ( 2 ), 144 – 155 . Google Scholar CrossRef Search ADS Schilling M. A. , Vidal P. , Ployhart R. E. , Marangoni A. ( 2003 ), ‘ Learning by doing something else: variation, relatedness, and the learning curve ,’ Management Science , 49 ( 1 ), 39 – 56 . Google Scholar CrossRef Search ADS Schmidt J. B. , Calantone R. J. ( 1998 ), ‘ Are really new product development projects harder to shut down? ,’ Journal of Product Innovation Management , 15 ( 2 ), 111 – 123 . Google Scholar CrossRef Search ADS Schneider C. , Veugelers R. ( 2010 ), ‘ On young highly innovative companies: why they matter and how (not) to policy support them ,’ Industrial and Corporate Change , 19 ( 4 ), 969 – 1007 . Google Scholar CrossRef Search ADS Segal E. ( 2004 ), ‘ Incubation in insight problem solving ,’ Creativity Research Journal , 16 ( 1 ), 141 – 148 . Google Scholar CrossRef Search ADS Shepherd D. A. , Cardon M. S. ( 2009 ), ‘ Negative emotional reactions to project failure and the self‐compassion to learn from the experience ,’ Journal of Management Studies , 46 ( 6 ), 923 – 949 . Google Scholar CrossRef Search ADS Shepherd D. A. , Patzelt H. , Williams T. A. , Warnecke D. ( 2014 ), ‘ How does project termination impact project team members? Rapid termination,‘Creeping death’, and Learning from failure ,’ Journal of Management Studies , 51 ( 4 ), 513 – 546 . Google Scholar CrossRef Search ADS Staw B. , Ross J. ( 1987 ), ‘Behavior in escalation situations: antecedents, prototypes, and solutions,’ in Cummings L.L. , Staw B.M. (eds), Research in Organization Behavior . JAI : Greenwhich, CT . Sterlacchini A. ( 1999 ), ‘ Do innovative activities matter to small firms in non-R&D-intensive industries? An application to export performance ,’ Research Policy , 28 ( 8 ), 819 – 832 . Google Scholar CrossRef Search ADS Tidd J. , Bessant J. , Pavitt K. ( 1997 ), Managing Innovation. Integrating Technological, Market and Organizational Change . John Wiley & Sons Ltd : Chichester, UK . Theil H. ( 1971 ), Principles of Econometrics . JohnWiley : New York, NY . Thomke S. ( 2001 ), ‘ Enlightened experimentation: the new imperative for innovation ,’ Harvard Business Review , 79 ( 2 ), 66 – 75 . Trigo A. , Vence X. ( 2012 ), ‘ Scope and patterns of innovation cooperation in Spanish service enterprises ,’ Research Policy , 41 ( 3 ), 602 – 613 . Google Scholar CrossRef Search ADS Tucker A. L. , Edmondson A. C. ( 2003 ), ‘ Why hospitals don't learn from failures ,’ California Management Review , 45 ( 2 ), 55 – 72 . Google Scholar CrossRef Search ADS Vega-Jurado J. , Gutierrez-Gracia A. , Fernandez-de-Lucio I. ( 2009 ), ‘ Does external knowledge sourcing matter for innovation? Evidence from the Spanish manufacturing industry ,’ Industrial and Corporate Change , 18 ( 4 ), 637 – 670 . Google Scholar CrossRef Search ADS Von Hippel E. , Tyre M. J. ( 1995 ), ‘ How learning by doing is done: problem identification in novel process equipment ,’ Research Policy , 24 ( 1 ), 1 – 12 . Google Scholar CrossRef Search ADS Wooldridge J. M. ( 2010 ), Econometric Analysis of Cross Section and Panel Data . MIT press : Cambridge, Massachusetts. Zangwill W. I. , Kantor P. B. ( 1998 ), ‘ Toward a theory of continuous improvement and the learning curve ,’ Management Science , 44 ( 7 ), 910 – 920 . Google Scholar CrossRef Search ADS Zelner B. A. ( 2009 ), ‘ Using simulation to interpret results from logit, probit, and other nonlinear models ,’ Strategic Management Journal , 30 ( 12 ), 1335 – 1348 . Google Scholar CrossRef Search ADS Zollo M. , Winter S. G. ( 2002 ), ‘ Deliberate learning and the evolution of dynamic capabilities ,’ Organization Science , 13 ( 3 ), 339 – 351 . Google Scholar CrossRef Search ADS Appendix Table A1. Proportion of companies experiencing failure (overall, conception, and implementation) for different types of innovation-related activities Types of innovation-related activities Failure overall (%) Failure conception (%) Failure implementation (%) Total Intramural R&D (only) 27.94 20.83 17.52 816 Other innovation activities (only) 19.33 10.33 13.33 600 Both 34.62 27.62 22.68 4052 Types of innovation-related activities Failure overall (%) Failure conception (%) Failure implementation (%) Total Intramural R&D (only) 27.94 20.83 17.52 816 Other innovation activities (only) 19.33 10.33 13.33 600 Both 34.62 27.62 22.68 4052 The category “other innovation activities” includes the following: extramural R&D; acquisition of machinery, equipment, and software; acquisition of other external knowledge; training; market introduction of innovations; other preparations. Table A1. Proportion of companies experiencing failure (overall, conception, and implementation) for different types of innovation-related activities Types of innovation-related activities Failure overall (%) Failure conception (%) Failure implementation (%) Total Intramural R&D (only) 27.94 20.83 17.52 816 Other innovation activities (only) 19.33 10.33 13.33 600 Both 34.62 27.62 22.68 4052 Types of innovation-related activities Failure overall (%) Failure conception (%) Failure implementation (%) Total Intramural R&D (only) 27.94 20.83 17.52 816 Other innovation activities (only) 19.33 10.33 13.33 600 Both 34.62 27.62 22.68 4052 The category “other innovation activities” includes the following: extramural R&D; acquisition of machinery, equipment, and software; acquisition of other external knowledge; training; market introduction of innovations; other preparations. Table A2. Innovation failures determinants: past exploration in stock Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.041*** 0.050*** 0.060*** 0.088*** 0.030** 0.032* (0.011) (0.018) (0.013) (0.021) (0.012) (0.019) R&D exploration stock-1 0.008 0.015 0.007 0.031* 0.007 0.009 (0.010) (0.014) (0.012) (0.017) (0.011) (0.015) R&D exploration × R&D exploration stock-1 −0.002 −0.005* −0.000 (0.003) (0.003) (0.003) R&D exploitation 0.027** 0.026** 0.040*** 0.038*** 0.006 0.006 (0.012) (0.012) (0.014) (0.014) (0.012) (0.012) Knowledge breadth 0.013 0.013 0.062*** 0.061*** 0.008 0.008 (0.012) (0.012) (0.014) (0.014) (0.013) (0.013) Knowledge depth 0.019 0.019 0.036 0.038 0.016 0.017 (0.023) (0.023) (0.025) (0.026) (0.024) (0.024) Other innovation expenditures 0.001 0.001 −0.001 −0.001 −0.001 −0.001 (0.014) (0.014) (0.016) (0.016) (0.015) (0.015) Firm age 0.151** 0.151** 0.113 0.113 0.124* 0.124* (0.068) (0.068) (0.077) (0.077) (0.074) (0.074) Group 0.040 0.041 0.021 0.022 0.136 0.136 (0.086) (0.086) (0.098) (0.098) (0.091) (0.091) Cost barriers −0.083 −0.085 −0.129 −0.134 −0.107 −0.107 (0.093) (0.093) (0.108) (0.108) (0.099) (0.099) Knowledge barriers 0.234 0.235 0.331 0.336 0.317 0.317 (0.308) (0.308) (0.345) (0.345) (0.316) (0.316) Market barriers—uncertainty 0.226*** 0.227*** 0.146 0.151 0.256*** 0.257*** (0.081) (0.081) (0.093) (0.093) (0.084) (0.084) Market barriers—dominated 0.101 0.101 0.177* 0.178* 0.028 0.028 (0.091) (0.091) (0.104) (0.104) (0.097) (0.097) Number of employees 0.054 0.053 0.070 0.067 0.027 0.027 (0.041) (0.041) (0.047) (0.047) (0.043) (0.043) Radical innovation 0.135** 0.135** 0.269*** 0.268*** 0.105 0.105 (0.068) (0.068) (0.076) (0.076) (0.072) (0.072) Org innovation 0.310*** 0.311*** 0.514*** 0.518*** 0.211*** 0.212*** (0.068) (0.068) (0.079) (0.079) (0.073) (0.073) Number of R&D researchers 0.048 0.054 −0.009 0.008 0.100 0.101 (0.059) (0.059) (0.064) (0.065) (0.061) (0.062) Wald Χ2 127.07 (19)*** 127.29 (20)*** 172.64 (19)*** 174.71 (20)*** 90.54 (19)*** 90.6 (20)*** Firm-period observations 2226 2226 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 4191 4191 Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.041*** 0.050*** 0.060*** 0.088*** 0.030** 0.032* (0.011) (0.018) (0.013) (0.021) (0.012) (0.019) R&D exploration stock-1 0.008 0.015 0.007 0.031* 0.007 0.009 (0.010) (0.014) (0.012) (0.017) (0.011) (0.015) R&D exploration × R&D exploration stock-1 −0.002 −0.005* −0.000 (0.003) (0.003) (0.003) R&D exploitation 0.027** 0.026** 0.040*** 0.038*** 0.006 0.006 (0.012) (0.012) (0.014) (0.014) (0.012) (0.012) Knowledge breadth 0.013 0.013 0.062*** 0.061*** 0.008 0.008 (0.012) (0.012) (0.014) (0.014) (0.013) (0.013) Knowledge depth 0.019 0.019 0.036 0.038 0.016 0.017 (0.023) (0.023) (0.025) (0.026) (0.024) (0.024) Other innovation expenditures 0.001 0.001 −0.001 −0.001 −0.001 −0.001 (0.014) (0.014) (0.016) (0.016) (0.015) (0.015) Firm age 0.151** 0.151** 0.113 0.113 0.124* 0.124* (0.068) (0.068) (0.077) (0.077) (0.074) (0.074) Group 0.040 0.041 0.021 0.022 0.136 0.136 (0.086) (0.086) (0.098) (0.098) (0.091) (0.091) Cost barriers −0.083 −0.085 −0.129 −0.134 −0.107 −0.107 (0.093) (0.093) (0.108) (0.108) (0.099) (0.099) Knowledge barriers 0.234 0.235 0.331 0.336 0.317 0.317 (0.308) (0.308) (0.345) (0.345) (0.316) (0.316) Market barriers—uncertainty 0.226*** 0.227*** 0.146 0.151 0.256*** 0.257*** (0.081) (0.081) (0.093) (0.093) (0.084) (0.084) Market barriers—dominated 0.101 0.101 0.177* 0.178* 0.028 0.028 (0.091) (0.091) (0.104) (0.104) (0.097) (0.097) Number of employees 0.054 0.053 0.070 0.067 0.027 0.027 (0.041) (0.041) (0.047) (0.047) (0.043) (0.043) Radical innovation 0.135** 0.135** 0.269*** 0.268*** 0.105 0.105 (0.068) (0.068) (0.076) (0.076) (0.072) (0.072) Org innovation 0.310*** 0.311*** 0.514*** 0.518*** 0.211*** 0.212*** (0.068) (0.068) (0.079) (0.079) (0.073) (0.073) Number of R&D researchers 0.048 0.054 −0.009 0.008 0.100 0.101 (0.059) (0.059) (0.064) (0.065) (0.061) (0.062) Wald Χ2 127.07 (19)*** 127.29 (20)*** 172.64 (19)*** 174.71 (20)*** 90.54 (19)*** 90.6 (20)*** Firm-period observations 2226 2226 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 4191 4191 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Columns 1 and 2, the probability to fail at the conception phase in Columns 3 and 4, and the probability to fail at the implementation phase in Columns 5 and 6. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels where the start date of the panel refers to the dependent variable (i.e., if t = 2007–2009, so t-1 = 2004–2006). * P < 0.10; ** P < 0.05; *** P < 0.01. Table A2. Innovation failures determinants: past exploration in stock Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.041*** 0.050*** 0.060*** 0.088*** 0.030** 0.032* (0.011) (0.018) (0.013) (0.021) (0.012) (0.019) R&D exploration stock-1 0.008 0.015 0.007 0.031* 0.007 0.009 (0.010) (0.014) (0.012) (0.017) (0.011) (0.015) R&D exploration × R&D exploration stock-1 −0.002 −0.005* −0.000 (0.003) (0.003) (0.003) R&D exploitation 0.027** 0.026** 0.040*** 0.038*** 0.006 0.006 (0.012) (0.012) (0.014) (0.014) (0.012) (0.012) Knowledge breadth 0.013 0.013 0.062*** 0.061*** 0.008 0.008 (0.012) (0.012) (0.014) (0.014) (0.013) (0.013) Knowledge depth 0.019 0.019 0.036 0.038 0.016 0.017 (0.023) (0.023) (0.025) (0.026) (0.024) (0.024) Other innovation expenditures 0.001 0.001 −0.001 −0.001 −0.001 −0.001 (0.014) (0.014) (0.016) (0.016) (0.015) (0.015) Firm age 0.151** 0.151** 0.113 0.113 0.124* 0.124* (0.068) (0.068) (0.077) (0.077) (0.074) (0.074) Group 0.040 0.041 0.021 0.022 0.136 0.136 (0.086) (0.086) (0.098) (0.098) (0.091) (0.091) Cost barriers −0.083 −0.085 −0.129 −0.134 −0.107 −0.107 (0.093) (0.093) (0.108) (0.108) (0.099) (0.099) Knowledge barriers 0.234 0.235 0.331 0.336 0.317 0.317 (0.308) (0.308) (0.345) (0.345) (0.316) (0.316) Market barriers—uncertainty 0.226*** 0.227*** 0.146 0.151 0.256*** 0.257*** (0.081) (0.081) (0.093) (0.093) (0.084) (0.084) Market barriers—dominated 0.101 0.101 0.177* 0.178* 0.028 0.028 (0.091) (0.091) (0.104) (0.104) (0.097) (0.097) Number of employees 0.054 0.053 0.070 0.067 0.027 0.027 (0.041) (0.041) (0.047) (0.047) (0.043) (0.043) Radical innovation 0.135** 0.135** 0.269*** 0.268*** 0.105 0.105 (0.068) (0.068) (0.076) (0.076) (0.072) (0.072) Org innovation 0.310*** 0.311*** 0.514*** 0.518*** 0.211*** 0.212*** (0.068) (0.068) (0.079) (0.079) (0.073) (0.073) Number of R&D researchers 0.048 0.054 −0.009 0.008 0.100 0.101 (0.059) (0.059) (0.064) (0.065) (0.061) (0.062) Wald Χ2 127.07 (19)*** 127.29 (20)*** 172.64 (19)*** 174.71 (20)*** 90.54 (19)*** 90.6 (20)*** Firm-period observations 2226 2226 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 4191 4191 Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.041*** 0.050*** 0.060*** 0.088*** 0.030** 0.032* (0.011) (0.018) (0.013) (0.021) (0.012) (0.019) R&D exploration stock-1 0.008 0.015 0.007 0.031* 0.007 0.009 (0.010) (0.014) (0.012) (0.017) (0.011) (0.015) R&D exploration × R&D exploration stock-1 −0.002 −0.005* −0.000 (0.003) (0.003) (0.003) R&D exploitation 0.027** 0.026** 0.040*** 0.038*** 0.006 0.006 (0.012) (0.012) (0.014) (0.014) (0.012) (0.012) Knowledge breadth 0.013 0.013 0.062*** 0.061*** 0.008 0.008 (0.012) (0.012) (0.014) (0.014) (0.013) (0.013) Knowledge depth 0.019 0.019 0.036 0.038 0.016 0.017 (0.023) (0.023) (0.025) (0.026) (0.024) (0.024) Other innovation expenditures 0.001 0.001 −0.001 −0.001 −0.001 −0.001 (0.014) (0.014) (0.016) (0.016) (0.015) (0.015) Firm age 0.151** 0.151** 0.113 0.113 0.124* 0.124* (0.068) (0.068) (0.077) (0.077) (0.074) (0.074) Group 0.040 0.041 0.021 0.022 0.136 0.136 (0.086) (0.086) (0.098) (0.098) (0.091) (0.091) Cost barriers −0.083 −0.085 −0.129 −0.134 −0.107 −0.107 (0.093) (0.093) (0.108) (0.108) (0.099) (0.099) Knowledge barriers 0.234 0.235 0.331 0.336 0.317 0.317 (0.308) (0.308) (0.345) (0.345) (0.316) (0.316) Market barriers—uncertainty 0.226*** 0.227*** 0.146 0.151 0.256*** 0.257*** (0.081) (0.081) (0.093) (0.093) (0.084) (0.084) Market barriers—dominated 0.101 0.101 0.177* 0.178* 0.028 0.028 (0.091) (0.091) (0.104) (0.104) (0.097) (0.097) Number of employees 0.054 0.053 0.070 0.067 0.027 0.027 (0.041) (0.041) (0.047) (0.047) (0.043) (0.043) Radical innovation 0.135** 0.135** 0.269*** 0.268*** 0.105 0.105 (0.068) (0.068) (0.076) (0.076) (0.072) (0.072) Org innovation 0.310*** 0.311*** 0.514*** 0.518*** 0.211*** 0.212*** (0.068) (0.068) (0.079) (0.079) (0.073) (0.073) Number of R&D researchers 0.048 0.054 −0.009 0.008 0.100 0.101 (0.059) (0.059) (0.064) (0.065) (0.061) (0.062) Wald Χ2 127.07 (19)*** 127.29 (20)*** 172.64 (19)*** 174.71 (20)*** 90.54 (19)*** 90.6 (20)*** Firm-period observations 2226 2226 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 4191 4191 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Columns 1 and 2, the probability to fail at the conception phase in Columns 3 and 4, and the probability to fail at the implementation phase in Columns 5 and 6. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels where the start date of the panel refers to the dependent variable (i.e., if t = 2007–2009, so t-1 = 2004–2006). * P < 0.10; ** P < 0.05; *** P < 0.01. Table A3. Innovation failures determinants: current exploration activities for SMEs (less than 50 employees) Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.037*** 0.047*** 0.036** (0.014) (0.015) (0.015) R&D exploitation 0.024 0.024 0.010 (0.015) (0.018) (0.017) Knowledge breadth −0.009 0.036* −0.010 (0.017) (0.019) (0.019) Knowledge depth 0.013 0.007 0.027 (0.035) (0.039) (0.038) Other innovation expenditures −0.012 −0.010 −0.009 (0.019) (0.022) (0.021) Firm age 0.165 0.236** 0.056 (0.105) (0.118) (0.122) Group −0.027 0.004 0.006 (0.135) (0.153) (0.153) Cost barriers −0.175 −0.188 −0.268** (0.120) (0.137) (0.134) Knowledge barriers 0.302 0.527 0.340 (0.334) (0.344) (0.367) Market barriers—uncertainty 0.340*** 0.235* 0.346*** (0.115) (0.129) (0.125) Market barriers—dominated 0.155 0.185 0.128 (0.125) (0.139) (0.140) Number of employees −0.129 −0.166 −0.105 (0.094) (0.110) (0.106) Radical innovation 0.133 0.294*** 0.025 (0.100) (0.109) (0.111) Org innovation 0.233** 0.411*** 0.155 (0.097) (0.110) (0.107) Number of R&D researchers 0.008 0.143 0.053 (0.126) (0.135) (0.138) Wald Χ2 52.6 (18)*** 70.23 (18)*** 33.52 (18)** Firm-period observations 1000 1000 1000 Firm observations 1805 1805 1805 Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.037*** 0.047*** 0.036** (0.014) (0.015) (0.015) R&D exploitation 0.024 0.024 0.010 (0.015) (0.018) (0.017) Knowledge breadth −0.009 0.036* −0.010 (0.017) (0.019) (0.019) Knowledge depth 0.013 0.007 0.027 (0.035) (0.039) (0.038) Other innovation expenditures −0.012 −0.010 −0.009 (0.019) (0.022) (0.021) Firm age 0.165 0.236** 0.056 (0.105) (0.118) (0.122) Group −0.027 0.004 0.006 (0.135) (0.153) (0.153) Cost barriers −0.175 −0.188 −0.268** (0.120) (0.137) (0.134) Knowledge barriers 0.302 0.527 0.340 (0.334) (0.344) (0.367) Market barriers—uncertainty 0.340*** 0.235* 0.346*** (0.115) (0.129) (0.125) Market barriers—dominated 0.155 0.185 0.128 (0.125) (0.139) (0.140) Number of employees −0.129 −0.166 −0.105 (0.094) (0.110) (0.106) Radical innovation 0.133 0.294*** 0.025 (0.100) (0.109) (0.111) Org innovation 0.233** 0.411*** 0.155 (0.097) (0.110) (0.107) Number of R&D researchers 0.008 0.143 0.053 (0.126) (0.135) (0.138) Wald Χ2 52.6 (18)*** 70.23 (18)*** 33.52 (18)** Firm-period observations 1000 1000 1000 Firm observations 1805 1805 1805 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Column 1, the probability to fail at the conception phase in Column 2, and the probability to fail at the implementation phase in Column 3. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels. * P < 0.10; ** P < 0.05; *** P < 0.01. Table A3. Innovation failures determinants: current exploration activities for SMEs (less than 50 employees) Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.037*** 0.047*** 0.036** (0.014) (0.015) (0.015) R&D exploitation 0.024 0.024 0.010 (0.015) (0.018) (0.017) Knowledge breadth −0.009 0.036* −0.010 (0.017) (0.019) (0.019) Knowledge depth 0.013 0.007 0.027 (0.035) (0.039) (0.038) Other innovation expenditures −0.012 −0.010 −0.009 (0.019) (0.022) (0.021) Firm age 0.165 0.236** 0.056 (0.105) (0.118) (0.122) Group −0.027 0.004 0.006 (0.135) (0.153) (0.153) Cost barriers −0.175 −0.188 −0.268** (0.120) (0.137) (0.134) Knowledge barriers 0.302 0.527 0.340 (0.334) (0.344) (0.367) Market barriers—uncertainty 0.340*** 0.235* 0.346*** (0.115) (0.129) (0.125) Market barriers—dominated 0.155 0.185 0.128 (0.125) (0.139) (0.140) Number of employees −0.129 −0.166 −0.105 (0.094) (0.110) (0.106) Radical innovation 0.133 0.294*** 0.025 (0.100) (0.109) (0.111) Org innovation 0.233** 0.411*** 0.155 (0.097) (0.110) (0.107) Number of R&D researchers 0.008 0.143 0.053 (0.126) (0.135) (0.138) Wald Χ2 52.6 (18)*** 70.23 (18)*** 33.52 (18)** Firm-period observations 1000 1000 1000 Firm observations 1805 1805 1805 Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.037*** 0.047*** 0.036** (0.014) (0.015) (0.015) R&D exploitation 0.024 0.024 0.010 (0.015) (0.018) (0.017) Knowledge breadth −0.009 0.036* −0.010 (0.017) (0.019) (0.019) Knowledge depth 0.013 0.007 0.027 (0.035) (0.039) (0.038) Other innovation expenditures −0.012 −0.010 −0.009 (0.019) (0.022) (0.021) Firm age 0.165 0.236** 0.056 (0.105) (0.118) (0.122) Group −0.027 0.004 0.006 (0.135) (0.153) (0.153) Cost barriers −0.175 −0.188 −0.268** (0.120) (0.137) (0.134) Knowledge barriers 0.302 0.527 0.340 (0.334) (0.344) (0.367) Market barriers—uncertainty 0.340*** 0.235* 0.346*** (0.115) (0.129) (0.125) Market barriers—dominated 0.155 0.185 0.128 (0.125) (0.139) (0.140) Number of employees −0.129 −0.166 −0.105 (0.094) (0.110) (0.106) Radical innovation 0.133 0.294*** 0.025 (0.100) (0.109) (0.111) Org innovation 0.233** 0.411*** 0.155 (0.097) (0.110) (0.107) Number of R&D researchers 0.008 0.143 0.053 (0.126) (0.135) (0.138) Wald Χ2 52.6 (18)*** 70.23 (18)*** 33.52 (18)** Firm-period observations 1000 1000 1000 Firm observations 1805 1805 1805 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Column 1, the probability to fail at the conception phase in Column 2, and the probability to fail at the implementation phase in Column 3. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels. * P < 0.10; ** P < 0.05; *** P < 0.01. Table A4. Innovation failures determinants: interaction effects for SMEs (less than 50 employees) Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.034** 0.059*** 0.047*** 0.079*** 0.031* 0.044* (0.015) (0.023) (0.017) (0.026) (0.016) (0.025) R&D exploration-1 0.007 0.029 −0.001 0.029 0.011 0.023 (0.015) (0.021) (0.017) (0.024) (0.017) (0.024) R&D exploration X R&D exploration-1 −0.005 −0.007* −0.003 (0.004) (0.004) (0.004) R&D exploitation 0.024 0.022 0.024 0.021 0.010 0.009 (0.018) (0.016) (0.018) (0.018) (0.017) (0.017) Knowledge breadth 0.036* −0.009 0.036* 0.036* −0.010 −0.010 (0.019) (0.017) (0.019) (0.019) (0.019) (0.019) Knowledge depth 0.007 0.015 0.007 0.010 0.028 0.028 (0.039) (0.035) (0.039) (0.039) (0.038) (0.038) Other innovation expenditures −0.010 −0.012 −0.010 −0.010 −0.010 −0.010 (0.022) (0.019) (0.022) (0.022) (0.021) (0.021) Firm age 0.236** 0.170 0.236** 0.240** 0.059 0.060 (0.118) (0.105) (0.118) (0.118) (0.122) (0.122) Group 0.004 −0.029 0.004 0.004 0.005 0.004 (0.153) (0.135) (0.153) (0.153) (0.153) (0.153) Cost barriers −0.188 −0.174 −0.188 −0.183 −0.272** −0.270** (0.137) (0.120) (0.137) (0.137) (0.135) (0.134) Knowledge barriers 0.527 0.310 0.528 0.542 0.337 0.343 (0.344) (0.335) (0.345) (0.345) (0.367) (0.367) Market barriers—uncertainty 0.235* 0.341*** 0.235* 0.237* 0.348*** 0.347*** (0.129) (0.115) (0.129) (0.129) (0.126) (0.125) Market barriers—dominated 0.185 0.160 0.184 0.190 0.131 0.133 (0.139) (0.125) (0.139) (0.139) (0.140) (0.140) Number of employees −0.166 −0.134 −0.166 −0.179 −0.100 −0.104 (0.110) (0.094) (0.110) (0.111) (0.107) (0.107) Radical innovation 0.294*** 0.134 0.294*** 0.296*** 0.025 0.025 (0.109) (0.100) (0.110) (0.109) (0.112) (0.111) Org innovation 0.411*** 0.234** 0.412*** 0.414*** 0.151 0.153 (0.110) (0.097) (0.110) (0.110) (0.108) (0.107) Number of R&D researchers 0.143 0.026 0.144 0.174 0.043 0.056 (0.135) (0.127) (0.136) (0.137) (0.140) (0.140) Wald Χ2 70.23 (18)*** 53.89 (20)*** 70.46 (19)*** 72.71 (20)*** 33.31 (19)** 33.69 (20)*** Firm-period observations 1000 1000 1000 1000 1000 1000 Firm observations 1805 1805 1805 1805 1805 1805 Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.034** 0.059*** 0.047*** 0.079*** 0.031* 0.044* (0.015) (0.023) (0.017) (0.026) (0.016) (0.025) R&D exploration-1 0.007 0.029 −0.001 0.029 0.011 0.023 (0.015) (0.021) (0.017) (0.024) (0.017) (0.024) R&D exploration X R&D exploration-1 −0.005 −0.007* −0.003 (0.004) (0.004) (0.004) R&D exploitation 0.024 0.022 0.024 0.021 0.010 0.009 (0.018) (0.016) (0.018) (0.018) (0.017) (0.017) Knowledge breadth 0.036* −0.009 0.036* 0.036* −0.010 −0.010 (0.019) (0.017) (0.019) (0.019) (0.019) (0.019) Knowledge depth 0.007 0.015 0.007 0.010 0.028 0.028 (0.039) (0.035) (0.039) (0.039) (0.038) (0.038) Other innovation expenditures −0.010 −0.012 −0.010 −0.010 −0.010 −0.010 (0.022) (0.019) (0.022) (0.022) (0.021) (0.021) Firm age 0.236** 0.170 0.236** 0.240** 0.059 0.060 (0.118) (0.105) (0.118) (0.118) (0.122) (0.122) Group 0.004 −0.029 0.004 0.004 0.005 0.004 (0.153) (0.135) (0.153) (0.153) (0.153) (0.153) Cost barriers −0.188 −0.174 −0.188 −0.183 −0.272** −0.270** (0.137) (0.120) (0.137) (0.137) (0.135) (0.134) Knowledge barriers 0.527 0.310 0.528 0.542 0.337 0.343 (0.344) (0.335) (0.345) (0.345) (0.367) (0.367) Market barriers—uncertainty 0.235* 0.341*** 0.235* 0.237* 0.348*** 0.347*** (0.129) (0.115) (0.129) (0.129) (0.126) (0.125) Market barriers—dominated 0.185 0.160 0.184 0.190 0.131 0.133 (0.139) (0.125) (0.139) (0.139) (0.140) (0.140) Number of employees −0.166 −0.134 −0.166 −0.179 −0.100 −0.104 (0.110) (0.094) (0.110) (0.111) (0.107) (0.107) Radical innovation 0.294*** 0.134 0.294*** 0.296*** 0.025 0.025 (0.109) (0.100) (0.110) (0.109) (0.112) (0.111) Org innovation 0.411*** 0.234** 0.412*** 0.414*** 0.151 0.153 (0.110) (0.097) (0.110) (0.110) (0.108) (0.107) Number of R&D researchers 0.143 0.026 0.144 0.174 0.043 0.056 (0.135) (0.127) (0.136) (0.137) (0.140) (0.140) Wald Χ2 70.23 (18)*** 53.89 (20)*** 70.46 (19)*** 72.71 (20)*** 33.31 (19)** 33.69 (20)*** Firm-period observations 1000 1000 1000 1000 1000 1000 Firm observations 1805 1805 1805 1805 1805 1805 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Columns 1 and 2, the probability to fail at the conception phase in Columns 3 and 4, and the probability to fail at the implementation phase in Columns 5 and 6. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels where the start date of the panel refers to the dependent variable (i.e., if t = 2007–2009, so t-1 = 2004–2006). * P < 0.10; ** P < 0.05; *** P < 0.01. Table A4. Innovation failures determinants: interaction effects for SMEs (less than 50 employees) Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.034** 0.059*** 0.047*** 0.079*** 0.031* 0.044* (0.015) (0.023) (0.017) (0.026) (0.016) (0.025) R&D exploration-1 0.007 0.029 −0.001 0.029 0.011 0.023 (0.015) (0.021) (0.017) (0.024) (0.017) (0.024) R&D exploration X R&D exploration-1 −0.005 −0.007* −0.003 (0.004) (0.004) (0.004) R&D exploitation 0.024 0.022 0.024 0.021 0.010 0.009 (0.018) (0.016) (0.018) (0.018) (0.017) (0.017) Knowledge breadth 0.036* −0.009 0.036* 0.036* −0.010 −0.010 (0.019) (0.017) (0.019) (0.019) (0.019) (0.019) Knowledge depth 0.007 0.015 0.007 0.010 0.028 0.028 (0.039) (0.035) (0.039) (0.039) (0.038) (0.038) Other innovation expenditures −0.010 −0.012 −0.010 −0.010 −0.010 −0.010 (0.022) (0.019) (0.022) (0.022) (0.021) (0.021) Firm age 0.236** 0.170 0.236** 0.240** 0.059 0.060 (0.118) (0.105) (0.118) (0.118) (0.122) (0.122) Group 0.004 −0.029 0.004 0.004 0.005 0.004 (0.153) (0.135) (0.153) (0.153) (0.153) (0.153) Cost barriers −0.188 −0.174 −0.188 −0.183 −0.272** −0.270** (0.137) (0.120) (0.137) (0.137) (0.135) (0.134) Knowledge barriers 0.527 0.310 0.528 0.542 0.337 0.343 (0.344) (0.335) (0.345) (0.345) (0.367) (0.367) Market barriers—uncertainty 0.235* 0.341*** 0.235* 0.237* 0.348*** 0.347*** (0.129) (0.115) (0.129) (0.129) (0.126) (0.125) Market barriers—dominated 0.185 0.160 0.184 0.190 0.131 0.133 (0.139) (0.125) (0.139) (0.139) (0.140) (0.140) Number of employees −0.166 −0.134 −0.166 −0.179 −0.100 −0.104 (0.110) (0.094) (0.110) (0.111) (0.107) (0.107) Radical innovation 0.294*** 0.134 0.294*** 0.296*** 0.025 0.025 (0.109) (0.100) (0.110) (0.109) (0.112) (0.111) Org innovation 0.411*** 0.234** 0.412*** 0.414*** 0.151 0.153 (0.110) (0.097) (0.110) (0.110) (0.108) (0.107) Number of R&D researchers 0.143 0.026 0.144 0.174 0.043 0.056 (0.135) (0.127) (0.136) (0.137) (0.140) (0.140) Wald Χ2 70.23 (18)*** 53.89 (20)*** 70.46 (19)*** 72.71 (20)*** 33.31 (19)** 33.69 (20)*** Firm-period observations 1000 1000 1000 1000 1000 1000 Firm observations 1805 1805 1805 1805 1805 1805 Failure overall Failure conception Failure implementation (1) (2) (3) (4) (5) (6) R&D exploration 0.034** 0.059*** 0.047*** 0.079*** 0.031* 0.044* (0.015) (0.023) (0.017) (0.026) (0.016) (0.025) R&D exploration-1 0.007 0.029 −0.001 0.029 0.011 0.023 (0.015) (0.021) (0.017) (0.024) (0.017) (0.024) R&D exploration X R&D exploration-1 −0.005 −0.007* −0.003 (0.004) (0.004) (0.004) R&D exploitation 0.024 0.022 0.024 0.021 0.010 0.009 (0.018) (0.016) (0.018) (0.018) (0.017) (0.017) Knowledge breadth 0.036* −0.009 0.036* 0.036* −0.010 −0.010 (0.019) (0.017) (0.019) (0.019) (0.019) (0.019) Knowledge depth 0.007 0.015 0.007 0.010 0.028 0.028 (0.039) (0.035) (0.039) (0.039) (0.038) (0.038) Other innovation expenditures −0.010 −0.012 −0.010 −0.010 −0.010 −0.010 (0.022) (0.019) (0.022) (0.022) (0.021) (0.021) Firm age 0.236** 0.170 0.236** 0.240** 0.059 0.060 (0.118) (0.105) (0.118) (0.118) (0.122) (0.122) Group 0.004 −0.029 0.004 0.004 0.005 0.004 (0.153) (0.135) (0.153) (0.153) (0.153) (0.153) Cost barriers −0.188 −0.174 −0.188 −0.183 −0.272** −0.270** (0.137) (0.120) (0.137) (0.137) (0.135) (0.134) Knowledge barriers 0.527 0.310 0.528 0.542 0.337 0.343 (0.344) (0.335) (0.345) (0.345) (0.367) (0.367) Market barriers—uncertainty 0.235* 0.341*** 0.235* 0.237* 0.348*** 0.347*** (0.129) (0.115) (0.129) (0.129) (0.126) (0.125) Market barriers—dominated 0.185 0.160 0.184 0.190 0.131 0.133 (0.139) (0.125) (0.139) (0.139) (0.140) (0.140) Number of employees −0.166 −0.134 −0.166 −0.179 −0.100 −0.104 (0.110) (0.094) (0.110) (0.111) (0.107) (0.107) Radical innovation 0.294*** 0.134 0.294*** 0.296*** 0.025 0.025 (0.109) (0.100) (0.110) (0.109) (0.112) (0.111) Org innovation 0.411*** 0.234** 0.412*** 0.414*** 0.151 0.153 (0.110) (0.097) (0.110) (0.110) (0.108) (0.107) Number of R&D researchers 0.143 0.026 0.144 0.174 0.043 0.056 (0.135) (0.127) (0.136) (0.137) (0.140) (0.140) Wald Χ2 70.23 (18)*** 53.89 (20)*** 70.46 (19)*** 72.71 (20)*** 33.31 (19)** 33.69 (20)*** Firm-period observations 1000 1000 1000 1000 1000 1000 Firm observations 1805 1805 1805 1805 1805 1805 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Columns 1 and 2, the probability to fail at the conception phase in Columns 3 and 4, and the probability to fail at the implementation phase in Columns 5 and 6. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels where the start date of the panel refers to the dependent variable (i.e., if t = 2007–2009, so t-1 = 2004–2006). * P < 0.10; ** P < 0.05; *** P < 0.01. Table A5. Innovation failures determinants: curvilinear effects for SMEs (less than 50 employees) Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.163*** 0.201*** 0.206*** (0.060) (0.070) (0.068) R&D exploration sq −0.016** −0.020** −0.022*** (0.008) (0.009) (0.008) R&D exploitation 0.018 0.017 0.002 (0.016) (0.018) (0.017) Knowledge breadth −0.006 0.038** −0.006 (0.017) (0.019) (0.019) Knowledge depth 0.016 0.011 0.030 (0.035) (0.039) (0.038) Other innovation expenditures −0.007 −0.004 −0.003 (0.020) (0.022) (0.022) Firm age 0.161 0.230** 0.049 (0.104) (0.116) (0.120) Group −0.020 0.009 0.016 (0.135) (0.152) (0.153) Cost barriers −0.177 −0.188 −0.271** (0.119) (0.135) (0.133) Knowledge barriers 0.308 0.530 0.344 (0.330) (0.338) (0.361) Market barriers—uncertainty 0.326*** 0.214* 0.325*** (0.115) (0.129) (0.125) Market barriers—dominated 0.166 0.198 0.142 (0.125) (0.138) (0.139) Number of employees −0.178* −0.225** −0.174 (0.096) (0.113) (0.109) Radical innovation 0.135 0.295*** 0.027 (0.100) (0.108) (0.111) Org innovation 0.229** 0.407*** 0.146 (0.097) (0.108) (0.107) Number of R&D researchers 0.100 0.256* 0.179 (0.133) (0.144) (0.147) Wald Χ2 58.36 (19)*** 75.76 (19)*** 40.02 (19) Firm-period observations 1000 1000 1000 Firm observations 1805 1805 1805 Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.163*** 0.201*** 0.206*** (0.060) (0.070) (0.068) R&D exploration sq −0.016** −0.020** −0.022*** (0.008) (0.009) (0.008) R&D exploitation 0.018 0.017 0.002 (0.016) (0.018) (0.017) Knowledge breadth −0.006 0.038** −0.006 (0.017) (0.019) (0.019) Knowledge depth 0.016 0.011 0.030 (0.035) (0.039) (0.038) Other innovation expenditures −0.007 −0.004 −0.003 (0.020) (0.022) (0.022) Firm age 0.161 0.230** 0.049 (0.104) (0.116) (0.120) Group −0.020 0.009 0.016 (0.135) (0.152) (0.153) Cost barriers −0.177 −0.188 −0.271** (0.119) (0.135) (0.133) Knowledge barriers 0.308 0.530 0.344 (0.330) (0.338) (0.361) Market barriers—uncertainty 0.326*** 0.214* 0.325*** (0.115) (0.129) (0.125) Market barriers—dominated 0.166 0.198 0.142 (0.125) (0.138) (0.139) Number of employees −0.178* −0.225** −0.174 (0.096) (0.113) (0.109) Radical innovation 0.135 0.295*** 0.027 (0.100) (0.108) (0.111) Org innovation 0.229** 0.407*** 0.146 (0.097) (0.108) (0.107) Number of R&D researchers 0.100 0.256* 0.179 (0.133) (0.144) (0.147) Wald Χ2 58.36 (19)*** 75.76 (19)*** 40.02 (19) Firm-period observations 1000 1000 1000 Firm observations 1805 1805 1805 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Column 1, the probability to fail at the conception phase in Column 2, and the probability to fail at the implementation phase in Column 3. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels. * P < 0.10; ** P < 0.05; *** P < 0.01. Table A5. Innovation failures determinants: curvilinear effects for SMEs (less than 50 employees) Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.163*** 0.201*** 0.206*** (0.060) (0.070) (0.068) R&D exploration sq −0.016** −0.020** −0.022*** (0.008) (0.009) (0.008) R&D exploitation 0.018 0.017 0.002 (0.016) (0.018) (0.017) Knowledge breadth −0.006 0.038** −0.006 (0.017) (0.019) (0.019) Knowledge depth 0.016 0.011 0.030 (0.035) (0.039) (0.038) Other innovation expenditures −0.007 −0.004 −0.003 (0.020) (0.022) (0.022) Firm age 0.161 0.230** 0.049 (0.104) (0.116) (0.120) Group −0.020 0.009 0.016 (0.135) (0.152) (0.153) Cost barriers −0.177 −0.188 −0.271** (0.119) (0.135) (0.133) Knowledge barriers 0.308 0.530 0.344 (0.330) (0.338) (0.361) Market barriers—uncertainty 0.326*** 0.214* 0.325*** (0.115) (0.129) (0.125) Market barriers—dominated 0.166 0.198 0.142 (0.125) (0.138) (0.139) Number of employees −0.178* −0.225** −0.174 (0.096) (0.113) (0.109) Radical innovation 0.135 0.295*** 0.027 (0.100) (0.108) (0.111) Org innovation 0.229** 0.407*** 0.146 (0.097) (0.108) (0.107) Number of R&D researchers 0.100 0.256* 0.179 (0.133) (0.144) (0.147) Wald Χ2 58.36 (19)*** 75.76 (19)*** 40.02 (19) Firm-period observations 1000 1000 1000 Firm observations 1805 1805 1805 Failure overall Failure conception Failure implementation (1) (2) (3) R&D exploration 0.163*** 0.201*** 0.206*** (0.060) (0.070) (0.068) R&D exploration sq −0.016** −0.020** −0.022*** (0.008) (0.009) (0.008) R&D exploitation 0.018 0.017 0.002 (0.016) (0.018) (0.017) Knowledge breadth −0.006 0.038** −0.006 (0.017) (0.019) (0.019) Knowledge depth 0.016 0.011 0.030 (0.035) (0.039) (0.038) Other innovation expenditures −0.007 −0.004 −0.003 (0.020) (0.022) (0.022) Firm age 0.161 0.230** 0.049 (0.104) (0.116) (0.120) Group −0.020 0.009 0.016 (0.135) (0.152) (0.153) Cost barriers −0.177 −0.188 −0.271** (0.119) (0.135) (0.133) Knowledge barriers 0.308 0.530 0.344 (0.330) (0.338) (0.361) Market barriers—uncertainty 0.326*** 0.214* 0.325*** (0.115) (0.129) (0.125) Market barriers—dominated 0.166 0.198 0.142 (0.125) (0.138) (0.139) Number of employees −0.178* −0.225** −0.174 (0.096) (0.113) (0.109) Radical innovation 0.135 0.295*** 0.027 (0.100) (0.108) (0.111) Org innovation 0.229** 0.407*** 0.146 (0.097) (0.108) (0.107) Number of R&D researchers 0.100 0.256* 0.179 (0.133) (0.144) (0.147) Wald Χ2 58.36 (19)*** 75.76 (19)*** 40.02 (19) Firm-period observations 1000 1000 1000 Firm observations 1805 1805 1805 Random effects probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail in any phase in Column 1, the probability to fail at the conception phase in Column 2, and the probability to fail at the implementation phase in Column 3. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels. * P < 0.10; ** P < 0.05; *** P < 0.01. Table A6. Innovation failures determinants: bivariate probit regressions (1) (2) (3) (4) Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation R&D exploration 0.037*** 0.022*** 0.034*** 0.019** 0.050*** 0.028** 0.146*** 0.103*** (0.008) (0.008) (0.008) (0.008) (0.012) (0.012) (0.029) (0.030) R&D exploration-1 0.006 0.005 0.023** 0.014 (0.008) (0.008) (0.012) (0.012) R&D exploration × R&D exploration-1 −0.004** −0.002 (0.002) (0.002) R&D exploration sq −0.014*** −0.011*** (0.004) (0.004) R&D exploitation 0.022*** −0.001 0.022*** −0.001 0.020** −0.002 0.016* −0.005 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.009) (0.008) Knowledge breadth 0.042*** 0.012 0.041*** 0.012 0.041*** 0.012 0.042*** 0.012 (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) Knowledge depth 0.012 0.001 0.012 0.001 0.013 0.001 0.012 0.001 (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) Other innovation expenditures −0.007 −0.004 −0.007 −0.003 −0.006 −0.003 0.000 0.002 (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.011) Firm age 0.079* 0.097** 0.079* 0.096** 0.079* 0.097** 0.073 0.092* (0.048) (0.049) (0.048) (0.049) (0.048) (0.049) (0.047) (0.049) Group 0.014 0.083 0.013 0.082 0.014 0.083 0.027 0.092 (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) Cost barriers −0.072 −0.080 −0.072 −0.081 −0.074 −0.082 −0.076 −0.083 (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) Knowledge barriers 0.073 0.150 0.071 0.148 0.076 0.151 0.071 0.147 (0.215) (0.213) (0.215) (0.213) (0.214) (0.214) (0.214) (0.213) Market barriers—uncertainty 0.138** 0.205*** 0.139** 0.206*** 0.140** 0.206*** 0.130** 0.199*** (0.058) (0.058) (0.058) (0.059) (0.058) (0.058) (0.058) (0.058) Market barriers—dominated 0.110* 0.059 0.112* 0.060 0.115* 0.062 0.114* 0.061 (0.065) (0.067) (0.066) (0.067) (0.066) (0.067) (0.065) (0.067) Number of employees 0.040 0.014 0.042 0.016 0.038 0.013 0.005 −0.011 (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.030) (0.030) Radical innovation 0.213*** 0.095* 0.212*** 0.094* 0.213*** 0.094* 0.215*** 0.096* (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) Org innovation 0.335*** 0.161*** 0.334*** 0.160*** 0.334*** 0.160*** 0.330*** 0.157*** (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) Number of R&D researchers −0.009 0.073* −0.013 0.069* 0.001 0.077* 0.049 0.116*** (0.040) (0.041) (0.040) (0.042) (0.041) (0.042) (0.043) (0.044) Wald Χ2 285.16 (36)*** 285.43 (38)*** 285.71 (40)*** 299.24 (38)*** Firm-period observations 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 (1) (2) (3) (4) Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation R&D exploration 0.037*** 0.022*** 0.034*** 0.019** 0.050*** 0.028** 0.146*** 0.103*** (0.008) (0.008) (0.008) (0.008) (0.012) (0.012) (0.029) (0.030) R&D exploration-1 0.006 0.005 0.023** 0.014 (0.008) (0.008) (0.012) (0.012) R&D exploration × R&D exploration-1 −0.004** −0.002 (0.002) (0.002) R&D exploration sq −0.014*** −0.011*** (0.004) (0.004) R&D exploitation 0.022*** −0.001 0.022*** −0.001 0.020** −0.002 0.016* −0.005 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.009) (0.008) Knowledge breadth 0.042*** 0.012 0.041*** 0.012 0.041*** 0.012 0.042*** 0.012 (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) Knowledge depth 0.012 0.001 0.012 0.001 0.013 0.001 0.012 0.001 (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) Other innovation expenditures −0.007 −0.004 −0.007 −0.003 −0.006 −0.003 0.000 0.002 (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.011) Firm age 0.079* 0.097** 0.079* 0.096** 0.079* 0.097** 0.073 0.092* (0.048) (0.049) (0.048) (0.049) (0.048) (0.049) (0.047) (0.049) Group 0.014 0.083 0.013 0.082 0.014 0.083 0.027 0.092 (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) Cost barriers −0.072 −0.080 −0.072 −0.081 −0.074 −0.082 −0.076 −0.083 (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) Knowledge barriers 0.073 0.150 0.071 0.148 0.076 0.151 0.071 0.147 (0.215) (0.213) (0.215) (0.213) (0.214) (0.214) (0.214) (0.213) Market barriers—uncertainty 0.138** 0.205*** 0.139** 0.206*** 0.140** 0.206*** 0.130** 0.199*** (0.058) (0.058) (0.058) (0.059) (0.058) (0.058) (0.058) (0.058) Market barriers—dominated 0.110* 0.059 0.112* 0.060 0.115* 0.062 0.114* 0.061 (0.065) (0.067) (0.066) (0.067) (0.066) (0.067) (0.065) (0.067) Number of employees 0.040 0.014 0.042 0.016 0.038 0.013 0.005 −0.011 (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.030) (0.030) Radical innovation 0.213*** 0.095* 0.212*** 0.094* 0.213*** 0.094* 0.215*** 0.096* (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) Org innovation 0.335*** 0.161*** 0.334*** 0.160*** 0.334*** 0.160*** 0.330*** 0.157*** (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) Number of R&D researchers −0.009 0.073* −0.013 0.069* 0.001 0.077* 0.049 0.116*** (0.040) (0.041) (0.040) (0.042) (0.041) (0.042) (0.043) (0.044) Wald Χ2 285.16 (36)*** 285.43 (38)*** 285.71 (40)*** 299.24 (38)*** Firm-period observations 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 Bivariate probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail at the conception phase in even and the probability to fail at the implementation phase in odd columns. Base sample in all columns is a cross-section (2004–2012). * P < 0.10; ** P < 0.05; *** P < 0.01. Table A6. Innovation failures determinants: bivariate probit regressions (1) (2) (3) (4) Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation R&D exploration 0.037*** 0.022*** 0.034*** 0.019** 0.050*** 0.028** 0.146*** 0.103*** (0.008) (0.008) (0.008) (0.008) (0.012) (0.012) (0.029) (0.030) R&D exploration-1 0.006 0.005 0.023** 0.014 (0.008) (0.008) (0.012) (0.012) R&D exploration × R&D exploration-1 −0.004** −0.002 (0.002) (0.002) R&D exploration sq −0.014*** −0.011*** (0.004) (0.004) R&D exploitation 0.022*** −0.001 0.022*** −0.001 0.020** −0.002 0.016* −0.005 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.009) (0.008) Knowledge breadth 0.042*** 0.012 0.041*** 0.012 0.041*** 0.012 0.042*** 0.012 (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) Knowledge depth 0.012 0.001 0.012 0.001 0.013 0.001 0.012 0.001 (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) Other innovation expenditures −0.007 −0.004 −0.007 −0.003 −0.006 −0.003 0.000 0.002 (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.011) Firm age 0.079* 0.097** 0.079* 0.096** 0.079* 0.097** 0.073 0.092* (0.048) (0.049) (0.048) (0.049) (0.048) (0.049) (0.047) (0.049) Group 0.014 0.083 0.013 0.082 0.014 0.083 0.027 0.092 (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) Cost barriers −0.072 −0.080 −0.072 −0.081 −0.074 −0.082 −0.076 −0.083 (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) Knowledge barriers 0.073 0.150 0.071 0.148 0.076 0.151 0.071 0.147 (0.215) (0.213) (0.215) (0.213) (0.214) (0.214) (0.214) (0.213) Market barriers—uncertainty 0.138** 0.205*** 0.139** 0.206*** 0.140** 0.206*** 0.130** 0.199*** (0.058) (0.058) (0.058) (0.059) (0.058) (0.058) (0.058) (0.058) Market barriers—dominated 0.110* 0.059 0.112* 0.060 0.115* 0.062 0.114* 0.061 (0.065) (0.067) (0.066) (0.067) (0.066) (0.067) (0.065) (0.067) Number of employees 0.040 0.014 0.042 0.016 0.038 0.013 0.005 −0.011 (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.030) (0.030) Radical innovation 0.213*** 0.095* 0.212*** 0.094* 0.213*** 0.094* 0.215*** 0.096* (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) Org innovation 0.335*** 0.161*** 0.334*** 0.160*** 0.334*** 0.160*** 0.330*** 0.157*** (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) Number of R&D researchers −0.009 0.073* −0.013 0.069* 0.001 0.077* 0.049 0.116*** (0.040) (0.041) (0.040) (0.042) (0.041) (0.042) (0.043) (0.044) Wald Χ2 285.16 (36)*** 285.43 (38)*** 285.71 (40)*** 299.24 (38)*** Firm-period observations 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 (1) (2) (3) (4) Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation R&D exploration 0.037*** 0.022*** 0.034*** 0.019** 0.050*** 0.028** 0.146*** 0.103*** (0.008) (0.008) (0.008) (0.008) (0.012) (0.012) (0.029) (0.030) R&D exploration-1 0.006 0.005 0.023** 0.014 (0.008) (0.008) (0.012) (0.012) R&D exploration × R&D exploration-1 −0.004** −0.002 (0.002) (0.002) R&D exploration sq −0.014*** −0.011*** (0.004) (0.004) R&D exploitation 0.022*** −0.001 0.022*** −0.001 0.020** −0.002 0.016* −0.005 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.009) (0.008) Knowledge breadth 0.042*** 0.012 0.041*** 0.012 0.041*** 0.012 0.042*** 0.012 (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) Knowledge depth 0.012 0.001 0.012 0.001 0.013 0.001 0.012 0.001 (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) Other innovation expenditures −0.007 −0.004 −0.007 −0.003 −0.006 −0.003 0.000 0.002 (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.011) Firm age 0.079* 0.097** 0.079* 0.096** 0.079* 0.097** 0.073 0.092* (0.048) (0.049) (0.048) (0.049) (0.048) (0.049) (0.047) (0.049) Group 0.014 0.083 0.013 0.082 0.014 0.083 0.027 0.092 (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) Cost barriers −0.072 −0.080 −0.072 −0.081 −0.074 −0.082 −0.076 −0.083 (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) Knowledge barriers 0.073 0.150 0.071 0.148 0.076 0.151 0.071 0.147 (0.215) (0.213) (0.215) (0.213) (0.214) (0.214) (0.214) (0.213) Market barriers—uncertainty 0.138** 0.205*** 0.139** 0.206*** 0.140** 0.206*** 0.130** 0.199*** (0.058) (0.058) (0.058) (0.059) (0.058) (0.058) (0.058) (0.058) Market barriers—dominated 0.110* 0.059 0.112* 0.060 0.115* 0.062 0.114* 0.061 (0.065) (0.067) (0.066) (0.067) (0.066) (0.067) (0.065) (0.067) Number of employees 0.040 0.014 0.042 0.016 0.038 0.013 0.005 −0.011 (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.030) (0.030) Radical innovation 0.213*** 0.095* 0.212*** 0.094* 0.213*** 0.094* 0.215*** 0.096* (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) Org innovation 0.335*** 0.161*** 0.334*** 0.160*** 0.334*** 0.160*** 0.330*** 0.157*** (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) Number of R&D researchers −0.009 0.073* −0.013 0.069* 0.001 0.077* 0.049 0.116*** (0.040) (0.041) (0.040) (0.042) (0.041) (0.042) (0.043) (0.044) Wald Χ2 285.16 (36)*** 285.43 (38)*** 285.71 (40)*** 299.24 (38)*** Firm-period observations 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 Bivariate probit regressions with robust standard errors clustered by firm are in parentheses. The degrees of freedom of Wald χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail at the conception phase in even and the probability to fail at the implementation phase in odd columns. Base sample in all columns is a cross-section (2004–2012). * P < 0.10; ** P < 0.05; *** P < 0.01. Table A7. Innovation failures determinants: probit seemingly unrelated regression (1) (2) (3) (4) Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation R&D exploration 0.037*** 0.022*** 0.034*** 0.019** 0.050*** 0.028** 0.146*** 0.103*** (0.008) (0.008) (0.008) (0.008) (0.012) (0.012) (0.029) (0.030) R&D exploration-1 0.006 0.005 0.023** 0.014 (0.008) (0.008) (0.012) (0.012) R&D exploration × R&D exploration-1 −0.004** −0.002 (0.002) (0.002) R&D exploration sq −0.014*** −0.011*** (0.004) (0.004) R&D exploitation 0.022*** −0.001 0.022*** −0.001 0.020** −0.002 0.016* −0.005 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.009) (0.008) Knowledge breadth 0.042*** 0.012 0.041*** 0.012 0.041*** 0.012 0.042*** 0.012 (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) Knowledge depth 0.012 0.001 0.012 0.001 0.013 0.001 0.012 0.001 (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) Other innovation expenditures −0.007 −0.004 −0.007 −0.003 −0.006 −0.003 0.000 0.002 (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.011) Firm age 0.079* 0.097** 0.079* 0.096** 0.079* 0.097** 0.073 0.092* (0.048) (0.049) (0.048) (0.049) (0.048) (0.049) (0.047) (0.049) Group 0.014 0.083 0.013 0.082 0.014 0.083 0.027 0.092 (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) Cost barriers −0.072 −0.080 −0.072 −0.081 −0.074 −0.082 −0.076 −0.083 (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) Knowledge barriers 0.073 0.150 0.071 0.148 0.076 0.151 0.071 0.147 (0.215) (0.213) (0.215) (0.213) (0.214) (0.214) (0.214) (0.213) Market barriers—uncertainty 0.138** 0.205*** 0.139** 0.206*** 0.140** 0.206*** 0.130** 0.199*** (0.058) (0.058) (0.058) (0.059) (0.058) (0.058) (0.058) (0.058) Market barriers—dominated 0.110* 0.059 0.112* 0.060 0.115* 0.062 0.114* 0.061 (0.065) (0.067) (0.066) (0.067) (0.066) (0.067) (0.065) (0.067) Number of employees 0.040 0.014 0.042 0.016 0.038 0.013 0.005 −0.011 (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.030) (0.030) Radical innovation 0.213*** 0.095* 0.212*** 0.094* 0.213*** 0.094* 0.215*** 0.096* (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) Org innovation 0.335*** 0.161*** 0.334*** 0.160*** 0.334*** 0.160*** 0.330*** 0.157*** (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) Number of R&D researchers −0.009 0.073* −0.013 0.069* 0.001 0.077* 0.049 0.116*** (0.040) (0.041) (0.040) (0.042) (0.041) (0.042) (0.043) (0.044) Wald Χ2 285.16 (36)*** 285.43 (38)*** 285.71 (40)*** 299.24 (38)*** Firm-period observations 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 (1) (2) (3) (4) Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation R&D exploration 0.037*** 0.022*** 0.034*** 0.019** 0.050*** 0.028** 0.146*** 0.103*** (0.008) (0.008) (0.008) (0.008) (0.012) (0.012) (0.029) (0.030) R&D exploration-1 0.006 0.005 0.023** 0.014 (0.008) (0.008) (0.012) (0.012) R&D exploration × R&D exploration-1 −0.004** −0.002 (0.002) (0.002) R&D exploration sq −0.014*** −0.011*** (0.004) (0.004) R&D exploitation 0.022*** −0.001 0.022*** −0.001 0.020** −0.002 0.016* −0.005 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.009) (0.008) Knowledge breadth 0.042*** 0.012 0.041*** 0.012 0.041*** 0.012 0.042*** 0.012 (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) Knowledge depth 0.012 0.001 0.012 0.001 0.013 0.001 0.012 0.001 (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) Other innovation expenditures −0.007 −0.004 −0.007 −0.003 −0.006 −0.003 0.000 0.002 (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.011) Firm age 0.079* 0.097** 0.079* 0.096** 0.079* 0.097** 0.073 0.092* (0.048) (0.049) (0.048) (0.049) (0.048) (0.049) (0.047) (0.049) Group 0.014 0.083 0.013 0.082 0.014 0.083 0.027 0.092 (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) Cost barriers −0.072 −0.080 −0.072 −0.081 −0.074 −0.082 −0.076 −0.083 (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) Knowledge barriers 0.073 0.150 0.071 0.148 0.076 0.151 0.071 0.147 (0.215) (0.213) (0.215) (0.213) (0.214) (0.214) (0.214) (0.213) Market barriers—uncertainty 0.138** 0.205*** 0.139** 0.206*** 0.140** 0.206*** 0.130** 0.199*** (0.058) (0.058) (0.058) (0.059) (0.058) (0.058) (0.058) (0.058) Market barriers—dominated 0.110* 0.059 0.112* 0.060 0.115* 0.062 0.114* 0.061 (0.065) (0.067) (0.066) (0.067) (0.066) (0.067) (0.065) (0.067) Number of employees 0.040 0.014 0.042 0.016 0.038 0.013 0.005 −0.011 (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.030) (0.030) Radical innovation 0.213*** 0.095* 0.212*** 0.094* 0.213*** 0.094* 0.215*** 0.096* (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) Org innovation 0.335*** 0.161*** 0.334*** 0.160*** 0.334*** 0.160*** 0.330*** 0.157*** (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) Number of R&D researchers −0.009 0.073* −0.013 0.069* 0.001 0.077* 0.049 0.116*** (0.040) (0.041) (0.040) (0.042) (0.041) (0.042) (0.043) (0.044) Wald Χ2 285.16 (36)*** 285.43 (38)*** 285.71 (40)*** 299.24 (38)*** Firm-period observations 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 Probit SURE with robust standard errors clustered by firm in parentheses. The degrees of freedom of Wald Χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions and Wald Χ2 test is reported with the joint significance. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail at the conception phase in odd columns and the probability to fail at the implementation phase in even columns. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels where the start date of the panel refers to the dependent variable (i.e., if t = 2007–2009, so t-1 = 2004–2006). * P < 0.10; ** P < 0.05; *** P < 0.01. Table A7. Innovation failures determinants: probit seemingly unrelated regression (1) (2) (3) (4) Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation R&D exploration 0.037*** 0.022*** 0.034*** 0.019** 0.050*** 0.028** 0.146*** 0.103*** (0.008) (0.008) (0.008) (0.008) (0.012) (0.012) (0.029) (0.030) R&D exploration-1 0.006 0.005 0.023** 0.014 (0.008) (0.008) (0.012) (0.012) R&D exploration × R&D exploration-1 −0.004** −0.002 (0.002) (0.002) R&D exploration sq −0.014*** −0.011*** (0.004) (0.004) R&D exploitation 0.022*** −0.001 0.022*** −0.001 0.020** −0.002 0.016* −0.005 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.009) (0.008) Knowledge breadth 0.042*** 0.012 0.041*** 0.012 0.041*** 0.012 0.042*** 0.012 (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) Knowledge depth 0.012 0.001 0.012 0.001 0.013 0.001 0.012 0.001 (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) Other innovation expenditures −0.007 −0.004 −0.007 −0.003 −0.006 −0.003 0.000 0.002 (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.011) Firm age 0.079* 0.097** 0.079* 0.096** 0.079* 0.097** 0.073 0.092* (0.048) (0.049) (0.048) (0.049) (0.048) (0.049) (0.047) (0.049) Group 0.014 0.083 0.013 0.082 0.014 0.083 0.027 0.092 (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) Cost barriers −0.072 −0.080 −0.072 −0.081 −0.074 −0.082 −0.076 −0.083 (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) Knowledge barriers 0.073 0.150 0.071 0.148 0.076 0.151 0.071 0.147 (0.215) (0.213) (0.215) (0.213) (0.214) (0.214) (0.214) (0.213) Market barriers—uncertainty 0.138** 0.205*** 0.139** 0.206*** 0.140** 0.206*** 0.130** 0.199*** (0.058) (0.058) (0.058) (0.059) (0.058) (0.058) (0.058) (0.058) Market barriers—dominated 0.110* 0.059 0.112* 0.060 0.115* 0.062 0.114* 0.061 (0.065) (0.067) (0.066) (0.067) (0.066) (0.067) (0.065) (0.067) Number of employees 0.040 0.014 0.042 0.016 0.038 0.013 0.005 −0.011 (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.030) (0.030) Radical innovation 0.213*** 0.095* 0.212*** 0.094* 0.213*** 0.094* 0.215*** 0.096* (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) Org innovation 0.335*** 0.161*** 0.334*** 0.160*** 0.334*** 0.160*** 0.330*** 0.157*** (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) Number of R&D researchers −0.009 0.073* −0.013 0.069* 0.001 0.077* 0.049 0.116*** (0.040) (0.041) (0.040) (0.042) (0.041) (0.042) (0.043) (0.044) Wald Χ2 285.16 (36)*** 285.43 (38)*** 285.71 (40)*** 299.24 (38)*** Firm-period observations 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 (1) (2) (3) (4) Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation Failure conception Failure implementation R&D exploration 0.037*** 0.022*** 0.034*** 0.019** 0.050*** 0.028** 0.146*** 0.103*** (0.008) (0.008) (0.008) (0.008) (0.012) (0.012) (0.029) (0.030) R&D exploration-1 0.006 0.005 0.023** 0.014 (0.008) (0.008) (0.012) (0.012) R&D exploration × R&D exploration-1 −0.004** −0.002 (0.002) (0.002) R&D exploration sq −0.014*** −0.011*** (0.004) (0.004) R&D exploitation 0.022*** −0.001 0.022*** −0.001 0.020** −0.002 0.016* −0.005 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) (0.009) (0.008) Knowledge breadth 0.042*** 0.012 0.041*** 0.012 0.041*** 0.012 0.042*** 0.012 (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) (0.008) (0.009) Knowledge depth 0.012 0.001 0.012 0.001 0.013 0.001 0.012 0.001 (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) Other innovation expenditures −0.007 −0.004 −0.007 −0.003 −0.006 −0.003 0.000 0.002 (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) (0.011) Firm age 0.079* 0.097** 0.079* 0.096** 0.079* 0.097** 0.073 0.092* (0.048) (0.049) (0.048) (0.049) (0.048) (0.049) (0.047) (0.049) Group 0.014 0.083 0.013 0.082 0.014 0.083 0.027 0.092 (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.061) Cost barriers −0.072 −0.080 −0.072 −0.081 −0.074 −0.082 −0.076 −0.083 (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) (0.068) Knowledge barriers 0.073 0.150 0.071 0.148 0.076 0.151 0.071 0.147 (0.215) (0.213) (0.215) (0.213) (0.214) (0.214) (0.214) (0.213) Market barriers—uncertainty 0.138** 0.205*** 0.139** 0.206*** 0.140** 0.206*** 0.130** 0.199*** (0.058) (0.058) (0.058) (0.059) (0.058) (0.058) (0.058) (0.058) Market barriers—dominated 0.110* 0.059 0.112* 0.060 0.115* 0.062 0.114* 0.061 (0.065) (0.067) (0.066) (0.067) (0.066) (0.067) (0.065) (0.067) Number of employees 0.040 0.014 0.042 0.016 0.038 0.013 0.005 −0.011 (0.029) (0.029) (0.029) (0.029) (0.029) (0.029) (0.030) (0.030) Radical innovation 0.213*** 0.095* 0.212*** 0.094* 0.213*** 0.094* 0.215*** 0.096* (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) (0.049) (0.051) Org innovation 0.335*** 0.161*** 0.334*** 0.160*** 0.334*** 0.160*** 0.330*** 0.157*** (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) (0.050) (0.051) Number of R&D researchers −0.009 0.073* −0.013 0.069* 0.001 0.077* 0.049 0.116*** (0.040) (0.041) (0.040) (0.042) (0.041) (0.042) (0.043) (0.044) Wald Χ2 285.16 (36)*** 285.43 (38)*** 285.71 (40)*** 299.24 (38)*** Firm-period observations 2226 2226 2226 2226 Firm observations 4191 4191 4191 4191 Probit SURE with robust standard errors clustered by firm in parentheses. The degrees of freedom of Wald Χ2 test for the overall significance of variables are reported in parentheses. Dummies for Pavitt (1984) four categories of industrial firms are included (but not shown) in all regressions and Wald Χ2 test is reported with the joint significance. Year dummies are not included, as the joint test of significance is never rejected at the standard confidence levels. Dependent variable is the probability to fail at the conception phase in odd columns and the probability to fail at the implementation phase in even columns. Base sample in all columns is an unbalanced panel, 2009–2012, with data at 3-year intervals in levels where the start date of the panel refers to the dependent variable (i.e., if t = 2007–2009, so t-1 = 2004–2006). * P < 0.10; ** P < 0.05; *** P < 0.01. © The Author 2017. Published by Oxford University Press on behalf of Associazione ICC. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Industrial and Corporate Change Oxford University Press http://www.deepdyve.com/lp/oxford-university-press/exploring-and-yet-failing-less-learning-from-past-and-current-KeT8BipcN3